text
stringlengths
254
1.16M
--- title: The signaling protein GIV/Girdin mediates the Nephrin-dependent insulin secretion of pancreatic islet β cells in response to high glucose authors: - Hao Wang - Ying-Chao Yuan - Cong Chang - Tetsuro Izumi - Hong-Hui Wang - Jin-Kui Yang journal: The Journal of Biological Chemistry year: 2023 pmcid: PMC10040812 doi: 10.1016/j.jbc.2023.103045 license: CC BY 4.0 --- # The signaling protein GIV/Girdin mediates the Nephrin-dependent insulin secretion of pancreatic islet β cells in response to high glucose ## Body The pancreatic islet β cells dynamically regulate insulin secretion to maintain glucose homeostasis [1]. The dysfunction of islet cells leads to hyperglycemia in type I and type II diabetes mellitus, the most common chronic disease affecting over 415 million people globally [2]. In physiological conditions, the rise in blood glucose levels is rapidly countered by insulin secretion from pancreatic islet β cells [3]. Subsequently, the released insulin can promote the uptake and storage of blood glucose in peripheral tissues, e.g., the liver, adipose tissue, and skeletal muscle [4]. The function of β cells for glucose-stimulated insulin secretion should be tightly regulated to ensure that a suitable amount of insulin is released. The molecular mechanism of glucose-stimulated insulin secretion remains complex, is not fully understood, and is currently a major research focus in the literature. Naturally, the process of glucose-stimulated insulin secretion of pancreatic islet β cells is biphasic [3]. In the first-phase secretion, a releasable pool of insulin secretory granules localizes close to the plasma membrane, and elevated glucose triggers the calcium influx to initiate the fusion of a small number of insulin-containing granules predocked at the plasma membrane; this occurs very quickly, typically in several minutes. In contrast, the second phase of insulin secretion is a more sustained process that lasts for hours under persisting glucose stimulation [5]. Under resting-state conditions, β cells have a dense web of cortical filamentous actin (F-actin) beneath the cell membrane to block the access of insulin granules to the cell periphery. In response to the elevated glucose level, β cells transduce the signaling to initiate the F-actin remodeling, which allows the movement of insulin granules to translocate and fuse with the plasma membrane for insulin release [5]. However, it remains unclear how the β cells could sense the elevated glucose and transmit proper intracellular signaling to tune insulin secretion. Recent evidence has suggested that the highly regulated intercellular junction between β cells via their cell surface receptors might be an important glucose-sensory interface for the proper regulation of insulin secretion [6, 7, 8]. For instance, the adherens junction, composed of E-cadherin and N-cadherin, is important for properly regulating the insulin secretion of islet β cells. Physiological or pathological conditions altering the expression or subcellular localization of adherens junctions might significantly impact the β-cell function in insulin secretion [7, 9]. Nevertheless, the signaling pathway coordination between the cell–cell junction and F-actin remodeling for insulin secretion remains largely unexplored. As a specialized component of adherens junction protein, *Nephrin is* an immunoglobulin-like protein that can combine with podocyte-associated membrane proteins, such as podocin, Neph proteins, and cadherin superfamily members, forming the slit diaphragm to bridge glomerular podocytes for regulated kidney filtration [10]. In addition to mediating cell–cell junction, Nephrin can also function as the key signaling molecule that maintains the filtration slits and podocyte survival by recruiting phosphoinositide 3-kinase (PI3K) for the subsequent activation of prosurvival Akt signaling [10, 11, 12]. Interestingly, *Nephrin is* also expressed in other tissues, such as the brain and pancreas, and it has been identified that Nephrin localizes in the cell–cell junction of β cells [13]. Nephrin has been recently reported to facilitate glucose-stimulated insulin release by pancreatic beta cells through dynamin-dependent Nephrin phosphorylation and endocytosis [11]. The mice with Nephrin pancreatic β cell–specific deletion have an impaired glucose sensing function of pancreatic β cells [14]. However, the signaling machinery mediating the Nephrin function on the cell surface to intracellular signaling pathway and F-actin remodeling for insulin secretion in islet β cells remains uncharacterized. GIV protein (also known as APE, Girdin, and HkRP1; gene, CCDC88A) has recently emerged as a multidomain signaling regulator to transduce the ligand-induced receptor activation at the plasma membrane and relay intracellular signaling pathways, e.g., PI3K-Akt, PKA/CREB, mTOR [15, 16, 17]. More specifically, GIV contains a unique Gα-binding and -activating (GBA) motif conferring guanine nucleotide exchange factor (GEF) activity on Gαi subunits to activate trimeric G proteins [18, 19]. In response to receptor tyrosine kinases and GPCRs stimulation, GIV-mediated G protein activation dissociates Gβγ subunits, which enhances the PI3K-Akt signaling [20]. The role of GIV in transmitting receptor signaling to promote the PI3K-Akt signaling pathway has been established in distinct cellular behaviors in various physiological and pathological conditions, such as liver fibrosis [21], tumor metastasis [22, 23], and kidney podocytes early injury [24]. However, the expression of GIV in pancreatic islets and its potential role in insulin secretion have yet to be investigated. In this study, we hypothesize that GIV may be a critical signaling modulator of Nephrin phosphorylation and endocytosis in response to extracellular high glucose concentration to regulate glucose-stimulated insulin secretion. For the first time, our work verifies the expression and subcellular localization of GIV at the cell–cell junctions of the mouse islet and islet β cells. We demonstrate that GIV can combine with Nephrin and recruit Src kinase to induce tyrosine phosphorylation of both proteins to transmit PI3K/Akt signaling. Consequently, the GIV/Nephrin/*Akt axis* facilitates the high glucose–induced endocytosis of Nephrin and the F-actin reorganization to promote insulin secretion. Finally, the levels of GIV and Nephrin are downregulated in diabetic mice, and the adenovirus-mediated overexpression of GIV could rescue the function of β cells in glucose-stimulated insulin secretion. Our study reveals the role of GIV in insulin secretion and identifies novel signaling machinery regulating glucose-stimulated insulin secretion, which may be a potential target for the therapy of diabetic mellitus. ## Abstract Glucose-stimulated insulin secretion of pancreatic β cells is essential in maintaining glucose homeostasis. Recent evidence suggests that the Nephrin-mediated intercellular junction between β cells is implicated in the regulation of insulin secretion. However, the underlying mechanisms are only partially characterized. Herein we report that GIV is a signaling mediator coordinating glucose-stimulated Nephrin phosphorylation and endocytosis with insulin secretion. We demonstrate that GIV is expressed in mouse islets and cultured β cells. The loss of function study suggests that GIV is essential for the second phase of glucose-stimulated insulin secretion. Next, we demonstrate that GIV mediates the high glucose-stimulated tyrosine phosphorylation of GIV and Nephrin by recruiting Src kinase, which leads to the endocytosis of Nephrin. Subsequently, the glucose-induced GIV/Nephrin/Src signaling events trigger downstream Akt phosphorylation, which activates Rac1-mediated cytoskeleton reorganization, allowing insulin secretory granules to access the plasma membrane for the second-phase secretion. Finally, we found that GIV is downregulated in the islets isolated from diabetic mice, and rescue of GIV ameliorates the β-cell dysfunction to restore the glucose-stimulated insulin secretion. We conclude that the GIV/Nephrin/Akt signaling axis is vital to regulate glucose-stimulated insulin secretion. This mechanism might be further targeted for therapeutic intervention of diabetic mellitus. ## GIV is expressed in pancreatic β cells We initially explored the tissue distribution of GIV. The results reveal that both mRNA and protein of GIV were widely expressed in many organs. Specifically, GIV was highly expressed in the brain and moderately in other tissues investigated, including endocrine tissues such as the adrenal gland, pancreas, and pituitary gland (Fig. S1, A and B). We investigated GIV expression in a normal murine pancreas. Immunostaining revealed that GIV was produced in both α and β cells in the islets of control mice (Fig. 1, A and B). GIV expression in murine islets, murine insulinoma cells, and rat insulinoma cells was confirmed by Western blot (Fig. 1C) and quantitative PCR (Fig. S1B). We then investigated the intracellular localization of GIV in β cells. Double-immunostaining analyses indicated that GIV was partially colocalized with insulin granules in murine β cells (Fig. 1D). Furthermore, we found that GIV was colocalized with F-actin close to the cell periphery (Fig. 1E), similar to the distribution of GIV in cos7 cells. These results indicate that GIV was expressed in pancreatic β cells, primarily localized on the plasma membrane. Figure 1GIV is expressed in pancreatic islets and β cells. A and B, immunofluorescence confocal images for GIV and insulin (A) or GIV and glucagon (B) in mouse pancreatic slices were shown. The scale bar represents 50 μm. C, Western blotting analysis for GIV and Nephrin in mouse islets and β cells. D and E, immunofluorescence staining of GIV and insulin (D) or GIV and Phalloidin (E) in MIN6 cells. The scale bar represents 5 μm. All the experiments were repeated at least three times. ## GIV/Girdin regulates glucose-stimulated insulin secretion in MIN6 cells and murine islets To further characterize the role of GIV in β-cell function, MIN6 cells were infected with lentivirus coding short hairpin RNA (shRNA) of GIV for downregulating the endogenous expression level of GIV, and about $66.2\%$ of GIV was reduced after knockdown (Fig. 2A). The depletion of GIV expression does not affect the total insulin content (Fig. 2B). Glucose-stimulated insulin secretion (GSIS) assay showed that GIV depletion profoundly decreased GSIS of MIN6 cells (Fig. 2C). Interestingly, downregulation of GIV did not affect KCl-stimulated insulin secretion (Fig. 2D). To overcome the limitation of using insulinoma cells to investigate the role of GIV in insulin secretion, we applied GIV shRNA downregulation to dissociated murine islet cells. About $80.5\%$ of GIV in shRNA-treated murine islets was robustly reduced (Fig. 2E). We confirmed that knockdown of GIV could decrease GSIS in murine islets (Fig. 2F). To directly examine the insulin secretion ability, we performed perifusion analyses in isolated islets, as described [25]. GIV shRNA affected glucose-induced insulin secretion in perifused islet cells compared with nontargeting shRNA (Fig. 2G). The area under the curve shows that only the second phase (6–30 min) of insulin secretion was markedly reduced in GIV-diminished islets (Fig. 2H). These results indicate that GIV is crucial for GSIS in MIN6 cells and mouse islets, especially for the second phase. Figure 2GIV regulated glucose-induced insulin secretion in MIN6 cells and mouse islets. MIN6 cells were infected with shRNA for 2 days. A, GIV protein levels were checked by immunoblotting ($$n = 3$$). B–D, the effect of GIV shRNA on total insulin content (B), glucose-induced insulin secretion (C), and KCl-induced insulin secretion in MIN6 cells (D) ($$n = 6$$). Islets from C57BL/6J mice (12–16 weeks old) were infected with shRNA for 2 days. E, GIV protein levels were checked by immunoblotting ($$n = 3$$). F, the effect of GIV shRNA on glucose-induced insulin secretion in mouse islets ($$n = 6$$). G, shRNA-treated mouse islets were perifused with Krebs–Ringer bicarbonate buffer containing 16.7 mM glucose for 30 min, followed by the standard buffer (2.8 mM glucose) for 10 min. Insulin secretion was normalized to the total insulin content. H, first phase (0–7 min), second phase (8–30 min), and total insulin secretion were calculated as the area under the curve (AUC) ($$n = 4$$). The statistical significance of differences between means was assessed by the Student’s t test. ∗∗$p \leq 0.01$; ∗∗∗$p \leq 0.001$; n.s. means not significant. ## GIV is required for glucose-induced phosphorylation of Nephrin and Akt To identify the molecular mechanism by which GIV regulates GSIS, we initially checked the subcellular expression of GIV in β cells. We isolated the cytosolic and crude membrane fractions, respectively. The results depict that GIV was almost located in the membrane fraction but not in the cytosolic fraction, similar to Nephrin, a well-known membrane protein in MIN6 cells (Fig. 3A). The colocalization of GIV and Nephrin shows that they might bind with each other to form a protein complex. To confirm this hypothesis, we explored the interaction between GIV and Nephrin by immunoprecipitation. The result showed that GIV could bind with Nephrin in MIN6 cells (Fig. 3B) and mouse pancreatic islets (Fig. S2A). In addition, high glucose had a negligible effect on the interaction between GIV and Nephrin compared with that in the control condition with low glucose (Fig. S2B), supporting that GIV constitutively binds Nephrin. Assuming that *Nephrin is* involved in insulin secretion caused by high glucose [13, 26], our data suggest that GIV and Nephrin might form a signaling complex to elicit intracellular downstream signaling to regulate insulin secretion. We further examined the effects of high glucose stimulation on the activation of GIV, Nephrin, and the downstream Akt protein by Western blotting. The results demonstrated that high glucose significantly increased the phosphorylation of GIV (Y1767) and Nephrin (Y1176/Y1193) in both MIN6 and INS1 cells. The phosphorylation of GIV and Nephrin has been reported to activate downstream Akt signaling [16]. We next assessed the role of Y1767 phosphorylation of GIV in the glucose-triggered signaling transduction by examining the changes in the levels of phosphorylated Akt and GIV in MIN6 and INS1 cells following glucose elevation. MIN6 and INS1 cells were treated with a Krebs–Ringer bicarbonate buffer containing 2.8 mM glucose (low glucose) or 16.7 mM glucose (high glucose) for 30 min. High-glucose treatment significantly increased the phosphorylated level of GIV-Y1767 (Figs. 3C and S2C). A significant increase in the level of Nephrin phosphorylation and Akt S473 phosphorylation was also evident after high-glucose treatment (Figs 3C and S2C). To investigate the molecular relationship between GIV and Nephrin, we then examined the expression level of phosphorylated Nephrin using GIV-shRNA-infected MIN6 cells. Interestingly, we found that both phosphorylated Nephrin and Akt were remarkably reduced in GIV-depleted MIN6 cells (Figs 3D and S2D). This result indicates that the tyrosine phosphorylation of GIV at Y1767 is essential for glucose-induced Nephrin and Akt phosphorylation. Glucose-induced phosphorylation of GIV at Y1767 was not changed by pretreatment with PI3K inhibitor LY294002 (Figs. 3, E and F and S2E), indicating that increases in GIV-Y1767 phosphorylation were not through the PI3K/Akt pathway in MIN6 cells, the effect of LY294002 was identified by unchanged expression of Akt phosphorylation after glucose stimulation (Figs. 3G and S2E). Interestingly, the insulin stimulation in a low-glucose condition significantly enhanced the Y1767 phosphorylation of GIV as the high glucose (Fig. S3, A–C), illustrating that the glucose-induced insulin may further serve as positive feedback stimuli to contribute the GIV phosphorylation event for signal amplification. Taken together, our results indicate that GIV-Y1767 phosphorylation is an upstream event of the PI3K/Akt pathway and positively correlates with the activation of Nephrin phosphorylation and Akt phosphorylation. Figure 3GIV is required for glucose-induced phosphorylation of Nephrin and Akt. A, subcellular location of GIV in MIN6 cells. The cytosolic and crude membrane fractions were isolated by ultracentrifugation. Postnucleation supernatant (PNS) indicates the total expression of proteins, β-actin and sodium and potassium pump (Na-K-ATPase) as the marker of cytosolic and crude membrane fractions, respectively. The experiments were repeated three times. B, total islet protein lysates (300 mg) from wildtype mice underwent immunoprecipitation with anti-GIV antibody or control IgG. The immunoprecipitates, as well as 1:20 of the original lysates, were immunoblotted with anti-GIV and anti-Nephrin antibodies ($$n = 3$$). C, MIN6 and INS1 cells were stimulated with 2.8 mM or 16.7 mM glucose for 30 min, and protein expression levels were analyzed by immunoblotting with antibodies toward the indicated proteins ($$n = 3$$). D, effects of insulin on GIV phosphorylation in MIN6 cells. MIN6 cells were stimulated with 2.8 mM or 16.7 mM glucose for 30 min in the presence or absence of insulin; 16.7 mM glucose treated was as a positive control of GIV phosphorylation ($$n = 3$$). E, effects of GIV on glucose-induced GIV, Nephrin, and Akt phosphorylation in MIN6 cells. MIN6 cells were treated with or without GIV shRNA or control shRNA lentivirus for 2 days, then were stimulated with 2.8 mM or 16.7 mM glucose for 30 min ($$n = 3$$ in each group). F, effects of PI3K inhibitors on glucose-induced GIV, Nephrin, and Akt phosphorylation in MIN6 cells. MIN6 cells were stimulated with 2.8 mM or 16.7 mM glucose for 30 min in the presence of dimethyl sulfoxide (DMSO) ($0.05\%$) or Ly294002 (0.5 μM or 5 μM); DMSO was treated as a control of the inhibitor (shown as 0 μM) ($$n = 3$$). F and G, the expression levels of phosphorylated proteins normalized by their total protein levels were measured by densitometry and the total protein levels were normalized by those of α-tubulin. The statistical significance of differences between means was assessed by one-way ANOVA with a Tukey’s test. ∗$p \leq 0.05$; ∗∗$p \leq 0.01$ versus 2.8 mM glucose, ##$p \leq 0.01$ versus 16.7 mM glucose with DMSO in each group). ## GIV phosphorylation (Y1767) mediates phosphorylation of Nephrin via recruitment of Src kinase Several members of the Src family kinases could induce Nephrin phosphorylation; Fyn was consistently found to be coimmunoprecipitated with Nephrin [27]. A previous study has revealed that GIV lacks kinase activity and that tyrosine phosphorylation of GIV at Y1767 by the nonreceptor tyrosine kinase Src promotes activation of PI3K signaling during cell migration [28]. Accordingly, we assumed Nephrin might interact with Src family kinases through GIV. To investigate what type of Src kinase family was recruited by GIV, Nephrin was immunoprecipitated from MIN6 cells treated with Ctrl-shRNA and GIV-shRNA. We found that, in GIV knockdown MIN6 cells, the expression of the indicated proteins in the cell did not change (Figs. 4A and S4A). However, Src kinase, but not Fyn and Lck, as well as Lyn kinases, was coimmunoprecipitated with Nephrin in control MIN6 cells (Fig. 4B). Interestingly, neither GIV nor Src was coimmunoprecipitated with Nephrin in GIV-deficient MIN6 cells, suggesting that Nephrin interacted with Src kinase through GIV (Fig. 4B). These findings indicate that, under normal conditions, GIV was used as a vector to recruit Src kinase interacting with Nephrin to form a complex and induced phosphorylation of Nephrin. Figure 4GIV-Y1767 phosphorylation–mediated phosphorylation of Nephrin through recruitment of Src family kinase. A and B, total islet protein lysates (1 mg) from MIN6 cells underwent immunoprecipitation with anti-Nephrin antibody or control IgG. The immunoprecipitates (B), as well as 1:50 of the original lysates (A), were immunoblotted with indicated antibodies ($$n = 3$$). C, effects of Src family kinase on glucose-induced GIV, Nephrin, and Akt phosphorylation in MIN6 cells. MIN6 cells were stimulated with 2.8 mM or 16.7 mM glucose for 30 min in the presence of dimethyl sulfoxide (DMSO) ($0.1\%$) or Saracatinib (10 μM). DMSO was treated as a control of Saracatinib (shown as 0 μM) ($$n = 3$$). D and E, MIN6 cells (D) or mouse islets (E) were stimulated with 2.8 mM or 16.7 mM glucose for 60 min in the presence of DMSO ($0.1\%$) or Saracatinib (10 μM), and the amount of insulin secretion was normalized to the total insulin content followed with being normalized with the amount of insulin secretion of standard condition ($$n = 6$$). The statistical significance of differences between means was assessed by the Student’s t test. ∗$p \leq 0.05$; ∗∗∗$p \leq 0.001.$ F, MIN6 cells treated with lentivirus encoding control shRNA or shRNA against GIV were infected with adenovirus encoding control LacZ, or wildtype or GIV-Y1767A mutant. The cell extracts were immunoblotted with anti-GIV and anti-a-tubulin antibodies ($$n = 3$$). G, MIN6 cells were subjected to glucose-stimulated insulin secretion assays as in Figure 2C ($$n = 3$$). The statistical significance of differences between means was assessed by one-way ANOVA with a Tukey’s test. ∗∗∗$p \leq 0.001$; n.s. means not significant. To further validate the role of Src in the phosphorylation of GIV and Nephrin, the effects of saracatinib, a selective Src family tyrosine kinase inhibitor [29], were examined in MIN6 cells. MIN6 cells were preincubated with vehicle (dimethyl sulfoxide) or saracatinib for 1 h before stimulation with low and high glucose. High glucose increased phosphorylation of GIV and Nephrin in addition to Akt in vehicle-treated MIN6 cells (Figs. 4C and S4B). Furthermore, phosphorylation of Src was also induced after high-glucose stimulation (Figs. 4C and S4B). However, such glucose-induced phosphorylation vanished in saracatinib-treated MIN6 cells (Figs. 4C and S4B). In addition, the glucose-induced high levels of total GIV and Nephrin were also abolished after adding Src kinase inhibitor (Fig. S4C), which indicated that such increment was due to a direct effect of increased phosphorylation of GIV and Nephrin. Concerning that Src kinase inhibitor erased the signaling enhancement, these findings suggest that Src kinase initially activated GIV-Y1767 phosphorylation after high glucose stimulation, subsequently promoting the GIV/Nephrin/PI3K/Akt pathway. To further evaluate the effect of Src on insulin secretion in β cells, MIN6 cells or islet cells were stimulated with low or high glucose with preincubation of saracatinib or vehicle. Saracatinib-treated MIN6 cells demonstrated significantly decreased insulin secretion compared with vehicle-treated MIN6 cells (Fig. 4D). Similarly, saracatinib-treated islet cells also exhibited markedly reduced insulin secretion compared with vehicle-treated islet cells (Fig. 4E). Since Src kinase initially activated the Y1767 phosphorylation of GIV after high-glucose stimulation, we speculate that the phosphomodification of this tyrosine site of GIV was essential for glucose-stimulated insulin secretion. To substantiate this speculation, we performed function rescue experiments to analyze the effect of Y1767 phosphorylation of GIV on the exocytosis of insulin granules in MIN6 cells. When the wildtype or GIV-Y1767A mutant was expressed in GIV-knockdown cells as the endogenous level of GIV (Fig. 4F), only the wildtype rescued the decreased insulin secretion (Fig. 4G), supporting the importance of Y1767 phosphorylation of GIV on insulin exocytosis. ## GIV facilitates the endocytosis of Nephrin upon high-glucose stimulation Numerous reports reveal that Nephrin's subcellular distribution is altered by glucose concentration, as *Nephrin is* localized to the plasma membrane under low-glucose conditions and translocated to the cytoplasm after high-glucose stimulation [13, 26]. High glucose-induced Nephrin phosphorylation accumulates more endocytosis-related proteins, such as Dynamin [11], to execute subsequent endocytosis. Since GIV regulated glucose-induced Nephrin phosphorylation, we speculate that GIV mediates the endocytosis of Nephrin after high-glucose stimulation in β cells. To further verify this possibility, we measured the expression level of Nephrin at the plasma membrane in murine β cells using a cell-surface biotinylation assay. Nephrin was detectable in plasma membrane fractions from MIN6 cells cultured in low glucose, but its intensity was markedly reduced in high glucose (Fig. 5, A and C). In contrast, the total level of Nephrin remained changed significantly (Fig. 5, A and D), suggesting that the decrease in the surface level of Nephrin could be attributed to endocytosis on the plasma membrane after high-glucose stimulation rather than the reduction in Nephrin synthesis, which is consistent with the previous studies [12, 14]. To exclude the possibility of high-glucose-induced nonspecific cell surface protein endocytosis, we examined the expression level of Na-K-ATP, a cell surface protein. We found that its expression level was unaltered after high-glucose stimulation (Fig. 5, A and E). However, when analyzing the expression of Nephrin at the plasma membrane in GIV knockdown MIN6 cells (Fig. 5, A and B), Nephrin expression was explicitly present in both low and high glucose, and the high-glucose-induced *Nephrin endocytosis* was completely abolished (Fig. 5, A and C). These findings suggested that GIV mediated high-glucose-induced *Nephrin endocytosis* in murine β cells. Figure 5GIV modulated F-actin remodeling through mediating endocytosis of Nephrin after high-glucose stimulation. A, subcellular expression of Nephrin on the plasma membrane in glucose-induced GIV downregulated MIN6 cells was checked by cell-surface biotinylation assay. B–E, the protein levels normalized by those of β-actin were measured by densitometry ($$n = 3$$). F, MIN6 cells were infected with shRNA for 48 h, and 2.8 mM glucose and 16.7 mM glucose were added for 30 min before cells were lysed. Cell lysates were subjected to GST-PAK1-PBD (GTP bound [active form] Rac1 interaction binding) or GST-Rhotekin-RBD (GTP bound [active form] RhoA interaction binding) pull-down followed by immunoblotting with anti-Rac1 and anti-RhoA. Normalized to total Rac1 and RhoA. G–J, the expression levels of active Rac1 and active RhoA normalized by their total protein levels were measured by densitometry, and the total protein levels of Rac1 and RhoA were normalized by those of β-actin ($$n = 3$$). The statistical significance of differences between means was assessed by the Student’s t test. ∗$p \leq 0.05$; ∗∗$p \leq 0.01$; ∗∗∗$p \leq 0.001$; n.s. means not significant. Nephrin endocytosis regulated PI3K/Akt-mediated actin reorganization through increasing Rac1 activity in podocytes [10]. In addition, it has been reported that GIV reorganized the actin cytoskeleton in cancer cells [16]. However, how GIV/Nephrin pathway regulates actin remodeling under elevated glucose in β cells remains unclear. Glucose-mediated actin remodeling in β cells highly depends on several actin-modulatory proteins, including the Rac1 and RhoA [5]. Rac1 directly stimulates actin depolymerization, and β cell–specific deletion of Rac1 inhibits F-actin disassembly and impairs glucose tolerance and GSIS in mice [30]. By contrast, RhoA regulates actin polymerization. The RhoA/ROCK pathway is responsible for the stabilization of actin cytoskeleton and inhibition of insulin secretion under physiological conditions [31]. To further validate the role of GIV in regulating cytoskeletal remodeling, we performed low-glucose and high-glucose treatments on β cells infected with control and GIV shRNA. Using GST-PAKI-PBD pull-down assay and GST-Rhotekin-RBD pull-down assay, we observed that high-glucose treatment significantly increased GTP-bound Rac1 (active Rac1) levels and decreased GTP-bound RhoA (active RhoA) levels in β cells (Fig. 5, F–H). However, after the downregulation of GIV, elevated glucose treatment could not activate GTP-bound Rac1 levels, while the GTP-bound RhoA levels were maintained at a low glucose level (Fig. 5, F–H). Since the total levels of Rac1 and RhoA were unchanged (Fig. 5, F, I and J), this result suggests that GIV is required for glucose-induced Rac1 activation and inhibition of RhoA activity. These results indicate that GIV is involved in regulating the transformation of RhoA activity into Rac1 activity in β cells, thus mediating the remodeling of the cytoskeleton. ## GIV mediates actin cytoskeleton remodeling in insulin granule exocytosis We employed ultrastructural analyses to further probe the requirement for GIV activity in insulin secretion. Electron microscopy revealed an increase in insulin granule localization at the β cell plasma membrane, in response to the high-glucose challenge, compared with the low-glucose state in normal murine islets (Fig. 6, A, C and E). However, GIV knockdown islets exhibited suppressed glucose-induced recruitment of insulin granules to the cell periphery but did not affect insulin docking under basal conditions (Fig. 6, B, D and E). GIV knockdown β cells had fewer docked insulin granules (quantified as granules within 200 nm of the plasma membrane) in the presence of high glucose. Since insulin content was unaffected by GIV-shRNA treatment (see Fig. 2A), the reduction in surface-localized insulin granules is not due to the defects in insulin biogenesis. These results suggest that GIV activity is crucial for glucose-dependent insulin granule positioning at the β-cell surface. Figure 6GIV signaling promotes insulin granule exocytosis via F-Actin remodeling. A–D, electron micrographs of β cells from GIV knockdown islets with 2.8 mM low-glucose- or 16.7 mM high-glucose-containing Krebs–Ringer bicarbonate buffer incubation. Insulin granules whose centers were located within 200 nm of the plasma membrane were categorized as docked granules and were indicated by red arrowheads. The scale bar represents 1 μm. E, the number of docked granules was calculated from 18 individual β cells by Student’s t test. ∗$p \leq 0.05$; n.s. means not significant. F–I, shRNA-treated mouse pancreatic β cells were incubated with 2.8 mM glucose and 16.7 mM glucose for 30 min, then cells were fixed and stained with phalloidin. The scale bar represents 20 μm. J, F-actin patterns in MIN6 cells were quantified as the fluorescent intensity of phalloidin by Student’s t test ($$n = 17$$). ∗∗∗$p \leq 0.001.$ Actin cytoskeleton rearrangement is a critical acute determinant of GSIS [5]. Actin microfilaments are organized as a dense meshwork beneath the β-cell plasma membrane that restricts insulin granule access to the docking and fusion machinery [32]. Glucose stimulation rapidly promotes filamentous actin (F-actin) remodeling to mobilize insulin granules to the cell periphery. However, how glucose stimulation regulates actin reorganization remains unclear. Therefore, we intended to explore whether GIV signaling promotes actin rearrangements in β cells, a prerequisite step in insulin granule positioning at the plasma membrane. Isolated Ctrl-shRNA and GIV-shRNA-treated β cells were stimulated with either low or high glucose, and F-actin was visualized using Alexa 546–labeled phalloidin. Under basal conditions, both Ctrl-shRNA and GIV-shRNA-treated β cells demonstrated a thick F-actin pattern (Fig. 6, F and G). However, high-glucose treatment elicited a striking reduction in F-actin (Fig. 6, H and J), consistent with previous reports. By contrast, the glucose-induced dissolution of F-actin was prevented by GIV-shRNA-mediated silencing of GIV signaling (Fig. 6, I and J). Therefore, GIV activity is required for glucose-triggered actin reorganization in β cells. ## Rescue of GIV restores the impaired glucose-stimulated insulin secretion from db/db islets We studied GIV expression in murine islets isolated from db/db (type 2 diabetes model) and control mice. The protein expression level of GIV in db/db islets was remarkably lower than in the control islets (Fig. 7, A and B), which indicated that the reduction of GIV is associated with the glucolipotoxicity of the islet. To understand whether high glucose regulated GIV expression as observed in the diabetic animal model, we studied the GIV expression of wildtype islets after chronic exposure to glucose. The result showed that chronic exposure to 20 mmol/l glucose significantly decreased the mRNA level of GIV of murine islet cells (Fig. 7C), consistent with the data from db/db islets. Next, we performed rescue experiments by adenovirus-mediated expression of wildtype and GIV-Y1767A mutant in db/db islets to match the endogenous level of GIV protein in control islets (Fig. 7, D and E). The GSIS results revealed that the db/db islets exhibited markedly reduced insulin secretion than the wildtype islets, whereas wildtype GIV significantly restored GSIS to the level found in wildtype islets expressing the control LacZ protein. In contrast, the GIV-Y1767A mutant could not rescue the impaired GSIS in db/db islets (Fig. 7F). These findings demonstrate the potent role of GIV in restoring impaired β-cell function in diabetic islets. Figure 7GIV restored impaired glucose-induced insulin secretion. A, protein expression level of GIV and Nephrin in islets from type 2 diabetic (db/db) mice and nondiabetic (control) mice. B, the band intensity of each protein from db/db islets was normalized by those of control islets ($$n = 3$$). C, MIN6 cells “starved” in 2.8 mM glucose for 7 days, followed by incubation with either 2.8 mM low glucose (LG) or 25 mM high glucose (HG) for 10 days. Relative quantification (RQ) of GIV mRNA expression was detected by quantitative PCR ($$n = 4$$). D, the control and db/db islets were infected with adenoviruses encoding control LacZ, wildtype, or GIV-Y1767A mutant. After a 1-h infection, the islets were rinsed and incubated for 48 h at 37 °C. The protein level of GIV in db/db islets was adjusted to that of endogenous GIV in control islets by immunoblotting with the anti-GIV antibody. E, the protein levels of GIV normalized by α-tubulin were measured by densitometry. The dashed line indicated the endogenous expression level of GIV in control islets ($$n = 3$$). F, the control and db/db islets were infected with adenoviruses with the condition described in D. The islets were preincubated in 2.8 mM glucose-containing Krebs–Ringer bicarbonate buffer for 1 h and were incubated in LG or HG buffer for 1 h. Insulin secretion was shown as the ratio of HG and LG ($$n = 6$$). The statistical significance of differences between means was assessed by the Student’s t test for B and C and by one-way ANOVA with a Tukey’s test for E and F. ∗$p \leq 0.05$, ∗∗$p \leq 0.01$; ∗∗∗$p \leq 0.001$; n.s. means not significant. ## Discussion This study unveiled that GIV affects the function of pancreatic β cells and promotes insulin secretion in response to glucose. This function of GIV is independent of its well-known and described role in kidney podocytes, where it regulates the structure and function of the glomerular slit diaphragm [24]. This study, for the first time, demonstrates that GIV not only is a master regulator of cell–cell adhesion in podocytes but also affects insulin signaling in islet β cells by the regulatory role of GIV regulating Nephrin phosphorylation and endocytosis. Generally, *Nephrin is* distributed on the plasma membrane of islet cells. It anchors stress fibers to form a mesh-like barrier that prevents insulin vesicles from contacting and fusing with the plasma membrane [13]. Previous studies have revealed that Nephrin phosphorylation and endocytosis are core signaling events governing GSIS [11, 14]. Nevertheless, the detailed mechanism of the glucose-induced *Nephrin endocytosis* for insulin secretion remains unclear. In this study, we found that GIV functions as a binding partner of Nephrin to colocalize on the plasma membrane in β cells. Upon high-glucose stimulation, GIV is phosphorylated at Y1767 by the activated Src kinase. As an intermediator, the phosphorylated GIV recruited Src kinase to phosphorylate Nephrin since both phosphorylation and Src binding of Nephrin was disrupted after the reduction of GIV. We also found glucose-induced Y1767 phosphorylation of GIV, followed by PI3K/Akt signaling stimulation. Our rescue experiment data validated that deficient phosphorylation of GIV at Y1767 could also effectively inhibit insulin secretion of MIN6 cells and in islets (Fig. 4G). Furthermore, we found that Src kinase, but not Fyn and Lck, as well as Lyn kinase, is required for interactions of GIV with Nephrin in murine islet β cells. Accordingly, we propose the following scenario demonstrating the role of phosphorylation of GIV at Y1767 during glucose-stimulated insulin secretion, and the recruitment of Src to interact with *Nephrin is* essential for *Nephrin endocytosis* and phosphorylation. Subsequently, the activated GIV functions in triggering the PI3K-dependent Akt phosphorylation, which facilitates the GSIS by activating Rac1 and RhoA to reorganize the actin cytoskeleton in islet β cells (Fig. 8).Figure 8GIV acutely promotes Nephrin-dependent glucose-stimulated insulin secretion via actin cytoskeleton reorganization in β cells. [ 1] Src kinase induced GIV-Y1767 phosphorylation in response to elevated glucose. [ 2] GIV-Y1767 phosphorylation recruited Src kinase to Nephrin for phosphorylating Nephrin. [ 3] Nephrin phosphorylation induced *Nephrin endocytosis* pathway. [ 4] Cytoplasm Nephrin from endocytosis and PI3K/Akt signaling regulated activities of actin-modulatory protein Rac1 and RhoA, to [5] permeabilize a peripheral F-actin barrier, and [6] promoted insulin granule exocytosis. In this proposed model, we argue that GIV-mediated phosphorylation of Nephrin causes the aggregation of the regulator of endocytosis, such as Dynamin, to regulate its endocytosis [11]. Consequently, the endocytosis of Nephrin activates the function of the essential regulator of the cytoskeleton, including the Rac1 and RhoA, which reduce the stress fibers that act as a barrier by actin remodeling [30, 31]. Therefore, insulin vesicles can fuse with the cell membrane and then release insulin [3]. We also found that the Src kinase inhibitor significantly retarded high-glucose-induced phosphorylation of the Nephrin–GIV complex, blocking *Nephrin endocytosis* and insulin release. The role of GIV in recruiting the Src kinase to bind and phosphorylate Nephrin/GIV complexes has also been identified using the loss of function experiment by the knockdown of GIV in pancreatic islet β cells. Therefore, we propose that GIV mediates the recruitment of nonreceptor kinase Src to amplify the tyrosine phosphorylation of Neprhin, which plays a role in the effect of sensitive glucose concentration. Recently, GIV has been found to be localized on E-cadherin-mediated cell–cell junctions, contributing to Wnt/β-catenin activation to reorganize the cytoskeleton and migration of mammalian cells [33]. On the other hand, the multifunctional GIV protein has been well recognized as an intracellular signaling mediator to propagate various signals across the plasma membrane [17]. In this study, we manipulated GIV expression in insulinoma MIN6 cells and dissociated murine islets and established the novel function of GIV in GSIS. Upon insulin stimulation or high-glucose condition, the total level and Y1767-phosphorylation of GIV were significantly increased. Our results demonstrate that the increased Y1767-phosphorylation of GIV positively correlates with the Src kinase activation, as shown that the Src kinase inhibitor, saracatinib, could effectively abolish the GIV phosphorylation (Fig. 4C). Using the adenovirus containing wildtype GIV and its mutant of Y1767A, we verified the essential role of Y1767 phosphorylation of GIV in GSIS via the rescue experiments in both GIV-depleted MIN6 cells and the islets from type 2 diabetic db/db mice (Figs 4G and 7F). Notably, GIV is known to act as a nonreceptor GEF to modulate trimeric G protein activity to affect Akt signaling [19, 20], which might also contribute to the regulatory machinery of GSIS. However, this possible mechanism remains to be explored in the future study, and it would be highly intriguing to further investigate whether GIV-mediated G protein activation plays a role in GSIS and how Y1767-phosphorylation structurally affects the GEF function of GIV for signal propagation. Moreover, the increased expression level of GIV may be the consequent events through a feedback mechanism upon high glucose, which led to the dissociation of the membrane-associated Nephrin from connected β cells to initiate endocytosis, thereby increasing the fraction of the Triton X-100 soluble complex of GIV/Nephrin in the lysates (Fig. 3C). Our research reveals that F-actin remodeling is induced upon the GIV-mediated endocytosis of Nephrin after glucose load in β cells. The role of GIV in cytoskeleton regulation has been reported that it can directly bind to actin filaments and stabilize F-actin [15]. Upon shRNA-mediated knockdown of GIV, stress fibers were disrupted to alter the cell morphology, suggesting the role of GIV in cytoskeleton-mediated cellular processes, such as insulin secretion. Our results showed that downregulation of GIV strongly impaired the F-actin depolymerization after glucose stimulation, thereby inhibiting subsequent insulin granule exocytosis in β cells. In comparison, GIV depletion did not affect the KCl-stimulated insulin secretion in MIN6 cells (Fig. 2D). Combining the electron microscopic observation, our work validates the function of GIV in properly localizing the insulin granules on the β-cell surface via cytoskeleton remodeling in the second phase of GSIS, however, not the granule fusion with the plasma membrane in the first phase of insulin secretion. Another interesting finding of this study is that GIV expression is decreased in islets from diabetic (db/db) mice compared with age-matched control (db/m) mice, which was different from what was previously reported in the kidney [24]. Consistent with these ideas, prolonged exposure to high glucose led to a downregulation of GIV expression in MIN6 cells, while acute exposure to high glucose increased GIV mRNA expression. Although the relevance of an in vitro model of acute and prolonged glucose exposure in insulinoma cells remains to be established, the opposite effect of chronic glucose exposure on GIV gene expression is contrary to what has been described in kidney cells [24], where chronic glucose exposure has been widely accepted as a model of glucotoxicity [13]. One possible interpretation is that GIV controls the endocytosis of Nephrin to facilitate insulin secretion, which may be impaired in diabetes because of the downregulation of GIV in pancreatic β cells. Further investigation would be highly desirable to define the precise mechanism whereby GIV augments insulin secretion. In conclusion, we have established that GIV is an important regulator of glucose-stimulated insulin secretion in pancreatic β cells. Our proposed working model suggests that GIV mediates the high glucose sensing of Nephrin and insulin secretion via the recruiting of Src kinase to trigger the tyrosine phosphorylation of Nephrin and GIV. The tyrosine phosphorylation of Nephrin initiates the Nephrin endocytosis, thereby affecting subsequent insulin exocytosis by PI3K/Akt-mediated actin remodeling. Therefore, this study identifies a novel signaling machinery regulating glucose-stimulated insulin secretion, which might be a potential target for the therapy of diabetic mellitus. Selective intervention to restore GIV expression or function may delay the need for insulin therapy in type 2 diabetes. Further exploration is crucial to characterize the role of GIV in regulating blood glucose homeostasis for a long-term study in vivo. ## Cell culture and islet culture All cells were cultured in a humidified incubator with $95\%$ air and $5\%$ CO2 at 37 °C. MIN6 cells [34] were cultured in Dulbecco's modified Eagle's medium containing $15\%$ fetal bovine serum supplemented with 50 μM 2-mercaptoethanol. INS1 $\frac{832}{13}$ cells [35] were cultured in RPMI1640 containing $10\%$ fetal bovine serum supplemented with 1 mM L-glutamine, 1 mM Hepes, 1 mM sodium pyruvate, and 50 μM 2-mercaptoethanol. HEK293A cells (Invitrogen) were cultured in Dulbecco’s modified Eagle’s medium containing $10\%$ fetal bovine serum supplemented with 1 mM L-glutamine. Pancreatic islets were isolated from mice sacrificed by cervical dislocation through injection of 500 units/ml collagenase solution (type XI; Sigma) into the pancreatic duct, followed by mild shaking digestion at 37 °C for 20 min, and isolated islets were manually selected under a dissecting microscope, as described elsewhere. Isolated islets were cultured overnight in RPMI 1640 medium containing $10\%$ fetal bovine serum, 100 units/ml penicillin-streptomycin. After overnight recovery, islets were treated with $0.05\%$ trypsin in PBS at 37 °C for 5 min, fully dispersed to primary β cells. The C57BL/6J mice, type 2 diabetic db/db mice, and nondiabetic control mice were purchased from Jiangsu Gempharmatech Laboratories. The mice were kept at constant temperature and humidity, with a 12-h light and dark cycle, and fed a regular unrestricted diet. Only male mice were used in our experiments. All animal experiments were performed in accordance with the national ethical guidelines implemented by our Institutional Animal Care and Use Committee and approved by the Ethical Review Committee of the Institute of Zoology, Capital Medical University, China. ## Construction of the lentivirus and adenovirus vectors Short hairpin RNA (shRNA) with nontargeting (CCTAAGGTTAAGTCGCCCTCG) or mouse Girdin targeting CDS region (CAGTCGATTCATCACCACCTA) was cloned into the PLKO.1 puro vector (Invitrogen). For the downregulation of GIV, MIN6 cells and islets were infected with GIV shRNA lentivirus. After 48-h infection, the cells or the islets were used for protein and functional assay. For the rescue experiment, shRNA with mouse Girdin targeting 3′-UTR sequence (GCAACTATAGGAACTATTAAA) was cloned into the PLKO.1 puro vector (Invitrogen). Full-length mouse GIV cDNA was cloned from MIN6 cells. Wildtype GIV was performed using the following primers: F1 5′-GGTCGACTCTAGAGGATCCAAGCCACCATGGAGAACGAAATCTTCACTCCCC-3′ and R1 5′-AATGACACGGTCTGCCTCTGC-3′ as well as F2 5′- CTCAGATTCTTGCACTGCAGAGG-3′ and R2 5′-TCCTTGTAGTCCATACCGGTAATACAACCATATTCATACCAAACAATTTGGCT-3′; site-directed mutagenesis of GIV-Y1767A was generated from wildtype of GIV as a template and performed using the following primers: forward 5′-AAGACTGAAGATGCCGCCACCAT-3′ and reverse 5′- AGAGCTGATGGTGGCGGCAT-3′. *To* generate recombinant adenoviruses, they were inserted into pENTR-3C (Invitrogen) and were transferred into pAd/CMV by LR Clonase recombination (Invitrogen). To express the exogenous protein, MIN6 and islets were infected with adenoviruses encoding GIV. For the rescue experiment of GIV downregulated MIN6 cells, MIN6 cells were first infected with lentivirus encoding GIV-shRNA targeting mouse 3′-UTR sequence and nontargeting sequence for 48 h, then they were infected with lentivirus encoding GIV-shRNA again followed with adenovirus encoding LacZ, wildtype GIV, or Y1767A mutant GIV for another 48 h. For the rescue experiment of db/db islets, db/db and control islets were infected with adenovirus encoding LacZ, wildtype GIV, or Y1767A mutant GIV for 48 h before being used for protein or functional assays. ## Immunoblotting and immunoprecipitation The sources of antibodies and their concentrations used are listed in Table S1. Cell lysate proteins, separated by gel electrophoresis, were transferred onto an Immobilon-P membrane (Millipore) and visualized by enhanced chemiluminescence (GE Healthcare Biosciences). Cells were lysed in lysis buffer containing 20 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM MgCl2, $1\%$ Triton X-100, 1 mM PMSF, and complete protease inhibitor cocktail (Roche). The lysates were cleared by centrifugation at 14,000 rpm for 15 min at 4 °C. The supernatants were subjected to immunoprecipitation with primary antibody and Protein G-Sepharose 4F (GE Healthcare Bioscience). After being washed five times with wash buffer containing 20 mM Tris-HCl pH 7.5, 150 mM NaCl, $10\%$ Glycerol, $0.1\%$ TritonX-100, the immunoprecipitates were subjected to SDS-PAGE and then transferred to a polyvinylidene difluoride membrane. The membrane was blocked with TBST (TBS plus $0.1\%$ Tween-20) containing $0.5\%$ nonfat dried milk powder and then incubated overnight at room temperature with the primary antibody. It was then washed three times with TBST and incubated for 1 h at room temperature with a 5000× dilution of horseradish peroxidase–conjugated secondary antibody (GE Healthcare Bioscience) in TBST containing $0.5\%$ nonfat dried milk powder, and was washed five times. Immunoreactive signals were then detected using ECL prime and an LAS-4000 chemiluminescence detection system (GE Healthcare Bioscience). ## Glucose-stimulated insulin secretion assay MIN6 cells plated on 24-well plates were cultured for 24 h. The cells were incubated in low-glucose Krebs–Ringer bicarbonate buffer (120 mM NaCl, 5 mM KCl, 24 mM NaHCO3, 1 mM MgCl2, 2 mM CaCl2, 15 mM Hepes pH 7.4, $0.1\%$ bovine serum albumin, 2.8 mM glucose) for 1 h followed by the same buffer or a high-glucose buffer containing 16.7 mM glucose for another 1 h. For mouse islets, batch assay and perfusion assay were performed as described [36]. Insulin levels were measured with a mouse insulin ELISA kit (Mercodia 10-1247-01) with an Infinite 200 Pro Reader (TECAN). ## Immunofluorescence and microscopy Murine pancreatic frozen sections and pancreatic β cells from C57BL/6N were fixed with $3\%$ paraformaldehyde in phosphate-buffered saline (PBS) for 30 min and permeabilized with $0.1\%$ Triton X-100 in PBS for 30 min. The cells were then treated with 50 mM NH4Cl-PBS for 10 min at room temperature and blocked with PBS containing $1\%$ bovine serum albumin for 15 min. The coverslips were incubated with primary antibody overnight, washed three times with PBS, and incubated with Alexa Fluor 488– or Alexa Fluor 568–conjugated secondary antibody (Invitrogen; 1:500 dilution) for 60 min. Samples were washed five times and mounted using SlowFade Gold (Invitrogen). The microscopic images were obtained with an A1 (Nikon) confocal laser scanning microscope equipped with a 100× oil immersion objective lens (1.49 NA) and NIS elements. The images were adjusted using NIS elements and ImageJ software. ## RNA isolation and expression analyses RNA was extracted using Trizol Reagent (Invitrogen) according to the manufacturer's instructions. Total RNA (1 μg) was reverse transcribed using oligo-(dT)12–18 primer and Superscript III (Invitrogen). Quantitative PCR was performed on a LightCycler 480 Real-Time PCR System (Roche) using SYBR Green I Master Mix reagent (Roche) with the primers. All reactions were run in triplicate. The relative mRNA expression level was calculated and normalized against Rplp$\frac{0}{36}$B4 mRNA expression. The primers for mouse genes were as follows: GIV forward primer 5′-GTGATCTCTACTGCTGAAGG-3′ and reverse primer 5′-TGTTGCTCCCTAGACCTGCT-3′, Rplp$\frac{0}{36}$B4 forward primer 5′-GGCCCTGCACTCTCGCTTTC-3′ and reverse primer 5′-TGCCAGGACGCGCTTGT-3′. ## Electron microscopic analysis of granule distribution Isolated islets were incubated in low-glucose (2.8 mM glucose) Krebs–Ringer buffer at 37 °C for 1 h followed with high-glucose (16.7 mM glucose) Krebs–Ringer buffer for another 1 h. They were fixed by immersion with $2\%$ paraformaldehyde, $2\%$ glutaraldehyde/$0.2\%$ picric acid in 0.1 M cacodylate buffer, pH 7.4, for 1.5 h at room temperature and embedded into $1\%$ agarose. They were then postfixed, embedded in plastic resin, and sectioned. Ultrathin sections (80 nm) were cut with microtome (Leica EM UC6) and analyzed under a transmission electron microscope (FEI Tecnai Spirit 120 kV). Micrographs were randomly taken at ×4000 magnification from 18 individual β cells from three mice for each genotype. The distance from the granule center to the plasma membrane was measured as described elsewhere [37]. ## Cell-surface biotinylation assay Start with cultured cells cooled to 4 °C. Place them on ice to maintain the temperature that is restrictive to endocytosis. Next, the membrane-impermeable sulfo-NHS-SS-biotin reagent is added, and cells are incubated in the dark for approximately 30 min. This allows sufficient time for biotin labels to covalently attach to the surface proteins. Cells are then removed from the ice and incubated at 37 °C for approximately 30 min. At this temperature, biotinylated surface proteins are endocytosed. Following incubation, the cells are cooled to 4 °C and a hydrophilic reducing agent like L-glutathione is added. This reacts with disulfide bonds and releases the biotin groups from labeled, nonendocytic proteins. Next, cells are lysed by centrifugation, thus breaking cell membranes and exposing biotinylated proteins. Following this, lysates are added to streptavidin-coated beads, and biotinylated proteins are allowed to bind. Beads are washed with cold PBS and eluted with a buffer containing detergents and reducing agents. These reagents denature bound proteins off beads and enable their recovery in the eluate. Proteins in the eluate are separated based on their molecular mass by gel electrophoresis. Lastly, Western blotting was carried out and probing the blot with protein-specific antibodies allowed the visualization of the target protein. The percentage of endocytosed protein can be quantified from the resulting band densities. ## Statistical analysis Data are presented as means ± SD values and compared by Student’s t test and one-way ANOVA with a Tukey’s test. A p value of <0.05 was considered significant. ## Data availability All data are contained within the article. ## Supporting information This article contains supporting information. Supporting Figures S1–S4 and Table S1 ## Conflict of interest The authors declare that they have no conflicts of interest with the contents of this article. ## Author contributions H. W., and H.-H. W. methodology; H. W., Y.-C. Y., and C. C. formal analysis; H. W., Y.-C. Y., and C. C. investigation; H. W., H.-H. W., and J.-K. Y. conceptualization; H. W., H.-H. W., and J.-K. Y. writing – original draft; T. I. writing – review & editing; H. W., H.-H. W., and J.-K. Y. supervision; H. W., H.-H. W., and J.-K. Y. funding acquisition. ## Funding and additional information This work was supported by grants from the $\frac{10.13039}{501100001809}$National Natural Science Foundation of China [82170809] to H. W., $\frac{10.13039}{501100001809}$National Natural Science Foundation of China [81930019] to J.-K. Y., $\frac{10.13039}{501100001809}$National Natural Science Foundation of China [22177030] and Natural Science Foundation of Hunan Province (2021JJ30093) to H.-H. W. It was also supported by the joint research program of the Institute for Molecular and Cellular Regulation, $\frac{10.13039}{501100007106}$Gunma University (grants 17007 to H.-H. W.). ## References 1. Sims E.K., Carr A.L.J., Oram R.A., DiMeglio L.A., Evans-Molina C.. **100 years of insulin: celebrating the past, present and future of diabetes therapy**. *Nat. Med.* (2021) **27** 1154-1164. PMID: 34267380 2. Roden M., Shulman G.I.. **The integrative biology of type 2 diabetes**. *Nature* (2019) **576** 51-60. PMID: 31802013 3. Rorsman P., Ashcroft F.M.. **Pancreatic beta-cell electrical activity and insulin secretion: of mice and men**. *Physiol. Rev.* (2018) **98** 117-214. PMID: 29212789 4. Tokarz V.L., MacDonald P.E., Klip A.. **The cell biology of systemic insulin function**. *J. Cell Biol.* (2018) **217** 2273-2289. PMID: 29622564 5. Kalwat M.A., Thurmond D.C.. **Signaling mechanisms of glucose-induced F-actin remodeling in pancreatic islet beta cells**. *Exp. Mol. Med.* (2013) **45** 6. Collares-Buzato C.B., Carvalho C.P.. **Is type 2 diabetes mellitus another intercellular junction-related disorder?**. *Exp. Biol. Med. (Maywood)* (2022) **247** 743-755. PMID: 35466731 7. Dissanayake W.C., Sorrenson B., Shepherd P.R.. **The role of adherens junction proteins in the regulation of insulin secretion**. *Biosci. Rep.* (2018) **38** 8. Konstantinova I., Nikolova G., Ohara-Imaizumi M., Meda P., Kucera T., Zarbalis K.. **EphA-Ephrin-A-mediated beta cell communication regulates insulin secretion from pancreatic islets**. *Cell* (2007) **129** 359-370. PMID: 17448994 9. Oakie A., Wang R.. **Beta-cell receptor tyrosine kinases in controlling insulin secretion and exocytotic machinery: c-kit and insulin receptor**. *Endocrinology* (2018) **159** 3813-3821. PMID: 30239687 10. Zhu J., Sun N., Aoudjit L., Li H., Kawachi H., Lemay S.. **Nephrin mediates actin reorganization via phosphoinositide 3-kinase in podocytes**. *Kidney Int.* (2008) **73** 556-566. PMID: 18033240 11. Jeon J., Leibiger I., Moede T., Walter B., Faul C., Maiguel D.. **Dynamin-mediated nephrin phosphorylation regulates glucose-stimulated insulin release in pancreatic beta cells**. *J. Biol. Chem.* (2012) **287** 28932-28942. PMID: 22718751 12. Verma R., Kovari I., Soofi A., Nihalani D., Patrie K., Holzman L.B.. **Nephrin ectodomain engagement results in Src kinase activation, nephrin phosphorylation, Nck recruitment, and actin polymerization**. *J. Clin. Invest.* (2006) **116** 1346-1359. PMID: 16543952 13. Fornoni A., Jeon J., Varona Santos J., Cobianchi L., Jauregui A., Inverardi L.. **Nephrin is expressed on the surface of insulin vesicles and facilitates glucose-stimulated insulin release**. *Diabetes* (2010) **59** 190-199. PMID: 19833886 14. Villarreal R., Mitrofanova A., Maiguel D., Morales X., Jeon J., Grahammer F.. **Nephrin contributes to insulin secretion and affects mammalian target of rapamycin signaling independently of insulin receptor**. *J. Am. Soc. Nephrol.* (2016) **27** 1029-1041. PMID: 26400569 15. Houssin E., Tepass U., Laprise P.. **Girdin-mediated interactions between cadherin and the actin cytoskeleton are required for epithelial morphogenesis in drosophila**. *Development* (2015) **142** 1777-1784. PMID: 25968313 16. Weng L., Enomoto A., Ishida-Takagishi M., Asai N., Takahashi M.. **Girding for migratory cues: roles of the Akt substrate girdin in cancer progression and angiogenesis**. *Cancer Sci.* (2010) **101** 836-842. PMID: 20132219 17. Garcia-Marcos M., Ghosh P., Farquhar M.G.. **GIV/Girdin transmits signals from multiple receptors by triggering trimeric G protein activation**. *J. Biol. Chem.* (2015) **290** 6697-6704. PMID: 25605737 18. Garcia-Marcos M., Ghosh P., Farquhar M.G.. **GIV is a nonreceptor GEF for G alpha i with a unique motif that regulates Akt signaling**. *Proc. Natl. Acad. Sci. U. S. A.* (2009) **106** 3178-3183. PMID: 19211784 19. Ghosh P., Garcia-Marcos M.. **Do all roads lead to Rome in G-protein activation?**. *Trends Biochem. Sci.* (2020) **45** 182-184. PMID: 31753703 20. DiGiacomo V., Marivin A., Garcia-Marcos M.. **When heterotrimeric G proteins are not activated by G protein-coupled receptors: structural insights and evolutionary conservation**. *Biochemistry* (2018) **57** 255-257. PMID: 29035513 21. Lopez-Sanchez I., Dunkel Y., Roh Y.S., Mittal Y., De Minicis S., Muranyi A.. **GIV/Girdin is a central hub for profibrogenic signalling networks during liver fibrosis**. *Nat. Commun.* (2014) **5** 4451. PMID: 25043713 22. Ghosh P., Beas A.O., Bornheimer S.J., Garcia-Marcos M., Forry E.P., Johannson C.. **A Galphai-GIV molecular complex binds epidermal growth factor receptor and determines whether cells migrate or proliferate**. *Mol. Biol. Cell* (2010) **21** 2338-2354. PMID: 20462955 23. Garcia-Marcos M., Jung B.H., Ear J., Cabrera B., Carethers J.M., Ghosh P.. **Expression of GIV/Girdin, a metastasis-related protein, predicts patient survival in colon cancer**. *FASEB J.* (2011) **25** 590-599. PMID: 20974669 24. Wang H., Misaki T., Taupin V., Eguchi A., Ghosh P., Farquhar M.G.. **GIV/girdin links vascular endothelial growth factor signaling to Akt survival signaling in podocytes independent of nephrin**. *J. Am. Soc. Nephrol.* (2015) **26** 314-327. PMID: 25012178 25. Wang H., Ishizaki R., Xu J., Kasai K., Kobayashi E., Gomi H.. **The Rab27a effector exophilin7 promotes fusion of secretory granules that have not been docked to the plasma membrane**. *Mol. Biol. Cell* (2013) **24** 319-330. PMID: 23223571 26. Palmén T., Ahola H., Palgi J., Aaltonen P., Luimula P., Wang S.. **Nephrin is expressed in the pancreatic beta cells**. *Diabetologia* (2001) **44** 1274-1280. PMID: 11692176 27. Lahdenpera J., Kilpelainen P., Liu X.L., Pikkarainen T., Reponen P., Ruotsalainen V.. **Clustering-induced tyrosine phosphorylation of nephrin by Src family kinases**. *Kidney Int.* (2003) **64** 404-413. PMID: 12846735 28. Lin C., Ear J., Pavlova Y., Mittal Y., Kufareva I., Ghassemian M.. **Tyrosine phosphorylation of the Galpha-interacting protein GIV promotes activation of phosphoinositide 3-kinase during cell migration**. *Sci. Signal.* (2011) **4** 29. Wu Z., Chang P.C., Yang J.C., Chu C.Y., Wang L.Y., Chen N.T.. **Autophagy blockade sensitizes prostate cancer cells towards Src family kinase inhibitors**. *Genes Cancer* (2010) **1** 40-49. PMID: 20811583 30. Asahara S., Shibutani Y., Teruyama K., Inoue H.Y., Kawada Y., Etoh H.. **Ras-related C3 botulinum toxin substrate 1 (RAC1) regulates glucose-stimulated insulin secretion via modulation of F-actin**. *Diabetologia* (2013) **56** 1088-1097. PMID: 23412604 31. Hammar E., Tomas A., Bosco D., Halban P.A.. **Role of the Rho-ROCK (Rho-associated kinase) signaling pathway in the regulation of pancreatic beta-cell function**. *Endocrinology* (2009) **150** 2072-2079. PMID: 19106222 32. Orci L., Gabbay K.H., Malaisse W.J.. **Pancreatic beta-cell web: its possible role in insulin secretion**. *Science* (1972) **175** 1128-1130. PMID: 4551150 33. Wang X., Enomoto A., Weng L., Mizutani Y., Abudureyimu S., Esaki N.. **Girdin/GIV regulates collective cancer cell migration by controlling cell adhesion and cytoskeletal organization**. *Cancer Sci.* (2018) **109** 3643-3656. PMID: 30194792 34. Miyazaki J., Araki K., Yamato E., Ikegami H., Asano T., Shibasaki Y.. **Establishment of a pancreatic beta cell line that retains glucose-inducible insulin secretion: special reference to expression of glucose transporter isoforms**. *Endocrinology* (1990) **127** 126-132. PMID: 2163307 35. Hohmeier H.E., Mulder H., Chen G., Henkel-Rieger R., Prentki M., Newgard C.B.. **Isolation of INS-1-derived cell lines with robust ATP-sensitive K+ channel-dependent and -independent glucose-stimulated insulin secretion**. *Diabetes* (2000) **49** 424-430. PMID: 10868964 36. Wang H., Mizuno K., Takahashi N., Kobayashi E., Shirakawa J., Terauchi Y.. **Melanophilin accelerates insulin granule fusion without predocking to the plasma membrane**. *Diabetes* (2020) **69** 2655-2666. PMID: 32994278 37. Wang H., Ishizaki R., Kobayashi E., Fujiwara T., Akagawa K., Izumi T.. **Loss of granuphilin and loss of syntaxin-1A cause differential effects on insulin granule docking and fusion**. *J. Biol. Chem.* (2011) **286** 32244-32250. PMID: 21768089
--- title: Disparities in health condition diagnoses among aging transgender and cisgender medicare beneficiaries, 2008-2017 authors: - Jaclyn M. W. Hughto - Hiren Varma - Gray Babbs - Kim Yee - Ash Alpert - Landon Hughes - Jacqueline Ellison - Jae Downing - Theresa I. Shireman journal: Frontiers in Endocrinology year: 2023 pmcid: PMC10040837 doi: 10.3389/fendo.2023.1102348 license: CC BY 4.0 --- # Disparities in health condition diagnoses among aging transgender and cisgender medicare beneficiaries, 2008-2017 ## Abstract ### Introduction The objective of this research is to provide national estimates of the prevalence of health condition diagnoses among age-entitled transgender and cisgender Medicare beneficiaries. Quantification of the health burden across sex assigned at birth and gender can inform prevention, research, and allocation of funding for modifiable risk factors. ### Methods Using 2009–2017 Medicare fee-for-service data, we implemented an algorithm that leverages diagnosis, procedure, and pharmacy claims to identify age-entitled transgender Medicare beneficiaries and stratify the sample by inferred gender: trans feminine and nonbinary (TFN), trans masculine and nonbinary (TMN), and unclassified. We selected a $5\%$ random sample of cisgender individuals for comparison. We descriptively analyzed (means and frequencies) demographic characteristics (age, race/ethnicity, US census region, months of enrollment) and used chi-square and t-tests to determine between- (transgender vs. cisgender) and within-group gender differences (e.g., TMN, TFN, unclassified) difference in demographics ($p \leq 0.05$). We then used logistic regression to estimate and examine within- and between-group gender differences in the predicted probability of 25 health conditions, controlling for age, race/ethnicity, enrollment length, and census region. ### Results The analytic sample included 9,975 transgender (TFN $$n = 4$$,198; TMN $$n = 2$$,762; unclassified $$n = 3$$,015) and 2,961,636 cisgender (male $$n = 1$$,294,690, female $$n = 1$$,666,946) beneficiaries. The majority of the transgender and cisgender samples were between the ages of 65 and 69 and White, non-Hispanic. The largest proportion of transgender and cisgender beneficiaries were from the South. On average, transgender individuals had more months of enrollment than cisgender individuals. In adjusted models, aging TFN or TMN Medicare beneficiaries had the highest probability of each of the 25 health diagnoses studied relative to cisgender males or females. TFN beneficiaries had the highest burden of health diagnoses relative to all other groups. ### Discussion These findings document disparities in key health condition diagnoses among transgender Medicare beneficiaries relative to cisgender individuals. Future application of these methods will enable the study of rare and anatomy-specific conditions among hard-to-reach aging transgender populations and inform interventions and policies to address documented disparities. ## Introduction Transgender people in the United States (US) experience significant health disparities throughout their life course relative to cisgender (non-transgender) people (1–6). These disparities stem from multilevel sources of stigma that serve as sources of chronic stress as well as barriers to accessing essential resources such as healthcare, employment, and income (7–9). While extensive community-based research drawn from convenience samples [10, 11], and research using state or national data has assessed the health of transgender youth and adults overall (12–14) and relative to cisgender people (15–18), a dearth of national studies have compared the health of an exclusively aging (i.e., age 65 or older) sample of transgender and cisgender adults. The risk of being diagnosed with a chronic health condition, such as cancer, HIV, depression, osteoporosis, and dementia, increases as one ages [19, 20]. Aging transgender adults are expected to be at an even higher risk of developing physical and mental health conditions than their cisgender peers due to stigma-related stress experienced throughout their lives and barriers to accessing quality healthcare (9, 21–25). Although researchers are increasingly studying the health of aging transgender adults (26–28), and some national studies have explored disparities among predominantly aging populations relative to cisgender groups (29–32), no national research, to our knowledge, has explored within- (e.g., trans feminine people vs. trans masculine people) and between- (e.g., transgender vs. cisgender) group gender differences among aging transgender adults and a general population of cisgender adults. Without comparative data on the health of aging transgender and cisgender subpopulations, it is difficult to know which subgroups are in greatest need of public health and policy interventions. Recent methodological advances have enabled the use of claims databases to study the health of transgender populations. Blosnich et al. [ 2013] innovatively used transgender-specific International Classification of Diseases, Ninth Edition (ICD-9) diagnosis codes (e.g., Gender Identity Disorder (GID)) to identify transgender veterans using national Veteran Health *Administration data* [14]. Proctor and colleagues later followed by applying transgender-related ICD-9 diagnosis codes to identify age- and disability-entitled transgender Medicare beneficiaries [33]. Using the same approach as Proctor et al., Dragon and colleagues found that, compared to cisgender Medicare recipients receiving care in 2015, a higher proportion of transgender beneficiaries receiving care during the sample year had been diagnosed with several major health conditions, including asthma, autism spectrum disorder, chronic obstructive pulmonary disease, depression, hepatitis, HIV, schizophrenia, and substance use disorders [30]. Building on the work of Proctor and Dragon, Progovac and colleagues examined disparities between transgender and cisgender Medicare beneficiaries receiving care between 2009 and 2014 [31]. They found that older and disabled transgender Medicare beneficiaries had more diagnoses for chronic health conditions than their cisgender counterparts [31]. Notably, however, none of these studies examined health diagnoses disparities by gender subgroup (trans feminine, trans masculine, cisgender male, cisgender female), despite prior survey-based research showing substantial within and between gender group variations in mental and physical health conditions [2, 6, 10, 11, 34, 35]. Our team has advanced algorithms identifying transgender beneficiaries in claims data and inferring their gender [12, 13, 36]. Using commercial insurance data from 2001-2019, we adapted a method developed by Jasuja and colleagues [13] that used diagnosis, procedure, and pharmacy claim codes to identify individuals with one or more transgender-related diagnoses or Endocrine Disorder Not Otherwise Specified [Endocrine NOS] in conjunction with prescriptions for gender-affirming hormones (e.g., estrogen, testosterone) or gender-affirming procedures (e.g., phalloplasty, vaginoplasty) to identify 38,598 transgender adults. We furthered the algorithm using an approach developed by Yee, Lind, & Downing [36] for Oregon Medicaid recipients to improve our ability to categorize transgender samples by inferred gender. 1 Specifically, we used a hierarchical approach to examine beneficiaries’ history of gender-affirming and reproductive anatomy-specific 2 procedures (e.g., prostate-related procedures, hysterectomy), diagnoses, and pharmacy claims for gender-affirming hormones to categorize the sample based on inferred gender identity: 3 trans masculine/nonbinary (TMN) or trans feminine/nonbinary (TFN) [12]. 4 By including reproductive anatomy-specific procedures and diagnoses in combination with gender-affirming care, we were able to infer the gender of $76\%$ of the sample ($50\%$ TMN; $26\%$ TFN), which represented a notable improvement from prior approaches [13]. When we applied our modified algorithm to study the health of younger, commercially-insured transgender individuals [12], we found that relative to TMN people, TFN people had significantly higher predicted probabilities of most health condition diagnoses, including HIV, atherosclerotic cardiovascular disorder, myocardial infarction, alcohol use disorder, and substance use disorder [4]. In contrast, TMN individuals had significantly higher predicted probabilities of diagnosed post-traumatic stress disorder and depression than TFN people. While our prior research provides insights into within-group gender-related disparities in health diagnoses among commercially-insured adults aged 18 and over, it is unknown whether these patterns are similar among aging transgender adults aged 65 and older. To more fully understand the health of aging transgender adults, research is needed to explore within- and between-group gender differences in health diagnoses among aging transgender adults, overall and relative to their cisgender counterparts. As the largest insurer of U.S adults aged 65 and older [37], *Medicare is* the ideal data source to utilize to document within- and between-group disparities in health diagnoses for aging transgender and cisgender adults. Building on prior work [13, 31, 36], in the present study, we sought to apply our claims-based method [12] to identify transgender and cisgender samples in Medicare data, stratify the samples by gender, and explore within- and between-group gender differences in health diagnoses among aging transgender and cisgender Medicare beneficiaries. Findings from this national study can help identify health diagnosis disparities among aging Medicare beneficiaries and the subgroups in greatest need of tailored interventions to prevent and treat adverse health outcomes. ## Study design/data source We conducted a retrospective cross-sectional analysis to identify transgender adults, stratify them into inferred gender subgroups, and compare these groups to cisgender people. Fee-for-service *Medicare data* were accessed through the Virtual Data Resource Center (VRDC) maintained by the Centers for Medicare & Medicaid Services (CMS) through a data use agreement (DUA 52772). We queried the Medicare Master Beneficiary Summary File and final paid claims for inpatient, physician, and other suppliers, and prescription services from 2008 to 2017. ## Identifying transgender individuals To identify transgender individuals, we adapted our algorithm [12] developed with commercial insurance claims for Medicare. These methods and the corresponding codes used to identify the transgender sample are described in detail elsewhere [12, 13]. Briefly, we included any person with a transgender-related diagnosis (e.g., GID); transgender-conclusive procedures (e.g., “operations for sex transformation, not elsewhere classified”); a diagnosis of endocrine NOS in conjunction with a transgender-suggestive procedure or gender-affirming hormone prescription (Supplementary Figure A). ## Stratifying the transgender sample by inferred gender We subsequently applied a previously-developed stepwise approach [12] to categorize the inferred gender of the transgender sample (Supplementary Figure B). As described in detail elsewhere [12], briefly, we first classified the inferred gender of transgender individuals based on the presence of claims for gender-affirming genital surgeries (e.g., “vaginal construction,” “construction of penis”). Then, within the remaining sample, we categorized the sample by gender if they had certain types of highly specific and highly sensitive reproductive anatomy-specific care and diagnoses (e.g., hysterectomy, pregnancy, prostate cyst, prostate screening). Next, we categorized individuals according to their receipt of gender-affirming hormones or procedures. Finally, using the remaining sample, individuals who had other reproductive anatomy-related diagnoses or procedures (e.g., vulvectomy for TMN or testicular hyperfunction for TFN) were categorized by inferred gender. Individuals who had not yet been assigned a gender category or those with conflicting codes at the final step remained unclassified. The unclassified group was comprised of people with a transgender-related diagnosis code (e.g., GID, transsexualism) and no gender-affirming hormones or procedures or reproductive-anatomy-related care or who had conflicting codes. ## Eligibility criteria for cisgender comparison cohort After identifying the transgender cohort, we selected a random $5\%$ sample from the remaining Medicare beneficiaries, whom we refer to here and going forth as cisgender. The sex of the beneficiaries classified as cisgender was taken from the Master Beneficiary Summary File. Because identification of the transgender cohort relied on engagement in care, we limited the cisgender cohort to beneficiaries who had at least one Part A, B, or D claim between 2008 and 2017. We excluded cisgender beneficiaries with missing data on sex and/or date of birth (about $5\%$ of the sample). ## Measures Sociodemographics. Age was categorized as 65-<70; 70-74; 75-79, 80-84, 85+. Race and ethnicity were categorized as Asian (non-Hispanic), Black (non-Hispanic), Hispanic, White (non-Hispanic), another race/ethnicity (non-Hispanic), or unknown. US Census regions included Northeast, Midwest, South, West, and Unknown. Since all Medicare beneficiaries receive fee-for-service (FFS) coverage and some may elect to pay for supplemental Part D prescription coverage, we created separate continuous months of insurance coverage variables for individuals with FFS coverage only and those with FFS plus Part D coverage. Health Condition Diagnoses. We used the CMS Chronic Condition Warehouse to identify diagnoses for 25 health conditions [38]. We grouped the health conditions diagnoses as follows: cancer (breast, colorectal, endometrial, lung, prostate); heart, lung, & kidney conditions (asthma, chronic kidney disease, chronic obstructive pulmonary disease [COPD], cardiac arrhythmia, congestive heart failure, coronary artery disease, hyperlipemia, hypertension, stroke); infectious diseases (hepatitis, HIV/AIDS); other health conditions (arthritis, diabetes, osteoporosis); mental or cognitive illness (dementia, depression, schizophrenia); and substance use disorders (alcohol, drug, tobacco use). ## Data analysis Since *Medicare data* include individuals who qualify for coverage based on age (i.e., 65 or older) and disability, and the current analysis focuses on the health of aging individuals, we restricted the transgender and cisgender samples to those eligible for Medicare based on age at enrollment. We then descriptively analyzed demographic characteristics and used χ2 tests and t-tests to assess between-group (transgender vs. cisgender) and within-group differences (e.g., TMN vs. TFN) in sociodemographics ($p \leq 0.05$). Next, we estimated the crude prevalence of each condition stratified by all gender subgroups. Since the distribution of sociodemographic characteristics varied by gender subgroup (TFN vs. TMN, cisgender male vs. cisgender female), we fit logistic regression models predicting the log odds of each condition while controlling for age at enrollment, race/ethnicity, months of enrollment, and Census region. To facilitate within- and between-group comparisons, we obtained the predicted probability of each condition for each of the 4 gender subgroups that could be classified (TFN people, TMN people, cisgender males, cisgender females), holding covariates at their means. Means differences in the predicted probabilities were also assessed ($p \leq 0.05$). All analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC). ## Reporting and discussion of results In reporting and discussing the results for the transgender and cisgender subgroups, we describe within- and between-group gender differences in the crude prevalence and means of the demographic characteristics and the predicted probabilities of the health diagnoses. For demographics, we report and discuss differences between transgender and cisgender people overall, as well as within-group differences among transgender (TFN vs. TMN) and cisgender (male vs. female) subgroups. For the predicted probabilities, we report all diagnoses and discuss the diagnoses with the widest within- and between-group disparities. For differences in the predicted probabilities, we first discuss between-group differences in individuals presumed to have been assigned the same sex at birth (i.e., cisgender males and TFN people; and cisgender females and TMN people). Although referring to a transgender person by their assigned birth sex is not affirming or appropriate, these comparisons were made because, on a population level, cisgender and transgender people assigned the same sex at birth are typically born with similar reproductive anatomy and endogenous hormones that could impact their risk of developing specific health conditions. People assigned the same sex at birth may also be more likely to experience similar social or developmental influences in childhood and adolescence that could influence their tendency to engage in behaviors that might increase or decrease their risk of developing specific conditions. We also discuss between-group differences between TFN people and cisgender females and between TMN people and cisgender males. We make these comparisons as the use of gender-affirming hormones and procedures may result in TFN people having hormone exposure levels and anatomies that are more closely aligned with cisgender females than cisgender males, and TMN individuals may come to have hormone exposure levels and anatomies that are more similar to cisgender males than cisgender females. Finally, making these comparisons in the results and discussion enables us to engage with the breadth of medical, public health, psychological, sociological, and other literature documenting the prevalence of and mechanisms underlying health diagnosis disparities between individuals who were assigned the same birth sex or who share similar gender identities or expressions. Specific limitations of this approach are denoted in the discussion. ## Demographic characteristics Table 1 summarizes the demographic characteristics of the sample. We identified 9,975 transgender FFS beneficiaries who qualified for Medicare based on age. Overall, 4,198 ($41.1\%$) were categorized as TFN, 2,762 ($27.7\%$) as TMN, and the gender could not be inferred and classified for 3,015 transgender beneficiaries ($30.2\%$ unclassified). Of the 2,961,636 cisgender individuals included in the comparison sample, 1,294,690 ($43.7\%$) were male, and 1,666,946 ($56.3\%$) were female. **Table 1** | Unnamed: 0 | TOTAL | TOTAL.1 | TOTAL.2 | TOTAL.3 | Unnamed: 5 | TRANSGENDER | TRANSGENDER.1 | TRANSGENDER.2 | TRANSGENDER.3 | TRANSGENDER.4 | TRANSGENDER.5 | Unnamed: 12 | CISGENDER | CISGENDER.1 | CISGENDER.2 | CISGENDER.3 | Unnamed: 17 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | N=2,971,611 | N=2,971,611 | N=2,971,611 | N=2,971,611 | | N=9,975 | N=9,975 | N=9,975 | N=9,975 | N=9,975 | N=9,975 | | N=2,961,636 | N=2,961,636 | N=2,961,636 | N=2,961,636 | | | | Transgender N=9,975 | Transgender N=9,975 | Cisgender N=2,961,636 | Cisgender N=2,961,636 | | TFN N=4,198 | TFN N=4,198 | TMN N=2,762 | TMN N=2,762 | Unclassified N=3,015 | Unclassified N=3,015 | | Male N=1,294,690 | Male N=1,294,690 | Female N=1,666,946 | Female N=1,666,946 | | | Age at Enrollment (y) | N | % | N | % | P-Value | N | % | N | % | N | % | P-value | N | % | N | $ | P-Value | | 65-<70 | 6003 | 60.2 | 1604745 | 54.2 | <0.0001 | 2643 | 63.0 | 1583 | 57.3 | 1777 | 58.9 | <0.0001 | 730794 | 56.4 | 873951 | 52.4 | <0.0001 | | 70-74 | 1565 | 15.7 | 464378 | 15.7 | | 684 | 16.3 | 502 | 18.2 | 379 | 12.6 | | 215803 | 16.7 | 248575 | 14.9 | | | 75-79 | 1132 | 11.3 | 362645 | 12.2 | | 481 | 11.5 | 333 | 12.1 | 318 | 10.5 | | 157583 | 12.2 | 205062 | 12.3 | | | 80-84 | 791 | 7.9 | 277783 | 9.4 | | 276 | 6.6 | 233 | 8.4 | 282 | 9.4 | | 109575 | 8.5 | 168208 | 10.1 | | | >= 85 | 484 | 4.9 | 252085 | 8.5 | | 114 | 2.7 | 111 | 4.0 | 259 | 8.6 | | 80935 | 6.3 | 171150 | 10.3 | | | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | Race/ethnicity | | Asian | 237 | 2.4 | 92031 | 3.1 | <0.0001 | 95 | 2.3 | 63 | 2.3 | 79 | 2.6 | 0.001 | 40329 | 3.1 | 51702 | 3.1 | <0.0001 | | Black | 655 | 6.6 | 239491 | 8.1 | | 234 | 5.6 | 197 | 7.1 | 224 | 7.4 | | 95361 | 7.4 | 144130 | 8.6 | | | Hispanic | 537 | 5.4 | 222517 | 7.5 | | 231 | 5.5 | 136 | 4.9 | 170 | 5.6 | | 96718 | 7.5 | 125799 | 7.5 | | | White | 8303 | 83.2 | 2340841 | 79.0 | | 3524 | 83.9 | 2322 | 84.1 | 2457 | 81.5 | | 1026070 | 79.3 | 1314771 | 78.9 | | | Another race/ ethnicity | 105 | 1.1 | 32060 | 1.1 | | 43 | 1.0 | 21 | 0.8 | 41 | 1.4 | | 14364 | 1.1 | 17696 | 1.1 | | | Unknown | 138 | 1.4 | 34696 | 1.2 | | 71 | 1.7 | 23 | 0.8 | 44 | 1.5 | | 21848 | 1.7 | 12848 | 0.8 | | | Census Region | Census Region | Census Region | Census Region | Census Region | Census Region | Census Region | Census Region | Census Region | Census Region | Census Region | Census Region | Census Region | Census Region | Census Region | Census Region | Census Region | Census Region | | Northeast | 2264 | 22.7 | 557113 | 18.8 | <0.0001 | 947 | 22.6 | 628 | 22.7 | 689 | 22.9 | <0.0001 | 237772 | 18.4 | 319341 | 19.2 | <0.0001 | | Midwest | 1962 | 19.7 | 669506 | 22.6 | | 788 | 18.8 | 521 | 18.9 | 653 | 21.7 | | 292635 | 22.6 | 376871 | 22.6 | | | South | 3252 | 32.6 | 1071084 | 36.2 | | 1388 | 33.1 | 997 | 36.1 | 867 | 28.8 | | 467104 | 36.1 | 603980 | 36.2 | | | West | 2468 | 24.7 | 628446 | 21.2 | | 1060 | 25.3 | 611 | 22.1 | 797 | 26.4 | | 281716 | 21.8 | 346730 | 20.8 | | | Unknown | --- | --- | 35487 | 1.2 | | --- | --- | --- | --- | --- | --- | | 15463 | 1.2 | 20024 | 1.2 | | | Months of Coverage | Mean | SD | Mean | SD | P-Value | Mean | SD | Mean | SD | Mean | SD | P-value | Mean | SD | Mean | SD | P-Value | | Mean months of FFS Coverage | 76.49 | 41.65 | 50.74 | 44.94 | <0.0001 | 79.99 | 39.47 | 88.32 | 37.08 | 60.76 | 43.79 | <0.0001 | 49.34 | 44.03 | 51.82 | 45.61 | <0.0001 | | Mean months of FFS + Part D Coverage | 50.04 | 41.64 | 32.47 | 38.06 | <0.0001 | 50.36 | 40.19 | 59.53 | 42.28 | 40.92 | 41.06 | <0.001 | 29.16 | 35.78 | 35.03 | 39.54 | <0.0001 | There were significant between-group gender differences in the demographics of the samples. Overall, a higher proportion of transgender beneficiaries enrolled in Medicare at a younger age, were non-Hispanic White, and lived in the Northeast or West at enrollment compared to cisgender beneficiaries. Transgender individuals also had a significantly longer mean period of continuous enrollment relative to cisgender individuals ($p \leq .0001$). There were also significant within-group gender differences. Regarding demographic differences between TFN and TMN people, although statistically significant, there were relatively small differences between these groups with regard to race/ethnicity and region. Notably, among the transgender sample, a larger proportion of TFN people enrolled before age 70, whereas a larger proportion of the unclassified group enrolled at age 85 or older; also, on average, TMN had significantly more months of continuous enrollment than other transgender groups ($p \leq .0001$). With regard to with-in-group differences for the cisgender sample, a higher proportion of cisgender males than females enrolled in Medicare before age 70, yet, on average, cisgender females had more continuous months of enrollment than cisgender males. Further, despite statistically significant differences in the distribution of race-ethnicity and geographic region, percent point differences were relatively small between cisgender males and females. ## Health condition diagnoses The unadjusted prevalence for each diagnosis among all transgender and cisgender subgroups is presented in Supplementary Table 1. Table 2 and Figure 1 present results from the adjusted models for TMN and TFN people and cisgender males and females, and Supplementary Table 2 presents the mean differences in the predicted probability of each diagnosis by gender subgroup. Overall, TFN or TMN people had the highest predicted probability (herein “probability”) of every diagnosis relative to cisgender males and females. Additionally, TFN individuals had the highest probability of being diagnosed with the majority of health conditions compared to TMN people, as well as cisgender males and females. When exploring within-group differences in the transgender population, TFN Medicare beneficiaries had a significantly and notably elevated probability of being diagnosed with HIV/AIDS and alcohol use disorder relative to TMN beneficiaries, whereas TMN beneficiaries had a significantly and notably elevated probability of being diagnosed with breast cancer, arthritis, and osteoporosis relative to TFN beneficiaries (all p’s<0.001). With the exception of breast cancer, asthma, arthritis, osteoporosis, and each mental health condition, which were significantly elevated among cisgender females, cisgender males had a higher probability of being diagnosed with all other conditions ($p \leq 0.01$). When between-group differences were assessed by those presumed to be assigned a male sex at birth, TFN beneficiaries had a significantly and notably elevated probability of being diagnosed with HIV/AIDS, hepatitis, depression, dementia, schizophrenia, and substance use disorder compared to cisgender male beneficiaries (all p’s<0.001). When exploring between-group differences by those presumed to be assigned a female sex at birth, the widest disparities were observed between TMN and cisgender female beneficiaries for all heart, lung, and kidney diseases, arthritis, hepatitis, depression, dementia, schizophrenia, and substance use disorder diagnoses, with TMN beneficiaries having a significantly elevated probability of all diagnoses (all p’s<0.001) except HIV/AIDS, which was comparable ($$p \leq 0.64$$). When assessing between-group disparities between TFN people and cisgender females, TFN people had a significantly higher probability of having all diagnoses (all p’s<0.001) except arthritis ($p \leq 0.55$), with the probability of the following diagnoses being particularly elevated: HIV/AIDS, hepatitis, and all substance use disorders. Conversely, TMN people had a significantly higher probability of being diagnosed with breast cancer, arthritis, osteoporosis, depression, and drug use disorder relative to cisgender males (all p’s<0.001). In contrast, cisgender males had a significantly higher probability of alcohol use disorder relative to TMN people ($p \leq 0.001$). ## Discussion This study advances the fields of endocrinology and population health research through the application of an algorithm that identifies transgender people in *Medicare data* and stratifies by inferred gender to explore within- and between-group disparities in health diagnoses among aging transgender and cisgender beneficiaries. We found that overall, aging TFN or TMN Medicare beneficiaries had a significantly higher probability of every health diagnosis studied relative to cisgender male or female beneficiaries, with TFN individuals experiencing some of the highest diagnosis burdens relative to other groups. Future application of these methods will enable the study of rare and hormone and anatomy-related conditions among hard-to-reach aging transgender populations and their cisgender counterparts and can inform interventions to address documented health diagnosis disparities [39]. In understanding the health diagnosis disparities observed here, it is essential to underscore the role of stigma and its stress-related sequelae that transgender individuals differentially experience throughout their lives relative to cisgender individuals [7]. Indeed, at the structural level, harmful state, federal, and organizational policies may intentionally or unintentionally restrict access to the essential resources (e.g., healthcare, education, employment, housing, public bathrooms, and other accommodations) that transgender people need to maintain their health [7]. Additionally, exposure to interpersonal discrimination, violence, and other sources of enacted stigma at the hands of healthcare providers, family members, employers, educators, sexual partners, and others further restricts access to essential resources and contributes to poor health directly and indirectly through chronic stress [7]. As a person ages, chronic activation of the body’s stress response system can compromise health over time via a process called allostatic load (40–42). Chronic stress is associated with adverse health outcomes, such as cancer, hypertension, diabetes, mood and substance use disorders, and even death [43, 44], and is therefore theorized to be a primary driver of the health diagnosis disparities observed among aging transgender individuals relative to cisgender individuals in this study [7]. Few large-scale studies have explored disparities within transgender populations, and relative to cisgender males and females for health conditions such as breast cancer that are known to differ by assigned birth sex. In the United States, breast cancer is approximately 70-100 times less common among cisgender males than females [45]. Building on prior cancer research with transgender samples (46–50), we found that both TMN people and cisgender females had a significantly elevated probability of a breast cancer diagnosis relative to TFN people and cisgender males. Additionally, although the probability of a breast cancer diagnosis in TFN people was significantly lower than that of cisgender females, it was significantly higher relative to that of cisgender males – a finding that aligns with data from a younger Dutch cohort study comparing the incidence of breast cancer in TFN people on estrogen to cisgender females and males [46]. Lifetime stress, a risk factor for cancer [51, 52], together with higher rates of screening among transgender people with breast tissue [53], may have contributed to the elevated probability of being diagnosed with breast cancer among TFN and TMN people relative to their cisgender counterparts. Higher levels of endogenous estradiol in the blood are also associated with a higher risk of breast cancer [54], whereas endogenous androgens are known to inhibit the progression of certain types of breast cancer [46, 55]. Decreased testosterone due to the use of antiandrogenic treatment and orchiectomy and denser breast tissue due to the use of exogenous hormones [56] could increase the risk for certain types of breast cancer in TFN people relative to cisgender males [46, 55]. Still, the risk of taking exogenous estrogen for gender affirmation is no greater for TFN people than it is for cisgender females taking such therapies for other indications (57–59). Moreover, our study found a significantly lower probability of being diagnosed with breast cancer in TFN people as compared to cisgender females, which may be due to TFN people having less breast tissue and lower lifetime exposure to estrogen than cisgender females. Notable differences in the probability of osteoporosis and arthritis were also observed among TFN and TMN individuals relative to their cisgender comparators. Specifically, TMN individuals had the highest probability of being diagnosed with osteoporosis and arthritis than all other groups. Further, although the probability of an osteoporosis diagnosis was substantially and significantly lower among TFN people than cisgender females, TFN people had a significantly higher probability of having an osteoporosis diagnosis relative to cisgender males. Our findings align with research showing that osteoporosis and most types of arthritis are more prevalent in older people assigned a female sex at birth than those assigned a male sex at birth (60–62). Obesity, heart disease, and smoking can also increase the risk for rheumatoid arthritis [60, 63], and our study and prior research suggest that TMN people may be at elevated risk of being diagnosed with these conditions (3, 64–66). Relatedly, known risk factors for osteoporosis in the general population include high alcohol consumption, tobacco use, anorexia nervosa, rheumatoid arthritis, chronic kidney disease, and HIV infection (59, 62, 67–69), all of which (except anorexia, which was not studied) were more commonly diagnosed among TFN or TMN people in our sample relative to cisgender males and females. Low physical activity has also been linked to osteoporosis risk in cisgender people [62] and transgender people, even before initiating hormones [59]. Research finds that some TFN and TMN individuals have lower levels of physical activity than their cisgender comparators due to various social and physical barriers to exercising, including inadequate changing facilities, revealing and heavily gendered sports clothing, body dissatisfaction, and fears of acceptance (70–72). Additionally, stress is a major risk factor for osteoporosis [62], and other chronic conditions (e.g., tobacco smoking) that increase osteoporosis and arthritis risk [61, 62, 73]; and as previously noted, transgender individuals experience greater stigma-related stress throughout their lives relative to cisgender people [7]. Wear and tear on one’s joints and changes in hormone levels in older age are associated with a greater risk of arthritis and osteoporosis, respectively [62, 74, 75]. Although there is no evidence to suggest a causal link between exogenous hormone use and arthritis risk in transgender people, in considering osteoporosis risk, the World Professional Association of Transgender Health notes that the use of gender-affirming medical and surgical interventions, such as hormone therapy, androgen blockade, and gonadectomy, have the potential to influence bone health in different ways [59]. While testosterone therapy has been associated with no change or even improvements in bone density among TMN people [59, 67, 76], the higher probability of being diagnosed with osteoporosis among TMN people in this sample may be due to natural reductions in estrogen among those who experienced menopause [62, 75]. Further, although TFN people have been shown to have improved bone density after initiating estrogen [59, 67, 77, 78], risk factors for osteoporosis include the absence of or underutilization of estrogen after gonadectomy or the use of androgen blockers without or with insufficient estrogen [59, 67, 79]. Further, although Gonadotropin-Releasing Hormone Agonists (GnRHa) are very effective in reducing testosterone levels to help TFN individuals achieve their gender-affirmation goals, the use of GnRHA can result in osteoporosis if concurrent doses of estrogen are insufficient [59, 67, 80]. In light of prior research, our findings underscore the necessity for clinicians to understand the various physiological, developmental, behavioral, and environmental risk and protective factors for osteoporosis and other health conditions in order to support aging transgender and cisgender individuals in optimizing their health and well-being. When exploring mental health, in alignment with prior research [5, 12, 13, 30, 59, 81], this study found a significantly higher probability of all mental health diagnoses among both aging TFN and TMN Medicare beneficiaries relative to their cisgender counterparts. Although TFN and TMN people had a similar probability of having a dementia diagnosis, the probability among TFN and TMN people was significantly elevated relative to cisgender males and females. Further, although depression diagnoses were also significantly elevated among TFN and TMN people compared to the cisgender subgroups, both TMN people and cisgender females had a significantly higher probability of depression than TFN people and cisgender males, respectively. Prior research finds that individuals assigned a female sex at birth are twice as likely to be diagnosed with depression than people assigned a male sex at birth [82, 83], which is consistent with our findings. Burgeoning research also finds an elevated probability of Alzheimer’s disease and related dementias [84] and schizophrenia [85] diagnoses among transgender people relative to cisgender individuals. Although the mechanisms underlying these disparities are not well understood, misdiagnoses by untrained or biased providers may contribute to more diagnoses for transgender people, particularly in the case of schizophrenia [59, 60]. Additionally, transgender people have historically been required to undergo psychotherapy to be approved for medical gender affirmation treatment (86–88); thus, the higher probability of mental health diagnoses observed here may be due to increased contact with mental health specialists rather than a reflection of the true probability of these conditions in aging transgender populations [85]. Nonetheless, extensive research has documented disparities in poor mental health among transgender vs. cisgender individuals [2, 59, 89]. Given that the onset of depression, dementia, cognitive decline, and schizophrenia have all been linked to stress in both transgender and cisgender samples (7, 90–96), it is quite probable that the elevated mental health burden observed among aging transgender people in this sample is due to the high levels of stigma-related stress that transgender people experience throughout the life course including in the context of receiving medical care [7, 26, 97]. Some of the most striking within- and between- gender group diagnostic disparities were observed with regard to infectious diseases. The probability of HIV/AIDS among TFN people was approximately 9-fold that of TMN people and cisgender females and 4.5-fold that of cisgender males. The high probability of HIV/AIDS among TFN people is well documented in the literature [4, 12, 13, 98] and has been attributed to multilevel factors. For example, TFN people may acquire HIV at higher rates due to networks-level factors such as the high prevalence of HIV within TFN people’s limited pool of potential sexual partners or anatomical considerations that predispose TFN people with a penis to engage in sexual acts that carry greater HIV risk (i.e., receptive anal sex) [99, 100]. A confluence of stigma-related, structural, and interpersonal factors also restrict access to essential human needs (i.e., employment, shelter, food, love, gender affirmation) for TFN people [7] and can lead to engagement in HIV risk behavior such as transactional sex for financial survival [99, 100], receptive anal sex (i.e., “bottoming”) as a means to be affirmed in one’s gender when access to other sources of affirmation are limited (100–102), and condomless sex to please one’s sexual partner in the context of relationship stigma and partner scarcity [100, 103, 104]. Similarly, the probability of hepatitis was significantly higher among TFN people than all other groups, and the probability of a hepatitis diagnosis among TMN people was roughly twice that of cisgender males and females. In addition to the aforementioned sexual pathways, the significantly higher probability of a hepatitis diagnosis among transgender people relative to cisgender individuals may also be driven by the sharing of syringes for injecting drugs, hormones, and silicone - behaviors that may be more prevalent among transgender people due to stigma-related barriers to healthcare (7, 99–101, 105–107). Consistent with prior research with Medicare- and commercially-insured individuals [3, 4, 12, 13, 30], aging transgender Medicare beneficiaries in this sample had a significantly higher probability of most substance use disorder diagnoses than their cisgender comparators. Research has consistently found elevated levels of substance use and diagnosed substance use disorders among transgender samples relative to cisgender samples [3, 5, 12, 13, 30, 108], due in part to the need to cope with the psychological toll of discrimination, violence, and other forms of stigma [1, 7, 34, 109]. In prior studies of younger, commercially-insured transgender individuals [3, 4, 12, 13], drug, alcohol, and tobacco use disorder diagnoses were particularly elevated among TFN people than TMN people. In contrast, the probability of having a drug use disorder diagnosis was fairly comparable between TFN and TMN people in the current study, though the probability of alcohol and tobacco use disorder diagnoses was significantly elevated among TFN people relative to TMN people in the sample. A similar trend was observed for cisgender individuals, such that cisgender males had a significantly higher probability of alcohol and tobacco use disorder diagnoses than cisgender females. The substance use disorder findings suggest that although stigma likely drives the higher prevalence of substance use disorders observed among transgender people [7], individuals assigned a male sex at birth may be more susceptible to some substance use disorders than those assigned a female sex at birth [110, 111]. It is also possible that developmental factors may lead individuals assigned a male sex at birth to cope with stress using certain substances more frequently than those assigned a female sex at birth [110, 111]. Although the receipt of a diagnosis is an important step in linkage to treatment, research finds that transgender people face difficulty finding affirming substance use treatment services (112–114). Thus, efforts are needed to effectively prevent and treat substance use disorders among transgender people by reducing societal stigma [7], helping transgender people to cope with stigma through health-promoting means [115, 116], and improving access to gender-affirming substance use treatment services [117]. ## Clinical and research implications Although TFN or TMN aging adults in our sample had a significantly higher probability of receiving specific health diagnoses relative to one or both cisgender subgroups, current, evidenced-based clinical care guidelines [59] show that the use of exogenous hormones carries similar risks for various health outcomes in both transgender and cisgender populations (e.g., risk of breast cancer in TFN people taking estrogen is no greater than for cisgender women taking estrogen for other indications) (57–59). Thus, research such as ours, which shows disparities in health diagnoses rather than the true prevalence of disease in transgender and cisgender people, should not be used to create differential access to medically-necessary treatment for transgender people [59, 87, 88]. Significant advancements in endocrinology and transgender medicine in recent years have helped to ensure the safe and effective delivery of medically-necessary gender-affirming hormones and surgical treatments to transgender people and have led to improvements in the psychological well-being and quality of life for transgender individuals accessing such care [59, 118]. Given that this study focuses on aging transgender individuals, it is possible that some of the health diagnosis disparities observed here are due to their past utilization of older treatment regimens prior to the advent of quality care guidelines, as well as a confluence of stigma-related social, behavioral, and environmental risk factors that shaped access to and use of quality gender-affirming medical care [7]. Our findings, together with prior research, underscore the need for high-quality, prospective research to identify the multitude of risk factors that may contribute to the health diagnosis disparities observed here. Such research should include ongoing, rigorous examinations of the risk and protective effects of gender-affirming hormone therapy and surgery and evaluations into whether transgender individuals are able to access guideline-concordant care. Such data can help to [1] improve healthcare providers’ ability to provide quality care to transgender patients; [2] ensure that transgender individuals receive the requisite information to provide fully informed consent when accessing medically-necessary and psychologically-beneficial medications and procedures to affirm their gender [59, 118, 119]; and [3] inform clinical and policy interventions aimed at improving the health and well-being of transgender people. ## Limitations The study has limitations. First, since transgender beneficiaries were included in our study based on their observed care, we are unable to validate whether individuals we identified as transgender were truly so or whether individuals we characterized as cisgender based on their absence of qualifying care were actually cisgender. Second, since our administrative data did not capture gender identity, we were forced to infer the gender identity of our sample and combine individuals who likely hold nonbinary and binary gender identities in the same category based on their shared use of certain gender-affirming hormones, procedures, or anatomy-specific care. Nonbinary transgender people have been shown to have differential healthcare utilization and risk of various health outcomes than binary transgender people [32, 120], and so the necessary combining of these groups in the present study may have obscured key differences in the probability of being diagnosed with one or more health conditions. Third, transgender and cisgender beneficiaries were included in our study based on their observed care, whereas individuals who did not access relevant care through fee-for-service Medicare at any point during the study period were not included; thus, our sample is unlikely to represent all aging transgender and cisgender Medicare beneficiaries. Fourth, our estimated burden of health diagnoses was based on a methodology that relies on healthcare utilization, and the longer individuals were enrolled in Medicare, the more time they had to engage in healthcare and receive a diagnosis. Since transgender individuals, on average, had more months of enrollment in fee-for-service Medicare than cisgender individuals, we adjusted for months of enrollment. However, undiagnosed or underdiagnosed conditions within any of the gender groups would result in an undercount of the true burden of disease. Further, as noted earlier, transgender individuals forced reliance on the healthcare system to receive gender-affirming care may have led to higher healthcare utilization and more opportunities to receive a health diagnosis relative to cisgender people. Fifth, our analyses may be prone to bias due to unmeasured confounding. Specifically, the effect of social risk factors such as education, housing stability, and income could not be assessed using our data. Still, our findings align with and extend other claims-based, survey, and clinical studies and act as a signal for future research and intervention efforts. ## Conclusion We adapted prior algorithms [3, 13, 33] to identify a large sample of aging transgender Medicare beneficiaries and examine within- and between-group differences in health diagnoses among transgender people and their cisgender counterparts. Extending prior research [3, 5, 12, 13, 30, 31, 33], we observed an elevated burden of health condition diagnoses among TFN or TMN people overall, with the greatest burden observed among TFN people relative to other groups. Our novel methods to identify a transgender sample using Medicare claims data and infer gender may be helpful for future researchers seeking to study the diagnosis of rare conditions, as well as identify transgender subgroups in need of preventive and treatment interventions aimed at reducing morbidity [26, 29, 30, 32] and mortality [42, 121] among aging transgender people in the U.S. ## Data availability statement The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author. ## Ethics statement This study was approved by the Brown University Institutional Review Board. Written informed consent from the participants was not required to participate in this study in accordance with national legislation and institutional requirements. ## Author contributions JH and TS conceived the project and wrote the grants that funded access to the data and supported the analyst's time. JH, TS, AA, JE, LH, KY, and JD were involved in developing the methods to identify the transgender sample and stratify the sample by inferred gender. All authors were involved in refining the algorithm. JH wrote and edited the manuscript and designed and edited the tables and supplementary figures. HV conducted the analyses and populated the tables. KY and LH reviewed the statistical code. GB contributed to the writing of the introduction and formatting of the paper. KY created Figure 1. All authors contributed to the article and approved the submitted version. ## Conflict of interest The authors declare that the research was conducted without any commercial or financial relationships that could be construed as a potential conflict of interest. ## Publisher’s note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. ## Supplementary material The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1102348/full#supplementary-material ## References 1. Reisner SL, White JM, Bradford J, Mimiaga MJ. **Transgender health disparities: Comparing full cohort and nested matched-pair study designs in a community health center**. *LGBT Health* (2014.0) **1**. DOI: 10.1089/lgbt.2014.0009 2. Su D, Irwin JA, Fisher C, Ramos A, Kelley M, Mendoza DAR. **Mental health disparities within the LGBT population: A comparison between transgender and nontransgender individuals**. *Transgend Health* (2016.0) **1** 12-20. DOI: 10.1089/trgh.2015.0001 3. Hughto JMW, Quinn EK, Dunbar M, Rose AJ, Shireman TI, Jasjua GK. **Prevalence and co-occurrence of alcohol, nicotine, and other substance use disorder diagnoses among US transgender and cisgender adults**. *JAMA Netw Open* (2021.0) **4**. DOI: 10.1001/jamanetworkopen.2020.36512 4. Hughes L, Shireman TI, Hughto J. **Privately insured transgender people are at elevated risk for chronic conditions compared with cisgender counterparts**. *Health Aff* (2021.0) **40**. DOI: 10.1377/hlthaff.2021.00546 5. Brown GR, Jones KT. **Mental health and medical health disparities in 5135 transgender veterans receiving healthcare in the veterans health administration: A case–control study**. *LGBT Health* (2016.0) **3**. DOI: 10.1089/lgbt.2015.0058 6. Day JK, Fish JN, Perez-Brumer A, Hatzenbuehler ML, Russell ST. **Transgender youth substance use disparities: Results from a population-based sample**. *J Adolesc Health* (2017.0) **61**. DOI: 10.1016/j.jadohealth.2017.06.024 7. White Hughto JM, Reisner SL, Pachankis JE. **Transgender stigma and health: A critical review of stigma determinants, mechanisms, and interventions**. *Soc Sci Med* (2015.0) **147**. DOI: 10.1016/j.socscimed.2015.11.010 8. Hendricks ML, Testa RJ. **A conceptual framework for clinical work with transgender and gender nonconforming clients: An adaptation of the minority stress model**. *Prof Psychology: Res Practice* (2012.0) **43**. DOI: 10.1037/a0029597 9. Kattari SK, Hasche L. **Differences across age groups in transgender and gender non-conforming people's experiences of health care discrimination, harassment, and victimization**. *J Aging Health* (2016.0) **28** 285-306. DOI: 10.1177/0898264315590228 10. Grant JM, Mottet LA, Tanis LA, Harrison J, Herman JL, Keisling M. *Injustice at every turn: A report of the national transgender discrimination survey* (2011.0) 11. James SE, Herman JL, Rankin S, Keisling M, Mottet L, Anafi M. *The report of the 2015 US transgender survey* (2016.0) 12. Hughto J, Hughes L, Yee K, Downing J, Ellison J, Alpert A. **Improving data-driven methods to identify and categorize transgender individuals by gender in insurance claims data**. *LGBT Health* (2022.0) **9**. DOI: 10.1089/lgbt.2021.0433 13. Jasuja GK, de Groot A, Quinn EK, Ameli O, Hughto JMW, Dunbar M. **Beyond gender identity disorder diagnoses codes: An examination of additional methods to identify transgender individuals in administrative databases**. *Med Care* (2020.0) **58**. DOI: 10.1097/MLR.0000000000001362 14. Blosnich JR, Brown GR, Shipherd P, Jillian C, Kauth M, Piegari RI. **Prevalence of gender identity disorder and suicide risk among transgender veterans utilizing veterans health administration care**. *Am J Public Health* (2013.0) **103**. DOI: 10.2105/AJPH.2013.301507 15. Conron KJ, Scott G, Stowell GS, Landers SJ. **Transgender health in Massachusetts: Results from a household probability sample of adults**. *Am J Public Health* (2012.0) **102**. DOI: 10.2105/AJPH.2011.300315 16. Brown GR, Jones KT. **Racial health disparities in a cohort of 5,135 transgender veterans**. *J Racial Ethnic Health Disparities* (2014.0) **1**. DOI: 10.1007/s40615-014-0032-4 17. Coulter RW, Blosnich JR, Bukowski LA, Herrick A, Siconolfi DE, Stall RD. **Differences in alcohol use and alcohol-related problems between transgender- and nontransgender-identified young adults**. *Drug Alcohol Depend* (2015.0) **154**. DOI: 10.1016/j.drugalcdep.2015.07.006 18. Downing JM, Przedworski JM. **Health of transgender adults in the US, 2014–2016**. *Am J Prev Med* (2018.0) **55**. DOI: 10.1016/j.amepre.2018.04.045 19. Prasad S, Sung B, Aggarwal BB. **Age-associated chronic diseases require age-old medicine: Role of chronic inflammation**. *Prev Med* (2012.0) **54**. DOI: 10.1016/j.ypmed.2011.11.011 20. 20 AAR. Chronic conditions among older Americans. Washington, DC: AARP (2013). Available at: http://assets.aarp.org/rgcenter/health/beyond_50_hcr_conditions.pdf.. *Chronic conditions among older Americans* (2013.0) 21. Fabbre V, Gaveras E. **The manifestation of multilevel stigma in the lived experiences of transgender and gender nonconforming older adults**. *Innovation Aging* (2019.0) **3**. DOI: 10.1093/geroni/igz038.2722 22. Witten TM. **End of life, chronic illness, and trans-identities**. *J Soc Work End Life Palliat Care* (2014.0) **10** 34-58. DOI: 10.1080/15524256.2013.877864 23. Witten TM. **The intersectional challenges of aging and of being a gender non-conforming adult**. *Generations* (2016.0) **40** 63-70 24. Addis S, Davies M, Greene G, MacBride-Stewart S, Shepherd M. **The health, social care and housing needs of lesbian, gay, bisexual and transgender older people: A review of the literature**. *Health Soc Care Community* (2009.0) **17**. DOI: 10.1111/j.1365-2524.2009.00866.x 25. Finkenauer S, Sherratt J, Marlow J, Brodey A. **When injustice gets old: A systematic review of trans aging**. *J Gay Lesbian Soc Services* (2012.0) **24**. DOI: 10.1080/10538720.2012.722497 26. White Hughto JM, Reisner SL. **Social context of depressive distress in aging transgender adults**. *J Appl Gerontol* (2016.0) **37**. DOI: 10.1177/0733464816675819 27. Cook-Daniels L, Munson M. **Sexual violence, elder abuse, and sexuality of transgender adults, age 50+: Results of three surveys**. *J GLBT Family Stud* (2010.0) **6**. DOI: 10.1080/15504281003705238 28. Porter KE, Ronneberg CR, Witten TM. **Religious affiliation and successful aging among transgender older adults: Findings from the trans MetLife survey**. *J Religion Spirituality Aging* (2013.0) **25**. DOI: 10.1080/15528030.2012.739988 29. Progovac AM, Mullin BO, Dunham E, Reisner SL, McDowell A, Roman MJS. **Disparities in suicidality by gender identity among Medicare beneficiaries**. *Am J Prev Med* (2020.0) **58**. DOI: 10.1016/j.amepre.2020.01.004 30. Dragon CN, Guerino P, Ewald E, Laffan AM. **Transgender Medicare beneficiaries and chronic conditions: Exploring fee-for-service claims data**. *LGBT Health* (2017.0) **4**. DOI: 10.1089/lgbt.2016.0208 31. Progovac AM, Cook BL, Mullin BO, McDowell A, Sanchez R MJ, Wang Y. **Identifying gender minority patients’ health and health care needs in administrative claims data**. *Health Aff* (2018.0) **37**. DOI: 10.1377/hlthaff.2017.1295 32. Pharr JR. **Health disparities among lesbian, gay, bisexual, transgender, and nonbinary adults 50 years old and older in the United States**. *LGBT Health* (2021.0) **8**. DOI: 10.1089/lgbt.2021.0009 33. Proctor K, Haffer SC, Ewald E, Hodge C, James CV. **Identifying the transgender population in the Medicare program**. *Trans Health* (2016.0) **1**. DOI: 10.1089/trgh.2016.0031 34. Reisner SL, Greytak EA, Parsons JT, Ybarra ML. **Gender minority social stress in adolescence: Disparities in adolescent bullying and substance use by gender identity**. *J Sex Res* (2015.0) **52**. DOI: 10.1080/00224499.2014.886321 35. Goldsen KIF, Romanelli M, Hoy-Ellis CP, Jung H. **Health, economic and social disparities among transgender women, transgender men and transgender nonbinary adults: Results from a population-based study**. *Prev Med* (2022.0) **156** 106988. DOI: 10.1016/j.ypmed.2022.106988 36. Yee K, Lind BK, Downing J. **Change in gender on record and transgender adults’ mental or behavioral health**. *Am J Prev Med* (2021.0) **62**. DOI: 10.1016/j.amepre.2021.10.016 37. Moon M. **What Medicare has meant to older Americans**. *Health Care Financ Rev* (1996.0) **18** 38. 38 CMS. Chronic conditions data warehouse: Centers for Medicare and Medicaid Services (2022). Available at: https://www2.ccwdata.org/web/guest/condition-categories-chronic.. *Chronic conditions data warehouse: Centers for Medicare and Medicaid Services* (2022.0) 39. Restar A, Jin H, Operario D. **Gender-inclusive and gender-specific approaches in trans health research**. *Transgend Health* (2021.0) **6**. DOI: 10.1089/trgh.2020.0054 40. McEwen BS, Seeman T. **Protective and damaging effects of mediators of stress: Elaborating and testing the concepts of allostasis and allostatic load**. *Ann N Y Acad Sci* (1999.0) **896** 30-47. DOI: 10.1111/j.1749-6632.1999.tb08103.x 41. McEwen BS. **Stress, adaptation, and disease: Allostasis and allostatic load**. *Ann N Y Acad Sci* (1998.0) **840** 33-44. DOI: 10.1111/j.1749-6632.1998.tb09546.x 42. Hughes LD, King WM, Gamarel KE, Geronimus AT, Panagiotou OA, Hughto JMW. **US Black–white differences in mortality risk among transgender and cisgender people in private insurance, 2011–2019**. *Am J Public Health* (2022.0) **112**. DOI: 10.2105/AJPH.2022.306963 43. Juster R-P, McEwen BS, Lupien SJ. **Allostatic load biomarkers of chronic stress and impact on health and cognition**. *Neurosci Biobehav Rev* (2010.0) **35** 2-16. DOI: 10.1016/j.neubiorev.2009.10.002 44. Duru OK, Harawa NT, Kermah D, Norris KC. **Allostatic load burden and racial disparities in mortality**. *J Natl Med Assoc* (2012.0) **104** 89-95. DOI: 10.1016/S0027-9684(15)30120-6 45. **Key statistics for breast cancer in men** 46. de Blok CJ, Wiepjes CM, Nota NM, van Engelen K, Adank MA, Dreijerink KM. **Breast cancer risk in transgender people receiving hormone treatment: Nationwide cohort study in the Netherlands**. *BMJ* (2019.0) **365** l1652. DOI: 10.1136/bmj.l1652 47. Quinn VP, Nash R, Hunkeler E, Contreras R, Cromwell L, Becerra-Culqui TA. **Cohort profile: Study of transition, outcomes and gender (STRONG) to assess health status of transgender people**. *BMJ Open* (2017.0) **7** e018121. DOI: 10.1136/bmjopen-2017-018121 48. Brown GR, Jones KT. **Incidence of breast cancer in a cohort of 5,135 transgender veterans**. *Breast Cancer Res Treat* (2015.0) **149**. DOI: 10.1007/s10549-014-3213-2 49. Brown GR. **Breast cancer in transgender veterans: A ten-case series**. *LGBT Health* (2015.0) **2** 77-80. DOI: 10.1089/lgbt.2014.0123 50. Braun H, Nash R, Tangpricha V, Brockman J, Ward K, Goodman M. **Cancer in transgender people: Evidence and methodological considerations**. *Epidemiol Rev* (2017.0) **39** 93-107. DOI: 10.1093/epirev/mxw003 51. Reiche EMV, Nunes SOV, Morimoto HK. **Stress, depression, the immune system, and cancer**. *Lancet Oncol* (2004.0) **5**. DOI: 10.1016/S1470-2045(04)01597-9 52. Moreno-Smith M, Lutgendorf SK, Sood AK. **Impact of stress on cancer metastasis**. *Future Oncol* (2010.0) **6**. DOI: 10.2217/fon.10.142 53. Weyers S, Villeirs G, Vanherreweghe E, Verstraelen H, Monstrey S, Van den Broecke R. **Mammography and breast sonography in transsexual women**. *Eur J Radiol* (2010.0) **74**. DOI: 10.1016/j.ejrad.2009.03.018 54. Brown SB, Hankinson SE. **Endogenous estrogens and the risk of breast, endometrial, and ovarian cancers**. *Steroids* (2015.0) **99** 8-10. DOI: 10.1016/j.steroids.2014.12.013 55. Chia K, O’Brien M, Brown M, Lim E. **Targeting the androgen receptor in breast cancer**. *Curr Oncol Rep* (2015.0) **17** 1-6. DOI: 10.1007/s11912-014-0427-8 56. McCormack VA, dos Santos Silva I. **Breast density and parenchymal patterns as markers of breast cancer risk: A meta-analysis**. *Cancer Epidemiol Biomark Prev* (2006.0) **15**. DOI: 10.1158/1055-9965.EPI-06-0034 57. D’Alonzo M, Bounous VE, Villa M, Biglia N. **Current evidence of the oncological benefit-risk profile of hormone replacement therapy**. *Medicina (Mex)* (2019.0) **55** 58. Gambacciani M, Cagnacci A, Lello S. **Hormone replacement therapy and prevention of chronic conditions**. *Climacteric* (2019.0) **22**. DOI: 10.1080/13697137.2018.1551347 59. Coleman E, Radix A, Bouman W, Brown G, De Vries A, Deutsch M. **Standards of care for the health of transgender and gender diverse people, version 8**. *Int J Transgender Health* (2022.0) **23**. DOI: 10.1080/26895269.2022.2100644 60. 60 CDC. Rheumatoid arthritis (RA). Atlanta, GA: Centers for Disease Control and Prevention (2022). Available at: https://www.cdc.gov/arthritis/basics/rheumatoid-arthritis.html#risk.. *Rheumatoid arthritis (RA)* (2022.0) 61. 61 CDC. Arthritis risk factors. Atlanta, GA: Centers for Disease Control and Prevention (2022). Available at: https://www.cdc.gov/arthritis/basics/risk-factors.htm.. *Arthritis risk factors* (2022.0) 62. Pouresmaeili F, Kamalidehghan B, Kamarehei M, Goh YM. **A comprehensive overview on osteoporosis and its risk factors**. *Ther Clin Risk Manage* (2018.0) **14** 2029. DOI: 10.2147/TCRM.S138000 63. Jiang P, Li H, Li X. **Diabetes mellitus risk factors in rheumatoid arthritis: A systematic review and meta-analysis**. *Clin Exp Rheumatol* (2015.0) **33** 64. Caceres BA, Jackman KB, Edmondson D, Bockting WO. **Assessing gender identity differences in cardiovascular disease in US adults: An analysis of data from the 2014–2017 BRFSS**. *J Behav Med* (2020.0) **43**. DOI: 10.1007/s10865-019-00102-8 65. Alzahrani T, Nguyen T, Ryan A, Dwairy A, McCaffrey J, Yunus R. **Cardiovascular disease risk factors and myocardial infarction in the transgender population**. *Circ Cardiovasc Qual Outcomes* (2019.0) **12** e005597. DOI: 10.1161/CIRCOUTCOMES.119.005597 66. Kyinn M, Banks K, Leemaqz S, Sarkodie E, Goldstein D, Irwig M. **Weight gain and obesity rates in transgender and gender-diverse adults before and during hormone therapy**. *Int J Obes* (2021.0) **45**. DOI: 10.1038/s41366-021-00935-x 67. Deutsch MB. *Guidelines for the primary and gender-affirming care of transgender and gender nonbinary people* (2016.0) 68. Cosman F, de Beur SJ, LeBoff M, Lewiecki E, Tanner B, Randall S. **Clinician’s guide to prevention and treatment of osteoporosis**. *Osteoporosis Int* (2014.0) **25**. DOI: 10.1007/s00198-014-2794-2 69. Mehler PS, Cleary BS, Gaudiani JL. **Osteoporosis in anorexia nervosa**. *Eating Disord* (2011.0) **19** 194-202. DOI: 10.1080/10640266.2011.551636 70. Jones BA, Arcelus J, Bouman WP, Haycraft E. **Barriers and facilitators of physical activity and sport participation among young transgender adults who are medically transitioning**. *Int J Transgend* (2017.0) **18**. DOI: 10.1080/15532739.2017.1293581 71. Jones BA, Haycraft E, Bouman WP, Arcelus J. **The levels and predictors of physical activity engagement within the treatment-seeking transgender population: A matched control study**. *J Phys Act Health* (2018.0) **15** 99-107. DOI: 10.1123/jpah.2017-0298 72. Oliveira J, Frontini R, Jacinto M, Antunes R. **Barriers and motives for physical activity and sports practice among trans people: A systematic review**. *Sustainability* (2022.0) **14**. DOI: 10.3390/su14095295 73. Lawless MH, Harrison KA, Grandits GA, Eberly LE, Allen SS. **Perceived stress and smoking-related behaviors and symptomatology in male and female smokers**. *Addict Behav* (2015.0) **51**. DOI: 10.1016/j.addbeh.2015.07.011 74. Almeida M, Laurent MR, Dubois V, Claessens F, O'Brien CA, Bouillon R. **Estrogens and androgens in skeletal physiology and pathophysiology**. *Physiol Rev* (2017.0) **97**. DOI: 10.1152/physrev.00033.2015 75. Manolagas SC. **Birth and death of bone cells: Basic regulatory mechanisms and implications for the pathogenesis and treatment of osteoporosis**. *Endocr Rev* (2000.0) **21**. DOI: 10.1210/edrv.21.2.0395 76. Irwig MS. **Testosterone therapy for transgender men**. *Lancet Diabetes Endocrinol* (2017.0) **5**. DOI: 10.1016/S2213-8587(16)00036-X 77. Singh-Ospina N, Maraka S, Rodriguez-Gutierrez R, Davidge-Pitts C, Nippoldt TB, Prokop LJ. **Effect of sex steroids on the bone health of transgender individuals: A systematic review and meta-analysis**. *J Clin Endocrinol Metab* (2017.0) **102**. DOI: 10.1210/jc.2017-01642 78. Wiepjes CM, Vlot MC, Klaver M, Nota NM, de Blok CJ, de Jongh RT. **Bone mineral density increases in trans persons after 1 year of hormonal treatment: A multicenter prospective observational study**. *J Bone Mineral Res* (2017.0) **32**. DOI: 10.1002/jbmr.3102 79. Motta G, Marinelli L, Barale M, Brustio PR, Manieri C, Ghigo E. **Fracture risk assessment in an Italian group of transgender women after gender-confirming surgery**. *J Bone Mineral Metab* (2020.0) **38**. DOI: 10.1007/s00774-020-01127-9 80. Klink D, Caris M, Heijboer A, van Trotsenburg M, Rotteveel J. **Bone mass in young adulthood following gonadotropin-releasing hormone analog treatment and cross-sex hormone treatment in adolescents with gender dysphoria**. *J Clin Endocrinol Metab* (2015.0) **100**. DOI: 10.1210/jc.2014-2439 81. Rich AJ, Scheim AI, Koehoorn M, Poteat T. **Non-HIV chronic disease burden among transgender populations globally: A systematic review and narrative synthesis**. *Prev Med Rep* (2020.0) **20** 101259. DOI: 10.1016/j.pmedr.2020.101259 82. Salk RH, Hyde JS, Abramson LY. **Gender differences in depression in representative national samples: Meta-analyses of diagnoses and symptoms**. *Psychol Bull* (2017.0) **143** 783-822. DOI: 10.1037/bul0000102 83. Eid RS, Gobinath AR, Galea LA. **Sex differences in depression: Insights from clinical and preclinical studies**. *Prog Neurobiol* (2019.0) **176** 86-102. DOI: 10.1016/j.pneurobio.2019.01.006 84. Guo Y, Li Q, Yang X, Jaffee MS, Wu Y, Wang F. **Prevalence of Alzheimer’s and related dementia diseases and risk factors among transgender adults, Florida, 2012−2020**. *Am J Public Health* (2022.0) **112**. DOI: 10.2105/AJPH.2022.306720 85. Barr SM, Roberts D, Thakkar KN. **Psychosis in transgender and gender non-conforming individuals: A review of the literature and a call for more research**. *Psychiatry Res* (2021.0) **306** 114272. DOI: 10.1016/j.psychres.2021.114272 86. Coleman E, Bockting W, Botzer M, Cohen-Kettenis P, DeCuypere G, Feldman J. **Standards of care for the health of transsexual, transgender, and gender-nonconforming people, version 7**. *Int J Transgend* (2012.0) **13** 165-232. DOI: 10.1080/15532739.2011.700873 87. Ashley F. **Gatekeeping hormone replacement therapy for transgender patients is dehumanising**. *J Med Ethics* (2019.0) **45**. DOI: 10.1136/medethics-2018-105293 88. Schulz SL. **The informed consent model of transgender care: An alternative to the diagnosis of gender dysphoria**. *J Humanistic Psychol* (2018.0) **58** 72-92. DOI: 10.1177/0022167817745217 89. Bränström R, Stormbom I, Bergendal M, Pachankis JE. **Transgender-based disparities in suicidality: A population-based study of key predictions from four theoretical models**. *Suicide Life Threat Behav* (2022.0) **52**. DOI: 10.1111/sltb.12830 90. Brennan G. **Stress vulnerability model and serious mental illness**. *Working with serious mental illness: A manual for clinical practice* (2000.0) 29-39 91. Fleming MP, Martin CR, Martin C, Preedy V, Patel V. **Trauma exposure, schizophrenia symptoms, and the stress vulnerability model**. *Comprehensive guide to post-traumatic stress disorders* (2015.0). DOI: 10.1007/978-3-319-08613-2_40-1 92. Greenberg MS, Tanev K, Marin MF, Pitman RK. **Stress, PTSD, and dementia**. *Alzheimers Dement* (2014.0) **10**. DOI: 10.1016/j.jalz.2014.04.008 93. Hammen C. **Stress and depression**. *Annu Rev Clin Psychology* (2005.0) **1**. DOI: 10.1146/annurev.clinpsy.1.102803.143938 94. McEwen BS. **Allostasis and allostatic load: Implications for neuropsychopharmacology**. *Neuropsychopharmacology* (2000.0) **22**. DOI: 10.1016/S0893-133X(99)00129-3 95. DuBois LZ, Juster R-P. **Lived experience and allostatic load among transmasculine people living in the United States**. *Psychoneuroendocrinology* (2022.0) **143** 105849. DOI: 10.1016/j.psyneuen.2022.105849 96. Lambrou NH, Gleason CE, Obedin-Maliver J, Lunn MR, Flentje A, Lubensky ME. **Subjective cognitive decline associated with discrimination in medical settings among transgender and nonbinary older adults**. *Int J Environ Res Public Health* (2022.0) **19** 9168. DOI: 10.3390/ijerph19159168 97. Valente PK, Dworkin JD, Dolezal C, Singh AA, LeBlanc AJ, Bockting WO. **Prospective relationships between stigma, mental health, and resilience in a multi-city cohort of transgender and nonbinary individuals in the United States, 2016–2019**. *Soc Psychiatry Psychiatr Epidemiol* (2022.0) **57**. DOI: 10.1007/s00127-022-02270-6 98. Stutterheim SE, van Dijk M, Wang H, Jonas KJ. **The worldwide burden of HIV in transgender individuals: An updated systematic review and meta-analysis**. *PLoS One* (2021.0) **16**. DOI: 10.1371/journal.pone.0260063 99. Becasen JS, Denard CL, Mullins MM, Higa DH, Sipe TA. **Estimating the prevalence of HIV and sexual behaviors among the US transgender population: A systematic review and meta-analysis, 2006–2017**. *Am J Public Health* (2019.0) **109**. DOI: 10.2105/AJPH.2018.304727 100. Poteat T, Scheim A, Xavier J, Reisner S, Baral S. **Global epidemiology of HIV infection and related syndemics affecting transgender people**. *J Acquir Immune Defic Syndr* (2016.0) **72**. DOI: 10.1097/QAI.0000000000001087 101. Sevelius JM. **Gender affirmation: A framework for conceptualizing risk behavior among transgender women of color**. *Sex Roles* (2013.0) **68**. DOI: 10.1007/s11199-012-0216-5 102. Melendez RM, Pinto R. **‘It's really a hard life’: Love, gender and HIV risk among male-to-female transgender persons**. *Cult Health Sex* (2007.0) **9**. DOI: 10.1080/13691050601065909 103. Blair KL, Hoskin RA. **Transgender exclusion from the world of dating: Patterns of acceptance and rejection of hypothetical trans dating partners as a function of sexual and gender identity**. *J Soc Pers Relationships* (2019.0) **36**. DOI: 10.1177/0265407518779139 104. Gamarel KE, Reisner SL, Laurenceau J-P, Nemoto T, Operario D. **Gender minority stress, mental health, and relationship quality: A dyadic investigation of transgender women and their cisgender male partners**. *J Fam Psychol* (2014.0) **28**. DOI: 10.1037/a0037171 105. Chandler CJ, Creasy SL, Adams BJ, Eaton LA, Bukowski LA, Egan JE. **Characterizing biomedical HIV prevention awareness and use among black transgender women in the United States**. *AIDS Behavior* (2021.0) **25**. DOI: 10.1007/s10461-021-03189-w 106. Wilson E, Rapues J, Jin H, Raymond HF. **The use and correlates of illicit silicone or “fillers” in a population-based sample of transwomen, San Francisco, 2013**. *J Sexual Med* (2014.0) **11**. DOI: 10.1111/jsm.12558 107. Benotsch EG, Zimmerman RS, Cathers L, Pierce J, McNulty S, Heck T. **Non-medical use of prescription drugs and HIV risk behaviour in transgender women in the mid-Atlantic region of the United States**. *Int J STD AIDS* (2016.0) **27**. DOI: 10.1177/0956462415595319 108. 108 SAMHSA. Key substance use and mental health indicators in the United States: Results from the 2017 National Survey on Drug Use and Health. Rockville, MD: Center for Behavioral Health Statistics and Quality (2018). Available at: https://www.samhsa.gov/data/.. *Key substance use and mental health indicators in the United States: Results from the 2017 National Survey on Drug Use and Health* (2018.0) 109. Wolfe HL, Biello KB, Reisner SL, Mimiaga MJ, Cahill SR, Hughto JMW. **Transgender-related discrimination and substance use, substance use disorder diagnosis and treatment history among transgender adults**. *Drug Alcohol Depend* (2021.0) **223** 108711. DOI: 10.1016/j.drugalcdep.2021.108711 110. McHugh RK, Votaw VR, Sugarman DE, Greenfield SF. **Sex and gender differences in substance use disorders**. *Clin Psychol Rev* (2018.0) **66** 12-23. DOI: 10.1016/j.cpr.2017.10.012 111. 111 NIDA. Sex and gender differences in substance use (2022). Available at: https://nida.nih.gov/publications/research-reports/substance-use-in-women/sex-gender-differences-in-substance-use.. *Sex and gender differences in substance use* (2022.0) 112. Adrian H, Collins A, Operario D, Wolfe H, Fernández Y, Briody V. **Providing gender-affirmative substance use disorder services for transgender and gender diverse people who use drugs in Rhode Island: A qualitative study**. (2022.0) 113. Lombardi E. **Substance use treatment experiences of transgender/transsexual men and women**. *J LGBT Health Res* (2007.0) **3** 37-47. DOI: 10.1300/J463v03n02_05 114. Lyons T, Shannon K, Pierre L, Small W, Krüsi A, Kerr T. **A qualitative study of transgender individuals’ experiences in residential addiction treatment settings: Stigma and inclusivity**. *Subst Abuse Treat Prev Policy* (2015.0) **10**. DOI: 10.1186/s13011-015-0015-4 115. Budge SL, Katz-Wise SL, Tebbe EN, Howard KA, Schneider CL, Rodriguez A. **Transgender emotional and coping processes: Facilitative and avoidant coping throughout gender transitioning**. *Couns Psychol* (2013.0) **41**. DOI: 10.1177/0011000011432753 116. White Hughto J, Pachankis J, Willie T, Reisner S. **Victimization and depressive symptomology in transgender adults: The mediating role of avoidant coping**. *J Couns Psychol* (2017.0) **64** 41-51. DOI: 10.1037/cou0000184 117. Glynn TR, van den Berg JJ. **A systematic review of interventions to reduce problematic substance use among transgender individuals: A call to action**. *Transgend Health* (2017.0) **2** 45-59. DOI: 10.1089/trgh.2016.0037 118. White Hughto J, Reisner S. **A systematic review of the effects of hormone therapy on psychological functioning and quality of life in transgender individuals**. *Transgend Health* (2016.0) **1** 21-31. DOI: 10.1089/trgh.2015.0008 119. Murad MH, Elamin MB, Garcia MZ, Mullan RJ, Murad A, Erwin PJ. **Hormonal therapy and sex reassignment: A systematic review and meta-analysis of quality of life and psychosocial outcomes**. *Clin Endocrinol (Oxf)* (2010.0) **72**. DOI: 10.1111/j.1365-2265.2009.03625.x 120. Reisner SL, Hughto JM. **Comparing the health of non-binary and binary transgender adults in a statewide non-probability sample**. *PLoS One* (2019.0) **14** e0221583. DOI: 10.1371/journal.pone.0221583 121. Boyer TL, Youk AO, Haas AP, Brown GR, Shipherd JC, Kauth MR. **Suicide, homicide, and all-cause mortality among transgender and cisgender patients in the Veterans Health Administration**. *LGBT Health* (2021.0) **8**. DOI: 10.1089/lgbt.2020.0235
--- title: Resolvin D1/N-formyl peptide receptor 2 ameliorates paclitaxel-induced neuropathic pain through the activation of IL-10/Nrf2/HO-1 pathway in mice authors: - Cun-Jin Su - Jiang-Tao Zhang - Feng-Lun Zhao - De-Lai Xu - Jie Pan - Tong Liu journal: Frontiers in Immunology year: 2023 pmcid: PMC10040838 doi: 10.3389/fimmu.2023.1091753 license: CC BY 4.0 --- # Resolvin D1/N-formyl peptide receptor 2 ameliorates paclitaxel-induced neuropathic pain through the activation of IL-10/Nrf2/HO-1 pathway in mice ## Abstract ### Introduction Paclitaxel is a chemotherapy drug that is commonly used to treat cancer, but it can cause paclitaxel-induced neuropathic pain (PINP) as a side effect. Resolvin D1 (RvD1) has been shown to be effective in promoting the resolution of inflammation and chronic pain. In this study, we evaluated the effects of RvD1 on PINP and its underlying mechanisms in mice. ### Methods Behavioral analysis was used to assess the establishment of the PINP mouse model and to test the effects of RvD1 or other formulations on mouse pain behavior. Quantitative real-time polymerase chain reaction analysis was employed to detect the impact of RvD1 on $\frac{12}{15}$ Lox, FPR2, and neuroinflammation in PTX-induced DRG neurons. Western blot analysis was used to examine the effects of RvD1 on FPR2, Nrf2, and HO-1 expression in DRG induced by PTX. TUNEL staining was used to detect the apoptosis of DRG neurons induced by BMDM conditioned medium. H2DCF-DA staining was used to detect the reactive oxygen species level of DRG neurons in the presence of PTX or RvD1+PTX treated BMDMs CM. ### Results Expression of $\frac{12}{15}$-Lox was decreased in the sciatic nerve and DRG of mice with PINP, suggesting a potential involvement of RvD1 in the resolution of PINP. Intraperitoneal injection of RvD1 promoted pain resolution of PINP in mice. Intrathecal injection of PTX-treated BMDMs induced mechanical pain hypersensitivity in naïve mice, while pretreatment of RvD1 in BMDMs prevented it. Macrophage infiltration increased in the DRGs of PINP mice, but it was not affected by RvD1 treatment. RvD1 increased IL-10 expression in the DRGs and macrophages, while IL-10 neutralizing antibody abolished the analgesic effect of RvD1 on PINP. The effects of RvD1 in promoting IL-10 production were also inhibited by N-formyl peptide receptor 2 (FPR2) antagonist. The primary cultured DRG neurons apoptosis increased after stimulation with condition medium of PTX-treated BMDMs, but decreased after pretreatment with RvD1 in BMDMs. Finally, Nrf2-HO1 signaling was additionally activated in DRG neurons after stimulation with condition medium of RvD1+PTX-treated BMDMs, but these effects were abolished by FPR2 blocker or IL-10 neutralizing antibody. ### Discussion In conclusion, this study provides evidence that RvD1 may be a potential therapeutic strategy for the clinical treatment of PINP. RvD1/FPR2 upregulates IL-10 in macrophages under PINP condition, and then IL-10 activates the Nrf2- HO1 pathway in DRG neurons, relieve neuronal damage and PINP. ## Introduction Paclitaxel (PTX) is widely used chemotherapy for the treatment of a range of malignancies, including breast, ovarian and lung cancer [1]. However, PTX often produces peripheral neuropathy and neuropathic pain in the distal extremities, persisting for months or years [2]. PTX-induced peripheral neuropathic pain (PINP) is a serious dose-limiting adverse effect during PTX treatment in cancer patients [3]. PINP greatly decreased the life quality of patients during and after chemotherapy. Because of PINP, life-saving cancer treatment usually has to be discontinued. To date, there are no Food and Drug Administration-approved drugs to prevent or treat PINP. Accordingly, there is an urgent need to develop novel therapies for preventing or treating PINP, improving both cancer treatment and life quality of afflicted patients. Mounting evidences indicate that macrophages contribute to initiation, maintenance, and resolution of chronic pain through neuro-immune interactions [4]. Under neuronal injury conditions, the injury neurons and resident macrophages are able to produce pro-inflammatory factors, chemokines and damage-associated molecular patterns (DAMPs), form a local inflammatory microenvironment, and further recruit circulating macrophages to infiltrate peripheral nerve tissue [5, 6]. The dorsal root ganglion (DRG), a structure that transmits pain signals from the periphery to the central nervous system, plays an important role in the development and maintenance of chronic pain [7]. A marked infiltration of macrophages was demonstrated in the DRGs of peripheral neuropathic pain model animals induced by multiple chemotherapeutic agents, including paclitaxel, platinum, and vincristine [8]. Classically, macrophages are subdivided into two major phenotypes, M1-like and M2-like. M1-like macrophages are pro-inflammatory, and M2-like macrophages are anti-inflammatory. Long-term activation of M1-like macrophages is considered to be important mechanisms for the chronic neuropathic pain, possible by activating the TRP channels and mitogen-activated protein kinase (MAPK) signaling [9]. M2-like macrophages can secrete a variety of anti-inflammatory factors, including IL-4, IL-10, TGF-β, inhibit local inflammatory response, and alleviate neuropathic pain [10]. It is worth noting that inhibition of the inflammatory response of macrophages or deleting macrophages with drugs can relieve pain in animal models of peripheral neuropathic pain [11, 12]. Collectively, these studies suggest that targeting macrophages may be an important strategy for alleviating neuropathic pain [13]. Resolvins are endogenous mediators that promote the inflammation resolution and are able to return the inflamed tissues to homeostasis [14]. Resolvin D1 (RvD1), a member of resolvins family, is mainly detected in macrophages and neutrophils [15]. RvD1 can enhance the phagocytosis of macrophages, enhance the clearance of aging cells, and promote macrophages to produce IL-10 [16, 17]. In our previous study, we demonstrated that RvD1 or RvE1 attenuated formalin-induced inflammatory pain in mice [18]. Although RvD1 is well-known to regulate macrophage function, it is unclear whether it can alleviate PINP through acting on macrophages. In the present study, we tested the hypothesis that RvD1 attenuates PINP possible through the regulation of macrophages in the DRGs. ## Animals Male ICR mice (8-10 weeks of age) from SLAC Company (Shanghai, China) were used in this study. Mice were housed four per cage with free access to food and water. All mice were kept in controlled room temperature (22 ± 2°C) and humidity (60-$80\%$). The illumination maintained on a 12h/12h light/dark cycle (lights on from 6:00 AM to 6:00 PM). Mice were numbered according to body weight, and then grouped by looking up the random number table. The number of mice in each group was shown in the corresponding figure legends. All animal experiments were blind to the operators during the allocation, the conduct of the experiment, the outcome assessment. All animal experiments and procedures were performed in accordance with the guidelines recommended by the International Association for the Study of Pain, and were approved by the University Committee on Animal Care of Soochow University. ## Paclitaxel-induced neuropathic pain model Paclitaxel (PTX) solution (Yangtze River Pharmaceutical Group, China) was diluted with saline before treatment. PTX was injected intraperiotoneally (i.p) in mice at a dose of 2 mg/kg on days 0, 2, 4 and 6 to generate PTX-induced neuropathic pain. The final PTX cumulative dose was 8 mg/kg per mouse. Mice in the control group received an intraperitoneal injection of saline as vehicle. ## Drug administration Mice received intraperitoneal (i.p.) injection of RvD1 (Cayman Chmeical, USA) at a dose of 5 μg/kg for 14 days or for a single injection. ## Intrathecal injection Mice were under a brief anesthesia with isoflurane, then we delivered drugs into cerebral spinal fluid (CSF) space around lumbosacral spinal cord through intrathecal (i.t.) injection. Spinal cord puncture was made with a 30G needle between the lumbar L5 and L6 to inject the IL-10 neutralizing antibody (Sigma, I5145, 10ug/10ul) to the CSF. A brisk tail-flick after the needle entry into subarachnoid space signed a successful spinal puncture. ## Behavioral tests Behavioral tests were performed in a quiet and temperature-controlled room between 9:00 AM and 5:00 PM, and were carried out by an operator blinded to drug treatments. Mechanical allodynia: *Mechanical allodynia* was assessed using the “up-and-down” methods as previously described [19]. Mice were placed beneath perspex boxes (10×10×7cm) set upon elevated wire mesh stands and acclimated for 30 min. Von Frey filament (0.008-1.4g) was applied to the mid-plantar area (avoiding the base of tori) with enough pressure to bend the hair. The filament was held for 5s. If the paw did not lift after 5s, an increased weight filament would be used next. Whereas, a subsequently weaker filament would be used if the paw lifted after filament stimuli. The $50\%$ mechanical paw withdrawal threshold was calculated as described previously [19]. The paw mechanical withdraw thresholds were recorded in grams (g). Thermal hyperalgesia: The thermal hyperalgesia was measured using tail flick test. Tail flick test was carried out by immersing the mouse tail in water (5 cm from the tip) maintained at 48°C.Tail flick latency time was measured as the time from the heat exposure to the withdraw of the tail. The cut off time for tail flick test was 15s. This test was carried out 3 times per mouse, and the average value was taken as latency time. Cold hyperalgesia test: *Cold hyperalgesia* was analyzed by the acetone stimulation test. Mice were placed into perspex boxes (10×10×7cm) with a wire mesh floor. Mice were allowed to habituate for 30min prior to the test. A drop (50 μl) of acetone was placed onto the center of planta skin. The responses to acetone were recorded in the following 30s after acetone application. Responses to acetone are divided into 4 grades: 0, no response; 1, quick withdraw, flick or stamp of the paw; 2, prolonged withdraw or repeat flicking of the paw; 3, repeated flicking of the paw with licking directed at the ventral side of the paw. Acetone was applied to each paw 3 times at a 10-15 min intervals, and the average score was calculated. ## Cell culture Bone marrow-derived macrophages (BMDMs) were isolated from the tibia and femoral bone marrow of 8-week-old male mice. BMDMs were cultured in Dulbecco’s modified Eagle’s medium (DMEM) containing $10\%$ fetal bovine serum, 20 ng/ml recombinant murine GM-CSF (novoprotein) and $1\%$ penicillin/streptomycin. The DRG tissue of 2-3 week old mice was isolated, digested with $0.15\%$ collagenase and $0.25\%$ trypsin until there was no visible tissue mass. After being filtered by 70 μm filter membrane, centrifuged and cultured in the neurobasal medium containing $10\%$ serum, $2\%$ B27 and $1\%$ penicillin/streptomycin. Cells were cultured as a monolayer under $5\%$ CO2 in a humidified incubator at 37°C. ## Quantitative real-time polymerase chain reaction Total RNA was extracted using Trizol Reagent (Invitrogen, Carlsbad, CA) according to the protocol supplied by the manufacturer. The RNA amount and quality were assessed by Microplate Reader (Thermo). RNA (500 ng) was converted to cDNA using RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, Waltham, USA). Real-time PCR was conducted using SYBR Green PCR Master Mix (Selleck, China) on Opticon real-time PCR Detection System (Applied Biosystems 7500, Grand Island, NY). Relative fold of differences in expression were calculated using the 2(-Delta Delta C(t)) method after normalization to GAPDH expression. The following primers for mouse were synthesized by Genewiz: GAPDH (forward: 5’-GAAGGTCGGTGTGAACGGAT-3’; reverse: 5’-AATCTCCACTTTGCCACTGC-3’), FPR2 (forward: 5’-GTCAAGATCAACAGAAGAAACC-3’; reverse: 5’-GGGCTCTCTCAAGACTATAAGG-3’), $\frac{12}{15}$-Lox (forward: 5’-GCGACGCTGC CCAATCCTAATC-3’; reverse: 5’-CATATGGCCACGCTGTTTTCTACC-3’), IL-4 (forward: 5’-ATGGATGTGCCAAACGTCCT-3’; reverse: 5’-CATATGGCCACGCT GTTTTCTACC-3’), IL-10 (forward: 5’-GGACTTTAAGGGTTACTTGGGTTGCC-3’; reverse: 5’-CATTTTGATCATCATGTATGCTTCT-3’), IL-1β (forward: 5’-TGTAA TGAAAGACGGCACACC-3’; reverse: 5’-TCTTCTTTGGGTATTGCTTGG-3’), TNF-α (forward: 5’-AGCCGATGGGTTGTACCTTG-3’; reverse: 5’-TTGGGCAGAT TGACCTCAGC-3’), TGF-β (forward: 5’-TGAACCAAGGAGACGGAATACAGG-3’; reverse: 5’-TACTGTGTGTCCAGGCTCCAAATG-3’). ## Western blotting Cells were washed 3 times with ice-cold PBS and lysed in RIPA supplemented with phenylmethylsulfonyl fluoride on ice for 30min. The samples were centrifuged at 12000rpm for 25min at 4°C, then the supernatant was collected. The protein concentration was assessed using a BCA protein assay (Beyotime, China). The protein samples were separated with $10\%$ resolving gel and electroblotted onto a polyvinylidene fluoride (PVDF) membrane. The membranes were blocked with $5\%$ non-fat milk for 1 h at room temperature, then incubated overnight at 4°C with primary antibodies (FPR2, Thermo Fisher, 720293; Nrf2, Proteintech, 16396-1-AP; HO-1, Proteintech, 10701-1-AP; Tubulin, Proteintech, 11224-1-AP; H3, Beyotime, AF7014), followed by HRP-conjugated secondary antibodies for 1 h at room temperature. Immunoreactive bands were detected by enhanced chemiluminescence (ECL) reagent using a chemiluminescence instrument. Image J was used to measure the grey intensity of the specific bands. ## Immunohistochemistry Mice were anesthetized using pentobarbital, and perfused through the ascending aorta with $4\%$ paraformaldehyde. L4 and L5 DRGs were removed and postfixed in $4\%$ paraformaldehyde overnight followed by dehydrating with $30\%$ sucrose solution in PBS for 3 days. Cryostat sections (15 μm) were cut and stained for IHC with primary antibody (CD68, Abcam) overnight at 4°C. Then the sections were incubated with goat anti-Rat IgG - H&L (Alexa Fluor 488) for 1 h at RT. ## TUNEL staining Neuronal apoptosis was detected by the TdT (terminal deoxyribonucleotidyl transferase)-mediated dUTP nick-end labeling (TUNEL) assay (Beyotime, China). Briefly, primary DRG neurons were washed twice with PBS, fixed with $4\%$ paraformaldehyde for 15 min, and then washed twice with PBS. TUNEL staining solution was prepared according to the instruction, dropped on the cells, reacted at 37°C for 1 hour, washed with PBS and then treated with DAPI staining solution for 3 minutes, washed with PBS for 3 times, and photographed with a microscope. ## Detection of intracellular reactive oxygen species The intracellular ROS of primary DRG neurons was detected by H2DCF-DA following the reference. After treatment, the DRG neurons were incubated with 25 mM H2DCF-DA for 30 min. After washing thrice with cold PBS, the cell images were acquired immediately via fluorescence microscopy. ## Statistical analysis The data were presented as the mean ± standard error of the mean (SEM). Student’s t-test was used to compare two groups, and one-way analysis of variance (ANOVA) was used to compare multiple samples. All statistical analyses were performed using the SPSS 13.0 software. $P \leq 0.05$ is considered as statistically significant. ## PTX treatment induces PINP and impairs RvD1 synthesis in mice PTX is a commonly used chemotherapy drug, but PINP is a common dose-limiting adverse drug reaction. PINP commonly exhibits an increased sensitivity to mechanical, heat and cold stimulation. Altered nociceptive thresholds to mechanical, heat and cold stimulation are hallmarks of the development of PINP. Adult male ICR mice received 4 injections of PTX (2 mg/kg, every other day, i.p.) for a total cumulative dose of 8 mg/kg. We assessed mechanical sensitivity in PTX-treated mice using von Frey filaments. PTX significantly reduced mechanical threshold to elicit a paw withdraw response. Increased mechanical sensitivity developed within 2 days post-initiation of PTX administration and was sustained for about 2 weeks (Figures 1A–C). The peak effects of mechanical hypersensitivity occurred between days 5 and 12 (Figure 1A). Thermal pain was detected by tail flick test at 48°C hot water. PTX-treated mice developed a slight and transient hypoalgesia to heat stimuli, which was observed on day 2 and 5 (Figures 1D–F). In contrast, PTX-treated mice were sensitive to cold stimuli. The acetone cold pain induced by PTX started on the 7th day and continued until the 15th day (Figures 1G–I). Together, these data indicated that mechanical and cold allodynia were successfully induced in PTX-treated mice (but not thermal hyperalgesia), and this phenotype was consistent with other reports [20]. **Figure 1:** *Paclitaxel induced peripheral neuropathic pain in mice, and decreased 12/15-Lox expression. PTX was injected intraperiotoneally (i.p) in mice at a dose of 2 mg/kg on days 0, 2, 4 and 6 to generate PTX-induced neuropathic pain. Nociceptive thresholds to mechanical (A), heat (D) and cold (G) stimulation of the hindpaws of mice (**** means ** Day 2 and Day 3 vs VEH, ****** means *** Day 5, 7, 10, 12 vs VEH). Individual mouse time course plots showing changes in mechanical (B), heat (E), and cold (H). Area under the time course cures (AUC) of mechanical (C), heat (F), and cold (I) thresholds. 12/15-Lox (J) and FPR2 (K) mRNAs in DRG, SN, skin tissues were determined by qPCR. *p < 0.05, **p < 0.01, ***p < 0.001 vs VEH group, n=8 (A–I), n=4 (J, K). PTX, paclitaxel; VEH, vehicle; AUC, area under curve.* Several studies reported that RvD1 could attenuate diverse inflammatory diseases [14], but the impact of RvD1 on PINP is still unclear. Herein, we detected $\frac{12}{15}$-Lox (a synthase of RvD1) and FPR2 (a receptor for RvD1) mRNA levels in sciatic nerve (SN) and dorsal root ganglia (DRG) tissues using qPCR analysis at day 14 after PTX administration. It was found that PTX treatment significantly reduced $\frac{12}{15}$-Lox mRNA expression by ~$20\%$ in SN and DRG tissues, but not in the skin from mice hindpaws (Figure 1J, $p \leq 0.05$). In contrast, PTX had no impact on FPR2 mRNA expression in SN, DRG and skin tissues (Figure 1K). The data indicated that RvD1 synthesis was impaired in the peripheral nervous system in PINP mice. ## Systemic administration of RvD1 reduces PTX-induced mechanical and cold hyperalgesia Next, we evaluated whether RvD1 could reduce mechanical and cold hyperalgesia induced by PTX in mice. First, we investigated the preventive efficacy of RvD1 on the development of PTX-induced pain hypersensitivity in mice. Mice received intraperitoneal (i.p.) injection of RvD1 (5 μg/kg/day) 24 h prior to PTX administration for a total of 15 days. RvD1 significantly reduced mechanical allodynia on day 5 after the first PTX injection (Figure 2A, $p \leq 0.01$). In RvD1 prevention group, the threshold of mechanical stimuli returned to baseline 15 days post-initiation of PTX administration, but it taken 18 days to recover without RvD1 prevention treatment (Figure 2A). Meanwhile, the area under curve (AUC) of Von *Frey analysis* also showed that RvD1 significantly reliefed the mechanical allodynia induced by RTX in mice (Figure 2B, $p \leq 0.01$). RvD1 significantly attenuated cold allodynia only on day 7 compared with PTX-treated mice (Figure 2C, $p \leq 0.01$). Although this effect was less pronounced than that of mechanical allodynia, the area under curve (AUC) of cold pain analysis showed RvD1 significantly decreased PTX-induced cold pain in mice (Figure 2D, $p \leq 0.05$). These data indicated that preventive application of RvD1 not only attenuated mechanical and cold hyperalgesia induced by PTX in mice. **Figure 2:** *Systemic administration of RvD1 reduces PTX-induced mechanical and cold hyperalgesia. Mice were injected with RvD1 (5 μg/kg, ip) 1 day before PTX injection, and injected continuously for 15 days. mechanical pain threshold tested using Von Frey and AUC (A, B), cold pain score tested using acetone stimulation and AUC (C, D); single injection of RvD1 (5 μg/kg, ip) on the 8th and 10th days after the first PTX injection, mechanical pain threshold and AUC of the hind paw of mice (E, F). *P < 0.05, **P < 0.01, ***P < 0.001 vs VEH group; #P < 0.05, ##P < 0.01 vs PTX group. n=8 in VEH and PTX groups, n=9 in RVD1+PTX group (A–D); n=8 in each group (E, F). RvD1, resolvin D1; PTX, paclitaxel; VEH, vehicle.* In addition, to investigate whether RvD1 could relieve pain when PINP was already established, we examined the potential therapeutic effect of RvD1 on PINP. An i.p. injection of RvD1 was given on day 8 following the first PTX injection. The mechanical allodynia was significantly inhibited at 2 h after RvD1 injection (Figure 2E, $p \leq 0.01$). This analgesic effect of RvD1 lasted for at least 2 days following RvD1 administration (Figure 2E). In addition, RvD1 still exhibited relief of mechanical pain with a single injection on day 10 (Figure 2E). Moreover, the area under curve of the Von *Frey analysis* also showed that the overall treatment with RvD1 significantly reliefed PINP (Figure 2F, $P \leq 0.01$). These data indicated that RvD1 may be useful not only for prevention of PINP, but also for treatment of PINP. ## RvD1 attenuates mechanical allodynia induced by PTX via regulating macrophages Recent findings indicate that the immune system, especially macrophages, plays a critical role in the development and maintenance of chemotherapy-induced peripheral neuropathy (CIPN). To test the hypothesis that macrophages contribute to the behavior signs of PINP, we intrathecally (i.t.) administrated PTX-treated BMDMs into spinal cerebrospinal fluid via lumbar puncture, and tested the mechanical pain threshold using Von Frey hairs. The experimental procedure was shown in Figure 3A. BMDMs (3.0×103 cells) were prepared in 10 μl PBS and collected for i.t. injection. As shown in Figures 3B, C, a single i.t. injection of normal BMDMs developed a rapid and transient (within 5 h) mechanical allodynia compared with PBS i.t. injection mice ($p \leq 0.05$). However, a single i.t. injection of PTX-treated BMDMs produced a remarkable decreased threshold of mechanical stimuli within 2 h (Figure 3B, $p \leq 0.001$). Notably, this nociceptive effect induced by PTX-treated BMDMs after i.t. injection lasted for 3 days (Figure 3B). **Figure 3:** *RvD1 attenuates mechanical allodynia induced by PTX via regulating macrophages. Mouse bone marrow-derived macrophages (BMDMs) were pretreated with 250 nm RVD1 for 3h, and then treated with 1 μM PTX for 24h. BMDMs of different treatment groups (3×103) were intrathecally injected into mice intervertebral foramen. The process is shown schematically in (A). Mechanical pain threshold was determined using Von Frey test (B, D), AUC analysis was shown in (C, E). *p < 0.05, **p < 0.01, ***p < 0.001 vs normal MC group; #p < 0.05, ##p < 0.01 vs MC-PTX group. n=6 in each group (B, C); n=8 in each group (D, E). MC, macrophage; RvD1, resolvin D1; PTX, paclitaxel.* We next examined the effect of BMDMs pretreated with RvD1 followed by PTX treatment on mechanical allodynia. BMDMs were pretreated with RvD1 (250 nM) for 24 h and then stimulated with PTX (1 μM) for 24 h. As shown in Figures 3D, E, the threshold for mechanical stimuli in RvD1+PTX-treated BMDMs group maintained at the same level compared with normal BMDMs group within 5 h after i.t. injection, and was significantly higher than PTX-treated BMDMs group ($p \leq 0.05$). These data indicated RvD1 alleviated PINP by acting on macrophages. We further detected the expressions of macrophage markers in DRG and evaluated the effect of RvD1 on macrophage infiltration. First, we detected F$\frac{4}{80}$ and CD68 mRNAs in DRG using qPCR method. The results showed that both F$\frac{4}{80}$ (Figure 4A, $p \leq 0.05$) and CD68 (Figure 4B, $p \leq 0.01$) mRNAs was significantly increased in mice DRG tissue after PTX treatment, but RvD1 had no effect on F$\frac{4}{80}$ and CD68 mRNAs in PTX-treated mice. We used immunofluorescence technique to label CD68 positive cells in DRG tissue to observe macrophage infiltration. The results were similar to the above qPCR experiment results (Figure 4C). These data indicated that the pain-relieving effect of RvD1 on PINP was not through inhibition of macrophage infiltration in the DRG. **Figure 4:** *RvD1 did not affect the infiltration of macrophages in DRG under PINP condition. The mRNA of mouse DRG tissue was extracted 14 days after PTX injection. The expression of macrophage markers F4/80 (A) and CD68 (B) mRNA was detected by qPCR, and the CD68 positive cells (C) in DRG were detected by immunofluorescence technique. *p < 0.05, **p < 0.01 vs CON group. n=4. Scale bar: 50 μm. RvD1, resolvin D1; PTX, paclitaxel.* ## RvD1 increases IL-10 production both in vitro and vivo Macrophages exhibit different phenotypes such as M0 (the resting phenotype), M1 (pro-inflammatory phenotype) and M2 (anti-inflammatory/pro-resolution phenotype). These phenotypes are characterized by distinct expression of pro-inflammation cytokines (IL-1β) or anti-inflammation cytokines (IL-10, TGF-β), and are closed related to macrophages functions. PINP involves a strong inflammation condition, and a series of studies have reported that RvD1 could regulate inflammation reaction. We speculate that RvD1 exerts analgesic effect by regulating the function of macrophages. We detected several inflammation factors mRNAs in mice tissues and BMDM samples. We found that anti-inflammation factor IL-10 was significantly decreased and pro-inflammation factor IL-1β was significantly increased on day 14 in DRG and SN tissues after PTX administration (Figure 5A). RvD1 prevention treatment significantly increased IL-10 mRNA both in DRG (Figure 5B, $p \leq 0.05$) and SN (Figure 5C, $p \leq 0.01$) tissues. However, RvD1 only decreased IL-1β mRNA in SN of PTX-treated mice, but not in DRG (Figure 5A). **Figure 5:** *The effect of RvD1 on the expression of inflammatory factors. The inflammatory factors IL-10, IL-4, TGF-β, IL-1β, TNF-α in DRG and SN tissues of mice were detected by qPCR 14 days after PTX injection (A–C). BMDMs were pretreated with 250 nm RvD1 for 3h and then treated with 1 μM PTX for 24 h, the expression of inflammatory factor mRNA was detected by qPCR (D). IL-10 protein was determined by ELISA (E). *p < 0.05, **p < 0.01 vs VEH or CON group; #P< 0.05, ##p< 0.01 vs PTX group. n=4. RvD1, resolvin D1; PTX, paclitaxel; VEH, vehicle.* Furthermore, we cultured BMDM cells to observe the effect of RvD1 on inflammatory factors after PTX treatment. PTX had no significant impact on IL-10 mRNA expression in BMDMs, but RvD1 pretreatment significantly increased IL-10 mRNA in RvD1+PTX group compared with PTX-treated BMDMs (Figure 5D, $p \leq 0.05$). RvD1 pretreatment failed to decrease IL-1β mRNA in BMDMs followed by PTX treatment, which was significantly increased after PTX treatment (Figure 5D). We detected IL-10 protein using ELISA in DRG tissue and BMDMs supernatant. The results showed that IL-10 protein had no significant change in DRGs after PTX treatment, but RvD1 treatment significantly increased IL-10 protein level in PTX-treated mice (Figure 5E, $p \leq 0.05$). Meanwhile, similar to tissue results, the concentration of IL-10 in the supernatant of BMDMs in RvD1+PTX-treated group significantly increased compared with PTX-treated group (Figure 5E, $p \leq 0.05$). These data suggested IL-10 was involved in the anti-allodynia effect of RvD1 in PINP. ## The analgesic effect of RvD1 on PINP is dependent of IL-10 Based on the above results, we speculated that IL-10 was required for RvD1 to attenuate PINP. Next, we used neutralizing antibody specific for mouse IL-10 to examine the contribution of IL-10 in the anti-allodynia effect of RvD1 in PINP mice. IL-10 neutralizing antibody or IgG control antibody was i.t. administrated to target spinal cord cells as well as DRG cells. Mice received a single i.t. injection of the antibodies on day 3 and day 8 after first PTX administration. The i.t. injection of control IgG had no significant impact on mechanical allodynia in RvD1+PTX-treated mice (Figure 6A). However, mechanical allodynia threshold rapidly decreased 1 h after i.t. injection of IL-10 neutralizing antibody on days 3 and 8 (Figure 6A, $p \leq 0.05$). This effect of IL-10 neutralizing antibody could last for 24 hours. The results suggested that IL-10 was required for the analgesic effect of RvD1 on PINP. **Figure 6:** *The analgesic effect of RvD1 on PINP is dependent of IL-10. On the 3rd and 8th day of PTX administration, mice were injected (i.t.) with IL-10 neutralizing antibody and IgG (2μg). Mechanical pain threshold was determined using von Frey test (A), AUC analysis was shown in (B). ###p<0.001 vs PTX group; &p<0.05, &&p<0.01 vs PTX+RVD1+IgG group. n=8 in each group. RvD1, resolvin D1; PTX, paclitaxel.* ## RvD1 increases macrophage IL-10 by activating FPR2 RvD1 is the endogenous agonist of formyl peptide receptor 2 (FPR2) [21]. It has been shown that activation of FPR2 by RvD1 can exert anti-inflammatory effects in many diseases [22, 23]. As shown in Figure 7A, the FPR2 expression in BMDMs had no significant change after PTX treatment only. However, the FPR2 expression in RvD1+PTX-treated BMDMs was significantly increased compared with PTX-treated group (Figures 7A, B). To further confirm the role of FPR2, we used FPR2 antagonist Boc1to observe its effect on IL-10 production. We found that Boc1 abolished the role of RvD1 in promoting IL-10 production (Figure 7C, $p \leq 0.05$). To test whether FPR2 participated in the relieving effect of RvD1 on PINP, mice were i.p. injected with Boc1. Inhibition of FPR2 signaling eliminated the pain relief effect of RvD1 on PINP (Figure 7D). These results suggested the RvD1-FPR2 axis was required for IL-10 production and pain relief effect. **Figure 7:** *RvD1 promoted the production of IL-10 and its analgesic effect on PINP by activating FPR2 receptor. BMDMs were treated with FPR2 receptor blocker BOC1 (1 μM) for 3 h, RvD1 (250 nm) for another 3 h, and then treated with PTX (1 μM)for 24h. The expression of FPR2 in BMDMs was detected using Western blot (A, B). The IL-10 protein in BMDMs supernatant was detected by ELISA (C). Mechanical pain threshold was determined using von Frey test (D). Boc1 (5 mg/kg/day) was administered intraperitoneally. *p < 0.05, **p < 0.01, ***p < 0.001 vs PTX group; #p<0.05, ##p<0.01 vs RvD1+PTX group. n=4 (A–C), n=8 (D). RvD1, resolvin D1; PTX, paclitaxel.* ## RvD1 attenuates the apoptosis of DRG neurons induced by BMDM conditioned medium Since macrophage activation is critical for initiating the neurons apoptosis in several neuropathy diseases [24], we then examined the effect of RvD1-treated BMDMs on DRG neurons apoptosis. The primary mouse DRG neurons were stimulated by different conditioned medium (CM) of BMDM, and the apoptosis neurons were determined using TUNEL staining. First, we prepared 4 groups of BMDMs CM, namely Con CM, PTX CM (treated with PTX only), RvD1+PTX CM (pretreated with RvD1 followed by PTX), RvD1 CM (treated with RvD1 only). As shown in Figure 8A, the number of TUNEL-positive DRG neurons significantly increased after the incubation of PTX CM of BMDMs. However, limited apoptosis neurons were observed following incubation of RvD1+PTX CM of BMDMs. The results indicated that RVD1-regulated macrophage could inhibit PTX-induced DRG neurons apoptosis. **Figure 8:** *RvD1 attenuates the apoptosis of DRG neurons induced by BMDM conditioned medium. BMDMs were treated with RvD1 (250 nm) for 3 h, and then treated with PTX (1 μM) for 24 h, the culture medium was collected and centrifuged, and the supernatant was taken as the condition medium. The primary DRG neurons were stimulated with the above condition medium for 24 h. The apoptosis of neurons was detected by TUNEL staining (A). The level of ROS was detected by DCFH-DA probe (B). Blue represents nuclei (DAPI staining), red represents apoptotic cells (TUNEL staining), green represents ROS (DCFH-DA staining). **p < 0.01, ***p < 0.001 vs PBS group; ; #p<0.05 vs PTX+PBS group. Scale bar: 100 μm. RvD1, resolvin D1; PTX, paclitaxel.* ## The BMDMs RvD1-FPR2-IL-10 axis plays a critical role in DRG neuronal antioxidant damage Several studies report that oxidative stress contributes for CIPN [25]. Under conditions of prolonged oxidative stress, the level of ROS is increased, leading to neuronal cells damage and apoptosis. DRG neurons were stimulated with collected different CMs of BMDMs, using DCFH-DA to detect intracellular ROS. As shown in Figure 8B, DCFH-DA fluorescence intensity significantly increased in DRG neurons stimulated by PTX-treated BMDMs CM. However, ROS fluorescence intensity was low when DRG neurons were stimulated by CM from RvD1+PTX-treated BMDMs. The results suggested that RvD1 could alleviate the oxidative stress induced by PTX in neurons by regulating macrophage function. The Nrf2-HO1 signaling pathway is an important pathway for protecting cells from oxidative damage. Although HO1 was increased in DRG neurons after stimulated with PTX CM (Figure 9A, $p \leq 0.05$), DRG neurons HO1 expression was further increased in RvD1+PTX CM stimulated group (Figure 9A, $p \leq 0.05$). In addition, transcript factor Nrf2 in nucleus had no change in neurons after PTX CM treatment (Figure 9B). However, Nrf2 in nucleus was significantly increased in neurons after RvD1+PTX CM treatment compared with PTX CM treatment group (Figure 9B, $p \leq 0.05$). To further study the role of RvD1-FPR2-IL-10 axis in activation of Nrf2-HO1 signaling pathway, we used FPR2 blocker Boc1 and IL-10 neutralizing antibody to observe the effect on HO1 expression and the sublocalization of Nrf2. We prepared 4 groups of BMDMs CM, namely PTX CM, RvD1+PTX CM, anti-IL-10+RvD1+PTX CM, Boc1+RvD1+PTX CM. As shown in Figure 9C, the expression of HO1 in neurons was significantly decreased in anti-IL-10+RvD1+PTX CM and Boc1+RvD1+PTX CM treatment group compared with PTX CM group ($p \leq 0.05$). On the other side, blocking FPR2 and IL-10 pathways inhibited Nrf2 nucleus translocation in DRG neurons (Figure 9D, $p \leq 0.01$). The results indicated that RvD1 activated FPR2-IL-10 axis in macrophage contributed for Nrf2-HO1 activation in DRG neurons. **Figure 9:** *The BMDMs RvD1-FPR2-IL-10 axis plays a critical role in DRG neuronal antioxidant damage. BMDMs were treated with BOC1 (1 μM) for 3 h, then treated with RvD1 (250 nM) for another 3 h, and then treated with PTX (1 μM) for 24 h, then the condition medium was collected. The primary DRG neurons were treated with IL-10 neutralizing antibody, then stimulated with the above condition medium for 24 h. The cytoplasmic and nuclear protein was isolated, and the expression of Nrf2 and HO-1 was detected by Western blot, the band gray was analyzed using Image J (A–D). *P<0.05 vs CON group; #p<0.05 vs PTX group; &p<0.05, &&p<0.01 vs PTX+RvD1 group. n=4. RvD1, resolvin D1; PTX, paclitaxel.* ## Discussion CIPN is a common cause of pain and poor quality of life in cancer patients and cancer survivors. Currently, there is a lack of effective drugs for the treatment of CIPN. RvD1 belongs to a unique family of lipid mediators called resolvins and has shown significant efficacy in the treatment of inflammation-related diseases. Our previous study showed that another resolvin, RvE1, can attenuate inflammatory pain [18]. The main objectives of the present study were to prove whether RvD1 had analgesic efficacy in the mouse PINP model, and to determine whether RvD1 could exert analgesic effect by regulating macrophages. In present study, we found RvD1 attenuated neuropathic pain induced by PTX, and these effects were mainly exerted by regulating the secretion of IL-10 in macrophages. Further, we found that the role of RvD1 in promoting IL-10 secretion by macrophages was dependent on its activation effect on FRP2. In addition, RvD1 could alleviate the damage of PTX to neurons by regulating macrophages via Nrf2-HO1 activation. Our study found RvD1 was beneficial for alleviating PINP in mice model, and proposed for the first time that its main mechanism of action was through activation of the RvD1/FPR2/IL-10 axis in macrophages. RvD1 is an endogenous lipid mediator derived from docosahexaenoic acid [26]. Studies have reported that RVD1 has a strong anti-inflammatory effect and is an important mediator in the process of inflammatory resolution (27–29). Our previous study showed that RvD1 attenuated inflammatory pain induced by formalin [18]. Other research team found RvD1 displayed potent analgesic properties in irritable bowel syndrome by inhibiting TRPV1 sensitisation [30]. Several groups reported that RVD1 could alleviate mechanical allodynia in spinal nerve ligation-induced neuropathic pain [31], noncompressive lumber disk herniation [32], spared nerve injury [33] model. However, the involvement of RvD1 in the resolution of neuropathic pain induced by PTX remains unknown. Interestingly, we found that $\frac{12}{15}$-LOX, the key synthetic enzyme of RvD1, decreased in the SN and DRG tissues in PINP mice. Moreover, treatment with exogenous RvD1 significantly promoted pain resolution in PINP. When PINP had been formed, RvD1 still had therapeutic effect. Therefore, these results reveal that RvD1 is benefit to alleviate PINP. Accumulating evidence suggests that macrophage plays important roles in the pathogenesis and resolution of neuropathic pain (34–36). In most instances, macrophages produce pain by releasing pro-inflammatory mediators such as TNFα and IL-1β, which enhance pain by modulating ion channels (37–39). It reported that PTX could induce macrophages infiltration into the DRG in a time course [8]. In present study, we found that mice mechanical pain threshold decreased after intrathecally injected PTX-treated macrophages. In addition, macrophage maker CD68 increased in DRGs in PINP mice. These finding further showed macrophages were involved in PINP. We also found that $\frac{12}{15}$-Lox, an endogenous synthase of RvD1, decreased in DRGs in PINP mice. This result indicated RvD1 synthesis was impaired under PINP condition. As a member of resolvins, RvD1 has been reported to act on macrophages to polarize toward a pro-resolution phenotype to alleviate inflammatory response (40–42). In present study, we found macrophages pretreated with RVD1 were able to alleviate PINP. However, RvD1 did not influence the CD68 expression level in PTX-treated mice. This suggested that RvD1 might exert analgesic effects on PINP by regulating macrophage function rather than inhibiting cellular infiltration. However, the exact mechanism was not clear. This requires linking the pharmacological effects of RvD1 with the function of macrophages to explore its role. Usually, macrophages are divided into two categories, M1-like and M2-like. M1-like macrophages, a classically activated phenotype, promote pain by producing pro-inflammatory cytokines and chemokines. In contrast, M2-like macrophages promote tissue repair and pain resolution via secreting anti-inflammatory cytokines. Several evidences have demonstrated that chemotherapy induces an increase in peripheral pro-inflammatory cytokines [43], such as TNF-α, IL-1β and IL-6 (44–46). Neuroinflammation has been considered as a potential common driver of CIPN, including PINP. Similar to the previous reports, we also found TNFα and IL-1β increased in SN and DRGs in PINP mice. RvD1 can play a protective role in many diseases by promoting M2 polarization of macrophages. RvD1 can attenuate gouty arthritis pain by reducing leukocyte recruitment and IL-1β production in the knee joint [47]. In addition, RvD1 can promote pulmonary inflammation resolution by enhancing M2 polarization. IL-10, IL-4 and TGF-β are classical anti-inflammatory factors secreted by M2-like macrophages. We found RvD1 significantly increased IL-10 and decreased IL-1β in the SN and DRGs tissues in PINP mice. These indicated RvD1 could attenuate PINP by promoting macrophages polarization toward M2 in peripheral neuronal tissues. It reported that macrophages in DRGs contribute to both the initiation and persistence of neuropathic pain [12], and the endogenous IL-10 from macrophages rather than other cells is required for CIPN resolution [48, 49]. It reported that activating the IL-10 signaling pathway by thalidomide or GLP-1 receptor agonist could attenuate neuropathic pain [50, 51]. In present study, we found RvD1 increased IL-10 protein level both in vivo and in BMDMs. However, blocking IL-10 signal abolished the analgesic effect of RvD1 on PINP. Our results confirmed IL-10 was required for the analgesic effect of RvD1 in neuropathic pain. RvD1 can bind to FPR2 with high affinity, and is the endogenous agonist for FPR2 [21]. FPR2 plays a crucial role in innate immune responses. Previous studies showed FPR2 was expressed on macrophages [52, 53]. It reported that FPR2 activation could attenuate inflammatory pain [54, 55]. In our study, RvD1 pretreatment increased FPR2 expression in PTX-treated macrophages. Several studies find FPR2 activation regulates macrophage polarization, and promotes inflammation resolution [56, 57]. We demonstrated that RvD1 upregulated IL-10 in macrophages via FPR2 activation. Our results showed blocking FPR2 by Boc-1 abolished the role of RvD1 in promoting IL-10 production. Meanwhile, Boc-1 abolished the analgesic effect of RvD1 on PINP. These suggested RvD1/FPR2 axis was required for IL-10 production, and was beneficial for PINP. It is well recognized that apoptosis of neurons in DRG is an important cause of the neuropathic pain in spinal nerve ligation injury and sciatic nerve injury model [58]. Also, other studies found neuronal apoptosis in DRG induced by PTX and vincristine contributed to CIPN [59, 60]. It is generally considered that activated macrophages play important roles in the initiation of neurons apoptosis in neuropathic pain [24]. Consistent with previous research reports, apoptosis was observed in primary DRG neurons after stimulation of condition medium from PTX-treated macrophages. However, apoptosis cell was decreased when the primary DRG neurons were stimulated by condition medium from RvD1+PTX treated macrophage. In addition, we found ROS in primary DRG neurons decreased in RvD1+PTX-treated macrophage CM group compared with PTX-treated macrophage CM group. Nrf2-HO1 pathway has a critical role against oxidative stress [61]. Recently, studies found enhancing Nrf2-HO1 signal can attenuate inflammatory pain by inhibiting MAPK pathway [62], and can alleviate vincristine and paclitaxel induced neuropathic pain (63–65). Nrf2 activation also attenuates chronic constriction injury-induced neuropathic pain via induction of PGC-1α-mediated mitochondrial biogenesis in the spinal cord [66]. So these evidences indicate that Nrf2 activation is beneficial in chronic pain [67]. The Nrf2-based therapy for chronic pain is a promising filed. Previous studies on IL-10 in pain mainly focused on inflammatory response pathways, but in recent years, studies have found that IL-10 can regulate a variety of signaling pathways. IL-10 can upregulate HO-1 expression and play a cardioprotective role in diabetic myocardial infarction [68]. IL-10 also promotes HO-1 expression in macrophages, thereby regulating the polarization of macrophagesc [69]. We found CM from RvD1+PTX treated macrophage could significantly promote Nrf2 nucleus translocation in DRG neurons. We speculated that IL-10 secreted from macrophages regulated by RvD1 mediates the dialogue between macrophages and neurons. Further, Nrf2 nucleus translocation was inhibited by FPR2 blocker Boc1 and IL-10 neutralizing antibody. Our results showed that RvD1 acted on FPR2 in macrophages to promote the secretion of IL-10, which in turn could activate the Nrf2-HO1 pathway in DRG neurons, exert anti-oxidative and anti-apoptotic effects, and then alleviate neuronal damage. Our study found RvD1 was beneficial for alleviating PINP in mice model, and proposed for the first time that its main mechanism of action was through activation of the RvD1/FPR2/IL-10 axis in macrophages. In addition, it is worth noting that some studies have reported that RvD1 can control tumor growth [70, 71], suggesting that RvD1 may not affect the therapeutic effect of paclitaxel on tumors when relieving PINP. The results indicated that systemic RvD1 supplementation would be a promising therapy strategy in the treatment of PINP. As for the roles of RvD1 in neuropathic pain induced by other chemotherapy drugs, including oxaliplatin and vincristine, more follow-up studies are needed. ## Data availability statement The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding authors. ## Ethics statement The animal study was reviewed and approved by Animal Ethics Committee of Soochow University. ## Author contributions CS: Investigation - lead, methodology - lead, data curation - lead, formal analysis - equal, writing-original draft - lead, writing-review & editing - equal. J-TZ: Conceptualization - equal, formal analysis - equal, writing-original draft - supporting. FZ: Investigation - supporting, methodology - supporting, equal. DX: Writing-original draft - supporting, writing-review & editing - equal. JP: *Formal analysis* - equal, writing-review editing - supporting. TL: Conceptualization - equal, data curation - lead, formal analysis - equal, methodology - equal, supervision - equal, writing-original draft - equal, writing-review & editing - equal. All authors contributed to the article and approved the submitted version. ## Conflict of interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ## Publisher’s note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. ## Supplementary material The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2023.1091753/full#supplementary-material ## References 1. Shi X, Sun X. **Regulation of paclitaxel activity by microtubule-associated proteins in cancer chemotherapy**. *Cancer Chemother Pharmacol* (2017) **80**. DOI: 10.1007/s00280-017-3398-2 2. Seretny M, Currie GL, Sena ES, Ramnarine S, Grant R, MacLeod MR. **Incidence, prevalence, and predictors of chemotherapy-induced peripheral neuropathy: A systematic review and meta-analysis**. *Pain* (2014) **155**. DOI: 10.1016/j.pain.2014.09.020 3. Banach M, Juranek JK, Zygulska AL. **Chemotherapy-induced neuropathies-a growing problem for patients and health care providers**. *Brain Behav* (2017) **7**. DOI: 10.1002/brb3.558 4. Domoto R, Sekiguchi F, Tsubota M, Kawabata A. **Macrophage as a peripheral pain regulator**. *Cells* (2021) **10**. DOI: 10.3390/cells10081881 5. Ren K, Dubner R. **Interactions between the immune and nervous systems in pain**. *Nat Med* (2010) **16**. DOI: 10.1038/nm.2234 6. Pinho-Ribeiro FA, Verri WA, Chiu IM. **Nociceptor sensory neuron-immune interactions in pain and inflammation**. *Trends Immunol* (2017) **38** 5-19. DOI: 10.1016/j.it.2016.10.001 7. Yeh TY, Luo IW, Hsieh YL, Tseng TJ, Chiang H, Hsieh ST. **Peripheral neuropathic pain: From experimental models to potential therapeutic targets in dorsal root ganglion neurons**. *Cells* (2020) **9**. DOI: 10.3390/cells9122725 8. Zhang H, Li Y, de Carvalho-Barbosa M, Kavelaars A, Heijnen CJ, Albrecht PJ. **Dorsal root ganglion infiltration by macrophages contributes to paclitaxel chemotherapy-induced peripheral neuropathy**. *J Pain* (2016) **17**. DOI: 10.1016/j.jpain.2016.02.011 9. Raoof R, Martin Gil C, Lafeber F, de Visser H, Prado J, Versteeg S. **Dorsal root ganglia macrophages maintain osteoarthritis pain**. *J Neurosci* (2021) **41**. DOI: 10.1523/JNEUROSCI.1787-20.2021 10. Celik MO, Labuz D, Keye J, Glauben R, Machelska H. **IL-4 induces M2 macrophages to produce sustained analgesia**. *JCI Insight* (2020) **5**. DOI: 10.1172/jci.insight.133093 11. Ntogwa M, Imai S, Hiraiwa R, Koyanagi M, Matsumoto M, Ogihara T. **Schwann cell-derived CXCL1 contributes to human immunodeficiency virus type 1 gp120-induced neuropathic pain by modulating macrophage infiltration in mice**. *Brain Behav Immun* (2020) **88**. DOI: 10.1016/j.bbi.2020.03.027 12. Yu X, Liu H, Hamel KA, Morvan MG, Yu S, Leff J. **Dorsal root ganglion macrophages contribute to both the initiation and persistence of neuropathic pain**. *Nat Commun* (2020) **11** 264. DOI: 10.1038/s41467-019-13839-2 13. Montague K, Malcangio M. **The therapeutic potential of Monocyte/Macrophage manipulation in the treatment of chemotherapy-induced painful neuropathy**. *Front Mol Neurosci* (2017) **10**. DOI: 10.3389/fnmol.2017.00397 14. de Faveri C, Fermino PMP, Piovezan AP, Volpato LK. **The inflammatory role of pro-resolving mediators in endometriosis: An integrative review**. *Int J Mol Sci* (2021) **22**. DOI: 10.3390/ijms22094370 15. Sun YP, Oh SF, Uddin J, Yang R, Gotlinger K, Campbell E. **Resolvin D1 and its aspirin-triggered 17R epimer. stereochemical assignments, anti-inflammatory properties, and enzymatic inactivation**. *J Biol Chem* (2007) **282**. DOI: 10.1074/jbc.M609212200 16. Rymut N, Heinz J, Sadhu S, Hosseini Z, Riley CO, Marinello M. **Resolvin D1 promotes efferocytosis in aging by limiting senescent cell-induced MerTK cleavage**. *FASEB J* (2020) **34** 597-609. DOI: 10.1096/fj.201902126R 17. Titos E, Rius B, Gonzalez-Periz A, Lopez-Vicario C, Moran-Salvador E, Martinez-Clemente M. **Resolvin D1 and its precursor docosahexaenoic acid promote resolution of adipose tissue inflammation by eliciting macrophage polarization toward an M2-like phenotype**. *J Immunol* (2011) **187**. DOI: 10.4049/jimmunol.1100225 18. Xu ZZ, Zhang L, Liu T, Park JY, Berta T, Yang R. **Resolvins RvE1 and RvD1 attenuate inflammatory pain**. *Nat Med* (2010) **16**. DOI: 10.1038/nm.2123 19. Cun-Jin S, Jian-Hao X, Xu L, Feng-Lun Z, Jie P, Ai-Ming S. **X-Ray induces mechanical and heat allodynia in mouse**. *Mol Pain* (2019) **15**. DOI: 10.1177/1744806919849201 20. Hama A, Takamatsu H. **Chemotherapy-induced peripheral neuropathic pain and rodent models**. *CNS Neurol Disord Drug Targets* (2016) **15** 7-19. DOI: 10.2174/1871527315666151110125325 21. Schmitz Nunes V, Rogerio AP, Abrahao O. **Insights into the activation mechanism of the ALX/FPR2 receptor**. *J Phys Chem Lett* (2020) **11**. DOI: 10.1021/acs.jpclett.0c02052 22. Molaei E, Molaei A, Hayes AW, Karimi G. **Resolvin D1, therapeutic target in acute respiratory distress syndrome**. *Eur J Pharmacol* (2021) **911**. DOI: 10.1016/j.ejphar.2021.174527 23. Filiberto AC, Ladd Z, Leroy V, Su G, Elder CT, Pruitt EY. **Resolution of inflammation**. *FASEB J* (2022) **36**. DOI: 10.1096/fj.202201114R 24. Yasuda H, Terada M, Maeda K, Kogawa S, Sanada M, Haneda M. **Diabetic neuropathy and nerve regeneration**. *Prog Neurobiol* (2003) **69**. DOI: 10.1016/s0301-0082(03)00034-0 25. Shim HS, Bae C, Wang J, Lee KH, Hankerd KM, Kim HK. **Peripheral and central oxidative stress in chemotherapy-induced neuropathic pain**. *Mol Pain* (2019) **15**. DOI: 10.1177/1744806919840098 26. Serhan CN, Chiang N. **Endogenous pro-resolving and anti-inflammatory lipid mediators: a new pharmacologic genus**. *Br J Pharmacol* (2008) **153 Suppl 1**. DOI: 10.1038/sj.bjp.0707489 27. Bhat TA, Kalathil SG, Bogner PN, Lehmann PV, Thatcher TH, Sime PJ. **AT-RvD1 mitigates secondhand smoke-exacerbated pulmonary inflammation and restores secondhand smoke-suppressed antibacterial immunity**. *J Immunol* (2021) **206**. DOI: 10.4049/jimmunol.2001228 28. Xiang SY, Ye Y, Yang Q, Xu HR, Shen CX, Ma MQ. **RvD1 accelerates the resolution of inflammation by promoting apoptosis of the recruited macrophages**. *Cell Death Discovery* (2021) **7** 339. DOI: 10.1038/s41420-021-00708-5 29. Isopi E, Mattoscio D, Codagnone M, Mari VC, Lamolinara A, Patruno S. **Resolvin D1 reduces lung infection and inflammation activating resolution in cystic fibrosis**. *Front Immunol* (2020) **11**. DOI: 10.3389/fimmu.2020.00581 30. Perna E, Aguilera-Lizarraga J, Florens MV, Jain P, Theofanous SA, Hanning N. **Effect of resolvins on sensitisation of TRPV1 and visceral hypersensitivity in IBS**. *Gut* (2021) **70**. DOI: 10.1136/gutjnl-2020-321530 31. Wang YH, Gao X, Tang YR, Chen FQ, Yu Y, Sun MJ. **Resolvin D1 alleviates mechanical allodynia**. *Neuroscience* (2022) **494** 12-24. DOI: 10.1016/j.neuroscience.2022.04.019 32. Liu ZH, Miao GS, Wang JN, Yang CX, Fu ZJ, Sun T. **Resolvin D1 inhibits mechanical hypersensitivity in sciatica by modulating the expression of nuclear factor-kappaB, phospho-extracellular signal-regulated kinase, and pro- and antiinflammatory cytokines in the spinal cord and dorsal root ganglion**. *Anesthesiology* (2016) **124**. DOI: 10.1097/ALN.0000000000001010 33. Tao X, Luo X, Zhang T, Hershey B, Esteller R, Ji RR. **Spinal cord stimulation attenuates mechanical allodynia and increases central resolvin D1 levels in rats with spared nerve injury**. *Front Physiol* (2021) **12**. DOI: 10.3389/fphys.2021.687046 34. Bang S, Xie YK, Zhang ZJ, Wang Z, Xu ZZ, Ji RR. **GPR37 regulates macrophage phagocytosis and resolution of inflammatory pain**. *J Clin Invest* (2018) **128**. DOI: 10.1172/JCI99888 35. da Silva MD, Bobinski F, Sato KL, Kolker SJ, Sluka KA, Santos AR. **IL-10 cytokine released from M2 macrophages is crucial for analgesic and anti-inflammatory effects of acupuncture in a model of inflammatory muscle pain**. *Mol Neurobiol* (2015) **51** 19-31. DOI: 10.1007/s12035-014-8790-x 36. Willemen HL, Eijkelkamp N, Garza Carbajal A, Wang H, Mack M, Zijlstra J. **Monocytes/Macrophages control resolution of transient inflammatory pain**. *J Pain* (2014) **15** 496-506. DOI: 10.1016/j.jpain.2014.01.491 37. Lv Z, Xu X, Sun Z, Yang YX, Guo H, Li J. **TRPV1 alleviates osteoarthritis by inhibiting M1 macrophage polarization**. *Cell Death Dis* (2021) **12** 504. DOI: 10.1038/s41419-021-03792-8 38. Kiguchi N, Kobayashi Y, Saika F, Sakaguchi H, Maeda T, Kishioka S. **Peripheral interleukin-4 ameliorates inflammatory macrophage-dependent neuropathic pain**. *Pain* (2015) **156**. DOI: 10.1097/j.pain.0000000000000097 39. El Karim I, McCrudden MT, Linden GJ, Abdullah H, Curtis TM, McGahon M. **TNF-alpha-induced p38MAPK activation regulates TRPA1 and TRPV4 activity in odontoblast-like cells**. *Am J Pathol* (2015) **185** 2994-3002. DOI: 10.1016/j.ajpath.2015.07.020 40. Hosseini Z, Marinello M, Decker C, Sansbury BE, Sadhu S, Gerlach BD. **Resolvin D1 enhances necroptotic cell clearance through promoting macrophage fatty acid oxidation and oxidative phosphorylation**. *Arterioscler Thromb Vasc Biol* (2021) **41**. DOI: 10.1161/ATVBAHA.120.315758 41. Kang JW, Lee SM. **Resolvin D1 protects the liver from ischemia/reperfusion injury by enhancing M2 macrophage polarization and efferocytosis**. *Biochim Biophys Acta* (2016) **1861**. DOI: 10.1016/j.bbalip.2016.06.002 42. Schmid M, Gemperle C, Rimann N, Hersberger M. **Resolvin D1 polarizes primary human macrophages toward a proresolution phenotype through GPR32**. *J Immunol* (2016) **196**. DOI: 10.4049/jimmunol.1501701 43. Brandolini L, d'Angelo M, Antonosante A, Allegretti M, Cimini A. **Chemokine signaling in chemotherapy-induced neuropathic pain**. *Int J Mol Sci* (2019) **20**. DOI: 10.3390/ijms20122904 44. Kim Y, Jung YH, Park SB, Kim H, Kwon JY, Kim HK. **TMI-1, TNF-alpha-Converting enzyme inhibitor, protects against paclitaxel-induced neurotoxicity in the DRG neuronal cells in vitro**. *Front Pharmacol* (2022) **13**. DOI: 10.3389/fphar.2022.842779 45. Al-Mazidi S, Alotaibi M, Nedjadi T, Chaudhary A, Alzoghaibi M, Djouhri L. **Blocking of cytokines signalling attenuates evoked and spontaneous neuropathic pain behaviours in the paclitaxel rat model of chemotherapy-induced neuropathy**. *Eur J Pain* (2018) **22**. DOI: 10.1002/ejp.1169 46. Huehnchen P, Muenzfeld H, Boehmerle W, Endres M. **Blockade of IL-6 signaling prevents paclitaxel-induced neuropathy in C57Bl/6 mice**. *Cell Death Dis* (2020) **11** 45. DOI: 10.1038/s41419-020-2239-0 47. Zaninelli TH, Fattori V, Saraiva-Santos T, Badaro-Garcia S, Staurengo-Ferrari L, Andrade KC. **: RvD1 disrupts nociceptor neuron and macrophage activation and neuroimmune communication, reducing pain and inflammation in gouty arthritis in mice**. *Br J Pharmacol* (2022) **179**. DOI: 10.1111/bph.15897 48. Krukowski K, Eijkelkamp N, Laumet G, Hack CE, Li Y, Dougherty PM. **CD8+ T cells and endogenous IL-10 are required for resolution of chemotherapy-induced neuropathic pain**. *J Neurosci* (2016) **36**. DOI: 10.1523/JNEUROSCI.3708-15.2016 49. Singh SK, Krukowski K, Laumet GO, Weis D, Alexander JF, Heijnen CJ. **CD8+ T cell-derived IL-13 increases macrophage IL-10 to resolve neuropathic pain**. *JCI Insight* (2022) **7**. DOI: 10.1172/jci.insight.154194 50. Deng MY, Ahmad KA, Han QQ, Wang ZY, Shoaib RM, Li XY. **Thalidomide alleviates neuropathic pain through microglial IL-10/beta-endorphin signaling pathway**. *Biochem Pharmacol* (2021) **192**. DOI: 10.1016/j.bcp.2021.114727 51. Tang X, Wu H, Mao X, Li X, Wang Y. **The GLP-1 receptor herbal agonist morroniside attenuates neuropathic pain**. *Biochem Biophys Res Commun* (2020) **530**. DOI: 10.1016/j.bbrc.2020.05.080 52. Petri MH, Laguna-Fernandez A, Gonzalez-Diez M, Paulsson-Berne G, Hansson GK, Back M. **The role of the FPR2/ALX receptor in atherosclerosis development and plaque stability**. *Cardiovasc Res* (2015) **105** 65-74. DOI: 10.1093/cvr/cvu224 53. Xu X, Gao W, Li L, Hao J, Yang B, Wang T. **Annexin A1 protects against cerebral ischemia-reperfusion injury by modulating microglia/macrophage polarization**. *J Neuroinflamm* (2021) **18** 119. DOI: 10.1186/s12974-021-02174-3 54. Colucci M, Stefanucci A, Mollica A, Aloisi AM, Maione F, Pieretti S. **New insights on formyl peptide receptor type 2 involvement in nociceptive processes in the spinal cord**. *Life (Basel)* (2022) **12**. DOI: 10.3390/life12040500 55. Vieira C, Salm DC, Horewicz VV, Ludtke DD, Emer AA, Koerich JF. **Electroacupuncture decreases inflammatory pain through a pro-resolving mechanism involving the peripheral annexin A1-formyl peptide receptor 2/ALX-opioid receptor pathway**. *Pflugers Arch* (2021) **473**. DOI: 10.1007/s00424-020-02502-1 56. Chen X, Zhuo S, Zhu T, Yao P, Yang M, Mei H. **Fpr2 deficiency alleviates diet-induced insulin resistance through reducing body weight gain and inhibiting inflammation mediated by macrophage chemotaxis and M1 polarization**. *Diabetes* (2019) **68**. DOI: 10.2337/db18-0469 57. Corminboeuf O, Leroy X. **FPR2/ALXR agonists and the resolution of inflammation**. *J Med Chem* (2015) **58**. DOI: 10.1021/jm501051x 58. Liao MF, Lu KT, Hsu JL, Lee CH, Cheng MY, Ro LS. **The role of autophagy and apoptosis in neuropathic pain formation**. *Int J Mol Sci* (2022) **23**. DOI: 10.3390/ijms23052685 59. Huang ZZ, Li D, Liu CC, Cui Y, Zhu HQ, Zhang WW. **CX3CL1-mediated macrophage activation contributed to paclitaxel-induced DRG neuronal apoptosis and painful peripheral neuropathy**. *Brain Behav Immun* (2014) **40**. DOI: 10.1016/j.bbi.2014.03.014 60. Chen XJ, Wang L, Song XY. **Mitoquinone alleviates vincristine-induced neuropathic pain through inhibiting oxidative stress and apoptosis**. *BioMed Pharmacother* (2020) **125**. DOI: 10.1016/j.biopha.2020.110003 61. Loboda A, Damulewicz M, Pyza E, Jozkowicz A, Dulak J. **Role of Nrf2/HO-1 system in development, oxidative stress response and diseases: an evolutionarily conserved mechanism**. *Cell Mol Life Sci* (2016) **73**. DOI: 10.1007/s00018-016-2223-0 62. Park J, Kim YT. **Erythronium japonicum alleviates inflammatory pain by inhibiting MAPK activation and by suppressing NF-kappaB activation**. *Antioxidants (Basel)* (2020) **9**. DOI: 10.3390/antiox9070626 63. Zhou L, Ao L, Yan Y, Li C, Li W, Ye A. **Levo-corydalmine attenuates vincristine-induced neuropathic pain in mice by upregulating the Nrf2/HO-1/CO pathway to inhibit connexin 43 expression**. *Neurotherapeutics* (2020) **17**. DOI: 10.1007/s13311-019-00784-7 64. Zhou YQ, Liu DQ, Chen SP, Chen N, Sun J, Wang XM. **Nrf2 activation ameliorates mechanical allodynia in paclitaxel-induced neuropathic pain**. *Acta Pharmacol Sin* (2020) **41**. DOI: 10.1038/s41401-020-0394-6 65. Zhou YQ, Liu DQ, Chen SP, Chen N, Sun J, Wang XM. **PPARgamma activation mitigates mechanical allodynia in paclitaxel-induced neuropathic pain**. *BioMed Pharmacother* (2020) **129**. DOI: 10.1016/j.biopha.2020.110356 66. Sun J, Li JY, Zhang LQ, Li DY, Wu JY, Gao SJ. **Nrf2 activation attenuates chronic constriction injury-induced neuropathic pain**. *Oxid Med Cell Longev* (2021) **2021**. DOI: 10.1155/2021/9577874 67. Zhou YQ, Mei W, Tian XB, Tian YK, Liu DQ, Ye DW. **The therapeutic potential of Nrf2 inducers in chronic pain: Evidence from preclinical studies**. *Pharmacol Ther* (2021) **225**. DOI: 10.1016/j.pharmthera.2021.107846 68. Gupta R, Liu L, Zhang X, Fan X, Krishnamurthy P, Verma S. **IL-10 provides cardioprotection in diabetic myocardial infarction**. *JCI Insight* (2020) **5**. DOI: 10.1172/jci.insight.133050 69. Li ZZ, Zhang ZW, Wang H, Chen YQ, Zhou XF, Duan LS. **Autologous transfusion of "old" red blood cells-induced M2 macrophage polarization through IL-10-Nrf2-HO-1 signaling complexes**. *Adv Clin Exp Med* (2020) **29**. DOI: 10.17219/acem/121924 70. Mattoscio D, Isopi E, Lamolinara A, Patruno S, Medda A, De Cecco F. **Resolvin D1 reduces cancer growth stimulating a protective neutrophil-dependent recruitment of anti-tumor monocytes**. *J Exp Clin Cancer Res* (2021) **40** 129. DOI: 10.1186/s13046-021-01937-3 71. Shan K, Feng N, Cui J, Wang S, Qu H, Fu G. **Resolvin D1 and D2 inhibit tumour growth and inflammation**. *J Cell Mol Med* (2020) **24**. DOI: 10.1111/jcmm.15436
--- title: 'Cyclopia extracts act as selective estrogen receptor subtype downregulators in estrogen receptor positive breast cancer cell lines: Comparison to standard of care breast cancer endocrine therapies and a selective estrogen receptor agonist and antagonist' authors: - Folasade R. Olayoku - Nicolette J. D. Verhoog - Ann Louw journal: Frontiers in Pharmacology year: 2023 pmcid: PMC10040842 doi: 10.3389/fphar.2023.1122031 license: CC BY 4.0 --- # Cyclopia extracts act as selective estrogen receptor subtype downregulators in estrogen receptor positive breast cancer cell lines: Comparison to standard of care breast cancer endocrine therapies and a selective estrogen receptor agonist and antagonist ## Abstract Breast cancer is the most diagnosed type of cancer amongst women in economically developing countries and globally. Most breast cancers express estrogen receptor alpha (ERα) and are categorized as positive (ER+) breast cancer. Endocrine therapies such as, selective estrogen receptor modulators (SERMs), aromatase inhibitors (AIs), and selective estrogen receptor downregulators (SERDs) are used to treat ER+ breast cancer. However, despite their effectiveness, severe side-effects and resistance are associated with these endocrine therapies. Thus, it would be highly beneficial to develop breast cancer drugs that are as effective as current therapies, but less toxic with fewer side effects, and less likely to induce resistance. Extracts of Cyclopia species, an indigenous South African fynbos plant, have been shown to possess phenolic compounds that exhibit phytoestrogenic and chemopreventive activities against breast cancer development and progression. In the current study, three well characterized Cyclopia extracts, SM6Met, cup of tea (CoT) and P104, were examined for their abilities to modulate the levels of the estrogen receptor subtypes, estrogen receptor alpha and estrogen receptor beta (ERβ), which have been recognized as crucial to breast cancer prognosis and treatment. We showed that the *Cyclopia subternata* Vogel (C. subternata Vogel) extracts, SM6Met and cup of tea, but not the C. genistoides extract, P104, reduced estrogen receptor alpha protein levels while elevating estrogen receptor beta protein levels, thereby reducing the ERα:ERβ ratio in a similar manner as standard of care breast cancer endocrine therapies such as fulvestrant (selective estrogen receptor downregulator) and 4-hydroxytamoxifen (elective estrogen receptor modulator). Estrogen receptor alpha expression enhances the proliferation of breast cancer cells while estrogen receptor beta inhibits the proliferative activities of estrogen receptor alpha. We also showed that in terms of the molecular mechanisms involved all the Cyclopia extracts regulated estrogen receptor alpha and estrogen receptor beta protein levels through both transcriptional and translational, and proteasomal degradation mechanisms. Therefore, from our findings, we proffer that the C. subternata Vogel extracts, SM6Met and cup of tea, but not the C. genistoides extract, P104, selectively modulate estrogen receptor subtypes levels in a manner that generally supports inhibition of breast cancer proliferation, thereby demonstrating attributes that could be explored as potential therapeutic agents for breast cancer. ## 1 Introduction Breast cancer (BC) is the most diagnosed type of cancer as well as the major source of cancer-associated deaths amongst women globally (DeSantis et al., 2015; DeSantis et al., 2019). The burden of disease is rapidly growing in economically developing countries with over half ($52\%$) of new BC cases and $62\%$ of mortalities occurring within this region (DeSantis et al., 2015). Roughly $70\%$ of BCs express ERα and are categorized as ER+ BC (Gonzalez et al., 2019). The most common endocrine treatments for ER+ BC thus target either ER signaling, via SERMs and SERDs, or the production of estrogen, via AIs (Rozeboom et al., 2019; Costa et al., 2020). The effects of estrogen in breast cancer are mediated by two ER subtypes, ERα and ERβ. ERα regulates the genes involved in cell proliferation, differentiation, and migration in mammary tissue via endocrine and paracrine mechanisms (Hartman et al., 2009; Leung et al., 2012; Huang et al., 2015; Saha et al., 2019). Interestingly, the role of ERβ in BC is still elusive since ERβ functions differently depending on the availability of ERα (Girgert et al., 2019). ERβ has generally been shown to facilitate apoptosis as well as to counter the proliferative activity of ERα in healthy mammary tissue (Huang et al., 2015). Furthermore, the level of ERβ and its co-expression with ERα has been suggested to modulate the cell’s response to estrogen in BC cell lines and may also modulate the response of ER+ BC to endocrine therapy (Song et al., 2019; Mal et al., 2020; Datta et al., 2022). Thus, ERβ should be considered as a potential target for the treatment of BC (Gustafsson and Warner, 2000; Nilsson et al., 2011; Hirao-Suzuki, 2021). The current study is motivated by the limitations associated with most adjuvant endocrine therapies developed to combat BC and the need to develop novel drugs that while effective, are less toxic, demonstrate fewer side effects, and are less likely to induce resistance (Clarke et al., 2015; Ramani et al., 2017; Sayed et al., 2019; Szostakowska et al., 2019; Franzoi et al., 2021). ERα has been identified as a viable drug target in resistant BC and thus the development of SERD therapies that specifically target the elimination of ERα is of considerable interest all the more so as fulvestrant, the only SERD currently approved by the FDA, suffers from poor oral bioavailability and has to be administered intramuscularly (Nathan et al., 2017; Shagufta et al., 2020; Downton et al., 2022; Farkas et al., 2022). Moreover, novel natural products or extracts provide possibilities for the discovery of new cancer therapies, especially for BC, as a substantial number of anticancer drugs currently used in the clinic are of natural origin (Zink and Traidl-Hoffmann, 2015; Wangkheirakpam, 2018; Yang et al., 2021). Traditional medicine involves the long historical use of natural products and their derivatives as herbal medicines or therapy for diseases based on ancient cultural theories and practices (Gurib-Fakim, 2006; Chintamunnee and Mahomoodally, 2012), with plants being the main source of medication (van Wyk and Prinsloo, 2018). The 2019 World Health Organization (WHO) global report on traditional and complementary medicine (T&CM) shows an increase in public interest and acceptance and indicates that the practice is mostly accepted in Africa (WHO Report, 2019), especially amongst the population in rural areas (Dalglish et al., 2019). Although T&CM has gained global recognition (Ekor, 2014; Tahvilian et al., 2014; Lopes et al., 2017; Wang K. et al., 2021) its use is still limited by a lack of quality evidence-based research (Pelkonen et al., 2014; Pal, 2021; Veziari et al., 2021). Often, traditional medicinal products are consumed as diet or as food supplements (Mbendana et al., 2019) and in South Africa, some dietary plants such *Aspalathus linearis* (rooibos tea), Cyclopia species (honeybush tea) and Athrixia phylicoides (bush tea) are considered medicinal herbal teas (Joubert et al., 2008). Extracts of A. linearis and A. phylicoides demonstrate assorted medicinal attributes, as do extracts from Cyclopia species, the major focus of the current study (Joubert et al., 2008; Louw et al., 2013; Joubert et al., 2019) Specifically, Cyclopia species, such as C. subternata Vogel, C. genistoides C. sessiliflora, C. intermedia, C. longifolia, and C. maculata, demonstrate anti-diabetic (Chellan et al., 2014; Schulze et al., 2016), anti-obesity (Pheiffer et al., 2013; Jack et al., 2018), and immune-stimulatory activities (Murakami et al., 2018) and osteoclast formation inhibition (Visagie et al., 2015); in addition to their useful application in nutraceutical, and cosmetic products (Joubert et al., 2019). Particularly of relevance to the current study, the C. subternata Vogel extract, SM6Met, was shown in several studies to possess phytoestrogenic activity, to display ERα antagonism and ERβ agonism, to antagonize estrogen-induced proliferation in ER+ BC cells (Mfenyana et al., 2008; Louw et al., 2013; Visser et al., 2013; Mortimer et al., 2015; van Dyk, 2018) and to ameliorate BC in rats (Visser et al., 2016; Oyenihi et al., 2018). Like SM6Met, the cup of tea (CoT) extract from C. subternata Vogel and the C. genistoides extract, P104, also exhibit phytoestrogenic properties and antagonize estrogen-induced proliferation in ER+ BC cells (Verhoog et al., 2007b; Visser et al., 2013; Roza et al., 2017). The current study focusses on the assessment of the potential SERD activities of the Cyclopia extracts, SM6Met, CoT and P104, via the ER subtypes, ERα and ERβ, in BC cell lines. We hypothesize that the Cyclopia extracts may function as selective ER subtype regulators, thus, selectively affecting the levels of ER subtypes. ## 2.1 Cell culture The human BC cell line, MCF7-BUS (Soto et al., 1995) was kindly donated by Ana Soto, department of Anatomy and Cell biology, Tufts University School of Medicine, and the T47D cell line (Keydar et al., 1979) was a generous donation from Iqbal Parker, Medical biochemistry division, University of Cape Town. The two cell lines were maintained at 37°C with $5\%$ CO2 and $90\%$–$95\%$ humidity in cell maintenance medium, which consisted of Dulbecco’s Modified Eagle’s Medium (DMEM) containing 4.5 g/mL glucose (Sigma-Aldrich, South Africa) supplemented with $5\%$ (v/v) heat-inactivated fetal calf serum (HI-FCS) (The Scientific Group, South Africa), 1.5 g/L sodium-bicarbonate, 0.11 g/L sodium-pyruvate and $1\%$ penicillin-streptomycin (100 IU/mL penicillin and 100 μg/mL streptomycin, Sigma-Aldrich) for MCF7 cells. For T47D cells the maintenance medium was the same except for $10\%$ FCS used. The cell lines were routinely tested for mycoplasma by Hoechst staining and found to be negative. Experiments were carried out on cell lines with passage numbers between 6–30. ## 2.2 Test panel The estrogenic compounds and Cyclopia extracts that make up the test panel include the endogenous hormone control, 17β-estradiol (E2), and the standard of care endocrine therapies (SOCs), [2]-4-hydroxytamoxifen (4-OHT) as a SERM control (Jordan, 2003) and fulvestrant (Ful) as a SERD control (Nathan et al., 2017), which were obtained from Sigma-Aldrich. The ER subtype selective ligands, methylpiperidinopyrazole (MPP), a ERα antagonist (Zhou et al., 2009), and liquiritigenin (Liq), a ERβ agonist (Mersereau et al., 2008), were purchased from Tocris Bioscience. The Cyclopia extracts, SM6Met, cup of tea (CoT), and P104 were obtained from cultivated and commercially harvested plant material and were previously prepared and characterized (Table 1). Retention samples of all the extracts have been preserved. **TABLE 1** | Polyphenolic compounds present in Cyclopia extracts | (g/100 g dry extract) a | (g/100 g dry extract) a.1 | (g/100 g dry extract) a.2 | | --- | --- | --- | --- | | Polyphenolic compounds present in Cyclopia extracts | SM6Met b | CoT b | P104 c | | Mangiferin | 1.899 | 1.000 | 3.606 | | Isomangiferin | 0.645 | 0.420 | 5.094 | | Luteolin | 0.040 | 0.018 | 0.096 | | Scolymoside (7-O-rutinosylluteolin) | 1.289 | 0.876 | nd f | | Vicenin-2 (6,8-di-β-D-glucopyranosylapigenin) | 0.089 | 0.065 | nd | | Eriocitrin (7-O-rutinosylerodictyol) | 0.846 | 0.600 | nd | | Hesperidin (7-O-rutinosylhesperetin) | 2.049 | 0.935 | nd | | 3′,5′-di-β-D-Glucopyranosylphloretin | 1.278 | 0.939 | nd | | 3′,5′-di-β-D-Glucopyranosyl-3-hydroxyphloretin d | 0.700 | 0.582 | nd | | 3-β-D-Glucopyranosyliriflophenone | 0.669 | 0.590 | nd | | 3-β-D-Glucopyranosyl-4-O-β-D-glucopyranosyliriflophenone e | 0.958 | 0.896 | nd | | Protocatechuic acid | 0.113 | 0.082 | nd | Stock solutions of the test panel were prepared in DMSO (Sigma-Aldrich) and stored at −20°C until use (see Supplementary Table S1). Stock solutions were diluted 1000X in treatment media, (phenol red-free DMEM (low glucose), 1.5 g/L sodium bicarbonate and 3.5 g/L glucose) to yield a final concentration of $0.1\%$ DMSO. ## 2.3 Western blot MCF7 and T47D cells were plated at 1.0 × 105 cells/well into 12 well plates and steroid starved in steroid starving media (phenol red-free DMEM (low glucose) supplemented with 1.5 g/L sodium bicarbonate, 3.5 g/L glucose, $1\%$ penicillin-streptomycin and $5\%$ heat-inactivated doubly dextran-coated charcoal-stripped FCS (2xDCCFCS) for MCF7 cells and $10\%$ 2xDCCFCS for T47D cells for 24 h. MCF7 and T47D cells were then washed in pre-warmed phosphate-buffered saline (PBS) and treated a with increasing concentrations of the test panel for 24 h in the treatment medium. Following treatment, the treated MCF7 and T47D cells were washed in 1 mL ice-cold PBS and lysed in 100 μL Radioimmunoprecipitation assay (RIPA) buffer [50 mM Tris-HCl, 150 mM NaCl, $1\%$ (v/v) NP40, $1\%$ (w/v) sodium deoxycholate and $0.1\%$ (w/v) SDS]. The lysates were transferred into Eppendorf tubes and 5x SDS reducing buffer [100 mM Tris-HCl pH 6.8, $50\%$ (v/v) SDS, $20\%$ (v/v) glycerol, $2\%$ (v/v) β-mercaptoethanol and $0.1\%$ (w/v) bromophenol blue] was added to enhance cell lysis. Thereafter the cell lysates were boiled at 95°C for 20 min. To separate the proteins, the sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) technique was employed, where 10 μL of the lysate was loaded onto a 15 well $10\%$ acrylamide gel containing $0.9\%$ (v/v), 2,2,2- trichloroethanol (TCE), both procured from Sigma-Aldrich. A protein molecular weight marker (color pre-stained protein standard, broad range from Inqaba Biotec) was loaded alongside the lysates to verify the sizes of ERα and ERβ protein. The gel was set to run at 75 V for 15 min and at 150 V for 60 min. The separated proteins on the acrylamide gel were imaged under UV light and the image acquired using the BioRad molecular imager, Gel DocTM XR+ with Image LabTM software. The acquired image of the total protein content was utilized for normalization (as detailed below). The proteins were then transferred to a Hybond-ECL nitrocellulose membrane (Separation Scientific) under 0.18 A electric current for 2 h. To ensure a successful protein transfer, the nitrocellulose membrane and the gels were imaged and acquired under UV light with the BioRad molecular imager, Gel DocTM XR+ with Image LabTM software after the transfer (see Supplementary Figure S1). The gels were then discarded, and the membranes were blocked in $10\%$ milk powder at room temperature for 90 min on a Stovall Belly dancer shaker. The membranes were then washed consecutively for 15 min and 5 min with 1x Tris-buffered saline tween (TBST) [50 mM Tris base, 150 mM NaCl and $0.1\%$ (v/v) Tween 20 dissolved in deionized water], followed by a 5 min wash with 1x Tris-buffered saline (TBS) (50 mM Tris base and 150 mM NaCl dissolved in deionized water). Thereafter, the membranes were probed with the primary antibody {anti-ERα [sc-8oo2 (F-10), Santa Cruz Biotechnology and anti- ERβ (MA524807/PPZ0506], Thermo Fisher Scientific} at 4°C overnight on a Stovall Belly dancer shaker. The membranes were then washed consecutively for 15 and 5 min in TBST, and in TBS for 5 min. After washing, the membranes were incubated with the secondary antibody (Rabbit anti-mouse IgG H&L (HRP) ab97046 from Abcam) for 90 min at room temperature on a Stovall Belly dancer shaker. Once more, the membranes were washed as described previously. The membranes were then incubated with BioRad ECL Western blotting reagent for 5 min and imaged using the iBrightTM Imaging System from Invitrogen (see Supplementary Figure S1). To quantify the intensity of the ERα and ERβ protein bands, MyImage Analysis software was used. Equal protein loading was ensured by normalization to the total protein content of each lane. The total protein content was obtainable by adding a sedative agent, TCE (Chopra et al., 2019), to the SDS-PAGE gels. The TCE attaches hydroxyethanone to the indole ring of tryptophan residues that results in the fluorescence of protein bands under UV light, which was quantified using transilluminator molecular imager (BioRad molecular imager, Gel DocTM XR+ with Image LabTM software). The total protein of the test panel-treated lysates was set relative to that of the vehicle. The normalization factor (NF) of the vehicle was set at 1, in which case a NF < 1 or NF > 1 indicates that the total protein content of the test panel-treated lysate is higher or lower than that of the vehicle, respectively. The intensity of the ERα and ERβ protein bands was then multiplied by the NFs to obtain the normalized intensity of the band. Normalized ER expression was plotted as a percentage (average ± SD) relative to the vehicle (DMSO) sample, which was set to $100\%$. Dose-response curves were generated by fitting experimental values to the three-parameter logistic curve fitting equation in GraphPad Prism with the maximal response constrained to $100\%$ to obtain the efficacy (maximal response) and potency (IC50). Supplementary Figure S1 contains an example of the full SDS-PAGE gel following protein separation, the full nitrocellulose membrane after protein transfer, the full SDS-PAGE gel after protein transfer and the full nitrocellulose membrane following immunoblotting (the Western blot). ## 2.3.1 Proteasomal and translational inhibition To study the effects of proteasomal and translational inhibition on the modulation of ERα and ERβ protein levels by the test panel the proteasomal inhibitor, MG132 (Fan et al., 2004) and the translational inhibitor, cycloheximide (CHX) (Schneider-Poetsch et al., 2010) were obtained from Sigma-Aldrich. Stock solutions for both inhibitors were prepared in DMSO and both were used at a final concentration of 1 nM. ## 2.4 Statistical analysis Statistical analysis was carried out using GraphPad Prism software version 5. Details of the individual statistical analysis used, including post-tests are described in the figure legends. Statistical difference is expressed as either a different letter or using symbols (*, # and $), as specified in the figure legends. Non-significant results are denoted by “ns.” For all Figures Average ± SD is of three independent biological experiments analyzed as such. ## 3.1 Subtype selective modulation of ER subtype protein levels by Cyclopia extracts in MCF7 and T47D cell lines MCF7 and T47D cell lines are considered acceptable models for ERα+ luminal A carcinomas (Lacroix & Leclercq, 2004). They require estrogen for proliferation and although both cell lines express ERα and ERβ, MCF7 has a high ERα/ERβ ratio and T47D has a low ERα/ERβ ratio (Nadal-Serrano et al., 2012) as confirmed in Figure 1. **FIGURE 1:** *Basal levels of ERα and ERβ protein in MCF7 and T47D cells. MCF7 and T47D cells were steroid starved for 24 h. The media was then changed to high glucose-DMEM only and the cells incubated for another 24 h, after which ERα and ERβ basal protein levels were determined using Western blot. The western blots shown are representatives of three independent experiments for (A) ERα and (B) ERβ. For quantification, the intensity of the ERα and ERβ bands were determined with MyImage Analysis software, after which the obtained values were normalized to total protein content and expressed as a percentage (AVG ± SD) of MCF7 ERα and ERβ protein levels, which was set at 100% (black dotted line). WB is not an absolutely quantitative technique (i.e., the absolute concentrations of proteins cannot be ascertained) only the relative amounts of a protein may be compared between cell lines. Furthermore, as two different antibodies were used for ERα and ERβ we cannot directly compare the absolute levels for the ER subtypes. Thus, we chose to normalize the ER subtype levels to that in MCF7 cells. Statistical analysis was done using unpaired t-test (*p < 0.05, ***p < 0.0001) to evaluate the statistical difference between the basal levels of ERα and ERβ protein in MCF7 and T47D cells.* Western blots were used to determine the efficacy and potency (Table 2) of the test panel in modulating the protein levels of ERα and ERβ in MCF7 (Figure 2) and T47D (Figure 3) cells. ERα protein levels were downregulated by all the Cyclopia extracts in a dose-dependent manner in the MCF7 and T47D cell lines. Specifically, for SM6Met the efficacy of the downregulation of ERα protein levels was $68.6\%$ in MCF7 and $73.7\%$ in T47D cells, for CoT it was $82.7\%$ in MCF7 and $75.0\%$ in T47D cells and for P104 it was $55.4\%$ in MCF7 and $71.3\%$ in T47D cells. Statistical comparison indicates that the efficacy of downregulation of ERα protein levels by the Cyclopia extracts was not significantly ($p \leq 0.05$) different (Table 2). ERβ protein levels were upregulated by the two C. subternata Vogel extracts, SM6Met and CoT, and downregulated by the C. genistoides extract, P104. Specifically, ERβ protein levels were upregulated by SM6Met to $145.4\%$ in MCF7 and $132.2\%$ in T47D cells, and to $124.5\%$ in MCF7 and $114.0\%$ in T47D cells by CoT. The upregulation of ERβ protein levels by CoT in T47D cells was lower, albeit not significantly lower than in MCF7 cells, however, it was significantly ($p \leq 0.05$) lower than the upregulation of ERβ protein levels by SM6Met (Table 2). In contrast, ERβ protein levels were downregulated in both cell lines by P$104\%$ to $90.56\%$ in MCF7 and $73.8\%$ in T47D cells, which was not statistically ($p \leq 0.05$) different. Both SM6Met and CoT downregulated ERα protein levels while simultaneously increasing ERβ protein levels in both cell lines, resulting in a decreased ERα:ERβ ratio (Table 2). However, the effect of SM6Met on the ERα:ERβ ratio was greater as it was more effective at both downregulating ERα protein levels and upregulating ERβ protein levels than CoT. Although, P104 downregulated both ERα and ERβ protein levels in both cell lines, its effect on the ERα protein levels in MCF7 cells was substantially more than on the ERβ protein levels resulting in ERα:ERβ ratio reduced to about that of CoT. However, in T47D cells, the efficacies for the downregulation of the ER subtype proteins were similar and thus, P104 did not have a major influence on the ERα:ERβ ratio in T47D cells. The potency (Table 2) of SM6Met in decreasing ERα protein levels was lower in MCF7 (3.1 × 10−9 mg/mL) than in T47D (1.2 × 10−12 mg/mL) cells as was the potency of SM6Met to increase ERβ proteins levels (1.7 × 10−12 mg/mL in MCF7 compared to 4.4 × 10−13 mg/mL in T47D cells). Similarly, the potency of CoT in decreasing ERα protein levels was lower in MCF7 (2.5 × 10−9 mg/mL) than in T47D (1.1 × 10−11 mg/mL) cells. However, in contrast, the potency of CoT in upregulating ERβ protein levels in MCF7 (1.0 × 10−13 mg/mL) was slightly higher than in T47D (9.9 × 10−13 mg/mL) cells. Additionally, the potency of P104 in decreasing ERα protein levels was significantly ($p \leq 0.05$) lower in MCF7 (4.6 × 10−7 mg/mL) than in T47D (7.1 × 10−15 mg/mL) cells, while the potency of P104 in downregulating ERβ protein levels in MCF7 (1.3 × 10−10 mg/mL) and in T47D (4.7 × 10−10 mg/mL) cells were similar. The potencies of the Cyclopia extracts in modulating either ERα or ERβ protein levels were not statistically different across the two cell lines, except for the potency of P104 in downregulating ERα protein levels in T47D cells. Comparison of the effects elicited by the Cyclopia extracts with those elicited by the ER subtype specific ligands (Figures 2, 3) suggests that ERα antagonism is unlikely to be the mechanism whereby the Cyclopia extracts exert their SERD activity against ERα as the ERα antagonist, MPP, did not downregulate ERα. ERβ agonist activity seems a more likely mechanism for the SERD activity against ERα as the ERβ agonist, liquiritigenin, did downregulate ERα. In comparing the relative effects of the full test panel in modulating the ERα:ERβ ratio (Table 2), we can distinguish three groups. Those that did not really affect the ratio (ERα:ERβ ratio around 1), which includes E2 and MPP in both cell lines, and liquiritigenin and P104 only in T47D cells. Those that had a marked effect on the ratio (ERα:ERβ ratio around 0.5), which includes the SOCs and the Cyclopia extracts, except for P104 in T47D cells, and those that had a major effect on the ratio (ERα:ERβ ratio below 0.2) such as liquiritigenin only in MCF7 cells. Overall comparison of the effects of the Cyclopia extracts with that of the SOCs indicates that the C. subternata Vogel extract, SM6Met, was slightly less effective than fulvestrant, but as effective as 4-OHT, in reducing the ERα:ERβ ratio, however with a markedly higher potency in increasing ERβ protein levels, while CoT was slightly less effective than both the SOCs and P104 was the least effective Cyclopia extract at reducing the ERα:ERβ ratio. ## 3.2 Exploration of the molecular mechanism whereby Cyclopia extracts modulate ERα and ERβ protein levels Estrogenic ligands regulate the expression and stability of ERα and ERβ in BC through diverse molecular mechanisms depending on the conformation change elicited in the ER subtypes (Pink and Jordan, 1996; Khissiin and Leclercq, 1999; Wijayaratne and McDonnell, 2001). These molecular mechanisms involve the transcriptional, translational as well as the post-translational stages (Kondakova et al., 2020), through a process that may be described as a “push” versus “pull” mechanism. The “push” is controlled by transcriptional and translational processes, while the “pull” is controlled by post-translational processes that result in the degradation of the receptor protein, mediated primarily by the ubiquitin-proteasome pathway (UPS) (Kondakova et al., 2020). To explore the molecular mechanisms underlying the modulation of ERα and ERβ protein turnover by the Cyclopia extracts in MCF7 and T47D BC cells, translation was inhibited using a protein synthesis inhibitor, cycloheximide (CHX) (Baliga et al., 1969; Perry et al., 1995), while degradation of the ER via the UPS was inhibited using a proteasome inhibitor, MG132 (Khissiin and Leclercq, 1999; Fan et al., 2004). ## 3.2.1 Effect of inhibition of protein synthesis on modulation of ER subtype protein levels by Cyclopia extracts SM6Met and CoT downregulate ERα and upregulate ERβ protein levels in MCF7 and T47D cells, while P104 downregulates the protein levels of both ER subtypes (Figures 4, 5). Addition of the translational inhibitor, CHX, reversed the downregulation of ERα by SM6Met to basal levels in both MCF7 (Figure 4F) and T47D (Figure 5F) cells, however, significantly ($p \leq 0.01$) so only at the higher concentration of SM6Met where the increase in protein levels was between 1.3 and 1.4-fold. Similarly, CHX also reversed the downregulation of ERα by P104 to basal levels in both cell lines (Figures 4H, 5H), however, significance ($p \leq 0.05$) was only observed in the T47D cells where the increase in protein levels was between 1.3 and 1.5-fold. In contrast, the addition of CHX had little effect on the downregulation of ERα by CoT (Figures 4G, 5G) and only increased levels by 1.1-fold. Generally, the addition of CHX did not have a significant effect on the modulation of ERβ protein levels, except in the case of SM6Met (Figure 5F) and CoT (Figure 5G) in T47D cells where the stabilizing effect of CHX on ERβ protein levels, though small (1.1 to 1.2-fold), is significant ($p \leq 0.05$). **FIGURE 4:** *Effect of CHX, a protein synthesis inhibitor, on the modulation of ERα and ERβ protein levels in MCF7 cells. MCF7 cells were steroid starved for 24 h and then treated with either vehicle (DMSO) or LogIC50 (from Figure 2; Table 2 with µg/ml converted to M) and saturating (1 μM) concentrations of (A) E2 or the SOCs, (B) fulvestrant (Ful) or (C) 4-OH-tamoxifen (4-OHT), or the ER subtype selective ligands, (D) liquiritigenin (Liq) or (E) methyl-piperidino-pyrazole (MPP), or LogIC50 (from Figure 2 in µg/ml) and saturating (10−6 μg/ml) concentrations of the Cyclopia extracts, (F) SM6Met, (G) cup of tea (CoT) or (H) P104 in the presence or absence of 1 nM CHX for another 24 h, after which the effect of ± CHX on ERα and ERβ protein levels were determined using Western blot. The western blots shown as insert are representatives of three independent experiments. For quantification, the intensity of the ERα and ERβ bands were determined with MyImage Analysis software, after which the obtained values were normalized to total protein content and expressed as a percentage (AVG ± SD) of DMSO, which was set at 100%. Fold-change is indicated above the bars. Statistical analysis was done using a two-tailed t-test to establish significant differences due to addition of CHX (# p < 0.05, ## p < 0.01 and ### p < 0.001).* **FIGURE 5:** *Effect of CHX, a protein synthesis inhibitor, on the modulation of ERα and ERβ protein levels in T47D cells. T47D cells were steroid starved for 24 h and then treated with either vehicle (DMSO) or LogIC50 (from Figure 3; Table 2 with µg/ml converted to M) and saturating (1 μM) concentrations of (A) E2 or the SOCs, (B) fulvestrant (Ful) or (C) 4-OH-tamoxifen (4-OHT), or the ER subtype selective ligands, (D) liquiritigenin (Liq) or (E) methyl-piperidino-pyrazole (MPP), or LogIC50 (from Figure 3 in µg/ml) and saturating (10−6 μg/ml) concentrations of the Cyclopia extracts, (F) SM6Met, (G) cup of tea (CoT) or (H) P104 in the presence or absence of 1 nM CHX for another 24 h, after which the effect of ± CHX on ERα and ERβ protein levels were determined using Western blot. The western blots shown as insert are representatives of three independent experiments. For quantification, the intensity of the ERα and ERβ bands were determined with MyImage Analysis software, after which the obtained values were normalized to total protein content and expressed as a percentage (AVG ± SD) of DMSO, which was set at 100%. Fold-change is indicated above the bars. Statistical analysis was done using a two-tailed t-test to establish significant differences due to addition of CHX (# p < 0.05, ## p < 0.01 and ### p < 0.001).* The addition of the translational inhibitor, CHX, caused no significant difference in the modulation of ERα and ERβ protein levels by liquiritigenin (Figures 4D, 5D) in both cell lines, and in the modulation by MPP (Figure 5E) in T47D cells. However, in MCF7 cells, there was a slight (1.1-fold), yet significant ($p \leq 0.05$), increase in the protein levels of ERα and ERβ upon the addition of CHX to MPP compared to MPP alone (Figure 4E). Similarly, the addition of the translational inhibitor, CHX, had no significant effect on the downregulation of either ERα or ERβ protein levels by E2 in either cell line (Figures 4A, 5A). Likewise, the effect of fulvestrant (Figure 4B) and 4-OHT (Figure 4C) on ERα and ERβ protein levels in MCF7 cells was not significantly altered by the addition of CHX. However, translational inhibition through the addition of CHX significantly ($p \leq 0.05$) increased the protein levels of ERα and ERβ modulated by fulvestrant in T47D cells (Figure 5B) by 1.3 to 1.4-fold and 1.2-fold, respectively. Similarly, the modulation of ERα, but not ERβ, protein levels by 4-OHT in T47D cells (Figure 5C) was significantly ($p \leq 0.0001$) reversed (1.7 to 1.8-fold) by the addition of CHX. ## 3.2.2 Effect of inhibition of proteasomal inhibition on modulation of ER subtype protein levels by Cyclopia extracts Inhibition of proteasomal degradation with MG132 generally counteracts the downregulatory effect of the Cyclopia extracts, SM6Met, CoT and P104, on ERα protein levels while enhancing the stabilization of ERβ protein levels by SM6Met and CoT (Figures 6, 7). Specifically, proteasomal inhibition counteracts the effects of the Cyclopia extracts, SM6Met (Figure 6F), CoT (Figure 6G) and P104 (Figure 6H) in downregulating ERα protein levels in MCF7 cells. Although not always significantly higher, the magnitude of the change due to the addition of MG132 was substantial (1.3 to 1.8- fold). In the T47D cells, the effects of MG132 on the downregulation of ERα protein levels by the Cyclopia extracts (Figures 7F–H) was substantially lower (1.1 to 1.3-fold) and mostly not significant. Stabilization of ERβ protein levels by SM6Met (Figures 6F, 7F) and CoT (Figures 6G, 7G) in both MCF7 and T47D cells was enhanced (1.1 to 1.3-fold), although not always significantly, by the addition of MG132. The effect of P104 on ERβ protein levels was significantly ($p \leq 0.05$) reversed (1.1 to 1.5-fold) by proteasomal inhibition in MCF7 cells (Figure 6H), while in T47D cells (Figure 7H) no effect was observed by adding MG132. **FIGURE 6:** *Effect of MG132, a UPS inhibitor, on the modulation of ERα and ERβ protein levels in MCF7 cells. MCF7 cells were steroid starved for 24 h and then treated with either vehicle (DMSO) or LogIC50 (from Figure 2; Table 2 with µg/ml converted to M) and saturating (1 μM) concentrations of (A) E2 or the SOCs, (B) fulvestrant (Ful) or (C) 4-OH-tamoxifen (4-OHT), or the ER subtype selective ligands, (D) liquiritigenin (Liq) or (E) methyl-piperidino-pyrazole (MPP), or LogIC50 (from Figure 2 in µg/ml) and saturating (10−6 μg/ml) concentrations of the Cyclopia extracts, (F) SM6Met, (G) cup of tea (CoT) or (H) P104 in the presence or absence of 1 nM MG132 for another 24 h, after which the effect of ± MG132 on ERα and ERβ protein levels were determined using Western blot. The western blots shown as insert are representatives of three independent experiments. For quantification, the intensity of the ERα and ERβ bands were determined with MyImage Analysis software, after which the obtained values were normalized to total protein content and expressed as a percentage (AVG ± SD) of DMSO, which was set at 100%. Fold-change is indicated above the bars. Statistical analysis was done using a two-tailed t-test to establish significant differences due to addition of MG132 (# p < 0.05, ## p < 0.01 and ### p < 0.001).* **FIGURE 7:** *Effect of MG132, a UPS inhibitor, on the modulation of ERα and ERβ protein levels in T47D cells. T47D cells were steroid starved for 24 h and then treated with either vehicle (DMSO) or LogIC50 (from Figure 3; Table 2 with µg/ml converted to M) and saturating (1 μM) concentrations of (A) E2 or the SOCs, (B) fulvestrant (Ful) or (C) 4-OH-tamoxifen (4-OHT), or the ER subtype selective ligands, (D) liquiritigenin (Liq) or (E) methyl-piperidino-pyrazole (MPP), or LogIC50 (from Figure 3 in µg/ml) and saturating (10−6 μg/ml) concentrations of the Cyclopia extracts, (F) SM6Met, (G) cup of tea (CoT) or (H) P104 in the presence or absence of 1 nM MG132 for another 24 h, after which the effect of ± MG132 on ERα and ERβ protein levels were determined using Western blot. The western blots shown as insert are representatives of three independent experiments. For quantification, the intensity of the ERα and ERβ bands were determined with MyImage Analysis software, after which the obtained values were normalized to total protein content and expressed as a percentage (AVG ± SD) of DMSO, which was set at 100%. Fold-change is indicated above the bars. Statistical analysis was done using a two-tailed t-test to establish significant differences due to addition of MG132 (# p < 0.05, ## p < 0.01 and ### p < 0.001).* Generally, proteasomal inhibition with MG132 did not significantly counteract the effects of the ER subtype selective ligands on the ERα and ERβ protein levels (Figures 6D, E, 7D, E), apart from liquiritigenin-induced downregulation of ERα protein levels in MCF7 cells (Figure 6D) and increased stabilization of ERβ protein levels by liquiritigenin (Figure 7D) in T47D cells. Although these effects were statistically significant, the magnitude of the fold-change due to MG132 was only substantial for ERα protein levels in MCF7 cells (1.5 to 1.7-fold) but not for ERβ protein levels in T47D cells (1.1 to 1.2-fold). Proteasomal inhibition with MG132 largely counteracts the effects of E2 (Figures 6A, 7A) in downregulating both ERα and ERβ in MCF7 and T47D cells. Although not always significantly higher, the fold change in ER levels due to MG132 addition are of a higher magnitude in MCF7 cells (1.3 to 1.5-fold) than in T47D cells (1.2 to 1.4-fold). Downregulation of ERα protein levels by the SOCs, fulvestrant (Figures 6B, 7B) and 4-OHT (Figures 6C, 7C) was counteracted to a statistically significant degree ($p \leq 0.05$) by the addition of MG132, except in the case of 4-OHT in MCF7 cells. Interestingly, the highest fold-change in ER levels due to the addition of MG132 (1.4 to 1.7-fold) was observed with 4-OHT in T47D cells, where addition of MG132 stabilized ERα protein levels to above that of basal levels. Upregulation of ERβ protein levels by the SOCs, fulvestrant (Figures 6B, 7B) and 4-OHT (Figures 6C, 7C), was slightly enhanced (between 1.1 and 1.3-fold) by the addition of MG132, although not always to a statistically significant degree ($p \leq 0.05$). ## 4 Discussion The SOCs for ER+ BC include SERDs such as fulvestrant that target and reduce the expression of ERα (Lu and Liu, 2020; Hernando et al., 2021). Specifically, fulvestrant, the only SERD currently used clinically, targets and degrades ERα protein through proteasomal degradation and is often used to combat tamoxifen and AI resistance (Mottamal et al., 2021). However, due to the poor pharmacokinetics associated with fulvestrant administration, which necessitates intramuscular injection, there are limitations on its bioavailability, which results in incomplete ERα repression by fulvestrant (Croxtall and McKeage, 2011; Nathan et al., 2017). Regardless, the positive attributes of fulvestrant, by degrading ERα protein, proffer insights for the development of novel oral SERDs with improved bioavailability to overcome endocrine therapy resistance in BC with improved efficacy and potency (Lu and Liu, 2020; Hernando et al., 2021). Furthermore, not only the absolute levels of ERα but rather the levels of ERα relative to that of ERβ, the ERα:ERβ ratio, has been shown to play an important role in the BC prognosis (Evers et al., 2014; Acconcia et al., 2017). ERα facilitates cell proliferation while ERβ enables cell apoptosis and counteracts the proliferative activity of ERα (Huang et al., 2015), and thus, an increased ERα:ERβ ratio is often associated with BC (Zhao et al., 2015; Acconcia et al., 2017). Therefore, the main objective in designing a novel SERD includes an oral pharmacokinetic profile superior to that of fulvestrant and a higher efficacy and potency of ERα degradation (Lu and Liu, 2020; Shagufta et al., 2020). Furthermore, if these novel SERDs were to selectively target ERα, but not ERβ, and thereby reduce the ERα:ERβ ratio that would be an added advantage (O’Boyle et al., 2018). From our results, it is clear that the C. subternata Vogel extracts, SM6Met and CoT, but not the C. genistoides extract, P104, display the most desirable attributes for BC prevention and treatment in downregulating ERα while upregulating ERβ and thereby reducing the ERα:ERβ ratio in both BC cell lines. Comparison of the effects on the ERα:ERβ ratio elicited by the C. subternata Vogel extracts, SM6Met and CoT, with those elicited by the SOCs, fulvestrant and 4-OHT, suggests that SM6Met is slightly less effective than the SERD, fulvestrant, but as effective as the SERM, 4-OHT, while CoT is less effective than both the SOCs. However, the potencies of the C. subternata Vogel extracts, SM6Met and CoT, are generally markedly higher than that of fulvestrant. We have previously shown that the C. subternata Vogel extracts, SM6Met and CoT, are absorbed when administered orally and elicit a biological effect in vivo, specifically by significantly reducing uterine weight and significantly delaying vaginal opening relative to solvent in the immature rat uterotrophic assay (Visser et al., 2013). Furthermore, SM6Met has demonstrated efficacy in reducing tumor mass and volume and increasing tumor free survival in a N-Methyl-N-nitrosourea (MNU)-induced rat mammary gland carcinogenesis model (Visser et al., 2016) and in suppressing tumor growth in an orthotopic model of LA7 cell-induced mammary tumors (Oyenihi et al., 2018). Thus the proven oral bioavailability of the C. subternata Vogel extracts coupled to the generally higher potency and comparable efficacy in vitro SERD activity suggest that these extracts are worthy of further investigation. The downregulation of ERα protein levels by E2 and fulvestrant in both MCF7 and T47D cell lines agrees with previous findings (Power and Thompson, 2003; Yeh et al., 2013; Garner et al., 2015; Joseph et al., 2016; Liu et al., 2016), while the downregulation of ERα protein levels by 4-OHT in the current study contradicts some previous findings (Power and Thompson, 2003; Garner et al., 2015), but is supported by others (Koibuchi et al., 2000; Joseph et al., 2016) in an estrogen-depleted environment as also used in the current study. Specifically, Garner et al. [ 2015] showed that at 48 h, ERα protein levels were completely ($100\%$) downregulated by 1 nM E2 and by 100 nM fulvestrant, while treatment with 1 μM of 4-OHT had no effect on ERα protein levels in MCF7 cells. Likewise, Yeh et al. [ 2013] demonstrated that after 6 h, ERα protein levels in MCF7 cells were downregulated to $35\%$ by 100 nM of E2 and 100 nM of fulvestrant. Also, Joseph et al. [ 2016] performed a dose-response assay and showed that ERα protein levels were downregulated by 1 μM of fulvestrant with an efficacy of $6.4\%$, while 1 μM 4-OHT displayed an efficacy of $51.9\%$ at 4 h in MCF7 cells. Furthermore, Power and Thompson [2003], demonstrated that at 24 h, ERα protein levels were downregulated by 1 nM of E2 in MCF7 cells, while no effect was seen in T47D cells. Also in the same study, 1 μM of 4-OHT was shown to upregulate ERα protein levels in both cell lines. Liu et al. [ 2016] showed that ERα protein levels were downgraded by more than $50\%$ in response to fulvestrant within the concentration range of 0.03–1 μM in T47D cells. Even though our results of the downregulation of ERα protein levels by E2 and fulvestrant, in MCF7 cells agree with the findings above, comparison of the extent of the downregulation of ERα protein levels (efficacy) is difficult due to the different time points used. The discrepancies in our results showing downregulation of ERα protein levels by 4-OHT and E2, with the no effect of 1 nM E2 treatment on ERα levels demonstrated by Power and Thompson (Power and Thompson, 2003) in T47D cells, and the findings of no effect on- and the upregulation of ERα protein levels by 4-OHT for both cell lines as shown by Garner et al. [ 2015] and Power and Thompson [2003], respectively, may be due to the variations in the genotypes of the cell lines used by the different laboratories, the difference in the concentrations of the test compounds and experimental procedures such as different time points used for test compound treatment, as well as different culture conditions and passage number and used by the diverse laboratories (Jones et al., 2000; Bahia et al., 2002; Wenger et al., 2004; Kleensang et al., 2016). Reports on the potencies of E2, fulvestrant and 4-OHT to modulate ER subtypes are rare as few researchers attempt dose-response curves, however, Joseph et al. [ 2016] demonstrated that ERα protein levels were downregulated by fulvestrant with a potency of 0.39 nM, while 4-OHT showed a potency of 0.14 nM in MCF7 cells, which differs slightly from our results showing a potency for fulvestrant of 6.94 × 10−13 M and 2.04 × 10−10 M for 4-OHT. The downregulation of ERβ protein levels by E2 in both MCF7 and T47D cell lines agrees with most previous findings (Peekhaus et al., 2004; Mishra et al., 2016), however, contradicts the findings of Power and Thompson [2003]. Specifically, Mishra et al. [ 2016] showed that ERβ protein levels were downregulated by 1 nM E2 in MCF7 cells, while Peekhaus et al. [ 2004] demonstrated that ERβ protein levels were downregulated by 10 nM E2 in MCF7 cells transfected with an ERβ expression vector. In contrast, Power and Thompson [2003] showed that ERβ protein levels were significantly upregulated by 1 nM E2 in both MCF7 and T47D cells. Furthermore, the upregulation of ERβ protein levels by fulvestrant in both MCF7 and T47D cell lines agrees with Mishra et al. [ 2016] and Peekhaus et al. [ 2004]. Specifically, Mishra et al. [ 2016] showed that ERβ protein levels were upregulated by 1 μM fulvestrant in MCF7 cells, while Peekhaus et al. [ 2004] demonstrated that ERβ protein levels were upregulated by 10 nM fulvestrant in MCF7 cells transfected with an ERβ expression vector. The upregulation of ERβ protein levels by 4-OHT in both MCF7 and T47D cell lines agrees with Peekhaus et al. [ 2004], however, contradicts the findings of Power and Thompson [2003] in T47D, but not MCF7 cells. Specifically, Peekhaus et al. [ 2004] demonstrated that ERβ protein levels were upregulated by 10 nM tamoxifen in MCF7 cells transfected with an ERβ expression vector. Although Power and Thompson [2003] also demonstrated that 24 h treatment of 1 μM 4-OHT significantly upregulated ERβ protein levels in MCF7 cells, they did, however, demonstrate significant downregulation in T47D cells. To recapitulate, E2, fulvestrant and 4-OHT all downregulated ERα protein levels in a concentration-dependent manner in both cell lines, however, the extent of downregulation by 4-OHT was considerably less. In contrast, although E2 downregulated ERβ protein levels, fulvestrant and 4-OHT both significantly elevated ERβ protein levels in both cell lines. Thus, the ERα:ERβ ratio was not greatly affected by E2, however, fulvestrant and 4-OHT greatly reduced the ERα:ERβ ratio confirming their beneficial effects in ER+ BC (Leclercq et al., 2006; Sotoca Covaleda et al., 2008; Pons et al., 2014; Acconcia et al., 2017). Of note, the potency of fulvestrant in upregulating ERβ protein levels was significantly lower in MCF7 than in T47D cells, while the potency of downregulation of ERα protein levels was significantly higher in MCF7 than in T47D cells, which may be because of the high ERα:ERβ ratio in MCF7 and low ERα:ERβ ratio in T47D cells (Pons et al., 2014). To the best of our knowledge, this is the first report of the dose-response modulation of ERα and ERβ protein levels by the ER subtype selective ligands, liquiritigenin and MPP, in BC cell lines. Liquiritigenin repressed ERα protein levels while concurrently increasing ERβ protein levels in both cell lines resulting in a decreased ERα:ERβ ratio. In contrast, MPP upregulated ERα and ERβ protein levels to the same extent in both cell lines and thus did not influence the ERα:ERβ ratio. Although liquiritigenin has been shown to bind to both ERα and ERβ with the same affinity, liquiritigenin specifically activates ERβ transcriptional activity and not that of ERα (Mersereau et al., 2008; Powell and Xu, 2008). Furthermore, the isomeric precursor of liquiritigenin, isoliquiritigenin (Ramalingam et al., 2018) and an extract from licorice root, which also consists of liquiritigenin, had been shown to downregulate ERα protein levels in MCF7 cells (Maggiolini et al., 2002; Hu et al., 2009), while liquiritigenin itself, as found in the current study, significantly downregulated ERα and upregulated ERβ levels in a BT-474 breast cancer cell-derived tumor xenograft model (Liang et al., 2022). The Cyclopia extracts all demonstrate ERα antagonism and ERβ agonism (Visser et al., 2013), however, a comparison of the effects of the ER subtype selective ligands, MPP (ERα antagonist) and liquiritigenin (ERβ agonist), suggests that the ERβ agonist rather than ERα antagonist activity of the Cyclopia extracts is responsible for the modulation of ER subtype levels observed. Regarding the Cyclopia extracts previous work by Visser [2013] showed that 9.8 μg/mL of all three Cyclopia extracts downregulates ERα protein levels in MCF7 cells with efficacies of $89.8\%$, $86.0\%$, and $70.1\%$ for SM6Met, CoT and P104, respectively. Visser did not do dose-response curves and thus potencies cannot be compared but as the concentration used by Visser corresponds to the highest concentration used during the current study, efficacies may be compared. Thus, results indicate that the efficacy for the downregulation of ERα protein levels in MCF7 cells by SM6Met at $68.6\%$ is higher in the current study than the $89.8\%$ shown by Visser, as is the $55.4\%$ downregulation by P104 in the current study compared to the $70.1\%$ shown by Visser. However, in contrast, the extent of downregulation of ERα protein levels in MCF7 cells by CoT in the current study ($82.7\%$) is similar to the $86.0\%$ shown by Visser [2013]. Likewise, Visser [2013] showed that 9.8 μg/mL of all three Cyclopia extracts upregulates ERβ protein levels in MCF7 cells. The efficacy of the upregulation of ERβ protein levels in MCF7 cells by SM6Met is slightly higher at $145.4\%$ in the current study than the $130.8\%$ shown by Visser, as is the efficacy of CoT at $124.5\%$ in the current study compared to the $110.9\%$ shown by Visser. In contrast to that seen by Visser [2013], ERβ protein levels were downregulated by P104 in the current study. Comparison of the attributes of the Cyclopia extracts revealed in the current study with that of other botanicals or plant extracts suggest some similarities. For example, the citrus plant-derived flavanone naringenin had been shown to have little effect on ERα (Acconcia et al., 2017) up to 1 µM but to decrease ERα protein levels at 200 µM (Xu et al., 2018), while increasing ERβ protein levels in MCF7 cells (Xu et al., 2018). Additionally, genistein, the major isoflavonoid found in soybeans, had little effect on ERα protein levels in MCF7 and T47D cells, while strongly increasing ERβ protein levels in T47D, but not MCF7 cells (Pons et al., 2014). Acetyltanshinone IIA (ATA), chemically modified from tanshinone IIA (TIIA), a major compound that was isolated from a medicinal plant, Salvia miltiorrhiza, specifically reduces the protein levels of ERα, but not ERβ, in MCF7 cells (Yu et al., 2014). Furthermore, triptolide, a diterpenoid isolated from the plant *Tripterygium wilfordii* Hook F also decreased ERα protein levels in MCF7 cells (Li et al., 2015), as did artemisinin, an antimalarial sesquiterpene lactone phytochemical isolated from the sweet wormwood plant, Artemisia annua, with the latter also shown to have no effect on ERβ protein levels in MCF7 cells (Sundar et al., 2008). In addition, assessment of the polyphenolic compounds quantified in the C. subternata Vogel extracts, SM6Met and CoT, and the C. genistoides extract, P104, (Table 1) may provide clues to their selective ER subtype downregulation. For instance, the xanthones, mangiferin and isomangiferin, and hespiridin that are present in both C. subternata and C. genistoides, were suggested to possess anti-cancer activities (Wang et al., 2018; Hsu et al., 2021; Yap et al., 2021). Specifically, hesperidin promotes MCF7 cell proliferation in the dose range of 12.5–100 μM (Liu et al., 2008) while displaying anti-proliferative activities above 100 μM (Hsu et al., 2021), downregulates ERα mRNA levels in MCF7 and T47D cells (Khamis et al., 2018), and increases ERβ protein levels in the hypothalamus of ovariectomized mice (Han et al., 2018). Therefore, hesperidin, which is the main (2.049 g/100 g dry extract) polyphenol quantified in SM6Met, and which is present at 2.2-fold higher levels than in CoT (0.935 g/100 g dry extract), may explain the fact that the efficacy of SM6Met in downregulating ERα protein and upregulating ERβ protein levels is generally greater than that of CoT. Hesperidin was not quantified in P104. Isomangiferin is the major polyphenol (Table 1) quantified in the C. genistoides extract, P104 ($\frac{5.094}{100}$ g dry extract), and is 7.9 to 12.1-fold higher than the levels in the C. subternata Vogel extracts, SM6Met and CoT, respectively, and although no work has been done on its effect on ERα or ERβ protein levels it has been shown to inhibit MCF7 cell proliferation and to suppress tumor growth in a mouse breast cancer mouse xenograft model using MDA-MB-231 cells (Wang et al., 2018). It would thus be interesting to evaluate the effects of isomangiferin on the ER subtype protein levels to ascertain if it is responsible for the downregulation of both ER subtypes by P104 as shown in the current study. Mangiferin (Table 1), which is the major polyphenol (Table 1) quantified in the C. subternata Vogel extract, CoT (1 g/100 g dry extract), but is 1.9 to 3.6-fold lower than the levels in the C. subternata Vogel extract, SM6Met and the C. genistoides extract, P104, respectively, have been shown to activate transcription via ERα but not via ERβ (Wilkinson et al., 2015), to inhibit proliferation of MCF7 cells (Li et al., 2013; Lv et al., 2013; Cuccioloni et al., 2016; Min Yap et al., 2021) and to increase ERβ, but not ERα, mRNA expression in bone marrow macrophage cells (Sekiguchi et al., 2017). It is thus difficult to speculate what the effect of mangiferin would be on the ER subtype protein levels and this would have to be investigated in future. Furthermore, luteolin and protocatechuic acid downregulate DHT-induced ERα protein expression and upregulate DHT-suppressed ERβ protein expression in a human prostatic epithelial cell line, BPH-1 (Tao et al., 2019), while luteolin reduces ERα protein expression in MCF7 cells (Wang et al., 2012), selectively transactivates via ERβ but not via ERα in SK-N-BE neuroblastoma cells (Innocenti et al., 2007), but not in HEK293 cells transfected with ER subtypes (Mortimer et al., 2015), inhibits E2-induced ERα transactivation in a yeast assay (Pinto et al., 2008), binds preferentially to ERβ (Verhoog et al., 2007b) and displays partial agonist activity in stimulating MCF7 cell proliferation (Resende et al., 2013) but inhibits E2-induced proliferation in MCF7 cells (Verhoog et al., 2007b). Luteolin is, however, present at very low concentrations in all Cyclopia extracts and is thus unlikely to alone be responsible for the effects of the Cyclopia extracts. In fact, we have previously shown that activity-guided fractionation does not retain all the desirable estrogenic attributes of the original SM6Met in one fraction (Mortimer et al., 2015) and thus it maybe the combinatorial effect of all or several of the compounds in the extracts that contribute to the selective modulation of the ER subtypes. Although some isolated pure phytoestrogen compounds are active against BC, it has been postulated that the range of their activity is less compared to that of crude extracts as the multifactorial reactions and synergy between phytoestrogenic compounds are only present in crude extracts (Gilbert and Alves, 2005; Rasoanaivo et al., 2011). Thus, phytoestrogenic extracts rather than isolated phytoestrogens may increase the likelihood of combining the attributes, such as the ability to downregulate ERα, upregulate ERβ and preferentially decrease the ERα:ERβ ratio, thought to be desirable for BC treatment and prevention. Taken together, our findings show that the molecular mechanism involved in the regulation of ERα and ERβ protein levels may be organized into several types. Those primarily regulated through proteasomal degradation such as E2 and liquiritigenin and those such as MPP primarily regulated through transcriptional and translational mechanisms. Other types involve a mixture of mechanisms, either equally or preferentially leaning towards one of the mechanisms. Specifically, 4-OHT and the Cyclopia extracts, CoT and P104, appear to equally favor proteasomal, and transcriptional and translational mechanisms, while fulvestrant and SM6Met generally favor proteasomal degradation. The regulation of ERα protein levels via proteasomal degradation by E2 and fulvestrant in MCF7 cells agrees with the findings of Zhao et al. [ 2015], and Wijayaratne and McDonnell [2001], while regulation of ERβ protein levels via proteasomal degradation by E2 agrees with the findings of Zhao et al. [ 2015]. Furthermore, Khissiin and Leclercq (Khissiin and Leclercq, 1999) showed that the downregulation of ERα protein levels by E2 in MCF7 cells was via both protein synthesis and proteasomal degradation. Additionally, although not in BC cells, Alarid et al. [ 1999] using CHX and MG132, and the transcription inhibitor, 5,6-DRB demonstrated that ERα protein levels downregulation by E2 was through proteasomal degradation and not via protein synthesis nor transcription in lactotrope cells, PR1. Our report is the first on the molecular mechanism of regulation of ER subtypes by the Cyclopia extracts and to our knowledge also by the ER subtype selective ligands, liquiritigenin and MPP. Despite many studies investigating the selective degradation of the ER for BC treatment (Lu and Liu, 2020; Shagufta et al., 2020; Wang Z. et al., 2021; Kumar et al., 2021; Mottamal et al., 2021), these have mostly focused on ERα with little mention of ERβ. Our study in elaborating on the molecular characteristics and mechanism of action of the Cyclopia extracts has in contrast explicitly evaluated selectively in terms of ER subtype levels. In conclusion, the current study indicates that the C. subternata Vogel extracts, SM6Met and CoT, rather than the C. genistoides extract, P104, display favorable attributes by degrading ERα while stabilizing ERβ. Coupled to the proven oral bioavailability of the C. subternata Vogel extracts (Visser et al., 2013; Visser et al., 2016; Oyenihi et al., 2018) the current study suggests that the C. subternata Vogel extracts may be of therapeutic benefit for BC prevention and treatment and provide the underpinning for the development of an ER-targeted phytopharmaceutical product from Cyclopia. ## Data availability statement The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author. ## Author contributions FO performed the experiments. AL wrote the first draft of the article. FO, NV, and AL interpreted the results and revised the article. All authors contributed to the article and approved the submitted version. ## Conflict of interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ## Publisher’s note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. ## Supplementary material The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar.2023.1122031/full#supplementary-material ## References 1. Acconcia F., Fiocchetti M., Marino M.. **Xenoestrogen regulation of ERα/ERβ balance in hormone-associated cancers**. *Mol. Cell. Endocrinol.* (2017) **457** 3-12. DOI: 10.1016/j.mce.2016.10.033 2. Alarid E. T., Bakopoulos N., Solodin N.. **Proteasome-mediated proteolysis of estrogen receptor: A novel component in autologous down-regulation**. *Mol. Endocrinol.* (1999) **13** 1522-1534. DOI: 10.1210/mend.13.9.0337 3. Bahia H., Ashman J. N., Cawkwell L., Lind M., Monson J. R., Drew P. J.. **Karyotypic variation between independently cultured strains of the cell line MCF-7 identified by multicolour fluorescence**. *Int. J. Oncol.* (2002) **20** 489-494. DOI: 10.3892/ijo.20.3.489 4. Baliga B. S., Pronczuk A. W., Munro H. N.. **Mechanism of cycloheximide inhibition of protein synthesis in a cell-free system prepared from rat liver**. *J. Biol. Chem.* (1969) **244** 4480-4489. DOI: 10.1016/s0021-9258(18)94343-7 5. Beelders T., de Beer D., Stander M. A., Joubert E.. **Comprehensive phenolic profiling of Cyclopia genistoides (L.) Vent. by LC-DAD-MS and -MS/MS reveals novel xanthone and benzophenone constituents**. *Molecules* (2014) **19** 11760-11790. DOI: 10.3390/molecules190811760 6. Chellan N., Joubert E., Strijdom H., Roux C., Louw J., Muller C. J. F.. **Aqueous extract of unfermented honeybush (cyclopia maculata) attenuates stz-induced diabetes and β-cell cytotoxicity**. *Planta Med.* (2014) **80** 622-629. DOI: 10.1055/s-0034-1368457 7. Chintamunnee V., Mahomoodally M. F.. **Herbal medicine commonly used against non-communicable diseases in the tropical island of Mauritius**. *J. Herb. Med.* (2012) **2** 113-125. DOI: 10.1016/j.hermed.2012.06.001 8. Chopra A., Willmore W. G., Biggar K. K.. **Protein quantification and visualization via ultraviolet-dependent labeling with 2,2,2-trichloroethanol**. *Sci. Rep.* (2019) **9** 13923. DOI: 10.1038/s41598-019-50385-9 9. Clarke R., Tyson J. J., Dixon J. M.. **Endocrine resistance in breast cancer-An overview and update**. *Mol. Cell. Endocrinol.* (2015) **418** 220-234. DOI: 10.1016/j.mce.2015.09.035 10. Costa B., Amorim I., Gärtner F., Vale N.. **Understanding breast cancer: From conventional therapies to repurposed drugs**. *Eur. J. Pharm. Sci.* (2020) **151** 105401. DOI: 10.1016/j.ejps.2020.105401 11. Croxtall J. D., McKeage K.. **Fulvestrant: A review of its use in the management of hormone receptor-positive metastatic breast cancer in postmenopausal women**. *Fulvestrant.* (2011) **71** 363-380. DOI: 10.2165/11204810-000000000-00000 12. Cuccioloni M., Bonfili L., Mozzicafreddo M., Cecarini V., Scuri S., Cocchioni M.. **Mangiferin blocks proliferation and induces apoptosis of breast cancer cells: Via suppression of the mevalonate pathway and by proteasome inhibition**. *Food Funct.* (2016) **7** 4299-4309. DOI: 10.1039/c6fo01037g 13. Dalglish S. L., Straubinger S., Kavle J. A., Gibson L., Mbombeshayi E., Anzolo J.. **Who are the real community health workers in Tshopo Province, Democratic Republic of the Congo?**. *BMJ Glob. Health* (2019) **4** e001529-9. DOI: 10.1136/bmjgh-2019-001529 14. Datta J., Willingham N., Manouchehri J. M., Schnell P., Sheth M., David J. J.. **Activity of estrogen receptor β agonists in therapy-resistant estrogen receptor-positive breast cancer**. *Front. Oncol.* (2022) **12** 857590. DOI: 10.3389/fonc.2022.857590 15. DeSantis C. E., Bray F., Ferlay J., Lortet-Tieulent J., Anderson B. O., Jemal A.. **International variation in female breast cancer incidence and mortality rates**. *Cancer Epidemiol. Biomarkers Prev.* (2015) **24** 1495-1506. DOI: 10.1158/1055-9965.EPI-15-0535 16. DeSantis C. E., Ma J., Gaudet M. M., Newman L. A., Miller K. D., Goding Sauer A.. **Breast cancer statistics, 2019**. *CA Cancer J. Clin.* (2019) **69** 438-451. DOI: 10.3322/caac.21583 17. Downton T., Zhou F., Segara D., Jeselsohn R., Lim E.. **Oral selective estrogen receptor degraders (SERDs) in breast cancer: Advances, challenges, and current status**. *Drug Des. Devel Ther.* (2022) **16** 2933-2948. DOI: 10.2147/DDDT.S380925 18. Ekor M.. **The growing use of herbal medicines: Issues relating to adverse reactions and challenges in monitoring safety**. *Front. Neurol.* (2014) **4** 1-10. DOI: 10.3389/fphar.2013.00177 19. Evers N. M., van den Berg J. H. J., Wang S., Melchers D., Houtman R., de Haan L. H. J.. **Cell proliferation and modulation of interaction of estrogen receptors with coregulators induced by ERα and ERβ agonists**. *J. Steroid Biochem. Mol. Biol.* (2014) **143** 376-385. DOI: 10.1016/j.jsbmb.2014.06.002 20. Fan M., Nakshatri H., Nephew K. P.. **Inhibiting proteasomal proteolysis sustains estrogen receptor-α activation**. *Mol. Endocrinol.* (2004) **18** 2603-2615. DOI: 10.1210/me.2004-0164 21. Farkas S., Szabó A., Hegyi A. E., Török B., Fazekas C. L., Ernszt D.. **Estradiol and estrogen-like alternative therapies in use: The importance of the selective and non-classical actions**. *Biomedicines* (2022) **10** 861. DOI: 10.3390/biomedicines10040861 22. Franzoi M. A., Agostinetto E., Perachino M., del Mastro L., de Azambuja E., Vaz-Luis I.. **Evidence-based approaches for the management of side-effects of adjuvant endocrine therapy in patients with breast cancer**. *Lancet Oncol.* (2021) **22** e303-e313. DOI: 10.1016/S1470-2045(20)30666-5 23. Garner F., Shomali M., Paquin D., Lyttle C. R., Hattersley G.. **RAD1901: A novel, orally bioavailable selective estrogen receptor degrader that demonstrates antitumor activity in breast cancer xenograft models**. *Anticancer Drugs* (2015) **26** 948-956. DOI: 10.1097/CAD.0000000000000271 24. Gilbert B., Alves L.. **Synergy in plant medicines**. *Curr. Med. Chem.* (2005) **10** 13-20. DOI: 10.2174/0929867033368583 25. Girgert R., Emons G., Gründker C.. **Estrogen signaling in erα-negative breast cancer: ERβ and GPER**. *Front. Endocrinol. (Lausanne)* (2019) **10** 781. DOI: 10.3389/fendo.2018.00781 26. Gonzalez T. L., Rae J. M., Colacino J. A.. **Implication of environmental estrogens on breast cancer treatment and progression**. *Toxicology* (2019) **421** 41-48. DOI: 10.1016/j.tox.2019.03.014 27. Gurib-Fakim A.. **Medicinal plants: Traditions of yesterday and drugs of tomorrow**. *Mol. Asp. Med.* (2006) **27** 1-93. DOI: 10.1016/j.mam.2005.07.008 28. Gustafsson J. A., Warner M.. **Estrogen receptor beta in the breast: Role in estrogen responsiveness and development of breast cancer**. *J. Steroid Biochem. Mol. Biol.* (2000) **74** 245-248. DOI: 10.1016/s0960-0760(00)00130-8 29. Han N. R., Nam S. Y., Hong S., Kim H. Y., Moon P. D., Kim H. J.. **Improvement effects of a mixed extract of flowers of Pueraria thomsonii Benth. and peels of Citrus unshiu Markovich on postmenopausal symptoms of ovariectomized mice**. *Biomed. Pharmacother.* (2018) **103** 524-530. DOI: 10.1016/j.biopha.2018.04.070 30. Hartman J., Ström A., Gustafsson J. Å.. **Estrogen receptor beta in breast cancer-Diagnostic and therapeutic implications**. *Steroids* (2009) **74** 635-641. DOI: 10.1016/j.steroids.2009.02.005 31. Hernando C., Ortega-Morillo B., Tapia M., Moragón S., Martínez M. T., Eroles P.. **Oral selective estrogen receptor degraders (Serds) as a novel breast cancer therapy: Present and future from a clinical perspective**. *Int. J. Mol. Sci.* (2021) **22** 7812. DOI: 10.3390/ijms22157812 32. Hirao-Suzuki M.. **Estrogen receptor β as a possible double-edged sword molecule in breast cancer: A mechanism of alteration of its role by exposure to endocrine-disrupting chemicals**. *Biol. Pharm. Bull.* (2021) **44** 1594-1597. DOI: 10.1248/bpb.b21-00468 33. Hsu P. H., Chen W. H., Juanlu C., Hsieh S. C., Lin S. C., Mai R. T.. **Hesperidin and chlorogenic acid synergistically inhibit the growth of breast cancer cells via estrogen receptor/mitochondrial pathway**. *Life* (2021) **11** 950. DOI: 10.3390/life11090950 34. Hu C., Liu H., Du J., Mo B., Qi H., Wang X.. **Estrogenic activities of extracts of Chinese licorice (Glycyrrhiza uralensis) root in MCF-7 breast cancer cells**. *J. Steroid Biochem. Mol. Biol.* (2009) **113** 209-216. DOI: 10.1016/j.jsbmb.2008.12.019 35. Huang B., Warner M., Gustafsson J. Å.. **Estrogen receptors in breast carcinogenesis and endocrine therapy**. *Mol. Cell. Endocrinol.* (2015) **418** 240-244. DOI: 10.1016/j.mce.2014.11.015 36. Human C., Danton O., de Beer D., Maruyama T., Alexander L., Malherbe C.. **Identification of a novel di-C-glycosyl dihydrochalcone and the thermal stability of polyphenols in model ready-to-drink beverage solutions with Cyclopia subternata extract as functional ingredient**. *Food Chem.* (2021) **351** 129273. DOI: 10.1016/j.foodchem.2021.129273 37. Innocenti G., Vegeto E., Dall’Acqua S., Ciana P., Giorgetti M., Agradi E.. *Phytomedicine* (2007) **14** 147-152. DOI: 10.1016/j.phymed.2006.05.005 38. Joubert E., de Beer D., Malherbe C. J. J., Muller M., Louw A., Gelderblom W. C. A. C. A.. **Formal honeybush tea industry reaches 20-year milestone – progress of product research targeting phenolic composition, quality and bioactivity**. *South Afr. J. Bot.* (2019) **127** 58-79. DOI: 10.1016/j.sajb.2019.08.027 39. Jack B. U., Malherbe C. J., Willenburg E. L., De Beer D., Huisamen B., Joubert E.. **Polyphenol-enriched fractions of cyclopia intermedia selectively affect lipogenesis and lipolysis in 3T3-L1 adipocytes**. *Planta Med.* (2018) **84** 100-110. DOI: 10.1055/s-0043-119463 40. Jones C., Payne J., Wells D., Delhanty J. D. A., Lakhani S. R., Kortenkamp A.. **Comparative genomic hybridization reveals extensive variation among different MCF-7 cell stocks**. *Cancer Genet. Cytogenet* (2000) **117** 153-158. DOI: 10.1016/S0165-4608(99)00158-2 41. Jordan V. C.. **Tamoxifen: A most unlikely pioneering medicine**. *Nat. Rev. Drug Discov.* (2003) **2** 205-213. DOI: 10.1038/nrd1031 42. Joseph J. D., Darimont B., Zhou W., Arrazate A., Young A., Ingalla E.. **The selective estrogen receptor downregulator GDC-0810 is efficacious in diverse models of ER+ breast cancer**. *Elife* (2016) **5** e15828. DOI: 10.7554/eLife.15828 43. Joubert E., Gelderblom W. C. A. C. A., Louw A., de Beer D.. **South African herbal teas: Aspalathus linearis, Cyclopia spp. and Athrixia phylicoides-A review**. *J. Ethnopharmacol.* (2008) **119** 376-412. DOI: 10.1016/j.jep.2008.06.014 44. Keydar I., Chen L., Karby S., Weiss F. R., Delarea J., Radu M.. **Establishment and characterization of a cell line of human breast carcinoma origin**. *Eur. J. Cancer* (1979) **15** 659-670. DOI: 10.1016/0014-2964(79)90139-7 45. Khamis A. A. A., Ali E. M. M., El-Moneim M. A. A., Abd-Alhaseeb M. M., El-Magd M. A., Salim E. I.. **Hesperidin, piperine and bee venom synergistically potentiate the anticancer effect of tamoxifen against breast cancer cells**. *Biomed. Pharmacother.* (2018) **105** 1335-1343. DOI: 10.1016/j.biopha.2018.06.105 46. Khissiin A., Leclercq G.. **Implication of proteasome in estrogen receptor degradation**. *FEBS Lett.* (1999) **448** 160-166. DOI: 10.1016/S0014-5793(99)00343-9 47. Kleensang A., Vantangoli M. M., Odwin-Dacosta S., Andersen M. E., Boekelheide K., Bouhifd M.. **Genetic variability in a frozen batch of MCF-7 cells invisible in routine authentication affecting cell function**. *Sci. Rep.* (2016) **6** 28994. DOI: 10.1038/srep28994 48. Koibuchi Y., Iino Y., Uchida T., Andoh T., Horii Y., Nagasawa M.. **Regulation of estrogen receptor and epidermal growth factor receptor by tamoxifen under high and low estrogen environments in MCF-7 cells grown in athymic mice**. *Oncol. Rep.* (2000) **7** 135-140. DOI: 10.3892/or.7.1.135 49. Kondakova I. v., Shashova E. E., Sidenko E. A., Astakhova T. M., Zakharova L. A., Sharova N. P.. **Estrogen receptors and ubiquitin proteasome system: Mutual regulation**. *Biomolecules* (2020) **10** 500. DOI: 10.3390/biom10040500 50. Kumar N., Gulati H. K., Sharma A., Heer S., Jassal A. K., Arora L.. **Most recent strategies targeting estrogen receptor alpha for the treatment of breast cancer**. *Mol. Divers* (2021) **25** 603-624. DOI: 10.1007/s11030-020-10133-y 51. Lacroix M., Leclercq G.. **Relevance of breast cancer cell lines as models for breast tumours: An update**. *Breast Cancer Res. Treat.* (2004) **83** 249-289. DOI: 10.1023/B:BREA.0000014042.54925.cc 52. Leclercq G., Lacroix M., Laios I., Laurent G.. **Estrogen receptor alpha: Impact of ligands on intracellular shuttling and turnover rate in breast cancer cells**. *Curr. Cancer Drug Targets* (2006) **6** 39-64. DOI: 10.2174/156800906775471716 53. Leung Y. K., Lee M. T., Lam H. M., Tarapore P., Ho S. M.. **Estrogen receptor-beta and breast cancer: Translating biology into clinical practice**. *Steroids* (2012) **77** 727-737. DOI: 10.1016/j.steroids.2012.03.008 54. Li H., Huang J., Yang B., Xiang T., Yin X., Peng W.. **Mangiferin exerts antitumor activity in breast cancer cells by regulating matrix metalloproteinases, epithelial to mesenchymal transition, and β-catenin signaling pathway**. *Toxicol. Appl. Pharmacol.* (2013) **272** 180-190. DOI: 10.1016/j.taap.2013.05.011 55. Li H., Pan G. F., Jiang Z. Z., Yang J., Sun L. X., Zhang L. Y.. **Triptolide inhibits human breast cancer MCF-7 cell growth via downregulation of the ERα-mediated signaling pathway**. *Acta Pharmacol. Sin.* (2015) **36** 606-613. DOI: 10.1038/aps.2014.162 56. Liang Y., Besch-Williford C., Hyder S. M.. **The estrogen receptor beta agonist liquiritigenin enhances the inhibitory effects of the cholesterol biosynthesis inhibitor RO 48-8071 on hormone-dependent breast-cancer growth**. *Breast Cancer Res. Treat.* (2022) **192** 53-63. DOI: 10.1007/s10549-021-06487-y 57. Liu J., Zheng S., Akerstrom V. L., Yuan C., Ma Y., Zhong Q.. **Fulvestrant-3 boronic acid (ZB716): An orally bioavailable selective estrogen receptor downregulator (SERD)**. *J. Med. Chem.* (2016) **59** 8134-8140. DOI: 10.1021/acs.jmedchem.6b00753 58. Liu L., Xu D. M., Cheng Y. Y.. **Distinct effects of naringenin and hesperetin on nitric oxide production from endothelial cells**. *J. Agric. Food Chem.* (2008) **56** 824-829. DOI: 10.1021/jf0723007 59. Lopes C. M., Dourado A., Oliveira R.. **Phytotherapy and nutritional supplements on breast cancer**. *Biomed. Res. Int.* (2017) **2017** 7207983. DOI: 10.1155/2017/7207983 60. Louw A., Joubert E., Visser K.. **Phytoestrogenic potential of cyclopia extracts and polyphenols**. *Planta Med.* (2013) **79** 580-590. DOI: 10.1055/s-0032-1328463 61. Lu Y., Liu W.. **Selective estrogen receptor degraders (SERDs): A promising strategy for estrogen receptor positive endocrine-resistant breast cancer**. *J. Med. Chem.* (2020) **63** 15094-15114. DOI: 10.1021/acs.jmedchem.0c00913 62. Lv J., Wang Z., Zhang L., Wang H. L., Liu Y., Li C.. **Mangiferin induces apoptosis and cell cycle arrest in MCF-7 cells both**. *J. Animal Veterinary Adv.* (2013) **12** 352-359. DOI: 10.3923/javaa.2013.352-359 63. Maggiolini M., Statti G., Vivacqua A., Gabriele S., Rago V., Loizzo M.. **Estrogenic and antiproliferative activities of isoliquiritigenin in MCF7 breast cancer cells**. *J. Steroid Biochem. Mol. Biol.* (2002) **82** 315-322. DOI: 10.1016/S0960-0760(02)00230-3 64. Mal R., Magner A., David J., Datta J., Vallabhaneni M., Kassem M.. **Estrogen receptor beta (ERβ): A ligand activated tumor suppressor**. *Front. Oncol.* (2020) **10** 587386. DOI: 10.3389/fonc.2020.587386 65. Mbendana D., Mamabolo K., Truter M., Kritzinger Q., Ndhlala A. R.. **Practices at herbal (muthi) markets in gauteng, South Africa and their impact on the health of the consumers: A case study of KwaMai-mai and marabastad muthi markets**. *South Afr. J. Bot.* (2019) **126** 30-39. DOI: 10.1016/j.sajb.2019.05.004 66. Mersereau J. E., Levy N., Staub R. E., Baggett S., Zogric T., Chow S.. **Liquiritigenin is a plant-derived highly selective estrogen receptor β agonist**. *Mol. Cell. Endocrinol.* (2008) **283** 49-57. DOI: 10.1016/j.mce.2007.11.020 67. Mfenyana C., DeBeer D., Joubert E., Louw A.. **Selective extraction of Cyclopia for enhanced**. *J. Steroid Biochem. Mol. Biol.* (2008) **112** 74-86. DOI: 10.1016/j.jsbmb.2008.08.005 68. Min Yap K., Sekar M., Jing Seow L., Hua Gan S., Reddy Bonam S., Najihah Izzati Mat Rani N.. **Mangifera indica (mango): A promising medicinal plant for breast cancer therapy and understanding its potential mechanisms of action**. (2021). DOI: 10.2147/BCTT.S316667 69. Mishra A. K., Abrahamsson A., Dabrosin C.. **Fulvestrant inhibits growth of triple negative breast cancer and synergizes with tamoxifen in ERα positive breast cancer by up-regulation of ERβ**. *Oncotarget* (2016) **7** 56876-56888. DOI: 10.18632/oncotarget.10871 70. Mortimer M., Visser K., de Beer D., Joubert E., Louw A.. **Divide and conquer may not be the optimal approach to retain the desirable estrogenic attributes of the cyclopia nutraceutical extract, SM6Met**. *PLoS One* (2015) **10** e0132950. DOI: 10.1371/journal.pone.0132950 71. Mottamal M., Kang B., Peng X., Wang G.. **From pure antagonists to pure degraders of the estrogen receptor: Evolving strategies for the same target**. *ACS Omega* (2021) **6** 9334-9343. DOI: 10.1021/acsomega.0c06362 72. Murakami S., Miura Y., Hattori M., Matsuda H., Malherbe C. J., Muller C. J. F.. **Cyclopia extracts enhance Th1-Th2-and Th17-type T cell responses and induce Foxp3 + cells in murine cell culture**. *Planta Med.* (2018) **84** 311-319. DOI: 10.1055/s-0043-121270 73. Nadal-Serrano M., Sastre-Serra J., Pons D. G., Miró A. M., Oliver J., Roca P.. **The ERalpha/ERbeta ratio determines oxidative stress in breast cancer cell lines in response to 17Beta-estradiol**. *J. Cell. Biochem.* (2012) **113** 3178-3185. DOI: 10.1002/jcb.24192 74. Nathan M. R., Schmid P., Schmid M. R. N. P.. **A review of fulvestrant in breast cancer**. *Oncol. Ther.* (2017) **5** 17-29. DOI: 10.1007/s40487-017-0046-2 75. Nilsson S., Koehler K. F., Gustafsson J. Å.. **Development of subtype-selective oestrogen receptor-based therapeutics**. *Nat. Rev. Drug Discov.* (2011) **10** 778-792. DOI: 10.1038/nrd3551 76. O’Boyle N. M., Barrett I., Greene L. M., Carr M., Fayne D., Twamley B.. **Lead optimization of benzoxepin-type selective estrogen receptor (ER) modulators and downregulators with subtype-specific ERα and ERβ activity**. *J. Med. Chem.* (2018) **61** 514-534. DOI: 10.1021/acs.jmedchem.6b01917 77. Oyenihi O. R., Krygsman A., Verhoog N., de Beer D., Saayman M. J., Mouton T. M.. **Chemoprevention of LA7-induced mammary tumor growth by SM6Met, a well-characterized Cyclopia extract**. *Front. Pharmacol.* (2018) **9** 650. DOI: 10.3389/fphar.2018.00650 78. Pal K.. **Retracing our path towards mother nature: A cost-effective green therapy for the ameliorative of recalcitrant triple negative breast cancer - a brief report**. *Acad. Lett.* (2021) **1466**. DOI: 10.20935/al1466 79. Peekhaus N. T., Chang T., Hayes E. C., Wilkinson H. A., Mitra S. W., Schaeffer J. M.. **Distinct effects of the antiestrogen Faslodex on the stability of estrogen receptors-α and -β in the breast cancer cell line MCF-7**. *J. Mol. Endocrinol.* (2004) **32** 987-995. DOI: 10.1677/jme.0.0320987 80. Pelkonen O., Xu Q., Fan T. P.. **Why is research on herbal medicinal products important and how can we improve its quality?**. *J. Tradit. Complement. Med.* (2014) **4** 1-7. DOI: 10.4103/2225-4110.124323 81. Perry R. R., Kang Y., Greaves B.. **Effects of tamoxifen on growth and apoptosis of estrogen-dependent and -independent human breast cancer cells**. *Ann. Surg. Oncol.* (1995) **2** 238-245. DOI: 10.1007/BF02307030 82. Pheiffer C., Dudhia Z., Louw J., Muller C., Joubert E.. **Cyclopia maculata (honeybush tea) stimulates lipolysis in 3T3-L1 adipocytes**. *Phytomedicine* (2013) **20** 1168-1171. DOI: 10.1016/j.phymed.2013.06.016 83. Pink J. J., Jordan V. C.. **Models of estrogen receptor regulation by estrogens and antiestrogens in breast cancer cell lines**. *Cancer Res.* (1996) **56** 2321-2330. PMID: 8625307 84. Pinto B., Bertoli A., Noccioli C., Garritano S., Reali D., Pistelli L.. **Estradiol-antagonistic activity of phenolic compounds from leguminous plants**. *Phytotherapy Res.* (2008) **22** 362-366. DOI: 10.1002/ptr.2327 85. Pons D. G., Nadal-Serrano M., Blanquer-Rossello M. M., Sastre-Serra J., Oliver J., Roca P.. **Genistein modulates proliferation and mitochondrial functionality in breast cancer cells depending on ERalpha/ERbeta ratio**. *J. Cell. Biochem.* (2014) **115** 949-958. DOI: 10.1002/jcb.24737 86. Powell E., Xu W.. **Intermolecular interactions identify ligand-selective activity of estrogen receptor alpha/beta dimers**. *Proc. Natl. Acad. Sci.* (2008) **105** 19012-19017. DOI: 10.1073/pnas.0807274105 87. Power K. A., Thompson L. U.. **Ligand-induced regulation of ERalpha and ERbeta is indicative of human breast cancer cell proliferation**. *Breast Cancer Res. Treat.* (2003) **81** 209-221. DOI: 10.1023/A:1026114501364 88. Ramalingam M., Kim H., Lee Y. il, Lee Y. il. **Phytochemical and pharmacological role of liquiritigenin and isoliquiritigenin from radix glycyrrhizae in human health and disease models**. *Front. Aging Neurosci.* (2018) **10** 348. DOI: 10.3389/fnagi.2018.00348 89. Ramani K. v., Ramani H., Alurkar S., Ajaikumar B. S., Trivedi R. G.. *Breast cancer: Medical treatment, side effects, and complementary therapies* (2017) 90. Rasoanaivo P., Wright C. W., Willcox M. L., Gilbert B.. **Whole plant extracts versus single compounds for the treatment of malaria: Synergy and positive interactions**. *Malar. J.* (2011) **10** S4. DOI: 10.1186/1475-2875-10-S1-S4 91. Resende F. A., de Oliveira A. P. S., de Camargo M. S., Vilegas W., Varanda E. A.. **Evaluation of estrogenic potential of flavonoids using a recombinant yeast strain and MCF7/BUS cell proliferation assay**. *PLoS One* (2013) **8** e74881. DOI: 10.1371/journal.pone.0074881 92. Roza O., Lai W. C., Zupkó I., Hohmann J., Jedlinszki N., Chang F. R.. **Bioactivity guided isolation of phytoestrogenic compounds from Cyclopia genistoides by the pER8:GUS reporter system**. *South Afr. J. Bot.* (2017) **110** 201-207. DOI: 10.1016/j.sajb.2016.06.001 93. Rozeboom B., Dey N., De P.. **ER+ metastatic breast cancer: Past, present, and a prescription for an apoptosis-targeted future**. *Am. J. Cancer Res.* (2019) **9** 2821-2831. PMID: 31911865 94. Saha T., Makar S., Swetha R., Gutti G., Singh S. K.. **Estrogen signaling: An emanating therapeutic target for breast cancer treatment**. *Eur. J. Med. Chem.* (2019) **177** 116-143. DOI: 10.1016/j.ejmech.2019.05.023 95. Sayed R. E., Jamal L. E., Iskandarani S. E., Kort J., Abdel Salam M., Assi H.. **Endocrine and targeted therapy for hormone-receptor-positive, HER2-negative advanced breast cancer: Insights to sequencing treatment and overcoming resistance based on clinical trials**. *Front. Oncol.* (2019) **9** 510. DOI: 10.3389/fonc.2019.00510 96. Schneider-Poetsch T., Ju J., Eyler D. E., Dang Y., Bhat S., Merrick W. C.. **Inhibition of eukaryotic translation elongation by cycloheximide and lactimidomycin**. *Nat. Chem. Biol.* (2010) **6** 209-217. DOI: 10.1038/nchembio.304 97. Schulze A. E., De Beer D., Mazibuko S. E., Muller C. J. F., Roux C., Willenburg E. L.. **Assessing similarity analysis of chromatographic fingerprints of Cyclopia subternata extracts as potential screening tool for**. *Anal. Bioanal. Chem.* (2016) **408** 639-649. DOI: 10.1007/s00216-015-9147-7 98. Sekiguchi Y., Mano H., Nakatani S., Shimizu J., Kataoka A., Ogura K.. **Mangiferin positively regulates osteoblast differentiation and suppresses osteoclast differentiation**. *Mol. Med. Rep.* (2017) **16** 1328-1332. DOI: 10.3892/mmr.2017.6752 99. Shagufta I., Ahmad S., Mathew S.. **Recent progress in selective estrogen receptor downregulators (SERDs) for the treatment of breast cancer**. *RSC Med. Chem.* (2020) **11** 438-454. DOI: 10.1039/c9md00570f 100. Song P., Li Y., Dong Y., Liang Y., Qu H., Qi D.. **Estrogen receptor β inhibits breast cancer cells migration and invasion through CLDN6-mediated autophagy**. *J. Exp. Clin. Cancer Res.* (2019) **38** 354. DOI: 10.1186/s13046-019-1359-9 101. Soto A. M., Sonnenschein C., Chung K. L., Fernandez M. F., Olea N., Olea Serrano F.. **The E-SCREEN assay as a tool to identify estrogens: An update on estrogenic environmental pollutants**. *Environ. Health Perspect.* (1995) **103** 113-122. DOI: 10.1289/ehp.95103s7113 102. Sotoca Covaleda A. M., van den Berg H., Vervoort J., van der Saag P., Ström A., Gustafsson J.-Å.. **Influence of cellular ERalpha/ERbeta ratio on the ERalpha-agonist induced proliferation of human T47D breast cancer cells**. *Toxicol. Sci.* (2008) **105** 303-311. DOI: 10.1093/toxsci/kfn141 103. Sundar S. N., Marconett C. N., Doan V. B., Willoughby J. A., Firestone G. L.. **Artemisinin selectively decreases functional levels of estrogen receptor-alpha and ablates estrogen-induced proliferation in human breast cancer cells**. *Carcinogenesis* (2008) **29** 2252-2258. DOI: 10.1093/carcin/bgn214 104. Szostakowska M., Trębińska-Stryjewska A., Grzybowska E. A., Fabisiewicz A.. **Resistance to endocrine therapy in breast cancer: Molecular mechanisms and future goals**. *Breast Cancer Res. Treat.* (2019) **173** 489-497. DOI: 10.1007/s10549-018-5023-4 105. Tahvilian R., Shahriari S., Faramarzi A., Komasi A.. **Ethno-pharmaceutical formulations in Kurdish ethno-medicine**. *Iran. J. Pharm. Res.* (2014) **13** 1029-1039. PMID: 25276205 106. Tao R., Miao L., Yu X., Orgah J. O., Barnabas O., Chang Y.. **Cynomorium songaricum Rupr demonstrates phytoestrogenic or phytoandrogenic like activities that attenuates benign prostatic hyperplasia via regulating steroid 5-α-reductase**. *J. Ethnopharmacol.* (2019) **235** 65-74. DOI: 10.1016/j.jep.2019.01.038 107. van Dyk L.. **Combinatorial treatments of tamoxifen with SM6Met, a selective estrogen receptor subtype modulator (SERSM), from Cyclopia subternata are superior to current endocrine treatments in breast cancer cell models**. (2018) 108. van Wyk A. S., Prinsloo G.. **Medicinal plant harvesting, sustainability and cultivation in South Africa**. *Biol. Conserv.* (2018) **227** 335-342. DOI: 10.1016/j.biocon.2018.09.018 109. Verhoog N. J. D., Joubert E., Louw A.. **Evaluation of the phytoestrogenic activity of Cyclopia genistoides (honeybush) methanol extracts and relevant polyphenols**. *J. Agric. Food Chem.* (2007a) **55** 4371-4381. DOI: 10.1021/jf063588n 110. Verhoog N. J. D., Joubert E., Louw A.. **Screening of four Cyclopia (honeybush) species for putative phyto-oestrogenic activity by oestrogen receptor binding assays**. *S Afr. J. Sci.* (2007b) **103** 13-21 111. Veziari Y., Kumar S., Leach M.. **Addressing barriers to the conduct and application of research in complementary and alternative medicine: A scoping review**. *BMC Complement. Med. Ther.* (2021) **21** 201. DOI: 10.1186/s12906-021-03371-6 112. Visagie A., Kasonga A., Deepak V., Moosa S., Marais S., Kruger M. C.. **Commercial honeybush (cyclopia spp.) tea extract inhibits osteoclast formation and bone resorption in RAW264.7 murine macrophages—an**. *Int. J. Environ. Res. Public Health* (2015) **12** 13779-13793. DOI: 10.3390/ijerph121113779 113. Visser J. A. K.. **Phytoestrogenic extracts of cyclopia modulate molecular targets involved in the prevention and treatment of breast cancer**. (2013) 114. Visser K., Mortimer M., Louw A.. **Cyclopia extracts act as ERα antagonists and ERβ agonists,**. *PLoS One* (2013) **8** e79223. DOI: 10.1371/journal.pone.0079223 115. Visser K., Zierau O., Macejová D., Goerl F., Muders M., Baretton G. B. G. B.. **The phytoestrogenic Cyclopia extract, SM6Met, increases median tumor free survival and reduces tumor mass and volume in chemically induced rat mammary gland carcinogenesis**. *J. Steroid Biochem. Mol. Biol.* (2016) **163** 129-135. DOI: 10.1016/j.jsbmb.2016.04.019 116. Wang B., Shen J., Wang Z., Liu J., Ning Z., Hu M.. **Isomangiferin, a novel potent vascular endothelial growth factor receptor 2 kinase inhibitor, suppresses breast cancer growth, metastasis and angiogenesis**. *J. Breast Cancer* (2018) **21** 11-20. DOI: 10.4048/jbc.2018.21.1.11 117. Wang K., Chen Q., Shao Y., Yin S., Liu C., Liu Y.. **Anticancer activities of TCM and their active components against tumor metastasis**. *Biomed. Pharmacother.* (2021) **133** 111044. DOI: 10.1016/j.biopha.2020.111044 118. Wang L. M., Xie K. P., Huo H. N., Shang F., Zou W., Xie M. J.. **Luteolin inhibits proliferation induced by IGF-1 pathway dependent ERα in human breast cancer MCF-7 cells**. *Asian Pac. J. Cancer Prev.* (2012) **13** 1431-1437. DOI: 10.7314/APJCP.2012.13.4.1431 119. Wang Z., Ma Z., Shen Z.. **Selective degradation of the estrogen receptor in the treatment of cancers**. *J. Steroid Biochem. Mol. Biol.* (2021) **209** 105848. DOI: 10.1016/j.jsbmb.2021.105848 120. Wangkheirakpam S.. **Traditional and folk medicine as a target for drug discovery**. *Natural products and drug discovery* (2018) 29-56 121. Wenger S. L., Senft J. R., Sargent L. M., Bamezai R., Bairwa N., Grant S. G.. **Comparison of established cell lines at different passages by karyotype and comparative genomic hybridization**. *Biosci. Rep.* (2004) **24** 631-639. DOI: 10.1007/s10540-005-2797-5 122. WHO Report (2019). WHO Global report on traditional and complementary medicine 2019. Geneva, Switzerland: World Health Organization.. *WHO Global report on traditional and complementary medicine 2019* (2019) 123. Wijayaratne A. L., McDonnell D. P.. **The human estrogen receptor-α is a ubiquitinated protein whose stability is affected differentially by agonists, antagonists, and selective estrogen receptor modulators**. *J. Biol. Chem.* (2001) **276** 35684-35692. DOI: 10.1074/jbc.M101097200 124. Wilkinson A. S., Taing M. W., Pierson J. T., Lin C. N., Dietzgen R. G., Shaw P. N.. **Estrogen modulation properties of mangiferin and quercetin and the mangiferin metabolite norathyriol**. *Food Funct.* (2015) **6** 1847-1854. DOI: 10.1039/c5fo00133a 125. Xu Z., Huang B., Liu J., Wu X., Luo N., Wang X.. **Combinatorial anti-proliferative effects of tamoxifen and naringenin: The role of four estrogen receptor subtypes**. *Toxicology* (2018) **410** 231-246. DOI: 10.1016/j.tox.2018.08.013 126. Yang L. N., Wu Z. L., Yang Z. J., Li S. G., Ouyang C. S.. **Exploring mechanism of key Chinese herbal medicine on breast cancer by data mining and network Pharmacology methods**. *Chin. J. Integr. Med.* (2021) **27** 919-926. DOI: 10.1007/s11655-020-3422-y 127. Yap K. M., Sekar M., Seow L. J., Gan S. H., Bonam S. R., Mat Rani N. N. I.. **Mangifera indica (mango): A promising medicinal plant for breast cancer therapy and understanding its potential mechanisms of action**. *Breast Cancer Targets Ther.* (2021) **13** 471-503. DOI: 10.2147/BCTT.S316667 128. Yeh W. L., Shioda K., Coser K. R., Rivizzigno D., McSweeney K. R., Shioda T.. **Fulvestrant-induced cell death and proteasomal degradation of estrogen receptor α protein in MCF-7 cells require the CSK c-src tyrosine kinase**. *PLoS One* (2013) **8** e60889. DOI: 10.1371/journal.pone.0060889 129. Yu T., Zhou Z., Mu Y., de Lima Lopes G., Luo K. Q.. **A novel anti-cancer agent, acetyltanshinone IIA, inhibits oestrogen receptor positive breast cancer cell growth by down-regulating the oestrogen receptor**. *Cancer Lett.* (2014) **346** 94-103. DOI: 10.1016/j.canlet.2013.12.023 130. Zhao Z., Wang L., James T., Jung Y., Kim I., Tan R.. **Reciprocal regulation of ERα and ERβ stability and activity by diptoindonesin G**. *Chem. Biol.* (2015) **22** 1608-1621. DOI: 10.1016/j.chembiol.2015.10.011 131. Zhou H.-B., Carlson K. E., Stossi F., Katzenellenbogen B. S., Katzenellenbogen J. A.. **Analogs of methyl-piperidinopyrazole (MPP): Antiestrogens with estrogen receptor alpha selective activity**. *Bioorg Med. Chem. Lett.* (2009) **19** 108-110. DOI: 10.1016/j.bmcl.2008.11.006 132. Zink A., Traidl-Hoffmann C.. **Green tea in dermatology - myths and facts**. *JDDG J. der Deutschen Dermatologischen Gesellschaft* (2015) **13** 768-775. DOI: 10.1111/ddg.12737
--- title: Target-driven machine learning-enabled virtual screening (TAME-VS) platform for early-stage hit identification authors: - Yuemin Bian - Jason J. Kwon - Cong Liu - Enrico Margiotta - Mrinal Shekhar - Alexandra E. Gould journal: Frontiers in Molecular Biosciences year: 2023 pmcid: PMC10040869 doi: 10.3389/fmolb.2023.1163536 license: CC BY 4.0 --- # Target-driven machine learning-enabled virtual screening (TAME-VS) platform for early-stage hit identification ## Abstract High-throughput screening (HTS) methods enable the empirical evaluation of a large scale of compounds and can be augmented by virtual screening (VS) techniques to save time and money by using potential active compounds for experimental testing. Structure-based and ligand-based virtual screening approaches have been extensively studied and applied in drug discovery practice with proven outcomes in advancing candidate molecules. However, the experimental data required for VS are expensive, and hit identification in an effective and efficient manner is particularly challenging during early-stage drug discovery for novel protein targets. Herein, we present our TArget-driven Machine learning-Enabled VS (TAME-VS) platform, which leverages existing chemical databases of bioactive molecules to modularly facilitate hit finding. Our methodology enables bespoke hit identification campaigns through a user-defined protein target. The input target ID is used to perform a homology-based target expansion, followed by compound retrieval from a large compilation of molecules with experimentally validated activity. Compounds are subsequently vectorized and adopted for machine learning (ML) model training. These machine learning models are deployed to perform model-based inferential virtual screening, and compounds are nominated based on predicted activity. Our platform was retrospectively validated across ten diverse protein targets and demonstrated clear predictive power. The implemented methodology provides a flexible and efficient approach that is accessible to a wide range of users. The TAME-VS platform is publicly available at https://github.com/bymgood/Target-driven-ML-enabled-VS to facilitate early-stage hit identification. ## 1 Introduction Drug discovery is expensive. Considering a representative target portfolio, high-throughput screening (HTS) is presently the most widely applicable technology for delivering chemical entry points for drug discovery campaigns (Scannell et al., 2022), but despite its popularity, this high-cost method can result in low hit rates (Zeng et al., 2020). The attrition rates of identified hits are further increased during the validation phase and optimization stage due to inherent deficits in the absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties (Feinberg et al., 2020; Xiong et al., 2021). Such challenges emphasize the demand for additional approaches that can, in parallel, perform a low-cost and efficient screening to identify potential hits and discard inappropriate structures. Thus, the strategy of exploiting the computational power of in silico virtual screening (VS) was proposed as a coherent solution. In VS efforts, structure-based and ligand-based approaches serve as two commonly used strategies. Structural data of proteins can aid in computational approaches to infer receptor–ligand interactions within target binding pockets and enable structure-based virtual screening (SBVS) (Shimada et al., 2019; Alon et al., 2021; Jumper et al., 2021; Akdel et al., 2022). SBVS can screen millions of molecules from large-scale compound libraries against protein structures (Lyu et al., 2019; Wang et al., 2019; Graff et al., 2021) and can be further augmented by integrating machine learning (ML) methods that unlock the capacity to screen an ultra-large chemical space (>1 billion compounds) (Lyu et al., 2019; Gentile et al., 2020; Gorgulla et al., 2020; Graff et al., 2021). Ligand-based virtual screening (LBVS) is another commonly used VS strategy where the chemical structures of known active compounds are used to generate a structure–activity model, which is then exploited to identify other molecules that potentially share similar bioactivity. *The* generation of large-scale chemical databases of bioactive molecules, like ChEMBL (Mendez et al., 2019), serves as a resource to further enable LBVS. Like SBVS, the integration of ML to boost LBVS capabilities has recently grown in popularity with rapid advancements in ML methods and the ever-increasing wealth of large datasets that have been generated (Jing et al., 2018; Vamathevan et al., 2019; Yang et al., 2019; Bian and Xie, 2021; Jiang et al., 2021). ML-integrated LBVS can provide a better understanding of chemical space through latent representations of the chemical properties to predict novel compound activity (Bian et al., 2019a; Bian et al., 2019b; Stokes et al., 2020; Bian and Xie, 2022). However, the generation of prerequisite datasets to enable VS is non-trivial. SBVS requires the structural information of the target, while protein preparation and crystallography are not facile tasks. LBVS necessitates known ligands with bioactivity data, which often do not exist. To address this demanding situation, we present the TArget-driven Machine learning-Enabled VS (TAME-VS) platform. The platform simply requires the input of a protein target ID and utilizes seven automated, customizable modules to assess compound libraries to identify potential hits. The platform expands the focus of the VS from the target of interest to a broader collection of proteins that share target functions or certain sequence homology. Augmented cheminformatics data are assessed against the expanded protein collection. Supervised machine learning classifiers are subsequently trained after labeling the fetched data and are used to screen future compounds. Herein, we provide further details on method implementation and discuss the results from retrospective case studies across a diverse set of protein targets. Our platform is built to be flexible, simple to use, and enable rapid evaluation of compound databases in a comprehensive manner. This methodology offers an opportunity to augment drug discovery efforts and can increase the accessibility of VS methods for both big and small organizations, and for both computational and experimental scientists. ## 2.1 Overall workflow The overall implementation of the TAME-VS is illustrated in Figure 1. There are three alternative starting points and seven modules in sequence. With only the UniProt ID of the target of interest as the input, the workflow can be initiated from Starting Point #1. The first module, Target Expansion, performs a global protein sequence homology search through the Basic Local Alignment Search Tool (BLAST) (Altschul et al., 1990) and expands the target list by identifying proteins with high sequence similarities within categorical protein family members. The second module, Compound Retrieval, extracts corresponding compounds with activity against the proteins in the target list by querying the ChEMBL database. The extracted compounds are grouped into active and inactive ligands according to assay types and activity cutoffs. The descriptive features of extracted compounds are subsequently converted to chemical fingerprints in the third module, Vectorization. The fourth module, ML Model Training, trains supervised ML classification models, by default, random forest (RF) and multilayer perceptron (MLP), based on the calculated fingerprints. In the fifth module, Virtual Screening, the trained machine learning models are applied to screen user-defined compound collections. By default, an Enamine diversity 50K library is screened, and compounds are ranked according to the prediction scores. Module 6, Post-VS Analysis, evaluates quantitative drug-likeness (QED) and calculates key physical–chemical properties for screened compounds. Finally, module 7, Data Processing, encapsulates all the data and presents the virtual hits in addition to the evaluation outcome of the entire chemical library in a summary report. Users can also initiate the workflow with their own customized target lists (pre-selected positive targets based on biological rationale) or compound lists (pre-selected active and inactive compounds) from Starting Point #2 and Starting Point #3, respectively. In addition, each module can be used individually, and the output of each module is exported to the corresponding folders for users to review. **FIGURE 1:** *Scheme of the workflow implemented in the target-driven, ML-enabled VS platform.* The platform is comprehensive yet flexible. It is designed to provide an ML-enabled solution for handling early-stage hit identification. The open-source package of the TAME-VS platform is publicly available on GitHub, together with instructions on setting up the system (https://github.com/bymgood/Target-driven-ML-enabled-VS). The details for each module are discussed in the following paragraphs. ## 2.2 Module 1: Target Expansion The purpose of Target *Expansion is* to broaden the cheminformatics investigation from the single-query target protein to a broader group of sequence-similar target proteins, based on the hypothesis that proteins with high sequence similarity may possess related structural features and may have an increased likelihood of sharing active ligands. A protein BLAST (BLASTp suite) global search is used to identify proteins with high sequence similarity to the query target through the Biopython package (Cock et al., 2009). The function NCBIWWW is imported from Bio. Blast. The default sequence similarity cutoff is set at $40\%$ but can be user-defined for a custom similarity cutoff. The arguments program and entrez_query are set to BLASTp and txid9606[ORGN] (homo sapiens). A table of expanded proteins, including collected target gene names, UniProt IDs, and percent identities are shown in the folder. ## 2.3 Module 2: Compound Retrieval The purpose of Compound *Retrieval is* to extract reported active and inactive ligands for the expanded target list from publicly available cheminformatics databases, such as ChEMBL, which documents 2.3 M compounds across 13 K targets. ChEMBL is utilized in this module and is accessed using the Python package chembl_webresource_client. The largest experimental datatype for a given protein is utilized to distinguish active and inactive compounds. The default activity cutoff is 1,000 nM for biochemical or biophysical activity (K i, IC 50, and EC 50), and the default activity cutoff for percentage inhibition (%INH) is $50\%$, with the option for users to define specific cutoff values. It is recommended to check if the %INH came from consistent compound concentrations. The folder contains a table summarizing the extracted compounds and their experimental value types, in addition to tables with standard activity values, standard activity value units, SMILES strings, InChI keys, and the associated protein UniProt ID for active and inactive compounds. ## 2.4 Module 3: Vectorization Vectorization is deployed to compute the selected types of molecular fingerprints for the extracted compounds. Different types of fingerprints evaluate the properties of the compounds from various aspects. The platform is designed to enable users to explore various types of fingerprints to evaluate the performance of the trained models using the Cheminformatics package RDKit (Landrum, 2006). Four types of fingerprints -Morgan, AtomPair, Topological and Torsion, and MACCS - are available to choose from in this module. Morgan fingerprints enumerate all circular fragments from each selected center-heavy atom up to the given radius of two atoms. The calculation is realized through get_morganfp. AtomPair fingerprints encode each atom as a type to enumerate all distances between pairs. The calculation is realized through get_AtomPairfp. Topological and Torsion fingerprints describe a linear sequence of four consecutively bonded non-hydrogen atoms, each described by its atomic type, the number of non-hydrogen branches attached to it, and its number of x electron pairs. Topological and torsion fingerprints are calculated with get_TopologicalTorsionfp. The MACCS fingerprint consists of 166 MDL substructure keys, which are calculated from the molecular graph. The calculation is realized through get_MACCS. The number of bits, which is an adjustable parameter, is set to 1,024 by default to hash the string into a fixed-length bit-vector for Morgan, AtomPair, and Topological and Torsion fingerprints. The folder contains tables of calculated fingerprints in bit-vector form for active and inactive compounds. ## 2.5 Module 4: ML Model Training The ML Model Training module is utilized to build the RF and MLP models using the calculated fingerprints from module 3 as input features. These two methods were selected to represent both classic ML algorithms and neural networks, and additional add-on features may be appended in future updates. The Python package scikit-learn (Pedregosa et al., 2011) is employed for RF and MLP model implementation. The function of RandomUnderSampler from the package imblearn.under_sampling is adopted to perform undersampling to counter potential imbalanced training data of active and inactive compounds (Lemaître et al., 2017). The function GridSearchCV from sklearn.model_selection is used to determine a preferred set of hyperparameters for trained models. The hyperparameter grid for RF includes n_estimators (50, 100, and 200) and max_depth (4, 6, 10, and 12). The hyperparameter grid for MLP includes hidden_layer_sizes [[50, 50, 50], [50, 50], and [50]]; activation (tanh and relu); and alpha (0.01 and 0.0001). As a concise evaluation for the trained models, ten-fold cross-validation is integrated, and figures of the receiver operating characteristic curve (ROC) for each model are exported for visual inspection. By default, both RF and MLP models are trained, but the user may select a specific model to be prepared. The trained prediction models are shown in the folder. ## 2.6 Module 5: Virtual Screening The purpose of Virtual *Screening is* to screen the user-defined compound collection using trained machine-learning models. By default, the Enamine diversity 50K library will be screened. An extra Python script, Library_preparation.py, is also attached in module 5 for preparing any user-defined libraries into a standard format for this platform. Trained models are loaded and screened in sequence. Compound prediction scores are written out separately for each model. ## 2.7 Module 6: Post-VS Analysis The Post-VS Analysis module evaluates and compares the screening library, with an emphasis on the top $1\%$ of virtual hits to the training set from the perspective of drug-likeness and physical–chemical properties. Distributions of prediction scores from both RF and MLP models are plotted. Quantitative estimate of drug-likeness (QED) (Bickerton et al., 2012), molecular weight (M.W.), LogP, number of H-bond acceptors, number of H-bond donors, and number of rotatable bonds are calculated using functions Descriptors. TPSA, Descriptors. MolWt, Descriptors. MolLogP, Descriptors. NumHAccepto, Descriptors. NumHDonors, and Descriptors. NumRotatableBonds in RDKit, respectively. Data tables including these calculated properties are exported, and distribution plots are prepared to facilitate an intuitional visual inspection. ## 2.8 Module 7: Data Processing In the final module, Data Processing, the selected compounds from the previous modules are consolidated and summarized, and a final list of suggested top virtual hits is reported. An ensemble ranking of molecules is calculated by averaging two individual rankings by RF and MLP. The top $1\%$ of compounds from RF and MLP models and the ensemble ranking are merged. Duplicates are removed as some molecules can be selected as top-ranked by more than one algorithm. Both the full compound list and the top $1\%$ virtual hit list are shown in the folder. ## 3.1 A case study of applying the platform to stromelysin-2 As an exemplified use case of the TAME-VS platform, we chose stromelysin-2 to illustrate the performance and output results of the modules (Figure 2). Stromelysin-2, also known as MMP10, is a proteolytic enzyme belonging to the matrix metalloproteinase (MMP) family that is known to break down extracellular matrix proteins and is involved in tissue remodeling, angiogenesis, and inflammation (Vaalamo et al., 1998; Saghizadeh et al., 2001; Krampert et al., 2004; Koller et al., 2012; Rohani et al., 2015). We entered the UniProt ID, P09238, as an input in Starting Point #1. Seven targets that share sequence similarity of over $40\%$ were identified and written down (Figure 2A). A total of 17,467 chemical records were retrieved with activity across all collected targets (Figure 2B). The retrieved chemical records were distributed among a variety of experimental data types, including biochemical and competitive binding. The biochemical assay data type IC 50, which contains the most records (10,727 records; see Figure 2C), was utilized to split curated compounds into active and inactive ones at a default concentration of 1000 nM. After the vectorization, the RF (Figure 2D) and MLP (Figure 2E) classification models were trained to distinguish active molecules from inactive ones. ROC curves of ten-time cross-validation provided an intuitional visualization of the robustness of the training process. The Enamine Diversity 50K library was then screened and scored separately by the trained RF (Figure 2F) and MLP (Figure 2G) models. After the virtual screening, properties like QED (Figure 2H), MW (Figure 2I), and LogP (Figure 2J) were calculated and plotted. Eventually, the scored full compound list and top $1\%$ virtual hit list were written on the disk. The overall process took approximately 30 min to finish on a MacBook Pro equipped with a 2.6 GHz 6-Core Intel Core i7 processor. **FIGURE 2:** *Applying target-driven, ML-enabled VS toward stromelysin-2 (UniProt ID: P09238) as a case study to exemplify outcomes from each module. (A). Protein list after target expansion. (B). Number of extracted compounds for each target in the protein list. (C). Distribution of experimental value types. (D). ROC curve for RF model training. (E). ROC curve for MLP model training. (F). Distribution of prediction scores on the Enamine Diversity 50K library using the RF model. (G). Distribution of prediction scores on the Enamine Diversity 50K library using the MLP model. (H). Distribution of calculated QED. (I). Distribution of calculated MW. (J). Distribution of calculated LogP. (K). Exemplified final reports.* ## 3.2 Retrospective studies on diverse protein targets In addition to evaluating the efficiency of our platform, we sought to address the effectiveness and performance of our pipeline across a range of protein types. Ten targets representing divergent protein categories, including GPCRs, ligases, oxidoreductases, proteases, kinases, phosphatases, and voltage-gated ion channels, were selected for these studies (Table 1). Using Starting Point #1, we performed retrospective VS studies on ten diverse protein targets using their UniProt IDs as the input and assessed if the platform could determine ex post facto known active compounds of targets over the broad range of chemical matter represented in the Enamine diversity 50K library. After target expansion and compound retrieval, we observed a wide range of identified homologous targets, retrieved known chemicals, and miscellaneous experimental assay types. **TABLE 1** | Target | UniProt ID | Category | # of homologous targets identified | # of reported molecules retrieved | Assay type that gives most records | | --- | --- | --- | --- | --- | --- | | A2b | P29275 | GPCR | 2 | 36629 | K i | | ACC1 | Q13085 | Ligase | 1 | 4268 | IC 50 | | AKR1B10 | O60218 | Oxidoreductase | 9 | 4449 | IC 50 | | CTSG | P08311 | Protease | 3 | 3076 | IC 50 | | JAK3 | P52333 | Kinase | 3 | 32951 | IC 50 | | MMP10 | P09238 | Protease | 7 | 17467 | IC 50 | | PRKD1 | Q15139 | Kinase | 2 | 6981 | Inhibition | | PTN6 | P29350 | Phosphatase | 1 | 2434 | IC 50 | | RPS6KA3 | P51812 | Kinase | 5 | 16383 | Inhibition | | SCN4A | P35499 | Voltage-gated ion channel | 9 | 13594 | IC 50 | A hit identification campaign for a novel target typically lacks reported active compounds or probes. To simulate this scenario, known hits of the query protein were withheld during the model training stage but reintroduced for scoring once the models had been trained with chemical data from the expanded protein target list (Figure 3A). The Enamine Diversity 50K library and known active compounds of the query target were evaluated by the trained predictive models, and the outcomes were assessed. Given that compounds in the Enamine diversity 50K library sparsely represent a general drug-like chemical space, the majority of these molecules are anticipated to be assigned relatively low VS scores by the predictive models compared to active compounds. Indeed, we observe that both RF and MLP assign higher VS scores to known active compound sets as compared to the Enamine 50K chemicals (Figures 3B, C), with a high degree of agreement between the two models (Supplementary Figure S1). Specifically, we observe a significant difference in $\frac{9}{10}$ targets for RF and $\frac{7}{10}$ targets for MLP. MLP was unable to detect a significant difference in known active compounds compared to Enamine 50K compounds for PRKD1 and RPS6KA3 due to the assay type of inhibition (Table 1) utilized for model training, which can suffer from inconsistent compound testing concentrations. MLP had a 3-fold greater variability in the scoring of known active compounds than RF. However, MLP, on average, provided a two-fold greater differential VS score between the Enamine 50K and known active compounds across targets compared to RF (Supplementary Figure S2). As anticipated, the ability of models to discern a difference between the Enamine 50K and known active compounds was correlated with the number of targets identified during the target expansion phase in addition to the number of molecules identified in the compound retrieval phase (Supplementary Figure S1). **FIGURE 3:** *Retrospective validations across ten different protein targets. (A) Schematic illustration of a retrospective study. Averaged VS scores reported by the RF model (B) and the MLP model (C) for the entire Enamine diversity 50K library (black bar) and known active compounds (gray bar) of the query target. The error bar represents the standard error of the mean (SEM), *p < 0.05; **p < 0.01; ***p < 0.001.* To enable the evaluation of model training and performance, six metrics, namely, AUC, precision, recall, specificity, F1 score, geometric mean, and index of balanced accuracy (IBA), were embedded in the TAME-VS platform for evaluating the performance from various aspects. The calculation of these metrics is detailed in the supplementary information. To reflect the imbalanced training data that active compounds are usually minorities, the training process adopted the down-sampling of inactive compounds with 10-time cross-validation. ROC curves with calculated AUC values for cross-validation were plotted automatically after model training for both RF (Supplementary Figure S3) and MLP (Supplementary Figure S4) models. The values of reported precision, recall, specificity, F1 score, geometric mean, and index of balanced accuracy are summarized into tables for both RF (Supplementary Table S1) and MLP (Supplementary Table S2). Across the ten diversified protein targets, calculated metrics suggested that trained RF and MLP models gave robust and equivalent performances on classifications (Supplementary Figure S5). We observed greater variability in AUC as measured by standard deviation and model precision, which were inversely proportional to the number of compounds available within the training set and the number of targets within target expansion, respectively (Supplementary Figure S6). To better understand the latent chemical space and structural insights that can be revealed from the process of virtual screening, we performed structural clustering and analysis for hits relating to stromelysin-2 (MMP10) from our retrospective studies (Figure 4). The Enamine 50K library was classified into one hundred structurally diversified clusters based on k-means clustering of encoded fingerprints, and we identified cluster #20 as having the greatest mean VS scores (Figure 4A). Interestingly, cluster #20 stood out as its upper extremes achieved comparable VS scores to known MMP10 active compounds (Figure 3B). Cluster #20 had a significant increase in mean RF-based VS score (0.42) compared to the remaining Enamine 50K library (0.35) (Figure 4B). From the perspective of physical–chemical properties, compounds in cluster #20 remained within the zone that follows the “rule of 5” (Lipinski et al., 1997) (Figure 4B). Subsequently, we performed t-SNE analysis to better visualize the chemical space coverage (Figure 4C). The compounds in the Enamine 50K library defined the overall boundary. Compounds in cluster #20 were largely concentrated as expected and partially overlapped with training molecules that were retrieved from the expanded target list. The known active MMP10 molecules were proximal to other training molecules but mostly independent from the compounds in cluster #20. Upon further investigation of specific chemical structures, it was found that compounds from cluster #20 that scored as highly active maintained a benzenesulfonamide group, which is a reported moiety of some known MMP10 inhibitors (Nara et al., 2016) (Figure 4D). This is an important finding, as known MMP10 active compounds were not included in the model training in retrospective studies. As seen in the retrieved molecules from the expanded target list, the TAME-VS platform detected chemical patterns in the training sets to construct a chemical understanding of the structure of the inhibitors. A Morgan fingerprint-based structural similarity search using the same known active MMP10 molecule as the query compound was conducted in parallel. Suggested molecules from our TAME-VS platform do not simply recur compounds with top Tanimoto coefficient (Tc) scores from the classical structural similarity search (Supplementary Figure S7A). TAME-VS can propose chemicals that align with the acquired structural patterns, even if they do not have high Tc similarity scores, which differs from the traditional approach of using structural similarity search (Supplementary Figure S7B, C). This observation further supported the claim that our TAME-VS platform can provide an alternative approach to tackle early-stage hit findings. **FIGURE 4:** *Structural insights revealed from the screening. (A). Box and Whisker plots for RF-based VS scores across structurally clustered groups of Enamine 50K library. (B). Comparison of VS scores and properties between cluster #20 and the remaining part of the Enamine 50K library. (C). t-SNE analysis to visualize covered chemical spaces by known MMP10 actives (blue), cluster #20 (orange), active molecules in the training set (green), inactive molecules in the training set (red), and full Enamine 50K compounds (purple). (D). One known MMP10 inhibitor and exemplified compounds in cluster #20. The overlapping benzenesulfonamide group is highlighted in green. Dissimilar moieties are colored pink.* ## 4 Discussion The use of large-scale, high-throughput screening has been a major cornerstone of modern drug discovery efforts to identify chemical hits for novel protein targets and will be an important resource for the foreseeable future. Our novel TAME-VS platform enables users to survey chemical libraries rapidly and cost-effectively for ab initio drug discovery campaigns at a very early stage. This plug-and-play system provides a high degree of customizability and enables a broad range of users to explore desired chemical spaces to rapidly identify potential starting points for further chemical evaluation. Indeed, our retrospective validation across different protein types demonstrates a clear value in our platform with reliable predictive performance in the majority of cases. We acknowledge that the use of the ChEMBL database limits the utility of the TAME-VS platform as low-homology or orphan proteins may not be represented within the database. However, this issue can be remedied by using the optional starting points, which allow users to flexibly supply their customized internal data, which may not be immediately available from public databases. For example, users can employ their own domain expertise to provide a more specified list of relevant targets for aggregating compound data in module 2, with Compound Retrieval serving as the optional Starting Point #2 to begin the platform. Alternatively, if users have prepared their own compound lists from their internal experimental testing, module 3, Vectorization, can function as the optional Starting Point #3 to leverage the remaining parts of the platform. The TAME-VS platform serves as a relevant and flexible tool to efficiently perform virtual screening across a broad range of drug discovery stages. Furthermore, the platform can be deployed in a piecemeal fashion by running an individual module or a combination of multiple modules depending on user needs. Inside each module, there is a stand-alone Python script that can run independently with customized inputs and outputs. A Jupyter notebook for each module is also provided in case users prefer a more interactive experience. The following are a few examples. Module 1 provides an immediate solution to automated BLASTp search, which can improve sequence-focused bioinformatics studies (Figure 5A). Module 2 searches a large-scale compound collection for activity against given targets, which enables the creation of a focused chemical library for particular protein targets or target groups (Figure 5B). If a user requires a quick tool for fingerprints and key physical–chemical property calculation, module 3 (Figure 5C) and module 6 (Figure 5D) can be used for the task. For a just-initiated drug discovery project on a protein target, combining module 1 and module 2 can provide a bioinformatics overview of related, similar proteins and their corresponding interacting molecules (Figure 5E). Another example is combining module 2 and module 6 to calculate the chemical properties of molecules with activity against a given target (Figure 5F). **FIGURE 5:** *This platform has the flexibility for multi-purpose adaptation. Using an individual module (A–D) or a combination of modules (E, F) to realize various functions.* In addition to its flexibility in use, TAME-VS is also highly adaptable. There have been rapid advancements in ML applications and an ever-increasing expansion in AI methodologies for drug discovery (Vamathevan et al., 2019; Bian and Xie, 2021). Although we provide RF and MLP models to represent both classic ML algorithms and neural networks, respectively, additional add-on features can be appended in future updates to accommodate advancements in ML methods. Additionally, our platform can accommodate the use of novel molecular features and datatypes. Screening libraries can be customized, and post-screening analysis can integrate extra dimensions. The platform is highly customizable, easily integrated, and can be used to analyze data from multiple sources. Our methodology provides a comprehensive, efficient, and flexible platform for virtual screening in target-driven drug discovery campaigns. It simplifies the process by streamlining the data processing, analysis, and visualization of results. This platform enables researchers to target novel proteins with limited starting information to rapidly evaluate and triage a large chemical space based on homology-expanded target lists and may help reduce the time and cost associated with launching a full drug discovery campaign. With its user-friendly programming environment, the TAME-VS platform can serve as an initial tool for early drug discovery and can increase the accessibility of these ML methods to a broad range of users. ## Data availability statement The open-source, freely available package of the TAME-VS platform is documented at https://github.com/bymgood/Target-driven-ML-enabled-VS. ## Author contributions YB, JK, and MS conceptualized the idea. YB designed, developed, and programmed the platform. YB and CL performed code review and debugging. YB and JK conducted retrospective validations and interpreted the outcomes. EM, MS, and AG contributed insights from drug discovery and medicinal chemistry perspectives. YB and JK wrote the manuscript. YB, JK, CL, EM, MS, and AG reviewed, revised, and approved the manuscript. ## Conflict of interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ## Publisher’s note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors, and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. ## Supplementary material The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmolb.2023.1163536/full#supplementary-material ## References 1. Akdel M., Pires D. E. V., Pardo E. P., Jänes J., Zalevsky A. O., Mészáros B.. **A structural biology community assessment of AlphaFold2 applications**. *Nat. Struct. Mol. Biol.* (2022) **29** 1056-1067. DOI: 10.1038/s41594-022-00849-w 2. Alon A., Lyu J., Braz J. M., Tummino T. A., Craik V., O’Meara M. J.. **Structures of the σ2 receptor enable docking for bioactive ligand discovery**. *Nature* (2021) **600** 759-764. DOI: 10.1038/s41586-021-04175-x 3. Altschul S. F., Gish W., Miller W., Myers E. W., Lipman D. J.. **Basic local alignment search tool**. *J. Mol. Biol.* (1990) **215** 403-410. DOI: 10.1016/S0022-2836(05)80360-2 4. Bian Y., Jing Y., Wang L., Ma S., Jun J. J., Xie X.-Q.. **Prediction of orthosteric and allosteric regulations on cannabinoid receptors using supervised machine learning classifiers**. *Mol. Pharm.* (2019a) **16** 2605-2615. DOI: 10.1021/acs.molpharmaceut.9b00182 5. Bian Y., Wang J., Jun J. J., Xie X.-Q.. **Deep convolutional generative adversarial network (dcGAN) models for screening and design of small molecules targeting cannabinoid receptors**. *Mol. Pharm.* (2019b) **16** 4451-4460. DOI: 10.1021/acs.molpharmaceut.9b00500 6. Bian Y., Xie X.-Q.. **Artificial intelligent deep learning molecular generative modeling of scaffold-focused and cannabinoid CB2 target-specific small-molecule sublibraries**. *Cells* (2022) **11** 915. DOI: 10.3390/cells11050915 7. Bian Y., Xie X.-Q.. **Generative chemistry: Drug discovery with deep learning generative models**. *J. Mol. Model.* (2021) **27** 71-18. DOI: 10.1007/s00894-021-04674-8 8. Bickerton G. R., Paolini G. V., Besnard J., Muresan S., Hopkins A. L.. **Quantifying the chemical beauty of drugs**. *Nat. Chem.* (2012) **4** 90-98. DOI: 10.1038/nchem.1243 9. Cock P. J., Antao T., Chang J. T., Chapman B. A., Cox C. J., Dalke A.. **Biopython: Freely available Python tools for computational molecular biology and bioinformatics**. *Bioinformatics* (2009) **25** 1422-1423. DOI: 10.1093/bioinformatics/btp163 10. Feinberg E. N., Joshi E., Pande V. S., Cheng A. C.. **Improvement in ADMET prediction with multitask deep featurization**. *J. Med. Chem.* (2020) **63** 8835-8848. DOI: 10.1021/acs.jmedchem.9b02187 11. Gentile F., Agrawal V., Hsing M., Ton A.-T., Ban F., Norinder U.. **Deep docking: A deep learning platform for augmentation of structure based drug discovery**. *ACS central Sci.* (2020) **6** 939-949. DOI: 10.1021/acscentsci.0c00229 12. Gorgulla C., Boeszoermenyi A., Wang Z.-F., Fischer P. D., Coote P. W., Padmanabha Das K. M.. **An open-source drug discovery platform enables ultra-large virtual screens**. *Nature* (2020) **580** 663-668. DOI: 10.1038/s41586-020-2117-z 13. Graff D. E., Shakhnovich E. I., Coley C. W.. **Accelerating high-throughput virtual screening through molecular pool-based active learning**. *Chem. Sci.* (2021) **12** 7866-7881. DOI: 10.1039/d0sc06805e 14. Jiang D., Wu Z., Hsieh C.-Y., Chen G., Liao B., Wang Z.. **Could graph neural networks learn better molecular representation for drug discovery? A comparison study of descriptor-based and graph-based models**. *J. cheminformatics* (2021) **13** 12-23. DOI: 10.1186/s13321-020-00479-8 15. Jing Y., Bian Y., Hu Z., Wang L., Xie X.-Q. S.. **Deep learning for drug design: An artificial intelligence paradigm for drug discovery in the big data era**. *AAPS J.* (2018) **20** 58. DOI: 10.1208/s12248-018-0210-0 16. Jumper J., Evans R., Pritzel A., Green T., Figurnov M., Ronneberger O.. **Highly accurate protein structure prediction with AlphaFold**. *Nature* (2021) **596** 583-589. DOI: 10.1038/s41586-021-03819-2 17. Koller F. L., Dozier E. A., Nam K. T., Swee M., Birkland T. P., Parks W. C.. **Lack of MMP10 exacerbates experimental colitis and promotes development of inflammation-associated colonic dysplasia**. *Lab. Investig.* (2012) **92** 1749-1759. DOI: 10.1038/labinvest.2012.141 18. Krampert M., Bloch W., Sasaki T., Bugnon P., Rulicke T., Wolf E.. **Activities of the matrix metalloproteinase stromelysin-2 (MMP-10) in matrix degradation and keratinocyte organization in wounded skin**. *Mol. Biol. Cell* (2004) **15** 5242-5254. DOI: 10.1091/mbc.e04-02-0109 19. Landrum G.. *RDKit: Open-source cheminformatics* (2006) 20. Lemaître G., Nogueira F., Aridas C. K.. **Imbalanced-learn: A python toolbox to tackle the curse of imbalanced datasets in machine learning**. *J. Mach. Learn. Res.* (2017) **18** 559-563 21. Lipinski C. A., Lombardo F., Dominy B. W., Feeney P. J.. **Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings**. *Adv. drug Deliv. Rev.* (1997) **23** 3-26. DOI: 10.1016/s0169-409x(00)00129-0 22. Lyu J., Wang S., Balius T. E., Singh I., Levit A., Moroz Y. S.. **Ultra-large library docking for discovering new chemotypes**. *Nature* (2019) **566** 224-229. DOI: 10.1038/s41586-019-0917-9 23. Mendez D., Gaulton A., Bento A. P., Chambers J., de Veij M., Félix E.. **ChEMBL: Towards direct deposition of bioassay data**. *Nucleic acids Res.* (2019) **47** D930-D940. DOI: 10.1093/nar/gky1075 24. Nara H., Sato K., Kaieda A., Oki H., Kuno H., Santou T.. **Design, synthesis, and biological activity of novel, potent, and highly selective fused pyrimidine-2-carboxamide-4-one-based matrix metalloproteinase (MMP)-13 zinc-binding inhibitors**. *Bioorg. Med. Chem.* (2016) **24** 6149-6165. DOI: 10.1016/j.bmc.2016.09.009 25. Pedregosa F., Varoquaux G., Gramfort A., Michel V., Thirion B., Grisel O.. **Scikit-learn: Machine learning in Python**. *J. Mach. Learn. Res.* (2011) **12** 2825-2830 26. Rohani M. G., Mcmahan R. S., Razumova M. V., Hertz A. L., Cieslewicz M., Pun S. H.. **MMP-10 regulates collagenolytic activity of alternatively activated resident macrophages**. *J. Investigative Dermatology* (2015) **135** 2377-2384. DOI: 10.1038/jid.2015.167 27. Saghizadeh M., Brown D. J., Castellon R., Chwa M., Huang G. H., Ljubimova J. Y.. **Overexpression of matrix metalloproteinase-10 and matrix metalloproteinase-3 in human diabetic corneas: A possible mechanism of basement membrane and integrin alterations**. *Am. J. pathology* (2001) **158** 723-734. DOI: 10.1016/S0002-9440(10)64015-1 28. Scannell J. W., Bosley J., Hickman J. A., Dawson G. R., Truebel H., Ferreira G. S.. **Predictive validity in drug discovery: What it is, why it matters and how to improve it**. *Nat. Rev. Drug Discov.* (2022) **21** 915-931. DOI: 10.1038/s41573-022-00552-x 29. Shimada I., Ueda T., Kofuku Y., Eddy M. T., Wüthrich K.. **GPCR drug discovery: Integrating solution NMR data with crystal and cryo-EM structures**. *Nat. Rev. Drug Discov.* (2019) **18** 59-82. DOI: 10.1038/nrd.2018.180 30. Stokes J. M., Yang K., Swanson K., Jin W., Cubillos-Ruiz A., Donghia N. M.. **A deep learning approach to antibiotic discovery**. *Cell* (2020) **180** 475-483. DOI: 10.1016/j.cell.2020.04.001 31. Vaalamo M., Karjalainen-Lindsberg M.-L., Puolakkainen P., Kere J., Saarialho-Kere U.. **Distinct expression profiles of stromelysin-2 (MMP-10), collagenase-3 (MMP-13), macrophage metalloelastase (MMP-12), and tissue inhibitor of metalloproteinases-3 (TIMP-3) in intestinal ulcerations**. *Am. J. pathology* (1998) **152** 1005-1014 32. Vamathevan J., Clark D., Czodrowski P., Dunham I., Ferran E., Lee G.. **Applications of machine learning in drug discovery and development**. *Nat. Rev. Drug Discov.* (2019) **18** 463-477. DOI: 10.1038/s41573-019-0024-5 33. Wang E., Sun H., Wang J., Wang Z., Liu H., Zhang J. Z.. **End-point binding free energy calculation with MM/PBSA and MM/GBSA: Strategies and applications in drug design**. *Chem. Rev.* (2019) **119** 9478-9508. DOI: 10.1021/acs.chemrev.9b00055 34. Xiong G., Wu Z., Yi J., Fu L., Yang Z., Hsieh C.. **ADMETlab 2.0: An integrated online platform for accurate and comprehensive predictions of ADMET properties**. *Nucleic Acids Res.* (2021) **49** W5-W14. DOI: 10.1093/nar/gkab255 35. Yang K., Swanson K., Jin W., Coley C., Eiden P., Gao H.. **Analyzing learned molecular representations for property prediction**. *J. Chem. Inf. Model.* (2019) **59** 3370-3388. DOI: 10.1021/acs.jcim.9b00237 36. Zeng W., Guo L., Xu S., Chen J., Zhou J.. **High-throughput screening technology in industrial biotechnology**. *Trends Biotechnol.* (2020) **38** 888-906. DOI: 10.1016/j.tibtech.2020.01.001
--- title: Iontophoresis-driven microneedle patch for the active transdermal delivery of vaccine macromolecules authors: - Ying Zheng - Rui Ye - Xia Gong - Jingbo Yang - Bin Liu - Yunsheng Xu - Gang Nie - Xi Xie - Lelun Jiang journal: Microsystems & Nanoengineering year: 2023 pmcid: PMC10040928 doi: 10.1038/s41378-023-00515-1 license: CC BY 4.0 --- # Iontophoresis-driven microneedle patch for the active transdermal delivery of vaccine macromolecules ## Abstract COVID-19 has seriously threatened public health, and transdermal vaccination is an effective way to prevent pathogen infection. Microneedles (MNs) can damage the stratum corneum to allow passive diffusion of vaccine macromolecules, but the delivery efficiency is low, while iontophoresis can actively promote transdermal delivery but fails to transport vaccine macromolecules due to the barrier of the stratum corneum. Herein, we developed a wearable iontophoresis-driven MN patch and its iontophoresis-driven device for active and efficient transdermal vaccine macromolecule delivery. Polyacrylamide/chitosan hydrogels with good biocompatibility, excellent conductivity, high elasticity, and a large loading capacity were prepared as the key component for vaccine storage and active iontophoresis. The transdermal vaccine delivery strategy of the iontophoresis-driven MN patch is “press and poke, iontophoresis-driven delivery, and immune response”. We demonstrated that the synergistic effect of MN puncture and iontophoresis significantly promoted transdermal vaccine delivery efficiency. In vitro experiments showed that the amount of ovalbumin delivered transdermally using the iontophoresis-driven MN patch could be controlled by the iontophoresis current. In vivo immunization studies in BALB/c mice demonstrated that transdermal inoculation of ovalbumin using an iontophoresis-driven MN patch induced an effective immune response that was even stronger than that of traditional intramuscular injection. Moreover, there was little concern about the biosafety of the iontophoresis-driven MN patch. This delivery system has a low cost, is user-friendly, and displays active delivery, showing great potential for vaccine self-administration at home. ## Introduction Coronavirus disease 2019 (COVID-19) has evolved into a pandemic and poses a great threat to public health1,2. Vaccination provides an effective way to prevent pathogen infection3,4. The development of effective vaccines is one key to preventing epidemic transmission and achieving herd immunity. To date, preventive vaccines have eliminated many diseases, such as smallpox and poliomyelitis5, and have the potential to curb the occurrence of many other common infectious diseases and halt the global spread of the emerging COVID-19 pandemic6. Most vaccines are given by intramuscular (IM) injection, but this method has many inherent limitations: fear of needles in children and adolescents7, procedural pain8, and infections caused by the repeated use of needles9. Due to the rich network of immune cells in the skin, transdermal immunization is becoming an attractive alternative10. However, owing to the stratum corneum (SC) barrier, only small molecule vaccines (~<500 Da) can be effectively administered through the skin and enter the systemic circulation, so macromolecular vaccines have poor skin permeability and low bioavailability11–13. To break through the barrier of the SC and improve the transdermal permeation efficiency of vaccines, transdermal vaccination strategies have been developed to enhance the transdermal immune effect. Microneedles (MNs) provide a new solution in the field of transdermal vaccine delivery due to their unique advantages of painless minimally invasive delivery, self-administration, and improved permeability14. The length of the MN is specially designed to give it the ability to penetrate the SC without stimulating nerve endings. The epidermis and dermis have a dense network of immune cells, such as Langerhans cells (LCs) and dermal dendritic cells (DDCs), whose anatomical sites can be reached by the MN during puncture15. MNs produce microchannels through the SC by skin penetration, allowing the vaccine macromolecules to permeate the anatomical sites of the immune cells16,17, so transdermal immunity using MNs is promising. However, transdermal vaccine delivery using MNs usually relies on passive diffusion via poked microchannels, which severely limits the transport speed and efficiency into the skin18. On the other hand, the iontophoresis technique can actively drive ionized and hydrophilic small molecules through the SC layer using a mild current (usually <0.5 mA/cm2), and therefore, this strategy has been applied in transdermal vaccine delivery and dermatological treatment19,20. Iontophoresis can control the transport process of vaccine molecules via the applied electricity because of electromigration and electroosmosis21–25. Iontophoresis has been successful in promoting the transdermal delivery of small hydrophilic molecules, but it usually fails in the transdermal delivery of macromolecules (e.g., proteins with a molecular weight >13 kDa) due to the barrier of SC26–28. However, most vaccines, such as nucleic acid vaccines, are macromolecules, which are difficult to administer transdermally into the skin using iontophoresis. Therefore, it is promising to utilize the synergistic advantages of macromolecule delivery using MNs and active delivery using iontophoresis to overcome the limitations of the low delivery efficiency of MNs and the failure of macromolecule delivery of iontophoresis. In this work, we developed a wearable iontophoresis-driven MN system for the efficient and active transdermal delivery of macromolecular vaccines, as shown in Fig. 1 and Video S1. This system mainly consists of an iontophoresis-driven MN patch and an iontophoresis-driven device. The transdermal vaccine delivery strategy of the iontophoresis-driven MN patch is “press and poke, iontophoresis-driven delivery, and immune response” in which solid MNs are first pressed onto the skin to create microchannels through the SC and then automatically retracted. The vaccine macromolecules are then delivered through these created microchannels via passive diffusion and iontophoresis and captured by antigen-presenting cells in the epidermis and dermis, finally activating the cells to exert immune effects. The iontophoresis-driven MN patch combines the advantages of the MN and iontophoresis techniques and significantly promotes the transdermal delivery efficiency of vaccine macromolecules. Moreover, a flexible polyacrylamide/chitosan hydrogel with high loading ability and good electrical conductivity was selected as the vaccine storage chamber, which not only addressed the limitation of low vaccine loading of dissolvable MNs but also guaranteed stable conductivity between the electrodes and skin during iontophoresis. In vivo transdermal immunization demonstrated that the iontophoresis-driven MN patch could achieve an effective immune response that was even stronger than that using traditional intramuscular injection. Therefore, our iontophoresis-driven MN system is low-cost and user-friendly, showing promise as an alternative for vaccine self-administration at home. Fig. 1Schematic illustration of the wearable iontophoresis-driven MN delivery system.a Schematic illustration of the wearable delivery system, which mainly consists of the iontophoresis-driven MN patch and the iontophoresis-driven device. The iontophoresis-driven MN patch is composed of Ag/AgCl electrodes on a flexible PI, solid MNs, hydrogel blocks, and an adhesive impermeable gasket. The iontophoresis-driven device can supply power and output a constant current for iontophoresis-driven vaccine delivery. b The transdermal vaccine delivery strategy of the iontophoresis-driven MN patch is “press and poke, iontophoresis-driven delivery, and immune response” ## Transdermal vaccine delivery strategy of iontophoresis-driven MN patches A wearable iontophoresis-driven MN delivery system composed of an iontophoresis-driven MN patch and an iontophoresis-driven device is proposed for the transdermal delivery of vaccine macromolecules, as shown in Fig. 1a and Video S1. The iontophoresis-driven MN patch mainly consists of Ag/AgCl electrodes on the flexible PI, solid MNs, hydrogel blocks, and a double-sided adhesive impermeable gasket. The circular hydrogel blocks are conductive and are used to load the vaccine antigen. The MN base is attached to the working electrode (WE), and the microneedles penetrate the circular hydrogel. The flexible electrodes, MNs, and hydrogel blocks are assembled in an impermeable gasket. The adhesive impermeable gasket can hold each component firmly in place, ensuring a stable iontophoresis current during vaccine delivery. Iontophoresis-driven MN patches are compressible due to the high elasticity of the hydrogel and gasket. In the iontophoresis-driven MN patches, the vaccine is loaded in the circular hydrogel, the skin is compressed and poked to create microchannels via the MNs, and the vaccine is delivered through the microchannels from the hydrogel via active iontophoresis and passive diffusion. The wearable iontophoresis-driven device can supply power and output a constant current for active iontophoresis-driven vaccine delivery. The transdermal vaccine delivery strategy of the Iontophoresis-driven MN patch is “press and poke, iontophoresis-driven delivery, and immune response”, as shown in Fig. 1b and Video S1.Press and poke. The iontophoresis-driven MN patch is conformably adhered to the skin. When the patch is pressed using a finger, the solid MN penetrates through the hydrogel and the SC, creating transient microchannels in the skin. Upon removal of the compression, the MN detaches from the skin and retracts into the hydrogel due to the elastic rebound energy of the gasket and hydrogels. Owing to the viscoelasticity and self-healing ability of the skin, the MN-induced microchannels will gradually close and heal upon retraction of the MNs. Moreover, the patch can be repeatedly pressed to reopen the microchannels in the skin and initiate another cycle of transdermal vaccine delivery. Iontophoresis-driven vaccine delivery. The vaccine in the hydrogel passively diffuses into skin via the poked microchannels along the concentration gradient according to Fick’s diffusion law. The passive diffusion rate is mainly determined by the concentration of the vaccine in the hydrogel and created microchannels. A mild electric current is applied between the WE and counter electrode (CE) to initiate iontophoresis and actively drive the charged vaccine macromolecules to permeate into the skin through the poked microchannels under the main driving forces of electromigration and electroosmosis29. Most of the electroosmotic flow during iontophoresis migrates along the low-resistance and preferential pathways associated primarily with the microchannels30–32. The iontophoresis-driven delivery rate of the vaccine is mainly determined by the formulation characteristics of the vaccine, the microchannels poked by the MNs, and the iontophoresis current and duration33. Moreover, the combination of passive diffusion and active iontophoresis-driven delivery may lead to a synergistic enhancement in the transdermal delivery efficiency of vaccines. Immune response. A dense network of immune cells is distributed in the epidermis and dermis of the skin. Upon delivery into the skin, the vaccine is captured by antigen-presenting cells (APCs), such as LCs and DDCs. After antigen stimulation, APCs migrate to the draining lymph nodes and activate Th lymphocytes to play a role in immunity34. APCs located in the epidermis and dermis activate the cells to exert an immune effect. Therefore, the agent stimulates the body’s immune system to recognize the agent as a threat, destroy it, and further recognize and destroy any of the microorganisms associated with that agent that it may encounter in the future. ## Fabrication and characterization of the wearable iontophoresis-driven MN delivery system A wearable iontophoresis-driven MN delivery system was developed, as shown in Fig. 2a. It consists of the iontophoresis-driven MN patch and the iontophoresis-driven device. The iontophoresis-driven MN patch is adhered to the wrist with a size of 28 × 18 × 2 mm3. The iontophoresis-driven device is encapsulated in a white 3D-printed insulating shell with a size of 53 × 19 × 10 mm3, which can be worn on the wrist using a custom watchband. The iontophoresis-driven device outputs the iontophoresis current to the patch using a flexible PCB connector. The wearable iontophoresis-driven MN delivery system is small and light weight (only 18 g), enabling self-administration in daily life. Fig. 2Characterization of the iontophoresis-driven MN delivery system.a Image of the iontophoresis-driven MN delivery system worn on a wrist. b The flexible PCB circuit of the iontophoresis-driven device. c Schematic diagram of the constant current output module. LM334 is the core chip of the constant current output circuit. d Image of a flexible iontophoresis-driven MN patch. e Optical image of the solid MNs. f SEM image of the solid MNs. g Optical image of the circular hydrogel block loaded with the vaccine. h SEM image of the freeze-dried hydrogel. i Optical image of the Ag/AgCl electrodes on the flexible PI The PCB diagram and detailed circuit principle of the iontophoresis-driven device are shown in Fig. 2b, c and Figs. S1–7, respectively. The iontophoresis-driven circuit is mainly composed of the charging module, the boosted module, and the constant current output module, as shown in Fig. S2. The iontophoresis-driven circuit is powered by a lithium battery that can be recharged via a micro-USB port. The boosted module adjusts the output voltage of the lithium battery at the designed voltage for the constant current chip. The constant current output module provides a constant iontophoresis-driven current for vaccine delivery. LM334 is the core chip of the constant current output circuit, as shown in Fig. 2c. The charging and constant current output performance of the iontophoresis-driven device were verified (Fig. S8). The output voltage of the iontophoresis-driven device is linearly proportional to the load resistance, indicating that the current output is independent of the load resistance from 5 to 20 kΩ (typical skin impedance is ~10 kΩ). Moreover, the iontophoresis-driven device can be stably maintained at 0.5 mA for 1 h with little fluctuation. A flexible iontophoresis-driven MN patch was assembled with Ag/AgCl electrodes on the flexible PI, solid MNs, hydrogel blocks, and a gasket, as shown in Fig. 2d. The impermeable gasket can conform and tightly stick to the curved skin surface, thereby avoiding interstitial fluid leakage and preventing hazardous infection. Figure 2e, f shows the optical and SEM images of the solid MNs fabricated by micromachining from stainless steel 316 L, which has high mechanical strength and biocompatibility for skin penetration35. Sixty-nine microneedles with conical tips and cylindrical bodies are uniformly arranged on the Φ12.4 mm cylindrical substrate. The average height, tip radius, and base diameter of the MNs are ~800, 20, and 400 µm, respectively. The adjacent microneedle distance is ~1200 µm. The solid MNs possess sharp tips for effective skin penetration. Figure 2g shows a cylindrical hydrogel block as the vaccine reservoir made of polyacrylamide (PAM)/chitosan for vaccine loading. The hydrogel block is compressible and conductive, can come in close contact with the curved skin and maintains good conductivity between the skin and the electrode. Figure 2h presents the morphology of the freeze-dried hydrogel. The hydrogel is porous, in which the vaccine and water are captured and stored. Figure 2i presents a pair of iontophoresis Ag/AgCl electrodes on a flexible PI. The Ag/AgCl electrodes were prepared by sputter coating Ag film and chlorination of Ag36. ## Basic performance of the iontophoresis-driven MN patch Since the iontophoresis-driven MN patch promotes vaccine delivery through poked microchannels, the skin penetration performance of the MN patch was investigated, as shown in Fig. 3a. During the “press and poke” stage, the resistance force increased with loading displacement until the maximum stress of the skin at the microneedle tips exceeded the rupture limit of skin, causing a sudden drop in force at point “P”. The critical penetration force of the MN patch was ~5.6 N, which is lower than typical thumb pressure37, indicating easy skin penetration to create microchannels by compression of the iontophoresis-driven MN patch. It was further demonstrated that neatly arranged microchannels formed in the poked skin (Fig. 3b). The distribution of the poked microchannels was consistent with that of MNs, demonstrating a high skin penetration rate (almost $100\%$). The depth and base diameter of the poked microchannels were ~500 and 250 μm, respectively, as shown in Fig. 3c, d. This result indicated that the MN patch could penetrate through the stratum corneum layer (~10–20 μm38) to promote transdermal vaccine delivery. Upon removal of the compression of the MN patch, the resistance force decreased, which was mainly determined by the elastic recovery of the skin and MN patch and the friction between the MN and skin. Once the MNs were detached from the skin, the friction force became zero, causing a force increase at point “Q” (~3.5 N)39. Moreover, the above “press and release” actions on rat skin using the MN patch were repeated for 50 cycles, and the morphology of MNs varied little without damage (Fig. 3e), indicating that the MN patch can repeatedly poke the skin. The fracture performance of the MN patch was further investigated, as shown in Fig. 3 f. The resistance force increased with loading displacement. The MN tips were bent at almost 90° without breakage at point “a” (2.6 N/needle)40, as shown in Fig. 3g. The fracture force of the MN patch was much higher than the skin penetration force (0.08 N/needle) due to the high mechanical strength of the solid stainless steel MNs, indicating that these MN patches can penetrate the skin without bending. Therefore, the MN patches can easily penetrate skin, produce microchannels, and retract from skin without any damage, avoiding MN breakage in the skin, foreign body sensation, and the possibility of inflammation. Fig. 3Basic performance of the iontophoresis-driven MN patch.a The relationship between the resistance force and loading displacement during the “press and poke” stage of MN patch application. The penetration force per microneedle was ~81.2 mN. b Optical image of rat skin poked by the MN patch. The MN-poked microchannels are marked with red dye. c OCT image of rat skin poked by the MN patch, demonstrating successful skin penetration. d SEM image of a microchannel in rat skin poked by the MN patch. e SEM image of a magnified microneedle after (e1) the 25th cycle and (e2) the 50th cycle of skin penetration. f The relationship between the resistance force and loading displacement during the MN patch fracture test. g SEM image of the microneedle after the fracture test. The microneedle was bent but not broken. h The hydrogel was connected to a circuit, and a green LED became lit. i Swelling rate of the hydrogel. The swelling rate reached ~$530\%$ in 24 h (data are mean ± SD, $$n = 3$$). j The cumulative release of OVA from the hydrogel in PBS with/without application of a 1 mA/cm2 iontophoresis current (data are mean ± SD, $$n = 3$$). k The compression stress−strain curves of the hydrogel for the 1st, 40th, and 100th cycles. l The cyclic compression force curves of the vaccine-loaded hydrogel block with a maximum strain of $70\%$ The hydrogel was connected with a circuit, and a green light-emitting diode became lit (Fig. 3h), demonstrating good conductivity. The conductivity of the hydrogel was 0.16 S/m measured by the two-point probe method. The hydrogel, as a conductive medium, effectively guarantees conductivity between the electrode and skin. Ovalbumin (OVA, a model antigen) was loaded into the hydrogel via its swelling behavior. The weight of the hydrogel rapidly increased in the first 4 h and reached swelling equilibrium in ~8 h. The final swelling rate was 530.2 ± $3.1\%$ in 24 h (Fig. 3i). The OVA loading performance of the hydrogel was investigated. The maximum OVA loading capacity of the hydrogel (40 mg) reached 151.0 ± 16.5 μg with a loading efficiency of ~$60\%$ when the hydrogel was soaked in 250 µL of OVA solution (1 mg/mL). Moreover, OVA-FITC was uniformly distributed in the hydrogel (Fig. S9). The OVA vaccine release performance of the hydrogel was examined, and the schematic diagram of the test is shown in Fig. S10. Figure 3j shows the cumulative release of OVA from the hydrogel with/without application of a 1 mA/cm2 iontophoresis current. The OVA release rates with and without the application of a 1 mA/cm2 iontophoresis current for 30 min were 8.90 ± 0.51 μg/mL and 2.66 ± 0.53 μg/mL, respectively. The current increased 3.3-fold upon application of the iontophoresis current compared with only passive diffusion. Moreover, the electrostatic adsorption of cationic chitosan on the negatively charged proteins is also a possible reason for the low passive diffusion rate of the hydrogel41. Figure 3k, l presents the cyclic compression stress‒strain and force‒time curves of the hydrogels with a maximum strain of $70\%$, respectively. The elasticity of the hydrogels varied little after 100 compression cycles. The compression stress reached ~130 kPa at $70\%$ strain, and the compression force of the vaccine-loaded hydrogel block was ~30 N at $70\%$ strain. As shown in Fig. S11, the shape of the hydrogel block displayed little variation, and no obvious damage was observed after 100 cycles of repeated compressions. Therefore, the hydrogel can be compressed with good elasticity for skin penetration and MN detachment. ## In vitro transdermal vaccine delivery performance To explore the transdermal vaccine delivery performance of the iontophoresis-driven MN patch, vertical Franz diffusion cells were designed in-house for in vitro permeation tests, as shown in Fig. 4a. The cathode iontophoresis method was employed in this work since OVA in PBS (pH 7.4) carries negative charges due to its isoelectric point of 4.7542. The effects of skin penetration by the MNs and iontophoresis on the in vitro OVA delivery were systematically investigated, as shown in Fig. 4b. The cumulative concentration of OVA increased with experimental time, and a significant difference in cumulative permeation was clearly observed among these groups after 30 min of administration. The cumulative amounts permeated in the control, MN, 1 ITP, and MN/1 ITP groups were 2.80 ± 0.59 μg, 22.35 ± 2.32 μg, 7.61 ± 0.78 μg, and 47.57 ± 6.31 μg, respectively. Under free diffusion, OVA gradually passed through intact skin and was significantly facilitated (almost 8-fold) via the created microchannels. The microchannels created by the MNs effectively overcame the skin barrier and were beneficial for increasing the permeation rate. Conventional iontophoresis could also promote OVA delivery (2.7-fold) compared with free diffusion through intact skin. Transdermal OVA delivery could be further be enhanced via the combination of skin penetration and iontophoresis. The cumulative OVA permeation amount in the MN/1 ITP group was ~17-fold that in the control group. The microchannels created by the MNs provide transdermal delivery routes for iontophoresis, and the MNs and iontophoresis can cooperate to further improve transdermal permeation. Fig. 4In vitro and numerically calculated transdermal OVA delivery performance of the iontophoresis-driven MN patches.a Schematic illustration of in vitro transdermal OVA delivery using the iontophoresis-driven MN patch assembled on a Franz diffusion cell. b In vitro cumulative amount of permeated OVA in various groups, including the control, 1 ITP, MN, and MN/1 ITP groups. c The effect of iontophoresis-driven current on in vitro transdermal OVA delivery (data are the mean ± SD, $$n = 3$$). d Schematic illustration of the FEA model for transdermal OVA delivery using an iontophoresis-driven MN patch. e The calculated cumulative amount of permeated OVA in the control, 1 ITP, MN, and MN/1 ITP groups. f The effect of iontophoresis-driven current on the calculated cumulative amount of permeated OVA in the MN/0.5 ITP, MN/1 ITP, MN/1.5 ITP, and MN/2 ITP groups. g Comparison of the final cumulative amount of permeated OVA between the in vitro experimental and calculated groups (data are mean ± SD, $$n = 3$$). h The calculated transdermal delivery speed of OVA in the control, 1 ITP, MN, and MN/1 ITP groups. i The effect of iontophoresis-driven current on the calculated transdermal OVA delivery speed in the MN/0.5 ITP, MN/1 ITP, MN/1.5 ITP, and MN/2 ITP groups. j The calculated transdermal OVA delivery process in the control, 1 ITP, MN, MN/0.5 ITP, MN/1 ITP, MN/1.5 ITP, and MN/2 ITP groups The effect of the iontophoresis-driven current on the permeation rate was further investigated, as shown in Fig. 4c. The cumulative amounts permeated in the MN/0.5 ITP, MN/1 ITP, MN/1.5 ITP, and MN/2 ITP groups were 38.35 ± 3.04 μg, 47.57 ± 6.31 μg, 61.39 ± 6.61 μg, and 74.53 ± 10.82 μg, respectively. The cumulative amount of permeated OVA increased almost linearly with the iontophoresis-driven current (Fig. S12a). This significant increase is attributed to the positive effects of electromigration. The cumulative amount permeated in the MN/2 ITP group was the highest, which was approximately twofold that in the MN/0.5 ITP group. The linear relationship between the iontophoresis current and the permeation amount indicates that the amount of vaccine permeated through the skin can be controlled by the iontophoresis current. ## Numerical analysis of transdermal vaccine delivery The synergistic permeation mechanism of skin penetration by the MNs and iontophoresis was further analyzed using the FEA method. The FEA model of the iontophoresis-driven MN patch for transdermal OVA delivery was established using COMSOL Multiphysics, as shown in Fig. 4d and Fig. S13. This model was analyzed using the potential coupling of Transport of Diluted Species and Electric Current Physics. The model parameters were set based on the in vitro transdermal delivery of OVA. Figure 4j and Video S2 show the concentration distribution during the permeation process for each simulation group. The OVA flow distribution of the MN and MN/ITP groups can be clearly observed. In particular, the OVA concentration in the MN/ITP groups was distributed along the electric field, demonstrating the transport function of iontophoresis. Based on the simulation results, the cumulative amount of permeated material and permeation rate were further calculated, as shown in Fig. 4e–i. The cumulative amounts of permeated material in the control, 1 ITP, MN, and MN/1 ITP groups were 0.17, 2.28, 6.32, and 55.08 μg, respectively, as shown in Fig. 4e. The MN/1 ITP group combined the MN and iontophoresis technologies and showed the highest permeability. Moreover, the cumulative amount permeated in the MN/ITP group also increased with the application of iontophoresis current (Fig. 4f and Fig. S12b), which is consistent with the in vitro experimental results, demonstrating the effectiveness and controllability of active iontophoresis on vaccine delivery. According to Fig. 4e, the permeation rate in the above groups could be further calculated, as shown in Fig. 4h. The permeation rates in the control and 1 ITP groups were low due to the ideal skin barrier of the stratum corneum without skin appendages (such as hair follicles) in the simulation. The permeation rate in the MN group rapidly decreased and then reached a balance under passive diffusion via the created microchannels. According to Fick’s diffusion law, the diffusion rate is proportional to the concentration gradient, and the concentration gradient gradually decreased between the MN patch and receptor chamber. The permeation rate in the MN/1 ITP group first decreased at the initial stage because passive diffusion plays the main role during OVA delivery at this point, then gradually increased because of the greater ability of iontophoresis to drive the vaccine via the microchannels created by the MNs, and finally decreased owing to the significant drop in the amount of vaccine stored in the patch. The MN/1 ITP group showed the highest permeation rate due to the combined effect of the MNs and iontophoresis compared with the control, MN, and 1 ITP groups. The permeation rate in the MN/ITP groups increased with iontophoresis current, as shown in Fig. 4i. The simulation and experimental permeation curves exhibited very similar tendencies, as shown in Fig. 4e, f and Fig. 4b, c. Therefore, we compared the calculated and in vitro experimental cumulative amount of permeated OVA (Fig. 4g), and the MN/ITP groups showed a very similar tendency, further demonstrating that vaccine delivery via iontophoresis-driven MN patches can be controlled by the iontophoresis current. ## Immunization performance The levels of antigen-specific IgG and its subtypes (IgG1 and IgG2a) are positively correlated with the intensity of the immune response43. IgG1 and IgG2a antibody levels well reflect the intensity of Th2 and Th1 responses, respectively44. To study the antibody response induced by vaccines administered by the iontophoresis-driven MN delivery system, OVA was selected as the model antigen to induce the antibody response in BALB/c mice (Fig. 5a). The intensity of the immune response of the mice in the different groups (including intradermal injection, intramuscular injection, control, MN, 0.5 ITP, and MN/0.5 ITP group) was tested. The control group received transdermal vaccination using an iontophoresis-driven MN patch without MN poke and iontophoresis. The levels of OVA-specific IgG, IgG1, and IgG2a in mouse serum were measured two weeks after booster vaccination (Fig. 5b–d). As shown in Fig. 5b–d, all antibody levels of OVA-specific IgG, IgG1, and IgG2a in the MN/0.5 ITP group were slightly higher than those in the intramuscular injection group, indicating that transdermal vaccination with MN/0.5 ITP could produce similar or even higher immune response than typical intramuscular injection. However, typical intramuscular injection is slightly painful and requires professional operation by medical staff. The antibody levels of OVA-specific IgG, IgG1, and IgG2a in the MN/0.5 ITP group were significantly higher than those in the intradermal injection, control, MN and 0.5 ITP groups, indicating that the combination of the MN and iontophoresis techniques not only promoted transdermal vaccine delivery but also contributed to promoting both Th1- and Th2-type humoral immune responses. The antibody levels of OVA-specific IgG, IgG1, and IgG2a in the MN group were superior to those in the 0.5 ITP group due to the limited transdermal permeability of the macromolecule OVA (~43 kDa) driven by iontophoresis through intact skin. MNs can destroy the barrier layer of the stratum corneum to produce microchannels for passive diffusion and active iontophoresis. Above all, the synergistic effect of skin penetration and iontophoresis can enhance transdermal vaccine delivery. Elevated ALT and AST levels are associated with hepatocyte injury45. Transdermal vaccine delivery using an iontophoresis-driven MN patch was performed in mice for 30 min, serum was collected 24 h later, and the ALT and AST levels were analyzed, as shown in Fig. 5e. The ALT and AST levels of mice were within the normal range pre- and postvaccination without significant differences, indicating good biocompatibility of the iontophoresis-driven MN patch. Fig. 5In vivo immunization studies of the iontophoresis-driven MN patch and its iontophoresis-driven device.a Transdermal OVA delivery for the vaccination of BALB/c mice using an iontophoresis-driven MN patch and its device. Iontophoresis-driven MN patches loaded with OVA were bound to the right abdominal skin of mice with shaved hair. The OVA-specific b IgG, c IgG1, and d IgG2a levels in the serum of the control, intradermal injection, intramuscular injection, MN, 0.5 ITP, and MN/0.5 ITP groups (data are mean± SD, $$n = 5$$ per group). e ALT and AST levels in the serum of MN/0.5 ITP group mice pre- and postvaccination (data are mean ± SD, $$n = 3$$). f, g IVIS images and the corresponding fluorescent area in the control, MN, 0.5 ITP, and MN/0.5 ITP groups (data are mean ± SD, $$n = 3$$). h Mouse skin recovery after vaccination with the iontophoresis-driven MN patch (MN puncture and 0.5 mA/cm2 iontophoresis for 30 min). The treated skin self-recovered within 15 min. i Hematoxylin and eosin (H&E) staining of the main organs (heart, liver, spleen, lung and kidney) of the mice in the MN/0.5 ITP and control groups In vivo transdermal OVA delivery via the iontophoresis-driven MN patch was further studied using a small animal in vivo imaging system (IVIS). The near-infrared (NIR) wavelength range is favorable for in vivo imaging46, so Cy7-labeled OVA (Cy7-OVA) was selected and loaded into hydrogels for transdermal delivery. The in vivo fluorescence images of four groups (control, MN, 0.5 ITP, and MN/0.5 ITP group) after 30 min of treatment were observed, as shown in Fig. 5f, g. The fluorescence intensity in the MN/0.5 ITP group was the strongest, and its fluorescence area was the largest, suggesting that the combination of the MN and iontophoresis techniques could effectively deliver macromolecular OVA into the skin. Both the MN and 0.5 ITP groups promoted OVA transport into the skin. However, the control group without skin penetration or iontophoresis showed extremely low fluorescence intensity, demonstrating that transdermal delivery of the macromolecule OVA by passive diffusion was inefficient. Skin penetration using MNs and iontophoresis may cause some skin damage and irritation24,47, so the ability of the skin to recover after treatment with iontophoresis-driven MN patches was evaluated. As shown in Fig. S14, the micropores in the mouse back skins that had been poked by the MNs healed within 40 min without erythema or lesions. Moreover, skin treated with MN puncture and 30 min of 0.5 mA/cm2 iontophoresis was examined, as shown in Fig. 5h. Upon iontophoresis, erythema could be observed on the skin poked by MNs, which might be attributed to the decrease in skin tolerance after MN puncture. However, these mild symptoms completely resolved within 15 min; thus, treatment combining MN puncture with iontophoresis is relatively safe. The biosafety of the iontophoresis-driven MN patch in terms of the skin, heart, liver, spleen, lungs and kidneys of the mice after patch administration was further investigated, as shown in Fig. 5i and Fig. S15. Histopathological examination (H&E staining) was conducted. Compared with the control group, pathological sections of the mouse organs in the MN/0.5 ITP group showed that the tissues of each organ were normal, including intact myocardial fiber tissue, normal liver parenchyma cells, dense rounding of the renal corpuscles, filling of the lungs with alveoli, and no obvious inflammatory cell infiltration in the skin. Therefore, transdermal vaccine delivery caused no significant pathological changes or toxicity to visceral tissues. These experimental results demonstrate that iontophoresis-driven MN patches provide a safe platform for transdermal vaccine delivery. ## Conclusion We developed an iontophoresis-driven MN patch and a portable iontophoresis-driven device for efficient and controllable transdermal immunization. The iontophoresis-driven MN patch integrated the technologies of MNs and iontophoresis well, and the transdermal vaccine delivery strategy was “press and poke, iontophoresis-driven delivery, and immune response”. Hydrogel blocks with good biocompatibility, excellent conductivity, high elasticity, and a large loading capacity were prepared as the key component of the MN patches for vaccine storage and active iontophoresis. In vitro experiments demonstrated that iontophoresis-driven MN patches could control the transdermal OVA delivery process by tuning the iontophoresis current. Moreover, both in vitro and in vivo experiments demonstrated that the synergistic effect of skin penetration using MNs and iontophoresis could enhance the transdermal delivery efficiency of vaccine macromolecules. Iontophoresis-driven MN patches applied with a mild iontophoresis current raised few safety concerns. Above all, the wearable iontophoresis-driven MN system combining MNs with iontophoresis offers a promising strategy for achieving transdermal immunity in a painless and actively controlled manner. ## Ethics statement All animal procedures conducted in this work were reviewed, approved, and supervised by the Institutional Animal Care and Use Committee at Sun Yat-Sen University (Approval Number: IACUC–DD–16–0904). ## Materials and animals Chitosan, acrylamide, ammonium persulfate and N,N′-methylene bisacrylamide (Macklin, China) were purchased for the fabrication of the hydrogel. Ovalbumin (OVA) was purchased from Sigma, USA. OVA-FITC was purchased from Solarbio (China). OVA-Cy7 (Qiyuebio, China) was purchased for in vivo imaging of the mice. Sprague–Dawley rats (male, 200 ± 30 g) were provided by the Xinhua Experimental Animal Farm (Guangzhou, China). Fresh rat skin was prepared by removal of hair and subcutaneous fat for mechanical tests and in vitro transdermal vaccine permeation tests. BALB/c mice (female, 18 ± 2 g, 6-week old) were purchased for transdermal immunity from Beijing Vital River Laboratory Animal Technology Co., Ltd. ## Fabrication and characterization of iontophoresis-driven MN patches Iontophoresis-driven MN patches were assembled with Ag/AgCl electrodes on a flexible polyimide (PI), MNs, a conductive hydrogel and medical impermeable double-sided adhesive gasket, as shown in Fig. 1a. The Ag/AgCl electrode was prepared by chlorination of the Ag electrode with FeCl3 solution (0.1 M)36. Solid MNs were fabricated by micromilling 316 L stainless steel, which has been widely applied in medical implants, as it shows good mechanical strength and biocompatibility48. Polyacrylamide/chitosan vaccine-loaded hydrogels were prepared by UV polymerization and replica molding49,50. The morphology of the MNs was observed using scanning electron microscopy (SEM, Quanta 400F, Oxford, Holland). The freeze-dried hydrogel was sprayed with gold, and the cross-sectional morphology was observed by SEM. The conductivity of the hydrogel was tested using the two-point probe method50. The swelling rate of the hydrogel was measured by soaking the hydrogel in sufficient PBS until the hydrogels reached swelling equilibrium at room temperature. ## Fabrication of the iontophoresis-driven device A portable iontophoresis-driven device was developed to output a constant current for the iontophoresis-driven delivery of vaccines. A miniature printed circuit board (PCB) (size: 36.5 × 13.2 × 1.0 mm3) with ultralow power consumption for iontophoresis was designed using Altium Designer (Altium, Australia). The schematic circuit of the iontophoresis-driven device is shown in Fig. S1. The circuit was mainly composed of a charging module, a boosted module and a constant current output module, as shown in Figs. S2–7. The performance tests of the iontophoresis-driven device are shown in Fig. S8. The output current of the iontophoresis-driven device (0.5, 1, 1.5, and 2 mA) can be adjusted with a switch. Moreover, the insulating shell of the device was fabricated by a 3D printer (Sindon, 3DWOX, Korea) to encapsulate the PCB of the iontophoresis-driven circuit. ## Mechanical tests of the iontophoresis-driven MN patches The skin penetration performance of the iontophoresis-driven MN patch was investigated using a universal material testing machine (Instron, 5543 A, Boston, USA), as shown in Fig. S16a. Rat skin was fixed on polystyrene foam. The iontophoresis-driven MN patch was gradually pressed to 4 mm, held for 10 s, and subsequently released. The press and release speed were 0.1 mm/s. The penetrated rat skin was soaked in tissue fixative solution ($4\%$ paraformaldehyde), and the skin holes were observed by optical coherence tomography (OCT, HSO-2000, TEK SQRAY, China) and SEM. The mechanical stability of the MNs was studied by repeating the press-release actions at different locations on the rat skin for 50 cycles. The morphology of MNs after 25 and 50 cycles was observed by SEM.The fracture performance of MNs was tested using a universal material testing machine (Instron, 5967, Boston, USA), as shown in Fig. S16b. The MNs were moved toward and pressed on a stainless steel plate at a speed of 0.1 mm/s until the resistance force reached 3.5 kN. The pressed MNs were observed using SEM and an ultra-deep field microscope (Keyence, VHX-5000, Osaka, Japan), as shown in Fig. S17.The mechanical properties of the hydrogel were tested using a universal material testing machine. The cylindrical hydrogel block was compressed at a speed of 0.1 mm/s. The hydrogel block was repeatedly compressed for 100 cycles at a limit strain of $70\%$. ## In vitro OVA loading and delivery tests The dried hydrogel (40 mg) was immersed in 250 µL of OVA solution (1 mg/mL) and placed at 4 °C for 48 h. The loading efficiency of the hydrogel was defined as the ratio of the actual amount of drug loaded to the total amount of drug applied. The in vitro OVA transdermal permeation efficiency of the iontophoresis-driven MN patch was tested using an intelligent transdermal testing instrument (TP-3A, Albert Tech., China). OVA-FITC was loaded into the hydrogel, and the distribution of OVA-FITC in the hydrogel was observed using confocal laser scanning microscopy (CLSM, FV3000, Olympus, Japan). The in vitro OVA delivery performance of the iontophoresis-driven MN patch was tested using a self-designed Franz diffusion cell, as shown in Figs. S18, 19. The receptor chamber (17 mL) was filled with phosphate-buffered saline (PBS, pH = 7.4). Rat skin samples with a size of 3 × 4 cm2 were fixed on the receptor chamber. The iontophoresis-driven MN patch was assembled on the donor chamber, and the iontophoresis-driven device was used to promote permeation. In vitro transdermal delivery tests were divided into 7 groups ($$n = 3$$), as listed in Table 1. [ 1] Control group: intact skin; [2] 1 ITP group: intact skin and iontophoresis with a current density of 1 mA/cm2; [3] MN group: skin penetration using MNs; [4] MN/0.5 ITP group: skin penetration using MNs and iontophoresis with a current density of 0.5 mA/cm2; [5] MN/1 ITP group: skin penetration using MNs and iontophoresis with a current density of 1 mA/cm2; [6] MN/1.5 ITP group: skin penetration using MNs and iontophoresis with a current density of 1.5 mA/cm2; and [7] MN/2 ITP group: skin penetration using MNs and iontophoresis with a current density of 2 mA/cm2. The OVA concentration in the diffusion cell was measured using a micro-BCA protein quantification kit (Leagene, pt0006-500t, Beijing, China).Table 1In vitro OVA transdermal delivery in the 7 groupsGroup nameSkin treatmentIontophoresisControl groupIntact skin0 mA/cm21 ITP groupIntact skin1 mA/cm2MN groupSkin penetration0 mA/cm2MN/0.5 ITP groupSkin penetration0.5 mA/cm2MN/1 ITP groupSkin penetration1 mA/cm2MN/1.5 ITP groupSkin penetration1.5 mA/cm2MN/2 ITP groupSkin penetration2 mA/cm2 ## Numerical simulations of transdermal vaccine delivery The transdermal vaccine delivery process by the iontophoresis-driven MN patch combining skin penetration using MNs and iontophoresis was analyzed by the finite element analysis (FEA) method. The FEA model of the iontophoresis-driven MN patch was established using COMSOL Multiphysics, as shown in Fig. S13. The FEA models and parameters are listed in Table S1. The synergistic permeation effect of skin penetration using MNs and iontophoresis with the iontophoresis-driven MN patches was calculated using potential coupling multiphysics, including transport of diluted species and electric current physics. The simulation conditions of the seven groups were the same as those of the OVA in vitro diffusion experiment. Ion transport was governed by the Nernst–Planck flux equation51. The diffusion rate, total amount diffused, and concentration distribution of OVA in the different simulation groups were calculated. ## Immunization studies The in vivo immunization performance of the iontophoresis-driven MN patch was investigated. Female BALB/c mice (18 ± 2 g, 6-week old) were vaccinated and divided into six groups: control group, intradermal injection group, intramuscular injection group, MN group, 0.5 ITP group, and MN/0.5 ITP group ($$n = 5$$ per group). The hair on the right abdomen of each mouse was shaved 24 h before vaccination. The dried hydrogels (40 mg) were immersed in 250 µL of OVA solution (1 mg/mL) for 48 h at 4 °C to load OVA into the hydrogel for transdermal vaccination. The control, MN, 0.5 ITP and MN/0.5 ITP groups were vaccinated with iontophoresis-driven MN patches for 30 min. The control group was treated using iontophoresis-driven MN patches without skin penetration and iontophoresis. This procedure was repeated again after 2 weeks for the booster dose. Two weeks after the booster dose, blood from the retro-orbital plexus was collected, and serum was separated for antibody concentration analysis. OVA-specific IgG, IgG1, and IgG2a levels in the serum samples were measured using ELISA kits (MEIMIAN Industrial Co., Ltd., Jiangsu, China). ## Biosafety assessment The abdominal skin recovery in the MN group and MN/0.5 ITP group after vaccination were observed. In MN/0.5 ITP group, blood from the retro-orbital plexus was collected before vaccination and 24 h after vaccination with the iontophoresis-driven MN patch. ALT and AST levels in the serum samples were measured using ELISA kits (Changchun Huili Biotech Co., Ltd., Jilin, China). Moreover, the skin, heart, liver, spleen, lungs and kidneys of the mice in the control group and MN/0.5 ITP group were harvested and stained with hematoxylin-eosin (H&E). ## Statistical analysis The experimental data were calculated and are expressed as the mean ± standard deviation (SD). The differences between two groups were determined by Student’s t test. A one-sample t test was used to determine significant differences between the experimental and FEA values. p values < 0.05 indicated that the results were considered statistically significant (*p value < 0.05, **p value < 0.01, ***p value < 0.001). ## Supplementary information Supplementary Information Video S1 Video S2 The online version contains supplementary material available at 10.1038/s41378-023-00515-1. ## References 1. Wang C, Wang Z, Wang G. **COVID-19 in early 2021: current status and looking forward**. *Signal Transduct. Target. Ther.* (2021.0) **6** 114. DOI: 10.1038/s41392-021-00527-1 2. Xiong Y, Liu Y, Cao L, Wang D. **Transcriptomic characteristics of bronchoalveolar lavage fluid and peripheral blood mononuclear cells in COVID-19 patients**. *Emerg. Microbes Infect.* (2020.0) **9** 76-770. DOI: 10.1080/22221751.2020.1747363 3. Bastola R. **Vaccine adjuvants: smart components to boost the immune system**. *Arch. Pharm. Res.* (2017.0) **40** 1238-1248. DOI: 10.1007/s12272-017-0969-z 4. Nguyen TT. **Progress in microneedle array patch (MAP) for vaccine delivery**. *Hum. Vaccin Immunother.* (2021.0) **17** 316-327. DOI: 10.1080/21645515.2020.1767997 5. Doherty M, Buchy P, Standaert B, Giaquinto C, Prado-Cohrs D. **Vaccine impact: benefits for human health**. *Vaccine* (2016.0) **34** 6707-6714. DOI: 10.1016/j.vaccine.2016.10.025 6. Caudill C. **Transdermal vaccination via 3D-printed microneedles induces potent humoral and cellular immunity**. *Proc. Natl Acad. Sci. USA* (2021.0) **118** e2102595118. DOI: 10.1073/pnas.2102595118 7. Orenius T, LicPsych H, Saila K, Mikola L. **Fear of injections and needle phobia among children and adolescents: an overview of psychological, behavioral, and contextual factors**. *SAGE Open Nurs.* (2018.0) **4** 2377960818759442. PMID: 33415191 8. Yilmaz G, Alemdar DK. **Using buzzy, shotblocker, and bubble blowing in a pediatric emergency department to reduce the pain and fear caused by intramuscular injection: a randomized controlled trial**. *J. Emerg. Nurs.* (2019.0) **45** 502-511. DOI: 10.1016/j.jen.2019.04.003 9. Singh B. **Needle-free immunization with chitosan-based systems**. *Int. J. Mol. Sci.* (2018.0) **19** 3639. DOI: 10.3390/ijms19113639 10. Pattarabhiran SP. **Dissolvable microneedle-mediated transcutaneous delivery of tetanus toxoid elicits effective immune response**. *AAPS PharmSciTech* (2019.0) **20** 257. DOI: 10.1208/s12249-019-1471-3 11. Balmert SC. **Dissolving undercut microneedle arrays for multicomponent cutaneous vaccination**. *J. Control Release* (2020.0) **317** 336-346. DOI: 10.1016/j.jconrel.2019.11.023 12. Du G. **Intradermal vaccination with hollow microneedles: a comparative study of various protein antigen and adjuvant encapsulated nanoparticles**. *J. Control Release* (2017.0) **266** 109-118. DOI: 10.1016/j.jconrel.2017.09.021 13. Chowdhury MR, Moshikur RM, Wakabayashi R, Moniruzzaman M, Goto M. **Biocompatible ionic liquids assisted transdermal co-delivery of antigenic protein and adjuvant for cancer immunotherapy**. *Int J. Pharm.* (2021.0) **601** 120582. DOI: 10.1016/j.ijpharm.2021.120582 14. Azmana M. **Transdermal drug delivery system through polymeric microneedle: a recent update**. *J. Drug Deliv. Sci. Technol.* (2020.0) **60** 101877. DOI: 10.1016/j.jddst.2020.101877 15. Chen M. **Cold to hot: binary cooperative microneedle array-amplified photoimmunotherapy for eliciting antitumor immunity and the Abscopal effect**. *ACS Appl. Mater. Interfaces* (2020.0) **12** 32259-32269. DOI: 10.1021/acsami.0c05090 16. Halder J, Gupta S, Kumari R, Gupta GD, Rai VK. **Microneedle array: applications, recent advances, and clinical pertinence in transdermal drug delivery**. *J. Pharm. Innov.* (2021.0) **16** 558-565. DOI: 10.1007/s12247-020-09460-2 17. Nagarkar R, Singh M, Nguyen HX, Jonnalagadda S. **A review of recent advances in microneedle technology for transdermal drug delivery**. *J. Drug Deliv. Sci. Technol.* (2020.0) **59** 101923. DOI: 10.1016/j.jddst.2020.101923 18. Lopez-Ramirez MA. **Built-in active microneedle patch with enhanced autonomous drug delivery**. *Adv. Mater.* (2020.0) **32** e1905740. DOI: 10.1002/adma.201905740 19. Takeuchi I, Suzuki T, Makino K. **Iontophoretic transdermal delivery using chitosan-coated PLGA nanoparticles for transcutaneous immunization**. *Colloids Surf. A: Physicochem. Eng. Asp.* (2021.0) **608** 125607. DOI: 10.1016/j.colsurfa.2020.125607 20. Fukuta T, Oshima Y, Michiue K, Tanaka D, Kogure K. **Non-invasive delivery of biological macromolecular drugs into the skin by iontophoresis and its application to psoriasis treatment**. *J. Control Release* (2020.0) **323** 323-332. DOI: 10.1016/j.jconrel.2020.04.044 21. Saepang K, Li SK, Chantasart D. **Passive and iontophoretic transport of pramipexole dihydrochloride across human skin microchannels created by microneedles in vitro**. *Int. J. Pharmaceutics* (2021.0) **609** 121092. DOI: 10.1016/j.ijpharm.2021.121092 22. Hettinga J, Carlisle R. **Vaccination into the dermal compartment: techniques, challenges, and prospects**. *Vaccines* (2020.0) **8** 534. DOI: 10.3390/vaccines8030534 23. Wang Y, Thakur R, Fan Q, Michniak B. **Transdermal iontophoresis: combination strategies to improve transdermal iontophoretic drug delivery**. *Eur. J. Pharm. Biopharm.* (2005.0) **60** 179-191. DOI: 10.1016/j.ejpb.2004.12.008 24. Bakshi P, Vora D, Hemmady K, Banga AK. **Iontophoretic skin delivery systems: success and failures**. *Int. J. Pharm.* (2020.0) **586** 119584. DOI: 10.1016/j.ijpharm.2020.119584 25. Yamada M, Prow TW. **Physical drug delivery enhancement for aged skin, UV damaged skin and skin cancer: Translation and commercialization**. *Adv. Drug Deliv. Rev.* (2020.0) **153** 2-17. DOI: 10.1016/j.addr.2020.04.008 26. Prausnitz MR, Langer R. **Transdermal drug delivery**. *Nat. Biotechnol.* (2008.0) **26** 1261-1268. DOI: 10.1038/nbt.1504 27. Dhote V, Bhatnagar P, Mishra PK, Mahajan SC, Mishra DK. **Iontophoresis: a potential emergence of a transdermal drug delivery system**. *Sci. Pharm.* (2012.0) **80** 1-28. DOI: 10.3797/scipharm.1108-20 28. Marwah H, Garg T, Goyal AK, Rath G. **Permeation enhancer strategies in transdermal drug delivery**. *Drug Deliv.* (2016.0) **23** 564-578. DOI: 10.3109/10717544.2014.935532 29. Wang Y, Zeng L, Song W, Liu J. **Influencing factors and drug application of iontophoresis in transdermal drug delivery: an overview of recent progress**. *Drug Deliv. Transl. Res.* (2022.0) **12** 15-26. DOI: 10.1007/s13346-021-00898-6 30. Li Y. **Iontophoresis-driven porous microneedle array patch for active transdermal drug delivery**. *Acta Biomater.* (2021.0) **121** 349-358. DOI: 10.1016/j.actbio.2020.12.023 31. Kale M, Kipping T, Banga AK. **Modulated delivery of donepezil using a combination of skin microporation and iontophoresis**. *Int. J. Pharm.* (2020.0) **589** 119853. DOI: 10.1016/j.ijpharm.2020.119853 32. Kusama S. **Transdermal electroosmotic flow generated by a porous microneedle array patch**. *Nat. Commun.* (2021.0) **12** 658. DOI: 10.1038/s41467-021-20948-4 33. 33.Kunal Pal et al. in Bioelectronics and Medical Devices, Ch. 16 (Nayak, 2019). 34. Sheng T. **Microneedle-mediated vaccination: innovation and translation**. *Adv. Drug Deliv. Rev.* (2021.0) **179** 113919. DOI: 10.1016/j.addr.2021.113919 35. Krieger KJ. **Development and evaluation of 3D‐printed dry microneedle electrodes for surface electromyography**. *Adv. Mater. Technol.* (2020.0) **5** 2000518. DOI: 10.1002/admt.202000518 36. Teymourian H. **Microneedle-based detection of ketone bodies along with glucose and lactate: toward real-time continuous interstitial fluid monitoring of diabetic ketosis and ketoacidosis**. *Anal. Chem.* (2020.0) **92** 2291-2300. DOI: 10.1021/acs.analchem.9b05109 37. Lim SH, Ng JY, Kang L. **Three-dimensional printing of a microneedle array on personalized curved surfaces for dual-pronged treatment of trigger finger**. *Biofabrication* (2017.0) **9** 015010. DOI: 10.1088/1758-5090/9/1/015010 38. van der Maaden K, Jiskoot W, Bouwstra J. **Microneedle technologies for (trans)dermal drug and vaccine delivery**. *J. Control Release* (2012.0) **161** 645-655. DOI: 10.1016/j.jconrel.2012.01.042 39. Yang J. **Touch-actuated microneedle array patch for closed-loop transdermal drug delivery**. *Drug Deliv.* (2018.0) **25** 1728-1739. DOI: 10.1080/10717544.2018.1507060 40. Chen K. **Fabrication of micro-needle electrodes for bio-signal recording by a magnetization-induced self-assembly method**. *Sensors* (2016.0) **16** 1533. DOI: 10.3390/s16091533 41. Ahmadi F, Oveisi Z, Samani SM, Amoozgar Z. **Chitosan based hydrogels: characteristics and pharmaceutical applications**. *Res. Pharm. Sci.* (2015.0) **10** 1-16. PMID: 26430453 42. Chen Y. **Interactions and emulsifying properties of ovalbumin with tannic acid**. *Lwt* (2018.0) **95** 282-288. DOI: 10.1016/j.lwt.2018.04.088 43. Li Z, He Y, Deng L, Zhang ZR, Lin Y. **A fast-dissolving microneedle array loaded with chitosan nanoparticles to evoke systemic immune responses in mice**. *J. Mater. Chem. B* (2020.0) **8** 216-225. DOI: 10.1039/C9TB02061F 44. Niu L, Chu LY, Burton SA, Hansen KJ, Panyam J. **Intradermal delivery of vaccine nanoparticles using hollow microneedle array generates enhanced and balanced immune response**. *J. Control Release* (2019.0) **294** 268-278. DOI: 10.1016/j.jconrel.2018.12.026 45. Chen X. **Hepatoprotective effects of lactobacillus on carbon tetrachloride-induced acute liver injury in mice**. *Int. J. Mol. Sci.* (2018.0) **19** 2212. DOI: 10.3390/ijms19082212 46. Li DH, Schreiber CL, Smith BD. **Sterically shielded heptamethine cyanine dyes for bioconjugation and high performance near-infrared fluorescence imaging**. *Angew. Chem. Int. Ed. Engl.* (2020.0) **59** 12154-12161. DOI: 10.1002/anie.202004449 47. Chu S, Foulad DP, Atanaskova Mesinkovska N. **Safety profile for microneedling: a systematic review**. *Dermatol. Surg.* (2021.0) **47** 1249-1254. DOI: 10.1097/01.DSS.0000790428.70373.f6 48. Zhu W. **Preparation and characterization of diamond-like carbon (DLC) film on 316L stainless steel by microwave plasma chemical vapor deposition (MPCVD)**. *Diam. Relat. Mater.* (2022.0) **122** 108820. DOI: 10.1016/j.diamond.2022.108820 49. Merkel TJ. **Scalable, shape-specific, top-down fabrication methods for the synthesis of engineered colloidal particles**. *Langmuir* (2010.0) **26** 13086-13096. DOI: 10.1021/la903890h 50. Gan D. **Conductive and tough hydrogels based on biopolymer molecular templates for controlling in situ formation of polypyrrole nanorods**. *ACS Appl. Mater. Interfaces* (2018.0) **10** 36218-36228. DOI: 10.1021/acsami.8b10280 51. Wang W, Chiang TY, Velegol D, Mallouk TE. **Understanding the efficiency of autonomous nano- and microscale motors**. *J. Am. Chem. Soc.* (2013.0) **135** 10557-10565. DOI: 10.1021/ja405135f
--- title: 'Scoping review of sexual and reproductive healthcare for men in the MENA (Middle East and North Africa) region: a handful of paradoxes?' authors: - Walid El Ansari - Mohamed Arafa - Haitham Elbardisi - Ahmad Majzoub - Mohammed Mahdi - Ahmed Albakr - Khalid AlRumaihi - Abdulla Al Ansari journal: BMC Public Health year: 2023 pmcid: PMC10040932 doi: 10.1186/s12889-022-14716-2 license: CC BY 4.0 --- # Scoping review of sexual and reproductive healthcare for men in the MENA (Middle East and North Africa) region: a handful of paradoxes? ## Abstract ### Background No study appraised the knowledge gaps and factors impacting men’s sexual and reproductive health (SRH) in MENA (Middle East and North Africa). The current scoping review undertook this task. ### Methods We searched PubMed and Web of Science (WoS) electronic databases for original articles on men’s SRH published from MENA. Data was extracted from the selected articles and mapped out employing the WHO framework for operationalising SRH. Analyses and data synthesis identified the factors impacting on men’s experiences of and access to SRH. ### Results A total of 98 articles met the inclusion criteria and were included in the analysis. The majority of studies focused on HIV and other sexually transmissible infections ($67\%$); followed by comprehensive education and information ($10\%$); contraception counselling/provision ($9\%$); sexual function and psychosexual counselling ($5\%$); fertility care ($8\%$); and gender-based violence prevention, support/care ($1\%$). There were no studies on antenatal/intrapartum/postnatal care and on safe abortion care ($0\%$ for both). Conceptually, there was lack of knowledge of the different domains of men’s SRH, with negative attitudes, and many misconceptions; as well as a deficiency of health system policies, strategies and interventions for SRH. ### Conclusion Men’s SRH is not sufficiently prioritized. We observed five ‘paradoxes’: strong focus on HIV/AIDS, when MENA has low prevalence of HIV; weak focus on both fertility and sexual dysfunctions, despite their high prevalence in MENA; no publications on men’s involvement in sexual gender-based violence, despite its frequency across MENA; no studies of men’s involvement in antenatal/intrapartum/postnatal care, despite the international literature valuing such involvement; and, many studies identifying lack of SRH knowledge, but no publications on policies and strategies addressing such shortcoming. These ‘mismatches’ suggest the necessity for efforts to enhance the education of the general population and healthcare workers, as well as improvements across MENA health systems, with future research examining their effects on men’s SRH. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12889-022-14716-2. ## Introduction Historically, sexual and reproductive health (SRH) and rights were viewed as a woman’s issue [1]. One of the United Nation’s Sustainable Development *Goals is* ensuring universal access to SRH [2]. However, despite the growing recognition that men also need SRH care, they remain underrepresented in the SRH social debate, care, and research [3]. Interestingly, the impact of male SRH on men’s welfare was only considered as function of women’s SRH rights [4]. Generally, barriers to men’s engagement in SRH include the lack of health insurance, masculinity ideas that conflict with SRH care, stigma related to accessing services, and lack of knowledge regarding available services [5]. For instance, across high- and low-income countries, shortcomings of men’s SRH at the policy level, availability and accessibility of clinical services and acceptability by society are evident [2, 6]. The Middle East and North Africa (MENA) region is no exception. The Reproductive Health Working Group established in 1988 in Cairo to advance research in the Arab countries and Turkey began with a focus on women [7]. With time, the focus widened to include men’s lives and their impact on women’s health [8]. Three decades later, the situation remains not much changed [9, 10]. Some Arab nations have attempted health system reforms recently, however further efforts are still needed to include SRH [11]. Challenges that appear to hinder the implementation of new strategies that facilitate research of SRH issues among men in MENA include attitudes of male dominancy, and traditional myths that led to stigma and biases when dealing with SRH. In addition, sociopolitical instabilities in MENA have resulted in one of the highest numbers and range of refugees/humanitarian settings globally, associated with collapsed health systems, lack of essential medications and contraceptives, as well as absence of and low access to skilled health care providers (HCP) [12]. Therefore, the current scoping review aimed to outline the knowledge gaps and considerations that impact on men regarding SRH in MENA. Specifically, we appraised the range of factors that bear on men’s SRH in terms of clients/users, healthcare providers (HCPs), healthcare system and sociopolitical environment. ## Scoping review The purpose of a scoping review is not to localize and account for every published information on the topic [13]. Rather, the goal is intentionally wider, to interrogate the literature, discover the important features of the topic, unearth potential gaps, display crucial examples and synthesize research evidence, particularly when the subject has not been meticulously reviewed or is complicated, heterogeneous and assorted [14–16]. Therefore, the scoping review was selected to appraise SRH care for men in MENA. The current scoping review was undertaken in line with The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) extension for scoping reviews [17]. We employed a six-step framework in line with Arksey and O’Malley [18], with procedural and methodological rigor, and clarity/transparency relating to methodology (Table 1) [19–21]. Table 2 outlines the definitions used in this review. The search terms employed are depicted in a supplementary file (Supplementary Box 1). The inclusion criteria employed by the current review included: [1] peer-reviewed empirical studies, all designs were taken into account; [2] published between January 2010 and May 2020; and [3] appraising the experiences of men in SRHC or the healthcare providers’ (HCPs) perspectives on men’s SRHC; and [4] undertaken in the nations of the MENA region. Items that did not meet the inclusion criteria were excluded. Table 1Six-step framework employed in the present scoping reviewStepDescriptionResearch questions *What is* the current status of the literature published from MENA regarding men and SRH?; How men in MENA experience SRH? These queries were broken into four specific objectives relating to clients/users, HCPs, healthcare system and sociopolitical environment. Search strategies Structured literature search using electronic databases that included PubMed and WoS, limited to English language using search terms. ( online supplemental Box 1)Charting the Data Data extracted consisted of items relevant to specific factors examined. Articles were mapped employing the eight domains of WHO framework for operationalising SRH [19]Collating, Summarizing, and Reporting ResultsReview team assembled, grouped, synthesized and condensed the findings. Theoretical frame work for analysis involved two interrelated schemes [20, 21] (detailed below). Potential gaps were mapped. Consultation ExerciseTwo senior experts specialized in SRH reviewed the findings to provide opinion on and substantiate the findingsAdopted from Arksey and O’Malley [18]Table 2Terms and definitions used in the current scoping reviewTermDefinitionSexual healthIntegration of the somatic, emotional, intellectual, and social aspects of sexual being, in ways that are positively enriching and that enhance personality, communication, and love [19].Reproductive healthState of complete physical, mental and social well-being and not merely the absence of disease or infirmity, in all matters relating to the reproductive system and to its functions and processes [2].MENAAlgeria, Bahrain, Djibouti, Egypt, Iraq, Jordan, Kuwait, Lebanon, Libya, Malta, Morocco, Oman, Qatar, Saudi Arabia, Syria, Tunisia, United Arab Emirates, Palestine, Yemen, Sudan, Western Sahara. Scoping review Type of research synthesis that aims to ‘map the literature on a particular topic or research area and provide an opportunity to identify key concepts; gaps in the research; and types and sources of evidence to inform practice, policymaking, and research [16].MENA Middle East and North Africa The review team systematically synthesized the findings and summarized and presented the charting results as they relate to the review questions and objectives. This included extracting the aims, populations, findings and conclusions of each included study development of an excel sheet. Then groupings were created based on the eight domains of SRH as outlined by WHO [19] framework for operationalising sexual health and its linkages to reproductive health, and further inspired by the findings that were emerging. Any disagreements were resolved by consensus between the team members. ## Theoretical frameworks The theoretical framework for analysis we employed involved two interrelated schemes. The first is WHO’s [2017] framework for operationalising sexual health and its linkages to reproductive health, comprising 8 domains: antenatal, intrapartum and postnatal care; comprehensive education and information; contraception counselling and provision; gender-based violence prevention, support and care; fertility care; prevention and control of HIV and other sexually transmissible infections (STIs); safe abortion care; sexual function and psychosexual counselling [19]. Then, for each of these domains, we further employed Kilbourne et al’s [2006] framework to outline the health service perspectives on the appreciation of health and healthcare disparities, highlighting the factors influencing men’s experiences, including: individual (HCPs and users); interpersonal (healthcare encounter and contact characteristics); organisational (healthcare system); and the larger influence of the community and public policies (sociopolitical). Our searches of men’s SRH yielded very sparse articles on healthcare encounter and contact characteristics, therefore our domains are organized under 3 categories, namely clients/ HCPs; healthcare system factors; and sociopolitical factors [22]. ## Search results Figure 1 shows the PRISMA flow chart of the search results of men’s experiences in sexual and reproductive healthcare in MENA countries. After searching and screening, a total of 98 articles were finally included in the present review. Fig. 1 PRISMA flow chart on search results of men’s experiences in sexual and reproductive healthcare in MENA countries [23] Men’s SRH topics were grouped using the WHO framework, illustrating the intertwined features between sexual health and reproductive health [1]. The great majority of studies focused on the prevention and control of HIV and other sexually transmissible infections ($67\%$). Conversely, only 1 study ($1\%$) discussed gender-based violence; and no studies attended to the two domains of antenatal, intrapartum and postnatal care and safe abortion care ($0\%$ for both) (Fig. 2). Fig. 2Description of identified studies in terms of WHO’s Framework [19] ## Clients and populations The identified studies included children/adolescents, youth, school/university students, HCPs, particular patient groups, general public or special populations e.g., industrial/tourist workers, refugees, truck drivers, refugees, seafarers, alcohol/drug abusers, men having sex with men and female/male sex workers, and people living with HIV/AIDS (PLWHA) or healthy children/ adolescents of HIV-positive parents (Supplementary Table 1). ## Comprehensive education and information A total of $10\%$ of the studies we identified were dedicated per se to comprehensive education. However, education was generally discussed in the greater majority of the articles that this review identified, focusing on discussions of the knowledge levels of different population groups towards several aspects of SRH in men, exploring the determinants of knowledge gaps among clients and HCPs, as well as health systems and sociopolitical factors (detailed below). Generally, these studies identified a knowledge lack regarding issues relating to men’s SRH in MENA. ## Prevention and control of HIV and other sexually transmissible infections Most studies of the review assessed the knowledge of a range of population groups pertaining to prevention and control of STIs and HIV/AIDS. More focus was directed towards knowledge level among clients/ users than of HCPs. Overall, there was low knowledge level across most of the groups that were assessed. Knowledge level was influenced by factors including age, sex, level of education and experience with age. Knowledge deficits often translated into negative attitudes towards HIV/AIDS or PLWHA. In addition, little is known about the perception of the magnitude of HIV/AIDS in MENA (Table 3). Table 3HIV and STIs in MENA: characteristics of users and health care professionalsCharacteristicClients/ usersHealth Care ProfessionalsKnowledge▸ School/ High school students: HIV/AIDS knowledge and prevention: inadequate; better among high school students; needed more understanding to prevent stigmatization/ discrimination of infected persons, knowledge varied significantly by country and gender [24–27]. Most boys knew about AIDS but rarely other STIs [28]. ▸ University students• HIV/AIDS: medical university students were aware of HIV, its transmission and prevention [29], with few misconceptions [30]. Conversely, dental students had low to moderate knowledge with high misconceptions and paramedical students had low knowledge [31, 32], knowledge was sometimes associated with being male and higher years of study [32, 33].• STIs: male medical students and dental students had low HPV knowledge [34]; clinical-years associated with better knowledge [35]. ▸ General population• HIV/AIDS: deficient knowledge, with misconceptions about prevention [36, 37]. Knowledge was positively associated with education, age, residence, experience, and socioeconomic status [38, 39].• STIs: low HPV awareness, which was better among older clients and females [40]. ▸ Special populations: HIV knowledge was high among PLWHA and alcohol/ drug abusers especially men with high education [41, 42], satisfactory in seafarers, but with some misconceptions, and low in refugees and dental patients [43–45]. ▸ Physicians: PHC physicians sometimes had never managed an AIDS case; had low HIV/AIDS transmission, treatment and risk behaviour knowledge [46]. Knowledge was associated with years of experience, status/specialty and practice location [47]. ▸ Dentists: moderate knowledge about oral HIV manifestations and transmission [48]. ▸ Nurses: low HIV/AIDS disease and prevention, however, had high knowledge in risk groups identification [49].Attitude▸ School/ high school students: negative attitudes toward AIDS and PLWHA but were willing to be HIV tested [31]. ▸ University students: undergraduates displayed moderate acceptance of PLWHA, and most were willing to care for an HIV-infected person, although attitudes fluctuated between equivocal or negative which was related to lack of HIV knowledge [29, 30, 50]. HPV vaccination was acceptable by male medical students and dental students [34], more among clinical-year students, those vaccinated for hepatitis B, and with higher HPV knowledge [35]. ▸ General population: although individuals expressed eagerness to know more about HIV/AIDS [51], a sense of fatalism regarding HIV acquisition was common [36], with negative attitude toward PLWHA. Factors affecting attitude were age, sex, marital or social status, educational level, experience, and nationality [52]. ▸ Special populations• PLWHA: low adherence to treatment [53].• Seafarers, sex workers and refugees: high risk behaviors [43, 44, 54].• Most alcohol/ drug abusers: negative attitudes towards PLWHA, but $55.5\%$ felt sympathy for them [42]. ▸ Physicians: most PHC physicians suggested isolating PLWHA in isolated places/hospitals [46]. ▸ Nurses: negative attitudes toward PLWHA/ suspected HIV cases (injecting drug users, MSM, sex workers), refusing to provide care or get blood sample; most reported that HIV patients should be ashamed of themselves [49, 55]. Attitude barriers to care included fear of getting infected with HIV, disbelief in effectiveness of infection control measures, misconceptions, fear of stigmatization, and moral judgments [56].Perceptions▸ Kuwait: majority of participants were satisfied with the government’s policy for AIDS prevention; and proposed that religion is important in dealing with HIV infection [38]. ▸ Egypt: compared to industrial workers, tourism workers had a better perception of the magnitude of the HIV/AIDS problem worldwide and in Egypt, and the likelihood of it worsening [57].HCPs Health care professionals, PHC Primary healthcare, HPV *Human papilloma* virus, PLWHA People living with HIV/AIDS, MSM Men who have sex with men, STIs Sexually transmitted infections In terms of the healthcare system, for HIV, much had been achieved in linking to and retention in care, antiretroviral therapy coverage and viral suppression, despite obstacles in prevention programs e.g., deficient funding and infection control due to lack in supplies and procedures as well as insufficient data/surveillance [58, 59]. The financial burden/ healthcare costs of PLWHA varied, depending on the presenting illness, clinical stage, developed opportunistic infection, co-morbidity, and pharmacological therapy [60], with empirical results illustrating a negative relationship between both public and private healthcare spending and HIV [61]. Policies and protocols regarding dealing with PLWHA were also absent [55]. In addition, war had restricted the surveillance activities e.g., access to voluntary counseling and treatment (VCT) centers in Syria [62]. Therefore, many opportunities for HIV testing, based on at-risk behaviors or clinical signs, were missed [63]. Several HIV awareness programs have been implemented. At hospital level, introduction of multi-disciplinary team (MDT) approach in managing HIV patients resulted in statistically significant control of the disease [64]. On the population level, school-based HIV education interventions exhibited mixed effectiveness in improving knowledge of HIV transmission/ prevention [24, 65, 66]. The key enabling factors were high quality of training for peer educators, supportive school principals, and parental acceptance of the intervention [67]. Community prevention in Sudan and Yemen was directed to the general public as well as to men who have sex with men (MSM) and female sex worker (FSW), focusing on behavioural change, enabling supportive environments and providing support for PLWHA [58]. Likewise, community-based educational interventions targeting truck drivers was effective in increasing coverage of HIV testing and counseling [68]. NGOs In Lebanon adopted HIV self-testing but uptake was low due to noncompliance of beneficiaries and lack of human/financial resources. Interestingly, self-testing was much improved during COVID-19 because of the absence of on-site activities, shifting more efforts towards HIV self-testing [69]. Generally, the apparent trend that this review observed was a deficiency of studies addressing the healthcare system and policies pertaining to the prevention and control of HIV and other STIs. As for the social factors, high stigma and discrimination toward PLWHA were present [26, 39], rooted in values and fears, and manifesting in reluctance to use the same health facilities as PLWHA [70]. PLWHA faced such stigma in their homes and at work, forcing them to seek support from NGOs or close family. This stigma affected their disclosure to the wider community due to uncertainty of the repercussions, leading to a lonely life and financial difficulties [41]. In addition, HIV testing uptake was limited by concerns about confidentiality and fear of repercussions on health and employment [71]. Stigmatization of PLWHA was inversely related to HIV/AIDS knowledge [47, 72]. Stigma extended to physicians providing care for PLWHA, caused by fear of infection, to the extent of community unwillingness to use those physicians’ services. On the other hand, stigma toward physicians who refused to provide care was linked to perceptions of unethical behavior [70]. Victimization was also evident, e.g., most Saudi students believed that PLWHA were responsible for their infection and that AIDS was a Godly punishment [72]. Collectively, the apparent trend we observed in terms of the studies in this review was the generalized stigmatization of the PLWHA as well as in some cases the HCWs dealing with them. ## Fertility care/ sexual function and psychosexual counselling Some end-users displayed limited/inadequate knowledge about the concept, availability and benefits of SRH, voiced by the need for more information and quality services [73]. Healthcare workers sometimes exhibited deficient knowledge with regards to male SRH services [73–76]. This led to different personal attitudes towards the problem that was affected by age, sex and level of education [77] (Table 4). As for sexual health, although the knowledge level was acceptable, however, the sociopolitical norms affected the proper attitude towards the topic. This goes for general population as well as HCP (Table 4). Table 4Fertility care, sexual function and psychosexual counselling in MENA: characteristics of users and health care professionalsCharacteristicClients/usersHealth Care ProfessionalsKnowledge▸ RH: inadequate knowledge about the concept, availability and benefits [73]. ▸ Premarital checkup:• Egypt: lack of knowledge among general population even among educated respondents [78].• KSA: university students aware of its importance in preventing transmission of hereditary diseases to offsprings and ensuring their partner’s health [79]. ▸ Sexual Health:• Younger boys: more aware of physiological/emotional puberty changes of their own sex; but not of opposite sex [28].• Adults: dialogue between patients and their treating physicians regarding ED assists patients to seek proper/safe medical advice [80]. ▸ Healthcare workers sometimes displayed low knowledge e.g., about ICSI [73, 75, 76]. ▸ Clinical practitioners: more likely to have accurate knowledge of FP options than oncologists [74]. ▸ Sexual Health:• Urologists: more knowledgeable about ED, but gynecologists had better attitude towards ED patients [81].• Nurses: most were not very knowledgeable about/confident to address sexuality, viewing it as not within their responsibilities [82, 83].• HCP: lacked confidence in their sex education skills and knowledge [84].Attitude▸ RH services: users not always satisfied with HCP attitudes, stating it was unpleasant, with poor communication and inappropriate management approach [73]. ▸ Premarital checkup:• Egypt: among general population, most respondents, except unmarried males, had favorable attitude [78].• KSA: most university students had generally positive attitude and good intended practices toward PMS. Most participants demanded implementing a law that prohibits incompatible marriages [79]. ▸ Sexual Health:• Many adolescent boys found female genital cutting necessary, favoured polygamous marriage at younger age, but not consanguineous marriages [28].• ED: sensitive issue among older clients, hence in rarely consulting. Conversely, university students were more liberal toward sex, had more risky behaviours [80, 85].• Gay communities: highly knowledgeable but had high-risk behavior (low condom use/ HIV testing), most disclosed their sexual orientation only to their partners and not to their HCP even if needed [86]. ▸ Physicians and HCP with previous SRH had better youth-friendly attitudes [87]. ▸ Family physicians: favorable attitudes toward infertility management, but attitude varied with age, gender and experience [77]. ▸ Oncologists: low perception of importance of FP, leading to poor referral to specialists; gender bias in informing males about FP options prior to cancer treatment compared to females [74, 75]. ▸ Sexual Health:• Nurses: negative attitude influenced by their beliefs about sex/sex education (viewing early sex education as problematic), negative attitude was associated with sex (female) and no previous training on sexuality [82–84].• Lebanon: HCP had positive attitudes towards LGBT patients; mental health providers less likely to believe that homosexuality is mental health disorder, but more likely a natural variation on the sexual orientation spectrum [88].HCP Health care professionals, KSA Kingdom of Saudi Arabia, PMS Pre-marital screening, ED Erectile dysfunction, LGBT Lesbian, gay, bisexual, and transgender, FP Fertility preservation, SRH Sexual and reproductive health, RH Reproductive health, ICSI Intracytoplasmic sperm injection Healthcare system-related factors suggested that improvements in quality of infertility management required evidence-based training, supplies, laboratory/radiology support, improved communications with specialists, and availability of guidelines [77]. Youth also felt that SRH services needed to be easily accessible and have equal geographical distribution [73]. Reproductive tourism attracted patients from countries with deficient invitro fertilization (IVF) services or policy restrictions, and required high-tech medical settings, with visa regulations allowing users to complete an entire IVF cycle [89]. ## Contraception counselling and provision The knowledge and acceptance of family planning varied across MENA. Generally, good awareness of contraceptive approaches was mainly for women’s methods but not for male contraception. However, such awareness was not translated into increased application/use of family planning in many MENA countries due to religious and sociocultural norms surrounding this topic. The lack of knowledge about male contraception was also including HCP e.g., pharmacists (Table 5). Table 5Contraception counselling and provision in MENA: characteristics of users and health care professionalsCharacteristicClients/usersHealth Care ProfessionalsKnowledge• UAE: most men aware of availability of male contraceptive methods, only few currently used them, mainly condoms, and only $1.1\%$ were sterilized. Few thought that contraceptive pills/ monthly injection for men are available [90].• Jordan: most men heard about family planning, most commonly intrauterine device and oral contraceptives [91].• Iraq: decreased knowledge regarding correct condom use and its effectiveness for contraception/ STIs prevention [92].• Egypt: most secondary-school pupils knew about contraception, girls had more information [93].Pharmacists: decreased knowledge about male OCPs and their mechanism of action, with negative perceptions towards them. Barriers to male OCPs were cultural norms, side effects, and poor compliance [94].Attitude• UAE: majority of men rejected male contraception, due to religious reasons, followed by cultural barriers, personal beliefs, medical disorders and economic factors. Male contraception use significantly associated with high education level of partners, low family size and family income [90].• Jordan: married men had good attitudes/knowledge about family planning, but only $45.1\%$ currently used contraception. However, most men agreed about a minimum 2 years’ child spacing and starting contraception after childbirth and that husband and wife should share decisions about family planning and number of children [91].• Sudan: three-fifths of men with reproductive age wives wished to use family planning services but only about one-fifth currently used an effective method. Men were more interested in learning more about female than male sterilization [95].• Iraq: condoms were rarely used for family planning due to lack of need, fertility-related reasons or use of female contraception methods [92].• Egypt: secondary-school pupils agreed about using contraceptive methods in the future [93].HCP Health care professionals, UAE United Arab Emirates, STIs Sexually transmitted infections, OCPs Oral contraceptive pills ## Sex-based violence Little information exists on gender-based violence in MENA, and our search yielded only 1 article. In Egypt, the majority of street children experienced more than one risk including harassment or abuse by police and other street children, drug abuse, and, among sexually active 15–17-year-olds, most reported multiple partners and never using condoms, and most girls had experienced sexual abuse [96]. Such behaviors put them in substantial overlap with populations at highest risk for HIV, namely men who have sex with men, commercial sex workers, and injection drug users [96]. ## Discussion Addressing SRH of men alongside that of women’s is essential. However, it has not received the attention it deserves worldwide. We outlined the current knowledge, knowledge gaps and considerations that impact on men’s SRH in MENA, and appraised the HCPs’, users’, healthcare systems’ and social factors affecting such services. To our knowledge, this is the first comprehensive scoping review of men’s SRH in MENA. Our main findings unearthed a strong HIV/AIDS focus of the published outputs, but a much weaker focus on issues related to fertility care, sexual dysfunctions/ counselling, and gender-based violence. *The* general population, different clientele groups, and a range of HCPs exhibited many SRH knowledge gaps, that subsequently lead to a high prevalence of unfavorable attitudes towards men’s SRH conditions, stigmatization, and the emergence of many misconceptions. Generally, across the range of countries under examination, the quality of SRH services could be improved. Surprisingly, we could not find published data on legislation, government policies or national SRH strategies. More importantly, we observed several paradoxes in terms of the lack of congruence between many of the domains that the published outputs addressed on the one hand; and the actual ‘on the ground’ situation across MENA on the other. The first paradox pertained to the strong HIV/AIDS focus across the published literature, despite the low prevalence/ burden in MENA region ($0.1\%$) [97, 98]. While it is difficult to speculate the reasons behind such discrepancy, perhaps it might be explained by the wide international interest and availability of funding to explore epidemiological and behavioural HIV/AIDS research, as evidenced by that most studies were funded by multi-lateral bodies e.g., UNICEF or philanthropic agencies [56, 67, 99]. Notwithstanding, war and political instabilities may increase the vulnerability of the region to HIV by reducing access to prevention services, destroying health care infrastructure, disrupting social support networks, increasing exposure to sexual violence, and expanding immigration and displacement [100]. Despite the tremendous efforts made in the global cognition and epidemiology of HIV infection, knowledge in MENA remains limited and controversial [99]. The large number of published papers on HIV/AIDS we observed concurs with the findings of a scoping review of men’s SRH in Nordic countries, where out of 68 studies that were identified, 15 papers dealt with STIs, mainly HIV (12 papers) and MSM (9 papers) [101]. The second paradox was the weak focus of the published articles on the two topics of fertility care and sexual dysfunctions/counselling, despite the high prevalence of fertility and sexual problems in MENA ($22.6\%$) [102]. Such misfit might by due to the complex cultural, religious, community gender and social norms prevalent among MENA populations that render them reluctant to disclose their SRH concerns [80]. Likewise, MENA has low knowledge of sexual relationships, attributed to a lack of sex education in schools and the conservative culture of the community, factors that might contribute to the increasing prevalence of e.g., premature ejaculation in the region [103]. Similarly, erectile dysfunction (ED) is quite prevalent among Arab men, probably explained by the high prevalence of endothelial dysfunction risk [104]. Again, our findings support a review that scoped men’s SRH across Scandinavia, where sexual functioning/ counselling studies covered a very small proportion ($\frac{2}{68}$) of the studies that the review identified [101]. The third paradox we observed was related to gender-based violence. The current review found that publications of gender-based violence prevention, support and care represented only $2\%$ of the retrieved studies. This is despite that women’s exposure to male domination has long been normalized in the Arab world [105, 106]. This could be due to possible gender disparity and male predominance in MENA representing barriers to such research [107]. Elsewhere, female researchers lag behind their male counterparts in successfully receiving grants [108]. Even though the prevalence of intimate partner violence (IPV) is high across Arab countries, evidence on its correlates remains limited [109]. The situation is complicated by the fact that the West’s social acceptance of divorce is not shared by Arab nations, where the centrality of marriage and family culture persists, and divorce continues to be stigmatized [109]. A scoping review of men’s SRH in the Nordic countries found that published studies about sexual violence comprised a very small minority ($\frac{2}{68}$) of the studies [101], concurring with our findings. Hence, efforts to mitigate gender gap and promote equity, diversity, and inclusion of females in research may improve any gender-based parity in research topics. The fourth paradox was that the present review found no studies pertaining to men’s SRH that addressed the domain of antenatal, intrapartum and postnatal care ($0\%$), despite that men’s involvement in maternal health programs is a key to increase utilization of maternal health services [110]. This could be due to prohibition of gender mixing in antenatal, intrapartum and postnatal care as well as in women’s hospital settings, hence disallowing male presence in such encounters. Such lack supports that in spite of the growing recognition of father’s importance for early family health/well-being, there has been very limited attention to men’s own experiences and developmental needs during their partner’s antenatal visits [111, 112]. Nevertheless, our observations contrasts with the Nordic study, where more than one-third of the papers were related to experiences of expectant fathers during antenatal, intrapartum and postnatal care [101]. Empowering men with antenatal care knowledge and joint decision-making with their spouses increases male involvement [113], particularly that complex community sociocultural norms and social stigma are barriers to men’s attendance at antenatal care services with their partners [114]. The present maternity health policies in Arab countries might need revision to allow fathers’ inclusion [115]; and our findings suggest a need for communication, education, and information-based health promotion programs that empower men in these domains. The fifth paradox was that a great proportion of studies discussed knowledge levels and gaps pertaining to men’s SRH in MENA, identifying a range of factors that influence knowledge. Surprisingly, there was no parallel body of literature debating effective interventions and their implementation in order to remedy/overcome such gaps. Similarly, we found no articles dedicated to policies and strategies to address such shortcomings at state, health care system and public health policy making level, when certainly improving men’s access to SRH requires state, health care system and health care providers interventions and policies [4, 101]. Likewise, the present review found no studies of men’s SRH that addressed safe abortion care ($0\%$), not surprising given that abortion is illegal across most of the region, and there are various legal and societal barriers to the practice in the countries where it is allowed [116]. Very few publications exist on abortion in MENA, and those that do exist tend to give an overview of the legislation of different states or evaluate Islam’s position on abortion [117–121]. Detailed studies on actual medical practices, political debates, local legal implementation, moral/social norms, and trajectories of MENA women are very rare [122–126], let alone those on men’s participation in safe abortion care. A unique characteristic across MENA is the prevalent political instabilities and refugee situations [44]. MENA has a current 16 million forcibly displaced and stateless people [127], situations that increase risky behavior for HIV and STIs [44]. Our findings resonate with Uganda, where refugee adolescents and displaced youth were a key population left behind in HIV prevention efforts [128]. Young refugees have limited access to sexual health information and resources in their resettlement places, highlighting a need for sexual health education programs for men and women as part of resettlement services [129]. Notwithstanding, Egypt has taken welcome steps: policies in progress include commitment to give refugees access to primary health care and education within national systems, and refugees are currently covered by the universal health care insurance scheme on equal footing with Egyptians [130]. The current scoping review has limitations. Surveys and reports about MENA health issues are mainly published in local languages and hardly accessible through electronic databases [102]. The study has many strengths. To our knowledge, this study is the first scoping review focusing on men’s experiences in SRH care across MENA. There was no time limitation for our literature search. In line with others [14, 15], the review was driven by a strict peer reviewed protocol. For an appropriate search, we examined the search strategy used in a similar published article on men’s SRH in Nordic countries [101] and modified the search terms they used. We searched four electronic bibliographic databases and reference lists of articles. Despite that our search was conducted using only English terms, our review included any articles published in English or Arabic; given that *Arabic is* the major language in MENA. The screening and data characterization forms that were employed were pretested by all members of the reviewer team and modified as appropriate before the review. Three training sessions included the completion of the screening and data characterization forms, using articles that were randomly selected from the literature. Data extraction of each article was undertaken by two independent members of the review team (WEA, MA) who worked simultaneously together on each article, and any disagreements were resolved by discussion. The review team systematically synthesized the findings by extracting the aims, populations, findings and conclusions of each included study and categorizing them based on the eight WHO SRH domains. ## Conclusion It is important for men to access SRH care. The available literature from the MENA region suggests that men’s SRH is not sufficiently prioritized. A detailed landscape of men’s experiences in SRH care across these countries remains to be explored. A number of pertinent ‘mismatches’ were evident in the literature. There was strong focus on HIV/AIDS, when MENA has the lowest prevalence of HIV in the world; a much weaker focus on fertility and sexual dysfunctions, despite that these conditions were much more prevalent in MENA countries; no publications on men’s involvement in sexual gender-based violence, despite women’s exposure to its various forms across the MENA region; and no studies pertaining to men’s SRH in terms of their involvement in antenatal, intrapartum and postnatal care, despite the prevailing international literature to the value of such involvement. Such incongruencies might serve to provide future direction for the formulation and implementation of comprehensive strategies to help tackle men’s SRH challenges in MENA region. These could include strengthening the current policies, strategies and interventions to enhance and improve the attitudes, behaviors and education of the general public, the youth, men in general as well as HCPs. Furthermore, strengthening the health systems over time in terms of political commitment, structures, organization, funding and interventions to address core issues and to more formally respond to men’s SRH challenges in MENA would be welcomed. Future research should examine the influence of policies and the healthcare service delivery and organization on men’s access and experiences in SRH care. ## Supplementary Information Additional file 1: Supplementary Box 1. Search terms used in the current scoping review. Supplementary Table 1. Included articles on men's experiences of sexual and reproductive healthcare in MENA countries [131–154]. ## References 1. Grandahl M, Bodin M, Stern J. **In everybody’s interest but no one’s assigned responsibility: midwives’ thoughts and experiences of preventive work for men’s sexual and reproductive health and rights within primary care**. *BMC Public Health* (2019.0) **19** 1423. DOI: 10.1186/s12889-019-7792-z 2. 2.Report of the International Conference on Population and Development (Cairo, 5–13. September 1994). New York (NY): United Nations; 1994 available at https://www.un.org/en/development/desa/population/migration/generalassembly/docs/globalcompact/A_CONF.171_13.pdf. Accessed 1 July 2022. 3. Kalmuss D, Austrian K. **Real men do... real men don’t: Young Latino and African American men’s discourses regarding sexual health care utilization**. *Am J Mens Health* (2010.0) **4** 218-30. DOI: 10.1177/1557988309331797 4. 4.Makarow M, Hojgaard L. Male reproductive health: Its impacts in relation to general well-being and low European fertility rates. 2010. Retrieved from http://www.esf.org/fileadmin/Public_documents/Publications/SPB40_MaleReproductiveHealth.pdf. 5. Burns JC, Reeves J, Calvert WJ, Adams M, Ozuna-Harrison R, Smith MJ, Baranwal S, Johnson K, Rodgers CRR, Watkins DC. **Engaging young black males in sexual and Reproductive Health Care: a review of the literature**. *Am J Mens Health* (2021.0) **15** 15579883211062024. DOI: 10.1177/15579883211062024 6. 6.WHO Europe. Strategy on the health and well-being of men in the WHO European region. Copenhagen: WHO Regional Office for Europe, Regional Committee for Europe; 2018. 7. DeJong J, Zurayk H, Myntti C, Tekçe B, Giacaman R, Bashour H, Ghérissi A, Gaballah N. **Health research in a turbulent region: the Reproductive Health Working Group**. *Reprod Health Matters* (2017.0) **25** 4-15. DOI: 10.1080/09688080.2017.1379864 8. Ghannam F. *Live and die like a man: gender dynamics in urban Egypt* (2013.0) 9. Abu-Rmeileh NME, Ghandour R, Tucktuck M, Obiedallah M. **Research priority-setting: reproductive health in the occupied palestinian territory**. *Reprod Health* (2018.0) **15** 27. DOI: 10.1186/s12978-018-0472-0 10. Shalash A, Alsalman HM, Hamed A, Abu Helo M, Ghandour R, Albarqouni L, Abu Rmeileh NM. **The range and nature of reproductive health research in the occupied palestinian territory: a scoping review**. *Reprod Health* (2019.0) **16** 41. DOI: 10.1186/s12978-019-0699-4 11. Kabakian-Khasholian T, Quezada-Yamamoto H, Ali A, Sahbani S, Afifi M, Rawaf S, El Rabbat M. **Integration of sexual and reproductive health services in the provision of primary health care in the Arab States: status and a way forward**. *Sex Reprod Health Matters* (2020.0) **28** 1773693. DOI: 10.1080/26410397.2020.1773693 12. 12.Inter-agency Working Group on Reproductive Health in Crises. Inter-agency field manual on reproductive health in humanitarian settings: 2010 revision for field review. Inter-agency Working Group on Reproductive Health in Crises; 2018. Available at https://iawg.wpengine.com/wp-content/uploads/2019/07/IAFM-English.pdf. Accessed 29 June 2022. 13. El Ansari W, Elhag W. **Weight regain and Insufficient Weight loss after bariatric surgery: definitions, prevalence, Mechanisms, Predictors, Prevention and Management Strategies, and knowledge Gaps-a scoping review**. *Obes Surg* (2021.0) **31** 1755-66. DOI: 10.1007/s11695-020-05160-5 14. El Ansari W, El-Ansari K. **Missing something? A scoping review of venous thromboembolic events and their associations with bariatric surgery. Refining the evidence base**. *Ann Med Surg (Lond)* (2020.0) **59** 264-73. DOI: 10.1016/j.amsu.2020.08.014 15. El Ansari W, El-Ansari K. **Missing something? Comparisons of effectiveness and outcomes of bariatric surgery procedures and their Preferred Reporting: Refining the evidence base**. *Obes Surg* (2020.0) **30** 3167-77. DOI: 10.1007/s11695-020-04659-1 16. Daudt HM, van Mossel C, Scott SJ. **Enhancing the scoping study methodology: a large, inter-professional team’s experience with Arksey and O’Malley’s framework**. *BMC Med Res Methodol* (2013.0) **13** 48. DOI: 10.1186/1471-2288-13-48 17. Tricco AC, Lillie E, Zarin W, O’Brien KK, Colquhoun H, Levac D, Moher D, Peters MD, Horsley T, Weeks L, Hempel S. **PRISMA extension for scoping reviews (PRISMA-ScR): checklist and explanation**. *Ann Intern Med* (2018.0) **169** 467-73. DOI: 10.7326/M18-0850 18. Arksey H, O’Malley L. **Scoping studies: towards a methodological framework**. *Int J Soc Res Methodol* (2005.0) **8** 19-32. DOI: 10.1080/1364557032000119616 19. 19.WHO (World Health Organization). 2017 Sexual health and its linkages to reproductive health: an operational approach. WHO: Geneva, Switzerland. Available at http://apps.who.int/iris/bitstream/handle/10665/258738/9789241512886-eng.pdf;jsessionid=A20D089FFCBCA9F97264F98E57881DA1?sequence=1. Accessed 2 Mar 2022. 20. Davis K, Drey N, Gould D. **What are scoping studies? A review of the nursing literature**. *Int J Nurs Stud* (2009.0) **46** 1386-400. DOI: 10.1016/j.ijnurstu.2009.02.010 21. Pham MT, Rajić A, Greig JD, Sargeant JM, Papadopoulos A, McEwen SA. **A scoping review of scoping reviews: advancing the approach and enhancing the consistency**. *Res Synth Methods* (2014.0) **5** 371-85. DOI: 10.1002/jrsm.1123 22. Kilbourne AM, Switzer G, Hyman K, Megan Crowley-Matoka M, Fine MJ. **Advancing Health Disparities Research within the Health Care System: a conceptual Framework**. *Am J Public Health* (2006.0) **96** 2113-21. DOI: 10.2105/AJPH.2005.077628 23. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD. **The PRISMA 2020 statement: an updated guideline for reporting systematic reviews**. *BMJ* (2021.0) **372** n71. DOI: 10.1136/bmj.n71 24. Abolfotouh MA. **The impact of a lecture on AIDS on knowledge, attitudes and beliefs of male school-age adolescents in the Asir Region of southwestern Saudi Arabia**. *J Community Health* (1995.0) **20** 271-81. DOI: 10.1007/BF02260410 25. Alhasawi A, Grover SB, Sadek A, Ashoor I, Alkhabbaz I, Almasri S. **Assessing HIV/AIDS knowledge, awareness, and attitudes among senior High School students in Kuwait**. *Med Princ Pract* (2019.0) **28** 470-6. DOI: 10.1159/000500307 26. Al-Iryani B, Raja’a YA, Kok G, Van Den Borne B. **HIV knowledge and stigmatization among adolescents in yemeni schools**. *Int Q Community Health Educ 2009* (2010.0) **30** 311-20. DOI: 10.2190/IQ.30.4.c 27. Boneberger A, Rückinger S, Guthold R, Kann L, Riley L. **HIV/AIDS related knowledge among school-going adolescents from the Middle East and North Africa**. *Sex Health* (2012.0) **9** 196-8. DOI: 10.1071/SH11054 28. Jaffer YA, Afifi M, Al Ajmi F, Alouhaishi K. **Knowledge, attitudes and practices of secondary-school pupils in Oman: II. Reproductive health**. *EMHJ* (2006.0) **12** 50-60. PMID: 17037221 29. Al-Jabri AA, Al-Abri JH. **Knowledge and attitudes of undergraduate medical and non-medical students in Sultan Qaboos University toward acquired immune deficiency syndrome**. *Saudi Med J* (2003.0) **24** 273-7. PMID: 12704503 30. Al-Rabeei NA, Dallak AM, Al-Awadi FG. **Knowledge, attitude and beliefs towards HIV/AIDS among students of health institutes in Sana’a city**. *EMHJ* (2012.0) **18** 221-6. PMID: 22574474 31. Al-Mazrou YY, Abouzeid MS, Al-Jeffri MH. **Knowledge and attitudes of paramedical students in Saudi Arabia toward HIV/AIDS**. *Saudi Med J* (2005.0) **26** 1183-9. PMID: 16127510 32. Nasir EF, Astrøm AN, David J, Ali RW. **HIV and AIDS related knowledge, sources of information, and reported need for further education among dental students in Sudan–a cross sectional study**. *BMC Public Health* (2008.0) **8** 86. DOI: 10.1186/1471-2458-8-286 33. Premadasa G, Sadek M, Ellepola A, Sreedharan J, Muttappallymyalil J. **Knowledge of and attitudes towards HIV/AIDS: a survey among dental students in Ajman, UAE**. *J Investig Clin Dent* (2015.0) **6** 147-55. DOI: 10.1111/jicd.12080 34. Farsi NJ, Baharoon AH, Jiffri AE, Marzouki HZ, Merdad MA, Merdad LA. **Human papillomavirus knowledge and vaccine acceptability among male medical students in Saudi Arabia**. *Hum Vaccin Immunother* (2021.0) **17** 1968-74. DOI: 10.1080/21645515.2020.1856597 35. Farsi NJ, Al Sharif S, Al Qathmi M, Merdad M, Marzouki H, Merdad L. **Knowledge of human papillomavirus (HPV) and Oropharyngeal Cancer and Acceptability of the HPV Vaccine among Dental Students**. *Asian Pac J Cancer Prev* (2020.0) **21** 3595-603. DOI: 10.31557/APJCP.2020.21.12.3595 36. Alkhasawneh E, McFarland W, Mandel J, Seshan V. **Insight into jordanian thinking about HIV: knowledge of jordanian men and women about HIV prevention**. *J Assoc Nurses AIDS Care* (2014.0) **25** e1-9. DOI: 10.1016/j.jana.2013.06.001 37. Al-Ghanim SA. **Exploring public knowledge and attitudes towards HIV/AIDS in Saudi Arabia. A survey of primary health care users**. *Saudi Med J* (2005.0) **26** 812-8. PMID: 15951875 38. al-Owaish R, Moussa MA, Anwar S, al-Shoumer H, Sharma P. **Knowledge, attitudes, beliefs, and practices about HIV/AIDS in Kuwait**. *AIDS Educ Prev* (1999.0) **11** 163-73. PMID: 10214499 39. Al-Serouri AW, Takioldin M, Oshish H, Aldobaibi A, Abdelmajed A. **Knowledge, attitudes and beliefs about HIV/AIDS in Sana’a**. *Yemen EMHJ* (2002.0) **8** 706-15. PMID: 15568447 40. Alshehri AM. **The Awareness of the Human Papillomavirus Infection and Oropharyngeal Cancer in People to Improve the Health Care System at Al Qunfudhah Region, Kingdom of Saudi Arabia**. *J Healthc Eng* (2021.0) **2021** 5185075. DOI: 10.1155/2021/5185075 41. Kulane A, Owuor JOA, Sematimba D, Abdulahi SA, Yusuf HM, Mohamed LM. **Access to HIV Care and Resilience in a long-term conflict setting: a qualitative Assessment of the Experiences of living with diagnosed HIV in Mogadishu**. *Somali IJERPH* (2017.0) **14** 721. DOI: 10.3390/ijerph14070721 42. Salama II, Kotb NK, Hemeda SA, Zaki F. **HIV/AIDS knowledge and attitudes among alcohol and drug abusers in Egypt**. *J Egypt Public Health Assoc* (1998.0) **73** 479-500. PMID: 17217020 43. Laraqui S, Laraqui O, Manar N, Ghailan T, Belabsir M, Deschamps F, Laraqui CH. **The assessment of seafarers’ knowledge, attitudes and practices related to STI/HIV/AIDS in northern Morocco**. *Int Marit Health* (2017.0) **68** 26-30. DOI: 10.5603/IMH.2017.0005 44. Holt BY, Effler P, Brady W, Friday J, Belay E, Parker K, Toole M. **Planning STI/HIV prevention among refugees and mobile populations: situation assessment of sudanese refugees**. *Disasters* (2003.0) **27** 1-15. DOI: 10.1111/1467-7717.00216 45. Nasir EF, Marthinussen MC, Åstrøm AN. **HIV/AIDS-related attitudes and oral impacts on daily performances: a cross-sectional study of sudanese adult dental patients**. *BMC Health Serv Res* (2013.0) **13** 335. DOI: 10.1186/1472-6963-13-335 46. Mahfouz AA, Alakija W, al-Khozayem AA, al-Erian RA. **Knowledge and attitudes towards AIDS among primary health care physicians in the Asir Region, Saudi Arabia**. *J R Soc Health* (1995.0) **115** 23-5. DOI: 10.1177/146642409511500108 47. Memish ZA, Filemban SM, Bamgboyel A, Al Hakeem RF, Elrashied SM, Al-Tawfiq JA. **Knowledge and attitudes of doctors toward people living with HIV/AIDS in Saudi Arabia**. *J Acquir Immune Defic Syndr* (2015.0) **69** 61-7. DOI: 10.1097/QAI.0000000000000550 48. Arheiam A, El Tantawi M, Al-Ansari A, Ingafou M, El Howati A, Gaballah K, AbdelAziz W. **Arab dentists’ refusal to treat HIV positive patients: a survey of recently graduated dentists from three Arab dental schools**. *Acta Odontol Scand* (2017.0) **75** 355-360. DOI: 10.1080/00016357.2017.1316867 49. Hassan ZM, Wahsheh MA. **Knowledge and attitudes of jordanian nurses towards patients with HIV/AIDS: findings from a nationwide survey**. *Issues Ment Health Nurs* (2011.0) **32** 774-84. DOI: 10.3109/01612840.2011.610562 50. Gańczak M, Barss P, Alfaresi F, Almazrouei S, Muraddad A, Al-Maskari F. **Break the silence: HIV/AIDS knowledge, attitudes, and educational needs among arab university students in United Arab Emirates**. *J Adolesc Health* (2007.0) **40** 572.e1-8. DOI: 10.1016/j.jadohealth.2007.01.011 51. Shama M, Fiala LE, Abbas MA. **HIV/AIDS perceptions and risky behaviors in squatter areas in Cairo, Egypt**. *J Egypt Public Health Assoc* (2002.0) **77** 173-200. PMID: 17219897 52. Alwafi HA, Meer AMT, Shabkah A, Mehdawi FS, El-Haddad H, Bahabri N, Almoallim H. **Knowledge and attitudes toward HIV/AIDS among the general population of Jeddah, Saudi Arabia**. *J Infect Public Health* (2018.0) **11** 80-4. DOI: 10.1016/j.jiph.2017.04.005 53. Al Tall YR, Mukattash TL, Sheikha H, Jarab AS, Nusair MB, Abu-Farha RK. **An assessment of HIV patient’s adherence to treatment and need for pharmaceutical care in Jordan**. *Int J Clin Pract* (2020.0) **74** e13509. DOI: 10.1111/ijcp.13509 54. Valadez JJ, Berendes S, Jeffery C, Thomson J, Ben Othman H, Danon L, Turki AA, Saffialden R, Mirzoyan L. **Filling the knowledge gap: measuring HIV Prevalence and Risk factors among men who have sex with men and female sex workers in Tripoli, Libya**. *PLoS ONE* (2013.0) **8** e66701. DOI: 10.1371/journal.pone.0066701 55. Kabbash IA, Abo Ali EA, Elgendy MM, Abdrabo MM, Salem HM, Gouda MR, Elbasiony YS, Elboshy N, Hamed M. **HIV/AIDS-related stigma and discrimination among health care workers at Tanta University Hospitals, Egypt**. *Environ Sci Pollut Res Int* (2018.0) **25** 30755-62. DOI: 10.1007/s11356-016-7848-x 56. Abdelrahman I, Lohiniva AL, Kandeel A, Benkirane M, Atta H, Saleh H, El Sayed N, Talaat M. **Learning about barriers to care for people living with HIV in Egypt: a qualitative exploratory study**. *J Int Assoc Provid AIDS Care* (2015.0) **14** 141-7. DOI: 10.1177/2325957413488180 57. El-Sayyed N, Kabbash IA, El-Gueniedy M. **Knowledge, attitude and practices of egyptian industrial and tourist workers towards HIV/AIDS**. *EMHJ* (2008.0) **14** 1126-35. PMID: 19161085 58. Bashir F, Ba Wazir M, Schumann B, Lindvall K. **The realities of HIV prevention. A closer look at facilitators and challenges faced by HIV prevention programmes in Sudan and Yemen**. *Glob Health Action* (2019.0) **12** 1659098. DOI: 10.1080/16549716.2019.1659098 59. Elgalib A, Shah S, Al-Habsi Z, Al-Fouri M, Lau R, Al-Kindi H, Al-Rawahi B, Al-Abri S. **The cascade of HIV care in Oman, 2015–2018: a population-based study from the Middle East**. *Int J Infect Dis* (2020.0) **90** 28-34. DOI: 10.1016/j.ijid.2019.09.017 60. Barry M, Ghonem L, Albeeshi N, Alrabiah M, Alsharidi A, Al-Omar HA. **Resource utilization and caring cost of people living with human immunodeficiency virus (PLHIV) in Saudi Arabia: a Tertiary Care University Hospital Experience**. *Healthc (Basel)* (2022.0) **10** 118. DOI: 10.3390/healthcare10010118 61. Bein M, Coker-Farrell EY. **The association between medical spending and health status: a study of selected african countries**. *Malawi Med J* (2020.0) **32** 37-44. DOI: 10.4314/mmj.v32i1.8 62. Khamis J, Ghaddar A. **HIV/AIDS in Syria and the response of the National AIDS Program during the war**. *Sex Transm Infect* (2018.0) **94** 173. DOI: 10.1136/sextrans-2017-053438 63. Marih L, Sawras V, Pavie J, Sodqi M, Malmoussi M, Tassi N, Bensghir R, Nani S, Lahsen AO, Laureillard D, El Filali KM, Champenois K, Weiss L. **Missed opportunities for HIV testing in patients newly diagnosed with HIV in Morocco**. *BMC Infect Dis* (2021.0) **21** 48. DOI: 10.1186/s12879-020-05711-2 64. Elgalib A, Al-Sawafi H, Kamble B, Al-Harthy S, Al-Sariri Q. **Multidisciplinary care model for HIV improves treatment outcome: a single-centre experience from the Middle East**. *AIDS Care* (2018.0) **30** 1114-9. DOI: 10.1080/09540121.2018.1479028 65. Al-Iryani B, Basaleem H, Al-Sakkaf K, Crutzen R, Kok G, van den Borne B. **Evaluation of a school-based HIV prevention intervention among yemeni adolescents**. *BMC Public Health* (2011.0) **11** 279. DOI: 10.1186/1471-2458-11-279 66. Saleh MA, Al-Ghamdi YS, Al-Yahia OA, Shaqran TM, Mosa AR. **Impact of health education program on knowledge about AIDS and hiv transmission in students of secondary schools in buraidah city, saudi arabia: an exploratory study**. *J Family Community Med* (1999.0) **6** 15-21. PMID: 23008592 67. Al-Iryani B, Basaleem H, Al-Sakkaf K, Kok G, van den Borne B. **Process evaluation of school-based peer education for HIV prevention among yemeni adolescents**. *SAHARA J* (2013.0) **10** 55-64. PMID: 23777570 68. Himmich H, Ouarsas L, Hajouji FZ, Lions C, Roux P, Carrieri P. **Scaling up combined community-based HIV prevention interventions targeting truck drivers in Morocco: effectiveness on HIV testing and counseling**. *BMC Infect Dis* (2015.0) **15** 208. DOI: 10.1186/s12879-015-0936-6 69. Maatouk I, Nakib ME, Assi M, Farah P, Makso B, Nakib CE, Rady A. **Community-led HIV self-testing for men who have sex with men in Lebanon: lessons learned and impact of COVID-19**. *Health Res Policy Syst* (2021.0) **19** 50. DOI: 10.1186/s12961-021-00709-x 70. Lohiniva AL, Kamal W, Benkirane M, Numair T, Abdelrahman M, Saleh H, Zahran A, Talaat M, Kandeel A. **HIV Stigma toward People living with HIV and Health Providers Associated with their care: qualitative interviews with Community Members in Egypt**. *J Assoc Nurses AIDS Care* (2016.0) **27** 188-98. DOI: 10.1016/j.jana.2015.11.007 71. Aunon FM, Wagner GJ, Maher R, Khouri D, Kaplan RL, Mokhbat J. **An exploratory study of HIV Risk Behaviors and Testing among male sex workers in Beirut, Lebanon**. *Soc Work Public Health* (2015.0) **30** 373-84. DOI: 10.1080/19371918.2014.979274 72. Badahdah AM. **Stigmatization of persons with HIV/AIDS in Saudi Arabia**. *J Transcult Nurs* (2010.0) **21** 386-92. DOI: 10.1177/1043659609360873 73. Khalaf I, Abu Moghli F, Froelicher ES. **Youth-friendly reproductive health services in Jordan from the perspective of the youth: a descriptive qualitative study**. *Scand J Caring Sci* (2010.0) **24** 321-31. DOI: 10.1111/j.1471-6712.2009.00723.x 74. Ghazeeri G, Zebian D, Nassar AH, Harajly S, Abdallah A, Hakimian S, Skaiff B, Abbas HA, Awwad J. **Knowledge, attitudes and awareness regarding fertility preservation among oncologists and clinical practitioners in Lebanon**. *Hum Fertil (Camb)* (2016.0) **19** 127-33. DOI: 10.1080/14647273.2016.1193636 75. Arafa MA, Rabah DM. **Attitudes and practices of oncologists toward fertility preservation**. *J Pediatr Hematol Oncol* (2011.0) **33** 203-7. DOI: 10.1097/MPH.0b013e3182068047 76. Rabah DM, Wahdan IH, Merdawy A, Abourafe B, Arafa MA. **Oncologists’ knowledge and practice towards sperm cryopreservation in arabic communities**. *J Cancer Surviv* (2010.0) **4** 279-83. DOI: 10.1007/s11764-010-0140-z 77. Eldein HN. **Family physicians’ attitude and practice of infertility management at primary care–Suez Canal University, Egypt**. *Pan Afr Med J* (2013.0) **15** 106. PMID: 24244792 78. Eshra DK, Dorgham LS, el-Sherbini AF. **Knowledge and attitudes towards premarital counselling and examination**. *J Egypt Public Health Assoc* (1989.0) **64** 1-15. PMID: 2520141 79. Melaibari M, Shilbayeh S, Kabli A. **University students’ knowledge, attitudes, and practices towards the National Premarital Screening Program of Saudi Arabia**. *J Egypt Public Health Assoc* (2017.0) **92** 36-43. DOI: 10.21608/EPX.2018.6648 80. Abdulah Al Turki Y. **Should an inquiry about sexual health, as a reflection of vascular health, be part of routine physicals for young men? Results from an outpatient study**. *Int J Impot Res* (2009.0) **21** 362-5. DOI: 10.1038/ijir.2009.36 81. Abdulmohsen MF, Abdulrahman IS, Al-Khadra AH, Bahnassy AA, Taha SA, Kamal BA, Al-Rubaish AM, Ai-Elq AH. **Physicians’ knowledge, attitude and practice towards erectile dysfunction in Saudi Arabia**. *EMHJ* (2004.0) **10** 648-54. PMID: 16335658 82. Abu Ali RM, Abed MA, Khalil AA, Al-Kloub MI, Ashour AF, Alnsour IA. **A survey on sexual counseling for patients with Cardiac Disease among Nurses in Jordan**. *J Cardiovasc Nurs* (2018.0) **33** 467-73. DOI: 10.1097/JCN.0000000000000472 83. Bdair IA, Maribbay GL. **Practices, Attitudes and Beliefs of jordanian nurses toward sexual Health Assessment of Patients with Coronary Artery Diseases**. *Sex Disabil* (2020.0) **38** 491-502. DOI: 10.1007/s11195-020-09639-y 84. Farrag S, Hayter M. **A qualitative study of egyptian school nurses’ attitudes and experiences toward sex and relationship education**. *J Sch Nurs* (2014.0) **30** 49-56. DOI: 10.1177/1059840513506941 85. Salameh P, Zeenny R, Salamé J, Waked M, Barbour B, Zeidan N, Baldi I. **Attitudes towards and practice of sexuality among University students in Lebanon**. *J Biosoc Sci* (2016.0) **48** 233-48. DOI: 10.1017/S0021932015000139 86. Elmahy AG. **Reaching egyptian gays using social media: a Comprehensive Health Study and a Framework for Future Research**. *J Homosex* (2018.0) **65** 1867-76. DOI: 10.1080/00918369.2017.1395658 87. Gausman J, Othman A, Al-Qotob R, Shaheen A, Abu Sabbah E, Aldiqs M, Hamad I, Dabobe M, Langer A. **Health care professionals’ attitudes towards youth-friendly sexual and reproductive health services in Jordan: a cross-sectional study of physicians, midwives and nurses**. *Reprod Health* (2021.0) **18** 84. DOI: 10.1186/s12978-021-01137-4 88. Naal H, Abboud S, Harfoush O, Mahmoud H. **Examining the Attitudes and Behaviors of Health-care Providers toward LGBT Patients in Lebanon**. *J Homosex* (2020.0) **67** 1902-19. DOI: 10.1080/00918369.2019.1616431 89. Inhorn MC, Shrivastav P. **Globalization and reproductive tourism in the United Arab Emirates**. *Asia Pac J Public Health* (2010.0) **22** 68S-74S. DOI: 10.1177/1010539510373007 90. Ghazal-Aswad S, Zaib-Un-Nisa S, Rizk DE, Badrinath P, Shaheen H, Osman N. **A study on the knowledge and practice of contraception among men in the United Arab Emirates**. *J Fam Plann Reprod Health Care* (2002.0) **28** 196-200. DOI: 10.1783/147118902101196559 91. Hamdan-Mansour AM, Malkawi AO, Sato T, Hamaideh SH, Hanouneh SI. **Men’s perceptions of and participation in family planning in Aqaba and Ma’an governorates**. *Jordan EMHJ* (2016.0) **22** 124-32. PMID: 27180740 92. Ismael AS, Sabir Zangana JM. **Knowledge, attitudes and practice of condom use among males aged (15–49) years in Erbil Governorate**. *Glob J Health Sci* (2012.0) **4** 27-36. DOI: 10.5539/gjhs.v4n4p27 93. Kabbash IA, Hassan NM, Al-Nawawy AN, Attalla AA, Mekheimer SI. **Evaluation of HIV voluntary counselling and testing services in Egypt. Part 1: client satisfaction**. *EMHJ* (2010.0) **16** 481-90. PMID: 20799546 94. Barakat M, Thiab S, Thiab S, Al-Qudah RA, Akour A. **Knowledge and perception regarding the development and acceptability of male contraceptives among pharmacists: a mixed sequential method**. *Am J Mens Health* (2022.0) **16** 15579883221074855. DOI: 10.1177/15579883221074855 95. Mustafa MA, Mumford SD. **Male attitudes towards family planning in Khartoum, Sudan**. *J Biosoc Sci* (1984.0) **16** 437-49. DOI: 10.1017/S0021932000015273 96. Nada KH, Suliman el DA. **Violence, abuse, alcohol and drug use, and sexual behaviors in street children of Greater Cairo and Alexandria, Egypt**. *AIDS* (2010.0) **24 Suppl 2** S39-44. DOI: 10.1097/01.aids.0000386732.02425.d1 97. Setayesh H, Roudi-Fahimi F, Feki SE. *HIV in the Middle East: low prevalence but not low risk* (2013.0) 98. Gokengin D, Doroudi F, Tohme J. **HIV/AIDS: trends in the Middle East and North Africa region**. *Int J Infect Dis* (2016.0) **44** 66-73. DOI: 10.1016/j.ijid.2015.11.008 99. Al-Iryani B, Al-Sakkaf K, Basaleem H, Kok G, van den Borne B. **Process evaluation of a three-year community-based peer education intervention for HIV prevention among yemeni young people**. *Int Q Community Health Educ* (2010.0) **31** 133-54. DOI: 10.2190/IQ.31.2.c 100. Abu-Raddad LJ, Hilmi N, Mumtaz G. **Epidemiology of HIV infection in the Middle East and North Africa**. *AIDS* (2010.0) **24** 5-23. DOI: 10.1097/01.aids.0000386729.56683.33 101. Baroudi M, Stoor JP, Blåhed H, Edin K, Hurtig AK. **Men and sexual and reproductive healthcare in the nordic countries: a scoping review**. *BMJ Open* (2021.0) **11** e052600. DOI: 10.1136/bmjopen-2021-052600 102. Eldib A, Tashani OA. **Infertility in the middle East and North Africa Region: a systematic review with meta-analysis of prevalence surveys**. *Libyan J Med Sci* (2018.0) **2** 37-44. DOI: 10.4103/LJMS.LJMS_24_18 103. Albakr A, Arafa M, Elbardisi H, ElSaid S, Majzoub A. **Premature ejaculation: an investigative study into assumptions, facts and perceptions of patients from the Middle East (PEAP STUDY)**. *Arab J Urol* (2021.0) **19** 303-9. DOI: 10.1080/2090598X.2021.1948159 104. El-Sakka AI. **Erectile dysfunction in arab countries. Part I: prevalence and correlates**. *Arab J Urol* (2012.0) **10** 97-103. DOI: 10.1016/j.aju.2012.01.004 105. Ammar NH. **Beyond the shadows: domestic spousal violence in a “democratizing”**. *Egypt Trauma Violence & Abuse* (2006.0) **7** 244-59. DOI: 10.1177/1524838006292520 106. Yount KM, Li L. **Women’s “justification” of domestic violence in Egypt**. *J Marriage Fam* (2009.0) **71** 1125-40. DOI: 10.1111/j.1741-3737.2009.00659.x 107. Elghossain T, Bott S, Akik C. **Prevalence of intimate partner violence against women in the arab world: a systematic review**. *BMC Int Health Hum Rights* (2019.0) **19** 29. DOI: 10.1186/s12914-019-0215-5 108. Safdar B, Naveed S, Chaudhary AMD, Saboor S, Zeshan M, Khosa F. **Gender disparity in grants and awards at the National Institute of Health**. *Cureus* (2021.0) **13** e14644. PMID: 34046277 109. Abouelenin M. **Resources, and intimate Partner Violence Against Women in Egypt before and after the Arab Spring**. *Violence Against Women* (2022.0) **28** 347-74. DOI: 10.1177/1077801221992877 110. Asmare G, Nigatu D, Debela Y. **Factors affecting men’s involvement in maternity waiting home utilization in North Achefer district, Northwest Ethiopia: a cross-sectional study**. *PLoS ONE* (2022.0) **17** e0263809. DOI: 10.1371/journal.pone.0263809 111. 111.Kotelchuck M, Levy RA, Nadel HM. Voices of fathers during pregnancy: the MGH prenatal care obstetrics Fatherhood Study Methods and results. Matern Child Health J 2022 Jun 29. 10.1007/s10995-022-03453-y. Epub ahead of print. PMID: 35768674. 112. Guspianto IIN, Asyary A. **Associated Factors of Male Participation in Antenatal Care in Muaro Jambi District, Indonesia**. *J Pregnancy* (2022.0) **2022** 6842278. DOI: 10.1155/2022/6842278 113. Nyamai PK, Matheri J, Ngure K. **Prevalence and correlates of male partner involvement in antenatal care services in eastern Kenya: a cross-sectional study**. *Pan Afr Med J* (2022.0) **41** 167. DOI: 10.11604/pamj.2022.41.167.31535 114. Boniphace M, Matovelo D, Laisser R, Yohani V, Swai H, Subi L, Masatu Z, Tinka S, Mercader HFG, Brenner JL, Mitchell JL. **The fear of social stigma experienced by men: a barrier to male involvement in antenatal care in Misungwi District, rural Tanzania**. *BMC Pregnancy Childbirth* (2022.0) **22** 44. DOI: 10.1186/s12884-022-04383-x 115. Bawadi HA, Qandil AM, Al-Hamdan ZM, Mahallawi HH. **The role of fathers during pregnancy: a qualitative exploration of arabic fathers’ beliefs**. *Midwifery* (2016.0) **32** 75-80. DOI: 10.1016/j.midw.2015.10.006 116. 116.Lucente A. Three countries where abortion is legal in the Middle East and North Africa. 2022. Available at https://www.al-monitor.com/originals/2022/05/three-countries-where-abortion-legal-middle-east-and-north-africa. 117. Hessini L. **Abortion and Islam: policies and practice in the Middle East and North Africa**. *Reprod Health Matters* (2007.0) **15/29** 75-84. DOI: 10.1016/S0968-8080(06)29279-6 118. Musallam B. *Sex and society in Islam: birth control before the nineteenth century* (1989.0) 119. Bowen DL. **Abortion, Islam, and the 1994 Cairo population conference**. *Int J Middle East Stud* (1997.0) **29/2** 161-184. DOI: 10.1017/S002074380006445X 120. Asman O. **Abortion in islamic countries: legal and religious aspects**. *Med Law* (2004.0) **23/1** 73-89 121. Shapiro GK. **Abortion law in muslim-majority countries: an overview of the islamic discourse with policy implications**. *Health Policy and Plan* (2014.0) **29/4** 483-94. DOI: 10.1093/heapol/czt040 122. Hajri S, Raifman S, Gerdts C. **This is real misery: experiences of women denied legal abortion in Tunisia**. *PLoS ONE* (2015.0) **10/12** e0145338. DOI: 10.1371/journal.pone.0145338 123. Maffi I, Affes M. **La santé sexuelle et reproductive en tunisie: institutions médicales, lois et itinéraires thérapeutiques des femmes après la revolution**. *L’Année du Maghreb* (2017.0) **17/2** 151-68 124. Gruenais M. **La publicisation du débat sur l’avortement au Maroc: L’Etat marocain en action**. *L’Année du Maghreb* (2017.0) **17/2** 219-34. DOI: 10.4000/anneemaghreb.3271 125. Maffi I. **Abortion in Tunisia after the revolution: bringing a new morality into the old reproductive order**. *Glob Public Health* (2018.0) **13/6** 680-91. DOI: 10.1080/17441692.2017.1284879 126. Shahawy S, Diamond MB. **Perspectives on induced abortion among palestinian women: Religion, culture and access in the occupied palestinian territories**. *Cult Health Sex* (2018.0) **20/3** 289-305. DOI: 10.1080/13691058.2017.1344301 127. 127.UNHCR. Middle East and North Africa. 2021a. Available at https://reporting.unhcr.org/mena. 128. Logie CH. **Exploring associations between adolescent sexual and reproductive health stigma and HIV testing awareness and uptake among urban refugee and displaced youth in Kampala**. *Uganda SRHM* (2019.0) **27** 86-106. PMID: 31880507 129. Kaczkowski W, Swartout KM. **Exploring gender differences in sexual and reproductive health literacy among young people from refugee backgrounds**. *Cult Health Sex* (2020.0) **22** 369-84. DOI: 10.1080/13691058.2019.1601772 130. 130.UNHCR. Middle East and North Africa. 2021b. Available at https://reporting.unhcr.org/globalreport2021/mena. 131. Stycos JM, Sayed HA, Avery R. **An evaluation of the population and development program in Egypt**. *Demography* (1985.0) **22** 431-43. DOI: 10.2307/2061070 132. Sallam SA, Mahfouz AA, Alakija W, al-Erian RA. **Continuing medical education needs regarding AIDS among egyptian physicians in Alexandria, Egypt and in the Asir Region, Saudi Arabia**. *AIDS Care* (1995.0) **7** 49-54. DOI: 10.1080/09540129550126957 133. Warren CW, Hiyari F, Wingo PA, Abdel-Aziz AM, Morris L. **Fertility and family planning in Jordan: results from the 1985 Jordan husbands’ fertility survey**. *Stud Fam Plann* (1990.0) **21** 33-9. DOI: 10.2307/1966589 134. Fido A, Al Kazemi R. **Survey of HIV/AIDS knowledge and attitudes of Kuwaiti family physicians**. *Fam Pract* (2002.0) **19** 682-4. DOI: 10.1093/fampra/19.6.682 135. El-Sony AI. **The cost to health services of human immunodeficiency virus (HIV) co-infection among tuberculosis patients in Sudan**. *Health Policy* (2006.0) **75** 272-9. DOI: 10.1016/j.healthpol.2005.01.006 136. Kobeissi L, Inhorn MC, Hannoun AB, Hammoud N, Awwad J, Abu-Musa AA. **Civil war and male infertility in Lebanon**. *Fertil Steril* (2008.0) **90** 340-5. DOI: 10.1016/j.fertnstert.2007.06.061 137. Kahhaleh JG, El Nakib M, Jurjus AR. **Knowledge, attitudes, beliefs and practices in Lebanon concerning HIV/AIDS, 1996–2004**. *EMHJ* (2009.0) **15** 920-33. PMID: 20187544 138. Al-Serouri AW, Anaam M, Al-Iryani B, Al Deram A, Ramaroson S. **AIDS awareness and attitudes among yemeni young people living in high-risk areas**. *EMHJ* (2010.0) **16** 242-50. PMID: 20795436 139. Kabbash IA, Mekheimer Sl, Hassan NM, Al-Nawawy AN, Attalla AA. **Evaluation of HIV voluntary counselling and testing services in Egypt. Part 2: service providers’ satisfaction**. *EMHJ* (2010.0) **16** 491-7. PMID: 20799547 140. Ellepola AN, Joseph BK, Sundaram DB, Sharma PN. **Knowledge and attitudes towards HIV/AIDS amongst Kuwait University dental students**. *Eur J Dent Educ* (2011.0) **15** 165-71. DOI: 10.1111/j.1600-0579.2010.00652.x 141. Saleh WF, Gamaleldin SF, Abdelmoty HI, Raslan AN, Fouda UM, Mohesen MN, Youssef MA. **Reproductive health and HIV awareness among newly married egyptian couples without formal education**. *Int J Gynaecol Obstet* (2014.0) **126** 209-12. DOI: 10.1016/j.ijgo.2014.02.027 142. Othman SM. **Knowledge about HIV/AIDS among high school students in Erbil city/Iraq**. *Glob J Health Sci* (2014.0) **7** 16-23. DOI: 10.5539/gjhs.v7n1p16 143. Akhu-Zaheya LM, Masadeh AB. **Sexual information needs of arab-muslim patients with cardiac problems**. *Eur J Cardiovasc Nurs* (2015.0) **14** 478-85. DOI: 10.1177/1474515115597353 144. AlMuzaini AA, Yahya AS, Ellepola AN, Sharma PN. **HIV/AIDS: dental assistants’ self-reported knowledge and attitudes in Kuwait**. *Int Dent J* (2015.0) **65** 96-102. DOI: 10.1111/idj.12136 145. Haroun D, El Saleh O, Wood L, Mechli R, Al Marzouqi N, Anouti S. **Assessing knowledge of, and Attitudes to, HIV/AIDS among University students in the United Arab Emirates**. *PLoS ONE* (2016.0) **11** e0149920. DOI: 10.1371/journal.pone.0149920 146. Lohiniva AL, Benkirane M, Numair T, Mahdy A, Saleh H, Zahran A, Okasha O, Talaat M, Kamal W. **HIV stigma intervention in a low-HIV prevalence setting: a pilot study in an egyptian healthcare facility**. *AIDS Care* (2016.0) **28** 644-52. DOI: 10.1080/09540121.2015.1124974 147. Sexty RE, Hamadneh J, Rösner S, Strowitzki T, Ditzen B, Toth B, Wischmann T. **Cross-cultural comparison of fertility specific quality of life in german, hungarian and jordanian couples attending a fertility center**. *Health Qual Life Outcomes* (2016.0) **14** 27. DOI: 10.1186/s12955-016-0429-3 148. Ashry M, Ziady H, Hameed M, Mohammed F. **Health-related quality of life in healthy children and adolescents of HIV-infected parents in Alexandria, Egypt**. *J Egypt Public Health Assoc* (2017.0) **92** 212-9. DOI: 10.21608/EPX.2018.22042 149. Spindler E, Bitar N, Solo J, Menstell E, Shattuck D. **Jordan’s 2002 to 2012 fertility stall and parallel USAID Investments in Family Planning: Lessons from an Assessment to Guide Future Programming**. *Glob Health Sci Pract* (2017.0) **5** 617-29. DOI: 10.9745/GHSP-D-17-00191 150. Zaidouni A, Ouasmani F, Benbella A, Ktiri F, Abidli Z, Bezad R. **What are the needs of infertile moroccan couples in assisted Reproductive Technology?: exploratory qualitative study in the first fertility public center in Morocco**. *Bangladesh J Medical Sci* (2020.0) **19** 697-704. DOI: 10.3329/bjms.v19i4.46628 151. Shah S, Elgalib A, Al-Wahaibi A, Al-Fori M, Raju P, Al-Skaiti M, Al-Mashani HN, Duthade K, Omaar I, Muqeetullah M, Mitra N, Shah P, Amin M, Morkos E, Vaidya V, Al-Habsi Z, Al-Abaidani I, Al-Abri SS. **Knowledge, Attitudes and Practices related to HIV Stigma and discrimination among Healthcare Workers in Oman**. *Sultan Qaboos Univ Med J* (2020.0) **20** e29-36. DOI: 10.18295/squmj.2020.20.01.005 152. Sait M, Aljarbou A, Almannie R, Binsaleh S. **Knowledge, attitudes, and perception patterns of contraception methods: cross-sectional study among saudi males**. *Urol Ann* (2021.0) **13** 243-53. DOI: 10.4103/UA.UA_42_20 153. Karim SI, Irfan F, Saad H, Alqhtani M, Alsharhan A, Alzhrani A, Alhawas F, Alatawi S, Alassiri M, Ahmed MA. **Men’s knowledge, attitude, and barriers towards emergency contraception: a facility based cross-sectional study at King Saud University Medical City**. *PLoS ONE* (2021.0) **16** e0249292. DOI: 10.1371/journal.pone.0249292 154. Kapoor NR, Langer A, Othman A, Gausman J. **Healthcare practitioners experiences in delivering sexual and reproductive health services to unmarried adolescent clients in Jordan: results from a cross-sectional survey**. *BMC Health Serv Res* (2022.0) **22** 31. DOI: 10.1186/s12913-021-07415-y
--- title: A multiscale mechanistic model of human dendritic cells for in-silico investigation of immune responses and novel therapeutics discovery authors: - Sara Sadat Aghamiri - Bhanwar Lal Puniya - Rada Amin - Tomáš Helikar journal: Frontiers in Immunology year: 2023 pmcid: PMC10040975 doi: 10.3389/fimmu.2023.1112985 license: CC BY 4.0 --- # A multiscale mechanistic model of human dendritic cells for in-silico investigation of immune responses and novel therapeutics discovery ## Abstract Dendritic cells (DCs) are professional antigen-presenting cells (APCs) with the unique ability to mediate inflammatory responses of the immune system. Given the critical role of DCs in shaping immunity, they present an attractive avenue as a therapeutic target to program the immune system and reverse immune disease disorders. To ensure appropriate immune response, DCs utilize intricate and complex molecular and cellular interactions that converge into a seamless phenotype. Computational models open novel frontiers in research by integrating large-scale interaction to interrogate the influence of complex biological behavior across scales. The ability to model large biological networks will likely pave the way to understanding any complex system in more approachable ways. We developed a logical and predictive model of DC function that integrates the heterogeneity of DCs population, APC function, and cell-cell interaction, spanning molecular to population levels. Our logical model consists of 281 components that connect environmental stimuli with various layers of the cell compartments, including the plasma membrane, cytoplasm, and nucleus to represent the dynamic processes within and outside the DC, such as signaling pathways and cell-cell interactions. We also provided three sample use cases to apply the model in the context of studying cell dynamics and disease environments. First, we characterized the DC response to Sars-CoV-2 and influenza co-infection by in-silico experiments and analyzed the activity level of 107 molecules that play a role in this co-infection. The second example presents simulations to predict the crosstalk between DCs and T cells in a cancer microenvironment. Finally, for the third example, we used the Kyoto Encyclopedia of Genes and Genomes enrichment analysis against the model’s components to identify 45 diseases and 24 molecular pathways that the DC model can address. This study presents a resource to decode the complex dynamics underlying DC-derived APC communication and provides a platform for researchers to perform in-silico experiments on human DC for vaccine design, drug discovery, and immunotherapies. ## Introduction Dendritic cells (DCs) comprise a diverse set of antigen-presenting cells that are responsible for the recognition of foreign and self-antigens and the subsequent regulation and initiation of specialized adaptive and innate immune responses [1, 2]. Via pattern recognition receptors, DCs can sense a wide range of epitopes expressed by pathogens and damaged cells [3]. The sophisticated ontogeny of DCs enables them to maintain tolerance in the presence of foreign and self-antigens or to initiate an inflammatory response (4–6). Striking the right balance to antigen response puts DCs in a critical pathway for disease management [7, 8]. An insufficient immune response to an antigen can suppress downstream cell differentiation leading to an increased risk of infection and malignancy [9, 10]. An over-reactive or chronic immune response, however, can lead to auto-immune diseases, allergies, and chronic inflammation [11, 12]. DCs mediate adaptive responses through cell-cell interactions (e.g., antigen presentation via the major histocompatibility complex (MHC) classes), the increase of co-stimulatory immune checkpoint ligands/receptors, and through the secretion of pro-and anti-inflammatory interleukins, growth factors, and chemokines [13, 14]. Antigen recognition triggers a cascade of signaling pathways that switch the DC cellular state from tolerant (immature) to inflammatory (mature) [15, 16]. DCs comprise three major subtypes with distinct immunogenicity and plasticity: conventional DCs (cDC1 and cDC2), plasmacytoid DCs (pDCs), and monocyte-derived DCs (MoDCs). As mature DCs, they can prime effectors and suppressors cells (e.g., lymphocyte T and B cells) to stimulate a wide range of immune responses [17, 18]. The significance of DCs in identifying and initiating an adaptive response to foreign and self-antigens has stimulated interest in isolating DCs as a potential therapeutic tool to program specific immune responses to pathogens and malignant cells (19–21). For example, in 2010, a DC-based vaccine was approved for the prevention of prostate cancer [22]. DC-based vaccine development for other diseases has not been as successful; achieving full maturation of DCs and a limited ability for DCs to activate T cells are some of the challenges that have been encountered [23]. Improved methods for characterizing and perturbing the complex mechanisms underlying DC maturation in the context of the broader immune system may help translate biological knowledge to clinical applications. Computational models, for example, have been gaining traction as a means to study the dynamics of immune responses in the context of homeostasis and diseased states (24–26) by utilizing a variety of mathematical frameworks (27–29) to represent multiple levels of biological regulation (e.g., genome-scale metabolic network regulation, signal transduction, cell-to-cell communication, etc.). Despite previous modeling efforts of immune-related biological diseases, a large-scale model of major DC functions and its communication with other immune cells is still lacking. Multiscale models have the potential to uncover the underlying mechanisms behind emergent behaviors at various scales such as intracellular, cellular, and systemic levels. Such models can consider various temporal and organizational scales including signal transduction, gene regulation, metabolism, cellular behaviors, and cytokine transport [25]. Different multiscale models have been developed to study the dynamic response of DC under different extracellular environments. For example, Klinke II DJ. developed a multiscale model to investigate the impact of the lung microenvironment on the education of DC for optimal T cell polarization. The model considered DC trafficking and education in the lung while taking into consideration the time, maturation, spatial distribution and IL12 response [30]. Mei Y. and colleagues created a multiscale platform, the ENteric Immune Simulator (ENISI), to study the mucosal immune response during colonic inflammation. The multiscale tool has the advantage of connecting three different scales - intracellular, cellular, and tissular - using different mathematical languages [31]. Lai et al. developed a multiscale model of DC-based vaccine by considering the signaling pathways underlying DC maturation, the bio-distribution of DCs in multiple organs, and the DC-T-cell response to identify optimal targets for enhancing anti-cancer DC vaccination in the context of melanoma [28]. Although those multiscale models included a tissue scale to study the dynamic distribution of DC in different organs, so far, these models have only focused on DC functionality under specific disease conditions or specific signaling pathways. However, DC functions involve complex intracellular and cellular networks that are critical for regulating cell activation and initiating immune responses. Logical modeling formalism has emerged as a particularly effective approach to modeling large-scale biological systems due to scalability and independence of kinetic parameters that are largely unknown (32–35). Logical models of different scales and complexity (a few to hundreds of components) have been applied to study various biological and translational questions [36], such as studying cellular crosstalk [37] and predicting cellular phenotypes [24, 38] and drug targets [39]. Here, we present a multiscale mechanistic model of human DCs that captures the complex interplay of intracellular molecular signaling to intercellular cell-cell communications. DC model enables researchers to easily modify, expand and test new hypotheses of the immune system. Our aim is to provide the researchers with computational tools to gain insight into DC mechanisms and disease pathology. The mechanistic model uses the logical mathematical framework [40] and focuses on signal transduction networks responsible for regulating DCs’ antigen-presenting cellular function, cellular interactions, maturation process, and immune cell population dynamics. It captures the dynamic biological events in response to diverse stimuli (pathogens, malignancy) and the downstream biological coordination between surface molecules (receptors, integrins, lectins), signal transduction (kinases, enzymes, transcription factors), and secretory factors (cytokines, chemokines). Two diseases are highlighted to demonstrate the utility of the model under diverse conditions. Lastly, receptor-ligand interactions between DCs and four immune cell types that DCs commonly interact with (T cells, B cells, natural killer (NK) cells, and neutrophils) have also been represented. The results of in-silico simulations of the model under various environmental conditions and network perturbations were validated using peer-reviewed published literature. ## Model construction The computational model is a mechanistic, logic-based model. Each component of the model can assume an active [1] or inactive [0] state at any time t. The activity state of the model’s internal components is determined by the regulatory mechanisms of other directly interacting components. These regulatory mechanisms are described with Boolean functions comprised of AND, OR, NOT operators [40]. To gain a comprehensive understanding of the molecular pathways involved in dendritic cells and antigen-presenting cells, we conducted a systematic search of the literature using PubMed. Our search was specifically focused on exploring the molecular pathways involved in each DC subtype: pDC, cDC1, cDC2, and MoDC. To limit the search results, we utilized a combination of search terms, including: “dendritic cells AND antigen-presenting cell AND MoDC AND molecular pathway,” “dendritic cells AND antigen-presenting cell AND pDC AND molecular pathway,” “dendritic cells AND antigen-presenting cell AND cDC1 AND molecular pathway,” and “dendritic cells AND antigen-presenting cell AND cDC2 AND molecular pathway.” This comprehensive search allowed us to obtain a wealth of information related to the molecular pathways involved in each DC subtype, providing a foundation for our investigation into the function and activation of these important immune cells. In the development of our model, we followed strict data inclusion criteria, limiting our selection to original research articles focused on healthy human subjects. Studies using mice and clinical trials were excluded from our manual literature mining process. The first draft of the model was constructed using the information obtained from the manual literature mining of the original studies. Upon reviewing the initial draft, we conducted a supplementary search of the literature utilizing both review and original studies to obtain well-established biological information related to regulators of the unconnected components. This thorough and systematic approach allowed us to develop a comprehensive model using 92 publications (83 original and 9 reviews) that represents the molecular pathways involved in dendritic cells. We defined subtype-specific markers to differentiate between pDC, cDC1, cDC2, MoDC (Results and Supplemental Figure S1A). To validate the model, we collected literature reporting specific DC response to different extracellular conditions and simulated emergent behaviors that were not directly programmed into the model [41]. Because logical models are of qualitative nature, model validations focus on the ability of the model to reproduce qualitative behaviors seen in wet-lab experiments (e.g., change in activity level of a component(s) under specific extracellular conditions) - a standard process for logical models (24, 41–43). From the publications, we retrieved information related to DC-specific stimuli, the effect of the studied environment, and comprehensive signaling pathways (receptors, kinases, transcription factors). The model consists of 281 components. These components are categorized into various classifications and compartments. There are 178 proteins, 87 RNAs, and 16 components representing phenotypes and cells. The proteins are organized in cell membrane (64 components), cytoplasm (40 components), nucleus (22 components), extracellular space (52 components). Figure 1 Created with BioRender.com. To standardize the naming convention of the components in the model, we used protein and gene names from the HUGO Gene Nomenclature Committee (HGNC) [44]. The model was built in the web-based modeling and analysis platform, Cell Collective, and manually curated using the aforementioned literature [45]. All components used to build the regulatory mechanisms have been annotated in Cell Collective with the exact quote from the reference literature. The model is publicly available in Cell Collective (under Published Models) where it can be simulated as well as downloaded (and other logical models published by the community) in several file formats (such as SBML-qual, text file of logical functions, and truth tables) [46, 47]. ## Model simulations and analyses Cell Collective was used to perform all computational simulations and analyses of the model. Cell Collective uses discrete mathematics to construct the model, but the simulated output values are semi-continuous, ranging from 0 to $100\%$ activity levels [48, 49]. The activity levels of external components are unitless and defined as a percent chance (probability * 100) of the component being active in a specific time t [24]. Depending on the desired experiment, the activity levels of external components can be set by the user to specific values, or they can be set to ranges from which values during each simulation are selected randomly (e.g., to simulate dose-response experiments). We used Cell Collective for two types of analyses: real-time and dose-response using asynchronous updates such that all genes take different times to make a transition, which is closer to biological phenomena [50]. The initial condition of the model was set to immature cellular phenotype as 1 (active) and all other components were set to zero since DCs are considered immature under the resting condition and before stimuli activation [51]. The immature DCs are recruited to the inflamed site by pathogen signals, capture foreign antigens and undergo maturation to DC subtypes [52]. While simulating the model in Cell Collective, the user can define the activity levels of external components to a specific point or provide ranges (e.g., varying between $0\%$ to $100\%$). When a range is defined for external components, their activity levels are selected randomly in each simulation. In the real-time simulation, we showed the activity of components at different times (steps), which was presented using the mean activity level of multiple simulations. For dose-response analysis, we conducted each simulation consisting of 800 steps. The activity levels output components are fractions of ones over the last 300 iterations (500 to 800 steps) that describe the model’s steady behavior as described by [48, 49]. Under each environment set for a biological scenario, we used 1,000 simulations. ## Global sensitivity analysis We used sensitivity analysis in Cell Collective to determine the association between external components (e.g., in-vitro inducers) and internal components (such as TLRs, cells, cytokines, and phenotypes). We used probabilistic global sensitivity analysis based on standardized regression coefficient (SRC) using the “sensitivity” package in R [24, 53] on input and output data of Cell Collective. In a single-input setting, we used SRC, which measures the strength of association between dependent and independent variables [53]. We performed Cell Collective simulations under input activity levels ranging from 0 to $100\%$. The activity levels of inputs and outputs were independent and dependent variables in the statistical model. A higher SRC value represents a higher strength of association between input and output variables. We used SRC and k-means clustering algorithms (900 samples, specified three clusters for low, medium, and high activity levels) methods to visualize the simulation results. ## Kyoto Encyclopedia of Genes and Genomes pathways analysis The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway [54] enrichment analysis was conducted using the DAVID Bioinformatics Resources (2021 Update) [55] to explore the model components at the functional level. DAVID is a gene functional classification tool in which we used a p-value <0.05 with a false discovery rate (< $5\%$) as the cutoff criterion for KEGG pathway enrichment. We used the ggplot2 R package to visualize the fold enrichment and P-values of the top 20 enriched KEGG pathways. ## A large-scale multicellular, mechanistic model of signal transduction regulation of dendritic cell immune responses We constructed a mechanistic multiscale model of signal transduction networks governing the proper function of human DCs, spanning biological scales from molecular to cell-to-cell communication. The model comprises 281 components and 702 interactions between these components that regulate the DC responses. The multiscale nature of our model is based on the intercellular and intracellular communication dynamics between DCs and four other immune cell types, serving as a bridge between innate and adaptive immunity [56]. To facilitate the model’s utility, its architecture across various biological scales is first depicted in a biological illustration of the pathways, communication molecules, cell markers, and receptors involved in regulating DCs from immature cells to mature phenotypes (Figure 1). **Figure 1:** *Schematic representation summarizing the main components of the model and the connection between each biological layer of the system.* Further, the model’s architecture is depicted in Figure 2. Herein, the organization follows communication from the DC’s extracellular space to the plasma membrane (ligand-receptors, markers), to the cytoplasm (kinases and signaling cascades), to the nucleus (transcription factors and gene regulations) and to the secretory compartments that communicate with various DC phenotypes and other immune cell types that interact with DCs (Figure 2; Table S1). The model is freely available in the Cell Collective modeling software and repository [45, 57] (DC model Link and workflow to Cell Collective environment in Supplemental Figure S2). Each component and interaction of the model have been fully annotated to facilitate data transparency and reusability. In Cell Collective, the community can also directly simulate and analyze the model, further improve it, or download it as an SBML file [46]. **Figure 2:** *Visualization of the large-scale DC model in Cell Collective. The network view of the model. Dots represent signaling molecules; edges represent interactions between the model components. Red edges represent the inhibitions, while green and gray edges show activatory interactions. External (stimuli) and internal components are colored yellow and gray, respectively.* The model includes main signaling pathways, immune checkpoints, cytokines, and DC response mechanisms, and is able to represent the DC antigen presentation and maturation functions. The intra-cellular scale includes receptors that can sense the cellular environment and the downstream pathways that regulate DC responses that result in different cellular phenotypes. A diverse group of pattern recognition receptors (PRRs) is included in the model, including toll-like receptors (TLRs; TLR1, 2, 3, 4, 7, 8, and 9), C-type lectin receptors (CLRs; CLEC4C, CLEC7A, CLEC9A, CLEC10A, CD209), nucleotide-binding domain/leucine-rich repeat-containing receptors (NLRs; NOD2). The model also contains regulatory proteins (NF-κB, PYCARD, NLRP3, SYK, LILRA4, ISG20, ADAR, BST2, DDIT3, ATF4, PPP1R15A), maturation molecules (CD80, CD83, CD86, CD40, HLA-DR), and cytokines (IL6, TNF, IL12, IFNA1, IL12A, IL1B, IL12B, IL23A, IL10, IFNB1) [58]. In our study, we differentiated between various DC subtypes using a combination of subtype-specific markers. For pDCs, we employed CLEC4C (C-type lectin domain family 4 member C), and NRP1 (neuropilin 1) (59–61). For cDC1, we utilized CLEC9A (Dendritic cell C-type lectin receptor 9A), XCR1 (XC chemokine receptor 1), and THBD (thrombomodulin) (59, 60, 62–64). cDC2 was characterized using CLEC10A (C-type lectin domain containing 10A), and CD163 (CD163 molecule) [59, 60, 65]. For MoDCs, we applied MRC1 (mannose receptor C-type 1), and CD1A (CD1a molecule) [59, 60, 66]. The intracellular molecular cascades stimulate interaction with other immune cells, including effector and exhausted T cells, B cells, NK cells, and neutrophils. For example, DCs capture and display antigen protein fragments on their plasma membrane through the antigen presentation process [67] then antigenic peptides are bound to appropriate molecules of the MHC, also known in humans as the human leukocyte antigen (HLA). T cells can recognize the antigens at the T cell-APC interface. DCs’ highly stimulatory and versatile APC function produces cytokines, interferons (IFNB), and tumor necrosis factor superfamily (TNF) to stimulate naive T cells to differentiate into effector subsets [68]. As such, cytokines (IL1, IL6, IL10, IL12, IL23) and IFNB are also included in the model. DCs increase the expression of the MHC, the adhesion molecules, and the co-stimulators upon maturation, further stimulating T-cell proliferation and cytokine release [69]. DC immune checkpoints are also included in the model (ICOS-LG, TNFSF9, TNFSF4, CD70, PVR, Nectin-2, BTLA, and PD-L1); the checkpoints regulate stimulatory and inhibitory pathways capable of maintaining self-tolerance and facilitating the immune response [70]. ## In-silico model validation To validate the mechanistic model of DC functions, we collected experimental data from 30 different studies (Supplemental Data, Table S2, Figure S1) and reproduced them via in-silico experiments. Below we describe four in-silico experiments that showcase how the model was used to validate well-known (extensively published) in-vitro experiments spanning intracellular communication dynamics, intercellular communication dynamics, and both inter and intracellular communication dynamics. The first experiment assessed the model’s ability to reproduce the behavior of TLRs. Namely, Grandclaudon [26] studied a range of DC molecular states expressing various patterns of communication signals. The authors present that DCs were treated for 24 hours with lipopolysaccharide (LPS), which activated TLR4 signaling pathways and induced DC communication molecules, including IL1β, IL6, TNF-α, and IL12 cytokines. Similar studies [58, 71, 72] present the LPS-induced secretion of DC cytokines as a non-trivial test to investigate the TLR4 cooperation in response to infections. A better understanding of the mechanisms of host resistance can provide a basis for the development of more effective adjuvants and immunotherapeutic regimens. To validate the ability of the model to reproduce TLR behavior, we first validated that the model contained all components and regulatory pathways to support this experiment. Next, we ran 1,000 dose-response simulations while defining three activity levels of LPS (0, 50, and 100). We then evaluated and compared the secretion of inflammatory cytokines – IL1β, IL6, IL12, IFN-α, and TNF – upon activation by LPS at each of the three different dose responses against well-established cytokine responses. Figure 3A displays the secretion level for each of the five cytokines at LPS doses 0, 50, and 100. As expected, there was no cytokine secretion at dose 0 and subsequent secretion and elevated secretion at doses 50 and 100, respectively. **Figure 3:** *In-silico model validations. (A) Inflammatory cytokine activity level in response to LPS using dose-response analysis. (B) Standard regression coefficient of main factors activated in the HIV infection environment. (C) Activity level of IL2, IFN-α and NK under CpG-containing oligonucleotides and poly(I:C) stimulation. (D) Markers of maturation at different time points in the presence of neutrophils. (E) Time course distribution of DC immature (im-DC) and mature (mDC) states with (+) or without (-) neutrophils.* The second experiment assessed the model’s ability to mount an appropriate immune response to the presence (and initiation) of the human immunodeficiency virus (HIV) infection. HIV initiates viral transcription through TLR8 and promotes the maturation of DCs (from immature (imDC) to mature plasmacytoid (pDC)) as defined by the expression of CD83 and CCR7 surface markers and the production of IFN-α and TNF [73, 74]. We used Cell Collective’s global sensitivity analysis (refer to the “Methods” section, “Global sensitivity analysis”) as a method to determine the association between HIV and each of the internal components. Figure 3B displays the correlation of activity for imDC, pDC, CD83, CCR7, IFN-α, and TNF in the presence of HIV infection. As expected, immature DCs (imDC) exhibit a negative correlation in the presence of HIV, which shows HIV-bearing immature DCs can differentiate into mature DCs in response to the infection, presenting HIV antigens to T cells and initiating viral immune responses. Further, mature DCs (pDC), as well as surface markers CD83 and CCR7, and IFN-α and TNF exhibit a positive standard regression coefficient (SRC), which means an increase in HIV load results in increased activity of these components. The simulation results are consistent with biological experiments that describe pDCs exposed to HIV strongly upregulating the expression of CD83 and functional CCR7 maturation markers, IFN-α, and TNF cytokines [73]. The third experiment assessed the model’s ability to simulate known intercellular dynamic crosstalk between DCs and other immune cells in tandem with intracellular communication dynamics. Gerosa [75] showed that human peripheral pDC and MoDCs are necessary to induce NK cell function depending on the type of microbial stimulus. In this experiment, pDCs and MoDCs were stimulated in response to CpG-containing oligonucleotides (CpG) and poly(I:C), and evaluated the mean activity level for NK cells, IL2, IFN-α, as a result of (CpG)/poly(I:C)-induced release of IL2 and IFN-α and subsequent activation of NK cells. Figure 3C displays the expected activity of IL2 and IFN-α as well as NK cells when CpG/poly(I:C) is inactive compared to an active state. Last, we validated intercellular communication dynamics between imDCs and neutrophils. Neutrophils stimulate imDCs to become competent antigen-presenting cells. This maturation phenotype is characterized by the expression of specific surface markers (e.g., HLA-DR, CD86, and CD40) and the secretion of IL12 in response to DC-neutrophil interactions (76–78). Figure 3D displays the activity of model components IL12, CD209, CD40, CD86, and HLA-DR in response to the presence of neutrophils over time, demonstrating the pathways responsible for neutrophil-induced DC maturation. Figure 3E shows the activity level of immature and mature DCs in the presence and absence of neutrophils. On the left, when neutrophils are absent, immature DCs continue to increase in activity over time, whereas mature DCs do not become active. On the right, as neutrophils become present, immature DC activity tapers, and mature DC activity increases. The aforementioned experiments illustrate the ability of the model to reproduce major experiments spanning complex inter- and intracellular communication dynamics. ## Case studies To aid researchers in identifying how to use this model, we showcase three case studies by presenting a brief application background, the method we used to apply the model in this context, and model results. ## Case 1: Intracellular communication dynamics. Characterization of DC response to a combinatorial COVID-19 and Influenza infection environment. In this case study, we integrated Covid-19 and Influenza pathogens into the model to characterize the molecular response of DC under single and co-infection conditions. Coronavirus disease 2019 (COVID-19) and Influenza respiratory disease, caused by Sars-CoV-2 and influenza virus, respectively, share similarities in seasonal manifestations, viral transmission method, symptoms, and immunopathogenesis [79, 80]. Co-infection with Sars-CoV-2 and influenza virus increases disease severity and impairs neutralizing antibody and CD4+ T cell responses [81]. Patients can develop both infections, and in some cases, co-infection leads to a poor prognosis (82–84). Despite the comprehensive investigation of DC behavior in single infections with Sars-CoV-2 or influenza, the comparative understanding of DC programming under co-infection is not fully explored due to limited patient cohorts and case studies (85–88). Thus, the purpose of this model-based study was to investigate the molecular behavior of DCs in three infectious states: infection with i) Influenza type A virus (IAV), ii) Sars-CoV-2, and ii) co-infection with IAV and Sars-CoV-2. We ran 900 dose-response simulations and identified three molecular patterns (Figure 4A, Table S3) based on similarity in activity levels of DCs molecular components. Each pattern presents a list of molecules that play a role in the co-infection. We reported the mean activity level of DC model molecules ranging from low activity (green) to fully activated (red) molecular state in single infection and co-infection cellular environments. **Figure 4:** *In-silico predictions of molecular activity across the whole DC model comparing Sars-CoV-2 and Influenza A virus (IAV) co-infection to the single infection. (A) The differential molecule expressions are grouped into three main patterns in response to each environmental setting. The first pattern grouped molecules that are regulated similarly in all three conditions. The second pattern is related to similar regulation between Influenza and co-infection, and the third one grouped similar behavior between Sars-CoV-2 and co-infection conditions. The scale represented the activity level ranging between 0 to 100%, 100 being the highest activity level. (B) The molecular signatures in pattern 1 in single and co-infection cellular environments. (C) An example of the second pattern shows that under different activity levels of Sars-CoV-2 (green, low; yellow, medium; purple, high), ICOSLG is inactive while it is upregulated in both co-infection and single IAV. (D) CLEC9A is categorized as the third pattern, and for both Sars-CoV-2 and co-infection, it has low activity levels compared to the IAV single infection.* In the first pattern (Figure 4A – Pattern 1 and 4B), we identified molecular signatures with similar activities in single infection and co-infection. Figure 4B presents molecular signatures following this pattern, including markers of DC differentiation (CD86, CD1A, CD40, CD83, ITGAM), PRRs (TLR8), immune checkpoint molecules (PVR, Nectin2), chemokines/chemokine receptor (CCR7, CCL19, CXCL8), cytokines (IL6, TNF, IL12, IL1B, IL12A, IL12B, IL23, IL10, IFNA1, IFNB1), signaling molecules (NF-κB, PYCARD, NLRP3, SYK, LILRA4, ISG20, ADAR, BST2, DDIT3, ATF4, PPP1R15A), and CLRs (CLEC4C, CLEC10A). Several of these signatures are expressed during the single infection studies on human samples (89–91). For example, separate studies on Sars-CoV-2 and influenza virus infections show expression of IL1B, IL10, TNF, CD86, CCR7, IL6, CXCL8, IFN [79, 89, 92, 93]. In the second pattern (Figure 4A – Pattern 2), the molecular signature characterizes the similarity between IAV single infection and co-infection. Previous studies indicated that immune checkpoints are increased in influenza single infection [94] but not in Sars-CoV-2 single infection [89, 95, 96]. Thus this experiment focuses on the significance of the immune checkpoint signatures. The immune checkpoints (TNFSF4, CD70, ICOSLG, PDCD1LG2), followed by cytokines (IL2, IFNL2, CXCL10), markers of DC differentiation (CD80, CD86), and signaling signature (SEMAD4), are upregulated in both co-infection and single IAV but downregulated under Sars-CoV-2 infection. As an example, Figure 4C shows the activity level of the ICOSLG immune checkpoint in 300 simulations per each infection condition (single and co-infection), which is higher in the presence of both viruses. In the third pattern (Figure 4A – Pattern 3), the molecular signatures of Sars-CoV-2 and co-infection were similar. The major signature includes a decreased expression of signaling and decreased expression of transcription factors in both Sars-CoV-2 and co-infection, suggesting a disruption of the signaling network associated with Sars-CoV-2 infection. Neuropilin-1 (NRP1), the only signaling protein to be highly expressed in the third pattern, was previously shown to facilitate Sars-CoV-2 entry by interacting with spike protein [97, 98]. Additional signatures are related to the decrease of pathogen sensors and maturation marker expressions, such as TLRs (TLR1, TLR7, MYD88), CLRs (CLEC9A, CLEC7A, CD209), and MHC class signatures (HLA-DQA, HLADPB1, HLA-DM, HLA-DRB1, HLA-DR, HLA-DQB1), suggesting the loss of DC function to sense and present antigen to other immune cells properly. For example, Figure 4D presents the simulation results of the C-type lectin domain containing 9A (CLEC9A) with a low activity level in co-infection. Several studies indicated that DCs displayed a defect in maturation and are depleted in COVID-19 patients, and as our in-silico simulations predicted, one of the mechanisms might be due to the defect of the signaling compartment and pathogen sensors [85, 86, 99]. Nevertheless, further experimental investigations are needed to explore these hypotheses. ## Case 2: Intercellular communication dynamics. Crosstalk between DCs and T cells in a cancer microenvironment. DCs play a crucial role in initiating a protective anti-tumoral response by presenting tumor antigens and providing co-stimulatory immune checkpoint to T cells [100]. However, tumor microenvironments sustain DCs in an immature/tolerant phenotype, thereby altering antigen presentation, co-stimulatory signals, and thus the ability to effectively activate T cells. Therefore, T-cells become exhausted due to continuous exposure to antigens and increase in multiple inhibitory immune checkpoints that further benefit the mechanism of resistance to immunotherapies [101]. Several factors with immunoregulatory properties are involved in DC-T cell interplay. For example, the cytokine HMGB1 released by cancer cells contributes to cancer development by promoting tolerogenic DC differentiation and the suppression of anti-tumoral T cells (102–104). Moreover, a study conducted in-vivo reported the role of HMGB1 in promoting T-cell exhaustion in the condition of trauma [105]. However, the role of cancer-derived HMGB1 in promoting exhaustion through the modulation of immune checkpoint expression has not been investigated. Modern immunotherapy approaches aim to reverse T cell exhaustion by blocking inhibitory immune checkpoint receptors [106]. Combinatorial treatments using approved inhibitors of PD-L1 immune checkpoint and two receptors PD-1 and CTLA-4, showed promising results. Additional immune checkpoint inhibitors are under clinical trial investigations [107]. However, not all cancer types respond equally, and patients can acquire resistance to immune checkpoint inhibitors (ICI) [108, 109]. Because many experimental studies have investigated the role of immune checkpoints individually, a computational approach can help to better understand the dynamic distribution of inhibitory and stimulatory immune checkpoints that can aid in identifying ideal checkpoint candidates and facilitate combinatorial therapeutic strategies. In this case study, we examined the impact of cancer-derived HMGB1 on the DC-T cell synapse interaction. We included the cancerous cytokine HMGB1 as environmental (cancer) stimulus. The DC model includes two groups of immune checkpoints: stimulatory and inhibitory ligands/receptors that are enable us to study the impact of HMGB1 on DC-T immune checkpoints. The ligands are expressed on DCs, while receptors are particularly expressed by T cells. In the model’s plasma membrane compartment (Figure 5A, Table S1) are included six stimulatory ligands (e.g., ICOS-LG, TNFSF9, TNFSF4, CD70), two inhibitory molecules (PD-L1, BTLA) and three molecules with a dual function depending on the receptors they are binding (CD80-CD86, PVR, and Nectin-2). From the T cell side, we included six stimulatory receptors (CD28, ICOS, TNFRSF9, TNFRSF4, CD27, CD226) to define effector T cells and three inhibitory receptors (PD-1, TNFRSF14, CTLA-4) that define exhausted T cells. Figure 5C showed the interaction between ligands with their respective receptors. Of note, CD80-CD86, used as main maturation markers, binds two different immune checkpoint receptors with opposite functions (the stimulatory receptor CD28 and the inhibitory receptor CTLA4), and the two ligands PVR and Nectin-2 share the same stimulatory receptor CD226 (Figure 5A). **Figure 5:** *Investigation of DCs-T cells crosstalk under an HMGB1 tumor environment. (A) The table indicates the immune checkpoint pairing between ligands and associated receptors along with the type of functions (stimulatory or inhibitory). (B) Assessment of IL6, IL8, IL12, and TNF cytokine expression under HMGB1 environment. (C) The activity level of MHC classes I and II are between 0 and 10, with the time 100 added at the final expression of both classes. (D) Standardized regression coefficient (SRC) between ligands and associated receptors on DC and T cells with HMGB1 environment. High and low SRC presented with yellow and purple, respectively. Arrows link the ligands to their respective receptors. (E) Time course of effector T and exhausted T cells activity level expression at a probability of activation at time steps 0, 1, 10, and 100 in HMGB1 simulation.* We compared the model’s simulation results under the HMGB1/cancer environment with published experimental data [110]. Messmer D. et al. showed that HMGB1 promotes the secretion of inflammatory cytokines (IL6, IL12, IL8, TNFα) and our in-silico simulation is consistent with the experimental data (Figure 5B). Then, in the HMGB1 environment, we evaluated: i) the activation of MHC Class I and II, ii) the distribution of ligands/receptors for co-stimulatory and inhibitory immune checkpoints expressed at the DC-T cell interface, and iii) the dynamic distribution of effector and exhausted T cells. We evaluated the distribution of the mean percentage activity level of MHC class I and II in five in-silico experiments at time steps 0, 1, 10, and 100. The mean activity level started from 0, and at time step 10 reaches $32.74\%$ for class I and $61.8\%$ for class II. MHC class I and II increase to maximum activity level at step 100 (Figure 5C). Our simulation indicated that MHC class I and II expression increased in response to HMGB1 simulation. Because our model does not include specific tumor antigens that can be presented by MHC to TCR, we cannot conclude that the increase of MHC expression is due to antigen overload, however, our simulation indicates an increase of MHC classes under HMGB1 simulation. Next, we investigated the dynamics of immune checkpoint pairs under HMGB1 environmental stimulation using dose-response and sensitivity analyses (Figure 5D). Our in-silico results showed that the stimulatory molecules such as CD27 (receptor for CD70), ICOS (receptor for ICOS-LG), TNFRSF4 (receptor for TNFSF4), and TNFRSF9 (receptor of TNFSF9) showed no significant correlation with HMGB1 stimulation and shared similar distribution with their receptors expressed by T cells. The PVR and Nectin-2 displayed high correlation as well as their receptor CD226. CD80-CD86 showed no significant correlation in response to HMGB1 stimulation (Figure 5D). Regarding the dual receptors of CD80-CD86, the stimulatory CD28 receptor, and the inhibitory receptor CTLA-4 showed no correlation under HMGB1 stimuli. The inhibitory pairing PD-L1 ligand with its receptor PD-1, the main target for immune checkpoint inhibitors, is highly represented. Moreover, the additional inhibitory receptor TNFRSF14 and its ligand BTLA ligand don’t show a significant distribution in response to HMGB1 stimulation (Figure 5D). We simulated the model under HMGB1 environment, and we evaluated the mean activity level of effector and exhausted T cells at time 0’, 1’, 10’, and 100’ from five in-silico experiments (Figure 5E). The exhausted T cell activity level is faster than the effector T cells at 10’ (Figure 5E). At the maximum time of the simulation, both T cell phenotypes do reach maximum accumulation (activity). The results demonstrate HMGB1 promotes both effectors and exhausted T cells and exhausted phenotype accumulated faster than effector. Immune checkpoint immunomodulatory functions are initiated by ligand-receptor interaction that can either promote or suppress T cell function [111]. CD226 is important in generating an anti-tumor response. While CD226 expression is required as a co-stimulatory factor for T cells during antigen presentation by APCs, the loss of CD226 can lead to impaired effector T activation and increased susceptibility to tumor development (112–114). Hence, in our model, the activity of CD226-PVR/Nectin2 contributed to the increase of effector T cells and is associated with the MHC expression. Among inhibitor pairs, only PD-L1-PD-1 displays a high correlation with an HMGB1 simulation. The interaction between PD-L1 and PD-1 drives T-cell dysfunction and exhaustion to prevent an efficient anti-tumor T-cell response [115, 116]. Previous studies indicated that HMGB1 increases PD-L1 expression in cancer cells; however, the modulation of PD-L1-PD-1 by HMGB1 in immune cells remains unknown. Our in-silico simulation suggests that HMGB1 can also promote PD-L1-PD-1 expression at the DC-T cell interface, thereby explaining the increase of exhausted T cells. In summary, using the example of analysis of multiple ligand/receptor-mediated cellular programming at the time, our in-silico experiments illustrated the capacity of the model to provide complex and dynamic insight into biological processes at the molecular and cellular scales. ## Case 3: The scope of the DC models offers potential applications in several immune-related diseases. The crosstalk between the disease environment and DCs highly contributes to the organization of the immune response (11, 12, 117–119). Because each disease environment is unique and complex, a multiscale model can be an effective tool to investigate the complexities underlying multiscale, systemic diseases. Given the DCs’ role in initiating both innate and adaptive immune responses, we sought to explore the links between the DC model’s core disease pathways we identified for IAV, Sars-CoV-2, and tumor microenvironment and additional diseases to identify the extensibility of our model. To do this, we performed a Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis against the model’s components (refer to “Methods” section, “Kyoto Encyclopedia of Genes and Genomes KEGG pathways analysis”). Using a cut-off p-value of <0.05, we identified 69 enriched pathways (Table S4) for the top 20 diseases (Figure 6). The fold enrichment analysis of the top 20 diseases revealed multiple categories, such as autoimmune disease, infection, and transplantation. **Figure 6:** *The top 20 human diseases and signaling pathways associated with the DC model. P-values and fold enrichment of KEGG pathways in DC. -log 10 of P-values was used for visualization. Thus, larger dot sizes correspond to lower P-values. The cutoffs P-value < 0.05 and a false discovery rate (FDR) < 5% were set for significant enrichment.* The two highest scores are represented by inflammatory bowel disease and tuberculosis infection. The enrichment for those two diseases can be explained by the presence of TLRs, lectins, and cytokines in supporting pathology [120]. For example, in the context of tuberculosis, TLRs and lectins can recognize different motifs of Mycobacterium tuberculosis, which in turn can trigger pro- or anti-inflammatory cytokine response [121]. The DC model can be utilized to identify pathways related to many diseases, including infections, cancer, and autoimmune diseases. The high fidelity of the model predictions will depend on the extension of the pathways related to the diseases of study. ## Discussion We have developed a mechanistic multiscale model of human DCs that spans biological scales from molecular interactions to cell-cell communication. We included biological events that occur between DCs’ environmental stimuli and their receptors, followed by activation of signal transduction in response to each signal. Moreover, we constructed the molecular network that links the downstream signal transduction of kinases and transcription factors to secreted cytokines/chemokines and growth factors. We extended the model further by integrating a cellular compartment that includes the communication between DCs and several innate/adaptive immune cells through direct (ligand-receptor) and indirect (cytokine, chemokines releases) interactions. Our model can be used to study DC maturation, differentiation to each subset, APCs function, and the bidirectional crosstalk between DCs and other immune cells. Because the model incorporates pathways that regulate and facilitate many key functions of DCs, it can be applied to study several diseases as well as the basic mechanism of DCs’ functions. The presented DC model leverages the widely used logical modeling formalism [40]. The advantages of this modeling approach include its scalability (efficient simulations) as evidenced by the fact that some of the largest computational models have been constructed using this formalism (e.g (122–125). Another advantage is that logical models do not rely on kinetic parameters that are mostly unknown [40, 126]. On the other hand, if one is interested in modeling relatively small and well-studied pathways (with known parameters), a kinetic modeling approach may be more appropriate. The model is limited by the missing data in the literature about any unknown interactions. Our model includes major pathways involved in DC immunobiology. Nevertheless, the model is limited in scope as it does not include all known DC signaling and cell-cell communication. The model is being provided in a readily exchangeable format (SBML) and easy-to-use modeling software (Cell Collective), making it relatively easy for the community to build on the model and continue to expand as needed by different applications. For example, to specifically investigate DC-T cell communication, T cell subsets such as CD4 and CD8 can be integrated by adding molecular and cellular components of the immunological synapse. We previously published a logical model of signal transduction networks governing CD4+ T cell differentiation in response to various cytokines [24]. Those same cytokines are also included in our DC model, creating the possibility of integrating both systems to study how DCs might influence CD4+ T cell fate and plasticity. As another example, HMGB1 interacts with several TLRs (e.g., TLR2, TLR4, and TLR9), which have been included in the DC model. HMGB1 also interacts with RAGE - a receptor for advanced glycation end-products - that is not currently included in our model. Adding RAGE to the system would increase the complex interplay between receptors and signaling pathways to mediate cytokine release and immune response (127–129). The model would then be able to simulate the different molecular intersections during single or multiple TLRs/RAGE activation and predict the multiple environmental conditions for efficient DCs maturation without compromising the adaptive response (e.g., T and B cells). Therefore, the multiscale model could be further used to characterize APC function in response to a stochastic tumor micro-environment with multiple components simultaneously. In our cancer in-silico simulations, our model-generated hypotheses suggested a list of potential immune checkpoints to explore for studying the effect of single and multiple combinatorial ICI on DC-T cell interaction outcome (Figure 5D). We showed the dynamics of immune checkpoint pairs under a tumor HMGB1 environment. Recent therapeutic approaches include the optimization of DC-based therapies by combining DC vaccines with immune checkpoint inhibitors (ICI), such as anti-CTLA-4 and anti-PD1/PDL1 [130, 131], or by silencing immune checkpoint signaling pathways [132]. Despite being in early clinical phases, combinatorial therapy holds a potential to balance toxicity, safety, and clinical outcomes [130, 131]. Additional ICI to restore T cell or APC activation is currently under investigation to expand therapeutic options and optimize the efficacy of the immune checkpoint targeting strategy [107, 133]. Nevertheless, the complexity of immune checkpoint ligands resides in their capacity to bind several different receptors with opposite functions, therefore switching between stimulatory and inhibitory signals. As the model prediction suggested, PVR and Nectin-2 showed a high activity similar to their receptor, CD226. Of note, PVR and Nectin-2 can trigger opposite signals whether they bind the stimulatory receptor (CD226) or the inhibitory receptors (TIGIT and CD96, not included in the model) [134]. Moreover, the optimal combination can depend on the ligands/receptors’ availability and the balance between stimulatory and inhibitory expression. Our model simulations suggest that the inhibitory receptor CTLA-4 has no activity under HMGB1 stimuli. At the same time, PD-1 and PD-L1 are highly correlated, suggesting that the use of anti-CTLA-4 might not be as effective as the use of anti-PD-1 or anti-PD-L1 to restore DC-T cell function in a cancerous HMGB1 environment [135]. The development of computational models that recapitulate complex human disease behavior can be a resource for scientists and clinicians to simulate thousands of possibilities for studying the complex biological process at multiple scales. The disease enrichment analysis highlighted the potential of our model to incorporate additional pathological events as some disease modules are already implemented. For example, Type I diabetes (T1DM), an auto-immune disease characterized by immune-mediated destruction of insulin-producing beta cells, is enriched in our model [136]. The loss of tolerance to self-antigens and the increase of autoreactive T cells instead of immunosuppressive T cells are the main cause of insulin deficiency. Several studies indicated that DCs presented self-antigen generated from degraded b-islet to prime autoreactive T cells via dysfunctional NF-κB and MAPK pathways [137, 138]. Current therapies focus on generating tolerant DCs and immunosuppressive T cells to target the auto-immune disease and restore the imbalance of tolerance [139]. To address those mechanisms, incorporating tolerogenic DCs and immunosuppressive T cell phenotype components under the stimulation of a self-antigen input could predict molecular conditions by which immunosuppressive cells are amplified to respond to disease pathology [140]. In summary, we have demonstrated the potential for a multiscale DC model to investigate the immunobiology of DCs and identify potential targets for improving the effectiveness of DC-based cell therapies. Lastly, the model can be further expanded to support additional mechanistic and therapeutic questioning related to DC ontogeny. ## Data availability statement The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding authors. ## Author contributions SSA, RA and TH conceived the study. SSA, RA and TH designed the study. SSA performed literature mining and collected the data. SSA constructed the models. SSA, and RA performed refinement of the constructed models. SSA, RA and BLP analyzed the data, performed the experimental work and analyzed the experimental results. SSA, BLP, RA and TH wrote the manuscript. RA and TH supervised the study. All authors contributed to the article and approved the submitted version. ## Conflict of interest TH is the majority stakeholder in Discovery Collective, Inc. with proprietary rights to Cell Collective. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ## Publisher’s note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. ## Supplementary material The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2023.1112985/full#supplementary-material ## References 1. Joffre O, Nolte MA, Spörri R, Reis e Sousa C. **Inflammatory signals in dendritic cell activation and the induction of adaptive immunity**. *Immunol Rev* (2009.0) **227**. DOI: 10.1111/j.1600-065X.2008.00718.x 2. Pasare C, Medzhitov R. **Toll-like receptors: linking innate and adaptive immunity**. *Microbes Infect* (2004.0) **6**. DOI: 10.1016/j.micinf.2004.08.018 3. Pandey S, Singh S, Anang V, Bhatt AN, Natarajan K, Dwarakanath BS. **Pattern recognition receptors in cancer progression and metastasis**. *Cancer Growth Metastasis* (2015.0) **8** 25-34. DOI: 10.4137/CGM.S24314 4. de Jong JMH, Schuurhuis DH, Ioan-Facsinay A, Welling MM, Camps MGM, van der Voort EIH. **Dendritic cells, but not macrophages or b cells, activate major histocompatibility complex class II-restricted CD4+ T cells upon immune-complex uptake**. *vivo Immunol* (2006.0) **119** 499-506. DOI: 10.1111/j.1365-2567.2006.02464.x 5. Kambayashi T, Laufer TM. **Atypical MHC class II-expressing antigen-presenting cells: can anything replace a dendritic cell**. *Nat Rev Immunol* (2014.0) **14**. DOI: 10.1038/nri3754 6. Figdor CG, de Vries IJM, Lesterhuis WJ, Melief CJM. **Dendritic cell immunotherapy: mapping the way**. *Nat Med* (2004.0) **10**. DOI: 10.1038/nm1039 7. Fricke I, Gabrilovich DI. **Dendritic cells and tumor microenvironment: a dangerous liaison**. *Immunol Invest* (2006.0) **35**. DOI: 10.1080/08820130600803429 8. Sesti-Costa R, de Moraes-Vieira PMM, Cervantes-Barragan L. **Dendritic cells: immune response in infectious diseases and autoimmunity**. *Mediators Inflammation* (2020.0) **2020**. DOI: 10.1155/2020/2948525 9. Audiger C, Rahman MJ, Yun TJ, Tarbell KV, Lesage S. **The importance of dendritic cells in maintaining immune tolerance**. *J Immunol* (2017.0) **198**. DOI: 10.4049/jimmunol.1601629 10. Domogalla MP, Rostan PV, Raker VK, Steinbrink K. **Tolerance through education: How tolerogenic dendritic cells shape immunity**. *Front Immunol* (2017.0) **8**. DOI: 10.3389/fimmu.2017.01764 11. Ganguly D, Haak S, Sisirak V, Reizis B. **The role of dendritic cells in autoimmunity**. *Nat Rev Immunol* (2013.0) **13**. DOI: 10.1038/nri3477 12. Chow TG, Gill MA. **Regulation of allergic inflammation by dendritic cells**. *Curr Opin Allergy Clin Immunol* (2020.0) **20** 56-63. DOI: 10.1097/ACI.0000000000000603 13. Lanzavecchia A, Sallusto F. **Regulation of T cell immunity by dendritic cells**. *Cell* (2001.0) **106**. DOI: 10.1016/s0092-8674(01)00455-x 14. Thaiss CA, Semmling V, Franken L, Wagner H, Kurts C. **Chemokines: a new dendritic cell signal for T cell activation**. *Front Immunol* (2011.0) **2**. DOI: 10.3389/fimmu.2011.00031 15. Wieczorek M, Abualrous ET, Sticht J, Álvaro-Benito M, Stolzenberg S, Noé F. **Major histocompatibility complex (MHC) class I and MHC class II proteins: conformational plasticity in antigen presentation**. *Front Immunol* (2017.0) **8**. DOI: 10.3389/fimmu.2017.00292 16. Reis e Sousa C. **Activation of dendritic cells: translating innate into adaptive immunity**. *Curr Opin Immunol* (2004.0) **16**. DOI: 10.1016/j.coi.2003.11.007 17. Geginat J, Nizzoli G, Paroni M, Maglie S, Larghi P, Pascolo S. **Immunity to pathogens taught by specialized human dendritic cell subsets**. *Front Immunol* (2015.0) **6**. DOI: 10.3389/fimmu.2015.00527 18. Eisenbarth SC. **Dendritic cell subsets in T cell programming: location dictates function**. *Nat Rev Immunol* (2019.0) **19** 89-103. DOI: 10.1038/s41577-018-0088-1 19. Banchereau J, Paczesny S, Blanco P, Bennett L, Pascual V, Fay J. **Dendritic cells: controllers of the immune system and a new promise for immunotherapy**. *Ann N Y Acad Sci* (2003.0) **987**. DOI: 10.1111/j.1749-6632.2003.tb06047.x 20. Mastelic-Gavillet B, Balint K, Boudousquie C, Gannon PO, Kandalaft LE. **Personalized dendritic cell vaccines-recent breakthroughs and encouraging clinical results**. *Front Immunol* (2019.0) **10**. DOI: 10.3389/fimmu.2019.00766 21. Yu J, Sun H, Cao W, Song Y, Jiang Z. **Research progress on dendritic cell vaccines in cancer immunotherapy**. *Exp Hematol Oncol* (2022.0) **11** 3. DOI: 10.1186/s40164-022-00257-2 22. Jähnisch H, Füssel S, Kiessling A, Wehner R, Zastrow S, Bachmann M. **Dendritic cell-based immunotherapy for prostate cancer**. *Clin Dev Immunol* (2010.0) **2010**. DOI: 10.1155/2010/517493 23. Cintolo JA, Datta J, Mathew SJ, Czerniecki BJ. **Dendritic cell-based vaccines: barriers and opportunities**. *Future Oncol* (2012.0) **8**. DOI: 10.2217/fon.12.125 24. Puniya BL, Todd RG, Mohammed A, Brown DM, Barberis M, Helikar T. **A mechanistic computational model reveals that plasticity of CD4+ T cell differentiation is a function of cytokine composition and dosage**. *Front Physiol* (2018.0) **9**. DOI: 10.3389/fphys.2018.00878 25. Wertheim KY, Puniya BL, La Fleur A, Shah AR, Barberis M, Helikar T. **A multi-approach and multi-scale platform to model CD4+ T cells responding to infections**. *PloS Comput Biol* (2021.0) **17**. DOI: 10.1371/journal.pcbi.1009209 26. Grandclaudon M, Perrot-Dockès M, Trichot C, Karpf L, Abouzid O, Chauvin C. **A quantitative multivariate model of human dendritic cell-T helper cell communication**. *Cell* (2019.0) **179** 432-447.e21. DOI: 10.1016/j.cell.2019.09.012 27. Verma M, Bassaganya-Riera J, Leber A, Tubau-Juni N, Hoops S, Abedi V. **High-resolution computational modeling of immune responses in the gut**. *Gigascience* (2019.0) **8**. DOI: 10.1093/gigascience/giz062 28. Lai X, Keller C, Santos G, Schaft N, Dörrie J, Vera J. **Multi-level computational modeling of anti-cancer dendritic cell vaccination utilized to select molecular targets for therapy optimization**. *Front Cell Dev Biol* (2021.0) **9**. DOI: 10.3389/fcell.2021.746359 29. Qiao L, Phipps-Yonas H, Hartmann B, Moran TM, Sealfon SC, Hayot F. **Immune response modeling of interferon beta-pretreated influenza virus-infected human dendritic cells**. *Biophys J* (2010.0) **98**. DOI: 10.1016/j.bpj.2009.10.049 30. Klinke DJ. **A multi-scale model of dendritic cell education and trafficking in the lung: implications for T cell polarization**. *Ann BioMed Eng* (2007.0) **35**. DOI: 10.1007/s10439-007-9318-6 31. Mei Y, Abedi V, Carbo A, Zhang X, Lu P, Philipson C. **Multiscale modeling of mucosal immune responses**. *BMC Bioinf* (2015.0) **16**. DOI: 10.1186/1471-2105-16-S12-S2 32. Schwab JD, Kühlwein SD, Ikonomi N, Kühl M, Kestler HA. **Concepts in boolean network modeling: What do they all mean**. *Comput Struct Biotechnol J* (2020.0) **18**. DOI: 10.1016/j.csbj.2020.03.001 33. Samaga R, Klamt S. **Modeling approaches for qualitative and semi-quantitative analysis of cellular signaling networks**. *Cell Commun Signal* (2013.0) **11**. DOI: 10.1186/1478-811X-11-43 34. Albert R, Thakar J. **Boolean modeling: a logic-based dynamic approach for understanding signaling and regulatory networks and for making useful predictions**. *Wiley Interdiscip Rev Syst Biol Med* (2014.0) **6**. DOI: 10.1002/wsbm.1273 35. Le Novère N. **Quantitative and logic modelling of molecular and gene networks**. *Nat Rev Genet* (2015.0) **16**. DOI: 10.1038/nrg3885 36. Hemedan AA, Niarakis A, Schneider R, Ostaszewski M. **Boolean modelling as a logic-based dynamic approach in systems medicine**. *Comput Struct Biotechnol J* (2022.0) **20**. DOI: 10.1016/j.csbj.2022.06.035 37. Siegle L, Schwab JD, Kühlwein SD, Lausser L, Tümpel S, Pfister AS. **A boolean network of the crosstalk between IGF and wnt signaling in aging satellite cells**. *PloS One* (2018.0) **13**. DOI: 10.1371/journal.pone.0195126 38. Béal J, Pantolini L, Noël V, Barillot E, Calzone L. **Personalized logical models to investigate cancer response to BRAF treatments in melanomas and colorectal cancers**. *PloS Comput Biol* (2021.0) **17**. DOI: 10.1371/journal.pcbi.1007900 39. Puniya BL, Allen L, Hochfelder C, Majumder M, Helikar T. **Systems perturbation analysis of a Large-scale signal transduction model reveals potentially influential candidates for cancer therapeutics**. *Front Bioeng Biotechnol* (2016.0) **4**. DOI: 10.3389/fbioe.2016.00010 40. Abou-Jaoudé W, Traynard P, Monteiro PT, Saez-Rodriguez J, Helikar T, Thieffry D. **Logical modeling and dynamical analysis of cellular networks**. *Front Genet* (2016.0) **7**. DOI: 10.3389/fgene.2016.00094 41. Niarakis A, Helikar T. **A practical guide to mechanistic systems modeling in biology using a logic-based approach**. *Brief Bioinf* (2021.0) **22**. DOI: 10.1093/bib/bbaa236 42. Niarakis A, Bounab Y, Grieco L, Roncagalli R, Hesse A-M, Garin J. **Computational modeling of the main signaling pathways involved in mast cell activation**. *Curr Top Microbiol Immunol* (2014.0) **382** 69-93. DOI: 10.1007/978-3-319-07911-0_4 43. Biane C, Delaplace F, Klaudel H. **Networks and games for precision medicine**. *BioSystems* (2016.0) **150** 52-60. DOI: 10.1016/j.biosystems.2016.08.006 44. Tweedie S, Braschi B, Gray K, Jones TEM, Seal RL, Yates B. **Genenames.org: the HGNC and VGNC resources in 2021**. *Nucleic Acids Res* (2021.0) **49**. DOI: 10.1093/nar/gkaa980 45. Helikar T, Kowal B, McClenathan S, Bruckner M, Rowley T, Madrahimov A. **The cell collective: toward an open and collaborative approach to systems biology**. *BMC Syst Biol* (2012.0) **6**. DOI: 10.1186/1752-0509-6-96 46. Chaouiya C, Bérenguier D, Keating SM, Naldi A, van Iersel MP, Rodriguez N. **SBML qualitative models: a model representation format and infrastructure to foster interactions between qualitative modelling formalisms and tools**. *BMC Syst Biol* (2013.0) **7**. DOI: 10.1186/1752-0509-7-135 47. Keating SM, Waltemath D, König M, Zhang F, Dräger A, Chaouiya C. **SBML level 3: an extensible format for the exchange and reuse of biological models**. *Mol Syst Biol* (2020.0) **16**. DOI: 10.15252/msb.20199110 48. Helikar T, Konvalina J, Heidel J, Rogers JA. **Emergent decision-making in biological signal transduction networks**. *Proc Natl Acad Sci USA* (2008.0) **105**. DOI: 10.1073/pnas.0705088105 49. Helikar T, Rogers JA. **ChemChains: a platform for simulation and analysis of biochemical networks aimed to laboratory scientists**. *BMC Syst Biol* (2009.0) **3**. DOI: 10.1186/1752-0509-3-58 50. Garg A, Di Cara A, Xenarios I, Mendoza L, De Micheli G. **Synchronous versus asynchronous modeling of gene regulatory networks**. *Bioinformatics* (2008.0) **24**. DOI: 10.1093/bioinformatics/btn336 51. Steinman RM, Hawiger D, Liu K, Bonifaz L, Bonnyay D, Mahnke K. **Dendritic cell function**. *Ann N Y Acad Sci* (2003.0) **987** 15-25. DOI: 10.1111/j.1749-6632.2003.tb06029.x 52. Kim R, Emi M, Tanabe K. **Functional roles of immature dendritic cells in impaired immunity of solid tumour and their targeted strategies for provoking tumour immunity**. *Clin Exp Immunol* (2006.0) **146**. DOI: 10.1111/j.1365-2249.2006.03215.x 53. 53 CRAN - package sensitivity. Available at: https://cran.r-project.org/web/packages/sensitivity/index.html (Accessed February 9, 2023).. *CRAN - package sensitivity* 54. Kanehisa M, Furumichi M, Tanabe M, Sato Y, Morishima K. **KEGG: new perspectives on genomes, pathways, diseases and drugs**. *Nucleic Acids Res* (2017.0) **45**. DOI: 10.1093/nar/gkw1092 55. Huang DW, Sherman BT, Tan Q, Kir J, Liu D, Bryant D. **DAVID bioinformatics resources: expanded annotation database and novel algorithms to better extract biology from large gene lists**. *Nucleic Acids Res* (2007.0) **35**. DOI: 10.1093/nar/gkm415 56. Palucka K, Banchereau J. **Dendritic cells: a link between innate and adaptive immunity**. *J Clin Immunol* (1999.0) **19** 12-25. DOI: 10.1023/a:1020558317162 57. Helikar T, Kowal B, Rogers JA. **A cell simulator platform: the cell collective**. *Clin Pharmacol Ther* (2013.0) **93**. DOI: 10.1038/clpt.2013.41 58. Trinchieri G, Sher A. **Cooperation of toll-like receptor signals in innate immune defence**. *Nat Rev Immunol* (2007.0) **7**. DOI: 10.1038/nri2038 59. Schlitzer A, Zhang W, Song M, Ma X. **Recent advances in understanding dendritic cell development, classification, and phenotype. [version 1; peer review: 2 approved]**. *F1000Res* (2018.0) **7**. DOI: 10.12688/f1000research.14793.1 60. Collin M, Bigley V. **Human dendritic cell subsets: an update**. *Immunology* (2018.0) **154** 3-20. DOI: 10.1111/imm.12888 61. Alculumbre S, Raieli S, Hoffmann C, Chelbi R, Danlos F-X, Soumelis V. **Plasmacytoid pre-dendritic cells (pDC): from molecular pathways to function and disease association**. *Semin Cell Dev Biol* (2019.0) **86** 24-35. DOI: 10.1016/j.semcdb.2018.02.014 62. Schraml BU, Reis e Sousa C. **Defining dendritic cells**. *Curr Opin Immunol* (2015.0) **32** 13-20. DOI: 10.1016/j.coi.2014.11.001 63. Swartz AM, Nair SK. **The**. *Curr Protoc* (2022.0) **2**. DOI: 10.1002/cpz1.410 64. Sasaki I, Kato T, Hemmi H, Fukuda-Ohta Y, Wakaki-Nishiyama N, Yamamoto A. **Conventional type 1 dendritic cells in intestinal immune homeostasis**. *Front Immunol* (2022.0) **13**. DOI: 10.3389/fimmu.2022.857954 65. Rhodes JW, Tong O, Harman AN, Turville SG. **Human dendritic cell subsets, ontogeny, and impact on HIV infection**. *Front Immunol* (2019.0) **10**. DOI: 10.3389/fimmu.2019.01088 66. Patente TA, Pinho MP, Oliveira AA, Evangelista GCM, Bergami-Santos PC, Barbuto JAM. **Human dendritic cells: their heterogeneity and clinical application potential in cancer immunotherapy**. *Front Immunol* (2018.0) **9**. DOI: 10.3389/fimmu.2018.03176 67. Heath WR, Carbone FR. **Cross-presentation in viral immunity and self-tolerance**. *Nat Rev Immunol* (2001.0) **1**. DOI: 10.1038/35100512 68. Pulendran B, Tang H, Manicassamy S. **Programming dendritic cells to induce T(H)2 and tolerogenic responses**. *Nat Immunol* (2010.0) **11**. DOI: 10.1038/ni.1894 69. Cella M, Scheidegger D, Palmer-Lehmann K, Lane P, Lanzavecchia A, Alber G. **Ligation of CD40 on dendritic cells triggers production of high levels of interleukin-12 and enhances T cell stimulatory capacity: T-T help**. *J Exp Med* (1996.0) **184**. DOI: 10.1084/jem.184.2.747 70. Kubo T, Hirohashi Y, Torigoe T. **[Dendritic cell and cancer immune checkpoint]**. *Nihon Rinsho Meneki Gakkai Kaishi* (2016.0) **39**. DOI: 10.2177/jsci.39.468 71. Stamatos NM, Carubelli I, van de Vlekkert D, Bonten EJ, Papini N, Feng C. **LPS-induced cytokine production in human dendritic cells is regulated by sialidase activity**. *J Leukoc Biol* (2010.0) **88**. DOI: 10.1189/jlb.1209776 72. Yu L, Wang L, Chen S. **Endogenous toll-like receptor ligands and their biological significance**. *J Cell Mol Med* (2010.0) **14**. DOI: 10.1111/j.1582-4934.2010.01127.x 73. Fonteneau J-F, Larsson M, Beignon A-S, McKenna K, Dasilva I, Amara A. **Human immunodeficiency virus type 1 activates plasmacytoid dendritic cells and concomitantly induces the bystander maturation of myeloid dendritic cells**. *J Virol* (2004.0) **78**. DOI: 10.1128/jvi.78.10.5223-5232.2004 74. Gringhuis SI, van der Vlist M, van den Berg LM, den Dunnen J, Litjens M, Geijtenbeek TBH. **HIV-1 exploits innate signaling by TLR8 and DC-SIGN for productive infection of dendritic cells**. *Nat Immunol* (2010.0) **11**. DOI: 10.1038/ni.1858 75. Gerosa F, Gobbi A, Zorzi P, Burg S, Briere F, Carra G. **The reciprocal interaction of NK cells with plasmacytoid or myeloid dendritic cells profoundly affects innate resistance functions**. *J Immunol* (2005.0) **174**. DOI: 10.4049/jimmunol.174.2.727 76. Schuster S, Hurrell B, Tacchini-Cottier F. **Crosstalk between neutrophils and dendritic cells: a context-dependent process**. *J Leukoc Biol* (2013.0) **94**. DOI: 10.1189/jlb.1012540 77. Megiovanni AM, Sanchez F, Robledo-Sarmiento M, Morel C, Gluckman JC, Boudaly S. **Polymorphonuclear neutrophils deliver activation signals and antigenic molecules to dendritic cells: a new link between leukocytes upstream of T lymphocytes**. *J Leukoc Biol* (2006.0) **79**. DOI: 10.1189/jlb.0905526 78. van Gisbergen KPJM, Sanchez-Hernandez M, Geijtenbeek TBH, van Kooyk Y. **Neutrophils mediate immune modulation of dendritic cells through glycosylation-dependent interactions between mac-1 and DC-SIGN**. *J Exp Med* (2005.0) **201**. DOI: 10.1084/jem.20041276 79. Khorramdelazad H, Kazemi MH, Najafi A, Keykhaee M, Zolfaghari Emameh R, Falak R. **Immunopathological similarities between COVID-19 and influenza: Investigating the consequences of Co-infection**. *Microb Pathog* (2021.0) **152**. DOI: 10.1016/j.micpath.2020.104554 80. Maltezou HC, Theodoridou K, Poland G. **Influenza immunization and COVID-19**. *Vaccine* (2020.0) **38**. DOI: 10.1016/j.vaccine.2020.07.058 81. Kim E-H, Nguyen T-Q, Casel MAB, Rollon R, Kim S-M, Kim Y-I. **Coinfection with Sars-CoV-2 and influenza a virus increases disease severity and impairs neutralizing antibody and CD4+ T cell responses**. *J Virol* (2022.0) **96**. DOI: 10.1128/jvi.01873-21 82. Dadashi M, Khaleghnejad S, Abedi Elkhichi P, Goudarzi M, Goudarzi H, Taghavi A. **COVID-19 and influenza Co-infection: A systematic review and meta-analysis**. *Front Med (Lausanne)* (2021.0) **8**. DOI: 10.3389/fmed.2021.681469 83. Swets MC, Russell CD, Harrison EM, Docherty AB, Lone N, Girvan M. **Sars-CoV-2 co-infection with influenza viruses, respiratory syncytial virus, or adenoviruses**. *Lancet* (2022.0) **399**. DOI: 10.1016/S0140-6736(22)00383-X 84. Alosaimi B, Naeem A, Hamed ME, Alkadi HS, Alanazi T, Al Rehily SS. **Influenza co-infection associated with severity and mortality in COVID-19 patients**. *Virol J* (2021.0) **18** 127. DOI: 10.1186/s12985-021-01594-0 85. Borcherding L, Teksen AS, Grosser B, Schaller T, Hirschbühl K, Claus R. **Impaired dendritic cell homing in COVID-19**. *Front Med (Lausanne)* (2021.0) **8**. DOI: 10.3389/fmed.2021.761372 86. Winheim E, Rinke L, Lutz K, Reischer A, Leutbecher A, Wolfram L. **Impaired function and delayed regeneration of dendritic cells in COVID-19**. *PloS Pathog* (2021.0) **17**. DOI: 10.1371/journal.ppat.1009742 87. Ghanem MH, Shih AJ, Khalili H, Werth EG, Chakrabarty JK, Brown LM. **Proteomic and single-cell transcriptomic dissection of human plasmacytoid dendritic cell response to influenza virus**. *Front Immunol* (2022.0) **13**. DOI: 10.3389/fimmu.2022.814627 88. Wimmers F, Donato M, Kuo A, Ashuach T, Gupta S, Li C. **The single-cell epigenomic and transcriptional landscape of immunity to influenza vaccination**. *Cell* (2021.0) **184** 3915-3935.e21. DOI: 10.1016/j.cell.2021.05.039 89. Onodi F, Bonnet-Madin L, Meertens L, Karpf L, Poirot J, Zhang S-Y. **Sars-CoV-2 induces human plasmacytoid predendritic cell diversification**. *J Exp Med* (2021.0) **218**. DOI: 10.1084/jem.20201387 90. Lee JS, Park S, Jeong HW, Ahn JY, Choi SJ, Lee H. **Immunophenotyping of COVID-19 and influenza highlights the role of type I interferons in development of severe COVID-19**. *Sci Immunol* (2020.0) **5**. DOI: 10.1126/sciimmunol.abd1554 91. Zhu L, Yang P, Zhao Y, Zhuang Z, Wang Z, Song R. **Single-cell sequencing of peripheral mononuclear cells reveals distinct immune response landscapes of COVID-19 and influenza patients**. *Immunity* (2020.0) **53** 685-696.e3. DOI: 10.1016/j.immuni.2020.07.009 92. Zolfaghari Emameh R, Nosrati H, Eftekhari M, Falak R, Khoshmirsafa M. **Expansion of single cell transcriptomics data of SARS-CoV infection in human bronchial epithelial cells to COVID-19**. *Biol Proced Online* (2020.0) **22** 16. DOI: 10.1186/s12575-020-00127-3 93. Ding Q, Lu P, Fan Y, Xia Y, Liu M. **The clinical characteristics of pneumonia patients coinfected with 2019 novel coronavirus and influenza virus in wuhan, China**. *J Med Virol* (2020.0) **92**. DOI: 10.1002/jmv.25781 94. Alculumbre SG, Saint-André V, Di Domizio J, Vargas P, Sirven P, Bost P. **Diversification of human plasmacytoid predendritic cells in response to a single stimulus**. *Nat Immunol* (2018.0) **19** 63-75. DOI: 10.1038/s41590-017-0012-z 95. Kong Y, Wang Y, Wu X, Han J, Li G, Hua M. **Storm of soluble immune checkpoints associated with disease severity of COVID-19**. *Signal Transduct Target Ther* (2020.0) **5** 192. DOI: 10.1038/s41392-020-00308-2 96. Vanderbeke L, Van Mol P, Van Herck Y, De Smet F, Humblet-Baron S, Martinod K. **Monocyte-driven atypical cytokine storm and aberrant neutrophil activation as key mediators of COVID-19 disease severity**. *Nat Commun* (2021.0) **12** 4117. DOI: 10.1038/s41467-021-24360-w 97. Cantuti-Castelvetri L, Ojha R, Pedro LD, Djannatian M, Franz J, Kuivanen S. **Neuropilin-1 facilitates Sars-CoV-2 cell entry and infectivity**. *Science* (2020.0) **370**. DOI: 10.1126/science.abd2985 98. Gudowska-Sawczuk M, Mroczko B. **The role of neuropilin-1 (NRP-1) in Sars-CoV-2 infection: Review**. *J Clin Med* (2021.0) **10**. DOI: 10.3390/jcm10132772 99. Saichi M, Ladjemi MZ, Korniotis S, Rousseau C, Ait Hamou Z, Massenet-Regad L. **Single-cell RNA sequencing of blood antigen-presenting cells in severe COVID-19 reveals multi-process defects in antiviral immunity**. *Nat Cell Biol* (2021.0) **23**. DOI: 10.1038/s41556-021-00681-2 100. Alfei F, Ho P-C, Lo W-L. **DCision-making in tumors governs T cell anti-tumor immunity**. *Oncogene* (2021.0) **40**. DOI: 10.1038/s41388-021-01946-8 101. Waldman AD, Fritz JM, Lenardo MJ. **A guide to cancer immunotherapy: from T cell basic science to clinical practice**. *Nat Rev Immunol* (2020.0) **20**. DOI: 10.1038/s41577-020-0306-5 102. Kang R, Zhang Q, Zeh HJ, Lotze MT, Tang D. **HMGB1 in cancer: good, bad, or both**. *Clin Cancer Res* (2013.0) **19**. DOI: 10.1158/1078-0432.CCR-13-0495 103. Demoulin S, Herfs M, Somja J, Roncarati P, Delvenne P, Hubert P. **HMGB1 secretion during cervical carcinogenesis promotes the acquisition of a tolerogenic functionality by plasmacytoid dendritic cells**. *Int J Cancer* (2015.0) **137**. DOI: 10.1002/ijc.29389 104. Tripathi A, Shrinet K, Kumar A. **HMGB1 protein as a novel target for cancer**. *Toxicol Rep* (2019.0) **6**. DOI: 10.1016/j.toxrep.2019.03.002 105. Muire PJ, Schwacha MG, Wenke JC. **Systemic T cell exhaustion dynamics is linked to early high mobility group box protein 1 (HMGB1) driven hyper-inflammation in a polytrauma rat model**. *Cells* (2021.0) **10**. DOI: 10.3390/cells10071646 106. Zarour HM. **Reversing T-cell dysfunction and exhaustion in cancer**. *Clin Cancer Res* (2016.0) **22**. DOI: 10.1158/1078-0432.CCR-15-1849 107. Marin-Acevedo JA, Kimbrough EO, Lou Y. **Next generation of immune checkpoint inhibitors and beyond**. *J Hematol Oncol* (2021.0) **14**. DOI: 10.1186/s13045-021-01056-8 108. Li B, Chan HL, Chen P. **Immune checkpoint inhibitors: basics and challenges**. *Curr Med Chem* (2019.0) **26**. DOI: 10.2174/0929867324666170804143706 109. Robert C. **A decade of immune-checkpoint inhibitors in cancer therapy**. *Nat Commun* (2020.0) **11** 3801. DOI: 10.1038/s41467-020-17670-y 110. Messmer D, Yang H, Telusma G, Knoll F, Li J, Messmer B. **High mobility group box protein 1: an endogenous signal for dendritic cell maturation and Th1 polarization**. *J Immunol* (2004.0) **173**. DOI: 10.4049/jimmunol.173.1.307 111. Zhang Y, Zheng J. **Functions of immune checkpoint molecules beyond immune evasion**. *Adv Exp Med Biol* (2020.0) **1248**. DOI: 10.1007/978-981-15-3266-5_9 112. Iguchi-Manaka A, Kai H, Yamashita Y, Shibata K, Tahara-Hanaoka S, Honda S. **Accelerated tumor growth in mice deficient in DNAM-1 receptor**. *J Exp Med* (2008.0) **205**. DOI: 10.1084/jem.20081611 113. Braun M, Aguilera AR, Sundarrajan A, Corvino D, Stannard K, Krumeich S. **CD155 on tumor cells drives resistance to immunotherapy by inducing the degradation of the activating receptor CD226 in CD8+ t cells**. *Immunity* (2020.0) **53** 805-823.e15. DOI: 10.1016/j.immuni.2020.09.010 114. Weulersse M, Asrir A, Pichler AC, Lemaitre L, Braun M, Carrié N. **Eomes-dependent loss of the Co-activating receptor CD226 restrains CD8+ T cell anti-tumor functions and limits the efficacy of cancer immunotherapy**. *Immunity* (2020.0) **53** 824-839.e10. DOI: 10.1016/j.immuni.2020.09.006 115. Simon S, Labarriere N. **PD-1 expression on tumor-specific T cells: Friend or foe for immunotherapy**. *Oncoimmunology* (2017.0) **7**. DOI: 10.1080/2162402X.2017.1364828 116. Shi L, Chen S, Yang L, Li Y. **The role of PD-1 and PD-L1 in T-cell immune suppression in patients with hematological malignancies**. *J Hematol Oncol* (2013.0) **6**. DOI: 10.1186/1756-8722-6-74 117. Bigley V, Barge D, Collin M. **Dendritic cell analysis in primary immunodeficiency**. *Curr Opin Allergy Clin Immunol* (2016.0) **16**. DOI: 10.1097/ACI.0000000000000322 118. Bandola-Simon J, Roche PA. **Dysfunction of antigen processing and presentation by dendritic cells in cancer**. *Mol Immunol* (2019.0) **113**. DOI: 10.1016/j.molimm.2018.03.025 119. Bieber K, Autenrieth SE. **Dendritic cell development in infection**. *Mol Immunol* (2020.0) **121**. DOI: 10.1016/j.molimm.2020.02.015 120. Bates J, Diehl L. **Dendritic cells in IBD pathogenesis: an area of therapeutic opportunity**. *J Pathol* (2014.0) **232**. DOI: 10.1002/path.4277 121. Sasindran SJ, Torrelles JB. **Mycobacterium tuberculosis infection and inflammation: what is beneficial for the host and for the bacterium**. *Front Microbiol* (2011.0) **2**. DOI: 10.3389/fmicb.2011.00002 122. Helikar T, Kochi N, Kowal B, Dimri M, Naramura M, Raja SM. **A comprehensive, multi-scale dynamical model of ErbB receptor signal transduction in human mammary epithelial cells**. *PloS One* (2013.0) **8**. DOI: 10.1371/journal.pone.0061757 123. Hernandez C, Thomas-Chollier M, Naldi A, Thieffry D. **Computational verification of Large logical models-application to the prediction of T cell response to checkpoint inhibitors**. *Front Physiol* (2020.0) **11**. DOI: 10.3389/fphys.2020.558606 124. Saez-Rodriguez J, Simeoni L, Lindquist JA, Hemenway R, Bommhardt U, Arndt B. **A logical model provides insights into T cell receptor signaling**. *PloS Comput Biol* (2007.0) **3**. DOI: 10.1371/journal.pcbi.0030163 125. Montagud A, Béal J, Tobalina L, Traynard P, Subramanian V, Szalai B. **Patient-specific boolean models of signalling networks guide personalised treatments**. *eLife* (2022.0) **11**. DOI: 10.7554/eLife.72626 126. Wynn ML, Consul N, Merajver SD, Schnell S. **Logic-based models in systems biology: a predictive and parameter-free network analysis method**. *Integr Biol (Camb)* (2012.0) **4**. DOI: 10.1039/c2ib20193c 127. Yang H, Wang H, Andersson U. **Targeting inflammation driven by HMGB1**. *Front Immunol* (2020.0) **11**. DOI: 10.3389/fimmu.2020.00484 128. Lee S-A, Kwak MS, Kim S, Shin J-S. **The role of high mobility group box 1 in innate immunity**. *Yonsei Med J* (2014.0) **55**. DOI: 10.3349/ymj.2014.55.5.1165 129. Yu M, Wang H, Ding A, Golenbock DT, Latz E, Czura CJ. **HMGB1 signals through toll-like receptor (TLR) 4 and TLR2**. *Shock* (2006.0) **26**. DOI: 10.1097/01.shk.0000225404.51320.82 130. Mougel A, Terme M, Tanchot C. **Therapeutic cancer vaccine and combinations with antiangiogenic therapies and immune checkpoint blockade**. *Front Immunol* (2019.0) **10**. DOI: 10.3389/fimmu.2019.00467 131. Zhao J, Chen Y, Ding Z-Y, Liu J-Y. **Safety and efficacy of therapeutic cancer vaccines alone or in combination with immune checkpoint inhibitors in cancer treatment**. *Front Pharmacol* (2019.0) **10**. DOI: 10.3389/fphar.2019.01184 132. Wang D, Huang XF, Hong B, Song X-T, Hu L, Jiang M. **Efficacy of intracellular immune checkpoint-silenced DC vaccine**. *JCI Insight* (2018.0) **3**. DOI: 10.1172/jci.insight.98368 133. Kong BY, Bolton H, Kim JW, Silveira PA, Fromm PD, Clark GJ. **On the other side: manipulating the immune checkpoint landscape of dendritic cells to enhance cancer immunotherapy**. *Front Oncol* (2019.0) **9**. DOI: 10.3389/fonc.2019.00050 134. Jin H-S, Park Y. **Hitting the complexity of the TIGIT-CD96-CD112R-CD226 axis for next-generation cancer immunotherapy**. *BMB Rep* (2021.0) **54** 2-11. DOI: 10.5483/BMBRep.2021.54.1.229 135. Moon J, Oh YM, Ha S-J. **Perspectives on immune checkpoint ligands: expression, regulation, and clinical implications**. *BMB Rep* (2021.0) **54**. DOI: 10.5483/BMBRep.2021.54.8.054 136. Clark M, Kroger CJ, Tisch RM. **Type 1 diabetes: A chronic anti-Self-Inflammatory response**. *Front Immunol* (2017.0) **8**. DOI: 10.3389/fimmu.2017.01898 137. Mollah ZUA, Pai S, Moore C, O’Sullivan BJ, Harrison MJ, Peng J. **Abnormal NF-kappa b function characterizes human type 1 diabetes dendritic cells and monocytes**. *J Immunol* (2008.0) **180**. DOI: 10.4049/jimmunol.180.5.3166 138. Babon JAB, DeNicola ME, Blodgett DM, Crèvecoeur I, Buttrick TS, Maehr R. **Analysis of self-antigen specificity of islet-infiltrating T cells from human donors with type 1 diabetes**. *Nat Med* (2016.0) **22**. DOI: 10.1038/nm.4203 139. Grohová A, Dáňová K, Špíšek R, Palová-Jelínková L. **Cell based therapy for type 1 diabetes: should we take hyperglycemia into account**. *Front Immunol* (2019.0) **10**. DOI: 10.3389/fimmu.2019.00079 140. Hotta-Iwamura C, Tarbell KV. **Type 1 diabetes genetic susceptibility and dendritic cell function: potential targets for treatment**. *J Leukoc Biol* (2016.0) **100** 65-80. DOI: 10.1189/jlb.3MR1115-500R
--- title: 'Impact of race on outcomes from catheter ablation of ventricular tachycardia in structural heart disease: A prospective registry from south metropolitan Chicago' authors: - Nathan W. Kong - Dalise Y. Shatz - Stephanie A. Besser - Gaurav A. Upadhyay - Roderick Tung journal: Heart Rhythm O2 year: 2023 pmcid: PMC10041081 doi: 10.1016/j.hroo.2023.01.007 license: CC BY 4.0 --- # Impact of race on outcomes from catheter ablation of ventricular tachycardia in structural heart disease: A prospective registry from south metropolitan Chicago ## Body Key Findings ▪Black patients undergoing ventricular tachycardia (VT) ablation had higher unadjusted risk of VT recurrence at follow-up.▪Black patients undergoing VT ablation had higher rates of hypertension and chronic kidney disease, and were more likely to present in VT storm.▪After adjustment for comorbidities including hypertension, chronic kidney disease, and VT storm, there was no difference in VT recurrence between racial groups. ## Abstract ### Background Whether racial disparities in outcomes are present after catheter ablation for scar-related ventricular tachycardia (VT) is not known. ### Objective The purpose of this study was to examine whether racial differences exist in outcomes for patients undergoing VT ablation. ### Methods From March 2016 through April 2021, consecutive patients undergoing catheter ablation for scar-related VT at the University of Chicago were prospectively enrolled. The primary outcome was VT recurrence, with secondary outcome of mortality alone and composite endpoint of left ventricular assist device placement, heart transplant, or mortality. ### Results A total of 258 patients were analyzed: 58 ($22\%$) self-identified as Black, and 113 ($44\%$) had ischemic cardiomyopathy. Black patients had significantly higher rates of hypertension (HTN), chronic kidney disease (CKD), and VT storm at presentation. At 7 months, Black patients experienced higher rates of VT recurrence ($$P \leq .009$$). However, after multivariable adjustment, there were no observed differences in VT recurrence (adjusted hazard ratio [aHR] 1.65; $95\%$ confidence interval [CI] 0.91–2.97; $$P \leq .10$$), all-cause mortality (aHR 0.49; $95\%$ CI 0.21–1.17; $$P \leq .11$$), or composite events (aHR 0.76; $95\%$ CI 0.37–1.54; $$P \leq .44$$) between Black and non-Black patients. ### Conclusion In this diverse prospective registry of patients undergoing catheter ablation for scar-related VT, Black patients experienced higher rates of VT recurrence compared to non-Black patients. When adjusted for highly prevalent HTN, CKD, and VT storm, Black patients had comparable outcomes as non-Black patients. ## Introduction Catheter ablation is the treatment of choice for symptomatic, sustained monomorphic ventricular tachycardia (VT).1 Randomized clinical trials and prospective cohort studies have demonstrated efficacy with catheter ablation for VT in the setting of structural heart disease.2, 3, 4 Previous studies have highlighted important differences between men and women undergoing VT ablation, and racial differences have been described in outcomes after catheter ablation for other arrhythmias such as atrial fibrillation.5,6 Previous studies of VT ablation have not included a sufficient number of racial minorities, particularly patients who self-identify as Black, to examine whether any meaningful differences in outcomes exist between races. We analyzed a single-center prospective registry from south metropolitan Chicago to determine whether racial disparities exist between patients who identify as Black and other groups undergoing catheter ablation for scar-related VT. ## Patient selection Consecutive patients between March 2016 and April 2021 with structural heart disease and evidence of electroanatomic scar, defined as low voltage <1.5 mV, were included for the present analysis. Patients with ischemic cardiomyopathy (ICM) and nonischemic cardiomyopathy (NICM) (including arrhythmogenic right ventricular cardiomyopathy) were included. All patients were prospectively enrolled in the University of Chicago VT Ablation Registry (IRB16-0272), which was established to examine the safety and outcomes during and after catheter ablation. If patients underwent multiple ablations, outcomes after the most recent procedure were reported. The University of Chicago Medical Center Institutional Review Board approved the creation, maintenance, and review of this prospective registry. All subjects provided informed consent. The research reported adhered to the Declaration of Helsinki as revised in 2013. Patients without any follow-up were excluded from analysis. ## Catheter ablation Noninvasive stimulation was performed with patients under light conscious sedation before induction of anesthesia to assess the morphology of a clinical or targeted VT and hemodynamic tolerance. General anesthesia with intubation was administered in all cases. Epicardial mapping and ablation was performed at the discretion of the operator, typically for cases with a history of previously failed endocardial ablation or cardiac magnetic resonance imaging that suggested epicardial delayed enhancement. Intravenous dopamine infusions (5–10 μg/min) were initiated if hypotension occurred or in cases with systolic blood pressure <110 mm Hg before programmed stimulation. High-density electroanatomic maps were created in sinus rhythm or paced rhythm with multielectrode catheters using CARTO (PentaRay, Biosense Webster, Diamond Bar, CA), EnSite Precision (2-2-2 Livewire, Abbott, Abbott Park, IL), or Rhythmia (Orion, Boston Scientific, Natick, MA) with standard low-voltage bipolar settings (0.5–1.5 mV; 0.5–1.0 mV for Rhythmia). Radiofrequency ablation was performed using an open-irrigated catheter (ThermoCool or ThermoCool SF, 3.5 mm, Biosense-Webster; FlexAbility SE, Abbott; IntelliNav, Boston Scientific). As previously described, high-density mapping with multielectrode catheters was performed to identify wavefront discontinuities in regions with isochronal crowding. Isochronal late activation mapping was performed to annotate local electrograms at offset of latest component. Ablation was performed until local electrograms were reduced or eliminated within deceleration zones with complete noninducibility as the procedural endpoint.7 ## Clinical data and follow-up Sex was defined as the one assigned at birth (either male or female). Race was self-identified at the time of registration. Height and weight were determined at the time of VT ablation. Body mass index (BMI) was calculated as the weight (in kilograms) divided by the height (in meters) squared. Obesity was defined as BMI ≥30 kg/m2. Left ventricular ejection fraction (LVEF) was determined from the most recent transthoracic echocardiogram before the VT ablation date. VT storm was defined as ≥3 episodes of an episode of VT requiring termination with cardioversion or antitachycardia pacing at the time of presentation. The etiology of the patient’s cardiomyopathy was determined by the treating physician based on clinical history. Patients were followed routinely with clinical history, physical examination, and implantable cardioverter-defibrillator interrogation within the first month after ablation and every 3–6 months thereafter. VT recurrence was defined as documented sustained monomorphic VT >30 seconds in duration or any appropriate implantable cardioverter-defibrillator therapy with antitachycardia pacing or delivery of shock. Left ventricular assist device (LVAD) implantation, heart transplant, and all-cause death were confirmed with electronic health records, the referring physician, or family members. Patients were followed until VT recurrence, LVAD implantation, heart transplant, death, or most recent clinic visit, as time to first event analysis. Prespecified subgroup analysis by ICM, NICM, and LVEF <$40\%$ also was performed. The primary outcome of interest was VT recurrence. The secondary outcome was composite and individual occurrence of LVAD implantation, heart transplant, and all-cause death. ## Statistical analysis For baseline and clinical follow-up outcomes, categorical variables are given as count (percentage), and continuous variables are given as either mean ± SD (if normally distributed) or median [interquartile ranges] (if non-normally distributed). Variables were compared using the χ2 test of association or Fisher exact test for categorical variables and Student t test for continuous variables with normal distributions. Continuous variables with non-normal distributions were compared using the Mann-Whitney U test. Cox proportion hazard models were created for primary and secondary outcomes. The proportion hazards assumption was checked using the Schoenfeld residual method. Results were adjusted for age, sex, history of hypertension, history of chronic kidney disease (CKD), procedural duration, and presentation of VT storm. Kaplan-Meier survival analysis was performed on primary and secondary outcomes as well as prespecified subgroup analysis for the year following catheter ablation. All 2-tailed $P \leq .05$ were considered significant. All analyses were performed using Stata Version 17 (StataCorp., College Station, TX), and visualizations were performed with ‘tidyverse’ and ‘survival’ packages in R 4.1.2 (R Core Team, 2021). ## Baseline characteristics Between March 2016 and April 2021, 289 ablation procedures for scar-related VT were performed at the University of Chicago Medical Center. After exclusion of repeat procedures, 258 patients ($89.3\%$ of total procedures) remained for final analysis. Median age was 65 [58–71] years, and 39 patients ($15\%$) were female. One hundred eighty-nine patients ($73\%$) self-identified as non-Hispanic White, 58 ($22\%$) self-identified as Black, 6 ($2\%$) self-identified as Hispanic White, and 5 ($2\%$) self-identified as Asian/Pacific Islander. Median LVEF was $31\%$ [$25\%$–$42\%$]. One hundred thirteen patients ($44\%$) had ICM as the etiology of their cardiomyopathy. One hundred twenty-two patients ($47\%$) previously had undergone a VT ablation procedure, and 37 ($14\%$) had undergone ≥2 previous ablation attempts. One hundred t patients ($43\%$) presented as VT storm. Median follow-up time was 6.5 [1–19] months. Stratified by self-identified race, Black patients were more likely to have hypertension ($66\%$ vs $40\%$; $$P \leq .001$$), CKD ($55\%$ vs $25\%$; $P \leq .001$), and present as VT storm ($57\%$ vs $40\%$; $$P \leq .028$$) (Table 1). There were no significant differences with regard to age, sex, BMI, baseline LVEF, etiology of cardiomyopathy, previous antiarrhythmic medications, number of previous VT ablations, or total procedural time. Table 1Baseline characteristics by self-identified raceBlackNon-BlackP valueNo. of patients58200Age (y)62.5 [53.5–69.0]65.0 [59.0–71.0].128Male46 (79.3)173 (86.5).255Body mass index (kg/m2)28.7 [25.4–32.4]28.4 [25.2–32.2].696Obesity (BMI >30 kg/m2)23 (43.4)74 (38.7).650Left ventricular ejection fraction (%)30.5 [24.1–40.1]31.9 [25.0–44.2].368Ischemic cardiomyopathy26 (44.8)87 (43.5).977Nonischemic cardiomyopathy32 (55.2)113 (56.5).977Arrhythmogenic right ventricular cardiomyopathy2 (3.4)16 (8.0).365Hypertension38 (65.5)79 (39.5).001Diabetes mellitus14 (24.1)43 (21.5).805Coronary artery disease32 (55.2)86 (43.0).137Chronic kidney disease32 (55.2)49 (24.5)<.001Stroke or transient ischemic attack5 (8.6)11 (5.5).366Atrial fibrillation or atrial flutter18 (31.0)72 (36.0).588VT storm33 (56.9)79 (39.5).028No. of previous VT ablations.831 034 (58.6)102 (51.0) 117 (29.3)68 (34.0) 24 (6.9)17 (8.5) ≥33 (5.2)13 (6.5)Previous β-blocker42 (72.4)121 (60.5).133Previous antiarrhythmic medication33 (56.9)114 (57.0)1 Previous amiodarone25 (43.1)75 (37.5).536Implantable cardioverter-defibrillator.224 Biventricular10 (17.2)54 (27.0) Dual-chamber30 (51.7)80 (40.0) Single-chamber12 (20.7)44 (22.0) Subcutaneous0 (0.0)7 (3.5) None6 (10.3)15 (7.5)Mapping access.556 Endocardial only31 (53.4)92 (46.0) Epicardial only2 (3.4)11 (5.5) Endocardial and epicardial25 (43.1)97 (48.5)Procedural time (min)300 [255–382]289.0 [225–362].140Fluoroscopy time (min)25.0 [14.5–37.2]18.6 [10.6–30.6].123Ablation time (s)1672 [1081–2352]1423 [1024–2101].337I-VT score–VT recurrence risk.383 High risk13 (22.4)35 (17.5) Intermediate risk29 (50.0)91 (45.5) Low risk16 (27.6)74 (37.0)I-VT score–Mortality risk.184 High risk18 [31]43 (21.5) Low risk40 (69.0)157 (78.5)Categorical variables are given as count (percent). Normally distributed continuous variables are given as mean. Non-normally distributed continuous variables are given as median [interquartile range].Bold values indicate statistically significant. BMI = body mass index; I-VT = International Ventricular Tachycardia; VT = ventricular tachycardia. ## VT recurrence In total, 66 patients ($26\%$) experienced VT recurrences. Black patients were significantly more likely to have VT recurrence compared to other races ($40\%$ vs $22\%$; $$P \leq .009$$). In unadjusted time-to-event analysis, Black patients had significantly a higher rate of VT recurrence ($$P \leq .013$$) in the year after ablation (Figure 1A). Kaplan-Meier subgroup analysis showed no significant difference in VT recurrence in patients with LVEF ≤$40\%$ (Figure 1B) and those with NICM (Figure 1D). In those with ICM, Black patients experienced a lower rate of VT-free survival compared to other races ($$P \leq .021$$) (Figure 1C). Black patients were at higher risk for VT recurrence compared to other races in unadjusted hazard analysis (hazard ratio 1.83; $95\%$ confidence interval [CI] 1.10–3.04; $$P \leq .020$$) but had no significant difference after multivariate adjustment (adjusted hazard ratio 1.65; $95\%$ CI 0.91–2.97; $$P \leq .1.00$$) (Table 2).Figure 1Freedom from ventricular tachycardia (VT) recurrence by self-identified race (A) and subgroup analysis of ejection fraction ≤$40\%$ (B), ischemic cardiomyopathy (C), and nonischemic cardiomyopathy (D).Table 2Primary and secondary outcomes stratified by self-identified raceBlackNon-BlackP valueNo. of patients58200VT recurrence23 (39.7)43 (21.5).009Left ventricular assist device4 (6.9)7 (3.5).273Heart transplant4 (6.9)4 (2.0).079All-cause death7 (12.1)30 (15.1).718Composite events14 (24.1)36 (18.2).413Crude hazard ratio for VT recurrence1.83 (1.10–3.04)Ref.020Adjusted∗ hazard ratio for VT recurrence1.65 (0.91–2.97)Ref.100Crude hazard ratio for all-cause death0.56 (0.25–1.34)Ref.200Adjusted∗ hazard ratio for all-cause death0.49 (0.21–1.17)Ref.107Crude hazard ratio for composite event1.28 (0.69–2.38)Ref.431Adjusted∗ hazard ratio for composite event0.76 (0.37–1.54)Ref.444Cox proportional hazard models for ventricular tachycardia (VT) recurrence and composite events by self-identified race. Values are given as n (%) or hazard ratio ($95\%$ confidence interval) unless otherwise indicated.∗Results are adjusted for age, sex, history of hypertension, history of chronic kidney disease, presentation as VT storm, and total procedural time. ## All-cause death, LVAD, and heart transplant Black patients had no difference in rates of all-cause mortality ($12\%$ vs $15\%$; $$P \leq .718$$) or rates of composite events ($24\%$ vs $18\%$; $$P \leq .413$$) compared to non-Black patients (Table 2). There was no statistically significant difference in crude or multivariate adjusted hazard ratios of all-cause death or composite events between races (Table 2). There was no significant difference in all-cause mortality between self-identified races in the year after ablation in the overall cohort and subgroup analysis (Figure 2). No differences were observed in the overall cohort or subgroups in the year after ablation between races for risk of composite events (Figure 3).Figure 2Freedom from all-cause mortality by self-identified race (A) and subgroup analysis of ejection fraction ≤$40\%$ (B), ischemic cardiomyopathy (C), and nonischemic cardiomyopathy (D).Figure 3Composite event-free survival by self-identified race (a) and subgroup analysis of ejection fraction ≤$40\%$ (B), ischemic cardiomyopathy (C), and nonischemic cardiomyopathy (D). ## Discussion In this single-center prospective registry of consecutive patients undergoing scar-related VT catheter ablation at a tertiary academic medical center in south metropolitan Chicago, we found that patients who self-identified as Black were significantly more likely to experience VT recurrence after catheter ablation. However, after multivariate adjustment, no racial disparities in outcomes were observed. Additionally, the increased rate of VT recurrence was not at the expense of higher mortality or composite outcomes of death, transplant, or LVAD. This is one of the first analyses of VT ablation outcomes stratified by race and adds to the growing body of literature highlighting racial disparities in outcomes across the spectrum of cardiovascular disease.8, 9, 10, 11 The results of this study are similar to previous studies showing differences in outcomes by races from other complex arrhythmia ablations, particularly atrial fibrillation.5 There are a few potential explanations for the relationship seen in this study. First, Black patients may have been more likely to present or be referred late, as they more often presented with VT storm compared to non-Black patient ($57\%$ vs $40\%$; $$P \leq .028$$). However, the number of previous VT ablations and previous medication usage were not different between the races. Second, VT ablation in Black patients seemed to be more technically challenging with more extensive substrate, with trends toward longer total procedural times (300 vs 289 minutes), longer fluoroscopy time (25 vs 19 minutes), and longer ablation times (1672 vs 1423 seconds). There was no statistically significant difference when stratified by International Ventricular Tachycardia (I-VT) risk score (Table 1).12 However, $22\%$ of Black patients were considered at high risk for VT recurrence by the I-VT score compared to $18\%$ of non-Black patients. Similarly, $31\%$ of Black patients were considered at high risk for mortality compared to $22\%$ of non-Black patients. Additionally, the proportion of NICM, which traditionally portends worse outcomes, was similar between the races. Importantly, Black patients had a significantly higher prevalence of comorbidities, specifically hypertension ($66\%$ vs $40\%$) and CKD ($55\%$ vs $25\%$). This finding suggests that comorbid conditions such as hypertension and CKD may play a role in VT substrates as well as outcomes after VT ablation. This is especially true given that after adjustment for rates of hypertension, CKD, presentation of VT storm, and total procedural time, there was no statistically significant difference in VT recurrence. It is worth highlighting that there were no differences in other clinical outcomes such as all-cause mortality or rates of heart transplant or LVAD implantation. In fact, Black patients had a trend toward improved survival compared to non-Black patients (adjusted hazard ratio 0.49; $95\%$ CI 0.21–1.17; $$P \leq .107$$). When stratified by etiology of cardiomyopathy, Black patients were more likely to have VT recurrence in the ICM group only. We hypothesize that the higher rates of hypertension and CKD in Black patients were more likely to contribute to VT recurrence in the ICM group because both comorbidities are known factors that increase morbidity and mortality in patients with ICM.13,14 Additionally, comorbidities such as hypertension, CKD, and obesity have been shown to increase the risk of recurrence after other catheter ablations, such as pulmonary vein isolation for atrial fibrillation.15, 16, 17 In patients with NICM, no differences by race were observed in VT recurrence or all-cause death. However, there was a trend toward decreased freedom from composite events in Black patients vs non-Black patients ($$P \leq .067$$) (Figure 3). Additionally, there were trends toward higher rates of heart transplant ($7\%$ vs $2\%$) and LVAD implantation ($7\%$ vs $4\%$) between Black patients and other races within the overall cohort. These findings likely reflect the hypothesis that Black patients had more advanced cardiomyopathies at the time of catheter ablation. Given that race in VT study populations is rarely reported, to the best of our knowledge this is the first cohort analysis to examine VT outcome differences by race. The University of Chicago Medical *Center is* distinctive in that is serves the local south metropolitan community with a high proportion of Black patients and whose primary insurer is Medicaid. Median household income of the top 5 zip codes served at the University of Chicago Medical *Center is* $37,948. Additionally, previous estimates indicate that >$60\%$ of health care encounters at the University of Chicago Medical Center are with Black patients and $64\%$ are with patients whose primary insurer is Medicaid.18 Despite this, the percentage of Black patients in our cohort was $22\%$, which is significantly lower than across the University of Chicago Medical Center, with previous estimates indicating that >$60\%$ of health care encounters are with Black patients.18 The reason for this disproportionate referral pattern warrants further investigation. Overall, these data highlight the need for improved access to advanced VT therapies to reduce disparities in outcome, even with higher rates of comorbidities. ## Study limitations First, our study was conducted from a single center with a unique patient demographic at a tertiary referral site, thus making the external generalizability limited. Larger multicenter series have not previously captured race. Second, patients were enrolled in the study at the time of catheter ablation but had variable and limited follow-up. Thirty-one patients ($13\%$) were lost to follow-up, defined as <7 days of follow-up without an event. We attempted to mitigate loss to follow-up through regular phone calls and remote implantable device monitoring. Lastly, although we attempted to adjust for known covariates, there likely remain unknown confounders that may play a part in the differences observed in our study population. Multivariate adjustment does not necessarily separate the impact of medical and biological comorbidities from race and socioeconomic equity. ## Conclusion In this diverse prospective registry of patients undergoing catheter ablation for scar-related VT, Black patients experienced higher rates of VT recurrence compared to non-Black patients. When adjusted for highly prevalent HTN, CKD, and VT storm, Black patients had comparable outcomes as non-Black patients. ## Funding Sources The authors have no funding sources to disclose. ## Disclosures The authors have no conflicts of interest to disclose. ## Authorship All authors attest they meet the current ICMJE criteria for authorship. ## Patient Consent All patients provided written informed consent. ## Ethics Statement The University of Chicago Medical Center Institutional Review Board approved the creation, maintenance, and review of this prospective registry. The research reported adhered to the Declaration of Helsinki as revised in 2013. ## Data availability The datasets generated and used during the current study are available from the corresponding author on reasonable request. ## References 1. Cronin E.M., Bogun F.M., Maury P.. **2019 HRS/EHRA/APHRS/LAHRS expert consensus statement on catheter ablation of ventricular arrhythmias**. *Europace* (2019) **21** 1143-1144. PMID: 31075787 2. Kuck K.H., Schaumann A., Eckardt L.. **Catheter ablation of stable ventricular tachycardia before defibrillator implantation in patients with coronary heart disease (VTACH): a multicentre randomised controlled trial**. *Lancet* (2010) **375** 31-40. PMID: 20109864 3. Reddy V.Y., Reynolds M.R., Neuzil P.. **Prophylactic catheter ablation for the prevention of defibrillator therapy**. *N Engl J Med* (2007) **357** 2657-2665. PMID: 18160685 4. Stevenson W.G., Wilber D.J., Natale A.. **Irrigated radiofrequency catheter ablation guided by electroanatomic mapping for recurrent ventricular tachycardia after myocardial infarction: the multicenter ThermoCool ventricular tachycardia ablation trial**. *Circulation* (2008) **118** 2773-2782. PMID: 19064682 5. Patel N., Deshmukh A., Thakkar B.. **Gender, race, and health insurance status in patients undergoing catheter ablation for atrial fibrillation**. *Am J Cardiol* (2016) **117** 1117-1126. PMID: 26899494 6. Frankel D.S., Tung R., Santangeli P.. **Sex and catheter ablation for ventricular tachycardia: an international ventricular tachycardia ablation center collaborative group study**. *JAMA Cardiol* (2016) **1** 938-944. PMID: 27556589 7. Aziz Z., Shatz D., Raiman M.. **Targeted ablation of ventricular tachycardia guided by wavefront discontinuities during sinus rhythm: a new functional substrate mapping strategy**. *Circulation* (2019) **140** 1383-1397. PMID: 31533463 8. Magnani J.W., Norby F.L., Agarwal S.K.. **Racial differences in atrial fibrillation-related cardiovascular disease and mortality: the Atherosclerosis Risk in Communities (ARIC) study**. *JAMA Cardiol* (2016) **1** 433-441. PMID: 27438320 9. Pool L.R., Ning H., Lloyd-Jones D.M., Allen N.B.. **Trends in racial/ethnic disparities in cardiovascular health among US adults from 1999-2012**. *J Am Heart Assoc* (2017) **6** 10. Yong C.M., Ungar L., Abnousi F., Asch S.M., Heidenreich P.A.. **Racial differences in quality of care and outcomes after acute coronary syndrome**. *Am J Cardiol* (2018) **121** 1489-1495. PMID: 29655881 11. Carnethon M.R., Pu J., Howard G.. **Cardiovascular health in African Americans: a scientific statement from the American Heart Association**. *Circulation* (2017) **136** e393-e423. PMID: 29061565 12. Vergara P., Tzou W.S., Tung R.. **Predictive score for identifying survival and recurrence risk profiles in patients undergoing ventricular tachycardia ablation: the I-VT score**. *Circ Arrhythm Electrophysiol* (2018) **11** 13. Soltero E.R., Petersen N.J., Earle N.R., Glaeser D.H., Urbauer D.L., Deswal A.. **Long-term results of coronary artery bypass grafting in patients with ischemic cardiomyopathy: the impact of renal insufficiency and noncardiac vascular disease**. *J Card Fail* (2005) **11** 206-212. PMID: 15812749 14. Johansson S., Rosengren A., Young K., Jennings E.. **Mortality and morbidity trends after the first year in survivors of acute myocardial infarction: a systematic review**. *BMC Cardiovasc Disord* (2017) **17** 53. PMID: 28173750 15. Santoro F., Di Biase L., Trivedi C.. **Impact of uncontrolled hypertension on atrial fibrillation ablation outcome**. *JACC Clin Electrophysiol* (2015) **1** 164-173. PMID: 29759360 16. Winkle R.A., Mead R.H., Engel G.. **Impact of obesity on atrial fibrillation ablation: patient characteristics, long-term outcomes, and complications**. *Heart Rhythm* (2017) **14** 819-827. PMID: 28232261 17. Yanagisawa S., Inden Y., Kato H.. **Impaired renal function is associated with recurrence after cryoballoon catheter ablation for paroxysmal atrial fibrillation: a potential effect of non-pulmonary vein foci**. *J Cardiol* (2017) **69** 3-10. PMID: 27499270 18. Sun M., Oliwa T., Peek M.E., Tung E.L.. **Negative patient descriptors: documenting racial bias in the electronic health record**. *Health Aff (Millwood)* (2022) **41** 203-211. PMID: 35044842
--- title: Chronic CD27-CD70 costimulation promotes type 1-specific polarization of effector Tregs authors: - Natalia Bowakim-Anta - Valérie Acolty - Abdulkader Azouz - Hideo Yagita - Oberdan Leo - Stanislas Goriely - Guillaume Oldenhove - Muriel Moser journal: Frontiers in Immunology year: 2023 pmcid: PMC10041113 doi: 10.3389/fimmu.2023.1023064 license: CC BY 4.0 --- # Chronic CD27-CD70 costimulation promotes type 1-specific polarization of effector Tregs ## Abstract ### Introduction Most T lymphocytes, including regulatory T cells, express the CD27 costimulatory receptor in steady state conditions. There is evidence that CD27 engagement on conventional T lymphocytes favors the development of Th1 and cytotoxic responses in mice and humans, but the impact on the regulatory lineage is unknown. ### Methods In this report, we examined the effect of constitutive CD27 engagement on both regulatory and conventional CD4+ T cells in vivo, in the absence of intentional antigenic stimulation. ### Results Our data show that both T cell subsets polarize into type 1 Tconvs or Tregs, characterized by cell activation, cytokine production, response to IFN-γ and CXCR3-dependent migration to inflammatory sites. Transfer experiments suggest that CD27 engagement triggers Treg activation in a cell autonomous fashion. ### Conclusion We conclude that CD27 may regulate the development of Th1 immunity in peripheral tissues as well as the subsequent switch of the effector response into long-term memory. ## Graphical Abstract ## Introduction The discovery of costimulation in the 1990s has tremendously increased our understanding of immune regulation and opened promising avenues for immune intervention. Several costimulatory molecules from two families have been identified which may display both specific and redundant functions. The immunoglobulin superfamily comprises CD80 and CD86, discovered in 1991, as well as ICOS-L. CD80 and CD86 play a critical role in initial priming and reinforce the TCR-induced signaling pathway, leading to increased activation, IL-2 production and survival. TNF superfamily ligands include CD70, GITRL, 4-1BBL and OX40L. A recent report suggests that TNF-family ligands may be expressed at higher levels on monocyte-derived inflammatory antigen-presenting-cells (APCs), i.e. at a later stage of the response, and may therefore essentially control the post-priming accumulation/function of T lymphocytes [1]. However, the intrinsic and extrinsic factors that govern the expression of the various costimulatory pathways and their respective roles in the course of immune responses remain unclear. Among the costimulatory pathways, CD27/CD70 has gained increasing interest in 2008, when it was shown to act as a switch between immunity and tolerance in vivo. CD27 is constitutively expressed in most T cells, including naïve and activated CD4+ and CD8+ T cells, and provides a second signal for T cell activation (2–4). CD27 signaling is controlled by its unique ligand, CD70, which is transiently expressed upon activation on dendritic cells, B cells and T cells in a tightly regulated fashion (5–8). Keller et al. generated mice that constitutively express CD70 in conventional dendritic cells and showed that the sole expression of CD70 by immature DCs was sufficient to convert CD8+ T cell tolerance into immunity [9]. The CD27/CD70 interaction seems to induce bidirectional intracellular signaling, with CD27 interacting with members of the TRAF family, leading to activation of NFκB and MAP kinases, and CD70 inducing PI3K and MEK activation [7, 10]. There is evidence that CD27 engagement may have different outcomes, depending on the strength, duration and timing of the stimulation. The CD27/CD70 interaction has been shown to provide a potent second signal for differentiation of CD4+ T cells into Th1 effectors [11, 12]. CD8+ cells into cytotoxic T lymphocytes (13–16). By contrast, persistent triggering of CD27 resulted in exhaustion and activation-induced cell death [17]. It is intriguing that regulatory T lymphocytes (Tregs) express higher levels of CD27 at steady state, as compared to conventional CD4 T cells (Tconvs), raising the question of the role of this costimulatory pathway in cell subsets displaying opposite functions. Coquet et al. demonstrated that CD27 signaling enhanced positive selection of Tregs (but not Tconvs) within the thymus, resulting in decreased Treg numbers in CD70 or CD70-deficient mice [18]. Of note, Tregs have been shown to express common master transcription factors with the T helper lineage they control (19–22). We hypothesized that the CD27/CD70 pathway may regulate the activation/polarisation of both Tconvs and Tregs and more specifically the plasticity (stability versus conversion) of Tregs. To test the impact of CD27 engagement on either subset in vivo, we took advantage of mice that constitutively express (or not) CD70 in conventional DCs and analyzed the functional and molecular features of both populations. ## CD70 overexpression in resting DCs induces activation of Tregs and Tconvs To evaluate the outcome of CD27 engagement on Tconvs and Tregs, a model of supra-optimal CD27 engagement in vivo was devised, by crossing mice carrying a CD70 transgene under the control of the CD11c promoter (CD11c-Cd70tg;CD27-/- mice [9, 23];) with Foxp3eGFP mice [24]. The heterozygous CD11c-Cd70tg;CD27+/- mice developed a spontaneous pathology starting at about 8-9 wk of age, and died prematurely (from 13 wk of age). A phenotypic analysis of splenic CD4+ T lymphocytes ex vivo from mice of 8 weeks of age revealed an increased proliferation of both Tconvs and Tregs, with a progressive conversion into activated cells expressing Th1-type markers, i.e. Eomes (Eomesodermin), CXCR3 and TIGIT (for T cell immunoglobulin and ITIM domain) (Figures 1A, B). T-bet and RORγt expression was upregulated on a minor population of Tregs, Of note, the proportion of proliferating cells, which increased by 2-fold in all CD4+ T cells, was much higher among Tregs, reaching $41.9\%$, as compared to T convs ($13.3\%$). The proportion of Eomes+ cells was increased by 4-fold in both subsets, reaching approximately $15.5\%$, whereas the expression of T-bet was increased in Tregs only (from 2.6 to $7.9\%$). CXCR3 was expressed by approximately $20\%$ of lymphocytes in either cell subset in CD27+/- control mice, and by $60.7\%$ and $31.3\%$ of Tregs and Tconvs in CD11c-CD70tg;CD27+/- mice, respectively. The expression of ICOS was also strongly increased, reaching about $70\%$ of Tregs. Of note, the proportion of (CD44+ CD62L-) effector Tregs was increased by 2-fold in CD70Tg mice, within a range between 20 and $35\%$ (not depicted), whereas the proportion of Tregs expressing CXCR3 or ICOS reached 60 and $70\%$, respectively, indicating that the type 1 phenotype was not restricted to the effector Treg subset. The memory phenotype of Tconvs remained unchanged in the same mice. Finally, TIGIT was expressed by a much higher proportion of Tregs in control mice ($34.3\%$ versus $5.1\%$ of Tconvs) and reached almost $60\%$ in CD11c-Cd70tg;CD27+/- mice, as compared to less than $20\%$ of Tconvs. The proportion of Tregs among CD4+ T lymphocytes was lower in CD11c-CD70tg;CD27+/- than in control mice, whereas the level of Foxp3 expression remained unchanged (Figure 1B, right panels). The distribution of Tregs remained unchanged in the liver, the mesenteric lymph nodes and the adipose tissue (Supplemental Figure 1A) with a notable drop in Treg frequency in the lamina propria, which correlated with intestinal dysfunction (not depicted). Note that the absolute numbers of CD4+ T lymphocytes were similar in both strains of mice (Supplemental Figure 1B). **Figure 1:** *Activation of Tregs and Tconvs in CD70 transgenic mice. Spleen cells from CD27+/- mice expressing or not a CD70tg were harvested at 6-8 wk of age and stained ex vivo for indicated proteins. (A) Spleen cells were stained for CD4, Foxp3 and the indicated markers. Singlets were selected by gating events in the diagonal of FSC-H vs FSC-A plots. Representative flow cytometry plots of indicated marker on gated viable CD4+ cells.(B) Proportion of positive cells among viable CD4+ Tregs and Tconvs (left panels; proportion of Tregs and intensity of Foxp3 expression (right panels). Data are from 3-4 independent experiments with 5-8 mice per group. Bars represent median ± SD. Unpaired t-test was used to determine statistical differences followed by FDR correction for multiple comparisons (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant. (C) Representative merged (n =8) t-distributed stochastic neighbor embedding (t-SNE) plot after dimensionality reduction and unsupervised clustering of flow cytometry data from CD4+ Tregs and Tconvs.* A tSNE analysis, based on the same four markers (Eomes, T-bet, CXCR3 and TIGIT), revealed two main changes in the phenotype of CD4+ T lymphocytes in CD11c-Cd70tg;CD27+/- as compared to CD27+/- controls (Figure 1C and Supplemental Figure 2). First, a population of Tregs emerged ($9.5\%$ of parent), which expressed high levels of both Eomes and TIGIT, and partly CXCR3 (26,$7\%$) and/or T-bet (5,$5\%$). Second, the expression of CXCR3 and/or TIGIT was also increased in a significant proportion of other cells, whereas T-bet and Eomes were largely restricted to this expanded population. Similarly, a population of Tconvs ($31.9\%$ of parent) differentiated upon CD27 engagement, and displayed similar features as activated Tregs, i.e. high levels of Eomes ($89.9\%$) and TIGIT ($62.1\%$), and lower levels of CXCR3 (38,$6\%$) or T-bet (3,$9\%$). The expression of CXCR3, but not TIGIT, was also increased on a separate population of Tconvs in CD70tg mice. As expected [25], the CD27 engagement resulted in strongly reduced expression of this receptor on all T lymphocytes (not shown). Thus, the CD27 engagement without intentional TCR stimulation resulted in the expansion of a Eomeshi TIGIThi cell population (representing $10\%$ and $30\%$ of Treg and Tconv, respectively) and increased the expression of CXCR3 on a large, distinct subpopulation in both subsets. TIGIT expression was upregulated on a large proportion of Tregs. Of note, both Tconvs and Tregs from CD11c-Cd70tg;CD27+/- mice also expressed higher levels of co-inhibitory receptors ICOS, CTLA4 and PD-1 (Figures 1A, B). To prevent an exhaustion potentially due to chronic CD70-mediated stimulation, we triggered CD27 on Tregs in vivo using agonistic anti-CD27 mAbs [26, 27]. The data in Figures 2A, C indicate that Tregs from WT mice injected twice with agonistic anti-CD27 mAbs displayed partly similar phenotype changes as those observed in CD11c-Cd70tg;CD27+/- mice; i.e. increased proliferation of Foxp3+ T cells, and enhanced expression of ICOS, PD-1, CTLA-4, Eomes, but not T-bet, CXCR3 nor TIGIT. Of note, the proportion of Tregs among CD4+ T cells increased as well as the level of Foxp3 expression (Figure 2B). The administration of agonistic mAbs also enhanced the proliferation of Tconvs and increased the expression of ICOS and CXCR3 on a minor population of cells. Finally, this treatment enhanced the expression of IFN-γ by both subsets. Collectively, these data indicate that CD27 may regulate the differentiation/survival of type 1 cells of regulatory and conventional lineages. **Figure 2:** *Injection of agonistic anti-CD27 mAb induces some features of Th1-type Tregs. WT mice were injected i.p. with agonistic anti-CD27 mAb (at days 0 and 3) and spleen cells were analyzed ex vivo by flow cytometry at day 6. (A) Spleen cells were stained for CD4, Foxp3 and the indicated markers? Singlets were selected by gating events in the diagonal of FSC-H vs FSC-A plots. Representative flow cytometry plots of indicated marker on gated viable CD4+ cells. (B) Proportion of Tregs and intensity of Foxp3 expression. (C) Proportion of CD4+ Tregs and Tconvs expressing the proliferation marker Ki67, transcription factors (Eomes, T-bet), cytokines (IFN-γ and IL-10) and inhibitory receptors (TIGIT, ICOS, CTLA-4 and PD-1). Data are from 3 independent experiments with 4 mice per group. Bars represent median ± SD. Unpaired t-test was used to determine statistical differences followed by FDR correction for multiple comparisons (*p < 0.05; **p < 0.01; ****p < 0.0001; ns, not significant).* The phenotypic changes of Tregs are reminiscent of a few studies showing the development of T-bet-dependent CXCR3+ Treg cells in response to IFN-γ produced by effector cells [19, 28, 29]. CXCR3+ Tregs have been shown to display unique phenotypic features and nonredundant functional properties to control Th1-related inflammation and autoimmune diseases [21, 28]. We therefore sought to determine (i) the functional status of Tregs in CD11c-CD70tg;CD27+/-; (ii) the transcriptional signature of CD27 engagement in vivo in *Tregs versus* Tconvs ## Transgenic expression of CD70 on DCs induces the activation of type 1- Tconvs and Tregs We next examined the capacity of both CD4+ T cell subsets to express IFN-γ and IL-10. Ex vivo intracellular FACS staining revealed that a range of 0-$40\%$ (median: 11,$1\%$) of Tconvs from CD11c-Cd70tg;CD27+/- mice expressed IFN-γ in response to calcium ionophore and PMA, as compared to less than $7\%$ (median: 6.25) from CD27+/- mice (Figure 3A). The proportion of Tregs expressing IFN-γ ex vivo was also increased in mice constitutively expressing CD70, reaching about $7.45\%$, as compared to $2\%$ in control mice. Both subsets displayed an increased expression of IL-10, from 2.75 to $8.5\%$ for Tregs and 0.5 to 1.6 for Tconvs. Thus, CD27 engagement in vivo resulted in increased expression of IFN-γ and IL-10 in both subsets, with Tconvs and Tregs as the major IFN-γ and IL-10 producers among CD4+ cells, respectively. The analysis of IFN-γ and IL-10 expression with the selected combination of markers (tSNE Figure 3B) indicated that the majority of the IFN-γ and IL-10 producers were located in the expanded EomeshiTIGIThi populations (as defined in Figure 1C). Thus, this subpopulation of Tregs included $36\%$ IFN-γ+ and $35\%$ IL-10+ cells, whereas the equivalent subset of Tconvs comprised 11,$2\%$ IFN-γ+ and $3.4\%$ IL-10+ cells (Supplemental Figure 3; note that the expression of some markers was altered by the permeabilization step and PMA/calcium ionophore activation). **Figure 3:** *Transgenic expression of CD70 on dendritic cells induces the differentiation of type 1 effectors. (A) Proportion of Tregs (left) and Tconvs (right) expressing IFN-γ or IL-10 after short stimulation in vitro with phorbol myristate acetate (PMA)-ionomycin in the presence of brefeldin A. Data are from 3 independent experiments with 3-8 mice per group. Bars represent median ± SD. Unpaired t-test was used to determine statistical differences followed by FDR correction for multiple comparisons (*p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant). (B) Merged (n = 8) tSNE plot after dimensionality reduction and unsupervised clustering of flow cytometry data from Tconvs and Tregs. The tSNE was built on the 3 markers and both cytokines.* The increased expression of IFN-γ and IL-10 by CD4+ T lymphocytes raised the question of their functionality. To evaluate the function of Tregs, we co-cultured Tconvs from WT mice with increasing numbers of Tregs from CD27+/- or CD11c-Cd70tg;CD27+/- mice in the presence of anti-CD3 and irradiated APCs. The data in Figure 4 suggest that CD27 engagement did not alter the suppressive capacity of Tregs, as assessed by the similar decrease in proliferation of Tconvs co-cultured with Tregs isolated from either strain. **Figure 4:** *Unchanged suppressive capacity of Tregs in CD11c-Cd70tg;CD27+/- mice. CD4+ Foxp3+ cells were sorted from CD27+/- or CD11c-Cd70tg;CD27+/- mice and co-cultured with CFSE-labeled, naive conventional CD4+ T cells from CD45.1 mice in the presence of soluble anti-CD3 mAb and irradiated splenocytes. After 3 days, flow cytometry profiles of CFSE were analyzed. Percent of suppression of proliferation as compared to cultures in which Treg cells were omitted. Data are representative of 4 independent experiments with n = 5 per group. Values are presented as the median ± SD and were compared by two-tailed unpaired student’s t-test. ns, not significant.* ## CD27 engagement results in Treg activation in a cell autonomous fashion An important question is whether the effect of CD27 engagement on *Tregs is* cell autonomous, or a consequence of pro-inflammatory factors produced by CD70-activated Tconvs. To address this question, we transferred purified Tregs from CD90.1 Foxp3eGFP mice into CD11c-Cd70tg or control mice, i.e. WT (Figure 5A) or CD27-/- (Figure 5B) recipients. Six days after transfer, the proportion (median) of CD90.1+ Tregs expressing KI67 (70.35 versus $22.1\%$), CXCR3 (45.9 versus $32.15\%$), T-bet (8.7 versus $0.8\%$), CTLA-4 (48.9 versus $42.3\%$), ICOS (56.7 versus $36.5\%$) and/or PD-1 (44 versus $31.1\%$) was increased in CD11c-Cd70tg;CD27-/-, as compared to WT hosts (Figure 5A). Note that the proportion of donor Tregs and their relative Foxp3 expression were much higher when transferred in recipients expressing a CD70 transgene. The CD27 expression was abrogated on transferred Tregs, a likely consequence of its engagement by its natural ligand expressed by most DCs (Figure 5 right panels), as previously demonstrated [25]. To confirm these observations, we transferred Tregs into CD27-/- hosts (which display impaired Treg differentiation) expressing or not CD70tg and found similar in vivo cell expansion/survival and phenotypic changes, resulting in a 10-fold increase in the number of transferred Tregs when CD70tg was present (Figure 5B). CD27 engagement was required, as adoptive transfer of CD27-deficient Tregs into CD70tg hosts resulted in lower expression of “type 1 markers” and lower numbers of donor Tregs detected in the host, as compared to CD27-sufficient donor Tregs (Figure 5C). These observations indicate that CD27 engagement on Tregs was sufficient to induce their proliferation and differentiation into effectors, demonstrating an intrinsic effect. **Figure 5:** *Cell autonomous activation/differentiation of Tregs upon CD27 engagement. (A, B) 5 x 105 Tregs purified from Foxp3eGFP CD90.1 mice were injected i.v. into CD11c-Cd70tg;CD27-/- recipients and either WT (A) or CD27-/- (B) as control recipients. (C) Tregs were purified from Foxp3eGFP CD90.1 mice either CD27 competent or deficient (CD27-/-) and 5 x 105 cells were injected i.v. into CD11c-Cd70tg;CD27-/- recipients. Spleen cells were analyzed ex vivo by flow cytometry 7 days after injection. Data show the proportion of transferred Tregs (Foxp3+ CD90.1+) expressing the proliferation marker Ki67, transcription factors (eomes, T-bet), chemokine receptor CXCR3 and inhibitory receptors (ICOS, CTLA-4 and PD-1) as well as the absolute number of CD90.1+ Tregs recovered. Controls include CD27 and CD70 staining. Data are representative of 3 independent experiments with 4 mice per group. Bars represent median ± SD. Unpaired t-test was used to determine statistical differences followed by FDR correction for multiple comparisons (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant).* Finally, to compare the effect of CD27 engagement on *Tconvs versus* Tregs, we transferred purified CD4+ T lymphocytes from CD90.1 Foxp3eGFP mice into CD11c-Cd70tg, CD27-/- or WT recipients. A phenotypic analysis performed 7 days after transfer (Supplemental Figure 4) showed a more than 2-fold increase in the proportion of Tregs among transferred cells (panel A), which correlated with higher proliferation of Tregs as compared to Tconvs (panel B). The expression of Foxp3 was increased on cells transferred into a CD70Tg host, as compared to WT host (panel A). Of note, the proliferation and CXCR3 expression of transferred, but not host, cells were increased, strongly suggesting that the altered phenotype was driven by CD27-CD70 interactions. ## Common transcriptional signature of CD27 engagement in Tregs versus Tconvs We next performed global transcriptional profiling on sorted Tregs and Tconvs (Supplemental Figure 5) from CD27+/- mice expressing or not the CD70 transgene (Supplemental File 2). $\frac{471}{92}$ and $\frac{280}{92}$ genes were significantly up/downregulated in Tconvs and Tregs, respectively (at a Log2 fold change>1, min 60 CPM reads per gene, FdR value <0.05)(see volcano plot in Supplemental Figure 6). As expected, Tregs and Tconvs were clearly distinct in principal components analysis. Cells from CD11c-Cd70tg;CD27+/- mice appeared to cluster separately from their CD27+/- counterparts (Figure 6A). Among the differentially expressed genes, we defined 6 gene clusters based on their behaviors in both comparisons (Figure 6B). Among the 758 genes upregulated upon CD27 engagement in either Tconvs or Tregs, cluster 1 genes were significantly upregulated in Tregs only and encode some components of inflammation (Stat1, S100a$\frac{8}{9}$); cluster 2 includes transcripts enhanced in both Tregs and Tconvs and encodes molecules specific for inflammatory/cytotoxic responses (Ifng, Nkg7, Prf1, Gzma/b, Cxcr3, Tbx21, …); cluster 3 genes were preferentially upregulated in Tconvs (with a tendency in Tregs), reaching similar expression levels in both subsets, and are involved in immune regulation (Zeb2, Prdm1; Tigit, Cebpb). Of note, $45\%$ of transcripts upregulated in Tregs upon CD27 engagement were similarly enriched in Tconvs. As expected [30], genes downregulated upon CD27 engagement included genes of the IL-17 pathway: Il17rb (in Tconvs) and Il6ra (in Tregs). **Figure 6:** *Common transcriptional signature of CD27 engagement in Tregs versus Tconvs. (A) Left panel: principal component analysis (PCA), a dimensionality reduction method, show the separation among Tregs and Tconvs due to CD70Tg. Variance in PC1 and PC2 is shown. (B) Left panel: differentially expressed genes of Tconvs and Tregs are clustered based on occurrence. Clusters 1 and 3 are upregulated specifically in Treg F1 CD70Tg and Tconv F1 CD70Tg, respectively, whereas clusters 4 and 6 consist of genes downregulated specifically in Treg F1 CD70Tg and Tconv F1 CD70Tg, respectively. Shared up/down regulated genes are found in cluster 2 and 5. Values are represented as Log2 fold-change obtained from median CPM of each gene. Selected genes for each cluster are displayed in the right margin. The number of genes in each cluster is shown in the left margin. Right panel, selected pathways enriched in clusters 1, 2, and 3 using clusterProfiler R package with default parameters and presented as −Log10 of p-value. (C) GSEA performed on F1 CD70Tg data set and Tconv F1 CD70Tg differentially expressed genes as gene sets. Normalized enrichment score (NES) and false discovery rate (FDR) are indicated. F1 mice: CD27+/-; F1 CD70Tg mice: CD11c-Cd70tg;CD27+/-.* A gene ontology analysis (Figure 6B, right panel) revealed a specific enrichment of common genes involved in T cell activation and migration, production of cytokines and cytokine-mediated signaling pathway in both Tconvs and Tregs (cluster 2). The cluster 1 includes pathways upregulated mainly in Tregs, related to inflammation, cell adhesion and extracellular matrix organization, whereas the cluster 3 comprises pathways upregulated essentially in Tconvs (response to IFN-γ, regulation of cell cycle transition and T cell activation). The overlay of the CD27-dependent Tconv signature on the comparison of CD11c-Cd70tg;CD27+/- versus CD27+/- Treg signature revealed a strong similarity between transcriptional changes in both subsets (GSEA in Figure 6C). These data indicate that the transcriptional program induced by CD27 engagement is similar in Tregs and Tconvs and is reminiscent of the Th1-type signature. ## Eomes overexpression and CXCR3 expression partially recapitulate CD27-induced transcriptional program in Tregs In order to assess the role of Eomes under steady-state conditions independently of TCR signaling, we used a transgenic mouse line that overexpresses this transcription factor in developing thymocytes under the control of the human CD2 (hCD2) regulatory elements (promoter and Locis Control Region) [31]. We sorted splenic CD4+ Tregs from these mice and from WT controls and analyzed their transcriptome (Supplemental File 2). Eomes-induced and repressed genes were very highly enriched in Tregs from CD11c-Cd70tg;CD27+/- mice, supporting the notion that this transcription factor is sufficient to drive a similar program (Figure 7A). A flow cytometry analysis confirmed the type 1 phenotype of Tregs from EomesTg, as assessed by increased expression of Tbet, CXCR3, TIGIT and PD-1, as compared to cells from WT mice (Supplemental Figure 7). **Figure 7:** *Common enriched pathways in Tregs from CD11c-Cd70tg;CD27+/-, EomesTg and CXCR3+ Tregs. (A, B) GSEA performed on F1 CD70Tg data set and Treg EomesTg (A) or Treg Cxcr3+ (B) differentially expressed genes as gene sets. Normalized enrichment score (NES) and false discovery rate (FDR) are indicated. (C) Bubble plot for selected pathways (database MsigDB; https://www.gsea-msigdb.org/gsea/msigdb/) enriched from genes upregulated in Treg F1 CD70Tg, Treg Cxcr3+ and Treg EomesTg using clusterProfiler R package with default parameters and presented as −Log10 of p-value. Color intensity indicates p-value and bubble size indicates number of enriched genes.* We next compared the signature of Tregs from CD11c-Cd70tg;CD27+/- mice with CXCR3+ Tregs [22] and found strong similarities with CD27-induced changes (Figure 7B). Pathway analysis of signature genes from CD11c-Cd70tg;CD27+/-, EomesTg and CXCR3+ Tregs from WT mice revealed a strong convergence among the 3 subsets (Figure 7C). Common enriched pathways are involved in T cell activation, response to IFN-γ, regulation of leukocyte differentiation, positive regulation of cytokine production and leukocyte migration. In addition to these common signatures, the leukocyte mediated cytotoxicity pathway was enriched only in Tregs from CD11c-Cd70tg;CD27+/- and EomesTg. Finally, we performed one experiment to evaluate the role for the CD27/*Eomes axis* in driving the activation of the Treg subpopulation. We injected anti-CD27 mAbs into Eomes sufficient or deficient mice. The data in Supplemental Figure 8, although preliminary, indicate that Eomes was dispensable for the induction of ICOS and PD-1 expression on Tregs and Tconvs in response to CD27 agonism, suggesting a functional redundancy, presumably with the related transcription factor T-bet. Additional experiments using single and double KO mice will be required to better define the respective role of these T-box transcription factors. ## Discussion The main finding of this work is that constitutive CD27 engagement in vivo, in the absence of intentional TCR engagement, results in a gradual process of differentiation into type 1-Tconvs and Tregs. These data complete a previous report showing that transgenic expression of CD70 by dendritic cells is sufficient to induce spontaneous conversion of conventional T lymphocytes into effector cells [9] and indicate that Tregs undergo a similar process of differentiation into type 1 cells. Our data reveal a dual effect of CD27 engagement on Tconvs and Tregs: (i) an increase in proliferation/survival and (ii) phenotypic and functional changes with the selective expansion of a population of Eomeshi TIGIThi cells ($30\%$ of Tconvs and $10\%$ of Tregs) that display the capacity to produce IFN-γ and/or IL-10 and the increased expression of the chemokine receptor CXCR3 on a large, separate subset ($30\%$ of Tconvs and $60\%$ of Tregs). In the absence of deliberate antigenic stimulation, it is likely that TCR signals are provided by environmental and/or auto-antigens [9, 17]. ## Effect of CD27 engagement on Tregs The observation that highly purified CD27+ Tregs become activated (as assessed by proliferation of 60-$70\%$ of cells and increased expression of Tbet, CXCR3, ICOS, PD-1) upon transfer into CD70 transgenic hosts (Figure 5) suggests that CD27 signals autonomously on Tregs and does not solely rely on IFN-γ produced by surrounding Tconvs as previously suggested [19]. Although a potential role of IFN-γ secreted by Tregs themselves cannot be excluded, as a significant proportion of Tregs from CD11c-Cd70tg;CD27+/- mice have the capacity to express IFN-γ, our observations are consistent with a cell autonomous role of CD27 engagement on Treg differentiation. Our data further show that transient CD27 engagement induced a selective expansion of Tregs and likely increased the stability of Tregs, as assessed by higher levels of Foxp3 expression (Figures 2 and 5A; Supplemental Figure 4A), suggesting that CD27 costimulation may have a stronger impact on Tregs than on Tconvs. Finally, the lack of activation/expansion of (CD27-deficient) host T cells upon transfer of CD27-competent CD4+ T cells into CD11c-Cd70tg;CD27-/- mice (Supplemental Figure 4B) further highlights the role of CD27 engagement, as opposed to systemic inflammation, in the activation CD4+ T lymphocytes. Of note, our data show that the CD27 engagement profoundly impacts the phenotype of Tregs and induces similar transcriptomic changes as in Tconvs (see later discussion). Several reports have highlighted the role of CXCR3 in the recruitment of Tregs to local inflammatory sites. CXCR3+ type-1 Tregs have been shown to play a critical and non-redundant role in the control of Th1-type (auto)-immunity. T-bet- dependent CXCR3+ Tregs accumulated within the pancreatic islets (but not the spleen) in the NOD mouse model of type 1 diabetes, at a frequency that correlated inversely with the size of the inflammatory infiltrate [28]. Mice that lack CXCR3 in Tregs specifically displayed an aggravated course of the experimental nephritis, that correlated with reduced Treg recruitment to the kidney and an overwhelming Th1 immune response [32]. Elegant studies by Levine and Rudensky [21] indicated that elimination of T-bet expressing Tregs resulted in severe autoimmunity and suggested that T-bet regulated their stability and their spatial positioning with T-bet+ effector cells. Littringer et al. further showed that, during Th1 immune responses in mouse and human, Tregs that expressed a set of Th1-specific co-inhibitory receptors and cytotoxic molecules arose [22]. The relevance of CD27 engagement on Tregs in vivo in more physiological conditions remains to be determined, but is suggested by a recent study demonstrating, that CD27 expression on Tregs was necessary to maintain tolerance and to suppress immune responses to tumours [33]. It is interesting that, in addition to their well-known anti-inflammatory function, Tregs may support the development of memory T cells, in particular via IL-10 or CTLA-4 (34–37). A recent report [37] shows that type 1-Tregs, expressing CXCR3, home in close proximity to CD8+ T and promote the generation of tissue-resident memory cells in multiple tissues, providing life-long immunosurveillance. Therefore, it is tempting to speculate that CD27 engagement may trigger the differentiation of type-1 Tregs involved in the transition of activated effector cells into quiescent memory cells in peripheral tissues. Furthermore, our data indicate that chronic CD27 engagement triggers the emergence of a Eomes+ TIGIT+ Treg population that proliferated and included most of IFN-γ and IL-10 producers. TIGIT is an inhibitory receptor expressed on Tregs as well as activated and memory T cells. TIGIT has been shown to increase the stability and the immune suppressive role of Tregs [38, 39]. T-bet was expressed in only $5.5\%$ of this population, suggesting a major role of the transcription factor Eomes, a paralogue of T-bet, in inducing IFN-γ expression. Although Eomes expression was shown to limit Foxp3 induction, and thereby the suppressive function of Tregs [40], our data show that chronic CD27 engagement in CD11c-Cd70tg;CD27+/- mice did not alter the expression of Foxp3 (Figure 1B) nor their suppressive capacity, whereas transient engagement enhanced its expression (Figures 2B, 5A and Supplemental Figure 4A). The function of Eomes+ TIGIT+ Tregs remains to be determined but could be related to an increased cytotoxic mechanisms of suppression in vivo, as Eomes expression has been associated with enhanced cytotoxicity [41]. Data in the literature show that Tregs, although dedicated to the inhibition of inflammatory responses, may produce IFN-γ. Although the consequence of this production is not entirely clear, a few reports indicate that Treg cell intrinsic IFN-γ production was required for their immunosuppressive function (42–44). ## Effect of CD27 engagement on Tconvs The effect of CD27/CD70 interaction on the function of *Tconvs is* in agreement with previous studies, showing a preferential differentiation toward the Th1/cytotoxic lineages, as assessed by the expression of cytokines, transcription factors, chemokines and chemokine receptors involved in type 1 inflammation (9, 12, 18, 42–44). Of note, our data further show that a population of Eomes+ TIGIT+ Tconvs expands, representing $30\%$ of Tconvs, and produces IFN-γ, and some IL-10. Our data are in accordance with a previous report showing that TIGIT+ CD4+ T cells exhibited defects in effector function, i.e. were poor producers of IL-2 and TNF-α, but produced high levels of IFN-γ [45]. The role of the inhibitory receptor TIGIT remains undetermined but could be related to the exhaustion of CD4+ T cells chronically activated via CD27. Eomes was strongly upregulated in CD4+ Tconvs of CD11c-Cd70tg;CD27+/- mice (3,2 log2 FC). This transcription factor, a paralogue of T-bet, appears to complement T-bet in triggering IFN-γ in T lymphocytes [46] and was involved in the differentiation of cytotoxic CD4+ T cells. Curran et al. [ 41] have shown that the activation of TNFR family receptors, in particular 4-1BB and to a lesser extent CD27, upregulated Eomes on CD4+ T lymphocytes. Eomes+ KLRG1+ CD4 T cells displayed cytotoxic properties and expressed Granzyme B. Similarly, Tconvs in CD11c-Cd70tg;CD27+/- mice expressed granzyme B, natural killer cell granule protein 7 (NKG7), a regulator of granule exocytosis and a promoter of IFN-γ production in Th1 cells [47]. There is some evidence that these killer CD4+ T cells may control the development of virus associated malignancies, an observation in line with the increased susceptibility of EBV-proliferative diseases in CD27-deficient patients [48, 49]. Accordingly, CD70 has been shown to play a dominant role in both CD4 and CD8 EBV-mediated CTL generation [50]. ## Pathophysiology of deregulated CD70 expression The overt inflammation in CD11c-Cd70tg;CD27+/- is reminiscent of similar data by Borst and colleagues who reported that mice with transgenic expression of CD70 on dendritic cells or B cells displayed a progressive pathology, i.e. a shift of Tconvs to an effector phenotype in absence of deliberate immunization and ultimately a lethal combined T and B cell immunodeficiency [9, 17, 51]. These observations are intriguing, as a series of findings concur with a role for CD27 in increasing the regulatory function of Tregs. Indeed, we and others have shown that CD27 co-stimulation increases the proliferation/stability (as assessed by enhanced Foxp3 expression) of Tregs [52]. In agreement with this concept, CD27 engagement led to an increased expression of inhibitory receptors that should in principle favor their suppressive capacity, accompanied by a moderate (1,5 fold) upregulation of IRF8, a transcription factor known to act as a “stabilisator” for Th1-suppressing Tregs [53]. It is therefore presently unclear whether Tregs from CD11c-CD70tg;CD27+/- contribute to the overall pathology (via the production of pro-inflammatory cytokines such as IFN-γ) or fail to adequately control the activity of CD27-activated Tconvs in these mice. Altered expression of chemokine receptors affecting the tissue sublocalisation of Tregs and/or a partial resistance of Tconvs to Tregs conferred by CD27 engagement represent possible explanations for the overt inflammation observed in CD11c-Cd70tg;CD27+/- mice. Additional experiments will be required to better characterize the capacity of Tregs to prevent inflammatory reactions in vivo following chronic and acute CD27 stimulation. ## Importance of the CD27/CD70 pathway in Tregs Costimulatory signals play an important role in Treg development and function. However, while CD28 appears to be required for both development [54] and expression of effector function in Tregs [55]. CD27 engagement appears to promote the expansion, survival, stability and migration of Tregs into inflammatory sites, with little to no effect on intrinsic Treg suppressive capacity (see [56] for a review and our own observations indicating a lack of autoimmune disorder in mice selectively lacking CD27 expression in Tregs (data not shown)). The precise identification of the CD70-expressing cell population delivering these survival signals remains to be firmly established, since CD70 can be expressed by a wide variety of immune cells including APCs and activated T and B lymphocytes. The present data (using a transgenic mouse strain in which CD70 is selectively overexpressed by CD11c-positive cells) strongly suggest that dendritic cells, known to be a preferred target of Treg-mediated regulation [56, 57], represent an important source of CD27 signaling [9]. Importantly, the sole expression of CD70 by immature DCs was shown to regulate immunity versus tolerance, highlighting the critical role of CD27-CD70 interactions at the interface between T cell and DC Our observations are in line with previous reports showing that the CD27/CD70 pathway triggers Treg development in thymic niches by rescuing differentiating cells from apoptosis [18], induces Treg accumulation in solid tumors in mice [58] and supports the generation of Tregs involved in the control of type 1 diabetes in NOD mice [59]. However, these results differ from a previous report by Jannie Borst’s group showing that CD27 agonism did not induce Treg expansion [27] in a murine model of therapeutic vaccination to tumor. Most studies, including ours, found no difference in the in vitro suppressive function of Tregs from CD70-intact and deficient mice [18] or upon CD27 triggering [59]. In addition to this intrinsic effect on Tregs, we have shown previously that CD27+ Tregs have the capacity to inhibit the expression of CD70 on DCs, a mechanism reminiscent of the trans-endocytosis reported for CTLA-4/CD80-CD86, thereby restricting the availability of costimulatory signals in the local environment [57]. The respective roles of both costimulatory pathways on Tregs remain elusive. Some similarity includes their non redundant role in thymic development and homeostasis in the periphery. Of note, there seems to be a consensus in the literature indicating a predominant role of CD27 signaling in Treg development and survival over function, in agreement with our own observations. ## Conclusion In conclusion, the data shown herein demonstrate that CD27 engagement favors the differentiation of CXCR3+ type 1-Tconvs and Tregs. Global transcriptional profiling of Tregs from Eomes Tg and CD11c-Cd70tg;CD27+/- mice revealed a strong convergence between them as well as with CXCR3+ Tregs [22]. However, the transcription factor Eomes appears dispensable for anti-CD27-induced Treg activation, an observation in line with reports suggesting functional redundancy and/or cooperativity between T-bet and Eomes (for review, see [60, 61]). The biological role of these type-1 Tregs in vivo remains to be determined but could be related to a unique, non-redundant function to control inflamed peripheral tissues and to support memory T cell differentiation in peripheral tissues. Thus, the CD27/CD70 pathway may contribute to the onset of inflammation and its resolution, i.e. the transition from effector to memory responses. ## Limitations of the study Our observations suggest that CD27 engagement potentiates the anti-inflammatory capacity of Tregs. However, this statement is based essentially on their phenotype, gene expression and in vitro tests of suppressive activity. Future investigations should include an evaluation of their capacity to control inflammatory responses in vivo in various inflammatory models. Another approach could be based on the use of CD27fl/fl mice (available in our laboratory) to selectively deplete CD27+ Tregs in vivo and assess their role in physiological and pathological immune responses. Finally, data should be interpreted with caution because the constitutive CD27-CD70 interaction affected T cell homeostasis and induced a progressive state of lethal immunodeficiency. ## Mice Wild-type C57Bl/6 mice were purchased from Envigo. CD11c-Cd70tg;CD27-/- and CD27-/- mice were kindly provided by Jannie Borst (NKI, Amsterdam) and crossed to Foxp3eGFP mice from Alexander Rudensky (Memorial Sloan-Kettering Cancer Center), kindly provided by Professor Adrian Liston, to generate CD11c-Cd70tg;CD27+/- Foxp3 reporter mice and CD27+/- control mice. hCD2-Eomestg mice were generated as previously described [31]. The Eomes floxed mice (B6.129S1(Cg)-Eomestm1.1Bflu /J -Strain #:017293) were crossed onto CD4Cre mice, both purchased from The Jackson Laboratory. Mice were bred and maintained in a temperature-controlled (23 °C) animal care facility with free access to food and water and used at 6 to 8 weeks of age. The experiments were carried out in accordance with the relevant European laws and institutional guidelines. All experiments were performed in compliance with the relevant laws and institutional guidelines from the Animal Care and Use Committee of the Institute for Molecular Biology and Medicine (IBMM, ULB). ## Isolation of immune cells and cell sorting Spleens were dissected from mouse spleen using 70µm cell strainers (Falcon) and further processed under sterile conditions. Single-cell suspensions were rinsed-out with RPMI-1640 (Lonza) supplemented with $10\%$ (vol/vol) fetal calf serum (FCS), 2 mM L-glutamine, 1 mM sodium pyruvate, 0.1 mM non-essential amino acids, 40 mM β-mercaptoethanol, 100 U ml−1of penicillin, and 100 U ml−1of streptomycin (all reagents from Lonza). CD4+ cells were purified by negative selection with magnetic depletion of B cells, macrophages, DCs, NK cells, granulocytes and CD8+ cells using a cocktail of biotinylated antibodies (anti-CD49b, DX5, eBiosciences; anti-GR1, RB6-8C5, produced in house; anti-Ter119, BioXCell; anti-CD11c, N418, produced in house; anti-CD19, BioXCell; anti-CD8β, H35, produced in house; anti-CD25, PC61.5, eBiosciences (used for Tconv-but not Treg- purification); anti-MHCII, M$\frac{5}{114.15.2}$, eBiosciences). Cells were recuperated after flow-through the magnetic column, with previous incubation with anti-biotin Microbeads (Miltenyi Biotec). Untouched cells were stained to exclude dead cells and incubated with Fc receptor-blocking antibodies CD$\frac{16}{32}$ (Fc block; BD Pharmingen) and surface staining antibody CD3+ and CD4+. Tconvs and Tregs were identified in FSC/SSC-low-to-moderate and sorted as GFP- or GFP+ respectively using a BD FACSAria III. ## Flow cytometry Single-cell suspensions were stained to exclude dead cells with live/dead fixable violet dead cell stain kit (Life technologies), incubated with Fc receptor-blocking antibodies CD$\frac{16}{32}$ (Fc block; BD Pharmingen), to block non-specific binding, followed by standard surface staining with fluorochrome-conjugated antibodies listed in Table 1. For intracellular staining, cells were fixed and permeabilized for 25 min with eBioscience™Foxp3/Transcription Factor Staining Buffer Set, Life Technologies) before intranuclear/intracytoplasmic staining. **Table 1** | REAGENT or RESOURCE | SOURCE | IDENTIFIER | | --- | --- | --- | | Antibodies | Antibodies | Antibodies | | Agonistic anti-CD27mAb (Rat anti-mouse RM7-3E5) | (Sakanishi &Yagita, 2010) | | | CD4 Alexa Fluor 700 (Rat anti-mouse RM4-5) | BD Biosciences | Cat #:557956 | | CD4 Pacific Blue (Rat anti-mouse(RM4-5) | BD Biosciences | Cat #: 558107 | | CD27 Pe-Cyanine7 (Rat anti-mouse LG7F9) | Thermo Fisher | Cat #: 25-0271-82 | | CD70 PerCP-eFluor710 (Rat anti-mouse FR70) | Thermo Fisher | Cat #:46-0701-82 | | CD90.1 PerCP-Cyanin5.5 (Mouse anti-mouse OX-7) | Biolegend | Cat #: 202516 | | CD152 APC (Rat anti-mouse UC10-489) | Thermo Fisher | Cat #: 17-1522-8 | | CD183 APC (Hamster anti-mouse CXCR3-183) | BD Biosciences | Cat #: 562266 | | CD278 PE (Armenian Hamster anti-mouse C398.4A) | BD Biosciences | Cat #: 565669 | | CD279 APC (Hamster anti-mouse J43) | BD Biosciences | Cat #: 562671 | | EOMES PE (Rat anti-mouse Dan11mag) | Thermo Fisher | Cat #: 12-4875-82 | | FOXP3 APC (Rat anti-mouse FJK-16s) | Thermo Fisher | Cat #: 17-5773-82 | | FOXP3 FITC (anti-mouse/rat FJK-16s) | Thermo Fisher | Cat #: 11-5773-82 | | IgG Purified Immunoglobulin | Sigma Aldrich | Cat#: I-8015 | | IL10 BV421 (Rat anti-mouse JES5-16E3) | BD Biosciences | Cat #: 566295 | | IL10 PE (Rat anti-mouse JES5-16E3) | BD Biosciences | Cat #: 554467 | | IFNγ APC (Rat anti-mouse XMG1.2) | BD Biosciences | Cat #: 554413 | | IFNγ BrillantViolet605 (Rat anti-mouse XMG1.2) | Biolegend | Cat #: 505839 | | Ki67 Alexa Fluor 700 (Rat anti-mouse B56) | BD Biosciences | Cat #: 561277 | | Tbet Pe (mouse anti-mouse/human eBio4B10) | Thermo Fisher | Cat #: 12-5825-82 | | TCRβ Chain PerCP-Cyanine 5.5 (Hamster anti-mouse) H57-597 | BD Biosciences | Cat #: 560657 | | TIGIT PerCP-eFluor710 (Rat anti-mouse GIGD7) | Thermo Fisher | Cat #: 46-9501-80 | | Chemicals, Peptides, and Recombinant Proteins | Chemicals, Peptides, and Recombinant Proteins | Chemicals, Peptides, and Recombinant Proteins | | Live/dead stain | Thermo Fisher | Cat#: L34976 | | PMA | Sigma Aldritch | Cat#: P8139 | | Ionomycin | Sigma Aldritch | Cat#: I0634 | | Brefeldin A | Thermo Fisher | Cat#: 00-4506-51 | | Critical Commercial Assays | Critical Commercial Assays | Critical Commercial Assays | | eBioscience Fixation/Perm diluent | Thermo Fisher | Cat#:00-05223-56 | | eBioscience Fixation/Permeabilization concentrate | Thermo Fisher | Cat#: 00-5213-43 | | Permeabilization Buffer 10x | Thermo Fisher | Cat#: 00-8333-56 | | BD Perm/Wash | BD Biosciences | Cat#: 51-2091KZ | | BD CytoFix/cytoperm | BD Biosciences | Cat#: 51-2090KZ | | Experimental Models: Organisms/Strains | Experimental Models: Organisms/Strains | Experimental Models: Organisms/Strains | | Mouse: CD11c-Cd70tg;CD27-/- | (9) | | | Mouse: CD27-/- | (9) | | | Mouse: C57BL/6JOlaHsd | Envigo | Stock#: 5704F | | Mouse: Foxp3eGFP | (63) | | | Software and Algorithms | Software and Algorithms | Software and Algorithms | | FlowJo software version 9.6.4 | Tree Star | | | GraphPad Prism 6 | GraphPad software | | | Other | Other | Other | | BD FACS Aria III | BD Biosciences | | | BD Canto II | BD Biosciences | | ## Intracellular cytokine detection For intracellular staining of IFN-γ and IL-10, cells were restimulated in triplicates in 96-well tissue culture plate for 4 h at 37°C, $5\%$ CO2 with 50ng/ml phorbol 12-myristate 13-acetate (PMA) (Sigma) and 1μg/ml ionomycin in the presence of 3μg/ml of an inhibitor of intracellular protein transport BrefeldinA (eBioscience) prior to staining. After 4h, cells were stained for dead cells and surface markers as described previously and then fixed (PFA $2\%$) during 30min and permeabilized and intracellular stained in Triton $0.1\%$ (diluted on BSA $0.5\%$). Cells were incubated with directly-conjugated cytokine-specific antibodies diluted in the corresponding permeabilization buffer for 30min and were washed in PBS before FACS analysis. ## In vivo treatment Agonistic anti-CD27 mAb treatment: mice were injected i.p. with 100 or 200µg of agonistic anti-CD27 mAb (BioXCell BE0348), at days 0 and 3, or isotype control (BioXCell BE0089). Spleen cells were analyzed ex vivo by flow cytometry at day 6. ## Treg cell transfer CD4+ Treg cell were enriched from the splenocyte suspension using magnetic microbeads-based CD4+ T cell isolation kit (MiltenyiBiotec) and MACS LS Columns (MiltenyiBiotec) according to the manufacturer’s instructions. Following separation, CD4+ T cells were stained with anti-CD4 Pacific Blue (BD Pharmingen). 5 x 105 Tregs sorted as CD4+ FOXP3+ from FOXP3GFP CD90.1 mice were injected i.v. into CD11c-CD70tg;CD27-/-, WT or CD27-/- (C57BL/6) recipients. Spleen cells were analyzed by flow cytometry 7 days later. ## CFSE-proliferation suppression assay Fresh splenic conventional T cells were sorted as CD4+CD25− from C57BL/6 mice. Tregs were sorted from Foxp3eGFP as CD4+CD90.1+GFP+. For carboxyfluorescein succinimidyl ester (CFSE) labelling, purified CD4+ T cells were resuspended in 10μg/ml of CFSE (Molecular Probes) for 107 cells for 10min at 37°C in dark in RPMI-1640 $0\%$ FBS and were washed in cold RPMI-FCS $10\%$ to neutralize CFSE action. 4×104 CFSE-labeled CD4+CD25− GFP- T cells were incubated with 105 irradiated splenocytes (2000 rad) with or without addition of Treg cells at the indicated ratios, and stimulated with 0.5μg/ml soluble anti-CD3 (2C11) for 72 h. Dividing cells were identified by CFSE dilution on FACS analysis. ## RNA purification and RNA sequencing We extracted RNA from 2x105 sorted CD4+ Foxp3- or CD4+ Foxp3+ populations in CD11c-Cd70tg;CD27+/- Foxp3 reporter mice and CD27+/- mice (in duplicates), and RNA from 2x105 sorted CD4+ CD25+ in hCD2-Eomestg mice and wild-type C57Bl/6 mice (in triplicate) using RNeasy Plus Mini kit according to manufacturer’sinstructions (Qiagen), and sample quality was tested on a 2100 Bioanalyzer (Agilent). Libraries were prepared using Ovation SoLo RNA-Seq System (NuGEN Technologies) and underwent paired-end sequencing (25 × 106 paired-end reads/sample, NovaSeq 6000 platform) performed by BRIGHTcore ULB-VUB, Belgium (http://www.brightcore.be). Adapters were removed with Trimmomatic-0.36. Reads were mapped to the reference genome mm10 using STAR_2.5.3 software with default parameters and sorted according to chromosome positions and indexed the resulting BAM files. Read counts were obtained using HTSeq-0.9.1. Genes with no raw read count, less than or equal to 10 in at least 1 sample were filtered out with an R script. Raw read counts were normalized, and a differential expression analysis was performed with DESeq2 by applying an adjusted $P \leq 0.05$ and an absolute log2 ratio larger than 1. ## Data availability RNA-*Seq data* that support the findings reported in this study have been deposited in the GEO Repository with the accession code no. GSE214395. ## Gene ontology analysis We introduced gene lists resulting from differential analysis between different groups to clusterProfiler v3.16 R package [62]. We used the comparison function to compare gene lists and determined any kind of gene-ontology association. ## Statistical analysis Statistical analyses were performed using Prism6 (GraphPad Software, La Jolla, CA) and R (version 4.2.0). Unpaired t-test was used to determine statistical differences followed by FDR correction for multiple comparisons Flow cytometry t-SNE plots of Figures 1B, 3B show a pool of 8 mice. Other data are shown as boxplot displaying the distribution of data (the minimum, first quartile, median, third quartile, and maximum). In Figures 1A, 2, 3A, 4, 5, ‘‘n’’ indicates the number of mice per group. For each figure, the number of experiments performed is indicated. A p-value ≤ 0.05 was considered significant and is denoted in figures as follows: ∗, $p \leq 0.05$; ∗∗, $p \leq 0.01$; ∗∗∗, $p \leq 0.001.$ No animal or sample was excluded from the analysis. ## Data availability statement The data presented in the study are deposited in the GEO repository, accession number GSE214395. ## Ethics statement The animal study was reviewed and approved by Animal Care and Use Committee of the Institute for Molecular Biology and Medicine (IBMM, ULB). ## Author contributions NB-A, VA, AA, GO, OL, SG and MM designed the study and analyzed the data. NB-A and VA performed most experiments with the help of AA. HY provided reagents. OL, SG, GO and MM supervised the study. MM wrote the manuscript with the help of OL, SG and GO. All authors contributed to the article and approved the submitted version. ## Conflict of interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ## Publisher’s note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. ## Supplementary material The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2023.1023064/full#supplementary-material ## References 1. Chang YH, Wang KC, Chu KL, Clouthier DL, Tran AT, Torres Perez MS. **Dichotomous expression of TNF superfamily ligands on antigen-presenting cells controls post-priming anti-viral CD4+ T cell immunity**. *Immunity* (2017) **47** 943-958.e9. DOI: 10.1016/j.immuni.2017.10.014 2. Hintzen RQ, Lens SM, Lammers K, Kuiper H, Beckmann MP, van Lier RA. **Engagement of CD27 with its ligand CD70 provides a second signal for T cell activation**. *J Immunol* (1995) **154**. DOI: 10.4049/jimmunol.154.6.2612 3. Hendriks J, Xiao Y, Borst J. **CD27 promotes survival of activated T cells and complements CD28 in generation and establishment of the effector T cell pool**. *J Exp Med* (2003) **198**. DOI: 10.1084/jem.20030916 4. Yamada A, Salama AD, Sho M, Najafian N, Ito T, Forman JP. **CD70 signaling is critical for CD28-independent CD8 + T cell-mediated alloimmune responses**. *J Immunol* (2005) **174**. DOI: 10.4049/jimmunol.174.3.1357 5. Tesselaar K, Gravestein LA, Van Schijndel GMW, Borst J, Van Lier RAW. **Characterization of murine CD70, the ligand of the TNF receptor family member CD27**. *J Immunol* (1997) **159**. DOI: 10.4049/jimmunol.159.10.4959 6. Tesselaar K, Xiao Y, Arens R, van Schijndel GM, Schuurhuis DH, Mebius RE. **Expression of the murine CD27 ligand CD70**. *JImmunol* (2003) **170** 33-40. DOI: 10.4049/jimmunol.170.1.33 7. Arens R, Nolte MA, Tesselaar K, Heemskerk B, Reedquist KA, van Lier RAW. **Signaling through CD70 regulates b cell activation and IgG production**. *J Immunol* (2004) **173**. DOI: 10.4049/jimmunol.173.6.3901 8. Sanchez PJ, McWilliams JA, Haluszczak C, Yagita H, Kedl RM. **Combined TLR/CD40 stimulation mediates potent cellular immunity by regulating dendritic cell expression of CD70 in vivo**. *J Immunol [Internet]* (2007) **178**. DOI: 10.4049/jimmunol.178.3.1564 9. Keller AM, Schildknecht A, Xiao Y, van den Broek M, Borst J. **Expression of costimulatory ligand CD70 on steady-state dendritic cells breaks CD8+ T cell tolerance and permits effective immunity**. *Immunity* (2008) **29**. DOI: 10.1016/j.immuni.2008.10.009 10. Akiba H, Nakano H, Nishinaka S, Shindo M, Kobata T, Atsuta M. **CD27, a member of the tumor necrosis factor receptor superfamily, activates NF-kappaB and stress-activated protein kinase/c-jun n-terminal kinase**. *J Biol Chem* (1998) **273**. DOI: 10.1074/jbc.273.21.13353 11. Soares H, Waechter H, Glaichenhaus N, Mougneau E, Yagita H, Mizenina O. **A subset of dendritic cells induces CD4+ T cells to produce IFN-gamma by an IL-12-independent but CD70-dependent mechanism in vivo**. *J Exp Med* (2007) **204**. DOI: 10.1084/jem.20070176 12. van Oosterwijk MF, Juwana H, Arens R, Tesselaar K, van Oers MHJ, Eldering E. **CD27-CD70 interactions sensitise naive CD4+ T cells for IL-12-induced Th1 cell development**. *Int Immunol* (2007) **19**. DOI: 10.1093/intimm/dxm033 13. Peperzak V, Veraar EAM, Keller AM, Xiao Y, Borst J. **The pim kinase pathway contributes to survival signaling in primed CD8+ T cells upon CD27 costimulation**. *J Immunol* (2010) **185**. DOI: 10.4049/jimmunol.1000159 14. Van Gisbergen KPJM, Klarenbeek PL, Kragten NAM, Unger P-PA, Nieuwenhuis MBB, Wensveen FM. **The costimulatory molecule CD27 maintains clonally diverse CD8 + T cell responses of low antigen affinity to protect against viral variants**. *Immunity* (2011) **35** 97-108. DOI: 10.1016/j.immuni.2011.04.020 15. Feau S, Garcia Z, Arens R, Yagita H, Borst J, Schoenberger and SP. **The CD4 + T-cell help signal is transmitted from APC to CD8 + T- cells**. *J Immunother.* (2012) **35**. DOI: 10.1038/ncomms1948 16. Taraban VY, Rowley TF, Kerr JP, Buchan SL. **CD27 costimulation contributes substantially to the expansion of functional memory CD8 + T cells after peptide immunization**. *Eur J Immunol* (2013) **43**. DOI: 10.1002/eji.201343579 17. Tesselaar K, Arens R, van Schijndel GMW, Baars PA, van der Valk MA, Borst J. **Lethal T cell immunodeficiency induced by chronic costimulation**. *Nat Immunol* (2003) **4** 49-54. DOI: 10.1038/ni869 18. Coquet JM, Ribot JC, Bąbała N, Middendorp S, van der Horst G, Xiao Y. **Epithelial and dendritic cells in the thymic medulla promote CD4+Foxp3+ regulatory T cell development**. *J Exp Med* (2013) **210**. DOI: 10.1084/jem.20112061 19. Koch MA, Tucker-Heard G, Perdue NR, Killebrew JR, Urdahl KB, Campbell DJ. **The transcription factor T-bet controls regulatory T cell homeostasis and function during type 1 inflammation**. *Nat Immunol* (2009) **10** 595-602. DOI: 10.1038/ni.1731 20. Kim B, Lu H, Ichiyama K, Jin W, Chang SH, Dong C. **Generation of ROR g t + antigen-specific T regulatory 17 cells from Foxp3 + precursors in autoimmunity**. *CellReports* (2017) **21** 195-207. DOI: 10.1016/j.celrep.2017.09.021 21. Levine AG, Medoza A, Hemmers S, Moltedo B, Niec RE, Schizas M. **Stability and function of regulatory T cells expressing the transcription factor T-bet**. *Nature* (2017) **546**. DOI: 10.1038/nature22360 22. Littringer K, Moresi C, Rakebrandt N, Zhou X, Schorer M, Dolowschiak T. **Common features of regulatory T cell specialization during Th1 responses**. *Front Immunol* (2018) **9**. DOI: 10.3389/fimmu.2018.01344 23. Brocker T, Riedinger M, Karjalainen K. **Targeted expression of major histocompatibility complex (MHC) class II molecules demonstrates that dendritic cells can induce negative but not positive selection of thymocytes in vivo**. *J Exp Med* (1997) **185**. DOI: 10.1084/jem.185.3.541 24. Rubtsov YP, Niec R, Josefowicz S, Li L, Darce J, Benoist C. **Stability of the regulatory T cell lineage in vivo**. *Sci (80- )* (2010) **329**. DOI: 10.1126/science.1191996 25. Burchill MA, Tamburini BA, Kedl RM. **T Cells compete by cleaving cell surface CD27 and blocking access to CD70 bearing APCs**. *Eur J Immunol* (2015) **45**. DOI: 10.1002/eji.201545749 26. Sakanishi T, Yagita H. **Anti-tumor effects of depleting and non-depleting anti-CD27 monoclonal antibodies in immune-competent mice**. *Biochem Biophys Res Commun* (2010) **393**. DOI: 10.1016/j.bbrc.2010.02.092 27. Ahrends T, Babała N, Xiao Y, Yagita H, Van Eenennaam H, Borst J. **CD27 agonism plus PD-1 blockade recapitulates CD4+T-cell help in therapeutic anticancer vaccination**. *Cancer Res* (2016) **76**. DOI: 10.1158/0008-5472.CAN-15-3130 28. Guan Tan T, Mathis D, Benoist C. **Singular role for T-BET + CXCR3 + regulatory T cells in protection from autoimmune diabetes**. *PNAS* (2016) **113**. DOI: 10.1073/pnas.1616710113 29. Koch MA, Thomas KR, Perdue NR, Smigiel KS, Srivastava S, Campbell DJ. **T-Bet+ treg cells undergo abortive Th1 cell differentiation due to impaired expression of IL-12 receptor β2**. *Immunity* (2012) **37**. DOI: 10.1016/j.immuni.2012.05.031 30. Coquet JM, Middendorp S, van der Horst G, Kind J, Veraar EAM, Xiao Y. **The CD27 and CD70 costimulatory pathway inhibits effector function of T helper 17 cells and attenuates associated autoimmunity**. *Immunity* (2013) **38** 53-65. DOI: 10.1016/j.immuni.2012.09.009 31. Istaces N, Splittgerber M, Lima Silva V, Nguyen M, Thomas S, Le A. **EOMES interacts with RUNX3 and BRG1 to promote innate memory cell formation through epigenetic reprogramming**. *Nat Commun* (2019) **10** 3306. DOI: 10.1038/s41467-019-11233-6 32. Paust HJ, Riedel JH, Krebs CF, Turner JE, Brix SR, Krohn S. **CXCR3+ regulatory T cells control TH1 responses in crescentic GN**. *J Am Soc Nephrol.* (2016) **27**. DOI: 10.1681/ASN.2015020203 33. Muth S, Klaric A, Radsak M, Schild H, Probst HC. **CD27 expression on treg cells limits immune responses against tumors**. *J Mol Med* (2022) **100**. DOI: 10.1007/s00109-021-02116-9 34. Pace L, Tempez A, Arnold-Schrauf C, Lemaitre F, Bousso P, Fetler L. **Regulatory T cells increase the avidity of primary CD8+ T cell responses and promote memory**. *Sci (80- ).* (2012) **338**. DOI: 10.1126/science.1227049 35. Kalia V, Penny LA, Yuzefpolskiy Y, Baumann FM, Sarkar S. **Quiescence of memory CD8+ T cells is mediated by regulatory T cells through inhibitory receptor CTLA-4**. *Immunity* (2015) **42**. DOI: 10.1016/j.immuni.2015.05.023 36. Laidlaw BJ, Craft J, Kaech SM. **The multifaceted role of CD4+ T cells in the regulation of CD8+ T cell memory maturation**. *Nat Rev Immunol* (2016) **16**. DOI: 10.1038/nri.2015.10 37. Ferreira C, Barros L, Baptista M, Blankenhaus B, Barros A, Figueiredo-Campos P. **Type 1 treg cells promote the generation of CD8+ tissue-resident memory T cells**. *Nat Immunol* (2020) **21**. DOI: 10.1038/s41590-020-0674-9 38. Chen F, Xu Y, Chen Y, Shan S. **TIGIT enhances CD4+ regulatory T-cell response and mediates immune suppression in a murine ovarian cancer model**. *Cancer Med* (2020) **9**. DOI: 10.1002/cam4.2976 39. Sato K, Yamashita-Kanemaru Y, Abe F, Murata R, Nakamura-Shinya Y, Kanemaru K. **DNAM-1 regulates Foxp3 expression in regulatory T cells by interfering with TIGIT under inflammatory conditions**. *Proc Natl Acad Sci U.S.A.* (2021) **118**. DOI: 10.1073/pnas.2021309118 40. Lupar E, Brack M, Garnier L, Laffont S, Rauch KS, Schachtrup K. **Eomesodermin expression in CD4**. *J Immunol* (2015) **195**. DOI: 10.4049/jimmunol.1501159 41. Curran MA, Geiger TL, Montalvo W, Kim M, Reiner SL, Al-Shamkhani A. **Systemic 4-1BB activation induces a novel T cell phenotype driven by high expression of eomesodermin**. *J Exp Med* (2013) **210**. DOI: 10.1084/jem.20121190 42. Sawitzki B, Kingsley CI, Oliveira V, Karim M, Herber M, Wood KJ. **IFN-γ production by alloantigen-reactive regulatory T cells is important for their regulatory function in vivo**. *J Exp Med* (2005) **201**. DOI: 10.1084/jem.20050419 43. Koenecke C, Lee C-W, Thamm K, Föhse L, Schafferus M, Mittrücker H-W. **IFN-γ production by allogeneic Foxp3+ regulatory T cells is essential for preventing experimental graft-versus-Host disease**. *J Immunol* (2012) **189**. DOI: 10.4049/jimmunol.1200413 44. Dikiy S, Rudensky AY. **Principles of regulatory T cell function**. *Immunity* (2023) **56**. DOI: 10.1016/j.immuni.2023.01.004 45. Zhang W, Anyalebechi JC, Ramonell KM, Chen CW, Xie J, Liang Z. **TIGIT modulates sepsis-induced immune dysregulation in mice with preexisting malignancy**. *JCI Insight* (2021) **6**. DOI: 10.1172/jci.insight.139823 46. Pearce EL, Mullen AC, Martins GA, Krawczyk CM, Hutchins AS, Zediak VP. **Control of effector CD8+ T cell function by the transcription factor eomesodermin**. *Sci (80- ).* (2003) **302**. DOI: 10.1126/science.1090148 47. Ng SS, De Labastida Rivera F, Yan J, Corvino D, Das I, Zhang P. **The NK cell granule protein NKG7 regulates cytotoxic granule exocytosis and inflammation**. *Nat Immunol* (2020) **21**. DOI: 10.1038/s41590-020-0758-6 48. Abolhassani H, Edwards ESJ, Ikinciogullari A, Jing H, Borte S, Buggert M. **Combined immunodeficiency and Epstein-Barr virus–induced b cell malignancy in humans with inherited CD70 deficiency**. *J Exp Med* (2017) **214** 91-106. DOI: 10.1084/jem.20160849 49. Izawa K, Martin E, Soudais C, Bruneau J, Boutboul D, Rodriguez R. **Inherited CD70 deficiency in humans reveals a critical role for the CD70-CD27 pathway in immunity to Epstein-Barr virus infection**. *J Exp Med* (2017) **214** 73-89. DOI: 10.1084/jem.20160784 50. Choi IK, Wang Z, Ke Q, Hong M, Qian Y, Zhao X. **Signaling by the Epstein–Barr virus LMP1 protein induces potent cytotoxic CD4+ and CD8+ T cell responses**. *Proc Natl Acad Sci United States America.* (2018) **115**. DOI: 10.1073/pnas.1713607115 51. Arens R, Tesselaar K, Baars PA, Van Schijndel GMW, Hendriks J, Pals ST. **Constitutive CD27/CD70 interaction induces expansion of effector-type T cells and results in IFNγ-mediated b cell depletion**. *Immunity* (2001) **15**. DOI: 10.1016/S1074-7613(01)00236-9 52. Winkels H, Meiler S, Lievens D, Engel D, Spitz C, Bürger C. **CD27 co-stimulation increases the abundance of regulatory T cells and reduces atherosclerosis in hyperlipidaemic mice**. *Eur Heart J* (2017) **38**. DOI: 10.1093/eurheartj/ehx517 53. Lee W, Kim HS, Baek SY, Lee GR. **Transcription factor IRF8 controls Th1-like regulatory T-cell function**. *Cell Mol Immunol* (2016) **13**. DOI: 10.1038/cmi.2015.72 54. Tang Q, Henriksen KJ, Boden EK, Aaron J, Ye J, Subudhi SK. **Cutting edge: CD28 controls peripheral homeostasis of CD4 + CD25 + regulatory T cells**. *J Immunol* (2003) **171)**. DOI: 10.4049/jimmunol.171.7.3348 55. Zhang R, Huynh A, Whitcher G, Chang JH, Maltzman JS, Turka LA. **An obligate cell-intrinsic function for CD28 in tregs**. *J Clin Invest.* (2013) **123**. DOI: 10.1172/JCI65013 56. Wing J. **Co-Signaling molecules in neurological diseases**. *Adv Exp Med Biol* (2019) **1189**. DOI: 10.1007/978-981-32-9717-3_9 57. Dhainaut M, Coquerelle C, Uzureau S, Denoeud J, Acolty V, Oldenhove G. **Thymus-derived regulatory T cells restrain pro- inflammatory Th 1 responses by downregulating CD 70 on dendritic cells**. *EMBO J* (2015) **34**. DOI: 10.15252/embj.201490312 58. Claus C, Riether C, Schürch C, Matter MS, Hilmenyuk T, Ochsenbein AF. **CD27 signaling increases the frequency of regulatory T cells and promotes tumor growth**. *Cancer Res* (2012) **72**. DOI: 10.1158/0008-5472.CAN-11-2791 59. Ye C, Low BE, Wiles MV, Brusko TM, Serreze DV, Driver JP. **CD70 inversely regulates regulatory T cells and invariant NKT cells and modulates type 1 diabetes in NOD mice**. *J Immunol* (2020) **205**. DOI: 10.4049/jimmunol.2000148 60. Pritchard GH, Kedl RM, Hunter CA. **The evolving role of T-bet in resistance to infection**. *Nat Rev Immunol* (2019) **19** 398-410. DOI: 10.1038/s41577-019-0145-4 61. Takeuchi A, Saito T. **CD4 CTL, a cytotoxic subset of CD4+ T cells, their differentiation and function**. *Front Immunol* (2017) **8**. DOI: 10.3389/fimmu.2017.00194 62. Yu G, Wang LG, Han Y, He QY. **ClusterProfiler: An r package for comparing biological themes among gene clusters**. *Omi A J Integr Biol* (2012) **16**. DOI: 10.1089/omi.2011.0118 63. Fontenot JD, Rasmussen JP, Williams LM, Dooley JL, Farr AG, Rudensky AY. **Regulatory T cell lineage specification by the forkhead transcription factor Foxp3**. *Immunity* (2005) **22**. DOI: 10.1016/j.immuni.2005.01.016
--- title: 'Short-Term Outcomes of the First-Session Prone Position in Patients With Severe Coronavirus Disease 2019: A Retrospective Chart Review' journal: Cureus year: 2023 pmcid: PMC10041127 doi: 10.7759/cureus.35437 license: CC BY 3.0 --- # Short-Term Outcomes of the First-Session Prone Position in Patients With Severe Coronavirus Disease 2019: A Retrospective Chart Review ## Abstract Introduction Prone positioning during ventilation is recommended for patients with severe coronavirus disease 2019 (COVID-19). However, the efficacy of first-session prone positioning in improving short-term outcomes remains unclear. Therefore, we aimed to investigate the impact of the rate of change in partial pressure of oxygen/fraction of inspired oxygen (P/F) ratio before and after initial prone positioning on activities of daily living (ADL) and outcomes at discharge. Methods *In this* retrospective chart review, 22 patients with severe COVID-19 who required ventilator management between April and September 2021 were analyzed. Patients with an improvement in the P/F ratio (after initial prone positioning, compared to that before the session) by > 16mHg and < 16mmHg were defined as responders and non-responders, respectively. Results Compared with non-responders, responders had a significantly shorter ventilator duration, a higher Barthel Index at discharge, and a higher proportion of discharged patients. There was a significant between-group difference in chronic respiratory comorbidities, with one case ($7.7\%$) among responders and six cases ($66.7\%$) among non-responders. Conclusions This study is the first of its kind to investigate short-term outcomes in patients with COVID-19 requiring ventilator management after initial prone positioning. After initial prone positioning, responders had higher P/F ratios as well as improved ADLs and outcomes at discharge. ## Introduction Patients with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection (coronavirus disease 2019 (COVID-19)) experience acute respiratory failure soon after the onset of dyspnea and hypoxemia, generally meeting the criteria for acute respiratory distress syndrome (ARDS) [1]. Treatment options for ARDS include oxygen therapy, ventilatory therapy, and extracorporeal membrane oxygenation (ECMO), as well as antiviral, anti-inflammatory, and anticoagulant medications. Furthermore, it is recommended that patients be placed in the prone position for 12-16 hours per day as an adjunctive treatment [2], depending on their condition, to help improve oxygenation, ventilation-perfusion ratio mismatch, and hemodynamic stability [3]. Early implementation of the prone position in non-COVID-19 patients with ARDS is associated with reduced mortality [4]. Additionally, prone positioning can reduce ventral-dorsal transpulmonary pressure differences and lung compression by the heart [5] and diaphragm, as well as improve the partial pressure of oxygen/fraction of inspired oxygen (P/F) ratio [6]. Currently, prone positioning is actively implemented in >$70\%$ of critically ill patients [3]. Several studies have demonstrated the effectiveness of awake-prone positioning under high-flow nasal cannula therapy [7,8]. Intubated patients with COVID-19 present with more severe symptoms. Further, most reports on prone positioning in patients with COVID-19 have focused on physiological changes [9,10], muscle strength and activity after intensive care unit (ICU) discharge [11], and mortality [3]. This study focused on the responsiveness to first-session prone positioning. Several studies have investigated the rate of change in the P/F ratio during the first-session prone positioning; however, they only investigated physiological parameters [9,10] and not muscle strength or activities of daily living (ADLs). Further, studies focusing on muscle strength and activity levels after ICU discharge applied supine positioning as the control [11]. Moreover, alpha and delta variants were rampant [12] in Japan during this period, leading to an increased number of severe patients. In Japan's aging society, the loss of muscle strength and ADLs due to the treatment of severe COVID-19 is a major concern. Currently, the responsiveness of patients to first-session prone positioning and its impact on short-term outcomes remain unclear. Therefore, it is important to examine the short-term effects of first-session prone positioning during early weaning from ventilation in patients with severe COVID-19 requiring ventilator management. This study aimed to investigate the impact of the rate of change in the P/F ratio before and after first-session prone positioning on ADLs and outcomes at discharge. ## Materials and methods Research design and sampling method This retrospective chart review study included Japanese subjects at a single center (Kitakyushu Municipal Medical Center, Japan). The medical records of 26 patients diagnosed with COVID-19 by polymerase chain reaction testing and placed on ventilator management between April and September 2021 were reviewed, considering CT scans and respiratory symptoms. Among them, four patients were excluded, including two patients who died during hospitalization, one who refused treatment, and one who was suddenly transferred to another hospital for ECMO support. Accordingly, 22 cases were included in the final analysis (Figure 1). Patients who died were excluded from the analysis as they could not be placed in the prone position on admission due to poor hemodynamics. **Figure 1:** *Patient selection flow diagram* Measurement items The primary outcome was the Barthel Index (BI) [13] after extubation and at discharge. BI ranged from fully independent (100 points) to fully assisted (0 points), determined according to the subject's movement status. Age, sex, body mass index (BMI), Charlson Comorbidity Index, and type of comorbidity (chronic respiratory disease, cerebrovascular disease, cardiac disease, diabetes, hypertension, cancer, and obesity) were evaluated as basic characteristics. The P/F ratio in each position and its rate of change, ventilator settings (fraction of inspired oxygen (FiO2) and positive end-expiratory pressure (PEEP)), intubation duration, number of days from onset to intubation, number of days in the hospital after extubation, the total number of days in the hospital, and blood data (lactate dehydrogenase, serum albumin, hemoglobin, and C-reactive protein levels) were evaluated as treatment-related factors. Other rehabilitation-related factors included the Medical Research Council (MRC) sum score at post-extubation and the presence of frailty at discharge. Obesity was defined as a BMI ≥ 28 kg/m2 [14]. Frailty was defined as a Clinical Frailty Scale [15] score ≥ 4. Definition of P/F ratio parameters and responders At our hospital, arterial blood gas measurements are routinely taken before the start of prone positioning, two hours, six hours, and 14 hours after the start of prone positioning, and after prone positioning. We evaluated the following parameters: [1] the P/F ratio in the supine position before the first-session prone position (supine position 1: SP1); [2] the highest P/F ratio value in the prone position (prone position: PP); and [3] the P/F ratio in the supine position after the first-session prone position (supine position 2: SP2). Responders were defined as those with a P/F ratio improvement of ≥16 mmHg from SP1 to SP2 [9]. Statistical analyses Data are presented as the median (interquartile range), regardless of the normality of the distribution. For continuous variables, the unpaired t-test and Mann-Whitney U test were used to compare normally and non-normally distributed variables, respectively. For categorical variables, the chi-squared test was used. Statistical significance was set at $P \leq 0.05.$ All statistical analyses were performed using EZR on R Commander version 1.55 (Saitama Medical Center, Jichi Medical University, Saitama, Japan) [16]. Introduction of the prone position The prone position was applied at the discretion of the attending physician, based on a P/F ratio of <200 mmHg immediately after ventilator management, blood test results, CT findings, and treatment response. Sedation was adjusted to achieve a Richmond Agitation-Sedation Scale score of -5 during management in the prone position. The procedure for changing the patient’s position to a prone position at our hospital was as follows. All routes, including ventilators, arterial lines, and intravenous routes, were placed on the same side. A towel large enough to cover the whole body was placed under the patient. Five staff were deployed, including one at the patient's head and two on each side. The staff member at the patient’s head secured the intubation tube and the patient's face while the other four staff members repositioned the patient. First, the patient’s body was moved sideways to the side opposite the routes. Subsequently, the patient was repositioned to the side-lying position. The patient’s upper body was lifted slightly; further, the upper limbs, which were on the underside of the body, were pulled to the dorsal side. Next, the body was rolled over to a prone position to complete the positioning. To prevent adverse events, including pressure ulcers related to the supine position, we checked for nerve compression and wrinkles in the clothing. Depending on the pathology, the prone position was held for 16-18 hours, from the evening until the following morning. Ethical considerations In accordance with the Declaration of Helsinki, an opt-out option was provided on our website, and the study outline was open to the public to allow patients to refuse inclusion in the study. The latest guidelines in Japan allow the use of clinical information for observational studies using an opt-out model [17]. This study was approved by the Institutional Review Board of Kitakyushu Municipal Medical Center (approval number: 202110004). ## Results Basic characteristics Table 1 shows the basic demographic characteristics of the 22 patients. The median age was 58 years. The proportions of males and females were $54.5\%$ and $45.5\%$, respectively. The median BMI was 25.5 kg/m2. The most common comorbidity was obesity ($36.4\%$), followed by chronic respiratory disease ($31.8\%$). There were no significant differences in age, sex, and BMI between responders and non-responders. There was a significant between-group difference in comorbidities, with chronic respiratory disease in one patient ($7.7\%$) in the responder group and six patients ($66.7\%$) in the non-responder group ($$p \leq 0.003$$). Additionally, the proportion of patients with obesity was higher among responders than among non-responders ($53.8\%$ vs. $11.1\%$, $$p \leq 0.074$$). **Table 1** | Unnamed: 0 | Overall (n = 22) | Responders (n = 13) | Non-responders (n = 9) | P-values | | --- | --- | --- | --- | --- | | Age, years | 58 (49–66) | 53 (46–64) | 61 (53–66) | 0.216 | | Male/female, n (%) | 12 (54.5)/10 (45.5) | 6 (46.2)/7 (53.8) | 6 (66.7)/3 (33.3) | 0.415 | | BMI (kg/m2), n (%) | 25.5 (22.6–31.7) | 29.4 (22.1–34.9) | 25.2 (24.2–25.6) | 0.186 | | <18.5 | 3 (13.6) | 1 (7.7) | 2 (22.2) | | | 18.5–25 | 6 (27.3) | 4 (30.8) | 2 (22.2) | | | 25–30 | 6 (27.3) | 2 (15.4) | 4 (44.4) | | | 30–35 | 5 (22.7) | 4 (30.8) | 1 (11.1) | | | ˃35 | 2 (9.1) | 2 (15.4) | 0 (0) | | | Obesity (BMI > 28 kg/m2) (%) | 8 (36.4) | 7 (53.8) | 1 (11.1) | 0.074 | | Charlson Comorbidity Index, n (%) | | | | 1.0 | | Low | 13 (59.1) | 7 (53.8) | 6 (66.7) | | | Medium | 8 (36.4) | 5 (38.5) | 3 (33.3) | | | High | 1 (4.5) | 1 (7.7) | 0 (0) | | | Comorbidity, n (%) | 32 | 15 | 17 | | | Respiratory disease | 7 (31.8) | 1 (7.7) | 6 (66.7) | 0.003 | | Cerebrovascular disease | 1 (4.5) | 0 (0) | 1 (11.1) | 0.219 | | Heart disease | 1 (4.5) | 0 (0) | 1 (11.1) | 0.219 | | Diabetes | 6 (27.3) | 2 (15.4) | 4 (44.4) | 0.131 | | Hypertension | 6 (27.3) | 3 (23.1) | 3 (33.3) | 0.592 | | Cancer | 3 (13.6) | 2 (15.4) | 1 (11.1) | 0.771 | | Length of time between COVID-19 diagnosis and intubation, days | 9 (40.9) | 10 (10–13) | 9 (7–11) | 0.079 | | Number of times in prone position, times | 3.5 (3–6) | 3 (3–4) | 6 (3–7) | 0.086 | | Mechanical ventilation, days | 7 (7–10.5) | 7 (5–8) | 11 (7–12) | 0.035 | | Length of hospital stay after extubation, days | 16 (14–21) | 16 (15–21) | 16 (14–25) | 0.867 | | Total length of hospital stay, days | 24 (21–35) | 24 (21–28) | 28 (21–49) | 0.547 | Rate of change in the P/F ratio and ventilator settings Figure 2 and Table 2 show the changes in the P/F ratio and ventilator settings. There were significant between-group differences in the PP and SP2 P/F ratios ($$p \leq 0.028$$ and $p \leq 0.01$, respectively), with greater values among responders than among non-responders. For all postures, there were no significant between-group differences in the FiO2 and PEEP. There were significant between-group differences in the percent change in the P/F ratio for SP1 → PP and SP1 → SP2 ($$p \leq 0.007$$ and $p \leq 0.01$, respectively), with greater values in responders than in non-responders. **Figure 2:** *Comparison of the change in the P/F ratio between responders and non-respondersP/F: partial pressure of oxygen/fraction of inspired oxygen; SP1 and SP2: supine position before and after prone positioning, respectively; PP: prone positioning.* TABLE_PLACEHOLDER:Table 2 Comparisons in treatment-related and rehabilitation-related factors Treatment- and rehabilitation-related factors are shown in Table 3. Responders had a significantly shorter ventilator duration (seven days [5-8] vs. 11 days [7-12], $$p \leq 0.035$$), higher BI at discharge (100 points [90-100] vs. 70 points [40-95], $$p \leq 0.029$$), and a higher proportion of discharged patients ($92.3\%$ vs. $33.3\%$, $$p \leq 0.006$$) compared to that in non-responders. **Table 3** | Unnamed: 0 | Overall (n = 22) | Responders (n = 13) | Non-responders (n = 9) | P-values | | --- | --- | --- | --- | --- | | Barthel Index | | | | | | After extubation | 7.5 (0–25) | 15 (0–25) | 0 (0–15) | 0.439 | | At hospital discharge | 95 (70–100) | 100 (90–100) | 70 (40–95) | 0.029 | | MRC after extubation | 33 (24–40) | 36 (30–40) | 26 (16–40) | 0.203 | | Rehabilitation after extubation (with assistance) | | | | | | Length of time until sitting, days | 1 (1–2) | 1 (1–2) | 2 (1–3) | 0.38 | | Length of time until standing, days | 2 (1–5) | 2 (1–3) | 4 (1–7) | 0.631 | | Length of time until ambulation, days | 4 (3–10) | 4 (3–9) | 6 (2–14) | 0.787 | | Laboratory data | | | | | | LDH (U/L) | 568 (466–682) | 593 (464–698) | 535 (488–611) | 0.896 | | CRP (mg/dL) | 9.0 (4.8–12.8) | 6.6 (3.7–10.4) | 12.7 (9.3–13.7) | 0.14 | | Hb (g/dL) | 12.5 (10.9–13.5) | 12.6 (10.7–13.3) | 12.4 (11.4–13.5) | 0.531 | | Alb (g/dL) | 2.2 (2.1–2.3) | 2.3 (2.1–2.4) | 2.1 (2.0–2.2) | 0.104 | | Frailty at hospital discharge, n (%) | 18 (81.8) | 10 (76.9) | 8 (88.9) | 0.471 | | Discharged to home, n (%) | 13 (59.1) | 12 (92.3) | 3 (33.3) | 0.006 | ## Discussion This is the first study to investigate the effects of first-session prone positioning on short-term outcomes in critically ill patients with COVID-19 who required ventilator management. We found that responders had significantly shorter intubation periods, as well as higher BI scores and higher home discharge rates. This suggests that short-term outcomes can be predicted from an earlier stage and thus can guide treatment and physiotherapy. Responders showed a significantly shortened duration of intubation. The improvement in oxygenation upon return from prone to a supine position may have resulted in effective correction of the ventilation-perfusion ratio mismatch. Additionally, compared with responders, non-responders showed a higher proportion of patients with pre-existing chronic respiratory disease ($7.7\%$ vs. $66.7\%$, $$p \leq 0.003$$). Accordingly, compared with responders, non-responders may have shown a worse response to prone positioning with respect to improvements in gas exchange capacity and ventilation-perfusion ratio mismatch due to organic problems, including reduced lung compliance and impaired diffusion. Notably, obesity tended to be more prevalent among responders than among non-responders ($53.8\%$ vs. $11.1\%$, $$p \leq 0.074$$), which is consistent with a previous report on patients with obesity and ARDS [18]. Since the prone position is effective in gravity-dependent alveolar collapse, the release of pressure from the abdominal organs is more pronounced in patients with obesity. Moreover, it may ameliorate the decrease in functional residual capacity caused by increased abdominal pressure [18]. Although obesity is a risk factor for severe COVID-19 [19], patients with severe COVID-19 who have obesity may benefit from prone therapy as the preferred positioning strategy. Other studies have reported no evidence of efficacy [20], and understanding COVID-19-specific symptoms and sample size requires future research. The disease course of severe COVID-19 is considered secondary to a systemic hyperinflammatory “cytokine storm” rather than reflective of lung damage directly caused by the virus [21]. Furthermore, muscle protein imbalance occurs when critically ill patients are forced to take to bed rest [22]. Excessive release of inflammatory cytokines and stress hormones leads to decreased protein synthetic capacity and increased degradation, which decreases muscle mass [23,24]. In our study, the mean total MRC score after extubation was 33 points, which was lower than the cut-off value for ICU-acquired weakness (48 points) [25], with no significant between-group difference (36 vs. 26 points, $$p \leq 0.203$$) (Table 3). A previous study reported that $70\%$ of patients had limb muscle weakness after extubation [26], which is a lower incidence than that in our study. Mobilization and electrical muscle stimulation under intubation performed in this previous study may have led to better results than those in the present study. The accumulated doses of sedatives and muscle relaxants vary according to the duration of intubation. Large doses can cause hemodynamic instability [27] and rapid development of myopathy [22]. Therefore, their prolonged use may adversely influence ADLs after extubation. The shorter duration of intubation among responders may have minimized the accumulation of sedatives as well as the deterioration in respiratory and limb muscle strength. Accordingly, this allowed a faster recovery in ADLs and better outcomes than those in non-responders. Systemic inflammation, hypoxemia, prolonged bed rest, and extensive medications in critically ill patients with COVID-19 can cause muscle weakness, fatigue, and decreased exercise tolerance [28]. Therefore, low-impact, high-frequency physical therapy interventions in a confined space within the red zone are recommended early after extubation. This study has several limitations. First, this was a single-center, small-scale retrospective study. The sample size was limited by our study design, given the exclusion of $84.2\%$ ($\frac{139}{165}$) of non-ventilated patients with COVID-19 admitted to our hospital. Therefore, this may have an effect on the statistical results. Second, genomic analysis of SARS-CoV-2 viruses was unavailable. The study period overlapped with the peak of the prevalence of the delta strain in our country [29]. Since different mutant variants have different severities and pathologies, the clinical outcomes of patients may vary according to the mutant variant. Finally, we did not measure cytokine levels among patients with severe COVID-19 since the required facilities were unavailable at our center. Collaboration with specialized facilities should be pursued in the future. Extensive prevention of severe disease leads to better outcomes; therefore, further research is warranted to inform the prevention of severe disease in the moderate stage [7,30]. It is imperative to re-examine the type of physiotherapy intervention administered during intubation and perform objective assessments within the red zone. ## Conclusions This study provides new information on the effects of first-session prone positioning on short-term outcomes in patients with severe COVID-19 requiring ventilator management. We found that improved response to the first-session prone positioning was related to better ADLs and outcomes at discharge in patients with severe COVID-19. Therefore, the prognosis of treatment and rehabilitation in patients with severe COVID-19 may be predicted at an early treatment stage. Further research is needed to confirm these results in patients with different mutant strains of SARS-CoV-2 and elucidate the role of cytokine levels in the relationships between responsiveness to first-session prone positioning and short-term outcomes. ## References 1. Berlin DA, Gulick RM, Martinez FJ. **Severe Covid-19**. *N Engl J Med* (2020) **383** 2451-2460. PMID: 32412710 2. Alhazzani W, Møller MH, Arabi YM. **Surviving Sepsis Campaign: guidelines on the management of critically ill adults with coronavirus disease 2019 (COVID-19)**. *Intensive Care Med* (2020) **46** 854-887. PMID: 32222812 3. Kharat A, Simon M, Guérin C. **Prone position in COVID 19-associated acute respiratory failure**. *Curr Opin Crit Care* (2022) **28** 57-65. PMID: 34750322 4. Guérin C, Reignier J, Richard JC. **Prone positioning in severe acute respiratory distress syndrome**. *N Engl J Med* (2013) **368** 2159-2168. PMID: 23688302 5. Behesht Aeen F, Pakzad R, Goudarzi Rad M, Abdi F, Zaheri F, Mirzadeh N. **Effect of prone position on respiratory parameters, intubation and death rate in COVID-19 patients: systematic review and meta-analysis**. *Sci Rep* (2021) **11** 14407. PMID: 34257366 6. Sryma PB, Mittal S, Madan K, Mohan A, Hadda V, Tiwari P, Guleria R. **Reinventing the wheel in ARDS: awake proning in COVID-19**. *Arch Bronconeumol (Engl Ed)* (2020) **56** 747-749 7. Ehrmann S, Li J, Ibarra-Estrada M. **Awake prone positioning for COVID-19 acute hypoxaemic respiratory failure: a randomised, controlled, multinational, open-label meta-trial**. *Lancet Respir Med* (2021) **9** 1387-1395. PMID: 34425070 8. Touchon F, Trigui Y, Prud'homme E. **Awake prone positioning for hypoxaemic respiratory failure: past, COVID-19 and perspectives**. *Eur Respir Rev* (2021) **30** 210022. PMID: 33952601 9. Coppo A, Bellani G, Winterton D. **Feasibility and physiological effects of prone positioning in non-intubated patients with acute respiratory failure due to COVID-19 (PRON-COVID): a prospective cohort study**. *Lancet Respir Med* (2020) **8** 765-774. PMID: 32569585 10. Zarantonello F, Andreatta G, Sella N, Navalesi P. **Prone position and lung ventilation and perfusion matching in acute respiratory failure due to COVID-19**. *Am J Respir Crit Care Med* (2020) **202** 278-279. PMID: 32437251 11. Binda F, Rossi V, Gambazza S. **Muscle strength and functional outcome after prone positioning in COVID-19 ICU survivors**. *Intensive Crit Care Nurs* (2022) **69** 103160. PMID: 34789437 12. Tani-Sassa C, Iwasaki Y, Ichimura N. **Viral loads and profile of the patients infected with SARS-CoV-2 delta, alpha, or R.1 variants in Tokyo**. *J Med Virol* (2022) **94** 1707-1710. PMID: 34825717 13. Shah S, Vanclay F, Cooper B. **Improving the sensitivity of the Barthel index for stroke rehabilitation**. *J Clin Epidemiol* (1989) **42** 703-709. PMID: 2760661 14. Yang Y, Wang L, Liu J, Fu S, Zhou L, Wang Y. **Obesity or increased body mass index and the risk of severe outcomes in patients with COVID-19: a protocol for systematic review and meta-analysis**. *Medicine (Baltimore)* (2022) **101** 0 15. Church S, Rogers E, Rockwood K, Theou O. **A scoping review of the Clinical Frailty Scale**. *BMC Geriatr* (2020) **20** 393. PMID: 33028215 16. Kanda Y. **Investigation of the freely available easy-to-use software 'EZR' for medical statistics**. *Bone Marrow Transplant* (2013) **48** 452-458. PMID: 23208313 17. Saeki S. **Author's postscript: update on the guidance of the "Amendments to the Act of the Protection of Personal Information" for Japanese medical facilities in May, 2022**. *J Epidemiol* (2022) **32** 439. PMID: 35718444 18. De Jong A, Molinari N, Sebbane M. **Feasibility and effectiveness of prone position in morbidly obese patients with ARDS: a case-control clinical study**. *Chest* (2013) **143** 1554-1561. PMID: 23450309 19. Ninomiya T, Otsubo K, Hoshino T. **Risk factors for disease progression in Japanese patients with COVID-19 with no or mild symptoms on admission**. *BMC Infect Dis* (2021) **21** 850. PMID: 34419004 20. Wolf M, Alladina J, Navarrete-Welton A. **Obesity and critical illness in COVID-19: respiratory pathophysiology**. *Obesity (Silver Spring)* (2021) **29** 870-878. PMID: 33533193 21. Kim JS, Lee JY, Yang JW. **Immunopathogenesis and treatment of cytokine storm in COVID-19**. *Theranostics* (2021) **11** 316-329. PMID: 33391477 22. Sagarra-Romero L, Viñas-Barros A. **COVID-19: short and long-term effects of hospitalization on muscular weakness in the elderly**. *Int J Environ Res Public Health* (2020) **17** 8715. PMID: 33255233 23. McClafferty B, Umer I, Fye G. **Approach to critical illness myopathy and polyneuropathy in the older SARS-CoV-2 patients**. *J Clin Neurosci* (2020) **79** 241-245. PMID: 33070904 24. Phillips SM, Dickerson RN, Moore FA, Paddon-Jones D, Weijs PJ. **Protein turnover and metabolism in the elderly intensive care unit patient**. *Nutr Clin Pract* (2017) **32** 112-120 25. Patel BK, Pohlman AS, Hall JB, Kress JP. **Impact of early mobilization on glycemic control and ICU-acquired weakness in critically ill patients who are mechanically ventilated**. *Chest* (2014) **146** 583-589. PMID: 25180722 26. Medrinal C, Prieur G, Bonnevie T. **Muscle weakness, functional capacities and recovery for COVID-19 ICU survivors**. *BMC Anesthesiol* (2021) **21** 64. PMID: 33653264 27. Guérin C, Albert RK, Beitler J. **Prone position in ARDS patients: why, when, how and for whom**. *Intensive Care Med* (2020) **46** 2385-2396. PMID: 33169218 28. Soares MN, Eggelbusch M, Naddaf E. **Skeletal muscle alterations in patients with acute Covid-19 and post-acute sequelae of Covid-19**. *J Cachexia Sarcopenia Muscle* (2022) **13** 11-22. PMID: 34997689 29. **National Institute of Infectious Diseases. Detection of coronavirus disease-2019 by strain through genomic surveillance**. (2022) 30. Perez-Nieto OR, Escarraman-Martinez D, Guerrero-Gutierrez MA. **Awake prone positioning and oxygen therapy in patients with COVID-19: the APRONOX study**. *Eur Respir J* (2022) **59** 2100265. PMID: 34266942
--- title: Prevalence and Associated Clinical Features of Type 1 Diabetes Mellitus Among Children Presented to a Tertiary Health Care Center of Himalayan Foothills journal: Cureus year: 2023 pmcid: PMC10041128 doi: 10.7759/cureus.35435 license: CC BY 3.0 --- # Prevalence and Associated Clinical Features of Type 1 Diabetes Mellitus Among Children Presented to a Tertiary Health Care Center of Himalayan Foothills ## Abstract Introduction Diabetes Mellitus (DM) is a complex metabolic disorder characterized by chronic hyperglycemia. Knowing its prevalence, associated clinical features, and complications is essential for diagnosing children having diabetes-like clinical features. Since there is a limited study from India and no similar study from this geographical part, the present study was carried out. Material and method *It is* a cross-sectional study, which includes children aged 1-18 years presented to the pediatric outpatient department (OPD), inpatient department (IPD), and emergency with clinical features of Type 1 Diabetes Mellitus (T1DM). The enrolled cases were assessed for confirmation of T1DM, and clinical features and associated complications were recorded in the case record form. Result A total of 218 children with clinical features of T1DM were enrolled, out of which 32 ($14.7\%$) had T1DM. Among the 32 T1DM patients, 31 ($96.9\%$) of the participants presented with polyuria, 29 ($90.6\%$) had polydipsia, and 13 ($40.6\%$) had polyphagia. Out of 32 children, 3 ($9.38\%$) had diabetic neuropathy, and 1 ($3.1\%$) had diabetic retinopathy. Conclusion We found that many children with diabetes have clinical features of T1DM and uncontrolled blood sugar. This emphasizes the need for early detection and treatment to prevent long-term complications. ## Introduction Diabetes Mellitus (DM) is a complex metabolic disorder characterized by chronic hyperglycemia. The American Diabetes Association (ADA) classified diabetes into four major types: Type 2 diabetes, type 1 diabetes, and other specific types, including monogenic diabetes, drug or chemical-induced diseases of the exocrine pancreas, infection-induced, and gestational diabetes. The two most common forms of diabetes are Type 1 Diabetes Mellitus (T1DM), characterized primarily by deficiency of insulin secretion because of autoimmune pancreatic β-cell damage, and Type 2 Diabetes Mellitus (T2DM), which occurs due to insulin resistance along with β-cell impairment [1]. According to International Diabetes Federation Atlas 2019, the incidence of T1DM in children and adolescents aged less than 15 years is increasing in nearly all parts of the world, with an estimated increase of $3\%$ per annum with geographic differences. India has the highest number of cases of T1DM in children and adolescents in terms of incidence and prevalence, with incident cases being 15900 per annum and estimated prevalent cases being 95600 per annum in 2019 [2]. The prevalence of T1DM has been studied by several methods; school records, national registries, hospital records, and surveys in different age groups in the developing world. The prevalence varies from as low as 0.09 per 1000 in China to as high as 3.40 in the United Kingdom [3]. " Incidence studies are available from many countries and reveal rates ranging from $\frac{36.5}{100}$,000 in Finland and $\frac{36.6}{100}$,000 in Sardinia (Italy) to 0.1-$\frac{4.6}{100}$,000 in China and $\frac{0.4}{100}$,000 in Thailand" [4-6]. Studies have been done in different parts of India to determine the prevalence of T1DM and associated complications. However, those studies are primarily based in southern India and are very few. Since there are limited studies from India and no similar study was done from this geographical part, the present study was carried out. ## Materials and methods It is a hospital-based observational study. The Institute Ethics Committee approved the study protocol (AIIMS/IEC/$\frac{20}{706}$). All children (1-18 years) presenting to AIIMS Rishikesh pediatric Inpatient Department (IPD), Outpatient Department (OPD), and Emergency department with clinical features of Type 1 Diabetes Mellitus were included in the study. The participants were provided with a participant information document; written consent was obtained and included in the study as per the inclusion and exclusion criteria. Inclusion criteria included: Children of age 1 year to 18 years; *Prior diagnosis* of Type 1 Diabetes Mellitus; Children with clinical features suggestive of Type 1 Diabetes Mellitus *Exclusion criteria* includes: Refusal of consent; Children <1 year of age; Other types of diabetes- Type 2 diabetes, Drug-induced, Pancreatic diabetes Once children were included in the study, a baseline assessment was done, which included the following: Relevant history as per the predesigned proforma; Relevant Examination. The cases were diagnosed as T1DM according to the International Society for Pediatric and Adolescent Diabetes (ISPAD) 2018 guidelines, clinical features of diabetes mellitus, and plasma glucose concentration ≥ 200 mg/dL. The socioeconomic status was categorized by using the updated version [2020] of the modified Kuppuswamy socioeconomic scale(based on occupation, education of family head, and family income); score ranges from < 5, 5-10, 11-15, 16-25, and 26-29 for lower, upper lower, lower middle, upper middle, and upper socioeconomic class, respectively. In confirmed cases, laboratory investigations were done for albumin creatinine ratio, lipid profile, free T4 (FT4) and Thyroid-stimulating hormone (TSH), tissue transglutaminase IgA (tTGA), HbA1c, and serum glutamic oxaloacetic transaminase/ serum glutamate pyruvate transaminase/ gamma-glutamyl transferase (SGOT/SGPT/GGT). The confirmed cases of T1DM were screened for associated complications; diabetic neuropathy, retinopathy, and nephropathy, *Statistical analysis* Data was entered using a Microsoft Excel spreadsheet and analyzed using professional statistics/SPSS version 23 for windows. Descriptive statistics were shown as means ± standard deviations for normal distribution values and medians/IQRs for continuous skewed distribution values and frequencies and percentages/proportions for categorical variables. Chi-square or Fisher exact test was used for categorical variables, and an independent t-test for numerical variables. A $5\%$ probability (p-value less than 0.05) was considered statistically significant for all comparisons. ## Results Two hundred eighteen children presented with clinical features of diabetes mellitus were included in the study over 18 months. Out of 218 study participants, $59.6\%$ [130] were from Uttarakhand, $39.4\%$ [86] of the participants belonged to Uttar Pradesh, and $0.5\%$ were from Delhi and Haryana. Out of 218 participants, $14.7\%$ [32] had T1DM, and $85.3\%$ [186] of the study participants did not have confirmed T1DM. $58.3\%$ [127] of the participants were males, and $41.7\%$ [91] were females, with male to female ratio of 1.4:1. Among confirmed cases of T1DM, 17($53.1\%$) were male, and 15($46.9\%$)were female with male to female ratio of 1.13:1. Majority of the study participants ($46.3\%$) belong to the lower middle class. Among confirmed T1DM cases, 18($56.2\%$) and 11($34.4\%$) cases belonged to Lower middle and upper lower socioeconomic status, respectively. Table 1 shows the baseline demographics of the study participants. **Table 1** | Parameters | Type 1 Diabetes | Type 1 Diabetes.1 | p-value | | --- | --- | --- | --- | | Parameters | Yes (n = 32) | No (n = 186) | p-value | | Age (Years) | 10.53 ± 4.50 | 8.99 ± 5.22 | 0.1221 | | Age | | | 0.0922 | | 1-5 Years | 5 (15.6%) | 61 (32.8%) | | | 5-12 Years | 16 (50.0%) | 62 (33.3%) | | | 12-18 Years | 11 (34.4%) | 63 (33.9%) | | | Gender | | | 0.5242 | | Male | 17 (53.1%) | 110 (59.1%) | | | Female | 15 (46.9%) | 76 (40.9%) | | | Socio-Economic Status | | | 0.0503 | | Upper | 1 (3.1%) | 0 (0.0%) | | | Upper Middle | 2 (6.2%) | 39 (21.0%) | | | Lower Middle | 18 (56.2%) | 83 (44.6%) | | | Upper Lower | 11 (34.4%) | 57 (30.6%) | | | Lower | 0 (0.0%) | 7 (3.8%) | | Among the 32 T1DM patients, 31 ($96.9\%$) of the participants presented with polyuria, 29 ($90.6\%$) had polydipsia, and 13 ($40.6\%$) had polyphagia. 3 ($9.4\%$) of the participants had fatigue, and 19 ($59.4\%$) patients presented with weakness. 14 ($43.8\%$) participants complained of weight loss. 8 ($25.0\%$) had nausea, and 13 ($40.6\%$) had vomiting at presentation. 12 ($37.5\%$) had lethargy, and 5 ($15.6\%$) presented with altered mental status. 5 ($15.6\%$) were presented with DKA. 2 ($6.2\%$) of the participants had presented with a fruity odor. None of the patients presented with malaise or were in a coma, as shown in Figure 1. **Figure 1:** *Summary of Clinical Presentation of children with type 1 Diabetes mellitus* The mean age of diagnosis among children with T1DM was 8.80 years. The Majority of the children had a high level of blood sugar at the time of enrolment in the study, with a median (IQR) value of 496 mg/dL (444.25-545). Of 32 cases with T1DM, 2 ($7.7\%$) also took oral hypoglycemic agents. The mean HbA1C (%) value in children with T1DM was 13.5±2.8. Most children {$$n = 20$$ ($62.5\%$} had HbA1C between 10-$15\%$. 3($9.4\%$) and 9($28.1\%$) children had HbA1C between 7-$10\%$ and >$15\%$, respectively. All 32 patients were screened for diabetic-associated complications; diabetic retinopathy and neuropathy were present in 1($3.1\%$) and 3($9.38\%$) cases, respectively. No children had features of diabetic nephropathy and macrovascular disease. The cases were screened for the associated autoimmune conditions; none had evidence of autoimmune hypothyroidism or autoimmune hepatitis. Out of 32 cases of T1DM, 2 ($6.25\%$) had raised tTGA levels (>3 times the upper limit), although duodenal biopsy could not be done in those cases. ## Discussion This study is a hospital-based observational study conducted in a tertiary health care center in Uttarakhand state of India. T1DM incidence is increasing in children and adolescents, as per the International Diabetes Federation Atlas 2019, in all parts of the world. The prevalence of T1DM in school children was $1.467\%$ out of 92,047 in a survey conducted by the government of India under the National Program [7]. Studies have been done in different parts of India to determine the prevalence of T1DM and associated complications. However, those studies are mostly based in southern India and are very few. Hence, we assessed the prevalence of T1DM in our center. Out of 218 participants, $14.7\%$ [32] of the participants had T1DM; a total of 16,344 children were presented to the Paediatrics department. This reveals that the hospital prevalence of T1DM is 1.95 per 1000 children. There are few studies and registries available in India which show the prevalence; the study done by Kalra S et al. in Haryana state reported a prevalence of $\frac{22.22}{100}$,000 population (5-16 years) while in the 0-5 years age group prevalence is $\frac{3.82}{100}$,000 [8]. Similarly, a study by Ramachandran A et al. in 1992 in South India found a prevalence of $\frac{0.26}{1000}$ in children <15 years of age [9]. In our study among diabetic children ($$n = 32$$), the mean age at the diagnosis was 8.8 ± 4.09 years. This is low compared to the finding by Praveen PA et al., with a mean age of diagnosis of 12.9 ± 6.5 years, and similar to a study done in Western India by Sharma B et al. with a mean age of 10.0 ± 3.63 years [10,11]. SEARCH registry of the US also shows lower mean age of diagnosis, i.e., 10 ± 4.5 years [12,13]. Our study's most common clinical presentation was osmotic symptoms (polyuria ($96.9\%$) and polydipsia ($90.6\%$)) followed by weakness ($59.4\%$), weight loss ($43.8\%$), polyphagia and vomiting ($40.6\%$) both and lethargy ($37.5\%$). Another less common presentation was nausea, altered mental status, fatigue, and fruity odor. In the study conducted by Praveen PA et al. at the Indian Council of Medical Research, the Majority of children presented with polyuria, polydipsia, and weight loss ($28.8\%$) as compared to osmotic symptoms alone, which was ($26.5\%$) [10]. This shows the importance of awareness about the common clinical features of T1DM among society and treating physicians for screening and early identification of T1DM. Among diagnosed cases of T1DM, $53.1\%$ and $46.9\%$ were male and female, respectively. Other studies from India, like the YDR registry and the SEARCH registry from the US, show similar findings, with a male proportion of $52.9\%$ and $53.3\%$, respectively [12,13]. On screening for associated complications, we found 1($3.1\%$) child had retinopathy, and 3 had neuropathy. No children had features of diabetic nephropathy and macrovascular disease. The mean (SD) LDL level was 94.38 (34.43) mg/dL, and 2 children had increased LDL levels of >100 mg/dL and were given dietary advice and repeat LDL levels on follow-up. In India, Sudhanshu S et al., in their study in Lucknow, found diabetic nephropathy in $3\%$, diabetic retinopathy in $3.6\%$, and raised LDL (>100 mg/dL) in $34\%$ subjects out of 164 who were screened for complications [14]. Another study from South India by Ramachandran A et al. revealed nephropathy in $7.1\%$, sensory neuropathy in $3\%$, and 13.4 % of children with diabetic retinopathy [15]. All children were screened for autoimmune conditions associated with T1DM, i.e., autoimmune hypothyroidism, autoimmune hepatitis, and celiac disease. All the patients were screened negative for autoimmune hypothyroidism and hepatitis. 2 ($6.25\%$) children had raised tTGA levels (>3 times the upper limit). The study by Sharma B et al., which enrolled 150 children, showed that $24.8\%$ of children had celiac disease and $14.1\%$ and $3.3\%$ had hypothyroidism and Grave's disease, respectively [11]. The prevalence of T1DM-associated autoimmune conditions in the Indian context is still poorly studied. In our study majority of children having T1DM belong to lower middle {$$n = 18$$($56.2\%$)} and upper lower class {$$n = 11$$ ($34.4\%$)}. This is different from the Indian YDR and SEARCH registry, which found that $60.8\%$ and $53.6\%$ of children with T1DM are from high socioeconomic status, respectively. A possible explanation for this; is increased diagnosis among these children being frequently referred to our government institutions. The Diabetes Control and Complications Trial and its follow-up Epidemiology of Diabetes Interventions and Complications study reported that intensified insulin therapy, along with support and education, results in better long-term glucose control and delays the complications of T1DM [16-18]. The study's possible limitations are a small sample size and a hospital-based study. Since it is a time-bound study, no follow-up was done for subsequent associated complications. ## Conclusions We found a significant number of children present with diabetes having clinical features like T1DM. It even involves lower and middle socioeconomic children, emphasizing the need for a surveillance system, early detection, and treatment to prevent long-term complications. ## References 1. Weber DR, Jospe N.. **Type 1 Diabetes Mellitus**. *Nelson Textbook Of Pediatrics* (2019) **21** 3019-3023 2. **IDF Diabetes Atlas 9th edition**. (2022) 3. **Diabetes in America**. (2022) 4. Karvonen M, Viik-Kajander M, Moltchanova E, Libman I, LaPorte R, Tuomilehto J. **Incidence of childhood type 1 diabetes worldwide. Diabetes Mondiale (DiaMond) Project Group**. *Diabetes Care* (2000) **23** 1516-1526. PMID: 11023146 5. Patterson CC, Dahlquist G, Soltész G, Green A. **Variation and trends in incidence of childhood diabetes in Europe**. *Lancet* (2000) **11** 873-876 6. Unachak K, Tuchinda C. **Incidence of type 1 diabetes in children under 15 years in northern Thailand, from 1991 to 1997**. *J Med Assoc Thai* (2001) **84** 923-928. PMID: 11759972 7. Kumar KM. **Incidence trends for childhood type 1 diabetes in India**. *Indian J Endocrinol Metab* (2015) **19** 0-5 8. Kalra S, Kalra B, Sharma A. **Prevalence of type 1 diabetes mellitus in Karnal district, Haryana state, India**. *Diabetol Metab Syndr* (2010) **2** 14. PMID: 20214794 9. Ramachandran A, Snehalatha C, Abdul Khader OM, Joseph TA, Viswanathan M. **Prevalence of childhood diabetes in an urban population in South India**. *Diabetes Res Clin Pract* (1992) **17** 227-231. PMID: 1425162 10. Praveen PA, Madhu SV, Viswanathan M. **Demographic and clinical profile of youth onset diabetes patients in India-results from the baseline data of a clinic based registry of people with diabetes in India with young age at onset-[YDR-02]**. *Pediatr Diabetes* (2021) **22** 15-21. PMID: 31885113 11. Sharma B, Nehara HR, Saran S, Bhavi VK, Singh AK, Mathur SK. **Coexistence of autoimmune disorders and type 1 diabetes mellitus in children: an observation from Western Part of India**. *Indian J Endocrinol Metab* (2019) **23** 22-26. PMID: 31016148 12. Hockett CW, Praveen PA, Ong TC. **Clinical profile at diagnosis with youth-onset type 1 and type 2 diabetes in two pediatric diabetes registries: SEARCH (United States) and YDR (India)**. *Pediatr Diabetes* (2021) **22** 22-30. PMID: 31953884 13. Hamman RF, Bell RA, Dabelea D. **The SEARCH for diabetes in youth study: rationale, findings, and future directions**. *Diabetes Care* (2014) **37** 3336-3344. PMID: 25414389 14. Sudhanshu S, Nair VV, Godbole T. **Glycemic control and long-term complications in pediatric onset type 1 diabetes mellitus: a single-center experience from Northern India**. *Indian Pediatr* (2019) **56** 191-195. PMID: 30954988 15. Ramachandran A, Snehalatha C, Sasikala R. **Vascular complications in young Asian Indian patients with type 1 diabetes mellitus**. *Diabetes Res Clin Pract* (2000) **48** 51-56. PMID: 10704700 16. Lauritzen T. **Pharmacokinetic and clinical aspects of intensified subcutaneous insulin therapy**. *Dan Med Bull* (1985) **32** 104-118. PMID: 3924487 17. Frid A, Gunnarsson R, Güntner P, Linde B. **Effects of accidental intramuscular injection on insulin absorption in IDDM**. *Diabetes Care* (1988) **11** 41-45. PMID: 3276476 18. Frid A, Ostman J, Linde B. **Hypoglycemia risk during exercise after intramuscular injection of insulin in thigh in IDDM**. *Diabetes Care* (1990) **13** 473-477. PMID: 2190773
--- title: 'Squishing sound heard following an intra-articular shoulder injection with fluid and air is associated with higher efficacy: A retrospective analysis' authors: - Jan M.A. Mens - Ronald T.M. van Kalmthout journal: Journal of Back and Musculoskeletal Rehabilitation year: 2022 pmcid: PMC10041411 doi: 10.3233/BMR-210360 license: CC BY 4.0 --- # Squishing sound heard following an intra-articular shoulder injection with fluid and air is associated with higher efficacy: A retrospective analysis ## Abstract ### BACKGROUND: Accuracy of blind intra-articular injections for the shoulder is rather low. It is unclear whether accurate injections for capsulitis of the shoulder are more effective than inaccurate injections. ### OBJECTIVE: It has been hypothesized that a squishing sound following an intra-articular injection with a mixture of air and fluid means that the injection was accurately placed and that the efficacy of accurately placed injections is greater than that of inaccurate injections. The aim of the present study was to test the hypothesis that a squishing sound following an injection predicts a better clinical result. ### METHODS: Files were selected of patients with capsulitis of the shoulder, who were treated with an intra-articular injection containing a mixture of triamcinolone, lidocaine, and air. After the injection, the shoulder was moved to determine whether a squishing sound could be produced. Efficacy was measured after two weeks according to the Patient Global Impression of Change scale. Differences in efficacy between injections with and without a squishing sound were expressed as an odds ratio. ### RESULTS: Sixty-one patients were selected. Squishing was heard after 47 injections ($77\%$). Two weeks after the injection, a positive outcome was reported by 49 patients ($80\%$). When squishing was heard, the effect was positive in 42 of the 47 patients ($89\%$) and when no squishing was heard, the effect was positive in 7 of the 14 patients ($50\%$). The odds ratio was 8.4 ($95\%$ CI 2.1–34.0; $$p \leq 0.003$$). ### CONCLUSION: Efficacy of injections with a mixture of triamcinolone, lidocaine, and air for capsulitis of the shoulder is significantly greater when a squishing sound was heard after the injection. We hypothesize that squishing is related to accuracy and accuracy to efficacy. A future study with X-ray arthrography is needed to verify both hypotheses. ## Introduction Capsulitis of the shoulder is a common condition in middle-aged people. Capsulitis is characterized by pain, especially at night, and limitation of shoulder movement in all directions [1]. Risk factors for developing capsulitis of the shoulder are: higher age, female sex, and diabetes [2]. The natural course of capsulitis has been described as evolving through three phases: pain, stiffness, and recovery [3]. According to old studies the duration of the first phase ranges from 10–36 weeks, the stiffness phase from 4–12 months, and the recovery phase from 5–26 months and the entire course 12–48 months [3]. However, recent studies show that full recovery may take multiple years [3]. Most studies recommend intra-articular injections for treatment of pain, and exercise for treatment of stiffness [4, 5, 6]. A recent overview summarized eight meta-analyses of randomized controlled trials regarding conservative treatment options for capsulitis of the shoulder [7]. The authors concluded that an intra-articular injection with a corticosteroid confers greater pain relief in the short-term (0–12 weeks) than a placebo injection. However, accuracy of blind intra-articular injections for the shoulder joint is often rather low [8]. In a systematic review, it was shown that in about half the selected studies the score was lower than $90\%$ and sometimes far below $90\%$. In one study, the accuracy was only $26.8\%$ [9]. In all those studies contrast dye was added to the injected fluid, and X-rays or MRI were used in order to determine accuracy. In 1991, Jacobs et al. suggested another method for determination of accuracy [10]. They wrote: “… intra-articular injection of a small volume of steroid into the shoulder may be less reliable than when a larger volume containing air is injected. Air in the shoulder results in a ‘squelch’ when the joint is subsequently moved.” According to these authors, the squishing sound following injection of a mixture of air and fluid indicates that the injection has been given correctly. After reading the article by Jacobs et al., we always add 2 ml of air to the fluid when injecting corticosteroids into the shoulder joint. Since then, we also make a note in the patient’s file indicating whether or not a squishing sound was heard. The aim of the present study was to test the hypothesis that squishing after an injection predicts a better clinical result. More specifically the aim was to investigate the relation between squishing (yes/no) and the Patient Global Impression of Change two weeks after the injection. ## Design We retrospectively analyzed the files of patients with phase-1 capsulitis of the shoulder who were treated with an intra-articular injection in the shoulder in our clinic between July 1, 2010 and July 1, 2016. This study was approved by the Institutional Review Board (IRB) of Erasmus University Rotterdam, MEC-2019-0138. The IRB concluded that written informed consent was not needed since the study was retrospective and the data were collected and analyzed anonymously. ## Selection of files Inclusion of patients was based on medical history, clinical examination, and ultrasound investigation. If indicated, an X-ray was taken. Data were collected from patients meeting all of the following criteria: 1) pain in the acromial region (with or without radiation into the arm); 2) restriction of passive anteflexion, and active internal rotation; 3) at least a 20-degree restriction of passive external rotation and of glenohumeral abduction; 4) pain at passive anteflexion as well as external and internal rotation; 5) no clinical signs of subacromial pain syndrome (SAPS) or osteoarthritis. We excluded patients who had a history of fracture of the humerus and/or scapula, patients with a rheumatic disease, bilateral shoulder pain, and those who had had an intra-articular injection into the studied shoulder within the past 3 months. Patients with full-thickness rotator-cuff rupture and/or peri-articular calcifications larger than 3 mm were also excluded. Sample size calculation with alpha 0.05 and beta 0.80 showed that the number of needed dossiers was 88. This value was based upon the estimated proportion of patients with squishing was $67\%$, efficacy was $85\%$ in case of squishing and $50\%$ in those without. Moreover, with the assumption that $25\%$ of the dossiers had to be excluded. ## Protocol The shoulder joint was treated with a blind injection using the method described by Cyriax [11]: “… the patient sitting and the shoulder internally rotated. The needle punctured the skin 2 cm below the point where the lateral edge of the acromion and the lower edge of spine of the scapula meet. Then the needle was moved in the direction of the point of the coracoid process until the humeral head was felt with the needle.” Some resistance is felt as the capsule is pierced, and the patient usually complains of pain at that point [10]. A mixture containing 0.5 ml triamcinolone 40 mg/ml, 3.5 ml lidocaine $1\%$, and 2 ml air was then injected. A 50 mm 22G needle was used. After the injection and removal of the needle, the physician rotated the slightly abducted arm internally and externally a few times to check for a squishing sound. The efficacy of the injection was measured after 2 weeks by asking the patients to give their impression on a six-point Patient Global Impression of Change (PGIC), which was documented in their files. A response of ‘much better’ or ‘completely recovered’ was defined as positive; ‘somewhat better,’ ‘unchanged,’ ‘somewhat worse’ and ‘much worse’ as negative. Some patients scored their improvement by giving a percentage. In such cases, $50\%$ improvement or higher was scored as positive and a score < $50\%$ as negative. In addition to the extent of restriction of range of motion (in degrees) and squishing (present or absent) data were collected regarding factors that might influence the success of the intervention: age, sex, diabetes (present or absent), and duration of symptoms in months. ## Statistical analysis Normally distributed continuous variables are presented as mean and standard deviation; non-normally distributed continuous variables as median and interquartile range. Differences between responders and non-responders are analyzed with an independent samples t-test for normally distributed continuous variables (age), with a Mann-Whitney test for non-normal distributions (duration of complaints, restriction of external rotation and abduction) and with a chi-square test for categorical variables (gender and diabetes). Differences in efficacy between injections with and without a squishing sound are expressed as odds ratios computed in a univariate logistic regression with the clinical effect (positive or negative) as the dependent variable. Odds ratios are presented with a $95\%$ confidence interval ($95\%$ CI). Analyses were performed using SPSS version 27. $P \leq 0.05$ is considered statistically significant. ## Squishing In the 6-year period, 88 patients fulfilled all inclusion criteria (Fig. 1 and Table 1). Two patients were excluded. Efficacy data were not recorded in 25 dossiers. Mostly because patients cancelled the appointment for follow-up. Hence, 61 cases were available for analysis. A squishing sound was heard following 47 of these 61 injections ($77\%$). Figure 1.Flow diagram. ## Efficacy Two weeks after the injection, none of the 61 patients reported worsening of their symptoms. A positive effect was reported by 49 patients ($80\%$) and a negative effect by 12 patients ($20\%$). Table 1Characteristics of the 86 included patientsVariableResponders $$n = 61$$Non-responders $$n = 25$$Age in years, mean (standard deviation)60 [10]54 [6]aGender, percentage women5156Diabetes, percentage84Duration of complaints in months, median (IQR)5 [6]4 (3.8)Restriction of external rotation in degrees, median (IQR)40 [20]40 [20]Restriction of abduction in degrees, median (IQR)40 [15]40 [23]Squishing (%)7780IQR = Interquartile range. aDifferences between groups significant $$p \leq 0.006.$$ ## Squishing and efficacy When a squishing sound was heard, the effect was positive in 42 of the 47 patients ($89\%$) and when no squishing sound was heard, the effect was positive in 7 of the 14 patients ($50\%$). A comparison of these outcomes showed an odds ratio of 8.4 ($95\%$ CI 2.1–34.0; $$p \leq 0.003$$) in the univariate analysis with Nagelkerke pseudo R2= 0.22. ## Check for selection bias due to missing data Twenty-five cases were lost to follow up. Besides the fact that the non-responders were on average 6 years younger than the responders, no significant differences were found between responders and non-responders (Table 1). ## Discussion The present study showed that patient scores on the PGIC scale reported two weeks after a corticosteroid injection were strongly related to squishing. Selection bias due to missing data seemed unlikely. To explain this strong association between squishing and efficacy, we hypothesize that squishing is related to accuracy and accuracy to efficacy. ## Squishing and accuracy It is our hypothesis that squishing is related to accuracy. Two previous publications used the squishing sound as evidence that the injected mixture had reached the joint cavity of the shoulder [10, 12]. The association between squishing and accuracy has never been proven for the shoulder, as far as we know. The following facts support our hypothesis: A study in the knee showed that after injection of a mixture of local anesthetic, corticosteroid, contrast dye, and 1 to 2 cc of air, a squishing sound was heard after $85\%$ of intra-articular injections and in none after injections placed extra-articularly [13]. Glattes e.a. hypothesized that the noise in the knee “… is produced as the air passes from the femoral-tibial compartment to the suprapatellar cavity.” [ 13] We theorize that the noise in the shoulder arises when the air will be moved during rotations from one part of the joint cavity to another part. Sounds following injection of a mixture of air and fluid into a joint after double-contrast arthrography are well-known. The American College of Radiology writes on their patient-information site that after arthrography “… the patient may … hear gurgling when the joint is moved” [14], and, as far as we know, squishing sounds have never been heard after intra-articular injections with fluid only. In case of inaccuracy or leakage of blind posterior injections, the fluid appears almost always in the infraspinatus and/or teres minor muscle belly [15, 16, 17, 18]. It is theoretically unlikely that an injection of a mixture of fluid and air in the belly of those muscles could result in a squishing sound when the shoulder is moved by the physician. ## Accuracy and efficacy It is our hypothesis that accuracy is related to efficacy. That inaccuracy reduces efficacy seems obvious, but a firm association between accuracy and efficacy was, as far as we know, not been assessed in previous studies [19, 20, 21, 22]. In two of those studies, a strong conclusion could not be drawn because in the analysis, the results of the patients with capsulitis were pooled with those of patients with various other diagnoses [19, 20]. In two other studies, the efficacy of accurate injections was not significantly greater than that of inaccurate injections [21, 22]. These results could have been due to the rather small number of inaccurate injections (seven and eleven). Another explanation for the weak association between accuracy and efficacy in the aforementioned studies could be that leakage of fluid outside the joint was not considered as a possible explanation for inefficacy. Leakage after injection is quite common. A systematic review comprising 21 publications mentioned the appearance of extra-articular contrast in $1.0\%$–$51.0\%$ of the cases [8]. Rutten et al. injected 50 shoulders guided by ultrasound and 50 by fluoroscopy, reporting a minimal amount of extra-articular contrast in $24\%$ of all procedures, but a massive amount in $27\%$ of them [23]. In most studies on accuracy, the appearance of contrast fluid in the joint was defined as a successful injection. However, this definition is debatable. Rutten et al. showed that even significant leakage of contrast did not compromise the diagnostic quality of the arthrogram [23]. However, significant leakage may reduce efficacy. It is theoretically possible that the production of a squishing sound after injection could mean that a large part of the fluid has reached the joint cavity without significant leakage. This could be a better indicator for correct placement than sufficiency of fluid reaching the joint, for purposes of producing an acceptable arthrogram. The present study has much in common with a former study comparing the effect of ultrasonography guided and blindly given intra-articular injections [25]. In that study ultrasonographic given injections were more effective to reduce pain and to improve range of motion and function two weeks after the injections. The authors suggest that the injection technique in their study was related to accuracy and accuracy to efficacy, but a check for accuracy was missing. ## Limitations One limitation of the study is that the patients were not ‘blinded’ for the squishing sound. Theoretically they could have been more prone to report a positive result when they heard the squishing sound. Another limitation is that we used only one effect measure (PGIC) and only one evaluation. Most prospective studies use two or three measures to determine clinical changes. Although the use of only one effect measure is a limitation, patient’s opinion is very responsive and, at least in phase 1, seems to be more relevant for the patient than surrogate measures such as improvement of range of motion or strength [26, 27, 28]. Moreover, the retrospective design and the small amount of only 12 patients with a negative result are limitations. A blinded control group with an injection of 4 ml placebo-fluid with 2 ml air could have made the conclusions more robust. ## Recommendations The present study indicates that a prospective study should be performed, with a blinded observer for squishing and a blinded radiographic control for intra-articular air. It is a challenge to find a method to make the researcher more objective in the assessment of squishing. ## Conclusion Hearing a squishing sound following an injection with a mixture of triamcinolone, lidocaine, and air correlates with a better short-term efficacy in patients with phase-1 capsulitis of the shoulder. It is quite probable that squishing is related to accuracy, and accuracy to a positive result. New studies are needed to assess the definitive causalities between squishing, accuracy and efficacy. ## Ethical approval This study was approved by the Institutional Review Board (IRB) of Erasmus University Rotterdam, MEC-2019-0138. ## Funding None to report. ## Informed consent Not applicable. ## Author contributions Both authors contributed to the formulating of the hypothesis. Both authors contributed to the collection of the data. JM was responsible for the statistical analysis, the raw writing of the manuscript and the correspondence. RK was responsible for the final version of the manuscript. ## Conflict of interest Both authors had no conflict of interest. ## References 1. Sun Y, Zhang P, Liu S, Li H, Jiang J, Chen S, Chen J. **Intra-articular steroid injection for frozen shoulder: A systematic review and meta-analysis of randomized controlled trials with trial sequential analysis**. *Am J Sports Med* (2017.0) **45** 2171-2179. PMID: 28298050 2. Huang YP, Fann CY, Chiu YH, Yen MF, Chen LS, Chen HH, Pan SL. **Association of diabetes mellitus with the risk of developing adhesive capsulitis of the shoulder: A longitudinal population-based follow up study**. *Arthritis Care Res (Hoboken)* (2013.0) **65** 1197-202. PMID: 23281342 3. Wong CK, Levine WN, Deo K, Kesting RS, Mercer EA, Schram GA, Strang BL. **Natural history of frozen shoulder: Fact or fiction? A systematic review**. *Physiotherapy* (2017.0) **103** 40-47. PMID: 27641499 4. Ando A, Sugaya H, Hagiwara Y, Takahashi N, Watanabe T, Kanazawa K, Itoi E. **Identification of prognostic factors for the nonoperative treatment of stiff shoulder**. *Int Orthop* (2013.0) **37** 859-64. PMID: 23503671 5. Challoumas D, Biddle M, McLean M, Millar NL. **Comparison of treatments for frozen shoulder: A systematic review and meta-analysis**. *JAMA Netw Open* (2020.0) **3** e2029581. PMID: 33326025 6. Hanchard NCA, Goodchild L, Brealey SD, Lamb SE, Rangan A. **Physiotherapy for primary frozen shoulder in secondary care: Developing and implementing stand-alone and post operative protocols for UK FROST and inferences for wider practice**. *Physiotherapy* (2020.0) **107** 150-160. PMID: 32026815 7. Lädermann A, Piotton S, Abrassart S, Mazzolari A, Ibrahim M, Stirling P. (2021.0) 8. Simoni P, Grumolato M, Malaise O, Preziosi M, Pasleau F, de Lemos Esteves F. **Are blind injections of gleno-humeral joint (GHJ) really less accurate imaging-guided injections? A narrative systematic review considering multiple anatomical approaches**. *Radiol Med* (2017.0) **122** 656-675. PMID: 28523634 9. Sethi PM, Kingston S, Elattrache N. **Accuracy of anterior intra-articular injection of the glenohumeral joint**. *Arthroscopy* (2005.0) **21** 77-80. PMID: 15650670 10. Jacobs LG, Barton MA, Wallace WA, Ferrousis J, Dunn NA, Bossingham DH. **Intra-articular distension and steroids in the management of capsulitis of the shoulder**. *BMJ* (1991.0) **302** 1498-501. PMID: 1855018 11. Cyriax J, Russel G. (1977.0) 164-166 12. Jacobs LG, Smith MG, Khan SA, Smith K, Joshi M. **Manipulation or intra-articular steroids in the management of adhesive capsulitis of the shoulder? A prospective randomized trial**. *J Shoulder Elbow Surg* (2009.0) **18** 348-53. PMID: 19393928 13. Glattes RC, Spindler KP, Blanchard GM, Rohmiller MT, McCarty EC, Block J. **A simple, accurate method to confirm placement of intra-articular knee injection**. *Am J Sports Med* (2004.0) **32** 1029-31. PMID: 15150053 14. [14] American College of Radiology. Radiologyinfo.org for patients. https://www.radiologyinfo.org/en/info.cfm?pg=arthrog. 15. Catalano OA, Manfredi R, Vanzulli A, Tomei E, Napolitano M, Esposito A, Resnick D. **MR arthrography of the glenohumeral joint: Modified posterior approach without imaging guidance**. *Radiology* (2007.0) **242** 550-4. DOI: 10.1148/radiol.2422051964 16. Gokalp G, Dusak A, Yazici Z. **Efficacy of ultrasonography-guided shoulder MR arthrography using a posterior approach**. *Skeletal Radiol* (2010.0) **39** 575-9. DOI: 10.1007/s00256-009-0793-8 17. Koivikko MP, Mustonen AO. **Shoulder magnetic resonance arthrography: A prospective randomized study of anterior andposterior ultrasonography-guided contrast injections**. *Acta Radiol* (2008.0) **49** 912-7. DOI: 10.1080/02841850802203585 18. Ogul H, Bayraktutan U, Ozgokce M, Tuncer K, Yuce I, Yalcin A. **Ultrasound-guided shoulder MR arthrography: Comparison of rotator interval and posterior approach**. *Clin Imaging* (2014.0) **38** 11-17. PMID: 24119385 19. Eustace JA, Brophy DP, Gibney RP, Bresnihan B, FitzGerald O. **Comparison of the accuracy of steroid placement with clinical outcome in patients with shoulder symptoms**. *Ann Rheum Dis* (1997.0) **56** 59-63. PMID: 9059143 20. Hegedus EJ, Zavala J, Kissenberth M, Cook C, Cassas K, Hawkins R, Tobola A. **Positive outcomes with intra-articular glenohumeral injections are independent of accuracy**. *J Shoulder Elbow Surg* (2010.0) **19** 795-801. PMID: 20655766 21. Raeissadat SA, Rayegani SM, Langroudi TF, Khoiniha M. **Comparing the accuracy and efficacy of ultrasound-guided versus blind injections of steroid in the glenohumeral joint in patients with shoulder adhesive capsulitis**. *Clin Rheumatol* (2017.0) **36** 933-940. PMID: 27566475 22. White AET, Tuite JD. **The accuracy and efficacy of shoulder injections in restrictive capsulitis**. *J Orthop Rheumatol* (1996.0) **9** 37-40 23. Rutten MJ, Collins JM, Maresch BJ, Smeets JH, Janssen CM, Kiemeney LA, Jager GJ. **Glenohumeral joint injection: A comparative study of ultrasound and fluoroscopically guided techniques before MR arthrography**. *Eur Radiol* (2009.0) **19** 722-30. PMID: 18958474 24. Kim YS, Lee HJ, Lee DH, Choi KY. **Comparison of high- and low-dose intra-articular triamcinolone acetonide injection for treatment of primary shoulder stiffness: A prospective randomized trial**. *J Shoulder Elbow Surg* (2017.0) **26** 209-215. DOI: 10.1016/j.jse.2016.09.034 25. Lee HJ, Lim KB, Kim DY, Lee KT. **Randomized controlled trial for efficacy of intra-articular injection for adhesive capsulitis: Ultrasonography-guided versus blind technique**. *Arch Phys Med Rehabil* (2009.0) **90** 1997-2002. DOI: 10.1016/j.apmr.2009.07.025 26. de Jong BA, Dahmen R, Hogeweg JA, Marti RK. **Intra-articular triamcinolone acetonide injection in patients with capsulitis of the shoulder: A comparative study of two dose regimens**. *Clin Rehabil* (1998.0) **12** 211-5. PMID: 9688036 27. Lorbach O, Anagnostakos K, Scherf C, Seil R, Kohn D, Pape D. **Nonoperative management of adhesive capsulitis of the shoulder: Oral cortisone application versus intra-articular cortisone injections**. *J Shoulder Elbow Surg* (2010.0) **19** 172-9. PMID: 19800262 28. Tveitå EK, Ekeberg OM, Juel NG, Bautz-Holter E. **Responsiveness of the shoulder pain and disability index in patients with adhesive capsulitis**. *BMC Musculoskelet Disord* (2008.0) **9** 161. PMID: 19055757
--- title: A scoping review on the impact of the COVID-19 pandemic on physical activity and sedentary behavior in Saudi Arabia authors: - Kelly R. Evenson - Shaima A. Alothman - Christopher C. Moore - Mariam M. Hamza - Severin Rakic - Reem F. Alsukait - Christopher H. Herbst - Baian A. Baattaiah - Reem AlAhmed - Hazzaa M. Al-Hazzaa - Saleh A. Alqahtani journal: BMC Public Health year: 2023 pmcid: PMC10041481 doi: 10.1186/s12889-023-15422-3 license: CC BY 4.0 --- # A scoping review on the impact of the COVID-19 pandemic on physical activity and sedentary behavior in Saudi Arabia ## Abstract ### Background In Saudi Arabia, stay-at-home orders to address the coronavirus disease 2019 (COVID-19) pandemic between March 15 and 23, 2020 and eased on May 28, 2020. We conducted a scoping review to systematically describe physical activity and sedentary behavior in Saudi Arabia associated with the timing of the lockdown. ### Methods We searched six databases on December 13, 2021 for articles published in English or Arabic from 2018 to the search date. Studies must have reported data from Saudi Arabia for any age and measured physical activity or sedentary behavior. ### Results Overall, 286 records were found; after excluding duplicates, 209 records were screened, and 19 studies were included in the review. Overall, 15 studies were cross-sectional, and 4 studies were prospective cohorts. Three studies included children and adolescents (age: 2–18 years), and 16 studies included adults (age: 15–99 years). Data collection periods were < = 5 months, with 17 studies collecting data in 2020 only, one study in 2020–2021, and one study in 2021. The median analytic sample size was 363 (interquartile range 262–640). Three studies of children/adolescents collected behaviors online at one time using parental reporting, with one also allowing self-reporting. All three studies found that physical activity was lower during and/or following the lockdown than before the lockdown. Two studies found screen time, television watching, and playing video games were higher during or following the lockdown than before the lockdown. Sixteen adult studies assessed physical activity, with 15 utilizing self-reporting and one using accelerometry. Physical activity, exercise, walking, and park visits were all lower during or following the lockdown than before the lockdown. Six adult studies assessed sedentary behavior using self-report. Sitting time (4 studies) and screen time (2 studies) were higher during or following the lockdown than before the lockdown. ### Conclusions Among children, adolescents, and adults, studies consistently indicated that in the short-term, physical activity decreased and sedentary behavior increased in conjunction with the movement restrictions. Given the widespread impact of the pandemic on other health behaviors, it would be important to continue tracking behaviors post-lockdown and identify subpopulations that may not have returned to their physical activity and sedentary behavior to pre-pandemic levels to focus on intervention efforts. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12889-023-15422-3. ## Background On March 11, 2020, the World Health Organization declared coronavirus disease 2019 (COVID-19) a pandemic. To tackle viral spread, person-to-person exposure was limited by imposing public movement restrictions. As a result, individuals' physical activity was impacted; there was a drastic change in exercising, walking, and bicycling for transportation and leisure [1]. Sedentary behavior often replaced time spent engaging in physical activity, as the stay-at-home restrictions reduced such opportunities. Even when restrictions were lifted, public facilities (indoor and outdoor) were closed or limited to curtail crowding [2]. The immediate impact of stay-at-home orders on physical activity has been documented through self-reported and device-based metrics. For example, in a convenience sample of 13,503 adults from 14 countries, self-reported moderate-to-vigorous physical activity declined by $41\%$ following pandemic-related restrictions [3]. The decline was greater for work activities than leisure activities, for those with more baseline physical activity compared to those with lesser physical activity, and for younger adults compared to older adults. Other studies using activity trackers indicated immediate declines in step counts (i.e., an indicator of walking levels) attributable to the pandemic in Japan [4], Singapore [5], the United States [6], and worldwide [7–10], although the magnitude of results varied between countries. There is evidence of similar impacts on physical activity among children [1, 11]. The immediate impact of stay-at-home orders has also been documented for sedentary behaviors, characterized as activities while awake with an energy expenditure of 1.5 metabolic equivalents or less while sitting, reclining, or lying [12]. A systematic review found five studies of apparently healthy children and 26 studies of apparently healthy adults, all of which reported increased sedentary behavior primarily due to the pandemic [11]. Another review included 19 studies of children/adolescents and 45 studies of adults and found consistent increases in sedentary behavior during the pandemic period, with larger gains among children compared to adults [13]. The acute global declines in physical activity and increases in sedentary behavior are of concern since engagement in physical activity improves bone health and weight status for children (age 3 to 5 years), improves cognitive function for children/adolescents (age 6 to 13 years), and reduces the risk of mortality, chronic diseases (e.g., certain cancers, cardiovascular disease, obesity), excessive weight gain, fall-related injuries, and dementia for adults [14]. Engagement in sedentary behavior acutely induces vascular dysfunction [15] and, in the long term, increases the risk of mortality, cardiovascular disease, and type 2 diabetes [14, 16]. Physical activity is one of the priority goals of Saudi Arabia’s Vision 2030 given its importance in chronic disease prevention and health benefits [17]. The Saudi Sports for All Federation outlines the vision, framework, goals, and corresponding strategies to help people of all ages become more physically active [18]. The 2021 Household Sports Practice Survey indicated that $48\%$ of the Saudi Arabia population engaged in at least 30 min/week of physical activity, which was higher than the 2019 prevalence of $45\%$ [19]. Two reviews that included studies through early 2018 found that the prevalence of physical inactivity in Saudi Arabia ranged from $55\%$-$96\%$ among children/adolescents, $73\%$-$91\%$ among female adults, and $50\%$-$85\%$ among male adults [20]. A third review included studies published between 2018–2021 that used population-based sampling in the Saudi Arabia population [21]. Among children and adolescents, approximately 80–$90\%$ did not attain at least 60 min/day of moderate-to-vigorous physical activity, while for adults approximately 50–$95\%$ had a low or insufficient physical activity that did not meet the World Health Organization’s recommendations [22, 23]. In this same review, about 50–$80\%$ of children and adolescents engaged in at least two hours/day of screen time or sedentary behavior, while for adults about half had a sitting time of five hours/day or more. Due to the COVID-19 pandemic, in Saudi Arabia stay-at-home orders were implemented on March 15, 2020, with a suspension of travel for non-essential work, followed by a nationwide curfew from March 23 to April 5, 2020 [24]. The curfews were extended until May 28, 2020, when most regions began easing the curfews. We conducted a scoping review to systematically describe physical activity and sedentary behavior (i.e., physical behaviors) among people of all ages in Saudi Arabia from the pre-COVID-19 pandemic period to the post-movement restriction period. A review focused on Saudi Arabia can bring an understanding of the impact of the pandemic on physical behaviors, identify groups that may not have returned to their pre-pandemic levels, and highlight potential needs for future research and surveillance. ## Search methods The scoping review protocol was developed in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) statement [25]. Since this review focused on documenting the pandemic-induced change in the prevalence of physical activity and sedentary behavior, and was a scoping rather than a systematic review [26], the protocol was not required to be registered with any platforms. The completed PRISMA-ScR checklist can be found in Supplement 1 [25]. We searched six databases (Cochrane Library, Global Health [EBSCO], PubMed, Scopus, SPORTDiscus [EBSCO], and the World Health Organization Global Index Medicus) on March 3, 2022, with the search strategy detailed in Supplement 2. After manually removing duplicate citations with reference management software, two authors independently screened all titles/abstracts and full-text articles for inclusion using Covidence systematic review software (www.covidence.org; Veritas Health Innovation; Melbourne, Australia), with discrepancies resolved by consensus. ## Inclusion and exclusion criteria Inclusion criteria were studies that either assessed physical activity or sedentary behavior before and during the pandemic, or asked participants to recall how their physical behaviors changed due to the pandemic. We included studies published between March 1, 2020 and March 3, 2022 that reported on physical activity or sedentary behavior in Saudi Arabia. We included observational studies published in either English or Arabic. Both self-reported and device-based measures of physical activity and sedentary behavior were included in the review. We excluded studies that did not collect data from March 2020 to March 2022 or did not report the impact of the COVID-19 pandemic on physical behaviors. We also excluded studies that did not report data specifically for Saudi Arabia. We excluded studies that did not include a measure of physical activity or sedentary behavior. For example, studies that discussed “intention to exercise” were excluded since that was not a direct measure of physical behaviors, such as in Alshareef et al. [ 27]. We excluded studies of hospitalized or institutionalized adults. Grey literature, dissertations, commentaries, and conference proceedings were also excluded. ## Abstraction and analysis Once the study inclusion was confirmed, one rater abstracted study details and a second rater checked the abstraction, with discrepancies resolved by consensus. The abstraction tool included the study name, study purpose, data collection period, region, sampling methods, target population, inclusion and exclusion criteria, and sample size. Information abstracted on the sample included age, gender, and nationality. We classified age groups based on the predominant age included in the study: children 1 to 12 years, adolescents 13 to 17 years, and adults 18 years and older [28]. For physical activity and sedentary behavior, we collected results at various time points (e.g., before and during lockdown) and the methods used (e.g., questionnaire and definitions). The quality of each study was assessed to identify strengths and weaknesses. This was performed by having two reviewers answer ten questions about each study, with disagreements between the raters resolved by consensus. We used the Joanna Briggs Institute Prevalence Critical Appraisal Tool Checklist for Prevalence Studies to assess study quality [29], making modifications and additions to fit the purposes of this review (Supplement 3). In recognition that objective quality assessment tools treat each threat to validity equally, [30] we did not intend to provide a total score for each study. Instead, we used the quality assessment results to focus on the specific threats to validity identified across the included studies. ## Results A total of 286 records were found; after manually removing 77 duplicates across databases, 209 records were screened for inclusion (Fig. 1). In the title and abstract screening stage, 162 records were excluded as irrelevant. After a full-text review of 47 reports, we included 19 studies [24, 31–48], all published in English. Fig. 1PRISMA flow diagram of the search strategy and results for the scoping review. From: Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71. https://doi.org/10.1136/bmj.n71. For more information, visit: http://www.prisma-statement.org/ Most studies were cross-sectional in design, except for four studies that collected data at two [35, 36, 45] or seven [43] time points (Table 1). Data collection mainly occurred in 2020, except for the study by Abdulaziz et al.[32] (data collection: September 2020 to February 2021) and Almugti et al. [ 38] (data collection: July 2021 to August 2021). Most data collection occurred for three months or fewer, except for Abdulaziz et al. [ 32] that collected data over five months. Analytic sample sizes ranged from 65 [46] to 30,134 [43], with a median of 363 (interquartile range 262 to 640).Table 1Description of each study included in the review ($$n = 19$$)First Author, YearDates of Data CollectionData Collection Period in MonthsData Collection Time PointsLocation in Saudi ArabiaSampling ProcedureAnalytic Sample SizeAge in Years Mean (SD) or Median [IQR]Age Range in yearsFemale PercentNationality PercentChildren/AdolescentsAlmugti, 2021 [38]July to August 202121Jizan $30.0\%$ Riyadh $19.4\%$, Eastern Province $12.7\%$, Medina $11.4\%$, Asir $9.8\%$, Makkah $8.6\%$, Bahah $4.1\%$, Qassim $1.8\%$, Hail $0.7\%$, Najran $0.5\%$, Tabuk $0.5\%$, Al Jawf $0.3\%$, Northern Borders $0.2\%$Through social media in Saudi Arabia (e.g., Twitter, WhatsApp, and Facebook)6519 [4]3 to $1541.0\%$All parents were SaudiHanbazaza, 2021 [44]June 22 to July 22, 202011All different regions across Saudi ArabiaConducted using online survey distributed via social media (WhatsApp, Twitter, and Snapchat) in Arabic280NR6 to $1548.9\%$NRAl Agha, 2021 [34]April 2020 to June 202031JeddahPatients with type 1 diabetes were contacted via an online virtual pediatric endocrine outpatient clinic15012.5 (4.6)2 to $1872.0\%$NRAdultsAbd El-Fatah, 2021 [31]October 202011Makkah region $69.7\%$, Eastern region $20.4\%$, Riyadh region $9.9\%$Mass email via collaborating authors networks, social media engagement (WhatsApp and Twitter), and snowball sampling36336.3 (8.5)20 to $5965.6\%$NRAbdulaziz, 2021 [32]September 2020 through February 202151Qassim province (cities of Buraidah, Unaizah, and AlRass)Each city selected $10\%$ of the PHCC in each city using simple random sampling (10P HCCs total; 5 from Buraidah, 3 from Unaizah, and 2 from AlRass). An average of 20 attendees were selected daily at each PHCC and interviewed29938.6 (13.1)18 to missing$72.6\%$NRAbdulsalam, 2021 [33]NRNR1JeddahOnline questionnaire was distributed using social media (Facebook, Twitter, Instagram, and WhatsApp) and email communication472NR18 to $5968.0\%$NRAl Fagih, 2020 [35]Pre-lockdown (February 2 to March 12, 2020) and lockdown (March 12 to April 19, 2020)2 pre and 2 post2RiyadhContacted patients at a cardiac center8265 [58 to 72]NR$35.4\%$NRAlfawaz, 2021 [24]Two weeks during and after Ramadan (May 11 to June 6, 2020)11NR, implies all regionsOnline questionnaire was cascaded to different social media outlets throughout Saudi Arabia196535.2 (13.1)15 to $7553.0\%$$83.1\%$ Saudi, $16.9\%$ non-SaudiAl-Musharaf, 2021 [36]Pre-lockdown (February to April 2019) and lockdown (April to May 2020)3 pre and 2 post2RiyadhRandomly selected women aged between 19 and 30 years with no history of medical issues from several colleges29720.7 (1.4)19 to $30100.0\%$All SaudiAlharthi, 2021 [37]NRNR1All regionsNR; enrolled only Saudis between the ages of 18 and 60384NR18 to $6050.6\%$All SaudiAlotaibi, 2021 [39]March 1 to April 30, 202021All regions*Canvassing on social media, local radio stations, and through university mailing lists22,053NR18 to $4044.5\%$NRAlqurashi, 2021 [40]March to May 202031Region: Eastern $60.1\%$, Western $23.5\%$, Central $15.9\%$, Northern $0.5\%$Google forms of the questionnaire were sent to participants to complete via email and social media platforms (Twitter, Telegram, and WhatsApp)208NR18 to $5688.9\%$Saudi $99.5\%$, Non-Saudi $0.5\%$Bakhsh, 2021 [41]Two week period between June and early July 202011Region: Western $70\%$, Central $16\%$, Eastern $8\%$, Southern $5\%$, Northern $1\%$Online questionnaire distributed on various platforms (WhatsApp, Twitter, and email). Questionnaire link was sent to the authors’ relatives, friends, and neighbors to participate in the study and to share the link with their contacts2,255NR18 to missing$64.0\%$Saudi $91\%$, Non-Saudi $9\%$Barwais, 2020 [42]April 9 to April 25, 202011Region: Makkah $73.4\%$ and Medinah $26.6\%$Convenience sample recruited through email invitations and on social media sites (Twitter, Telegram, and WhatsApp groups)24433.8 (7.7)18 to $5036.9\%$NRBinDhim, 2021 [43]Total of 7 waves of collection in 2020grouped into 4 waves of ~ 3 months each7 waves; grouped into 4 quartersAll 13 administrative regionsProportional quota sampling using phone interviews with an age- and gender-stratified random selection of phone numbers from a list generated from the Sharik Association for Health Research (a database of > 80,000 individuals interested in participating in health research, covering all 13 administrative regions)30,134 over all waves; Quarter 1 = 7050, Quarter 2 = 11,289, Quarter 3 = 5183, Quarter 4 = 661236.5 (13.5)18 to $9951.2\%$ overallAll SaudiJalal, 2021 [45]Before lockdown (March 2020) and during lockdown (June 2020)1 pre and 1 post2Al-AhsaStudents of undergraduate programs were selected from their registration numbers by using a simple random technique62820.5 (1.9)18 to $3070.9\%$NRMagliah, 2021 [46]Three days after the lockdown ended in Saudi Arabia (June 21–23, 2020)11JeddahWeb survey (Google Forms) distributed via social media to patients who were actively attending the specialized insulin pump clinic6530 (7.9)18 to missing$70.8\%$All SaudiŠagát, 2020 [47]May 10 to May 17, 202011RiyadhSimple randomization to select 1000 potential participants on the Riyadh municipality forum groups that were available on social media who were then sent the online questionnaire46335.6 (9.8)18 to $6444.1\%$$71\%$ Saudi citizens and $29\%$ foreignersSultan, 2021 [48]August to September 202021NR, implies all regionsNon-probability convenient sample; online survey distributed using social media33840 [IQR not reported]30 to $4479.0\%$NRAbbreviations: IQR Interquartile range, NA Not applicable, NR Not reported, PHCC Primary health care clinics, SD Standard deviation*obtained directly from authors Studies were conducted in certain regions of Saudi Arabia such as Al-Ahsa [45], Jeddah [33, 34, 46], Qassim [32], Riyadh [35, 36, 47], or in multiple regions [31, 40–42]. Other studies included participants from the entire country, either stated explicitly [37, 38, 43, 44] or implied [24, 39, 48]. The most common sampling procedure was some combination of convenience sampling through social media platforms, email, radio, or mailing lists [24, 31, 33, 38–42, 44, 46, 48]. Other studies used country-level proportional quota [43] or simple random sampling through health clinics [32, 34, 35], universities [36, 45], or municipal forum groups [47]. The sampling technique was not reported for one study [37]. Three studies included children and adolescents [34, 38, 44], 1 study included participants aged 15 to 75 years (which we assigned to the adult group) [24], and the remaining 15 studies included adults at least 18 years of age [31–33, 35–37, 39–43, 45–48]. All studies enrolled both males and females, except Al-Musharaf et al. [ 36] who enrolled females only. Some studies enrolled only participants of Saudi nationality [36–38, 43, 46], while others included Saudi and non-Saudi nationalities residing in the country [24, 40, 41, 47]. Ten studies did not report the nationality of participants [31–35, 39, 42, 44, 45, 48]. ## Impact on physical behaviors of children/adolescents Three studies recruited children/adolescents through social media platforms [38, 44] or a pediatric endocrine clinic [34]. The age range was wide for all three studies, from a minimum age of 2 [34], 3 [38], or 6 [44] to a maximum age of 15 [38, 44] or 18 years [34]. All relied on parental reports, although one study also allowed self-reporting [34] (Table 2). The questionnaires were administered online and asked about the daily duration of physical activity [34], daily duration of moderate-to-vigorous physical activity [38], or frequency of participation in physical activity [44] before and during the lockdown. All three studies conducted measurements at a single time point. Table 2Physical activity results from the scoping review ($$n = 19$$)First Author, YearPhysical Activity Assessment MethodPhysical Activity Definitions UsedPhysical Activity Before LockdownPhysical Activity During and/or Following LockdownChange ReportedSummaryChildren/Adolescents Almugti, 2021 [38]Parent-reported questionnaire modified by an expert panelBased on Canadian 24-Hour Movement Guidelines for Children and Youth: At least 1 h/day of MVPAMVPA < 1 h/day $29\%$, ≥ 1 h/day $71\%$MVPA < 1 h/day $49\%$, ≥ 1 h/day $51\%$P for difference 0.001Decreased MVPA Hanbazaza, 2021 [44]Parent-reported questionnaireChange in children’s PA (increased, decreased, or remained unchanged); The number of d/wk their children participated in PA before and during lockdown (5 response options ranging from “None at all” to “5–6 times a week”) with PA categorized as ≤ 4 times/week as "not physically active" and ≥ 5 times a week as "physically active"$19.3\%$ were physically active$16.1\%$ were physically activeProportion of physically active decreased $3.2\%$ but was not statistically significant ($$p \leq 0.30$$)Decreased PA Al Agha, 2021 [34]Parent- or self-reported questionnaireDuration of PADaily duration of PA before lockdown: < 30 min $40.5\%$, 30–60 min $28.0\%$, Not practicing $27.4\%$, missing $4.1\%$*NRPA during lockdown: Decreased $66.1\%$, Increased $19.0\%$, Not affected $14.9\%$Decreased PAAdults Abd El-Fatah, 2021 [31]IPAQ-SFChanges in PA reported as no change, positive change, or negative change, see footnote below table for definitionsMedian PA 380 min/wk; Low $62.3\%$, Moderate $37.5\%$, High $0.3\%$Median PA 320 min/wk; Low $63.6\%$, Moderate $34.2\%$, High $2.2\%$Moderate PA days/wk: No change $54.5\%$, Positive change (increased PA) $25.1\%$, Negative change (decreased PA) $20.4\%$; Moderate PA min/day: No change $48.2\%$, Positive change $32.2\%$, Negative change $19.6\%$Decreased median PA Abdulaziz, 2021 [32]QuestionnaireEngagement in any PA, such as walking, going to the gym, and playing sportsNR$58.9\%$ engaging in PAChange in exercise: Increased $14.0\%$, Decreased $30.1\%$, No change $37.8\%$; Change in park visits: Increased $4.7\%$, Decreased $76.7\%$, No change $18.6\%$Decreased exercise; Decreased park visits Abdulsalam, 2021 [33]Questionnaire translated, tested, and validated by experts at the universityCategorized PA hr/week and PA level per dayUsual daily PA level very low $11.7\%$, low $17.4\%$, normal $51.5\%$, high $17.8\%$, very high $1.7\%$Usual daily PA level very low $22.2\%$, low $29.0\%$, normal $34.7\%$, high $13.1\%$, very high $0.8\%$Usual daily PA level and PA in hr/wk significantly decreased during the COVID-19 periodDecreased usual daily PA level; decreased PA hr/wk Al Fagih, 2020 [35]Uniaxial accelerometer embedded in patients' cardiac implantable device (Medtronic ICD/CRT)PA defined in h/dayMedian 2.4 h/day PAMedian 1.8 h/day PA$27.1\%$ decline in PA; change in PA occurred in the first week of March 2020 which coincides with the implementation of social distancing measuresDecreased total PA Alfawaz, 2021 [24]Questionnaire designed/revised by multidisciplinary experts and pilotedCategorized daily walking, home physical activities, weight lifting, and swimming: never, 1–2, 3–4, or > 4 daysDaily walking never $21.0\%$, 1–2 d/wk $23.6\%$, 3–4 d/wk $24.9\%$, > 4 d/wk $30.5\%$; Home PA never $42.8\%$, 1–2 d/wk $19.0\%$, 3–4 d/wk $18.0\%$, > 4 d/wk $20.1\%$Daily walking never $23.6\%$, 1–2 d/wk $22.9\%$, 3–4 d/wk $24.4\%$, > 4 d/wk $29.1\%$; Home PA never $44.6\%$, 1–2 d/wk $18.8\%$, 3–4 d/wk $17.1\%$, > 4 d/wk $19.5\%$Significant changes in walking, home physical activities with weights, and swimming (p values < 0.001)Decreased daily walking; Decreased home activities; Increased swimming Al-Musharaf, 2021 [36]GPAQ Arabic versionMeeting recommendation of ≥ 600 MET-min/week$47.5\%$ meeting recommendations$40.7\%$ meeting recommendationsTest for difference in meeting PA recommendations $$p \leq 0.08$$Decreased meeting PA recommendations Alharthi, 2021 [37]Modified IPAQ (New Zealand PAQ)Reported did or did not do exerciseDid exercise $64.2\%$Did exercise $48.9\%$Increased exercise $48.9\%$ ($$p \leq 0.01$$)Decreased "did exercise" Alotaibi, 2021 [39]QuestionnaireActive or inactive based on the WHO guidelines (150–300 min/week of moderate or 75–150 min/week of vigorous intensity PA, or some equivalent combination). MVPA further divided into < 3 d/wk or ≥ 3 d/wkInactive $77.1\%$, Active $22.9\%$; MVPA > 3 times/wk $13.0\%$, < 3 times/wk $9.9\%$Inactive $80.0\%$, Active $20.0\%$; MVPA > 3 times/wk $10.2\%$, < 3 times/wk $9.8\%$Inactive + $2.9\%$, Active -$2.9\%$; > 3 times/wk -$2.9\%$, < 3 times/wk -$0.09\%$Decreased active group; increased inactive group Alqurashi, 2021 [40]Questionnaire developed and pilotedPA before and during lockdown, asking about: 1) engagement in PA or sports before the pandemic, 2) their gym attendance, 3) whether they exercised at home during the lockdown, 4) whether their time spent exercising increased during the quarantineEngaged in PA before pandemic $59.1\%$; exercised at a gym before pandemic $28.8\%$Practice exercise at home during lockdown: never $36.5\%$, 1–3 times/wk $39\%$, 4–6 times/wk $10.1\%$, every day $14.4\%$During the quarantine period, increased exercise time at home: strongly agree $27.5\%$, agree $25.5\%$, neutral $29.8\%$, disagree $12.5\%$Increased time in exercise at home Bakhsh, 2021 [41]QuestionnaireAsked participants about changes in their level of PA, weekly frequency of PA, duration of PA per day, and types of PA performed during quarantineNRFrequency of PA: none $40\%$, 1–2 d/wk $22\%$, 3 d/wk $11\%$, 4–6 d/wk $14\%$, daily $13\%$; Duration of PA: none $40\%$, < 30 min/d $13\%$, 30 min/d $15\%$, 1 h/d $24\%$, > 1 h/d $8\%$; Type of PA: walking $65\%$ (most common), cardiorespiratory exercise $11\%$, and resistance training $7\%$Change in PA level: increased $27\%$, decreased $52\%$, no change $21\%$Decreased PA Barwais, 2020 [42]IPAQ-SFTotal MET-min/wkMean 903 (SD 755.6) MET-min/wkMean 387 (SD 397.8) MET-min/wkp for paired difference 0.001 with a large effect size ($d = 0.89$); Social contexts: significant decreases in PA performed alone ($p \leq 0.001$), with family ($p \leq 0.05$), with friends ($p \leq 0.05$), and with groups ($p \leq 0.001$)Decreased PA overall and for men and women BinDhim, 2021 [43]Questionnaire refined through linguistic validation, reliability testing, and focus group evaluationWHO/US PA Guidelines: low level of PA or acceptable PA level (≥ 150 min/wk of MPA and/or ≥ 75 min/wk VPA)Acceptable PA level Q1 (January to mid-March) $41.0\%$Acceptable PA level Q2 (mid-March to June) $26.5\%$, Q3 (July to September) $24.6\%$, Q4 (October to December) $24.6\%$Significant decline in prevalence odds of having an acceptable PA level between Q1 and each of Q2-4 (all $p \leq 0.001$) for both unadjusted and adjusted (age, gender, and region) analysisDecreased PA Jalal, 2021 [45]GPAQMeeting recommendation of > = 600 total MET-min/wkTotal PA mean 1149.2 (SD 120.08) MET-min/week; Attaining ≥ 600 MET-min/wk $52.1\%$Total PA mean 1116.5 (SD 125.3) MET-min/week; Attaining ≥ 600 MET-min/wk $47.9\%$differences: MET-min/week $$p \leq 0.0001$$, attaining ≥ 600 MET-min/wk $$p \leq 0.03$$Decreased total PA; Decreased meet PA recommendations Magliah, 2021 [46]Web survey (Google Forms) with a section asking about the impact of lockdown on different self-management behaviors, which included rating their ability to maintain PA in comparison with the pre-lockdown periodReport change in PA during vs before lockdown in five categories (Greatly decreased to greatly increased)NRNRGreatly decreased $41.5\%$, somewhat decreased $26.2\%$, no change $9.2\%$, somewhat increased $15.4\%$, greatly increased $7.7\%$Decreased PA Šagát, 2020 [47]QuestionnaireCategorize their weekly frequency of PA before and after the pandemic: none, once, 2–3 times, 4–5 times, or 6–7 times per weekDid not practice PA $7.3\%$, PA 1 time/wk $10.3\%$, PA 2–3 times/wk $35.6\%$, PA 4–5 times/wk $24.1\%$, PA 6–7 times/wk $22.7\%$Did not practice PA $20.0\%$, PA 1 time/wk $15.2\%$, PA 2–3 times/wk $25.1\%$, PA 4–5 times/wk $25.8\%$, PA 6–7 times/wk $13.9\%$Significant differences in proportions who "did not practice PA" ($$p \leq 0.001$$; increase), "practiced PA once a wk" ($$p \leq 0.02$$; increase), "practiced PA 2–3 times a wk" ($$p \leq 0.001$$; decrease), and "practiced PA 6–7 time a wk" ($p \leq 0.001$; decrease)Decreased PA Sultan, 2021 [48]Questionnaire developed after a literature review; tested reliabilityCategorized participants as not active, light activity, active, or very active (categories not further defined)Not active $5.3\%$, light activity $31.2\%$, active $54.9\%$, very active $6.8\%$Not active $19.0\%$, light activity $36.2\%$, active $37.4\%$, very active $7.4\%$"Not active" category increased significantly $p \leq 0.001$Increased "not active" categoryAbbreviations: d days, ICD/CRT Implantable cardioverter-defibrillator / cardiac resynchronization therapy, IPAQ International PA Questionnaire, GPAQ Global PA Questionnaire, MET Metabolic equivalent of task, min minutes, MPA Moderate PA, MVPA Moderate to vigorous PA, NA Not applicable, NR Not reported, PA Physical activity, wk week, VPA Vigorous physical activity, WHO World Health OrganizationIPAQ-SF November 2005 scoring protocol: Low [not moderate or high]; Moderate [either A) ≥ 3 days of vigorous activity of ≥ 20 min/day, B) ≥ 5 days of moderate-intensity activity or walking of ≥ 30 min/day, or C) ≥ 5 days of any combination of walking, moderate-intensity or vigorous intensity activities achieving ≥ 600 MET-min/week]; High [either A) vigorous-intensity activity on ≥ 3 days achieving ≥ 1500 MET-min/week or B) ≥ 7 days of any combination of walking, moderate-intensity or vigorous intensity activities achieving ≥ 3000 MET-min/week]*obtained directly from authors The lockdown was associated with a 20-percentage point decrease in moderate-to-vigorous physical activity of at least one hour/day [38] and a three-percentage point decrease in the percent classified as physically active [44]. The third study among participants aged 2 to 18 years with diabetes reported that the lockdown was associated with decreased physical activity for $66.1\%$ of the sample, increased physical activity for $19.0\%$, and no change for $14.9\%$ [34]. Two studies assessed sedentary behavior by asking parents to report time on digital screens [38] or time spent playing video games and watching television [44] (Table 3). Compared to the pre-lockdown period, the lockdown period was associated with 24.0, 35.0, and 22.5 percentage point increases in the proportion of children/adolescents on screens for > 2 h/day [38], video games for > = 3 h/day [44], and watching television for > = 4 h/day [44], respectively. Table 3Sedentary behavior results from the scoping review ($$n = 8$$)First Author, YearSedentary Behavior Assessment MethodSedentary Behavior Definitions UsedSedentary Behavior Before LockdownSedentary Behavior During and or Following LockdownChange ReportedSummaryChildren/Adolescents Almugti, 2021 [38]Parent-reported questionnaire modified by expert panelHow much time their child spent viewing digital screens, including TV, tablets, and phones; categories based on Canadian 24-Hour Movement Guidelines for Children and Youth: < = 2 h/day of screen timeUse of screens ≤ 2 h/day $37\%$, > 2 h/day $63\%$Use of screens ≤ 2 h/day $13\%$, > 2 h/day $87\%$difference $$p \leq 0.001$$Increased screen time Hanbazaza, 2021 [44]Parent-reported questionnaireHow long child played video games and watched TV per day; categorized as > = 3 h/day playing video games (considered high) and > = 4 h/day watching TV (considered high)Playing video games > = 3 h/day $40.4\%$; watching TV > = 4 h/day $21.1\%$Playing video games > = 3 h/day $75.4\%$; watching TV > = 4 h/day $43.6\%$Significant increases (p-values < 0.001) in proportions that were playing video games > = 3 h/day and were watching TV > = 4 h/dayIncreased video game time, Increased TV timeAdults Abd El-Fatah, 2021 [31]IPAQ-SFRoutine sitting in the day as no change, positive change, or negative changeDaily sitting: 1–2 h/day $20.4\%$, 3–4 h/day $27\%$, 5–6 h/day $21.5\%$, More than 6 h/day $31.1\%$,Daily sitting: 1–2 h/day $12.9\%$, 3–4 h/day $13.8\%$, 5–6 h/day $18.2\%$, More than 6 h/day $55.1\%$,Daily sitting in h/day: No change $45.5\%$, Decrease $8.3\%$, Increase $46.2\%$Increased sitting time Abdulsalam, 2021 [33]Questionnaire translated, tested, and validated by experts at the universityTime spent in front of the computer, mobile devices, television, etc < 1 h/day $9.1\%$, 1–2 h/day $24.2\%$, 3–4 h/day $33.1\%$, 5–6 h/day $21.2\%$, > 6 h/day $12.5\%$ < 1 h/day $4.7\%$, 1–2 h/day $7.6\%$, 3–4 h/day $20.3\%$, 5–6 h/day $31.1\%$, > 6 h/day $36.2\%$Significant increase (e.g., before $12.5\%$ spent > 6 h/day, but during the pandemic it became the most prevalent category ($36.2\%$)Increased screen time Al-Musharaf, 2021 [36]GPAQ Arabic versionContinuous min/day sitting or recliningMean 451.4 (SD 242.1) min/dayMean 484.9 (SD 257.2) min/dayTest for difference $$p \leq 0.07$$Increase sitting time Jalal, 2021 [45]GPAQContinuous min/day sitting or recliningMean 448.7 (SD 73.6) min/dayMean 517.8 (SD 83.0) min/dayDifferences in min/day $$p \leq 0.0001$$Increased sitting time Šagát, 2020 [47]QuestionnaireAsked to categorize their physical behavior at their job/occupationOccupation: sitting always or most of the time $30.5\%$, sitting and moving equally $27.9\%$, moving always or most of the time $42.4\%$Occupation: sitting always or most of the time $50.9\%$, sitting and moving equally $24.2\%$, moving always or most of the time $24.9\%$Significant differences in proportions "sitting always or most of the time" ($p \leq 0.001$; increase) and "moving always or most of the time" ($p \leq 0.001$; decrease) during their job/occupationIncreased sitting time at work Sultan, 2021 [48]Questionnaire tested for reliabilityDaily screen time categorized as < 1 h/day, 1–3 h/day, 4–5 h/day, or ≥ 6 h/day < 1 h/day $13.4\%$, 1–3 h/day $48.7\%$, 4–5 h/day $23.1\%$, ≥ 6 h/day $14.8\%$ < 1 h/day $7.1\%$, 1–3 h/day $29.4\%$, 4–5 h/day $28.2\%$, ≥ 6 h/day $35.3\%$Increases in proportions with ≥ 6 h/day of screen time and of social media time (p ≤ 0.001 for both)Increased screen timeAbbreviations: GPAQ Global Physical Activity Questionnaire, IPAQ -SF International Physical Activity Questionnaire short form, h hours, min minutes, SD Standard deviation, TV Television*obtained directly from authors ## Impact on physical behaviors of adults Three studies recruited adults through health clinics [32, 35], and the remaining studies recruited through social media platforms [24, 31, 33, 39–42, 46, 48], universities [36, 45], municipal forum groups [47], proportional quota sampling from throughout the country [43], or was not reported [37]. All but one study relied on self-reported physical activity assessed using questionnaires (Table 2). These included the International Physical Activity Questionnaire [31, 37], the Global Physical Activity Questionnaire [36, 45], or some other questionnaires that experts designed and pilot tested [24, 33, 40, 43], assessed for reliability [48], or not pilot tested (or unreported as such) [32, 39, 41, 46, 47]. The exception was Al Fagih et al. study [35] that enrolled patients with cardiac implantable devices and relied on the accelerometer embedded in those devices for assessing the duration of physical activity for just over a month immediately before and after lockdown. Among 82 patients, median total physical activity declined from pre-lockdown (2.4 h/day) to lockdown (1.8 h/day). The other clinic-oriented study found that $30.1\%$ and $76.7\%$ reported decreases in exercise and park visits, respectively, during the lockdown than before the lockdown [32]. Studies found that fewer adults met recommendations for physical activity [36, 39, 45] or fewer classified themselves as “active or very active” compared to before the lockdown [48]. Studies also found that the lockdown was associated with lower daily or weekly physical activity levels [31, 33, 41–43, 45–47], lower exercise [37], and less daily walking and participation in household activities [24]. Other studies found an increase in swimming [24] and an increase in time spent exercising at home [40] associated with the lockdown. One study explored the results by gender and found that lower physical activity during the lockdown compared to the pre-lockdown period was similar for both women ($$n = 90$$) and men ($$n = 154$$) [42]. Six studies assessed sedentary behavior using a questionnaire (Table 3). They found that sitting time increased, both overall [31, 33, 36, 45] and while at work [47], and daily screen time increased, all attributable to the lockdown [48]. Specifically, two studies using continuous measures reported an increase in sitting time of 33.5 min/day [36] or 69.1 min/day [45]. Three studies using a categorical measure reported that the proportion of adults with > 6 or ≥ 6 h/day of sitting time [31] or screen time [33, 48] increased by 20.5 to 24.0 percentage points. ## Quality assessment The quality assessment tool, comprising ten questions and applied to the 19 studies, is provided in Table 4, with the corresponding questions itemized in Supplement 3. All studies had data analysis with sufficient coverage ($$n = 19$$, question 5), and most measured physical behaviors in a standard way for all participants ($$n = 17$$, question 8). Most studies described study subjects and the setting in adequate detail ($$n = 14$$, question 6) and used a valid method to assess the volume of physical behaviors ($$n = 13$$, question 7). Twelve studies summarized physical behaviors using appropriate analytic methods (question 10), and about half of the studies provided sample size justification ($$n = 10$$, question 3). However, few studies sampled participants appropriately ($$n = 4$$, question 2) or provided an appropriate sampling frame to address the target population ($$n = 1$$, question 1), with most studies recruiting participants through social media platforms, professional networks, health clinics, or universities. Few studies also assessed physical behaviors at least once before and once after lockdown ($$n = 4$$, question 9) or reported an adequate response rate or appropriately managed non-response ($$n = 3$$, question 4; 15 did not report on response rate).Table 4Quality assessment results with studies listed in alphabetical order by the first author's last name ($$n = 19$$)First Author, YearQ1Q2Q3Q4Q5Q6Q7Q8Q9Q10Children/Adolescents Almugti, 2021 [38]NNNUYYYYNY Hanbazaza, 2021 [44]NNYUYYNYNY Al Agha, 2021 [34]NUNUYNNNNYAdults Abd El-Fatah, 2021 [31]NNYUYYYYNN Abdulaziz, 2021 [32]NYYYYYUUNU Abdulsalam, 2021 [33]NNNUYYYYNY Al Fagih, 2020 [35]NYNUYNYYYY Alfawaz, 2021 [24]NNYUYYYYNY Al-Musharaf, 2021 [36]NYNUYYYYYN Alharthi, 2021 [37]UUYUYYYYNN Alotaibi, 2021 [39]NNNUYYNYNY Alqurashi, 2021 [40]NNNUYYYYNY Bakhsh, 2021 [41]NNYUYYYYNY Barwais, 2020 [42]NNNUYYYYNN BinDhim, 2021 [43]YYYNYNYYYY Jalal, 2021 [45]NUYUYNYYYN Magliah, 2021 [46]NNNYYYNYNY Šagát, 2020 [47]NNYYYNNYNY Sultan, 2021 [48]NNYUYYYYNU Total Yes1410319141317412 Total Unclear13015001102 Total No1712910551155The quality assessment is available in Supplement 3 of this paperAbbreviations: N No, U Unclear, Y Yes ## Discussion This scoping review, based on 19 studies from Saudi Arabia, found consistent evidence across the available literature indicating that physical activity declined and sedentary behavior increased during the COVID-19 lockdown period compared with before. This was most consistent for children, adolescents, and adults across all studies, and was similar for men and women, as reported in one study [42]. For adults, physical activity was lower with the lockdown by approximately 5 to 15 percentage points when considering the studies that classified the proportion as either “active” or “meeting physical activity guidelines” and sedentary behavior was higher by approximately 20 to 25 percentage points for studies that classified the proportion with ~ 6 h/day or more of sitting or screen time. For children/adolescents, all three studies indicated lower physical activity or moderate-to-vigorous physical activity with the lockdown [34, 38, 44], while two studies indicated a higher proportion of children/adolescents spent time on digital screens [38], playing video games [44], and watching television during the lockdown [44]. There were a couple of notable exceptions among adults, wherein the lockdown was associated with a self-reported increase in swimming [24] and an increase in time spent exercising at home [40]. The higher time spent exercising at home was expected, since time for exercising elsewhere may have been spent at home. Another study reported home exercise among adults but found fewer home activities with the lockdown [24]. We identified four reviews of the global impact of COVID-19 on physical behaviors to compare our results. These reviews found that studies consistently reported lower self-report or device-measured physical activity [1, 11, 49] and higher sedentary behavior [1, 11, 13] associated with lockdown policies. People who were more active prior to the pandemic had larger declines in physical activity [11]. As noted by Stockwell et al.[11], these findings are despite health practitioners and various government organizations guiding how to stay active in self-quarantine and during a lockdown. The detrimental impact of the lockdown on physical activity was also documented among children/adolescent patients with diabetes [34] and adult patients with heart failure [35] in Saudi Arabia. Other reviews identified studies from different countries that indicated similar declines in physical activity as a result of the lockdown among patients with diabetes, heart failure, congenital heart disease, obesity, and neuromuscular disease [1, 11, 49]. Our review findings are also consistent with an online cross-sectional survey of 2970 adults conducted in April 2020 from the Middle East and North Africa (MENA) region, wherein no physical activity engagement increased from before the pandemic ($34.9\%$) to during the pandemic ($39.1\%$) [50]. Additionally, they found other adverse health impacts over the short term, including weight gain, longer sleep time, and higher reporting of irritability, physical and emotional exhaustion, and tension. Other studies in Saudi Arabia or the MENA region found detrimental impacts of the pandemic on physical health, including changes in eating habits [51], weight gain [51, 52], diabetes [27, 34, 53], and mental health [54–56]. Awareness of increasing physical activity is needed for children, adolescents, and adults in Saudi Arabia [21] since there has been a recent upward trend in obesity and diabetes [57–59]. Taken together with the findings from this review, there is concern over the long-term impact of these observations on physical behaviors. Increasing physical activity and decreasing sedentary behavior is a worthwhile endeavor, given its potential benefit in reducing hospitalizations, admission to an intensive care unit, and death among those with COVID-19 [60, 61]. Physical activity is impacted by the physical environment, including built and natural surroundings [62]. For example, parks provide a crucial place for physical activity and general recreation, particularly in more temperate areas of Saudi Arabia and seasonably cooler times of the year. In our review, one study in the Qassim region found that adults self-reported fewer park visits during the lockdown than before the lockdown [32]. A prospective study in the United States found an increase in park visits at the start of the pandemic, followed by a marked decline in lockdown near closed parks but not near open parks [63]. Once closed parks opened again, their usage increased to levels found at the start of the pandemic. From a worldwide perspective, park visits increased with the start of the pandemic and were lower when the government levied stay-at-home restrictions [2]. If park observation or usage data exist in Saudi Arabia, it would be valuable to investigate whether the patterns followed global trends. Furthermore, an investigation into the impact of the built environment on changing physical behaviors during the pandemic would be worthwhile [64]. ## Limitations of the studies included We found several notable limitations in the current literature that were reviewed, both from the quality assessment (Table 4) and from our observations. First, most studies were cross-sectional, relying on participants to self-report activities before and during the lockdown. Recency bias is a threat to these studies, whereby activities occurring more recently might be easier to report than those occurring more distant in time. Prospective measurement, accomplished in four studies, reduces the risk of bias [35, 36, 43, 45]. This limitation is especially pertinent for two studies that collected data in 2021 [32, 38]. Second, many studies used convenience and nonrepresentative samples, particularly relying on social media platforms for recruitment. While these studies offer the advantage of lower costs and quick access to participants, they also limit participation to those without access to social media platforms and those unwilling to participate using those recruitment channels. This further limits the generalizability of the results. Third, assessments were mostly based on self-reporting or parental-reporting using a wide variety of questionnaires, precluding our ability to summarize findings with meta-analytic techniques accurately. In fact, many questionnaires appeared non-standardized and did not provide information about the total volume of physical activity. Using valid and reliable metrics for this population would be preferred. Fourth, many studies lacked information on the nationality of their sample, and most studies did not report findings by potential modifiers, such as gender, age, socioeconomic status, region of the country, nationality, or health-associated metrics, possibly due to sample sizes. Fifth, since the period before the COVID-19 pandemic and during the lockdown were not the same seasons of the year, seasonality could confound the relationships observed. Finally, although we included three studies on children and adolescents, the age range was wide and primarily based on parental reporting, which is limited [65]. It would be helpful to document the impacts by narrower age groups to discern any differences that might have a lasting effect on physical behaviors in the future. However, despite these limitations, the findings of the lockdown’s impact on physical behaviors were largely consistent. Future studies are needed to prospectively document physical activity and sedentary behavior changes from pre-pandemic to post-lockdown. ## Strengths and limitations of this review This scoping review was comprehensive, with searches conducted in six databases, which included Arabic studies, although there were none. To our knowledge, this is the first review to describe the impacts of the COVID-19 pandemic on physical activity and sedentary behavior in Saudi Arabia among children, adolescents, and adults [66]. Based on our inclusion criteria, we accepted all papers regardless of study quality but quantified study quality using a previously developed tool. Despite the strengths of this review, several limitations also exist. The questionnaires used across studies were heterogeneous, as was how they were analyzed, limiting our ability to meta-analyze findings to summarize results. The lack of reporting of findings by sociodemographic and health-related metrics also precluded our ability to summarize across subpopulations. ## Conclusions In 2021, the Saudi Sports for All Federation set a target to decrease the prevalence of physical inactivity by $30\%$ in adults by 2030 [18]. The findings from this scoping review stress the need to improve physical activity and curtail sedentary behavior in Saudi Arabia, particularly in light of the apparent decline in physical activity and increase in sedentary behavior during and following the COVID-19 lockdown period. This is in agreement with recent worldwide reports on physical activity among children, adolescents, and adults [67–69]. Colleagues have identified the global pattern of unhealthy lifestyle behaviors (including physical behaviors) and the COVID-19 pandemic as a “syndemic”, wherein two or more health conditions or diseases negatively interact [70]. Major areas of focus to support physical activity were designated by the World Health Organization [71] and the International Society for Physical Activity and Health [72]. Some of the Gulf Cooperation Council countries have national policies and strategies to promote physical activity, but implementation is generally low [73]. Others have called for a coordinated regional effort to promote physical activity and reduce sedentary behavior [73]. Consideration should also be given to the specific barriers and facilitators of physical activity and sedentary behavior in Saudi Arabia [74, 75] and the socio-ecologic correlates relevant to this unique time period [76]. Given the widespread impact of the COVID-19 pandemic on other health behaviors, it would be important to continue tracking behaviors and identify subpopulations that may not have returned their physical activity and sedentary behavior to pre-pandemic levels to focus on intervention efforts. ## Supplementary Information Additional file 1: Table S1. Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) Checklist. Table S2. Databases searched, search structure, and search terms. Table S3. Quality assessment tool applied to each included study. ## Role of the funding/sponsor The funders of this study had no role in this scoping review. ## References 1. Park AH, Zhong S, Yang H, Jeong J, Lee C. **Impact of COVID-19 on physical activity: A rapid review**. *J Glob Health* (2022.0) **12** 05003. DOI: 10.7189/jogh.12.05003 2. Geng DC, Innes J, Wu W, Wang G. **Impacts of COVID-19 pandemic on urban park visitation: a global analysis**. *J For Res (Harbin)* (2022.0) **32** 553-67. DOI: 10.1007/s11676-020-01249-w 3. Wilke J, Mohr L, Tenforde AS, Edouard P, Fossati C, Gonzalez-Gross M, Sanchez Ramirez C, Laino F, Tan B, Pillay JD. **A Pandemic within the Pandemic? Physical Activity Levels Substantially Decreased in Countries Affected by COVID-19**. *Int J Environ Res Public Health* (2021.0) **18** 2235. DOI: 10.3390/ijerph18052235 4. Obuchi SP, Kawai H, Ejiri M, Ito K, Murakawa K. **Change in outdoor walking behavior during the coronavirus disease pandemic in Japan: A longitudinal study**. *Gait Posture* (2021.0) **88** 42-46. DOI: 10.1016/j.gaitpost.2021.05.005 5. Ong JL, Lau T, Massar SAA, Chong ZT, Ng BKL, Koek D, Zhao W, Yeo BTT, Cheong K, Chee MWL. **COVID-19-related mobility reduction: heterogenous effects on sleep and physical activity rhythms**. *Sleep* (2021.0) **44** zsaa179. DOI: 10.1093/sleep/zsaa179 6. Hunter RF, Garcia L, de Sa TH, Zapata-Diomedi B, Millett C, Woodcock J, Pentland A, Moro E. **Effect of COVID-19 response policies on walking behavior in US cities**. *Nat Commun* (2021.0) **12** 3652. DOI: 10.1038/s41467-021-23937-9 7. Tison GH, Avram R, Kuhar P, Abreau S, Marcus GM, Pletcher MJ, Olgin JE. **Worldwide Effect of COVID-19 on Physical Activity: A Descriptive Study**. *Ann Intern Med* (2020.0) **173** 767-770. DOI: 10.7326/M20-2665 8. 8.FItbit News. The impact of Coronavirus on Global Activity. Published March 23, 2020. Accessed at https://blog.fitbit.com/covid-19-global-activity/. 9. 9.The Effect of the Global Pandemic on Active Lifestyles. Written April 9, 2020. Accessed at https://www.garmin.com/en-US/blog/general/the-effect-of-the-global-pandemic-on-active-lifestyles/ 10. 10.2020: How Garmin Users Prioritized Movement in a Global Pandemic. Published February 1, 2021. Accessed at https://www.garmin.com/en-GB/blog/2020-how-garmin-users-prioritized-movement-in-a-global-pandemic/ 11. Stockwell S, Trott M, Tully M, Shin J, Barnett Y, Butler L, McDermott D, Schuch F, Smith L. **Changes in physical activity and sedentary behaviours from before to during the COVID-19 pandemic lockdown: a systematic review**. *BMJ Open Sport Exerc Med* (2021.0) **7** e000960. DOI: 10.1136/bmjsem-2020-000960 12. **Standardized use of the terms "sedentary" and "sedentary behaviours"**. *Appl Physiol Nutr Metab* (2012.0) **37** 540-542. DOI: 10.1139/h2012-024 13. Runacres A, Mackintosh KA, Knight RL, Sheeran L, Thatcher R, Shelley J, McNarry MA. **Impact of the COVID-19 pandemic on sedentary time and behaviour in children and adults: a systematic review and meta-analysis**. *Public Health* (2021.0) **18** 11286 14. 14.US Department of Health and Human Services: Physical Activity Guidelines for Americans. 2nd edition. Washington, DC: US Department of Health and Human Services. 2018. Accessed at https://health.gov/paguidelines/second-edition/. 15. Carter S, Hartman Y, Holder S, Thijssen DH, Hopkins ND. **Sedentary behavior and cardiovascular disease risk: mediating mechanisms**. *Exerc Sport Sci Rev* (2017.0) **45** 80-86. DOI: 10.1249/JES.0000000000000106 16. Katzmarzyk PT, Powell KE, Jakicic JM, Troiano RP, Piercy K, Tennant B. **Physical activity guidelines advisory c: sedentary behavior and health: update from the 2018 physical activity guidelines advisory committee**. *Med Sci Sports Exerc* (2019.0) **51** 1227-1241. DOI: 10.1249/MSS.0000000000001935 17. 17.Vision 2030. Accessed May 12, 2022 at https://www.vision2030.gov.sa/ 18. 18.Saudi Sports for All Federation: Sports for All: Overall Strategy. Accessed at www.sportsforall.com.sa. 2021: 1–29. 19. 19.Kingdom of Saudi Arabia, General Authority for Statistics: According to Households Sport Practice Survey in Saudi Arabia. Accessed June 27, 2022 at https://www.stats.gov.sa/en/news/315. 2022. 20. Alasqah I, Mahmud I, East L, Usher K. **Patterns of physical activity and dietary habits among adolescents in Saudi Arabia: A systematic review**. *Int J Health Sci (Qassim)* (2021.0) **15** 39-48. PMID: 33708043 21. 21.Evenson KR, Alhusseini N, Moore CC, Hamza MM, Al-Qunaibet A, Rakic S, Alsukait RF, Herbst CH, AlAhmed R, Al-Hazzaa HM, Alqahtani SA: Scoping review of population-based studies of physical activity and sedentary behavior in Saudi Arabia. J Phys Act Health 2023:in press. 22. Bull FC, Al-Ansari SS, Biddle S, Borodulin K, Buman MP, Cardon G, Carty C, Chaput JP, Chastin S, Chou R. **World health organization 2020 guidelines on physical activity and sedentary behaviour**. *Br J Sports Med* (2020.0) **54** 1451-1462. DOI: 10.1136/bjsports-2020-102955 23. 23.World Health Organization: WHO Guidelines on Physical Activity and Sedentary Behavior. Accessed at https://apps.who.int/iris/bitstream/handle/10665/336656/9789240015128-eng.pdf. In. Geneva, Switzerland: World Health Organization; 2020: 1–531. 24. Alfawaz H, Amer OE, Aljumah AA, Aldisi DA, Enani MA, Aljohani NJ, Alotaibi NH, Alshingetti N, Alomar SY, Khattak MNK. **Effects of home quarantine during COVID-19 lockdown on physical activity and dietary habits of adults in Saudi Arabia**. *Sci Rep* (2021.0) **11** 5904. DOI: 10.1038/s41598-021-85330-2 25. Tricco AC, Lillie E, Zarin W, O'Brien KK, Colquhoun H, Levac D, Moher D, Peters MDJ, Horsley T, Weeks L. **PRISMA extension for scoping reviews (PRISMA-ScR): checklist and explanation**. *Ann Intern Med* (2018.0) **169** 467-473. DOI: 10.7326/M18-0850 26. Munn Z, Peters MDJ, Stern C, Tufanaru C, McArthur A, Aromataris E. **Systematic review or scoping review? Guidance for authors when choosing between a systematic or scoping review approach**. *BMC Med Res Methodol* (2018.0) **18** 143. DOI: 10.1186/s12874-018-0611-x 27. Alshareef R, Al Zahrani A, Alzahrani A, Ghandoura L. **Impact of the COVID-19 lockdown on diabetes patients in Jeddah**. *Saudi Arabia Diabetes Metab Syndr* (2020.0) **14** 1583-1587. DOI: 10.1016/j.dsx.2020.07.051 28. 28.Christiansen S, Iverson C, Flanagin A, al. e: AMA Manual of Style: A Guide for Authors and Editors. Accessed August 31, 2022 at https://www.amamanualofstyle.com/, 11th edn: Oxford University Press; 2020. 29. Munn Z, Moola S, Lisy K, Riitano D, Tufanaru C. **Methodological guidance for systematic reviews of observational epidemiological studies reporting prevalence and cumulative incidence data**. *Int J Evid Based Healthc* (2015.0) **13** 147-153. DOI: 10.1097/XEB.0000000000000054 30. Savitz DA, Wellenius GA, Trikalinos TA. **The Problem with mechanistic risk of bias assessments in evidence synthesis of observational studies and a practical alternative: assessing the impact of specific sources of potential bias**. *Am J Epidemiol* (2019.0) **188** 1581-1585. DOI: 10.1093/aje/kwz131 31. Abd El-Fatah NK, Alshehri AA, Alsulami FH, Alasmari N, Osman NA. **Association between mental health outcomes and changes in lifestyle behavior index among saudi adults 16 weeks after COVID-19 pandemic lockdown release**. *Front Public Health* (2021.0) **9** 728117. DOI: 10.3389/fpubh.2021.728117 32. Abdulaziz A, Kalevaru CS. **Lifestyle changes of attendees at primary health care centers during the COVID-19 pandemic in qassim province, Saudi Arabia**. *J Family Med Primary Care* (2021.0) **10** 3838-3844. DOI: 10.4103/jfmpc.jfmpc_930_21 33. Abdulsalam NM, Khateeb NA, Aljerbi SS, Alqumayzi WM, Balubaid SS, Almarghlani AA, Ayad AA, Williams LL. **Assessment of dietary habits and physical activity changes during the full COVID-19 curfew period and its effect on weight among adults in Jeddah, Saudi Arabia**. *Int J Environ Res Public Health* (2021.0) **18** 8580. DOI: 10.3390/ijerph18168580 34. Al Agha AE, Alharbi RS, Almohammadi OA, Yousef SY, Sulimani AE, Alaama RA. **Impact of COVID-19 lockdown on glycemic control in children and adolescents**. *Saudi Med J* (2021.0) **42** 44-48. DOI: 10.15537/smj.2021.1.25620 35. Al Fagih A, Al Onazi M, Al Basiri S, Al-Kaf F, Dagriri K, Al Hebaishi Y, Samargandy S, Al Shengeiti L. **Remotely monitored inactivity due to COVID-19 lockdowns: Potential hazard for heart failure patients**. *Saudi Med J* (2020.0) **41** 1211-1216. DOI: 10.15537/smj.2020.11.25449 36. Al-Musharaf S, Aljuraiban G, Bogis R, Alnafisah R, Aldhwayan M, Tahrani A. **Lifestyle changes associated with COVID-19 quarantine among young Saudi women: A prospective study**. *PLoS ONE* (2021.0) **16** e0250625. DOI: 10.1371/journal.pone.0250625 37. Alharthi MH. **Weight control practices among the Saudi Arabian population during the Covid-19 lockdown**. *J Family Med Prim Care* (2021.0) **10** 3797-3802. DOI: 10.4103/jfmpc.jfmpc_702_21 38. Almugti HS, Alotaibi A, Almohammed A, Abuhadi R, Baeshen R, Alharthi Z, Alsharari A, Alotaibi S, Omar Y, Alturki N. **Impact of COVID-19 on Saudi Children: Special Focus on Behavioral, Social, and Emotional Aspects, 2020–2021**. *Cureus* (2021.0) **13** e19856. PMID: 34963861 39. Alotaibi AS, Boukelia B. **The effect of pre-quarantine physical activity on anxiety and depressive symptoms during the covid-19 lockdown in the kingdom of saudi arabia**. *Int J Environ Res Public Health* (2021.0) **18** 7771. DOI: 10.3390/ijerph18157771 40. Alqurashi RM. **Changes in Nutritional Habits and lifestyles during the COVID-19 Lockdown period in Saudi Arabia: A Cross-Sectional Study**. *Current Res Nutr Food Sci* (2021.0) **9** 934-944. DOI: 10.12944/CRNFSJ.9.3.20 41. Bakhsh MA, Khawandanah J, Naaman RK, Alashmali S. **The impact of COVID-19 quarantine on dietary habits and physical activity in Saudi Arabia: a cross-sectional study**. *BMC Public Health* (2021.0) **21** 1487. DOI: 10.1186/s12889-021-11540-y 42. Barwais FA. **Physical activity at home during the covid-19 pandemic in the two mostaffected cities in Saudi Arabia**. *Open Public Health Journal* (2020.0) **13** 470-476. DOI: 10.2174/1874944502013010470 43. BinDhim NF, Althumiri NA, Basyouni MH, AlMousa N, AlJuwaysim MF, Alhakbani A, Alrashed N, Almahmoud E, AlAloula R, Alqahtani SA. **Exploring the Impact of COVID-19 Response on Population Health in Saudi Arabia: Results from the "Sharik" Health Indicators Surveillance System during 2020**. *Int J Environ Res Public Health* (2021.0) **18** 5291. DOI: 10.3390/ijerph18105291 44. Hanbazaza M, Wazzan H. **Changes in eating habits and lifestyle during COVID-19 curfew in children in Saudi Arabia**. *Nutr Res Pract* (2021.0) **15** S41-s52. DOI: 10.4162/nrp.2021.15.S1.S41 45. Jalal SM, Beth MRM, Al-Hassan HJM, Alshealah NMJ. **Body mass index, practice of physical activity and lifestyle of students during COVID-19 lockdown**. *J Multidiscip Healthc* (2021.0) **14** 1901-1910. DOI: 10.2147/JMDH.S325269 46. Magliah SF, Zarif HA, Althubaiti A, Sabban MF. **Managing Type 1 diabetes among Saudi adults on insulin pump therapy during the COVID-19 lockdown**. *Diabetes Metab Syndr* (2021.0) **15** 63-68. DOI: 10.1016/j.dsx.2020.12.013 47. Šagát P, Bartík P, González PP, Tohănean DI, Knjaz D. **Impact of COVID-19 quarantine on low back pain intensity, prevalence, and associated risk factors among adult citizens residing in Riyadh (Saudi Arabia): a cross-sectional study**. *Int J Environ Res Public Health* (2020.0) **17** 7302. DOI: 10.3390/ijerph17197302 48. Sultan I, Alobaidi RA, Sewaid KK, Bader MU, Almuwallad NT, Mohammed RA. **Assessment of the effect of the COVID-19 pandemic on the lifestyle of the population in saudi arabia: a cross-sectional online survey study**. *Cureus* (2021.0) **13** e19796. PMID: 34956785 49. Caputo EL, Reichert FF. **Studies of physical activity and covid-19 during the pandemic: a scoping review**. *J Phys Act Health* (2020.0) **17** 1275-1284. DOI: 10.1123/jpah.2020-0406 50. Cheikh Ismail L, Osaili TM, Mohamad MN, Al Marzouqi A, Jarrar AH, Zampelas A, Habib-Mourad C, Omar Abu Jamous D, Ali HI, Al Sabbah H. **Assessment of eating habits and lifestyle during the coronavirus 2019 pandemic in the Middle East and North Africa region: a cross-sectional study**. *Br J Nutr* (2021.0) **126** 757-766. DOI: 10.1017/S0007114520004547 51. Mumena W. **Impact of COVID-19 Curfew on Eating Habits, Eating Frequency, and Weight According to Food Security Status in Saudi Arabia: A Retrospective Study**. *Progress in Nutrition* (2021.0) **22** 1-9 52. Alshahrani SM, Alghannam AF, Taha N, Alqahtani SS, Al-Mutairi A, Al-Saud N, Alghnam S. **The impact of COVID-19 pandemic on weight and body mass index in Saudi Arabia: A longitudinal study**. *Front Public Health* (2021.0) **9** 775022. DOI: 10.3389/fpubh.2021.775022 53. Alsalman AA, Aldossari MR, Alomani ZD, Alkhunaizi SI, Aljardah ZA, Almousa FA, Kalalah ZA. **Impact of coronavirus disease lockdown on children with type 1 diabetes mellitus in Al-Khobar, Saudi Arabia**. *Cureus* (2022.0) **14** e21350. PMID: 35198267 54. Alzahrani F, Alshahrani NZ, Abu Sabah A, Zarbah A, Abu Sabah S, Mamun MA. **Prevalence and factors associated with mental health problems in Saudi general population during the coronavirus disease 2019 pandemic: A systematic review and meta-analysis**. *Psych J* (2022.0) **11** 18-29. DOI: 10.1002/pchj.516 55. Al Dhaheri AS, Bataineh MF, Mohamad MN, Ajab A, Al Marzouqi A, Jarrar AH, Habib-Mourad C, Abu Jamous DO, Ali HI, Al Sabbah H. **Impact of COVID-19 on mental health and quality of life: Is there any effect? A cross-sectional study of the MENA region**. *PLoS ONE* (2021.0) **16** e0249107. DOI: 10.1371/journal.pone.0249107 56. Almhizai RA, Almogren SH, Altwijery NA, Alanazi BA, Al Dera NM, Alzahrani SS, Alabdulkarim SM. **Impact of COVID-19 on children's and adolescent's mental health in Saudi Arabia**. *Cureus* (2021.0) **13** e19786. PMID: 34963826 57. Al-Quwaidhi AJ, Pearce MS, Critchley JA, Sobngwi E, O'Flaherty M. **Trends and future projections of the prevalence of adult obesity in Saudi Arabia, 1992–2022**. *East Mediterr Health J* (2014.0) **20** 589-595. DOI: 10.26719/2014.20.10.589 58. **Burden of obesity in the Eastern Mediterranean Region: findings from the Global Burden of Disease 2015 study**. *Int J Public Health* (2018.0) **63** 165-176. DOI: 10.1007/s00038-017-1002-5 59. Salem V, AlHusseini N, Razack H, Naoum A, Sims O, Alqahtani S. **Prevalence, risk factors, and interventions for obesity in Saudi Arabia: A systematic review**. *Obes Rev* (2022.0) **23** e13448. DOI: 10.1111/obr.13448 60. Sallis R, Young DR, Tartof SY, Sallis JF, Sall J, Li Q, Smith GN, Cohen DA. **Physical inactivity is associated with a higher risk for severe COVID-19 outcomes: a study in 48 440 adult patients**. *Br J Sports Med* (2021.0) **55** 1099-1105. DOI: 10.1136/bjsports-2021-104080 61. Cho DH, Lee SJ, Jae SY, Kim WJ, Ha SJ, Gwon JG, Choi J, Kim DW, Kim JY. **Physical activity and the risk of COVID-19 infection and mortality: a nationwide population-based case-control study**. *J Clin Med* (2021.0) **10** 1539. DOI: 10.3390/jcm10071539 62. Aytur S, Jones S, Stransky M, Evenson K. **Measuring physical activity in outdoor community recreational environments: Implications for research, policy, and practice**. *Curr Cardiovasc Risk Rep* (2015.0) **9** 1-13 63. Volenec ZM, Abraham JO, Becker AD, Dobson AP. **Public parks and the pandemic: How park usage has been affected by COVID-19 policies**. *PLoS ONE* (2021.0) **16** e0251799. DOI: 10.1371/journal.pone.0251799 64. Gu X, Keller J, Zhang T, Dempsey D, Roberts H, Jeans K, Stevens W, Borchard J, VanPelt J, Tulchin-Francis K. **Disparity in built environment and its impacts on youths' physical activity behaviors during COVID-19 pandemic restrictions**. *J Racial Ethn Disparities.* (2022.0) **14** 1-11. DOI: 10.1007/s40615-022-01341-3 65. Chinapaw MJ, Mokkink LB, van Poppel MN, van Mechelen W, Terwee CB. **Physical activity questionnaires for youth: a systematic review of measurement properties**. *Sports Med (Auckland, NZ* (2010.0) **40** 539-563. DOI: 10.2165/11530770-000000000-00000 66. Zhang F, Zhang Y, Yu Y, Lu W, Zhang H. **Bibliometric analysis of the top-50 cited articles on COVID-19 and physical activity**. *Front Public Health* (2022.0) **10** 926244. DOI: 10.3389/fpubh.2022.926244 67. Ramirez Varela A, Hallal PC, Mejia Grueso J, Pedisic Z, Salvo D, Nguyen A, Klepac B, Bauman A, Siefken K, Hinckson E. **Status and trends of physical activity surveillance, policy, and research in 164 countries: findings from the global observatory for physical activity-GoPA! 2015 and 2020 surveys**. *J Phys Act Health* (2023.0) **20** 112-128. DOI: 10.1123/jpah.2022-0464 68. Reilly JJ, Barnes J, Gonzalez S, Huang WY, Manyanga T, Tanaka C, Tremblay MS. **Recent secular trends in child and adolescent physical activity and sedentary behavior internationally: analyses of active healthy kids global alliance global matrices 1.0 to 4.0**. *J Phys Activity Health* (2022.0) **19** 729-736. DOI: 10.1123/jpah.2022-0312 69. Aubert S, Barnes JD, Demchenko I, Hawthorne M, Abdeta C, Abi Nader P, Adsuar Sala JC, Aguilar-Farias N, Aznar S, Bakalar P. **Global matrix 4.0 physical activity report card grades for children and adolescents: results and analyses from 57 countries**. *J Phys Activity Health* (2022.0) **19** 700-728. DOI: 10.1123/jpah.2022-0456 70. Arena R, Pronk NP, Laddu D, Whitsel LP, Sallis JF, Lavie CJ, Network H-P. **Mapping one million COVID-19 deaths and unhealthy lifestyle behaviors in the United States: recognizing the syndemic pattern and taking action**. *Am J Med* (2022.0) **135** 1288-1295. DOI: 10.1016/j.amjmed.2022.06.006 71. **Edited by the world health organization**. *Geneva, Switzerland* (2022.0) **2022** 1-112 72. Milton K, Cavill N, Chalkley A, Foster C, Gomersall S, Hagstromer M, Kelly P, Kolbe-Alexander T, Mair J, McLaughlin M. **Eight investments that work for physical activity**. *J Phys Act Health* (2021.0) **18** 625-630. DOI: 10.1123/jpah.2021-0112 73. Al-Hazzaa HM. **Physical activity research in the gulf cooperation council countries: progress made but work still to do**. *J Phys Act Health* (2022.0) **19** 769-770. DOI: 10.1123/jpah.2022-0484 74. 74.Alqahtani AS, Baattaiah BA, Alharbi MD, Khan F, Aldhahi MI. Barriers and facilitators affecting physical activity among adults in Saudi Arabia during COVID-19 quarantine. Health Promot Int 2021;daab191. 10.1093/heapro/daab191. 75. Chaabane S, Chaabna K, Doraiswamy S, Mamtani R, Cheema S. **Barriers and facilitators associated with physical activity in the middle east and north africa region: a systematic overview**. *Int J Environ Res Public Health* (2021.0) **18** 1647. DOI: 10.3390/ijerph18041647 76. Knight RL, McNarry MA, Sheeran L, Runacres AW, Thatcher R, Shelley J, Mackintosh KA. **Moving forward: understanding correlates of physical activity and sedentary behaviour during Covid-19-an integrative review and socioecological approach**. *Int J Environ Res Public Health* (2021.0) **18** 10910. DOI: 10.3390/ijerph182010910
--- title: Kinetics of CD169, HLA-DR, and CD64 expression as predictive biomarkers of SARS-CoV2 outcome authors: - Arianna Gatti - Paola Fassini - Antonino Mazzone - Stefano Rusconi - Bruno Brando - Giovanni Mistraletti journal: Journal of Anesthesia, Analgesia and Critical Care year: 2023 pmcid: PMC10041484 doi: 10.1186/s44158-023-00090-x license: CC BY 4.0 --- # Kinetics of CD169, HLA-DR, and CD64 expression as predictive biomarkers of SARS-CoV2 outcome ## Abstract ### Introduction Discriminating between virus-induced fever from superimposed bacterial infections is a common challenge in intensive care units. Superimposed bacterial infections can be detected in severe SARS-CoV2-infected patients, suggesting the important role of the bacteria in COVID-19 evolution. However, indicators of patients’ immune status may be of help in the management of critically ill subjects. Monocyte CD169 is a type I interferon-inducible receptor that is up-regulated during viral infections, including COVID-19. Monocyte HLA-DR expression is an immunologic status marker, that decreases during immune exhaustion. This condition is an unfavorable prognostic biomarker in septic patients. Neutrophil CD64 upregulation is an established indicator of sepsis. ### Methods In this study, we evaluated by flow cytometry the expression of cellular markers monocyte CD169, neutrophil CD64, and monocyte HLA-DR in 36 hospitalized patients with severe COVID-19, as possible indicators of ongoing progression of disease and of patients’ immune status. Blood testings started at ICU admission and were carried on throughout the ICU stay and extended in case of transfer to other units, when applicable. The marker expression in mean fluorescence intensity (MFI) and their kinetics with time were correlated to the clinical outcome. ### Results Patients with short hospital stay (≤15 days) and good outcome showed higher values of monocyte HLA-DR (median 17,478 MFI) than long hospital stay patients (>15 days, median 9590 MFI, $$p \leq 0.04$$) and than patients who died (median 5437 MFI, $$p \leq 0.05$$). In most cases, the recovery of the SARS-CoV2 infection-related signs was associated with the downregulation of monocyte CD169 within 17 days from disease onset. However in three surviving long hospital stay patients, a persistent upregulation of monocyte CD169 was observed. An increased neutrophil CD64 expression was found in two cases with a superimposed bacterial sepsis. ### Conclusion Monocyte CD169, neutrophil CD64, and monocyte HLA-DR expression can be used as predictive biomarkers of SARS-CoV2 outcome in acutely infected patients. The combined analysis of these indicators can offer a real-time evaluation of patients’ immune status and of viral disease progression versus superimposed bacterial infections. This approach allows to better define the patients’ clinical status and outcome and may be useful to guide clinicians’ decisions. Our study focused on the discrimination between the activity of viral and bacterial infections and on the detection of the development of anergic states that may correlate with an unfavorable prognosis. ## Introduction CD169 (Sialoadhesin or Siglec-1) is a type I interferon-inducible receptor, and as reported by several studies, its expression is upregulated on the surface of monocytes and dendritic cells during viral infections [1–3]. CD169 is normally expressed on resident macrophages, whereas it is virtually undetectable on quiescent monocytes in the peripheral blood of healthy subjects [4]. CD169 monocyte expression (moCD169) is increased upon the release of antiviral molecules, such as Interferon Type I [4]. It is also reported that moCD169 expression is increased in early SARS-CoV2 infection [5]. Among patients with mild COVID-19, a time-dependent expression of moCD169 with the highest values within the first 3 days after the onset of symptoms has been described, with expression levels returning to the normal range within the subsequent 3–4 weeks [6]. On the other hand, some studies showed a strong inflammatory response to SARS-CoV2 infection in severe cases, with also quantitative alterations of the monocyte and macrophage compartments [7, 8]. Monocytes/macrophages play a key role in the immune control of infections. Namely, monocyte HLA-DR (moHLA-DR) expression is an immunologic status marker, correlating with an efficient antigen-presenting function [9, 10]. Persistent over-stimulation or immune exhaustion induces the decrease of HLA-DR expression on monocytes, determining the drift to an anergic state. In several studies, the reduced moHLA-DR expression has been identified as an unfavorable prognostic biomarker during severe infections and sepsis [9, 11, 12]. Downregulation of moHLA-DR was also reported in critical COVID-19 patients, admitted to intensive care unit (ICU) [13, 14]. During prolonged stays in ICU for COVID-19, a common clinical problem is the discrimination between virus-induced findings and superimposed bacterial infections, which tend to occur in a later phase. The ordinary clinical parameters, i.e., fever, acute-phase reactant levels, lactate, blood, and respiratory tract cultures, do not always provide clues of the necessary sensitivity and specificity [15, 16]. It is well known that the high-affinity immunoglobulin Fc-gamma receptor type I CD64 is constitutively expressed at very low density on the surface of blood neutrophils in healthy subjects. The neutrophil CD64 density (neCD64) is promptly upregulated upon stimulation by bacteria and their soluble products, thus being considered a sensitive and specific marker of severe bacterial infections and sepsis [12, 17]. Sepsis is the most common cause of death among hospitalized patients in ICU [18]. The initial phase of sepsis is characterized by a hyperinflammatory status with an increase of pro-inflammatory markers (C-reactive protein, procalcitonin, IL- 6) and is followed by a immunosuppressive phase [19]. These markers are routinely used in the diagnosis and management of sepsis. As reported by several studies, these markers are also increased in patients with severe COVID-19, defining a condition known as “COVID-19 viral sepsis” [20, 21]. In addition, the co-occurrence of bacterial superinfections can be detected in severely SARS-CoV2 infected patients, suggesting the important role of bacteria in COVID-19 evolution [22, 23]. The aim of our study was therefore to evaluate by a quick, real-time flow cytometric assay if the combined expression of moCD169, moHLA-DR, and neCD64 in ICU patients with severe COVID-19 could allow a better definition of patients’ clinical status and outcome, to guide clinicians’ decisions. Our study focused on the discrimination between the activity of viral and bacterial infections and in the detection of developing anergic states, which may correlate with an unfavorable prognosis. The kinetics of moCD169, moHLA-DR, and neCD64 during the clinical course was also prospectively evaluated in some representative patients. ## Patients and control group Thirty-six hospitalized patients from December 2020 to April 2021 with severe COVID-19 were included in the study. Twenty-two patients ($61\%$) were admitted to ICU and overall $\frac{6}{36}$ ($16.6\%$) died within a median of 44 days from admission. The characteristics of patients are reported in Table 1. Ten patients did not show any unfavorable risk factor at admission (i.e., diabetes, hypertension, obesity) nor major comorbidities (i.e., cardiovascular diseases, chronic obstructive pulmonary disease, or neoplasia) and one of them died. In five cases, only one risk factor was present. In other four patients, two unfavorable risk factors were present and two of them died. The remaining seventeen individuals showed at least three unfavorable comorbidities and three patients did not survive. Table 1 also summarizes the clinical severity scores (SOFa, SAPSII, P/F grade, P/F, and CCI) of all the patients enrolled in this study. Table 1Characteristics of 36 patients included in the studyCOPD chronic obstructive pulmonary disease Ground-glass opacities at chest x-ray were present at admission in $64\%$ of cases and in 8 patients pulmonary embolism was subsequently evidenced. In $\frac{33}{36}$ cases, steroid treatment was undertaken in the early phase, along with support therapy. Blood testings started at hospital admission and were carried on throughout the ICU unit transfer and stay. In 15 subjects ($42\%$), the prospective evaluation of moCD169, neCD64, and moHLA-DR during the clinical course was carried out, for a median of 12 days after the first analysis. In addition, 10 age-matched healthy donors were included as the normal control group (NC). The approval of this observational study was obtained from the local scientific committee, within the extended informed consent procedures activated during the early COVID-19 epidemic. The immunophenotypic testings were performed using the leftovers of routine full blood counts, and the study results were not used to undertake any clinical decision. In our institution, healthy blood donors sign informed consent for research studies on their routine samples, which are used to represent normal reference values in a variety of laboratory testings, including cell phenotyping. ## Flow cytometric analysis Fifty microliters of EDTA-whole blood samples were taken from the fresh routine tubes ordered for the full blood count. Aliquots were mixed with 10 μl IOTest Myeloid Activation antibody cocktail (Beckman Coulter, Milano, Italy), containing anti-CD169-PE (R-Phycoerythrin, clone 7-239), anti-CD64-PB (Pacific Blue, clone 22), and anti-HLA-DR-APC (Allophycocyanin, clone Immu357) and incubated for 20 min at room temperature in the dark. The samples were processed with a lyse-no-wash technique using ammonium chloride and no fixation. Flow cytometric analysis was immediately performed on a FACS Lyric (Becton Dickinson-BD, Milano, Italy), daily calibrated using BD CS&T beads according to the recommended Euroflow procedures [24], with a specifically designed application setting and defined target values, following the manufacturer’s instructions. The samples were acquired immediately after the lysis at a medium rate. Data from 100,000 total white cell events were acquired. Monocytes were identified with a Side Scatter (SSC)/CD64 dot plot. Granulocytes and lymphocytes were gated based on the respective SSC, CD64, and forward scatter (FSC) features (Fig. 4). The degree of surface marker expression was quantitated by the mean fluorescence intensity (MFI) ratio. The MFI ratio is calculated by dividing the geometric mean of the positive fluorescence peak of the relevant population (i.e., moCD169) by the geometric mean of the negative control or reference population (i.e., lymphocytes) that is not expressing that marker. Monocyte CD169 expression levels were calculated as the ratio between CD169 geometric mean fluorescence intensity (MFI) units on monocytes and MFI units on lymphocytes, which acted as the internal negative reference population. CD64 expression level on granulocytes was similarly calculated as the ratio between CD64 MFI geometric mean units on granulocytes and CD64 MFI units on lymphocytes, which acted as the internal negative reference population. MoHLA-DR expression level was calculated as plain geometric mean MFI, since no internal purely HLA-DR-negative reference cell populations can be found. Patient moHLA-DR is displayed in comparison to normal control subjects, as depicted in the examples in Fig. 5. The robustness and stability of MFI measurements over time were ensured by the daily calibration with CS&T beads. ## Statistical analysis The distributions of surface antigen expression levels and ratios were checked for normality using the Kolmogorov-Smirnov test. In the case of non-normally distributed variates, statistical comparisons were performed using the non-parametric Mann-Whitney test. Comparisons between groups with normally distributed variates were performed using the Student t-test. Statistically significant differences were established by a p value < 0.05 in both cases. ## Results The moCD169 MFI ratios and neCD64 MFI ratios were not normally distributed, as assessed with the Kolmogorov-Smirnov test. The Mann-Whitney test was therefore used for comparisons. MoHLA-DR expression levels showed a normal distribution both in patients and normal controls (NC), and Student’s t-test was used for comparisons. SARS-CoV2-infected patients invariably expressed a strongly upregulated moCD169 during the early disease phase (Figs. 1 and 4) (mean 10 days from the beginning of symptoms, as reported at hospital admission by patients, where applicable, or by their relatives) as compared to quiescent monocytes in the NC group (mean moCD169 MFI ratio: 125 vs 5; $$p \leq 0.00084$$).Fig. 1Expression of monocyte CD169 in acutely SARS-CoV2-infected patients in the early phase of the disease and during the follow-up, according to the length of hospital stay. The overall trend to moCD169 downregulation in the long-term did not reach statistical significance, due to some patients with persistent positive swabs for SARS-CoV2 During the follow-up, moCD169 expression displayed a downward but individually variable trend, associated to viral clearance. We did not find any significant difference of expression of moCD169 between SARS-CoV2-infected patients admitted to ICU and those admitted to other clinical units. In addition, during the early disease phase, SARS-CoV2-infected patients showed a slightly increased expression of neCD64, as compared to quiescent granulocytes in the healthy NC group (mean MFI ratio in SARS-CoV2 patients: 10.9 vs 2.7 in NC; $$p \leq 0.003$$) (Fig. 2). Nevertheless in most cases, no significant microbiological agents were detected in the culture samples. Fig. 2Expression of neutrophil CD64 at disease onset and during the follow-up. A mild upregulation of neCD64 is observed in the majority of patients, while very high values were detectable only in patients with overt bacteremia and sepsis Patients admitted to ICU showed significantly lower values of moHLA-DR than those admitted to other clinical units (mean 7,841 MFI vs 11,754; $$p \leq 0.017$$). During the early SARS-CoV2 infection, no significant differences of moHLA-DR expression were found between the six patients who died and those surviving in ICU (mean moHLA-DR expression in ICU patients who died: 6692 MFI vs surviving ICU patients: 9623, p= n.s) (Fig. 3).Fig. 3Expression of monocyte HLA-DR as an indicator of immune competence in SARS-CoV2 infected patients. The lowest level of expression was observed in patients who died for multiple complications, whereas patients with prolonged hospital stay showed a lower HLA-DR expression level, as compared with patients with shorter and more favorable clinical course and normal control (NC) subjects Patients with short hospital stay (≤15 days, median 9 days) and good outcome showed higher values of moHLA-DR (17,478 MFI) than long hospital stay patients (>15 days, median 47 days, with mean moHLA-DR MFI: 9590; $$p \leq 0.04$$) and than patients who died within 44 days (5437 MFI; $$p \leq 0.05$$). No significant differences in moCD169 MFI ratios were found between long and short hospital stay patients. In most cases, the clearance of the SARS-CoV2 infection-related signs was associated with the normalization of moCD169 within 17 days from disease onset (mean moCD169 MFI ratio returning close to 5). However, in three surviving long hospital stay patients (median 56 days), a persistent upregulation of moCD169 was observed (Fig. 4), with a mean moCD169 MFI ratio =26. In these three cases, the persistent molecular positivity for SARS-CoV2 in upper and lower airway swabs was also detected. Fig. 4Peripheral blood white cell gating strategy (left panel). Monocytes, neutrophils, and lymphocytes are identified according to their CD64 expression and Side Scatter (SSC-A) features. The diagrams to the right show an example of very high monocyte CD169 upregulation in a patient with active COVID-19, mild neutrophil activation (CD64), and preserved monocyte HLA-DR expression A total of twenty-two patients were admitted to ICU. Nine surviving patients had short hospital stay (<15 days), and 21 surviving patients had long hospital stay (mean 47 days). Six subjects died within 44 days from ICU admission. During the follow-up, a general reduction of neCD64 expression was found as compared to disease onset (mean MFI ratio from 10.9 to 5; $$p \leq 0.01$$). In two cases, a superimposed bacterial sepsis was diagnosed during the hospital follow-up. In one case with neCD64 MFI ratio = 15.7 an overt sepsis was defined with positive blood cultures for Serratia marcescens. In the other case with neCD64 MFI ratio = 11.5 and a severely downregulated moHLA-DR MFI, averaging 327, an overt sepsis supported by positive blood cultures for *Staphylococcus capitis* was diagnosed, shortly before the patient’s death (Fig. 5).Fig. 5Examples of some of the many possible findings in the monitoring of SARS-CoV2-infected patients undergoing sepsis during their clinical follow-up. Upper row (a): Moderate neCD64 upregulation, normal moHLA-DR expression, and cleared SARS-CoV2, with normal moCD169 expression. This patient showed normal immune competence and recovered from sepsis. Lower row (b): Moderate neCD64 upregulation, severe moHLA-DR downregulation, and just slightly increased moCD169. This patient died for infectious complications, despite having effectively cleared the SARS-CoV2 In two patients, the progressive and significantly decreased expression of moHLA-DR was associated to fatal outcome in the long term (5834 MFI and 12,320 MFI at onset, respectively, and 2991 MFI and 2865 MFI, respectively, after 10 days from ICU admission). On the other hand, in another ICU patient, the progressive recovery of moHLA-DR was associated to an improving clinical condition that allowed his discharge after 63 days. ## Discussion The aim of our study was to evaluate retrospectively the changes in the expression of moCD169, moHLA-DR, and neCD64 by flow cytometry in 36 hospitalized patients with severe COVID-19. The marker expression findings with time were correlated to the clinical outcome and the length of hospital stay. In 15 cases, the prospective analysis of these markers with time during the clinical course was also evaluated. Several studies reported that the presence of risk factors such as obesity, hypertension, cardiovascular, or pulmonary disease might lead to a more severe outcome of SARS-CoV2 infection [25, 26]. In our limited study, we were unable to demonstrate such correlations. Seventeen individuals showed at least three unfavorable comorbidities, but only three of them did not survive. Regarding the other three patients who died, one patient did not show any comorbidity and the remaining two patients showed only two unfavorable comorbidities. On the other hand and in accordance with other recent studies [27, 28], we found a strong correlation between three flow cytometric indicators with different clinical meaning: moCD169, moHLA-DR, and neCD64 expression and the degree of COVID-19 critical illness, the clinical outcome, and the length of hospital stay. Patients admitted to ICU showed significantly lower values of moHLA-DR than subjects admitted to other clinical units, indicating that a dysfunctional immune response may contribute to the development of acute respiratory failure [13]. *In* general, COVID-19 patients in ICU display a more critical and dynamic clinical picture that often gets complicated with time. For this reason, reliable biological indicators that may enable the quick interpretation of the different ongoing signs and symptoms are needed. The flow cytometric monitoring of the combined expression of moCD169, moHLA-DR, and neCD64 seems fitting the purpose. In a report on a group of 110 patients admitted to the surgical ICU [29], those with a reduced moHLA-DR had a longer median ICU stay (6 days vs. 3 days), independent of the cause of critical illness. We found that patients with short hospital stay (≤15 days, median 9 days) and good outcome showed higher values of moHLA-DR than long hospital stay patients (>15 days, median 47 days), and than patients who died within 44 days. To our knowledge, these observations were not reported before in patients with SARS-CoV2 infection. The definition of “long hospital stay” was set after 15 days taking into account the changes in viral load. As reported by Doehn et al. [ 6] between 13 and 15 days since the onset of symptoms, both in mild and severe SARS-CoV2 infection a significant decrease of viral load and moCD169 expression can be detected, along with an increase of serum anti-SARS-CoV2 IgG concentration. In $\frac{33}{36}$ cases, steroid treatment was undertaken along with support therapy in both short hospital stay patients and in ICU long stay patients. The blood testings were performed during the steroid treatment. The steroid administration did not seem however to influence the moHLA-DR expression. As reported by Boomer et al., no significant difference was seen in cytokine production in septic patients treated or not treated with corticosteroids [30]. In the study by Spinetti et al. [ 13], low moHLA-DR expression remained unchanged within the first 5 days after the ICU admission in critical COVID-19 patients. Conversely, in our 15 cases with a prospective flow cytometric monitoring, the changes of moHLA-DR occurred over a more extended period of time (about 12 days after the first testing) and correlated to the clinical outcome. It was also reported that moCD169 expression is increased in early SARS-CoV2 infection [5] and returns to normal range within the subsequent 3–4 weeks. CD169 is a type I interferon-inducible receptor; however in the present study, we were unable to evaluate the levels of circulating IFN-gamma. As reported by other authors we found that in most cases, the clearance of the SARS-CoV2 infection was associated with the downregulation of moCD169 within 17 days from disease onset [6]. However in three surviving long hospital stay patients (mean 56 days), a persistent upregulation of moCD169 was observed, with sustained positivity for molecular SARS-CoV2 swabs. It is reported that severe COVID-19 disease is associated to a high SARS-CoV2 viral load and a longer virus-shedding time [31]. Therefore, a persistent upregulation of moCD169 can be considered as an unfavorable prognostic factor in SARS-CoV2 infection, correlating with a lack of an effective viral clearance. Therefore, the kinetics of moCD169, moHLA-DR, and neCD64 expression may be useful to discriminate febrile events and to stratify patients according to their risk of complications and length of hospital stay. The quick and real-time characteristic of this flow cytometric assay makes it suitable for on-demand orders with a very short turn-around time (less than 60 min). Superimposed bacterial infections are the commonest complications in ICU patients. An upregulated neCD64 expression has long been demonstrated as a reliable indicator of severe bacterial infections and sepsis [12]. In our patient cohort, increased values of neCD64 were demonstrated at disease onset, however in most cases without significant positive cultures for bacteria. A significantly low rate of bacteremia was found among COVID-19 patients in the early disease phase [32]. However in another study, the cumulative risk of developing an episode of ICU-acquired bloodstream infection in critical SARS-CoV2 patients significantly increased after 15 days from ICU admission [33]. During our patients’ follow-up, a general reduction of neCD64 expression was observed, although in two cases the neCD64 MFI ratio reached high values in association with positive blood cultures. Our study has several limitations: the small patient cohort, the initial flow cytometric monitoring not strictly performed at regular intervals during the early phase of the study, and the lost to follow-up of some patients due to their transfer to external ICUs or medical wards during the early SARS-CoV2 epidemic. ## Conclusion The monocyte CD169 and HLA-DR expression can be used as predictive biomarkers of SARS-CoV2 outcome in acutely infected patients, since they can reliably indicate the status of viral persistence or clearance and the occurrence of superimposed biological phenomena that negatively affect the immune response. The simultaneous assessment of neCD64 expression is of help in discriminating virus-induced fever from severe bacterial infections and sepsis. This real-time flow cytometric assay can be performed on demand, with a short turnaround time (1 h). We believe that the prospective combined monitoring of moCD169, moHLA-DR, and neCD64 can be useful for the management of COVID-19 patients and possibly also for other critical infectious diseases and sepsis. ## Authors’ information All authors are members of the medical or biologist staff at the Western Milan Area Hospital Consortium, Legnano General Hospital, 20025 Legnano (Milano), Italy, in their respective wards and units. ## References 1. Kim WK, McGary CM, Holder GE. **Increased expression of CD169 on blood monocytes and its regulation by virus and CD8 T cells in macaque models of HIV infection and AIDS**. *AIDS Res Hum Retroviruses* (2015.0) **31** 696-706. DOI: 10.1089/aid.2015.0003 2. Van der Kuyl AC, Van den Burg R. **Sialoadhesin (CD169) expression in CD14+ cells is upregulated early after HIV-1 infection and increases during disease progression**. *PloS One* (2007.0) **2** e257. DOI: 10.1371/journal.pone.0000257 3. Pino M, Erkizia I, Benet S. **HIV-1 immune activation induces Siglec-1 expression and enhances viral**. *Retrovirology* (2015.0) **12** 37. DOI: 10.1186/s12977-015-0160-x 4. Bourgoin P, Biéchelé G, Ait Belkacem I, Morange PE, Malergue F. **Role of the interferons in CD64 and CD169 expressions in whole blood: Relevance in the balance between viral- or bacterial-oriented immune responses**. *Immun Inflamm Dis* (2020.0) **8** 106-123. DOI: 10.1002/iid3.289 5. Bedin AS, Makinson A, Picot MC. **Monocyte CD169 expression as a biomarker in the early diagnosis of coronavirus disease 2019**. *J Infect Dis* (2021.0) **223** 562-567. DOI: 10.1093/infdis/jiaa724 6. Doehn JM, Tabeling C, Biesen R. **CD169/SIGLEC1 is expressed on circulating monocytes in COVID-19**. *Infection* (2021.0) **49** 757-762. DOI: 10.1007/s15010-021-01606-9 7. Chevrier S, Zurbuchen Y, Cervia C. **A distinct innate immune signature marks progression from mild to severe COVID-19**. *Cell Rep Med* (2020.0) **2** 100166. DOI: 10.1016/j.xcrm.2020.100166 8. Gatti A, Radrizzani D, Viganò P. **Decrease of non-classical and intermediate monocyte subsets in severe acute SARS-CoV-2 infection**. *Cytometry A* (2020.0) **97A** 887-890. DOI: 10.1002/cyto.a.24188 9. Zhuang Y, Peng H, Chen Y. **Dynamic monitoring of monocyte HLA-DR expression for the diagnosis, prognosis, and prediction of sepsis**. *Front Biosci (Landmark Ed)* (2017.0) **22** 1344-1354. DOI: 10.2741/4547 10. Monneret G, Venet F. **Sepsis-induced immune alterations monitoring by flow cytometry as a promising tool for individualized therapy**. *Cytometry B Clin Cytom* (2016.0) **90B** 376-386. DOI: 10.1002/cyto.b.21270 11. Meisel C, Schefold JC, Pschowski R. **Granulocyte-macrophage colony-stimulating factor to reverse sepsis-associated immunosuppression: a double-blind, randomized, placebo-controlled multicenter trial**. *Am J Respir Crit Care Med* (2009.0) **180** 640-648. DOI: 10.1164/rccm.200903-0363OC 12. Monneret G, Gossez M, Aghaeepour N. **How clinical flow cytometry rebooted sepsis immunology**. *Cytometry A* (2019.0) **95A** 431-441. DOI: 10.1002/cyto.a.23749 13. Spinetti T, Hirzel C, Fux M. **Reduced monocytic human leukocyte antigen-DR expression indicates immunosuppression in critically ill COVID-19 patients**. *Anesth Analg* (2020.0) **131** 993-999. DOI: 10.1213/ANE.0000000000005044 14. Benlyamani I, Venet F, Coudereau R. **Monocyte HLA-DR measurement by flow cytometry in COVID-19 patients: an interim review**. *Cytometry A* (2020.0) **97A** 1217-1221. DOI: 10.1002/cyto.a.24249 15. Cavaillon JM, Singer M, Skirecki T. **Sepsis therapies: learning from 30 years of failure of translational research to propose new leads**. *EMBO Mol Med* (2020.0) **12** e10128. DOI: 10.15252/emmm.201810128 16. Yeh CF, Wu CC, Liu SH. **Comparison of the accuracy of neutrophil CD64, procalcitonin, and C-reactive protein for sepsis identification: a systematic review and meta-analysis**. *Ann Intensive Care* (2019.0) **9** 5. DOI: 10.1186/s13613-018-0479-2 17. Trzeciak A, Pietropaoli AP, Kim M. **Biomarkers and associated immune mechanisms for early detection and therapeutic management of sepsis**. *Immune Netw* (2020.0) **20** e23. DOI: 10.4110/in.2020.20.e23 18. Genga KR, Russell JA. **Update of sepsis in the intensive care unit**. *J Innate Immun* (2017.0) **9** 441-455. DOI: 10.1159/000477419 19. Rhodes A, Evans LE, Alhazzani W, Levy MM, Antonelli M. **Surviving sepsis campaign: international guidelines for management of sepsis and septic shock: 2016**. *Intensive Care Med* (2017.0) **43** 304-377. DOI: 10.1007/s00134-017-4683-6 20. Li H, Liu L, Zhang D, Xu J, Dai H. **SARS-CoV-2 and viral sepsis: observations and hypotheses**. *Lancet* (2020.0) **395** 1517-1520. DOI: 10.1016/S0140-6736(20)30920-X 21. Prescott HC, Girard TD. **Recovery from severe COVID-19: leveraging the lessons of survival from sepsis**. *JAMA* (2020.0) **324** 739-740. DOI: 10.1001/jama.2020.14103 22. 22.Mazzariol A, Benini A, Unali I, Nocini R, Smania M, Bertoncelli A, De Sanctis F, Ugel S, Donadello K, Polati E, Gibellini D (2021) Dynamics of SARS-CoV2 infection and multi-drug resistant bacteria superinfection in patients with assisted mechanical ventilation. Front Cell Infect Microbiol. 10.3389/fcimb.2021.683409 23. Tetaj N, Capone A, Stazi GV, Marini MC, Garotto BD, Scarci S, Ilaria Caravella I, Macchione M, De Angelis MG, Di Lorenzo R, Carucci A, Antonica MV, Gaviano I, Inversi C, Agostini E, Canichella F, Taloni G, Evangelista F, Onnis I, Mogavero G, Lamanna ME, Rubino D, Di Frischia M, Porcelli C, Cesi E, Antinori A, Palmieri F, D’Offizi G, Taglietti F, Nisii C, Cataldo MA, Ianniello S, Campioni P, Vaia F, Nicastri E, Girardi E, Marchioni L. **Epidemiology of ventilator-associated pneumonia in ICU COVID-19 patients: an alarming high rate of multidrug-resistant bacteria**. *J Anesth Analg Crit Care* (2022.0) **2** 36. DOI: 10.1186/s44158-022-00065-4 24. Glier H, Novakova M, Te Marvelde J. **Comments on EuroFlow standard operating procedures for instrument setup and compensation for BD FACS canto II, Navios and BD FACS Lyric instruments**. *J Immunol Methods* (2019.0) **475** 112680. DOI: 10.1016/j.jim.2019.112680 25. Al Heialy S, Hachim MY, Hachim IY. **Combination of obesity and co-morbidities leads to unfavorable outcomes in COVID-19 patients**. *Saudi J Biol Sci* (2021.0) **28** 1445-1450. DOI: 10.1016/j.sjbs.2020.11.081 26. Hernández-Garduño E. **Obesity is the comorbidity more strongly associated for Covid-19 in Mexico. A case-control study**. *Obes Res Clin Pract* (2020.0) **14** 375-379. DOI: 10.1016/j.orcp.2020.06.001 27. Ortillon M, Coudereau R, Cour M. **Monocyte CD169 expression in COVID-19 patients upon intensive care unit admission**. *Cytometry A* (2021.0) **99** 466-471. DOI: 10.1002/cyto.a.24315 28. Comins-Boo A, Larrañaga MG, Roa-Bautista A. **Validation of a quick flow cytometry-based assay for acute infection based on CD64 and CD169 expression. New tools for early diagnosis in COVID-19 pandemic**. *Front Med* (2021.0) **8** 655785. DOI: 10.3389/fmed.2021.655785 29. Tamulyte S, Kopplin J, Brenner T. **Monocyte HLA-DR assessment by a novel Point-of-Care Device is feasible for early identification of ICU patients with complicated courses - a proof-of-principle study**. *Front Immunol* (2019.0) **10** 432. DOI: 10.3389/fimmu.2019.00432 30. Boomer JS, To K, Chang KC, Takasu O, Osborne MDF, Messr BS, Walton AH, Bricker TL, Jarman SD, Kreisel D, Krupnick AS, Srivastava A, Swanson PE, Green JM, Hotchkiss SR. **Immunosuppression in patients who die of sepsis and multiple organ failure**. *JAMA.* (2011.0) **306** 2594-2605. DOI: 10.1001/jama.2011.1829 31. Liu Y, Yan LM, Wan L. **Viral dynamics in mild and severe cases of COVID-19**. *Lancet Infect Dis* (2020.0) **20** 656-657. DOI: 10.1016/S1473-3099(20)30232-2 32. Sepulveda J, Westblade LF, Whittier S. **Bacteremia and blood culture utilization during COVID-19 surge in New York City**. *J Clin Microbiol* (2020.0) **58** e00875-e00820. DOI: 10.1128/JCM.00875-20 33. Giacobbe DR, Battaglini D, Ball L. **Bloodstream infections in critically ill patients with COVID-19**. *Eur J Clin Invest* (2020.0) **50** e13319. DOI: 10.1111/eci.13319
--- title: Exploring biopsychosocial correlates of pregnancy risk and pregnancy intention in women with chronic kidney disease authors: - Elizabeth R. Ralston - Priscilla Smith - Katherine Clark - Kate Wiles - Joseph Chilcot - Kate Bramham journal: Journal of Nephrology year: 2023 pmcid: PMC10041500 doi: 10.1007/s40620-023-01610-2 license: CC BY 4.0 --- # Exploring biopsychosocial correlates of pregnancy risk and pregnancy intention in women with chronic kidney disease ## Abstract ### Introduction Women with Chronic Kidney Disease (CKD) are at increased risk of adverse pregnancy and renal outcomes. It is unknown how women with CKD understand their pregnancy risk. This nine-centre, cross-sectional study aimed to explore how women with CKD perceive their pregnancy risk and its impact on pregnancy intention, and identify associations between biopsychosocial factors and perception of pregnancy risk and intention. ### Methods Women with CKD in the UK completed an online survey measuring their pregnancy preferences; perceived CKD severity; perception of pregnancy risk; pregnancy intention; distress; social support; illness perceptions and quality of life. Clinical data were extracted from local databases. Multivariable regression analyses were performed. Trial registration: NCT04370769. ### Results Three hundred fifteen women participated, with a median estimated glomerular filtration rate (eGFR) of 64 ml/min/1.73m2 (IQR 56). Pregnancy was important or very important in 234 ($74\%$) women. Only 108 ($34\%$) had attended pre-pregnancy counselling. After adjustment, there was no association between clinical characteristics and women’s perceived pregnancy risk nor pregnancy intention. Women’s perceived severity of their CKD and attending pre-pregnancy counselling were independent predictors of perceived pregnancy risk. Importance of pregnancy was an independent predictor of pregnancy intention but there was no correlation between perceived pregnancy risk and pregnancy intention (r = − 0.002, $95\%$ CI − 0.12 to 0.11). ### Discussion Known clinical predictors of pregnancy risk for women with CKD were not associated with women’s perceived pregnancy risk nor pregnancy intention. Importance of pregnancy in women with CKD is high, and influences pregnancy intention, whereas perception of pregnancy risk does not. ### Supplementary Information The online version contains supplementary material available at 10.1007/s40620-023-01610-2. ## Introduction Three percent of women of reproductive age are estimated to be affected by Chronic Kidney Disease (CKD) [1], with prevalence anticipated to rise with increasing rates of obesity [2], diabetes [3] and advancing maternal age [4]. Pregnancies with CKD are complicated by increased risk to both mother and baby including superimposed pre-eclampsia [5, 6], preterm birth [6, 7], small for gestational age infants [6, 8], admission to neonatal care [6], and acceleration of CKD progression [5, 6]. Pre-pregnancy counselling clinics are an opportunity to discuss potential pregnancy risk with provision of individualised care and psychological preparation for a complex pregnancy [9, 10]. Pre-pregnancy counselling aims to help set realistic expectations, discuss potential maternal and fetal risk, adjust medication, and optimise timings [9]. It is recommended by the UK Kidney Association and within the Confidential Enquiries into Maternal Deaths in the United Kingdom that women with pre-existing medical conditions are referred for pre-pregnancy counselling by a multidisciplinary team [4, 11]. However, provision of information related to objective risk should not assume that women will perceive the same risk [12–14]. An individual’s perceived risk and perceptions pertaining to their health condition can be used to understand and predict health-related behaviour and decisions [15, 16]. In women without CKD, perceived pregnancy risk has been reported to influence women’s pregnancy-related behaviour and decision making [17, 18] but knowledge of risk perception of pregnancy and how it may influence pregnancy intentions in women with CKD is limited. Improved understanding of how women with CKD perceive pregnancy risk and factors influencing pregnancy intention is likely to enhance risk communication and provision of individualised care. The aims of this study were to understand how women with CKD perceive pregnancy risk, to examine the relationship between biopsychosocial factors and perception of pregnancy risk and pregnancy intentions, and to understand the relationship between perceived pregnancy risk and pregnancy intention. ## Design This was an online cross-sectional study of women of reproductive age with CKD in the United Kingdom, registered online at ClinicalTrials. Gov (Study ID: NCT04370769, 30.04.2020). Ethical approval was given by the Research Ethics Committee and Health Research Authority (London Bloomsbury Research Ethics Committee, Ref: 20/LO/0257). ## Participants and procedure Women who had attended routine renal clinics and/or renal pre-pregnancy counselling at nine sites in the United Kingdom between October 2020 to December 2021 were recruited. Women were eligible if they were between 18 to 50 years old, diagnosed with CKD Stages 1–5 according to the Kidney Disease Improving Global Outcomes guidelines [19], had attended a pre-pregnancy or renal clinic within the last 2 years (2018–2021) and English speaking due to lack of validation of several standardised measures in non-English languages. Women were excluded if they were established on haemodialysis or peritoneal dialysis as risk to their kidney function was no longer relevant, currently pregnant or unable to provide informed consent. Clinical care teams recruited women face-to-face or by telephone. A study invitation, information sheet, data protection document and study hyperlink was sent to their e-mail address. Paper versions were also offered to reduce bias. Informed consent was obtained before any data were collected. Clinical characteristics recorded outside of pregnancy within the last 2 years including most recent estimated glomerular filtration rate (eGFR) using Chronic Kidney Disease Epidemiology Collaboration equation (CKD-EPI) [20] without ethnicity adjustment, primary CKD diagnosis, number of inpatient admissions in the past 5 years, dialysis and transplant history, proteinuria, blood pressure and antihypertensive history were extracted from local databases by clinical care teams. Single measurements of eGFR were used as women were under the care of their local renal team and had an established CKD diagnosis. Clinically relevant cut points were used to identify proteinuria as participating sites recorded either Albumin: creatinine ratio (ACR) or Protein: creatinine ratio (PCR). The cut points were: ACR > 70 mg/mmol or PCR > 100 mg/mmol. Variables are described in Supplementary Table 1. ## Measures Demographic and social information, and perspectives of pregnancy were collected via a self-report questionnaire. Perception of pregnancy risk was assessed by a modified version of the Perception of Pregnancy Risk Questionnaire (PPRQ) [21]. Results are reported as an overall mean score, and two subscale scores: Risk to Self and Risk to Baby. The greater the score, the greater the perceived pregnancy risk. Modifications of the PPRQ for the purposes of this study included alteration of tense to enable completion outside of pregnancy and adaptation to include kidney disease. Content validity of the modified PPRQ was assessed. Summarised changes to the PPRQ and content validity results are presented in supplementary file 1 and supplementary table 2. Pregnancy intention was measured by the Desire to Avoid Pregnancy Scale (DAP) [22]. For application in this study a high score indicated stronger pregnancy intention and a lower score indicated stronger pregnancy avoidance. The following psychological attributes were also measured; depression and anxiety using the Patient Health Questionnaire for Depression and Anxiety (PHQ-4) [23], illness perceptions using Brief Illness Perceptions Questionnaire (B-IPQ) [24], social support using Multidimensional Scale of Perceived Social Support (MSPSS) [25], quality of life using the Kidney Disease Quality of Life Instrument Short Form (KDQOL-SF version 1.3)[26] with sub-scales Physical Component Score (PCS), Mental Component Score (MCS) and Kidney Disease Component Score (KDCS) and lastly COVID-19 risk perception was established and controlled for in the analyses using the COVID-19 Risk Perception Score [27]. These measures are described further in Supplementary Table 1. ## Statistical analysis A priori power analysis was conducted to determine the necessary sample size. Standardised measures were scored following authors’ instructions and summarised alongside baseline and demographic data. Groupwise comparisons between CKD stage and the summarised variables were made (Supplementary Table 3). Correlation between PPRQ subscales was assessed using Pearson r and latent factors modelled. Reliability of the PPRQ scale measure and one summary KDQOL score was determined by Cronbach’s alpha scores (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha$$\end{document}α). The KDQOL subscale was chosen based on the Pearson r correlation between the three summary scores and other psychological attributes measured, the subscale score with weakest association with the other attributes was used. Significance was set at P ≤ 0.05, using the set alpha (α) value of 0.05. Missing data were described and imputed using multiple imputation with 20 imputations. For sensitivity, the analyses were repeated using listwise deletion and results were compared (Supplementary materials: Tables 4, 5, 6, 7, 8). There were two outcome measures; perception of pregnancy risk and pregnancy intention. Initial independent sample t-tests were performed examining perception of pregnancy risk and pregnancy intention between women with different severities of CKD (stage 1–2 compared with CKD stages 3–5) and transplant recipients compared with non-transplant recipients. Univariate linear regression was used to explore the association between variables and perceived pregnancy risk and pregnancy intention. Hierarchical linear regression was used for variable selection for separately assessing association between perception of pregnancy risk and pregnancy intention. The first block adjusted for demographic characteristics that had significant univariate associations, and clinically relevant characteristics (eGFR, history of previous dialysis, transplant, clinically relevant proteinuria, and chronic hypertension), attendance to pre-pregnancy counselling was included as a confounding factor. The second block adjusted for psychological attributes that were significantly associated in the univariate regression. The variance explained (R2) was reported in all models, and multivariate Wald test was performed to evaluate the difference between nested models. Pearson r correlation was performed to assess whether there is an association between perception of pregnancy risk and pregnancy intention with the overall cohort and subgroups with varying severities of CKD: stage 1–2, stages 3–5, non-transplant recipients and transplant recipients. ## Results A total of 716 women were contacted to participate in the study, there was a $44\%$ response rate with 315 women included in the analysis, the majority of whom were recruited from London clinics (251, $80\%$) (Supplementary Table 9). The mean age was 35 years (SD 7.1) (Table 1) and the median pre-pregnancy eGFR was 64 (IQR 56) mL/min/1.73m2. Ninety-four ($30\%$) were renal transplant recipients, 53 ($56\%$) of whom were pre-emptive. Over half had at least one pregnancy (186, $59\%$) and 46 ($14.6\%$) were currently trying to conceive at time of participating. Most women reported pregnancy being important or very important to themselves (234; $74.3\%$) and to their family (211; $67\%$).Table 1Summary of demographic, clinical and psychological variablesVariableOverall ($$n = 315$$)Age (years) (mean (SD))35.0 (7.1) Missing0Ethnicity (%) Asian36 (11.4) Black47 (14.9) White203 (64.4) Mixed13 (4.1) Other15 (4.8) Missing1 (0.3)Education (%) No formal education4 (1.3) GCSE (full time education to age 16)62 (19.7) A Level (full time education to age 18)60 (19.0) Undergraduate degree102 (32.4) Postgraduate degree87 (27.6)Online completion (%)315 ($100\%$)Employment (%) Full-time152 (48.3) Part-time78 (24.8) Home maker26 (8.3) Retired1 (0.3) Unemployed30 (9.5) Student13 (4.1) Other14 (4.4) Missing1 (0.3)Living arrangement (%) Living with partner203 (64.4) Living with relatives/friends63 (20.0) Living alone47 (14.9) Missing2 (0.6) Socio-economic statusa (median [IQR])5.0 [3.0,7.0] Missing12Religion (%) No religion142 (45.1) Christian131 (41.6) Buddhist1 (0.3) Hindu11 (3.5) Jewish2 (0.6) Muslim17 (5.4) Sikh2 (0.6) Other8 (2.5) Missing1 (0.3) Actively practising religion (%)Yes89 (28.3) Missing2 (0.6)Self-reported pregnancy preferences Pregnancy importance (%) Unimportant31 (9.8) Slightly important14 (4.4) Moderately important36 (11.4) Important75 (23.8) Very important159 (50.5) Missing0 Importance of pregnancy to family (%) Unimportant38 (12.1) Slightly important17 (5.4) Moderately important49 (15.6) Important99 (31.4) Very important112 (35.6) Missing0 Attended pre pregnancy counselling (%)Yes108 (34.3) Missing1 (0.3) Currently trying for pregnancy (%)Yes46 (14.6) Missing2 (0.6) Previous pregnancybYes186 (59.0) Missing3 (1.0) Perceived severity of kidney disease (0–100) (median [IQR])50 [25, 70] Missing5Clinical summary Cause of CKD (%) Glomerulonephritis64 (20.3) Chronic progressive nephropathy/ vesicoureteral reflux51 (16.2) Autosomal dominant polycystic kidney disease47 (14.9) Diabetic nephropathy24 (7.6) Congenital/inherited30 (9.5) Transplant9 (2.9) Other38 (12.1) *Systemic lupus* erythematosus35 (11.1) Unknown6 (1.9) Missing11 (3.5) eGFRc (ml/min/1.73m2) (median [IQR])64 [37.0, 93.0] Missing4 No. of inpatient admissions in the past five years (median [IQR])1.00 [0.00, 3.00] *Previous dialysis* (%)Yes50 (15.9) Missing3 (1.0) Previous transplant (%) No220 (69.8) Yes–pre-emptive74 (23.6) Yes–not pre-emptive20 (6.3) Missing1 (0.3) Chronic hypertensiond (%)Yes167 (53.0) Missing23 (7.3) Clinically relevant proteinuriae (%)Yes56 (17.8) Missing105 (33.3)Mean (SD)Median (IQR)Min and maxSkewnessKurtosisMissingCronbach’s \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha$$\end{document}αPsychological summary Perceived pregnancy riskf Overall46 [23]46 (34.2)0–1000.2− 0.519 ($6\%$)0.92 Risk to baby44 [25]43 (36.6)0–1000.3− 0.617 ($5.4\%$)0.91 Risk to mother48 [23]48 (32.8)0–1000.1− 0.59 ($2.9\%$)0.81 COVID-19 risk perceptiong5 [1]5 (1.3)1–7− 0.40.52 ($0.6\%$)0.78 Pregnancy intentionh2 [1]2 (1.8)0–4− 0.1− 1.06 ($1.9\%$)0.96 Distressi3 [3]2 [4]0–121.20.84 ($1.3\%$)0.90 Perceived social supportj Overall perceived social support6 [1]6 (1.4)1–7− 1.63.27 ($2.2\%$)0.95 From significant other6 [1]6 (1.2)1–7− 1.83.11($0.3\%$)0.96 From friends6 [1]6 (2.0)1–7− 1.62.65 ($1.6\%$)0.96 From family6 [1]6 (1.6)1–7− 1.62.33 ($1\%$)0.94 Illness perceptionsk47 [13]48 [16]0–78− 0.40.43 ($1\%$)0.71 Quality of Lifel Physical component summary66 [25]76 (36.7)2–100− 0.9− 0.417 ($5.4\%$)0.95 Mental component summary63 [24]71 (40.1)2–100− 0.7− 0.732 ($10.2\%$)0.91 Kidney disease component summary76 [15]79 (19.4)32–97− 1.00.5123 ($39\%$)0.94eGFR estimated glomerular filtration rate; CKD chronic kidney diseaseaSocioeconomic status is measured using the Index of Multiple Deprivation (IMD). The IMD measures relative deprivation across each small area in England in deciles, where 1 represents the most deprived $10\%$ to 10 which represents the least deprived 10 percent. Deprivation is measures across seven domains; income, employment, education, health, crime, barriers to housing and services, and living environmentbSelf-reported pregnancy historycCalculated using Chronic Kidney Disease Epidemiology Collaboration 2009 equation without ethnicity adjustmentdHistory of antihypertensives and/or > 140 mmHg/ > 90 mmHgeAlbumin: creatinine ratio > 70 mg/mmol or protein: creatinine ratio > 100 mg/mmolfMeasured using a modified version of Perception of Pregnancy Risk Questionnaire, increased score indicates greater perceived risk (0–100)gMeasured using COVID-19 Risk Perception Score, increased score indicates greater perceived COVID risk (1—7)hMeasured using Desire to Avoid Pregnancy Scale, increased score indicates greater pregnancy intention (0–4)iMeasured using Patient Health Questionnaire 4 Item Scale, high score indicates greater anxiety and depression (0–12)jMeasured using Multidimensional Scale of Perceived Social Support, increased score indicates greater social support (1–7)kMeasured using Brief Illness Perceptions Questionnaire, increased score indicates more negative illness beliefs (0–78)lMeasured using Kidney Disease Quality of Life Scale, increased score indicates greater quality of life (0–100) The majority of women had not attended pre-pregnancy counselling (206, $65.4\%$). Those who had attended had more advanced CKD (median pre-pregnancy eGFR 54 mL/min/1.73m2 versus 68 mL/min/1.73m2, $$p \leq 0.007$$), perceived greater pregnancy risk (mean overall perceived pregnancy risk score 51.7 versus 42.5, $$p \leq 0.001$$) and had higher education ($71.3\%$ versus $54.4\%$ university graduates, $$p \leq 0.001$$) (Supplementary Table 10). The subscales of perception of pregnancy risk questionnaire (PPRQ) were strongly correlated ($r = 0.81$). When latent factors were modelled correlation increased ($r = 0.90$), thus the overall perceived pregnancy risk scale was used in the analysis. Correlations between KDQOL summary scores and the other psychological constructs confirmed that the MCS (r = − 0.75, $95\%$ CI − 79 to − 0.69) and KDCS had strong negative correlations with distress (r − 0.66, $95\%$ CI − 0.73 to − 0.57, $p \leq 0.00$). The PCS did not correlate with the psychological attributes measured (r = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\le$$\end{document}≤ 0.50) so the PCS was used as the QoL indicator in regression analyses to avoid collinearity with other variables (Supplementary Table 11). ## Perception of pregnancy risk There was a significant difference in the overall perceived pregnancy risk of women with CKD stages one to two (PPRQ mean 38.7, SD = 22.9) and stages three to five (PPRQ mean 53.2, SD = 23.1); t[280] = − 5.6, $95\%$ CI − 19.6 to − 9.4, $p \leq 0.001.$ There was also a difference in perceived pregnancy risk between women who had received a kidney transplant (PPRQ mean 55.5, SD = 22.5) compared to those who had not (PPRQ mean 41.3, SD = 23.5): t[295] = -5.2, $95\%$ CI − 19.7 to − 8.7, $p \leq 0.001.$ Univariate analysis identified older age, pre-pregnancy counselling attendance, greater perceived severity of CKD, previous dialysis, kidney transplantation, clinically relevant proteinuria, and chronic hypertension as significantly associated with higher perceived pregnancy risk. Preserved kidney function, greater perceived quality of life, and being employed were significantly associated with a lower perception of pregnancy risk. Women who perceived their CKD with more negative beliefs, experienced greater anxiety and depression (distress), or reported a perceived greater risk of COVID-19 had increased perceived pregnancy risk (Table 2).Table 2Assessing the univariate relationships between individual variables with perception of pregnancy risk and pregnancy intention as dependent variablesVariablePerception of pregnancy riskPregnancy intentionCoefficientspCoefficientspAge0.5 (0.1–0.8)0.017− 0.01 (− 0.03 to 0.01)0.22Ethnicity–Asian7.4 (− 1.2 to 16.0)0.0910.9 (0.5–1.3) < 0.001Ethnicity–Black1.5 (− 6.0 to 8.9)0.700.4 (0.1–0.7)0.026Ethnicity–Other3.8 (− 5.7 to 13.4)0.430.2 (− 0.3 to 0.6)0.48Education–GCSE− 7.9 (− 31.4 to 15.7)0.51− 0.2 (− 1.4 to 0.9)0.73Education – A level− 14.5 (− 38.0 to 9.0)0.22− 0.02 (− 1.2 to 1.1)0.97Education–Undergraduate− 18.4 (− 41.6 to 4.8)0.12− 0.2 (− 1.3 to 1.0)0.79Education – Postgraduate− 16.1 (− 39.3 to 7.2)0.18− 0.2 (− 1.3 to 1.0)0.78Employed – Yes− 7.4 (− 13.3 to 1.5)0.0150.1 (− 0.2 to 0.4)0.62Living arrangement – relatives/friends− 1.7 (− 8.6 to 5.1)0.62− 0.3 (− 0.6 to 0.0)0.072Living arrangement – alone0.6 (− 6.9 to 8.1)0.87− 0.3 (− 0.7 to 0.1)0.089Socio-economic statusa− 0.7 (− 1.8 to 0.3)0.18− 0.04 (− 0.1 to 0.1)0.085Religious–yes1.8 (− 3.5 to 7.1)0.500.5 (0.3–0.8) < 0.001Importance of pregnancy0.5 (− 1.5 to 2.5)0.640.4 (0.3–0.5) < 0.001Importance of pregnancy to family0.1 (− 1.9 to 2.1)0.920.3 (0.2–0.4) < 0.001Attended PPC8.9 (3.4–14.4)0.0020.6 (0.3–0.8) < 0.001Pregnancy history1.3 (− 0.2 to 2.8)0.08− 0.04 (− 0.1 to 0.03)0.27Children0.7 (− 2.2 to 3.5)0.64− 0.2 (− 0.4 to − 0.1)0.001Perceived CKD severity0.4 (0.3–0.5) < 0.0010.002 (0.006 to 0.003)0.51Clinical characteristicsPrevious Dialysis–yes11.1 (4.0–18.1)0.0020.1 (− 0.3 to 0.4)0.65Previous Transplant–yes14.2 (8.7–19.7) < 0.0010.1 (− 0.2 to 0.3)0.70eGFRb− 0.2 (− 0.3 to − 0.2) < 0.0010.0001 (− 0.004 to 0.004)0.97Clinically relevant proteinuriac9.1 (1.8–16.4)0.016− 0.1 (− 0.4 to 0.3)0.64Chronic hypertensiond7.9 (2.4–13.4)0.0050.1 (− 0.1 to 0.4)0.29Psychological characteristicsCOVID – risk perceptione5.2 (2.7–7.8) < 0.001− 0.2 (− 0.4 to − 0.1) < 0.001Pregnancy intentionf− 0.04 (− 2.5 to 2.4)0.97–-Pregnancy risk perceptionh––0.0001 (− 0.01 to 0.01)0.97Distressi1.3 (0.5–2.2)0.002− 0.1 (− 0.09 to − 0.01)0.009Social supportj− 0.8 (− 3.1 to 1.5)0.520.1 (− 0.03 to 0.2)0.15Illness perceptionsk0.8 (0.6–1.0) < 0.001− 0.004 (− 0.01 to 0.005)0.39Quality of life11 – physical− 0.2 (− 0.3 to − 0.04)0.0060.01 (0.001–0.01)0.013Reported to one significant figure after decimalIMD index of multiple deprivation; PPC pre-pregnancy counselling; eGFR estimated glomerular filtration rate; CKD chronic kidney disease. reference categories: ethnicity = white ethnicity, education = no formal education, living arrangement = living with partneraSocioeconomic status is measured using the Index of Multiple DeprivationbeGFRcalculated using Chronic Kidney Disease Epidemiology Collaboration 2009 equation without ethnicity adjustmentcAlbumin:creatinine ratio > 70 mg/mmol or Protein: creatinine ratio > 100 mg/mmoldHistory of antihypertensives and/or > 140 mmHg/ > 90 mmHgeMeasured using COVID-19 Risk Perception ScorefMeasured using Desire to Avoid Pregnancy ScalegMeasured using a modified version of Perception of Pregnancy Risk QuestionnairehMeasured using Patient Health Questionnaire 4 Item ScaleiMeasured using Multidimensional Scale of Perceived Social SupportjMeasured using Brief Illness Perceptions QuestionnairekMeasured using Kidney Disease Quality of Life Scale – physical component summary After inclusion of demographic data, psychological attributes and relevant clinical characteristics in the model, pre-pregnancy counselling attendance, greater perceived severity of CKD, more negative illness beliefs and greater COVID-19 risk perception remained significantly associated with greater perceived pregnancy risk (Table 3). There was no association between clinical characteristics and perceived pregnancy risk. Inclusion of psychological variables within the model improved explained variance from $21\%$ (R2 = 0.21, $95\%$ CI 0.13–0.30) to $33\%$ (R2 = 0.33, $95\%$ CI 0.24–0.42). The Wald test was significant ($p \leq 0.001$) indicating that psychological attributes significantly contribute to the model of perceived pregnancy risk. Table 3Adjusted linear regression models investigating the association with perception of pregnancy riskPerception of pregnancy risk modelModel 1Model 2CoefficientspCoefficientspIntercept46.5 (30.4–62.5) < 0.001− 7.9 (-33.5 to 17.6)0.54Demographic factorsAge0.1 (− 0.2 to 0.5)0.430.1 (− 0.3 to 0.4)0.68Employed–Yes− 6.1 (− 11.6 to -0.7)0.028− 3.7 (− 9.1 to 1.7)0.17PPC attended6.8 (1.6–12.0)0.0116.5 (1.5–11.4)0.011Clinical characteristicseGFRa− 0.2 (− 0.2 to − 0.1) < 0.001− 0.03 (− 0.1 to 0.1)0.49Previous dialysis–yes3.5 (− 4.2 to 11.2)0.371.5 (-5.7 to 8.8)0.68Previous transplant–yes7.8 (1.4–14.3)0.0174.7 (− 1.5 to 10.9)0.14Clinically relevant proteinuriab–yes3.9 (− 3.9 to 11.6)0.335.0 (− 2.9 to 12.9)0.21Chronic hypertensionc–yes4.2 (− 1.2 to 9.6)0.132.0 (− 3.3 to 7.3)0.47Psychological attributesCOVID risk perceptiond––2.7 (0.3–5.2)0.031Distresse––0.7 (− 0.2 to 1.6)0.14Illness perceptionsf––0.4 (0.2–0.7)0.001Quality of lifeg–PCS––0.1 (− 0.03 to 0.2)0.13Perceived CKD severity––0.1 (0.04–0.3)0.010Model summaryR20.21 (0.13–0.30)0.33 (0.24–0.42)ΔR20.19 (0.12–0.28)0.30 (0.21–0.39)Reported to one significant figure after decimal. Model 1 adjusts for demographic and clinical factors only. Model 2 adjusts for demographic, clinical and psychological factorsPPC pre-pregnancy counselling; eGFR estimated glomerular filtration rate; CKD chronic kidney disease; R2 R squared; ΔR2 Adjusted R squareaeGFRcalculated using Chronic Kidney Disease Epidemiology Collaboration 2009 equation without ethnicity adjustmentbAlbumin:creatinine ratio > 70 mg/mmol or Protein: creatinine ratio > 100 mg/mmolcHistory of antihypertensives and/or > 140 mmHg/ > 90 mmHgdMeasured using COVID-19 Risk Perception ScoreeMeasured using Patient Health Questionnaire 4 Item ScalefMeasured using Brief Illness Perceptions QuestionnairegMeasured using Kidney Disease Quality of Life Scale physical component summary ## Pregnancy intention There was no significant difference in pregnancy intentions between women with CKD stages one to two (pregnancy intention mean score = 2, SD = 1.1) and stages three to five (pregnancy intention mean score = 2.1, SD = 1.2); t[303] = -0.5, $95\%$ CI -0.3 to 0.2, $$p \leq 0.647.$$ There was no significant difference in pregnancy intentions between women who had received a kidney transplant (pregnancy intention mean score = 2.1, SD = 1.2) compared to those who had not (pregnancy intention mean score = 2.0, SD = 1.1): t[307] = − 0.4, $95\%$ CI − 0.33 to 0.22, $$p \leq 0.697.$$ In the univariate analyses religious identity, Black or Asian ethnicity, attendance at pre-pregnancy counselling, regarding pregnancy as important to themselves and their families, and greater quality of life were all measurably associated with greater pregnancy intention. Conversely, greater perceived COVID-19 risk, greater distress, and increased parity were associated with avoidance of pregnancy. Clinical characteristics were not associated with pregnancy intention (Table 2). In the multivariable model, pregnancy counselling attendance, religious identity and regarding pregnancy with greater importance were significantly associated with greater pregnancy intention (Table 4). An increase in number of children, and greater perceived risk of COVID-19 were associated with avoidance of pregnancy. No association was identified between clinical characteristics and pregnancy intention. Inclusion of psychological variables within the model improved explained variance from $33\%$ (R2 = 0.33, $95\%$ CI 0.25–0.42) to $36\%$ (R2 = 0.36, $95\%$ CI 0.28–0.45) with significant Wald test ($$p \leq 0.005$$) indicating that psychological attributes contribute to the model explaining pregnancy intention. Table 4Adjusted linear regression models investigating associations with pregnancy intention as outcomeVariableModel 1Model 2CoefficientspCoefficientspDemographic factorsIntercept0.4 (− 0.1 to 0.9)0.110.9 (0.1–1.7)0.035Ethnicity–Asian0.3 (− 0.1 to 0.7)0.130.3 (− 0.1 to 0.7)0.13Ethnicity–Black0.02 (− 0.3 to 0.4)0.89− 0.003 (− 0.3 to 0.3)0.99Ethnicity–Other− 0.01 (− 0.4 to 0.4)0.940.1 (− 0.3 to 0.5)0.69Religion–yes0.3 (0.05–0.5)0.0190.3 (0.03–0.5)0.028Importance of pregnancy0.4 (0.2 to 0.5) < 0.0010.4 (0.2–0.5) < 0.001Importance of pregnancy to family0.01 (− 0.1 to 0.1)0.870.02 (− 0.1 to 0.2)0.76Attended PPC0.3 (0.05–0.5)0.0190.3 (0.01–0.5)0.041Children− 0.31 (− 0.4 to − 0.2) < 0.001− 0.3 (− 0.4 to − 0.2) < 0.001Clinical characteristicseGFRa0.0002 (− 0.003 to 0.004)0.920.0005 (− 0.004 to 0.003)0.79Previous dialysis–yes0.3 (− 0.1 to 0.6)0.120.3 (− 0.1 to 0.6)0.096Previous transplant–yes− 0.02 (− 0.3 to 0.3)0.870.005 (− 0.3 to 0.3)0.97Clinically relevant proteinuriab–yes− 0.2 (− 0.5 to 0.1)0.11− 0.2 (− 0.5 to 0.1)0.11Chronic hypertensionc–yes0.2 (− 0.1 to 0.4)0.160.2 (− 0.001 to 0.5)0.051Psychological attributesCOVID risk perceptiond––− 0.1 (− 0.2 to -0.01)0.029Quality of life–physicale––0.003 (− 0.002 to 0.01)0.28Distressf––− 0.02 (− 0.1 to 0.02)0.25Model summaryR20.33 (0.25–0.42)0.36 (0.28 to 0.45)ΔR20.3 (0.22–0.39)0.33 (0.24 to 0.41)Reported to one significant figure after decimal. Model 1 adjusts for demographic and clinical factors only. Model 2 adjusts for demographic, clinical and psychological factorsPPC pre-pregnancy counselling; eGFR estimated glomerular filtration rate; CKD chronic kidney disease; R2 R squared; ΔR2 adjusted R squareaeGFR calculated using Chronic Kidney Disease Epidemiology Collaboration 2009 equation without ethnicity adjustmentbAlbumin: creatinine ratio > 70 mg/mmol or Protein: creatinine ratio > 100 mg/mmolcHistory of antihypertensives and/or > 140 mmHg/ > 90 mmHgdMeasured using COVID-19 Risk Perception ScoreeMeasured using Kidney Disease Quality of Life Scale physical component summaryfMeasured using Patient Health Questionnaire 4 Item Scale There was no association between perception of pregnancy risk and pregnancy intention (r = − 0.002, $95\%$ CI − 0.12 to 0.11, $$p \leq 0.97$$). No association was consistent in the subgroup analyses amongst: CKD stages 1 to 2 ($r = 0.09$, $95\%$ CI − 0.08 to 0.25, $$p \leq 0.30$$), CKD stages 3 to 5 (r = − 0.08, $95\%$ CI -0.24 to 0.09, $$p \leq 0.35$$), kidney transplant recipients (r = − 0.1, $95\%$ CI − 0.3 to 0.12, $$p \leq 0.38$$) and non-kidney transplant recipients ($r = 0.04$, $95\%$ CI − 0.1 to 0.18, $$p \leq 0.57$$). ## Discussion Clinical risk factors for adverse pregnancy outcomes including eGFR, hypertension and proteinuria did not affect perception of pregnancy risk or future pregnancy intention in women with CKD in the adjusted analyses. Severity of CKD and previous transplant history had no association with pregnancy intention. Women’s own perception of CKD severity was associated with increased perceived pregnancy risk, but this did not influence women’s pregnancy intentions, thus highlighting the priority of pregnancy for many women. Women who have previously attended pre-pregnancy counselling have greater perceived pregnancy risk and pregnancy intention. Women with CKD have greater pregnancy risk perception scores in comparison to other tested cohorts. Previous risk perception scores in uncomplicated pregnancies were 24.0 (SD 14.5) [21], substantially lower than the overall scores reported in this study (PPRQ overall: 46, SD 23). Women with reduced kidney function (CKD stage 3–5) reported to perceive greater pregnancy risk in comparison to women with CKD stages 1–2. Overall, we observed that measured kidney function has no association with either perceived pregnancy risk or pregnancy intention, and women’s own perception of their CKD severity had a stronger impact upon their perceived pregnancy risk than measured kidney function. This is in agreement with previous studies that report differences between women’s perceived risk and that of a healthcare professional or an objective risk assessment [13, 14]. Qualitative studies report the decisional conflict between balancing pregnancy desire and pregnancy risk related to CKD [10]. This study demonstrates that for many women with CKD pregnancy desire outweighs perceived risk as perception of pregnancy risk did not impact pregnancy intention across the cohort nor in subgroups with varying severities and transplant history. This finding is not aligned with psychological theories that propose behaviour intention can be explained and predicted by perceived risk along with other psychological constructs [15, 16]. The psychological cost that may be incurred by avoiding pregnancy should also be acknowledged [15]. For many women, pregnancy is valued as a life ambition and thus the psychological cost of avoiding pregnancy may exceed potential risk. Similar to other reports, attendance to pre-pregnancy counselling was low [10], but women who had attended pre-pregnancy counselling reported greater risk perception, which may be due to being more informed. Alternatively, women who attend pre-pregnancy counselling may have pre-existing greater perceived pregnancy risk, which underlies their motivation to attend. Assessment of risk perception before and after pre-pregnancy counselling would provide further insight. Pre-pregnancy counselling was attended by women with more clinical risk factors, which likely reflects referral pathways for women with increased clinical risk. However, pre-pregnancy counselling was also attended by women with higher levels of educational achievement, perhaps highlighting inequality of access. Although the lack of an association between clinical risk factors and a woman’s perception of pregnancy risk presented here may be due to the importance of pregnancy to women regardless of clinical status, it may also demonstrate limited awareness of actual risk that could be improved with pre-pregnancy counselling. All women with CKD should therefore be offered pre-pregnancy counselling, with attendance facilitated and language barriers addressed to reduce inequality. The study findings need to be interpreted considering its limitations. Factors that may influence women’s perceptions of pregnancy risk have not been adjusted for in the analysis due to restricted access to data or unreliable reporting. This includes individual complexities of uncontrolled disease and medication history which may influence perceptions of pregnancy risk due to possible teratogenic risks [28]. In addition, the availability heuristic proposes that women’s previous experiences of pregnancy, including complications such as pre-eclampsia and pre-term deliveries, as well as what stage they are at in their pregnancy journey, may also be used to formulate their perception of pregnancy risk and pregnancy intention [29]. It is important to consider the possible bias underpinning attendance to pre-pregnancy counselling clinics and how this attendance may subsequently influence pregnancy risk perception. The PPRQ was modified for this research and whilst the content validity of the instrument was assessed, further psychometric validation assessment is recommended. Secondly, women participated during the COVID-19 pandemic and so the data collected was a snapshot at a time where pregnancy intention and risk perception may have changed. Furthermore, the majority of the cohort had an undergraduate degree or above, in comparison to 40.6 percent in the United Kingdom having post-secondary school educational qualification [30], thus is unlikely to be representative of all women with CKD. Digital exclusion may be contributory, although postal questionnaires were offered. In addition, women were recruited from metropolitan areas within the United Kingdom with access to pre-pregnancy counselling, which would differ for women from rural areas and may either increase or reduce their perceptions of pregnancy risk. It is important to note a large proportion of the cohort were White women. It is possible that women from different ethnicities may have different perceptions of pregnancy risk, for instance people of African or Afro-Caribbean ancestry are at greater risk of CKD progression [31], adverse pregnancy outcomes and mortality [32] which may heighten their perception of pregnancy risk. Lastly, the findings cannot be generalised to a non-English speaking population. This is the first study that has investigated perceptions of pregnancy risk and pregnancy intentions in women with CKD and has highlighted that pregnancy desire may be unwavering regardless of perceived or objective pregnancy risk. Pregnancy is important to women with CKD, but frequently women are unaware of their risk of adverse pregnancy outcomes. The clinical implication of these findings is that there needs to be greater efforts to support physical and psychological optimisation prior to pregnancy for women with CKD. This includes greater emphasis for women with CKD to receive pre-pregnancy counselling as recommended by the mentioned guidelines to facilitate informed decision making [4, 11]. Pre-pregnancy counselling should include the successful communication of clinical risk including adverse pregnancy and renal outcomes. However, shared-decision making regarding pregnancy requires an understanding of perceived risk, and the importance of pregnancy. ## Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 150 KB) ## References 1. Piccoli GB, Attini R, Vasario E. **Pregnancy and chronic kidney disease: a challenge in all CKD stages**. *Clin J Am Soc Nephrol* (2010.0) **5** 844-855. DOI: 10.2215/CJN.07911109 2. MacLaughlin HL, Hall WL, Sanders TA, Macdougall IC. **Risk for chronic kidney disease increases with obesity: health survey for England 2010**. *Public Health Nutr* (2015.0) **18** 3349-3354. DOI: 10.1017/S1368980015000488 3. Wu B, Bell K, Stanford A. **Understanding CKD among patients with T2DM: prevalence, temporal trends, and treatment patterns—NHANES 2007–2012**. *BMJ Open Diabetes Res Care* (2016.0) **4** e000154. DOI: 10.1136/bmjdrc-2015-000154 4. Wiles K, Chappell L, Clark K. **Clinical practice guideline on pregnancy and renal disease**. *BMC Nephrol* (2019.0) **20** 1-43. DOI: 10.1186/s12882-019-1560-2 5. Piccoli GB, Cabiddu G, Attini R. **Risk of adverse pregnancy outcomes in women with CKD**. *J Am Soc Nephrol* (2015.0) **26** 2011-2022. DOI: 10.1681/ASN.2014050459 6. Wiles K, Webster P, Seed P. **The impact of chronic kidney disease Stages 3–5 on pregnancy outcomes**. *Nephrol Dial Transpl* (2020.0). DOI: 10.1093/ndt/gfaa247 7. Harel Z, Park A, McArthur E. **Prepregnancy renal function and risk of preterm birth and related outcomes**. *CMAJ* (2020.0) **192** 851-857. DOI: 10.1503/cmaj.200089 8. al Khalaf S, O’Reilly É, McCarthy F. **Pregnancy outcomes in women with chronic kidney disease and chronic hypertension: a National cohort study**. *Am J Obstet Gynecol* (2021.0) **225** 1-20. DOI: 10.1016/j.ajog.2021.03.045 9. Hall M, Lightstone L. **Prepregnancy counseling and risk assessment**. *Renal disease in pregnancy* (2018.0) 13-24 10. Jesudason S, Tong A. **The patient experience of kidney disease and pregnancy**. *Best Pract Res Clin Obstet Gynaecol* (2019.0) **57** 77-88. DOI: 10.1016/j.bpobgyn.2018.12.003 11. 11.Centre for Maternal and Child Enquiries (2011) Saving mothers’ lives: reviewing maternal deaths to make motherhood safer: 2006–08. In: The Eighth Report on Confidential Enquiries into Maternal Deaths in the United Kingdom 12. Lee S. **Risk perception in women with high-risk pregnancies**. *Br J Midwifery* (2014.0) **22** 8-13. DOI: 10.12968/bjom.2014.22.1.8 13. Heaman M, Beaton A, Gupton A, Sloan J. **A comparison of childbirth expectations in high-risk and low-risk pregnant women**. *Clin Nurs Res* (1992.0) **1** 252-265. DOI: 10.1177/105477389200100305 14. Gray B. **Hospitalization history and differences in self-rated pregnancy risk**. *West J Nurs Res* (2006.0) **28** 216-229. DOI: 10.1177/0193945905282305 15. Janz N, Becker M. **The health belief model. A decade later**. *Health Educ Q* (1984.0) **11** 1-47. DOI: 10.1177/109019818401100101 16. Rogers R. **A protection motivation theory of fear appeals and attitude change**. *J Psychol* (1975.0) **91** 93-114. DOI: 10.1080/00223980.1975.9915803 17. Widnes SF, Schjøtt J, Granas AG. **Risk perception and medicines information needs in pregnant women with epilepsy: a qualitative study**. *Seizure* (2012.0) **21** 597-602. DOI: 10.1016/j.seizure.2012.06.007 18. Kim C, McEwen L, Piette J. **Risk perception for diabetes among women with histories of gestational diabetes mellitus**. *Diabetes Care* (2007.0) **30** 2281-2286. DOI: 10.2337/dc07-0618 19. **KDIGO 2012 clinical practice guideline for evaluation and management of CKD**. *Kidney Int Suppl* (2012.0). DOI: 10.1038/kisup.2012.76 20. Levey A, Stevens L, Schmid C. **A new equation to estimate glomerular filtration rate**. *Ann Intern Med* (2009.0) **150** 604-612. DOI: 10.7326/0003-4819-150-9-200905050-00006 21. Heaman M, Gupton A. **Psychometric testing of the perception of pregnancy risk questionnaire**. *Res Nurs Health* (2009.0) **32** 493-503. DOI: 10.1002/nur.20342 22. Rocca CH, Ralph LJ, Wilson M. **Psychometric evaluation of an instrument to measure prospective pregnancy preferences**. *Med Care* (2019.0) **57** 152-158. DOI: 10.1097/MLR.0000000000001048 23. Kroenke K, Spitzer RL, Williams J, Lowe B. **An ultra-brief screening scale for anxiety and depression: the PHQ-4**. *Psychosomatics* (2009.0) **50** 613-621. DOI: 10.1176/appi.psy.50.6.613 24. Broadbent E, Petrie KJ, Main J, Weinman J. **The brief illness perception questionnaire**. *J Psychosom Res* (2006.0) **60** 631-637. DOI: 10.1016/j.jpsychores.2005.10.020 25. Zimet G, Dahlem N, Zimet S, Farley G. **The multidimensional scale of perceived social support**. *J Pers Assess* (1988.0) **52** 30-41. DOI: 10.1207/s15327752jpa5201_2 26. Hays R, Kallich J, Mapes D. **Development of the kidney disease quality of life KDQOL instrument**. *Qual Life Res* (1994.0) **3** 329-338. DOI: 10.1007/BF00451725 27. Dryhurst S, Schneider CR, Kerr J. **Risk perceptions of COVID-19 around the world**. *J Risk Res* (2020.0) **23** 994-1006. DOI: 10.1080/13669877.2020.1758193 28. Widnes S, Schjøtt J. **Risk perception regarding drug use in pregnancy**. *Am J Obstet Gynecol* (2017.0) **216** 375-378. DOI: 10.1016/j.ajog.2016.12.007 29. Tversky A, Kahneman D. **Availability: a heuristic for judging frequency and probability**. *Cogn Psychol* (1973.0) **5** 207-232. DOI: 10.1016/0010-0285(73)90033-9 30. 30.Eurostat (2022) Population by educational attainment level, sex and age (%) main indicators. https://appsso.eurostat.ec.europa.eu/nui/submitViewTableAction.do. Accessed 5 May 2022 31. **K/DOQI clinical practice guidelines for chronic kidney disease: evaluation, classification, and stratification**. *Am J Kidney Dis* (2002.0) **39** 1-266. PMID: 11774095 32. MacDorman MF, Thoma M, Declcerq E, Howell EA. **Racial and ethnic disparities in maternal mortality in the United States using enhanced vital records, 2016–2017**. *Am J Public Health* (2021.0) **111** 1673-1681. DOI: 10.2105/AJPH.2021.306375
--- title: A comprehensive platform for analyzing longitudinal multi-omics data authors: - Suhas V. Vasaikar - Adam K. Savage - Qiuyu Gong - Elliott Swanson - Aarthi Talla - Cara Lord - Alexander T. Heubeck - Julian Reading - Lucas T. Graybuck - Paul Meijer - Troy R. Torgerson - Peter J. Skene - Thomas F. Bumol - Xiao-jun Li journal: Nature Communications year: 2023 pmcid: PMC10041512 doi: 10.1038/s41467-023-37432-w license: CC BY 4.0 --- # A comprehensive platform for analyzing longitudinal multi-omics data ## Abstract Longitudinal bulk and single-cell omics data is increasingly generated for biological and clinical research but is challenging to analyze due to its many intrinsic types of variations. We present PALMO (https://github.com/aifimmunology/PALMO), a platform that contains five analytical modules to examine longitudinal bulk and single-cell multi-omics data from multiple perspectives, including decomposition of sources of variations within the data, collection of stable or variable features across timepoints and participants, identification of up- or down-regulated markers across timepoints of individual participants, and investigation on samples of same participants for possible outlier events. We have tested PALMO performance on a complex longitudinal multi-omics dataset of five data modalities on the same samples and six external datasets of diverse background. Both PALMO and our longitudinal multi-omics dataset can be valuable resources to the scientific community. The analysis of longitudinal bulk and single-cell multi-omics data is a highly complex task. Here, the authors introduce PALMO, a software platform with five modules to analyse longitudinal bulk and single-cell multi-omics data, which is extensively tested in external datasets that include multiple omics modalities. ## Introduction Applying multi-omics technologies to measure longitudinal specimens of human participants provides unprecedented insights on disease such as COVID-191–3, diabetes4 and lymphoma5. Single-cell technologies such as single-cell ribonucleic acid sequencing (scRNA-seq) and single-cell assay for transposase-accessible chromatin sequencing (scATAC-seq) can offer granular details on disease mechanisms and are increasingly utilized in biological and clinical research6–8. It is anticipated that more and more longitudinal bulk and single-cell omics data will be generated by the scientific community. Different statistical methods are used to analyze longitudinal data to account for the diversities in research interest, study design, and/or data type (continuous or categorical)9,10. Generalized linear mixed model (GLMM) is a popular approach for analyzing continuous longitudinal data. It is common that the same dataset can be examined from multiple perspectives with different methods. Complications such as human heterogeneity, interdependency between multiple samples of same participant, missing and/or incomplete data, unbalanced dataset, and unexpected outlier events (e.g., severe adverse events in clinical trials) are all intrinsic to longitudinal data. The usage of single-cell technologies brings additional complications such as dropout, sparseness, interdependency between cells of same sample, and unbalanced cell counts in individual samples11,12. Advanced methods have been applied to analyze longitudinal bulk omics data with customized codes for specific projects4,13. Sophisticated methods for analyzing cross-sectional single-cell omics data have also been developed with mixed performance14–18. While software tools such as variancePartition19 and tcR20 can be repurposed to examine longitudinal omics data either from a single perspective and/or collected on a single technical platform, we are not aware of any well-accepted software package that is specifically designed to analyze longitudinal bulk and single-cell omics data. Instead, researchers rely on customized codes to analyze such data, which is time-consuming, error-prone and a non-small challenge to many people. A comprehensive yet simple-to-use software tool to extract insightful information from longitudinal omics data is desired. Here, we present PALMO (https://github.com/aifimmunology/PALMO), a software package designed to analyze longitudinal bulk and single-cell omics data (Fig. 1a). Five analytical modules are implemented in PALMO (Fig. 1b): (i) Variance decomposition analysis (VDA) evaluates contributions of factors of interest to the total variance of individual features (Fig. 1c). ( ii) Coefficient of variation (CV) profiling (CVP) assesses intra-participant variation over time in bulk data and identifies consistently stable or variable features among participants (Fig. 1d). ( iii) Stability pattern evaluation across cell types (SPECT) assesses longitudinal stability patterns of features in single-cell omics data and identifies stable or variable features that are unique to individual cell types but consistent among participants (Fig. 1e). ( iv) Outlier detection analysis (ODA) examines the possibility of abnormal events occurring during a longitudinal study (Fig. 1f). ( v) Time course analysis (TCA) evaluates transcriptomic changes over time based on longitudinal scRNA-seq data of the same participant and identifies genes that exhibit significant temporal changes (Fig. 1g). Together these five modules provide unique insights on longitudinal omics data from multiple perspectives. We also developed functions to display CVs of features of interest in circos plots (Fig. 1h). We test PALMO performance on a complex longitudinal multi-omics dataset of five data modalities and six external datasets of diverse background. Fig. 1General workflow and analysis schema of PALMO.a PALMO can work with complex longitudinal data, including clinical data, bulk omics data, and single-cell omics data. b Overview of five analytical modules implemented in PALMO. c Variance decomposition analysis (VDA) applies generalized linear mixed model to assess contributions of factors of interest (such as disease status, sex, individual participant, cell type, experimental batch, etc.) to the total variance of individual features in the data. d Coefficient of variation (CV) profiling (CVP) is designed for bulk longitudinal data, calculates CV of repeated measurements on the same participant to assess the corresponding longitudinal stability, and compares CVs of different participants to identify consistently stable or variable features. e Stability pattern evaluation across cell types (SPECT) is the CVP counterpart for single-cell omics data, analyzes stability patterns of features across different cell types and different participants, classifies features based on how often they are stable or variable in cell type-donor combinations, and identifies features that are unique to individual cell types and consistent among participants. f Outlier detection analysis (ODA) evaluates how many features in a sample are outliers when compared with the corresponding features in other samples of same participant, assesses whether the number of outlier features in the sample is significantly higher than expectation, and identifies possible abnormal events occurred during a longitudinal study. g Time course analysis (TCA) uses the hurdle model to evaluate transcriptomic changes over time based on longitudinal scRNA-seq data of same participants, models time as a continuous variable for data with at least three timepoints, and identifies up- or down-regulated genes over time. h PALMO uses circos plots to display CVs of features of interest and reveal stability patterns across features, participants, cell types, and data modalities. Adobe Illustrator (version 27.1.1; https://www.adobe.com/products/illustrator.html) was used to draw (a), arrange panels, and edit text. PowerPoint (version 16.69; https://www.microsoft.com/en-us/microsoft-365/powerpoint) was used to draw (b). ## A complex longitudinal multi-omics dataset to demonstrate PALMO performance To demonstrate PALMO performance, we collected sixty blood samples (plasma and peripheral blood mononuclear cells (PBMCs)) from six healthy, non-smoking Caucasian donors (three females and three males) between 25 to 38 years old over a 10-week period (Supplementary Fig. 1a). Complete blood count (CBC) was collected on all these samples (Supplementary Data 1a). The abundance of 1,156 plasma proteins were measured on these samples as well (Supplementary Data 1c), but only 1,042 ($68\%$) proteins had reliable quantification results (Supplementary Data 2a). High-dimensional flow cytometry and droplet-based scRNA-seq assays were performed on a subset of 24 PBMC samples from four donors (one female and three males) over Week 2 to 7. A total of 27 cell types were identified from flow cytometry data (Supplementary Fig. 2, Supplementary Data 1b). Droplet-based scATAC-seq assay was also performed on 18 out of the 24 PBMC samples. This multi-omics dataset of five data modalities on the same samples can be a valuable resource to the scientific community for immune health study and/or software development. We retrieved high quality scRNA-seq data of 472,464 cells and labeled them to 31 different cell types using Seurat level2 labelling16 (Supplementary Fig. 3a, b, Supplementary Data 3a). Among the nineteen overlapping cell types identified by both scRNA-seq and flow cytometry, the corresponding cell frequencies as measured by the two data modalities were highly correlated (two-sided $p \leq 0.05$ on Pearson correlation as evaluated by R function “cor.test()”) except for those of double negative T (dnT) cells (Supplementary Fig. 3c). Unless specified otherwise, we filtered out low frequent cell types (average frequency <$0.5\%$) and kept 19 out of the 31 cell types for downstream analysis (Supplementary Data 3b). We also kept only 11,191 genes that had an average (across timepoints) expression of 0.1 or higher in at least one cell type of one donor. scATAC-seq data was analyzed using the ArchR21 package. We observed 294,623 peaks in 135,566 cells after removing doublets. Cells were labeled to 28 different cell types using genescore matrix as implemented in ArchR (Supplementary Fig. 3d, e). We noticed the labeling scores on scATAC-seq data were much lower than the corresponding scores on scRNA-seq data, likely reflecting the challenge in cell labeling on scATAC-seq data. We filtered out low quality cells (labeling score <0.5), removed cell types having less than 50 remaining cells, and kept 14 out of the 28 cell types for downstream analysis (Supplementary Data 3b). We also kept only 24,769 genes that had an average (across timepoints) gene score of 0.1 or higher in at least one cell type of one donor. In addition to our own data, we also evaluated PALMO performance against six external omics datasets of diverse complexities, different sample types and/or different technical platforms (Supplementary Fig. 1b). More examples of PALMO usage beyond those presented here can be found in PALMO vignettes (https://github.com/aifimmunology/PALMO/blob/main/Vignette-PALMO.pdf), including performance on unbalanced data, data with replicates, and data of a single donor with multiple timepoints. ## Application of VDA to assess sources of variations We applied VDA to evaluate inter- and intra-donor variations in our bulk data (CBC, PBMC frequencies from flow cytometry, and plasma proteomics data), using donor and week (timepoint) as factors of interest. CBC measurements showed strong inter-donor variations and minuscule intra-donor variations (Supplementary Fig. 4a, b). PBMC frequencies from flow cytometry showed very strong inter-donor variations (Supplementary Fig. 4c, d) with intra-class correlation (ICC) ranging from $51\%$ (IgD CD27− B cells) to $98\%$ (CD4 Temra: CD4+ effector memory T cells re-expressing CD45RA). In comparison, the highest ICC on intra-donor variations was $19\%$ (cDC1: conventional type 1 dendritic cells). Plasma proteins followed a similar trend with some exceptions (Supplementary Fig. 4e, f, Supplementary Data 2a). Inter-donor variations of 621 ($60\%$) out of the 1042 quantified proteins contributed more than $50\%$ to the corresponding total variance. Only 75 proteins ($7\%$) had more intra-donor variation than inter-donor variation, but none contributed more than $50\%$ to the total. A previous study22 identified 155 proteins having high inter-donor variations, $81\%$ [126] of which overlapped with the 621 inter-donor variable proteins. We added cell type as a factor of interest in the VDA of our scRNA-seq and scATAC-seq data. Inter-cell-type variations were more prominent than inter- and intra-donor variations in both single-cell data modalities. Based on our scRNA-seq data, 10, 0, and 4384 genes had more than $50\%$ of total variance from inter-donor, intra-donor, and inter-cell-type variations, respectively (Fig. 2a, Supplementary Data 3c). Nine of the top ten inter-cell-type variable genes (ICC: 98–$99\%$, Fig. 2b) have known immune functions (Supplementary Data 3d). The top gene, LILRA4, is predominantly expressed in plasmacytoid dendritic cells (pDCs) and prevents pDCs from overblown reaction to viral infections23. Six of the top ten inter-donor variable genes (ICC: 53–$94\%$, Fig. 2c) are linked to the X or Y chromosome and seven of them showed differential expression between ovary and prostate/testis, reflecting the sex difference between male and female donors. Contributions from intra-donor variations to the total variance were small (ICC ≤ $3\%$, Fig. 2d), indicating the immune systems of the four healthy donors were quite stable over the study period. Fig. 2Variance decomposition on longitudinal single-cell omics data.a Overall distributions of variance explained by inter-donor variations (Donor), longitudinal intra-donor variations (Week), variations among cell types (Celltype), or residual variations (Residual) based on scRNA-seq data. The scRNA-seq data was collected on 24 independent peripheral blood mononuclear cell (PBMC) samples from $$n = 4$$ healthy participants with each participant contributing one sample a week for 6 weeks. The distributions were evaluated based on pseudo-bulk intensities of $$n = 11$$,191 genes in 19 cell types. b, c Examples of genes whose total expression variance was most explained by inter-cell-type variations (b) or inter-donor variations (c). d Examples of genes that had the most but still minuscular intra-donor variations in expression. b–d Pseudo-bulk intensities of the corresponding genes in 19 cell types were displayed in boxplots. e Same as (a) but based on scATAC-seq data from $$n = 18$$ out of the 24 PBMC samples with 2 participants contributing 6 samples while other 2 participants contributing 3 samples. The distributions were evaluated based on gene scores of $$n = 24$$,769 genes in 14 cell types. f, g The top list of genes whose inter-cell-type (f) or inter-donor (g) variations contributed most to the total variance in scATAC-seq data. h The top list of genes that had the most intra-donor variations in scATAC-seq data. a–e Each boxplot displays the median (centerline), the first and third quartiles (the lower and upper bound of the box), and the 1.5x interquartile range (whiskers) of the data. ICC: intra-class correlation. Adobe Illustrator (version 27.1.1; https://www.adobe.com/products/illustrator.html) was used to arrange panels and edit text. Source data are provided as a Source Data file. The VDA results on our scATAC-seq data, using genescore matrix, showed similar trends as that on our scRNA-seq data (Fig. 2e). A total of 33, 0, and 7847 genes had more than $50\%$ of total variance from inter-donor, intra-donor, and inter-cell-type variations, respectively (Supplementary Data 3e). All the top ten inter-cell-type variable genes (ICC: 95–$97\%$, Fig. 2f) have known immune functions (Supplementary Data 3f). The top gene, SPIB, is an enhancer regulating pDC development24. Among the top ten inter-donor variable genes (ICC: 58–$89\%$, Fig. 2g), XIST, ZNF705D, GTF2IRD2, and USP32P2 have differential expression between ovary and prostate/testis; RHD encodes a key protein in the Rh blood group system; and GSTM1 belongs to a highly polymorphic supergene family and affects heterogeneous response to toxicity25. *These* genes appeared to capture more diverse types of differences among donors than their counterparts from scRNA-seq data. The ICCs of the top five intra-donor variable genes (ICC: 32–$34\%$, Fig. 2h) were about 10-fold higher than that of the corresponding top gene, JUN, by scRNA-seq data, suggesting chromatin accessibility might be more sensitive to biological changes than gene expression. variancePartition19 was previously developed to study variations in gene expression data and can be applied to longitudinal omics data for the same purpose. VDA generated almost identical results as variancePartition on two tested datasets after removing missing values (Supplementary Fig. 5), which was needed to run variancePartition but not VDA. VDA can be used to study T-cell receptor (TCR) repertoires. Previously sorted CD4+ and CD8+ non-naïve T cells were isolated from PBMC samples of four systemic sclerosis (SSc) donors and analyzed to obtain sequencing data of TCR β-chains26. The data was originally analyzed using tcR20, which was developed specifically for TCR data with functions either providing sample-level views on the whole repertories or treating clonotype data as binary (present or absent). We downloaded the TCRβ data (GSE156980) and calculated the frequency of unique clonotypes from both CD4+ and CD8+ T cells. A total of 288,597 unique clonotypes were obtained from CD4+ T cells and 11,739 from CD8+ T cells, respectively. We treated the clonotype data as continuous and used donor, time, and subtype (limited SSc versus diffuse SSc) as factors of interest in VDA. We identified from CD4+ T cells 6,625, 3, and 41 clonotypes having more than $50\%$ of total variance from inter-donor, intra-donor, and inter-subtype variations, respectively (Supplementary Fig. 6a–d, Supplementary Data 4a). The corresponding counts from CD8+ T cells were 650, 0, and 1 (Supplementary Fig. 6e–h, Supplementary Data 4b). As illustrated in Supplementary Fig. 6b, f, many inter-donor variable clonotypes were donor-specific and stable over time, making them potential candidates responsible for SSc pathogenesis. The identification of inter-subtype variable clonotypes (Supplementary Fig. 6d, h) is interesting since some of them might be specific to either limited SSc or diffuse SSc. VDA provided additional insights on the TCR data beyond the original study26. ## Application of CVP to evaluate longitudinal stability We applied CVP to identify longitudinally stable and variable proteins from our proteomics data (Fig. 3a). The distribution of median CV (among donors) peaked near $5\%$ (Supplementary Fig. 7a), which we used as a cut-off to separate variable (median CV > $5\%$) and stable (median CV < $5\%$) proteins (Supplementary Data 2b–d). A total of 413 proteins were longitudinally variable, among which SNAP23, GRAP2, ARG1, AIFM1, and MESD had the highest median CV (24.6-$27.7\%$, Fig. 3b). Such moderate CV values are consistent with the observed low intra-donor variations by VDA. A total of 629 proteins were longitudinally stable, among which SOD2, NRP2, OSCAR, NRCAM, and MIA had the lowest median CV (0.6–$0.8\%$, Fig. 3c). These stable proteins may be interesting biomarker candidates if they change under some disease conditions. They can also be used to bridge proteomics data of different experimental batches. Fig. 3Longitudinal stability of plasma proteome.a Scatter plots of coefficient of variation (CV) versus mean of normalized protein expression (NPX) over 10 timepoints in $$n = 6$$ participants. One plasma sample per week was collected from $$n = 6$$ participants over 10 weeks. The evaluation for each participant was based on measurements on 1042 proteins in the corresponding 10 plasma samples. The longitudinal stable and variable proteins are represented in blue and red, respectively. b, c Heatmap of CV of top 50 longitudinally variable (b CV > $5\%$) or stable (c CV < $5\%$) plasma proteins. d Top panel: Number of proteins with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$z > 2.5$$\end{document}z>2.5 (red) or \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$z < -2.5$$\end{document}z<−2.5 (blue) in individual samples, where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$z=({NPX}-\overline{{NPX}})/{SD}$$\end{document}z=(NPX−NPX¯)/SD with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{{NPX}}$$\end{document}NPX¯ and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${SD}$$\end{document}SD being the mean and the standard deviation, respectively, of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${NPX}$$\end{document}NPX across samples of the same participant. Bottom panel: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-{\log }_{10}({p})$$\end{document}−log10(p) for individual samples being possible outliers, where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${p}$$\end{document}p is calculated based on a binomial test (two-sided). e Protein examples clearly demonstrate that Week 6 of participant PTID3 was an outlier. b, c, e Each boxplot displays the median (centerline), the first and third quartiles (the lower and upper bound of the box), and the 1.5x interquartile range (whiskers) of the data. Adobe Illustrator (version 27.1.1; https://www.adobe.com/products/illustrator.html) was used to arrange panels and edit text. Source data are provided as a Source Data file. ## Application of ODA to discover a possible abnormal event We noticed that proteomics data of donor PTID3 exhibited higher CV values than those of other donors (Fig. 3a) and weaker intra-donor correlations at week 6 than at other weeks (Supplementary Fig. 7b). We applied ODA to check whether donor PTID3 had an abnormal event at week 6. We selected \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left|z\right| > 2.5$$\end{document}z>2.5 as the criterion for outliers so that just above $1\%$ of all quantifiable proteins are expected to be outliers. More accurately, we expected $1.24\%$ of proteins, i.e., 19 proteins per donor per time point, to be outliers by chance (Methods). A total of 71 outlier proteins were identified at Week 6 on donor PTID3 (adjusted $$p \leq 2.7$$ × 10−26, Fig. 3d, e, Supplementary Data 2e, f). Eight of the top ten proteins having the highest z scores (2.84–2.85) play important roles in immune response and immunity (Supplementary Data 2g). Gene set enrichment analysis (GSEA) revealed the outlier proteins were enriched in immunological processes such as adaptive immune responses, antigen processing and presentation via major histocompatibility complex (MHC) class II, T cell activation, etc. ( Supplementary Fig. 7c). Single-sample GSEA (ssGSEA)27 on all PTID3 samples identified Week 6 as an outlier and revealed increased activity at Week 6 in important immune processes (Supplementary Fig. 7d), including MYC targets (v1 and v2)28, interferon-alpha and gamma responses29, androgen response30, pancreas beta cells31, and peroxisome32. Although further validation is required, these results suggest the possible occurrence of an immunological perturbation event (such as infection) experienced by PTID3 at week 6. Such outlier phenotypes can be obscured by analyses focusing on differences between sample groups. ## Application of SPECT to reveal diverse gene stability patterns We applied SPECT to analyze our scRNA-seq data. Noticing the two well-known housekeeping genes, ACTB and GAPDH, had CVs (across timepoints) just above $10\%$ in some cell types (Supplementary Fig. 8), we used a CV cut-off of $10\%$ to separate longitudinally variable (CV > $10\%$) or stable (CV < $10\%$) genes in individual cell types of individual donors. We then counted how many times individual genes were variable and/or stable in the 76 combinations between donor ($$n = 4$$) and cell type ($$n = 19$$). A gene was denoted as super variable (SUV) or super stable (SUS) if it was variable or stable in at least 40 donor-cell type combinations. A gene was denoted as variable across time in cell-types (VATIC) or stable across time in cell-types (STATIC) if it was variable or stable in at least one cell type across all donors but in less than 40 donor-cell type combinations. We identified a total of 700 SUV genes (Supplementary Fig. 9a), 2129 SUS genes (Supplementary Fig. 9b), 5750 VATIC genes, and 4004 STATIC genes from the dataset. Since a gene can be consistently variable in one cell type and consistently stable in another, VATIC and STATIC genes are not mutually exclusive (Supplementary Fig. 9c). The SUV genes were enriched in 57 pathways, many of which are associated with cellular proliferation and activity (Supplementary Data 3g). Eight of the top ten SUV genes (Supplementary Data 3h) have distinct roles in gene regulation, including four transcription factors (FOS, FOSB, JUN, and KLF9), two phosphatases (DUSP1 and PPP1R15A), one regulator of mTOR pathway (DDIT4)33, and one inhibitor of NF-κB pathway (TNFAIP3)34. In comparison, the SUS genes were enriched in 501 pathways of rather diverse, basic cellular processes (Supplementary Data 3i). Among the top ten SUS genes (Supplementary Data 3j), five (RPS12, RPL10, RPL13, RPLP1, and RPL41) encode ribosomal proteins and two (FTL and FTH1) encode ferritin for iron storage. Many SUS genes are more stable than ACTB and GAPDH and may be good candidates for estimating batch effects in scRNA-seq data35. ## STATIC genes as potential biomarkers for cell types or biological conditions We collected up to 25 top STATIC genes from each cell type and obtained 220 unique genes (Fig. 4a, Supplementary Data 5a). These 220 STATIC genes are enriched in pathways such as innate (adjusted $$p \leq 1.43$$ × 10−9) and adaptive (adjusted $$p \leq 1.33$$ × 10−9) immune response, allograft rejection (adjusted $$p \leq 3.06$$ × 10−16), lymphocyte mediated immunity (adjusted $$p \leq 3.72$$ × 10−8), myeloid mediated immunity (adjusted $$p \leq 2.71$$ × 10−5), B/T-cell proliferation (adjusted $p \leq 1.46$ × 10−3), acute inflammatory response (adjusted $$p \leq 7.48$$ × 10−3), hematopoietic cell lineage (adjusted $$p \leq 2.44$$ × 10−4), etc. ( Supplementary Data 5b). Examples of top STATIC genes for major cell types were shown in Fig. 4b, including: GIMAP7, LEF1, CD27, CCR7, and TSHZ2 for T cells; CD79A, MS4A1, TCL1A, CD79B, and TNFRSF13C for B cells; PRF1, FGFBP2, SPON2, CST7, and KLRD1 for natural killer (NK) cells; CD14, FCN1, MNDA, SEPINA1, and SPI1 for monocytes; and LILRA4, IRF7, FCER1A, SERPINF1, and SPIB for dendritic cells (DCs). All these genes demonstrated cell type-specific stability patterns and have well-documented roles in the corresponding cell types (Supplementary Data 5c).Fig. 4Properties of 220 STATIC genes of PBMC.a Heatmap of coefficient of variation (CV) evaluated on 93 out of the 220 stable across time in cell-types (STATIC) genes that were identified from 19 cell types in the longitudinal scRNA-seq data of $$n = 4$$ healthy participants. The CVs for each of the $$n = 4$$ participants were evaluated based on pseudo-bulk intensities in the corresponding 6 independent peripheral blood mononuclear cell (PBMC) samples. The 93 STATIC genes include up to ten top STATIC genes from individual cell types. b Circos plots displaying CV of five example STATIC genes identified from each of five major cell types: T cells, B cells, natural killer (NK) cells, monocytes (Mono), and dendritic cells (DCs). c Uniform Manifold Approximation and Projection (UMAP) using only the 220 STATIC genes as input features (sUMAP) on the same longitudinal scRNA-seq data. d–f sUMAP using the same 220 STATIC genes on three external PBMC datasets ((d) CNP00011023, (e) GSE1496892, (f) GSE16437816), where cells are labeled as in the original studies. g Distributions of Pearson correlation coefficient between gene expression (pseudo-bulk intensity) in scRNA-seq data and gene score in scATAC-seq data, one for the 220 STATIC genes (median correlation 0.70), one for the top 500 highly variable genes (HVGs, median correlation 0.40), one for the 10,608 reliable genes (average expression ≥0.1, median correlation 0.21), and one for random gene pairs ($95\%$ upper confidence bound at 0.399). The correlations were calculated across 14 cell types in 18 PBMC samples ($$n = 252$$ data points). h, i Venn diagrams showing the overlaps between the 220 STATIC genes and biomarkers distinguishing either healthy controls (Normal) versus participants infected with influenza (FLU, left panel) or *Normal versus* participants infected with SARS-CoV-2 (COVID19, right panel). The biomarkers were identified from either (h) CNP00011023 or (i) GSE1496892. Adobe Illustrator (version 27.1.1; https://www.adobe.com/products/illustrator.html) was used to arrange panels and edit text. Source data are provided as a Source Data file. We used the 220 STATIC genes as input features and projected PBMCs in our scRNA-seq data onto a two-dimensional Uniform Manifold Approximation and Projection11 (UMAP; Fig. 4c), which we refer to as sUMAP from now on. We also generated sUMAPs using the same 220 STATIC genes on three independent scRNA-seq datasets2,3,16 of PBMCs (Fig. 4d–f) in which cells were labeled as in the original studies. In all four cases, the 220 STATIC genes separated major cell types and most of their subtypes very well, suggesting that some STATIC genes are potentially good markers for cell types. Gene scores are routinely computed from scATAC-seq data to infer expression of the corresponding genes and used to label cells in scATAC-seq data based on a scRNA-seq reference21. We calculated the Pearson correlation between expression in scRNA-seq data and gene score in scATAC-seq data of the same genes across cell types and samples. Due to data sparseness, incomplete reference assembly, non-coding RNAs, and uncharacterized sequences, Pearson correlation could be calculated only on 10,608 ($94.7\%$) of the 11,191 reliable genes (Fig. 4g). Interestingly, among genes with strong correlations (Supplementary Fig. 10), the correlation was mainly influenced by differences between cell types, which partially justifies the use of gene score for cell labeling on scATAC-seq data. Within individual cell types, the correlation however appeared to be poor across different samples, likely reflecting the complexity of gene regulation. Pearson correlation was obtained on 206 ($93.6\%$) of the 220 STATIC genes with a median value of 0.70. In comparison, Pearson correlation was obtained on 403 ($80\%$) of the top 500 highly variable genes (HVGs), which are widely used in dimension reduction on scRNA-seq data11, with a significantly lower median value of 0.40 ($$p \leq 1.98$$ × 10−13, Mann–Whiney test; Supplementary Data 5d). We randomly paired unrelated genes, calculated the corresponding correlations between expression and gene score, and found that the obtained distribution had a $95\%$ upper confidence bound at R0 = 0.399 (Methods). Thus, any correlations below R0 were not statistically better than those between random, unrelated gene pairs. A total of 7252 ($68\%$) out of the 10,608 reliable genes and 201 ($49.8\%$) out of the 403 HVGs had a correlation below R0, in comparison with 40 ($19\%$) out of the 206 STATIC genes. To properly label cells in scATAC-seq data based on gene score approach, one should only use genes whose expression versus gene score correlations are above R0. Some STATIC genes might be good candidates for this purpose. We further investigated how the 220 STATIC genes fared as potential disease biomarkers. Previously, two studies2,3 applied scRNA-seq to analyze PBMCs of healthy controls (Normal) and of patients infected with either influenza (FLU) or SARS-CoV-2 (COVID19). We reanalyzed the data using methods described in the original studies and identified differential expression genes (DEGs) distinguishing *Normal versus* FLU or *Normal versus* COVID19. For simplicity, DEGs from individual cell types were combined when compared with the 220 STATIC genes. Out of the 18,824 genes measured in the first study (CNP0001102)3, 681 and 632 DEGs were identified for distinguishing *Normal versus* FLU and *Normal versus* COVID19, respectively. The corresponding overlap with the STATIC genes was 49 for *Normal versus* FLU (one-side hypergeometric $$p \leq 4.8$$ × 10−26) and 50 for *Normal versus* COVID19 (one-side hypergeometric $$p \leq 1.7$$ × 10−28, Fig. 4h). A total of 33,538 genes were measured in the second study (GSE149689)2. A total of 126 STATIC genes (one-side hypergeometric $$p \leq 4.8$$ × 10−74) overlapped with the 3040 DEGs for *Normal versus* FLU while 86 STATIC genes (one-side hypergeometric $$p \leq 2.1$$ × 10−61) overlapped with the 1396 DEGs for *Normal versus* COVID19 (Fig. 4i). In all cases, the 220 STATIC genes were significantly enriched as DEGs, suggesting their potential for monitoring some disease conditions. To illustrate that SPECT can handle scRNA-seq data of diverse sample types, we applied it to scRNA-seq data from a mouse brain study (GSE129788)36. In the study scRNA-seq data was collected from brain tissues of eight young (2–3 months) and eight old (21–23 months) mice, from which 37,069 cells of high-quality data were labeled to 25 cell types, 14,699 genes were detected, marker genes for each of the 25 cell types were collected, and 1113 DEGs distinguishing young versus old mouse brains were identified from a subset of 15 cell types. The study was not longitudinal per se. We treated data from the eight samples of each age group as repeated measurements for the group, just like repeated measurements at different timepoints in a longitudinal study. Since SPECT does not utilize the ordering of timepoints, its usage to the data is justified. We collected up to 25 STATIC genes per cell type and obtained 304 unique genes from all 25 cell types (Fig. 5a, Supplementary Data 6a). sUMAP using these 304 STATIC genes was able to separate the cell types as labeled in the original study very well (Fig. 5b). Out of the 304 STATIC genes, 299 genes were identified in the original study as marker genes for the corresponding cell types (Fig. 5c, Supplementary Data 6b). From the 15 cell types having DEGs, we collected 234 STATIC genes that were significantly overlapped with the 1113 young versus old DEGs ($$n = 123$$, one-side hypergeometric $$p \leq 6.2$$ × 10−77, Fig. 5d). These results further demonstrated that some STATIC genes are good markers for cell types or biological conditions in the mouse brain study. Fig. 5Properties of 304 STATIC genes of mouse brain tissue.a Heatmap of coefficient of variation (CV) of the 304 stable across time in cell-types (STATIC) genes that were identified from 25 cell types in the scRNA-seq data of a mouse brain study (GSE12978836). The CVs were evaluated based on pseudo-bulk intensities in brain tissues from either $$n = 8$$ young or $$n = 8$$ old mice. b Uniform Manifold Approximation and Projection (UMAP) using only the 304 STATIC genes as input features (sUMAP) on the same scRNA-seq data. Cells are labeled as in the original study. c Percentage of top STATIC genes that overlap with cell-type marker genes identified in the original study. Up to 25 top STATIC genes from each cell type are compared with the corresponding marker genes of the same cell type. d Venn diagram showing the overlap between the 234 STATIC genes identified from 15 out of the 25 cell types and biomarkers distinguishing young versus old mice that were identified in the original study from the same 15 cell types. Adobe Illustrator (version 27.1.1; https://www.adobe.com/products/illustrator.html) was used to arrange panels and edit text. Source data are provided as a Source Data file. ## Circos plots to reveal stability patterns of protein families PALMO implements circos plots to display stability patterns from multiple single-cell data modalities together. We displayed the stability pattern of gene expression and gene score of six protein families that are essential for immunity in Fig. 6, including human leukocyte antigens (HLAs, Fig. 6a), interferon regulatory factors (IRFs, Fig. 6b), interleukins (ILs, Fig. 6c), chemokine (C-X-C motif) receptor/ligand (CXCR/L) family (Fig. 6d), Janus kinases (JAKs) and signal transducer and activator of transcription proteins (STATs, Fig. 6e), and tumor necrosis factor receptor superfamily (TNFRSF, Fig. 6f). All these protein families showed diverse stability patterns among members and across cell types, with HLAs and ILs having the most striking contrasts. The rich variety in such stability patterns suggests that different members of same protein superfamilies may play different roles in individual cell types. We noticed that gene expression and gene score generally did not exhibit the same stability patterns despite the rather strong correlations between them (Supplementary Fig. 11). It turns out that strong correlations were mainly driven by difference between cell types rather than difference between samples, likely reflecting the complexity of gene regulation as mentioned before. Fig. 6Circos plots showing stability patterns of five protein families.a Circos plot displaying stability patterns of gene expression (outer circles) and gene score (inner circles) of human leukocyte antigen (HLA) protein family (member: HLA-A, HLA-B, HLA-C, HLA-DRA, HLA-DPA1, and HLA-DRB1). Samples with missing data or cell types with low cell counts are shown in grey. b–f Same as (a) but for (b) interferon regulatory factors (IRFs; member: IRF1, IRF2, IRF3, IRF4, IRF5, and IRF8), (c) interleukins (ILs; member: IL32, IL7R, IL10RA, IL2RB, IL1B and IL18), (d) chemokine (C-X-C motif) receptor/ligand (CXCR/L) protein family (member: CXCR4, CXCR5, CXCR6, CXCL8, CXCL10, and CXCL16), (e) Janus kinase (JAK) and signal transducer and activator of transcription (STAT) protein family (member: JAK1, JAK2, JAK3, STAT3, STAT4, and STAT6), and (f) tumor necrosis factor receptor superfamily (TNFRSF; member: TNFRSF1B, TNFRSF13C, TNFRSF10B, TNFRSF25, TNFRSF11A, and TNFRSF17). The CV of gene expression for each of $$n = 4$$ participants was calculated from pseudo-bulk intensities in the corresponding 6 independent peripheral blood mononuclear cell (PBMC) samples. The CV of gene score for each participant was based on either 6 (for $$n = 2$$ participants) or 3 (for other $$n = 2$$ participants) PBMC samples. Adobe Illustrator (version 27.1.1; https://www.adobe.com/products/illustrator.html) was used to arrange panels and edit text. Source data are provided as a Source Data file. ## Application of TCA to reveal heterogenous immune responses among COVID-19 patients We applied TCA to analyze longitudinal scRNA-seq data of four COVID-19 patients, each having data of at least three timepoints, in a previous study3 and identified significantly up- or down-regulated genes over time (adjusted $p \leq 0.05$ and slope magnitude >0.1, Fig. 7a–d, Supplementary Data 7a) and the corresponding pathways (Supplementary Data 7b). We observed rather heterogeneous immune responses by these patients during recovery (Fig. 7e), which was not presented in the original study. Fig. 7Heterogeneous immune responses by COVID19 patients during recovery.a Volcano plot showing temporal expression changes of individual genes in different cell types during the recovery of patient COV-3 (female, 41 years old, mild symptoms, data on day D1/D4/D16), based on longitudinal scRNA-seq data in CNP00011023. The x-axis shows the slope (coefficient) of gene expression change as a linear function of time. The y-axis shows the corresponding adjusted p value of the slope. b–d Same as (a) but for patients (b) COV-2 (male, 45 years old, mild symptoms, data on D1/D4/D7/D10/D16), (c) COV-1 (male, 15 years old, mild symptoms, data on D1/D4/D16), and (d) COV−5 (female, 85 years old, severe symptoms, data on D1/D7/D13). a–d Each plot contains results on up to 18,824 genes in 13 cell types (up to 244,712 data points). e Counts of significantly upregulated (adjusted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$$p \leq 0.05$$$\end{document}$p \leq 0.05$ and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${slope} > 0.1$$\end{document}slope>0.1, red) and significantly downregulated (adjusted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$$p \leq 0.05$$$\end{document}$p \leq 0.05$ and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${slope} < -0.1$$\end{document}slope<−0.1, blue) genes during the recovery of the four COVID-19 patients in individual cell types. a–d The p-value for slope was calculated based on two-sided likelihood-ratio test and adjusted by Benjamini and Hochberg procedure for testing many genes. Adobe Illustrator (version 27.1.1; https://www.adobe.com/products/illustrator.html) was used to arrange panels and edit text. Source data are provided as a Source Data file. Patient COV-3 had barely any significant genes except that IFI27 decreased in DCs, IFI44L decreased in naïve B cells, and IGLC3 decreased in plasma cells, suggesting possible dampening of immune modulation. The significant genes of patient COV-2 included eighteen upregulated genes in monocytes, four genes each in memory B cells and naïve B cells, and twelve genes split among other six cell types. Gene enrichment analysis on the eighteen upregulated genes in monocytes revealed only one significant pathway: myeloid leukocyte mediated immunity (adjusted $$p \leq 0.044$$). The significant genes of COV-1 included eleven upregulated and six downregulated genes in cycling plasma cells, seven upregulated and sixteen downregulated genes in cycling T cells, six downregulated genes in naïve B cells, and fifteen genes split among other seven cell types. The significant genes in cycling plasma cells were significantly enriched in five pathways including regulation of humoral immune response (adjusted $$p \leq 3.92$$ × 10−3), Fc receptor mediated stimulatory signaling pathway (adjusted $$p \leq 3.92$$ × 10−3), and immunoglobulin production (adjusted $$p \leq 0.011$$), indicating a predominant role of humoral immunity in the recovery of the patient. Patient COV-5 had significant genes in almost all cell types except for DCs and monocytes, including eight upregulated and eight downregulated genes in memory B cells, six upregulated and six downregulated genes in naïve B cells, one upregulated and ten downregulated genes in activated CD4+ T cells, two upregulated and eight downregulated genes in plasma cells, and 43 genes split among other seven cell types. Seven ($58\%$) of the twelve significant genes in naïve B cells were also significant in memory B cells and in the same direction of change, suggesting common responses by the two cell types. The significant genes in memory B cells were enriched in interferon gamma (adjusted $$p \leq 3.28$$ × 10−6) and alpha (adjusted $$p \leq 4.86$$ × 10−5) response, antigen processing and presentation (adjusted $$p \leq 0.036$$), and antigen processing and presentation of peptide or polysaccharide antigen via MHC class II (adjusted $$p \leq 0.044$$). The significant genes in naïve B cells were enriched in interferon alpha (adjusted $$p \leq 1.96$$ × 10−5) and gamma (adjusted $$p \leq 1.96$$ × 10−5) response. The significant genes in plasma cells were enriched in innate and humoral immune responses ($$p \leq 3.46$$ × 10−4 and $$p \leq 5.79$$ × 10−4, respectively) although both with an adjusted $$p \leq 0.084.$$ These results aligned to the patient’s disease severity and advanced age. For comparison, we also used Seurat to analyze patient COV-5 data of activated CD4+ T cells. To satisfy Seurat’s requirement of selecting two contrast groups, we did the analysis in two iterations, i.e., day 1 (D1) versus D7 + D13 and D1 + D7 versus D13. We obtained 942 and 1018 DEGs (adjusted $p \leq 0.05$), respectively, with an overlap of 813 DEGs (Supplementary Fig. 12a). TCA identified 921 significantly up- or down-regulated genes (adjusted $p \leq 0.05$), only 21 of which overlapped with both Seurat results. *The* genes obtained from TCA or Seurat were quite different. We collected top ten up- and top ten down-regulated genes from all three approaches and plotted the corresponding gene expression in heatmaps (Supplementary Fig. 12b–d). TCA results showed better dynamic changes over time than Seurat results in our opinion. ## Discussion The five modules in PALMO analyze longitudinal omics data from multiple perspectives as continuous data. VDA provides a global view on the sources of variance within the whole dataset. TCA studies the time series of individual participants. CVP and SPECT first examine data of individual participants separately and then summarize the observations across different participants. All these four methods focus on individual features. ODA is the only method to provide a sample-level analysis. Which module(s) to use on a specific dataset depends on the research question of interest. Additional methods need to be developed for research interest not covered here. We observed that a small set of STATIC genes, 220 for PBMC and 304 for mouse brain tissues, distinguished cell types well and captured some biological differences. The PBMC STATIC genes showed better correlation between gene expression in scRNA-seq data and gene score in scATAC-seq data than HVGs. It would be interesting to see whether these observations can be extended to scRNA-seq data of other sample types. We selected up to 25 STATIC genes per cell type in our analysis. It is possible that a better set of genes can be selected with a more sophisticated selection procedure. Plasma proteins are often targeted as disease biomarkers, thus understanding their temporal stability is of particular interest. Conceptually, highly variable proteins are poor biomarker candidates since their values likely have very high sampling variations. The rather moderate CV values of the most variable proteins in our study suggest sampling variations are not a big concern on these proteins. The small CV values of the most stable proteins, on the other hand, indicate they do not change much under normal, healthy conditions. So, if they ever change under some disease conditions, they should be closely explored as potential biomarkers. We condensed single-cell data into pseudo-bulk data in VDA, SPECT and ODA. Recent literature14,17,18 revealed that many single-cell methods fail to properly account for variations in cross-sectional scRNA-seq data and generate many false DEGs as a result. In comparison, pseudo-bulk approaches mostly generate reliable results although they may be underpowered. Longitudinal single-cell omics data is even more complicated than cross-sectional scRNA-seq data and may require new statistical methods to properly handle its many types of variations. Furthermore, memory and CPU requirements for using GLMMs to analyze longitudinal single-cell omics data at single-cell level may be challenging even for cloud-based computing. We adopted the pseudo-bulk approach in VDA, SPECT and ODA as a practical compromise. In TCA we bypassed some of the complications by analyzing data of individual cell types and of individual participants separately. The lack of a well-accepted software package for longitudinal omics data makes it difficult to benchmark PALMO performance. We compared PALMO with variancePartition19, tcR20, and Seurat16, which is summarized in Supplementary Fig. 1c. VDA can handle missing data but variancePartition cannot, which is an advantage of VDA since missing values in longitudinal omics data are almost inevitable. The two tools generated almost identical results on two tested datasets after removing missing values. PALMO was not developed specifically for TCR data. When we applied VDA to the TCR data of SSc donors, we obtained results that are potentially interesting but not reported in the original study using tcR. We believe PALMO complements TCR specific tools (such as tcR) on TCR data. Seurat requires users to select two contrast groups in DEG analysis and thus is not appropriate for analyzing longitudinal data of more than two timepoints. Nevertheless, when we applied both TCA and Seurat to the longitudinal scRNA-seq data of activated CD4+ T cells of a COVID-19 patient, the two methods generated rather different results on up- or down-regulated genes. Heatmaps of the corresponding top genes revealed that TCA results showed better dynamic changes over time than Seurat results. PALMO has been published as an R package in CRAN with a detailed reference manual and vignettes to demonstrate its usage. It can be easily installed and executed in R or RStudio. As we demonstrated, it can be used to analyze longitudinal bulk and single-cell omics data generated on diverse technical platforms and/or of diverse sample types, including but not limited to: clinical lab test results, cell type composition, gene expression, protein abundance, bulk or single-cell omics data, TCR sequencing data, etc. We believe it can facilitate the analysis of some longitudinal omics data. In addition, our longitudinal multi-omics dataset of five data modalities on the same samples can also be a valuable resource for immune health study and software development. ## Healthy donors Blood samples were obtained from Bloodworks Northwest (Seattle, WA) through protocols approved by the Bloodworks Northwest institutional review board and complying with all relevant ethical regulations. We enrolled $$n = 6$$ clinically healthy participants (no diagnosis of active or chronic disease) with age between 25 to 38 years with equal self-report sex ratio. Viable peripheral blood mononuclear cells (PBMCs) and plasma samples were collected from each participant over $t = 10$ weeks. Complete blood count (CBC) was measured to evaluate overall health of all donors over all timepoints ($$n = 6$$, $t = 10$). Minimal biometric data were collected on these participants which were handled following the Health Insurance Portability and Accountability Act (HIPAA) guidelines. Informed consent to participate in the study and to publish data from the research was obtained from all participants. ## Sample handling A volume of 30 mL of blood was drawn into BD NaHeparin vacutainer tubes (for PBMC; BD #367874) or K2-EDTA vacutainer tubes (for plasma; BD #367863). Upon arrival at the processing lab all NaHeparin tubes for each donor were pooled into a sterile plastic receptacle to establish one common pool and stored at room temperature until processing (4 h or less from draw). For PBMC isolation, at each time point the pool of blood was gently swirled until fully mixed, about 30 times, and a volume of blood was removed and combined with an equivalent volume of room temperature PBS (ThermoFisher #14190235). PBMC were isolated using one or more Leucosep tubes (Greiner Bio-One #227290) loaded with 15 mL of Ficoll Premium (GE Healthcare #17-5442-03) to which a 3 mL cushion of PBS had been slowly added on top of the Leucosep barrier. The 24–30 mL diluted whole blood was slowly added to the tube and spun at 1000xg for 10 min at 20 °C with no brake. PBMC were recovered from the Leucosep tube by quickly pouring all volume above the barrier into a sterile 50 mL conical tube; 15 mL cold PBS + $0.2\%$ BSA (Sigma #A9576; “PBS + BSA”) was added, and the cells were pelleted at 400xg for 5–10 min at 4–10 °C. The supernatant was quickly decanted, the pellet dispersed by flicking the tube, and the cells washed with 25–50 mL cold PBS + BSA. Cell pellets were combined, if applicable, the cells were pelleted as before, supernatant quickly decanted, and residual volume was carefully aspirated. The PBMC were resuspended in 1 mL cold PBS + BSA per 15 mL whole blood processed and counted with a Cellometer Spectrum (Nexcelom) using Acridine Orange/Propidium Iodide solution. PBMC were cryopreserved $90\%$ FBS (ThermoFisher #10438026)/$10\%$ DMSO (Fisher Scientific #D12345) at 1–5 × 106 cells/mL by slow freezing in a Coolcell LX (VWR #75779-720) overnight in a −80 °C freezer followed by transfer to liquid nitrogen. For plasma isolation, the K2-EDTA source tube was gently inverted 10 times, and the appropriate volume of whole blood was extracted using an 18-gauge needle and syringe and transferred to a similar plastic tube with no additives (Greiner Bio-One #456085). The blood was centrifuged at 2000xg for 15 min at 20 °C with a brake of 1, and $80\%$–$90\%$ of the plasma supernatant were removed by careful pipetting for immediate freezing at −80 °C. Plasma was assayed after the first freeze/thaw. Thawed PBMC of four donors over six timepoints ($$n = 4$$, $t = 6$) were assayed by flow cytometry, scRNA-seq and scATAC-seq in two batches (donors PTID5 and PTID6, donors PTID2 and PTID4) by a team of operators. Plasma of all donors over all timepoints ($$n = 6$$, $t = 10$) was isolated and cryopreserved by a team of operators37. ## Flow cytometry PBMC were removed from liquid nitrogen storage and immediately thawed in a 37 °C water bath. Cells were diluted dropwise into 37 °C AIM V media (Thermo Fisher Scientific #12055091) up to a final volume of 10 mL. A single wash was performed in 10 mL of PBS + BSA, pelleting cells at 400xg for 5–10 min at 4–10 °C. PBMC were resuspended 2 mL in PBS + BSA and counted using a Cellometer Spectrum. 1–2 × 106 cells were incubated with Human TruStain FcX (BioLegend #422302) and Fixable Viability Stain 510 (BD #564406) prior to staining with a 25-color cell surface panel on ice for 25 min. Cells were washed and fixed with $4\%$ paraformaldehyde (Electron Microscopy Sciences #15713) prior to acquisition on a BD Symphony cytometer. Raw data were compensated and curated to remove unrepresentative events due to instrument fluidics variability (time gating), doublets (by FSC-H and FSC-W), and cells exhibiting membrane permeability (live/dead gating) prior to quantification using BD FlowJo software v10.6.1. ## Proteomics Plasma samples were submitted to Olink (Uppsala, Sweden) for assay using the Olink Proximity Extension assay, run on the Fluidigm Biomark system. Patient samples were distributed evenly across two plates, and all time points per patient were run on the same plate, with randomized well locations. Samples were assayed using the Olink Discovery Assay which encompasses a total of 1156 proteins across 13 panels (Cardiometabolic [V.3603], Cardiovascular II [V.5006], Cardiovascular III [V.6113], Cell Regulation [V.3701], Development [V.3512], Immune Response [V.3202], Inflammation [V.3021], Metabolism [V.3402], Neuro Exploratory [V.3901], Neurology [V.8012], Oncology II [V.7004], Oncology III [V.4001], Organ Damage [V.3311]). Quality assessment, limit of detection, and normalization were performed by Olink using the plate bridging control, two positive controls, and three background controls. ## Sample preparation, hashing, and pooling Single-cell RNA-seq libraries were generated on PBMC prepared as above using the 10x Genomics Chromium 3’ Single Cell Gene Expression assay (#1000121) and Chromium Controller Instrument according to the manufacturer’s published protocol with modifications for cell hashing38. To block off-target antibody binding, Blocking Solution (5 µL of Human Trustain FcX (BioLegend #422302), and 13.7 µL of a $10\%$ Bovine Serum Albumin (BSA)) was added to 500,000 cells suspended in 50 µL Dulbecco’s Phosphate Buffered Saline (DPBS; Corning Life Sciences #21-031-CM) and incubated for 10 min on ice. To stain samples, 0.5 µg (1 µL) of a TotalSeq™-A anti-human Hashtag Antibody was suspended in 31.3 µL DPBS/$2\%$ BSA, then added to each sample. For each batch of samples, 100,000 cells from 12 hashed samples with a distinct Hashtag Antibody were pooled into the hashed pool. Roughly 20,000 cells from a Leukopak healthy control were also labeled with a distinct TotalSeq™-A Hashtag Antibody and were spiked into each pool to serve as a batch control. ## Droplet encapsulation and reverse transcription From each pool, 64,000 cells were loaded into each well of a Chromium Single Cell Chip G (10x Genomics #1000073) (8 wells per chip), targeting a recovery of 20,000 singlets from each well. Gel Beads-in-emulsion (GEMs) were then generated using the 10x Chromium Controller. The resulting GEM generation products were then transferred to semi-skirted 96-well plates and reverse transcribed on a C1000 Touch Thermal Cycler (Bio-Rad) programmed at 53 °C for 45 min, 85 °C for 5 min, and a hold at 4 °C. Following reverse transcription, GEMs were broken, and the pooled single-stranded cDNA and Hashtag Oligo fractions were recovered using Silane magnetic beads (Dynabeads MyOne SILANE #37002D). ## Library generation and separation Barcoded, full-length cDNA including the Hashtag Oligos (HTOs) from the TotalSeq™-A Hashtag Antibodies were then amplified with a C1000 Touch Thermal Cycler programmed at 98 °C for 3 min, 11 cycles of (98 °C for 15 s, 63 °C for 20 s, 72 °C for 1 min), 72 °C for 1 min, and a hold at 4 °C. Amplified cDNA was purified and separated from amplified HTOs using a 0.6x size selection via SPRIselect magnetic bead (Beckman Coulter #22667) and a 1:10 dilution of the resulting cDNA was run on a Fragment Analyzer (Agilent Technologies #5067-4626) to assess cDNA quality and yield. HTO libraries were purified further with SPRIselect magnetic bead (Beckman Coulter #22667) and amplified and indexed with a custom HTO i7 index on a C1000 Touch Thermal Cycler programmed at 95 °C for 3 min, 10 cycles of (95 °C for 20 s, 64 °C for 30 s, 72 °C for 20 s), 72 °C for 1 min, and a hold at 4 °C. The resulting HTO libraries were purified with SPRIselect magnetic bead (Beckman Coulter #22667) post-amplification and a 1:10 dilution of the resulting HTO libraries were run on a Fragment Analyzer (Agilent Technologies #5067-4626) to assess HTO quality and yield. A quarter of the cDNA sample (10 ul) was used as input for library preparation. Amplified cDNA was fragmented, end-repaired, and A-tailed is a single incubation protocol on a C1000 Touch Thermal Cycler programmed at 4 °C start, 32 °C for min, 65 °C for 30 min, and a 4 °C hold. Fragmented and A-tailed cDNA was purified by performing a dual-sided size-selection using SPRIselect magnetic beads (Beckman Coulter #22667). A partial TruSeq Read 2 primer sequence was ligated to the fragmented and A-tailed end of cDNA molecules via an incubation of 20 °C for 15 min on a C1000 Touch Thermal Cycler. The ligation reaction was then cleaned using SPRIselect magnetic beads (Beckman Coulter #22667). PCR was then performed to amplify the library and add the P5 and indexed P7 ends (10x Genomics #1000084) on a C1000 Touch Thermal Cycler programmed at 98 °C for 45 sec, 13 cycles of (98 °C for 20 sec, 54 °C for 30 sec, 72 °C for 20 sec), 72 °C for 1 min, and a hold at 4 °C. PCR products were purified by performing a dual-sided size-selection using SPRIselect magnetic beads (Beckman Coulter #22667) to produce final, sequencing-ready libraries. ## Quantification and sequencing Final libraries were quantified using Picogreen and their quality was assessed via capillary electrophoresis using the Agilent Fragment Analyzer HS DNA fragment kit and/or Agilent Bioanalyzer High Sensitivity chips. Libraries were sequenced on the Illumina NovaSeq platform using S4 flow cells. Read lengths were 28 bp read1, 8 bp i7 index read, 91 bp read2. Final libraries were quantified using a Quant-iT PicoGreen dsDNA Assay Kit (Thermo Fisher Scientific, P7589) on a SpectraMax iD3 (Molecular Devices). Library quality and average fragment size was assessed using a Bioanalyzer (Agilent, G2939A) High Sensitivity DNA chip (Agilent, 5067-4626). Libraries were sequenced on the Illumina NovaSeq platform with the following read lengths: 51nt read 1, 8nt i7 index, 16nt i5 index, 51nt read 2. ## scRNA-seq data pre-processing scRNA-seq data of four donors were generated in two batches, each containing data of two donors. Each batch of data was pre-processed separately37. Briefly, binary base call (BCL) files were demultiplexed using the mkfastq function in the 10x Cell Ranger software (version 3.1.0), producing fastq files. Fastq files were then checked for quality (FastQC version 0.11.3) and run through the 10x Cell Ranger alignment function (cell ranger count) against the human reference annotation (Ensembl GRCh38). Mapping was performed using default parameters. Upon completion, Cell Ranger produced an output directory per file that contains the following: bam file (binary alignment file), HDF5 file (Hierarchical Data Format) with all reads, HDF file containing just the filtered reads, summary report (html and csv), and cloupe.cloupe (a file for the 10x Loupe visual browser). ## scRNA-seq data analysis Individual HDF5 files (filtered) were loaded into the R statistical programming language (version 3.6.0) using Bioconductor (version 3.1.0) and the Seurat package (version 3.1.5)37. For simplicity, sample names were captured as a list in R and iteratively processed within a loop (refer to https://satijalab.org/seurat/ for more information). Within the loop, samples were normalized with the NormalizeData function followed by the FindVariableFeatures function with parameters: vst selection method and 2000 features. Label transfer was performed using previously published procedures39 and with the Seurat reference dataset. Labeling included the FindTransferAnchors and TransferData functions performed in the Seurat package. We merged the two batches of data using the Seurat merge function. We calculated read depth, mitochondrial percentage, and number of UMIs per sample. Cells were filtered with nFeature_RNA > 200 and percent.mt <10. The merged data structure was normalized (using NormalizeData and FindVariableFeatures functions) and then saved as an RDS for further analysis. The top 3000 variable genes were used for PCA and UMAP based dimension-reduction maps using 30 principal components (PCs). We checked for possible batch effects using the bridging controls but did not observe any obvious batch effects. Cell labels obtained from the original batches were kept. Doublets were removed from further analysis. In total we retrieved high quality data of 472,464 cells from 4 donors and labeled them to 31 cell types from Seurat level 2 labelling. The cell type frequencies in each sample were calculated and compared with flow-based cell frequencies. Nineteen cell types (CD4_Naive, CD4_TEM, CD4_TCM, CD4_CTL, CD8_Naive, CD8_TEM, CD8_TCM, Treg, MAIT, gdT, NK, NK_CD56bright, B_naive, B_memory, B_intermediate, CD14_Mono, CD16_Mono, cDC2, pDC) were selected for further analysis after filtering out cell types with a low frequency (<$0.5\%$). ## Sample preparation Permeabilized-cell scATAC-seq was performed. A $5\%$ w/v digitonin stock was prepared by diluting powdered digitonin (MP Biomedicals, 0215948082) in DMSO (Fisher Scientific, D12345), which was stored in 20 µL aliquots at −20 °C until use. To permeabilize, 1 × 106 cells were added to a 1.5 mL low binding tube (Eppendorf, 022431021) and centrifuged (400×g for 5 min at 4 °C) using a swinging bucket rotor (Beckman Coulter Avanti J-15RIVD with JS4.750 swinging bucket, B99516). Cells were resuspended in 100 µL cold isotonic Permeabilization Buffer (20 mM Tris-HCl pH 7.4, 150 mM NaCl, 3 mM MgCl2, $0.01\%$ digitonin) by pipette-mixing 10 times, then incubated on ice for 5 min, after which they were diluted with 1 mL of isotonic Wash Buffer (20 mM Tris-HCl pH 7.4, 150 mM NaCl, 3 mM MgCl2) by pipette-mixing five times. Cells were centrifuged (400 × g for 5 min at 4 °C) using a swinging bucket rotor, and the supernatant was slowly removed using a vacuum aspirator pipette. Cells were resuspended in a chilled TD1 buffer (Illumina, 15027866) by pipette-mixing to a target concentration of 2300–10,000 cells per µL. Cells were filtered through 35 µm Falcon Cell Strainers (Corning, 352235) before counting on a Cellometer Spectrum Cell Counter (Nexcelom) using ViaStain acridine orange/propidium iodide solution (Nexcelom, C52-0106-5). ## Tagmentation and fragment capture scATAC-seq libraries were prepared according to the Chromium Single Cell ATAC v1.1 Reagent Kits User Guide (CG000209 Rev B) with several modifications. 19,000 cells were loaded into each tagmentation reaction. Permeabilized cells were brought up to a volume of 12 µl in TD1 buffer (Illumina, 15027866) and mixed with 3 µl of Illumina TDE1 Tn5 transposase (Illumina, 15027916). Transposition was performed by incubating the prepared reactions on a C1000 Touch thermal cycler with 96–Deep Well Reaction Module (Bio-Rad, 1851197) at 37 °C for 60 min, followed by a brief hold at 4 °C. A Chromium NextGEM Chip H (10x Genomics, 2000180) was placed in a Chromium Next GEM Secondary Holder (10x Genomics, 3000332) and $50\%$ Glycerol (Teknova, G1798) was dispensed into all unused wells. A master mix composed of Barcoding Reagent B (10x Genomics, 2000194), Reducing Agent B (10x Genomics, 2000087), and Barcoding Enzyme (10x Genomics, 2000125) was then added to each sample well, pipette-mixed, and loaded into row 1 of the chip. Chromium Single Cell ATAC Gel Beads v1.1 (10x Genomics, 2000210) were vortexed for 30 s and loaded into row 2 of the chip, along with Partitioning Oil (10x Genomics, 2000190) in row 3. A 10x Gasket (10x Genomics, 370017) was placed over the chip and attached to the Secondary Holder. The chip was loaded into a Chromium Single Cell Controller instrument (10x Genomics, 120270) for GEM generation. At the completion of the run, GEMs were collected, and linear amplification was performed on a C1000 Touch thermal cycler with 96–Deep Well Reaction Module: 72 °C for 5 min, 98 °C for 30 sec, 12 cycles of: 98 °C for 10 sec, 59 °C for 30 sec and 72 °C for 1 min. ## Sequencing library preparation GEMs were separated into a biphasic mixture through addition of Recovery Agent (10x Genomics, 220016), the aqueous phase was retained and removed of barcoding reagents using Dynabead MyOne SILANE (10x Genomics, 2000048) and SPRIselect reagent (Beckman Coulter, B23318) bead clean-ups. Sequencing libraries were constructed by amplifying the barcoded ATAC fragments in a sample indexing PCR consisting of SI-PCR Primer B (10x Genomics, 2000128), Amp Mix (10x Genomics, 2000047) and Chromium i7 Sample Index Plate N, Set A (10x Genomics, 3000262) as described in the 10x scATAC User Guide. Amplification was performed in a C1000 Touch thermal cycler with 96–Deep Well Reaction Module: 98 °C for 45 sec, for 11 cycles of: 98 °C for 20 sec, 67 °C for 30 sec, 72 °C for 20 sec, with a final extension of 72 °C for 1 min. Final libraries were prepared using a dual-sided SPRIselect size-selection cleanup. SPRIselect beads were mixed with completed PCR reactions at a ratio of 0.4x bead:sample and incubated at room temperature to bind large DNA fragments. Reactions were incubated on a magnet, the supernatant was transferred and mixed with additional SPRIselect reagent to a final ratio of 1.2x bead:sample (ratio includes first SPRI addition) and incubated at room temperature to bind ATAC fragments. Reactions were incubated on a magnet, the supernatant containing unbound PCR primers and reagents was discarded, and DNA bound SPRI beads were washed twice with $80\%$ v/v ethanol. SPRI beads were resuspended in Buffer EB (Qiagen, 1014609), incubated on a magnet, and the supernatant was transferred resulting in final, sequencing-ready libraries. ## scATAC data pre-processing scATAC-seq data were available for donor PTID2 and PTID4 at week 2–7 (6 timepoints) and for PTID5 and PTID6 at week 2, 4, and 7. scATAC-seq libraries were processed. In brief, cellranger-atac mkfastq (10x Genomics v1.1.0) was used to demultiplex BCL files to FASTQ. FASTQ files were aligned to the human genome (10x Genomics refdata-cellranger-atac-GRCh38-1.1.0) using cellranger-atac count (10x Genomics v1.1.0) with default settings. scATAC fragments were submitted to the ArchR package to create the ArchR object21. Per-cell quality control (QC) was performed using methods as mentioned in ArchR. The QC analysis showed FRiP score (the fraction of reads that fall into a peak) >0.25. The TSS enrichment and log10(nFrags) data showed comparable range across all samples. Doublets were removed using filterDoublets() function. In total we observed 294,623 peaks in 135,566 cells. ## scATAC-seq data analysis Using plotEmbedding function in ArchR, embedded IterativeLSI was used to perform UMAP based dimension reduction. Unconstrained integration was used to align scATAC-seq gene score matrix in ArchR object with the corresponding scRNA-seq gene expression matrix, from which cells were labeled to 28 cell types along with labeling scores to measure the quality of the cell-label transfer. We filtered out low quality cells (labeling score <0.5), removed cell types having less than 50 remaining cells, and kept 14 (B_intermediate, B_naive, CD14_Mono, CD16_Mono, CD4_Naive, CD4_TCM, CD8_Naive, CD8_TEM, cDC2, gdT, MAIT, NK, NK_CD56bright, and pDC) out of the 28 cell types for downstream analysis. *The* gene score matrix was retrieved using the getGroupSE() function in ArchR21 and used for downstream analysis by PALMO. ## Reagents and resources Critical reagents and resources used in our experiments are listed in Supplementary Data 8. ## Overview The current version of PALMO contains five analytical modules to analyze longitudinal omics data from multiple perspectives. It treats longitudinal omics data as continuous variables. PALMO has been published as an R package in CRAN with a detailed reference manual and vignettes to demonstrate its usage (https://cran.r-project.org/web/packages/PALMO/index.html). It can be easily installed and executed in R or RStudio. ## PALMO S4 object PALMO is a R based package that uses the setClass function to create an S4 object oriented system. The S4 object consists of a list of data structures with different types of elements such as strings, numbers, vectors, embedded lists, etc. It stores input expression data, input metadata, and output results into separate data structures for easy retrieval and interpretation. More details can be found in Section 3.9 of PALMO vignettes (https://raw.githubusercontent.com/aifimmunology/PALMO/main/Vignette-PALMO.pdf). Function createPALMOobject() takes two inputs (anndata and data) to create an PALMO S4 object: anndata is a data frame containing sample annotations. For longitudinal bulk data, data is a data frame with features (such as genes or proteins) as rows, samples as columns, and expression values as elements. For longitudinal single-cell omics data, data is a Seurat object. For single-cell omics data without a Seurat object, function createPALMOfromsinglecellmatrix() first creates a Seurat object from an expression matrix or data frame and then creates a PALMO S4 object. Function annotateMetadata() assigns columns in the original sample annotation data to designated variables (sample_column, donor_column, and time_column) of the PALMO object for longitudinal analysis. Function mergePALMOdata() cleans up the PLAMO object by filtering out data missing essential information on sample_column, donor_column, or time_column. Function checkReplicates() first checks whether there are replicated samples at the same time points and of the same participants and, if yes, takes the median values among replicated samples. Function avgExpCalc() carries out pseudo-bulking on single-cell omics data. Function naFilter() filters out data whose fraction of NAs is above na_cutoff (default: 0.4). ## Variance decomposition analysis (VDA) For variance decomposition, we want to evaluate contributions from factors of interest \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\{ {F}_{i}\}$$\end{document}{Fi} to the total variance of analyte Y with or without the influence of fixed effects \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\{{X}_{j}\}$$\end{document}{Xj}. Some \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\{ {F}_{i}\}$$\end{document}{Fi} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\{{X}_{j}\}$$\end{document}{Xj} may be the same variables. We treat \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\{{F}_{i}\}$$\end{document}{Fi} as random effects in a linear mixed model, that is, with fixed effects,1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Y\, \sim \,{X}_{1}+{X}_{2}+\ldots+{X}_{m}+\,\left(1{{{{{\rm{|}}}}}}{F}_{1}\right)+\left({1{{{{{\rm{|}}}}}}F}_{2}\right)+\ldots+(1{{{{{\rm{|}}}}}}{F}_{n}).$$\end{document}Y~X1+X2+…+Xm+1∣F1+1∣F2+…+(1∣Fn). Or, without fixed effects,2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Y\, \sim \,\left(1{{{{{\rm{|}}}}}}{F}_{1}\right)+\left({1{{{{{\rm{|}}}}}}F}_{2}\right)+\ldots+(1{{{{{\rm{|}}}}}}{F}_{n}).$$\end{document}Y~1∣F1+1∣F2+…+(1∣Fn). Using lme440, one can obtain the corresponding variance \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma }_{i}^{2}$$\end{document}σi2, including the residual variance \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma }_{R}^{2}$$\end{document}σR2. Then the total variance of Y is given by3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma}_{total}^{2}={\sigma}_{1}^{2}+{\sigma}_{2}^{2}+\ldots+{\sigma}_{n}^{2}+{\sigma}_{R}^{2}.$$\end{document}σtotal2=σ12+σ22+…+σn2+σR2. The proportion of variance explained by factor \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${F}_{i}$$\end{document}*Fi is* then \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma }_{i}^{2}/{\sigma }_{{total}}^{2}$$\end{document}σi2/σtotal2. Similar approach was used in variancePartition19 where the percentage of variance explained was interpreted as the intra-class correlation (ICC). VDA can be performed with the function lmeVariance(). VDA results can be displayed with functions variancefeaturePlot() and gene_featureplot(). ## Coefficient of variation (CV) profiling (CVP) CVP is designed for bulk longitudinal data and contains two functions: [1] Function cvCalcBulkProfile() calculates CV of all features and generates the corresponding CV profile. [ 2] Function cvCalcBulk() identifies consistently stable and variable features, which has two important parameters: Parameter cvThreshold (default: $5\%$) specifies the CV cutoff for distinguishing stable (CV < cvThreshold) or variable (CV > cvThreshold) features. Parameter donorThreshold (default: the total number of donors) defines the minimum number of donors on which a feature needs to be stable or variable to be considered as consistently stable or variable. One may choose cvThreshold as the mode of the corresponding CV distribution. ## Stability pattern evaluation across cell types (SPECT) SPECT is the CVP counterpart for single-cell data and contains the following functions: [1] Function cvCalcSCProfile() calculates the CVs of all features in individual cell types and of individual donors and generates the corresponding CV profile. [ 2] Function cvSCsampleprofile() calculates the CVs of all features of individual donors regardless of difference in cell types and generates the corresponding CV profile. [ 3] Function cvCalcSC() determines whether individual features are stable (CV < cvThreshold) or variable (CV > cvThreshold) in individual cell types and of individual donors. One may choose cvThreshold as the mode of the corresponding CV distribution or a convenient value based on the CVs of housekeeping genes. [ 4] Function VarFeatures() first counts how many times individual features are variable in cell type-donor combinations and then classifies variable features as follows: Features whose counts are above parameter groupThreshold are classified as super variable (SUV). Features whose counts are below groupThreshold but which are consistently variable across all donors in at least one cell type are classified as variable across time in cell-types (VATIC). The default groupThreshold value is set to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${N}_{{donor}}*{N}_{{celltype}}/2$$\end{document}Ndonor*Ncelltype/2 where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${N}_{{donor}}$$\end{document}*Ndonor is* the number of donors and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${N}_{{celltype}}$$\end{document}*Ncelltype is* the number of cell types. [ 5] Function StableFeatures() is similar to VarFeatures() but classifies stable features as super stable (SUS) or stable across time in cell-types (STATIC). [ 6] Function dimUMAPPlot() generates a UMAP plot using a set of selected genes as input. ## Outlier detection analysis (ODA) ODA applies both graphic and statistical methods to examine the temporal behavior of longitudinal data. Function sample_correlation() calculates intra- and inter-donor correlations (across analytes) and displays the results in a heatmap. Timepoints showing obvious weaker correlations with other timepoints are potential outliers. To detect abnormal timepoints, function outlierDetect() first calculates the mean and the standard deviation (SD) of each analyte from samples of the same donor across all timepoints, calculates \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$z=\frac{{value}-{mean}}{{SD}}$$\end{document}z=value−meanSD for the analyte at individual timepoints, and then counts at individual timepoints how many analytes are outliers with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left|z\right| > {z}_{0}$$\end{document}z>z0, where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${z}_{0}$$\end{document}z0 is a user selected cutoff value. Assuming \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$z$$\end{document}z follows a normal distribution, it is straightforward to calculate the expected rate \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r$$\end{document}r of analytes having \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left|z\right| > {z}_{0}$$\end{document}z>z0 (two-sided) or having \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$z > {z}_{0}$$\end{document}z>z0 or \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$z < -{z}_{0}$$\end{document}z<−z0 (one-sided). Afterwards function outlierDetectP() uses binomial tests to evaluate the p values for the counts of outliers at individual timepoints and applies Benjamini and Hochberg procedure to adjust the p values since multiple timepoints are tested. A donor-specific abnormal timepoint is identified if the corresponding adjusted p value is less than 0.05. In this study we chose \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${z}_{0}=2.5$$\end{document}z0=2.5 and thus \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$$r = 1.24$\%$$\end{document}$r = 1.24$% for \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left|z\right| > 2.5$$\end{document}z>2.5 or \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$$r = 0.62$\%$$\end{document}$r = 0.62$% for \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$z > 2.5$$\end{document}z>2.5 or \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$z < -2.5$$\end{document}z<−2.5. While the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$z$$\end{document}z-score method described here can handle data with only three timepoints, Dixon’s test may be a better alternative for such a small dataset. ## Time course analysis (TCA) Function sclongitudinalDEG() uses the hurdle model implemented in the MAST package (https://github.com/RGLab/MAST/) to study temporal changes in longitudinal scRNA-seq data. The data is first split into subsets of individual cell types and individual participants and then analyzed independently. If the data has at least three timepoints, the function models normalized expression of each gene as a linear function of time and evaluates the slope of time and the corresponding p value (likelihood ratio test). If the data has only two timepoints, the function performs DEG analysis between the two timepoints as implemented in MAST and obtains fold change and the corresponding p value. Potential confounding factors (such as experimental batch, sex, age, etc.) can be specified by parameter adjfac which are adjusted in the analysis. Genes that are expressed in less than a certain fraction of cells (specified by parameter mincellsexpressed, default 0.1) are filtered out from the analysis. Obtained p-values are adjusted for multiple comparisons using the Benjamini and Hochberg procedure. Adjusted p-value <0.05 were considered significant in this study. ## Circos plots for displaying stability patterns PALMO has two functions to show the stability patterns of single-cell omics data. *Function* genecircosPlot() displays the CV values of features of interest in individual cell types and across individual donors based on a single data modality. Function multimodalView() displays the CV values of features of interest in individual cell types and across individual donors based on two independent data modalities. ## Random correlation between gene expression and gene score *To* generate the distribution of random correlation between gene expression in scRNA-seq data and gene score in scATAC-seq data, we randomly shuffled the order of reliable genes, calculated the correlations between expression of pre-shuffle genes and gene score of post-shuffle genes at the same positions, and repeated the process 1000 times. The obtained distribution of correlations provided a good estimate on the correlation between random, unrelated gene pairs, which had a $95\%$ upper confidence bound at R0 = 0.399. Any correlations below R0 were no better than that between random, unrelated gene pairs and thus not statistically meaningful. ## Published single cell datasets We retrieved scRNA-seq data from published PBMC datasets CNP00011023, GSE1496892, and GSE16437816. Datasets CNP0001102 and GSE164378 were from longitudinal studies. Single-cell data objects were created in Seurat v4.0.0 and cells were labeled as in the original studies. Dataset CNP0001102 consists of three healthy controls (normal), two participants infected with influenza (Flu) and five participants infected with SARS-CoV-2 (COVID-19). Dataset GSE149689 consists of four normal, five Flu, and eleven COVID-19 participants. Dataset GSE164378 dataset consists of eight participants with PBMC samples collected at three timepoints. Mouse brain scRNA-seq data was obtained from published dataset GSE12978836. The dataset contains single cell RNA data from brain tissues of eight young (2–3 months) and eight old (21–23 months) mice. The dataset consists of a total 37,069 cells labeled to 25 cell types. ## TCRß repertoire dataset We downloaded the TCRβ sequencing data of 4 systemic sclerosis patients from GSE15698026. First, we merged the TCR repertoire data from the 4 patients with 3 timepoints into a single file. Second, we calculated the frequency of each unique CDR3 peptide in each patient sample as the ratio between the observed reads of the peptide to the total peptide reads in the sample. Third, we termed unique CDR3 peptides as clonotypes and labeled them from 1 to the total number of clonotypes. In total, we collected 288,597 (out of 355,024) unique clonotypes from CD4+ T cells and 11,739 (out of 14,883) from CD8+ T cells, respectively. The frequency data matrix from CD4+ or CD8+ T cells was then submitted to PALMO as input data frame. ## Differential expression gene (DEG) analysis on scRNA-seq data DEG analysis on datasets (CNP0001102 and GSE149689) was performed using the FindMarkers function from the Seurat package (version 4.0.0). The groups were specified using “ident.1” and “ident.2” in the function. The Benjamini and Hochberg (BH) procedure as implemented in the Seurat package was applied to adjust p-values, controlling the false discovery rate (FDR) in multiple testing. DEGs were identified if the corresponding average log2-Fold change was greater than 0.1 and the corresponding adjusted p value was less than 0.05. ## Seurat differential analysis on longitudinal scRNA-seq data of a COVID19 patient Seurat based differential analysis was performed on the longitudinal scRNA-seq data of activated CD4+ T cells of patient COV-5 in dataset CNP00011023, using the function FindMarkers() with parameters test.use = “MAST” and logfc.threshold = 0. The groups were defined by parameters ident.1 and ident.2. For example, to capture differential genes between day 1 (D1) versus day 7 (D7) and day 13 (D13), we selected ident.1 = D1 and ident.2 = (D7 and D13). Similar approach was carried out for comparing D13 versus D1 and D7 (ident.1 = (D1 and D7) and ident.2 = D13). The significant genes were identified by adjusted p value <0.05. ## Pathway enrichment analysis Fast Gene Set Enrichment Analysis (fgsea) was performed to identify enriched pathways among targeted genes41. A custom collection of gene sets that included the GO v7.2, KEGG v7.2 and Hallmark v7.2 from the Molecular Signatures Database (MSigDB, v7.2) were used as the pathway database. Genes were pre-ranked by the decreasing order of their correlation coefficients. The running sum statistics and Normalized Enrichment Scores (NES) were calculated for each comparison. The pathway enrichment p-values were adjusted using the Benjamini and Hochberg procedure and pathways with adjusted p-values <0.05 were considered significantly enriched. Over representation analysis was performed using the Fisher test. For a single sample GSEA (ssGSEA), we used the GSVA v1.40 R package27. ## Data analysis and visualization Data analysis was performed in R, a statistical computing language (https://www.R-project.org/). Basic data visualization was performed using ggplot2 v3.3, ggpubr 0.4, and circular plots by circlize v0.4. The UMAP visualization was performed using Seurat v4.0.0. Statistical tests were performed as mentioned in each section. Multi-test correction was applied to the p-values to control the FDR using the Benjamini and Hochberg procedure and adjusted $p \leq 0.05$ were considered significant. ## Supplementary information Supplementary Information Transparent Peer Review File Editorial Assessment Report Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Supplementary Data 6 Supplementary Data 7 Supplementary Data 8 The online version contains supplementary material available at 10.1038/s41467-023-37432-w. ## Source data Source Data ## Peer review information Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. This article has been peer reviewed as part of Springer Nature’s Guided Open Access initiative. ## References 1. Bernardes JP. **Longitudinal multi-omics analyses identify responses of megakaryocytes, erythroid cells, and plasmablasts as hallmarks of severe COVID-19**. *Immunity* (2020.0) **53** 1296-1314.e9. DOI: 10.1016/j.immuni.2020.11.017 2. 2.Lee, J. S. et al. Immunophenotyping of COVID-19 and influenza highlights the role of type I interferons in development of severe COVID-19. Sci. Immunol.5, eabd1554 (2020). 3. Zhu L. **Single-cell sequencing of peripheral mononuclear cells reveals distinct immune response landscapes of COVID-19 and influenza patients**. *Immunity* (2020.0) **53** 685-696.e3. DOI: 10.1016/j.immuni.2020.07.009 4. Zhou W. **Longitudinal multi-omics of host-microbe dynamics in prediabetes**. *Nature* (2019.0) **569** 663-671. DOI: 10.1038/s41586-019-1236-x 5. Zhang S. **Longitudinal single-cell profiling reveals molecular heterogeneity and tumor-immune evolution in refractory mantle cell lymphoma**. *Nat. Commun.* (2021.0) **12** 2877. DOI: 10.1038/s41467-021-22872-z 6. Buenrostro JD. **Single-cell chromatin accessibility reveals principles of regulatory variation**. *Nature* (2015.0) **523** 486-490. DOI: 10.1038/nature14590 7. Sharma A. **Longitudinal single-cell RNA sequencing of patient-derived primary cells reveals drug-induced infidelity in stem cell hierarchy**. *Nat. Commun.* (2018.0) **9** 4931. DOI: 10.1038/s41467-018-07261-3 8. Wilk AJ. **A single-cell atlas of the peripheral immune response in patients with severe COVID-19**. *Nat. Med.* (2020.0) **26** 1070-1076. DOI: 10.1038/s41591-020-0944-y 9. 9.Diggle, P. J., Heagerty, P., Liang, K.-Y. & Zeger, S. L. Analysis of Longitudinal Data. (Oxford Univ. Press, 2002). 10. Fitzmaurice GM, Ravichandran C. **A primer in longitudinal data analysis**. *Circulation* (2008.0) **118** 2005-2010. DOI: 10.1161/CIRCULATIONAHA.107.714618 11. Gustafsson J. **Sources of variation in cell-type RNA-Seq profiles**. *PLoS One* (2020.0) **15** e0239495. DOI: 10.1371/journal.pone.0239495 12. Hafemeister C, Satija R. **Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression**. *Genome Biol.* (2019.0) **20** 296. DOI: 10.1186/s13059-019-1874-1 13. Sperisen P, Cominetti O, Martin F-PJ. **Longitudinal omics modeling and integration in clinical metabonomics research: challenges in childhood metabolic health research**. *Front. Mol. Biosci.* (2015.0) **2** 44. DOI: 10.3389/fmolb.2015.00044 14. Dal Molin A, Baruzzo G, Di Camillo B. **Single-cell RNA-Sequencing: assessment of differential expression analysis methods**. *Front. Genet.* (2017.0) **8** 62. DOI: 10.3389/fgene.2017.00062 15. Finak G. **MAST: a flexible statistical framework for assessing transcriptional changes and characterizing heterogeneity in single-cell RNA sequencing data**. *Genome Biol.* (2015.0) **16** 278. DOI: 10.1186/s13059-015-0844-5 16. Hao Y. **Integrated analysis of multimodal single-cell data**. *Cell* (2021.0) **184** 3573-3587.e29. DOI: 10.1016/j.cell.2021.04.048 17. Squair JW. **Confronting false discoveries in single-cell differential expression**. *Nat. Commun.* (2021.0) **12** 5692. DOI: 10.1038/s41467-021-25960-2 18. Zimmerman KD, Espeland MA, Langefeld CD. **A practical solution to pseudoreplication bias in single-cell studies**. *Nat. Commun.* (2021.0) **12** 738. DOI: 10.1038/s41467-021-21038-1 19. Hoffman GE, Schadt E. **E. variancePartition: interpreting drivers of variation in complex gene expression studies**. *BMC Bioinforma.* (2016.0) **17** 483. DOI: 10.1186/s12859-016-1323-z 20. Nazarov VI. **tcR: an R package for T cell receptor repertoire advanced data analysis**. *BMC Bioinforma.* (2015.0) **16** 175. DOI: 10.1186/s12859-015-0613-1 21. Granja JM. **ArchR is a scalable software package for integrative single-cell chromatin accessibility analysis**. *Nat. Genet.* (2021.0) **53** 403-411. DOI: 10.1038/s41588-021-00790-6 22. Uhlen M. **Towards a knowledge-based human protein atlas**. *Nat. Biotechnol.* (2010.0) **28** 1248-1250. DOI: 10.1038/nbt1210-1248 23. Cao W. **Regulation of TLR7/9 responses in plasmacytoid dendritic cells by BST2 and ILT7 receptor interaction**. *J. Exp. Med.* (2009.0) **206** 1603-1614. DOI: 10.1084/jem.20090547 24. Schotte R, Nagasawa M, Weijer K, Spits H, Blom B. **The ETS transcription factor Spi-B is required for human plasmacytoid dendritic cell development**. *J. Exp. Med.* (2004.0) **200** 1503-1509. DOI: 10.1084/jem.20041231 25. Chbili C. **Glutathione S-transferase M1 and T1 polymorphisms and the risk of mild hepatotoxicity induced by carbamazepine in a tunisian population study**. *BMC Neurol.* (2018.0) **18** 24. DOI: 10.1186/s12883-018-1013-8 26. Servaas NH. **Longitudinal analysis of T-cell receptor repertoires reveals persistence of antigen-driven CD4+ and CD8+ T-cell clusters in systemic sclerosis**. *J. Autoimmun.* (2021.0) **117** 102574. DOI: 10.1016/j.jaut.2020.102574 27. Hänzelmann S, Castelo R, Guinney J. **GSVA: gene set variation analysis for microarray and RNA-seq data**. *BMC Bioinforma.* (2013.0) **14** 7. DOI: 10.1186/1471-2105-14-7 28. 28.Gnanaprakasam, J. N. R. & Wang, R. MYC in Regulating Immunity: Metabolism and Beyond. Genes8, 88 (2017). 29. Hu X, Ivashkiv LB. **Cross-regulation of signaling pathways by interferon-gamma: implications for immune responses and autoimmune diseases**. *Immunity* (2009.0) **31** 539-550. DOI: 10.1016/j.immuni.2009.09.002 30. Trigunaite A, Dimo J, Jørgensen TN. **Suppressive effects of androgens on the immune system**. *Cell. Immunol.* (2015.0) **294** 87-94. DOI: 10.1016/j.cellimm.2015.02.004 31. Burrack AL, Martinov T, Fife BT. **T cell-mediated beta cell destruction: autoimmunity and alloimmunity in the context of type 1 diabetes**. *Front. Endocrinol.* (2017.0) **8** 343. DOI: 10.3389/fendo.2017.00343 32. 32.Di Cara, F. et al. Peroxisomes in immune response and inflammation. Int. J. Mol. Sci. 20, 3877 (2019). 33. Sofer A, Lei K, Johannessen CM, Ellisen LW. **Regulation of mTOR and cell growth in response to energy stress by REDD1**. *Mol. Cell. Biol.* (2005.0) **25** 5834-5845. DOI: 10.1128/MCB.25.14.5834-5845.2005 34. Verstrepen L. **Expression, biological activities and mechanisms of action of A20 (TNFAIP3)**. *Biochem. Pharmacol.* (2010.0) **80** 2009-2020. DOI: 10.1016/j.bcp.2010.06.044 35. Lin Y. **scMerge leverages factor analysis, stable expression, and pseudoreplication to merge multiple single-cell RNA-seq datasets**. *Proc. Natl. Acad. Sci. U. S. A.* (2019.0) **116** 9775-9784. DOI: 10.1073/pnas.1820006116 36. Ximerakis M. **Single-cell transcriptomic profiling of the aging mouse brain**. *Nat. Neurosci.* (2019.0) **22** 1696-1708. DOI: 10.1038/s41593-019-0491-3 37. Savage AK. **Multimodal analysis for human ex vivo studies shows extensive molecular changes from delays in blood processing**. *iScience* (2021.0) **24** 102404. DOI: 10.1016/j.isci.2021.102404 38. Stoeckius M. **Cell Hashing with barcoded antibodies enables multiplexing and doublet detection for single cell genomics**. *Genome Biol.* (2018.0) **19** 224. DOI: 10.1186/s13059-018-1603-1 39. Stuart T. **Comprehensive integration of single-**. *Cell Data Cell* (2019.0) **177** 1888-1902.e21. PMID: 31178118 40. Bates D, Mächler M, Bolker B, Walker S. **Fitting linear mixed-effects models using lme4**. *J. Stat. Softw.* (2015.0) **67** 1-48. DOI: 10.18637/jss.v067.i01 41. 41.Korotkevich, G. et al. Fast gene set enrichment analysis. bioRxiv 060012 10.1101/060012 (2021). 42. 42.Vasaikar, S. V. et al. A comprehensive platform for analyzing longitudinal multi-omics data (v1.2). Zenodo, 10.5281/zenodo.7549226 (2023).
--- title: 'Silmitasertib (CX-4945), a Clinically Used CK2-Kinase Inhibitor with Additional Effects on GSK3β and DYRK1A Kinases: A Structural Perspective' authors: - Przemyslaw Grygier - Katarzyna Pustelny - Jakub Nowak - Przemyslaw Golik - Grzegorz M. Popowicz - Oliver Plettenburg - Grzegorz Dubin - Filipe Menezes - Anna Czarna journal: Journal of Medicinal Chemistry year: 2023 pmcid: PMC10041529 doi: 10.1021/acs.jmedchem.2c01887 license: CC BY 4.0 --- # Silmitasertib (CX-4945), a Clinically Used CK2-Kinase Inhibitor with Additional Effects on GSK3β and DYRK1A Kinases: A Structural Perspective ## Abstract A clinical casein kinase 2 inhibitor, CX-4945 (silmitasertib), shows significant affinity toward the DYRK1A and GSK3β kinases, involved in down syndrome phenotypes, Alzheimer’s disease, circadian clock regulation, and diabetes. This off-target activity offers an opportunity for studying the effect of the DYRK1A/GSK3β kinase system in disease biology and possible line extension. Motivated by the dual inhibition of these kinases, we solved and analyzed the crystal structures of DYRK1A and GSK3β with CX-4945. We built a quantum-chemistry-based model to rationalize the compound affinity for CK2α, DYRK1A, and GSK3β kinases. Our calculations identified a key element for CK2α’s subnanomolar affinity to CX-4945. The methodology is expandable to other kinase selectivity modeling. We show that the inhibitor limits DYRK1A- and GSK3β-mediated cyclin D1 phosphorylation and reduces kinase-mediated NFAT signaling in the cell. Given the CX-4945’s clinical and pharmacological profile, this inhibitory activity makes it an interesting candidate with potential for application in additional disease areas. ## Introduction The human kinome contains 518 protein kinases accounting for $1.7\%$ of all human genes.1 Kinases control a variety of physiological processes, including cell growth, differentiation, proliferation, angiogenesis, apoptosis, cytoskeleton rearrangement, metabolism, and others.2 Deregulation of specific kinases has been linked to virtually all major disease areas.3 Consequently, kinases became one of the most important targets for drug discovery,4 with 73 kinase inhibitor drugs authorized to date by the FDA.5 CX-4945 (silmitasertib) is the first orally bioavailable inhibitor of casein kinase 2 (CK2) with acceptable pharmacological properties.6 CX-4945 is a molecule with a relatively low polar surface area, few rotatable bonds, and low aqueous solubility, bearing one carboxylic acid and two weakly basic aromatic nitrogen residues. Its molecular properties are summarized in the Supporting Material, Table S1. It was granted orphan drug designation by the FDA for cholangiocarcinoma (bile duct cancer) in 2017 and is developed as a drug candidate in other solid tumors.7 Phosphorylation of cellular targets by CK2 was reported to facilitate SARS-CoV-2 spread, and subsequently, antiviral activity for CX-4945 was reported.8 Clinical trials to demonstrate putative benefits are ongoing.9 CX-4945 is an ATP-competitive inhibitor characterized by a Ki of 380 pM for CK2α. Although it inhibits a few off-target kinases (e.g., PIM1, HIPK3, or CLK3),10 the reported adverse effects are moderate. This illustrates a current paradigm shift, where targeting multiple kinases with related substrate phosphorylation patterns is considered advantageous over highly specific kinase inhibitors.11−13 The dual-specificity tyrosine phosphorylation-regulated kinase 1A (DYRK1A) encoding gene is located in the Down syndrome critical region and was originally associated with neurodegenerative diseases, including Alzheimer’s and Down syndrome.14 More recent findings also implicated the role of DYRK1A in cancer.15,16 DYRK1A phosphorylation primes the substrates for subsequent phosphorylation by a processive kinase, the constitutively active glycogen synthase kinase-3β (GSK3β).17,18 GSK3β phosphorylates more than a hundred different substrates19 and has been implicated in diverse cellular processes, including embryogenesis, immune response, inflammation, apoptosis, autophagy, wound healing, neurodegeneration, and carcinogenesis.20 Abnormal regulation of GSK3β has been linked to the onset and progression of chronic conditions, including cancer, diabetes, neurodegenerative, and behavioral diseases.21 The processive kinase GSK3β and its priming kinase DYRK1A are critical for the regulation of β-cell function. The inhibition of DYRK1A and GSK3β brings a synergistic effect, leading to an increase in insulin release and β-cell proliferation. The synergism has been attributed, among others, to the regulation of ion channels responsible for the exocytosis of insulin from storage granules, and its benefits on β cell health could already impressively be demonstrated by an aminopyrazine series of inhibitors.22 The latter possess about equipotent DYRK1A and GSK3β inhibitory efficacy. This improvement of glucose homeostasis by the synergistic inhibition of DYRK1A and GSK3β hints at an opportunity for the curative therapy of diabetes.23−25 DYRK1A and GSK3β are serine–threonine kinases of the CMGC family and share many structural features.26 This opens the possibility for developing dual inhibitors, which is especially attractive since the latter may benefit from the synergistic effects mentioned above. Many ATP-competitive inhibitors and allosteric modulators of GSK3β have been reported.27 Several DYRK1A inhibitors have also been discovered and investigated in the context of neurodegenerative disorders28,29 and autism.30 While inhibiting multiple CMGC kinases,31 the inhibitors reported to date are characterized by relatively poor affinities. Prior studies have demonstrated CX-4945’s inhibition of DYRK1A,32 prompting us to evaluate its utility as an antidiabetic. As silmitasertib is already undergoing clinical evaluation in patients, its overall safety profile seemed promising. Furthermore, the reported primary activity as a potent CK2 inhibitor can possibly exert additional beneficial effects in diabetes,33 thus further increasing the proposed synergism. Here, we present the results of a structural and computational evaluation of CX-4945 as a dual DYRK1A and GSK3β kinase inhibitor. The crystal structures of the inhibitor in complex with DYRK1A and both the phosphorylated and non-phosphorylated forms of GSK3β kinase reveal the binding mode of CX-4945 in those kinases. Rational justification of CX-4945’s binding and inhibitory power from the crystal structures required the involvement of quantum chemical calculations. This allowed the identification and assignment of crucial contributions and roles for each functional group of the inhibitor for binding. This method, which translates structural data into different energy contributions, can be extrapolated to other kinases. Moreover, we show functional implications of the inhibition of DYRK1A and GSK3β at the cellular levels. ## CX-4945 is a Potent Inhibitor of DYRK1A and GSK3β CX-4945 (Figure 1A) is a potent ATP-competitive small molecule inhibitor of CK2, with the unusual structural feature of having a free carboxylic acid. This is rarely seen in kinase inhibitors. For targeting CK2α, however, this functional group seems to be beneficial, as it is a characteristic shared with other inhibitors, like TTP22 or CX-5011. CX-4945 also inhibits kinases from the CMGC family, including the dual-specificity protein kinases CLK2, CLK3, and serine/threonine homeodomain-interacting protein kinase 3 (HIPK3) or others.34,35 Additionally, CX-4945 was reported earlier to block DYRK1A.32 In our effort to identify dual DYRK1A/GSK3β inhibitors, we decided to evaluate the efficacy of CX-4945 as a potential multitarget inhibitor. **Figure 1:** *(A) Chemical structure of CX-4945 (silmitasertib). (B, C) Direct interaction determined by microscale thermophoresis (MST) of an inhibitor with DYRK1A (upper panel) and GSK3β (lower panel), with the Kd value summarized in the table. (D) Inhibitory activity of CX-4945 against DRYRK1A (upper panel) and GSK3β (lower panel) as determined in the Cook kinase activity assay. (E) CX-4945 inhibits DYRK1A and GSK3β-mediated phosphorylation of Cyclin D1 in the transiently transfected HEK293T cells with plasmids encoding HA-Cyclin D1 and FLAG-DYRK1A or FLAG-GSK3β. The protein profile was analyzed in cell lysate with Western Blot using specific monoclonal antibodies anti-FLAG (DYRK1A and GSK3β), anti-HA (Cyclin D1), and anti-phosho-Cyclin D1.* First, we investigated the direct interaction of three compounds with DYRK1A and GSK3β using microscale thermophoresis (MST). Our tests included CX-4945, harmine, a known inhibitor of DYRK1A,36 and 1-azakenpaullone, an inhibitor of GSK3β.37 The recombinant kinase domains of DYRK1A (126–490) and GSK3β (26–383), both with C-terminal His-tag extension, were expressed in E. coli, purified to homogeneity, and labeled with a relevant fluorescent dye (cf. Experimental Section). Then, the dissociation constants were determined in the direct binding assay using constant concentrations of DYRK1A or GSK3β kinase domains (20 and 62.5 nM, respectively) titrated with increasing concentrations of the compounds. Increasing the concentration of the small molecules dose-dependently affected the thermophoretic profile of both DYRK1A and GSK3β, suggesting physical interaction with both kinases (Figure 1C). Expectedly, harmine was significantly more effective toward DYRK1A than GSK3β. On the other hand, 1-azakenpaullone affected only its known target, GSK3β, but not DYRK1A, demonstrating the specificity of the assay. Our results indicate that CX-4945 strongly binds to both kinases with nanomolar dissociation constants (Kd) (Figure 1B). The affinity of CX-4945 toward DYRK1A was higher compared to the reference compound harmine (Kd 1.8 and 13.4 nM, respectively), while for GSK3β it was comparable to that of the reference compound, 1-azakenpaullone (37.8 nM). In a global analysis of 243 clinically evaluated kinase drugs,38 the reported Kd,app (apparent) of DYRK1A-CX-4945 and GSK3β-CX-4945 were 35 and 4800 nM, respectively. This Kd,app was calculated based on EC50 values from the Kinobeads assay of the cell lysates titrated with compounds. The values obtained were 35.2 ± 20.74 nM for DYRK1A and 10 429.37 ± 32 287.91 nM for GSK3β (Table S2 of ref [38]). The observed discrepancy may be due to the nature of the proteins analyzed: in our MST assay, truncated kinases (active kinase domains) were used, while in the Kinobeads assay, the whole proteins were analyzed. Moreover, the activation status of an endogenous kinase in cells may not be the same as that of a recombinant protein in a cell-free system. Nevertheless, both assays identified DYRK1A and GSK3β as CX-4945 targets. Next, we checked whether CX-4945 was able to impair DYRK1A- and GSK3β-mediated phosphorylation. The inhibitory potency (IC50) was evaluated using the Cook activity assay.39 For this, a fixed amount of the kinase (0.25 μM) was titrated with the tested inhibitors (1 nM to 10 μM) in the presence of ATP (128 μM) and a substrate peptide (0.5 mM). The ATP concentration was selected based on the experimentally determined KM of 118 and 128 μM for DYRK1A and GSK3β, respectively. The DYRKtide peptide (RRRFRPASPLRGPPK) was used as a substrate for DYRK1A, while the GYS1 peptide (YRRAAVPPSPSLSRHSSPHQ(pS)EDEEE) with the phosphoSer residue in the +4 position was used for GSK3β. CX-4945 turned out to be a very potent inhibitor of both kinases with potency (IC50) in the nanomolar range (Figure 1D). For DYRK1A, once more, CX-4549 proved to be more active than harmine (IC50 of 0.16 and 0.27 μM, respectively) (Figure 1D upper panel), while no inhibitory activity was detected for 1-azakenpaullone. Our results from the Cook assay correlate well with the previously published data for DYRK1A,32 where CX-4945 inhibits this kinase in the nanomolar range and with higher potency than harmine. The small discrepancies observed are due to assay sensitivity. CX-4945 reveals furthermore similar inhibitory activity against other kinases from the DYRK family, like DYRK2.10 Moreover, CX-4945 strongly inhibits GSK3β with IC50 values of 0.19 μM and shows an inhibitory power comparable to 1-azakenpaullone (IC50 0.15 μM) (Figure 1D lower panel). In the kinome scan with a single point inhibition readout presented by Battistutta and co-workers, CX-4945 showed $55\%$ inhibition of GSK3β at 500 nM,10 while in our assay, with a concentration range of 1 nM to 10 μM, we observed $50\%$ inhibition at 190 nM. Our CX-4945 values are slightly shifted toward higher activity. To further confirm our findings, we analyzed the effect of CX-4945 binding on the thermal stability of both kinases. The protein’s melting temperatures in the presence of the inhibitors were determined using the dye-based thermal-shift assay (Figure S3).40 The thermal-shift assay indicated that binding of CX-4945 significantly stabilized DYRK1A and GSK3β kinase domains and induced shifts of 12 °C and 9.5 °C, respectively. For DYRK1A, the observed stabilizing effect of CX-4945 was, once more, stronger than that of harmine (10.5 °C), while 1-azakenpaullone induced only slight changes in the protein’s melting temperature (2 °C). For GSK3β, the interaction with 1-azakenpaullone had a more prominent effect (13 °C), but no temperature stabilization effect was observed after incubation with harmine. Our data establish CX-4945 as a dual DYRK1A and GSK3β inhibitor with in vitro potency allowing us to expect a biological effect. To evaluate our hypothesis, we tested whether CX-4945 would inhibit DYRK1A and GSK3β in mammalian cell lines. GSK3β is a constitutively active kinase whose activity in a cell is modulated by various kinases through phosphorylation at Ser9. Therefore, we used only GSK3β (26–383) kinase domain constructs in our cellular studies. Consequently, the observed effects are caused by the inhibitors and not by other cellular events. Among the common substrates of both kinases, we selected Cyclin D1, which is a positive regulator of the cell cycle. It controls the transition from a proliferative to a quiescent state and determines the fate of the cell. Both DYRK1A and GSK3β were shown to be required for Cyclin D1 phosphorylation. This, in turn, leads to the nuclear export of Cyclin D1 and its subsequent degradation in the proteasome.41,42 Moreover, Cyclin D1 belongs to a limited group of GSK3β substrates for which priming is not mandatory,43 and thus allows the direct investigation of GSK3β inhibition. The effect of CX-4945 on Cyclin D1’s phosphorylation status was evaluated in HEK293T cells, where the tested kinases and Cyclin D1 were transiently overexpressed. The protein profile was analyzed in cell lysate with Western Blot (Figure 1E) using specific monoclonal antibody anti-FLAG (DYRK1A and GSK3β), anti-HA (Cyclin D1), and anti-phospho-Cyclin D1 to exclude endogenous proteins. Thr289 was chosen because it is a known phosphorylation site of both tested kinases but not for Cyclin D-dependent kinases.44 When DYRK1A or GSK3β alone was overexpressed, the signal from phosphorylated Cyclin D1 was not detected, demonstrating that the endogenous expression of Cyclin D1 does not interfere with the assay’s results. Additionally, when Cyclin D1 alone was overexpressed, the amount of the endogenous phosphorylated form was negligible. The effect of DYRK1A overexpression on the phosphorylation of Cyclin D1 was fully abolished in the presence of harmine. The same effect was obtained with 1-azakenpaullone when GSK3β was overexpressed together with Cyclin D1. This demonstrates that the observed phosphorylation is indeed mediated by the respective kinases. Furthermore, for both kinases, the treatment with CX-4945 strongly inhibits Cyclin D1 phosphorylation, and complete inhibition was observed for 10 μM concentration. Collectively, these results clearly demonstrate that CX-4945 is a potent DYRK1A and GSK3β inhibitor in vitro and in mammalian cells. ## Structural Basis of DYRK1A Inhibition by CX-4945: Binding to the ATP Pocket To determine the molecular mechanism of DYRK1A inhibition by CX-4945, we solved the crystal structure of the kinase inhibitor complex at 2.77 Å resolution. The complex was crystallized in the C121 space group, with eight protein molecules found in the asymmetric unit. For all molecules in the asymmetric unit, the entire protein was well-ordered and comprised a long hairpin-like structure of an N-terminal DH box followed by a catalytic domain in the active kinase conformation (Figure S1A). Mass spectrometry analysis revealed heterogenous phosphorylation of our DYRK1A preparation (Figure S1C). However, the electron density maps showed only phosphorylation of DYRK1A at Tyr321, the second tyrosine of the YxY motif of the activation loop. This suggests that other phosphorylation spots are either of low frequency or located in regions undefined by the electron density. The electron density clearly defines the inhibitor in all eight protein–ligand complexes contained in the asymmetric unit (Figure S2A–H). The inhibitor-containing molecules superimpose with an average root-mean-square deviation (RMSD) below 0.4 Å over 320 Cα atoms. Because the inhibitor’s binding mode is equivalent in all complexes in the asymmetric unit, further discussion relates to molecule A unless indicated otherwise. CX-4945 occupies the ATP-binding site, sandwiched between the N- and C-lobes of the kinase domain (Figure 2A). In analogy to the protein–ligand structure docked by Kim and co-workers,32 the inhibitor is stabilized by hydrogen bonds involving functional groups in opposing parts of the benzo-naphthyridine moiety. Furthermore, we note that this is a well-conserved network of interactions among several kinases (Figure 5). In their work, Kim and colleagues mention the formation of 4 hydrogen bonds between the ligand and the protein, which slightly contrasts with our structure, where we count 3 instead. We attribute this difference to solvation waters, which seem to have been excluded from the docking strategy. On the other hand, our crystal structure improves on the relative orientation of the chlorophenyl group inside the pocket and lipophilic contacts, which were not entirely well captured in the docked structure. Besides the hydrogen-bond network and lipophilic interactions, water-mediated contacts are visible in our crystal model. The carboxyl moiety of the inhibitor contributes with a direct hydrogen bond with the main chain amide of Asp307 from the DFG motif (Figure 2B). Calculations based on model systems built from the crystal structure show that this is the strongest ligand–residue interaction with a direct hydrogen bond, amounting to −13.5 kcal/mol of stabilization. The interaction with the side chain of Lys188 is also particularly strong, and this is due to the coupled effect of a hydrogen bond with an ionic bridge between the two groups. Our calculations estimate that in the ligand–lysin contact, the ionic bridge amounts to $48\%$ of the −12.6 kcal/mol interaction energy. The hydrogen bond for that same pair corresponds to $41.5\%$, and $10.5\%$ is the interaction between the ligand and the alkyl chain of Lys188. Nearby carboxylates should decrease the strength of the interaction. Nonetheless, the synergy of Lys188 with Asp307 is a key element for binding, as it was previously used to anchor other DYRK1A inhibitors. This includes harmine, a potent and specific inhibitor of DYRK1A (PDB 3ANR).45 The carboxylate group of CX-4945 may additionally participate in a hydrogen bond with the main chain amide of Phe308 and a water-mediated contact with the side chain of Glu203 (Figure 2B). However, we estimate both interactions to be rather weak. In the latter, the carboxylate groups yield too much electrostatic repulsion (stabilization of −1.2 kcal/mol). Furthermore, note that the water-mediated bridge is not present in all complexes, further strengthening the observations from our calculations. For the interaction with Phe308, the large distance to the proton should be the main factor determining an enthalpic gain below 1 kcal/mol. **Figure 2:** *Crystal structure of CX-4945 bound to the active sites of DYRK1A and GSK3β. (A) The overall fold of DYRK1A (violet purple) in cartoon representation with CX-4945 (green sticks) at the ATP-binding pocket. (B) The insert showing CX-4945-DYRK1A interaction in the ATP-binding pocket. (C) Hydrophobic interactions stabilizing CX-4945 at the DYRK1A ATP-binding pocket. (D) Overview of the crystal structure of GSK3β (pink)–CX-4945 (green) complex. (E) Hydrogen-bond interactions of the inhibitor bound to the ATP pocket of GSK3β. (F) Hydrophobic interactions of CX-4945 at the ATP-binding pocket of GSK3β.* One of the nitrogen atoms on the benzo-naphthyridine fragment of the inhibitor (atom 8) contributes with a hydrogen bond with the main chain amide of Leu241 within the hinge region. Calculations on the isolated CX-4945-Leu241 fragment yield the estimate of −9.6 kcal/mol for this contact. The hydrogen bond alone is responsible for as much as $53\%$ of the interaction energy in this ligand–residue pair. The adjacent main chain carboxyl oxygens of Glu239, Leu241, and Ser242 are involved in π-stacking interactions with the benzo-naphthyridine moiety. The latter stabilizes the complex with about −6.5 kcal/mol, and Glu239 is expected to be the main contributor. Hydrophobic interactions involve the benzo-naphthyridine and the side chains of Ile165, Ala186, Val173, Val222, Val306, and Phe238 from the N-lobe, Met240 of the hinge region, and Leu294 and Val306 from the C-lobe (Figure 2C). Model systems based solely on all of the side chains of those amino acids resulted in a contribution of −14.8 kcal/mol to the binding energy. The chlorine atom of the chlorophenyl moiety of CX-4945 resides in a shallow pocket formed by the side chains of Phe170, Val173, and the main chain atoms of residues 166–168 (Figure 2C). These yield a stabilization of −7.5 kcal/mol to the complex. A sulfate ion is tightly coordinated in the close vicinity of the inhibitor by direct hydrogen bonds contributed by the side chains of Asn292 and Asp307 and a water-mediated hydrogen bond with the main chain of Glu291 and the side chain of Asn244. This sulfate contributes with an oxygen−π interaction also involving the chlorophenyl group of CX-4945. The second sulfate ion is coordinated by Ser169 from the glycine-rich loop and Lys289 from the catalytic loop. The second ion is in a position similar to a hydrolyzed γ-phosphate from ATP bound to PKA (PDB 1RDQ)46 or a bound phosphate in the structure of Haspin with a 5-iodotubercidin ligand (PDB 3IQ7).47 *It is* plausible that targeting this conserved binding pocket containing positively charged amino acids by more druggable bioisosteres could improve the selective inhibitor design. Alternatively, the present sulfate ions could be exploited for the formation of additional polar interactions. ## Structural Basis of GSK3β Inhibition by CX-4945: Binding to the ATP Pocket The binding of CX-4945 to both phosphorylated and non-phosphorylated (Tyr216) forms of GSK3β was characterized by protein crystallography. The kinase phosphorylated at the activation loop crystallized in the P3121 space group with two protein molecules in the asymmetric unit, while the non-phosphorylated kinase crystallized in the P43212 space group with a single protein molecule in the asymmetric unit. The structures were refined at 3.00 and 2.85 Å, respectively. Interestingly, both structures were determined from identically prepared kinase samples. Only the crystallization conditions allowed the separation of the phosphorylated from the non-phosphorylated form. The overall structure of GSK3β adopts a classical bilobal kinase fold. The structures of the phosphorylated and non-phosphorylated forms of GSK3β superimpose with an RMSD of 0.41 Å over 252 Cα atoms, and in both cases, the ATP pocket represents the type I active kinase, DFG-in conformation. Irrespective of the structure, CX-4945 is located at the ATP-binding pocket (Figure S2I–L), and the binding mode is similar to that observed for the active site of the DYRK1A kinase (Figure 2E). The inhibitor is stabilized by three direct hydrogen bonds, water-mediated contacts, and hydrophobic interactions. The carboxyl moiety of the inhibitor contributes with a direct hydrogen-bond interaction with the amide of Asp200 (equivalent to Asp307 in DYRK1A) from the DFG motif (Figure 2E). This interaction mirrors quite closely the case of DYRK1A with the calculated ligand–residue binding energies between −13.1 and −14.7 kcal/mol depending on the selected molecular complex. In GSK3β, there is a hydrogen bond coupled with an ionic bridge to the side chain ammonium of Lys85 (Lys188 in DYRK1A, Figure 2E). Again, the interaction strength goes hand in hand with what was observed for DYRK1A, though in one of the protein–ligand complexes of GSK3β, the residue Lys85 is bridging between the ligand and Glu97. Such a bridging situation, in which two carboxylates compete for different hydrogen atoms of the same ammonium group, weakens the ligand–Lys85 contact by almost 4.0 kcal/mol. In the other complex of the asymmetric unit, the interaction between the ligand’s carboxylate, Glu97 (Glu203 in DYRK1A), and the main chain amide of Phe201 (Phe308 in DYRK1A) is mediated by a molecule of water. Using our in-pocket optimization algorithm on the ligand–residue complex to optimize the hydrogen-bond network indicates that the molecule of water is primarily mediating the interaction between Glu97 (donor) and Phe201 (acceptor).48 The strongest interactions involving CX-4945’s carboxylate are thus expected to involve Asp200 and Lys85, generating an anchoring motif in GSK3β similar to what we observed for DYRK1A. One of the nitrogen atoms of the benzo-naphthyridine group (atom 8) establishes a hydrogen bond with the main chain amide of Val135 (Leu241 in DYRK1A). This interaction seems to be weaker in GSK3β than in DYRK1A: −7.8 kcal/mol instead of −9.6 kcal/mol. Such a gain of 1.8 kcal/mol is to some extent correlated with the increase of the donor–acceptor distance in the hydrogen bond, which results from the natural dynamics of the system. However, differences in the respective side chains (the removal of methylene from leucine-to-valine) contribute with a decrease of the interaction strength by 1.0 kcal/mol. Nevertheless, the weight of the hydrogen bond for this ligand–residue interaction remains identical in both protein complexes ($55\%$). The adjacent main chain carboxyl oxygens (Asp133 and Val135; Glu239 and Leu241 in DYRK1A) establish π-stacking interactions, which, according to our calculations, should stabilize the structure with about −4.8 kcal/mol. We stress that our evaluation is based on the total ligand–residue interactions. Consequently, also in this case, the leucine-to-valine conversion in the pocket affects the total interaction energies. The hydrophobic interactions with the inhibitor mainly involve the following side chains of the kinase (corresponding residues of DYRK1A are indicated in parentheses): Ile62 (Ile165), Ala83 (Ala186), Val70 (Val173), Leu132 (Phe238) of the N-lobe and Leu188 (Leu294) of the C-lobe. Tyr134 participates in a ring-stacking interaction with the inhibitor, reminiscent of a hydrophobic interaction provided by the side chain of Met240 in DYRK1A. Cys199 contributes with a sulfur−π interaction instead of the hydrophobic interaction with the side chain of Val306 at an equivalent site of DYRK1A (Figure 2F). Overall, the sum of all hydrophobic side chain contributions is estimated to be approximately −15.7 kcal/mol. This is lower in GSK3β than in DYRK1A, which may be rationalized by the cysteine residue. Comparing the crystal structures of Tyr216 phosphorylated and non-phosphorylated forms of GSK3β (Figure S2M,N) shows that major differences take place in the activation loop, more specifically, between residues 200 and 226 (Figure 3A). In the structures analyzed, the side chain of Tyr216 is seen in two distinct conformations. When Tyr216 is phosphorylated (pTyr216), the side chain of the residue is stabilized in an anti-conformation, which directs it out of the substrate binding site (Figure 3A). The phosphate moiety of pTyr216 makes interactions with Arg220 and Arg223, which helps in stabilizing the activation loop in the active conformation (Figure 3B). On the other hand, in the non-phosphorylated form of GSK3β, the side chain of Tyr216 is shifted toward the substrate binding groove between the two lobes of the kinase domain, thus adopting a gauche conformation. This directs the group downwards toward the bottom of the peptide binding cleft. The crystal structure for the non-phosphorylated form of GSK3β (crystalized in sodium acetate, imidazole, and disodium malonate) shows two malonate ions in the vicinity of Val214 (Figure 3C). These malonate ions form hydrogen bonds with three residues: Arg96, Arg180, and Lys205 from the substrate binding groove. The intense positive potential generated by the cluster of basic side chains is consequently neutralized. This neutralization of the positive charge and most significantly the interaction with Arg96 from the N-terminal lobe positions the catalytic residues in their active conformation. Calculations on model systems show furthermore that the presence of the malonates is critical for stabilizing the attachment of Tyr216 to the neighborhood of residues Arg96, Arg180, and Lys205. Once more, using in-pocket optimization to optimize the hydrogen-bond network, we observe that the malonates are bridging the tyrosine with the nearby arginine residues. Removing the malonates from the quantum chemical model system to better resemble the cell environment leads to an increase of the interaction energy by 1.9 kcal/mol. This means that the dissociation constant becomes more than 20 times higher. To better understand the phosphorylation of Tyr216, we used in-pocket optimization to generate a hypothetical structure of pTyr216 in the gauche conformation, thus interacting with Val214, Arg96, Arg180, and Lys205. Though, of course, our calculations do not incorporate the conformational changes accompanying the phosphorylation of gauche-fixed Tyr216, we observe a significant difference in the interaction with the neighboring residues in comparison to the anti-conformation. This suggests a strong thermodynamic driving force for flipping Tyr216 as soon as phosphorylation takes place. **Figure 3:** *Activation loop comparison between phosphorylated and non-phosphorylated Tyr216 GSK3β forms. (A) Superimposition of phosphorylated (pTyr216, pink) and non-phosphorylated (Tyr216, teal) GSK3β with the CX-4945 inhibitor. The activation loop Asp200-Glu226 is shown in yellow. The stabilization of the activation loop in phosphorylated GSK3β (B) and non-phosphorylated GSK3β (C) form.* In conclusion, both the phosphorylated and non-phosphorylated forms of GSK3β that we crystallized are in their active conformation. This is either supported by a hydrogen-bond network around pTyr216 or by exogenous oxyanions. The role of Tyr216 phosphorylation in GSK3β function has been uncertain, with contradictory results from in vivo studies.49,50 However, GSK3β kinase activity studies with phospho-primed peptide substrates revealed that Tyr216-phosphorylated GSK3β is only 5-fold more active than the corresponding non-phosphorylated enzyme.51 *This is* a very modest effect in comparison with related kinases, where activation segment tyrosine phosphorylation produces >1000-fold stimulation, suggesting that this particular phosphorylation has a modulatory, rather than a direct regulatory role in GSK3β function. CX-4945 Restores DYRK1A/GSK3β-Mediated Inhibition of the Calcineurin/NFAT Pathway. One of the important cellular targets of DYRK1A and GSK3β is the NFAT (Nuclear factor of activated T-cells) transcription factor, which plays a major role in regulating the cell cycle.52 The calcineurin/NFAT signaling controlled by DYRK1A and GSK3β activity is an important target for neurodegenerative processes and β-cell proliferation. Both kinases inactivate NFAT’s transcription factor by phosphorylation of its nuclear pool, which leads to cytosolic export. We therefore decided to check if treatment with CX-4945 could restore NFAT signaling via inhibition of DYRK1A or GSK3β. To assess the effect of CX-4945 on the calcineurin/NFAT/DYRK1A or GSK3β pathway, we imaged the nuclear translocation of the EGFP-NFAT fusion protein. In unstimulated cells, EGFP-NFAT stayed predominantly in the cytosol, and only after stimulation with ionomycin, the nuclear translocation and accumulation of EGFP-NFAT were observed (Figure 4). Overexpression of any of the kinases blocked the nuclear accumulation of EGFP-NFAT upon cell stimulation, proving the negative regulation of the calcineurin/NFAT pathway by DYRK1A or GSK3β. Treatment with either CX-4945 or harmine could restore the calcineurin/NFAT pathway and lead to the nuclear translocation and accumulation of EGFP-NFAT, despite the presence of overexpressed DYRK1A. Additionally, CX-4945 and 1-azakenpaullone reversed the effect of GSK3β-controlled NFAT cellular localization. These results further reinforce that CX-4945 efficiently inhibits DYRK1A and GSK3β in the cell and can restore the functionality of kinase-affected pathways. **Figure 4:** *Representative fluorescence images of CX-4945’s impact on DYRK1A- and GSK3β-mediated inhibition on NFAT signaling. EGFP-NFATc1 (green) was cotransfected with either a mock vector, DYRK1A (red), or GSK3β (red) into HEK293T cells. Cells pretreated for 3 h with CX-4945 or harmine (5 μM) or DMSO before stimulation for 1 h with ionomycin (IM; 6 μM).* ## Kinase Selectivity Determinants of CX-4945 CX-4945 was previously characterized with respect to its selectivity profile. At a concentration of 500 nM, it affected the activity of 49 from the 235 kinases tested by more than $50\%$.10 However, only 10 kinases were inhibited by more than $90\%$. Four of them, CK2α, CLK3, DYRK2, and HIPK3, belong to the CMGC kinase family. DYRK1A was not included in the kinase panel selected for a single-concentration kinase screen during the initial evaluation, but its close isoform, DYRK2, was shown to be inhibited by more than $95\%$. For GSK3β, a $55\%$ drop in activity was observed at 500 nM.10 Our study clearly demonstrates that CX-4945 directly inhibits DYRK1A and GSK3β at comparable nanomolar concentrations. Interestingly, the inhibitory activity of CX-4945 against DYRK1A was even stronger than that of harmine, an alkaloid obtained from plants and widely used as a selective and potent inhibitor of DYRK1A. The potent inhibitory activity of CX-4945 against CLKs,35 HIPKs, PIM1, and our two kinases, DYRK1A and GSK3β, is supported by structural data that provided detailed information on the binding mode of the inhibitor (Figure 5 and references therein). In all kinase inhibitor crystal structures, CX-4945 is firmly positioned in an ATP-binding pocket. Of the kinases with available crystal structures complexed with CX-4945, only PIM1 does not belong to the CMGC family but to CAMK. The different kinase family memberships are reflected in the binding mode of the inhibitor. In the CMGC family, CX-4945 forms three direct hydrogen bonds with the kinase protein, which involves a catalytic Lys, the amino group of Asp within the DFG motif, and the amino group of Leu or Val within the hinge region. In the PIM1 structure, no binding to the hinge region is observed because, contrary to the CMGC family members, PIM1 contains a Pro insertion in the hinge region. This reduces connections between the inhibitor and the kinase backbone. The anchoring of CX-4945’s carboxylate by a Lys residue is however conserved. Despite the similar binding poses of CX-4945 in these kinases, in particular, the ones in the CMGC family, protein–ligand complexes show different affinities. **Figure 5:** *Binding mode of CX-4945 in ATP-binding pockets of CMGC and CAMK kinases. (A) Overlay of the crystal structures of kinases from CMGC and CAMK families crystalized with CX-4945. CX-4945 bound to DYRK1A (PDB ID 7Z5N, purple), GSK3β (PDB ID 7Z1F, pink), HIPK2 (PDB ID 6P5S,34 ruby), CK2α (PDB ID 3PE1,10 orange), CLK1 (PDB ID 6KHD,35 wheat), and PIM1 (PDB ID 5O11,53 slate). (B) Closeup of the binding mode of CX-4945 to the indicated kinases.* To better understand the origin of such differences, we applied our newly developed semiempirical energy decomposition analysis (EDA),54 which is suitable for in-depth analysis of binding energies. This was used to examine the binding modes of CX-4945 to DYRK1A and GSK3β and compare them to the original target of the inhibitor, CK2α. For better understanding, different contributions to the interaction energy are represented in the form of maps (Figure 6). These include electrostatics, dispersion forces (lipophilicity), and all summed contributions available in our EDA. Additional maps describing implicit solvation effects, exchange-polarization, overlap-repulsion, and charge transfer are presented in Figure S4. Values for each contribution to the binding energy are available in Table S5. **Figure 6:** *Energy decomposition analysis of the complexes between CX-4945 and GSK3β and DYRK1A and CK2α. (A) Maps for electrostatics; (B) lipophilicity maps based on dispersion interactions; (C) total interaction maps; and (D) the magnitude of different contributions to the energy decomposition. For each subfigure, the leftmost part relates to GSK3β, and in the middle, we have DYRK1A and then CK2α. Modeling was based on the PDB files 7Z1G (GSK3β), 7Z5N (DYRK1A), and 3PE1 (CK2α).* The first remark we make is that all interaction maps are very similar. This reflects the subtle differences expected for kinase selectivity.55 Electrostatic maps are strongly dominated by the ligand’s carboxylate group. This is due to the very strong hydrogen-bond interactions. There are additionally minor attractive contributions from aromatic nitrogen atoms. The remaining are untargeted long-range electrostatics. From the many calculations we performed on these three protein–ligand complexes, we verified that the partial charges on the carboxylate group are always well preserved: +0.35 electrons for the carbon and −0.60/–0.66 electrons for the oxygen atoms. Similar values were obtained for HIPK2, CLK1, and PIM1 and for the free ligand. This shows that despite its clear importance for binding, the different affinities all of these kinases show toward CX-4945 cannot possibly be accounted for by the carboxylate group and the respective hydrogen bonds. Finally, the magnitude of electrostatics allows us to infer that GSK3β offers the strongest (global) electrostatic effect over the ligand, followed closely by CK2α. DYRK1A offers a much weaker electrostatic stabilization of the ligand in the pocket. This further attests to our observation from above that electrostatics do not correlate with different binding affinities observed experimentally. Interestingly, the dispersion maps for DYRK1A and GSK3β are indistinguishable (Figure S7 for a direct comparison), though the total lipophilic stabilization differs by almost 7 kcal/mol between proteins (DYRK1A is favored). Dispersion maps do however show quite clearly the delineation of the proteins’ shallow pockets as such close contacts result in stronger lipophilic interactions. In the case of DYRK1A and GSK3β, the benzo-naphthyridine ring is clearly the center of lipophilicity (cf. Table S6), though the strongest interactions per atom are on the chlorophenyl group (chlorine). Interactions with the benzo-naphthyridine are primarily dominated by ring C, in particular the atoms around nitrogen atom 8 (Ile165, Leu241, and Leu294 for DYRK1A; Ala83, Tyr134, and Leu188 for GSK3β). This is followed by ring A (Phe238, Val222, and Val306 in the case of DYRK1A; Cys199, Val110, and Leu132 for GSK3β), and finally by the amino-pyridine part, i.e., ring B. The total contribution of dispersion is also revealing since it follows the trend of experimental affinities. Comparing total dispersion energies against the respective stabilizing effect of each functional group (Tables S6 and S7) reveals that, except for the phenyl group, CK2α and DYRK1A show quite similar behaviors. As London dispersion forces correlate with lipophilicity, we expect to a good extent for CK2α and DYRK1A to show identical lipophilic power. The CK2α and DYRK1A maps differ however in the relative contribution of the chlorophenyl group to the protein–ligand interactions. This is indicative of a very specific interaction that occurs in CK2α and is absent in DYRK1A. We then expect in CK2α an interaction targeting (or targeted by) the chlorophenyl moiety of the ligand. Comparatively, the global lipophilic character of GSK3β is significantly decreased. This is clear in the systematically lower interaction energies between each group and the protein, which is furthermore translated in the significantly lower total dispersion stabilization. Consequently, we conclude that GSK3β offers a weaker lipophilic environment than the other two proteins. We stress furthermore that, consistent with all of these observations, there is a steep jump in the lipophilic contribution per atom in the ligand’s phenyl group when going from GSK3β to DYRK1A and then CK2α (Table S7). The information we extract from the dispersion maps extends however beyond the interactions between the π system of the inhibitor and the protein: it allows us to distinguish the different behavior evidenced by the carboxylate group. Despite its negative formal charge, carboxyls show reasonably soft oxygen atoms due to the delocalized double bond. It is therefore to expect that, contrary to alcohol, the carboxylate is also sensitive to the lipophilic environment of the protein. Though differences are not as astonishing as in the case of the previously discussed phenyl group, the carboxylate group of CX-4945 demonstrates stronger lipophilic affinity to DYRK1A and CK2α than it does for GSK3β. It is very intriguing to note that DYRK1A stabilizes this carboxylate even better by lipophilic interactions than CK2α. This hints at stronger lipophilicity of the protein in the part of the pocket associated with the carboxylate and is in good agreement with the potential targeted interaction between CK2α and the chlorophenyl group. When summed up, the protein–ligand interactions are clearly dominated by the hydrogen bonds, which are identical in all complexes studied (cf. Figure 6C). This is in great agreement with thermodynamic binding data for CK2α bound to CX-4945,56 and it strengthens the anchoring picture built for the Asp and Lys residues conserved in all structures.45 In conclusion, the total interaction maps reinforce the critical role played by the carboxylate group and the nitrogen atom opposite to it in CX-4945’s scaffold (atom 8). Accounting for the different binding affinities requires however looking further into the strength of each interaction. Overall, our energy decompositions reveal that GSK3β and CK2α offer stronger electrostatic environments than DYRK1A. This is visible in Figure 6D and may furthermore be followed in Table S5. Figure S5 stresses the predominance of positively charged (blue) residues in the area around the inhibitor and provides a structural justification for the results of our calculations. Long-range electrostatics however lack the specificity to account for kinase selectivity. Our calculations support that the main forces that distinguish the binding of CX-4945 to GSK3β from binding to CK2α or DYRK1A are the dispersion interactions and the lipophilic environments of the proteins. This however does not account for the 5-fold difference in binding between CK2α and DYRK1A observed in our data. Exploiting the short-range nature of London dispersion forces (RAB–6), we searched for the protein fragments with the strongest dispersion-like interactions with the chlorophenyl group of the inhibitor. Calculations on the first layer of amino acids of the pocket revealed that π-stacking with the main chain is an important factor for all proteins (for residues, see Table S8). In the case of CK2α, there is also a quite prominent T-stack contact with His160 (Figure 7) and the interaction with Leu45. Though the lipophilic contact with the alkyl chain of Leu45 is stronger in CK2α (than the contacts with the Ile residues in GSK3β and DYRK1A), these interactions are not specific to CK2α. Furthermore, these lipophilic contacts are rather unspecific and not directional. On the other hand, the T-stacked interaction between His160 and the ortho-chlorophenyl is exclusive to CK2α (i.e., this interaction is absent in DYRK1A, GSK3β, HIPK2, all 4 CLKs, and PIM1) and it is quite direction-specific. We note that Battistutta and colleagues already detected the unusual orientation of His160 in this protein–ligand complex,10 but they assigned the effect to a water-mediated hydrogen bond between His160 and nitrogen 13 of ring B. Our calculations evidenced the role played by a T-stack interaction between His160 and the chlorophenyl moiety of the ligand. This inspired us to run a series of calculations on model systems built directly from the CK2α-CX-4945 crystal structure to infer which is the dominating contribution to the interaction between the two groups. All of the calculations resulted in a modest stabilization of approximately −1 kcal/mol for the water-mediated hydrogen bond between the imidazole ring and nitrogen 13. This is less than the stabilization achieved by the orientation-specific T-stack between the two aromatic rings (−2.2 kcal/mol). Furthermore, energy decomposition analysis of the complex with CK2α, including the crystal’s explicit waters, retains the exact same shape and weights observed for the interaction maps without explicit water molecules. This is particularly relevant to ascertain the weight attributable to the stabilizing effect of the water-mediated interaction with His160. Nevertheless, all structures of CK2α complexed with the CX-4945 position of the imidazole ring of His160 in a way that promotes the hydrogen-bond interaction.10,56,57 *This is* experimental evidence that the interaction is of importance to the unusual orientation of the side chain of His160. To ascertain the robustness of our calculations, we ran high-level ab initio calculations on the two complexes. Those calculations evidenced that the relative strength of the two interactions is dependent on the medium surrounding the protein–ligand complex. Cautiously, we state that in water, the two interactions are equally strong and that we expect the T-stack to dominate in more polar media. Further tests run on the same systems show furthermore that the T-stack interaction is very close to the optimal interaction mode between the two groups. More details on these two aspects may be found in the Supporting Material. The theoretical data that we collected points to the fact that the picomolar affinity of CX-4945 to CK2α is largely due to His160. The latter may thus be seen as an additional anchoring point of the ligand to the protein. **Figure 7:** *Interactions between His160 and CX-4945 in CK2α, as available from the PDB file 3PE1. Protein carbon atoms are represented in light blue, whereas the inhibitor’s carbon atoms are colored in light brown. Hydrogen bonds are marked in green dashed lines. We selected the anchoring of CX-4945’s carboxylate to Lys68 and the water-mediated hydrogen bond with His160. The orange dashed line marks the T-stack interaction.* ## Discussion Small molecule kinase inhibitors are one of the most pursued goals in drug discovery. CX-4945, known as silmitasertib, was developed by Cylene Pharmaceuticals in 2011 and is one of the most promising drug candidates of this class. It also recently entered phase I/II clinical trials for cholangiocarcinoma58 and multiple myeloma.6 The therapeutic potential of CX-4945 arises due to its well-documented in vitro and in vivo efficiency and is also supported by its desirable pharmacokinetic profile (long half-life, oral bioavailability, limited toxicity).59,60 Here, we show that CX-4945 strongly binds and inhibits DYRK1A and GSK3β. Consequently, both kinases could be considered exploitable therapeutic targets. Since DYRK1A and GSK3β are a pair of priming and processive kinases, which sequentially phosphorylate common substrates, it is difficult to distinguish which kinase is the inhibitor target within the cell. For the assays, we selected Cyclin D1, which is a unique substrate of both kinases but not linked by priming activity. For both kinases, treatment with CX-4945 inhibits Cyclin D1 (Thr286) phosphorylation. Additionally, in the cells treated with CX-4945, we observed the recovery of DYRK1A- and GSK3β-controlled NFAT signaling. This proved that CX-4945’s inhibitory activity against the tested kinases is not limited to biochemical assays, and also CX-4945 is able to impair DYRK1A- and GSK3β-mediated phosphorylation in mammalian cells. The structural evaluation revealed subtle but deciding factors that determine the inhibitor specificity. However, it was possible only with the support of quantum chemical calculations to perform an in-depth analysis that pointed us toward kinase selectivity elements. We believe that this approach can be used broadly to predict and tune kinase inhibitor selectivity. Our calculations indicate that the different affinities of CX-4945 toward GSK3β and DYRK1A are related to the environment offered by each protein. In GSK3β, electrostatics dominate, a reflection not only of the nature of residues in the kinase’s pocket but especially of the overall charge distribution/environment provided by the whole protein. In DYRK1A, lipophilicity is stronger. Our calculations show that this is not a consequence of any residue in particular but rather a feature of the whole pocket. Further strengthening this observation is the increased shape complementarity offered by DYRK1A’s pocket, reflected in the increased REP contribution. This is essential for the strength of lipophilic contacts between the protein and the inhibitor. The calculations show that CX-4945 favors binding to lipophilic environments, which accounts for the relative affinity exhibited toward GSK3β, DYRK1A, and CK2α. This provides further support to observations made by others10 that lipophilicity is a key feature in targeting the ATP-binding pocket of CK2α. Besides two anchoring hydrogen bonds between the ligand and the protein, no other interaction came out as particularly strong from our analysis. Our calculations suggest that if kinase-specific inhibitors are to be developed, then investing in the electrostatic prowess of the inhibitor will be favorable to strengthen binding to GSK3β. On the other hand, further endowing the ligand with lipophilicity is expected to promote binding to DYRK1A. However, it is important that the two key hydrogen-bond interactions remain unaffected. To account for the subnanomolar affinity of CX-4945 to CK2α, we invoke a targeted interaction with a histidine residue exclusive to this protein. Our calculations suggest that the T-stacked interaction of this residue with the chlorophenyl fragment of the inhibitor is quasi-optimal and dominates over water-mediated hydrogen bonds in polar media. Nonetheless, we feel that both interactions are important to create an additional anchor of the ligand to the protein. Also, according to the calculations, further development of the inhibitor should focus on this contact. The interaction maps show that investing in the lipophilic character of the chlorophenyl group should promote the affinity of the ligand to CK2α. The quantum chemical calculations on model systems reveal that the His-chlorophenyl T-stacked interaction is optimal when the Hε1 proton of *His is* positioned in the center of chlorophenyl’s aromatic ring. Though we did not include entropic effects in these calculations, a rough estimate indicates that modifications of CX-4945 that achieve that structural arrangement could improve binding by a factor of 3. Overall, our results further suggest that CX-4945 may be considered as an interesting, proof-of-concept molecule to study both in vitro and in vivo regulation of metabolic pathways dependent on DYRK1A and GSK3β kinase activity. For example, related to diabetes or neurodegenerative diseases. The confirmed safety, high affinity, and reasonable selectivity make this a viable candidate for development up to clinical trials. In addition, future modifications of the CX-4945 scaffold are desirable to achieve higher efficacy and selectivity. We are convinced that our structural and quantum mechanical data will serve as a solid basis for further medicinal chemistry development. ## Conclusions We investigated the affinity of a clinical casein kinase 2 inhibitor, CX-4945, toward the DYRK1A and GSK3β kinases implicated in the biology of several diseases. The results confirmed that CX-4945 strongly binds to DYRK1A and GSK3β kinases with a dissociation constant (Kd) of 1.8 and 37.8 nM while being a very potent inhibitor of both kinases with IC50 in the nanomolar range (160 and 190 nM). Inhibitory activity toward both kinases was not only limited to in vitro assays but also extended to cellular models. The activity of this inhibitor toward other kinases offered the opportunity to study the effect of DYRK1A and GSK3β kinases on the NFAT pathway, allowing us to predict a positive effect toward β-cell expression and diabetes. The crystal structures of CX-4945 complexed with DYRK1A and GSK3β were solved by X-ray crystallography and analyzed using additional quantum chemical models. Extending our analysis to CK2α, we built a quantum chemical selectivity model using our new energy decomposition analysis. Only with the help of extensive quantum chemical calculations could we identify a key element for CK2α’s subnanomolar affinity to CX-4945, which may be exploited in future drug discovery ventures. This is a particular contact between the chlorophenyl group of CX-4945 and His160 of the protein (Figure 7). The methodology we employed is expandable to other kinase selectivity modeling. ## Compound Purity All compounds are >$95\%$ pure by HPLC. Harmine, acquired from Sigma-Aldrich, catalog number 286044, purity of $98\%$; 1-azakenpaullone, acquired from Sigma-Aldrich, catalog number A3734, purity of $96.5\%$; CX-4945, acquired from MedChemExpress, catalog number HY-50855, purity of $99.3\%$. ## Cell Culture The human HEK293T cell line was obtained from the European Collection of Cell Culture. Cells were cultured in minimal DMEM medium (Invitrogen) supplemented with $10\%$ fetal bovine serum (Lonza). Cells were maintained at 37 °C in a humidified atmosphere containing $5\%$ CO2. ## Plasmid Construction The DNA encoding kinase domains of both DYRK1A (126–490) and GSK3β (26–383) with N-FLAG (MDYKDDDDK) and NFATc1 and Cyclin D1 with N-HA (MYPYDVPDYS) tags were synthesized by Genscript and cloned into pcDNA3.1 for eukaryotic cell overexpression. Fluorescent protein fusions were prepared by appending genes encoding mCherry and EGFP using restriction-free cloning.61,62 For bacterial expression, the fragment of the gene encoding kinase domain of DYRK1A (126–490) was PCR amplified and subcloned into pET24a, using a restriction-free method.61,62 The kinase domain of DYRK1A was expressed together with a non-cleavable C-terminal hexahistidine tag. The fragment of the gene encoding kinase domain of GSK3β (26–383) was codon optimized and synthesized by Genscript, and then the gene was cloned into a pET24a expression plasmid. The kinase domain of GSK3β was expressed with a C-terminal hexahistidine tag and proceeded with tobacco etch virus protease cleavage site (ENLYFQ*GHHHHHH). ## Protein Expression and Purification DYRK1A was expressed in E. coli LOBSTR strain (Kerafast) in LB medium supplemented with kanamycin (50 μg/mL) at 17 °C for 16 h. The pellet was resuspended in cold lysis buffer (20 mM HEPES pH 7.5, 500 mM NaCl, $5\%$ glycerol, 15 mM imidazole, and 5 mM 2-mercaptoethanol supplemented with EDTA-free Protease Inhibitor Cocktail (Roche)) and the cells were disintegrated by sonication. Clarified lysate was passed through HisPur Cobalt resin (Thermo Fisher Scientific, Waltham, MA, United States), and the protein of interest was eluted with stepwise increments of imidazole concentration (50–300 mM). The fraction corresponding to DYRK1A was pulled and dialyzed against 20 mM HEPES, pH 7.5, containing 50 mM NaCl and 5 mM 2-Mercaptoethanol. Further purification was obtained by ion-exchange chromatography on a HiTrap Q FF column (Cytiva) followed by size exclusion chromatography on a HiLoad $\frac{16}{600}$ Superdex 75 pg column (Cytiva) in 20 mM HEPES, pH 7.5, containing 150 mM NaCl and 5 mM 2-mercaptoethanol. Purified DYRK1A kinase was flash-frozen in liquid nitrogen and stored at −80 °C for further analysis. The initial expression and purification steps for GSK3β were identical to DYRK1A expression, except that the TB medium was used instead of LB, and the buffers used were of pH 7.2 instead of 7.5. The His tag was removed by TEV protease cleavage during dialysis, subsequent to affinity resin elution. The separated His tag was removed by negative chromatography on HisPur Cobalt resin. Further purification and buffer exchange was obtained on a HiLoad $\frac{16}{600}$ Superdex 75 pg column (Cytiva). Purified GSK3β was flash-frozen in liquid nitrogen and stored at −80 °C for further analysis. For the His-tagged GSK3β, the purification was similar to DYRK1A, except that the buffers used were of pH 7.2 instead of 7.5, and the final buffer for size exclusion chromatography contained 300 mM NaCl instead of 150 mM. Purified His-tagged GSK3β was flash-frozen in liquid nitrogen and stored at −80 °C for further analysis. ## Microscale Thermophoresis and Ratiometric Analysis Human kinases (GSK3β and DYRK1A) were labeled using Monolith His-Tag Labeling Kit RED-tris-NTA 2nd Generation (MO-L018, NanoTemper Technologies) according to the manufacturer’s instructions, and labeled proteins are referred to as targets. The affinity of the dye to the his-tagged proteins was estimated prior to the binding assay according to the manufacturer’s instructions. Minimal target concentrations in the assay were set to a constant concentration of 20 nM for DYRK1A and 62.5 nM for GSK3β-based on the evaluation of the dye-target interaction analysis. In the binding assays, small molecules were used as ligands in three parallel two-fold dilution series, with starting concentrations 5, 5, and 100 μM in the case of harmine, CX-4945, and 1-azakenpaullone for DYRK1A, respectively, and 200, 12.5, and 50 μM in the case of harmine, CX-4945, and 1-azakenpaullone for GSK3β, respectively. Microscale thermophoresis and ratiometric measurements were performed simultaneously on a Monolith X (NanoTemper Technologies), at wavelengths 670 and 650 nm, with medium MST power.63 Dissociation constants (Kd) were analyzed using the MO Control software (NanoTemper Technologies). Data were visualized in the OriginPro 2022 software package.64 ## Cook Activity Assay The inhibitory potency (IC50) of all compounds was determined in the Cook activity assay39,62 in which ADP production was coupled to NADH oxidation by pyruvate kinase and lactate dehydrogenase. The peptide substrates, DYRKtide (RRRFRPASPLRGPPK) for DYRK1A and GYS1 (YRRAAVPPSPSLSRHSSPHQ(pS)EDEEE) for GSK3β, were chemically synthesized by Caslo ApS. The assay mixture contained 100 mM MOPS (pH 6.8), 100 mM KCl, 10 mM MgCl2, 1 mM phosphoenolpyruvate, 1 mM peptide substrate, 1 mM 2-mercaptoethanol, 15 U/mL lactate dehydrogenase with 10 U/mL pyruvate kinase, and 10.7 mM NADH. Seventy-five microliters of the assay mixture was mixed with 10 μL of 2.5 μM kinase and 5 μL of a compound in DMSO with a concentration ranging from 20 nM to 200 μM and incubated for 10 minutes at room temperature. Then, the reaction was started by the simultaneous addition of 10 μL of 1280 μM ATP. The enzyme velocity was measured at 340 nm over a time period of 300 s at room temperature. Control reactions in the absence of the peptide substrate were used to detect ATPase activity for basal concentrations. All measurements were done in triplicate, and IC50 was determined using GraphPad Prism software. ## Dye-Based Thermal-Shift Assay DYRK1A and GSK3β stability in the presence of harmine, CX-4945, and 1-azakenpaullone were analyzed by the proteins’ melting temperature determination using the thermal-shift assay (TSA) as described previously.40 Both proteins (1.5 mg/mL) were incubated with a 1:200 diluted Sypro Orange dye, 20 mM HEPES, 100 mM KCl, 10 mM MgCl2, 1 mM 2-mercaptoethanol (pH 8.0) and compound (10 μM) or DMSO. The fluorescence signal of Sypro Orange was determined as a function of temperature between 5 and 95 °C in increments of 0.5 °C/min (λex 492 nm, λem 610 nm). The melting temperature was calculated as the inflection point of the fluorescence as a function of temperature. Each experiment was carried out in triplicates. ## Protein Crystallization, Data Collection, and Structure Determination For crystallization, DYRK1A was concentrated to 12–15 mg/mL and GSK3β to 6–8 mg/mL. The proteins were incubated overnight with 3–10 molar excess of CX-4945 at 4 °C. The preparation was mixed 1:1 (v/v) with the crystallization solutions. Crystallization experiments were carried out at 20 °C. Crystals appeared within 1–3 days at room temperature. The DYRK1A/CX-4945 complex (PDB ID 7Z5N) was obtained in 0.1 M Tris HCl (pH 7.7) containing 0.1 M lithium sulfate and $40\%$ PEG400. The phosphorylated GSK3β/CX-4945 complex (PDB ID 7Z1F) was obtained in 0.1 M imidazole (pH 6.5) containing 1.0 M sodium acetate trihydrate and 0.2 M and 10 mM yttrium (III) chloride. The non-phosphorylated (Tyr216) GSK3β/CX-4945 complex (PDB ID 7Z1G) was obtained in 0.1 M imidazole (pH 6.5) containing 1.0 M sodium acetate trihydrate and 0.2 M disodium malonate. Crystals were cryoprotected with mother liquor containing $25\%$ glycerol and flash-frozen in liquid nitrogen. The diffraction data were collected at BESSY (Berlin) and ESRF (Grenoble). The diffraction data was indexed and integrated in XDS.65 Data was scaled in AIMLESS66 from the CCP4 software package.67 Following steps were performed in Phenix.68 The structures of DYRK1A and both GSK3β were solved by molecular replacement using PHASER69 and 6EIS and 6GN1, respectively, as search models. Models were refined by interchanging cycles of automated refinement using phenix.refine70 and manual building in Coot.71 Data collection and refinement statistics are summarized in Table S2. Restraints for the inhibitors were created in the GradeServer.72 ## Cyclin D1 Phosphorylation Profile HEK293T cells were seeded in 24-well plates at the density of 2×105 cells/well. Twenty-four hours later, the cells were cotransfected with plasmids encoding HA-Cyclin D1 (0.5 μg) with Flag-DYRK1A or FLAG-GSK3β or an empty vector (0.2 μg) using PEI Prime (Sigma-Aldrich). Forty-eight hours later, the cells were treated with inhibitors CX-4945 (10 μM), harmine (10 μM), 1-azakenpaullone (10 μM), or DMSO for 3 h and lysed in RIPA buffer containing protease inhibitors (Sigma-Aldrich) and phosphatase inhibitors (Calbiochem). Total cell proteins (10 μg) were separated by $12\%$ SDS-PAGE and transferred to the PVDF membrane (ThermoScientific). Proteins were analyzed by Western Blot using the anti-FLAG monoclonal antibody (Sigma-Aldrich; F3165) for DYRK1A and GSK3β detection, the anti-HA monoclonal antibody (Cell Signaling; C29F4) for Cyclin D1, and anti-phospho-Cyclin D1 (Thr286) (Cell Signaling; D29B3) for phospho-Cyclin D1 relevant HRP-conjugated secondary antibodies. ## NFAT Translocation Assay HEK293T cells were grown on a μ-Slide 8 well (IBIDI) to $50\%$–$70\%$ confluency. The plasmids expressing the desired proteins, mCherry-DYRK1A or mCherry-GSK3β and eGFP-NFATc1, were transiently cotransfected for 24 h with PEI Prime (Sigma-Aldrich). Cells were pretreated with inhibitors CX-4945 (5 μM), harmine (5 μM), or DMSO for 3 h and then stimulated with ionomycin (Thermo Fisher Scientific) for 1 h. Cells were washed with 1 mL of PBS, and the nuclei were stained with Hoechst 33258 (ThermoScientific) for 10 min at 37 °C and fixed with $4\%$ paraformaldehyde in phosphate-buffered saline (PBS) for 10 min at 25 °C. Images were collected with a Zeiss Axio Observer 3 fluorescence microscope and analyzed in ZEN Blue edition software. ## Computational Studies Most quantum chemical calculations were performed with the ULYSSES package.73 The method of choice was GFN2-xTB74 together with the ALPB solvation model,75 which we showed in a recent publication to be adequate to accurately account for nonbonded interactions in systems of biological interest.54 See also the Supporting Material for further discussion, taking also into consideration the recent work of Villot and co-workers76 on the suitability of GFN2-xTB when applied to biological systems. When mentioned, electronic populations were estimated using Mulliken population analysis. In the optimization of fragments of the experimental structures, we used our recently developed in-pocket optimization, and the energy decomposition analysis was performed according to our recent method.48,54 The proteins were prepared using Schrödinger’s MAESTRO,77 followed by OPLS3e structural minimization.78 The pockets were all cut manually to understand and ensure their balance with respect to the full protein systems. DLPNO-CCSD(T)79 calculations were run with ORCA 5.0 with libint2 using the default options.80 Calculations were run using the def2-TZVP81 basis set with the resolution of identity.82 ## References 1. Manning G., Whyte D. B., Martinez R., Hunter T., Sudarsanam S.. **The protein kinase complement of the human genome**. *Science* (2002.0) **298** 1912. DOI: 10.1126/SCIENCE.1075762 2. Buljan M., Ciuffa R., van Drogen A., Vichalkovski A., Mehnert M., Rosenberger G., Lee S., Varjosalo M., Pernas L. E., Spegg V., Snijder B., Aebersold R., Gstaiger M.. **Kinase Interaction Network Expands Functional and Disease Roles of Human Kinases**. *Mol. Cell* (2020.0) **79** 504. DOI: 10.1016/J.MOLCEL.2020.07.001 3. Lahiry P., Torkamani A., Schork N. J., Hegele R. A.. **Kinase mutations in human disease: interpreting genotype–phenotype relationships**. *Nat. Rev. Genet.* (2010.0) **11** 60-74. DOI: 10.1038/nrg2707 4. Cohen P., Cross D., Jänne P. A.. **Kinase drug discovery 20 years after imatinib: progress and future directions**. *Nat. Rev. Drug Discovery* (2021.0) **20** 551. DOI: 10.1038/s41573-021-00195-4 5. Attwood M. M., Fabbro D., Sokolov A. V., Knapp S., Schiöth H. B.. **Trends in kinase drug discovery: targets, indications and inhibitor design**. *Nat. Rev. Drug Discovery* (2021.0) **20** 839. DOI: 10.1038/s41573-021-00252-y 6. ClinicalTrials.gov identifier NCT02128282. 7. Siddiqui-Jain A., Drygin D., Streiner N., Chua P., Pierre F., O’Brien S. E., Bliesath J., Omori M., Huser N., Ho C., Proffitt C., Schwaebe M. K., Ryckman D. M., Rice W. G., Anderes K.. **CX-4945, an Orally Bioavailable Selective Inhibitor of Protein Kinase CK2, Inhibits Pro-survival and Angiogenic Signaling and Exhibits Antitumor Efficacy Antitumor Mechanisms of the Selective CK2 Inhibitor CX-4945**. *Cancer Res.* (2010.0) **70** 10288. DOI: 10.1158/0008-5472.CAN-10-1893 8. Bouhaddou M., Memon D., Meyer B., White K. M., Rezelj V. V., Marrero M. C., Polacco B. J., Melnyk J. E., Ulferts S., Kaake R. M., Batra J., Richards A. L., Stevenson E., Gordon D. E., Rojc A., Obernier K., Fabius J. M., Soucheray M., Miorin L., Moreno E., Koh C., Tran Q. D., Hardy A., Robinot R., Vallet T., Nilsson-Payant B. E., Hernandez-Armenta C., Dunham A., Weigang S., Knerr J., Modak M., Quintero D., Zhou Y., Dugourd A., Valdeolivas A., Patil T., Li Q., Hüttenhain R., Cakir M., Muralidharan M., Kim M., Jang G., Tutuncuoglu B., Hiatt J., Guo J. Z., Xu J., Bouhaddou S., Mathy C. J. P., Gaulton A., Manners E. J., Félix E., Shi Y., Goff M., Lim J. K., McBride T., O’Neal M. C., Cai Y., Chang J. C. J., Broadhurst D. J., Klippsten S., De Wit E., Leach A. R., Kortemme T., Shoichet B., Ott M., Saez-Rodriguez J., tenOever B. R., Mullins R., Fischer E. R., Kochs G., Grosse R., García-Sastre A., Vignuzzi M., Johnson J. R., Shokat K. M., Swaney D. L., Beltrao P., Krogan N. J.. **The Global Phosphorylation Landscape of SARS-CoV-2 Infection**. *Cell* (2020.0) **182** 685. DOI: 10.1016/j.cell.2020.06.034 9. ClinicalTrials.gov identifier NCT04668209. 10. Battistutta R., Cozza G., Pierre F., Papinutto E., Lolli G., Sarno S., O’Brien S. E., Siddiqui-Jain A., Haddach M., Anderes K., Ryckman D. M., Meggio F., Pinna L. A.. **Unprecedented selectivity and structural determinants of a new class of protein kinase CK2 inhibitors in clinical trials for the treatment of cancer**. *Biochemistry* (2011.0) **50** 8478. DOI: 10.1021/bi2008382 11. Morphy R., Kay C., Rankovic Z.. **From magic bullets to designed multiple ligands**. *Drug Discovery* (2004.0) **9** 641. DOI: 10.1016/S1359-6446(04)03163-0 12. Roth B. L., Sheffer D. J., Kroeze W. K.. **Magic shotguns versus magic bullets: selectively non-selective drugs for mood disorders and schizophrenia**. *Nat. Rev. Drug Discovery* (2004.0) **3** 353. DOI: 10.1038/nrd1346 13. Youdim M. B. H., Buccafusco J. J.. **Multi-functional drugs for various CNS targets in the treatment of neurodegenerative disorders**. *Trends Pharmacol. Sci.* (2005.0) **26** 27. DOI: 10.1016/J.TIPS.2004.11.007 14. Stotani S., Giordanetto F., Medda F.. **DYRK1A inhibition as potential treatment for Alzheimer’s disease**. *Future Med. Chem.* (2016.0) **8** 681. DOI: 10.4155/FMC-2016-0013 15. Lindberg M. F., Meijer L.. **Dual-Specificity, Tyrosine Phosphorylation-Regulated Kinases (DYRKs) and cdc2-Like Kinases (CLKs) in Human Disease, an Overview**. *Int. J. Mol. Sci.* (2021.0) **22** 6047. DOI: 10.3390/IJMS22116047 16. Pucelik B., Barzowska A., Dąbrowski J. M., Czarna A.. **Diabetic Kinome Inhibitors—A New Opportunity for β-Cells Restoration**. *Int. J. Mol. Sci.* (2021.0) **22** 9083. DOI: 10.3390/IJMS22169083 17. Hergovich A., Lisztwan J., Thoma C. R., Wirbelauer C., Barry R. E., Krek W.. **Priming-Dependent Phosphorylation and Regulation of the Tumor Suppressor pVHL by Glycogen Synthase Kinase 3**. *Mol. Cell. Biol.* (2006.0) **26** 5784. DOI: 10.1128/MCB.00232-06 18. Liu F., Liang Z., Wegiel J., Hwang Y.-W., Iqbal K., Grundke-Iqbal I., Ramakrishna N., Gong C.-X.. **Overexpression of Dyrk1A contributes to neurofibrillary degeneration in Down syndrome**. *FASEB J.* (2008.0) **22** 3224. DOI: 10.1096/FJ.07-104539 19. Beurel E., Grieco S. F., Jope R. S.. **Glycogen synthase kinase-3 (GSK3): regulation, actions, and diseases**. *Pharmacol. Ther.* (2015.0) **148** 114. DOI: 10.1016/J.PHARMTHERA.2014.11.016 20. Cohen P., Frame S.. **The renaissance of GSK3**. *Nat. Rev. Mol. Cell Biol.* (2001.0) **2** 769. DOI: 10.1038/35096075 21. Maqbool M., Hoda N.. **GSK3 Inhibitors in the Therapeutic Development of Diabetes, Cancer and Neurodegeneration: Past, Present and Future**. *Curr. Pharm. Des.* (2017.0) **23** 4332. DOI: 10.2174/1381612823666170714141450 22. Shen W., Taylor B., Jin Q., Nguyen-Tran V., Meeusen S., Zhang Y.-Q., Kamireddy A., Swafford A., Powers A. F., Walker J., Lamb J., Bursalaya B., DiDonato M., Harb G., Qiu M., Filippi C. M., Deaton L., Turk C. N., Suarez-Pinzon W. L., Liu Y., Hao X., Mo T., Yan S., Li J., Herman A. E., Hering B. J., Wu T., Seidel H. M., McNamara P., Glynne R., Laffitte B.. **Inhibition of DYRK1A and GSK3B induces human β-cell proliferation**. *Nat. Commun.* (2015.0) **6**. DOI: 10.1038/NCOMMS9372 23. Barzowska A., Pucelik B., Pustelny K., Matsuda A., Martyniak A., Stępniewski J., Maksymiuk A., Dawidowski D., Rothweiler U., Dulak J., Dubin G., Czarna A.. **DYRK1A Kinase Inhibitors Promote β-Cell Survival and Insulin Homeostasis**. *Cells* (2021.0) **10** 2263. DOI: 10.3390/CELLS10092263 24. Dirice E., Walpita D., Vetere A., Meier B. C., Kahraman S., Hu J., Dančík V., Burns S. M., Gilbert T. J., Olson D. E., Clemons P. A., Kulkarni R. N., Wagner B. K.. **Inhibition of DYRK1A stimulates human β-cell proliferation**. *Diabetes* (2016.0) **65** 1660. DOI: 10.2337/db15-1127 25. Wang P., Alvarez-Perez J.-C., Felsenfeld D. P., Liu H., Sivendran S., Bender A., Kumar A., Sanchez R., Scott D. K., Garcia-Ocaña A., Stewart A. F.. **A high-throughput chemical screen reveals that harmine-mediated inhibition of DYRK1A increases human pancreatic beta cell replication**. *Nat. Med.* (2015.0) **21** 383. DOI: 10.1038/nm.3820 26. Varjosalo M., Keskitalo S., van Drogen A., Nurkkala H., Vichalkovski A., Aebersold R., Gstaiger M.. **The Protein Interaction Landscape of the Human CMGC Kinase Group**. *Cell Rep.* (2013.0) **3** 1306. DOI: 10.1016/j.celrep.2013.03.027 27. Bhat R. V., Andersson U., Andersson S., Knerr L., Bauer U., Sundgren-Andersson A. K.. **The Conundrum of GSK3 Inhibitors: Is it the Dawn of a New Beginning?**. *J. Alzheimer’s Dis.* (2018.0) **64** S547. DOI: 10.3233/JAD-179934 28. Pathak A., Rohilla A., Gupta T., Akhtar M. J., Haider M. R., Sharma K., Haider K., Yar M. S.. **DYRK1A kinase inhibition with emphasis on neurodegeneration: A comprehensive evolution storycum-perspective**. *Eur. J. Med. Chem.* (2018.0) **158** 559. DOI: 10.1016/J.EJMECH.2018.08.093 29. Smith B., Medda F., Gokhale V., Dunckley T., Hulme C.. **Recent Advances in the Design, Synthesis, and Biological Evaluation of Selective DYRK1A Inhibitors: A New Avenue for a Disease Modifying Treatment of Alzheimer’s?**. *ACS Chem. Neurosci.* (2012.0) **3** 857. DOI: 10.1021/CN300094K 30. Dang T., Duan W. Y., Yu B., Tong D. L., Cheng C., Zhang Y. F., Wu W., Ye K., Zhang W. X., Wu M., Wu B. B., An Y., Qiu Z. L., Wu B. L.. **Autism-associated Dyrk1a truncation mutants impair neuronal dendritic and spine growth and interfere with postnatal cortical development**. *Mol. Psychiatry* (2018.0) **23** 747. DOI: 10.1038/MP.2016.253 31. Henderson S. H., Sorrell F., Bennett J., Hanley M. T., Robinson S., Hopkins Navratilova I., Navratilova I. H., Elkins J. M.. **Mining Public Domain Data to Develop Selective DYRK1A Inhibitors**. *ACS Med. Chem. Lett.* (2020.0) **11** 1620. DOI: 10.1021/ACSMEDCHEMLETT.0C00279 32. Kim H., Lee K.-S., Kim A.-K., Choi M., Choi K., Kang M., Chi S.-W., Lee M.-S., Lee J.-S., Lee S.-Y., Song W.-J., Yu K., Cho S.. **A chemical with proven clinical safety rescues Down-syndrome-related phenotypes in through DYRK1A inhibition**. *Dis. Model. Mech.* (2016.0) **9** 839. DOI: 10.1242/DMM.025668 33. Ampofo E., Nalbach L., Menger M. D., Montenarh M., Götz C.. **Protein Kinase CK2—A Putative Target for the Therapy of Diabetes Mellitus?**. *Int. J. Mol. Sci.* (2019.0) **20** 4398. DOI: 10.3390/ijms20184398 34. Agnew C., Liu L., Liu S., Xu W., You L., Yeung W., Kannan N., Jablons D., Jura N.. **The crystal structure of the protein kinase HIPK2 reveals a unique architecture of its CMGC-insert region**. *J. Biol. Chem.* (2019.0) **294** 13545. DOI: 10.1074/JBC.RA119.009725 35. Lee J. Y., Yun J.-S., Kim W.-K., Chun H.-S., Jin H., Cho S., Chang J. H.. **Structural Basis for the Selective Inhibition of Cdc2-Like Kinases by CX-4945**. *Biomed Res. Int.* (2019.0) **2019**. DOI: 10.1155/2019/6125068 36. Göckler N., Jofre G., Papadopoulos C., Soppa U., Tejedor F. J., Becker W.. **Harmine specifically inhibits protein kinase DYRK1A and interferes with neurite formation**. *FEBS J.* (2009.0) **276** 6324. DOI: 10.1111/j.1742-4658.2009.07346.x 37. Kunick C., Lauenroth K., Leost M., Meijer L., Lemcke T.. **1-Azakenpaullone is a selective inhibitor of glycogen synthase kinase-3β**. *Bioorg. Med. Chem. Lett.* (2004.0) **14** 413. DOI: 10.1016/j.bmcl.2003.10.062 38. Klaeger S., Heinzlmeir S., Wilhelm M., Polzer H., Vick B., Koenig P.-A., Reinecke M., Ruprecht B., Petzoldt S., Meng C., Zecha J., Reiter K., Qiao H., Helm D., Koch H., Schoof M., Canevari G., Casale E., Re Depaolini S., Feuchtinger A., Wu Z., Schmidt T., Rueckert L., Becker W., Huenges J., Garz A.-K., Gohlke B.-O., Zolg D. P., Kayser G., Vooder T., Preissner R., Hahne H., Tõnisson N., Kramer K., Götze K., Bassermann F., Schlegl J., Ehrlich H.-C., Aiche S., Walch A., Greif P. A., Schneider S., Felder E. R., Ruland J., Médard G., Jeremias I., Spiekermann K., Kuster B.. **The target landscape of clinical kinase drugs**. *Science* (2017.0) **358**. DOI: 10.1126/science.aan4368 39. Czarna A., Wang J., Zelencova D., Liu Y., Deng X., Choi H. G., Zhang T., Zhou W., Chang J. W., Kildalsen H., Seternes O. M., Gray N. S., Engh R. A., Rothweiler U.. **Novel Scaffolds for Dual Specificity Tyrosine-Phosphorylation-Regulated Kinase (DYRK1A) Inhibitors**. *J. Med. Chem.* (2018.0) **61** 7560. DOI: 10.1021/ACS.JMEDCHEM.7B01847 40. Reinhard L., Mayerhofer H., Geerlof A., Mueller-Dieckmann J., Weiss M. S.. **Optimization of protein buffer cocktails using Thermofluor**. *Acta Crystallogr., Sect. F: Struct. Biol. Commun.* (2013.0) **69** 209. DOI: 10.1107/S1744309112051858 41. Diehl J. A., Cheng M., Roussel M. F., Sherr C. J.. **Glycogen synthase kinase-3β regulates cyclin D1 proteolysis and subcellular localization**. *Genes Dev.* (1998.0) **12** 3499. DOI: 10.1101/GAD.12.22.3499 42. Thompson B. J., Bhansali R., Diebold L., Cook D. E., Stolzenburg L., Casagrande A.-S., Besson T., Leblond B., Désiré L., Malinge S., Cris J. D.. **DYRK1A controls the transition from proliferation to quiescence during lymphoid development by destabilizing Cyclin D3**. *J. Exp. Med.* (2015.0) **212** 953. DOI: 10.1084/JEM.20150002 43. Sutherland C.. **What Are the bona fide GSK3 Substrates?**. *Int. J. Alzheimer’s Dis.* (2011.0) **2011**. DOI: 10.4061/2011/505607 44. Diehl J. A., Zindy F., Sherr C. J.. **Inhibition of cyclin D1 phosphorylation on threonine-286 prevents its rapid degradation via the ubiquitin-proteasome pathway**. *Genes Dev.* (1997.0) **11** 957. DOI: 10.1101/GAD.11.8.957 45. Ogawa Y., Nonaka Y., Goto T., Ohnishi E., Hiramatsu T., Kii I., Yoshida M., Ikura T., Onogi H., Shibuya H., Hosoya T., Ito N., Hagiwara M.. **Development of a novel selective inhibitor of the Down syndrome-related kinase Dyrk1A**. *Nat. Commun.* (2010.0) **1**. DOI: 10.1038/NCOMMS1090 46. Yang J., Eyck L. F. T., Xuong N. H., Taylor S. S.. **Crystal Structure of a cAMP-dependent Protein Kinase Mutant at 1.26 Å: New Insights into the Catalytic Mechanism**. *J. Mol. Biol.* (2004.0) **336** 473. DOI: 10.1016/J.JMB.2003.11.044 47. Eswaran J., Patnaik D., Filippakopoulos P., Knapp S.. **Structure and functional characterization of the atypical human kinase haspin**. *Proc. Natl. Acad. Sci. U.S.A.* (2009.0) **106** 20198. DOI: 10.1073/pnas.0901989106 48. Menezes F., Napolitano V., Fröhlich T., Siebenmorgen T., Rioton S., Olmos J. D. J., Dubin G., Popowicz G. M.. **Quantum Chemistry in a Pocket**. (2022.0) 49. Itoh K., Tang T. L., Neel B. G., Sokol S. Y.. **Specific modulation of ectodermal cell fates in Xenopus embryos by glycogen synthase kinase**. *Development* (1995.0) **121** 3979. DOI: 10.1242/dev.121.12.3979 50. Murai H., Okazaki M., Kikuchi A.. **Tyrosine dephosphorylation of glycogen synthase kinase-3 is involved in its extracellular signal-dependent inactivation**. *FEBS Lett.* (1996.0) **392** 153. DOI: 10.1016/0014-5793(96)00806-x 51. Dajani R., Fraser E., Roe S. M., Yeo M., Good V. M., Thompson V., Dale T. C., Pearl L. H.. **Structural basis for recruitment of glycogen synthase kinase 3beta to the axin-APC scaffold complex**. *EMBO J.* (2003.0) **22** 494. DOI: 10.1093/emboj/cdg068 52. Liu H., Wang K., Chen S., Sun Q., Zhang Y., Chen L., Sun X.. **NFATc1 phosphorylation by DYRK1A increases its protein stability**. *PLoS One* (2017.0) **12**. DOI: 10.1371/JOURNAL.PONE.0172985 53. Bogusz J., Zrubek K., Rembacz K. P., Grudnik P., Golik P., Romanowska M., Wladyka B., Dubin G.. **Structural analysis of PIM1 kinase complexes with ATP-competitive inhibitors**. *Sci. Rep.* (2017.0) **7**. DOI: 10.1038/s41598-017-13557-z 54. Menezes F., Popowicz G. M.. **Understanding the Nature of Interactions in Macromolecules**. (2022.0) 55. Davis M. I., Hunt J. P., Herrgard S., Ciceri P., Wodicka L. M., Pallares G., Hocker M., Treiber D. K., Zarrinkar P. P.. **Comprehensive analysis of kinase inhibitor selectivity**. *Nat. Biotechnol.* (2011.0) **29** 1046. DOI: 10.1038/nbt.1990 56. Ferguson A. D., Sheth P. R., Basso A. D., Paliwal S., Gray K., Fischmann T. O., Le H. V.. **Structural basis of CX-4945 binding to human protein kinase CK2**. *FEBS Lett.* (2011.0) **585** 104. DOI: 10.1016/j.febslet.2010.11.019 57. Lindenblatt D., Nickelsen A., Applegate V. M., Hochscherf J., Witulski B., Bouaziz Z., Marminon C., Bretner M., Borgne M. L., Jose J., Niefind K.. **Diacritic Binding of an Indenoindole Inhibitor by CK2α Paralogs Explored by a Reliable Path to Atomic Resolution CK2α′ Structures**. *ACS Omega* (2019.0) **4** 5471. DOI: 10.1021/acsomega.8b03415 58. Zakharia K., Miyabe K., Wang Y., Wu D., Moser C. D., Borad M. J., Roberts L. R.. **Preclinical In Vitro and In Vivo Evidence of an Antitumor Effect of CX-4945, a Casein Kinase II Inhibitor, in Cholangiocarcinoma**. *Transl. Oncol.* (2019.0) **12** 143. DOI: 10.1016/J.TRANON.2018.09.005 59. Pierre F., Chua P. C., O’Brien S. E., Siddiqui-Jain A., Bourbon P., Haddach M., Michaux J., Nagasawa J., Schwaebe M. K., Stefan E., Vialettes A., Whitten J. P., Chen T. K., Darjania L., Stansfield R., Anderes K., Bliesath J., Drygin D., Ho C., Omori M., Proffitt C., Streiner N., Trent K., Rice W. G., Ryckman D. M.. **Discovery and SAR of 5-(3-Chlorophenylamino)benzo[c][2,6]naphthyridine-8- carboxylic Acid (CX-4945), the first clinical stage inhibitor of protein kinase CK2 for the Treatment of Cancer**. *J. Med. Chem.* (2011.0) **54** 635. DOI: 10.1021/jm101251q 60. Son Y. H., Song J. S., Kim S. H., Kim J.. **Pharmacokinetic characterization of CK2 inhibitor CX-4945**. *Arch. Pharm. Res.* (2013.0) **36** 840-845. DOI: 10.1007/s12272-013-0103-9 61. Bond S. R., Naus C. C.. **RF-Cloning.org: an online tool for the design of restriction-free cloning projects**. *Nucleic Acids Res.* (2012.0) **40** W209. DOI: 10.1093/NAR/GKS396 62. Cook P. F., Neville M. E., Vrana K. E., Hartl F. T., Roskoski R.. **Adenosine Cyclic 3′,5′-Monophosphate Dependent Protein Kinase: Kinetic Mechanism for the Bovine Skeletal Muscle Catalytic Subunit**. *Biochemistry* (1982.0) **21** 5794. DOI: 10.1021/BI00266A011 63. López-Méndez B., Baron B., Brautigam C. A., Jowitt T. A., Knauer S. H., Uebel S., Williams M. A., Sedivy A., Abian O., Abreu C., Adamczyk M., Bal W., Berger S., Buell A. K., Carolis C., Daviter T., Fish A., Garcia-Alai M., Guenther C., Hamacek J., Holková J., Houser J., Johnson C., Kelly S., Leech A., Mas C., Matulis D., McLaughlin S. H., Montserret R., Nasreddine R., Nehmé R., Nguyen Q., Ortega-Alarcón D., Perez K., Pirc K., Piszczek G., Podobnik M., Rodrigo N., Rokov-Plavec J., Schaefer S., Sharpe T., Southall J., Staunton D., Tavares P., Vanek O., Weyand M., Wu D.. **Reproducibility and accuracy of microscale thermophoresis in the NanoTemper Monolith: a multi laboratory benchmark study**. *Eur. Biophys. J.* (2021.0) **50** 411. DOI: 10.1007/s00249-021-01532-6 64. Origin(Pro), version 2022; OriginLab Corporation: Northampton, MA, USA.. *Origin(Pro)* 65. Kabsch W.. **XDS**. *Acta Crystallogr., Sect. D: Struct. Biol.* (2010.0) **66** 125. DOI: 10.1107/S0907444909047337 66. Evans P. R., Murshudov G. N.. **How good are my data and what is the resolution?**. *Acta Crystallogr., Sect. D: Struct. Biol.* (2013.0) **69** 1204. DOI: 10.1107/S0907444913000061 67. Winn M. D., Ballard C. C., Cowtan K. D., Dodson E. J., Emsley P., Evans P. R., Keegan R. M., Krissinel E. B., Leslie A. G. W., McCoy A., McNicholas S. J., Murshudov G. N., Pannu N. S., Potterton E. A., Powell H. R., Read R. J., Vagin A., Wilson K. S.. **Overview of the CCP4 suite and current developments**. *Acta Crystallogr., Sect. D: Struct. Biol.* (2011.0) **67** 235. DOI: 10.1107/S0907444910045749 68. Liebschner D., Afonine P. V., Baker M. L., Bunkóczi G., Chen V. B., Croll T. I., Hintze B., Hung L.-W., Jain S., McCoy A. J., Moriarty N. W., Oeffner R. D., Poon B. K., Prisant M. G., Read R. J., Richardson J. S., Richardson D. C., Sammito M. D., Sobolev O. V., Stockwell D. H., Terwilliger T. C., Urzhumtsev A. G., Videau L. L., Williams C. J., Adams P. D.. **Macromolecular structure determination using X-rays, neutrons and electrons: recent developments in Phenix**. *Acta Crystallogr., Sect. D: Struct. Biol.* (2019.0) **75** 861. DOI: 10.1107/S2059798319011471 69. McCoy A. J., Grosse-Kunstleve R. W., Adams P. D., Winn M. D., Storoni L. C., Read R. J.. **Phaser crystallographic software**. *J. Appl. Crystallogr.* (2007.0) **40** 658-674. DOI: 10.1107/S0021889807021206 70. Afonine P. V., Grosse-Kunstleve R. W., Echols N., Headd J. J., Moriarty N. W., Mustyakimov M., Terwilliger T. C., Urzhumtsev A., Zwart P. H., Adams P. D.. **Towards automated crystallographic structure refinement with phenix.refine**. *Acta Crystallogr., Sect. D: Struct. Biol.* (2012.0) **68** 352. DOI: 10.1107/S0907444912001308 71. Emsley P., Cowtan K.. **Coot: model-building tools for molecular graphics**. *Acta Crystallogr., Sect. D: Struct. Biol.* (2004.0) **60** 2126. DOI: 10.1107/S0907444904019158 72. Smart O. S., Womack T. O., Sharff A., Flensburg C., Keller P., Paciorek W., Vonrhein C., Bricogne G.. *Grade, version 1.2.20* (2011.0) 73. Menezes F., Popowicz G. M.. **ULYSSES: An Efficient and Easy to Use Semiempirical Library for C++**. *J. Chem. Inf. Model.* (2022.0) **62** 3685. DOI: 10.1021/acs.jcim.2c00757 74. Bannwarth C., Ehlert S., Grimme S.. **GFN2-xTB - An Accurate and Broadly Parametrized Self-Consistent Tight-Binding Quantum Chemical Method with Multipole Electrostatics and Density-Dependent Dispersion Contributions**. *J. Chem. Theory Comput.* (2019.0) **15** 1652. DOI: 10.1021/acs.jctc.8b01176 75. Ehlert S., Stahn M., Spicher S., Grimme S.. **Robust and Efficient Implicit Solvation Model for Fast Semiempirical Methods**. *J. Chem. Theory Comput.* (2021.0) **17** 4250. DOI: 10.1021/acs.jctc.1c00471 76. Villot C., Ballesteros F., Wang D., Lao K. U.. **Coupled Cluster Benchmarking of Large Noncovalent Complexes in L7 and S12L as Well as the C60 Dimer, DNA–Ellipticine, and HIV–Indinavir**. *J. Phys. Chem. A* (2022.0) **126** 4326. DOI: 10.1021/acs.jpca.2c01421 77. MAESTRO Schrödinger Release 2022-3; LLC: New York, NY, 2021.. *MAESTRO Schrödinger Release 2022-3* (2021.0) 78. Roos K., Wu C., Damm W., Reboul M., Stevenson J. M., Lu C., Dahlgren M. K., Mondal S., Chen W., Wang L., Abel R., Friesner R. A., Harder E. D.. **OPLS3e: Extending Force Field Coverage for Drug-Like Small Molecules**. *J. Chem. Theory Comput.* (2019.0) **15** 1863-1874. DOI: 10.1021/acs.jctc.8b01026 79. Riplinger C., Pinski P., Becker U., Valeev E. F., Neese F.. **Sparse maps - A systematic infrastructure for reduced-scaling electronic structure methods. II. Linear scaling domain based pair natural orbital coupled cluster theory**. *J. Chem. Phys.* (2016.0) **144**. DOI: 10.1063/1.4939030 80. Neese F., Valeev E. F.. **Software update: The ORCA program system - Version 5.0**. *WIREs Comput. Mol. Sci.* (2022.0) **12**. DOI: 10.1002/wcms.1606 81. Weigend F., Ahlrichs R.. **Balanced basis sets of split valence, triple zeta valence and quadruple zeta valence quality for H to Rn: Design and assessment of accuracy**. *Phys. Chem. Chem. Phys.* (2005.0) **7** 3297. DOI: 10.1039/B508541A 82. Weigend F., Hellweg A., Hattig C., Hofener S., Klopper W.. **Accurate Coulomb-fitting basis sets for H to Rn**. *Phys. Chem. Chem. Phys.* (2006.0) **8** 1057. DOI: 10.1039/B515623H
--- title: Resequencing of a Pekin duck breeding population provides insights into the genomic response to short-term artificial selection authors: - Simeng Yu - Zihua Liu - Ming Li - Dongke Zhou - Ping Hua - Hong Cheng - Wenlei Fan - Yaxi Xu - Dapeng Liu - Suyun Liang - Yunsheng Zhang - Ming Xie - Jing Tang - Yu Jiang - Shuisheng Hou - Zhengkui Zhou journal: GigaScience year: 2023 pmcid: PMC10041536 doi: 10.1093/gigascience/giad016 license: CC BY 4.0 --- # Resequencing of a Pekin duck breeding population provides insights into the genomic response to short-term artificial selection ## Abstract ### Background Short-term, intense artificial selection drives fast phenotypic changes in domestic animals and leaves imprints on their genomes. However, the genetic basis of this selection response is poorly understood. To better address this, we employed the Pekin duck Z2 pure line, in which the breast muscle weight was increased nearly 3-fold after 10 generations of breeding. We denovo assembled a high-quality reference genome of a female Pekin duck of this line (GCA_003850225.1) and identified 8.60 million genetic variants in 119 individuals among 10 generations of the breeding population. ### Results We identified 53 selected regions between the first and tenth generations, and $93.8\%$ of the identified variations were enriched in regulatory and noncoding regions. Integrating the selection signatures and genome-wide association approach, we found that 2 regions covering 0.36 Mb containing UTP25 and FBRSL1 were most likely to contribute to breast muscle weight improvement. The major allele frequencies of these 2 loci increased gradually with each generation following the same trend. Additionally, we found that a copy number variation region containing the entire EXOC4 gene could explain $1.9\%$ of the variance in breast muscle weight, indicating that the nervous system may play a role in economic trait improvement. ### Conclusions Our study not only provides insights into genomic dynamics under intense artificial selection but also provides resources for genomics-enabled improvements in duck breeding. ## Background Artificial selection experiments can help to understand the mechanism that allows populations to adapt to strong selection pressure [1, 2]. In combination with population-level genome sequencing, attempts have been made to identify alleles whose frequencies change systematically during selection experiments [3–5]. However, most studies were focused on lower organisms and were retrospective or used a single time point to characterize the dynamic changes in allele frequencies [1, 6–8]. Selecting a suitable domestic animal model for continuous high-intensity artificial selection trials may improve our understanding of the genetic basis of complex traits in domestic animals. Studies of 2 representative bidirectional selected resource populations in which chicken body weight [9–13] and abdominal fat [14–18] were selected have shown that continuous artificial selection results in obvious phenotypic and genomic differentiation in poultry. Artificial selection signals have also been identified in animals, such as rabbits [19], sheep [20], goats [21, 22], and pigs [23]. However, many domestic animal populations established for selection experiments are not as complete as chicken populations because of the long generation interval and sequencing costs in these species. Ducks (Anas platyrhynchos) (NCBI:txid8839) are among the most economically important waterfowl; they can provide meat, eggs, and down for humans and show important characteristics, such as a short generation interval, high reproductive ability, and a long but traceable history of artificial selection [24]. Moreover, ducks, like other birds, have smaller genomes than nonavian terrestrial vertebrates. These characteristics make the duck an effective model for studying genomic footprints of artificial selection. High-quality reference genomes are the foundation for genetic research and molecular marker breeding, which can support innovations in sustainable animal production [25, 26]. To date, reference genomes have been published for a variety of domestic animals such as ducks, chickens, pigs, cattle, sheep, and goats [24, 27–31], providing important resources for livestock and poultry genetic breeding. To completely capture the variation present in the genome of our study population, we chose a female individual from our population for PacBio long-read sequencing and genome assembly to obtain a high-quality chromosome-level reference genome of Pekin duck (GCA_003850225.1) (Fig. 1A). **Figure 1::** *Overview of the assembly quality and characteristics of the Pekin duck genome. (A) Circular diagram depicting the characteristics of the GCA_003850225.1 assembly. The tracks from the outer to inner circles represent the following: chromosomes, gene density (window size of 200 kb), SNP density (window size of 200 kb), TE density (window size of 200 kb), CNV density (window size of 3 Mb), and GC content (%) (window size of 200 kb). (B, C) Tree maps of fragmentation differences between long-read (GCA_003850225.1) and short-read (BGI_duck1.0) Pekin duck genome assemblies. The size of each rectangle of each chromosome is scaled to that of the contig sequence. The larger and fewer the internal boxes are, the more contiguous the contigs. (D) Comparison of the sequence contig length distribution between long-read (GCA_003850225.1) and short-read (BGI_duck_1.0) Pekin duck genome assemblies. (E) Comparison of mapping rates when the Pekin duck population (119 birds from 10 generations) whole-genome resequencing data were mapped to GCA_003850225.1 and the short-read genome assembly (BGI_duck_1.0). A 2-tailed paired t test was used for statistical assessment. ****P < 0.0001.* Herein, we utilized a Pekin duck pure line selected for breast muscle weight for 10 generations. The foundation stock was the local conserved population in Beijing, China. After 10 generations of high-intensity artificial selection, the breast muscle weight of the Pekin duck Z2 line increased from 80 g to 220 g. These data were investigated to elucidate the dynamic response patterns in the Pekin duck genome under artificial selection. ## Data Description To understand the genetic basis of short-term, intense artificial selection, we utilized a Pekin duck pure line selected for breast muscle weight at 6 weeks of age. We implemented the whole-genome resequencing of 119 ducks across 10 generations (1 generation per year, 2005–2014) and mapped the resequencing data to the high-quality reference genomes assembled in this study. Using 8.60 million of single-nucleotide polymorphisms (SNPs), we tested for dynamic changes in population structure under short-term intense selection, and we assessed the genome for signatures of short-term intense selection associated with breast muscle weight. ## An improved Pekin duck genome assembly To carry out the de novo assembly of the Pekin duck genome, we adopted a combination of PacBio long-read sequencing, BioNano optical mapping, and Hi-C technologies. We first generated 65.9 Gb of PacBio long reads with 50× genome coverage, 49 Gb of BioNano high-quality reads with 41× genome coverage, and 106 Gb of Hi-C reads with 82× genome coverage (Supplementary Table S1). Then, these data were used to assemble the new duck genome (Fig. 1A). Assembly was performed in a stepwise fashion (Supplementary Fig. S1), to generate assemblies with improvements for each process. First, we used the initial PacBio subreads to construct 2,682 contigs, yielding a contig N50 of 4.17 Mb (Table 1). Second, we used optical mapping (BioNano Genomics Irys) data to link the PacBio contigs into scaffolds, resulting in the population of 1,788 scaffolds, and the N50 was 6.24 Mb (Table 1). Then, the Hi-C data (Supplementary Table S2) were used to cluster the scaffolds at the chromosomal scale, resulting in 1,852 scaffolds (Supplementary Fig. S2). The final assembled genome length was 1.12 Gb, with scaffold N50 and contig N50 values of 76.13 Mb and 4.10 Mb, respectively (Table 1). **Table 1:** | Assembly | Number of scaffolds | Scaffold N50 (Mb) | Genome size | | --- | --- | --- | --- | | PacBio | ─ | ─ | 1154570803 | | PacBio + BioNano | 1788 | 6.24 | 1153069495 | | PacBio + BioNano + Hi-C | 1852 | 41.14 | 1134893859 | | GCA_003850225.1 | 1330 | 76.13 | 1134894103 | We next evaluated the quality of the new assembly. BUSCO (RRID:SCR 015 008) [32] (v5.1.2) assessments of the new assembly revealed $93.3\%$ completeness (Supplementary Fig. S3). The continuity of our new assembly yielded a 62-fold improvement compared to that of BGI_duck_1.0 (76.13 vs. 1.23 Mb) (Supplementary Table S3). GCA_003850225.1 (IASCAAS_PekinDuck_PBH1.5) had fewer gaps than BGI_duck_1.0 ($0.26\%$ vs. $3.17\%$), which also indicated that our new assembly presented a higher level of integrity (Fig. 1B-D, Supplementary Table S3). After the annotation of the newly assembled reference genome, we obtained 22,079 annotated genes, representing an increase of $34.22\%$ relative to BGI_duck_1.0. All of the above results indicated that the quality of our newly assembled Pekin duck genome was greatly improved compared with that of the BGI_duck_1.0 assembly. Furthermore, we mapped the resequencing data of 119 Pekin duck individuals (Supplementary Table S4) to BGI_duck_1.0 and GCA_003850225.1, and the mapping rate was significantly improved (Fig. 1E) (t test, $P \leq 2.22$e-16) when using the new assembly. To track the dynamic change process in the Pekin duck genome driven by intense selection, we conducted the resequencing of 30 individuals (15 males, 15 females) from G1 to G10 in intervals of 3 generations. We mapped all of the paired-end reads to the GCA_003850225.1 assembly with an average coverage rate of $94.77\%$ and an average depth of 7.08× (6.26–8.01×) (Supplementary Table S4). These sequencing data enabled us to identify a total of 8.60 million SNPs and 81 high-confidence copy number variation regions (CNVRs) among the 119 individuals. ## Breeding process in the Pekin duck Z2 line Breast muscle weight was calculated according to the following equations. First, breast muscle volume (BMV) = BB × KL × BMT, and BMW = 0.6228 × BMV + 17.042 [33], where BB is breast breadth, KL is keel length, and BMT is breast muscle thickness. Herein, we used a Vernier caliper to measure BB and KL, while BMT was measured with B-ultrasound scanning technology (Fig. 2A). BMW was then estimated according to the BMV derived from the BB, KL, and BMT values. The statistical results showed that after 10 generations of high-intensity artificial selection, the breast muscle weight of Pekin duck increased from 80 to 220 g (Fig. 2B, Supplementary Fig. S4). **Figure 2::** *Phenotypic and population genetic structure variation over 10 generations. (A) Measurement of duck breast muscle volume (BMV) in vivo at 6 weeks of age. The 3 measured values were breast breadth (BB), keel length (KL), and breast muscle thickness (BMT). The weight of breast muscle was calculated based on the following formulas: BMV = BB × KL × BMT, and BMW = 0.6228 × BMV + 17.042. (B) Changes in breast muscle weight in the Pekin duck Z2 line over 10 generations. (C) Principal component analysis (PCA) of 10 generations. The red circles represent first-generation (G1) individuals, the yellow circles represent fourth-generation (G4) individuals, the green circles represent seventh-generation (G7) individuals, and the blue circles represent tenth-generation (G10) individuals.* ## Dynamic changes in population structure under short-term intense selection To examine genetic differentiation at the whole genome level in 10 generations, we performed principal component analysis (PCA) [34] using the whole-genome SNP data of G1, G4, G7, and G10. Individuals across generations separated along the 2 principal component dimensions, and ducks between the G1 and G10 generations could be clearly separated into 2 clusters (Fig. 2C). In the G10 generation, population diversity showed a decreasing trend (Fig. 2C, bottom right). However, there were no significant differences in the proportions of nonsynonymous mutations and synonymous mutations (dN/dS) in the coding region between the 4 generations (Supplementary Table S5). The dN/dS ratio of all generations was below 1, indicating that the Pekin duck population was subjected to continuous negative selection. To identify the dynamic changes in duck genomic variations, we calculated the allele frequency difference (ΔAF) between the G1 and G10 generations for each SNP and sorted these values into $5\%$ bins (ΔAF = 0 to 0.05, etc.). We evaluated the enrichment of the SNPs in each bin in exons, introns, and untranslated regions (UTRs) to illustrate the numbers and distributions of sites that played a major role during artificial selection in the sequenced genomes. We observed a large number of allele frequency shifts in the entire data set, but no SNPs with ΔAF > 0.55 were identified (Fig. 3), implying that the directional selection event related to the breast muscle weight of Pekin ducks was in accord with polygenetic and soft selective sweep patterns [35]. We found significant enrichments of high ΔAF SNPs (ΔAF > 0.3) in both UTRs and introns (χ2 test, $P \leq 0.05$), whereas in exons, the excess was only 2 SNPs (Fig. 3, Supplementary Table S6). We found that exonic SNPs tended to be significantly enriched in bins with ΔAF < 0.1 (Supplementary Table S6). Therefore, changes in noncoding regions played an important role during the breeding process in the Pekin duck Z2 line. **Figure 3::** *Allele frequency difference (ΔAF) analyses. The majority of SNPs showed low ΔAF values between the first- and tenth-generation ducks. The black line indicates the number of SNPs in nonoverlapping ΔAF bins (left y-axis). The colored lines denote the M values (log2-fold changes) of the relative frequencies of SNPs in coding regions (yellow), UTRs (green), and introns (blue), according to ΔAF bins (right y-axis).* ## Identification of signals under artificial selection We employed a joint analysis strategy to calculate fixation index (Fst) [36] values and cross-population extended haplotype homozygosity (XP-EHH) [37] values (10-kb window, 5-kb step) to identify potential selected regions between the G1 and G10 populations (Fig. 4A and B, Supplementary Fig. S5). Using the empirical quantiles of the top $1\%$ of SNPs and taking the intersection of the Fst (Fst > 0.09) and XP-EHH (XP-EHH >1.47, XP-EHH <−1.48) analysis values, we identified a total of 187 regions as potentially containing selective signals (Supplementary Table S7). Among these candidate regions, 59 genes were found across ∼1.22 Mb (Supplementary Table S8). In addition, the results illustrated that the allele frequency of the potential selected regions identified by different approaches showed an upward trend over the 10 generations (Supplementary Fig. S6A and B). *The* genetic diversity of the potential selected regions exhibited the opposite change trend among different generations (Supplementary Fig. S7). To exclude the potential selected regions that may be due to genetic drift, we then employed a genome-wide association study (GWAS) to identify the overlapping selection signatures associated with breast muscle weight. We applied a Bonferroni threshold of a −log10 $P \leq 8.94$ value for outliers to identify selected SNPs associated with breast muscle weight, and a total of 22 SNPs reached the association analysis threshold (Fig. 4C, Supplementary Table S9). The allele frequencies of the 22 SNPs increased gradually over 10 generations (Fig. 4D, Supplementary Fig. S6C), indicating that these loci were subjected to continuous selection. After overlapping the associated SNPs with candidate selected regions, we identified 2 regions associated with breast muscle weight. These signals were located on chromosomes 3 and 16 (Chr3: 0.18–0.44 Mb; Chr16: 3.40–3.50 Mb). *The* genetic diversity of the 2 selected regions was significantly different between generations (Fig. 4E). **Figure 4::** *Overlapping selection signals in the genomes of the first and tenth generations. (A) Manhattan plot of selected regions between the first generation (G1) and the tenth generation (G10). Pairwise fixation index values (Fst) are calculated in 10-kb sliding windows and 5-kb steps. The significance threshold for Fst is 0.103 (1%). (B) Manhattan plot of selected regions between the first generation (G1) and the tenth generation (G10). Cross-population extended haplotype homozygosity (XP-EHH) values were calculated in 10-kb sliding windows and 5-kb steps. The significance thresholds for XP-EHH were 1.475 and −1.480 (1%). (C) Manhattan plots for a GWAS of breast muscle weight. The gray horizontal dashed lines indicate the Bonferroni-corrected significance threshold of the GWAS (−log10  P = 8.94), and the selection signals are indicated with a gray background. (D) Allele frequency trajectories of 22 SNPs. (E) The variation trend of genetic diversity in selected regions among generations (Chr3:0.18–0.44 Mb; Chr16:3.40–3.50 Mb). (F) Regions containing loci associated with breast muscle weight ranging from 0.18 to 0.44 Mb along chromosome 3 and 3.40 to 3.50 Mb along chromosome 16. All genotyped SNPs are color coded according to their pairwise LD, with the leader SNP (Chr3:416,692; Chr16:3,467,244) calculated by comparing the first generation (G1) and tenth-generation (G10) populations. SNPs are colored based on the strength of the LD values (r2 values) considering the most strongly associated SNP and the other SNPs in the region. (G) The blue line diagrams refer to the fixation indexes (Fst) on selected regions (Chr3:0.18–0.44 Mb; Chr16:3.40–3.50 Mb) between G1 and G10. Fst values are calculated in 10-kb sliding windows in 5-kb steps. The selection signals that overlapped with characterized GWAS loci are indicated with a gray background. The green line diagrams refer to XP-EHH for selected regions (Chr3:0.18–0.44 Mb; Chr16:3.40–3.50 Mb) between G1 and G10. XP-EHH values are calculated in 10-kb sliding windows in 5-kb steps. The selection signals that overlapped with characterized GWAS loci are indicated with a gray-blue background. (H) Associations between genotypes of 2 leader SNPs in candidate regions and breast muscle weight. Boxplots indicate the median (centerline), 25th to 75th percentiles (limits), and minimum and maximum values (whiskers). The indicated P values are based on 1-way ANOVA. ****P < 0.0001, ***P < 0.001, **P < 0.01, and ns indicates that the P value was not significant. (I) Schematic diagram showing the genes distributed within the candidate regions (Chr3:0.34–0.43 Mb and Chr16:3.46–3.49 Mb).* To accurately detect the genomic footprints left by artificial selection, we examined the linkage disequilibrium (LD, expressed as r2) of the top SNPs (Chr3:416,692 bp; Chr16:3,467,244 bp) and surrounding SNPs within these candidate regions. Then, we detected corrected signals (r2 > 0.4) in the 0.34 to 0.43 Mb region of chromosome 3 and the 3.46 to 3.49 Mb region of chromosome 16, which were associated with breast muscle weight (Fig. 4F and G). The extent of LD between the variants within the candidate regions and the lead SNPs increased gradually over 10 generations (Supplementary Fig. S8). These regions spanned 120 kb and contained 2,072 SNPs. Notably, we found that a large proportion ($93.87\%$, 1,$\frac{945}{2}$,072) of the SNPs in the candidate regions were located in noncoding regions (Supplementary Fig. S9). Thus, continuous breeding for breast muscle weight in Pekin ducks has a polygenic basis with many loci responding to continuous artificial selection, and noncoding sequences may play an important role in Pekin duck breast muscle weight improvement. Genotyping the ducks using the lead SNP located at Chr3:416,692 (T>C) revealed that individuals carrying the variant T alleles exhibited heavier breast muscles (Fig. 4H). The top SNP on chromosome 16 (Chr16:3,467,244 T>G) showed the same trend (Fig. 4H). In addition, 2 genes (UTP25 and FBRSL1) were identified in the 2 putative selected regions (Fig. 4I). Combined analysis with the global transcriptomic data of ducks revealed that UTP25 and FBRSL1 were widely expressed in various tissues of Pekin ducks (Supplementary Fig. S10). We then tracked the expression levels of UTP25 and FBRSL1 genes in the breast muscles of different developmental periods of the Pekin duck Z2 line. The results illustrated that the expression level of FBRSL1 decreased with the increase of days, while UTP25 was continuously expressed in breast muscle (Supplementary Fig. S11). ## Identification of CNVRs under artificial selection Copy number variations (CNVs) show higher mutational rates than SNPs [38], typically involve larger genomic regions, and potentially affect a wide range of phenotypic traits [39–41]. Based on our new assembly, we obtained 81 CNVRs with high credibility and accuracy on autosomes (Fig. 5A, Supplementary Table S10). In total, we identified 2 duplicated CNVRs and 1 deletion with significant allele frequency changes between the G1 and G10 populations (Fig. 5A and B). The CNVRs detected in most individuals were duplications located on chromosome 1 at 200.83 to 201.27 Mb (CNV1) and chromosome 2 at 123.14 to 123.16 Mb (CNV2) (Fig. 5B), and these 2 CNVRs were annotated in 2 genes, EXOC4 and TRPA1 (Fig. 5C). In addition, the CNVR identified in most individuals ($78\%$) was a homozygous copy number loss or hemizygous copy number loss on chromosome 4 at 54.81 to 54.82 Mb (CNV3) (Supplementary Table S11), and the allele frequency of this CNVR increased gradually over the 10 generations (Fig. 5D). **Figure 5::** *Genome-wide screening of selected copy number variations (CNVs) between the first and tenth generations. (A) The relative frequency difference (RFD) between the first generation (G1) and the tenth generation (G10) is plotted against the position on each of the autosomes. The 2 horizontal dashed lines indicate the genome-wide thresholds of selection signals, which showed the highest absolute RFD value of 5% (>4.1). (B) Examples of copy number variant form of selected CNVs. The abscissa represents the selected copy number variation region, and the ordinates on both sides represent standardized absolute copy numbers (0, homozygous deletion; 1, normal copy number, i.e., normal diploid; 0.5, loss of heterozygosity; 1.5, heterozygous repetition; 2, homozygous repetition; more than 2 denotes complex multicopy). CNV1 (Chr1:200.83–201.27 Mb) and CNV2 (Chr2:123.14–123.16 Mb) mainly consisted of multiple copies, and CNV3 (Chr4:54.81–54.82 Mb) was a deletion. (C) Schematic diagram showing the genes distributed within the candidate regions. (D) Frequency changes in candidate CNVs over 10 generations.* Interestingly, the EXOC4 gene was completely encompassed by CNV1. Combining this analysis with transcriptome data showed that EXOC4 was widely expressed in various periods and tissues, and the expression level in breast muscle, sebum, and brain was higher than that in other tissues (Supplementary Fig. S12A). Similar to CNV1, only 1 gene, TRPA1, was located in CNV2. A study has shown that TRPA1 is associated with growth traits (including body weight, body length, body height, etc.) in bovine [42]. In Pekin ducks, we did not observe the high expression level of TRPA1 in breast muscle (Supplementary Fig. S12B), but the body weight increased with the increase of breast muscle weight (Supplementary Fig. S4). ## Discussion Sequencing technology and comparative genomics have furthered our understanding of selection and variation. Based on long-term selection and phenotypic differentiation in livestock populations, several major genes controlling economically important traits in livestock have been identified [43–46]. However, the genetic mechanism of short-term, intense artificial selection remains unclear. Our work demonstrated how the response to short-term intense artificial selection targeting a complex trait—8-week breast muscle weight—in the Pekin duck Z2 line has been predominantly achieved by recruiting a large number of loci in the genome to undergo frequency shifts. This result was supported by the conclusions of previous studies on chicken experimental models [3, 4, 13]. Based on our new assembly, we analyzed the SNP information of each generation. We found that noncoding sequences have played a prominent role during Pekin duck breeding. This is can potentially be explained as, on the one hand, compared with coding sequences, cis-regulatory elements of pleiotropic loci are now considered the main source of phenotypic differentiation [47–49]. And, on the other hand, mutation bias reduces the mutation rate of functionally constrained regions [50]. After a new gene is formed by gene duplication and establishes a clear function, the mutation frequency of its coding region is restricted [51]. Thus, most of the mutations found in coding regions are synonymous mutations, and the mutation frequency in these regions is lower than that in noncoding region. We did not find completely fixed loci in the genome of the Pekin duck Z2 line, although this population was subjected to high-intensity artificial selection for 10 generations. However, we detected allele frequency shifts at many loci throughout the Pekin duck genome. Two possible explanations may account for this finding. First, it is more common for complex quantitative traits to involve a combination of multiple standing variations at many loci. Such polygenic adaptive patterns achieve fast phenotypic optimization through allele frequency shifts at many loci but do not necessarily lead to the fixation of any variation [13, 35]. Once selection pressure relaxes, the phenotype regresses toward the original (preselection) stat [52]. Second, no single genetic variation was shown to be either a necessary or sufficient condition for population breeding in this work. In this study, we identified 2 genes, UTP25 and FBRSL1, significantly associated with breast muscle weight. FBRSL1 belongs to the Polycomb group (PcG) gene and is essential for many biological processes in mammals, including stem cell maintenance and differentiation [53, 54]. Our results showed that the expression level of FBRSL1 in the breast muscle of Pekin duck Z2 line decreased with the increase of age. This may hint that FBRSL1 was strongly selected to promote the development of Pekin duck breast muscle. UTP25 (also named DIEXF in human) reportedly affects digestive organ expansion by regulating the p53 pathway [55, 56]. However, the function of UTP25 has not been characterized in animal muscles. UTP25 was strongly selected in the continuous breeding process of the Pekin duck Z2 line, and further investigation would be necessary to elucidate its functions. It can be anticipated that CNVs may be more important hotspots to reveal selection-induced molecular changes in Pekin ducks. First of all, CNVs are widespread in domesticated animals, such as pigeon, sheep, pig, chicken, cattle, horse, and dog [57–66]. Moreover, the identified CNVs were largely found in genes that encode growth factors and receptors, as well as genes related to development, and they play a role mainly through duplication [67]. Second, CNVs may have a role in rapid adaptation under strong selective pressure. This phenomenon was found in our experimental population that 81 CNVs were identified in the Pekin duck genome under 10 generations of high-intensity selection. Similar phenomena were found in experimental evolution studies in microbes under nutrient limitation and multicellular systems [68–70]. Finally, and most important, there is evidence for CNVs under selection in domesticated species. During domestication, CNVs underlying domestication traits increase in frequency in the population in response to selection, and genomic signatures for selection can sometimes be detected associated with these CNVs [57, 71, 72]. In this study, we identified 2 duplication CNVRs that were annotated in 2 genes, EXOC4 and TRPA1, respectively. EXOC4 may affect the development of breast muscle in Pekin ducks by affecting glucose transport and insulin synthesis. Studies have found that EXOC4 was involved in insulin synthesis and glucose transport in skeletal muscle [73–76]. As a component of the exocyst complex, EXOC4 is required for targeting of Glut4 to the plasma membrane by insulin [75]. Our results showed that after 10 years of artificial selection, the number of individuals with multiple copies of EXOC4 gene increases in the population of Pekin ducks. We speculate that the duplication makes fold increase of Exoc4 proteins, and a large number of Exoc4 proteins facilitate glucose transport to cells. Since cells become more efficient at taking up glucose, the excess glucose can be converted to fat and amino acids and stored by the body. Therefore, this may be one of the factors affecting the change of breast muscle weight of Pekin ducks. However, this requires confirmation. There is solid evidence demonstrating that TRPA1 is expressed throughout the mammalian body and has potential beneficial effects on systemic metabolism, including glucose metabolism [77]. Growing experimental evidence suggests that TRPA1 plays an important role in weight gain, obesity, and insulin secretion [42, 77–81]. In Pekin ducks, we did not observe the high expression level of TRPA1 in breast muscle, but the body weight increased with the increase of breast muscle weight. Previously, our group has also demonstrated that there is a high genetic and phenotypic correlation between breast muscle weight and body weight in Pekin ducks (0.83 and 0.80) [33]. We surmise that TRPA1 indirectly affects the breast muscle weight of Pekin ducks by affecting body weight. In summary, our research enabled us to understand the genetic variation mechanism of farm animal genomes under intense artificial selection and will provide useful information for the establishment of an efficient molecular breeding system for livestock. ## Subject details and sampling All duck samples for this study were collected from Pekin Duck Breeding Base, Changping District, Beijing. The Z2 line originated from the initially conserved population of Pekin duck in Beijing. The Pekin duck Z2 line was selected as the research object because it has many characteristics, including (i) it is bred in a closed group, with a pure pedigree and a clear genetic background; (ii) the selection pressure of this line is constant, which is favorable for the accumulation of alleles gradually; (iii) at the age of 6 weeks each generation, 15 male and 15 female ducks were randomly selected from a large population for a slaughter test to measure their breast muscle weight and retain blood samples; and (iv) the breeding of all generations was completed in Pekin duck breeding base, Changping District, Beijing, and the performance measurement was done in spring, resulting in little difference in environmental effect between generations. Furthermore, all ducks were kept in a similar environment and had free access to water and feed pellets [33]. We measured the breast muscle weight of the whole population in vivo at the age of 6 weeks of each generation and selected ducks with the heavier breast muscle weight as parents to produce the next generation. Roughly 750 individuals in each generation were retained for breeding, with a $35\%$ to $40\%$ retention rate for female ducks and a $7\%$ to $8\%$ retention rate for male ducks. The inbreeding was strictly avoided by calculating the inbreeding coefficient of each generation. The breast muscle weight (BMW) trait was estimated by breast muscle volume (BMV), breast width (BB), keel length (KL), and breast muscle thickness (BMT) (Fig. 2A, Supplementary Fig. S4). The correlation equation between these traits was as follows: BMV = BB × KL × BMT; BMW = 0.6228 × BMV + 17.042 [19]. BB and KL were measured by vernier calipers, while BMT was measured by ultrasound scanning technology. The correlation parameters of BMW and BMV are real and reliable numerical results based on the measured breast muscle weight of years of slaughter experiments and fitted by a linear regression equation model. In this study, we collected the blood of 30 ducks and phenotypic data of each individual (15 males and 15 females per generation) in the first, fourth, seventh, and tenth generations, respectively. A total of 120 duck samples were obtained (one sample was lost in the fourth generation), and a total of 119 samples were obtained finally. In addition, we randomly selected an adult female Pekin duck from the Pekin duck Z2 line to collect its liver for PacBio sequencing. We also collected breast muscle tissue from a male Pekin duck for BioNano and Hi-C sequencing. Furthermore, phenotype and pedigree data from all samples were collated for subsequent analysis. All individuals used for resequencing were collected wing vein blood and rapidly frozen at −20 °C. The phenol–chloroform method was used to extract blood DNA. The quality and quantity of the DNA were examined via Nanodrop and agarose gel electrophoresis. Then, the Illumina (San Diego, CA, USA) HiSeq X Ten platform (RRID:SCR_016385) was used to sequence the paired-end sequencing libraries with an inserted fragment length of approximately 500 bp in 8×. ## Reference genome assembly and annotation We used the combined strategy of long-reads single-molecule sequencing (PacBio, Beijing, China, RRID:SCR_017988) [82, 83], optical mapping (Bionano Genomics, Beijing, China) [84, 85], and chromosome interaction mapping (Hi-C) [86], which improved contiguity and completeness relative to the BGI_duck_1.0 [87]. We first used Canu (RRID:SCR_015880) [88] (v1.7.1) to correct and trim the subreads of PacBio with the default parameters. Then we assembled the high-quality sequence obtained in the previous step into contigs and adjusted the “correctedErrorRate” parameter to 0.05. Pilon (RRID:SCR_014731) [89] (v1.23) was used to polish the assembled contigs twice. Then scaffolds were assembled using Irys optical mapping data. We first adopted IrysSolve (Bionano Genomics) to assemble the raw *Bionano data* into an optical map with default parameters. Next, the runBNG pipelines [90] (v1.02) were used to construct the scaffolds based on the overlapping information between the optical map and PacBio contigs. Then we adopted Hi-C technology to anchor the scaffolds near the chromosome level. We first used bowtie2 to align the clean Hi-C raw reads to scaffolds. *We* generated ∼1.06 Gb pair-end reads, and ∼593.5 million were uniquely mapped to the scaffolds (Supplementary Table S10). After filtering out reads with low mapping, multiple hits, duplications, and singletons, only valid pairs were retained for subsequent analysis. After that, HiC-Pro (RRID:SCR_017643) [91] (v2.10.0) was used to construct an interaction matrix for valid interaction pairs of ∼371.6 million, and HiCPlotter [92] (v0.8.1) was used to draw the interaction heatmap. Then, Juicer was used to align the clean Hi-C reads to the draft assembly, and then the extracted data were automatically generated into a nearly chromosomal length assembly using the 3-dimensional DNA pipeline. The final draft was corrected using PBJelly (RRID:SCR_012091) [93]. The genome assembly was annotated by the NCBI Eukaryotic Genome Annotation Pipeline [66], an automated pipeline that annotates genes, transcripts, and proteins on draft and finished genome assemblies. ## Variant calling and filtering The raw reads from Illumina sequencing were filtered before downstream analyses by removing adapter sequences, contaminated reads, and low-quality reads. Then the reads were mapped to the assembly (GCA_003850225.1) with Burrows–Wheeler alignment (RRID:SCR_010910) [94] (v0.7.17-r1198) using the default parameters. SAMtools (RRID:SCR_002105) [95] (v1.13–14) software was used to convert mapping results into the BAM format and to filter the unmapped and nonunique reads. The paired reads that were mapped to the exact same position on the reference genome were identified with MarkDuplicates in Picard [96] (RRID:SCR_006525) to avoid any influence on variant detection. After the comparative evaluation of the depth and coverage of the results, we used the HaplotypeCaller program of GATK (RRID:SCR_001876) [97] (v1.90) software to call SNPs and indels to ensure the accuracy. In addition, this method can avoid the interference of false-positive sites in the follow-up analysis. For SNPs and indels, we restricted the variant form to biallelic variants by setting the option of GATK to "-T SelectVariants -SelectType SNP – RestrictTallelesto Biallelic." For the total variants, we set the GATK option “-T SelectVariants -select 'AF < 1.00'” to limit the allele frequency. We then filtered the output by using VCFtools (RRID:SCR_001235) [98] (v0.1.14). SNPs that did not meet the following criteria were excluded: (i) 3× < mean sequencing depth (over all included individuals) < 30×, (ii) a minor allele frequency >0.05 and a max allele frequency <0.99, (iii) maximum missing rate <0.1, and (iv) only 2 alleles. We used the CNVcaller (RRID:SCR_015752) [99] software to detect CNVs across 119 individuals, and this method also took into account the depth of reads and pair relationship, so as to identify the CNV interval. We first specified a 1,000-bp sliding window and a 500-bp step to count the GC, repeat, and gap contents of each window in the reference genome to generate the reference genome database. Then we calculated the absolute number of copies per window. Third, we used the “CNV.Discovery.sh” script to detect the CNV of the genomes with the parameter settings “-f 0.05 -h 5 -r 0.01 -p primaryCNVR -m mergeCNVR.” Finally, we used the “Genotype.py” script to genotype the copy numbers of each sample and generate the VCF file. ## Analysis of population genetic differences The Smartpca program of EIGENSOFT (RRID:SCR_004965) [100] (v4.2) software was used for PCA of whole-genome SNPs. We plotted the first 2 eigenvectors in 2 dimensions with our own R script for G1 and G4, G4 and G7, G7 and G10, and G1 and G10 populations, respectively. For estimations of allele frequencies of single SNPs, we used VCFTools [98] (v0.1.14) to filter the raw sequencing data of G1 and G10 generations. The parameters are set to be “–max-missing 0.9 –maf 0.01 –min-meanDP 5 –max-meanDP 30.” After filtering, 8,433,767 reliable SNPs were obtained for allele frequency estimation. The per-SNP absolute allele frequency difference (ΔAF) between the G1 and G10 generations was then calculated using the following formula: ΔAF = abs (RefAFG10 − RefAFG1). We next binned SNPs by ΔAF in steps of 0.05 (i.e., ΔAF = 0–0.05, 0.05–0.10, etc. until 0.95–1.00) and intersected these binned SNPs with coding exons, introns, and UTRs. ## Genome-wide association analysis of traits A GWAS was performed using a mixed linear model of EMMAX [101] program using genome-wide SNP data and breast muscle weight of 119 individuals from the resequenced population. The analysis model was where y is the phenotypic value (the breast muscle weight of per duck), X is the matrix corresponding to the fixed effect, and b is the fixed effect size. Fixed effects include sex effects. G is the genetic matrix corresponding to the population kinship, and e is the random residual. PCA was performed based on all SNPs, and the top 3 components were set as fixed effects in the mixed model to correct for population stratification. We defined a Bonferroni correction threshold of 0.01/N (−log10 $$P \leq 8.94$$) to identify the significant loci of the GWAS results, where N was the number of whole-genome SNPs. For the identified associated genes, we also referred to the Pekin duck Panoramic transcription map (transcriptome data of all tissues in the 3 development stages) established by our previous study [46] to check whether these genes were expressed in the development stage of breast weight muscle, so as to further confirm that they did participate in the regulation of breast muscle development. ## Identification of selected regions We used the VCFtools [98] (v0.1.14) software to calculate Fst between G1 and G10 generations by selecting parameters of 10-kb windows with a 5-kb step size. We used selscan software [102] (v1.2.0a) to calculate the XP-EHH values of G1 and G10 groups with the same parameter settings of 10-kb sliding windows and 5-kb step size, in which the G1 generation was taken as the reference group and the G10 generation as the query group. The overlapping regions of 2 windows with statistical values above the top $1\%$ of the quantile would be selected as preliminary candidate regions. ## Linkage disequilibrium analysis In addition, to further narrow the candidate interval, we also used the Haploview (RRID:SCR_003076) [103] to analyze the linkage disequilibrium of the candidate region (related to Fig. 4). We adopted the squared correlation coefficient (r2) as the coefficient to measure the linkage disequilibrium between the leader SNP and the surrounding SNPs. The parameters settings were “–ld-window 99999 –ld-window-kb 1000 –ld-window-r2 0 –r2.” Then we integrated and plotted the LD data with GWAS results (related to Supplementary Fig. S8). ## Identification of selected CNVRs Based our new assembly, we used the CNV caller [99] to identify genome-wide CNVs and CNVRs. To avoid false positives, 2 parameters, silhouette coefficient (>0.7) and minor allele frequency (>0.05) were adopted to filter the CNVRs obtained, and then we got 81 CNVRs with high credibility and accuracy on autosomes. Subsequently, we used the relative frequency difference (RFD) [104] to detect the CNVs that occurred on the Pekin duck genome during population differentiation. After that, we adopted the topmost $5\%$ RFD value (absolute RFD value >4.1) as the threshold to screen the potentially selected CNVRs. ## Additional Files Supplemental Fig. S1. The pipeline for multilevel chromosome assembly. Canu was used for constructing initial contigs. Then, polishing was performed with Pilon using PacBio-only long reads. Hybrid scaffolding of the PacBio-corrected contigs and the BioNano-based consensus map was performed using the hybrid scaffolding module within runBNG software. The Hi-C sequencing data were first aligned to the assembled contigs/scaffolds using the Bowtie end-to-end algorithm, and then the assembled scaffolds were clustered, ordered, and directed into a chromosome level using 3-dimensional DNA. The final draft was corrected using PBJelly. Supplemental Fig. S2. Hi-C interactions among 29 chromosomes with a 40-kb resolution. Supplemental Fig. S3. BUSCO completeness assessment for the new assembly. In summary, it covered $93.3\%$ (7,$\frac{780}{8}$,338) of complete BUSCO genes and $1.3\%$ ($\frac{111}{8}$,338) of fragmented BUSCO genes. Supplemental Fig. S4. Generational average phenotypic values of prime traits of the Pekin duck Z2 line. The abscissa denotes the generation. Supplemental Fig. S5. Distribution of Fst and XP-EHH of 10-kb windows size for whole-genome-wide variants between the G1 and G10 generations. Bins of Fst and XP-EHH are presented along the x-axes. μ, mean; δ, standard deviation. Supplemental Fig. S6. The variation of allele frequency of putative regions over 10 generations (corresponding to Fig. 2). ( A, B) Red and green lines indicate the frequency variation of representative loci in selective regions of Fst and XP-EHH, respectively (corresponding to a $1\%$ significance level of Fst in 10-kb sliding windows and a $1\%$ significance level of XP-EHH in 10-kb sliding windows; each line represents the top SNP of a sliding window). ( C) Blue lines indicate the frequency variation of 22 SNPs that reached the Bonferroni significance threshold of the GWAS (−log10 $P \leq 8.94$). Supplementary Fig. S7. The variation trend of genetic diversity in potentially selective regions among generations. This figure shows the trend of genetic diversity in the common selection signatures obtained by Fst and XP-EHH tests across generations. The indicated P values are based on 1-way ANOVA. ** $P \leq 0.01$, *$P \leq 0.05$, and ns indicates that the P value was not significant. Supplemental Fig. S8. Linkage disequilibrium (LD) in 2 candidate regions of 10 generations. Each diamond contains a level of LD (r2) between all SNP pairs. Supplemental Fig. S9. SNP distribution in candidate regions. Supplemental Fig. S10. Expression levels of the UTP25 and FBRSL1 genes in different tissues of Pekin duck. ( A) FBRSL1 expression levels in Pekin duck tissues. ( B) UTP25 expression levels in Pekin duck tissues. The expression data were obtained from a global gene expression database for ducks generated from transcriptome analyses. The database lists the expression levels of all genes in breast muscle, skin, liver, fat (abdominal fat), brain, heart, kidney, lung, spleen, sternum, and shank tissues at different developmental periods in Pekin ducks. Supplemental Fig. S11. Expression levels of the UTP25 and FBRSL1 genes at different developmental periods of the Pekin duck Z2 line. Supplemental Fig. S12. Expression levels of the EXOC4 and TRPA1 genes in different tissues of Pekin duck. ( A) EXOC4 expression levels in Pekin duck tissues. ( B) TRPA1 expression levels in Pekin duck tissues. The expression data were obtained from a global gene expression database for ducks generated from transcriptome analyses. The database lists the expression levels of all genes in breast muscle, skin, liver, fat (abdominal fat), brain, heart, kidney, lung, spleen, sternum, and shank tissues at different developmental periods in Pekin ducks. Supplemental Table S1. Summary of sequencing data. Supplemental Table S2. Summary of Hi-C reads mapping results. Supplemental Table S3. Comparison of genome assemblies between the GCA_003850225.1 and BGI_duck_1.0. Supplemental Table S4. Number of ducks used for 10 generations for the genome resequencing and mapping summary. Supplemental Table S5. Genomic SNP statistics. Supplemental Table S6. Distributions of SNP counts in the different delta allele frequency bins for coding sequences, introns, and UTRs. Supplemental Table S7. Overlapping regions between the analysis results of Fst and XP-EHH. Pairwise fixation index values (Fst) and cross-population extended haplotype homozygosity (XP-EHH) values were calculated in 10-kb windows sliding in 5-kb steps. The top $1\%$ of the overlap regions between the analysis results of Fst and XP-EHH were considered candidate regions. Supplemental Table S8. Summary of putative regions on the genome. These regions sum to 2.06 Mb, and ∼2.14 Mb were annotated into 65 genes. Supplemental Table S9. The significant SNPs of GWAS analyses. We performed GWAS analyses of breast muscle weight. The Bonferroni significance threshold of the GWAS was −log10 $P \leq 8.94.$ There were 22 SNPs that passed the Bonferroni significance threshold. These SNPs are located in noncoding sequences. Supplemental Table S10. Putative CNVRs under artificial selection. Supplemental Table S11. The copy number variation genotypes of 3 candidate CNVRs in each individual. ## Abbreviations BB: breast breadth; BMT: breast muscle thickness; BMV: breast muscle volume; BMW: breast muscle weight; bp: base pairs; BUSCO: Benchmarking Universal Single-Copy Orthologs; BWA: Burrows–Wheeler aligner; CNV: copy number variation; CNVR: copy number variation region;Fst: fixation index; g: gram; G: generation; Gb: gigabase pairs; GC: guanine-cytosine; GWAS: genome-wide association study; Hi-C: High-throughput Chromosome Conformation Capture; kb: kilobase pairs; KL: keel length; LD: linkage disequilibrium; Mb: megabase pairs; NCBI: National Center for Biotechnology Information; PCA: principal component analysis; RFD: relative frequency difference; SNP: single-nucleotide polymorphism; TE: transposable element; UTR: untranslated regions; XP-EHH: population extended haplotype homozygosity; ΔAF: allele frequency difference. ## Funding This work was supported by grants from the National Natural Science Foundation of China [31972523], the Young Top-notch Talent Project of the National Ten Thousand Talent Program, the China Agriculture Research System of MOF and MARA (CARS-42-5), and the CAAS Innovation Team Project (ASTIP-IAS-9, CAAS-ZDRW202104). ## Data Availability The assembly and annotation of Pekin duck has been deposited in GenBank under the Bioproject accession code PRJNA496533 (accession No. RHJV01000000). All supporting data are available in the GigaScience GigaDB database [105]. ## Competing Interests The authors have declared no competing interests. ## Ethics Statement All animals used in the study were treated following the guidelines for the experimental animals established by the Council of China Animal Welfare. Protocols of the experiments were approved by the Science Research Department of the Institute of Animal Sciences, Chinese Academy of Agricultural Sciences (CAAS) (Beijing, China). ## Authors’ Contributions Z.Zhou and S.H. conceived the project, designed the research, and managed the project. S.H., Z.Zhou, Z.G., J.Hu, M.Xie, W.Huang, Y.Zhang, and Q.Zhang constructed the population. M.Xie, J.T., W.Huang, J.Hu, Y.Zhang, Z.G., G.Xing, W.F., Y.Xu, S.Liang, and D.Liu. collected the phenotype data. Y. Jiang, Z.Liu, M.Li, H.C., and Z.Zhou performed the genome assembly. S.Y., Z.Zhou, Z.Liu, D.Zhou, and P.Hua performed bioinformatics analysis. S.Y. and Z.Zhou wrote the manuscript. ## References 1. Fuller RC, Baer CF, Travis J.. **How and when selection experiments might actually be useful**. *Integr Comp Biol* (2005) **45** 391-404. PMID: 21676785 2. Lou RN, Therkildsen NO, Messer PW. **The effects of quantitative trait architecture on detection power in short-term artificial selection experiments**. *G3 (Bethesda)* (2020) **10** 3213-27. PMID: 32646912 3. Zhang H, Liang Q, Wang N. **Microevolutionary dynamics of chicken genomes under divergent selection for adiposity**. *iScience* (2020) **23** 101193. PMID: 32554187 4. Zan Y, Sheng Z, Lillie M. **Artificial selection response due to polygenic adaptation from a multilocus, multiallelic genetic architecture**. *Mol Biol Evol* (2017) **34** 2678-89. PMID: 28957504 5. Marks HL.. **Long-term selection for body weight in Japanese quail under different environments**. *Poult Sci* (1996) **75** 1198-203. PMID: 8893294 6. Castro JP, Yancoskie MN, Marchini M. **An integrative genomic analysis of the Longshanks selection experiment for longer limbs in mice**. *Elife* (2019) **8** 42014 7. Kelly JK, Hughes KA.. **Pervasive linked selection and intermediate-frequency alleles are implicated in an evolve-and-resequencing experiment of Drosophila simulans**. *Genetics* (2019) **211** 943-61. PMID: 30593495 8. Seabra SG, Fragata I, Antunes MA. **Different genomic changes underlie adaptive evolution in populations of contrasting history**. *Mol Biol Evol* (2018) **35** 549-63. PMID: 29029198 9. Siegel PB.. **Selection for body weight at eight weeks of age: 1. Short term response and heritabilities**. *Poult Sci* (1962) **41** 954-62 10. Johansson AM, Pettersson ME, Siegel PB. **Genome-wide effects of long-term divergent selection**. *PLos Genet* (2010) **6** e1001188. PMID: 21079680 11. Dunnington EA, Siegel PB.. **Long-term divergent selection for eight-week body weight in white Plymouth Rock chickens**. *Poult Sci* (1996) **75** 1168-79. PMID: 8893291 12. Lillie M, Sheng ZY, Honaker CF. **Genomic signatures of 60 years of bidirectional selection for 8-week body weight in chickens**. *Poult Sci* (2018) **97** 781-90. PMID: 29272516 13. Lillie M, F. Honaker Christa, Siegel PB.. **Bidirectional selection for body weight on standing genetic variation in a chicken model**. *G3 (Bethesda)* (2019) **9** 1165-1173. PMID: 30737239 14. Wu X, Zhang Q, Xu S. **Differential expression of six chicken genes associated with fatness traits in a divergently selected broiler population**. *Mol Cell Probes* (2016) **30** 1-5. PMID: 26746358 15. Wang HB, Li H, Wang QG. **Profiling of chicken adipose tissue gene expression by genome array**. *BMC Genomics* (2007) **8** 193. PMID: 17594506 16. Zhang H, Hu X, Wang Z. **Selection signature analysis implicates the PC1/PCSK1 region for chicken abdominal fat content**. *PLoS One* (2012) **7** e40736-e. PMID: 22792402 17. Zhang H, Wang S-Z, Wang Z-P. **A genome-wide scan of selective sweeps in two broiler chicken lines divergently selected for abdominal fat content**. *BMC Genomics* (2012) **13** 704. PMID: 23241142 18. Yang Z-M, Wang W, Du Z-Q. **Expression profiling of preadipocyte microRNAs by deep sequencing on chicken lines divergently selected for abdominal fatness**. *PLoS One* (2015) **10** e0117843. PMID: 25675096 19. Carneiro M, Rubin CJ, Di Palma F. **Rabbit genome analysis reveals a polygenic basis for phenotypic change during domestication**. *Science* (2014) **345** 1074-9. PMID: 25170157 20. Li X, Yang J, Shen M. **Whole-genome resequencing of wild and domestic sheep identifies genes associated with morphological and agronomic traits**. *Nat Commun* (2020) **11** 2815. PMID: 32499537 21. Zheng Z, Wang X, Li M. **The origin of domestication genes in goats**. *Sci Adv* (2020) **6** eaaz5216. PMID: 32671210 22. Dong Y, Zhang X, Xie M. **Reference genome of wild goat (capra aegagrus) and sequencing of goat breeds provide insight into genic basis of goat domestication**. *BMC Genomics* (2015) **16** 431. PMID: 26044654 23. Ai H, Fang X, Yang B. **Adaptation and possible ancient interspecies introgression in pigs identified by whole-genome sequencing**. *Nat Genet* (2015) **47** 217-25. PMID: 25621459 24. Li J, Zhang J, Liu J. **A new duck genome reveals conserved and convergently evolved chromosome architectures of birds and mammals**. *GigaScience* (2021) **10** giaa142. PMID: 33406261 25. Zhang L, Hu J, Han X. **A high-quality apple genome assembly reveals the association of a retrotransposon and red fruit colour**. *Nat Commun* (2019) **10** 1494. PMID: 30940818 26. Peace CP.. **DNA-informed breeding of rosaceous crops: promises, progress and prospects**. *Horticulture Res* (2017) **4** 17006 27. Li M, Tian S, Jin L. **Genomic analyses identify distinct patterns of selection in domesticated pigs and Tibetan wild boars**. *Nat Genet* (2013) **45** 1431-8. PMID: 24162736 28. Bickhart DM, Rosen BD, Koren S. **Single-molecule sequencing and chromatin conformation capture enable de novo reference assembly of the domestic goat genome**. *Nat Genet* (2017) **49** 643-50. PMID: 28263316 29. Zhu F, Yin Z-T, Wang Z. **Three chromosome-level duck genome assemblies provide insights into genomic variation during domestication**. *Nat Commun* (2021) **12** 5932. PMID: 34635656 30. Li M, Sun C, Xu N. **De novo assembly of 20 chicken genomes reveals the undetectable phenomenon for thousands of core genes on microchromosomes and subtelomeric regions**. *Mol Biol Evol* (2022) **39** msac066. PMID: 35325213 31. Wang K, Hu H, Tian Y. **The chicken pan-genome reveals gene content variation and a promoter region deletion in IGF2BP1 affecting body size**. *Mol Biol Evol* (2021) **38** 5066-81. PMID: 34329477 32. Simao FA, Waterhouse RM, Ioannidis P. **BUSCO: assessing genome assembly and annotation completeness with single-copy orthologs**. *Bioinformatics* (2015) **31** 3210-2. PMID: 26059717 33. Xu Y, Hu J, Zhang Y. **Selection response and estimation of the genetic parameters for multidimensional measured breast meat yield related traits in a long-term breeding Pekin duck line**. *Asian-Australas J Anim Sci* (2018) **31** 1575-80. PMID: 29642677 34. Reich D, Price AL, Patterson N.. **Principal component analysis of genetic data**. *Nat Genet* (2008) **40** 491-2. PMID: 18443580 35. Pritchard JK, Pickrell JK, Coop G.. **The genetics of human adaptation: hard sweeps, soft sweeps, and polygenic adaptation**. *Curr Biol* (2010) **20** R208-15. PMID: 20178769 36. Wright S.. **The genetical strucyure of populations**. *Ann Eugenics* (1949) **15** 323-54 37. Sabeti PC, Varilly P, Fry B. **Genome-wide detection and characterization of positive selection in human populations**. *Nature* (2007) **449** 913-8. PMID: 17943131 38. Cooper GM, Nickerson DA, Eichler EE.. **Mutational and selective effects on copy-number variants in the human genome**. *Nat Genet* (2007) **39** S22-9. PMID: 17597777 39. Bickhart DM, Hou Y, Schroeder SG. **Copy number variation of individual cattle genomes using next-generation sequencing**. *Genome Res* (2012) **22** 778-90. PMID: 22300768 40. Weischenfeldt J, Symmons O, Spitz F. **Phenotypic impact of genomic structural variation: insights from and for human disease**. *Nat Rev Genet* (2013) **14** 125-38. PMID: 23329113 41. Xu Y, Shi T, Cai H. **Associations of MYH3 gene copy number variations with transcriptional expression and growth traits in Chinese cattle**. *Gene* (2014) **535** 106-11. PMID: 24316128 42. Wu J, Wang J, Yue B. **Research on association between variants and haplotypes of TRPV1 and TRPA1 genes with growth traits in three cattle breeds**. *Anim Biotechnol* (2019) **30** 202-11. PMID: 30595081 43. Van Laere A-S, Nguyen M, Braunschweig M. **A regulatory mutation in IGF2 causes a major QTL effect on muscle growth in the pig**. *Nature* (2003) **425** 832-6. PMID: 14574411 44. Clop A, Marcq F, Takeda H. **A mutation creating a potential illegitimate microRNA target site in the myostatin gene affects muscularity in sheep**. *Nat Genet* (2006) **38** 813-8. PMID: 16751773 45. Karim L, Takeda H, Lin L. **Variants modulating the expression of a chromosome domain encompassing PLAG1 influence bovine stature**. *Nat Genet* (2011) **43** 405-13. PMID: 21516082 46. Zhou Z, Li M, Cheng H. **An intercross population study reveals genes associated with body size and plumage color in ducks**. *Nat Commun* (2018) **9** 2648. PMID: 30018292 47. Carroll SB.. **Evo-devo and an expanding evolutionary synthesis: a genetic theory of morphological evolution**. *Cell* (2008) **134** 25-36. PMID: 18614008 48. Stern DL, Orgogozo V.. **The loci of evolution: how predictable is genetic evolution?**. *Evolution* (2008) **62** 2155-77. PMID: 18616572 49. Wittkopp PJ, Kalay G.. **Cis-regulatory elements: molecular mechanisms and evolutionary processes underlying divergence**. *Nat Rev Genet* (2012) **13** 59-69 50. Monroe JG, Srikant T, Carbonell-Bejerano P. **Mutation bias reflects natural selection in Arabidopsis thaliana**. *Nature* (2022) **602** 101-5. PMID: 35022609 51. Nei M, Suzuki Y, Nozawa M.. **The neutral theory of molecular evolution in the genomic era**. *Annu Rev Genomics Hum Genet* (2010) **11** 265-89. PMID: 20565254 52. Dunnington EA, Honaker CF, McGilliard ML. **Phenotypic responses of chickens to long-term, bidirectional selection for juvenile body weight–historical perspective**. *Poult Sci* (2013) **92** 1724-34. PMID: 23776258 53. Bathla S, Rawat P, Baithalu R. **Profiling of urinary proteins in Karan Fries cows reveals more than 1550 proteins**. *J Proteomics* (2015) **127** 193-201. PMID: 26021477 54. Chen Z, Yao Y, Ma P. **Haplotype-based genome-wide association study identifies loci and candidate genes for milk yield in Holsteins**. *PLoS One* (2018) **13** e0192695. PMID: 29447209 55. Zhang S-J, Liu C-J, Yu P. **Evolutionary interrogation of human biology in well-annotated genomic framework of rhesus macaque**. *Mol Biol Evol* (2014) **31** 1309-24. PMID: 24577841 56. Chen J, Ruan H, Ng SM. **Loss of function of def selectively up-regulates Delta113p53 expression to arrest expansion growth of digestive organs in zebrafish**. *Genes Dev* (2005) **19** 2900-11. PMID: 16322560 57. Bruders R, Van Hollebeke H, Osborne EJ. **A copy number variant is associated with a spectrum of pigmentation patterns in the rock pigeon (Columba livia)**. *PLos Genet* (2020) **16** e1008274. PMID: 32433666 58. Ghosh S, Qu Z, Das PJ. **Copy number variation in the horse genome**. *PLos Genet* (2014) **10** e1004712. PMID: 25340504 59. Guo Y, Gu X, Sheng Z. **A complex structural variation on chromosome 27 leads to the ectopic expression of HOXB8 and the muffs and beard phenotype in chickens**. *PLos Genet* (2016) **12** e1006071. PMID: 27253709 60. Lee YL, Takeda H, Costa Monteiro Moreira G. **A 12 kb multi-allelic copy number variation encompassing a GC gene enhancer is associated with mastitis resistance in dairy cattle**. *PLos Genet* (2021) **17** e1009331. PMID: 34288907 61. Qiu Y, Ding R, Zhuang Z. **Genome-wide detection of CNV regions and their potential association with growth and fatness traits in Duroc pigs**. *BMC Genomics* (2021) **22** 332. PMID: 33964879 62. Liu M, Fang L, Liu S. **Array CGH-based detection of CNV regions and their potential association with reproduction and other economic traits in Holsteins**. *BMC Genomics* (2019) **20** 181. PMID: 30845913 63. Stafuzza NB, Silva RMO, Fragomeni BO. **A genome-wide single nucleotide polymorphism and copy number variation analysis for number of piglets born alive**. *BMC Genomics* (2019) **20** 321. PMID: 31029102 64. Yuan C, Lu Z, Guo T. **A global analysis of CNVs in Chinese indigenous fine-wool sheep populations using whole-genome resequencing**. *BMC Genomics* (2021) **22** 78. PMID: 33485316 65. Salmon Hillbertz NH, Isaksson M, Karlsson EK. **Duplication of FGF3, FGF4, FGF19 and ORAOV1 causes hair ridge and predisposition to dermoid sinus in Ridgeback dogs**. *Nat Genet* (2007) **39** 1318-20. PMID: 17906623 66. Pruitt KD, Brown GR, Hiatt SM. **RefSeq: an update on mammalian reference sequences**. *Nucleic Acids Res* (2014) **42** D756-63. PMID: 24259432 67. Lye ZN, Purugganan MD.. **Copy number variation in domestication**. *Trends Plant Sci* (2019) **24** 352-65. PMID: 30745056 68. DeBolt S.. **Copy number variation shapes genome diversity in Arabidopsis over immediate family generational scales**. *Genome Biol Evol* (2010) **2** 441-53. PMID: 20624746 69. Gresham D, Desai MM, Tucker CM. **The repertoire and dynamics of evolutionary adaptations to controlled nutrient-limited environments in yeast**. *PLos Genet* (2008) **4** e1000303. PMID: 19079573 70. Farslow JC, Lipinski KJ, Packard LB. **Rapid Increase in frequency of gene copy-number variants during experimental evolution in Caenorhabditis elegans**. *BMC Genomics* (2015) **16** 1044. PMID: 26645535 71. Axelsson E, Ratnakumar A, Arendt M-L. **The genomic signature of dog domestication reveals adaptation to a starch-rich diet**. *Nature* (2013) **495** 360-4. PMID: 23354050 72. Reiter T, Jagoda E, Capellini TD. **Dietary variation and evolution of gene copy number among dog breeds**. *PLoS One* (2016) **11** e0148899. PMID: 26863414 73. Ewart M-A, Clarke M, Kane S. **Evidence for a role of the exocyst in insulin-stimulated Glut4 trafficking in 3T3-L1 adipocytes**. *J Biol Chem* (2005) **280** 3812-6. PMID: 15550383 74. Fulcher F, Smith B, Russ M. **Dual role for myosin II in GLUT4-mediated glucose uptake in 3T3-L1 adipocytes**. *Exp Cell Res* (2008) **314** 3264-74. PMID: 18773891 75. Inoue M, Chang L, Hwang J. **The exocyst complex is required for targeting of Glut4 to the plasma membrane by insulin**. *Nature* (2003) **422** 629-33. PMID: 12687004 76. Kee AJ, Yang L, Lucas CA. **An actin filament population defined by the tropomyosin Tpm3.1 regulates glucose uptake**. *Traffic* (2015) **16** 691-711. PMID: 25783006 77. Derbenev AV, Zsombok A.. **Potential therapeutic value of TRPV1 and TRPA1 in diabetes mellitus and obesity**. *Semin Immunopathol* (2016) **38** 397-406. PMID: 26403087 78. Mahajan N, Khare P, Kondepudi KK. **TRPA1: pharmacology, natural activators and role in obesity prevention**. *Eur J Pharmacol* (2021) **912** 174553. PMID: 34627805 79. Kagawa Y, Ozaki-Masuzawa Y, Hosono T. **Garlic oil suppresses high-fat diet induced obesity in rats through the upregulation of UCP-1 and the enhancement of energy expenditure**. *Exp Ther Med* (2020) **19 2** 1536-40. PMID: 32010335 80. Khare P, Jagtap S, Jain Y. **Cinnamaldehyde supplementation prevents fasting-induced hyperphagia, lipid accumulation, and inflammation in high-fat diet-fed mice**. *Biofactors* (2016) **42** 201-11. PMID: 26893251 81. Kim MJ, Son HJ, Song SH. **The TRPA1 agonist, methyl syringate suppresses food intake and gastric emptying**. *PLoS One* (2013) **8** e71603. PMID: 23990963 82. Eid J, Fehr A, Gray J. **Real-time DNA sequencing from single polymerase molecules**. *Science* (2009) **323** 133-8. PMID: 19023044 83. Gordon D, Huddleston J, Chaisson MJ. **Long-read sequence assembly of the gorilla genome**. *Science* (2016) **352** aae0344. PMID: 27034376 84. Schwartz DC, Li X, Hernandez LI. **Ordered restriction maps of Saccharomyces cerevisiae chromosomes constructed by optical mapping**. *Science* (1993) **262** 110-4. PMID: 8211116 85. Hastie AR, Dong L, Smith A. **Rapid genome mapping in nanochannel arrays for highly complete and accurate de novo sequence assembly of the complex Aegilops tauschii genome**. *PLoS One* (2013) **8** e55864. PMID: 23405223 86. Lieberman-Aiden E, van Berkum NL, Williams L. **Comprehensive mapping of long-range interactions reveals folding principles of the human genome**. *Science* (2009) **326** 289-93. PMID: 19815776 87. Huang Y, Li Y, Burt DW. **The duck genome and transcriptome provide insight into an avian influenza virus reservoir species**. *Nat Genet* (2013) **45** 776-83. PMID: 23749191 88. Koren S, Walenz BP, Berlin K. **Canu: scalable and accurate long-read assembly via adaptivek-mer weighting and repeat separation**. *Genome Res* (2017) **27** 722-36. PMID: 28298431 89. Walker BJ, Abeel T, Shea T. **Pilon: an integrated tool for comprehensive microbial variant detection and genome assembly improvement**. *PLoS One* (2014) **9** e112963. PMID: 25409509 90. Yuan Y, Bayer PE, Lee H-T. **runBNG: a software package for BioNano genomic analysis on the command line**. *Bioinformatics* (2017) **33** 3107-9. PMID: 28605539 91. Servant N, Varoquaux N, Lajoie BR. **HiC-Pro: an optimized and flexible pipeline for Hi-C data processing**. *Genome Biol* (2015) **16** 259. PMID: 26619908 92. Akdemir KC, Chin L.. **HiCPlotter integrates genomic data with interaction matrices**. *Genome Biol* (2015) **16** 198. PMID: 26392354 93. English AC, Richards S, Han Y. **Mind the gap: upgrading genomes with Pacific Biosciences RS long-read sequencing technology**. *PLoS One* (2012) **7** e47768. PMID: 23185243 94. Li H, Durbin R.. **Fast and accurate short read alignment with Burrows-Wheeler transform**. *Bioinformatics* (2009) **25** 1754-60. PMID: 19451168 95. Danecek P, Bonfield JK, Liddle J. **Twelve years of SAMtools and BCFtools**. *Gigascience* (2021) **10** giab008. PMID: 33590861 96. 96. Picard Toolkit . 2019; Broad Institute, GitHub Repository. https://broadinstitute.github.io/picard/.. (2019) 97. McKenna A, Hanna M, Banks E. **The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data**. *Genome Res* (2010) **20** 1297-303. PMID: 20644199 98. Danecek P, Auton A, Abecasis G. **The variant call format and VCFtools**. *Bioinformatics* (2011) **27** 2156-8. PMID: 21653522 99. Xihong W, Zhuqing Z, Yudong C. **CNVcaller: highly efficient and widely applicable software for detecting copy number variations in large populations**. *Gigascience* (2017) **6** 1-12 100. Price AL, Patterson NJ, Plenge RM. **Principal components analysis corrects for stratification in genome-wide association studies**. *Nat Genet* (2006) **38** 904-9. PMID: 16862161 101. Kang HM, Sul JH, Service SK. **Variance component model to account for sample structure in genome-wide association studies**. *Nat Genet* (2010) **42** 348-54. PMID: 20208533 102. Szpiech ZA, Hernandez RD.. **selscan: an efficient multithreaded program to perform EHH-based scans for positive selection**. *Mol Biol Evol* (2014) **31** 2824-7. PMID: 25015648 103. Barrett JC, Fry B, Maller J. **Haploview: analysis and visualization of LD and haplotype maps**. *Bioinformatics* (2005) **21** 263-5. PMID: 15297300 104. Zhou Z, Jiang Y, Wang Z. **Resequencing 302 wild and cultivated accessions identifies genes related to domestication and improvement in soybean**. *Nat Biotechnol* (2015) **33** 408-14. PMID: 25643055 105. Yu S, Liu Z, Li M. *GigaScience Database* (2023). DOI: 10.5524/102361
--- title: Molecular origins of mutational spectra produced by the environmental carcinogen N-nitrosodimethylamine and SN1 chemotherapeutic agents authors: - Amanda L Armijo - Pennapa Thongararm - Bogdan I Fedeles - Judy Yau - Jennifer E Kay - Joshua J Corrigan - Marisa Chancharoen - Supawadee Chawanthayatham - Leona D Samson - Sebastian E Carrasco - Bevin P Engelward - James G Fox - Robert G Croy - John M Essigmann journal: NAR Cancer year: 2023 pmcid: PMC10041537 doi: 10.1093/narcan/zcad015 license: CC BY 4.0 --- # Molecular origins of mutational spectra produced by the environmental carcinogen N-nitrosodimethylamine and SN1 chemotherapeutic agents ## Abstract DNA-methylating environmental carcinogens such as N-nitrosodimethylamine (NDMA) and certain alkylators used in chemotherapy form O6-methylguanine (m6G) as a functionally critical intermediate. NDMA is a multi-organ carcinogen found in contaminated water, polluted air, preserved foods, tobacco products, and many pharmaceuticals. Only ten weeks after exposure to NDMA, neonatally-treated mice experienced elevated mutation frequencies in liver, lung and kidney of ∼35-fold, 4-fold and 2-fold, respectively. High-resolution mutational spectra (HRMS) of liver and lung revealed distinctive patterns dominated by GC→AT mutations in 5’-Pu-G-3’ contexts, very similar to human COSMIC mutational signature SBS11. Commonly associated with alkylation damage, SBS11 appears in cancers treated with the DNA alkylator temozolomide (TMZ). When cells derived from the mice were treated with TMZ, N-methyl-N-nitrosourea, and streptozotocin (two other therapeutic methylating agents), all displayed NDMA-like HRMS, indicating mechanistically convergent mutational processes. The role of m6G in shaping the mutational spectrum of NDMA was probed by removing MGMT, the main cellular defense against m6G. MGMT-deficient mice displayed a strikingly enhanced mutant frequency, but identical HRMS, indicating that the mutational properties of these alkylators is likely owed to sequence-specific DNA binding. In sum, the HRMS of m6G-forming agents constitute an early-onset biomarker of exposure to DNA methylating carcinogens and drugs. ## Graphical Abstract Graphical AbstractDistinct DNA methylating agents, including N-nitrosodimethylamine and temozolomide generate nearly identical mutational spectra in repair proficient and deficient mice, and in cell culture, suggesting a convergent chemical mechanism and mutational process. ## INTRODUCTION N-Nitrosamines such as N-nitrosodimethylamine (NDMA; Figure 1) are DNA alkylating agents that are abundant in the human environment [1,2]. The toxic, mutagenic and carcinogenic effects of NDMA have been demonstrated in numerous animal studies [3,4], leading to classification of this agent by the International Agency for Research on Cancer as a probable (Group 2A) human carcinogen [5]. While human exposure can occur through several routes, industrial contamination of watersheds has been a major source of concern, particularly for NDMA [6]. This N-nitrosamine has been detected as a contaminant of drinking water at locations on the Environmental Protection Agency National Priorities List, including a Superfund Site in Wilmington, Massachusetts, where the agent was recently associated with a cluster of childhood malignancies [7,8]. Risk of exposure, however, is not limited to industrial sites because N-nitrosamines are also found in cured meats [9,10], tobacco combustion products [11], and contaminated pharmaceuticals, including valsartan, metformin, and ranitidine (12–15). **Figure 1.:** *Experimental workflow from toxicant exposure to high-resolution mutational spectra (HRMS). (A) Compounds evaluated for mutagenic properties include N-nitrosodimethylamine (NDMA), N-methyl-N-nitrosourea (MNU), streptozotocin (STZ) and temozolomide (TMZ), all of which form a putative methyldiazonium ion (B) prior to reaction with DNA to form adducts such as O6-methylguanine (m6G) (C), a mutagenic adduct that mispairs with thymine during DNA synthesis. Portions of MNU and STZ are differentially colored in blue to show shared structural features, and the red functional group on each structure denotes the methyl group transferred to DNA. (D) Neonatal C57BL/6J-gptΔ mice were treated with NDMA on days 8 and 15 post-birth for a total dose of 10.5 mg/kg. Lung, liver and kidney were harvested 10 weeks following the second injection. In some experiments, derivatives of the C57BL/6J-gptΔ mouse were used, where the RaDR locus was incorporated with or without the Mgmt gene, which encodes a protein that repairs m6G. (E) In parallel, mouse embryo fibroblasts (MEFs) generated from C57BL/6J- gptΔ mice were treated with the direct-acting mutagens, MNU, STZ or TMZ, as indicated. DNA isolated from these samples was analyzed: (F) for point mutations in the gpt gene, or (G) subjected to duplex sequencing to generate an HRMS for each agent.* Like many carcinogens, NDMA requires biochemical activation to a DNA damaging electrophile as a chemical prelude to its potent biological effects. As such, DNA alkylation by NDMA and other N-nitroso compounds is believed to be mediated by reactive alkyl diazonium ions. The electrophilic methyldiazonium ion (Figure 1B) derived from NDMA is the biotransformation product of the parent carcinogen by mammalian cytochrome P450s, primarily isoenzyme 2E1 (CYP2E1) [16,17]. Reaction of the methyldiazonium ion with DNA produces 7-methylguanine (m7G; ∼$70\%$ of all DNA adducts), methyl phosphotriester (a backbone adduct; ∼$15\%$), O6-methylguanine (m6G; ∼$7\%$), 3-methyladenine (m3A; ∼$3\%$), and several other quantitatively minor DNA base adducts (18–20). DNA damaging electrophiles can be classified by the degree of selectivity they display for oxygens during nucleophilic substitution reactions, with SN1 agents showing more oxyphilicity than agents that react via an SN2 mechanism [21]. When one considers methylation reactions, SN1 agents react not only with the most nucleophilic targets (e.g. the N7 of guanine), but also with base and phosphate oxygens, including the O6- of guanine and the O4- and O2- of thymine. Reactions at the oxygen atoms on nucleobases, especially the O6- of guanine, are commonly associated with potent mutagenicity and carcinogenicity (22–26). Among the many methylated DNA bases, m6G has a special significance in that it plays two distinct and important roles in the cellular response to DNA alkylating agents. First, its ability to pair with thymine during DNA synthesis (Figure 1C) [27] leads to an m6G:T base pair, which is a target for mismatch repair enzymes that convert the T-containing strand into a nick that eventually matures into a lethal strand break [28,29]. This property explains the mechanism underlying several cytotoxic agents that are used or have been used in cancer chemotherapy, including the N-methyl nitrosamides N-methyl-N-nitrosourea (MNU) and streptozotocin (STZ), and the widely used methyl imidazotetrazine, temozolomide (TMZ) (30–32). Second, if a cell escapes the toxic effect of the m6G:T pairing event, an additional round of replication generates GC→AT mutations (29,33–36), which are often seen in the wake of exposure to cancer-causing DNA alkylating agents, as shown in the current work with the environmental carcinogen, NDMA. The distinctive roles m6G plays in environmental carcinogenesis, as well as in cancer therapy, underscore the importance of finding ways to classify DNA alkylating agents by the patterns of DNA adduct-driven mutations they form in living cells. Several biochemical systems dictate sensitivity or resistance to agents such as NDMA. As one example, DNA repair systems offer defenses against several DNA adducts and the loss or exceeded capacity of these defenses enhances cancer risk (37–39). Many studies with repair-deficient cells or animals show that the levels of specific methylated DNA adducts typically track with mutagenic and carcinogenic endpoints, pointing to the role that these adducts could have in malignant transformation and the role that specific repair systems have to prevent cancer initiation [40,41]. With agents such as NDMA, a second factor that determines risk is the level of expression of enzymatic systems that generate the DNA-interactive metabolites. For example, in liver tissue, which is a prime target for neoplastic transformation by NDMA, hepatocytes in the centrilobular regions are notably vulnerable to DNA adduct formation, in part because they are among the first cells encountered following most methods of exposure to alkylating agents and in part because they typically have high CYP2E1 expression [42,43]. Lung and kidney also express this CYP, albeit to a lower extent, and are also target tissues [44,45]. The current work compared mutations induced by NDMA, MNU, STZ and TMZ (Figure 1A) in various murine organs or in cell culture (Figure 1D-G). While structurally dissimilar, each of these agents is believed to generate a methyldiazonium ion as a common intermediate (Figure 1B), which reacts with DNA by an SN1 chemical mechanism to generate m6G. The present work defined the high-resolution mutagenic fingerprints of NDMA, TMZ, STZ and MNU across 96 3-base contexts in the mouse genome. As expected [40,41,46,47], the mutations of NDMA were primarily GC→AT transitions and were strongly enhanced in cells lacking O6-methylguanine-methyltransferase (MGMT), a direct reversal DNA repair protein that removes the methyl group from m6G, restoring undamaged guanine. The detailed mutational landscape across all 3-base contexts was the same in the presence or absence of MGMT, indicating that the repair protein did not favor some contexts over others. The data are most consistent with the conclusion that the appearance of rugged, but nearly identical, mutagenic landscapes for the four agents studied here was driven more by the sequence-selective reaction with DNA of the electrophilic methyldiazonium ion than by the sequence specificity of the repair protein MGMT. It was also found that the pattern of mutations observed for the four agents studied shows high computational similarity to the pattern of mutations seen in tissues of cancer patients treated with TMZ, a pattern known as COSMIC mutational signature SBS11 [48]. Lastly, it is noteworthy that mechanistically informative data are yielded from the mouse model within ten weeks of exposure to NDMA, underscoring the value of high-resolution mutational analysis for probing the very earliest steps on the pathway to malignancy. ## Animals and in vivo procedures Most experiments utilized C57BL/6J gptΔ transgenic mice [49,50], a gift from T. Nohmi. Some experiments used RaDRR/R;gptg/g;Mgmt−/− mice, which were created by crossing RaDRR/R;gptg/gmice [37] with Mgmt−/− mice [51]. NDMA was synthesized and characterized as previously described [52] (Fig. S1). A total dose of 10.5 mg/kg NDMA, reported to be the maximum tolerated dose [53], diluted in $0.9\%$ saline was administered via intraperitoneal injection to neonatal mice (Figure 1D) [53]. The first dose (3.5 mg/kg in 10 μl) was administered at 8 days of age, and the second dose (7 mg/kg in 20 μl) was administered at 15 days of age. Vehicle treated mice were administered equivalent volumes of saline. Mice were euthanized at ten weeks post-injection by CO2 inhalation. Both male and female mice were used in all experimental groups ($$n = 5$$) in all mouse experiments. Mice were provided irradiated pelleted diet (Isopro RMH 3000; Purina Mills, Inc., St. Louis, MO) and filtered water ad libitum and were maintained in an AAALAC accredited facility. Mice were housed in static microisolator cages with autoclaved hardwood chip bedding and a nestlet. Macroenvironmental conditions included temperature maintenance at 70 ± 2°F, 30–$70\%$ humidity, and a 12:12 h light:dark cycle. Mice were free of the following murine pathogens: ecto- and endoparasites, mouse parvovirus, mouse hepatitis virus, Mouse Rotavirus, Minute Virus of Mice, Ectromelia, Sendai virus, *Pneumonia virus* of mice, Reovirus, Theilovirus, Lymphocytic Choriomeningitis Virus, Mycoplasma pulmonis, Filobacterium rodentium, *Polyoma virus* and Mouse Adenovirus. Pathogen status was verified by full serology testing using sentinel mice every six months, and with an abbreviated serology panel every two months. All animal procedures were performed with approval from the Massachusetts Institute of Technology (MIT) Committee on Animal Care (Protocols 0520-035-23 and 0821-075-24) in accordance with the Guide for the Care and Use of Laboratory Animals 8th Edition and AVMA Guidelines for Euthanasia of Animals: 2020 Edition. ## Cell culture C57BL/6J gptΔ mouse embryonic fibroblasts (MEFs) were established and maintained in high glucose, GlutaMAX DMEM (Gibco) supplemented with $10\%$ FBS (VWR Life Sciences), sodium pyruvate (1 mM, Gibco), penicillin (100 IU) and streptomycin (100 μg/ml) (Corning) as previously described [40]. Cells were shown to be mycoplasma-free by PCR. For growth-inhibition assays, 2.5 × 104 cells were seeded in 6-well plates and incubated overnight. Cells were treated with SN1 alkylating agents as follows. TMZ (Sigma-Aldrich): 50, 100 and 200 μM (stock solutions in DMSO (Sigma-Aldrich)) in FBS-free complete media for 24 h. STZ (Sigma-Aldrich): 500, 1000 and 2000 μM (stock solutions in 0.1 M citrate buffer, pH 4.5) in HBSS, pH 7.4 for 6 h. MNU (Sigma-Aldrich): 125, 250 and 500 μM (stock solutions in 0.1 M citrate buffer, pH 4.5) in HBSS, pH 7.4 for 1 h. After the treatments, cells were washed in PBS (Gibco) and cultured in complete media for another 48 h. Cell growth was analyzed by determining percentages of cell numbers from treatment and untreated control groups by Coulter counter. For gpt assays, 4 × 105 cells were seeded into 15 cm tissue culture dishes (4 dishes per group) and treated with TMZ, STZ and MNU, as described above, at 60–$70\%$ confluence. Viable cells were harvested and washed twice in PBS. ## GptΔ assay: Detection of mutational events that functionally inactivate the gpt reporter gene in the mouse or mouse-derived MEF genomes The gptΔ assay is the technology traditionally used with the gptΔ mouse to probe for small mutations in the 459 bp gpt transgene, of which approximately 160 copies exist in the mouse diploid genome [50]. The gpt gene, which is not expressed in the mammalian cells, is contained in a λ-EG10 viral cassette that can be recovered ex vivo through a lambda packaging reaction. The λ-EG10 phage was packaged in vitro from genomic DNA using a Transpack packaging extract (Agilent Technologies) and then transfected into E. coli YG6020 expressing Cre-recombinase, generating a 6.4 kb plasmid carrying the gpt and chloramphenicol acetyltransferase genes. These bacteria were cultured on selective media containing chloramphenicol (25 μg/ml, Sigma-Aldrich) and 6-TG (25 μg/ml, Sigma-Aldrich) or chloramphenicol alone. 6-TG resistance was confirmed by regrowth of colonies on plates containing chloramphenicol and 6-TG. The mutant frequency of each group was calculated (ratio of total 6-TG-resistant colonies to the average of chloramphenicol resistant colonies). The samples were processed and analyzed in a blinded fashion. In this study, the gptΔ assay was performed on several tissues isolated from the mouse experiments. Specifically, the liver, kidneys, and lungs were collected at ten weeks following the second NDMA injection, flash frozen in liquid nitrogen, and stored at −80°C until analysis. Genomic DNA was extracted from ∼25 mg of liver tissue, one kidney, and whole lungs using the RecoverEase DNA Isolation Kit (Agilent Technologies). The gptΔ assay as described above was also performed on MEFs derived from the gptΔ mouse (see above) that were treated with the toxicants studied here. For the cell culture studies, genomic DNA from 2 × 106 cells per group was prepared using the MegaLong DNA Isolation Kit (G-Biosciences) according to the manufacturer's instructions. ## Duplex sequencing (DS) To determine the identity of mutations and the sequence contexts in which they occur, we utilized a form of error-corrected sequencing called duplex sequencing (DS). Previously, we described how DS can be used directly on genomic samples derived from the gptΔ mouse to yield high-resolution mutational spectra in an unbiased (unselected) manner [49]. Unlike the traditional gptΔ assay, DS evaluates the mutations throughout the entire 6382 bp transgenic region (which contains the gpt gene), without the need for a phenotypic selection. In the present work, 6.4 kb gpt-containing plasmids from gptΔ mouse livers and cell culture were isolated using a Miniprep Kit (Qiagen) following instructions of the manufacturer. Approximately 6 μg of plasmid DNA was diluted in 1× TE buffer-low EDTA, pH 8 (Thermo Scientific), transferred to a microTUBE AFA Fiber Pre-Slit Snap-Cap (Covaris) and sonicated using a Covaris E220 Focused-ultrasonicator. The following program was used to obtain 300–350 bp fragments: peak incident power = 140 W, duty factor = $10\%$, cycles per burst = 200, treatment time = 80 s, temperature = 4–7°C. DS libraries for gptΔ C57BL/6J mice and MEFs were prepared using the NEBNext Ultra II DNA Library Prep (E7103, NEB) or as described in Chawanthayatham et al. [ 49] and sequenced on a NextSeq500 or NovaSeq 6000 Illumina platform by using a 150 bp paired-end protocol (MIT BioMicro Center). DS libraries for the gptΔ mouse lungs and Mgmt−/− mice were prepared using an input of 1 μg or 500 ng genomic DNA, respectively, using a TwinStrand Biosciences, Inc. Mouse Mutagenesis Kit according to the manufacturer's protocol. This kit uses probes that hybridize to 20 genomic regions, each 2.4 kb long. The genomic coordinates are provided in Supplementary Table S1. These probes were purposely selected to map in intronic and intergenic regions, to avoid selection bias. The libraries were sequenced on an Illumina NovaSeq 6000 DNA sequencer on an S4 flow cell using a 150 bp paired-end protocol (MIT BioMicro Center). ## Data processing For each sample, the two fastq files generated by the Illumina sequencer were converted into an unaligned bam file, using Picard Tools. The resulting bam file was processed using UnifiedConsensusMaker.py (version 3.0) as available from the Kennedy Lab at University of Washington (https://github.com/Kennedy-Lab-UW/Duplex-Sequencing). Briefly, this program queries the molecular tags on each read and assembles them into families (strands with the same tag). These are subsequently collapsed into single-strand consensus sequences (SSCS). Then, the software identifies the SSCS corresponding to the complementary strand; these are grouped together as a double-strand consensus sequence (DCS). Unpaired SSCSs were flagged and ignored for the rest of the analysis. Sequence information was accepted only when the information from the two complementary DNA strands was in perfect agreement. The DCSs (as fastq files) were subsequently aligned to the target region (6382 nt in the λ-EG10 genome) using Burrows-Wheeler aligner. Unmapped reads were filtered out. Finally, the properly aligned DCS reads were trimmed (bases 1–8 and 120–137) and collapsed into a pileup file using SamTools. The pileup file was then analyzed to identify the mutation count (CountMuts.py) and their sequence location (mut-position.py). Custom Python scripts were then used to generate a list of unique mutations (a mutation at a given genomic location is counted only once), construct, normalize and plot mutational spectra. Normalization was done by dividing the mutational frequency at each trinucleotide sequence context to the frequency of that trinucleotide sequence context in the target region. For the samples generated using the TwinStrand Biosciences Inc. kit, the analysis was performed using the manufacturer's recommended pipeline running on DNANexus. The resulting mutation-position files (.mut) were then analyzed as described above to generate the normalized mutational spectra. The relative trinucleotide frequencies in the 6.4kb target from the gptΔ mouse and from the TwinStrand muta-mouse1.0 probe set are shown in Figures S2 and S3 in the SI Appendix. ## Histological analysis Sections of kidney and liver from mice ten weeks post treatment were fixed in $10\%$ buffered formalin, embedded in paraffin, and sectioned at 5 μm thickness, stained with hematoxylin and eosin (H&E) by the MIT Division of Comparative Medicine Comparative Pathology Laboratory. H&E-stained sections of liver and kidney were scored by a Board-certified veterinary pathologist blinded to sample identity. Sections of liver were evaluated for the presence of hepatocellular degeneration (cell swelling with cytoplasmic alteration), fibrosis, and lipidosis using distribution-based ordinal scoring: 0 = normal, 1, uncommon detection in < $5\%$ of liver fields (200×); 2, detectable in up to 5–$20\%$ of liver fields; 3, detectable in up to 20–$65\%$ of liver fields; 4, detectable in >$65\%$ of liver fields. The following criteria were used for scoring hepatic inflammation, necrosis, nuclear enlargement (karyomegaly), extramedullary hematopoiesis, Kupffer cell hyperplasia, Ito cell hyperplasia, and bile duct hyperplasia: 0 = normal; 1 = minimal (1–5 foci), 2 = mild (6–12 foci), 3 = moderate (13–18 foci), 4 = severe (>18 foci) (adapted from [54]). A total inflammation score for each liver was generated by combining individual scores of portal, midzonal, and centrilobular inflammation from each submitted section. Sections of kidney were evaluated for the presence of inflammation, glomerulonephropathy, tubular necrosis, tubular degeneration, tubular regeneration, tubular casts and interstitial fibrosis. Slides were examined using an Olympus BX41 microscope attached with an iKona digital camera and photographed. ## Statistical analysis GraphPad Prism 9 was used for statistical analyses of all data. Gpt mutant frequencies and histopathology scores were compared by the Mann–Whitney U-test. The differences between groups were considered significant when the P value was <0.05. ## Illustration tools Graphical images were created with CorelDRAW 2019. The pLOGO plots were generated using Schwartz Lab probability logo generator (https://plogo.uconn.edu/). ## NDMA-induced mutations and mutational patterns To assess the mutagenic effects of NDMA in vivo, we used an established carcinogenic protocol [37,53] in which neonatal male and female gptΔ C57BL/6J mice were administered two intraperitoneal injections of NDMA (10.5 mg/kg total, Figure 1D). Animals were aged for ten weeks following the second dose. The gptΔ C57BL/6J mice carry the λ EG10 transgene with multiple copies of the *Escherichia coli* gpt gene, which was used as a reporter to detect point mutations induced by NDMA. Positive selection of mutant colonies with 6-TG allowed for quantitation of the mutant frequencies of NDMA in various tissues (Figure 1F). NDMA, at ten weeks post-exposure, resulted in increased numbers of mutations in the liver, lung and kidney (Figure 2A–C), all of which are target organs in various animal models of NDMA-induced cancer. There were significant increases of ∼30- and 45-fold in the mutant frequencies in the livers of NDMA-treated males and females, respectively, as compared to controls (Figure 2A, **$$P \leq 0.0079$$). Lung point mutations were also increased by >4-fold in NDMA-treated mice compared to controls (Figure 2B, males **$$P \leq 0.0079$$, females *$$P \leq 0.0159$$). Kidneys experienced an approximate doubling in the mutant frequencies in both males and females; however, only the males were found to be significantly increased (Figure 2C, **$$P \leq 0.0079$$). **Figure 2.:** *Organ-specific mutagenicity of NDMA and HRMS of NDMA-treated livers and lungs. (A–C) NDMA induced mutant frequencies of the liver, lung and kidney in gptΔ C57BL/6J wild-type mice. (D) The proportion of single-nucleotide substitution mutations in the liver and lung was measured directly by using duplex sequencing (DS). (E) DS revealed the NDMA-induced high resolution mutational spectrum in liver and lung. The spectrum was dominated by GC→AT mutations with a lower frequency of AT→GC transitions. The sequencing data were plotted as the mutant base (X) accompanied by its 5’ and 3’ neighbors (N); that is, 5’-NXN-3’. There are 16 combinations for each of the six types of base substitution mutations, resulting in a total of 96 possible triplex contexts. The data shown are averaged from the livers of five males and five females and the lungs of two males and two females treated with NDMA. (F) The probability LOGO (pLOGO) was generated from all 15-base pair sequence contexts adjacent to the mutated base, GC→AT. Guanine in gray highlight represents the fixed G position. For panels A-C, statistical comparisons were done with Mann–Whitney U-test, *P = 0.0159, **P = 0.0079. For panel E, bars reflect averages, error bars denote 1 SD. For panel F, the 15-base sequence contexts with G→A mutations fixed at the zero position were extracted and from all datasets. Shown is the compilation of all sequence contexts. Inter-individual replicate sequences were included in the analysis. The red bar (log-odds value of ± 3.05) indicates the P = 0.05 statistically significant threshold following Bonferroni correction. The foreground sequences (fg) represent NDMA-induced mutations (liver n(fg) = 3751; lung n(fg) = 3451). The background sequences (bg) represent the genome sequenced (liver n(bg) = 13 041; lung n(bg) = 10 437).* Duplex sequencing, a very sensitive DNA sequencing method that is four orders of magnitude more accurate than conventional next-generation sequencing [49,55], was used to generate high-resolution mutational spectra from the lungs and livers of NDMA-treated gptΔ mice. The base-substitution profile (Figure 2D) and high-resolution mutational spectrum (Figure 2E) of NDMA obtained by deep sequencing was dominated by GC→AT transition mutations, which were presumably caused by the mutagenic DNA adduct m6G pairing with thymine during replication in vivo [27,56,57]. As seen in Figure 2D–E, these transition mutations were present as early as ten weeks following NDMA administration. These early-onset GC→AT mutations occurred in a distinctive pattern, with the sequence context dependency of 5’-PuGN-3’ (Fig. 2E; where Pu = a purine and N = any base). The strongly contoured mutational spectrum of NDMA was highly reproducible and did not appear to be influenced by the sex of the animal. By contrast, the HRMS of the vehicle control animals featured a diverse pattern of transition and transversion mutations with no apparent sequence context tropism, except for C→T mutations in CpG sites, which are typically associated with 5-methylcytosine deamination (SI Appendix Fig. S4A and C). The HRMS of the lungs displayed a similar pattern (Figure 2E and Fig. S5), despite only a 4-fold increase of NDMA induced mutations compared to controls (Figure 2B). To understand the influence of neighboring bases on NDMA induced mutagenesis, we extracted the 15-base sequence contexts with G→A mutations fixed at the zero position from all datasets (Figure 2F). The probability LOGO (pLOGO) [58] was utilized to visualize the biased appearance of adjacent bases. Guanine and adenine bases were significantly overrepresented at the 5’-proximal position to the observed G→A mutations in both the liver and lung spectra. Other significantly overrepresented bases are summarized in Tables 1 and 2. By contrast, the most frequent and significant base found at the proximal 5’-position in control (vehicle-treated) animals was cytosine (SI Appendix Figure S6A, B and Supplementary Table S2), which is a distinguishing molecular feature of COSMIC SBS1 [48], a common background mutational signature in which C→T mutations occur in CpG sites due to the deamination of 5-methylcytosine. ## Histopathologic changes in liver consistent with xenobiotic insult NDMA induced carcinogenesis in animal models has a higher frequency in males compared to females. However, our results demonstrate that NDMA induces an equivalent mutational burden and spectra in both sexes. Histological analysis of the liver was performed to evaluate any potential differences. At the time of euthanasia, there was no difference in body weights of treated animals compared to vehicle controls (Figure 3A). In NDMA-treated males, the liver-to-body weight ratio decreased significantly, whereas the kidney-to-body weight ratio increased (Figure 3B and D). Females treated with NDMA did not show the same trends as males, but they did show increased lung-to-body weight ratios (Figure 3C). **Figure 3.:** *Macroscopic and histological changes at 10 weeks post NDMA treatment. (A) Neonatal gptΔ C57BL/6J mice were administered a total dose of 10.5 mg/kg NDMA split between day 8 (1/3 dose) and day 15 (2/3 dose). Samples were collected 10 weeks post-NDMA treatment. (B–D) Organ to body weight ratios. (E–J) Representative hematoxylin and eosin (H&E) sections of liver and (K–M) histopathology scores from male and female C57BL/6J gptΔ mice treated with saline (control) or NDMA as described in material and methods. (F) NDMA-treated male mice exhibited areas of periacinar hepatocellular degeneration (swelling) with or without karyomegaly (arrowheads) and (G) bizarre mitotic figures (arrows). (I) Sections of the liver from an NDMA-treated female mice contained multifocal areas of hepatocellular degeneration with or without karyomegaly (arrowheads), bizarre mitoses (J, arrow), and lymphocytic and histiocytic infiltrates (open arrow). Original magnification x400, scale bars = 50 μm. Whitney U-test, *P = 0.0159, **P = 0.0079.* Macroscopic changes were not observed in any organs. Histological examination of livers from NDMA-treated mice, however, revealed multifocal areas of periacinar (centrilobular to midzonal) hepatocellular degeneration (Figure 3F–G, I–J) evidenced by hydropic degeneration (swelling) and cytoplasmic alteration (Figure 3F and 3L; **$$P \leq 0.0079$$) and karyomegaly (Figure 3I and 3M; **$$P \leq 0.0079$$). The location of these changes within the liver is consistent with the high concentration of P450s in the pericentral hepatocytes resulting in the greatest exposure to the reactive metabolite of NDMA. Occasionally, hepatocytes from NDMA-treated mice exhibited atypical mitoses (Figure 3G and J) possibly resulting in chromosome missegregation and aneuploidy associated with oncogenesis [59]. Scattered within the hepatic parenchyma, there were low numbers of lymphocytic and histiocytic infiltrates in NDMA-treated mice (Figure 3I). ## Role of MGMT as a protection against NDMA in vivo The DNA adduct m6G is subject to direct reversal repair by MGMT (Figure 4A), and pharmacological inhibition of this protein in cells derived from the gptΔ C57BL/6J mouse results in an increase in m6G levels in DNA following methylating agent exposure [40]. To assess the impact of MGMT on mutational frequency in the liver, MGMT-deficient mice [51] containing the gptΔ and Rosa26 Direct Repeat (RaDR) [60] transgenes were exposed to NDMA (Figure 1D) [37]. We observed a 3-fold (females) to 4-fold (males) enhanced mutant frequency in NDMA-treated Mgmt−/− mice compared with NDMA-treated wild-type (WT) mice (Figure 4B). The mutant frequencies of NDMA in the WT gptΔ and WT gptΔ/RaDR mice, the two strains used in this study, were indistinguishable (SI Appendix Figure S7). To understand the role of DNA repair in the sequence context-dependent formation of NDMA mutations, DS was performed on genomic DNA isolated from livers of NDMA-treated Mgmt−/- mice. As was seen in the NDMA-treated WT mice, GC→AT mutations were the dominant single-base substitution recorded at 10 weeks post-NDMA exposure, with a preferential distribution once again in the 5’-PuGN-3’ context (Figure 4D–F). Overall, the mutational patterns of NDMA in Mgmt−/−and WT mice were nearly identical, displaying a cosine similarity of 0.98 (Figure 5). In both genotypes, but most conspicuously seen in the Mgmt−/- mice, mutations were more abundant in the 5’-GGN-3’ sequence context, as compared with the 5’-AGN-3’ context (Figure 4C, E–F; Figure 2D–E). The reason for this context specificity of mutations is unclear, but some possibilities are considered in the Discussion. The HRMS of vehicle control-treated Mgmt−/− mice (SI Appendix Fig. S4B and D) showed a diverse array of background mutations like that observed in WT mice (SI Appendix Figure S4A and C). Taken together, the evidence points to the spectra seen in Figures 2E and 4E–F as defining features reflecting exposure to NDMA or, as shown below, to other agents that generate a methyldiazonium ion prior to reacting with DNA. Similar to the WT mice, analysis of neighboring bases revealed that guanine and adenine were significantly overrepresented at the most proximal 5’-position of the G→A mutation (Figure 4D and Table 3), whereas the control spectra frequently featured C→T mutations at the 5’ C of CpG hotpots (Figure S6C and Table S2). **Figure 4.:** *NDMA-induced mutations and mutational spectra in livers of MGMT-deficient mice. (A) The biochemical mechanism of MGMT, the repair protein for m6G. NDMA generates an electrophile in vivo that methylates DNA to form, among other DNA lesions, m6G. A nucleophilic cysteine thiol residue on MGMT attacks the methyl group of m6G, resulting in a methylated MGMT protein and an undamaged guanine. (B) Mgmt−/− gptΔ C57BL/6J mice were administered NDMA (10.5 mg/kg total dose) using the regimen of Figure 1D, and liver samples were collected 10 weeks following treatment. NDMA-induced point mutant frequencies in the livers of male and female MGMT-deficient and -proficient mice are shown. (C) The proportion of single-nucleotide mutations in livers of saline control and NDMA-treated Mgmt−/− mice measured directly by using DS. (D) The pLOGO analysis produced from all 15-base pair sequence contexts adjacent to the mutated base, GC→AT. Guanine in gray highlight represents the fixed G position. (E, F) HRMS from the livers of three male and three female MGMT-deficient mice treated with NDMA. For panel B, statistical comparisons were done with the Mann–Whitney U-test, *$$P \leq 0.0159$$, **$$P \leq 0.0079.$$ For panel D, the 15-base sequence contexts with G→A mutations fixed at the zero position were extracted from all datasets. Shown is the compilation of all sequence contexts. Inter-individual replicate sequences were included in the analysis. The red bar (log-odds value of ±3.05) indicates the $$P \leq 0.05$$ statistically significant threshold following Bonferroni correction. The foreground sequences (fg) represent NDMA-induced mutations (liver n(fg) = 52 253). The background sequences (bg) represent the genome sequenced (liver n(bg) = 10 437). For panels E and F, bars reflect averages ($$n = 3$$ for each group), error bars denote 1 SD.* **Figure 5.:** *Cosine similarity matrix of the mutational spectra of NDMA treated WT liver (NDMA-WT-liver), WT lung (NDMA-WT-lung) and MGMT-deficient mice (NDMA-Mgmt); MEFs treated with temozolomide (TMZ), N-methyl-N-nitrosourea (MNU), and streptozotocin (STZ); and human mutational signature SBS11. Before performing the cosine similarity comparisons, all mutational spectra were baseline corrected by subtracting the corresponding background (vehicle-treated) spectrum, and then normalized to reflect mutation frequency per trinucleotide.* TABLE_PLACEHOLDER:Table 3. ## Comparison of mutational patterns of NDMA with other SN1 alkylating agents: SN1 agents show cosine similarity to human mutational signature SBS11 Animal models are an indispensable tool for recapitulating the toxicological effects of environmental agents in humans. However, to reduce the number of animals utilized in research and make our research more efficient, we have generated a mouse embryonic fibroblast cell line from the gptΔ mice [40]. With the understanding that cell culture models lack multiple in vivo features, such as xenobiotic metabolism, and tissue distribution and exposure, we set out to analyze several SN1 alkylating agents of clinical relevance that do not require metabolic activation (MNU, STZ, TMZ, all of which are or have been used in cancer therapy). To investigate these structurally diverse, but similar-acting alkylating agents, we first performed a dose-response toxicity study in the MEFs. Cells treated with doses generating no more than $50\%$ growth inhibition were subsequently collected for mutational analyses, and HRMS were generated for each of the compounds (Figure 6A–C). The mutational spectra of all three alkylating agents predominantly displayed GC→AT mutations, with a pattern strongly resembling the mutational spectrum of NDMA (Figures 2E and 4E, F). The mutations of MNU, STZ and TMZ were primarily in the 5’-PuGN-3’ sequence context (Figure 6A–C); their detailed spectra displayed a high cosine similarity (>0.9) with each other, with the NDMA spectra from WT animals (in both liver and lung), with the spectra from the livers of MGMT-deficient animals, as well as with COSMIC mutational signature SBS11 (Figure 5). COSMIC mutational signature SBS11 (Figure 6D) is a mathematically extracted pattern from the sequences of thousands of human tumors, and is found in samples from patients that were treated with SN1 therapeutic agents, such as temozolomide and dacarbazine [48]. These spectra are also markedly different (cosine similarity < 0.5) than the background mutational spectra in each of the experimental systems described here (i.e. mouse liver, lung, MEFs) (Figure S8). **Figure 6.:** *HRMS of MEFs treated with alkylating agents MNU, STZ and TMZ. The plots depict the HRMS data and their dose-dependent mutagenicities at the gpt locus (insets) for MNU (A), STZ (B) and TMZ (C). (D) COSMIC mutational signature SBS11 of human cancer.* ## DISCUSSION There is a great need for biomarkers that appear in the initial stages of the carcinogenic process. From a practical perspective, early onset biomarkers could invoke increased monitoring or intervention by surgery, or other forms of treatment that serve to thwart tumor development. From a mechanistic perspective, such biomarkers provide insight into the biochemical and genetic events that lead to disease. Understanding those events is an important step towards establishing mechanistically informed novel cancer prevention and treatment strategies. The present work is based on a classical view of chemical carcinogenesis as involving an early phase, during which DNA adducts of the carcinogen trigger a founder or initiation spectrum of mutations, followed by a long period, during which further genetic and non-genetic events lead to cancer development. During tumor outgrowth, promotional events, such as an activated immune system, cause additional diversity among the heterogeneous population of cells in the malignant mass. As shown here with NDMA, and in work on other agents [49,61,62], the founder spectrum can be a highly distinctive genetic fingerprint. Indeed, that fingerprint could help to identify which agents, among the many to which people are exposed, contribute to the causation of specific human cancers. Of note is the fact that with NDMA, and in earlier work on the liver carcinogen, aflatoxin B1 [49], the founder spectrum appears as early as ten weeks into the carcinogenic process—well before morphological abnormalities develop—offering the possibility that the founder spectrum can be used as a harbinger of subsequent disease. In an Adverse Outcome Pathway (AOP) framework, the early appearance of the founder spectrum constitutes a key event that correlates with the increased risk of the adverse outcome, which is cancer. Liquid biopsies, where circulating mutant DNA or cells are analyzed, offer one avenue by which this predictive goal could be realized. Examination of the high-resolution mutational spectrum of NDMA reveals a rugged, distinctive mutational landscape (Figures 2 and 4). For NDMA, as well as for other mutagenic carcinogens, the distribution of point mutations across sequence space reflects the chemical steps that generate the observed mutational spectrum. We have proposed that, in general, three chemical factors mold the sequence-dependent mutational patterns: 1) DNA lesion formation, 2) lesion repair, and 3) error-prone replication across the DNA lesion [63]. When considering these mechanistic elements, it is not surprising that the mutational spectra of NDMA and the three clinical agents (MNU, STZ and TMZ) reported in this study were, in fact, highly similar. Each is a methylating agent capable of forming the same reactive electrophile that reacts with DNA by an SN1 chemical mechanism. In sum, the mutational spectra we observed reflect the properties of the methyldiazonium ion in cells. It is notable that a reaction of DNA with this ion can generate many different DNA lesions (detailed in the Introduction), but the genetic evidence seen in the present work is that the chemical modification responsible for the bulk of mutagenesis is m6G. As discussed in more detail below, this notion is reinforced by the observation that the mutational frequency in liver increases strikingly in Mgmt−/− animals treated with NDMA, and the primary client adduct for this repair protein is m6G. The impact of MGMT on mutational frequency, the dominance of GC→AT mutations, and the scarcity of all other types of point mutations in the HRMS indicate that the mutagenic contributions of other DNA lesions known to be present (e.g. m3A, O4-methylthymine (m4T), and the abasic site formed by m7G depurination) are very minor under the conditions of this experiment. As indicated above, comparison of mutational spectra also provides mechanistic insight into the contribution of the repair protein MGMT to the NDMA-induced mutational landscape. Since the distribution of the mutations in the GC→AT region across all 16 m6G 3-base contexts is largely unchanged in Mgmt−/− and WT mice (Figure 5, cosine similarity 0.98), it is inferred that the biochemical activity of MGMT does not have a strong sequence context specificity. Moreover, the qualitative agreement of the mutational spectra for SN1 methylating compounds between our mouse model and in vitro systems indicates that the MEF cell model provides an accurate representation of mutagenic processes in vivo. It is noteworthy that the mutational spectra we recorded for MNU and TMZ in MEFs are markedly different than the spectra reported for the same compounds (at similar doses) in several other studies [64,65]. The mutational spectra of MNU and TMZ in human induced pluripotent stem cells were dominated by AT→GC mutations, while featuring very few GC→AT mutations [64]. This difference may reflect high levels of MGMT in those cell lines. A second cell-based study of TMZ-induced mutations found an SBS11-like mutational spectrum only in tumor-derived cells that were deficient in mismatch repair (MMR) [65], suggesting that MMR deficiency may be a mechanistic requirement for SBS11. While MMR deficiency is an established survival mechanism by which cancer cells evolve resistance to alkylating agents such as TMZ [65], it is unlikely to be the only pathway that allows cells to survive alkylation damage [66,67]. In the present work, while the mouse livers or MEFs were exposed to substantial doses of alkylating agents, cell toxicity was not high enough to select for MMR deficiency. Additionally, our DS methodology is applied directly to a heterogenous population of cells, without any clonal amplification, which once again prevents the selection of an unusually resistant, MMR-deficient clone. Moreover, the mutational spectra we recorded do not show discernible contributions from MMR-associated COSMIC mutational signatures (i.e. SBS6, 14, 15, 20, 21, 26 and 44), supporting the view that the SBS11-like mutational spectra we observed were unlikely to reflect MMR deficiency in addition to alkylating agent exposure. Because earlier work in E. coli indicates that mutant frequency from polymerase-mediated replication past m6G is essentially identical in all 16 three-base contexts [57], and we see here that expression of MGMT affects the amount, but not the distribution of mutations, we conclude that the molecular origin of the patterns of mutations observed in the NDMA spectrum is primarily the sequence context-dependent formation of m6G. This formation follows the preferential binding and reactivity of the putative methyldiazonium ion with guanines in different DNA sequence contexts. An examination of the literature on the reaction of electrophilic methylating agents with naked DNA (68–70) is roughly in accord with the current observations in vivo. Given this rationale, even among the mutational hotspots at 5’-PuGN-3’ sequences, the observed increased probability of mutations at the 5’-GGN-3’ sites relative to the 5’-AGN-3’ sites could be attributed to differential chemical reactivity. The molecular basis for this effect, however, remains to be elucidated and warrants further investigation. While the mutational spectrum of NDMA is dominated by the GC→AT mutations, it does feature a small fraction of AT→GC mutations in a pattern that cannot be explained by the background spectra (mutational spectra in vehicle-treated mice). After background subtraction, AT→GC mutations constitute $4.6\%$ of the total number of mutations. Among the adenine and thymine lesions formed by SN1 alkylating agents, m3A and m4T are chemically the most likely to contribute to the observed, but infrequent, AT→GC mutations. While m3A is primarily known as a strong replication blocker, it can also be mutagenic; recent work from this laboratory shows that mice knocked out for AAG (a.k.a MPG), a glycosylase known to have m3A as its principal target, display an enhanced mutant frequency over wild-type mice when exposed to NDMA [37]. Future examination of the mutational spectra of AAG-deficient mice may reveal an increase in the fraction of mutations at A:T base pairs, which would support the possible role of AAG as a protection against those mutations. A second plausible candidate for the lesion responsible for AT→GC mutations is m4T. Compared to other methylated bases formed by SN1 agents, the amount of m4T formed by SN1 alkylating agents is small (<$1\%$). However, in vitro and in vivo studies have shown that m4T primarily leads to AT→GC mutations [71,72], suggesting that this adduct, qualitatively, has the mutagenic specificity to explain the AT→GC pattern in the NDMA-treated mice. Interestingly, as compared to WT mice, the mutational spectrum of NDMA in Mgmt−/- mice has a significantly reduced proportion of AT→GC mutations ($1.7\%$). This lower fraction, however, is primarily a reflection of the increased proportion of GC→AT mutations attributable to an enhanced yield of m6G in the MGMT-deleted mice. One variable investigated in this work was the role of sex in the susceptibility of mice to the early-stage mutagenic properties of NDMA. As with most animal models, as well as humans, males are more susceptible than females to the eventual development of cancer (53,73–75). It is noteworthy that the ten-week post-dosing mutation frequencies in the livers of male and female mice treated with NDMA were similar (Figure 2), which is in contrast to the high incidence of hepatic tumors in males and their lower frequency in females reported at much later time points in the literature [53,73,75]. Using a similar model system and dosing regimen, we observed the lack of a sex-based difference in mutational load with the environmental hepatocarcinogen aflatoxin B1, which also shows a male-selective carcinogenic tropism [76]. The reasons for the greater carcinogenic potency of NDMA and aflatoxin to males as compared to females, despite identical mutation frequencies at ten weeks post initiation, is unknown, but hints are offered from work on a structurally similar nitrosamine, diethylnitrosamine [77]. In that case, the sex-based disparity has been attributed to increased levels of the pleiotropic cytokine interleukin-6 (IL-6) in males, and an estrogen-mediated inhibition of IL-6 in females (78–80). With regard to the present work, it is possible that a differential inflammatory response in adult males versus females explains the sex-based differential in cancer generation in NDMA-treated animals. Cognizant of the possible role of inflammation in tumorigenesis, we looked for, but found no histologic evidence of a greater inflammatory response to NDMA in males as compared to females at ten weeks following treatment (Figure 3). This result points to the importance of post-initiation intrinsic and extrinsic factors that are probably necessary for tumor development once the stage is set by the initial mutations produced by NDMA exposure at a young age. During a hypothetical inflammatory stage, DNA-damaging products such as reactive oxygen species (ROS) and other endogenous mutagens may be generated, and enhanced production of these species could superimpose a layer of inflammatory genomic damage on top of the damage previously acquired from toxicant-DNA interactions. As one example, ROS produce 7,8-dihydro-8-oxoguanine lesions, which, when left unrepaired, result in GC→TA mutations that accumulate in a pattern known as mutational signature SBS18 [48]. Analysis of the NDMA mutational patterns at ten weeks did not show evidence of SBS18 or other mutational patterns typically associated with inflammation [48,64], suggesting that additional mutational processes are needed for full malignant transformation to occur following puberty [81]. It is reasoned that the combination of NDMA alkylation and subsequent inflammatory DNA lesions, or the induction of inflammation-associated deaminases [82], results in the high susceptibility of male mice during carcinogenesis. Thus, while we have identified a pattern of mutations consistent with the miscoding properties of m6G at ten weeks after exposure, we would expect to see an increasing complexity in the mutational pattern during tumor development [83]. That complexity was indeed was seen in our earlier work on aflatoxin B1 carcinogenesis [49]. While our data do not explain the well-known sex-based difference in susceptibility of males to cancer, they do help to explain the organotropism of NDMA, which induces cancer in several different organ systems. We employed a neonatal mouse model [53], in which NDMA administration to C57BL/6J male mice results in nearly $80\%$ of the mice developing hepatic tumors, along with a $7\%$ incidence of lung tumors. This organotropism for NDMA-induced tumors is consistent in the gpt mutant frequency observed in our transgenic gptΔ C57BL/6J mice (Figure 2A–C), with a much higher frequency of mutations in the liver as compared to the lungs and kidneys [44,45]. These differences in organ susceptibility probably reflect the differences in CYP2E1 expression, with expression significantly higher in the liver compared to kidneys and lung [84,85], and the details of the methods used for NDMA administration, including timing, dose, and route. Regarding CYP2E1 expression, it is noteworthy that this enzyme is significantly over-expressed in Type 2 diabetics [86], potentially putting them at increased risk to the genetic effects of NDMA. To compound the problem, NDMA has been reported to contaminate some batches of metformin [13], a leading drug used to control Type 2 and gestational diabetes. ## CONCLUSION The present work on NDMA was inspired by the potential role that this toxicant plays in environmental carcinogenesis. We observed, however, that the mutational spectrum of NDMA in mice shows high similarity to human mutational signature SBS11, which has been linked to the aftereffects of cancer chemotherapy in humans [48] and to exposure of animals to certain other alkylating agents [87,88]. Using MEFs derived from the same mouse used for the NDMA study, we confirmed that several agents that are used or have been used in chemotherapy, MNU, STZ and TMZ, produce the same spectrum as NDMA. When viewed as a whole, our work and that of others supports the model that SBS11 reflects a mutational process driven by SN1 alkylating agents that produce m6G. Hence, the observation of an SBS11-like spectrum could reflect exposure to a host of different methylating agents, including those present as environmental contaminants as well as those used in therapy. An additional observation is that the distinctive mutational spectrum of NDMA is evident very early, only ten weeks after carcinogen administration. Accordingly, it certainly reflects past exposure to DNA alkylating agents and, if it is a durable pattern and appears at the time of tumor development, it could be useful as a biomarker of later-life disease. Further, the work underscores the quantitative importance of the repair protein MGMT as a protection against these agents. And, lastly, the data are most consistent with a model whereby the distinctive mutational landscape of these agents is dictated by the preference for reaction of the DNA-reactive putative methyldiazonium ion with specific sequence contexts, most notably 5’-PuG-3’. ## DATA AVAILABILITY The data underlying this article are available in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) repository, under the BioProject PRJNA910200 (http://www.ncbi.nlm.nih.gov/bioproject/910200), and on FAIRDOMHub (https://fairdomhub.org/studies/1130). The code used for data analysis is available at the Zenodo DOI: 10.5281/zenodo.7711071. ## SUPPLEMENTARY DATA Supplementary Data are available at NAR Cancer Online. ## FUNDING National Institute of Environmental Health Sciences [P42-ES027707, P30-ES002109, T32-ES007020 to A.L.A., J.E.K., and an under-represented minority supplement P42-ES027707-05S1 to A.L.A.]; National Cancer Institute [R01-CA080024]; MIT Jameel Water and Food Systems Laboratory. Funding for open access charge: National Institute of Environmental Health Sciences [P42-ES027707, R01-CA080024]. Conflict of interest statement. None declared. ## References 1. Brunnemann K.D., Hecht S.S., Hoffmann D.. **N-nitrosamines: environmental occurrence, in vivo formation and metabolism**. *J. Toxicol. Clin. Toxicol.* (1982.0) **19** 661-688. PMID: 6761448 2. Tricker A.R., Spiegelhalder B., Preussmann R.. **Environmental exposure to preformed nitroso compounds**. *Cancer Surv.* (1989.0) **8** 251-272. PMID: 2696580 3. Bogovski P., Bogovski S.. **Animal species in which N-nitroso compounds induce cancer**. *Int. J. Cancer* (1981.0) **27** 471-474. PMID: 7275353 4. Hecht S.S.. **Approaches to cancer prevention based on an understanding of N-nitrosamine carcinogenesis**. *Proc. Soc. Exp. Biol. Med.* (1997.0) **216** 181-191. PMID: 9349687 5. **IARC monographs on the evaluation of carcinogenic risks to humans. Ingested nitrate and nitrite, and cyanobacterial peptide toxins**. *IARC Monogr. Eval. Carcinog. Risks Hum.* (2010.0) **94** v-vii, 1–412. PMID: 21141240 6. Mitch W.A., Sharp J.O., Trussell R.R., Valentine R.L., Alvarez-Cohen L., Sedlak D.L.. **N-nitrosodimethylamine (NDMA) as a drinking water contaminant: a review**. *Environ. Eng. Sci.* (2003.0) **20** 389-404 7. 7. MA-DPH  The Wilmington Childhood Cancer Study: An Epidemiologic Investigation of Childhood Cancer from 1990-2000. 2021; United States Bureau of Environmental Health.. *The Wilmington Childhood Cancer Study: An Epidemiologic Investigation of Childhood Cancer from 1990-2000* (2021.0) 8. Engelward B.P.. **Implications of an epidemiological study showing an association between in utero NDMA exposure and childhood cancer**. *Environ. Mol. Mutagen.* (2021.0) **62** 288-292. PMID: 33963777 9. Sen N.P., Iyengar J.R., Miles W.F., Panalaks T.. **Nitrosamines in cured meat products**. *IARC Sci. Publ.* (1976.0) **14** 333-342 10. McCutcheon J.W.. **Nitrosamines in bacon: a case study of balancing risks**. *Public Health Rep.* (1984.0) **99** 360-364. PMID: 6431483 11. Hecht S.S., Hoffmann D.. **Tobacco-specific nitrosamines, an important group of carcinogens in tobacco and tobacco smoke**. *Carcinogenesis* (1988.0) **9** 875-884. PMID: 3286030 12. Snodin D.J., Elder D.P.. **Short commentary on NDMA (N-nitrosodimethylamine) contamination of valsartan products**. *Regul. Toxicol. Pharmacol.* (2019.0) **103** 325-329. PMID: 30629969 13. White C.M.. **Understanding and preventing (N-nitrosodimethylamine) NDMA contamination of medications**. *Ann. Pharmacother.* (2020.0) **54** 611-614. PMID: 31771343 14. Zmysłowski A., Książek I., Szterk A.. **N-nitrosodimethylamine contamination in the metformin finished products**. *Molecules* (2020.0) **25** 5304. PMID: 33202951 15. Adamson R.H., Chabner B.A.. **The finding of N-nitrosodimethylamine in common medicines**. *Oncologist* (2020.0) **25** 460-462. PMID: 32267983 16. Yang C.S., Smith T., Ishizaki H., Hong J.Y.. **Enzyme mechanisms in the metabolism of nitrosamines**. *IARC Sci. Publ.* (1991.0) **105** 265-274 17. Sohn O.S., Ishizaki H., Yang C.S., Fiala E.S.. **Metabolism of azoxymethane, methylazoxymethanol and N-nitrosodimethylamine by cytochrome P450IIE1**. *Carcinogenesis* (1991.0) **12** 127-131. PMID: 1988172 18. Singer B., Grunberger D.. **Intrinsic Properties of Nucleic Acids**. *Molecular Biology of Mutagens and Carcinogens* (1983.0) 15-44 19. Beranek D.T.. **Distribution of methyl and ethyl adducts following alkylation with monofunctional alkylating agents**. *Mutat. Res.* (1990.0) **231** 11-30. PMID: 2195323 20. O’Connor P.J.. **Interaction of chemical carcinogens with macromolecules**. *J. Cancer Res. Clin. Oncol.* (1981.0) **99** 167-186. PMID: 7251633 21. Dietze P.E., Jencks W.P.. **Swain-Scott correlations for reactions of nucleophilic reagents and solvents with secondary substrates**. *J. Am. Chem. Soc.* (1986.0) **108** 4549-4555 22. Loveless A.. **Possible relevance of O6 alkylation of deoxyguanosine to the mutagenicity and carcinogenicity of nitrosamines and nitrosamides**. *Nature* (1969.0) **223** 206-207. PMID: 5791738 23. Zhai Q., Wang P., Wang Y.. **Cytotoxic and mutagenic properties of regioisomeric O2-, N3- and O4-ethylthymidines in bacterial cells**. *Carcinogenesis* (2014.0) **35** 2002-2006. PMID: 24710626 24. Dosanjh M.K., Singer B., Essigmann J.M.. **Comparative mutagenesis of O6-methylguanine and O4-methylthymine in Escherichia coli**. *Biochemistry* (1991.0) **30** 7027-7033. PMID: 2069960 25. Shrivastav N., Li D., Essigmann J.M.. **Chemical biology of mutagenesis and DNA repair: cellular responses to DNA alkylation**. *Carcinogenesis* (2010.0) **31** 59-70. PMID: 19875697 26. Jasti V.P., Spratt T.E., Basu A.K.. **Tobacco-specific nitrosamine-derived O2-alkylthymidines are potent mutagenic lesions in SOS-induced Escherichia coli**. *Chem. Res. Toxicol.* (2011.0) **24** 1833-1835. PMID: 22029400 27. Loechler E.L., Green C.L., Essigmann J.M.. **In vivo mutagenesis by O6-methylguanine built into a unique site in a viral genome**. *Proc. Natl. Acad. Sci. U.S.A.* (1984.0) **81** 6271-6275. PMID: 6093094 28. Kaina B., Christmann M., Naumann S., Roos W.P.. **MGMT: key node in the battle against genotoxicity, carcinogenicity and apoptosis induced by alkylating agents**. *DNA Repair (Amst.)* (2007.0) **6** 1079-1099. PMID: 17485253 29. Fu D., Calvo J.A., Samson L.D.. **Balancing repair and tolerance of DNA damage caused by alkylating agents**. *Nat. Rev. Cancer* (2012.0) **12** 104-120. PMID: 22237395 30. Kolarić K.. **Combination chemotherapy with 1-methyl-1-nitrosourea (MNU) and cyclophosphamide in solid tumors**. *Z. Krebsforsch. Klin. Onkol. Cancer Res. Clin. Oncol.* (1977.0) **89** 311-319. PMID: 143813 31. Strobel H., Baisch T., Fitzel R., Schilberg K., Siegelin M.D., Karpel-Massler G., Debatin K.M., Westhoff M.A.. **Temozolomide and other alkylating agents in glioblastoma therapy**. *Biomedicines* (2019.0) **7** 69. PMID: 31505812 32. Aoki T., Kokudo N., Komoto I., Takaori K., Kimura W., Sano K., Takamoto T., Hashimoto T., Okusaka T., Morizane C.. **Streptozocin chemotherapy for advanced/metastatic well-differentiated neuroendocrine tumors: an analysis of a multi-center survey in Japan**. *J. Gastroenterol.* (2015.0) **50** 769-775. PMID: 25348496 33. Margison G.P., Santibáñez Koref M.F., Povey A.C.. **Mechanisms of carcinogenicity/chemotherapy by O6-methylguanine**. *Mutagenesis* (2002.0) **17** 483-487. PMID: 12435845 34. Basu A.K., Essigmann J.M.. **Establishing linkages among DNA damage, mutagenesis, and genetic diseases**. *Chem. Res. Toxicol.* (2022.0) **35** 1655-1675. PMID: 35881568 35. Gourzones C., Bret C., Moreaux J.. **Treatment may be harmful: mechanisms/prediction/prevention of drug-induced DNA damage and repair in multiple myeloma**. *Front. Genet.* (2019.0) **10** 861. PMID: 31620167 36. Peng Y., Pei H.. **DNA alkylation lesion repair: outcomes and implications in cancer chemotherapy**. *J. Zhejiang. Univ. Sci. B* (2021.0) **22** 47-62. PMID: 33448187 37. Kay J.E., Corrigan J.J., Armijo A.L., Nazari I.S., Kohale I.N., Torous D.K., Avlasevich S.L., Croy R.G., Wadduwage D.N., Carrasco S.E.. **Excision of mutagenic replication-blocking lesions suppresses cancer but promotes cytotoxicity and lethality in nitrosamine-exposed mice**. *Cell Rep.* (2021.0) **34** 108864. PMID: 33730582 38. Sharma S., Salehi F., Scheithauer B.W., Rotondo F., Syro L.V., Kovacs K.. **Role of MGMT in tumor development, progression, diagnosis, treatment and prognosis**. *Anticancer Res.* (2009.0) **29** 3759-3768. PMID: 19846906 39. Gerson S.L.. **MGMT: its role in cancer aetiology and cancer therapeutics**. *Nat. Rev. Cancer* (2004.0) **4** 296-307. PMID: 15057289 40. Thongararm P., Fedeles B.I., Khumsubdee S., Armijo A.L., Kim L., Thiantanawat A., Promvijit J., Navasumrit P., Ruchirawat M., Croy R.G.. **Modulation of N-methyl-N-nitrosourea mutagenesis in mouse embryo fibroblasts derived from the gpt delta mouse by an inhibitor of the O6-methylguanine methyltransferase, MGMT**. *Chem. Res. Toxicol.* (2020.0) **33** 625-633. PMID: 31841318 41. Sandercock L.E., Hahn J.N., Li L., Luchman H.A., Giesbrecht J.L., Peterson L.A., Jirik F.R.. **Mgmt deficiency alters the in vivo mutational spectrum of tissues exposed to the tobacco carcinogen 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK)**. *Carcinogenesis* (2008.0) **29** 866-874. PMID: 18281247 42. Souliotis V.L., van Delft J.H., Steenwinkel M.J., Baan R.A., Kyrtopoulos S.A.. **DNA adducts, mutant frequencies and mutation spectra in lambda lacZ transgenic mice treated with N-nitrosodimethylamine**. *Carcinogenesis* (1998.0) **19** 731-739. PMID: 9635857 43. Yoo J.S., Ishizaki H., Yang C.S.. **Roles of cytochrome P450IIE1 in the dealkylation and denitrosation of N-nitrosodimethylamine and N-nitrosodiethylamine in rat liver microsomes**. *Carcinogenesis* (1990.0) **11** 2239-2243. PMID: 2265475 44. Suzuki T., Itoh T., Hayashi M., Nishikawa Y., Ikezaki S., Furukawa F., Takahashi M., Sofuni T.. **Organ variation in the mutagenicity of dimethylnitrosamine in Big Blue mice**. *Environ. Mol. Mutagen.* (1996.0) **28** 348-353. PMID: 8991063 45. Nishikawa A., Furukawa F., Kasahara K., Lee I.S., Suzuki T., Hayashi M., Sofuni T., Takahashi M.. **Comparative study on organ-specificity of tumorigenicity, mutagenicity and cell proliferative activity induced by dimethylnitrosamine in Big Blue mice**. *Cancer Lett.* (1997.0) **117** 143-147. PMID: 9377541 46. Hahn J.N., Sandercock L.E., Giesbrecht J.L., McDonogh S., Luchman H.A., Jirik F.R.. **Investigating the mutagenic and carcinogenic effects of the tobacco-derived nitrosamine, NNK, in mice lacking specific DNA repair pathways**. *Cancer Res.* (2005.0) **65** 717 47. Esteller M., García-Foncillas J., Andion E., Goodman S., Hidalgo O., Vanaclocha V., Baylin S., Herman J.. **Inactivation of the DNA-repair gene MGMT and the clinical response of gliomas to alkylating agents**. *N. Engl. J. Med.* (2000.0) **343** 1350-1354. PMID: 11070098 48. Alexandrov L.B., Kim J., Haradhvala N.J., Huang M.N., Tian Ng A.W., Wu Y., Boot A., Covington K.R., Gordenin D.A., Bergstrom E.N.. **The repertoire of mutational signatures in human cancer**. *Nature* (2020.0) **578** 94-101. PMID: 32025018 49. Chawanthayatham S., Valentine C.C., Fedeles B.I., Fox E.J., Loeb L.A., Levine S.S., Slocum S.L., Wogan G.N., Croy R.G., Essigmann J.M.. **Mutational spectra of aflatoxin B1 in vivo establish biomarkers of exposure for human hepatocellular carcinoma**. *Proc. Natl. Acad. Sci. U.S.A.* (2017.0) **114** E3101-E3109. PMID: 28351974 50. Nohmi T., Katoh M., Suzuki H., Matsui M., Yamada M., Watanabe M., Suzuki M., Horiya N., Ueda O., Shibuya T.. **A new transgenic mouse mutagenesis test system using Spi- and 6-thioguanine selections**. *Environ. Mol. Mutagen.* (1996.0) **28** 465-470. PMID: 8991079 51. Glassner B.J., Weeda G., Allan J.M., Broekhof J.L., Carls N.H., Donker I., Engelward B.P., Hampson R.J., Hersmus R., Hickman M.J.. **DNA repair methyltransferase (Mgmt) knockout mice are sensitive to the lethal effects of chemotherapeutic alkylating agents**. *Mutagenesis* (1999.0) **14** 339-347. PMID: 10375003 52. He M., Croy R.G., Essigmann J.M., Swager T.M.. **Chemiresistive carbon nanotube sensors for N-nitrosodialkylamines**. *ACS Sens.* (2019.0) **4** 2819-2824. PMID: 31573183 53. Dass S.B., Hammons G.J., Bucci T.J., Heflich R.H., Casciano D.A.. **Susceptibility of C57BL/6 mice to tumorigenicity induced by dimethylnitrosamine and 2-amino-1-methyl-6-phenylimidazo [4,5-b]pyridine in the neonatal bioassay**. *Cancer Lett.* (1998.0) **124** 105-110. PMID: 9500198 54. Thoolen B., Maronpot R.R., Harada T., Nyska A., Rousseaux C., Nolte T., Malarkey D.E., Kaufmann W., Küttler K., Deschl U.. **Proliferative and nonproliferative lesions of the rat and mouse hepatobiliary system**. *Toxicol. Pathol.* (2010.0) **38** 5S-81S. PMID: 21191096 55. Schmitt M.W., Kennedy S.R., Salk J.J., Fox E.J., Hiatt J.B., Loeb L.A.. **Detection of ultra-rare mutations by next-generation sequencing**. *Proc. Natl. Acad. Sci. U.S.A.* (2012.0) **109** 14508-14513. PMID: 22853953 56. Ellison K.S., Dogliotti E., Connors T.D., Basu A.K., Essigmann J.M.. **Site-specific mutagenesis by O6-alkylguanines located in the chromosomes of mammalian cells: influence of the mammalian O6-alkylguanine-DNA alkyltransferase**. *Proc. Natl. Acad. Sci. U.S.A.* (1989.0) **86** 8620-8624. PMID: 2813414 57. Delaney J.C., Essigmann J.M.. **Effect of sequence context on O6-methylguanine repair and replication in vivo**. *Biochemistry* (2001.0) **40** 14968-14975. PMID: 11732917 58. O’Shea J.P., Chou M.F., Quader S.A., Ryan J.K., Church G.M., Schwartz D. **pLogo: a probabilistic approach to visualizing sequence motifs**. *Nat. Methods* (2013.0) **10** 1211-1212. PMID: 24097270 59. Ganem N.J., Pellman D.. **Linking abnormal mitosis to the acquisition of DNA damage**. *J. Cell Biol.* (2012.0) **199** 871-881. PMID: 23229895 60. Sukup-Jackson M.R., Kiraly O., Kay J.E., Na L., Rowland E.A., Winther K.E., Chow D.N., Kimoto T., Matsuguchi T., Jonnalagadda V.S.. **Rosa26-GFP direct repeat (RaDR-GFP) mice reveal tissue- and age-dependence of homologous recombination in mammals in vivo**. *PLoS Genet.* (2014.0) **10** e1004299. PMID: 24901438 61. Poon S.L., McPherson J.R., Tan P., The B.T., Rozen S.G.. **Mutation signatures of carcinogen exposure: genome-wide detection and new opportunities for cancer prevention**. *Genome Med.* (2014.0) **6** 24. PMID: 25031618 62. Shen J.C., Kamath-Loeb A.S., Kohrn B.F., Loeb K.R., Preston B.D., Loeb L.A.. **A high-resolution landscape of mutations in the BCL6 super-enhancer in normal human B cells**. *Proc. Natl. Acad. Sci. U.S.A.* (2019.0) **116** 24779-24785. PMID: 31748270 63. Fedeles B.I., Essigmann J.M.. **Chapter 25:mutational Spectra Provide Insight into the Mechanisms Bridging DNA Damage to Genetic Disease**. *DNA Damage, DNA Repair and Disease* (2020.0) 214-253 64. Kucab J.E., Zou X., Morganella S., Joel M., Nanda A.S., Nagy E., Gomez C., Degasperi A., Harris R., Jackson S.P.. **A compendium of mutational signatures of environmental agents**. *Cell* (2019.0) **177** 821-836. PMID: 30982602 65. Touat M., Li Y.Y., Boynton A.N., Spurr L.F., Iorgulescu J.B., Bohrson C.L., Cortes-Ciriano I., Birzu C., Geduldig J.E., Pelton K.. **Mechanisms and therapeutic implications of hypermutation in gliomas**. *Nature* (2020.0) **580** 517-523. PMID: 32322066 66. Nagel Z.D., Kitange G.J., Gupta S.K., Joughin B.A., Chaim I.A., Mazzucato P., Lauffenburger D.A., Sarkaria J.N., Samson L.D.. **DNA repair capacity in multiple pathways predicts chemoresistance in glioblastoma multiforme**. *Cancer Res.* (2017.0) **77** 198-206. PMID: 27793847 67. Shen Y., Grisdale C.J., Islam S.A., Bose P., Lever J., Zhao E.Y., Grinshtein N., Ma Y., Mungall A.J., Moore R.A.. **Comprehensive genomic profiling of glioblastoma tumors, BTICs, and xenografts reveals stability and adaptation to growth environments**. *Proc. Natl. Acad. Sci. U.S.A.* (2019.0) **116** 19098-19108. PMID: 31471491 68. Richardson F.C., Boucheron J.A., Skopek T.R., Swenberg J.A.. **Formation of O6-methyldeoxyguanosine at specific sites in a synthetic oligonucleotide designed to resemble a known mutagenic hotspot**. *J. Biol. Chem.* (1989.0) **264** 838-841. PMID: 2642906 69. Dolan M.E., Oplinger M., Pegg A.E.. **Sequence specificity of guanine alkylation and repair**. *Carcinogenesis* (1988.0) **9** 2139-2143. PMID: 3180351 70. Richardson K.K., Richardson F.C., Crosby R.M., Swenberg J.A., Skopek T.R.. **DNA base changes and alkylation following in vivo exposure of Escherichia coli to N-methyl-N-nitrosourea or N-ethyl-N-nitrosourea**. *Proc. Natl. Acad. Sci. U.S.A.* (1987.0) **84** 344-348. PMID: 3540961 71. Altshuler K.B., Hodes C.S., Essigmann J.M.. **Intrachromosomal probes for mutagenesis by alkylated DNA bases replicated in mammalian cells: a comparison of the mutagenicities of O4-methylthymine and O6-methylguanine in cells with different DNA repair backgrounds**. *Chem. Res. Toxicol.* (1996.0) **9** 980-987. PMID: 8870985 72. Andersen N., Wang J., Wang P., Jiang Y., Wang Y.. **In-vitro replication studies on O2-methylthymidine and O4-methylthymidine**. *Chem. Res. Toxicol.* (2012.0) **25** 2523-2531. PMID: 23113558 73. Anderson L.M., Hagiwara A., Kovatch R.M., Rehm S., Rice J.M.. **Transplacental initiation of liver, lung, neurogenic, and connective tissue tumors by N-nitroso compounds in mice**. *Fundam. Appl. Toxicol.* (1989.0) **12** 604-620. PMID: 2731672 74. Hassan M.M., Botrus G., Abdel-Wahab R., Wolff R.A., Li D., Tweardy D., Phan A.T., Hawk E., Javle M., Lee J.S.. **Estrogen replacement reduces risk and increases survival times of women with hepatocellular carcinoma**. *Clin. Gastroenterol. Hepatol.* (2017.0) **15** 1791-1799. PMID: 28579181 75. Vesselinovitch S.D.. **The sex-dependent difference in the development of liver tumors in mice administered dimethylnitrosamine**. *Cancer Res.* (1969.0) **29** 1024-1027. PMID: 4305498 76. Woo L.L., Egner P.A., Belanger C.L., Wattanawaraporn R., Trudel L.J., Croy R.G., Groopman J.D., Essigmann J.M., Wogan G.N.. **Aflatoxin B1-DNA adduct formation and mutagenicity in livers of neonatal male and female B6C3F1 mice**. *Toxicol. Sci.* (2011.0) **122** 38-44. PMID: 21507988 77. Rao K.V., Vesselinovitch S.D.. **Age- and sex-associated diethylnitrosamine dealkylation activity of the mouse liver and hepatocarcinogenesis**. *Cancer Res.* (1973.0) **33** 1625-1627. PMID: 4352793 78. Naugler W.E., Sakurai T., Kim S., Maeda S., Kim K., Elsharkawy A.M., Karin M.. **Gender disparity in liver cancer due to sex differences in MyD88-dependent IL-6 production**. *Science* (2007.0) **317** 121-124. PMID: 17615358 79. Galien R., Garcia T.. **Estrogen receptor impairs interleukin-6 expression by preventing protein binding on the NF-kappaB site**. *Nucleic Acids Res.* (1997.0) **25** 2424-2429. PMID: 9171095 80. Liu H., Liu K., Bodenner D.L.. **Estrogen receptor inhibits interleukin-6 gene expression by disruption of nuclear factor kappaB transactivation**. *Cytokine* (2005.0) **31** 251-257. PMID: 16043358 81. Shalapour S., Karin M.. **Immunity, inflammation, and cancer: an eternal fight between good and evil**. *J. Clin. Invest.* (2015.0) **125** 3347-3355. PMID: 26325032 82. Chan K., Roberts S.A., Klimczak L.J., Sterling J.F., Saini N., Malc E.P., Kim J., Kwiatkowski D.J., Fargo D.C., Mieczkowski P.A.. **An APOBEC3A hypermutation signature is distinguishable from the signature of background mutagenesis by APOBEC3B in human cancers**. *Nat. Genet.* (2015.0) **47** 1067-1072. PMID: 26258849 83. Gerstung M., Jolly C., Leshchiner I., Dentro S.C., Gonzalez S., Rosebrock D., Mitchell T.J., Rubanova Y., Anur P., Yu K.. **The evolutionary history of 2,658 cancers**. *Nature* (2020.0) **578** 122-128. PMID: 32025013 84. Akakpo J.Y., Ramachandran A., Orhan H., Curry S.C., Rumack B.H., Jaeschke H.. **2020) 4-methylpyrazole protects against acetaminophen-induced acute kidney injury**. *Toxicol. Appl. Pharmacol.* **409** 115317. PMID: 33157119 85. Chanas B., Wang H., Ghanayem B.I.. **Differential metabolism of acrylonitrile to cyanide is responsible for the greater sensitivity of male vs female mice: role of CYP2E1 and epoxide hydrolases**. *Toxicol. Appl. Pharmacol.* (2003.0) **193** 293-302. PMID: 14644629 86. Wang Z., Hall S.D., Maya J.F., Li L., Asghar A., Gorski J.C.. **Diabetes mellitus increases the in vivo activity of cytochrome P450 2E1 in humans**. *Br. J. Clin. Pharmacol.* (2003.0) **55** 77-85. PMID: 12534643 87. Dow M., Pyke R.M., Tsui B.Y., Alexandrov L.B., Nakagawa H., Taniguchi K., Seki E., Harismendy O., Shalapour S., Karin M.. **Integrative genomic analysis of mouse and human hepatocellular carcinoma**. *Proc. Natl. Acad. Sci. U.S.A.* (2018.0) **115** E9879-E9888. PMID: 30287485 88. Matsumura S., Sato H., Otsubo Y., Tasaki J., Ikeda N., Morita O.. **Genome-wide somatic mutation analysis via Hawk-Seq**. *Arch. Toxicol.* (2019.0) **93** 2689-2701. PMID: 31451845
--- title: MBOAT7-driven lysophosphatidylinositol acylation in adipocytes contributes to systemic glucose homeostasis authors: - William J. Massey - Venkateshwari Varadharajan - Rakhee Banerjee - Amanda L. Brown - Anthony J. Horak - Rachel C. Hohe - Bryan M. Jung - Yunguang Qiu - E. Ricky Chan - Calvin Pan - Renliang Zhang - Daniela S. Allende - Belinda Willard - Feixiong Cheng - Aldons J. Lusis - J. Mark Brown journal: Journal of Lipid Research year: 2023 pmcid: PMC10041558 doi: 10.1016/j.jlr.2023.100349 license: CC BY 4.0 --- # MBOAT7-driven lysophosphatidylinositol acylation in adipocytes contributes to systemic glucose homeostasis ## Body The rising prevalence of obesity closely parallels the rise of nonalcoholic fatty liver disease (NAFLD), which is a leading cause of liver disease mortality in developed countries [1, 2, 3, 4]. Because of this rapidly growing health crisis, there is an increasing need for mechanistic insight to allow a path for the rational design of new therapeutic strategies. One important clue in NAFLD research has recently emerged where multiple genome wide association studies identified the common rs641738 SNP located close to the lysophosphatidylinositol (LPI)-acylating enzyme, membrane bound O-acyltransferase 7 (MBOAT7), as a risk allele. The rs641738 (C>T) SNP is associated with increased susceptibility to NAFLD [5, 6, 7], liver diseases of alcoholic and viral etiologies [8, 9, 10], and other complex metabolic diseases [11, 12]. Using an antisense oligonucleotide (ASO)-mediated approach, we recently showed that Mboat7 knockdown exacerbated high fat diet (HFD)-induced hepatic steatosis and inflammation, hyperinsulinemia, and insulin resistance [13]. It is important to note that we also showed that genetic deletion of the neighboring gene transmembrane channel like 4 did not promote NAFLD [13]. Very similar findings were reported by Meroni et al. [ 14] using a morpholino oligonucleotide (MPO)-driven knockdown approach, further demonstrating that Mboat7 loss-of-function promoted fatty liver and importantly showed that Mboat7 expression is suppressed by insulin. More recently, several independent laboratories have shown that hepatocyte-specific deletion of Mboat7 (Mboat7HSKO) worsens hepatic steatosis, inflammation, and fibrosis, but unlike the oligonucleotide-based knockdown, does not impact insulin or glucose homeostasis [15, 16, 17]. Given the disparate results between the results of published ASO and MPO knockdown experiments [13, 14] and those of these recent studies of Mboat7HSKO [15, 16, 17], we hypothesized that the function of MBOAT7 in adipocytes may be critically important in regulating circulating insulin levels and action in target tissues. It is well appreciated that ASOs can effectively suppress target gene expression in adipocytes [13, 18] and adipose tissue plays a critically important role in the progression of NAFLD and systemic glucose homeostasis. To better understand the tissue-specific roles of MBOAT7, we generated Mboat7 adipocyte-specific KO (Mboat7ASKO) and Mboat7 hepatocyte-specific KO (Mboat7HSKO) mice to understand the cell-autonomous contributions of Mboat7 to HFD-induced metabolic disturbances. Here, we find that adipocyte Mboat7 contributes less so to hepatic steatosis and injury than hepatocyte Mboat7 but instead plays a critically important role in regulating local adipose tissue LPI/PI homeostasis, hyperinsulinemia, and systemic insulin sensitivity. ## Abstract We previously demonstrated that antisense oligonucleotide-mediated knockdown of Mboat7, the gene encoding membrane bound O-acyltransferase 7, in the liver and adipose tissue of mice promoted high fat diet-induced hepatic steatosis, hyperinsulinemia, and systemic insulin resistance. Thereafter, other groups showed that hepatocyte-specific genetic deletion of Mboat7 promoted striking fatty liver and NAFLD progression in mice but does not alter insulin sensitivity, suggesting the potential for cell autonomous roles. Here, we show that MBOAT7 function in adipocytes contributes to diet-induced metabolic disturbances including hyperinsulinemia and systemic insulin resistance. *We* generated Mboat7 floxed mice and created hepatocyte- and adipocyte-specific Mboat7 knockout mice using Cre-recombinase mice under the control of the albumin and adiponectin promoter, respectively. Here, we show that MBOAT7 function in adipocytes contributes to diet-induced metabolic disturbances including hyperinsulinemia and systemic insulin resistance. The expression of Mboat7 in white adipose tissue closely correlates with diet-induced obesity across a panel of ∼100 inbred strains of mice fed a high fat/high sucrose diet. Moreover, we found that adipocyte-specific genetic deletion of Mboat7 is sufficient to promote hyperinsulinemia, systemic insulin resistance, and mild fatty liver. Unlike in the liver, where Mboat7 plays a relatively minor role in maintaining arachidonic acid-containing PI pools, Mboat7 is the major source of arachidonic acid-containing PI pools in adipose tissue. Our data demonstrate that MBOAT7 is a critical regulator of adipose tissue PI homeostasis, and adipocyte MBOAT7-driven PI biosynthesis is closely linked to hyperinsulinemia and insulin resistance in mice. ## Mice and experimental diets *To* generate conditional Mboat7 KO mice, we obtained “KO first” (Mboat7tm1a(KOMP)Wtsi) mice from Dr Philip Hawkins [19] and crossed these mice with mice transgenically expressing FLP recombinase to remove the NEO cassette resulting in a conditional Mboat7 floxed allele. The FLP transgene was then bred out of the line and resulting Mboat7flox/WT mice were subsequently bred with transgenic mice expressing Cre recombinase under either the adiponectin promoter/enhancer (for adipocyte-specific deletion) or under the albumin promoter (for hepatocyte-specific deletion). Importantly, each substrain of mice was backcrossed >10 generations into the C57BL/6J background to produce congenic lines, and confirmation of sufficient backcrossing into the C57BL/6J background was confirmed by mouse genome SNP scanning at the Jackson Laboratory (Bar Harbor, ME). For all experiments, age-matched animals were put on either standard rodent chow (Teklad 2918) or $60\%$ kcal fat HFD (Research Diets D12492) for up to 20 weeks. Glucose and insulin tolerance tests (ITTs) were performed as previously described [20, 21, 22, 23, 24, 25] in mice following 12 and 14 weeks of concurrent chow and HFD-feeding, respectively. Quantitation of lean and fat mass were done using an EchoMRITM-130 Body Composition Analyzer (EchoMRI International). All mice were maintained in an Association for the Assessment and Accreditation of Laboratory Animal Care, International-approved animal facility, and all experimental protocols were approved by the IACUC of the Cleveland Clinic (IACUC protocols # 2018-2053 and # 00002499). ## Standardized necropsy conditions To keep results consistent, the experimental mice were fasted for 4 h (from 09:00 to 13:00) prior to necropsy for 20 weeks diet study and 16 weeks HFD study. For 12 weeks HFD study with fasting-refeeding, mice were fasted overnight (16:00–08:00), then bled via submandibular bleed and refed HFD for 3 h (until 11:00). At necropsy, all mice were terminally anesthetized with ketamine/xylazine (100–160 mg/kg ketamine-20–32 mg/kg xylazine) and a midline laparotomy was performed. Blood was collected by heart puncture. Following blood collection, a whole-body perfusion was conducted by puncturing the right atria and slowly delivering 10 ml of saline into the left ventricle of the heart to remove blood from tissues. Tissues were collected and immediately snap frozen in liquid nitrogen for subsequent biochemical analysis or fixed for morphological analysis. Various adipose depots were defined by their anatomical location. Gonadal white adipose tissue (gWAT) was dissected from the area surrounding the testes (males) or fallopian tubes (females), subcutaneous white adipose tissue was dissected from the subcutaneous surface of the overlying inguinal skin, retroperitoneal adipose tissue was collected from the retroperitoneum (beneath the kidneys), mesenteric adipose tissue was collected by dissecting away the pancreas and then careful dissection of fat from the remaining intestinal tissue, and brown adipose tissue (BAT) was carefully dissected from the subscapular space to remove overlying white adipose tissue (WAT). ## Hyperinsulinemic-euglycemic clamp All procedures required for the hyperinsulinemic–euglycemic clamp were approved by the Vanderbilt University IACUC and performed at the Vanderbilt Mouse Metabolic Phenotyping Center. For these studies, age-matched Mboat7flox/flox and Mboat7ASKO mice were maintained on HFD for 11–12 weeks. Thereafter, while being maintained on HFD, catheters were implanted into a carotid artery and a jugular vein of mice for sampling and infusions respectively, 5 days before the study as described previously [26]. Insulin clamps were performed on mice fasted for 5 h using a modification of the method described by Ayala et al. [ 27]. [ 3-3H]-glucose was primed (1.5 μCi) and continuously infused for a 90 min equilibration and basal sampling periods (0.075 μCi/min). [ 3-3H]-glucose was mixed with the nonradioactive glucose infusate (infusate specific-activity of 0.5 μCi/mg) during the 2 h clamp period. Arterial glucose was clamped using a variable rate of glucose (plus trace [3-3H]-glucose) infusion, which was adjusted based on the measurement of blood glucose at 10 min intervals. By mixing radioactive glucose with the nonradioactive glucose infused during a clamp, deviations in arterial glucose-specific activity are minimized and steady state conditions are achieved. The calculation of glucose kinetics is therefore more robust [28]. Baseline blood or plasma variables were calculated as the mean of values obtained in blood samples collected at −15 and −5 min. At time zero, insulin infusion (4 mU/kg body weight/min) was started and continued for 120 min. Mice received heparinized saline-washed erythrocytes from donors at 5 μl/min to prevent a fall in hematocrit. Blood was taken from 80 to 120 min for the determination of [3-3H]-glucose. Clamp insulin was determined at $t = 100$ and 120 min. At 120 min, 13 μCi of 2[14C]deoxyglucose ([14C]2DG) was administered as an intravenous bolus. Blood was taken from 2 to 25 min for determination of [14C]2DG. After the last sample, mice were anesthetized and tissues were freeze-clamped for further analysis. Radioactivity of [3-3H]-glucose and [14C]2DG in plasma samples and [14C]2DG-6-phosphate in tissue samples were determined by liquid scintillation counting. Glucose appearance (Ra) and disappearance (Rd) rates were determined using steady-state equations [29]. Endogenous glucose appearance (endoRa) was determined by subtracting the glucose infusion rate from total Ra. The glucose metabolic index (Rg) was calculated as previously described [30]. ## Histological analysis and imaging H&E staining of paraffin-embedded liver sections was performed as previously described [21, 22, 23, 24, 25]. Histopathologic evaluation was scored in a blinded fashion by a board-certified pathologist with expertise in gastrointestinal/liver pathology (Daniela S. Allende–Cleveland Clinic). H&E slides were scanned using a Leica Aperio AT2 Slide Scanner (Leica Microsystems, GmbH, Wetzlar, Germany; https://www.leicabiosystems.com/us/digital-pathology/manage/aperio-imagescope/) and images were processed using ImageScope (Aperio, Software Version 12.1). ## Quantification of adipose tissue immune cell populations by flow cytometry After 16 weeks of HFD-feeding, gWAT was excised, washed with 1× PBS, and immediately placed into RPMI with 1 mg/ml Type II Collagenase (Sigma Aldrich, St. Louis, MO) for 30 min at 37°C with gentle agitation. Digested clumps of tissue were pressed through a 100 μm strainer and washed with 1× PBS. Cells were centrifuged at 500 g for 10 min; supernatant was aspirated. Cell pellet containing the stromal vascular fraction was resuspended in ammonium-chloride-potassium Lysing Buffer (Life Technologies, Grandstand, NY) for 5 min at room temperature. Cells were washed with 1× PBS and centrifuged at 300 g for 10 min. For flow cytometry, cells were resuspended in freshly prepared fluorescence-activated cell sorting (FACS) buffer (1× PBS, $3\%$ FBS) and aliquoted into 96-well plates. Cells were centrifuged at 830 g for 4 min, resuspended in 50 μl FACS buffer containing 0.5 μg anti-mouse CD16/CD32 mAb 2.4G2 (Mouse BD Fc Block, BD Pharmingen, San Diego, CA), and incubated for 15 min 4°C. After blocking, cells were stained with a fluorochrome-conjugated antibody panel (antibodies described in the key resource table) for 30 min at 4°C in the dark. Cells were washed and centrifuged at 830 g for 4 min twice with FACS buffer. Stained cells are resuspended in 200 μl of $1\%$ paraformaldehyde and kept in the dark at 4°C overnight. Stained cells were centrifuged at 830 g for 5 min. Stained cells were resuspended in 300 μl of FACS buffer and data was collected on a Cytek Aurora full spectrum (365–829 nm range) cytometer using SpectroFlo® software (Cytek® Biosciences, Fremont, CA; https://cytekbio.com/pages/spectro-flo). Data collected on the Aurora were analyzed using FlowJo software (Tree Star, Inc., Ashland, OR; https://www.flowjo.com/). Gating strategies are shown in supplemental Fig. S10. ## Immunoblotting Whole tissue homogenates were made from tissues in a modified RIPA buffer as previously described [20, 21, 22, 23, 24, 25] that was supplemented with an additional $2\%$ (w/v) sodium dodecylsulfate, and protein was quantified using the BCA assay (Pierce). 10 μg of protein was separated by 4–$12\%$ SDS-PAGE, transferred to a PVDF membrane, and proteins were detected after incubation with specific antibodies as previously described [20, 21, 22, 23, 24, 25]. ## RNA isolation and bulk RNA sequencing gWAT RNA was isolated using Trizol-Chloroform extraction, gDNA removal by eliminator spin column (Qiagen), and isopropanol precipitation followed by ethanol clean-up. RNA quality was confirmed by Bioanalyzer (Agilent). RNA-SEQ libraries were generated using the Illumina mRNA TruSEQ Directional library kit and sequenced using an Illumina HiSEQ4000 (both according to the Manufacturer’s instructions). RNA sequencing was performed by the University of Chicago Genomics Facility. Raw sequencing data in the form of FASTQ files were transferred to and analyzed by the Bioinformatics Core at Case Western Reserve University. FASTQ files were trimmed for quality and adapter sequences using TrimGalore! ( version 0.6.5 Babraham Institute, https://github.com/FelixKrueger/TrimGalore), a wrapper script for CutAdapt and FastQC. Reads passing quality control were aligned to the mm10 mouse reference genome using STARAligner [31] (version 2.5.3a) and guided using the GENCODE gene annotation. Aligned reads were analyzed for differential gene expression using Cufflinks [32] (version 2.2.1) which reports the fragments per kilobase of exon per million fragments mapped for each gene. *Significant* genes were identified using a significance cutoff of q-value <0.05 (FDR) and used as input for downstream analysis. RNA sequencing data can be accessed at GEO Profiles at the NCBI: GSE203414 and full accession will be available upon acceptance of this work. In addition, the RNA expression data can be found in supplemental Table S2 associated with the online version of this article. ## Plasma and liver biochemistries To determine the level of hepatic injury in mice fed HFD, plasma was used to analyze alanine aminotransferase levels using enzymatic assays as previously described [13]. Extraction of liver lipids and quantification of total plasma and hepatic triglycerides, total cholesterol, and free cholesterol was conducted using enzymatic assays as described previously [21, 22, 23, 24, 25]. Esterified cholesterol was calculated by subtracting free cholesterol from total cholesterol. ## Targeted quantification of phosphatidylinositol (PI) and lysophosphatidylinositol lipids Quantitation of LPI and PI species was performed as previously described [13]. Briefly, LPI and PI standards (LPI-16:0, LPI-18:0, LPI-18:1, LPI-20:4, PI-38:4) and the two internal standards (LPI-17:1-d31, PI-34:1-d31) were purchased from Avanti Polar Lipids. HPLC grade water, methanol, and acetonitrile were purchased from Thermo Fisher Scientific. Standard LPI and PI species at concentrations of 0, 5, 20, 100, 500, and 2,000 ng/ml were prepared in $90\%$ methanol containing 2 internal standards at the concentration of 500 ng/ml. The volume of 5 μl was injected into the Shimadzu LCMS-8050 for generating the internal standard calibration curves. A triple quadrupole mass spectrometer (Thermo Fisher Scientific Quantiva, Waltham, MA) was used for analysis of LPI and PI species. The mass spectrometer was coupled to the outlet of an ultra high performance liquid chromatography system (Vanquish, Thermo Fisher Scientific, Waltham, MA), including an auto sampler with refrigerated sample compartment and inline vacuum degasser. The HPLC eluent was directly injected into the triple quadrupole mass spectrometer and the analytes were ionized at ESI negative mode. Analytes were quantified using Selected Reaction Monitoring and the Selected Reaction Monitoring transitions (m/z) were 571 → 255 for LPI-16:0, 599 → 283 for LPI-18:0, 597 → 281 for LPI-18:1, 619 → 303 for LPI-20:4, 885 → 241 for PI-38:4, 583 → 267 for internal standard LPI-17:1, and 866 → 281 for internal standard PI-34:1-d31. Xcalibur software (https://www.thermofisher.com/order/catalog/product/OPTON-30965) was used to get the peak area for both the internal standards and LPI and PI species. The internal standard calibration curves were used to calculate the concentration of LPI and PI species in the samples. ## Adipose tissue phosphoproteomic and pathway analyses to examine LPI-induced signaling events in WAT The goal of this experiment was to unbiasedly identify LPI-responsive signaling events in mouse WAT after an acute exposure (15 min) to a physiological level of LPI. Optimization of in vivo LPI dosing was previously described by Helsley et al. 2019 [13]. Briefly, mice were fasted for 4 h between the hours of 09:00–13:00, then we delivered 50 μl of either sterile saline or 400 μM (20 nmol, 12.3 μg) 18:1 LPI intraperitoneally. After 10 min, mice were anesthetized and tissue was dissected as described above. Adipose tissue samples were homogenized, the protein was precipitated with acetone, and the protein concentration was measured. A total of 350 μg of protein from each sample was digested with trypsin and the resulting tryptic peptides were subjected to phosphoserine and phosphothreonine enrichment using the Thermo Fisher Scientific Phosphopeptide Enrichment (Thermo Fisher Scientific High Select™ A32992). The enrichment was performed based on the manufacturer’s instructions. The enriched peptide samples were dried then reconstituted in 25 μl $0.1\%$ formic acid for LC-MS analysis. The LC-MS system was a Thermo Fisher Scientific Fusion Lumos mass spectrometer system. The HPLC column was a Dionex 15 cm × 75 μm id Acclaim Pepmap C18, 2 μm, 100 Å reversed-phase capillary chromatography column. Five μl volumes of the extract were injected and the peptides eluted from the column by an acetonitrile/$0.1\%$ formic acid gradient at a flow rate of 0.25 μl/min were introduced into the source of the mass spectrometer on-line. The microelectrospray ion source was operated at 1.9 kV. The digest was analyzed using the data-dependent multitask capability of the instrument acquiring full scan mass spectra to determine peptide molecular weights and product ion spectra to determine amino acid sequence in successive instrument scans. The LC-MS/MS data files were searched against the mouse UniProtKB databases using the program Proteome Discoverer 2.5 (https://www.thermofisher.com/us/en/home/industrial/mass-spectrometry/liquid-chromatography-mass-spectrometry-lc-ms/lc-ms-software/multi-omics-data-analysis/proteome-discoverer-software.html?gclid=EAIaIQobChMI9Kei3-rR_QIVYGxvBB3mnwJbEAAYASAAEgI-KvD_BwE&cid=E.23CMD.DL103.12911.01&ef_id=EAIaIQobChMI9Kei3-rR_QIVYGxvBB3mnwJbEAAYASAAEgI-KvD_BwE:G:s&s_kwcid=AL!3652!3!646724600131!p!!g!!proteome%20discoverer) considering Oxidation of Met, N-terminal Acetyl, S, T, and Y phosphorylation as a differential modification. A maximum of three missed cleavages were permitted. The peptide and protein false discovery rates were set to 0.01 using a target-decoy strategy. Phosphorylation sites were identified using ptmRS node in PD2.4. The relative abundance of the positively identified phosphopeptides was determined using the extracted ion intensities (Minora Feature Detection node) with Retention time alignment. All peptides were included in the quantitation, the peptide intensities were normalized to total peptide amount. Missing values were imputed in Perseus using a normal distribution. Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed to evaluate the biological relevance and functional pathways on the significant expressed genes of the LPI-related phosphoproteomics. The Entrez ID and official gene symbols were mapped based on GeneCards (http://www.genecards.org/). A total of 6,141 phosphopeptides were identified with 165 and 40 phosphopeptides determined to be |log2FoldChange| >0.5 different in the LPI and saline samples with a P-value <0.05 (two-tailed t test) in Mboat7ASKO and Mboat7flox/flox mice, respectively. The functional pathways were ranked by -log10FDR. Significant pathways were identified with FDR <0.05. The coverage of significant expressed genes among the total genes provided by KEGG were also calculated. All pathway analyses were performed by using Enrichr [33]. Phosphoproteomics data have been deposited to the ProteomeXchange *Consortium via* the PRIDE [34] partner repository with the dataset identifier PXD039894 and can be found in supplemental Table S3 associated with the online version of this article. ## Statistical analyses, key reagents, and data availability For the data shown in Fig. 1 [hybrid mouse diversity panel (HMDP)] correlations and associated P-values were calculated with the bi-weight midcorrelation, which is robust to outliers and associated P-value. Single comparisons between two groups were performed using two-tailed Student’s t tests with $95\%$ confidence intervals. Comparisons involving multiple time points were assessed using an ordinary one-way ANOVA followed by Tukey’s multiple comparisons test, or using a two-way ANOVA followed by Tukey’s multiple comparisons test. All data presented as mean ± SEM. Values were considered significant at $P \leq 0.05.$ ∗$P \leq 0.05$, ∗∗$P \leq 0.01$, ∗∗∗$P \leq 0.001$ and ∗∗∗∗$P \leq 0.0001.$ For all other Fig. panels each experiment consisted of a minimum of two biological replicates and data are presented as mean ± SD except were noted in the Fig. legend. All data were analyzed using either one-, two-, or three-way ANOVA where appropriate, followed by either Tukey’s, Bonferroni’s, or Student’s t-tests for post hoc analysis. Differences were considered significant at $P \leq 0.05.$ GraphPad Prism 9.4.0 (La Jolla, CA; https://www.graphpad.com/features) was used for data analysis. A list of key reagents can be found in supplemental Table S1. All materials, methods, and datasets included in this article are readily available upon request. Fig. 1Mboat7 Expression in White Adipose *Tissue is* Correlated with Adiposity in mice. We used a systems genetics approach to examine links between Mboat7 expression and metabolic traits in mice from the hybrid mouse diversity panel (HMDP). To induce obesity, all mouse strains represented in the HMDP were fed an obesity-promoting high fat and high sucrose diet. Across the different strains in the HMDP, the expression (RMA, Robust Multi-Array Average) of Mboat7 in adipose tissue has a strong negatively correlated with gonadal (A), subcutaneous (B), and retroperitoneal (C) white adipose tissue mass in females. There is a similar negative correlation in male (D) retroperitoneal adipose tissue. MBOAT7, membrane bound O-acyltransferase 7. ## Adipocyte-specific genetic deletion of Mboat7 promotes mild fatty liver but profound hyperinsulinemia and insulin resistance To understand the cell autonomous roles of Mboat7 in HFD-driven metabolic disturbance we generated parallel colonies of either Mboat7HSKO or Mboat7ASKO mice and subjected them to HFD feeding. Our initial rationale to study the role of Mboat7 outside of the liver came from studies where we used a systems genetics approach to examine tissue-specific links between Mboat7 expression and adiposity in mice represented in the HMDP when challenged with an obesity-promoting HFD and high sucrose diet [35, 36]. In our previous work, we found that Mboat7 expression in the liver was only modestly correlated (r = −0.244, $$P \leq 0.01$$) with adiposity [13]. However, here we show that Mboat7 expression in WAT is strongly negatively correlated with gWAT, scWAT, and retroperitoneal WAT in female mice and retroperitoneal WAT in male mice (Fig. 1A–D). It is interesting to note that WAT Mboat7 mRNA expression is not significantly correlated with adiposity measures in mice maintained on a standard rodent chow (data not shown), suggesting a role for WAT Mboat7 in selectively shaping diet-driven metabolic disturbance. *To* generate congenic Mboat7ASKO mice, we crossed mice harboring a post-FLP recombinase conditionally targeted Mboat7 floxed allele [19] to transgenic mice expressing Cre recombinase under the adiponectin promoter/enhancer [37], and then backcrossed these mice >10 generations into the C57BL/6J background. This is important since the genetic background can significantly impact metabolic traits. Compared to age-matched, chow fed control mice (Mboat7flox/flox), Mboat7ASKO mice had significantly reduced MBOAT7 protein expression in gWAT, scWAT, and subscapular BAT and a modest increase in skeletal muscle (Fig. 2A). It is important to note that liver MBOAT7 protein levels were not different in Mboat7ASKO mice (Fig. 2A). However, the molecular weight of the major MBOAT7 protein isoform in the liver is slightly smaller than the major isoform in WAT and BAT (Fig. 2A). Adipocyte-specific deletion of Mboat7 promotes marked reorganization of MBOAT7 substrate (LPI) and product (PI) lipids in gWAT (Fig. 2B–D). Particularly when challenged with a HFD, Mboat7ASKO mice have significantly elevated levels of 16:0 LPI, 18:0 LPI, and 18:1 LPI lipids in gWAT (Fig. 2B). Reciprocally, the predominant enzymatic product of MBOAT7 (38:4 PI) is reduced by >$60\%$ in Mboat7ASKO mice under both chow and HFD feeding conditions (Fig. 2C). Although less abundant, 36:4 PI and 38:5 PI lipids are also reduced in the gWAT of Mboat7ASKO mice and there is a reciprocal increase in other more saturated species of PI (34:2 PI, 36:2 PI, and 36:3 PI) (Fig. 2D). It is important to note that the levels of LPI and PI lipids in the plasma and liver were unaffected in Mboat7ASKO mice (supplemental Fig. S1), which is in contrast to what has been recently reported in Mboat7HSKO mice where both liver and plasma LPI and PI homeostasis is altered [15, 16, 17]. Collectively, these results demonstrate that MBOAT7 is a major regulator of the local LPI and PI lipidome in gWAT but unlike Mboat7HSKO mice, adipocyte-specific deletion of Mboat7 does not significantly alter circulating or liver LPI/PI homeostasis. Fig. 2Adipocyte-Specific Mboat7 Deletion (Mboat7ASKO) Promotes Mild Fatty Liver. Male control (Mboat7fl/fl) or adipocyte-specific Mboat7 knockout mice (Mboat7ASKO) were fed chow or high fat diet (HFD) for 20 weeks and metabolically phenotyped. A: Western blots from tissues collected from age-matched, chow fed Mboat7fl/fl or Mboat7ASKO mice. B–M: Gonadal white adipose tissue (gWAT), lysophosphatidylinositol (LPI), (B) and phosphatidylinositol (PI) species, including the MBOAT7 product PI-38:4 (C) and others (D) were quantified via LC-MSin Mboat7fl/fl or Mboat7ASKO mice were fed chow or high fat diet (HFD) for 20-week ($$n = 5$$–7; ∗∗∗∗P ≤ 0.0001; Two-way (C) or Three-way (B, D) ANOVA with Tukey’s post hoc test). E: Representative liver H&E stained sections. 10× magnification (scale bar = 200 μm). F: Liver weight measurements from Mboat7fl/fl or Mboat7ASKO mice fed Chow and HFD for 20 weeks ($$n = 6$$–8; Two-way ANOVA with Tukey’s post hoc test). G: *Percent steatosis* was quantified by a blinded pathologist ($$n = 5$$–7; Two-way ANOVA with Tukey’s post hoc test). Hepatic triglycerides (H) and hepatic esterified cholesterol (I) were measured enzymatically ($$n = 5$$–7; Two-way ANOVA with Tukey’s post hoc test). J: Representative gWAT H&E stained sections. 10× magnification (scale bar = 200 μm). K: Body weight was measured weekly. Percent fat (L) and % Lean (M) mass were determined via echo-MRI after 8 weeks of chow or HFD in Mboat7fl/fl or Mboat7ASKO mice ($$n = 5$$–7; Two-way ANOVA with Tukey’s post hoc test). All data are presented as mean ± S.D. Next, we compared and contrasted HFD-driven metabolic phenotypes in Mboat7ASKO and Mboat7HSKO mice. In agreement with what has been previously reported by three independent groups [15, 16, 17], we also find that Mboat7HSKO mice have significant reductions in 38:4 PI and elevation in its substrate LPI in the liver but not in gWAT (supplemental Fig. S2G–L). Mboat7HSKO mice have profound hepatic steatosis and elevated alanine aminotransferase (supplemental Fig. S2C–F), confirming the concept that MBOAT7 activity in hepatocytes opposes hepatic steatosis and liver injury [15, 16, 17]. Interestingly, adipocyte-specific deletion of Mboat7 also promotes mild hepatomegaly and hepatic steatosis after 20 weeks of HFD feeding but not shorter durations (Fig. 2E–G and supplemental Fig. S3A). However, even after 20 weeks of HFD, the significantly increased hepatic triglyceride and cholesterol ester levels apparent in Mboat7HSKO (supplemental Fig. S2E, [15, 16, 17]) are not seen in Mboat7ASKO mice (Fig. 2H, I). It is important to note that upon HFD feeding total body weight and WAT histology are similar in Mboat7HSKO when compared to their controls (supplemental Fig. S2B; and data not shown) as well as Mboat7ASKO compared to their appropriate controls (Fig. 2J, K). We also found that, after 8 weeks of HFD feeding, there were no differences between Mboat7ASKO and control mice in lean or fat mass as measured by EchoMRI (Fig. 2L, M). Collectively, these data suggest that MBOAT7 has clear cell autonomous roles in regulating circulating and tissue LPI and PI pools and MBOAT7-driven LPI acylation in hepatocytes is the primary driver of hepatic steatosis seen with MBOAT7 loss-of-function. ## Adipocyte-specific, but not hepatocyte-specific, deletion of Mboat7 promotes hyperinsulinemia and insulin resistance Obesity and T2D mellitus are commonly seen in subjects with NAFLD and it is well appreciated that excessive accumulation of lipids (i.e., “lipotoxicity) in the liver, skeletal muscle, and pancreatic beta cells can drive the overproduction of insulin and insulin resistance in target tissues [38, 39, 40]. Recent studies have demonstrated that the hepatic expression of MBOAT7 is suppressed in obese humans and rodents [13, 14] and insulin treatment can acutely suppress MBOAT7 mRNA and protein expression [14]. Furthermore, we demonstrated that ASO-mediated knockdown of Mboat7 in HFD fed mice elicits profound hyperinsulinemia and systemic insulin resistance [13]. However, several recent studies have demonstrated that genetic deletion of Mboat7 specifically in hepatocytes (i.e., Mboat7HSKO mice) does not alter insulin action or glucose tolerance [15, 17]. To further understand the cell autonomous roles of Mboat7 in insulin production and insulin action, we performed a series of studies comparing Mboat7ASKO and Mboat7HSKO mice (Fig. 3 and supplemental Fig. S2M–P). Despite having a severe fatty liver (supplemental Fig. S2I–K), Mboat7HSKO mice have normal glucose tolerance versus Mboat7flox/flox controls when challenged with a HFD (supplemental Fig. S2M). HFD fed Mboat7HSKO mice also have similar levels of fasting blood glucose and leptin but modestly increased fasting insulin when compared to Mboat7flox/flox controls (supplemental Fig. S2N–P). In contrast, when challenged with a HFD, Mboat7ASKO mice exhibit impaired systemic glucose tolerance and increased fasting blood glucose and insulin levels (Fig. 3A–L). It is important to note that most metabolic studies were performed in male Mboat7ASKO mice but many of the lipid changes and glucose intolerance phenotypes were also apparent in female Mboat7ASKO mice (supplemental Fig. S4). When we subjected Mboat7ASKO mice to an ITT to assess in vivo insulin sensitivity, there were some apparent differences between male and female mice (supplemental Fig. S5). When maintained on a chow diet, both male and female Mboat7ASKO mice had similar ITT responses to those seen in Mboat7flox/flox control mice (supplemental Fig. S5A–D). However, when challenged with a HFD, male Mboat7ASKO mice had an unexpected increase in blood glucose levels, whereas HFD fed Mboat7flox/flox control mice maintained glucose levels after an insulin challenge indicating clear HFD-induced insulin resistance (supplemental Fig. S5E, F). Whereas HFD fed female Mboat7flox/flox control mice showed appreciable insulin-induced lowering of blood glucose, HFD fed Mboat7ASKO mice showed significantly blunted insulin action (supplemental Fig. S5G, H). To further understand the underlying mechanism of insulin resistance, we performed euglycemic-hyperinsulinemic clamp studies in HFD fed control (Mboat7flox/flox) and Mboat7ASKO mice (Fig. 3E–L and supplemental Fig. S5I–K). HFD fed Mboat7ASKO mice exhibited reduced insulin sensitivity compared to HFD fed Mboat7flox/flox mice as reflected by lower glucose infusion rate during the clamp (Fig. 3E–G). Importantly, the insulin resistance seen in HFD fed Mboat7ASKO mice appears to be peripheral in nature given the blunted insulin-induced glucose disappearance (Fig. 3I, J) with no change to insulin-induced suppression of hepatic glucose production (supplemental Fig. S5I, J). Specifically, WAT insulin sensitivity is altered given gWAT specific uptake of 2-deoxyglucose (Rg) is reduced significantly and a similar trend is observed in scWAT (Fig. 3K, L). In contrast, there were no differences observed in glucose uptake in skeletal muscle, BAT, heart, or brain tissues (supplemental Fig. S5K). It is also interesting to note that gWAT tissue levels of 18:0- and 18:1-containing LPIs were significantly correlated with fasting blood glucose and glucose tolerance test area under the curve (supplemental Fig. S6). To look more broadly at whole body metabolism, we also subjected Mboat7ASKO and Mboat7ASKO mice fed HFD for 10 weeks to indirect calorimetry and found no differences in energy expenditure at thermoneutrality (30°C) or room temperature (23°C) (data not shown). Collectively, these data show for the first time that MBOAT7-driven LPI acylation in adipocytes, but not in hepatocytes, is critically important for the maintenance of circulating insulin and glucose levels as well as systemic insulin action. Fig. 3Adipocyte-Specific Mboat7 Deletion (Mboat7ASKO) Promotes Glucose Intolerance, Hyperinsulinemia, and Peripheral Insulin Resistance. A–C: Male control (Mboat7fl/fl) or adipocyte-specific Mboat7 knockout mice (Mboat7ASKO) were fed a chow or HFD for 12 weeks and then underwent an intraperitoneal glucose tolerance test (GTT). A: Plasma glucose levels were measured (in duplicate or triplicate at each time point) throughout the GTT ($$n = 5$$–7; Three-way ANOVA with Tukey’s post hoc test). B: *The area* under the curve was calculated for each mouse throughout the GTT ($$n = 5$$–7; Two-way ANOVA with Tukey’s post hoc test). C: Fasting blood glucose was measured after a 4-h fast (time = 0 min for GTT) ($$n = 5$$–7; Two-way ANOVA with Tukey’s post hoc test). D: *Fasting plasma* insulin was measured in Mboat7fl/fl or Mboat7ASKO mice that were fed a chow or HFD for 20 weeks ($$n = 5$$–7; Two-way ANOVA with Tukey’s post hoc test). E–L: Male control (Mboat7fl/fl) or adipocyte-specific Mboat7 knockout mice (Mboat7ASKO) were fed HFD for 12–13 weeks, underwent surgery for catheterization of carotid artery and a jugular vein, and were subjected to euglycemic-hyperinsulinemic clamping. E: Plasma glucose levels were measured throughout the clamp ($$n = 6$$–9; Two-way ANOVA with Bonferroni’s post hoc test). F: Glucose infusion rates (GIRs) were measured throughout the clamp ($$n = 6$$–9; Two-way ANOVA with Bonferroni’s post hoc test). G: Steady state glucose infusion rate was calculated by averaging GIRs from 80 to 120 min ($$n = 6$$–9; Student’s t test). H: Plasma glucose levels were determined by averaging the −10 and 0 min time points ($$n = 6$$–9; Student’s t test). I: The rate of glucose disappearance (Rd) was calculated for animals in the fasting state by averaging the Rd from −10 and 0 min time points and the clamped state by averaging Rd from 80 to 120 min ($$n = 6$$–9; Two-way ANOVA with Bonferroni’s post hoc test). J: Fold glucose disappearance was calculated by dividing clamp Rd by fasting Rd ($$n = 6$$–9; Student’s t test). Tissue specific uptake was measured in gonadal (K) and subcutaneous (L) white adipose tissue ($$n = 6$$–9; Student’s t test). ## MBOAT7 is a critical metabolic regulator of adipose tissue lipid homeostasis and adipose tissue function The proper storage of excess energy in the form of triacylglycerol (TAG) within adipose tissue is critically important for overall metabolic health. Although the major lipid class stored in WAT is TAG, adipocytes play a critically important role in regulating the levels of other lower abundance lipids that can shape local and systemic inflammatory and hormonal responses to influence T2D mellitus, NAFLD and other related metabolic diseases. To further understand how MBOAT7 impacts adipose tissue metabolism, we performed a series of comprehensive lipidomics studies in gWAT isolated from chow and HFD fed Mboat7flox/flox control mice and Mboat7ASKO mice. Although MBOAT7 has been previously shown to have exquisite substrate selectivity toward saturated LPIs and arachidonyl-CoA [41, 42, 43], we also analyzed the abundance of other free fatty acids, oxylipins, and diverse species of neutral lipids and glycerophospholipids. First, as a reference, it is notable that hepatocyte-specific deletion of Mboat7 (Mboat7HSKO mice) results in a large accumulation of substrate LPIs and a modest reduction in 38:4 PI in the liver but this is not apparent in gWAT (supplemental Fig. S2G–L). In contrast, Mboat7ASKO mice do not show alteration in LPI or PI lipids in the liver (supplemental Fig. S1D–F) but do have elevations in 18:0- and 18:1-containing LPIs and large reductions in 38:4 PI in gWAT, especially in the context of HFD (Fig. 2B–D). Mboat7ASKO mice also have reciprocal increases in more saturated PI species (34:2, 36:2, and 36:3) compared to Mboat7flox/flox control mice in gWAT (Fig. 2D), but not liver (supplemental Fig. S1F). These results demonstrate that MBOAT7 plays a critical role in maintaining LPI and PI levels within the local tissue context and shows that MBOAT7 is the major enzymatic source of 38:4 in WAT (Fig. 2C) but a more minor contributor in the liver (supplemental Fig. S2H) [15, 16, 17]. Given the critical role that MBOAT7 plays in WAT LPI/PI homeostasis, we wanted to more comprehensively understand whether other lipid classes were altered in the WAT of Mboat7ASKO mice. Other than modest alterations in 36:2 and 36:3 phosphatidylethanolamine (PE), adipocyte-specific deletion of Mboat7 did not result in significant alterations of free fatty acids, non-LPI lysophospholipids, phosphatidylcholines, sphingomyelins, or ceramides (supplemental Fig. S7). Given the critical role that MBOAT7 plays in the Land’s cycle remodeling pathway [44], we also wanted to analyze arachidonic acid (AA)-derived oxylipins. When we quantified the WAT levels of >60 species of oxylipins, none were significantly different between Mboat7flox/flox control mice and Mboat7ASKO mice (supplemental Fig. S8). Consistent with this finding, we did not see any changes in Pla2g4a expression between Mboat7flox/flox and Mboat7ASKO mice (supplemental Fig. S9A). Also, when we quantified the levels of neutral lipids such as diacylglycerols and TAGs there were no significant differences between Mboat7flox/flox control mice and Mboat7ASKO mice when fed an HFD, except that select species of TAG (52:3, 56:4, 50:5, etc.) were modestly reduced in chow fed Mboat7ASKO mice (supplemental Fig. S10). Despite the conservation seen in the major lipid classes found in WAT, Mboat7ASKO mice did have some significant alterations in some complex lipids containing either AA (20:4, n-6) or eicosapentaenoic acid (EPA, 20:5, n-3) in WAT (supplemental Fig. S11) Specifically, some AA-containing PE species were increased in the WAT from Mboat7ASKO mice (supplemental Fig. S11A); however, the expression of other AA-specific acyl transferase, Lpcat3 (also known as Mboat5) was not changed (supplemental Fig. S9B). Furthermore, EPA-containing PI (38:5 PI) was dramatically reduced in Mboat7ASKO mice, while other EPA-containing PS and PE species were slightly elevated (supplemental Fig. S11C). These data support the notion that adipocyte MBOAT7 plays a central role in WAT glycerophospholipid homeostasis locally but does not appreciably impact LPI and PI balance in the circulation or the liver. Given the alterations in WAT lipid homeostasis in Mboat7ASKO mice, we wanted to examine key aspects of adipose tissue function and cellularity within the adipose organ. Although body weight and total lean and fat mass were not significantly different between Mboat7flox/flox control mice and Mboat7ASKO mice after 8 weeks on diet (Fig. 2K–M), the weight of gWAT was significantly lower in Mboat7ASKO mice after 20 weeks of HFD feeding (Fig. 4A), inverse of the significant increase in liver weight of Mboat7ASKO animals at 20 weeks of HFD (Fig. 2F). However, less than 20 weeks of HFD feeding did not result in significant differences in liver or adipose tissue weights (supplemental Fig. S3A–D). Circulating levels of the adipocyte-derived hormone leptin were also reduced in HFD fed Mboat7ASKO mice after 16 and 20 weeks of HFD (Fig. 4B and supplemental Fig. S3E). Next, we wanted to examine whether basal or catecholamine-stimulated lipolysis (i.e., a critical physiological function of WAT) was altered inMboat7ASKO mice. Although in general the circulating levels of lipolysis product glycerol and NEFA were not altered under basal conditions or stimulated conditions in Mboat7-deficient mice, chow fed Mboat7ASKO mice did have elevations in circulating glycerol when treated with the β3-adrenergic receptor agonist CL-326,243 (Fig. 4C, D). Similarly, we did not observe significant changes in plasma glycerol, NEFA, or triglycerides in Mboat7ASKO versus control mice under fasting or refeeding conditions (supplemental Fig. S3F–H). Next, we performed bulk RNA sequencing in gWAT. Under HFD feeding conditions, Mboat7ASKO mice exhibited differential gene expression associated with altered lipid metabolism and immune cell populations (Fig. 4E–G). In particular, Mboat7ASKO mice had elevated WAT mRNA expression of genes primarily expressed in macrophages including a cluster of differentiation 68 (Cd68), integrin alpha X (Itgax, encoding CD11c), C-type lectin domain containing 7a (Clec7a), among others (Fig. 4F, G) indicating the potential for an altered abundance of adipose tissue macrophages. To further understand whether the abundance of macrophages and other immune cell populations were altered in Mboat7ASKO mice fed HFD for 16 weeks, we isolated the stromal vascular fraction from gWAT and performed flow cytometric analysis of immune cell populations (Fig. 4H–J and supplemental Fig. S12). Of the adipose tissue macrophage populations identified, Mboat7ASKO mice had elevated levels of Cd11c-/Cd206+ cell populations and a reciprocal decrease in Cd11c+/Cd206+ double-positive cells (Fig. 4G). Although the relative percentage of Cd11c-/Cd206+ cells was not significantly correlated with body weight (Fig. 4H), the Cd11c+/Cd206+ double-positive cell population was significantly positively correlated with body weight in this cohort (Fig. 4I, J). In addition to alterations in macrophage subsets, HFD fed Mboat7ASKO mice had elevated T cells and B cells compared to control (Mboat7flox/flox) mice (supplemental Fig. S12B–F). These data demonstrate that adipocyte MBOAT7-driven LPI acylation plays an important role in WAT lipid metabolic and immune cell homeostasis which can then shape systemic glucose tolerance. Fig. 4Adipocyte-Specific Mboat7 Deletion (Mboat7ASKO) Reorganizes White Adipose Tissue Gene Expression and Circulating Adipokine Levels. Male control (Mboat7fl/fl) or adipocyte-specific Mboat7 knockout mice (Mboat7ASKO) were fed chow or high fat diet (HFD) for 20-week. A: gWAT weight measurements from Mboat7fl/fl or Mboat7ASKO mice fed Chow and HFD for 20 weeks ($$n = 6$$–8; Two-way ANOVA with Tukey’s post hoc test). B: Plasma Leptin was measured in Mboat7fl/fl or Mboat7ASKO mice that were fed a chow or HFD for 20 weeks ($$n = 5$$–7; Two-way ANOVA with Tukey’s post hoc test). To assess β3–adrenergic stimulated lipolysis, plasma nonesterified fatty acids (NEFA), (C) and glycerol (D) were measured in Mboat7fl/fl or Mboat7ASKO mice fed a chow or HFD for 11 weeks 15 min after saline or CL316,243 injection ($$n = 2$$–5/group; Three-way ANOVA with Tukey’s post hoc test). E–G: gWAT RNA was used for RNA-sequencing from Mboat7fl/fl or Mboat7ASKO mice that were fed a chow or HFD for 20 weeks. E: Groups clustered primarily based on the diet by principal component analysis ($$n = 4$$/group). F: A Volcano plot of transcripts was used to determine differentially expressed genes (DEGs) in Mboat7fl/fl or Mboat7ASKO mice that were fed an HFD for 20 weeks. Plot summarizes log2 fold changes versus significance in response to Mboat7 inhibition ($$n = 4$$/group; genes with q-val < 0.05 and fold change > |0.5| were considered significantly differentially expressed). G: Row-normalized expression for the top 20 up and downregulated DEGs that reached a P-value <0.05 are shown by heat map in Mboat7fl/fl or Mboat7ASKO mice that were fed an HFD for 20 weeks. H: gWAT stromal vascular fraction was subjected to flow cytometry analysis of macrophage subpopulations. I, J: Correlation between gWAT macrophage subsets and body weight. All data are presented as mean ± S.D. unless otherwise noted. ## LPIs induce acute signaling events associated with insulin action in adipose tissue in a MBOAT7-dependent manner Given that 1) LPIs accumulate in gWAT of HFD fed and Mboat7ASKO mice (Fig. 2B and supplemental Fig. S4A), 2) Mboat7ASKO results in profound glucose intolerance and insulin resistance (Fig. 3, supplemental Figs. S4H, I and S5), and 3) LPIs correlate with fasting blood glucose and area under the curve during glucose tolerance testing (supplemental Fig. S6B–D), we hypothesized that LPIs may mediate signaling associated with insulin resistance. To test this hypothesis, we injected chow fed Mboat7flox/flox and Mboat7ASKO mice with saline or LPI-18:1 and collected gWAT acutely after 15 min. To unbiasedly identify LPI-driven signaling events we performed global phosphoproteomic analysis of isolated gWAT and found that. Mboat7ASKO mice upregulate 163 phosphopeptides in response to LPI but not in control Mboat7flox/flox mice. Using KEGG pathway analysis, we find that indeed LPI-18:1 significantly induces a phosphopeptide signature associated with insulin signaling and insulin resistance in mice lacking MBOAT7 function in adipocytes (Mboat7ASKO mice) (Fig. 5A–C). In contrast, mice with intact MBOAT7 function (Mboat7flox/flox mice) show only 40 significantly altered phosphopeptides and no significantly altered KEGG pathways when challenged with exogenous LPI (supplemental Fig. S13). Together, these data indicate that LPI-18:1 can alter the gonadal adipose tissue phosphoproteome in a MBOAT7-dependent manner and this is associated with pathways linked to insulin resistance. Fig. 5LPI-18:1 Induces Acute Signaling Events Associated with Insulin Action in Adipose Tissue of Mboat7ASKO mice. A: Volcano plot of phosphopeptides upregulated and downregulated in gonadal white adipose tissue (gWAT) of Mboat7ASKO mice. Phosphopeptides determined to be |log2FoldChange| >0.5 different in the LPI and saline samples with a P-value <0.05 (two-tailed t test) were considered significantly differentially phosphorylated. ( Average of $$n = 4$$/group; pink dots represent significantly upregulated phospho-peptides, green dots represent significantly downregulated phospho-peptides; and purple dots represent insulin resistance-associated phosphopeptides that are significantly different). B, C: KEGG pathway analysis of significantly differentially phosphorylated peptides. KEGG, Kyoto Encyclopedia of Genes and Genomes; LPI, lysophosphatidylinositol; Mboat7ASKO, Mboat7 adipocyte-specific knockout mice. ## Discussion Since the original genome wide association studies study by Buch et al. [ 8] linking the rs641738 SNP near MBOAT7 to liver disease, there has been rapid progress in our understanding of how MBOAT7 is mechanistically linked to the progression of alcohol-associated liver disease, NAFLD, and viral-driven liver injury. The clear association between MBOAT7 loss-of-function and diverse liver diseases serves as yet another example of how genetics can powerfully identify new pathways relevant to human disease. It also supports the long-standing notion that abnormal lipid metabolism initiates liver injury. Since 2019, several animal studies have likewise demonstrated that Mboat7 loss-of-function in mice is sufficient to drive NAFLD progression [13, 14, 15, 16, 17]. This article builds on our initial observation that ASO-mediated knockdown of Mboat7 promotes NAFLD progression, hyperinsulinemia, and insulin resistance in mice [13]. Here we have further clarified the cell autonomous roles of Mboat7 in HFD-driven metabolic disturbance by comparing metabolic phenotypes in Mboat7ASKO and Mboat7HSKO mice. The major findings of the current study include the following: 1) Mboat7 expression in WAT is negatively correlated with adiposity across the strains represented in the HMDP, 2) Hepatocyte-specific deletion of Mboat7 (Mboat7HSKO) promotes fatty liver and liver injury but does not alter tissue insulin sensitivity in HFD fed mice, 3) Adipocyte-specific deletion of Mboat7 (Mboat7ASKO) under chronic HFD feeding promotes mild fatty liver, clear hyperinsulinemia, and systemic insulin resistance in HFD fed mice, 4) MBOAT7 is the major source of AA containing PI (38:4 PI) and also indirectly impacts other glycerophospholipids in gWAT, 5) MBOAT7 function in gWAT does not alter the tissue abundance of AA-derived oxylipins, 6) MBOAT7 function in gWAT impacts circulating leptin levels, and the immune cell landscape in WAT under HFD feeding, and 7) acylation of LPIs by MBOAT7 in WAT limits proinsulin resistance signaling. Collectively, these data support a clear cell autonomous role for MBOAT7-driven acylation of LPI lipids as a key protective mechanism against obesity-linked NAFLD progression, hyperinsulinemia, and systemic insulin resistance. These data strongly suggest that MBOAT7 is an important contributor to multiple aspects of the metabolic syndrome, which is regulated by the combination of MBOAT7 function in hepatocytes, adipocytes, and likely other cell types that contribute to tissue inflammation and fibrosis. One of the key findings of this work is that MBOAT7 function in adipocytes is critically important for the maintenance of euglycemia in obese mice (Fig. 3). In contrast, genetic deletion of Mboat7 in hepatocytes is sufficient to drive hepatic steatosis but does not drastically alter hyperinsulinemia or insulin sensitivity in either chow or HFD fed mice (supplemental Fig. S2). Interestingly, our studies show drastic effects on systemic glucose tolerance and insulin resistance in HFD fed Mboat7ASKO mice yet our euglycemic-hyperinsulinemic studies indicate no change in suppression of hepatic glucose production by insulin and only modest decreases in insulin-stimulated glucose uptake by white adipose depots (Fig. 3 and supplemental Fig. S5). Because these depots show less glucose uptake activity than other tissues that account for the majority of glucose disposal, such as skeletal and cardiac muscle, BAT, and the brain, there is likely other mechanisms that contribute to these systemic effects. While glucose intolerance and insulin resistance was obvious in HFD fed Mboat7ASKO mice, these phenotypes are not significant in chow fed Mboat7ASKO mice, which may be explained in part by significant increases in β3-adrenergic activation (Fig. 4D), which has been previously shown to enhance fatty acid oxidation in adipose tissue of mice and humans [45, 46]. In addition to these findings regarding local WAT defects, these studies raise the question of how MBOAT7 in adipocytes controls insulin secretion in the pancreas and fat accumulation in the liver. The most straightforward explanation is that alterations in either the substrates (LPIs or fatty acyl-CoAs) or products (38:4 PI) of the MBOAT7 reaction initiate endocrine lipid signaling effects. We originally hypothesized that when adipocyte MBOAT7 function is lost, substrate LPIs would accumulate in both the WAT as well as the circulation to initiate a WAT-to-pancreas endocrine axis to stimulate insulin overproduction. In support of this concept, several published reports show that LPI lipids can stimulate glucose-stimulated insulin secretion in pancreatic beta cells in culture [47, 48]. Furthermore, 18:1 LPI has also been shown to stimulate the key metabolic hormone glucagon-like peptide in intestinal enteroendocrine L-cells to further shape insulin secretion [49, 50]. However, our results suggest that although WAT tissue of LPI is dramatically increased in Mboat7ASKO mice (Fig. 2B), circulating levels of LPI lipids are unaltered in Mboat7ASKO mice (supplemental Fig. S1). These data suggest this potential lipid-mediated endocrine axis is unlikely, although there is a clear LPI-mediated autocrine or paracrine axis within WAT (Fig. 5). It is important to note however that LPI lipids can exhibit altered signaling potential when bound to albumin versus being carried on plasma lipoproteins [51], so additional work is required to determine whether adipocyte MBOAT7 activity selectively impacts albumin-conjugated versus lipoprotein-associated LPI levels. In addition to an LPI-endocrine signaling axis, it is possible that MBOAT7 can impact adipose tissue homeostasis and systemic insulin sensitivity via other mechanisms. It remains possible that MBOAT7 product PI lipids including 38:4 PI and resulting phosphoinositides may also play a key role in the phenotypes observed. Another possibility to consider is that glycerophosphatidylinositol-linked proteins may be altered in Mboat7ASKO mice given that PIs are required for glycerophosphatidylinositol synthesis [52]. These possibilities deserve further exploration. Another interesting observation coming from the current study is the different molecular weights of MBOAT7 observed in crude tissue lysates (Fig. 2A) versus the consistent size in microsomal fractions (supplemental Fig. S2A), which could suggest that multiple MBOAT7 isoforms exist in different tissues and may localize to organelles other than the endoplasmic reticulum. In conclusion, this work shows that MBOAT7 function in adipocytes and hepatocytes play unique roles in shaping HFD-driven metabolic disturbance and further supports the notion that the LPI-MBOAT-PI axis may have untapped therapeutic potential in obesity-related insulin resistance and NAFLD progression. ## Data availability RNA sequencing data can be accessed at GEO Profiles at the NCBI: GSE203414, the mass spectrometry phospoproteomics data have been deposited to the ProteomeXchange *Consortium via* the PRIDE [34] partner repository with the dataset identifier PXD039894 and 10.6019/PXD039894 and full accession will be available upon acceptance of this work. All materials, methods, and datasets included in this article is readily available upon request. ## Supplemental data This article contains supplemental data [53, 54]. Supplemental Data Supplemental Tables S2–S7 Supplemental Figures S1–S13 and Table S1 ## Conflict of interest Dr. Daniela Allende reports serving as an Advisory Board Member for Incyte Corporation. All other authors declare no competing financial interests related to this work. ## Author contributions W. J. M. and J. M. B. planned the project; W. J. M., D. S. A., B. W., F. C., A. J. L., and J. M. B. designed experiments; W. J. M. and J. M. B. analyzed data; W. J. M. and J. M. B. wrote the original draft of the manuscript; D. S. A., B. W., F. C., and A. J. L. provided useful discussion directing collaborative aspects of the project; W. J. M., V. V., R. B., A. L. B., A. J. H., R. C. H., B. M. J., Y. Q., E. R. C., C. P., R. Z., D. S. A., B. W., F. C., A. J. L., and J. M. B. either conducted mouse experiments, performed biochemical workup of mouse tissues, or analyzed data; W. J. M., V. V., R. B., A. L. B., A. J. H., R. C. H., B. M. J., Y. Q., E. R. C., C. P., R. Z., D. S. A., B. W., F. C., A. J. L., and J. M. B. aided in manuscript preparation; W. J. M., V. V., R. B., A. L. B., A. J. H., R. C. H., B. M. J., Y. Q., E. R. C., C. P., R. Z., D. S. A., B. W., F. C., A. J. L., and J. M. B. edited the final manuscript. ## Funding and additional information This work was supported in part by $\frac{10.13039}{100000002}$National Institutes of Health grants R01 DK120679 (J. M. B.), P01 HL147823 (J. M. B.), P50 AA024333 (J. M. B.), U01 AA026938 (J. M. B.), R01 DK130227 (J. M. B.), R01 DK117850 (A. J. L.), and R01 HL148577 (A. J. L.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. ## References 1. Cohen J.C., Horton J.D., Hobbs H.H.. **Human fatty liver disease: old questions and new insights**. *Science* (2011) **332** 1519-1523. PMID: 21700865 2. Wree A., Broderick L., Canbay A., Hoffman H.M., Feldstein A.E.. **From NAFLD to NASH to cirrhosis-new insights into disease mechanisms**. *Nat. Rev. Gastroenterol. Hepatol.* (2013) **10** 627-636. PMID: 23958599 3. Rinella M.E., Sanyal A.J.. **Management of NAFLD: a stage-based approach**. *Nat. Rev. Gastroenterol. Hepatol.* (2016) **13** 196-205. PMID: 26907882 4. Sarwar R., Pierce N., Koppe S.. **Obesity and nonalcoholic fatty liver disease: current perspectives**. *Diabetes Metab. Syndr. Obes.* (2018) **11** 533-542. PMID: 30288073 5. Mancina R.M., Dongiovanni P., Petta S., Pingitore P., Meroni M., Rametta R.. **The MBOAT7-TMC4 variant rs641738 increases risk of nonalcoholic fatty liver disease in individuals of European descent**. *Gastroenterology* (2016) **150** 1219-1230. PMID: 26850495 6. Luukkonen P.K., Zhou Y., Hyotylainen T., Leivonen M., Arola J., Orho-Melander M.. **The MBOAT7 variant rs641738 alters hepatic phosphatidylinositols and increases severity of non-alcoholic fatty liver disease in humans**. *J. Hepatol.* (2016) **65** 1263-1265. PMID: 27520876 7. Teo K., Abeysekera K.W.M., Adams L., Aigner E., Anstee Q.M., Banales J.M.. **rs641738C>T near MBOAT7 is associated with liver fat, ALT and fibrosis in NAFLD: a meta-analysis**. *J. Hepatol.* (2021) **74** 20-30. PMID: 32882372 8. Buch S., Stickel F., Trepo E., Way M., Herrmann A., Nischalke H.D.. **A genome-wide association study confirms PNPLA3 and identifies TM6SF2 and MBOAT7 as risk loci for alcohol-related cirrhosis**. *Nat. Genet.* (2015) **47** 1443-1448. PMID: 26482880 9. Thabet K., Asimakopoulos A., Shojaei M., Romero-Gomez M., Mangia A., Irving W.L.. **MBOAT7 rs641738 increases risk of liver inflammation and transition to fibrosis in chronic hepatitis C**. *Nat. Commun.* (2016) **7** 12757. PMID: 27630043 10. Thabet K., Chan H.L.Y., Petta S., Mangia A., Berg T., Boonstra A.. **The membrane-bound**. *Hepatology* (2017) **65** 1840-1850. PMID: 28109005 11. Johansen A., Rosti R.O., Musaev D., Sticca E., Harripaul R., Zaki M.. **Mutations in MBOAT7, encoding lysophosphatidylinositol acyltransferase I, lead to intellectual disability accompanied by epilepsy and autistic features**. *Am. J. Hum. Genet.* (2016) **99** 912-916. PMID: 27616480 12. Jacher J.E., Roy N., Ghaziuddin M., Innis J.W.. **Expanding the phenotypic spectrum of MBOAT7-related intellectual disability**. *Am. J. Med. Genet. B Neuropsychiatr. Genet.* (2019) **180** 483-487. PMID: 31282596 13. Helsley R.N., Varadharajan V., Brown A.L., Gromovsky A.D., Schugar R.C., Ramachandiran I.. **Obesity-linked suppression of membrane-bound O-acyltransferase 7 (MBOAT7) drives non-alcoholic fatty liver disease**. *Elife* (2019) **8** 14. Meroni M.P., Dongiovanni M., Longo M., Carli F., Baselli G., Rametta R.. **Mboat7 down-regulation by hyper-insulinemia induces fat accumulation in hepatocytes**. *EBioMedicine* (2020) **52** 102658. PMID: 32058943 15. Tanaka Y., Shimanaka Y., Caddeo A., Kubo T., Mao Y., Kubota T.. **LPIAT1/MBOAT7 depletion increases triglyceride synthesis fueled by high phosphatidylinositol turnover**. *Gut* (2021) **70** 180-193. PMID: 32253259 16. Thangapandi V.R., Knittelfelder O., Brosch M., Patsenker E., Vvedenskaya O., Buch S.. **Loss of hepatic**. *Gut* (2021) **70** 940-950. PMID: 32591434 17. Xia M., Chandrasekaran P., Rong S., Fu X., Mitsche M.A.. **Hepatic deletion of**. *J. Lipid Res.* (2021) **62** 100031. PMID: 32859645 18. Crooke S.T., Baker B.F., Crooke R.M., Liang X.H.. **Antisense technology: an overview and prospectus**. *Nat. Rev. Drug Discov.* (2021) **20** 427-453. PMID: 33762737 19. Anderson K.E., Kielkowska A., Durrant T.N., Juvin V., Clark J.,, Stephens L.R.. **Lysophosphatidylinositol-acyltransferase-1 (LPIAT1) is required to maintain physiological levels of PtdIns and PtdInsP(2) in the mouse**. *PLoS One* (2013) **8** 20. Schugar R., Shih D.M., Warrier M., Helsley R.N., Burrows A., Ferguson D.. **The TMAO-producing enzyme flavin-containing monooxygenase 3 regulates obesity and the beiging of white adipose tissue**. *Cell Rep.* (2017) **19** 2451-2461. PMID: 28636934 21. Warrier M., Shih D.M., Burrows A.C., Ferguson D., Gromovsky A.D., Brown A.L.. **The TMAO-generating enzyme flavin monooxygenase 3 is a central regulator of cholesterol balance**. *Cell Rep.* (2015) **10** 326-338. PMID: 25600868 22. Gromovsky A.D., Schugar R.C., Brown A.L., Helsley R.N., Burrows A.C., Ferguson D.. **Δ-5 fatty acid desaturase FADS1 impacts metabolic disease by balancing proinflammatory and proresolving lipid mediators**. *Arterioscler. Thromb. Vasc. Biol.* (2017) **38** 218-231. PMID: 29074585 23. Brown J.M., Betters J.L., Lord C., Ma Y., Han X., Yang K.. **CGI-58 knockdown in mice causes hepatic steatosis but prevents diet-induced obesity and glucose intolerance**. *J. Lipid Res.* (2010) **51** 3306-3316. PMID: 20802159 24. Brown J.M., Chung S., Sawyer J.K., Degirolamo C., Alger H.M., Nguyen T.. **Inhibition of stearoyl-coenzyme A desaturase 1 dissociates insulin resistance and obesity from atherosclerosis**. *Circulation* (2008) **118** 1467-1475. PMID: 18794388 25. Brown J.M., Bell T.A., Alger H.M., Sawyer J.K., Smith T.L., Kelley K.. **Targeted depletion of hepatic ACAT2-driven cholesterol esterification reveals a non-biliary route for fecal sterol loss**. *J. Biol. Chem.* (2008) **16** 10522-10534 26. Berglund E.D., Li C.Y., Poffenberger G., Ayala J.E., Fueger P.T., Willis S.E.. **Glucose metabolism in vivo in four commonly used inbred mouse strains**. *Diabetes* (2008) **57** 1790-1799. PMID: 18398139 27. Ayala J.E., Bracy D.P., McGuinness O.P., Wasserman D.H.. **Considerations in the design of hyperinsulinemic-euglycemic clamps in the conscious mouse**. *Diabetes* (2006) **55** 390-397. PMID: 16443772 28. Finegood D.T., Bergman R.N., Vranic M.. **Estimation of endogenous glucose production during hyperinsulinemic-euglycemic glucose clamps. Comparison of unlabeled and labeled exogenous glucose infusates**. *Diabetes* (1987) **36** 914-924. PMID: 3297886 29. Steele R., Wall J.S., De Bodo R.C., Altszuler N.. **Measurement of size and turnover rate of body glucose pool by the isotope dilution method**. *Am. J. Physiol.* (1985) **187** 15-24 30. Kraegen E.W., James D.E., Jenkins A.B., Chisholm D.J.. **Dose-response curves for in vivo insulin sensitivity in individual tissues in rats**. *Am. J. Physiol.* (1985) **248** E353-362. PMID: 3883806 31. Dobin A., Davis C.A., Schlesinger F., Drenkow J., Zaleski C., Jha S.. **STAR: ultrafast universal RNA-seq aligner**. *Bioinformatics* (2013) **29** 15-21. PMID: 23104886 32. Trapnell C., Williams B.A., Pertea G., Mortazavi A., Kwan G., van Baren M.J.. **Transcript assembly and quantification by Rna-seq reveals unannotated transcripts and isoform switching during cell differentiation**. *Nat. Biotechnol.* (2010) **28** 511-515. PMID: 20436464 33. Kuleshov M.V., Jones M.R., Rouillard A.D., Fernandez N.F., Duan Q., Wang Z.. **Enrichr: a comprehensive gene set enrichment analysis web server 2016 update**. *Nucleic Acids Res.* (2016) **44** W90-W97. PMID: 27141961 34. Perez-Riverol Y., Bai J., Bandla C., Hewapathirana S., García-Seisdedos D., Kamatchinathan S.. **The PRIDE database resources in 2022: a hub for mass spectrometry-based proteomics evidences**. *Nucleic Acids Res.* (2022) **50** D543-D552. PMID: 34723319 35. Ghazalpour A., Rau C.D., Farber C.R., Bennett B.J., Orozco L.D., van Nas A.. **Hybrid mouse diversity panel: a panel of inbred mouse strains suitable for analysis of complex genetic traits**. *Mamm. Genome* (2012) **23** 680-692. PMID: 22892838 36. Parks B.W., Nam E., Org E., Kostem E., Norheim F., Hui S.T.. **Genetic control of obesity and gut microbiota composition in response to high-fat, high-sucrose diet in mice**. *Cell Metab.* (2013) **17** 141-152. PMID: 23312289 37. Eguchi J., Wang X., Yu S., Kershaw E.E., Chiu P.C., Dushay J.. **Transcriptional control of adipose lipid handling by IRF4**. *Cell Metab.* (2011) **13** 249-259. PMID: 21356515 38. Unger R.. **Lipotoxicity in the pathogenesis of obesity-dependent NIDDM. Genetic and clinical implications**. *Diabetes* (1995) **44** 863-870. PMID: 7621989 39. Unger R.H., Clark G.O., Scherer P.E., Orci L.. **Lipid homeostasis, lipotoxicity and the metabolic syndrome**. *Biochim. Biophys. Acta* (2010) **1801** 209-214. PMID: 19948243 40. Samuel V.T., Shulman G.I.. **Nonalcoholic fatty liver disease as a nexus of metabolic and hepatic diseases**. *Cell Metab.* (2018) **27** 22-41. PMID: 28867301 41. Gijón M.A., Riekhof W.R., Zarini S., Murphy R.C., Voelker D.R.. **Lysophospholipid acyltransferases and arachidonate recycling in human neutrophils**. *J. Biol. Chem.* (2008) **283** 30235-30245. PMID: 18772128 42. Lee H.C., Inoue T., Sasaki J., Kubo T., Matsuda S., Nakasaki Y.. **LPIAT1 regulates arachidonic acid content in phosphatidylinositol and is required for cortical lamination in mice**. *Mol. Biol. Cell* (2012) **23** 4689-4700. PMID: 23097495 43. Zarini S., Hankin J.A., Murphy R.C., Gijón M.A.. **Lysophospholipid acyltransferases and arachidonate recycling in human neutrophils**. *Prostaglandins Other Lipid Mediat.* (2014) **113-115** 52-61. PMID: 25175316 44. Valentine W.J., Hashidate-Yoshida T., Yamamoto S., Shindou H.. **Biosynthetic enzymes of membrane glycerophospholipid diversity as therapeutic targets for drug development**. *Adv. Exp. Med. Biol.* (2020) **1274** 5-27. PMID: 32894505 45. Kim H., Pennisi P.A., Gavrilova O., Pack S., Jou W., Setser-Portas J.. **Effect of adipocyte beta3-adrenergic receptor activation on the type 2 diabetic MKR mice**. *Am. J. Physiol. Endocrinol. Metab.* (2006) **6** E1227-E1236 46. Weyer C., Tataranni P.A., Snitker S., Danforth E., Ravussin E.. **Increase in insulin action and fat oxidation after treatment with CL 316,243, a highly selective beta3-adrenoceptor agonist in humans**. *Diabetes* (1998) **10** 1555-1561 47. Metz S.A.. **Lysophosphatidylinositol, but not lysophosphatidic acid, stimulates insulin release. A possible role for phospholipase A2 but not de novo synthesis of lysophospholipid in pancreatic islet function**. *Biochem. Biophys. Res. Commun.* (1986) **138** 720-727. PMID: 3527169 48. Liu B., Song S., Ruz-Maldonado I., Pingitore A., Huang G.C., Baker D.. **GPR55-dependent stimulation of insulin secretion from isolated mouse and human islets of Langerhans**. *Diabetes Obes. Metab.* (2016) **18** 1263-1273. PMID: 27561953 49. Arifin S.A., Paternoster S., Carlessi R., Casari I., Ekberg J.H., Maffucci T.. **Oleoyl-lysophosphatidylinositol enhances glucagon-like peptide-1 secretion from enteroendocrine L-cells through GPR119**. *Biochim. Biophys. Acta Mol. Cell Biol. Lipids* (2018) **1863** 1132-1141. PMID: 29883799 50. Harada K., Kitaguchi T., Kamiya T., Aung K.H., Nakamura K., Ohta K.. **Lysophosphatidylinositol-induced activation of the cation channel TRPV2 triggers glucagon-like peptide-1 secretion in enteroendocrine L-cells**. *J. Biol. Chem.* (2017) **292** 10855-10864. PMID: 28533434 51. Kurano M., Kobayashi T., Sakai E., Tsukamoto K., Yatomi Y.. **Lysophosphatidylinositol, especially albumin-bound form, induces inflammatory cytokines in macrophages**. *FASEB J.* (2021) **35** 52. Kanzawa N., Maeda Y., Ogiso H., Murakami Y., Taguchi R., Kinoshita T.. **Peroxisome dependency of alkyl-containing GPI-anchor biosynthesis in the endoplasmic reticulum**. *Proc. Natl. Acad. Sci. U.S.A.* (2009) **106** 17711-17716. PMID: 19815513 53. Folz J.S., Shalon D., Fiehn O.. **Metabolomics analysis of time-series human small intestine lumen samples collected in vivo**. *Food & Function* (2021) **12** 9405-9415. PMID: 34606553 54. Massey W., Osborn L.J., Banerjee R., Horak A., Fung K.K., Orabi D.. **Flavin-Containing Monooxygenase 3 (FMO3) Is Critical for Dioxin-Induced Reorganization of the Gut Microbiome and Host Insulin Sensitivity**. *Metabolites* (2022) **12**. DOI: 10.3390/metabo12040364
--- title: 'Multi-organ impairment and long COVID: a 1-year prospective, longitudinal cohort study' authors: - Andrea Dennis - Daniel J Cuthbertson - Dan Wootton - Michael Crooks - Mark Gabbay - Nicole Eichert - Sofia Mouchti - Michele Pansini - Adriana Roca-Fernandez - Helena Thomaides-Brears - Matt Kelly - Matthew Robson - Lyth Hishmeh - Emily Attree - Melissa Heightman - Rajarshi Banerjee - Amitava Banerjee journal: Journal of the Royal Society of Medicine year: 2023 pmcid: PMC10041626 doi: 10.1177/01410768231154703 license: CC BY 4.0 --- # Multi-organ impairment and long COVID: a 1-year prospective, longitudinal cohort study ## Abstract ### Objectives To determine the prevalence of organ impairment in long COVID patients at 6 and 12 months after initial symptoms and to explore links to clinical presentation. ### Design Prospective cohort study. ### Participants Individuals. ### Methods In individuals recovered from acute COVID-19, we assessed symptoms, health status, and multi-organ tissue characterisation and function. ### Setting Two non-acute healthcare settings (Oxford and London). Physiological and biochemical investigations were performed at baseline on all individuals, and those with organ impairment were reassessed. ### Main outcome measures Primary outcome was prevalence of single- and multi-organ impairment at 6 and 12 months post COVID-19. ### Results A total of 536 individuals (mean age 45 years, $73\%$ female, $89\%$ white, $32\%$ healthcare workers, $13\%$ acute COVID-19 hospitalisation) completed baseline assessment (median: 6 months post COVID-19); 331 ($62\%$) with organ impairment or incidental findings had follow-up, with reduced symptom burden from baseline (median number of symptoms 10 and 3, at 6 and 12 months, respectively). Extreme breathlessness ($38\%$ and $30\%$), cognitive dysfunction ($48\%$ and $38\%$) and poor health-related quality of life (EQ-5D-5L < 0.7; $57\%$ and $45\%$) were common at 6 and 12 months, and associated with female gender, younger age and single-organ impairment. Single- and multi-organ impairment were present in $69\%$ and $23\%$ at baseline, persisting in $59\%$ and $27\%$ at follow-up, respectively. ### Conclusions Organ impairment persisted in $59\%$ of 331 individuals followed up at 1 year post COVID-19, with implications for symptoms, quality of life and longer-term health, signalling the need for prevention and integrated care of long COVID. Trial Registration: ClinicalTrials.gov Identifier: NCT04369807 ## Background Symptoms of long COVID, also known as post-acute sequelae of COVID-19, are well documented,1,2 but natural history is poorly characterised, either by symptoms, organ impairment or function.3–5 Among 3762 individuals with suspected or confirmed COVID-19, debilitating symptoms lasted beyond 35 weeks, with fatigue, breathlessness and cognitive dysfunction being the most frequent. 6 In the UK’s largest long COVID clinic, non-hospitalised patients required specialist referral at similar rates to hospitalised patients and were more likely to report breathlessness and fatigue, with reduced health-related quality of life (HRQoL). 7 A US study of 270,000 individuals post COVID-19 showed that one-third had persistent symptoms at 3–6 months (more common than post-influenza symptoms based on a matched cohort with otherwise similar risk factors), possibly due to direct organ-specific rather than general viral effects, and potentially informing development of effective treatments. 8 Long COVID may be linked to severity of initial illness in some hospitalised patients, but prognostic factors are neither defined nor investigated systematically in non-hospitalised patients.7–9 To conduct trials of possible therapies for long COVID, we need stratification by symptoms or investigations. 3 Our interim magnetic resonance imaging (MRI) data in 201 individuals showed mild organ impairment in the heart, lungs, kidneys, liver, pancreas and spleen, with single- and multi-organ impairment in $70\%$ and $29\%$, respectively, 4 months after COVID-19. 9 Clinical utility of these MRI metrics for chronic and multi-system conditions has been shown. 10,11 More severe ongoing symptoms of breathlessness and fatigue were associated with myocarditis ($p \leq 0.05$) 9, but symptoms and multi-organ manifestations have not been correlated. In the UK and other countries, health system and research responses have begun at scale. 12 However, clinical patient pathways are unclear and there are still no proven, evidence-based therapies, either in subgroups or in the overall long COVID population. Single- and multi-organ impairments need investigation over the medium and long term to assess resource utilisation and health system needs. In individuals with long COVID, we therefore prospectively investigated the following: Symptoms, organ impairment and function over 1 year, particularly relating to ongoing breathlessness, cognitive dysfunction and HRQoL.Associations between symptoms and organ impairment. ## Patient population and study design This study took place during the COVID-19 emergency, before the formal definition of long COVID. Patients with evidence of COVID-19 who no longer had active SARS-CoV-2 infection, but had ongoing symptoms, could enter the study. A retrospective evaluation of the duration of the symptoms defined individuals who had long COVID based on continuing symptoms for ≥12 weeks, as informed by the earliest WHO, NICE and NHS England guidance regarding post-COVID complications that are now in current guidelines, policy and practice in the UK. 7 Recruitment was by response to advertisement or specialist referral to two non-acute imaging sites (Perspectum, Oxford and Mayo Clinic Healthcare, London) from April 2020 to August 2021 (the period of COVID-19 waves 1 and 2 in UK) and written informed consent was provided (see Supplementary Methods for inclusion and exclusion criteria). Those with evidence of organ impairment, based on bloods, MRI or incidental findings (organ abnormalities such as lesions, cysts, abnormal vessels and tumours incidentally detected during the MRI analysis), were invited for a 6-month follow-up (contacted on ≥2 occasions to minimise loss to follow-up). Each visit comprised MRI and blood investigations (full blood count, biochemistry) and online questionnaires completed beforehand. 9 ## Diagnostic assessment in non-acute settings Quantitative multi-organ MRI (COVERSCAN, Perspectum, Oxford) was used to assess organ impairment as previously reported (Figure S1), 9 using healthy controls (no prior COVID-19 diagnosis, no hospital discharge ≥4 months prior to enrolment and contraindications to MRI; 59 in Oxford, 33 in London). Participants underwent a 40-min MRI scan of the lungs, heart, kidney, liver, pancreas and spleen on 1.5-T or 3-T Siemens scanners at three imaging sites (Oxford: MAGNETOM Aera 1.5 T, Mayo Healthcare London: MAGNETOM Vida 3 T, Chenies Mews Imaging Centre London: MAGNETOM Prisma 3 T). MR metrics were standardised to deliver a single report interpretable by clinicians. Each report included 49 organ-specific metrics with reference ranges to determine impairment (updated from our prior study 10) after determining distribution of each metric in healthy controls matched for age and sex ($$n = 92$$) and for organ volumes from healthy controls representing complete sex and height subgroups ($$n = 1835$$) in this study and UK Biobank 13 (Tables S1a and S1b). Repeatability of the metrics was evaluated in the healthy controls using standardised performance testing criteria. Technical success was determined by reporting quality-assured measures for each variable reported herein, and overall, in delivering a report for each patient. ## Symptoms, function and organ impairment Assessment focused on commonly reported symptoms, HRQoL 9 and degree of breathlessness (Dyspnoea 12 scale 7). Participants were asked at follow-up about time off work due to COVID-19 (not done at baseline). Multi-organ impairment at baseline and follow-up was defined as ≥2 MRI measurements from different organs outside reference ranges. Further details are in Supplementary Methods. ## Statistical analysis The study was powered based on the primary outcome measure: to determine the prevalence of heart, kidney and liver injury in a cohort of patients recovering from COVID-19, using multi-parametric MRI, defined relative to pre-defined thresholds from a healthy cohort. Using a published method, 14 the required sample size for $10\%$ expected prevalence, 15 $5\%$ allowable margin of error and $95\%$ confidence and allowing for $10\%$ attrition rate, was 507. Analyses were conducted in R version 4.1.0, defining statistical significance by uncorrected p-value <0.05 (2-sided). Normally distributed-continuous variables are expressed as mean (standard deviation, SD); non-normally distributed-continuous are expressed as median (interquartile range, IQR); and categorical variables are expressed in frequency (percentage). Fisher’s exact or McNemar’s tests were used for unpaired and paired categorical data, and paired/unpaired t-test or Wilcoxon tests were used for paired/unpaired continuous data, depending on the normality of the underlying distribution. Stepwise multi-variable logistic regression was performed for associations with symptom groups, and a sensitivity analysis excluding metabolic syndrome was performed (Supplementary Methods). ## Study population for diagnostic assessment in non-acute settings A total of 536 individuals (mean age 45 years, $73\%$ female, median body mass index (BMI) 25 kg/m2, $13\%$ COVID-19 hospitalisation, $32\%$ healthcare workers) were included at baseline (Figure 1). Demographics were comparable to healthy controls (Table S1a). Most were ‘UK first wave’ (COVID-19 January–September 2020: $$n = 497$$) and 39 were ‘second wave’ (COVID-19 after September 2020). Over half of individuals ($$n = 296$$, $55\%$) had experienced acute COVID-19 confirmed by antibody or PCR tests and they had marginally higher BMI (25 kg/m2, $95\%$ confidence interval (CI) 22–28) compared to those with clinical diagnosis of COVID-19 (26 kg/m2, $95\%$ CI 23–30) (Table S2). Median time from initial COVID-19 symptoms to baseline assessment was 182 days (IQR: 132–222). **Figure 1.:** *Study population from recruitment to follow-up.* A total of 388 individuals ($72\%$) with organ impairment identified at baseline, from blood or MRI assessment, were eligible for follow-up, of which 331 ($62\%$ of baseline cohort) completed (the follow-up group). Median time from baseline to follow-up assessment was 196 days (IQR: 182–209). Age and BMI were higher in those who completed both visits, and more had been hospitalised for acute COVID-19, compared to those without organ impairment who were not eligible for follow-up (Table S3). In the follow-up group, demographics and risk factors showed no major differences between baseline and follow-up (Table 1). Technical success of MRI and integrated in-person assessment was $99.1\%$ and $98.3\%$ at baseline and follow-up assessments, respectively. Reports summarising findings were delivered to patients and primary care clinicians in all cases. **Table 1.** | Unnamed: 0 | Whole cohort | Follow-up group | Follow-up group.1 | Follow-up group.2 | | --- | --- | --- | --- | --- | | Characteristic | Baseline(n = 536) | Baseline(n = 331) | Follow-up(n = 331) | p value (baseline vs. follow-up visits) | | Age (years) | 45 (11) | 46 (11) | 47 (11) | – | | Sex (n females) | 389 (73%) | 241 (73%) | 241 (73%) | – | | BMI (kg/m2) | 25 (23, 29) | 26 (23, 30) | 26 (23,31) | 0.269 | | Ethnicity | | | | – | | White | 477 (89%) | 295 (89%) | 295 (89%) | | | Mixed | 21 (4%) | 12 (4%) | 3 (1%) | | | South Asian | 24 (4%) | 17 (5%) | 15 (5%) | | | Black | 13 (2%) | 6 (2%) | 5 (2%) | | | Healthcare worker | 172 (32%) | 112 (34%) | 112 (34%) | No change | | At least one COVID-19 vaccination | 10 (2%) | 5 (2%) | 197 (60%) | <0.001 | | Co-morbidities and risks | | | | | | Smoking | | | | | | Never | 349 (65%) | 218 (66%) | 222 (67%) | 0.221 | | Current | 14 (3%) | 7 (2%) | 7 (2%) | >0.999 | | Ex-smoker | 172 (32%) | 106 (32%) | 102 (31%) | 0.343 | | BMI | | | | | | ≥25 kg/m2 | 293 (55%) | 200 (60%) | 200 (61%) | >0.999 | | ≥30 kg/m2 | 120 (22%) | 91 (27%) | 92 (28%) | 0.789 | | Hypertension | 44 (8%) | 33 (10%) | 37 (11%) | 0.221 | | Diabetes | 10 (2%) | 7 (2%) | 9 (3%) | 0.683 | | Heart disease | 9 (2%) | 4 (1%) | 4 (1%) | No change | | Asthma | 101 (19%) | 62 (19%) | 56 (17%) | 0.114 | | Hospitalised during acute COVID-19 | 72 (13%) | 57 (17%) | 57 (17%) | – | | Time off work (days) | 56 (14, 180) | 58 (14, 150) | 125 (35, 296) | <0.001 | | 15 common symptoms | | | | | | Number reported (median (IQR)) | 10 (8, 11) | 10 (8, 11) | 3 (0, 5) | <0.001 | | None reported in history | 0 (0%) | 0 (0%) | 93 (28%) | <0.001 | | None reported in history/questionnaires | 0 (0%) | 0 (0%) | 60 (19%) | <0.001 | | Symptom groups | | | | | | Systemic | 245 (46%) | 159 (48%) | 2 (1%) | <0.001 | | Cardiopulmonary | 238 (44%) | 143 (43%) | 15 (5%) | <0.001 | | Severe breathlessness (dyspnoea 12 ≥ 10) | 187 (36%) | 120 (38%) | 93 (30%) | 0.016 | | Cognitive dysfunction | 268 (50%) | 160 (48%) | 127 (38%) | 0.005 | | Poor HRQoL | 281 (55%) | 181 (57%) | 138 (45%) | <0.001 | | Less common symptoms only | 66 (13%) | 37 (12%) | 108 (35%) | <0.001 | | Duration (days: median, (IQR)) | | | | | | Initial symptoms-to-assessment | 182 (132, 222) | 170 (126, 208) | 384 (350, 431) | – | | COVID-19 positive-to-assessment | 110 (53, 175) | 110 (53, 170) | 328 (265, 375) | – | | Organ impairment | | | | | | Liver | 151 (29%) | 119 (36%) | 106 (33%) | 0.153 | | Heart | 102 (19%) | 71 (22%) | 70 (21%) | >0.999 | | Kidney | 79 (15%) | 60 (18%) | 56 (17%) | 0.583 | | Pancreas | 100 (20%) | 80 (26%) | 56 (22%) | 0.201 | | Lungs | 12 (2%) | 7 (2%) | 5 (2%) | >0.999 | | Spleen | 43 (8%) | 33 (10%) | 29 (9%) | 0.453 | | ≥1 organ | 314 (59%) | 228 (69%) | 194 (59%) | <0.001 | | ≥2 organs | 122 (23%) | 97 (29%) | 88 (27%) | 0.336 | ## Symptoms over 1 year In the whole cohort, at baseline, all participants were symptomatic (Table 1, Figure 2(a)), including those not eligible for follow-up (Figure S2). Females and individuals with obesity were more likely to have ≥1 systemic symptoms, cardiopulmonary symptoms or poor HRQoL ($p \leq 0.001$, 0.006 and <0.001 for sex and $$p \leq 0.002$$, 0.012 and 0.004 for BMI) (Table S4). **Figure 2.:** *Proportion of individuals with long COVID and symptoms (a: at baseline in the whole cohort and in the follow-up groups; b: at baseline vs. at follow-up in the follow-up group) or impairment (c: in the whole cohort and in the follow-up groups vs. healthy controls; d: at baseline vs. at follow-up in the follow-up group). Significant differences (p < 0.05) are indicated with a star, numbers above columns indicate the sample size (n) for each group. Note for c: compared between healthy controls (grey) and at baseline for long COVID (blue).* In the follow-up group, at baseline and follow-up, participants reported a median of 10 (IQR: 8–11) symptoms and 3 (0–5) symptoms ($p \leq 0.001$), respectively. At baseline, all five symptom groups had similar prevalence (systemic: $48\%$, cardiopulmonary: $43\%$, severe breathlessness: $38\%$, cognitive dysfunction: $48\%$, poor HRQoL: $57\%$); $12\%$ reported none of these symptoms but other less common symptoms instead. At follow-up, symptoms were reduced, particularly systemic ($1\%$) and cardiopulmonary ($5\%$) ($p \leq 0.001$). Exceptions were fatigue, breathlessness and cognitive dysfunction, where prevalence remained high (Figure 2(b)). Some participants reported these symptoms only at follow-up. Most common symptoms improved by follow-up: fatigue ($98\%$ to $64\%$), myalgia ($89\%$ to $35\%$), shortness of breath ($90\%$ to $47\%$), headache ($85\%$ to $34\%$), chest pain ($81\%$ to $38\%$), fever ($73\%$ to $2\%$), cough ($75\%$ to $11\%$) and sore throat ($71\%$ to $11\%$). Of 331, 60 ($18\%$) had resolved all symptoms at follow-up (Table 2). **Table 2.** | Characteristic | Ongoing symptomatic (symptoms reported at baseline and follow-up)(n = 264) | Resolved symptoms (no symptoms reported at follow-up)(n = 60) | p-value | | --- | --- | --- | --- | | Follow-up group baseline characteristics | Follow-up group baseline characteristics | Follow-up group baseline characteristics | Follow-up group baseline characteristics | | Demographics | | | | | Age (years) | 46 (10) | 48 (12) | 0.340 | | Female sex (n) | 204 (77%) | 34 (57%) | 0.002 | | BMI (kg/m2) | 26 (23, 30) | 28 (23, 31) | 0.519 | | White | 239 (91%) | 49 (82%) | 0.066 | | Mixed | 3 (1%) | 0 (0%) | >0.999 | | South Asian | 8 (3%) | 7 (12%) | 0.010 | | Black | 3 (1%) | 2 (3%) | 0.232 | | Healthcare worker | 89 (34%) | 19 (32%) | 0.880 | | Co-morbidities and risks | | | | | Never smoked | 173 (66%) | 40 (67%) | >0.999 | | Current smoker | 4 (2%) | 2 (3%) | 0.308 | | Ex-smoker | 87 (33%) | 18 (30%) | 0.760 | | BMI>=25 kg/m2 | 159 (60%) | 38 (63%) | 0.770 | | BMI>=30 kg/m2 | 72 (27%) | 18 (30%) | 0.750 | | Hypertension | 25 (9%) | 6 (10%) | 0.813 | | Diabetes | 3 (1%) | 4 (7%) | 0.024 | | Asthma | 53 (20%) | 9 (15%) | 0.468 | | Pre-existing heart disease | 3 (1%) | 1 (2%) | 0.561 | | Hospitalised during acute COVID-19 | 46 (17%) | 10 (17%) | >0.999 | | Number of common symptoms reported | 10 (9, 11) | 9 (7, 10) | <0.001 | | MRI abnormality | | | | | Liver | | | | | cT1 or fat high | 95 (36%) | 23 (39%) | 0.766 | | cT1 (high) | 36 (14%) | 9 (15%) | 0.836 | | cT1 (ms) | 727 (681, 770) | 715 (680, 768) | 0.690 | | Fat (high) | 87 (33%) | 20 (33%) | >0.999 | | Fat (%) | 3 (2, 6) | 3 (1, 6) | 0.732 | | Volume (high) | 23 (9%) | 6 (10%) | 0.801 | | Volume (ml) | 1423 (1252, 1693) | 1463 (1277, 1653) | 0.650 | | Pancreas | | | | | srT1 or fat high | 64 (26%) | 14 (24%) | 0.868 | | srT1 high | 29 (12%) | 7 (12%) | >0.999 | | srT1 (ms) | 720 (686, 772) | 722 (684, 767) | 0.775 | | Fat (high) | 51 (20%) | 12 (20%) | >0.999 | | Fat (%) | 3 (2, 6) | 3 (3, 5) | 0.287 | | Kidney | | | | | Cortex T1 (high) | 44 (17%) | 14 (24%) | 0.193 | | Cortex T1 left 1.5T (ms) | 1084 (74) | 1094 (61) | 0.323 | | Cortex T1 right 1.5T (ms) | 1072 (70) | 1084 (68) | 0.260 | | Cortex T1 left 3T (ms) | 1427 (74) | 1399 (54) | 0.204 | | Cortex T1 right 3T (ms) | 1400 (85) | 1382 (70) | 0.509 | | Volume (high) | 21 (8%) | 7 (12%) | 0.318 | | Volume left (ms) | 146 (129, 165) | 158 (133, 179) | 0.044 | | Volume right (ms) | 146 (128, 166) | 150 (137, 172) | 0.185 | | Spleen | | | | | Splenomegaly | 28 (11%) | 5 (8%) | 0.813 | | Volume (ml) | 180 (141, 242) | 194 (166, 252) | 0.146 | | Lung | | | | | FAC (low) | 7 (3%) | 0 (0%) | 0.359 | | FAC (%) | 44 (36, 50) | 43 (35, 50) | 0.844 | | Heart | | | | | Myocardial Injury/T1 (high) | 24 (9%) | 11 (18%) | 0.062 | | Average T1 1.5T (ms) | 980 (965, 993) | 972 (949, 987) | 0.047 | | Average T1 3T (ms) | 1182 (31) | 1184 (37) | 0.903 | | LV EF (low) | 13 (5%) | 4 (7%) | 0.531 | | LV EF (%) | 60 (5) | 58 (5) | 0.029 | | RV EF (low) | 10 (4%) | 4 (7%) | 0.302 | | RV EF (%) | 59 (5) | 58 (6) | 0.108 | | LV EDV (high) | 2 (1%) | 0 (0%) | >0.999 | | LV EDV (ml) | 78 (71, 87) | 82 (75, 94) | 0.008 | | RV EDV (high) | 3 (1%) | 0 (0%) | >0.999 | | RV EDF (ml) | 75 (66, 84) | 80 (70, 93) | 0.005 | | Global longitudinal strain (high) | 11 (4%) | 0 (0%) | 0.230 | | Global longitudinal strain (%) | −14 (2) | −14 (2) | 0.474 | | Number of organs impaired on MRI | | | | | 0 | 84 (32%) | 17 (28%) | 0.646 | | ≥1 | 180 (68%) | 43 (72%) | 0.646 | | ≥2 | 75 (28%) | 20 (33%) | 0.438 | | ≥3 | 26 (10%) | 7 (12%) | 0.641 | | Follow-up characteristics | Follow-up characteristics | Follow-up characteristics | Follow-up characteristics | | Hospitalised between baseline and follow-up | 13 (5%) | 1 (2%) | 0.480 | | Time off work (days) | 172 (42, 300) | 54 (21, 120) | <0.001 | | MRI abnormality | | | | | Liver | | | | | cT1 or fat high | 88 (35%) | 17 (28%) | 0.447 | | cT1 (high) | 40 (16%) | 7 (12%) | 0.547 | | cT1 (ms) | 724 (693, 768) | 708 (676, 755) | 0.150 | | Fat (high) | 74 (29%) | 16 (27%) | 0.754 | | Fat (%) | 3 (2, 6) | 3 (2, 6) | 0.877 | | Volume (high) | 24 (9%) | 5 (8%) | >0.999 | | Volume (ml) | 1427 (1271, 1699) | 1474 (1316, 1675) | 0.429 | | Pancreas | | | | | srT1 or fat high | 43 (22%) | 10 (21%) | >0.999 | | cT1 high | 16 (8%) | 4 (8%) | >0.999 | | srT1 (ms) | 720 (684, 754) | 700 (673, 740) | 0.113 | | Fat (high) | 35 (18%) | 10 (20%) | 0.683 | | Fat (%) | 3 (2, 5) | 3 (2, 6) | 0.989 | | Kidney | | | | | Cortex T1 (high) | 45 (17%) | 9 (15%) | 0.848 | | Cortex T1 left 1.5T (ms) | 1086 (58) | 1086 (67) | 0.949 | | Cortex T1 right 1.5T (ms) | 1082 (53) | 1078 (67) | 0.702 | | Cortex T1 left 3T (ms) | 1422 (91) | 1424 (53) | 0.931 | | Cortex T1 right 3T (ms) | 1404 (85) | 1393 (59) | 0.605 | | Volume (high) | 21 (8%) | 6 (10%) | 0.610 | | Volume left (ms) | 145 (127, 165) | 158 (137, 180) | 0.011 | | Volume right (ms) | 149 (129, 167) | 156 (143, 172) | 0.028 | | Spleen | | | | | Splenomegaly | 24 (9%) | 5 (8%) | >0.999 | | Volume (ml) | 176 (139, 246) | 192 (161, 245) | 0.074 | | Lung | | | | | FAC (low) | 4 (2%) | 0 (0%) | 0.580 | | FAC (%) | 45 (38, 52) | 45 (38, 52) | 0.752 | | Heart | | | | | Myocardial Injury / T1 (high) | 25 (9%) | 7 (12%) | 0.632 | | Average T1 1.5T (ms) | 982 (969, 992) | 975 (960, 987) | 0.069 | | Average T1 3T (ms) | 1182 (38) | 1184 (48) | 0.895 | | LV EF (low) | 6 (2%) | 2 (3%) | 0.644 | | LV EF (%) | 60 (4) | 58 (5) | 0.013 | | RV EF (low) | 7 (3%) | 5 (8%) | 0.051 | | RV EF (%) | 59 (5) | 57 (5) | 0.014 | | LV EDV (high) | 2 (1%) | 0 (0%) | >0.999 | | LV EDF (ml) | 78 (70, 86) | 82 (75, 92) | 0.012 | | RV EDV (high) | 4 (2%) | 0 (0%) | >0.999 | | RV EDF (ml) | 74 (66, 86) | 81 (72, 91) | 0.015 | | Global longitudinal strain (high) | 15 (6%) | 5 (8%) | 0.551 | | Global longitudinal strain (%) | −15 (2) | −14 (2) | 0.017 | | Number of organs impaired on MRI | | | | | 0 | 112 (42%) | 23 (38%) | 0.664 | | ≥1 | 152 (58%) | 37 (62%) | 0.664 | | ≥2 | 73 (28%) | 13 (22%) | 0.419 | | ≥3 | 28 (11%) | 5 (8%) | 0.813 | ## Function over 1 year HRQoL was poor at baseline in the whole cohort of individuals with long COVID: median visual analog score (VAS) score $60\%$ (IQR: $40\%$–$70\%$) and median reported health utility index score 0.67 (IQR: 0.48–0.77). The most highly ranked sub-optimal health dimensions were problems completing usual activities of living and pain ($56\%$ and $45\%$, respectively, reporting moderate to extreme difficulties). These difficulties were also observed at baseline in the follow-up group (median VAS score $60\%$, IQR: $40\%$–$70\%$, and median reported health utility index score of 0.67, IQR: 0.48–0.77) (Figures S3 and S4). At follow-up, there was increased health utility index score to 0.71 (range 0.56–0.81) ($p \leq 0.001$), but $42\%$ still reported utility score <0.7, and $28\%$ still complained of severe breathlessness. Almost every individual at follow-up ($\frac{271}{302}$, $90\%$) had taken COVID-19-related time off work (median: 125 days, IQR: 35–296). At follow-up, $95\%$ of healthcare workers had taken time off work (median 180 days, range 41–308) and 62 ($63\%$) had taken >100 days off work. Those with ongoing symptoms reported taking more time off work (median 172 days, range 42–300) compared to those whose symptoms had resolved (median 54 days, range 21–120) (Table 2). ## Organ impairment over 1 year Most standard-of-care biochemical investigations were within normal range and not predictive of outcomes (Table S5), except lactate dehydrogenase ($\frac{59}{306}$ ($19\%$) and $\frac{70}{319}$ ($22\%$)), creatinine kinase ($\frac{26}{313}$ ($8\%$) and $\frac{41}{323}$ ($13\%$)); cholesterol ($\frac{152}{313}$ ($49\%$) and $\frac{157}{326}$ ($48\%$)); and mean cell haemoglobin concentration ($\frac{62}{313}$ ($20\%$) and $\frac{49}{324}$ ($15\%$); $$p \leq 0.05$$), which were elevated at both baseline and follow-up. On MRI of 536 participants at baseline, $59\%$ and $23\%$ had impairment in ≥1 and ≥2 organs, respectively (Figure 2(c)), although impairment was usually mild (Table S6), e.g. among participants with cardiac impairment, none had severe heart failure. Liver steatosis, kidney fibro-inflammation and splenomegaly at baseline were more frequent in all symptom groups (Table S4). Liver steatosis was associated with systemic symptoms and severe breathlessness (Figure S5). In the follow-up group, single-organ impairment between visits improved but remained high ($69\%$ at baseline and $59\%$ at follow-up had impairment in ≥1 organ; $p \leq 0.001$) without improvement in individual metrics between visits (Figure 2D). Multi-organ impairment did not improve ($29\%$ to $27\%$; $$p \leq 0.336$$). Individuals without organ impairment had lower symptom burden compared to those with at least one organ impairment (Figure S6). At baseline, lung impairment (lower fractional area change) and impairment in ≥3 organs had the highest symptom burden. Healthcare workers were more likely to have liver impairment ($$p \leq 0.015$$ at baseline and $$p \leq 0.034$$ at follow-up) than the rest of the cohort. In individuals without symptoms at follow-up ($$n = 60$$), organ impairment was present in 43 ($72\%$) at baseline and 37 ($62\%$) at follow-up (Table 2, Figure S6). In symptomatic participants at follow-up ($$n = 264$$), there were 84 ($32\%$) and 112 ($42\%$) individuals without organ impairment at baseline and follow-up, respectively. ## Associations between symptoms and organ impairment Looking at individual symptoms and by five symptom groups, neither abnormal biochemical investigations nor organ impairment were predictive of full symptom resolution at follow-up (Tables S4–S7). Several liver-specific parameters were associated with specific symptom burden (Figure 3, Table S7). High liver fat was present in $\frac{58}{187}$ with severe breathlessness but only $\frac{70}{328}$ without severe breathlessness at baseline. Conversely, low liver fat was more likely in those without severe breathlessness, at both timepoints, in the follow-up group. High liver volume at follow-up was associated with lower HRQoL. Hepatomegaly was present in $\frac{20}{138}$ with poor QoL but in only $\frac{8}{167}$ with better QoL at follow-up. **Figure 3.:** *Association between risk factors and: (a) severe breathlessness at baseline; (b) poor HRQoL at follow-up. Note: (a) associations predicted whole cohort symptom at baseline using baseline covariates and (b) predicted follow-up symptom using follow-up covariates in the follow-up group.* ## Discussion In this UK prospective study of largely non-hospitalised long COVID, we have four new findings. First, we confirm multi-organ impairment at 6 and 12 months in $29\%$ of individuals with long COVID, with persistent symptoms and reduced function. Second, despite some associations between organ impairment and symptoms, there is currently insufficient evidence for distinct long COVID subtypes. Blood biomarkers, the current standard of care, showed no relation to clinical presentation. Third, symptoms, blood investigations and quantitative, multi-organ MRI did not predict trajectory or recovery. Fourth, we demonstrate feasibility of scalable, multi-organ assessment in non-acute settings in the pandemic context. Several studies confirm persistence of symptoms in individuals with long COVID up to 1 year. 16 We now add that three in five people with long COVID have impairment in at least one organ, and one in four have impairment in two or more organs, in some cases without symptoms. Impact on quality of life and time off work, particularly in healthcare workers, is a major concern for individuals, health systems and economies. 17 Many healthcare workers had no prior illness ($2\%$ diabetes, $2\%$ heart disease and $22\%$ asthma, which may play a pathophysiologic role 18) but of 172 such participants, 19 were still symptomatic at follow-up and off work for a median of 180 days. We need comparison with similar analyses from other long COVID studies, and other long-term conditions, 19 with considerable workforce planning implications. There have been heterogenous methods to investigate long COVID, whether qualitative20,21 or quantitative, 5 or symptom surveys 6 versus cohort studies.4,22 Most research still focuses on individual organs.23–25 The scale of long COVID burden necessitates action to develop, evaluate and implement evidence-based investigation, treatment and rehabilitation 26 (e.g. STIMULATE-ICP and other studies 4). Metabolic diseases, including non-alcoholic liver disease and diabetes, are postulated to play a role in the pathogenesis of severe COVID-19 and possibly long COVID. 27 We observe associations with markers of liver dysfunction, and increased risk of symptoms in female and obese individuals with long COVID, which have been shown previously.28,29 Cognitive impairment appears to develop in the natural history of long COVID in some patients. However, the underlying mechanisms of the condition or syndrome remain elusive. We did not find evidence by symptoms, blood investigations or MRI to clearly define long COVID subtypes. We observed symptom resolution at 12 months in $29\%$ of individuals with long COVID, particularly with cardiopulmonary and systemic symptoms, aligning with other longitudinal studies.5,13,19 Our findings of reductions in HRQoL and time at work reinforce prior research and are concerning, despite improvement over time. Future research must consider associations between symptoms, multi-organ impairment and function in larger cohorts, enabling clearer stratification and evaluation of treatments. The COVERSCAN study was initiated in the early COVID-19 pandemic’s first wave when face-to-face assessment and investigation, and reduced health system capacity were major concerns for patients and health professionals. In the UK and other countries, long COVID carries high burden of investigations and healthcare utilisation across specialties, and definitive care pathways are lacking. We show feasibility, acceptability and scalability of a rapid (40-min), multi-organ MRI protocol for practice and research. Alongside routine clinical assessment and blood tests, COVERSCAN can exclude organ impairment in integrated, multidisciplinary care pathways. 30 Such MRI assessment has potential application beyond the pandemic for multi-system assessment and investigation, including in lower resource settings. ## Implications for research There are three research implications. First, complex intervention trials, including care pathways from investigation to rehabilitation are needed to evaluate therapies. Second, associations between symptoms, symptom groups, blood investigations and MRI must be investigated in larger populations. Third, long COVID pathophysiology is still unclear, and should ideally be studied at the same time as clinical trials. ## Implications for clinical practice and public health There are three practice and policy implications. First, COVERSCAN could be used to rule out organ impairment and to identify subgroups requiring specialist referral. Second, long COVID is a multi-organ condition, needing multi-organ assessment and multidisciplinary care. Third, poor 1-year post-COVID recovery rates highlight the need for rehabilitation and integrated care, relevant to other long-term conditions. ## Strengths and limitations To date, this is the largest, comprehensive, systematic, multi-organ, post-COVID study over 1 year. The study cohort is representative of long COVID resulting from the first and second waves in the UK in a mainly community setting in terms of risk factors and acute COVID-19 hospitalisation.7,8 There are limitations. We acknowledge that the study may not be representative of the condition in patients infected with other variants of SARS-CoV-2, or of long COVID in mainly hospitalised patients. Our study recruited patients who self-referred from online patient support forums and social media rather than a systematic screen of post-COVID patients, as long COVID clinics were not yet established. We did not have history and imaging prior to the pandemic and so it is difficult to determine if COVID-19/long COVID caused impairment. Despite low prevalence of self-reported co-morbidities in this cohort, in some participants the organ impairment we identified may be due to pre-existing, undiagnosed co-morbidities rather than direct post-COVID sequelae. There was no assessment of brain function. Not all participants had a laboratory-confirmed COVID-19 diagnosis. Generalisability of results from the UK’s first COVID-wave to the global population requires further exploration. We successfully delivered multi-organ MRI to assess organ health, but clinical utility remains to be determined. Health utilisation (e.g. primary care interactions) and economic burden of long COVID were not assessed. Time off work was not assessed at baseline when the UK was in lockdown with a national furlough scheme in operation. Patients with normal assessment at baseline were not followed up to minimise in-person contact and due to resource and funding constraints. Therefore, while $\frac{222}{536}$ ($41\%$) patients at baseline were defined as normal after assessment, we cannot say that they had better outcomes, and this may represent a selection bias. Nevertheless, we can model cost-saving of a single assessment versus multiple assessments to achieve a discharge decision. ## Conclusions Long COVID symptoms commonly persist at 12 months, even in those not severely affected by acute COVID-19. Diagnosis and follow-up of long COVID can be performed in non-acute settings. Continued research in multi-system assessment and pharmacotherapy for those reporting ongoing fatigue, breathlessness and cognitive problems is required to address long COVID burden, in parallel with mechanistic studies to understand pathophysiology. ## Key points Question: *What is* the prevalence of organ impairment in long COVID at 6 and 12 months post COVID-19?Findings: *In a* prospective study of 536 mainly non-hospitalised individuals, all were symptomatic at 6 months and $59\%$ had single-organ impairment. Although symptom burden decreased, organ impairment persisted in the 331 followed up at 12 months post COVID-19.Meaning: Organ impairment in long COVID has implications for symptoms, quality of life and longer-term health, signalling the need for prevention and integrated care of long COVID. ## Competing Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: AD, NE, SF, MP, AR-F, HT-B, MK, MR and RB are employees of Perspectum. All other authors have no competing or conflicting interests. ## Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The study was partly financed by an Amendment to a European Commission’s Horizon 2020 grant 719445 (AMENDMENT Reference No. AMD-719445-8): Non-invasive rapid assessment of chronic liver disease using Magnetic Resonance Imaging with LiverMultiScan (RADIcAL). AB has received funding from NIHR (including the STIMULATE-ICP study), AstraZeneca, European Union, UK Research and Innovation and British Medical Association. DW is funded by an NIHR advanced fellowship. MG is funded by ARC and DH&SC NIHR. ## Ethics approval The protocol was approved by South Central – Berkshire B Research Ethics Committee (20/SC/0185) and registered (https://clinicaltrials.gov/ct2/show/NCT04369807). ## Guarantor AB. ## Contributorship Study design: AD, RB, COVERSCAN team. Patient recruitment: RB, COVERSCAN team. Data collection: COVERSCAN team. Data analysis: AD, NE, HT-B, AR-F, AB, SM, COVERSCAN team. Data interpretation: AB, AD, NE, HT-B, RB. Initial article drafting: AB, AD, RB. Critical review of early and final versions of the article: all authors. Specialist input: AB (cardiology); RB, MH, DW, MC, DJC (general medicine); MH, MC, DW (long COVID); MB, RB, MR (imaging); AD (statistics); AB (epidemiology/public health); MG (primary care); LH, EA (patient and public involvement). ## Acknowledgements We are grateful to Alison Telford for statistical critique of this article. ## Provenance Not commissioned. ## Consent to participate All participants gave full informed consent. ## Availability of data and materials All relevant data to the study are included in the article and the supplementary material. ## ORCID iD Amitava Banerjee https://orcid.org/0000-0001-8741-3411 ## Supplemental Material Supplemental material for this article is available online. ## References 1. Groff D, Sun A, Ssentongo AE, Ba DM, Parsons N, Poudel GR. **Short-term and long-term rates of postacute sequelae of SARS-CoV-2 infection: a systematic review**. *JAMA Netw Open* (2021.0) **4** e2128568. PMID: 34643720 2. Soriano JB, Murthy S, Marshall JC, Relan P, Diaz JV.. **A clinical case definition of post-COVID-19 condition by a Delphi consensus**. *Lancet Infect Dis* (2022.0) **22** e102-e107. PMID: 34951953 3. Yong SJ, Liu S.. **Proposed subtypes of post-COVID-19 syndrome (or long-COVID) and their respective potential therapies**. *Rev Med Virol* (2022.0) **32** e2315. PMID: 34888989 4. Routen A, O’Mahoney L, Ayoubkhani D, Banerjee A, Brightling C, Calvert M. **Understanding and tracking the impact of long COVID in the United Kingdom**. *Nat Med* (2022.0) **28** 11-15. PMID: 34811549 5. Al-Aly Z, Xie Y, Bowe B.. **High-dimensional characterization of post-acute sequelae of COVID-19**. *Nature* (2021.0) **594** 259-264. PMID: 33887749 6. Davis HE, Assaf GS, McCorkell L, Wei H, Low RJ, Re’em Y. **Characterizing long COVID in an international cohort: 7 months of symptoms and their impact**. *eClinicalMedicine* (2021.0) **38** 101019. PMID: 34308300 7. Heightman M, Prashar J, Hillman TE, Marks M, Livingston R, Ridsdale HA. **Post-COVID-19 assessment in a specialist clinical service: a 12-month, single-centre, prospective study in 1325 individuals**. *BMJ Open Resp Res* (2021.0) **8** e001041 8. Taquet M, Dercon Q, Luciano S, Geddes JR, Husain M, Harrison PJ.. **Incidence, co-occurrence, and evolution of long-COVID features: a 6-month retrospective cohort study of 273,618 survivors of COVID-19**. *PLoS Med* (2021.0) **18** e1003773. PMID: 34582441 9. Dennis A, Wamil M, Alberts J, Oben J, Cuthberston DJ, Wootton D. **Multiorgan impairment in low-risk individuals with post-COVID-19 syndrome: a prospective, community-based study**. *BMJ Open* (2021.0) **11** e048391 10. Borlotti A, Thomaides-Brears H, Georgiopoulos G, Banerjee R, Robson MD, Fusco DN. **The additive value of cardiovascular magnetic resonance in convalescent COVID-19 patients**. *Front Cardiovasc Med* (2022.0) **9** 854750. PMID: 35463767 11. Heber SD, Hetterich H, Lorbeer R, Bayerl C, Machann J, Auweter S. **Pancreatic fat content by magnetic resonance imaging in subjects with prediabetes, diabetes, and controls from a general population without cardiovascular disease**. *PLoS One* (2017.0) **12** e0177154. PMID: 28520813 12. Augustin M, Schommers P, Stecher M, Dewald F, Gieselmann L, Gruell H. **Post-COVID syndrome in non-hospitalised patients with COVID-19: a longitudinal prospective cohort study**. *Lancet Reg Health Eur* (2021.0) **6** 100122. PMID: 34027514 13. Wilman HR, Kelly M, Garratt S, Matthews PM, Milanesi M, Herlihy A. **Characterisation of liver fat in the UK biobank cohort**. *PLoS One* (2017.0) **12** e0172921. PMID: 28241076 14. Naing L, Winn T, Rusli BN.. **Practical issues in calculating the sample size for prevalence studies**. *Arch Orofac Sci* (2006.0) **1** 9-14 15. Chen T, Wu D, Chen H, Yan W, Yang D, Chen G. **Clinical characteristics of 113 deceased patients with coronavirus disease 2019: retrospective study**. *BMJ* (2020.0) **368** m1091. PMID: 32217556 16. Han Q, Zheng B, Daines L, Shiekh A.. **Long-term sequelae of COVID-19: a systematic review and meta-analysis of one-year follow-up studies on Post-COVID symptoms**. *Pathogens* (2022.0) **11** 269. PMID: 35215212 17. Martin C, Luteijn M, Letton W, Robertson J, McDonald S.. **A model framework for projecting the prevalence and impact of long-COVID in the UK**. *PLoS One* (2021.0) **16** e0260843. PMID: 34855874 18. Cervia C, Zurbuchen Y, Taeschler P.. **Immunoglobulin signature predicts risk of post-acute COVID-19 syndrome**. *Nat Commun* (2022.0) **13** 446. PMID: 35078982 19. Gabbay M, Shiels C, Hillage J.. **Factors associated with the length of fit note-certified sickness episodes in the UK**. *Occup Environ Med* (2015.0) **72** 467-475. PMID: 25713158 20. Shah R, Ali FM, Nixon SJ, Ingram JR, Salek SM, Finlay AY.. **Measuring the impact of COVID-19 on the quality of life of the survivors, partners and family members: a cross-sectional international online survey**. *BMJ Open* (2021.0) **11** e047680 21. Macpherson K, Cooper K, Harbour J, Mahal D, Miller C, Nairn M.. **Experiences of living with long COVID and of accessing healthcare services: a qualitative systematic review**. *BMJ Open* (2022.0) **12** e050979 22. Tleyjeh IM, Saddik B, Ramakrishnan RK, AlSwaidan N, AlAnazi A, Alhazmi D. **Long term predictors of breathlessness, exercise intolerance, chronic fatigue and well-being in hospitalized patients with COVID-19: a cohort study with 4 months median follow-up**. *J Infect Public Health* (2022.0) **15** 21-28. PMID: 34861604 23. Badenoch JB, Rengasamy ER, Watson C, Jansen K, Chakraborty S, Sundaram RD. **Persistent neuropsychiatric symptoms after COVID-19: a systematic review and meta-analysis**. *Brain Commun* (2022.0) **4** fcab297. PMID: 35169700 24. Yan Z, Yang M, Lai CL.. **Long covid-19 syndrome: a comprehensive review of its effect on various organ systems and recommendation on rehabilitation plans**. *Biomedicines* (2021.0) **9** 966. PMID: 34440170 25. 25Xie Y, Xu E, Bowe B, Al-Aly Z. Long-term cardiovascular outcomes of COVID-19. Nat Med 2022; 28: 583–590. 26. Murray E, Goodfellow H, Bindman J, Blandford A, Bradbury K, Chaudhry T. **Development, deployment and evaluation of digitally enabled, remote, supported rehabilitation for people with long COVID-19 (living with COVID-19 recovery): protocol for a mixed-methods study**. *BMJ Open* (2022.0) **12** e057408 27. Khunti K, Davies MJ, Kosiborod MN, Nauck MA.. **Long COVID – metabolic risk factors and novel therapeutic management**. *Nat Rev Endocrinol* (2021.0) **17** 379-380. PMID: 33875855 28. Ayoubkhani D, Khunti K, Nafilyan V, Maddox T, Humberstone B, Diamond I. **Post-covid syndrome in individuals admitted to hospital with covid-19: retrospective cohort study**. *BMJ* (2021.0) **372** n693. PMID: 33789877 29. Sudre CH, Murray B, Varsavsky T, Graham MS, Penfold RS, Bowyer RC. **Attributes and predictors of long COVID**. *Nat Med* (2021.0) **27** 626-631. PMID: 33692530 30. Parker AM, Brigham E, Connolly B, McPeake J, Agranovich AV, Kenes MT. **Addressing the post-acute sequelae of SARS-CoV-2 infection: a multidisciplinary model of care**. *Lancet Respir Med* (2021.0) **9** 1328-1341. PMID: 34678213
--- title: Mental stress induces endothelial dysfunction by AT1R-mediated redox imbalance in overweight/obese men authors: - H.N.M. Rocha - G.M.S. Batista - A.S. Storch - V.P. Garcia - G.F. Teixeira - J. Mentzinger - E.A.C. Gomes - M.O. Campos - A.C.L. Nóbrega - N.G. Rocha journal: Brazilian Journal of Medical and Biological Research year: 2023 pmcid: PMC10041671 doi: 10.1590/1414-431X2023e12547 license: CC BY 4.0 --- # Mental stress induces endothelial dysfunction by AT1R-mediated redox imbalance in overweight/obese men ## Abstract The main goal of this study was to determine whether oxidative imbalance mediated by AT1 receptor (AT1R) is responsible for deleterious endothelial responses to mental stress (MS) in overweight/obese class I men. Fifteen overweight/obese men (27±7 years old; 29.8±2.6 kg/m2) participated in three randomized experimental sessions with oral administration of the AT1R blocker olmesartan (40 mg; AT1R blockade) or ascorbic acid (AA; 3g) infusion or placebo [both intravenously ($0.9\%$ NaCl) and orally]. After two hours, endothelial function was determined by flow-mediated dilation (FMD) before (baseline), 30 min (30MS), and 60 min (60MS) after a five-minute acute MS session (Stroop Color Word Test). Blood was collected before (baseline), during MS, and 60 min after MS for redox homeostasis profiling: lipid peroxidation (TBARS; thiobarbituric acid reactive species), protein carbonylation, and catalase activity by colorimetry and superoxide dismutase (SOD) activity by an ELISA kit. At the placebo session, FMD significantly decreased 30MS ($$P \leq 0.05$$). When compared to baseline, TBARS ($P \leq 0.02$), protein carbonylation ($P \leq 0.01$), catalase ($P \leq 0.01$), and SOD ($P \leq 0.01$) increased during the placebo session. During AT1R blockade, FMD increased 30 min after MS ($$P \leq 0.01$$ vs baseline; $P \leq 0.01$ vs placebo), while AA infusion increased FMD only 60 min after MS. No differences were observed during MS with the AT1R blockade and AA regarding TBARS, protein carbonylation, catalase, and SOD. AT1R-mediated redox imbalances played an important role in endothelial dysfunction to mental stress. ## Introduction Overweight/obesity is considered a major risk factor for noncommunicable diseases such as cardiovascular diseases, diabetes, and some types of cancer, contributing substantially to worldwide mortality [1]. According to the World Health Organization, $39\%$ of adults aged 18 years or over are overweight, while $13\%$ have been diagnosed with obesity [2]. In addition, psychological or mental stress (MS) is also an important risk factor for the development and progression of cardiovascular diseases [3], increasing the risk of acute coronary syndrome by $30\%$ and the risk of stroke by $24\%$ in men [4]. Acute MS seems to lead to a transitory endothelial dysfunction both in health and disease [5,6] together with an impaired endothelial repair mechanism [7]. Chronically, MS effects can be permanent for endothelial function, especially in individuals that already present risk factors such as overweight/obesity [8]. The mechanism by which MS leads to endothelial dysfunction in humans is still unknown. The renin-angiotensin system (RAS) activated by stress-mediated sympathoexcitation plays a central role in endothelial homeostasis [9]. Angiotensin II (Ang II) is the main active mediator of the RAS classic pathway, which acts on endothelial and smooth muscle cells, through the angiotensin type 1 (AT1R) and type 2 receptors [10]. In vivo and in vitro experimental studies have demonstrated that increases in Ang II - AT1R signaling leads to an imbalance of vasoactive substances, downregulating endothelial nitric oxide synthase, an enzyme that synthesizes nitric oxide (NO), and vasoconstrictor tone predominates [11]. Also, Ang II - AT1R pathway seems to activate NADPH oxidase, increasing reactive oxygen species (ROS) and inflammation in overweight/obese adults [9,12]. However, it is not clear whether AT1R-mediated oxidative stress is the underlying mechanism of endothelial responses to MS. It has been demonstrated that both adults and children with obesity present diminished flow-mediated dilation (FMD), a proxy of endothelial function, at resting conditions [13- 16]. Sales et al. [ 6] highlighted that MS evokes acute transient reductions in FMD in obese adults with metabolic syndrome. It is worth noting that chronic stress in obese adults seems to double the cardiovascular morbidity and mortality compared to healthy individuals [17,18]. Thus, it is critical to elucidate the impact of the Ang II-AT1R pathway on deleterious endothelial responses to MS in adults at increased cardiovascular risk, such as those with overweight/obesity. Considering that Ang II modulates MS responses, it is believed that imbalances in redox homeostasis mediated by AT1R may be the underlying mechanism related to impaired stress-induced endothelial dysfunction in overweight/obese adults. Also, we hypothesized that AT1R blockade and ascorbic acid - free radical scavenger (antioxidant) - would similarly restore endothelial function in response to MS. ## Study population and protocol Fifteen non-hypertensive overweight/obesity grade I men (27±7 years) were recruited from the local community. All individuals presented a BMI between 25 and 35 kg/m2 and body fat mass higher than $25\%$. Inclusion criteria included absence of any diagnosed disease, non-smoker status, and sedentary lifestyle (<150 min per week of moderate intensity cardiorespiratory exercise training) [19]. All data collection took place at the Laboratory of Exercise Sciences (Niteroi, Brazil) during August 2017 and October 2018. This study protocol was approved by the Ethics Committee of Fluminense Federal University (CAAE 76594217.0.0000.5243) and by the Brazilian Clinical Trials Registry (Rio de Janeiro, RJ; REQ: 9237; www.ensaiosclinicos.gov.br) and conformed to the standards set by the latest revision of the Declaration of Helsinki. All subjects gave written informed consent before their participation in the study. Biochemical analyses were conducted at the first visit. The subjects were then invited to a second visit that consisted of eligibility screening, i.e., clinical history assessment, anthropometric and arterial pressure measurements, resting electrocardiogram, and biochemical blood analysis interpretation. When the inclusion criteria were met, subjects were invited to three experimental sessions, with at least seven days between them. The protocol consisted of a randomized, 3-way crossover, blind, placebo-controlled study. Experimental sessions consisted of oral administration of angiotensin II type 1 receptor blocker (AT1R blockade; 40 mg, olmesartan (OLM), lot number: 60818, Pfizer, USA), ascorbic acid (AA; 3 g diluted in 500 mL of $0.9\%$ NaCl) administered intravenously for 30 min, or placebo [both intravenously ($0.9\%$ NaCl) and orally]. In the AT1R blockade session, an olmesartan pill was offered and saline infusion was performed; in the AA session, a placebo pill was offered and AA infusion was performed; in the placebo session, a placebo pill was offered and saline infusion was performed. All subjects were instructed to avoid alcohol, caffeine, and intense exercise in the 48 h prior to the visits. In addition, subjects were advised to follow a low-nitrate, low-nitrite diet prescribed by a nutritionist in the 24 h prior to the sessions. Mainly, subjects were advised to avoid red meat, fish, dark green vegetables, citrus fruits, oilseeds, and highly processed food. The experimental sessions took place in the morning in a climate-controlled environment (22-24°C). After the subjects arrived, blood pressure was measured in the seated position, and the subject was instructed to lie on the stretcher for drug administration. At this time, an intravenous catheter was placed in the antecubital cavity for blood sampling to evaluate endothelial biomarkers and oxidative stress, and a blind oral administration of AT1R blocker or an intravenous administration of AA or placebo was performed. All participants were submitted to the three conditions, with an interval of at least seven days between them. Subjects then rested supine for two hours, the time required for the blocker to reach the peak of action [20]. Following the resting period, brachial artery FMD was assessed in the dominant arm (baseline). Subsequently, subjects were submitted to a 5-min MS task. FMD was conducted again 30 (30MS) and 60 min (60MS) after MS. Venous blood samples were also collected before (baseline), during (MS), and 60 min (60MS) after MS (Figure 1). Immediately after sampling, each blood tube was centrifuged according to the specific requirements of each variable, and the plasma was aliquoted and snap-frozen. At the time of analyses, the aliquots were thawed at room temperature and discarded after use. **Figure 1:** *Experimental protocol. Asterisks indicate blood sampling. FMD: flow-mediated dilation; PL: placebo; AT1RB: angiotensin II type 1 receptor blockade; AA: ascorbic acid.* ## Biochemical analyses Blood samples were drawn from the antecubital vein after 12-h fasting for the following measurements: fasting glucose, total cholesterol, high-density lipoprotein (HDL)-cholesterol, triglycerides, and insulin using enzymatic colorimetric methods. Very low-density lipoprotein (VLDL)-cholesterol values were calculated based on triglyceride values, and low-density lipoprotein (LDL)-cholesterol was calculated by the Friedewald equation, which is based on total cholesterol, HDL-cholesterol, and triglyceride values. ## Bioimpedance Body composition by bioelectrical impedance analysis predicts the percentage of lean mass, fat mass, and total water volume (extracellular and intracellular) through an electrical current generated and detected by electrodes. Two electrodes each were positioned in the metacarpal and metatarsal, discharging an electric current of 50 kHz generated by an external source (Quantum II - Body Composition Analyzer; RJL Systems, USA) [21]. This current was detected by two other electrodes positioned in the wrist and ankle, evaluating the change in initial frequency. The impedance and reactance data provided by the source were analyzed using the RJL Systems Body Composition software. ## Mental stress The MS task applied was an adapted version of the Stroop Color Word Test [22], which consists of a slideshow projected on the ceiling above the subject that changes every two seconds. In addition, auditory conflicts were continuously inflicted via headphones using a standardized audio clip of three different people (two men and one woman) saying names of colors. The colors mentioned in the audio were in a different order then those presented in the slideshow. MS tasks consisted of two minutes of baseline measurements, five minutes of MS, and three minutes of recovery after the test, during which the subject rested quietly. Non-invasive beat-by-beat blood pressure and heart rate were recorded via photoplethysmography on the middle finger (Finometer, Finapres Medical Systems, The Netherlands). The level of perceived stress was assessed after each test using a subjective scale from zero to four, as follows: 0=non-stressful, 1=not very stressful, 2=stressful, 3=very stressful, and 4=extremely stressful. Blood flow measurements were performed at baseline, in the last 30 s of the first three minutes of MS, and in the last minute of recovery. Blood sampling for evaluation of endothelial biomarkers and oxidative stress was performed in the last two minutes of the MS task (Figure 1). ## Flow-mediated dilation Brachial artery FMD was measured on the dominant arm before and 30 and 60 min after the MS task. Of all fifteen subjects, FMD was performed on the left arm of only two. Subjects adopted the supine position with the shoulder abducted at 80°. The forearm position was determined and held to optimize brachial artery imaging. In accordance with the most recent FMD guidelines [21], a rapid inflation/deflation pneumatic cuff (E-20 Rapid Cuff Inflator, D.E. Hokanson, USA) of appropriate size was placed around the forearm immediately distal to the olecranon process. Brachial artery imaging was obtained on the distal third of the arm (2-12 cm above the antecubital fossa) using a multifrequency linear-array (8-12 MHz) probe coupled to a high-resolution Doppler ultrasound system (LogiQ P5, GE Medical Systems, USA). Diameter and blood velocity were simultaneously acquired in duplex mode at a pulsed frequency of 30 MHz and adjusted to the full vessel width (insonation angle ≤60°). Baseline diameter and mean blood velocity waveforms were continuously recorded for 30 s. The cuff was then rapidly inflated to 220 mmHg for five minutes. After this period, the cuff was rapidly deflated. Doppler recordings resumed 15 s before cuff deflation and continued for three minutes. Brachial artery diameter was analyzed offline with an automated edge-detection and wall-tracking software (Vascular Research Tools 5, Medical Imaging Applications, USA). Regions of interest were identified and kept for the remaining analyses [6,23]. ## Oxidative stress Oxidative stress was determined by the measurement of lipid peroxidation markers (thiobarbituric acid reactive species, TBARS), protein oxidation (protein carbonylation) concentrations, and the activity of catalase and superoxide dismutase (SOD) (antioxidant enzymes), in plasma isolated from venous blood samples collected in EDTA tubes, which were centrifuged at 1050 g for 15 min at 20°C. ## Lipid peroxidation The evaluation of lipid peroxidation was performed by determining the levels of TBARS. This method is based on the reaction between two molecules of thiobarbituric acid (TBA) and one of malondialdehyde (MDA) resulting from lipid peroxidation and producing a complex (MDA:TBA) of pink color. To this end, 100 μL of serum was homogenized with 50 μL of SDS ($8.1\%$), 550 μL phosphoric acid ($1\%$), and 300 μL of thiobarbituric acid ($0.6\%$). This solution was then heated to 95°C for 1 h in a dry bath and then centrifuged (2000 g) for 5 min at 25°C. The supernatant was used to quantify the TBARS levels. Plasma concentrations of lipoperoxides are reported in terms of MDA (nmol/mL) and determined in duplicate by TBARS measurement using a fluorimetric method (CV: $10.58\%$). The absorbance of each test was obtained in a 96-well microplate reader (Synergy H1 Hybrid Multi-mode, Biotek; USA) at 532 nm. This method used the substance 1,1,3,3-tetramethoxypropane to make the standard curve [24]. ## Protein carbonylation The quantification of protein carbonylation was accomplished through the reaction of 2,4-dinitrophenylhydrazine (DNPH) with the carbonyls of oxidized proteins. In this assay, the total protein concentration was determined in duplicate (CV: $22.02\%$) according to the method of Lowry et al. [ 25] using a standard curve of albumin. The carbonyl concentration values were normalized by mg of albumin and are reported as nmol/g. ## Catalase activity The catalase enzymatic activity was determined in duplicate (CV: $17.03\%$) by colorimetric assay (Catalase Assay kit, USA), using plasma isolated from venous blood samples collected in EDTA tubes, according to manufacturer's instructions. ## Superoxide dismutase activity SOD activity was determined by an enzyme-linked immunosorbent assay (ELISA) kit (Human SOD2/Mn-SOD DuoSet ELISA Kit, R&D Systems, USA), using plasma isolated from venous blood samples collected in EDTA tubes, according to manufacturer's instructions. ## Calculations and statistical analysis After analyzing brachial artery diameter and blood velocity, blood flow was calculated from the mean blood velocity and vessel area, considering 60 as a constant (i.e., Vmean × Area × 60). Shear rate (SR), a proxy of shear stress, was calculated as four times the ratio between mean blood velocity (Vmean) and the artery diameter [i.e., 4 × (Vmean/diameter)]. The area under the curve (AUC) was obtained from the cumulative SR during FMD from post-occlusion until peak diameter. Vascular conductance was calculated from mean blood flow and mean arterial pressure (mL·min·mmHg-1). Because baseline diameter (Dbase) could bias the FMD%, which is a ratio between peak diameter (Dpeak) and Dbase, the allometric scale proposed by Atkinson and Batterham [26] was used to account for possible baseline interferences. The regression's slope between logarithmically transformed Dpeak and Dbase was calculated [placebo, β=0.97 ($95\%$CI: 0.897 to 1.047); AT1R blockade, β=0.96 ($95\%$CI: 0.879 to 1.048); AA, β=0.96]. A regression slope smaller than one suggests that Dpeak and Dbase do not increase proportionally, meaning that the assumptions made based on FMD% might be biased. Then, logDbase, logDpeak, and the difference between them (Dpeak-Dbase) (logDdiff) were entered into a multivariate general linear model considering logDdiff as the dependent variable, session (placebo or AT1R blockade) as the fixed factor, and logDbase as the covariate. Adjusted means were then antilog-transformed, subtracted by a value of 1 and multiplied by 100 to facilitate interpretation as percentage. Considering the FMD results as the main outcome and the alpha error of 0.05, the power of the statistical test for a sample size of 14 individuals was 0.8. The Shapiro-Wilk test and homoscedasticity were performed by the Levene's test to verify the normal distribution of the variables. Two-way ANOVA was then performed for repeated measurements, where “condition” and “moment” were considered as the main factors. When significant differences were found for group, time, and/or interaction, Fisher's test was used as a post hoc procedure. The paired Student's t-test was carried out to compare the magnitude of response to MS between both sessions. Data are reported as means±SD. A probability less than or equal to $5\%$ was considered statistically significant in two-tailed analyses. The statistical package used was Statistica (version 10.0, StatSoft Inc. 2011, USA). ## Results The anthropometric, clinical, and biochemical profiles are presented in Table 1. As expected, all subjects presented a BMI between 26.7 and 34 kg/m2, characterizing the overweight/obesity grade I criteria, and a body fat mass higher than $27\%$. **Table 1** | Variables | Unnamed: 1 | Reference values | | --- | --- | --- | | Overweight (n) | 8 | - | | Obese (n) | 7 | - | | Age (years) | 27±7 | - | | Weight (kg) | 91.7±10.2 | - | | Height (cm) | 175±0.08 | - | | BMI (kg/m2) | 29.8±2.51 | 25-34.9 | | Body fat (%) | 31.7±3.62 | 12-20 | | Waist circumference (cm) | 99.0±5.9 | 90-110 | | SBP (mmHg) | 123±7 | ≤120 | | DBP (mmHg) | 80±8 | ≤80 | | Heart rate (bpm) | 72±11 | 60-100 | | Glucose (mg/dL) | 87.4±7.76 | 65-99 | | Insulin (uIU/mL) | 12.9±6.19 | 1.9-23 | | HOMA-IR | 2.98±1.42 | <4.5 | | HOMA-β | 192.70±105.08 | 167.0-175.0 | | Total cholesterol (mg/dL) | 185.45±41.17 | <190 | | HDL (mg/dL) | 41.18±7.54 | >40 | | LDL (mg/dL) | 120.54±35.32 | <130 | | VLDL (mg/dL) | 23.83±9.74 | 2-30 | | TG (mg/dL) | 109.36±57.1 | <150 | Regarding MS responses, according to the subjective scale used, the average level of perceived stress in all sessions was 2 (stressful). Table 2 shows systolic blood pressure (SBP), diastolic blood pressure (DBP), mean blood pressure (MBP), heart rate (HR), blood flow, and vascular conductance at baseline, during MS, and during recovery. There was a significant increase in the hemodynamic variables SBP, DBP, MBP, and HR ($P \leq 0.05$ vs baseline) during MS in all sessions. In recovery, these variables decreased to baseline levels ($P \leq 0.01$ vs MS). No differences were observed regarding blood flow and conductance. It is noteworthy that no differences were observed in the magnitude of response of hemodynamic variables between sessions, indicating that MS caused the same effect in all experimental sessions. Also, no differences were observed in regards to the direct effect of medication on hemodynamic variables, as we can attest by the lack of difference between the baseline moments of each session (Table 2). Moreover, none of the subjects enrolled in the present study reported adverse effects during any of the sessions. **Table 2** | Variables | Baseline | Mental Stress | Recovery | | --- | --- | --- | --- | | Placebo | | | | | SBP (mmHg) | 123±9 | 136±10* | 124±10 | | DBP (mmHg) | 76±8 | 88±10* | 77±7 | | MBP (mmHg) | 91±8 | 104±10* | 93±8 | | Heart rate (bpm) | 62±8 | 75±10* | 64±8 | | Blood flow (mL/min) | 179.89±77.49 | 224.37±102.31 | 232.63±141.96 | | Conductance (mL·min-1·mmHg-1) | 1.96±0.77 | 2.15±0.94 | 2.53±1.42 | | AT1R blockade | | | | | SBP (mmHg) | 120±8 | 133±10* | 123±8 | | DBP (mmHg) | 75±9 | 88±8* | 75±7 | | MBP (mmHg) | 87±17 | 99±19* | 92±6 | | Heart rate (bpm) | 63±8 | 76±10* | 65±9 | | Blood flow (mL/min) | 193.78±134.41 | 257.68±129.60 | 203.05±177.91 | | Conductance (mL·min-1·mmHg-1) | 2.07±1.38 | 2.47±1.24 | 2.35±1.86 | | Ascorbic acid | | | | | SBP (mmHg) | 124±9 | 140±6* | 126±10 | | DBP (mmHg) | 78±8 | 91±6* | 80±8 | | MBP (mmHg) | 94±8 | 107±6* | 95±8 | | Heart rate (bpm) | 65±11 | 76±11* | 65±8 | | Blood flow (mL/min) | 204.63±159.67 | 306.33±161.24 | 253.44±257.29 | | Conductance (mL·min-1·mmHg-1) | 2.22±1.38 | 2.95±1.41 | 2.73±2.52 | Regarding endothelial function, Dbase-adjusted FMD decreased 30 min after MS (baseline, 8.73±$1.03\%$ vs 30MS, 7.49±$1.03\%$; $$P \leq 0.05$$) during the placebo session but increased 60 min after the stress task (30MS, 7.49±$1.03\%$ vs 60MS, 9.93±$1.03\%$; $P \leq 0.02$). During the AT1R blockade session, FMD increased significantly in response to MS (baseline, 7.60±$1.02\%$ vs 30MS, 10.66±$1.03\%$; $P \leq 0.01$), and it was different from placebo (30MS, 10.66±$1.03\%$; $P \leq 0.01$). Baseline FMD was decreased compared to AA (OLM, 7.60±$1.02\%$ vs AA, 9.33±$1.03\%$; $P \leq 0.02$). Also, FMD decreased 60 min after MS (30MS, 10.66±$1.03\%$ vs 60MS, 9.89±$1.03\%$; $$P \leq 0.03$$), but it was still higher compared to baseline ($P \leq 0.01$). As for the AA session, Dbase-adjusted FMD at 60MS was improved compared to baseline (baseline, 9.25±$1.02\%$ vs 60MS, 11.03±$1.03\%$; $P \leq 0.04$) and 30MS (30MS, 9.41±$1.02\%$; $P \leq 0.01$) (Figure 2). **Figure 2:** *Flow-mediated dilation (FMD) before (baseline), 30 (30MS), and 60 min (60MS) after mental stress in overweight/obese individuals after oral administration of placebo, AT1R blockade, or AA. Vertical lines indicate means and SD. *P<0.05 vs baseline; †P<0.05 vs 30MS; ‡P<0.05 vs placebo at the same moment; §P<0.05 vs AT1R at the same moment (ANOVA). AT1RB: angiotensin II type 1 receptor blockade; AA: ascorbic acid.* Table 3 provides the results of resting diameter (cm), peak diameter (cm), FMD%, AUCSR, and FMD%/AUCSR. The resting diameter was smaller at the placebo session compared to AT1R blockade, leading to a baseline difference in resting diameter ($P \leq 0.01$), whilst baseline resting diameter at the AA session was smaller than baseline in the AT1R blockade. There was a baseline difference between FMD% during placebo and AT1R sessions ($P \leq 0.04$). As expected, peak diameter was significantly higher than resting diameter at all times during the three sessions ($P \leq 0.01$). At the placebo session, peak diameter at 30MS was lower compared to baseline ($P \leq 0.04$) and 60MS ($P \leq 0.02$), which was higher than baseline ($P \leq 0.02$). No differences regarding FMD% were observed in the placebo session. While AUCSR did not change during the placebo session, FMD%/AUCSR was significantly decreased at 30MS compared to baseline ($P \leq 0.02$). **Table 3** | Variables | Baseline | 30MS | 60MS | | --- | --- | --- | --- | | Placebo | | | | | Resting diameter (cm) | 0.407±0.57 | 0.408±0.56 | 0.409±0.53 | | Peak diameter (cm) | 0.441±0.55† | 0.436±0.55* † | 0.446±0.55* † § | | FMD (%) | 9.06±2.79 | 7.22±1.94 | 9.45±3.85 | | AUCSR (10-3·s1) | 20.79±10.36 | 24.78±15.02 | 20.98±9.11 | | FMD%/AUCSR (%·10-3·s1) | 5.17±4.01 | 3.46±1.98* | 4.14±2.00 | | AT1R blockade | | | | | Resting diameter (cm) | 0.416±0.64# | 0.403±0.62* | 0.408±0.59* | | Peak diameter (cm) | 0.443±0.63† | 0.446±0.57† | 0.444±0.57† | | FMD (%) | 6.69±3.54# | 11.20±5.03* # | 9.47±4.38* | | AUCSR (10-3·s1) | 23.93±18.62 | 22.05±9.13 | 23.79±9.07 | | FMD%/AUCSR (%·10-3·s1) | 4.79±5.18 | 5.88±4.02# | 4.07±2.25 | | Ascorbic acid | | | | | Resting diameter (cm) | 0.391±0.50‡ | 0.391±0.48 | 0.385±0.49* # § | | Peak diameter (cm) | 0.429±0.46† | 0.427±0.56† | 0.430±0.48† # | | FMD (%) | 9.39±3.34 | 9.54±4.56# ‡ | 12.20±3.88* § | | AUCSR (10-3·s1) | 20.20±10.70 | 18.56±7.47 | 22.47±9.77 | | FMD%/AUCSR (%·10-3·s1) | 5.55±3.55 | 5.19±2.550# | 6.53±3.26# ‡ | Regarding the AT1R blockade session, resting diameter presented the same behavior as peak diameter during the placebo session, although peak diameter did not change during AT1R blockade. In relation to FMD%, endothelial function was improved at 30 min compared to baseline ($P \leq 0.01$) and to the same moment in the placebo session ($P \leq 0.02$). At 60 min, FMD% was still increased compared to baseline ($P \leq 0.02$). AUCSR did not change during the AT1R blockade session. However, FMD%/AUCSR was increased at 30MS compared to placebo ($P \leq 0.02$). During the AA session, baseline resting diameter was lower compared to the same moment in the AT1R blockade session ($P \leq 0.01$). At 60 min, resting diameter was lower than baseline ($P \leq 0.01$) and at 30 min during the same session ($P \leq 0.02$), and lower than the same moment in the placebo session ($P \leq 0.01$). FMD% at 30 min was higher compared to the same moment in the placebo session ($P \leq 0.04$) but was lower compared to the AT1R blockade session ($P \leq 0.02$). At 60 min, FMD% was improved compared to baseline ($P \leq 0.02$) and 30 min ($P \leq 0.03$) during AA infusion. Similar to the AT1R blockade session, AUCSR did not change during the AA blockade session, but FMD%/AUCSR was increased at 30 min compared to placebo ($P \leq 0.02$). Moreover, at 60 min, FMD%/AUCSR was significantly greater in the AA session than in the placebo and AT1R blockade sessions. In the placebo session, lipid peroxidation was increased in response to MS (baseline, 4.97±1.02 nmol/mL vs MS, 6.06±1.95 nmol/mL, $P \leq 0.02$; placebo, 6.06±1.95 nmol/mL vs AT1R blockade, 1.08±0.43 nmol/mL, $P \leq 0.01$; placebo, 6.06±1.95 nmol/mL vs AA, 4.74±1.21 nmol/mL, $P \leq 0.01$) but greatly decreased 60 min after (baseline, 4.97±1.02 nmol/mL vs 60MS, 4.01±0.99 nmol/mL, $P \leq 0.03$; MS, 6.06±1.95 nmol/mL vs 60MS, 4.01±0.99 nmol/mL, $P \leq 0.01$). During the AT1R blockade session and AA sessions, no differences were observed, meaning that both prevented lipid peroxidation increase during MS (Figure 3A). Protein carbonylation was also increased during MS (baseline, 2.95±1.09 nmol/g vs MS, 4.52±1.95 nmol/g; $P \leq 0.01$) and returned to baseline levels 60 after MS during the placebo session (MS, 4.52±1.95 nmol/g vs 60MS, 3.16±2.45 nmol/g; $P \leq 0.03$). Similar to lipid peroxidation, AT1R blockade and AA prevented protein carbonylation increase during MS (placebo, 4.52±1.95 nmol/g vs AT1R blockade, 3.05±1.38 nmol/mL, $P \leq 0.01$; placebo, 4.52±1.95 nmol/g vs AA, 3.22±2.33 nmol/mL, $P \leq 0.01$) (Figure 3B). **Figure 3:** *Lipid peroxidation (A) and carbonylated proteins (B) before (baseline), during (MS), and 60 min (60MS) after mental stress in overweight/obesity individuals after oral administration of placebo, AT1R blockade, and AA. Vertical lines indicate means and SD. *P<0.05 vs baseline; †P<0.05 vs 60MS; ‡P<0.05 vs placebo at the same moment. FMD: flow-mediated dilation; AT1RB: angiotensin II type 1 receptor blockade; AA: ascorbic acid.* As for catalase activity, there was a baseline difference between placebo and AT1R blockade (placebo, 61.18±34.7 (nmol·min-1·mL-1) vs AT1R blockade, 120.27±96.2 pg/mL; $P \leq 0.05$). During the placebo session, catalase activity increased during MS (baseline, 61.18±34.7 pg/mL vs MS, 116.24±61.9 pg/mL; $P \leq 0.01$) and decreased to baseline levels 60 min after (MS, 116.24±61.9 pg/mL vs 60MS, 76.43±60.16 pg/mL; $P \leq 0.02$). Once again, no differences were observed with either the AT1R blockade or with AA (Figure 4A). Comparable to catalase, SOD activity increased during MS only in the placebo session (baseline, 0.80±0.49 pg/mL vs MS, 0.85±0.72 pg/mL, $P \leq 0.01$; placebo, 0.85±0.72 pg/mL vs AT1R blockade, 0.67±0.31 pg/mL, $P \leq 0.05$; placebo vs AA, 0.48±0.22 pg/mL, $P \leq 0.05$) and decreased to baseline levels 60 min after (MS, 0.85±0.72 pg/mL vs 60MS, 0.39±0.22 pg/mL; $P \leq 0.02$) (Figure 4B). **Figure 4:** *Catalase (A) and SOD (B) activity before (baseline), during (MS), and 60 min (60MS) after mental stress in overweight/obesity individuals after oral administration of placebo, AT1R blockade, and AA. Vertical lines indicate means and SD. *P<0.05 vs baseline; †P<0.05 vs MS; ‡P<0.05 vs placebo at the same moment (ANOVA). FMD: flow-mediated dilation; AT1RB: angiotensin II type 1 receptor blockade; AA: ascorbic acid.* ## Discussion The findings of the present study are three-fold: 1) AT1R blockade improved endothelial function after stressful situations in normotensive overweight/obese grade I men, reinforcing our hypothesis that the activation of the Ang II-AT1R pathway may be an important mechanism responsible for transient endothelial dysfunction; 2) ascorbic acid also improved endothelial function albeit only 1 h after exposure to stress; 3) stress provoked increases in both the oxidative profile (lipid peroxidation and protein carbonylation) and the antioxidant enzymes (catalase and SOD), while AT1R blockade and ascorbic acid prevented this response. Thus, the present study provided evidence that AT1R-mediated oxidative stress is an important underlying mechanism of transitory endothelial dysfunction induced by MS in overweight/obese adults. Tasks of mental stress have been largely used as a simulation of mental or psychological stress situations in a standardized and controlled environment under hemodynamic, vascular, and electrocardiographic monitoring. Several studies, including from our research group, have used this type of intervention to assess endothelial function in healthy subjects and patients under cardiometabolic risk [3,5- 7]. The Stroop task used in the present study was able to inflict the same stressful stimulus in the three sessions, as evidenced by similar increases in hemodynamic variables during the protocol. However, to the best of our knowledge, ours is the first study to provide direct evidence of the participation of the Ang II-AT1R pathway in the impairment of endothelial function in response to MS. In the present study, endothelial function was severely impaired 30 min after MS but recovered within 60 min in the placebo session. It is well documented that acute exposure to MS leads to transient endothelial dysfunction in healthy and pathological conditions [5,6]; however, the magnitude and extent of this response can be influenced by the duration of MS [5], subjects' responsiveness [27], and previous health conditions [28]. A previous study from our group showed that individuals with metabolic syndrome had reduced FMD at 30 and 60 min after acute MS [6]. Considering that the subjects in this research did not present any other comorbidity, the recovery process may have been more efficient. Moreover, the return of FMD to baseline levels was accompanied by a normalization of the oxidative stress levels. It is possible that the normalization of the redox homeostasis is the mechanism behind the improvement in endothelial function. On the other hand, AT1R blockade improved endothelial function after MS and this effect was maintained for up to 60 min, providing evidence that the Ang II-AT1R pathway is implicated in the transient endothelial function observed after MS. Indeed, the activation of the Ang II-AT1R pathway enhances the expression of ROCK1 and gp91 phox, the catalytic component of NAPDH oxidase [29], which promotes oxidative stress and imbalances among vasoactive substances. Ascorbic acid has been related to endothelial cell proliferation, apoptosis and smooth muscle-mediated vasodilation, among other endothelium-mediated effects [30]. Clinical studies have shown that intravenous infusion of ascorbate promotes endothelial-dependent dilation in patients with cardiovascular risk such as atherosclerosis [31] and diabetes [32], possibly by sparing endothelial cell-derived NO and scavenging superoxide that would otherwise react with free NO [33]. Corroborated by Plotnick et al. [ 34], AA did not influence endothelial function after mental stress, however, the maintenance of FMD at values similar to pre-stress could be interpreted as a protection/prevention mechanism against the transitory endothelial dysfunction observed during the placebo session. The improvement observed 60 min after MS could be a delayed effect of AA. It is important to highlight that Halliwill et al. [ 35] has shown that inhibition of the sympathetic system does not improve endothelial function during or after MS, reinforcing that oxidative stress may have a key role in modulating the endothelial response to MS. Endothelial cells generate superoxide and hydrogen peroxide (H2O2) as a result of both cytoplasmic and mitochondrial metabolism [33]. Moreover, NADPH oxidase activation, mainly by the stimulation of its subunits NOX1 and NOX4 [29], boosts the formation and accumulation of intracellular superoxide anion [36]. The superoxide anion is rapidly dismutated to hydrogen peroxide, provoking endothelial cell damage [29]. While ROS signaling is a key player in the maintenance of vascular tone, exposure to stressful situations seems to evoke imbalanced redox homeostasis, as observed in the present study. The MS task increased lipid peroxidation and protein carbonylation and increased catalase and SOD activity, possibly in response to a pro-oxidant environment. This phenomenon was neither observed when the Ang II- AT1R pathway was blocked nor when AA was infused. These findings supported our hypothesis that the transitory endothelial dysfunction observed after MS could be a result of A1TR-mediated redox imbalance. Some limitations must be considered when interpreting the results of the present study. The lack of women in our sample may be considered a limitation concerning the external validity of the results to the entire population. In order to avoid the established effects of sex hormones on the vascular function of women, we opted to enroll only men in the study. Therefore, the present results do not allow us to infer that the same responses would be observed in women. Also, a group with eutrophic subjects would enrich the study, so our results cannot be extrapolated to this specific population either. Additionally, the lack of plasma Ang II measurements is a limitation. Regardless of the Ang II levels, the AngII-AT1R pathway may have a role in stress-mediated endothelial dysfunction even in normotensive overweight/obese adults. However, the 3-way crossover, randomized, placebo-controlled protocol may attenuate this limitation. During the experimental sessions, blind evaluators assessed all measures; however, a single non-blind evaluator analyzed all FMD data. Moreover, endothelium-independent vasodilation and AT1R blockade could not be tested; however, Stangier et al. [ 37] showed that 40 mg of telmisartan produced $80\%$ inhibition of the receptor. Given that olmesartan presents higher binding affinity to the AT1R than telmisartan, it is unlikely that the inhibition induced in the present study was lower than that observed with telmisartan [38,39]. Lastly, our results did not reflect changes in vitro and other studies are necessary for a better understanding of the mechanisms involved in this phenomenon. In conclusion, the results of the present study provided compelling evidence regarding the transient endothelial dysfunction observed in response to acute MS. Moreover, the mentioned impairment in FMD seems to be directly influenced by the redox homeostasis imbalance. Blockade of the Ang II-AT1R pathway evoked a significant improvement in endothelial function after MS, while AA presented a delayed positive impact on flow-mediated dilation. Both interventions also prevented the deregulation of ROS signaling, providing evidence that A1TR-mediated oxidative stress is an important mediator of the FMD response after MS. Our study provides new insight into the mechanisms that underlie the deleterious response to MS and the MS consequence on the endothelial function of individuals with overweight/obesity. Thereby, defining new pathways that influence endothelial function is an important tool for understanding the pathological response to stressful situations. ## References 1. Trachta P, Drapalova J, Kavalkova P, Touskova V, Cinkajzlova A, Lacinova Z. **Three months of regular aerobic exercise in patients with obesity improve systemic subclinical inflammation without major influence on blood pressure and endocrine production of subcutaneous fat**. *Physiol Res* (2014.0) **63** S299-S308. DOI: 10.33549/physiolres.932792 2. Gonzalez-Muniesa P, Martinez-Gonzalez MA, Hu FB, Despres JP, Matsuzawa Y, Loos RJF. **Obesity**. *Nat Rev Dis Primers* (2017.0) **3** 17034. DOI: 10.1038/nrdp.2017.34 3. Broadley AJM, Korszun A, Abdelaal E, Moskvina V, Jones CJH, Nash GB. **Inhibition of cortisol production with metyrapone prevents mental stress-induced endothelial dysfunction and baroreflex impairment**. *J Am Coll Cardiol* (2005.0) **46** 344-350. DOI: 10.1016/j.jacc.2005.03.068 4. Musey PI, Schultebraucks K, Chang BP. **Stressing out about the heart: a narrative review of the role of psychological stress in acute cardiovascular events**. *Acad Emerg Med* (2020.0) **27** 71-79. DOI: 10.1111/acem.13882 5. Ghiadoni L, Donald AE, Cropley M, Mullen MJ, Oakley G, Taylor M. **Mental stress induces transient endothelial dysfunction in humans**. *Circulation* (2000.0) **102** 2473-2478. DOI: 10.1161/01.CIR.102.20.2473 6. Sales ARK, Fernandes IA, Rocha NG, Costa LS, Rocha HNM, Mattos JDM. **Aerobic exercise acutely prevents the endothelial dysfunction induced by mental stress among subjects with metabolic syndrome: the role of shear rate**. *Am J Physiol Heart Circ Physiol* (2014.0) **306** H963-H971. DOI: 10.1152/ajpheart.00811.2013 7. Rocha NG, Sales AR, Miranda RL, Silva MS, Silva JF, Silva BM. **Aerobic exercise modulation of mental stress-induced responses in cultured endothelial progenitor cells from healthy and metabolic syndrome subjects**. *Life Sci* (2015.0) **123** 93-99. DOI: 10.1016/j.lfs.2014.12.026 8. Geiker NRW, Astrup A, Hjorth MF, Sjödin A, Pijls L, Markus CR. **Does stress influence sleep patterns, food intake, weight gain, abdominal obesity and weight loss interventions and vice versa?**. *Obes Rev* (2018.0) **19** 81-97. DOI: 10.1111/obr.12603 9. Favero G, Paganelli C, Buffoli B, Rodella LF, Rezzani R. **Endothelium and its alterations in cardiovascular diseases: life style intervention**. *Biomed Res Int* (2014.0) **2014** 801896. DOI: 10.1155/2014/801896 10. Mehta PK, Griendling KK. **Angiotensin II cell signaling: physiological and pathological effects in the cardiovascular system**. *Am J Physiol Cell Physiol* (2007.0) **292** C82-C97. DOI: 10.1152/ajpcell.00287.2006 11. Liu H, Kitazato KT, Uno M, Yagi K, Kanematsu Y, Tamura T. **Protective mechanisms of the angiotensin II type 1 receptor blocker candesartan against cerebral ischemia:**. *J Hypertens* (2008.0) **26** 1435-1445. DOI: 10.1097/HJH.0b013e3283013b6e 12. Widlansky ME, Gokce N, Keaney JF, Vita JA. **The clinical implications of endothelial dysfunction**. *J Am Coll Cardiol* (2003.0) **42** 1149-1160. DOI: 10.1016/S0735-1097(03)00994-X 13. Apovian CM, Gokce N. **Obesity and cardiovascular disease**. *Circulation* (2012.0) **125** 1178-1182. DOI: 10.1161/CIRCULATIONAHA.111.022541 14. Barton M, Carmona R, Ortmann J, Krieger JE, Traupe T. **Obesity-associated activation of angiotensin and endothelin in the cardiovascular system**. *Int J Biochem Cell Biol* (2003.0) **35** 826-837. DOI: 10.1016/S1357-2725(02)00307-2 15. Celermajer DS, Sorensen KE, Gooch VM, Spiegelhalter DJ, Miller OI, Sullivan ID. **Non-invasive detection of endothelial dysfunction in children and adults at risk of atherosclerosis**. *Lancet* (1992.0) **340** 1111-1115. DOI: 10.1016/0140-6736(92)93147-F 16. Tsai TH, Chai HT, Sun CK, Yen CH, Leu S, Chen YL. **Obesity suppresses circulating level and function of endothelial progenitor cells and heart function**. *J Transl Med* (2012.0) **10** 137. DOI: 10.1186/1479-5876-10-137 17. Pi-Sunyer X. **The medical risks of obesity**. *Postgrad Med* (2009.0) **121** 21-33. DOI: 10.3810/pgm.2009.11.2074 18. Simon GE, Arterburn D, Rohde P, Ludman EJ, Linde JA, Operskalski BH. **Obesity, depression, and health services costs among middle-aged women**. *J Gen Intern Med* (2011.0) **26** 1284-1290. DOI: 10.1007/s11606-011-1774-x 19. **DATASUS. Mortalidade por doenças do aparelho circulatório (CID-10) 2011** 20. Schwocho LR, Masonson HN. **Pharmacokinetics of CS-866, a new angiotensin II receptor blocker, in healthy subjects**. *J Clin Pharmacol* (2001.0) **41** 515-527. DOI: 10.1177/00912700122010393 21. Thijssen DH, Black MA, Pyke KE, Padilla J, Atkinson G, Harris RA. **Assessment of flow-mediated dilation in humans: a methodological and physiological guideline**. *Am J Physiol Heart Circ Physiol* (2011.0) **300** H2-H12. DOI: 10.1152/ajpheart.00471.2010 22. Stroop JR. **Studies of interference in serial verbal reactions**. *J Experimental Psychol* (1935.0) **18** 643-662. DOI: 10.1037/h0054651 23. Fernandes IA, Sales ARK, Rocha NG, Silva BM, Vianna LC, da Nobrega ACL. **Preserved flow-mediated dilation but delayed time-to-peak diameter in individuals with metabolic syndrome**. *Clin Physiol Funct Imaging* (2014.0) **34** 270-276. DOI: 10.1111/cpf.12092 24. Tsikas D. **Assessment of lipid peroxidation by measuring malondialdehyde (MDA) and relatives in biological samples: analytical and biological challenges**. *Anal Biochem* (2017.0) **524** 13-30. DOI: 10.1016/j.ab.2016.10.021 25. Lowry OH, Rosebrough NJ, Farr AL, Randall RJ. **Protein measurement with the Folin phenol reagent**. *J Biol Chem* (1951.0) **193** 265-275. DOI: 10.1016/S0021-9258(19)52451-6 26. Atkinson G, Batterham AM. **Allometric scaling of diameter change in the original flow-mediated dilation protocol**. *Atherosclerosis* (2013.0) **226** 425-427. DOI: 10.1016/j.atherosclerosis.2012.11.027 27. Carter JR, Ray CA. **Sympathetic neural responses to mental stress: responders, nonresponders and sex differences**. *Am J Physiol Heart Circ Physiol* (2009.0) **296** H847-H853. DOI: 10.1152/ajpheart.01234.2008 28. Loures DL, Sant Anna I, Baldotto CSR, de Sousa EB, da Nobrega ACL. **Mental stress and cardiovascular system [in Portuguese]**. *Arq Bras Cardiol* (2002.0) **78** 525-530. DOI: 10.1590/S0066-782X2002000500012 29. Chung IM, Kim YM, Yoo MH, Shin MK, Kim CK, Suh SH. **Immobilization stress induces endothelial dysfunction by oxidative stress via the activation of the angiotensin II/its type I receptor pathway**. *Atherosclerosis* (2010.0) **213** 109-114. DOI: 10.1016/j.atherosclerosis.2010.08.052 30. May JM, Harrison FE. **Role of vitamin C in the function of the vascular endothelium**. *Antioxid Redox Signal* (2013.0) **19** 2068-2083. DOI: 10.1089/ars.2013.5205 31. Tousoulis D, Xenakis C, Tentolouris C, Davies G, Antoniades C, Crake T. **Effects of vitamin C on intracoronary L-arginine dependent coronary vasodilatation in patients with stable angina**. *Heart* (2005.0) **91** 1319-1323. DOI: 10.1136/hrt.2004.044537 32. Tousoulis D, Antoniades C, Vasiliadou C, Kourtellaris P, Koniari K, Marinou K. **Effects of atorvastatin and vitamin C on forearm hyperaemic blood flow, asymmentrical dimethylarginine levels and the inflammatory process in patients with type 2 diabetes mellitus**. *Heart* (2007.0) **93** 244-246. DOI: 10.1136/hrt.2006.093112 33. May JM. **How does ascorbic acid prevent endothelial dysfunction?**. *Free Radic Biol Med* (2000.0) **28** 1421-1429. DOI: 10.1016/S0891-5849(00)00269-0 34. Plotnick MD, D'Urzo KA, Gurd BJ, Pyke KE. **The influence of vitamin C on the interaction between acute mental stress and endothelial function**. *Eur J Appl Physiol* (2017.0) **117** 1657-1668. DOI: 10.1007/s00421-017-3655-4 35. Halliwill JR, Lawler LA, Eickhoff TJ, Dietz NM, Nauss LA, Joyner MJ. **Forearm sympathetic withdrawal and vasodilatation during mental stress in humans**. *J Physiol* (1997.0) **504** 211-220. DOI: 10.1111/j.1469-7793.1997.211bf.x 36. Nauseef WM. **Biological roles for the NOX family NADPH oxidases**. *J Biol Chem* (2008.0) **283** 16961-16965. DOI: 10.1074/jbc.R700045200 37. Stangier J, Su CA, van Heiningen PN, Meinicke T, van Lier JJ, de Bruin H. **Inhibitory effect of telmisartan on the blood pressure response to angiotensin II challenge**. *J Cardiovasc Pharmacol* (2001.0) **38** 672-685. DOI: 10.1097/00005344-200111000-00004 38. Meredith P. **Comparative ARB pharmacology**. *Br J Cardiol* (2010.0) **17** s3-s5 39. Miura S, Karnik SS, Saku K. **Review: angiotensin II type 1 receptor blockers: class effects**. *J Renin Angiotensin Aldosterone Syst* (2011.0) **12** 1-7. DOI: 10.1177/1470320310370852
--- title: 'Examining capabilities, opportunities, and motivations for healthy eating behaviors in Latin American restaurants: a quantitative application of the COM-B model to inform future interventions' authors: - Melissa Fuster - Maria P. Santos - Emily Dimond - Terry T. K. Huang - Margaret A. Handley journal: BMC Nutrition year: 2023 pmcid: PMC10041691 doi: 10.1186/s40795-023-00712-1 license: CC BY 4.0 --- # Examining capabilities, opportunities, and motivations for healthy eating behaviors in Latin American restaurants: a quantitative application of the COM-B model to inform future interventions ## Abstract ### Background Eating foods away from home has been associated with poor diet quality and adverse health outcomes. Research is needed to examine barriers and facilitators to making healthier eating choices in restaurant settings. We operationalized the Capability, Opportunity, and Motivation for Behavior Model (COM-B Model) to conduct a behavioral diagnosis for healthy eating behaviors at Latin American restaurants (LARs), an understudied yet increasingly important food environment with the potential to positively influence diets. ### Methods We conducted an online survey with adults in the United States that reported eating food from LARs at least once a month ($$n = 509$$) recruited via an online market research panel to examine capabilities – physical (e.g., skills) and psychological (e.g., knowledge), opportunities – social (e.g., norms) and physical (e.g., environmental), and motivations – reflective (e.g., self-conscious intentions) and automatic (e.g., emotions) associated with healthier choices at LARs. In a survey focused on LAR-associated behaviors, each COM-B domain was scored between 1–5, with scores ≥ 4 denoted as having high capability, opportunity, and motivation to eat healthfully at LARs (potential range of total score = 6–35). Regression analysis was used to examine the association between COM-B scores (total and by domain) and select demographic characteristics (age, gender, race, Latin heritage, income, education, marital status, and Latin majority state of residency). ### Results More than half of the participants ($57.1\%$) were classified as having high physical capability, followed by psychological capability ($43.9\%$) in the LAR environment. The proportions of participants with either high motivation or high opportunity were low, ranging from $37.3\%$ (reflective motivation) to physical opportunity ($15.6\%$). The overall mean COM-B total score was 19.8 ± 3.0. Higher total COM-B scores were associated with younger age, self-identifying as white, having Latin heritage, and having higher income ($p \leq 0.05$). ### Conclusions This study expands the application of the COM-B framework using quantitative inquiry to evaluate levels of capability, motivation, and opportunity for healthy eating in LAR settings and initial demographic associations with determinants for healthy eating in these settings. This work can aid in tailoring interventions and developing evaluation tools for LAR-related healthy eating interventions. ## Background While public health policies and interventions have been increasingly focusing on restaurants as a sector for intervention, these efforts often fail to include establishments serving minority communities, exacerbating existing diet-related health inequities. According to the National Restaurant Association, $80\%$ of consumers eat at a restaurant serving ethnic cuisine at least once a month [1]. Within these, there are over 120,000 Latin American restaurants (LARs) in the United States (US), most of which are independently owned. Mexican restaurants alone make up $8\%$ of all US restaurants [2, 3]. Yet, despite their importance, LARs (along with other ethnic restaurants) remain an understudied sector, where factors associated with consumer food choices in these settings are not well understood. This gap is important, given the increased consumption of foods away from home and its association with poor dietary outcomes [4]. Diet-related non-communicable diseases, such as heart disease and diabetes, are among the leading causes of death globally. In the U.S., minority populations, including Latins/Hispanics, are disproportionately affected by these conditions. Research examining consumer choices in restaurants show that several customer attributes are associated with nutrition considerations when choosing a meal, including knowledge of health issues, weight concerns, gender, age, and marital status [5]. This research has not taken place in ethnic eateries, where there may be additional factors influencing food choices and availability, given the importance of culture and notions of authenticity [6, 7]. Theory-driven research is needed to provide evidence for future intervention and policies that can best address the added level of complexity when seeking to promote behavior change related to establishments serving ethnic cuisines. This research need presents an opportunity for the application of implementation theories and frameworks to address a complex, community-based evidence-to-practice gap. Determinant frameworks, such as the Capability, Opportunity, and Motivation for Behavior (COM-B) Model can help identify key drivers of dietary behaviors for informing intervention design, as part of the Behavior Change Wheel, an intervention development framework that brings together theory-based tools to understand and change behaviors [8]. The COM-B model synthetizes theoretical frameworks to understand a given behavior within interacting factors – capability, opportunity, and motivation. Capability is addressed as physical (i.e., whether the individual has the needed skills to engage in the desired behavior) and psychological (i.e., an individual’s comprehension and knowledge). Opportunity encompasses the external factors influencing the behavior, including physical (i.e., aspects of the built and food environments that influence behavior) and social opportunity (i.e., cultural and social norms that influence the behavior). Lastly, motivation encompasses the internal processes that direct behavior, including reflective (i.e., evaluations and plans) and automatic (emotions) motivation [8]. The COM-B Model has been used to understand a wide range of behaviors, including eating behaviors [8–10]. However, the research has been mostly qualitative. Willmott and colleagues present an exception, operationalizing COM-B to examine general healthy eating behaviors among youth, demonstrating the predictive utility of COM-B for general healthy eating behaviors [11]. In this study, we built on these previous applications to [1] develop a short COM-B-based questionnaire to assess determinants of making healthier eating choices in LARs, and [2] examine the demographic factors associated with COM-B. ## Survey development We developed a cross-sectional survey that operationalized the six COM-B domains into a short scale. The items were developed building on previous research that operationalized the COM-B [11, 12]. The resulting scale was composed of 14 items, encompassing thirteen 5-point Likert-scale questions and one open-ended question. The 5-point Likert-scale questions asked respondents to rate the level of agreement with a given statement and were scored according to how the response supported the given COM-B domain. The psychological capability domain was examined using an open-ended question to assess knowledge about healthful dishes in Latin American cuisines, designed to accommodate the potential variety in responses, given the diversity within the Latin American community. The scale was pre-tested among a small number of respondents ($$n = 5$$) who met the survey inclusion criteria and revised for language clarification. ## Data collection Data were collected using a web survey between April and July 2021 distributed via social media and through Centiment, an online market research panel. The survey was available in English and Spanish. Participants were screened for eligibility prior to the full survey application. Inclusion criteria for the study were: being an adult (18 years of age or older), living in the 50 US states or Puerto Rico, and eating at or ordering from restaurants specializing in Latin American cuisines at least once a month. The survey took about 10–12 min to complete. In addition to the COM-B questions, we collected demographic information as follows: Income was assessed as individual annual income, using eight categories in the survey, ranging from $0 to $150,000 or greater. Education was assessed as the highest level attained (less than high school, high school/GED, some college, bachelor's degree, post-graduate education). Gender was assessed as male, female, transgender woman, transgender man, and non-binary. Latin/Hispanic background was self-reported through the question, “Do you have a Latin/Hispanic heritage?” ( Yes/No), and we included a question to collect specific heritage information. Race was collected using distinct categories (White, Black/African American, Asian, American Indian/Alaska Native, Pacific Islander, Mixed/Multi-racial) and a write-in option. Age was collected as a continuous variable. We also collected information about employment status, marital status, and place of residence. ## Data analysis The Likert-scale responses were scored according to level of support for COM-B domains, using a scale of 1–5. The qualitative responses for psychological capability were coded by two raters independently, initially using a scale of 0–2. Response that clearly showed a lack of knowledge regarding potentially healthful Latin foods were scored as “0”, including those who responded “I don’t know,” or who listed unhealthful options (e.g., fried foods). Responses that provided a potentially healthy Latin food but lacked specificity or clarity for why they were healthy were scored as "1". Examples included “chicken” and “a vegetarian dish.” Responses that clearly showcased knowledge of healthy Latin foods were scored as "2". Common examples included a specific food like ceviche (a seafood-based dish) or a healthier preparations of a usual dish like “baked chicken” (instead of fried). The raters were trained in qualitative coding, and they had nutrition training and knowledge of Latin American cuisines. Independent scores were reconciled by the raters, in collaboration with the lead author and study principal investigator, where the team discussed the responses in disagreement until a common score was reached. Responses were recoded into a 1–5 scale, to conform with the rest of the scale items. The six individual COM-B domain scores were calculated as the average of responses received within the individual items related to the domain, to avoid giving more weight to domains where more items were incorporated. The total COM-B score was calculated by summing up the resulting individual six scores, with a potential range of 6–30, where higher scores denoted greater capability, opportunity, and motivation to make healthier choices at LARs. Confirmatory factor analysis was used to validate the COM-B in our sample, using variance–covariance matrix with maximum likelihood estimation and varimax rotation. For descriptive, exploratory analysis, domain scores were transformed into a categorical variable to denote high scores, using the cut-off of 4 or higher. However, total and domain scores were analyzed as continuous variables given the resulting distribution of the exploratory categorical variables. Mean and standard deviations were computed for each item in the COM-B scale, as well as the total COM-B score and domain scores. Scores were normally distributed. We carried out descriptive statistical analysis as part of the sample description and preliminary bivariate analysis to examine the association between demographic characteristics and COM-B domains and total scores. We then carried out multilinear regression analysis to examine the association of select socio-demographic characteristics with the COM-B scores, namely, age, race, Latin heritage, gender, marital status, educational attainment, income, and place of residence. Age was analyzed as a continuous variable. Due to data distribution, race was collapsed as a binary variable, denoting respondents identifying as white, compared to those selecting other categories (non-white). Latin heritage was assessed as binary (yes/no), with specific heritage reported only as part of sample description, due to small cell counts for more specific analysis. Gender was analyzed as male and female, due to the small sample ($$n = 3$$) self-identifying within nonbinary categories, which were set to missing. The education variable was analyzed as a binary variable comparing respondents with less than a bachelor’s degree to those with a bachelor’s degree or higher. Income was examined as Low (< $25,000), Middle ($25,000-$74,999) and High (> $75,000) based on the 2020 US median personal income and the Pew Research Center income classification method [13, 14]. Marital status was analyzed as a binary variable (married/living with someone compared to single/divorced/separated) (no widowed person in sample). Place of residence was collapsed into a binary variable, denoting if the state had an above average percentage of Latin/Hispanic population ($12\%$ according to 2010 Census data) [15]. The analysis was conducted using SAS 9.4 and STATA 11.2. Significance was set at $p \leq 0.05$, but we also noted marginally significant associations at $p \leq 0.10$, given the exploratory nature of the analysis. ## Sample characteristics More than half of the survey sample was comprised of those who are white, female, married, and employed (Table 1). A larger percentage ($46\%$) was classified as middle income than lower ($23\%$). Regarding educational attainment, there was a close to even split between respondents with a bachelor’s degree or higher, and those with some college or below. More than half of the sample reported residing in a US state/location with above average percentage of Latin population. The top locations were California ($11.3\%$), Texas ($10.8\%$), New York ($9.3\%$), and Florida ($9.0\%$) – the states with the highest population of Latinos in the US. Close to half of the sample reported having Latin heritage. Regarding racial self-identification, most with Latin heritage identified as white ($41.5\%$) or mixed/other ($36.2\%$); only a few identified as Black ($6.3\%$). Table 1Sample Description ($$n = 509$$, except where noted)Sample Characteristicsn (%) or mean ± SDRace White341 (67.3) Black32 (6.3) American Indian / Alaska Native6 (1.2) Asian17 (3.4) Native Hawaiian / Pacific Islander3 (0.6) Multiracial84 (16.6)Latin/Hispanic heritage (% yes)210 (41.3)Age47.1 ± 18.3Gender Woman273 (53.5) Man230 (45.3) Gender minority/non-binary3 (0.6)*Marital status* Married/living with someone267 [53] Single155 (30.8) Widowed/Divorced/separated76 (15.1)Income ($$n = 468$$) Low (< $25,000)110 (23.4) Middle ($25,000–74,9999)217 (46.4) High (≥ $75,000)141 (30.1)Education ($$n = 492$$) No post-secondary education91 (18.4) Some college138 (27.9) Bachelor’s160 (32.3) Graduate degree106 (21.4)Employment Employed298 (62.2) Unemployed50 (10.4) Retired131 (27.4)% Living in state with above average Latin/Hispanic population271 (53.6) ## Determinants of making healthy choices in LARs: COM-B assessment Individual COM-B item mean scores ranged from 2.86 to 3.80, with the highest score found for being willing to try new healthier foods when eating at LARs (Motivation-Reflective). In CFA, using three factors, item loading ranged from 0.41 to 0.96, with all items above 0.4 (Table 2). The resulting Goodness of Fit Index and Adjusted Goodness of Fit Index were 0.98 and 0.92, respectively. Table 2COM-B scale item mean scores and factor loading, with resulting mean domain scores distributions ($$n = 504$$)COM-B DomainScale Item1Item Score (mean ± SD)Factor Loading in CFADomain Score (mean ± SD)% with High Domain Scores2 n (%)Capability-Psychological (Knowledge)3When you think of healthy, or “good for you” dishes in Latin cuisines, what comes to mind?3.02 ± 1.920.413.02 ± 1.92224 ($46.5\%$)Capability-Physical (Skill)*It is* easy to identify healthy options when visiting Latin restaurants3.57 ± 0.920.523.57 ± 0.92289 ($57.1\%$)Opportunity -Social (Social influences)Most of my friends and family tend to order healthy options when visiting Latin restaurants3.56 (0.87)0.553.08 ± 0.6452 ($10.3\%$)I expect to find healthy options when visiting Latin restaurants3.14 (0.98)0.63Authentic Latin dishes are not really healthy.a2.94 (1.03)0.57Opportunity-Physical (Environmental constraints)Latin restaurants tend to offer a good variety of healthy and appealing choices3.45 (0.95)0.963.25 ± 0.6279 ($15.6\%$)Healthier choices in Latin restaurants tend to be more expensive than other, less healthy choices. a2.86 (1.06)0.42Latin restaurants tend to serve too much food. a3.45 (0.95)0.96Motivation-Reflective (Self-efficacy, plans)I am willing to try new healthier foods when eating at Latin restaurants3.80 (0.96)0.673.65 ± 0.62189 ($37.3\%$)*It is* OK to indulge in foods that may not be healthy when eating at Latin restaurants3.66 (0.94)0.41Ordering healthy choices at Latin restaurants will have a big positive impact in my overall health3.62 (0.98)0.63I want to eat healthy dishes at Latin restaurants3.53 (0.92)0.71Motivation-Automatic (Emotions, reinforcements)Eating healthier choices at Latin restaurants makes me like feel I am restricting myself. a3.01 (1.09)0.423.29 ± 0.73122 ($24.1\%$)I tend to feel good physically when I select lighter meals in Latin restaurants3.58 (0.93)0.651With the exception of psychological capability, all items scored based on responses to Likert-scale responses (1 = Strongly disagree; 2 = Disagree; 3 = Neither agree nor disagree; 4 = Agree; 5 = Strongly agree). Items marked with (a) denote reverse scoring, where disagreement was scored higher;2Scores ≥ 4;3Smaller sample size for psychological capability, $$n = 482$$ The mean COM-B domain scores ranged from 3.02 (Capability-Psychological) to 3.65 (Motivation-Reflective) (Table 2). Within a possible range of 6–30, the total mean COM-B score was 19.8 ± 3.0, ranging from 11.6 to 30. When score distributions were examined as the proportion within the sample scoring high (≥ 4) in each COM-B domain, the domains with the highest percentage of high scores were those related to capability, where more than half or close to half of the sample were classified as having the physical and psychological capability, respectively, to select healthier choices at LARs (Table 2). The prevalence of high scores was much lower for the motivation and opportunity domains. While only approximately one-third of respondents had high motivation to select healthier choices at LARs, even fewer were classified as having the social and physical opportunities needed to engage in the desired behavior (Table 2). On average, respondents had a total of 1.9 ± 1.4 domains with high scores. Very few respondents had high scores across all or most of the six domains; more than half presented only one to two high scoring domains (Fig. 1).Fig. 1Distribution of total COM-B domains with high (≥ 4) scores, $$n = 489$$ We found overlaps among participants scoring high in specific domains. Figure 2 denotes the proportion of participants with overlapping high scoring domains, darker gray color denotes a higher proportion of participants with high scores in both corresponding domains in the y and x axis. Physical capacity and reflective motivation had the highest proportion of overlapping high scores ($35\%$ of participants), followed by automatic motivation and physical capacity ($25\%$ of participants) and reflective motivation and physical opportunity ($22\%$ of participants; Fig. 2).Fig. 2Proportion of participants with overlapping high scores by domain dyads ## Sociodemographic factors associated with higher COM-B scores Higher total COM-B scores were associated with younger age, being white, having Latin heritage, being a woman, higher income, and living in a state with above average percentage of Latin/Hispanic population (Table 3). Associations also across domain scores. Age and race were significantly associated across 5 of the 6 domains (except for psychological capability, which was significantly associated with gender—higher among women—and higher education). For age, younger customers had higher scores, except for physical opportunity, where increasing age was associated with increased score. Latin heritage was positively associated with higher social opportunity and automatic motivation. Aside from higher psychological capability, being a woman was associated with higher reflective motivation, but marginally lower automatic motivation ($p \leq 0.10$). Income was marginally associated with the automatic motivation domain, where medium income compared to low income was associated with lower scores ($p \leq 0.10$). Lastly, living in a location with above average percentage of Latin/Hispanic population was only associated with higher social opportunity. The total COM-B score showed the highest number of significantly associated demographic factors. Across the six domains, automatic motivation showed the most associations with demographic factors (5 out of 6), followed by social opportunity, with 4 out of 6 (Table 3).Table 3Multilinear Regression results examining COM-B scores (total and by domain) against sociodemographic factors ($$n = 429$$)COM-BCapability-PsychologicalCapability-PhysicalOpportunity-PhysicalOpportunity-SocialMotivation-ReflectiveMotivation-Automaticβ (se)β (se)β (se)β (se)β (se)β (se)β (se)Age-0.026 (0.009)**0.009 (0.006)-0.008 (0.003)**0.005 (0.002)*-0.008 (0.002)***-.006 (0.002)**-0.007 (0.002)**Race Non-WhiteREFREFREFREFREFREFREF White1.24 (0.37)**0.007 (0.242)0.313 (0.112)**0.124 (0.075)^0.253 (0.077)**0.222(0.077) **0.387 (0.091)***Latin Heritage NoREFREFREFREFREFREFREF Yes0.977 (0.342)**0.282 (0.224)0.151 (0.105)0.115 (0.070)0.227 (0.072)**0.052 (0.071)0.159 (0.084)^Gender WomanREFREFREFREFREFREFREF Man-0.517 (0.288)^-0.412 (0.188)*-0.096 (0.089)-0.001 (0.059)0.042 (0.061)-0.120 (0.060)*0.120 (0.071)^Marital Status Single/divorced/ separatedREFREFREFREFREFREFREF Married/Living with someone0.206 (0.305)0.123(0.200)-0.024 (0.094)0.012 (0.060)0.015 (0.065)0.0314 (0.064)0.072 (0.076)Educational Attainment Some college or lessREFREFREFREFREFREFREF Bachelor’s or higher0.317 (0.294)0.438 (0.193)*-0.052 (0.091)-0.028 (0.060)-0.042 (0.062)0.045 (0.062)-0.025 (0.073)Individual Income Low (< $25,000)REFREFREFREFREFREFREF Medium ($25,000-$74,999)-0.136 (0.369)0.021(0.242)-0.026 (0.114)-0.063 (0.075)-0.002 (0.078)0.046 (0.078)-0.162 (0.092)^ High ($75,000 +)0.918 (0.438)*0.437 (0.288)0.216 (0.136)0.081 (0.090)0.11 (0.093)0.105 (0.093)0.054 (0.109)Living in location with above average Latin/Hispanic population NoREFREFREFREFREFREFREF Yes0.592 (0.283)*0.168 (0.186)0.024 (0.087)0.020 (0.058)0.104 (0.060)^0.120 (0.059)0.049 (0.070)Levels of significance: ^$p \leq 0.10$, *$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$ ## Discussion This study applied a theoretical framework to examine capabilities, opportunities, and motivation associated with healthier eating behaviors in LARs, an understudied, yet increasingly important community food source. The application of the COM-B model allows a systematic exploration of behavior, in this case, a range of behaviors associated with selecting healthier foods at LARs. The examination revealed that, while interventions may be needed across all areas of the COM-B, there is a pressing need to address physical and social opportunities, as the domains found most lacking among the respondents, to make a strong impact on these LAR environment healthy eating behaviors. Physical opportunity is having an environment that affords the time, resources, location, and other physical affordances associated with the desired behavior. The association between food environments and healthier food consumption has been documented [16], noting linkages between healthful environments and cardiovascular disease risks [17]. In this study, we focused on the availability of appealing healthy options that are also affordable and in adequate portions. While food environment research to date underscores the need to increase availability and accessibility of healthy options, this paper illustrates the importance of norms and notions associated with foods. These factors are captured under the social opportunity domain, referring to interpersonal influences and sociocultural norms that enable the desired behavior [8]. Social norms can influence food consumption via peer influence and also by changing perceptions of certain foods, with the potential for making healthier options more appealing [18, 19]. In restaurant setting, social perceptions concerning establishment types have the potential to influence food choices, regardless of actual food availability. For instance, notions of fast food restaurants as places for junk food consumption have been associated as a barrier for healthy eating at these establishments [20]. Moreover, research has also started to examine the role of hedonic descriptions to motivate healthier food consumption in restaurant menus [21]. The application of the COM-B as part of the larger, related intervention design framework – the Behavior Change Wheel Model – allows for the identification of potential intervention functions to promote desired changes. Physical and social opportunity can be enhanced via environmental restructuring, restriction, and enablement. These changes imply interventions beyond individual-level strategies and the importance of changing consumer nutrition environments in LARs, not only by increasing the variety of healthier food options, but also by enabling healthier choices through, for example, menus highlights. Emerging evidence from restaurant intervention studies suggests that restaurants that increase the availability of healthier options and enable healthier choices can result from increased consumption [cite/expand]. These changes can also boost social opportunity, showcasing healthy options through innovative approaches that not only underscore the health benefit of these choices but also showcases these healthier options as traditional and palatable. While most interventions, to date, have focused on augmenting physical opportunity, more attention is needed to address social norms concerning healthier foods. These approaches can help increase interest in these offerings, while also normalizing these healthier options in the community, which, in turn, can help foster changes in social and cultural norms about which foods customers should expect to find in these establishments. Our study’s operationalization of the COM-B model facilitated a statistical analysis to examine demographic characteristics associated with having the capabilities, opportunities, and motivations necessary to engage in healthier eating behaviors at LARs. Age and race were consistently associated across all domains, except for psychological capability. The results suggest that younger customers may have more social opportunity, motivation, and skills (physical capability) to engage in healthier eating behaviors at LARs. This corroborates previous research showing greater interest in healthier eating among younger generations [22]. On the other hand, older age was associated with physical opportunity, which may point to age-related differences in perceived physical and economic access to healthy choices in LARs. The significant association with race shows the potential racial inequities in facilitators for healthier choices in LARs. While most studies tend to examine race and ethnicity, incorporating “Hispanic/Latino” as a category, our study examined race and Latin heritage separately, allowing for a more nuanced exploration of these demographic characteristics. Our results suggest that self-identifying as white was associated with more COM-B domains, compared with having Latin heritage. These findings showcase the importance of examining racial differences among Latin/Hispanic communities, an important research need documented in past research [23]. Gender was associated with higher psychological capability and motivation. This coincides with research that examines gender difference in interest in healthy eating behaviors. In restaurant settings, women tend to worry more about the caloric content of foods, compared to men [24]. Healthier foods, like vegetables, fruit, and fish, are typically associated with femininity, and women are usually more aware of the health-diet relationship than men [25]. Research examining food purchasing behaviors by gender is scarce, and it suggests that men tend to consume more foods away from home, compared with women, but no significant differences in diet quality were found [26]. More research is still needed to elucidate gender differences, including more studies that focus on men and gender minorities. Variables associated with socioeconomic status (income and education) did not show associations across most COM-B domains. Income was only significantly and positively associated with total COM-B score. Higher education was only significantly associated with greater psychological capability (knowledge). Hence, while socioeconomic factors have been documented as important factors associated with diet quality [27, 28], economic access appears to be only one part of the story for food choices in restaurant settings. However, more research is needed to elucidate differences in food consumption in restaurants by socioeconomic characteristics, building on emerging work examining dietary intakes by food source [29]. Lastly, we found that residence in a state with an above average percentage of Latin population, was significant only for the total COM-B score and social opportunity. We expected to find that a higher proportion of Latin/Hispanic population may result in higher exposure to LARs and Latin food in markets [30], and that this exposure may, in turn, influence aspects of the COM-B, such as through increased knowledge of Latin foods or greater social opportunity, for example. Our results suggest some association, but more research is needed. Our research was limited, as we assessed Latin/Hispanic population at the state level, which fails to capture neighborhood-level concentrations – an approach that can be explored in future research. While research documents the potential benefits of ethnic enclaves through greater availability of relevant cultural institutions, like restaurants, more work is needed to assess how such restaurants may be perceived by others in the area. Recent work examining how immigrant-run food establishments are perceived show different results, where in some cultural diversity in food store availability is viewed as a positive [31], whereas in another work these establishments are perceived to be of lower quality [32]. However, more research is still needed to examine potential associations between the availability of ethnic restaurants and food choices. The present assessment applied the COM-B in a complex, community-based setting, moving this area of research beyond mostly clinical settings and qualitative approaches in past studies. However, there are some limitations in the study. First, our measures were based on self-reports which might be biased by social desirability. Second, to keep the scale short, some aspects of the COM-B, particularly the capability measures, were assessed with only one item. In addition, knowledge was assessed through an open-ended question because we wanted to capture consumers’ interaction with a rich diversity of offerings in complex LAR settings. Future research is warranted to further validate and refine the survey tool. Third, while this study enhanced our understanding of demographic group differences in COM-B, future research should extend the research to objective measures of restaurant purchases and consumption as well as the consideration of other factors that influence eating behaviors at restaurants, such as prices, type of eating occasion, presence of others, and type of restaurant. Finally, our findings are not generalizable to the larger population. However, the distinct associations found in this initial assessment suggest that this may be a promising assessment for aiding in targeting interventions. ## Conclusion Foods away from home are an increasingly important food source, including from LARs. More research is needed to understand potential enablers and barriers for healthy eating behaviors in these settings and to develop effective multi-level interventions with the potential to positively influence dietary health. The examination of COM-B in relation to LAR-associated behaviors provides an advancement in this area of research, while also extending the application of the underlying theoretical framework. The survey developed in this research can be adapted and expanded for application in intervention design and evaluation studies, as a feasible tool that can be applied as part of other data collection efforts. Greater understanding of the role of ethnic cuisines and how consumers interact with these restaurants is critical to addressing diet-related chronic diseases and reducing health disparities. ## References 1. 1.Wolf B. Some Latin Love: Cuisine from South and Central America and the Caribbean provides plentiful menu opportunities. QSR Magazine. 2015. Available from: https://www.qsrmagazine.com/menu-innovations/some-latin-love. 2. 2.CHD Expert. 2018 Mexican restaurant industry trends report. Chicago: CHD Expert Company; 2018. 3. 3.Demeyer J. Latin American Cuisine: A Look at the South & Central American/Caribbean Menu Type. [2015, November 15, 2021]. Available from: https://www.chd-expert.com/latin-american-cuisine-a-look-at-the-south-central-american-caribbean-menu-type/. 4. Liu J, Rehm CD, Micha R, Mozaffarian D. **Quality of meals consumed by US adults at full-service and fast-food restaurants, 2003–2016: persistent low quality and widening disparities**. *J Nutr* (2020.0) **150** 873-883. DOI: 10.1093/jn/nxz299 5. 5.Choi J, Zhao J. Consumers’ behaviors when eating out. Brit Food J. 2014;116(3):494–509. 6. Fuster M, Guerrero K, Elbel B, Ray K, Huang TTK. **Engaging ethnic restaurants to improve community nutrition environments: a qualitative study with Hispanic Caribbean restaurants in New York City**. *Ecol Food Nutr* (2020.0) **59** 1-17. DOI: 10.1080/03670244.2020.1717481 7. Ray K. *The Ethnic Restaranteur* (2016.0) 8. Michie S, van Stralen MM, West R. **The behaviour change wheel: a new method for characterising and designing behaviour change interventions**. *Implement Sci* (2011.0) **6** 42. DOI: 10.1186/1748-5908-6-42 9. Beck AL, Iturralde E, Haya-Fisher J, Kim S, Keeton V, Fernandez A. **Barriers and facilitators to healthy eating among low-income Latino adolescents**. *Appetite* (2019.0) **138** 215-222. DOI: 10.1016/j.appet.2019.04.004 10. Muhwava LS, Murphy K, Zarowsky C, Levitt N. **Experiences of lifestyle change among women with gestational diabetes mellitus (GDM): A behavioural diagnosis using the COM-B model in a low-income setting**. *PLoS ONE* (2019.0) **14** e0225431. DOI: 10.1371/journal.pone.0225431 11. Willmott TJ, Pang B, Rundle-Thiele S. **Capability, opportunity, and motivation: an across contexts empirical examination of the COM-B model**. *BMC Public Health* (2021.0) **21** 1014. DOI: 10.1186/s12889-021-11019-w 12. Lydon S, Greally C, Tujjar O, Reddy K, Lambe K, Madden C. **Psychometric evaluation of a measure of factors influencing hand hygiene behaviour to inform intervention**. *J Hosp Infect* (2019.0) **102** 407-412. DOI: 10.1016/j.jhin.2019.02.003 13. 13.Pew Research Center. The American Middle Class Is Losing Ground: No longer the majority and falling behind financially. Washington, D.C; 2015. Available from: https://www.pewresearch.org/social-trends/2015/12/09/the-american-middle-class-is-losing-ground/. 14. 14.U.S. Census Bureau. Real Median Personal Income in the United States (MEPAINUSA672N). retrieved from FRED, Federal Reserve Bank of St. Louis. 2022. https://fred.stlouisfed.org/series/MEPAINUSA672N. 15. 15.Hernández-Nieto R, Gutiérrez MC. Hispanic Map of the United States. Cambridge: Instituto Cervantes at the Faculty of Arts and Sciences of Harvard University; 2017. Report No.: 035–11/2017EN. 16. Paquet C. **Environmental Influences on food behaviour**. *Int J Environ Res Public Health* (2019.0) **16** 2763. DOI: 10.3390/ijerph16152763 17. Malambo P, Kengne AP, De Villiers A, Lambert EV, Puoane T. **Built environment, selected risk factors and major cardiovascular disease outcomes: a systematic review**. *PLoS ONE* (2016.0) **11** e0166846. DOI: 10.1371/journal.pone.0166846 18. Çoker EN, Jebb SA, Stewart C, Clark M, Pechey R. **Perceptions of social norms around healthy and environmentally-friendly food choices: linking the role of referent groups to behavior**. *Front Psychol* (2022.0) **13** 974830. DOI: 10.3389/fpsyg.2022.974830 19. Higgs S. **Social norms and their influence on eating behaviours**. *Appetite* (2015.0) **86** 38-44. DOI: 10.1016/j.appet.2014.10.021 20. Anderson K, Mirosa M. **Revealing barriers to healthier fast food consumption choices**. *Brit Food J* (2014.0) **116** 821-831. DOI: 10.1108/BFJ-11-2012-0269 21. Gavrieli A, Attwood S, Wise J, Putnam-Farr E, Stillman P, Giambastiani S. **Appealing dish names to nudge diners to more sustainable food choices: a quasi-experimental study**. *BMC Public Health* (2022.0) **22** 2229. DOI: 10.1186/s12889-022-14683-8 22. Failla J. **Defining differences (Cover story)**. *Prepared Foods* (2016.0) **185** 46-51 23. Cuevas AG, Dawson BA, Williams DR. **Race and skin color in Latino health: an analytic review**. *Am J Public Health* (2016.0) **106** 2131-2136. DOI: 10.2105/AJPH.2016.303452 24. 24.Choi J, Zhao J. Consumers’ behaviors when eating out. Brit Food J. 2014;116(3):494–509. 25. Arganini C, Saba A, Comiato R, Virgili F, Turrini A, Maddock J. **Gender Differences in Food Choice and Dietary Intake in Modern Western Societies**. *Public Health - Social and Behavioral Health* (2012.0) 83-102 26. Crane MM, Tangney CC, French SA, Wang Y, Appelhans BM. **Gender comparison of the diet quality and sources of food purchases made by urban primary household food purchasers**. *J Nutr Educ Behav* (2019.0) **51** 199-204. DOI: 10.1016/j.jneb.2018.07.016 27. French SA, Tangney CC, Crane MM, Wang Y, Appelhans BM. **Nutrition quality of food purchases varies by household income: the Shopper study**. *BMC Public Health* (2019.0) **19** 231. DOI: 10.1186/s12889-019-6546-2 28. Orr C, Keyserling T, Ammerman A, Berkowitz S. **Diet quality trends among adults with diabetes by socioeconomic status in the US: 1999–2014**. *BMC Endocr Disord* (2019.0) **19** 54. DOI: 10.1186/s12902-019-0382-3 29. Drewnowski A, Rehm CD. **Energy intakes of U.S. children and adults by food purchase location and by specific food source**. *Nutr J* (2013.0) **12** 59. DOI: 10.1186/1475-2891-12-59 30. Martins MC, Diaz JE, Valiño R, Kwate NOA, Abraido-Lanza AF. **Havens of risks or resources? A study of two Latino neighborhoods in New York City**. *J Urban Health: Bulle N Y Acad Med* (2014.0) **91** 477-488. DOI: 10.1007/s11524-014-9868-5 31. Díez J, Conde P, Sandin M, Urtasun M, López R, Carrero JL. **Understanding the local food environment: a participatory photovoice project in a low-income area in Madrid**. *Spain Health Place* (2017.0) **43** 95-103. DOI: 10.1016/j.healthplace.2016.11.012 32. Rivera-Navarro J, Brey E, Franco M. **Immigration and use of public spaces and food stores in a large city: a qualitative study on urban health inequalities**. *J Migration Health* (2020.0) **1–2** 100019. DOI: 10.1016/j.jmh.2020.100019
--- title: A mixed-methods exploration of attitudes towards pregnant Facebook fitness influencers authors: - Melanie Hayman - Marian Keppel - Robert Stanton - Tanya L. Thwaite - Kristie-Lee Alfrey - Stephanie Alley - Cheryce Harrison - Shelley E. Keating - Stephanie Schoeppe - Summer S. Cannon - Lene A. H. Haakstad - Christina Gjestvang - Susan L. Williams journal: BMC Public Health year: 2023 pmcid: PMC10041693 doi: 10.1186/s12889-023-15457-6 license: CC BY 4.0 --- # A mixed-methods exploration of attitudes towards pregnant Facebook fitness influencers ## Abstract ### Background Exercise during pregnancy is associated with various health benefits for both mother and child. Despite these benefits, most pregnant women do not meet physical activity recommendations. A known barrier to engaging in exercise during pregnancy is a lack of knowledge about appropriate and safe exercise. In our current era of social media, many pregnant women are turning to online information sources for guidance, including social media influencers. Little is known about attitudes towards pregnancy exercise information provided by influencers on social media platforms. This study aimed to explore attitudes towards exercise during pregnancy depicted by social media influencers on Facebook, and user engagement with posted content. ### Methods A mixed-methods approach was used to analyse data from 10 Facebook video posts of social media influencers exercising during pregnancy. Quantitative descriptive analyses were used to report the number of views, shares, comments and emotive reactions. Qualitative analysis of user comments was achieved using an inductive thematic approach. ### Results The 10 video posts analysed were viewed a total of 12,117,200 times, shared on 11,181 occasions, included 13,455 user comments and 128,804 emotive icon reactions, with the most frequently used icon being ‘like’ ($81.48\%$). The thematic analysis identified three themes associated with attitudes including [1] exercise during pregnancy [2] influencers and [3] type of exercise. A fourth theme of community was also identified. Most user comments were associated with positive attitudes towards exercise during pregnancy and the influencer. However, attitudes towards the types of exercise the influencer performed were mixed (aerobic and body weight exercises were positive; resistance-based exercise with weights were negative). Finally, the online community perceived by users was mostly positive and recognised for offering social support and guidance. ### Conclusions User comments imply resistance-based exercise with weights as unsafe and unnecessary when pregnant, a perception that does not align with current best practice guidelines. Collectively, the findings from this study highlight the need for continued education regarding exercise during pregnancy and the potential for social media influencers to disseminate evidence-based material to pregnant women who are highly receptive to, and in need of reliable health information. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12889-023-15457-6. ## Background Exercise during pregnancy is associated with pre-, peri-, and post-partum health benefits for both mother and child [1, 2]. Despite comprehensive guidelines for exercise prescription during pregnancy [3], few pregnant women achieve adequate exercise for optimal health benefits [4–7]. Women report a range of factors that influence their engagement in physical activity during pregnancy, including physical discomforts of pregnancy, concerns about the safety of exercise, lack of motivation and/or confidence, lack of social support, and limited access to pregnancy specific exercise resources and programs [8]. Pregnancy is a time that prompts women to consider their health and associated health-related behaviours [9]. As such, this ‘health event’ may result in women adopting healthier behaviours in an attempt to improve their pregnancy outcomes [9]. To help guide these behaviour changes, pregnant women are increasingly sourcing information about exercise, nutrition, and gestational weight gain [10–12] from highly accessible online sources, including social media platforms such as Facebook and Instagram [13]. Social media has the potential to influence individual health [14, 15]. Whilst some users may actively seek out sources of information via social media platforms, others may also be exposed to information they have not explicitly sought out as a result of highly complex platform algorithms that analyse a myriad of user characteristics and online behaviours in an attempt to ‘feed’ users with highly relevant content. Although social media platforms are largely unregulated and can contribute to unhealthy behaviours due to the unintentional and intentional spread of misinformation [15], women’s use of social media has been found to be associated with increased confidence and engagement in exercise, independent of the quality of the information provided [13]. Pregnant women also use social media platforms for advice, social support, and connection with other women going through the same experience [12, 16, 17]. As a result, social media platforms have become important information sources for pregnant women, and have the potential to positively influence attitudes and practice of exercise and other health behaviours [12, 17, 18]. Increased demand and use of social media for information, support, and connection has also given rise to the phenomenon of ‘social media influencers’. A social media influencer can be described as a new genre of celebrity, an ordinary person who has gained fame and influence by sharing content (e.g., videos, photos, inspirational words) across social media platforms [19]. These platforms enable social media users to view, share, and interact with content posted by social media influencers’ via posting of comments (written and visual text) or reacting to posts with emotive icons (emojis) associated with recognised and established emotions, such as ‘like’, ‘love’, ‘care’, ‘sad’ and ‘angry’ [19]. It is common for social media influencers to be paid by companies to promote and/or endorse products [20], and social media ‘fitness influencers’ promote values or lifestyles such as health, physical activity, or wellbeing [21]. This study focussed on Facebook fitness influencers—henceforth referred to as ‘influencers’. Several recent studies highlight the ability of some influencers to promote and impact health behaviours and intentions such as healthy eating and exercise [20, 22–25]. Influencers can provide visually appealing content and may be perceived as ‘experts’ by their audiences [26]. Influencer credibility is often determined by factors such as success, attractiveness, trustworthiness, and relatability rather than use of academic literature or evidence of relevant influencer qualifications [20, 24]. In fact, research suggests that the more ‘physically attractive’ an influencer is, the greater the perceived credibility and expertise of the influencer, which in turn increases audience engagement and respect for the influencer’s content [27]. Moreover, users can develop parasocial relationships with influencers whereby users experience a sense of connectedness and express feelings of affection, encouragement, gratitude, and loyalty toward the influencer [25]. These parasocial relationships then have the potential power to also influence user behaviours, as users perceive influencers as role models. Despite influencer content being often underpinned by marketing tactics, such as the use of persuasive communication and product endorsements [24], research suggests that users view information provided by influencers as just as trustworthy and credible as information they receive from family and friends [28]. This is likely because influencers are perceived as ‘everyday people’ who are more ‘relatable’ to users, therefore users consider influencers more trustworthy and credible [28]. Thus, if the influencer is perceived as credible, users are more likely to accept the advice provided and attempt to mimic the influencer’s behaviour [22]. As such, influencers may play a role in promoting health behaviours [14], and are therefore potentially well positioned to target specific population groups such as pregnant women [29, 30]. In fact, approximately $30\%$ of women of childbearing age (between 18 and 35 years) regularly access influencer content [31]. There are several influencers who promote exercise during pregnancy. However, little is known about attitudes toward influencers and their messaging of exercise-related information. Since attitudes are an important determinant of exercise behaviour [32], and influencers have the potential to improve pregnant women’s engagement in exercise, this mixed-methods study aimed to: [1] explore user attitudes toward the influencer and the information provided, and [2] examine user engagement with information related to exercise during pregnancy (posted by influencers). ## Methods This study adopted a mixed-methods approach. Archival, user-generated data was collected from influencer posts on the social media platform Facebook between August and September 2020, and descriptively and thematically analysed. Facebook was chosen over other social media platforms, as at the time of the study, Facebook held the overwhelming market share of all social media platforms (Facebook $71.7\%$ compared to Twitter at $8.99\%$ and Instagram $7.48\%$) [33]. Additionally, video posts were chosen as the primary data source since Facebook video content is known to be more engaging than other digital content formats [34]. Ethical approval was obtained from the Central Queensland University Human Research Ethics Committee prior to commencement of the study (Approval number 2020-063). ## Data selection No previously published protocol was identified to inform data selection protocols for the present study. Hence, researchers experienced in internet and social media research were consulted and a novel two-stage data selection process was developed based on: [1] selection of influencers, and [2] selection of Facebook video content. ## Selection of influencers For the purpose of the present study, we defined an influencer as one who met the following criteria: [1] ‘Public Figure’ Facebook page, [2] minimum of 10,000 Facebook followers thus including both micro (10,000–50,000 followers) and macro (+ 50,000 followers) influencers [35], [3] was pregnant within the last five years, and [4] minimum of two videos posted on Facebook showing the influencer exercising while pregnant. Two independent researchers (MK & MH) manually searched Facebook and the internet (using Google) to identify a list of female influencers ($$n = 26$$). One researcher (MK) then screened the list of influencers for eligibility, based on these predetermined criteria. Of the 26 potentially eligible influencers, five met the inclusion criteria for the present study. See Table 1 for additional Influencer characteristics. ## Selection of Facebook video content Video posts were eligible for inclusion if the following criteria were met: [1] video content was posted within the last five years; [2] video content showed the influencer exercising while pregnant; and [3] the video post had a minimum of five user comments (including written and visual text). A minimum user comment limit was set to ensure a level of engagement with the video post and that a variety of comments would be available for analysis. Eligible video posts were then ranked in descending order of user engagement (calculated by totaling the number of emotive icon reactions, shares, and user comments on each post). The two videos with the highest engagement from each influencer were selected for data extraction. To identity potential video posts for inclusion in this study and limit the potential impact of personalization algorithms affecting the selection of video posts, two independent researchers (MK & MH) conducted a systematic search of the selected influencer’s Facebook pages using combinations of the following search terms: pregnant, pregnancy, exercise, and fitness. All results were screened for eligibility. ## Data extraction The final selection of video posts ($$n = 10$$) was assigned to nine independent researchers (MH, SC, KLA, CH, SK, SS, LH, CG, SA) for data extraction and screening. Eight researchers were allocated one video post each, and one researcher was allocated two posts (due to the low number of user comments present on each post). Each researcher was provided with a hyperlink to their assigned video post, a pre-formatted Microsoft Excel template to record the data, an instructional data extraction video, and written instructions outlining the process of accessing, extracting, recording, and screening the data [see Additional file 1]. Figure 1 outlines the process of data collection. Fig. 1Flowchart outlining the process of data collection The template recorded the following information from each included Facebook video post: hyperlink, influencer’s name, date of data extraction, type of exercise/s used in video, gestational age (where available), and any written commentary uploaded with the video post. Engagement data was also extracted, including the number of times the video was shared, the number of views the video had received, and the type and number of emotive icon reactions generated. Public user comments attached to each post were extracted by manually copying and pasting comments from the video post into the excel template. Extracted comments were screened using the following exclusion criteria: [1] non-English language comments, [2] comments with no context (such as tagging another Facebook user), [3] visual comments with no words (e.g., comment only included a photo, symbol, emotive icon reaction, and/or graphic) as the true meaning of visual comments without further context cannot be determined with any level of accuracy, and [4] comments that were not directed at the influencer, or related to the video content, exercise, or pregnancy. Importantly, interactions between the influencer and users were not explored, as this was outside of the scope of this study. Remaining data were then collated in a separate Excel spreadsheet and de-identified (influencer and commenter names were removed) in preparation for analysis. ## Data analysis An analysis of user interaction with posts on social media was conducted to provide insight into level of engagement with the content [36, 37]. For quantitative analysis, summed totals for each of the seven emotive icon reactions (like, love, care, haha, wow, sad, angry) were calculated across the selected Facebook video posts. Qualitative analysis of user comments was achieved using an inductive thematic approach, in accordance with Braun & Clarke’s [38] six-step process: [1] data familiarisation, [2] initial code generation, [3] theme identification, [4] theme review, [5] theme definition and naming, and [6] report production. Two reviewers (MK & TT) independently screened and coded the first 100 comments to generate initial codes, before coming together to review and discuss. The coding process was then completed in full by the first reviewer, with codes continually modified and refined during the process. The second reviewer was consulted during and at completion of coding until a consensus was reached, and a final codebook developed [see Additional file 2]. Codes were categorised to identify preliminary themes and sub-themes by the first reviewer, and the second reviewer was consulted to discuss and review the themes. Codes and preliminary themes were validated independently by two additional reviewers (MH & SW). Preliminary themes were then categorised into major themes until final overarching themes and sub-themes were identified. ## Results Four of the five influencers’ posts included in this study were linked to some sort of advertising, marketing or sponsorship. Two influencers included a hyperlink to their fitness app in both posts, while one influencer included references to the clothing the influencer was wearing. Another influencer included their Facebook page address in their two posts while the remaining influencer made no reference to any advertising, marketing or sponsorship in either of their posts. None of the influencers declared that their posts were sponsored, or that they had received any form of renumeration from an external source for any of their posts included in this study. See Table 1 for additional influencer characteristics below. At the time of data collection, the 10 selected video posts had been viewed a total of 12,117,200 times and were shared a total of 11,181 times. The total number of comments (including written and visual text) from each post ranged from at least 8 comments to 8,300. After removing comments that only consisted of visual text (emojis), written text comments ranged from 1 to 251. The number of emotive icon reactions generated by all video posts totalled 128,804, with the most frequently used icon being ‘like’ ($81.48\%$). Detailed Facebook user engagement information is shown in Table 2. A range of exercises including aerobic exercises (such as skipping and modified burpees), resistance-based exercises (consisting of either bodyweight exercises [squats, lunges, planks]) or exercises that encompassed weights (shoulder press, deadlift, squats with a kettlebell), were demonstrated in the video posts by the influencers. Some video posts included a combination of aerobic and resistance-based exercises. A total of 706 Facebook user comments were thematically analysed, and four overarching themes identified: [1] attitudes towards exercise during pregnancy, [2] attitudes towards the influencers, [3] attitudes towards the types of exercise, and [4] community. The first three themes characterise both positive and negative views, while the final theme, community, relates to the level of social support that was apparent among Facebook user comments. See Table 3 for a summary of themes and subthemes. Attitudes towards exercise during pregnancy in general, and attitudes towards the influencer were mostly positive. However, attitudes towards the types of exercise the influencer performed were mixed, with all negative comments found to be associated with posts that involved the influencer performing resistance-based exercises using weights. Table 1Influencer *Characteristics a* Total influencer followers according to Facebook as of 20 January 2023Influencer #Followers aPage CategoryFacebook IntroStage of pregnancyExercises within postsAdvertising / marketing / sponsorship128,000 000Fitness TrainerJoin my community of confident, healthy and fit women worldwide!3rd TrimesterBurpees, squats, bicep curls, side raises, kettlebell high pullHyperlink to influencer app3rd Trimesterreverse arm curls, resistance band high pulls, medicine ball squats, held/prolonged squat, resistance band donkey kicks’, bicep curlsHyperlink to influencer app29,800,000Public FigureHealth & Fitness Expert / Qualified PTNot discloseddeadliftsHyperlink to influencer appNot disclosedsquat, plank push up, sumo squat, fire hydrant (hip abduction), modified push up, clamHyperlink to influencer app32,000 000Public FigureBikini Pro, Fitness Model, Certified Personal Trainer, Mom & NYT BestsellingNot disclosedskipping, medicine ball, weights, battle rope, static row, boxingInfluencer FB addressNot disclosedweighted squats, kettlebell swing, jump ropeInfluencer FB address4125,000AthleteEmailInstagramFan mailing address40 weeksShoulder press, slam balls, jumping lunges, triceps extension, speed skaters, reverse flyNIL40 weeksSled push, hammer chops on tire, ball toss over the shoulders, lawnmower pulls, push-ups, Russian kettlebell swingsNIL538,800Athlete“[Influencer Name] Official Facebook Page”Instagram13 weeksparallel rows, pistol squat to reverse lunge, plank walksClothing brand14 weeksresistance band hack squat, squat to curtsyClothing brand Table 2Number of views, shares, comments, and emotive icon reactions per Facebook influencer video postPost #ViewsSharesUser comments – before screening aUser comments - after screening bEmotive icon reactionsLike (%)Love (%)Care (%)Haha (%)Wow (%)Sad (%)Angry (%)Total emotive icon reactions per post15,500,0006,2088,300 a25147,00079.18,20013.820.21190.24,0006.7330.1530.159,40726,300,0004,4004,300 a18048,00085.06,50011.5760.11,8003.2300.1540.156,460318,800883021042,00072.239014.110.036813.320.4110.42,7724239,000233214724,70081.386214.91382.440.1761.310.05,781528,700158175391,80074.552721.810.0803.340.240.22,416616,7005163311,10078.924617.640.3443.21,39471,9001247166769.82121.90.088.39683,3002036106561.93028.611.098.610594,40059210845.810845.841.7166.8236104,40069111181.02115.353.6137Totals12,117,20011,18113,455706104,95116,9051402106,40670122a Facebook restricts the free export of user comments to a maximum of 500, hence all comments for this post were unavailable to be included for screening.b All user comments were screened to remove visual texts and identify written texts for further thematic analysis. Table 3Themes and subthemes resulting from thematic analysis of Facebook user commentsThemeSubthemesAttitudes towards exercise during pregnancyHealth benefits to mother and babyExercise is beneficial and safe when pregnantExercise is unachievableAttitudes towards the influencersMotivational and inspiringRelating to the influencerComparison to selfExpertise and credibilityAttitudes towards the types of exerciseNot all exercise is safeExpertise and credibilityCommunitySocial support and guidance ## Health benefits to mother and baby *Users* generally held positive attitudes about the importance of exercising when pregnant and acknowledged the health benefits for the mother and child. Users identified improved mental health and mood, reduction in the physical discomforts of pregnancy, reduced labour complication, and healthy development of the baby as benefits of exercise during pregnancy. For many users, exercise was recognised as contributing to a quick and safe labour:You’ll find that labour is so much “easier” when you’ve been fit before and during your whole pregnancy!! It gives you the endurance you need! With my first I ran 3 miles on my due date! Users also acknowledged that exercise behaviours of the pregnant woman supported foetal development:Working out at moderate intensity is actually proven to be more beneficial. Increased blood, oxygen, and endorphins to the foetus help strengthen baby and actually help build a healthier heart in the baby. ## Exercise is beneficial and safe when pregnant Many users considered exercise during pregnancy to be an important factor in prenatal care, with one user suggesting:…it is encouraged by doctors to exercise and maintain a healthy lifestyle during pregnancy. Some users even sought advice directly from the influencer or other commenters, recognising the importance of exercise during pregnancy, and highlighting the perceived ‘expert role’ of influencers:…can I ask you what kind of exercises you did during your pregnancy? I know exercising during pregnancy is good, but I know there are specific types of exercises to avoid as they are harmful to the baby. I just wanted to know which exercises were safe. Many users alluded to the need for exercise during pregnancy to be ‘safe’ and be guided by healthcare providers, and that users should modify exercise and/or “listen” to their body. When the influencer was performing resistance-based exercises with weights and aerobic exercises, one user commented:[Influencer] is well conditioned to this exercise before her pregnancy and is using safe regressions to her training, at a very safe time in her pregnancy to be doing this training. ## Exercise is unachievable While users had positive attitudes about the influencer and the exercises depicted, they often perceived the exercises performed by influencers as unachievable for themselves and made comparisons between the capabilities of the influencer and themselves. A commonly expressed attitude by users towards the influencers was ‘she can, but I can’t’. Users also expressed negative beliefs and doubts about their own ability to partake in exercise when pregnant, due to past or current pregnancy experiences. Comparative comments were often linked to factors that prevented users from exercising such as, concerns about risk, lack of motivation, negative social influence, physical discomforts of pregnancy, health complications, lack of time, lack of knowledge, and fatigue. My second pregnancy in a row I would have been totally burned out after that far along! I was exhausted. I didn’t work out before or during my pregnancy. Even though I started under 130 lbs and only gained just under 30 lbs, I had some issues with feet and ankle swelling and my son was born early. I wish I was in better shape beforehand!…I puked 24 hrs a day 7 days a week for 8 months. I barely had the energy to walk to the bathroom so there wasn’t any exercise for me...except for vomiting...lol Overall, users held positive attitudes about the importance of exercising when pregnant, acknowledging the health benefits and need for safety when exercising during pregnancy. Users also perceived exercise an important aspect of prenatal care, and some users actively sought guidance from influencers. Some users also perceived the exercises of the influencer to be unachievable for themselves, resulting in comparisons between the capabilities of the influencer and themselves. These perceptions were often accompanied with personal user experiences often linked to factors that prevented the user from exercising. ## Motivational and inspiring Facebook users’ attitudes toward the influencer were typically positive, with influencers viewed as role models. The inspirational role of the influencer was characterised by comments providing praise and admiration:I’m almost 30 weeks and absolutely love following your workouts! You been an inspiration and this being my first pregnancy has been wonderful so far! A lot of it has to do with me staying so active! Thanks for your posts! Synonymous with expressions of praise and admiration were comments about the influencer’s physical attractiveness, strength, capabilities, and commitment to exercising while pregnant:Your dedication, positivity and beauty motivate and amaze me! Other users aspired to achieve the influencers level of fitness and the influencers physique:I wish I looked like her, she is so fit! ## Relating to the influencer Intertwined with user admiration for the influencer was a positive (albeit one-sided) relationship with the influencer. This one-sided parasocial relationship was portrayed when users expressed knowledge of intimate details about the influencer’s life (e.g., living location, children or partners’ names), aggressively defended the influencer against criticism from other users, gave personal well wishes to the influencer and their family, or found other ways to relate directly with the influencer:Congratulations, I know how you feel am just over the 26 week mark, still training but nowhere near what I use to do and I can’t wait to get back at it probably as it’s doing my head in not been able to lift my normal strength in weights lol.. x ## Expertise and credibility Influencers were mostly trusted and regarded as experts by users. This perceived expertise also meant users saw the influencers as credible sources of information. Some users commented, “She knows what she’s doing’, and “Her form is perfect and it is not like she is a beginner exerciser”, while others sought advice:Did you have a pregnancy workout daily routine we can follow? I am currently 22 weeks. I’ve never seen a modified burpee before! *This is* great! I have some joint issues and could totally do this version! Thanks for sharing. Overall, users’ attitudes toward the influencer were positive. Users considered the influencers trustworthy and credible. Comments mostly consisted of praise and admiration in addition to positive comments regarding the influencer’s physical attractiveness, strength, capabilities, and commitment to their exercise behaviours. Evidence of parasocial relationships between the user and the influencer emerged as users demonstrated connections with the influencer (e.g., citing personal details about the influencer), and defended the influencer behaviours to other users’ criticism. Negative user attitudes towards resistance-based exercises using weights during pregnancy was also associated with discreditation of the influencer’s status as an exercise professional and/or expert. Although some users appeared to admire an influencer, the same users also disapproved of the influencer’s choice in exercise:You are an amazing and smart person and a true inspiration to all out there, but this is too risky sorry. Influencer credibility was also brought into question after they chose to perform resistance-based exercise using weights when pregnant, as it was perceived to be against healthcare provider recommendations and exercise guidelines, as one user commented:I am under the impression that when you reach a certain point in pregnancy, you shouldn’t be lifting weights. My OBGYN told me I couldn’t lift weights but could do other exercises. The influencer’s motivations to exercise was also questioned and deemed as a selfish act of choosing appearance over health:…exactly figure or health of baby. Totally irresponsible. Overall, user attitudes were positive and supportive when influencers engaged in aerobic or bodyweight exercises. but users expressed concerns about the safety and need when influencers engaged in resistance-based exercises. Negative user attitudes tended to lower levels of influencer trust and credibility, with some users questioning the influencers expertise. ## Not all exercise is safe Although users expressed a desire to exercise, uncertainty about the safety of exercises during pregnancy was another barrier to performing the exercises themselves:I’m pregnant now and would love an easy list of what’s safe and when, so I can feel more confident in the exercise I do. Users held mixed attitudes towards the different types of exercise the influencer performed. When influencers engaged in aerobic or bodyweight exercises, user attitudes towards the influencer and the type of exercise the influencer was performing were positive and supportive. An example comment relating to these types of exercises was:YOUR body is MADE for burpees, so don’t sweat it. You kept doing what your body was used to. In contrast, when influencers engaged in resistance-based exercise using weights, the user attitudes were mixed. Users expressed a range of concerns and questioned the safety and necessity of the exercise being performed:...is that a good idea to be lifting heavy weights when you are this far along? Other users perceived exercises involving weights to be high risk, harmful to the baby, and contradictory to health provider recommendations:Any midwife or gyno doc will tell you don’t lift or strain during last trimester. Why risk both your lives. Specifically, almost all negative comments from users towards influencers were associated with the influencer engaging in resistance-based exercises using weights. Medical complications during pregnancy, and the advice of healthcare providers, were identified as some of the reasons why resistance-based exercises using weights was viewed as unsafe:I was absolutely huge out front and could NOT do any of these sorts of exercises whilst pregnant. Believe me I so wanted to, but genetics had other plans for me. I couldn’t even see a weight on the floor past the bump let alone pick it up. It would have been incredibly detrimental to me to have put any more strain on my abdominals. My midwives checked my abdomen throughout my pregnancy and strongly advised against anything like this. I ended up with a 6cm separation post-partum, so they were completely on the money. ## Social support and guidance A ‘community’ theme was identified however this theme is not directly related to attitudes about exercise during pregnancy. This theme suggests that by sharing personal experiences, seeking and offering advice, and supporting other users, a supportive community of users is created independent of the influencer. Within this community, users challenged attitudes and misconceptions about exercise during pregnancy and offered emotional support and guidance about exercise to each other:[user name] that’s how I was with my first. I knew that with my second pregnancy, I wanted to change that. If you have another baby in the future, definitely try to work out before and during! *Makes a* world of difference.[user name] congratulations you’re a wonderful woman, just like me, I have 7 beautiful children, 4 boys and 3 girls...hope that you’re in good health, because it’s all that matters.[user name] we live in a country with extremely poor habits. Exercise is one of the keys to long healthy life and very important while pregnant. Overall, a supportive community of users was created independent of the influencer. Users shared personal experiences, sought advice from each other and offered emotional support and guidance to one another. Attitudes and misconceptions about exercise during pregnancy were also challenged between users. ## Discussion The primary aim of this study was to explore attitudes towards influencer-posted pregnancy exercise content on social media. Various user attitudes were identified, including those towards exercise during pregnancy, the influencer, and different types of exercise. Positive and negative attitudes towards exercise during pregnancy emerged from the analysis, with acknowledgement of the positive health benefits contrasting with negative beliefs that exercises demonstrated by influencers would be unachievable for the average pregnant woman. Attitudes towards influencers were generally favourable, with users viewing them as motivational, relatable, and credible. However, attitudes towards the types of exercise performed by pregnant influencers displayed a degree of dissonance, with users perceiving aerobic or bodyweight exercises as safe for mother and child, but resistance-based exercise using weights as concerning, unnecessary, or unsafe. An additional theme of ‘community’ was also identified, highlighting the social networking/support aspect of social media platforms as found in previous studies [12, 17]. Facebook users generally expressed positive attitudes towards exercise during pregnancy. Users recognised the health benefits of exercise for the mother and baby, the importance of maintaining a level of activity during their pregnancy and expressed a desire to increase their exercise behaviours. These results are similar to findings of previous studies which report that women believe some form of exercise is essential to engage in when pregnant and beneficial for labour, mother, and child [8]. While evidence of positive attitudes towards exercise is a starting point for influencing pregnant women’s intentions to exercise [32], previous research findings of low participation in exercise during pregnancy suggests that positive attitudes alone are not sufficient to change the exercise behaviours of pregnant women [8]. Despite recognition of the importance and benefits of exercise during pregnancy, users identified several barriers to engaging in exercise similar to previous studies [8, 39]. Physical discomfort, health/pregnancy complications, lack of motivation, time constraints, limited availability of credible information, negative attitudes of family and friends, and concerns about the risks of harm to themselves or their unborn child were all identified as barriers to exercise during pregnancy. In this study, users also perceived the exercises performed by influencers as unachievable for themselves and made comparisons between the capabilities of the influencer and themselves [8, 39]. This contrasts somewhat with previous research that suggests the ability of an influencer to perform difficult or advanced exercises while pregnant inspires some users to engage in similar exercises [40]. Our findings demonstrated positive attitudes of trust in the influencer’s knowledge, with users generally perceiving influencers as having expertise in exercise during pregnancy, and commenting about the inspiration and motivation they drew from the influencer. According to the literature, perceived expertise and trustworthiness of the influencer [28] as well as personal relevance of the message and the user’s affective attitudes [41] can have a positive effect on users’ attitudes towards the influencer [20] and can also result in users being persuaded to make behaviour changes, at least in the short term [22, 41]. Further to this, we identified the presence of parasocial relationships where influencers were perceived by users as someone they felt personally and emotionally connected to and trusted as credible sources of advice about exercise during pregnancy [25]. As such, the influencer may be seen as a role model whereby users adopt exercise behaviours promoted by the influencer through the perceived trust and credibility the user feels for the influencer [25]. This was evidenced in our study by user comments that reflected praise, admiration, vehement defence of the influencer, and the suggestion of a one-sided relationship with the influencer. This combination of a positive attitude and emotional attachment to the influencer, expertise and trustworthiness, influences a person’s desire to mimic the behaviours of the influencer [22]. Therefore, pregnant women may be motivated to increase their engagement in exercise by watching positively perceived influencers exercising when they are pregnant [21]. Although influencers were perceived as experts by many, when they performed resistance-based exercises using weights (as opposed to aerobic or bodyweight exercises), their perceived expertise and credibility came into question. This attitude is consistent with literature which suggests that when influencers are not seen as trustworthy or knowledgeable, they lose credibility with their audience [24]. According to current guidelines, it is recommended that pregnant women engage in two sessions a week of muscle-strengthening resistance-based exercises at moderate intensity using bodyweight, resistance bands or light weights [42]. Despite evidence to the contrary, some users in the current study viewed resistance-based exercises using weights as a dangerous behaviour, risking harm to the mother and unborn child. Thus, with an influencer advocating for exercises perceived as unsafe to be performed during pregnancy, users can perceive them as not credible. Findings from this study are similar to previous reports which suggest that while many women understand and acknowledge the safety and health benefits of some types of exercise during pregnancy, there remains a segment of the population who are uncertain [2, 43]. Based on these current and past findings, it appears that women either do not understand the different types of resistance exercises, are not fully aware of the recommended guidelines for exercise during pregnancy, or may be receiving outdated and misguided information from healthcare professionals [44, 45]. In addition to seeking information from influencers about exercise during pregnancy, it was apparent that users in this study also experienced a sense of community and networking, using the platform as an opportunity to seek support and advice from other users, and to share common experiences. Social media is an accessible form of social support [18] that enables women to learn from each other [46]. Access to social support in like-minded communities may also reinforce social norms about exercise, and encourage women to participate in exercise during pregnancy [39, 47]. Findings of the current study indicate that the misconceptions about pregnancy safe exercises may be a contributing factor to some of the barriers that women experience, and the negative attitudes they hold about their ability to engage in exercise while pregnant. Access to credible information affects both the attitudes towards exercise during pregnancy and a pregnant woman’s perceived behavioural control, which is the confidence, resources, and ability they have to overcome the barriers they experience [48, 49]. Information about the risks, benefits, and prescription of exercise during pregnancy could improve the confidence of pregnant women, contribute to their sense of control to engage in exercise [49], improve attitudes towards exercise [32], and overall, encourage women to participate in exercise [40]. Advice received from healthcare providers about exercise during pregnancy is often conservative, likely owing to a lack of understanding or awareness on the part of the health professional [45, 50], and may not include information about the safety and efficacy of resistance-based exercise using weights [51]. This highlights an urgent need to equip healthcare providers with the knowledge, resources, and confidence to provide accurate exercise advice to encourage pregnant women to exercise [45, 50, 52]. Findings of the current study also add to the existing research that more health interventions are needed to address the information gaps about exercising during pregnancy [51], and that influencers and social media platforms may provide a possible avenue for health promoters to access wider audiences. ## Limitations To the best of the researchers’ knowledge, no other study has examined attitudes towards Facebook influencers targeting exercise during pregnancy. Previous studies have generally focused on attitudes towards exercise during pregnancy with women in offline environments [8]. Some studies have focused on exercise attitudes online [21]; however, none are specific to exercise during pregnancy and influencers. This study has some limitations which need to be highlighted. During data extraction and screening, users were deidentified, and demographic information was not collected, hence limiting the extent to which the findings are representative of pregnant women’s attitudes. Furthermore, due to the anonymity of the data, cultural backgrounds were unknown. Therefore, the generalisability of the findings and diversity across cultures is unknown. Another limitation is the choice of only one social media platform for data collection. Despite some Facebook posts having several thousand comments attached, access to the data was limited (Facebook only permits free export of 500 user comments), and only a small portion of these could be extracted for analysis. Other factors to consider when interpreting the findings include the potential presence of ‘online trolls’ who purposely leave negative comments on social media sites and the possibility that influencers may have purchased comments to boost their status and performance metrics and reach, as these factors could not be controlled for in the present study. Finally, although our findings are consistent with other studies exploring attitudes toward exercise during pregnancy, they may be confounded by sampling bias. Some comments may have come from users who were actively seeking out exercise related content online through influencers, or by users who were selectively exposed to the posts due to platform algorithms based on their prior online behaviour. Nonetheless, these users are likely to share common characteristics or behaviours, hence the reason they landed at the same content, and therefore may share similar attitudes toward the influencer. Conversely, it is possible that users with no interest in exercise related content or conflicting attitudes were not exposed to the post, and therefore were unable to provide any commentary for consideration. ## Directions for future research Very little research exists exploring the potential impact and/or influence of social media influences on exercise behaviours among pregnant women. Future studies could investigate the demographics of women who follow influencers, and how and why pregnant women engage with influencers and also other users, as this information might provide insight into how information is shared with other social networks and could also be used to develop strategies to better engage women with evidence-based information. Similar studies could also be extended to other social media platforms such as Instagram and YouTube or other specific platforms that women use to follow influencers. Moreover, future analysis could extend beyond user comments to also include the interaction between users and influencer, as influencer interaction may affect user engagement and user trust. ## Conclusion By examining the attitudes towards influencers and exercise during pregnancy on Facebook, this study found that users recognise the health benefits and importance of exercise during pregnancy. Additionally, the present study shows that women are seeking information and advice from social media influencers, who for the most part, are trusted and perceived as experts. However, despite exercise guidelines recommending that resistance-based exercise using weights is safe and beneficial during pregnancy, the perception still remains that these exercises are unsafe and unnecessary. The popularity of social media may provide a unique platform for evidence-based information dissemination. Further, health promoters should explore opportunities to engage with influencers, thus utilising their reach, perceived credibility, and expertise to disseminate evidence-based information to highly receptive and engaged audiences and positively influence the exercise behaviours of pregnant women. ## Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 ## References 1. Harrison CL, Brown WJ, Hayman M, Moran LJ, Redman LM. **The role of physical activity in preconception, pregnancy and postpartum health**. *Semin Reprod Med* (2016.0) **34** e28-e37. DOI: 10.1055/s-0036-1583530 2. Mudd LM, Pivarnik JM, Pfeiffer KA, Paneth N, Chung H, Holzman C. **Maternal physical activity during pregnancy, child leisure-time activity, and child weight status at 3 to 9 years**. *J Phys Activity Health* (2015.0) **12** 506-14. DOI: 10.1123/jpah.2013-0173 3. 3.Hayman M, Brown WJ, Brinson A, Budzynski-Seymour E, Bruce T, Evenson KR. Public health guidelines for physical activity during pregnancy from around the world: a scoping review.British Journal of Sports Medicine. 2023:bjsports-2022-105777. 4. Hayman M, Short C, Reaburn P. **An investigation into the exercise behaviours of regionally based australian pregnant women**. *J Sci Med sport* (2016.0) **19** 664-8. DOI: 10.1016/j.jsams.2015.09.004 5. Gjestland K, Bø K, Owe KM, Eberhard-Gran M. **Do pregnant women follow exercise guidelines? Prevalence data among 3482 women, and prediction of low-back pain, pelvic girdle pain and depression**. *Br J Sports Med* (2013.0) **47** 515-20. DOI: 10.1136/bjsports-2012-091344 6. Mielke GI, Crochemore-Silva I, Domingues MR, Silveira MF, Bertoldi AD, Brown WJ. **Physical activity and sitting time from 16 to 24 weeks of pregnancy to 12, 24, and 48 months Postpartum: findings from the 2015 Pelotas (Brazil) Birth Cohort Study**. *J Phys Activity Health* (2021.0) **18** 587-93. DOI: 10.1123/jpah.2020-0351 7. Hesketh KR, Evenson KR. **Prevalence of US pregnant women meeting 2015 ACOG physical activity guidelines**. *Am J Prev Med* (2016.0) **51** e87-9. DOI: 10.1016/j.amepre.2016.05.023 8. Harrison AL, Taylor NF, Shields N, Frawley HC. **Attitudes, barriers and enablers to physical activity in pregnant women: a systematic review**. *J physiotherapy* (2018.0) **64** 24-32. DOI: 10.1016/j.jphys.2017.11.012 9. Rockliffe L, Peters S, Heazell AEP, Smith DM. **Understanding pregnancy as a teachable moment for behaviour change: a comparison of the COM-B and teachable moments models**. *Health Psychol Behav Med* (2022.0) **10** 41-59. DOI: 10.1080/21642850.2021.2014851 10. Marcolino MS, Oliveira JAQ, D’Agostino M, Ribeiro AL, Alkmim MBM, Novillo-Ortiz D. **The impact of mHealth interventions: systematic review of systematic reviews**. *JMIR mHealth and uHealth* (2018.0) **6** e23. DOI: 10.2196/mhealth.8873 11. Sayakhot P, Carolan-Olah M. **Internet use by pregnant women seeking pregnancy-related information: a systematic review**. *BMC Pregnancy Childbirth* (2016.0) **16** e65. DOI: 10.1186/s12884-016-0856-5 12. Dalhaug EM, Haakstad LAH. **What the health? Information sources and maternal lifestyle behaviors**. *Interact J Med Res* (2019.0) **8** e10355. DOI: 10.2196/10355 13. Huberty J, Dinkel D, Beets MW, Coleman J. **Describing the use of the internet for health, physical activity, and nutrition information in pregnant women**. *Matern Child Health J* (2013.0) **17** 1363-72. DOI: 10.1007/s10995-012-1160-2 14. Abroms LC. **Public health in the era of social media**. *Am J Public Health Perspect* (2019.0) **109** 130-S1. DOI: 10.2105/AJPH.2018.304947 15. Chou W-YS, Oh A, Klein WMP. **Addressing health-related misinformation on social media**. *JAMA* (2018.0) **320** 2417-8. DOI: 10.1001/jama.2018.16865 16. Kraschnewski JL, Chuang CH, Poole ES, Peyton T, Blubaugh I, Pauli J. **Paging “Dr. Google”: does technology fill the gap created by the prenatal care visit structure? Qualitative focus group study with pregnant women**. *J Med Internet Res* (2014.0) **16** e147. DOI: 10.2196/jmir.3385 17. Lagan BM, Sinclair M, George Kernohan W. **Internet use in pregnancy informs women’s decision making: a web-based survey**. *Birth* (2010.0) **37** 106-15. DOI: 10.1111/j.1523-536X.2010.00390.x 18. Baker B, Yang I. **Social media as social support in pregnancy and the postpartum**. *Sex Reproductive Healthc* (2018.0) **17** 31-4. DOI: 10.1016/j.srhc.2018.05.003 19. 19.Giles DC, Edwards L. Instagram and the rise of the social media ‘influencer’. Twenty-first century celebrity: Fame in digital culture. Emerald Publishing Limited; 2018. 20. Lou C, Yuan S. **Influencer marketing: how message value and credibility affect consumer trust of branded content on social media**. *J Interact Advertising* (2019.0) **19** 58-73. DOI: 10.1080/15252019.2018.1533501 21. Sokolova K, Perez C. **You follow fitness influencers on YouTube. But do you actually exercise? How parasocial relationships, and watching fitness influencers, relate to intentions to exercise**. *J Retailing Consumer Serv* (2021.0) **58** e102276. DOI: 10.1016/j.jretconser.2020.102276 22. Ki C-WC, Cuevas LM, Chong SM, Lim H. **Influencer marketing: social media influencers as human brands attaching to followers and yielding positive marketing results by fulfilling needs**. *J Retailing Consumer Serv* (2020.0) **55** e102133. DOI: 10.1016/j.jretconser.2020.102133 23. Nafees L, Cook CM, Stoddard JE. **The impact of the Social Media Influencer Power on Consumer Attitudes toward the brand: the Mediating/Moderating Role of Social Media Influencer source credibility**. *Atl Mark J* (2020.0) **9** 3 24. Munnukka J, Uusitalo O, Toivonen H. **Credibility of a peer endorser and advertising effectiveness**. *J Consumer Mark* (2016.0) **33** 182-92. DOI: 10.1108/JCM-11-2014-1221 25. Enke N, Borchers NS. **Social media influencers in strategic communication: a conceptual framework for strategic social media influencer communication**. *Int J Strategic Communication* (2019.0) **13** 261-77. DOI: 10.1080/1553118X.2019.1620234 26. Durau J, Diehl S, Terlutter R. **Motivate me to exercise with you: the effects of social media fitness influencers on users’ intentions to engage in physical activity and the role of user gender**. *Digit Health* (2022.0) **8** 1-17. DOI: 10.1177/20552076221102769 27. Ki CWC, Kim YK. **The mechanism by which social media influencers persuade consumers: the role of consumers’ desire to mimic**. *Psychol Mark* (2019.0) **36** 905-22. DOI: 10.1002/mar.21244 28. Almahdi MH, Alsayed N, Alabbas A. *Influencers we trust? A model of trust transfer in social media influencer marketing* (2022.0) 159-73 29. Lutkenhaus RO, Jansz J, Bouman MP. **Tailoring in the digital era: stimulating dialogues on health topics in collaboration with social media influencers**. *Digit health* (2019.0) **5** 1-11. DOI: 10.1177/2055207618821521 30. Moorhead SA, Hazlett DE, Harrison L, Carroll JK, Irwin A, Hoving C. **A new dimension of health care: systematic review of the uses, benefits, and limitations of social media for health communication**. *J Med Internet Res* (2013.0) **15** e85. DOI: 10.2196/jmir.1933 31. Duplaga M. **The Use of Fitness Influencers’ Websites by Young Adult Women: a cross-sectional study**. *Int J Environ Res Public Health* (2020.0) **17** 6360. DOI: 10.3390/ijerph17176360 32. Cannella D, Lobel M, Monheit A. **Knowing is believing: information and attitudes towards physical activity during pregnancy**. *J Psychosom Obstet Gynecol* (2010.0) **31** 236-42. DOI: 10.3109/0167482X.2010.525269 33. 33.GlobalStats. Social media stats worldwide Sept 2021 - Sept 2022: https://gs.statcounter.com/social-media-stats. (Accessed Nov 2022). 34. Gabarron E, Larbi D, Dorronzoro E, Hasvold PE, Wynn R, Årsand E. **Factors engaging users of Diabetes Social Media channels on Facebook, Twitter, and Instagram: Observational Study**. *J Med Internet Res* (2020.0) **22** e21204. DOI: 10.2196/21204 35. Kim D, Kim H. **Social media influencers as human brands: an interactive marketing perspective**. *J Res Interact Mark* (2022.0) **17** 94-109. DOI: 10.1108/JRIM-08-2021-0200 36. Bai Q, Dan Q, Mu Z, Yang M. **A systematic review of emoji: current research and future perspectives**. *Front Psychol* (2019.0) **10** e2221. DOI: 10.3389/fpsyg.2019.02221 37. Kralj Novak P, Smailović J, Sluban B, Mozetič I. **Sentiment of emojis**. *PLoS ONE* (2015.0) **10** e0144296. DOI: 10.1371/journal.pone.0144296 38. Braun V, Clarke V. **Using thematic analysis in psychology**. *Qualitative Res Psychol* (2006.0) **3** 77-101. DOI: 10.1191/1478088706qp063oa 39. Coll CV, Domingues MR, Gonçalves H, Bertoldi AD. **Perceived barriers to leisure-time physical activity during pregnancy: a literature review of quantitative and qualitative evidence**. *J Sci Med sport* (2017.0) **20** 17-25. DOI: 10.1016/j.jsams.2016.06.007 40. Chana R, Haith-Cooper M. **Diet and physical activity in pregnancy: a study exploring women’s beliefs and behaviours**. *Br J Midwifery* (2019.0) **27** 297-304. DOI: 10.12968/bjom.2019.27.5.297 41. 41.Petty RE, Cacioppo JT. The elaboration likelihood model of persuasion. Communication and persuasion: Central and peripheral routes to attitude change.Springer: New York;1986.pp. 1–24. 42. Brown WJ, Hayman M, Haakstad LAH, Lamerton T, Mena GP, Green A. **Australian guidelines for physical activity in pregnancy and postpartum**. *J Sci Med Sport* (2022.0) **25** 511-9. DOI: 10.1016/j.jsams.2022.03.008 43. Duncombe D, Wertheim EH, Skouteris H, Paxton SJ, Kelly L. **Factors related to exercise over the course of pregnancy including women’s beliefs about the safety of exercise during pregnancy**. *Midwifery* (2009.0) **25** 430-8. DOI: 10.1016/j.midw.2007.03.002 44. Evenson KR, Bradley CB. **Beliefs about exercise and physical activity among pregnant women**. *Patient Educ Couns* (2010.0) **79** 124-9. DOI: 10.1016/j.pec.2009.07.028 45. Stengel MR, Kraschnewski JL, Hwang SW, Kjerulff KH, Chuang CH. **What my doctor didn’t tell me”: examining health care provider advice to overweight and obese pregnant women on gestational weight gain and physical activity**. *Womens Health Issues* (2012.0) **22** e535-40. DOI: 10.1016/j.whi.2012.09.004 46. Smith KP, Christakis NA. **Social networks and health**. *Ann Rev Sociol* (2008.0) **34** 405-29. DOI: 10.1146/annurev.soc.34.040507.134601 47. Latimer AE, Martin Ginis KA. **The importance of subjective norms for people who care what others think of them**. *Psychol Health* (2005.0) **20** 53-62. DOI: 10.1080/08870440412331300002 48. 48.Ajzen I. The theory of planned behavior. Organizational behavior and human decision processes. 1991; 50(2):179–211. doi:10.1016/0749-5978(91)90020-T. 49. De Vivo M, Hulbert S, Mills H, Uphill M. **Examining exercise intention and behaviour during pregnancy using the theory of Planned Behaviour: a meta-analysis**. *J Reproductive Infant Psychol* (2016.0) **34** 122-38. DOI: 10.1080/02646838.2015.1118022 50. Hopkinson Y, Hill DM, Fellows L, Fryer S. **Midwives understanding of physical activity guidelines during pregnancy**. *Midwifery* (2018.0) **59** 23-6. DOI: 10.1016/j.midw.2017.12.019 51. Hayman M, Reaburn P, Alley S, Cannon S, Short C. **What exercise advice are women receiving from their healthcare practitioners during pregnancy?**. *Women Birth* (2020.0) **33** e357-e62. DOI: 10.1016/j.wombi.2019.07.302 52. Hayman M, Short C, Reaburn P. **Regionally based medical practitioners may need support when prescribing exercise to pregnant women**. *Aust J Rural Health* (2017.0) **25** 62-3. DOI: 10.1111/ajr.12293
--- title: 'The cross-sectional area ratio of a specific part of the flexor pollicis longus tendon- a stable sonographic measurement for trigger thumb: a cross-sectional trial' authors: - Wenbin Zhu - Huan Zhou - Zhe Hu - Hongyan Chen - Juan Liu - Jin Li - Xiaoyuan Feng - Xueqin Li journal: BMC Musculoskeletal Disorders year: 2023 pmcid: PMC10041694 doi: 10.1186/s12891-023-06316-x license: CC BY 4.0 --- # The cross-sectional area ratio of a specific part of the flexor pollicis longus tendon- a stable sonographic measurement for trigger thumb: a cross-sectional trial ## Abstract ### Background Trigger thumb is a pathologic condition of the digital pulleys and flexor tendons. To find a cutoff value of the cross-sectional area ratio of specific parts of the flexor pollicis longus tendon to diagnosis trigger thumb in the high-frequency ultrasound examination. ### Methods We evaluated 271 healthy volunteers and 57 patients with clinical diagnosis of trigger thumb. The cross-sectional area of the metacarpophalangeal joint of flexor pollicis longus tendon (C1) and the cross-sectional area of the midpoint of the first metacarpal of flexor pollicis longus tendon (C2) were analyzed. ### Results There is no difference between gender, age and left and right hands in the ratio of C1 to C2 (C1/ C2). The mean of C1/ C2 in the healthy thumb was 0.983 ± 0.103, which was significantly smaller in comparison to the diseased thumb ($P \leq 0.05$). Based on the receiver operating characteristic curve, we chose the diagnostic cut-off value for the C1/ C2 to be 1.362 and 1.153 in order to differ a trigger thumb from children and adults. ### Conclusions The C1/ C2 of the healthy thumb was relatively stable, with a mean value of 0.983 ± 0.103. The cutoff value of C1/C2 to distinguish healthy thumb from diseased thumb in children and adults were 1.362 and 1.153, respectively. ## Background Trigger finger/ thumb is a pathologic condition of the digital pulleys and flexor tendons, with lifetime occurrence rates of $2.6\%$ in healthy individuals and $10\%$ in diabetics [1]. Two peaks in incidence occur: the first under the age of eight and the second (more common) in the fourth and fifth decades of life. When children get trigger finger/ thumb, it affects boys and girls equally and is most common in the thumb. In adults, women are much more likely to be affected by trigger finger, and typically, in their dominant hand [2]. The diagnosis is based on a characteristic clinical history and physical examination, such as flexion and extension disorders and tenderness on local palpation [3]. As an inexpensive and largely available imaging method, ultrasound can dynamically evaluate the surface structure of the hand and compare it with adjacent fingers and healthy contralateral [4]. The ultrasound findings of trigger finger/ thumb are as follows; a global or nodular hypoechoic thickening of the involved first annular (A1) pulley [5], increased thickness of flexor tendons, cysts, diffuse thickening of the synovial sheath, irregular internal echotexture, and tendon laceration [6]. Recently, Spirig et al. defined a cut-off value of the pulley thickness based on a simple-to-use value of 0.62 mm in order to distinguish between healthy and diseased A1 pulley [7]. In children with trigger thumb, no definite ultrasound abnormality of the A1 pulley has been found [8]. At the same time, it is difficult for us to check the A1 pulley with the 5–11 MHz transducer that we normally use [9]. Therefore, a more easily observed ultrasound parameter is needed to evaluate trigger thumb in children and adults. To our knowledge, pediatric trigger thumb presents with focal flexor tendon enlargement at the level of the A1 pulley [8]. The thickness of flexor tendon under the A1 pulley is related to the severity of trigger thumb [10]. Our team found that the cross-sectional area of flexor pollicis longus (FPL) tendon at the metacarpophalangeal (MCP) joint (C1) and at the midpoint of the first metacarpal bone (C2), and the ratio of C1 and C2 (C1/C2) can reflect the pathological changes of trigger thumb in children and adults [11]. Therefore, in this study, we further expanded the study sample to explore the diagnostic role of specific sectional area of FPL tendon in trigger thumb. ## Study population and clinical information This study was approved by the Ethics Committee of our hospital. High-resolution ultrasound examinations were performed by experienced ultrasound physicians with a L5-12 transducer (iU-22 system, Philips Medical Systems, Andover, MA, USA). During the examination, the patient should keep quiet, sit quietly on the chair, put hands flat on the examination bed, palm up, and keep both wrist joints, MCP joints and interphalangeal joints straight. First, we placed the transducer horizontally at the level of the MCP joint of the tested thumb, scanned the FPL tendon horizontally and continuously, and carefully confirmed that the tendon and tendon sheath have no lesions. Then, at the MCP joint and the midpoint of the first metacarpal bone, we adjusted the inspection angle of transducer to make the acoustic beam perpendicular to the FPL tendon, so as to clearly display the FPL tendon. Finally, we selected the "Zoom" key to enlarge the FPL tendon and measured the cross-sectional area of the FPL tendon by automatic wrapping method ("Trace" key)to obtain the cross-sectional area of the MCP joint of FPL tendon (C1) and the midpoint of the first metacarpal of FPL tendon (C2) were analyzed (Fig. 1). There were three experienced ultrasound physicians participating in ultrasound examination. The first step was shared by three ultrasound physicians; the last two steps were completed by a doctor who completed the C1 and C2 measurements of all the people included in the study. The study involved two groups. Fig. 1A-B Ultrasonic probe position (black line) and ultrasonic measurement image (dotted line) of the cross-sectional area (C1) of FPL tendon (white arrow) at the MCP joint. C-D Ultrasonic probe position (black line) and ultrasonic measurement image (dotted line) of the cross-sectional area (C2) of FPL tendon (white arrow) at the midpoint of the first metacarpal bone ## First group: Healthy volunteers Inclusion criteria: patients in the physical examination center of our hospital. Exclusion criteria: 1) Pregnancy or lactation; 2) Trauma history of FPL tendon; 3) History of tendon sheath diseases such as finger tendinosis, tendonitis, tenosynovitis, tendon sheath cyst, giant cell tumor of tendon sheath; 4) History of tuberculosis, skin, rheumatic immunity, metabolic or endocrine diseases, especially diabetes and gout;; 5) History of carpal tunnel syndrome; 6) Long-term use of steroids; 7) Workers or activists who need to repeatedly rub the FPL tendon sheath: such as carpenters, weightlifters, restaurant waiters, keyboard players, rock climbers or enthusiasts; 8) Left handedness; 9) Poor inspection compliance or tendon, tendon sheath or pulley disease found during the inspection. The following parameters were recorded during the evaluation: age, sex, C1 and C2 of both thumbs. 271 healthy volunteers were recruited in four different age groups (0- 6 year, 7–17 year, 18- 40 year and ≥ 41 year). The thumb of each volunteer's left and right hand was measured by ultrasound (a total of 542 thumbs). All samples were collected from March 2016 to October 2019. ( Table 1).Table 1Clinical characteristics of healthy volunteers and patients with trigger thumbCharacteristicNumber (male: female)Age (years)Normal thumb271(144: 127)24.0[1- 80]Group 164(39: 25)4.5[1- 7]Group 286(46: 40)11.0[7- 17]Group 362(30: 32)30.3[19- 40]Group 459(29: 30)57.6[41- 80]Trigger thumb65(13: 52)40.4[2- 65]Group A16(4: 12)3.8 [2- 6]Group B49(9: 40)52.4[41- 65]Data are presented as mean (range) where applicableGroup1: 0 – 6 year, Group 2: 7–17 year, Group3:18- 40 year and Group4: ≥ 41 yearGroup A: 0- 6 year) and Group B: ≥ 41 year ## Second group: Patients with trigger thumb Inclusion criteria: orthopedic doctors in our hospital diagnosed patients with trigger thumb for the first time based on typical clinical symptoms and physical examination. Diagnosis is usually made on a clinical basis, in which a painful popping or clicking sound and locking of a thumb is elicited by flexion and extension of the trigger thumb. The most common presenting sign in patients with trigger thumb is tenderness or pain over the A1 pulley. Exclusion criteria: 1) Those who have a history of trigger finger disease and have been treated with drugs or surgery before; 2)pregnancy or lactation; 3) Trauma history of FPL tendon; 4) History of tendon sheath diseases such as finger tendinosis, tendonitis, tenosynovitis, tendon sheath cyst, giant cell tumor of tendon sheath; 5) History of tuberculosis, skin, rheumatic immunity, metabolic or endocrine diseases, especially diabetes and gout;; 6) History of carpal tunnel syndrome; 7) Long-term use of steroids; 8) Left handedness; 9) Poor inspection compliance or tendon, tendon sheath or pulley disease found during the inspection. The following parameters were recorded during the evaluation: age, sex, C1 and C2 of both thumbs, and marked the diseased side. 57 patients with a clinical diagnosis of trigger thumb in our hospital from March 2012 to March 2016 were collected, and a total of 65 affected thumbs were examined by ultrasonography. The 57 patients were divided into two groups according to age, Group A (0–6 years old) and Group B (≥ 41 years old). ( Table 1) *The data* are from the patients collected in our previous trial [11]. In this trial, we regrouped them by age for statistical analysis. ## Statistical analysis According to the normality of the data, the independent two-sample t-test, the paired t-test and one-way ANOVA were used to analyze data. Descriptive statistics included frequency (percentage) and mean ± standard deviation (Mean ± SD). To determine a diagnostic cut-off value of the ratio of C1 to C2 (C1/C2), we used the receiver operating characteristic (ROC) curve, calculated from the sonographic measurements of the patients. The significance level for all statistical analyses was set at 0.05. ## Sonographic measurements of healthy volunteers In the investigated healthy volunteers, the average of C1 and C2 were 0.072 ± 0.024 cm2, 0.074 ± 0.023 cm2. ( Table 2) We found a significant difference in C1 and C2 in comparison to each different age. ( Fig. 2A) The Group 4 (≥ 41 year) had the largest values, followed by Group 3 (18- 40 year), Group 2 (7- 17 year) and finally, the Group 1 (0- 6 year). ( $P \leq 0.05$) There were differences in cross-sectional area between men and women. Males have a larger cross-sectional area than females. ( $P \leq 0.05$) (Fig. 2B) In contrast, no difference in the cross-sectional area of ​​left and right hands. ( $$P \leq 0.241$$, $$P \leq 0.132$$) (Fig. 2C). There is no difference between gender, age, and left and right hands in C1/ C2. ( Fig. 3) The average of C1/ C2 was 0.983 ± 0.103. ( Table 1). ## The relationship between clinical data and US measurements of patients In patients with trigger thumb, the average of C1, C2 and C1/C2 were no statistically different between men and women (Fig. 4A- C). The C1, C2 and C1/ C2 of ​​group A (0- 6 year) is larger than that of group B (≥ 41 year). ( $P \leq 0.05$) (Fig. 4D- F) The mean of C1/ C2 in different age groups were significantly larger than that of healthy people. ( $P \leq 0.05$) (Fig. 5). Based on the ROC curve, we chose the diagnostic cut-off value for the C1/ C2 to be 1.362 and 1.153 in order to differ a trigger thumb from children and adults. The sensitivity and specificity of this cut-off value were shown to be $98.0\%$ and $94.1\%$ in adults; $100\%$ and $100\%$ in children. ( Fig. 6). Table 2The sonographic information of measurement in normal thumb and trigger thumbC1 (cm2)C2(cm2)C1/C2Normal thumb0.072 ± 0.0240.074 ± 0.0230.983 ± 0.103Group 10.046 ± 0.0130.048 ± 0.0130.978 ± 0.100Group 20.073 ± 0.0180.075 ± 0.0180.967 ± 0.078Group 30.081 ± 0.0200.081 ± 0.0190.996 ± 0.104Group 40.092 ± 0.0220.092 ± 0.0181.000 ± 0.131Trigger thumb0.224 ± 0.3240.122 ± 0.1872.156 ± 1.362Group A0.312 ± 0.4310.146 ± 0.2223.112 ± 2.364Group B0.196 ± 0.2810.114 ± 0.1751.840 ± 0.081In each box of the measurements, the mean value ± standard deviation is described in the upper boxC1 = the cross-sectional area of metacarpophalangeal (MCP) joint of flexor pollicis longus (FPL) tendon; C2 = the cross-sectional area of the midpoint of the first metacarpal of FPL tendon; C1/C2 = the ratio of C1 to C2Group1: 0 – 6 year, Group 2: 7–17 year, Group3:18- 40 year and Group4: ≥ 41 yearGroup A: 0- 6 year) and Group B: ≥ 41 yearFig. 2In the normal thumb, comparisons of the C1 and C2 in different ages, genders, and left and right hands. The asterisk indicates a significant difference ($P \leq 0.05$)Fig. 3Comparisons of the C1/ C2 between gender, age, and left and right hands in the normal thumbFig. 4Comparisons of the C1, C2 and C1/ C2 in the trigger thumbFig. 5Comparisons of C1/ C2 in the normal thumb and trigger thumb. The asterisk indicates a significant difference ($P \leq 0.05$)Fig. 6The measurements of the C1/ C2 in normal thumb are plotted against the measurements in trigger thumb. A The ROC areas under curve of the group A (0- 6 year) pulley was 1; B the ROC areas under curve of group B (≥ 41 year) was 0.9895. ROC = Receiver operating characteristic; AUC= the area under the ROC curve ## Discussion The pathogenesis of trigger finger/thumb is not clear and is still under study. Many studies suggest that inflammation and age-related degeneration are very important in adult trigger finger/thumb. Kohler et al. found that the senescence of tendon stem cells in terms of cell size and functional adaptability may lead to tendon ageing and degeneration [12]. Therefore, the senescence of tendon stem cells caused their inability to maintain the tendon structure, thus interacting with mechanical factors, which may create a situation analogous to overuse. During this process, the balance between mechanical strain and tendon structure was disturbed, resulting in systemic inflammatory edema and mechanical (overuse) degradation of tendon. In this study, we found that C1 and C2 in the investigated healthy volunteers have significant differences between different ages. C1 and C2 showed an increasing trend with age, the Group 4 (≥ 41 year) had the largest values, which supported the above conclusion. At the same time, we also found that there was no significance between C1 and C2 of the left and right hands in the investigated healthy volunteers, which suggested that it was appropriate to use the normal contralateral thumb as the control in daily clinical work. Some studies have shown that trigger finger is more common in people with metabolic diseases (such as diabetes, hypothyroidism) [13]. In patients with diabetes, the lifetime incidence of trigger finger/thumb is $10\%$, while that of healthy people is $2.6\%$ [1]. Therefore, we exclude patients with tuberculosis, skin, rheumatic immunity, metabolic or endocrine diseases. In trigger finger/thumb, flexor tendons often become swollen and, on a transverse scan, their cross-sectional area is rounder than that of the adjacent unaffected tendons [14]. Our previous study found that the ratio (C1/C2) of the cross-sectional area of FPL tendon at the MCP joint (C1) to the cross-sectional area of the midpoint of the first metacarpal bone (C2) can be used to suggest trigger thumb [11]. Therefore, whether C1/C2 has a diagnostic cutoff value indicating trigger thumb, we further expanded the sample size for further exploration in this study. At present, studies have shown that diabetes has been identified as the main risk factor for trigger finger development, and it is also related to the duration of the disease [15]. Therefore, in the follow-up study, our team will consider analyzing the changes of C1/C2 in specific diseases and whether the diagnostic cutoff value we found is applicable. In the ultrasound examination of the FPL tendon in normal people, we found that C1/ C2 was not affected by age, gender, and the left/ right hands, with a mean value of 0.983 ± 0.103. In patients with trigger thumb, we found that C1/ C2 of both adults and children is much higher than that of normal people. This was consistent with our previous test results [11], indicating that the repeatability and operability of this parameter were very good. Moreover, there was a significant difference in C1/ C2 between different age, the children’s C1/ C2 much larger than adults. This may be related to the different pathogenesis of adults and children. Various causes of adult’s trigger finger/thumb have been proposed, including repetitive finger movements or compressive forces at the A1 pulley and repetitive local trauma [1]. However, the pediatric trigger thumb was a common thumb disease in children. It was considered to be an acquired disease because of the size mismatch between the enlarged FPL tendon and the narrow oblique pulley [16]. Through software analysis of ROC curve, we defined a cutoff value to distinguish healthy thumb from diseased thumb for adults (1.153) and children (1.362), which supplied us with reliable, clear-cut features indicating the diagnosis trigger thumb. It was worth noting that our cut-off value was $100\%$ sensitivity and specificity in diagnosing children’s trigger thumb. Our diagnostic method may play a more important role in the diagnosis of children's trigger thumb. There are several limitations to our study. First, tendon abnormalities are inconstant findings. Kim et al. found blurring or irregularity of the tendon margins, tendon sheath effusion, and loss of the normal fibrillar echogenic pattern in $62\%$, $16\%$, and $14\%$ of cases [6]. The C1/C2 may not be suitable for patients with normal flexor tendons. Second, the correlation between C1/C2 and prognosis is not involved in this study, which can be used as the next direction. Third, we did not analyze trigger finger patients with metabolic, immune, and other diseases, whose symptoms may be more serious. At the same time, we did not collect the severity and duration of patients. These data maybe very important for us to further analyze the correlation of multiple factors. The final limitation of the study arises from the unblinded sonography examiner, who knew clinically whether the involved digit was a trigger thumb or a healthy thumb. This is impossible to avoid since an experienced sonographer will diagnose this pathology sonographically as well as clinically. ## Conclusions Trigger thumb is a pathologic condition of the digital pulleys and flexor tendons. There is no clear and reliable measurement parameter to represent the flexor tendon in trigger thumb. This study suggests the cross-sectional area ratio (C1/C2) of a specific part of the flexor pollicis longus tendon in healthy people obtained by high-frequency ultrasound is relatively stable, with an average value of 0.983 ± 0.103. The C1/ C2 of patients with trigger thumb is significantly increased. In order to distinguish between healthy and diseased, we defined a cutoff value for adults (1.153) and children (1.362), which supplied us with ultrasonic parameters for diagnosing trigger thumb. ## References 1. Akhtar S, Bradley MJ, Quinton DN, Burke FD. **Management and referral for trigger finger/thumb**. *BMJ.* (2005.0) **331** 30-3. DOI: 10.1136/bmj.331.7507.30 2. 2.Jeanmonod R, Harberger S, Waseem M. Trigger Finger. [Updated 2022 Aug 7]. In: StatPearls. Treasure Island: StatPearls Publishing; 2023. Available from: https://www.ncbi.nlm.nih.gov/books/NBK459310/. 3. Nimigan AS, Ross DC, Gan BS. **Steroid injections in the management of trigger fingers**. *Am J Phys Med Rehabil.* (2006.0) **85** 36-43. DOI: 10.1097/01.phm.0000184236.81774.b5 4. Draghi F, Gitto S, Bianchi S. **Injuries to the Collateral Ligaments of the Metacarpophalangeal and Interphalangeal Joints: Sonographic Appearance**. *J Ultrasound Med.* (2018.0) **37** 2117-33. DOI: 10.1002/jum.14575 5. Guerini H, Pessis E, Theumann N. **Sonographic appearance of trigger fingers**. *J Ultrasound Med.* (2008.0) **27** 1407-13. DOI: 10.7863/jum.2008.27.10.1407 6. Kim H, Lee S. **Ultrasonographic assessment of clinically diagnosed trigger fingers**. *RHEUMATOL INT* (2010.0) **30** 1455-1458. DOI: 10.1007/s00296-009-1165-3 7. Spirig A, Juon B, Banz Y, Rieben R, Vögelin E. **Correlation Between Sonographic and In Vivo Measurement of A1 Pulleys in Trigger Fingers**. *Ultrasound Med Biol* (2016.0) **42** 1482-1490. DOI: 10.1016/j.ultrasmedbio.2016.02.017 8. Verma M, Craig CL, DiPietro MA. **Serial ultrasound evaluation of pediatric trigger thumb**. *J Pediatr Orthop.* (2013.0) **33** 309-13. DOI: 10.1097/BPO.0b013e318287f728 9. Rojo-Manaute JM, Soto VL, De Las HSJ, Del VSM, Del CM, Martín JV. **Percutaneous intrasheath ultrasonographically guided first annular pulley release: anatomic study of a new technique**. *J Ultrasound Med.* (2010.0) **29** 1517-29. DOI: 10.7863/jum.2010.29.11.1517 10. Sato J, Ishii Y, Noguchi H, Takeda M. **Sonographic Appearance of the Flexor Tendon, Volar Plate, and A1 Pulley With Respect to the Severity of Trigger Finger**. *The Journal of Hand Surgery* (2012.0) **37** 2012-2020. DOI: 10.1016/j.jhsa.2012.06.027 11. Zhu W, Chen H, Li X, Xing D. **High-frequency ultrasonographic assessment of tenosynovitis of flexor pollicis longus tendon**. *Chinese Journal of Hand Surgery* (2017.0) **33** 112-113 12. Kohler J, Popov C, Klotz B. **Uncovering the cellular and molecular changes in tendon stem/progenitor cells attributed to tendon aging and degeneration**. *AGING CELL.* (2013.0) **12** 988-99. DOI: 10.1111/acel.12124 13. Vasiliadis AV, Itsiopoulos I. **Trigger Finger: An Atraumatic Medical Phenomenon**. *J Hand Surg Asian Pac* (2017.0) **22** 188-193. DOI: 10.1142/S021881041750023X 14. Chuang XL, Ooi CC, Chin ST. **What triggers in trigger finger? The flexor tendons at the flexor digitorum superficialis bifurcation**. *J Plast Reconstr Aesthet Surg.* (2017.0) **70** 1411-9. DOI: 10.1016/j.bjps.2017.05.037 15. Kuczmarski AS, Harris AP, Gil JA, Weiss AC. **Management of Diabetic Trigger Finger**. *J Hand Surg Am* (2019.0) **44** 150-153. DOI: 10.1016/j.jhsa.2018.03.045 16. Fernandes C, Dong K, Rayan G. **Paediatric Trigger-Locked Thumb**. *J Hand Surg Asian Pac* (2022.0) **27** 2-9. DOI: 10.1142/S2424835522300018
--- title: Dietary choline and betaine intake, cardio-metabolic risk factors and prevalence of metabolic syndrome among overweight and obese adults authors: - Mohammad Sadegh Pour Abbasi - Ayda Zahiri Tousi - Yalda Yazdani - Sahar Vahdat - Farshad Gharebakhshi - Negin Nikrad - Ali Manzouri - Abnoos Mokhtari Ardekani - Faria Jafarzadeh journal: BMC Endocrine Disorders year: 2023 pmcid: PMC10041695 doi: 10.1186/s12902-023-01323-4 license: CC BY 4.0 --- # Dietary choline and betaine intake, cardio-metabolic risk factors and prevalence of metabolic syndrome among overweight and obese adults ## Abstract ### Background Choline is an important metabolite involved in phospholipids synthesis, including serum lipids, and is the immediate precursor of betaine. There are numerous studies with inconsistent results that evaluated the association between dietary choline intakes with cardiovascular risk factors. In addition, the association between dietary betaine and choline intakes with cardio-metabolic risk factors is not well studied. In the current study, our aim was to evaluate dietary choline and betaine intakes in the usual diet of obese individuals and to assess its association with serum lipids, blood pressure and glycemic markers among obese individuals. ### Methods We recruited a total number of 359 obese people aged between 20 and 50 years in the present study. A semi-quantitative food frequency questionnaire (FFQ) was used for dietary assessment; dietary choline and betaine intakes were calculated using the United States Department of Agriculture (USDA) database. National cholesterol education program adult treatment panel (NCEP-ATP)-III criteria was used metabolic syndrome (MetS) definition. Enzymatic methods were used to assess biochemical variables. Body composition was measured with the bioelectrical impedance analysis (BIA) method. ### Results Higher body mass index (BMI), waist to hip ratio (WHR), fat-free mass (FFM) and basal metabolic rate (BMR) were observed in higher tertiles of dietary choline intake ($P \leq 0.01$). There was no significant difference in terms of biochemical parameters among different tertiles of dietary choline intake, while systolic blood pressure (SBP) and diastolic blood pressure (DBP) were reduced in higher betaine tertiles ($P \leq 0.05$). For total dietary choline and betaine intakes, there was a reduction in DBP and low density lipoprotein (LDL) concentrations ($P \leq 0.05$). Also, a non-significant reduction in serum total cholesterol (TC), triglyceride (TG) and MetS prevalence was observed in higher tertiles of dietary choline and betaine intakes. After classification of the study population according to MetS status, there was no significant difference in biochemical variables in subjects with MetS ($P \leq 0.05$), while in the non-MetS group, SBP, DBP, TG and insulin levels reduced in higher tertiles of dietary betaine and choline ($P \leq 0.05$). ### Conclusion According to our findings, higher dietary intakes of choline and betaine were associated with lower levels of blood pressure and LDL concentrations among obese individuals. Further studies are warranted to confirm the results of the current study. ## Introduction Obesity is considered as one of the most important health problems worldwide and its prevalence is growing in different geographical regions [1]. The worldwide number of overweight and obese adults in 2014, was more than 1.9 billion and 600 million adults respectively [2]. Alongside of increased obesity prevalence, the occurrence of non-communicable disease (NCDs) is also increasing mostly because of changes in lifestyle and dietary behaviors [3]. In Iran, increased obesity prevalence mostly is attributed to nutrition transition and the combined prevalence of overweight and obesity may be as high as $76\%$ in some regions [4–6]. Diet, is a modifiable risk factor for chronic disease and in recent years, numerous studies have been published focusing on the role of healthy adequate diet in diet-disease relationships [7–10]. Numerous studies have focused on the relationship between single dietary ingredients (e.g. isolate effects of vitamins or minerals) [11–14], or the role of dietary patterns [15–19] dietary indices (e.g. glycemic indices, inflammatory indices, etc.) [ 17, 20–22] or herbal medicine [23–25] in developing obesity and related disorders; but, very limited number of studies have evaluated the role of dietary compounds like betaine and choline in obesity-related comorbidities. Choline and betaine are quaternary ammonium compounds that are synthesized from diet or de novo synthesis in tissues; although an insufficient diet can develop choline deficiency [26, 27]. Choline is considered as the primary source of methyl groups in the diet, and its major dietary sources are eggs, sea foods, milk, liver and beef [28], while betaine is mostly, obtained from cereals and grains, beets and spinach, shrimp, wheat germ, wheat bread, and raw mushrooms [29–31]. Choline has numerous roles in the body such as its role in membrane phospholipids, like phosphatidylcholine, choline plasmalogens, and sphingomyelin, acting as cholinergic neurotransmission, platelet-activating-factor formation, hepatic secretion of very low density lipoprotein cholesterol (VLDL), and methyl transport [32]. Choline is a potent methyl donor that produces betaine through oxidation and betaine functions as a compatible osmolyte and methyl donor in many pathways, including the homocysteine methylation [33]. Numerous studies, have investigated the beneficial effects of dietary betaine and choline on body composition or cardio-metabolic markers; the results of the studies evaluating the effects of dietary choline and betaine on anthropometric measurements like body mass index (BMI) or fat mass (FM) are inconsistent [34–37]. The results of the studies evaluating the effects of dietary choline or betaine intake on cardiovascular risk factors (e.g. blood pressure or lipid profile) are more consistent; while several studies showed that an increase in dietary choline intake was associated with a reduced prevalence of hypercholesterolemia [35] and reduced risk of ischemic stroke [38], some others showed no association between dietary betaine and choline intakes with cardiovascular disease (CVD) risk factors [39] and its incidence [40]. Other studies reported more favorable glycemic markers and lipid profile in higher dietary intakes of choline and betaine; in the study by Gao X et al. [ 41], higher dietary choline and betaine intakes were associated with lower insulin resistance. In a population-based study among 2332 male participants that was performed by Virtanen JK et al., [ 42] dietary choline and phosphatidylcholine intakes were associated with reduced diabetes risk; while in two other population-based studies, higher dietary choline and betaine intakes were associated with increased diabetes risk [37, 43]. Therefore, there is a great between-study heterogeneity regarding the association between dietary choline and betaine intakes and metabolic parameters in different studies that is possibly due to differences in the disease status or geographical distributions; moreover, no study is available to evaluate this hypothesis in obese individuals in Tabriz and Tehran cities of Iran. Obesity is the origin of numerous comorbidities and obese individuals are at greater risk of numerous diseases. Therefore, in the current study, we aimed to investigate the association between dietary choline and betaine intakes with metabolic parameters including lipid profile, glycemic markers, blood pressure and risk of metabolic syndrome among obese adults in Iran. ## Participants A cross-sectional study was conducted among 359 obese individuals in Tabriz and Tehran cities, Iran. Study subjects were invited by public announcements and were included if they met inclusion criteria (e.g. being aged 20 to 50 years old, BMI ≥ 30 kg/m2). The exclusion criteria were: being pregnant, lactating, menopause, having recent bariatric surgery, or CVD, cancer, hepatic and renal diseases, diabetes mellitus, and taking any weight-affecting medications. Full-informed approved written consent was taken from all of the participants and the study proposal was approved by the Ethics Committee of Tabriz University of Medical Sciences (Code: IR.TBZMED.REC.1401.648). ## General characteristics and anthropometric assessments Socio-demographic information including sex, age, smoking status, education attainment, marital status, occupation, medical histories, and family size were obtained via questionnaire; then, socioeconomic status (SES) score was calculated by quantifying the scores of occupation, educational status, family size and home ownership as individual indicators that were ranked from lowest to highest. Body composition measurements was done by bioelectrical impedance analysis (BIA) method (Tanita, BC-418 MA, Tokyo, Japan). Participant’s height and weight were measured using a wall-mounted stadiometer and a Seca scale (Seca co., Hamburg, Germany) to the nearest 0.5 cm and 0.1 kg respectively. Short form of the International Physical Activity Questionnaire (IPAQ) was used for physical activity assessment [44–46]. Waist circumference (WC) was measured at the midpoint between the lower costal margin and the iliac crest using a tape measure to the nearest 0.1 cm while hip circumference (HC) was measured over the widest part of the buttocks and was recorded to the nearest 0.1 cm. BMI and waist-to-hip ratio (WHR) were calculated. Blood pressure was measured with a standard mercury sphygmo-manometer twice in the same arm after at least 15 min of rest and then mean of the two measurements was used for analysis. Metabolic syndrome (MetS) was defined according to the national cholesterol education adult treatment panel (NCEP-ATP) - III criteria [47–49]. ## Dietary assessments Dietary information was collected using a validated semi-quantitative food frequency questionnaire (FFQ), adapted for Iranian population [50]. Participants were asked to report the frequency and amount of each food item consumed on a daily, weekly, monthly or yearly basis. Then, the reported frequency of consumed foods and portion sizes for each food item were converted to gram using household measures. Choline, glycero-phospho-choline, phospho-choline, phosphatidyl-choline, and betaine were calculated by multiplying each food item based on the United States Department of Agriculture (USDA) food content databases [51]. Total choline intake was calculated as the sum of choline intake from free choline, glycero-phospho-choline, phospho-choline, and phosphatidyl-choline. The sum of total choline and betaine together was used to calculate total choline and betaine intake. ## Biochemical assessment A 10 ml venous blood samples was obtained from each subject and centrifuged at 4500 rpm for 10 min to separate serum and plasma. Serum total cholesterol (TC), triglyceride (TG), high-density lipoprotein cholesterol (HDL-C), and fasting blood sugar (FBS) were evaluated using commercial kits (Pars Azmoon, Tehran, Iran). Furthermore, low-density lipoprotein cholesterol (LDL-C) level was estimated by the Friedewald equation [52]. Enzyme-linked immunosorbent assay kits were used to measure serum insulin, concentrations (Bioassay Technology Laboratory, Shanghai Korean Biotech, Shanghai City, China). Homeostatic model assessment for insulin resistance (HOMA-IR) was calculated using the formula: fasting insulin (µ IU/ml) × fasting glucose (mmol/l) /22.5 and quantitative insulin sensitivity check index (QUICKI) as 1/ [log fasting insulin (µU/mL) + log glucose (mmol/L)]. ## Statistical analyses Statistical Package for Social Sciences (version 21.0; SPSS Inc, Chicago IL) was used for data analysis. Data were represented as mean ± SD and frequency and percent for continuous and discrete quantitative variables. The comparison of continuous and discrete quantitative variables across tertiles of dietary choline, betaine and total choline and betaine intakes were performed using Chi-square and one-way analysis of variance (ANOVA) respectively. Analysis of co-variance (ANCOVA) was used for comparison of biochemical variables after adjustment for confounders (age, sex, BMI, PA, history of CVD, smoking and total energy intake). ## Results The comparison of general characteristics and anthropometric features among different tertiles of dietary choline, betaine and total choline and betaine intakes are presented in Table 1. There was a total of $57.9\%$ males and $41.5\%$ females in the current study. As shown, BMI, WHR, fat free mass (FFM) and BMR were higher in higher tertiles of dietary choline intake ($P \leq 0.01$). WHR was also higher in higher tertiles of dietary betaine intake than in lower tertile. For total dietary choline and betaine intakes, BMI, WC and FFM were higher in highest tertiles. The comparison of dietary energy and nutrient intakes across tertiles of dietary betaine and choline intakes is presented in Table 2. There was an increase in almost all of the dietary micronutrients’ intake in higher tertiles of dietary choline, betaine and total choline and betaine intakes ($P \leq 0.001$). In Table 3, the comparison of serum lipids and glycemic markers across different tertiles of dietary choline, betaine and total choline and betaine intakes is shown in Table 3. As shown, no significant difference in terms of biochemical parameters in different tertiles of dietary choline intake was observed, while there was a reduction in SBP and DBP in higher betaine tertiles ($P \leq 0.05$). For total dietary choline and betaine intakes, there was a decrease in DBP and LDL concentrations ($P \leq 0.05$). A clinically significant but statistically non-significant reduction in serum TC and TG was also observed by increased dietary choline, betaine and total choline and betaine intakes. As shown in Table 4, after classification of study population into two groups based on MetS status, no significant difference was observed in any of the biochemical variables in individuals with MetS by tertiles of dietary choline, betaine and total choline and betaine intakes (P.0.05), while in individuals without MetS, in higher tertiles of dietary choline, betaine and total choline and betaine intake, lower levels of SBP and TG were observed. In higher tertiles of dietary betaine and total choline and betaine intakes, lower levels of DBP was observed. Also, in non-MetS individuals, increased total choline and betaine intakes were accompanied with reduced serum insulin concentrations. Results of the biochemical variables were achieved after adjustment for age, BMI, physical activity level, smoking, history of CVD and total energy intake. Table 1General characteristics and anthropometric measurements of study participants across different tertiles of dietary choline, betaine and total choline and betaine intakeVariablesTotal cholineTotal betaineTotal choline and betaine1st tertile($$n = 112$$)2nd tertile($$n = 113$$)3rd tertile($$n = 113$$)P1st tertile($$n = 112$$)2nd tertile($$n = 113$$)3rd tertile($$n = 113$$)P1st tertile($$n = 112$$)2nd tertile($$n = 113$$)3rd tertile($$n = 113$$)P Age (year) 41.20 (9.54)40.41 (8.46)39.96 (9.16)0.58341.67 (9.24)40.53 (9.34)39.08(8.39)0.05641.94(9.50)40.57(8.68)39.07(8.81)0.059 Weight (kg) 89.26 (14.23)92.07(14.86)94.86 (14.03) 0.015 89.92 (16.80)92.20 (13.77)94.08(12.49)0.09988.97(15.60)91.38(13.89)95.83(13.25) 0.001 Height (cm) 165.90 (9.85)168.36 (9.78)169.44 (9.64) 0.022 167.53(10.36)168.05(9.56)167.91(9.67)0.881166.93(10.26)167.47(9.23)169.32(9.94)0.161 BMI (kg/m 2) 32.42 (4.97)32.50 (4.75)33.07 (4.85) 0.050 31.96 (5.37)32.65(4.13)33.38(4.91)0.08831.88(5.23)32.60(4.31)33.50(4.88) 0.043 WC (cm) 105.51 (9.15)106.34 (9.68)108.25 (10.02)0.092105.34 (10.51)106.82(8.73)107.95(9.59)0.127105.26(10.05)106.21(8.44)108.63(10.19) 0.026 Height (cm) 115.08 (9.37)114.82 (8.63)114.82 (9.87)0.975115.19 (9.32)113.31(9.04)116.13(9.30)0.095114.58(9.52)113.97(8.37)116.15(9.82)0.231 WHR 0.92 (0.09)0.93 (0.07)0.95 (0.07) 0.043 0.91 (0.08)0.95(0.08)0.93(0.07) 0.011 0.92(0.08)0.93(0.08)0.94(0.06)0.296 FM (kg) 34.51 (7.57)33.88 (10.57)33.12 (8.95)0.69935.74 (10.57)33.31(8.54)33.12(8.68)0.26834.24(8.01)34.43(9.80)33.04(9.29)0.626 FFM (kg) 58.82 (12.18)62.86 (12.83)64.71 (11.49) 0.026 60.86 (12.92)62.57(12.44)62.79(12.06)0.68459.16(12.31)61.50(12.75)64.80(11.672) 0.037 BMR (kcal) 7228.90 (1602.91)7996.58(1513.40)8097.51(1680.69) 0.049 7778.58(1520.56)7950.70(1473.04)7852.78(1787.09)0.8427547.92(1420.28)7755.70(1697.38)8130.95(1642.48)0.122 PA (min/week) 1653.96(2786.17)2405.45(3498.22)2371.40(3287.67)0.3542031.22(2784.99)1773.48(2654.86)2541.14(3797.19)0.3521694.892441.221793.542441.222754.252441.220.108 MetS [n(%)] Yes 49 (43.20)40 (35.39)46 (40.70)0.703*46 (41.10)52 [46]37 (32.70)0.201*48 (42.90)43 [38]44 (38.90)0.550*BMI, Body mass index; WC, Waist Circumference; WHR, waist to hip ratio; FM, Fat Mass; FFM, Fat Free Mass; BMR, Basal Metabolic Rate; PA, Physical Activity; MetS, metabolic syndrome; all data are mean (± SD) except MetS that is presented as number and percent. P values derived from One-Way ANOVA with Tukey’s post-hoc comparisons. P* values derived from chi-squared test Table 2Energy-adjusted dietary intakes of study participants across different tertiles of dietary choline, betaine and total choline and betaine intakeDietary component intakeTotal cholineTotal betaineTotal choline and betaine1st tertile($$n = 112$$)2nd tertile($$n = 113$$)3rd tertile($$n = 113$$)P1st tertile($$n = 112$$)2nd tertile($$n = 113$$)3rd tertile($$n = 113$$)P1st tertile($$n = 112$$)2nd tertile($$n = 113$$)3rd tertile($$n = 113$$)P Glycero- phospho-choline 36.10 (13.17)51.32(15.26)79.50(27.79)< 0.00146.02(19.10)54.88(25.33)66.11(30.74)< 0.00139.28(14.65)52.60(19.80)75.07(29.82)< 0.001 Phospho-choline 9.24 (3.99)12.30(3.84)18.82(6.41)< 0.00112.30(5.05)12.94(5.72)15.13(7.57)0.00210.34(4.57)12.67(4.76)17.35(7.12)< 0.001 Phosphatidyl-choline 78.59 (24.23)125.79(30.43)207.36(70.12)< 0.001108.52(44.46)139.43(76.47)164.06(74.80)< 0.00186.51(29.483)125.80(40.83)199.51(76.23)< 0.001 Sphingomyelin 6.75 (2.29)10.48(2.93)17.21(5.54)< 0.0019.55(3.97)11.92(6.41)12.99(6.18)< 0.0017.72(2.73)10.60(4.13)16.13(6.31)< 0.001 Protein (g/day) 70.82 (17.15)93.77(18.45)133.88(38.35)< 0.00181.5825(24.18)95.8602(30.48)121.6646(42.68)< 0.00172.85(19.22)94.10(21.55)131.76(39.04)< 0.001 Fat (g/day) 71.31 (26.92)93.54(34.23)136.36(51.15)< 0.00180.29(37.19)96.58(39.98)124.98(52.19)< 0.00171.6(26.38)97.30(42.80)132.55(47.99)< 0.001 Carbohydrate (g/day) 341.18)109.99)428.57(117.80)582.15(176.52)< 0.001379.20(138.95)410.65(126.43)564.70(179.39)< 0.001340.2(117.48)430.29(118.96)582.10(170.36)< 0.001 Total Fiber (g/day) 51.87 (27.32)63.39(31.45)99.03(53.08)< 0.00135.55(9.13)52.45(13.62)108.26(44.69)< 0.00137.51(10.95)58.94(21.12)104.56(48.56)< 0.001 Saturated fatty acids (mg/day) 20.20 (7.63)26.98(9.21)40.65(17.80)< 0.00125.04(12.95)28.27(12.35)34.73(17.66)< 0.00121.36(8.519)28.41(13.29)38.16(16.91)< 0.001 Iron (mg/day) 18.37 (11.05)22.11(6.32)30.83(10.31)< 0.00118.1(5.93)21.15(5.59)32.02(13.24)< 0.00116.86(5.17)22.63(10.58)31.76(9.81)< 0.001 Magnesium (mg/day) 392.22 (122.58)516.88(133.99)717.59(270.72)< 0.001457.36(146.88)510.42(171.51)660.61(294.92)< 0.001403.15(130.302)524.86(147.159)698.34(276.97)< 0.001 Zinc (mg/day) 10.36 (2.92)13.86(3.26)20.10(8.63)< 0.00112.18(3.89)14.05(4.92)18.1(9.19)< 0.00110.83(3.30)13.96(3.80)19.51(8.80)< 0.001 Phosphorus (mg/day) 1281.90(323.25)1714.54(377.19)2407.33(651.53)< 0.0011535.21(484.41)1722.86(532.91)2151.82(773.77)< 0.0011348.53(390.59)1725.58(423.680)2328.73(699.16)< 0.001 Calcium (mg/day) 887.89(285.26)1201.82(356.38)1774.44(602.80)< 0.0011059.18(425.06)1175.92(437.44)1633.38(648.53)< 0.001913.55(311.04)1210.49(395.10)1738.43(608.04)< 0.001 Potassium (mg/day) 3341.69(1169.24)4466.70(1314.78)6389.50(2193.42)< 0.0014300.97(1740.11)4456.82(1795.46)5458.71(2374.35)< 0.0013652.92(1446.56)4551.11(1596.34)5993.80(2278.87)< 0.001 VitaminB9 (µg/day) 541.49(157.50)665.22(192.78)956.52(323.49)< 0.001555.41(191.91)645.58(151.54)963.46(325.02)< 0.001516.69(149.97)659.75(153.84)984.76(307.67)< 0.001 VitaminB12 (µg/day) 3.02(2.093)5.46(7.38)7.52(6.22)< 0.0014.34(4.54)5.22(5.19)6.48(7.64)0.0273.43(2.64)5.23(6.05)7.35(7.52)< 0.001 Vitamin A (RAE/day) 557.88(289.11)895.65(738.33)1248.92(740.10)< 0.001799.14(589.25)882.21(631.58)1026.26(805.70)0.043604.39(362.23)914.70(672.12)1184.38(815.53)< 0.001 Vitamin D (µg/day) 1.33(1.05)1.85(1.26)2.91(1.69)< 0.0012.02(1.39)1.96(1.40)2.13(1.72)0.7181.60(1.187)1.92(1.39)2.58(1.73)< 0.001 Vitamin K (µg/day) 185.8814(188.38)224.9417(167.92)347.7095(321.03)< 0.001195.10(151.45)256.65(288.57)307.21(261.42)0.003163.14(135.42)241.93(195.37)352.61(324.70)< 0.001 Vitamin E (mg/day) 12.20(5.99)16.12(8.33)20.38(8.38)< 0.00113.40(6.065)15.50(7.794)19.84(9.49)< 0.00111.98(5.37)16.38(8.095)20.33(8.91)< 0.001 Table 3Biochemical parameters of study participants across different tertiles of dietary choline, betaine and total choline and betaine intakeVariablesTotal cholineTotal betaineTotal choline and betaineT1($$n = 112$$)T2($$n = 113$$)T3($$n = 113$$)P*P**T1($$n = 112$$)T2($$n = 113$$)T3($$n = 113$$)P*P**T1($$n = 112$$)T2($$n = 113$$)T3($$n = 113$$)P*P** SBP (mmHg) 123.29(15.33)122.95(14.60)121.84(18.87)0.785 < 0.001 125.63(14.99)122.61(13.42)119.86(19.58) 0.029 < 0.001 125.35(14.94)121.46(14.21)121.29(19.22)0.109 < 0.001 DBP (mmHg) 82.70(10.99)81.42(10.94)80.75(13.08)0.4510.28683.69(10.31)81.11(10.85)80.09(13.50) 0.049 0.19983.57(10.20)81.13(11.64)80.18(12.95) 0.051 0.248 TC (mg/dL) 196.12(41.08)193.06(33.31)186.22(35.15)0.1180.351195.64(40.92)191.20(35.37)188.51(33.61)0.3410.907195.79(41.88)193.18(33.02)186.41(34.49)0.1420.184 TG (mg/dL) 155.64(104.58)151.60(84.17)146.54(90.96)0.7660.065157.53(85.26)151.83(109.28)144.41(83.76)0.5730.061156.93(97.74)154.35(105.90)142.50(73.75)0.4670.059 HDL-C (mg/dL) 43.33(9.70)44.20(9.71)43.06(9.18)0.6430.39843.91(10.29)43.43(8.89)43.26(9.40)0.8690.45243.88(10.29)43.40(9.15)43.33(9.15)0.8970.466 LDL-C (mg/dL) 127.55(33.84)124.07(29.87)119.25(31.94)0.1490.157127.90(34.88)123.30(30.43)119.64(30.23)0.1520.162128.54(34.10)124.81(30.87)117.51(30.19) 0.031 0.055 Glucose (mg/dL) 90.44(12.71)92.26(14.52)95.50(27.36)0.1410.10491.78(15.67)94.63(24.31)91.85(17.22)0.4560.06690.73(15.98)94.04(23.84)93.48(17.56)0.3960.065 Insulin (µIU/mL) 15.51(10.06)16.30(10.84)16.25(17.87)0.9180.25617.11(11.37)16.43(11.52)14.85(16.20)0.5280.24715.87(10.15)17.09(12.39)15.25(16.41)0.6640.229 HOMA-IR 3.56(2.52)3.68(2.32)3.96(4.38)0.7210.2453.95(3.00)3.93(3.01)3.41(3.55)0.4480.2383.58(2.49)4.11(3.35)3.54(3.62)0.4530.233 QUICKI 0.33(0.04)0.33(0.03)0.33(0.04)0.6810.3350.32(0.03)0.33(0.04)0.33(0.03)0.2260.3800.33(0.0400.33(0.04)0.33(0.03)0.5140.302SBP, systolic blood pressure; DBP, diastolic blood pressure; TC, total cholesterol; TG, triglyceride; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; HOMA-IR, homeostatic model of insulin resistance; QUICKI, quantitative insulin sensitivity check index); P* values are obtained from ANCOVA model after adjustment for the confounding effects of age, sex, BMI and physical activity. P** values are obtained from ANCOVA model after adjustment for the confounding effects of age, sex, BMI, physical activity, history of CVD, smoking and total energy intake Table 4Biochemical parameters of study participants based on MetS status across different tertiles of dietary choline, betaine and total choline and betaine intakeSBP (mmHg)DBP (mmHg)TC (mg/dL)TG (mg/dL)HDL-C (mg/dL)LDL-C (mg/dL)Glucose (mg/dL)Insulin (µIU/mL)HOMA-IRQUICKI MetS Total cholineT1129.35 (17.82)85.16 (13.82)202.06 (39.00)179.87 (92.20)38.67 (7.90)129.74 (34.27)96.45 (17.01)16.13 (8.21)3.77 (1.83)0.32 (0.03)T2128.60 (16.41)83.26 (12.66)200.47 (33.01)201.43 (67.54)39.47 (6.37)122.68 (31.38)99.91 (21.57)18.83 (9.27)4.60 (2.24)0.31 (0.03)T3123.48 (26.34)79.18 (17.97)199.25 (40.13)164.11 (69.38)41.29 (8.08)127.22 (37.32)113.74 (45.75)21.84 (28.77)5.95 (6.88)0.31 (0.03) P* 0.3920.9970.6430.1460.0770.7190.0960.3580.2760.488Total betaineT1133.57 (18.45)85.96 (11.67)203.46 (46.68)194.84 (95.09)40.61 (10.19)130.06 (40.15)101.15 (22.90)20.10 (12.04)4.92 (3.25)0.31(0.02)T2123.15 (13.13)78.57 (11.23)197.73 (27.88)175.73 (60.02)40.07 (5.90)122.36 (28.38)108.30 (44.90)17.12 (9.36)4.67 (3.60)0.32 (0.03)T3125.06 (26.47)83.27 (19.73)200.82 (36.48)172.58 (78.84)38.75 (6.11)128.13 (34.07)100.44 (21.93)19.14 (26.71)4.63 (5.74)0.32 (0.03) P* 0.1310.4760.5090.1310.1870.5060.5030.3960.3000.447Total choline and betaineT1130.82 (19.41)84.28 (12.53)200.57 (40.25)186.39 (92.17)39.32 (8.56)126.55 (34.24)97.82 (23.95)16.67 (8.73)3.86 (1.84)0.32 (0.03)T2126.13 (13.53)83.86 (15.34)202.90 (35.59)174.54 (71.14)40.86 (7.74)130.01 (37.35)110.09 (47.23)23.21 (11.99)6.40 (4.29)0.30 (0.03)T3124.64 (25.46)80.25 (17.04)199.19 (37.00)180.03 (73.14)39.41 (6.56)125.00 (32.91)103.16 (21.47)17.59 (25.97)4.34 (5.62)0.32 (0.03) P* 0.3930.9380.6620.1390.1140.7070.1090.4970.3830.518 None-MetS Total cholineT1117.65 (11.60)76.56 (8.52)194.52 (40.03)122.43 (87.35)47.36 (8.52)124.74 (37.09)88.97 (9.42)15.08 (11.20)3.41 (2.90)0.33 (0.03)T2116.39 (13.35)77.47 (10.27)186.90 (34.41)117.91 (60.02)47.24 (10.06)121.58 (28.77)88.96 (9.33)15.34 (11.29)3.32 (2.26)0.33 (0.03)T3116.27 (16.29)77.18 (10.58)177.24 (31.45)108.15 (50.49)47.89 (8.74)112.94 (28.30)87.77 (12.41)13.64 (8.43)3.02(1.96)0.33 (0.03) P* 0.001 0.0720.222 0.001 0.0840.1490.2020.0510.0560.614Total betaineT1117.00 (12.53)78.02 (8.63)189.12 (41.43)132.75 (74.75)46.47 (10.57)122.69 (33.97)87.04 (8.84)15.52 (10.79)3.43 (2.75)0.33 (0.03)T2115.94 (12.36)77.00 (10.33)186.29 (34.08)100.21 (33.07)46.03 (8.58)122.26 (31.53)88.78 (10.02)16.07 (12.55)3.55 (2.61)0.33 (0.04)T3114.95 (16.15)76.53 (10.48)182.47 (32.01)115.10 (75.12)46.51 (8.65)114.74 (28.99)88.09 (12.03)12.93 (7.55)2.86 (1.70)0.33 (0.03) P* 0.001 0.038 0.129 0.005 0.1020.1440.1150.0710.0840.538Total choline and betaineT1117.41 (12.69)78.47 (8.70)193.06 (42.53)123.04 (63.09)48.86 (10.21)126.86 (35.55)88.00 (10.58)15.37 (10.98)3.40 (2.81)0.33 (0.03)T2114.48 (12.46)76.00 (9.90)186.93 (32.46)119.50 (79.71)47.13 (8.44)120.33 (29.86)88.48 (12.34)14.77 (11.82)3.24 (2.43)0.33 (0.03)T3116.01 (16.16)77.16 (10.69)178.78 (31.72)106.67 (51.79)48.88 (8.92)112.96 (28.41)87.27 (12.34)14.05 (8.17)3.13 (1.88)0.33 (0.03) P* < 0.001 0.030 0.148 0.002 0.0920.1300.420 0.046 0.0570.673MetS, metabolic syndrome; SBP, systolic blood pressure; DBP, diastolic blood pressure; TC, total cholesterol; TG, triglyceride; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; HOMA-IR, homeostatic model of insulin resistance; QUICKI, quantitative insulin sensitivity check index); P-values are obtained from ANCOVA model after adjustment for the confounding effects of age, sex, BMI, physical activity, history of CVD, smoking and total energy intake There was a reduction in the prevalence of MetS by increase in tertiles of dietary choline, betaine and total choline and betaine intakes among participants (Fig. 1). Fig. 1The prevalence of metabolic syndrome in different choline, betaine and total choline and betaine intake categories ($$P \leq 0.703$$, 0.201 and 0.550 respectively, by chi-square analysis) ## Discussion The results of the current study showed that higher dietary choline and betaine intakes was associated with increased BMI and WHR among obese individuals, although FFM and BMR were also greater in higher tertiles of dietary choline and betaine intakes. Moreover, reduced blood pressure and LDL concentrations and a non-significant reduction in TC and TG levels were also observed even after adjustment for the confounding effects of age, BMI, physical activity level, smoking, history of CVD and total energy intake. Similar to our findings, increased BMI and WHR by increased dietary choline intake were also observed in the study by Golzarand M et al., [ 36] and Dibaba D et al., [ 37] in general population. While in several other studies no significant difference or reduced BMI level was reported in different dietary betaine or choline categories [34, 35]. It seems that the inconsistency in results of different studies is due to difference in the general and demographic characteristics of the studies’ populations. We enrolled obese individuals and observed a difference in BMI between tertiles of dietary choline and total choline and betaine intakes after adjustment for dietary energy intake. In the study by Wu G et al. [ 53], feeding rats with choline-deficient diet led to body weight gain and reduced fat mass among eight-week-old male ob/ob mice; the observed weight gain was due to increased adipose tissue lipolytic activity and enhanced expression of active hormone-sensitive lipase by choline-deficient diet. In another study by Raubenheimer PJ et al., [ 54, 55] total weight gain after feeding choline-deficient diets in rats was lower than choline-supplemented diets. Although BMI increased, but it seems that body composition rather than BMI is a better reflection of anthropometric changes in our adult population, because increased dietary choline and betaine intakes was associated with increased FFM and BMR and a non-significant reduction in fat mas; this finding was very interesting and confirming the previous study by Gao X et al., reporting higher dietary choline and betaine intakes was associated with better body composition among the adult Canadian population [34]. Reduced blood pressure due to increased dietary betaine and total choline and betaine intakes in our study was similar to previous studies; in one population- based cross-sectional study among individuals aged more than 20 years old, dietary choline intake was inversely associated with incidence of hypertension among women [$$n = 4748$$; odds ratio (OR): 0.89; $95\%$ CI: 0.77, 1.02] [56]. In another study by Taesuwan S et al., [ 57], dietary choline intake was inversely associated with blood pressure in a cross-sectional study of National Health and Nutrition Examination Survey (NHANES). The proposed mechanisms for protective role of dietary choline and betaine against hypertension is endogenous production of a phosphatidylcholine (PC) molecule that exerts anti-hypertensive effects due to its high docosahexaenoic acid (DHA) content; it is shown that PC also reduces heart rate and improves vascular reactivity in human [57, 58]. Also, choline improves vagal activity and inhibits the inflammatory response in spontaneous hypertension and therefore, reduces the consequent cardiovascular damage in hypertension [59–61]. In our study, increased dietary choline and betaine intakes were also associated with reduced TC, TG and LDL concentrations. Although, reduced TG and TC were not statistically significant, but the reduction was clinically meaningful. Choline supplementation normalizes cholesterol metabolism and the expression of genes involved in cholesterol transport and esterification [62]. Similar to our study, in the study by Roe J et al. serum betaine but not choline was associated with favorable cardio-metabolic risk factors (e.g. lower LDL and TG) among older adults [63]. In another study choline supplementation reduced serum cholesterol and LDL concentrations in patients with type 2 diabetes mellitus (T2DM) [64]. While several other studies found a positive association between dietary choline intake or choline supplementation and serum lipids; in the study by Pary AV et al., [ 65], a weak positive association between dietary choline intake and serum LDL was reported only up to an intake of ± 250 mg/day. In an experimental model, choline deficiency reduced all kinds of serum lipids among female rats [66]. In one study, three eggs intake per day for four weeks, as the main dietary choline source, increased total cholesterol, HDL, and LDL cholesterol in healthy volunteers [67], while in another study, phosphatidylcholine supplementation in healthy humans did not alter serum cholesterol but increased TG levels [68]. These findings indicate that choline form (e.g. its biochemical structure, and dietary or supplemented choline) and dosage are important determinants of its health effects. Concerning the limitations of the current study, the study’s cross-sectional design makes it challenging to draw conclusions about causality; longitudinal investigations are required to clarify the cause-effect relationships between dietary choline and betaine intake, and cardio-metabolic risk factors. Also, we used semi-quantitative FFQ for dietary assessment that because of its subjective nature, it might stem for recall bias; however, the FFQ’s validity and reliability was confirmed in the previous studies. The multiple variables investigated as well as the relatively high number of samples are other strengths of this study. In conclusion, dietary choline and betaine intakes in obese individuals were associated with lower levels of blood pressure and low density lipoprotein (LDL) concentrations. The summarized beneficial effects of choline and betaine is presented as graphical abstract in Fig. 2. Due to great between-study heterogeneity about the health effects of dietary choline and betaine in different populations, further studies are warranted to expand these findings to different geographical distributions. Fig. 2Summarized beneficial effects of choline and betaine on blood pressure observed in the current study ## References 1. Sangsefidi ZS, Salehi-Abarghouei A, Sangsefidi ZS, Mirzaei M, Hosseinzadeh M. **The relation between low carbohydrate diet score and psychological disorders among iranian adults**. *Nutr metabolism* (2021.0) **18** 1-9. DOI: 10.1186/s12986-021-00546-3 2. 2.WHO. Obesity and overweight 2015 [Available from: http://www.who.int. 3. Popkin BM. **The nutrition transition and obesity in the developing world**. *J Nutr* (2001.0) **131** 871S-3S. DOI: 10.1093/jn/131.3.871S 4. Sh J-A, Jouyandeh Z, Qorbani M, Soroush A, Larijani B, Hasani-Ranjbar S. **Prevalence of obesity and overweight in adults and children in Iran; a systematic review**. *J Diabetes Metab Disord* (2014.0) **13** 121-7. DOI: 10.1186/s40200-014-0121-2 5. Leman MA, Claramita M, Rahayu GR. **Predicting factors on modeling health behavior: a systematic review**. *Am J Health Behav* (2021.0) **45** 268-78. DOI: 10.5993/AJHB.45.2.7 6. Gasmi A, Noor S, Piscopo S, Menzel A. **Lifestyle Genetics-Based reports in the treatment of obesity**. *Arch Razi Inst* (2021.0) **76** 707. PMID: 35096307 7. Johansson K, Neovius M, Hemmingsson E. **Effects of anti-obesity drugs, diet, and exercise on weight-loss maintenance after a very-low-calorie diet or low-calorie diet: a systematic review and meta-analysis of randomized controlled trials**. *Am J Clin Nutr* (2014.0) **99** 14-23. DOI: 10.3945/ajcn.113.070052 8. Backus R, Wara A. **Development of obesity: mechanisms and physiology**. *Veterinary Clinics: Small Animal Practice* (2016.0) **46** 773-84. PMID: 27267438 9. Valerio A, Nisoli E, Rossi AP, Pellegrini M, Todesco T, El Ghoch M. **Obesity and higher risk for severe complications of Covid-19: what to do when the two pandemics meet**. *J Popul Ther Clin Pharmacol* (2020.0) **27** e31-e6. DOI: 10.15586/jptcp.v27iSP1.708 10. 10.Ghanbari E, Asgari P, Seraj-Khorrami N. Effectiveness of Transcranial Direct Current Stimulation on Cravings in Overweight Individuals. International Journal of Body, Mind and Culture. 2022. 11. Picklo MJ, Thyfault JP. **Vitamin E and vitamin C do not reduce insulin sensitivity but inhibit mitochondrial protein expression in exercising obese rats**. *Appl Physiol Nutr Metab* (2015.0) **40** 343-52. DOI: 10.1139/apnm-2014-0302 12. Brock K, Huang W-Y, Fraser D, Ke L, Tseng M, Stolzenberg-Solomon R. **Low vitamin D status is associated with physical inactivity, obesity and low vitamin D intake in a large US sample of healthy middle-aged men and women**. *J Steroid Biochem Mol Biol* (2010.0) **121** 462-6. DOI: 10.1016/j.jsbmb.2010.03.091 13. 13.Lu L, Chen C, Yang K, Zhu J, Xun P, Shikany JM et al. Magnesium intake is inversely associated with risk of obesity in a 30-year prospective follow-up study among American young adults.European Journal of Nutrition. 2020;59(8). 14. Khodarahmi M, Asghari-Jafarabadi M, Abbasalizad Farhangi M. **A structural equation modeling approach for the association of a healthy eating index with metabolic syndrome and cardio-metabolic risk factors among obese individuals**. *PLoS ONE* (2019.0) **14** e0219193. DOI: 10.1371/journal.pone.0219193 15. Paradis A-M, Godin G, Pérusse L, Vohl M-C. **Associations between dietary patterns and obesity phenotypes**. *Int J Obes* (2009.0) **33** 1419-26. DOI: 10.1038/ijo.2009.179 16. Song Y, Park MJ, Paik H-Y, Joung H. **Secular trends in dietary patterns and obesity-related risk factors in korean adolescents aged 10–19 years**. *Int J Obes* (2010.0) **34** 48-56. DOI: 10.1038/ijo.2009.203 17. Farhangi MA, Najafi M, Jafarabadi MA, Jahangiry L. **Mediterranean dietary quality index and dietary phytochemical index among patients candidate for coronary artery bypass grafting (CABG) surgery**. *BMC Cardiovasc Disord* (2017.0) **17** 1-8. DOI: 10.1186/s12872-017-0544-z 18. Vajdi M, Farhangi MA, Nikniaz L. **Diet-derived nutrient patterns and components of metabolic syndrome: a cross-sectional community-based study**. *BMC Endocr Disorders* (2020.0) **20** 1-13. DOI: 10.1186/s12902-020-0547-0 19. Farhangi MA, Jahangiry L, Asghari-Jafarabadi M, Najafi M. **Association between dietary patterns and metabolic syndrome in a sample of tehranian adults**. *Obes Res Clin Pract* (2016.0) **10** 64-S73. DOI: 10.1016/j.orcp.2015.05.011 20. Ludwig DS. **Dietary glycemic index and obesity**. *J Nutr* (2000.0) **130** 280S-3S. DOI: 10.1093/jn/130.2.280S 21. Ruiz-Canela M, Zazpe I, Shivappa N, Hébert JR, Sánchez-Tainta A, Corella D. **Dietary inflammatory index and anthropometric measures of obesity in a population sample at high cardiovascular risk from the PREDIMED (PREvencion con DIeta MEDiterranea) trial**. *Br J Nutr* (2015.0) **113** 984-95. DOI: 10.1017/S0007114514004401 22. Farhangi MA, Najafi M. **Dietary inflammatory index: a potent association with cardiovascular risk factors among patients candidate for coronary artery bypass grafting (CABG) surgery**. *Nutr J* (2018.0) **17** 1-10. DOI: 10.1186/s12937-018-0325-2 23. Farhangi MA, Dehghan P, Tajmiri S. **Powdered black cumin seeds strongly improves serum lipids, atherogenic index of plasma and modulates anthropometric features in patients with Hashimoto’s thyroiditis**. *Lipids Health Dis* (2018.0) **17** 1-7. DOI: 10.1186/s12944-018-0704-x 24. Aslani Z, Mirmiran P, Alipur B, Bahadoran Z, Farhangi MA. **Lentil sprouts effect on serum lipids of overweight and obese patients with type 2 diabetes**. *Health Promotion Perspectives* (2015.0) **5** 215. DOI: 10.15171/hpp.2015.026 25. Tajmiri S, Farhangi MA, Dehghan P. **Nigella Sativa treatment and serum concentrations of thyroid hormones, transforming growth factor β (TGF-β) and interleukin 23 (IL-23) in patients with Hashimoto’s Thyroiditis**. *Eur J Integr Med* (2016.0) **8** 576-80. DOI: 10.1016/j.eujim.2016.03.003 26. Sanders LM, Zeisel SH. **Choline: dietary requirements and role in brain development**. *Nutr Today* (2007.0) **42** 181. DOI: 10.1097/01.NT.0000286155.55343.fa 27. da Costa K-A, Niculescu MD, Craciunescu CN, Fischer LM, Zeisel SH. **Choline deficiency increases lymphocyte apoptosis and DNA damage in humans**. *Am J Clin Nutr* (2006.0) **84** 88-94. DOI: 10.1093/ajcn/84.1.88 28. Zeisel SH, Blusztajn JK. **Choline and human nutrition**. *Ann Rev Nutr* (1994.0) **14** 269-96. DOI: 10.1146/annurev.nu.14.070194.001413 29. Vennemann FBC, Ioannidou S, Valsta LM, Dumas C, Marga C, Mensink GBM. **Dietary intake and food sources of choline in european populations**. *Brit J Nutr* (2015.0) **114** 2046-55. DOI: 10.1017/S0007114515003700 30. 30.Howe JC, Williams JR, Holden JM. USDA database for the choline content of common foods—2004; Version current 4 May 2004 [Available from: http://www.nal.usda.gov/fnic/foodcomp/data/choline/choline.html 31. PAVLOS S. **Medicinal plants against obesity: a Met-Analysis of Literature**. *J Complement Med Res* (2022.0) **12** 244. DOI: 10.5455/jcmr.2021.12.04.36 32. 32.Dietary reference intakesFor thiamin, riboflavin, niacin, vitamin B 6, folate, vitamin B 12, pantothenic acid, biotin, and choline1998editor. Washington, DCIn: Institute of Medicine NAoS, The National Academies Press. *For thiamin, riboflavin, niacin, vitamin B 6, folate, vitamin B 12, pantothenic acid, biotin, and choline* (1998.0) 33. Craig SA. **Betaine in human nutrition**. *Am J Clin Nutr* (2004.0) **80** 539-49. DOI: 10.1093/ajcn/80.3.539 34. Gao X, Wang Y, Randell E, Pedram P, Yi Y, Gulliver W. **Higher dietary choline and betaine intakes are associated with better body composition in the adult population of Newfoundland, Canada**. *PLoS ONE* (2016.0) **11** e0155403. DOI: 10.1371/journal.pone.0155403 35. Detopoulou P, Panagiotakos DB, Antonopoulou S, Pitsavos C, Stefanadis C. **Dietary choline and betaine intakes in relation to concentrations of inflammatory markers in healthy adults: the ATTICA study**. *Am J Clin Nutr* (2008.0) **87** 424-30. DOI: 10.1093/ajcn/87.2.424 36. Golzarand M, Bahadoran Z, Mirmiran P, Azizi F. **Dietary choline and betaine intake and risk of hypertension development: a 7.4-year follow-up**. *Food Funct* (2021.0) **12** 4072-8. DOI: 10.1039/D0FO03208E 37. Dibaba DT, Johnson KC, Kucharska-Newton AM, Meyer K, Zeisel SH, Bidulescu A. **The association of dietary choline and betaine with the risk of type 2 diabetes: the atherosclerosis risk in Communities (ARIC) study**. *Diabetes Care* (2020.0) **43** 2840-6. DOI: 10.2337/dc20-0733 38. Millard HR, Musani SK, Dibaba DT, Talegawkar SA, Taylor HA, Tucker KL. **Dietary choline and betaine; associations with subclinical markers of cardiovascular disease risk and incidence of CVD, coronary heart disease and stroke: the Jackson Heart Study**. *Eur J Nutr* (2018.0) **57** 51-60. DOI: 10.1007/s00394-016-1296-8 39. Dalmeijer GW, Olthof MR, Verhoef P, Bots ML, van der Schouw YT. **Prospective study on dietary intakes of folate, betaine, and choline and cardiovascular disease risk in women**. *Eur J Clin Nutr* (2008.0) **62** 386-94. DOI: 10.1038/sj.ejcn.1602725 40. Nagata C, Wada K, Tamura T, Konishi K, Kawachi T, Tsuji M. **Choline and betaine intakes are not associated with cardiovascular disease mortality risk in japanese men and women**. *J Nutr* (2015.0) **145** 1787-92. DOI: 10.3945/jn.114.209296 41. Gao X, Wang Y, Sun G. **High dietary choline and betaine intake is associated with low insulin resistance in the Newfoundland population**. *Nutrition* (2017.0) **33** 28-34. DOI: 10.1016/j.nut.2016.08.005 42. Virtanen JK, Tuomainen T-P, Voutilainen S. **Dietary intake of choline and phosphatidylcholine and risk of type 2 diabetes in men: the Kuopio Ischaemic Heart Disease risk factor study**. *Eur J Nutr* (2020.0) **59** 3857. DOI: 10.1007/s00394-020-02223-2 43. Zhou L, Li X, Li S, Wen X, Peng Y, Zhao L. **Relationship between dietary choline intake and diabetes mellitus in the National Health and Nutrition Examination Survey 2007-2010**. *J Diabetes* (2021.0) **13** 554-61. DOI: 10.1111/1753-0407.13143 44. Vasheghani-Farahani A, Tahmasbi M, Asheri H, Ashraf H, Nedjat S, Kordi R. **The Persian, last 7-day, long form of the International Physical Activity Questionnaire: translation and validation study**. *Asian J Sports Med* (2011.0) **2** 106-16. DOI: 10.5812/asjsm.34781 45. Guerra ZC, Moore JR, Londoño T, Castro Y. **Associations of acculturation and gender with obesity and physical activity among Latinos**. *Am J Health Behav* (2022.0) **46** 324-36. DOI: 10.5993/AJHB.46.3.11 46. 46.Zelenović M, Kontro T, Dumitru RC, Aksovic N, Bjelica B, Alexe DI et al. Leisure-time physical activity and all-cause mortality: A systematic review.Revista de Psicología del Deporte. 2022;31(1). 47. Grundy SM, Cleeman JI, Daniels SR, Donato KA, Eckel RH, Franklin BA. **Diagnosis and management of the metabolic syndrome: an American Heart Association/National Heart, Lung, and Blood Institute scientific statement**. *Circulation* (2005.0) **112** 2735-52. DOI: 10.1161/CIRCULATIONAHA.105.169404 48. Gallegos-Gonzalez G, Pineda-García G, Serrano-Medina A, Martinez AL, Ochoa-Ruiz E. **Association between stress and metabolic syndrome and its mediating factors in University students**. *Am J Health Behav* (2021.0) **45** 1091-102. DOI: 10.5993/AJHB.45.6.12 49. Muqdamal AH, Abdulhameed A, Ali Mansour A. **Total testosterone to estradiol ratio as a predictor marker of metabolic syndrome in males**. *Arch Razi Inst* (2022.0) **77** 351-7. PMID: 35891738 50. Mirmiran P, Esfahani FH, Mehrabi Y, Hedayati M, Azizi F. **Reliability and relative validity of an FFQ for nutrients in the Tehran lipid and glucose study**. *Public Health Nutr* (2010.0) **13** 654-62. DOI: 10.1017/S1368980009991698 51. 51.USDA Database for the Choline Content of Common Foods., Release 2 2020 [Available from: https://data.nal.usda.gov/dataset/usda-database-choline-content-common-foods-release-2-2008. 52. Rerksuppaphol L, Rerksuppaphol S. **Comparison of equations for the calculation of low-density lipoprotein cholesterol in thai population**. *J Nat Sci Biol Med* (2021.0) **12** 224-9. DOI: 10.4103/jnsbm.jnsbm_53_21 53. 53.Wu G, Zhang L, Li T, Lopaschuk G, Vance DE, Jacobs RL. Choline deficiency attenuates body weight gain and improves glucose tolerance in ob/ob mice. Journal of obesity. 2012;2012. 54. Raubenheimer PJ, Nyirenda MJ, Walker BR. **A choline-deficient diet exacerbates fatty liver but attenuates insulin resistance and glucose intolerance in mice fed a high-fat diet**. *Diabetes* (2006.0) **55** 2015-20. DOI: 10.2337/db06-0097 55. Gani IH, Al-Obaidi Z. **Molecular docking studies of tyrosine kinase inhibitors: Exemplified protocol to advance pharmaceutical education in medicinal chemistry**. *Pharm Educ* (2022.0) **22** 110-4. DOI: 10.46542/pe.2022.224.110114 56. Taesuwan S, Vermeylen F, Caudill MA, Cassano PA. **Relation of choline intake with blood pressure in the National Health and Nutrition Examination Survey 2007–2010**. *Am J Clin Nutr* (2019.0) **109** 648-55. DOI: 10.1093/ajcn/nqy330 57. 57.Taesuwan S, Thammapichai P, Ganz AB, Jirarattanarangsri W, Khemacheewakul J, Leksawasdi N. Associations of choline intake with hypertension and blood pressure among older adults in cross-sectional 2011–2014 National Health and Nutrition Examination Survey (NHANES) differ by BMI and comorbidity status.British Journal of Nutrition. 2021:1–9. 58. Gokalp G, Berksoy E, Bardak S, Demir G, Demir S, Anil M. **Is there a relationship between thyroid hormone levels and suicide attempt in adolescents?**. *Archives of Clinical Psychiatry (São Paulo)* (2021.0) **47** 130-4 59. Liu L, Lu Y, Bi X, Xu M, Yu X, Xue R. **Choline ameliorates cardiovascular damage by improving vagal activity and inhibiting the inflammatory response in spontaneously hypertensive rats**. *Sci Rep* (2017.0) **7** 42553. DOI: 10.1038/srep42553 60. Paul R, Mukkadan J. **Modulation of blood glucose, oxidative stress, and anxiety level by controlled vestibular stimulation in prediabetes**. *J Nat Sci Biol Med* (2020.0) **11** 111-7 61. Aghajani R, Nemati N, Hojjati Zidashti Z, Bagherpour T. **Effect of aerobic program in the morning and afternoon on obestatin and the body composition of overweight and obese women**. *J Chem Health Risks* (2020.0) **10** 117-25 62. Al Rajabi A, Castro GS, da Silva RP, Nelson RC, Thiesen A, Vannucchi H. **Choline supplementation protects against liver damage by normalizing cholesterol metabolism in Pemt/Ldlr knockout mice fed a high-fat diet**. *J Nutr* (2014.0) **144** 252-7. DOI: 10.3945/jn.113.185389 63. Roe AJ, Zhang S, Bhadelia RA, Johnson EJ, Lichtenstein AH, Rogers GT. **Choline and its metabolites are differently associated with cardiometabolic risk factors, history of cardiovascular disease, and MRI-documented cerebrovascular disease in older adults**. *Am J Clin Nutr* (2017.0) **105** 1283-90. DOI: 10.3945/ajcn.116.137158 64. Rashvand S, Mobasseri M, Tarighat-Esfanjani A. **Effects of choline and magnesium concurrent supplementation on coagulation and lipid profile in patients with type 2 diabetes mellitus: a pilot clinical trial**. *Biol Trace Elem Res* (2020.0) **194** 328-35. DOI: 10.1007/s12011-019-01802-7 65. 65.Van Parys A, Brække MS, Karlsson T, Vinknes KJ, Tell GS, Haugsgjerd TR et al. Assessment of Dietary Choline Intake, Contributing Food Items, and Associations with One-Carbon and Lipid Metabolites in Middle-Aged and Elderly Adults: The Hordaland Health Study.The Journal of Nutrition. 2021. 66. Tinoco J, Shannon A, Lyman RL. **Serum lipids in choline-deficient male and female rats**. *J Lipid Res* (1964.0) **5** 57-62. DOI: 10.1016/S0022-2275(20)40260-3 67. Lemos BS, Medina-Vera I, Blesso CN, Fernandez ML. **Intake of 3 eggs per day when compared to a choline bitartrate supplement, downregulates cholesterol synthesis without changing the LDL/HDL ratio**. *Nutrients* (2018.0) **10** 1-12. DOI: 10.3390/nu10020258 68. Olthof MR, Van Vliet T, Verhoef P, Zock PL, Katan MB. **Effect of homocysteine-lowering nutrients on blood lipids: results from four randomised, placebo-controlled studies in healthy humans**. *PLoS Med* (2005.0) **2** e135. DOI: 10.1371/journal.pmed.0020135
--- title: Adequacy of energy and macronutrients intake in differently active slovenian adolescents authors: - Emanuela Čerček Vilhar - Petra Golja - Gregor Starc - Barbara Koroušić Seljak - Katja Zdešar Kotnik journal: BMC Nutrition year: 2023 pmcid: PMC10041699 doi: 10.1186/s40795-023-00708-x license: CC BY 4.0 --- # Adequacy of energy and macronutrients intake in differently active slovenian adolescents ## Abstract ### Objective Evaluate the adequacy of energy/macronutrient intake in adolescents according to the Slovenian national recommendations adopted from the recommendations of the German Nutrition Society and to identify differences in energy/macronutrient intake between differently active adolescents. ### Methods Data on energy and macronutrient intake (24-hour dietary recall), physical activity (SHAPES questionnaire), and anthropometric characteristics (body mass and height) of adolescents were obtained on a representative sample of first-year secondary school students (average (SD) age: 15.3 (0.5) years; $$n = 341$$), who were included in the national survey The Analysis of Children’s Development in Slovenia (ACDSi) in $\frac{2013}{14.}$ ### Results $75\%$ of adolescents met the national recommendations for carbohydrates and proteins and $44\%$ for fats, whereas only $10\%$ of adolescents met the recommendations for energy intake. Energy/macronutrient intakes were significantly higher in vigorously physically active (VPA) boys compared to moderately (MPA) and less (LPA) physically active boys. No such differences were observed between girls of different physical activity levels. ### Conclusion Adolescents need to be encouraged to meet their energy needs according to gender and physical activity (especially VPA girls) and to reach for higher quality foods in adequate macronutrient proportions. ## Introduction Adolescence is one of the most intense periods of growth and development in human life, during which major physical and hormonal changes occur [1]. Consequently, an increased energy and macro- and micronutrients intake is required [1]. In order to maintain existing body mass, energy balance must be maintained, which is achieved when a person’s total daily energy intake equals to total daily energy expenditure [2]. Inadequate dietary intake during adolescence can lead to delayed growth, endocrine dysfunction, impaired cognitive function, sexual development disorders, and bone mass loading disorders [2], [3]. In contrast, excessive dietary intake and physical inactivity can lead to a positive energy balance and the development of overweightness and obesity [4]. Although nutritional deficiency may seem paradoxical in terms of excessive food intake, diet may include micronutrients-poor food that lacks required amounts of micronutrients and therefore overweight and obese individuals may suffer from vitamin and mineral deficiencies. Indeed, several studies have already revealed the association between obesity and micronutrient deficiencies, which have been documented for iron [5], vitamin D [6], and zinc [7] deficiency. Obesity in children and adolescents can lead to both, short term consequences such as mental (i.e. low self-esteem, eating disorders, attention deficit hyperactivity disorder, depression), and physical (i.e. high blood pressure, dyslipidaemia, hyperinsulin aemia and/or insulin resistance, asthma, type 1 diabetes, chronic inflammation) health problems, as well as long term adverse consequences on physical morbidity (cardiometabolic diseases and cancer) and thus premature mortality in adulthood [8], [9], [10], [11]. Due to rapid physical changes during adolescence, adolescents may have difficulty accepting their bodies, and they are also more susceptible to the influences of the media and their peers than adults. These are all factors that can negatively affect adolescents’ eating habits [12]. They start skipping breakfast, while snacking, eating at restaurants and fast food chains, and consuming sweet and high-energy drinks remain common [12], [13]. Reports indicate that adolescents generally consume large amounts of total and saturated fats, salt and sugar, but too little complex carbohydrates and dietary fiber [3]. Of particular concern is the fact that the majority of obese adolescents remain obese into adulthood [14]. A key role in maintaining adequate energy consumption is physical activity, which has unfortunately been decreasing in adolescents in recent years [15]. Furthermore, studies have revealed that adolescents spend growing amounts of their free time in front of screens, which results in insufficient amount of sleep, which can contribute to the development of overeating behaviors and eventually to obesity [16]. In contrast, although less frequent, malnutrition has also been observed in adolescents due to inadequate energy intake over a long period of time [17]. Studies have shown that inadequate energy and nutrient intake due to increased physical activity is most prevalent in young athletes and physically more active individuals. Among athletes, inadequate nutrient intake can increase the incidence of sports injuries, negatively affect recovery, motor performance, and sports results [18], and long-term deficiency can lead to the development of Relative energy deficiency syndrome in sport (RED-S) among athletes [19], [20]. Maintaining an adequate energy and macronutrient balance is therefore crucial for normal growth and development of adolescents and for preventing long-term health consequences. Adolescents can maintain their energy balance primarily through healthy, balanced diet and adequate extent of physical activity [2]. Thus, the aim of our study was to determine the adequacy of energy and macronutrients intake (carbohydrates, proteins, and fats) in Slovenian adolescents, relative to their daily physical activity. We expected this would allow us to identify vulnerable subgroups of adolescents in terms of nutritional intake. ## Study design The present study is part of the larger cross-sectional study The Analysis of Children’s Development in Slovenia (ACDSi) 2014, conducted every ten years in Slovenia on a representative sample of children and adolescents [21]. The protocol of the study was approved by the National Medical Ethics Committee of the Republic of Slovenia ($\frac{52}{03}$/14 and $\frac{66}{11}$/12) [21]. ## Study sample The sample was selected in a two-stage procedure with clustered and stratified sampling. In the first stage, 16 out of 170 Slovenian secondary schools were selected based on the secondary school educational programme and geographical location. In order to ensure the national representativeness of the sample not only by gender, age (about 200 males and 200 females from each of the four years of secondary school education), and geographical location, but also by secondary school programme, the ratio of sampled students from different secondary school educational programmes was maintained equal to the ratio at the national level in Slovenia (data from the Statistical Office of the Republic of Slovenia). In the second stage of sampling, a required number of classes from each school were randomly selected for participation. If the number of students from the selected classes was insufficient due to lack of their responses or other reasons, additional students from other classes in the same school were randomly included in the sample. The final ACDSi survey sample included participants who were enrolled in 15 public high schools, provided informed consent signed by their parents, and completed the survey ($$n = 1479$$). The ACDSi sample represented $5\%$ of the total national high school population and was ensured to be representative in terms of geographic location, diversity of national school programs, gender, and age. The students, who were enrolled in the first-year of secondary school educational program, were presented with an additional section on nutrition in the questionnaire. Therefore, this subsample was included in the analysis of the present study. The inclusion criteria for the final analysis was that the participants had to participate in both 24-h recalls ($$n = 341$$). ## Assessment of dietary intake Daily energy, carbohydrate, protein, and fat intakes of participating adolescents were estimated using a 24-hour recall method following the recommendations of the European Food Safety Authority (EFSA). In each school, trained interviewers conducted a 24-hour recall computer-assisted interview. In accordance with EFSA standards, the method was performed twice with the same participant, two weeks to one month apart. Interviews at each school were conducted from Monday to Friday, so food intake of a weekend day was recorded for all participants who performed interviews on Mondays. During the interview, participants were asked to indicate exactly what foods they had consumed at all meals, including all snacks and beverages, the day before the interview. To estimate the amount of food consumed, a picture book of portion sizes [22] and national household measures (e.g., 1 tablespoon of honey, 1 cup of milk, 1 glass of water, 1 piece of bread, etc.) were used. All collected data were carefully reviewed by the trained interviewer (12 different interviewers conducted 24-recall on-site) and, if necessary, additional information on food intake was obtained from the adolescents. Subsequently, the interviewer entered all data into the Slovenian web-based tool for dietary assessment named Open Platform for Clinical Nutrition [23], which was used to obtain data on the energy and macronutrient composition of foods and beverages. All the entries were double checked by 2 trained interviewers. If the foods and recipes entered into OPEN did not yield a specific energy value and/or lacked data for macronutrients, the entries were replaced with similar foods and/or recipes for which the corresponding values were available. In turn, we considered the average dietary intakes calculated by the OPEN tool from the two 24-hour recalls, which was only feasible for those individuals who completed both 24-hour recalls. The energy and micronutrient data were then exported from OPEN to the MS Excel spreadsheet for each individual for further analysis. The obtained data on estimated daily intake of carbohydrates, fats, proteins, and energy were compared with the Slovenian national reference values [24], separately for boys and girls. The same reference values were used for all adolescents in our sample: those for the age group of 15–18 years. ## Assessment of physical activity Physical activity was assessed with a questionnaire School Health Action, Planning and Evaluation System (SHAPES) [25] questionnaire, that was originally developed for the Canadian population. For the ACDSi survey [21], it was translated into Slovenian and converted into an electronic form. Prior to the beginning our measurements, we tested it for reliability and validity at one of the schools that were later on not included in the sample, and established that it was suitable for the needs of our study. Originally, the questionnaire SHAPES contain 45 questions, but in the present study only questions on physical activity were used. When completing the questionnaire, participants reported the amount of time they were vigorously (VPA) and moderately (MPA) physically active in the last seven days. From the MPA and VPA data obtained, average daily energy expenditure for physical activity (DEEPA; kcal/kg·day) was calculated using the method of Wong and Leatherdale [26], separately for each participant. Shortly, based on the data on VPA and MPA, we calculated the average daily physical activity energy expenditure (DEEPA) for each subject (calculated separately for VPA and MPA). For this purpose, the total weekly amount of physical activity was first calculated by adding the number of hours a subject spent in physical activity on seven days separately for VPA and MPA. Then the sum for VPA or MPA was divided by seven (i.e., by the number of days in the week) to obtain the average amount of VPA or MPA expressed in hours/day. DEEPA was then expressed in kilocalories per kilogram of body mass per day (kcal/kg·d) using the method of Wong and Leatherdele [26], where the average daily duration of VPA (expressed in hours) was multiplied by 6 metabolic equivalents (MET) and the average daily duration of MPA (expressed in hours) was multiplied by 3 MET, according to the equation: average DEEPA = (hours spent daily with VPA x 6MET) + (hours spent daily with MPA x 3 MET); where 1 MET = 1 kcal / kg·hour. Adolescents were then classified into three categories, based on the calculated value of DEEPA: into less active (DEEPA below the 16th percentile), moderately active (DEEPA between the 16th and 84th percentiles), and vigorously active individuals (DEEPA above the 84th percentile) [26]. ## Anthropometric measurements and body mass index (BMI) Body mass and body height data were obtained with anthropometric measurements. Measurements were performed according to standard protocols described by Lohman et al. [ 27] using SECA 769 scale (Seca Gmbh & co., Hamburg, Germany) and anthropometer GPM 101 (Siber & Hegner, Zürich, Switzerland). BMI was subsequently calculated as the ratio of body mass (kg) per square of body height (m2). Subjects were classified into categories of normal weight, overweight, obese, and underweight using the cut-off points proposed by Cole et al. [ 28], [29]. ## Data analysis Normality of the data distribution was tested using the Shapiro-Wilk test. Because of the asymmetric distribution (food intake, age…), non-parametric tests were used to test for statistical differences, and data were presented as median values and 5th and 95th percentile values. The reference values for energy intake from the national recommendations (the Slovenian national recommendation was adopted by the Nutrition Societies of Germany, Austria, and Switzerland) are estimated based on physical activity level (low, moderate, and high) [24], so we used the DEEPA calculation to classify participants into these three groups. Mann-Whitney U-test (for two independent samples) was used for: (a) gender differences in body mass index (BMI; kg/m2), average DEEPA, energy, and macronutrient intake (protein, carbohydrate, and fat); (b) gender differences in energy and macronutrient intake (protein, carbohydrate, and fat) in groups of adolescents with different DEEPA; (c) comparison of BMI and DEEPA between adolescents who met and did not meet energy intake recommendations; (d) the assessment of gender differences in compliance with the existing recommendations for energy and macronutrients intake, for all adolescents combined and also for differently active adolescents. The nonparametric Kruskal-Wallis test was performed to compare energy and macronutrient intakes between differently physically active participants (i.e., less / moderately / vigorously physically active; LPA, MPA, and VPA). A post-hoc test to assess differences between the groups was performed where Kruskal-Wallis test showed statistically significant differences. The Pearson’s chi-square test of independence for 3 × 2 contingency tables was used to compare percentages of differently active adolescents who met or did not met Slovenian national reference values (RV) for daily intake recommendations for energy and macronutrients. The p value was adjusted to 0.008 with the Bonferroni correction due to multiple comparisons. All data were analyzed using IBM SPSS Statistics 22 (IBM, Armonk, New York, USA). The level of statistical significance was set at 0.05. ## Results Our sample included 341 adolescents (girls $$n = 179$$, $52.5\%$; boys $$n = 162$$, $47.5\%$) from the first year of secondary school. The average (± standard deviation) age of the adolescents was 15.3 (± 0.5) years (range 14 to 18 years). Boys were on average taller and heavier than girls. Height was 174.9 (± 7.0) and 165.3 (± 6.5) cm, respectively ($p \leq 0.001$). Body mass was 65.5 (± 10.9) and 59.3 (± 10.9) kg, respectively ($p \leq 0.001$). According to BMI, $75.4\%$ of adolescents were of normal weight (average BMI was 21.1), $16.9\%$ were overweight (average BMI was 26.7), $3.0\%$ were obese (average BMI was 34.0), and $4.7\%$ were underweight (average BMI was 17.6). Results demonstrated that $1.8\%$ of adolescents were not physically active at all, while the most physically active adolescent had an average DEEPA (based on self-reported values) of 30.3 kJ/kg·day. The calculated median (5. – 95. percentile) values for average DEEPA was 2.3 (0.0 -3.4) in LPA adolescents, 7.6 (3.9–12.9) in MPA adolescents, and 9.5 (13.3–23.6) in VPA adolescents. Boys were significantly more physically active than girls ($p \leq 0.001$). ## Absolute dietary intake in adolescents The calculated median intake values for energy and macronutrients, physical activity extent, and BMI are presented in Table 1 relative to gender. A comparison between genders demonstrated that girls consumed significantly less energy per day than boys ($p \leq 0.001$). The intake of all macronutrients (i.e. carbohydrates, proteins, and fats) was also significantly higher in boys than in girls ($p \leq 0.001$). *In* general, higher extent of physical activity was observed in boys than in girls, however, the result was statistically significant only in the MPA group of adolescents ($$p \leq 0.002$$). There were no statistical differences between boys and girls relative to BMI in any of differently physically active groups. Table 1Average daily energy intake, macronutrient intake, physical activity extent, and body mass index in adolescentsAlln = 341Boysn = 162Girlsn = 179 M 5. − 95. percentile M 5. − 95. percentile M 5. − 95. percentile p Energy (kcal)1733937–316121121179–35621481865–25410.000Energy (kJ)72563923–13,23488434936–14,91362013622–10,6390.000Carbohydrates (g)236123–434275154–481207104–3530.000Carbohydrates (%)5443–655443–655645–650.001Proteins (g)6831–1348446–1505428–910.000Fats (g)5425–1186931–1354922–910.000Fats (%)2920–392920–402919–390.30DEEPA (kcal/kg·day)7.71.6–1.39.52.2–18.06.21.3–16.30.000DEEPA LPA ($$n = 55$$)2.30.0–3.42.30.0–3.42.30.1–3.30.50DEEPA MPA ($$n = 224$$)7.63.9–12.98.93.9–13.06.83.8–12.60.002DEEPA VPA ($$n = 56$$)9.513.3–23.615.413.3–26.116.414.4–21.30.13BMI (kg/m2)20.717.3–27.720.717.3–26.720.817.3–28.40.55underweight ($$n = 13$$)16.914.8 - NA16.816.8 - NA16.814.8 - NA0.41normal weight ($$n = 259$$)20.317.8–23.420.217.6–23.120.417.8–23.60.31overweight ($$n = 55$$)25.623.6–28.424.723.4–27.426.123.7–28.50.03obese ($$n = 11$$)31.628.4 - NA33.231.6 - NA30.828.4 - NA0.32n, number of participants; M, median; NA, not available due to small sample; g, gram; kcal, kilocalorie; kJ, kilojoule; DEEPA, average daily energy expenditure for physical activity; LPA, less physically active; MPA, moderately physically active; VPA, vigorously physically active; BMI, body mass index; significance results obtained with Mann-Whitney U-test for two independent samples between boys and girls are presented with exact p-values As demonstrated in Table 2, the differently active boys had significantly different intake of energy ($$p \leq 0.02$$), carbohydrates ($$p \leq 0.03$$), proteins ($$p \leq 0.002$$), and fats ($$p \leq 0.02$$)). A more detailed post-hoc analysis demonstrated that VPA boys consumed significantly more energy compared to MPA boys ($$p \leq 0.01$$). There was no significant difference in energy intake between LPA and VPA boys ($$p \leq 0.21$$) and between LPA and MPA boys ($$p \leq 1.0$$). The same was true for carbohydrate intake ($$p \leq 0.03$$), although a larger intake was observed ($$p \leq 0.02$$) in MPA than in VPA boys. Differences between groups of differently physically active boys were also found in protein intake, namely, VPA boys consumed more proteins than LPA boys ($$p \leq 0.005$$), as well as MPA boys ($$p \leq 0.01$$). VPA boys consumed more fat than MPA boys ($$p \leq 0.02$$), while there was no significant difference in fat intake between LPA and VPA boys ($$p \leq 0.91$$) and MPA and LPA boys ($$p \leq 1.0$$). In girls, no statistically significant differences were found in energy ($$p \leq 0.91$$), carbohydrate ($$p \leq 0.83$$), protein ($$p \leq 0.96$$), or fat ($$p \leq 0.90$$) intake between girls with different physical activity extent. Additional gender comparisons revealed that the energy and all macronutrients (i.e. carbohydrates, proteins, and fats) intake was significantly higher in boys than in girls in all groups of physical activity ($p \leq 0.005$). Interestingly, there were no statistically significant differences in BMI between differently active adolescents. Table 2Body mass index, energy, and macronutrients intake in differently physically active adolescents. n, number of subjects; LPA, less physically active; MPA, moderately physically active; VPA, vigorously physically active adolescents; BMI, body mass index; M, median; significance results obtained with Kruskal-Wallis test between differently physically active adolescents are presented with exact p-valuesBMI (kg/m2)Energy (kcal)Carbohydrates (g)Proteins (g)Fats (g) M 5. − 95. percentile p M 5. − 95. percentile p M 5. − 95. percentile p M 5. − 95. percentile p M 5. − 95. percentile p All; $$n = 338$$ LPA ($$n = 55$$)20.517.0–35.00.1216541007–30170.000236123–3700.0026034–960.0005329–1090.01MPA ($$n = 224$$)20.917.5–27.61686909–3022230121–4216731–1275325–109VPA ($$n = 56$$)20.517.2–26.921101023–4142277122–5538323–1656924–142 Boys; $$n = 162$$ LPA ($$n = 19$$)20.416.8–24.60.4021051111–31690.02276137–4080.037346–980.00270312 − 1140.02MPA ($$n = 100$$)21.017.3–26.720521117–3222262140–4768145–1356630–127VPA ($$n = 40$$)20.517.3–27.023821332–4517304196–6339141–1838335–151 Girls; $$n = 176$$ LPA ($$n = 36$$)20.815.9–38.50.471498936–24680.91212104–3520.835527–910.964729–860.90MPA ($$n = 124$$)20.917.4–28.31470873–2588204112–3475528–934922–92VPA ($$n = 16$$)20.415.9–25.31589638–280421489–3765519–935122–105n, number of subjects; LPA, less physically active; MPA, moderately physically active; VPA, vigorously physically active adolescents; BMI, body mass index; M, median; significance results obtained with Kruskal-Wallis test between differently physically active adolescents are presented with exact p-values ## Meeting national recommendations for dietary intake in adolescents A comparison of adolescent energy and macronutrient intake with national recommendations [24] is presented as the percentage of adolescents who met / did not meet daily recommended value for energy and all macronutrients (carbohydrates, proteins, and fats) according to gender (Table 3) and physical activity (Table 4). Table 3Percentage of adolescents, who met/did not meet reference values (RV) for daily energy and nutrient intakeBoys($$n = 162$$)Girls($$n = 179$$)* Met RV Did not meet RV Met RV Did not meet RV pn%n%n%n% Energy 2012.314287.7158.516191.50.17 Carbohydrates 11470.44829.614279.33720.70.06 Proteins 13482.72817.312067.05933.00.001 Fats 7546.38753.77541.910458.10.41n, number of subjects; *, number of girls for energy intake is 176 due to missing data for physical activity; RV, reference values; significance results obtained with Pearson’s chi-square test between the percentage of boys and girls, who met or did not meet RV for daily energy and macronutrient intake, are presented with exact p-values Table 4Percentage of differently active adolescents, who met or did not meet reference value (RV) for daily energy and nutrient intakeBoys($$n = 162$$)Girls($$n = 176$$) Met RV Did not meet RV p Met RV Did not meet RV p RV [2016] n%n%n%n% Energy 0.320.31boys/girls (kcal)LPA$\frac{421.11578.9513.93186.12300}{2000}$MPA$\frac{109.7939.386.511693.52600}{2300}$VPA$\frac{615.03485212.51487.53000}{2600}$ Carbohydrates 0.750.89> $50\%$ of energy from carbohydratesLPA1263.2736.82877.8822.2MPA7370.93029.19979.82520.2VPA2972.51127.51275.0425.0 Proteins 0.070.55boys / girls(g protein/day)64 / 48 gLPA1368.4631.62775.0925.0MPA8481.61918.48266.14233.9VPA3792.537.51062.5637.5 Fats $0.050.6230\%$ of energy from fatLPA1368.4631.61336.12363.9MPA4139.86260.25342.77157.3VPA2152.51947.5850.0850.0n, number of subjects; LPA, less physically active; MPA, moderately physically active; VPA, vigorously physically active adolescents; significance results obtained with Pearson’s chi-square test between differently physically active adolescents are presented with exact p-values ($p \leq 0.008$ due to multiple comparisons in 3 × 2 contingency table (Bonferroni correction)) Referring to Table 3, results demonstrated that $75\%$ of adolescents were meeting the national recommendations for both carbohydrate and protein intakes, while only $44\%$ and $10\%$ of adolescents were meeting the national recommendations for fat and energy intakes, respectively. A significantly higher percentage of boys ($82.7\%$) were meeting recommendations for protein intake than girls ($67.0\%$) ($$p \leq 0.001$$). In contrast, no statistical differences in meeting the national recommendation were found between boys and girls for either energy ($$p \leq 0.18$$), carbohydrates ($$p \leq 0.06$$), or fats ($$p \leq 0.41$$). Results revealed (Table 4) that there were no statistically significant differences in the percentages of adolescents who met / did not meet energy and macronutrient recommendation between differently active boys and girls. Figure 1 presents the percentage values of recommended daily intake of energy and macronutrients, for all adolescents, as well as separately for boys and girls, for the two groups: those who met the recommendations and those who did not. Results demonstrated that in the group of adolescents, who met the recommendations, girls achieved statistically higher percentages of the recommended daily intake of carbohydrates than boys ($$p \leq 0.007$$), while boys achieved statistically higher percentages of the recommended daily intake of proteins than girls ($$p \leq 0.001$$). No similar differences were observed between boys and girls with comparable extent of physical activity (results not presented). Fig. 1Percentage values of recommended daily energy and macronutrient intake according to the existing recommendations, presented with box-plots (results presented as minimum, maximum, sample median, as well as 1st and 3rd quartile) for all adolescents (dark gray plots), as well as separately for boys (light-gray plots) and girls (open plots) in two groups: in those who did not meet the recommendations (left side) and those who did (right side). Significant differences between boys and girls are presented with exact p-values (tested with Mann Whitney U-test). M, median; n, number of subjects *Additional analysis* between adolescents who met and did not meet the energy intake recommendations revealed that adolescents who met the recommendations had a statistically higher BMI (21.0 kg/m2) than adolescents who did not meet the recommendations (19.0 kg/m2) ($$p \leq 0.000$$). In addition, the extent of physical activity according to DEEPA was not significantly different between adolescents who met recommendations (8.8 kJ/kg·day) and those who did not meet the recommendations for energy intake (8.4 kJ/kg·day) ($$p \leq 0.71$$). ## Discussion The results of the present study suggest that three-quarters of adolescents met the recommendations for carbohydrate ($75\%$) and protein ($75\%$) intake, while far lower percentage of those who met the recommendations was observed for fat ($44\%$) and energy ($10\%$) intake. A comparison between boys and girls demonstrated that the intake of energy and all macronutrients was significantly higher in boys than in girls in all three groups of physical activity extent, namely LPA, MPA, and VPA. In addition, protein intake recommendations were more frequently met by boys ($83\%$) than by girls ($67\%$), while no significant gender differences were found for other macronutrients and energy intake. The present study demonstrated that adolescent boys were significantly more physically active than girls ($35\%$ higher DEEPA median). In addition, significant differences in energy and macronutrient intake were observed in differently physically active boys, but not in differently physically active girls. Nevertheless, there were no statistically significant differences in the percentages of adolescents who met or did not meet energy and macronutrient recommendations between differently active boys and girls. However, in the group of adolescents who met the recommendations, boys achieved a higher percentage of the recommended daily protein intake, while girls consumed higher percentage of carbohydrates. The percentage of adolescents who were meeting the recommendations for energy intake was very low ($12\%$ of boys; $9\%$ of girls). In contrast, in a study by Kobe et al. [ 30] also conducted on Slovenian adolescents in 2012, a higher percentage of adolescents (aged 15 to 16 years) were meeting the energy intake recommendations than in our study ($46\%$ of boys and $32\%$ of girls). Similar results to Kobe et al. [ 30] was reported in a study of Barić et al. [ 31] in Croatian adolescents (average age = 16 years) performed in 2001, where a total of $38\%$ of adolescents were meeting the energy intake recommendations [32]. Different results between the three studies could be due to different energy assessment methods (24-h recall and the OPEN platform (Slovenia) used in our study, the 3-day weighed dietary protocol and the Prodi 5.2 Expert (Germany) in the study by Kobe et al. [ 30], and the food frequency questionnaire and the national food composition table (Croatia) in the study by Barić et al. [ 31]), and different cut-off points used for energy intake (Slovenian national recommendations adopted from the Nutrition Societies of Germany, Austria, and Switzerland [24] in our study and in the study by Kobe et al. [ 30], and US recommendation in the study by Barić et al. [ 31]). On average, the adolescents in our sample who did not meet the national recommendations for energy intake, reached less than three quarters of recommended value for energy intake. Assuming that the reported intake is a good approximation of the actual intake, chronic energy deficiency would impair growth, development, brain function, and the endocrine system [1], as well as health-related symptoms of RED-S [19]. Nevertheless, these results should be interpreted with caution, as 24-recall method has some well-known inherent limitations, which are further discussed below. However, it should not be excluded that the low values of energy intake and the low percentage of individuals meeting the recommendations for energy intake are also due to underreporting of energy intake. This is also supported by the fact that $75\%$ of adolescents in our study were classified as normal weight (according to BMI). Another factor, that may contribute to the observed low percentage of adolescents who met the national recommendations for energy intake is self-assessment of the amount of physical activity [33]. A comparison of energy intake in differently physically active adolescents revealed that VPA boys had higher energy intake than LPA boys, with no other differences observed between differently active adolescents. All girls, regardless of their physical activity, had similar energy intakes. Similar to the results of the present study for girls, Ottawere et al. [ 34] reported no significant differences in energy intake between differently physically active adolescents of both genders. This is of particular concern because research conducted by Peklaj et al. [ 19] in athletes found that only $13\%$ of athletes had no health-related symptoms of RED-S, which was particularly evident in girls. The total carbohydrate intake of adolescents in our study was adequate, which was also reported in some other studies conducted in Slovenia [35], [36]. VPA boys had a statistically higher carbohydrate intake than MPA boys, while no other differences were observed between differently active adolescents. Girls with different physical activity had similar carbohydrate intakes. Ottavere et al. [ 34] found no significant differences in carbohydrate intake according to physical activity level in both genders. The results of Zdešar Kotnik et al. [ 37], obtained on the same sample as in the present study, demonstrated that the intake of salty and sweet snacks was too high in adolescents of both genders. Other studies conducted in Slovenia, have also demonstrated that adolescents consume too much sugar, snacks, and sweet drinks [26], [30], [35]. Adolescents should therefore be encouraged to choose complex and unprocessed carbohydrates such as whole grains, fresh fruits and vegetables, and low-sugar products more often. The high percentage of adolescents ($83\%$ of boys; $79\%$ of girls) meeting protein recommendations in our sample, can be explained mainly by high meat consumption [37], while other studies found a high intake of milk and dairy products among adolescents [35]. Girls consumed on average $26\%$ less protein than boys. In addition, significantly fewer girls than boys met national recommendations for protein intake, so it is recommended that they reach for foods composed of high-quality proteins such as eggs, fish, and legumes. A comparison of adolescents with different physical activity extent demonstrated that VPA boys consumed more proteins than MPA and LPA boys, while no differences were observed for girls. Similar results were reported for both genders in the study of Otteawere et al. [ 34], in which more active adolescents had significantly higher protein intakes than moderately active ones. Although $80\%$ of adolescents in the present study met recommendations for protein intake, a lower percentage of such individuals was observed in VPA and MPA girls ($63\%$ and $66\%$, respectively), which should therefore strive for higher protein intake. Less than half of the adolescents in our study ($46\%$ of boys, $42\%$ of girls) met the recommendation to obtain at least $30\%$ of daily energy intake from fats. A similar percentage was observed in the studies of Sanchez et al. [ 38] ($33\%$ of adolescents) and Barić et al. [ 31] ($22\%$ of adolescents), while a study conducted in Australia reported a higher percentage of adolescents who met their total daily fat intake ($70\%$) [39]. It has also been reported that in Slovenia, the adolescent population consumed the lowest amounts of fat compared to adults and elderly adults [35]. Several studies have also highlighted a worrying fact about the inadequate composition of fat intake in adolescents [35], [36], [39], [40]. It is therefore recommended that all adolescents, regardless of gender, should include more foods with unsaturated fatty acids in their daily diet and avoid foods with saturated fats. Relative to physical activity, no significant differences in absolute fat intake were found in girls. In boys, differences were found between MPA and VPA boys, with the latter having a higher intake. Otteawere et al. [ 34] did not found any differences in total fat intake in both genders. Fats are a good source of energy, and to fulfill adolescent’s daily energy needs, it is necessary that they consume the right amount while considering fat composition. In the present study, the majority of adolescents had adequate BMI (average BMI of 21.1 kg/m2), which would not be expected in adolescents with chronic energy deficiency. Yet, the results of Peklaj et al. [ 19], who conducted a study on the prevalence of RED-S-related symptoms in a sample of Slovenian competitive athletes, indicated the same issue. In addition, results demonstrated that adolescents, who did not meet the energy intake recommendations according to their reported food intake, had higher BMI than adolescent who met the recommendation according to their reported food intake. Yet, these two groups had a comparable DEEPA. These two observations indicate to an underreporting of dietary intake, particularly in individuals with higher BMI. Finally, we would like to consider some of the potential methodological limitations of the present study. Although this may not be feasible in large scale studies, the use of accelerometry or heart rate monitors over self-report methods for the assessment of physical activity extent is of course preferential. Namely, self-assessment of physical activity with the questionnaire SHAPES may be subjective, leading to an overestimation of DEEPA [41], [34]. In contrast, the conservative MET values (6 MET for VPA and 3 MET for MPA) used in our calculation of DEEPA would have the opposite effect. Indeed, some guidelines recommend the use of higher MET values for VPA (7 MET) and MPA (4 MET) and a higher ratio between MET and kcal/kg·h (a ratio of more than 1 MET = 1 kcal/kg·h is recommended) [42], [43]. We therefore expect that at least some of the potential overreporting of physical activity extent has been neutralized with the use of conservative MET values in DEEPA calculation. It is also worth considering that the recommended values for energy intake may have been too low for some individuals, because physical activity classification in our study was based on DEPPA values (i.e. estimation of average daily energy expenditure for physical activity), whereas the recommendations for energy intake are based on classifying individuals according to their physical activity level (i.e. estimation of total daily energy expenditure). However, if this were true, an even smaller percentage of adolescents than reported would meet the recommendations. Our main consideration is that the majority of subjects, who did not meet the recommendations for energy intake according to the analysis of two 24-h recalls, had an appropriate BMI. This may suggest that the adolescents were either underreporting their daily food intake or that they selected some inadequate foods in the OPEN system. However, the latter does not seem to be a plausible explanation, as data in the OPEN system are regularly harmonized with the European Food Information Resource [44]. Thus, if only feasible, we suggest that a more accurate method for the assessment of total energy intake (e.g. a weighted food diary) should be used instead of 24-h recalls. The study provides an insight into how different levels of physical activity affect energy and macronutrient intake in adolescent boys and girls. With this knowledge, it is possible to establish preventive measures to minimize the negative health consequences of inadequate nutrient and energy intake. The results of the study also reveal, which issues should be focused on (gender, especially highly active girls) while providing advice on nutrition to adolescents with different levels of physical activity. ## Conclusions and implications for research and practice Adolescents need to ensure an adequate intake of energy and all macronutrients for normal growth and development, while maintaining regular and sufficient physical activity. Although results on energy intake obtained from 24-h recalls should be taken with caution (namely, although the majority of adolescents failed to meet recommendations for energy intake according to their reported food intake, they also had an appropriate BMI, which is an unlikely situation, if energy intake were consistently low), some important conclusions can be made. Present study demonstrated that according to their reported food intake, most Slovenian adolescents are well supplied with carbohydrates and proteins. In contrast, almost all of them failed to meet minimal recommendations for energy intake, and less than $50\%$ met the minimal recommendations for fat intake. Particularly alarming was the energy intake data in VPA girls, who consume the same amounts of energy than MPA and LPA girls, despite being physically very active and having higher nutrient and energy needs. Therefore, the attention should be paid not only to food intake of less active, but also to that of vigorously active adolescents. To prevent negative consequences of inadequate energy and nutrient intake, knowledge about proper nutrition must be transmitted to all groups of adolescents with different activity levels, which can be obtained through various educational activities at schools or in sports clubs. Parents and coaches should also be involved, as they may influence adolescents’ eating habits. ## References 1. Corkins MR, Daniels SR, Ferranti SD, Kim JH, Magge SN, Schwarzenberg SJ. **Nutrition in Children and Adolescents**. *Med Clin North Am* (2016.0) **100** 1217-35. DOI: 10.1016/j.mcna.2016.06.005 2. Romieu I, Dossus L, Barquera S, Blottière HM, Franks PW, Gunter M. **Energy balance and obesity: what are the main drivers?**. *Cancer Causes Control* (2017.0) **28** 247-58. DOI: 10.1007/s10552-017-0869-z 3. Savarino G, Corsello A, Corsello G. **Macronutrient balance and micronutrient amounts through growth and development**. *Ital J Pediatr* (2021.0) **47** 109. DOI: 10.1186/s13052-021-01061-0 4. **Dietary reference values for nutrients Summary report**. *EFSA Support Publ* (2017.0) **14** e15121E 5. Hutchinson C. **A review of iron studies in overweight and obese children and adolescents: a double burden in the young?**. *Eur J Nutr* (2016.0) **55** 2179-97. DOI: 10.1007/s00394-016-1155-7 6. Pereira-Santos M, Costa PRF, Assis AMO, Santos CAST, Santos DB. **Obesity and vitamin D deficiency: a systematic review and meta-analysis**. *Obes Rev* (2015.0) **16** 341-9. DOI: 10.1111/obr.12239 7. Gu K, Xiang W, Zhang Y, Sun K, Jiang X. **The association between serum zinc level and overweight/obesity: a meta-analysis**. *Eur J Nutr* (2019.0) **58** 2971-82. DOI: 10.1007/s00394-018-1876-x 8. Reilly JJ, Kelly J. **Long-term impact of overweight and obesity in childhood and adolescence on morbidity and premature mortality in adulthood: systematic review**. *Int J Obes* (2011.0) **35** 891-8. DOI: 10.1038/ijo.2010.222 9. Kang NR, Kwack YS. **An update on mental health problems and cognitive behavioral therapy in pediatric obesity**. *Pediatr Gastroenterol Hepatol Nutr* (2020.0) **23** 15-25. DOI: 10.5223/pghn.2020.23.1.15 10. Weihe P, Spielmann J, Kielstein H, Henning-Klusmann J, Weihrauch-Blüher S. **Childhood obesity and cancer risk in adulthood**. *Curr Obes Rep* (2020.0) **9** 204-12. DOI: 10.1007/s13679-020-00387-w 11. Weihrauch-Blüher S, Schwarz P, Klusmann JH. **Childhood obesity: increased risk for cardiometabolic disease and cancer in adulthood**. *Metabolism* (2019.0) **92** 147-52. DOI: 10.1016/j.metabol.2018.12.001 12. Das JK, Salam RA, Thornburg KL, Prentice AM, Campisi S, Lassi ZS. **Nutrition in adolescents: physiology, metabolism, and nutritional needs**. *Ann N Y Acad Sci* (2017.0) **1393** 21-33. DOI: 10.1111/nyas.13330 13. De Vet E, Stok FM, De Wit JBF, De Ridder DTD. **The habitual nature of unhealthy snacking: how powerful are habits in adolescence?**. *Appetite* (2015.0) **95** 182-7. DOI: 10.1016/j.appet.2015.07.010 14. Alberga AS, Sigal RJ, Goldfield G, Prud’homme D, Kenny GP. **Overweight and obese teenagers: why is adolescence a critical period? Pediatr**. *Obes* (2012.0) **7** 261-73 15. Hayes G, Dowd KP, MacDonncha C, Donnelly AE. **Tracking of physical activity and sedentary behavior from adolescence to young adulthood: a systematic literature review**. *J Adolesc Health* (2019.0) **65** 446-54. DOI: 10.1016/j.jadohealth.2019.03.013 16. Chaput J-P, Dutil C. **Lack of sleep as a contributor to obesity in adolescents: impacts on eating and activity behaviors**. *Int J Behav Nutr Phys Act* (2016.0) **13** 103. DOI: 10.1186/s12966-016-0428-0 17. 17.Golden NH, Schneider M, Wood C et al. Nutrition co, Adolescence co, Obesity so,. Preventing obesity and eating disorders in adolescents. Pediatrics. 2016;138(3):e20161649. 18. Petrie HJ, Stover EA, Horswill CA. **Nutritional concerns for the child and adolescent competitor**. *Nutr* (2004.0) **20** 620-31. DOI: 10.1016/j.nut.2004.04.002 19. Peklaj E, Reščič N, Koroušic Seljak B, Rotovnik Kozjek N. **Is RED-S in athletes just another face of malnutrition?**. *Clin Nutr ESPEN* (2022.0) **48** 298-307. DOI: 10.1016/j.clnesp.2022.01.031 20. Mountjoy M, Kaiander Sundgot-Borgen J, Burke M. **International Olympic Committee (IOC) Consensus Statement on relative Energy Deficiency in Sport (RED-S): 2018 update**. *IJSNEM* (2018.0) **28** 316-31. DOI: 10.1123/ijsnem.2018-0136 21. Starc GK, Kovač M, Strel J. **TheACDSi 2014—a decennial study on adolescents’ somatic, motor, psycho-social development and healthy lifestyle: study protocol**. *Anthropol Noteb* (2016.0) **21** 107-23 22. 22.Inštitut za varovanje zdravja. Republike Slovenije (Institute of Health Protection of Republic of Slovenia). Slikovno gradivo s prikazom velikosti porcij. Ljubljana, Slovenia., 2013. https://www.nijz.si/sites/www.nijz.si/files/publikacije-datoteke/slikovno_gradivo_s_prikazom_velikosti_porcij.pdf Accessed January 15, 2022. 23. 23.OPEN Platform for Clinical Nutrition. http://opkp.si/en Accessed March 1, 2015. 24. 24.Referenčne vrednosti za vnos hranil. Tabelarična priporočila za otroke (od 1. Leta starosti naprej), mladostnike, odrasle, starejše, nosečnice ter doječe matere. Ljubljana, Slovenija: Nacionalni inštitut za javno zdravje. National Institute of Public Health); 2016. 25. Leatherdale ST, Manske S, Wong SL, Cameron R. **Integrating research, policy, and practice in school-based physical activity prevention programming: the School Health Action, Planning, and evaluation system (SHAPES) physical activity module**. *Health Promot Pract* (2009.0) **10** 254-61. DOI: 10.1177/1524839906298499 26. Wong SL, Leatherdale ST. **Association between sedentary behavior, physical activity, and obesity: inactivity among active kids**. *Prev Chronic Dis* (2009.0) **6** A26. PMID: 19080032 27. Lohman TG, Roche AF, Martorell RA. *Anthropometric standardization reference manual* (1988.0) 28. Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. **Establishing a standard definition for child overweight and obesity worldwide: international survey**. *BMJ* (2000.0) **320** 1240. DOI: 10.1136/bmj.320.7244.1240 29. Cole TJ, Flegal KM, Nicholls D, Jackson AA. **Body mass index cut offs to define thinness in children and adolescents: international survey**. *BMJ* (2007.0) **335** 194. DOI: 10.1136/bmj.39238.399444.55 30. Kobe H, Štimec M, Ribič CH, Fidler Mis N. **Food intake in slovenian adolescents and adherence to the optimized mixed Diet: a nationally representative study**. *Public Health Nutr* (2012.0) **15** 600-8. DOI: 10.1017/S1368980011002631 31. Barić IC, Cvjetić S, Satalić Z. **Dietary intakes among croatian schoolchildren and adolescents**. *Nutr Health* (2001.0) **15** 127-38. DOI: 10.1177/026010600101500206 32. Havel DH, Calloway DH, Gussow JD, Mertz W, Nesheim MC. *National Research Council (US) Subcommittee on the tenth Edition of the recommended Dietary allowances. Recommended Dietary Allowances: 10th Edition* (1989.0) 33. Adamo KB, Prince SA, Tricco AC, Connor-Gorber S, Tremblay M. **A comparison of indirect versus direct measures for assessing physical activity in the pediatric population: a systematic review**. *Int J Pediatr Obes IJPO Off J Int Assoc Study Obes* (2009.0) **4** 2-27 34. Ottevaere C, Huybrechts I, Béghin L, Cuenca-Garcia M, De Bourdeaudhuij I, Gottrand F. **Relationship between self-reported dietary intake and physical activity levels among adolescents: the HELENA study**. *Int J Behav Nutr Phys Act* (2011.0) **8** 8. DOI: 10.1186/1479-5868-8-8 35. 35.Gregorič M, Blaznik U, Fajdiga Turk V, Delfar N et al. Različni vidiki prehranjevanja prebivalcev Slovenije v starosti od 3 mesecev do 74 let. NIJZ. 2019; 1-126. https://www.dlib.si/stream/URN:NBN:SI:DOC-GMUGUE1W/36643cb6-69b8-4323-ab59-4da5e3a2b98b/PDF. Accessed January 31, 2022. 36. Fidler Mis N, Kobe H, Štimec M. **Dietary dietary intake of macro- and micronutrients in slovenian adolescents: comparison with reference values**. *Ann Nutr Metab* (2012.0) **61** 305-13. DOI: 10.1159/000342469 37. 37.Zdešar Kotnik KJ, Gregorič G, Koroušić M, Seljak B, Golja P, editors, editors. Evaluation of nutrition among adolescents in Slovenia. Health of children and adolescents; 3rd scientific and professional international conference; 2019; Portorož: University of Primorska Press. 38. Sanchez A, Norman GJ, Sallis JF, Calfas KJ, Cella J, Patrick K. **Patterns and Correlates of Physical Activity and Nutrition Behaviors in adolescents**. *Am J Prev Med* (2007.0) **32** 124-30. DOI: 10.1016/j.amepre.2006.10.012 39. O’Sullivan TA, Ambrosini G, Beilin LJ, Mori TA, Oddy WH. **Dietary intake and food sources of fatty acids in australian adolescents**. *Nutr* (2010.0) **27** 153-9. DOI: 10.1016/j.nut.2009.11.019 40. Rippin HL, Hutchinson J, Jewell J, Breda JJ, Cade JE. **Child and adolescent nutrient intakes from current national dietary surveys of european populations**. *Nutr Res Rev* (2019.0) **32** 38-69. DOI: 10.1017/S0954422418000161 41. Wong SL, Leatherdale ST, Manske SR. **Reliability and validity of a school-based physical activity questionnaire**. *Med Sci Sports Exerc* (2006.0) **38** 1593-600. DOI: 10.1249/01.mss.0000227539.58916.35 42. Butte NF, Watson KB, Ridley K, Zakeri IF, McMurray RG, Pfeiffer KA. **A youth compendium of physical activities: activity codes and metabolic intensities**. *Med Sci Sports Exerc* (2018.0) **50** 246-56. DOI: 10.1249/MSS.0000000000001430 43. Melzer K, Heydenreich J, Schutz Y, Renaud A, Kayser B, Mäder U. **Metabolic equivalent in adolescents, active adults and pregnant women**. *Nutrients* (2016.0) **8** 438. DOI: 10.3390/nu8070438 44. 44.European Food Information Resource (EuroFIR). https://www.eurofir.org/ Accessed March 1, 2022.
--- title: 'The association between fasting plasma glucose variability and incident eGFR decline: evidence from two cohort studies' authors: - Niloofar Deravi - Yasaman Sharifi - Fatemeh Koohi - Seyed Saeed Tamehri Zadeh - Soroush Masrouri - Fereidoun Azizi - Farzad Hadaegh journal: BMC Public Health year: 2023 pmcid: PMC10041700 doi: 10.1186/s12889-023-15463-8 license: CC BY 4.0 --- # The association between fasting plasma glucose variability and incident eGFR decline: evidence from two cohort studies ## Abstract ### Background Glycemic variability (GV) is developing as a marker of glycemic control, which can be utilized as a promising predictor of complications. To determine whether long-term GV is associated with incident eGFR decline in two cohorts of Tehran Lipid and Glucose Study (TLGS) and Multi-Ethnic Study of Atherosclerosis (MESA) during a median follow-up of 12.2 years. ### Methods Study participants included 4422 Iranian adults (including 528 patients with T2D) aged ≥ 20 years from TLGS and 4290 American adults (including 521 patients with T2D) aged ≥ 45 years from MESA. The Multivariate Cox proportional hazard models were used to assess the risk of incident eGFR decline for each of the fasting plasma glucose (FPG) variability measures including standard deviation (SD), coefficient of variation (CV), average real variability (ARV), and variability independent of the mean (VIM) both as continuous and categorical variables. The time of start for eGFR decline and FPG variability assessment was the same, but the event cases were excluded during the exposure period. ### Results In TLGS participants without T2D, for each unit change in FPG variability measures, the hazards (HRs) and $95\%$ confidence intervals (CI) for eGFR decline ≥ $40\%$ of SD, CV, and VIM were 1.07(1.01–1.13), 1.06(1.01–1.11), and 1.07(1.01–1.13), respectively. Moreover, the third tertile of FPG-SD and FPG-VIM parameters was significantly associated with a 60 and $69\%$ higher risk for eGFR decline ≥ $40\%$, respectively. In MESA participants with T2D, each unit change in FPG variability measures was significantly associated with a higher risk for eGFR decline ≥ $40\%$.Regarding eGFR decline ≥ $30\%$ as the outcome, in the TLGS, regardless of diabetes status, no association was shown between FPG variability measures and risk of eGFR decline in any of the models; however, in the MESA the results were in line with those of GFR decline ≥ $40\%$.Using pooled data from the two cohorts we found that generally FPG variability were associated with higher risk of eGFR decline ≥ $40\%$ only among non-T2D individuals. ### Conclusions Higher FPG variability was associated with an increased risk of eGFR decline in the diabetic American population; however, this unfavorable impact was found only among the non-diabetic Iranian population. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12889-023-15463-8. ## Introduction Glycemic variability (GV), also known as blood glucose swings, includes a wide range of blood glucose variations that occur throughout the day, including hypoglycemic periods and postprandial spikes, as well as fluctuations that occur at the same time on different days. Diabetes complications are associated with GV in both type one diabetes and type 2 diabetes (T2D) patients [1–5]. Accordingly, GV was shown to be mostly associated with oxidative stress [6] and an increased incidence of hypoglycemia, a trigger of inflammatory processes [5, 7]. GV is measured on a short- and long-term basis [5]. Evidence supports the importance of this index both as a criterion for glycemic control and in the prevention of complications among patients with diabetes [1, 2, 4, 5, 8–12]. Short-term GV refers to between-days or within-day glycemic fluctuations, which are often measured via continuous glucose monitoring mostly in patients with type 1 diabetes [13]. Long-term GV refers to glycemic fluctuations over months to years, mostly measured by visit-to-visit variability in either fasting plasma glucose (FPG) or HbA1c in patients with diabetes [13]. Previous studies have shown that long-term GV may predict chronic kidney disease (CKD), the main microvascular complication of diabetes leading to both morbidity and mortality [14–18] with a high prevalence and incidence rate among both Iranian and American populations [19–22].The Food and Drug Administration (FDA) committee and National Kidney Foundation (NKF) have published a series of studies to investigate whether eGFR declines less than $50\%$ could be defined as an important kidney endpoint [23–26]. The committee reported that a $30\%$ decline in eGFR can be regarded as a reliable surrogate endpoint in some circumstances; however, a $40\%$ eGFR decline could present stronger evidence [27–29]. Accordingly, a meta-analysis of 1.7 million participants reported the average adjusted 10-year risk of end-stage renal disease (ESRD) for eGFR declines of $40\%$ and $30\%$, were $83\%$ and $64\%$, respectively [24]. Previous studies mainly conducted among T2D patients in the East Asian region have investigated the effect of HbA1c variation on eGFR decline [30–32], however, there is a paucity of information on the effect of FPG variability on eGFR decline in individuals with T2D [32]. Therefore, the current study for the first time investigated the association of long-term FPG variability with eGFR decline ≥ 30 and $40\%$ in non-CKD adults with and without T2D in the Tehran Lipid and Glucose Study (TLGS), as well as in the participants of the Multi-Ethnic Study of Atherosclerosis (MESA) study during about one decade of follow-up. ## Study population Study participants were selected from participants of TLGS, an ongoing large-scale population-based cohort study conducted on a representative population of Tehran city, the capital of Iran. TLGS aims to determine the risk factors for non-communicable diseases. The design of TLGS has been published before [33, 34]. In brief, in phase 1 (1999–2001) 15,005 participants ≥ 3 years entered the study; the data collection has been continued since then at approximately three-year intervals in follow-up phases (i.e., phases 2 (2002–2005), 3 (2005–2008), 4 (2008–2011), 5 (2011–2014), and 6 (2014–2017)). Moreover, 3555 participants entered the cohort in phase 2 of the study and were subsequently followed in phases 3, 4, 5, and 6. In the present study, we included 9137 participants of the TLGS cohort (1057 patients with T2D) aged ≥ 20 years who participated in phase 2 (as the baseline phase). We aimed to calculate the visit-to-visit variability (VVV) of FPG; therefore, we included those with available FPG values in phases 3, and 4.. Consequently, we considered 2002–2011 as the exposure period. We excluded individuals with no measurement of FPG at any of the phases 2–4, diagnosed CKD before phase 2 and eGFR decline ≥ $40\%$ during the measurement period (i.e., at any of the phases 3–4), lost to follow-up at any of the phases, and missing data on covariates. Among non-diabetes group, those with incident T2D at any of the phases 3–6 were also excluded. Finally, 4422 participants (including 528 patients with T2D) remained for data analysis and followed till March 2018. Using a similar approach, for the eGFR decline ≥ $30\%$, 4,181 individuals (483 patients with T2D) were entered in the data analysis. Figure 1 demonstrates the flowchart of the TLGS study participants. Fig. 1TLGS study participants’ flowchart The MESA dataset, from a longitudinal study, was also used. Participants aged 45 to 84 years at baseline, from six sites around the United States, were oversampled by four ethnic/racial groups. Design and objectives have been described in detail elsewhere [35]. Briefly, after the baseline examination, there have been five additional follow-up visits at biennial intervals, the most recent ongoing in 2016–18. Institutional Review Board approval was granted at each site and informed consent was obtained from each participant. MESA included 6814 participants (859 patients with T2D aged ≥ 45 years who participated in phase 1 (Sep 2002 to Feb 2004). Similar to the process for the TLGS cohort we measured the VVV of FPG in MESA participants; therefore, we included those with available FPG values in sequential phases 1, 2 (Mar 2004 to Sep 2005), and 3 (2005 to May 2007). We excluded individuals with no measurement of FPG at any of the phases 1–3, diagnosed CKD at phases 1, eGFR decline ≥ $40\%$ at any of the phases 2–3, lost follow-up at any of the phases, and missing data on covariates. Those with incident T2D at any of the phases 2–5 in the non-diabetic group were also excluded. Finally, 4290 participants (including 521 patients with T2D) remained for data analysis. Hence, the final samples were followed for incident eGFR decline ≥ $40\%$ after the measurement period. Using the similar approach, for the eGFR decline ≥ $30\%$, 4,290 individuals (521 patients with T2D) were entered in the data analysis. Figure 2 demonstrates the flowchart of the MESA study participants. Fig. 2MESA study participants’ flowchart ## Clinical and laboratory measurements We collected the baseline characteristics including data on age, sex, drug, family, past medical history, and smoking status (never or ever) through a standard questionnaire and measured body weight and height, diastolic blood (DBP), and systolic (SBP) pressure during a clinical examination. We calculated body mass index (BMI) as weight (kg) divided by the square of height (m2). Using a standard mercury sphygmomanometer, we measured blood pressure twice in a seating position after 15 min of rest. We also drew blood samples for the measurement of FPG, total cholesterol (TC), and creatinine levels, after 12–14 h of overnight fasting, from all participants, assayed serum creatinine levels by the Jaffe kinetic calorimetric method, and analyzed the samples once internal quality control met the standard satisfactory criteria. The standard 2-h post-load plasma glucose (2-HPG) test was also performed for those not on glucose lowering drugs. Physical activity level was assessed using the Modifiable Activity Questionnaire (Low, moderate-high) [36]. Moderate-high Physical activity level was defined as achieving a score ≤ 600 MET-minutes per week [37]. Details of laboratory assessments were reported previously [38]. Detailed description of MESA clinical and laboratory measurements has been published elsewhere [35]. We defined VVV as an intra individual variability in FPG levels recorded across three consecutive examinations in both TLGS and MESA. We used four indices of variability: (a) coefficient of variation (CV), (b) Standard deviation (SD), (c) average real variability (ARV), and (d) variability independent of the mean (VIM). We calculated VIM as 100 * SD/meanβ, where β is the regression coefficient based on the natural logarithm of SD on the natural logarithm of the mean. We calculated ARV according to the following formula, where n denotes the number of measures of FPG [39].\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{ARV}=\frac{1}{n-1}{\sum }_{$i = 1$}^{n-1}\left|Value(i+1\right)-Value(i)|$$\end{document}ARV=1n-1∑$i = 1$n-1Value(i+1-Value(i)| ARV is the average of absolute differences between consecutive values. ## Definition of outcomes and variables The main outcome for this study was a reduction in eGFR of ≥ 40 and ≥ $30\%$ from the baseline. eGFR was estimated by use of the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation. 141*min (Scr/ƙ, 1)α *max (Scr/ƙ, 1)-1.209 *0.993age *1.018 (if female)*1.159 (if black). In this equation, we measured serum creatinine (Scr) in mg/dL, and age in years. ƙ is 0.9 and 0.7 for women and men, respectively; α is -0.329 and -0.411 for women and men; min indicates the minimum of Scr/ƙ or 1, and max indicates a maximum of Scr/ƙ or 1. eGFR was expressed as mL/min/1.73 m2. The percentage of change in eGFR was calculated as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{\text{Follow up measurement}-\text{ Baseline measurement}}{\text{Baseline measurement}} \times 100$$\end{document}Follow up measurement-Baseline measurementBaseline measurement×100 We defined T2D as FPG of ≥ 126 mg/dl or current use of antidiabetic drugs. Moreover, prevalent cardiovascular disease (CVD) was a self-reported CVD history with a prior diagnosis of CVD by a physician. Smoking was categorized as ever smoker (current or past) versus non-smoker. Smoking was categorized as ever smoker (current or past) versus non-smoker. Hypertension, was defined as the presence of at least one of the following criteria: (a) having SBP ≥ 140 mm Hg, (b) having DBP ≥ 90 mm Hg, and (c) initiation of anti-hypertensive drugs usage. ## Statistical methods We showed the baseline characteristics of participants as frequency (%) for categorical variables and the mean ± standard deviation (SD) for continuous variables. We also categorized participants with and without T2D into 3 groups according to the tertiles of FPG-SD and compared the baseline characteristics across these tertiles. Chi-square test and analysis of variance were used to compare the clinical and demographic characteristics, as appropriate. The follow-up time used for time-to-event analyses was defined as the time from baseline date (phase 2) to either eGFR decline ≥ 40 or $30\%$ in phases 5 or 6, or date of last data collection or death, whichever occurred first. Only the first occurrence of each outcome was used for analysis. Each glucose variability measure was categorized by tertiles with a reference level of the lowest tertile. For individuals with incident eGFR decline, survival time was defined as the mid-time between the entered date and the event date. Linear trends across the tertiles were calculated by including the tertile as a continuous variable in the models. Moreover, to estimate adjusted HRs and $95\%$ CIs for incident eGFR decline associated with GV, we applied multivariable Cox proportional hazard (Cox PH) models, separately for participants with and without T2D. We used the measures of FPG variability including CV, SD, ARV, and VIM as a continuous variable (for each unit change) and tertiles in Cox PH models (the lowest tertile was considered as the reference). We also created four models in each dataset and adjusted for well known risk factors for CKD that were reported in a systematic review in this field. [ 40, 41]: Model 1: adjusted for age and sex at baseline (phase 2). Model 2: Model 1 + marital status, education, ever smoking, prevalent CVD, physical activity, anti-diabetic drug use, anti-hypertensive drug, lipid-lowering drug, BMI, WC, SBP, DBP, eGFR, and FPG at baseline (phase 2). Model 3: Model 1 + marital status, education, ever smoking, prevalent CVD, and physical activity at baseline (phase 2), and anti-diabetic drug use, anti-hypertensive drug, and lipid-lowering drug over phases 2–4, and average BMI, WC, SBP, DBP, and eGFR over phases 2. And Model 4: Model 3 + the average FPG during the measurement period. We conducted the above-mentioned analysis in the pooled data of TLGS and MESA cohorts using stratified Cox regression analysis and R version 3.6.2. Only, in the TLGS cohort, the above analysis was performed for 2-HPG among non-type 2 diabetic population; the association between this parameter with eGFR decline among newly diagnosed type 2 diabetic population was also examined only in Model 1. We calculated the median follow-up between 2002 (baseline phase) and 2018 (the end of the study) and assessed the PH assumptions in Cox models with the Schoenfeld residuals test and log–log plots, showing all proportional assumptions were appropriate. We performed the statistical analyses using STATA version 14. We also considered a P-value of < 0.05 as statistically significant. ## The TLGS cohort Participants included 528 patients with T2D (women = 326) with a mean (SD) age of 52.5 (10.6) years and 3894 participants without T2D (women = 2288) with a mean (SD) age of 40.5 (13.1) years. Baseline characteristics of the study population for participants with and without T2D across tertiles of FPG-SD are presented in Table 1. Generally, compared to the first tertile of FPG-SD, subjects with T2D at the third tertile had higher lipid-lowering drug use as well as higher FPG levels at baseline, phases 3, and 4. There was also a significant difference between ages of participants between teriles of FPG-SD. Moreover, among participants without T2D, those in the third tertile of FPG-SD generally had higher BMI, WC, and FPG levels at phases 3 and 4 compared to the reference group. There was also a significant difference between the levels of education of participants between teriles of FPG-SD.Table 1Baseline characteristics of participants across tertiles of SD for fasting plasma glucose in TLGS, Tehran Lipid and Glucose Study (2002–2018)Characteristics With diabetes ($$n = 528$$) T1 ($$n = 176$$) T2 ($$n = 176$$) T3 ($$n = 176$$) P value Age (year) 52.51 ± 10.64 52.37 ± 11.62 53.99 ± 9.84 51.18 ± 10.26 0.044 Sex (women)326 (61.74)112 (63.64)97 (55.11)117 (66.48)0.074Marital status (married)457 (86.55)144 (81.82)156 (88.64)157 (89.20)0.078Education (high school and more)124 (23.48)42 (23.86)42 (23.86)40 (22.73)0.959Antihypertensive drug124 (23.48)48 (27.27)37 (21.02)39 (22.16)0.338Lipid-lowering drug 81 (15.34) 18 (10.23) 25 (14.20) 38 (21.59) 0.011 Current smoking109 (20.64)34 (19.32)43 (24.43)32 (18.18)0.304Moderate-high physical activity315 (59.66)103 (58.52)101 (57.39)111 (63.07)0.516BMI (Kg/m2)29.43 ± 4.7529.61 ± 4.8929.02 ± 4.3029.65 ± 5.040.386WC (cm)98.87 ± 10.8599.09 ± 11.2298.67 ± 9.3498.84 ± 11.390.937SBP (mmHg)127.91 ± 20.12126.87 ± 19.61129.55 ± 19.78127.31 ± 20.960.408DBP (mmHg)78.35 ± 10.1978.35 ± 10.1979.87 ± 11.0278.77 ± 10.500.380eGFR (mL/min/1.73 m2)88.67 ± 15.1189.01 ± 16.0687.82 ± 14.6389.17 ± 14.630.659FPG (mg/dl) 156.45 ± 53.00 132.38 ± 36.75 154.43 ± 40.29 182.55 ± 64.92 < 0.001 FPG at phase 3 (mg/dl) 130.66 ± 54.9 130.66 ± 37.70 150.39 ± 38.41 188.64 ± 68.55 < 0.001 FPG at phase 4 (mg/dl) 166.2 ± 60.6 135.39 ± 36.81 163.13 ± 42.28 198.53 ± 76.18 < 0.001 Anti-diabetic drug at baseline 231 (43.75) 60 (34.09) 75 (42.61) 96 (54.55) 0.001 Without diabetes ($$n = 3$$,894)T1 ($$n = 1$$,331)T2 ($$n = 1$$,331)T3 ($$n = 1$$,331) P valueAge (year)40.49 ± 13.0940.18 ± 13.0440.28 ± 13.0241.02 ± 13.210.201Sex (women)2,288 (58.76)808 (60.71)740 (58.13)740 (57.36)0.190Marital status (married)3,144 (80.74)1,056 (79.34)1,027 (80.68)1,061 (82.25)0.168Education (high school and more) 1,318 (33.85) 436 (32.76) 466 (36.61) 416 (32.25) 0.039 Antihypertensive drug220 (5.65)79 (5.94)66 (5.18)75 (5.81)0.675Lipid-lowering drug113 (2.90)40 (3.01)26 (2.04)47 (3.64)0.052Current smoking812 (20.85)267 (20.06)258 (20.27)787 (61.01)0.318Moderate-high physical activity2458 (63.12)858 (64.46)813 (63.86)503 (38.99)0.149BMI (Kg/m2) 26.82 ± 4.44 26.64 ± 4.44 26.76 ± 4.39 27.08 ± 4.49 0.033 WC (cm) 89.22 ± 11.70 88.34 ± 11.72 89.19 ± 11.40 90.15 ± 11.91 < 0.001 SBP (mmHg)113.15 ± 16.00112.36 ± 15.48113.34 ± 16.16113.79 ± 16.340.066DBP (mmHg)73.54 ± 10.0773.11 ± 9.8773.55 ± 10.2573.98 ± 10.080.085eGFR (mL/min/1.73 m2)96.59 ± 16.5796.77 ± 16.5196.64 ± 16.2296.37 ± 16.980.818FPG (mg/dl)88.44 ± 8.0488.33 ± 6.7588.63 ± 7.7188.37 ± 9.470.579FPG at phase 3 (mg/dl) 87.66 ± 7.74 88.11 ± 6.64 87.45 ± 7.48 87.41 ± 8.93 0.033 FPG at phase 4 (mg/dl) 93,20 ± 8.10 89.66 ± 6.51 92.53 ± 6.86 97.50 ± 8.73 < 0.001 Data are represented as mean ± standard deviation for continuous variables and frequency (percent) for categorical variables T tertile, CVD cardiovascular disease, BMI body mass index, WC waist circumference, SBP systolic blood pressure, DBP diastolic blood pressure, eGFR estimated glomerular filtration rate; FPG *Fasting plasma* glucose *After a* median follow-up of 12.2 years (interquartile range: 11.1–13.3 years), 131 incident eGFR decline ≥ $30\%$ and 72 incident eGFR decline ≥ $40\%$ among subjects with T2D occurred; the corresponding values for non-diabetic participants were, 629 and 115, respectively. Table 2 shows the association of FPG variability as a continuous variable in four models with an eGFR decline ≥ $40\%$. Among participants without T2D, each unit change in FPG variability measures was significantly associated with a higher risk for eGFR decline ≥ $40\%$ in all models of FPG-SD, FPG-CV, and FPG-VIM; the corresponding HRs and $95\%$ CIs in the last models were 1.07 (1.01–1.13), 1.06 (1.01–1.11), and 1.07 (1.01–1.13), respectively. However, among participants with T2D, none of these measures were associated with eGFR decline ≥ $40\%$ events, even in model 1.Table 2HRs and $95\%$ CIs of incident eGFR decline ≥ $40\%$ according to each unit increase in FPG variability measures in Tehran Lipid and Glucose StudyVariability measuresModel 1Model 2Model 3Model 4HR ($95\%$ CI) P valueHR ($95\%$ CI) P valueHR ($95\%$ CI) P valueHR ($95\%$ CI) P valueSD With diabetes1.00(0.99-1.01)0.6450.99 (0.98-1.00)0.1670.99 (0.99-1.00)0.2580.99 (0.98-1.00)0.106 Without diabetes1.08(1.02-1.13) 0.006 1.06(1.01-1.12) 0.020 1.07 (1.01-1.13) 0.012 1.07 (1.01-1.13) 0.014 CV With diabetes0.99 (0.98-1.01)0.5110.99 (0.97-1.01)0.1580.99 (0.97-1.01)0.1900.99 (0.97-1.00)0.138 Without diabetes1.06(1.01-1.12) 0.012 1.05 (1.00-1.11) 0.033 1.06 (1.01-1.11) 0.017 1.06 (1.01-1.11) 0.017 ARV With diabetes1.00(0.99-1.01)0.8741.00 (0.99-1.00)0.3101.00 (0.99-1.00)0.5131.00 (0.99-1.00)0.309 Without diabetes1.03(0.99-1.07)0.1581.03(0.99-1.07)0.1371.03 (0.99-1.07)0.1831.03 (0.99-1.07)0.208VIM With diabetes1.00 (0.98-1.01)0.4630.99 (0.98-1,00)0.2520.99 (0.98-1.00)0.2330.99 (0.98-1.00)0.250 Without diabetes1.07(1.01-1.13) 0.016 1.06 (1.00-1.11) 0.041 1.07 (1.01-1.13) 0.019 1.07 (1.01-1.13) 0.018 Model 1: adjusted for age and sex at baseline (phase 2)Model 2: Model 1 + marital status, education, ever smoking, prevalent CVD, physical activity, anti-diabetic drug use, anti-hypertensive drug, lipid-lowering drug, BMI, WC, SBP, DBP, eGFR, and FPG at baseline (phase 2)Model 3: Model 1 + marital status, education, ever smoking, prevalent CVD, and physical activity at baseline (phase 2), and anti-diabetic drug use, anti-hypertensive drug, and lipid-lowering drug over phases 2–4, and average BMI, WC, SBP, DBP, and eGFR over phases 2–4Model 4: Model 3 + average FPG The associations of FPG variability as a categorical variable with incident eGFR decline ≥ $40\%$ are presented in Table 3. Generally, for SD and VIM in all of the models, a significant increasing trend was observed for incident eGFR decline among participants without T2D. Furthermore, among participants without T2D, the third tertiles of FPG-SD, and FPG-VIM were associated with higher risks in all models (the corresponding HRs and $95\%$ CIs in the last model were 1.60 (1.01–2.54), and 1.69(1.06–2.69), respectively). Whereas, no trends for incident eGFR decline were observed in participants with T2D in any of the FPG variability measures. Table 3HRs and $95\%$ CIs of incident eGFR decline ≥ $40\%$ according to tertiles of FPG variability measures in Tehran Lipid and Glucose StudyVariability measures Model 1 Model 2 Model 3 Model 4 HR ($95\%$ CI)HR ($95\%$ CI)HR ($95\%$ CI)HR ($95\%$ CI) With diabetes SD T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T21.20 (0.69–2.10)1.06 (0.59–1.90)0.93 (0.53–1.66)0.88 (0.49–1.58) T30.98 (0.54–1.77)0.76 (0.39–1.47)0.77 (0.42–1.44)0.67 (0.33–1.33) P trend 0.9680.4140.4170.254CV T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T21.04 (0.60–1.82)0.96 (0.53–1.73)0.80 (0.44–1.43)0.77 (0.42–1.39) T30.94 (0.53–1.66)0.74 (0.40–1.37)0.74 (0.41–0.34)0.70 (0.38–1.29) P trend 0.8280.3320.3270.253ARV T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T21.47 (0.84–2.58)1.29 (0.71–2.33)0.99 (0.55–1.79)0.94 (0.52–1.72) T31.09 (0.60–2.01)0.84 (0.43–1.65)0.83 (0.44–1.55)0.71 (0.35–1.44) P trend 0.7490.5880.5390.344VIM T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T21.10 (0.63–1.92)1.18 (0.67–2.09)1.09 (0.62–1.92)1.10 (0.62–1.93) T30.94 (0.53–1.66)0.84 (0.46–1.51)0.82 (0.46–1.47)0.83 (0.46–1.49) P trend 0.8350.5770.5120.746 Without diabetes SD T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T21.28 (0.80–2.07)1.37 (0.84–2.21)1.32 (0.81–2.13)1.31 (0.81–2.12) T3 1.61 (1.02–2.53) 1.67 (1.05–2.64) 1.61 (1.02–2.55) 1.60 (1.01–2.54) P trend 0.039 0.029 0.041 0.047 CV T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T21.36 (0.85–2.18)1.43 (0.89–2.30)1.43 (0.90–2.31)1.44 (0.89–2.30) T31.53 (0.97–2.42)1.56 (0.98–2.48)1.55 (0.98–2.47)1.55 (0.97–2.46) P trend 0.0710.0640.0650.068ARV T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T21.08 (0.68–1.70)1.18 (0.74–1.88)1.21 (0.76–1.93)1.21 (0.76–1.92) T31.27 (0.82–1.96)1.38 (0.89–2.15)1.30 (0.84–2.01)1.28 (0.82–1.99) P trend 0.2860.1470.2420.272VIM T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T21.42 (0.88–2.28)1.49 (0.93–2.41)1.49 (0.93- 2.39)1.49 (0.92–2.40) T3 1.65 (1.04–2.62) 1.66 (1.04–2.67) 1.69 (1.06–2.69) 1.69(1.06–2.69) P trend 0.034 0.035 0.029 0.029 Model 1: adjusted for age and sex at baseline (phase 2)Model 2: Model 1 + marital status, education, ever smoking, prevalent CVD, physical activity, anti-diabetic drug use, anti-hypertensive drug, lipid-lowering drug, BMI, WC, SBP, DBP, eGFR, and FPG at baseline (phase 2)Model 3: Model 1 + marital status, education, ever smoking, prevalent CVD, and physical activity at baseline (phase 2), and anti-diabetic drug use, anti-hypertensive drug, and lipid-lowering drug over phases 2–4, and average BMI, WC, SBP, DBP, and eGFR over phases 2–4Model 4: Model 3 + average FPG Supplementary Tables 1 and 2 show the association of FPG variability both as a continuous and categorical variable in four models with an eGFR decline ≥ $30\%$. Regardless of diabetes status, no association was shown between FPG variability measures and risk of eGFR decline in any of the models. As a sensitivity analysis, we examined the association of 2-HPG variability both as continuous and categorical variables with eGFR declines ≥ $30\%$ and $40\%$ among participants without T2D and newly diagnosed T2D not on glucose lowering medications (Supplementary Tables 3, 4, 5, 6, and 7, respectively). No association was shown between 2-HPG variability measures and risk of eGFR decline in any of the models in both non-diabetic individuals and newly diagnosed T2D participants. ## MESA cohort A population of 521 participants (women = 236) with T2D with a mean (SD) age of 63.2 (9.1) years and 3769 participants without T2D (women = 1968) with a mean (SD) age of 60.4 (9.9) years were evaluated. Baseline characteristics of the study population for participants with and without T2D across tertiles of FPG-SD are presented in Table 4. Compared to the first tertile of FPG-SD, subjects with T2D at the third tertile were generally younger, and had higher BMI as well as FPG levels at baseline, phases 2, and 3. Furthermore, among participants without T2D, those in the third tertile of FPG-SD (compared to the first tertile) were generally younger and had higher BMI, WC, and FPG levels in phases 3, and 4. There was also a significant difference between smoking status of participants between teriles of FPG-SD.Table 4Baseline characteristics of the participants across tertiles of SD for fasting plasma glucose in Multi-Ethnic Study of Atherosclerosis studyCharacteristics With diabetes ($$n = 521$$) T1 ($$n = 176$$) T2 ($$n = 172$$) T3 ($$n = 173$$) P value Age (year)63.21 ± 9.1364.85 ± 8.8264.67 ± 8.8560.10 ± 8.96 < 0.001 Sex (women)236 (45.30)83 (47.16)72 (41.86)81 (46.82)0.541Marital status (married)320 (61.42)108 (61.36)110 (63.95)102 (58.96)0.635Education (high school and more)281 (53.93)105 (59.66)86 (50.00)90 (52.02)0.161Antihypertensive drug317 (60.84)106 (60.23)114 (66.28)97 (56.07)0.148Lipid-lowering drug142 (27.26)47 (26.70)47 (27.33)48 (27.75)0.976Current smoking259 (49.71)82 (46.59)88 (51.16)89 (51.45)0.595Moderate-high physical activity279 (53.55)96 (54.55)91 (52.51)92 (53.18)0.947BMI (Kg/m2) 30.51 ± 5.63 29.59 ± 5.18 30.88 ± 6.05 31.08 ± 5.58 0.028 WC (cm)104.67 ± 14.19102.84 ± 13.17105.77 ± 15.06105.43 ± 14.200.107SBP (mmHg)130.45 ± 20.19129.15 ± 20.94133.31 ± 19.96128.95 ± 19.440.077DBP (mmHg)71.70 ± 9.3970.91 ± 10.1571.91 ± 8.6572.29 ± 9.290.366eGFR (mL/min/1.73 m2)91.86 ± 20.1988.19 ± 16.9989.37 ± 19.0298.06 ± 22.79 < 0.001 FPG (mg/dl) 150.22 ± 53.21 127.13 ± 31.31 142.42 ± 39.19 181.46 ± 66.68 < 0.001 FPG at phase 3 (mg/dl) 153.43 ± 58.79 126.86 ± 30.96 141.61 ± 36.31 192.23 ± 76.08 < 0.001 FPG at phase 4 (mg/dl) 141.00 ± 50.74 127.35 ± 30.08 137.37 ± 39.12 158.49 ± 69.49 < 0.001 Anti-diabetic drug at baseline390 (74.86)128 (72.73)134 (77.91)128 (73.99)0.511Without diabetes ($$n = 3$$,769)T1 ($$n = 1$$,278)T2 ($$n = 1$$,253)T3 ($$n = 1$$,238) P valueAge (year) 60.41 ± 9.89 60.78 ± 9.83 60.62 ± 9.91 59.83 ± 9.90 0.036 Sex (women)1,968 (52.22)685 (53.60)631 (50.36)652 (52.67)0.245Marital status (married)2,394 (63.52)804 (62.91)815 (65.04)775 (62.60)0.385Education (high school and more)2,630 (69.78)885 (69.25)873 (69.67)872 (70.44)0.806Antihypertensive drug1,059 (28.10)349 (27.31)338 (26.98)372 (30.05)0.173Lipid-lowering drug490 (13.00)154 (12.05)159 (12.69)177 (14.30)0.227Current smoking 1,858 (49.30) 591 (46.24) 642 (51.24) 625 (50.48) 0.025 Moderate-high physical activity2321 (61.58)797 (62.36)763 (60.89)761 (61.47)0.746BMI (Kg/m2) 27.57 ± 4.99 27.08 ± 4.80 27.40 ± 4.85 28.24 ± 5.25 < 0.001 WC (cm) 95.77 ± 13.52 94.56 ± 12.95 95.45 ± 13.27 97.34 ± 14.18 < 0.001 SBP (mmHg)122.92 ± 20.21123.13 ± 20.37122.65 ± 19.95122.97 ± 20.310.834DBP (mmHg)71.55 ± 10.0371.63 ± 10.0071.25 ± 9.8471.79 ± 10.240.388eGFR (mL/min/1.73 m2)82.29 ± 13.7682.31 ± 13.5682.14 ± 13.4082.42 ± 14.340.875FPG (mg/dl)87.51 ± 8.9088.80 ± 7.4587.67 ± 8.3887.06 ± 10.630.089FPG at phase 3 (mg/dl) 90.64 ± 8.58 88.56 ± 7.17 89.88 ± 7.63 93.57 ± 9.83 < 0.001 FPG at phase 4 (mg/dl) 90.22 ± 8.71 88.39 ± 7.33 89.81 ± 8.18 92.52 ± 9.96 < 0.001 Data are represented as mean ± standard deviation for continuous variables and frequency (percent) for categorical variablesT tertile, CVD cardiovascular disease, BMI body mass index, WC waist circumference, SBP systolic blood pressure, DBP diastolic blood pressure, eGFR estimated glomerular filtration rate, FPG fasting plasma glucose *After a* median follow-up of 9.2 years (interquartile range: 6.6–9.6 years), 106 incident eGFR decline ≥ $30\%$ and 49 incident eGFR decline ≥ $40\%$ among subjects with T2D occurred; the corresponding values for non-diabetic participants were, 157 and 62, respectively. Table 5 shows the association of FPG variability as a continuous variable in four models with an eGFR decline ≥ $40\%$. Among participants with T2D, each unit change in FPG variability measures was significantly associated with a higher risk for eGFR decline in model 4 of FPG-SD, FPG-CV, and FPG-VIM; the corresponding HRs and $95\%$ CIs were 1.01 (1.00–1.02), 1.02 (1.00–1.03), and 1.01(1.00–1.02), respectively. However, among participants without T2D, none of these measures were associated with an eGFR decline ≥ $40\%$. The associations of FPG variability as a categorical variable with incident eGFR decline ≥ $40\%$ are presented in Table 6. As shown in Table 6, a significant increasing trend was observed for incident eGFR decline among participants with T2D for FPG-SD in model 1, and FPG-ARV in both models 1 and 3. Furthermore, the third tertiles of FPG-SD (model 1), and FPG-ARV (models 1 and 3) were associated with significantly higher risks, the corresponding HRs and $95\%$ CIs were 2.15 (1.03–4.50), 2.45 (1.16–5.16), and 2.31 (1.08–4.96), respectively. However, no trends were observed in participants without T2D.Table 5HRs and $95\%$ CIs of incident eGFR decline ≥ $40\%$ according to each unit increase in FPG variability measures in Multi-Ethnic Study of AtherosclerosisVariability measures Model 1 Model 2 Model 3 Model 4 HR ($95\%$ CI) P valueHR ($95\%$ CI) P valueHR ($95\%$ CI) P valueHR ($95\%$ CI) P valueSD With diabetes1.01(1.01–1.02) 0.001 1.01(1.00–1.02) 0.040 1.01(1.01–1.02) 0.002 1.01 (1.00–1.02) 0.042 Without diabetes1.03(0.94–1.12)0.5161.02(0.94–1.11)0.6151.01(0.93–1.10)0.7621.01(0.93–1.10)0.790CV With diabetes1.02 (1.01–1.04) 0.008 1.02(1.00–1.03)0.0671.02(1.00–1.03) 0.020 1.02 (1.00–1.03) 0.041 Without diabetes1.02(0.94–1.10)0.6351.02(0.94–1.10)0.6311.01(0. 93–1.09)0.7971.01(0.93–1.09)0.799ARV With diabetes1.01(1.00–1.01) 0.006 1.01(1.00–1.01) 0.026 1.01(1.00–1.01) 0.007 1.01 (1.00–1.01)0.087 Without diabetes1.00(0.94–1.07)0.8891.00(0.94–1.07)0.9900.99(0.93–1.06)0.8270.99(0.93–1.06)0.794VIM With diabetes1.01(1.00–1.02)0.0521.00(1.01–1.02)0.1191.01(1.00–1.02)0.1441.01(1.00–1.02)0.055 Without diabetes1.02(0.93–1.11)0.6561.02(0.93–1.12)0.6331.01(0.93–1.10)0.8031.01(0.93–1.10)0.799Model 1: adjusted for age and sex at baseline (phase 2)Model 2: Model 1 + marital status, education, ever smoking, prevalent CVD, physical activity, anti-diabetic drug use, anti-hypertensive drug, lipid-lowering drug, BMI, WC, SBP, DBP, eGFR, and FPG at baseline (phase 2)Model 3: Model 1 + marital status, education, ever smoking, prevalent CVD, and physical activity at baseline (phase 2), and anti-diabetic drug use, anti-hypertensive drug, and lipid-lowering drug over phases 2–4, and average BMI, WC, SBP, DBP, and eGFR over phases 2–4Model 4: Model 3 + average FPGTable 6HRs and $95\%$ CIs of incident eGFR decline ≥ $40\%$ according to tertiles of FPG variability measures in Multi-Ethnic Study of AtherosclerosisVariability measures Model 1 Model 2 Model 3 Model 4 HR ($95\%$ CI)HR ($95\%$ CI)HR ($95\%$ CI)HR ($95\%$ CI) With diabetes SD T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T21.48 (0.69–3.19)1.28 (0.58–2.81)1.26 (0.57–2.77)1.18 (0.53–2.60) T3 2.15 (1.03–4.50) 1.38 (0.59–2.22)1.96 (0.93–4.17)1.46 (0.64–3.34) P trend 0.040 0.4590.0690.367CV T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T20.81 (0.37–1.79)0.65 (0.29–1.49)0.70 (0.31–1.57)0.66 (0.29–1.49) T31.78 (0.91–3.47)1.32 (0.65–2.70)1.60 (0.81–3.16)1.35 (0.67–2.75) P trend 0.0730.3260.1280.307ARV T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T21.48(0.67–3.30)1.16 (0.50–2.68)1.35 (0.60–3.06)1.20 (0.53–2.75) T3 2.45 (1.16–5.16) 1.79 (0.79–4.08) 2.31 (1.08–4.96) 1.78 (0.78–4.06) P trend 0.015 0.132 0.025 0.145VIM T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T20.54 (0.24–1.23)0.51 (0.22–1.18)0.50 (0.21–1.13)0.48 (0.21–1.12) T31.52 (0.80–2.87)1.39 (0.72–2.69)1.35 (0.70–2.61)1.49 (0.76–2.93) P trend 0.1550.2340.2640.178 Without diabetes SD T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T20.57 (0.30–1.09)0.59 (0.31–1.14)0.59(0.31–1.13)0.59 (0.30–1.13) T30.93 (0.53–1.64)0.95 (0.53–1.68)0.88 (0.49–1.56)0.87 (0.49–1.55) P trend 0.7440.8040.6230.603CV T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T20.74 (0.40–1.36)0.79 (0.43–1.45)0.76 (0.41–1.40)0.76 (0.41–1.40) T30.82 (0.45–1.49)0.86 (0.47–1.58)0.80 (0.44–1.46)0.80 (0.44–1.46) P trend 0.4830.6000.4450.444ARV T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T20.77 (0.42–1.39)0.74 (0.41–1.35)0.74 (0.41–1.35)0.74 (0.41–1.34) T30.70 (0.38–1.30)0.69 (0.37–1.28)0.65 (0.35–1.21)0.64 (0.34–1.19) P trend 0.2430.2150.1600.148VIM T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T20.67 (0.36–1.24)0.73 (0.39–1.36)0.70 (0.38–1.30)0.70 (0.38–1.29) T30.78 (0.43–1.41)0.83 (0.45–1.52)0.77 (0.43–1.40)0.77 (0.43–1.41) P trend 0.3840.5160.3720.375Model 1: adjusted for age and sex at baseline (phase 2)Model 2: Model 1 + marital status, education, ever smoking, prevalent CVD, physical activity, anti-diabetic drug use, anti-hypertensive drug, lipid-lowering drug, BMI, WC, SBP, DBP, eGFR, and FPG at baseline (phase 2)Model 3: Model 1 + marital status, education, ever smoking, prevalent CVD, and physical activity at baseline (phase 2), and anti-diabetic drug use, anti-hypertensive drug, and lipid-lowering drug over phases 2–4, and average BMI, WC, SBP, DBP, and eGFR over phases 2–4Model 4: Model 3 + average FPG The association between FPG variability measures and eGFR decline ≥ $30\%$ was in line with the reported results for GFR decline ≥ $40\%$, except that one unit increase in FPG-ARV was significantly associated with a higher risk for eGFR decline in model 4 among T2D participants (1.01 (1.00–1.01)) (Supplementary Tables 8 and 9). ## Pooled data of TLGS and MESA cohorts The association between FPG variability measures and eGFR decline ≥ $40\%$ as a continuous and categorical variable was also measured in pooled data of MESA and TLGS cohorts (Tables 7 and 8, respectively). Among participants without T2D, each unit change in FPG variability measures was significantly associated with a higher risk for eGFR decline in all models of FPG-SD, FPG-CV, and FPG-VIM; the corresponding HRs and $95\%$ CIs of model 4 were 1.05 (1.01–1.10), 1.05 (1.00–1.09), and 1.05 (1.00–1.10), respectively. Among participants with T2D, only model 1 of FPG-ARV was associated with eGFR decline ≥ $40\%$ (1.01(1.00–1.01)). However, no association was shown between FPG variability as a categorical variable and risk of eGFR decline in any of the models in both individuals with and without T2D in pooled data of TLGS and MESA cohorts. When we replaced eGFR decline ≥ $30\%$ in place of eGFR ≥ $40\%$ as the outcome of the study, no significant association was demonstrated for FPG variability measures in the fully adjusted model in both individuals and without T2D in the pooled data of two cohorts (data not shown).Table 7HRs and $95\%$ CIs of incident eGFR decline ≥ $40\%$ according to each unit increase in FPG variability measures in pooled data of Tehran Lipid and Glucose Study and Multi-Ethnic Study of Atherosclerosis cohortsVariability measures Model 1 Model 2 Model 3 Model 4 HR ($95\%$ CI) P valueHR ($95\%$ CI) P valueHR ($95\%$ CI) P valueHR ($95\%$ CI) P valueSD With diabetes1.01(1.00–1.01)0.0541.00 (0.99–1.01)0.7671.00 (1.00–1.01)0.1751.00 (0.99–1.01)0.881 Without diabetes 1.06(1.01–1.11) 0.010 1.05(1.01–1.10) 0.025 1.05 (1.01–1.10) 0.024 1.05 (1.01–1.10) 0.025 CV With diabetes1.01 (1.00–1.02)0.1471.00 (0.99–1.01)0.7991.00 (0.99–1.02)0.3941.00 (0.99–1.02)0.725 Without diabetes 1.05(1.01–1.09) 0.020 1.04 (1.00–1.09) 0.038 1.05 (1.00–1.09) 0.030 1.05 (1.00–1.09) 0.030 ARV With diabetes 1.01(1.00–1.01) 0.026 1.00 (1.00–1.01)0.4111.00 (1.00–1.01)0.1051.00 (1.00–1.01)0.590 Without diabetes1.02(0.99–1.06)0.2481.02 (1.09–1.06)0.2311.02 (0.98–1.05)0.3231.02 (0.98–1.05)0.343VIM With diabetes1.00 (1.00–1.01)0.4381.00 (0.99–1.01)0.8061.00 (0.99–1.01)0.8251.00 (0.99–1.01)0.580 Without diabetes 1.05(1.01–1.10) 0.025 1.05 (1.00–1.10) 0.043 1.05 (1.00–1.10) 0.032 1.05 (1.00–1.10) 0.032 Model 1: adjusted for age and sex at baseline (phase 2)Model 2: Model 1 + marital status, education, ever smoking, prevalent CVD, physical activity, anti-diabetic drug use, anti-hypertensive drug, lipid-lowering drug, BMI, WC, SBP, DBP, eGFR, and FPG at baseline (phase 2)Model 3: Model 1 + marital status, education, ever smoking, prevalent CVD, and physical activity at baseline (phase 2), and anti-diabetic drug use, anti-hypertensive drug, and lipid-lowering drug over phases 2–4, and average BMI, WC, SBP, DBP, and eGFR over phases 2–4Model 4: Model 3 + average FPGTable 8HRs and $95\%$ CIs of incident eGFR decline ≥ $40\%$ according to tertiles of FPG variability measures in pooled data of Tehran Lipid and Glucose Study and Multi-Ethnic Study of Atherosclerosis cohortsVariability measures Model 1 Model 2 Model 3 Model 4 HR ($95\%$ CI)HR ($95\%$ CI)HR ($95\%$ CI)HR ($95\%$ CI) With diabetes SD T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T21.29 (0.82–2.02)1.05 (0.66–1.66)1.08 (0.69–1.71)1.00 (0.63–1.59) T31.41 (0.89–2.21)0.92 (0.56–1.52)1.19 (0.75–1.90)0.92 (0.55–1.56) P trend 0.1410.7340.4620.766CV T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T20.95 (0.60–1.49)0.75 (0.47–1.19)0.80 (0.50–1.26)0.74 (0.47–1.18) T31.27 (0.83–1.95)0.92 (0.59–1.44)1.07 (0.69–1.65)0.93 (0.59–1.47) P trend 0.2680.7880.7300.837ARV T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T21.45 (0.92–2.29)1.09 (0.68–1.76)1.13 (0.71–1.81)1.05 (0.65–1.69) T31.58 (0.99–2.49)1.08 (0.65–1.79)1.31 (0.82–2.11)1.04 (0.62–1.76) P trend 0.0550.7810.2530.885VIM T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T20.87 (0.55–1.37)0.84 (0.53–1.33)0.83 (0.52–1.31)0.84 (0.53–1.33) T31.19 (0.78–1.81)1.06 (0.69–1.62)1.05 (0.68–1.62)1.11 (0.72–1.71) P trend 0.4230.7790.8050.629 Without diabetes SD T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T20.94 (0.65–1.37)0.99 (0.68–1.45)0.96 (0.66–1.41)0.96 (0.66–1.41) T31.28 (0.90–1.82)1.32 (0.93–1.88)1.26 (0.88–1.79)1.25 (0.88–1.79) P trend 0.1610.1230.1970.208CV T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T21.07 (0.74–1.54)1.13 (0.78–1.63)1.12 (0.77–1.62)1.12 (0.77–1.62) T31.20 (0.84–1.73)1.25 (0.87–1.79)1.20 (0.84–1.73)1.20 (0.84–1.72) P trend 0.3100.2350.3200.324ARV T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T20.94 (0.65–1.34)0.99 (0.69;1.42)0.99 (0.69–1.42)0.98 (0.68–1.42) T31.02 (0.72–1.45)1.06 (0.75–1.51)1.02 (0.71–1.45)1.01 (0.71–1.44) P trend 0.9180.7430.9380.973VIM T11.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) T21.06 (0.73–1.53)1.12 (0.78–1.63)1.11 (0.77–1.61)1.11 (0.77–1.61) T31.24 (0.87–1.78)1.28 (0.89–1.84)1.25 (0.87–1.79)1.25 (0.87–1.80) P trend 0.2340.1840.2240.223Model 1: adjusted for age and sex at baseline (phase 2)Model 2: Model 1 + marital status, education, ever smoking, prevalent CVD, physical activity, anti-diabetic drug use, anti-hypertensive drug, lipid-lowering drug, BMI, WC, SBP, DBP, eGFR, and FPG at baseline (phase 2)Model 3: Model 1 + marital status, education, ever smoking, prevalent CVD, and physical activity at baseline (phase 2), and anti-diabetic drug use, anti-hypertensive drug, and lipid-lowering drug over phases 2–4, and average BMI, WC, SBP, DBP, and eGFR over phases 2–4Model 4: Model 3 + average FPG ## Discussion For the first time, we examined the association between GV over 6 years assessed by SD, CV, ARV, and VIM and incident eGFR decline in both T2D and non-T2D individuals, separately in two well-known cohorts, namely TLGS and MESA during one a decade follow-up. Using eGFR decline ≥ $40\%$ as the outcome, in the MESA, in the fully adjusted model, higher GV using SD and CV measures were significantly associated with a higher risk of eGFR decline among those with T2D; however, among those without T2D, no associations were found. In the TLGS, the higher GV using SD, CV, and VIM measures were significantly associated with a higher risk of eGFR decline only among those without T2D. Applying eGFR decline ≥ $30\%$ as the outcome, in the TLGS, regardless of diabetes status, no association was shown between FPG variability measures and risk of eGFR decline; however, in the MESA the results were in line with those of GFR decline ≥ $40\%$. Moreover, using pooled data from the two cohorts we found that each unit increase in FPG variability with all GV measurements excluding ARV were associated with about $5\%$ higher risk of eGFR decline ≥ $40\%$ only among non-T2D individuals. To the best of our knowledge, several studies [42] have assessed the relationship between GV and incident CKD, ESRD, or diabetic kidney disease (DKD) and eGFR rate decline among T2D patients. GV was assessed by HbA1c variability in all of these studies except two studies [43, 44] that were assessed by both HbA1c and FPG variability. Currently, eGFR decline is considered a validated surrogate endpoint for ESRD in randomized clinical trials (RCTs) as well as in cohort studies [24, 45, 46]. It was shown that while $20\%$ and $30\%$ eGFR decline are extremely susceptible to the existence of acute effects, $40\%$ and $57\%$ are more robust [25]. In an international meta-analysis of more than 1.7 million individuals with incident 12,344 ESRD events, it was shown that a decline in eGFR ≥ $30\%$ and ≥ $40\%$ over a 3-year baseline period was associated with an adjusted HR of 7.0 (3.9–12.7) and 15.7 (7.4–33.4), respectively, compared to no changes in eGFR in those with baseline eGFR ≥ 60 mL/min/1.73 m2 [24]. The corresponding values for T2D Japanese patients were estimated at 18.4 (7.6–44.7) and 12.8 (5.2–32.2) [45]. The following studies [30–32], assessed GV and eGFR decline among T2D patients. All of these studies evaluated the annual changes in eGFR among T2D individuals. Takenouchi et al. found that higher HbA1C-CV was associated with a higher risk of eGFR decline, mainly among those with an albumin/creatinine ratio ≥ 30 mg/g. In this study, no association was found between FPG -GV and eGFR decline [32]. Based on eGFR decline, Low et al. also found that renal disease progression was more common among those with T2D with higher HbA1C-CV independent of mean HbA1C. However, compared with patients with better average glycemic control, T2D patients with sub-optimal average glycemic control (i.e. HbA1C > $8.0\%$)were more likely to develop renal disease at lower magnitudes of HbA1C variability [31]. However, Lin Lee et al. found that even among T2D patients with well-controlled HbA1C levels (< $7\%$), those with high HbA1C-CV still experienced faster eGFR decline[30]. In our data analysis, among T2D patients in MESA, with sub-optima mean FPG level about of 151 mg/dl, FPG variability had a greater likelihood for eGFR decline ≥ 30 and ≥ $40\%$. To the best of our knowledge, there is also only one study that investigates the correlation between VVV and macrovascular and microvascular events in the general population [14]. In a 10-year prospective cohort study, Jang et al. found that the HbA1C-CV tertile was associated with an increased risk for macro-and microvascular events in non-DM middle-aged participants. The higher HbA1C variability was an independent risk factor for microvascular events, however for macrovascular events, the risk was more prominent for variabilities in FPG and post 2-h blood glucose [14]. Our findings in MESA among non-T2D participants are consistent with this study, which found no associations between FPG variability measurements and eGFR decline ≥ $40\%$; however, FPG-SD, FPG-CV, and FPG-VIM were associated with higher eGFR decline ≥ $40\%$ among non-T2D participants in TLGS. Several studies have investigated the impact of ethnicity on eGFR decline [47], progression to ESRD, and development of CKD [48]. Peralta et al. assessed the ethnicity and racial disparities in kidney function decline in participants without CKD [47]. In age- and sex-adjusted models, black individuals had a higher risk of incident CKD among those with eGFR higher than 90 ml/min per 1.73 m2 as well as 60 < eGFR ≤ 90 ml/min per 1.73 m2, followed by Hispanics, while Chinese with 60 < eGFR ≤ 90 ml/min per 1.73 m2 had the lowest risk of incident CKD. The associations attenuated following adjustment for CKD risk factors, particularly hypertension and diabetes. Therefore, the authors concluded that the rate of kidney function decline before incident CKD could be different among various ethnicities, which cannot be fully explained by differences in CKD well-established risk factors [47]. The ethnicity disparity in CKD development in diabetic patients has also been revealed. Hull et al. found a higher rate of CKD development in South Asia relative to the White population [49]. Collectively, ethnicity can influence the rate of kidney function decline and its associations with different risk factors, which could be a reasonable explanation for our different association between GV and eGFR decline in the *American versus* Iranian cohort. Despite our best efforts, no previous studies addressed the impact of GV variability on eGFR decline in both T2D and non-T2D simultaneously. From a pathophysiological perspective, although several potential mechanisms have been proposed to have the potential to connect enhanced GV with a higher risk of incident micro- and macro-vascular complications of diabetes including renal impairment, ESRD, and CKD, the exact mechanism has yet to be determined. There is a bulk of evidence [9, 50–52] in support of the fact that short-term, as well as long-term GV can enhance inflammatory cytokines [53], oxidative stress[53], and induce endothelial dysfunction[44, 54–56], all of which have been shown to have mandatory roles in diabetes complications. GV can increase human tubule-interstitial cell growth, collagen production, and endothelial apoptosis rates compared with persistent exposure to high glucose levels [55, 57, 58]. ## Strengths and limitations The present study contains strengths that worth to be acknowledging. The primary strength of our study is that we examined the impact of exposure to GV on incident eGFR decline in two well-known cohorts among both participants with and without T2D during a long-term follow-up. Second, although the level of adjustment for potential risk factors is different among studies, we adjusted for well-known CKD risk factors in the current study. On the other hand, several limitations should be mentioned. First, we did not have access to the data on HbA1C variability. Of note, Yang et al. demonstrated that both FPG-CV and HbA1c-CV can predict the development of ESRD in diabetes [44]. Furthermore, Jang et al. found out that while HbA1C variability is a better predictor of insulin resistance and inflammatory responses, FPG and 2-HPG are better predictors of sympathoadrenal activation, which was shown to be associated with hypoglycemia [14]. Second, we also did not have access to the data on the urine sample and the albuminuria status. Moreover, it was shown that eGFR is unreliable in detecting renal function in those with diabetes as it overestimates and underestimates measured GFR (mGFR) at lower mGFR and higher mGFR values, respectively [59]; thus, we used eGFR decline as a valid surrogate for renal failure similar to many RCTs [46] and few cohort studies [24, 46, 60]. Third, our study lacked data concerning the episodes of hypoglycemia, which was revealed to enhance the risk of CKD in those with T2D [61]. Forth, we removed participants with eGFR decline ≥ $30\%$ during the period of FPG fluctuations; however, since our exposure period did not proceed outcome in the strict sense (i.e. the time of start for eGFR decline and FPG variability assessment was the same), the absence of mentioned time lag may lead to inverse causality. Finally, the TLGS cohort was performed among residents of the metropolitan city of Tehran; thus, our findings cannot be extrapolated into rural zones of Iran and other ethnicities. It is important to note that the MESA cohort represents four different ethnicities including the white population which constitute the greatest part of the cohort (about $40\%$) followed by African American, Hispanic and Chinese Americans, aged ≥ 45 years [62], so it cannot be extrapolated to the younger age population. ## Conclusion In summary, we found that higher FPG variability is associated with an increased risk of eGFR decline of ≥ 30 and ≥ $40\%$ in the American population with diabetes. However, the unfavorable impact of FPG-GV was found only among the non-diabetic Iranian population for incident eGFR decline of ≥ $30\%$. ## Supplementary Information Additional file 1:Supplementary Table 1. HRs and $95\%$ CIs of incident eGFR decline ≥ $30\%$ according to each unit increase in FPG variability measures in Tehran Lipid and Glucose Study. Supplementary Table 2. HRs and $95\%$ CIs of incident eGFR decline ≥ $30\%$ according to tertiles of FPG variability measures in Tehran Lipid and Glucose Study. Supplementary Table 3. HRs and $95\%$ CIs of incident eGFR decline ≥ $30\%$ according to each unit increase in 2-HPG variability measures in Tehran Lipid and Glucose Study. Supplementary Table 4. HRs and $95\%$ CIs of incident eGFR decline ≥ $30\%$ according to tertiles of 2-HPG variability measures in Tehran Lipid and Glucose Study. Supplementary Table 5. HRs and $95\%$ CIs of incident eGFR decline ≥ $40\%$ according to each unit increase 2-HPG variability measures in Tehran Lipid and Glucose Study. Supplementary Table 6. HRs and $95\%$ CIs of incident eGFR decline ≥ $40\%$ according to each unit increase 2-HPG variability measures in Tehran Lipid and Glucose Study. Supplementary Table 7. HRs and $95\%$ CIs of incident eGFR decline ≥ $30\%$ and $40\%$ according to each unit increase in 2-HPG variability measures in Tehran Lipid and Glucose Study. Supplementary Table 8. HRs and $95\%$ CIs of incident eGFR decline ≥ $30\%$ according to each unit increase in FPG variability measures in Multi-Ethnic Study of Atherosclerosis. Supplementary Table 9. HRs and $95\%$ CIs of incident eGFR decline ≥ $30\%$ according to tertiles of FPG variability measures in Multi-Ethnic Study of Atherosclerosis. ## References 1. Brownlee M, Hirsch IB. **Glycemic variability: a hemoglobin A1c–independent risk factor for diabetic complications**. *JAMA* (2006.0) **295** 1707-1708. DOI: 10.1001/jama.295.14.1707 2. 2.Lachin JM, Genuth S, Nathan DM, Zinman B, Rutledge BN, DCCT/EDIC Research Group. Effect of glycemic exposure on the risk of microvascular complications in the diabetes control and complications trial—revisited. Diabetes. 2008;57(4):995-1001. 3. Ayano-Takahara S, Ikeda K, Fujimoto S, Hamasaki A, Harashima SI, Toyoda K, Fujita Y, Nagashima K, Tanaka D, Inagaki N. **Glycemic variability is associated with quality of life and treatment satisfaction in patients with type 1 diabetes**. *Diabetes Care* (2015.0) **38** e1-e2. DOI: 10.2337/dc14-1801 4. DeVries JH. **Glucose variability: where it is important and how to measure it**. *Diabetes* (2013.0) **62** 1405. DOI: 10.2337/db12-1610 5. Suh S, Kim JH. **Glycemic variability: how do we measure it and why is it important?**. *Diabetes Metab J* (2015.0) **39** 273-282. DOI: 10.4093/dmj.2015.39.4.273 6. Cavalot F. **Do data in the literature indicate that glycaemic variability is a clinical problem? Glycaemic variability and vascular complications of diabetes**. *Diabetes Obes Metab* (2013.0) **15** 3-8. DOI: 10.1111/dom.12140 7. Smith-Palmer J, Brändle M, Trevisan R, Federici MO, Liabat S, Valentine W. **Assessment of the association between glycemic variability and diabetes-related complications in type 1 and type 2 diabetes**. *Diabetes Res Clin Pract* (2014.0) **105** 273-284. DOI: 10.1016/j.diabres.2014.06.007 8. Jin SM, Kim TH, Oh S, Baek J, Joung J, Park SM, Cho Y, Sohn S, Hur K, Lee MS. **Association between the extent of urinary albumin excretion and glycaemic variability indices measured by continuous glucose monitoring**. *Diabet Med* (2015.0) **32** 274-279. DOI: 10.1111/dme.12607 9. Wadén J, Forsblom C, Thorn LM, Gordin D, Saraheimo M. **variability predicts incident cardiovascular events, microalbuminuria, and overt diabetic nephropathy in patients with type 1 diabetes**. *Diabetes* (2009.0) **58** 2649-2655. DOI: 10.2337/db09-0693 10. Muggeo M, Zoppini G, Bonora E, Brun E, Bonadonna RC, Moghetti P, Verlato G. **Fasting plasma glucose variability predicts 10-year survival of type 2 diabetic patients: the verona diabetes study**. *Diabetes Care* (2000.0) **23** 45-50. DOI: 10.2337/diacare.23.1.45 11. **6. Glycemic Targets: Standards of Medical Care in Diabetes—2022**. *Diabetes Care* (2022.0) **45** S83. DOI: 10.2337/dc22-S006 12. Draznin B. **11. Chronic kidney disease and risk management: standards of medical care in diabetes-2022**. *Diabetes Care* (2022.0) **45** S175-S184. DOI: 10.2337/dc22-S011 13. Gorst C, Kwok CS, Aslam S, Buchan I, Kontopantelis E, Myint PK, Heatlie G, Loke Y, Rutter MK, Mamas MA. **Long-term glycemic variability and risk of adverse outcomes: a systematic review and meta-analysis**. *Diabetes Care* (2015.0) **38** 2354-2369. DOI: 10.2337/dc15-1188 14. Jang JY, Moon S, Cho S, Cho KH, Oh CM. **Visit-to-visit HbA1c and glucose variability and the risks of macrovascular and microvascular events in the general population**. *Sci Rep* (2019.0) **9** 1374. DOI: 10.1038/s41598-018-37834-7 15. Cardoso CRL, Leite NC, Moram CBM, Salles GF. **Long-term visit-to-visit glycemic variability as predictor of micro- and macrovascular complications in patients with type 2 diabetes: the Rio de Janeiro type 2 diabetes cohort study**. *Cardiovasc Diabetol* (2018.0) **17** 33. DOI: 10.1186/s12933-018-0677-0 16. Barzegar N, Ramezankhani A, Tohidi M, Azizi F, Hadaegh F. **Long-term glucose variability and incident cardiovascular diseases and all-cause mortality events in subjects with and without diabetes: Tehran Lipid and glucose study**. *Diabetes Res Clin Pract* (2021.0) **178** 108942. DOI: 10.1016/j.diabres.2021.108942 17. Penno G, Solini A, Bonora E, Fondelli C, Orsi E, Zerbini G, Morano S, Cavalot F, Lamacchia O, Laviola L. **HbA1c variability as an independent correlate of nephropathy, but not retinopathy, in patients with type 2 diabetes: the Renal Insufficiency And Cardiovascular Events (RIACE) Italian multicenter study**. *Diabetes Care* (2013.0) **36** 2301-2310. DOI: 10.2337/dc12-2264 18. Cummings DM, Larsen LC, Doherty L, Lea CS, Holbert D. **Glycemic control patterns and kidney disease progression among primary care patients with diabetes mellitus**. *J Am Board Fam Med* (2011.0) **24** 391-398. DOI: 10.3122/jabfm.2011.04.100186 19. Moazzeni SS, Arani RH, Hasheminia M, Tohidi M, Azizi F, Hadaegh F. **High incidence of chronic kidney disease among Iranian diabetic adults: using CKD-EPI and MDRD equations for estimated glomerular filtration rate**. *Diabetes Metab J* (2021.0) **45** 684-697. DOI: 10.4093/dmj.2020.0109 20. Feng XS, Farej R, Dean BB, Xia F, Gaiser A, Kong SX, Elliott J, Lindemann S, Singh R. **CKD prevalence among patients with and without type 2 diabetes: regional differences in the United States**. *Kidney Med* (2022.0) **4** 100385. DOI: 10.1016/j.xkme.2021.09.003 21. Bikbov B, Purcell CA, Levey AS, Smith M, Abdoli A, Abebe M, Adebayo OM, Afarideh M, Agarwal SK, Agudelo-Botero M. **Global, regional, and national burden of chronic kidney disease, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017**. *The lancet* (2020.0) **395** 709-733. DOI: 10.1016/S0140-6736(20)30045-3 22. Bouya S, Balouchi A, Rafiemanesh H, Hesaraki M. **Prevalence of chronic kidney disease in Iranian general population: a meta-analysis and systematic review**. *Ther Apher Dial* (2018.0) **22** 594-599. DOI: 10.1111/1744-9987.12716 23. Heerspink HJL, Tighiouart H, Sang Y, Ballew S, Mondal H, Matsushita K, Coresh J, Levey AS, Inker LA. **GFR decline and subsequent risk of established kidney outcomes: a meta-analysis of 37 randomized controlled trials**. *Am J Kidney Dis* (2014.0) **64** 860-866. DOI: 10.1053/j.ajkd.2014.08.018 24. Coresh J, Turin TC, Matsushita K, Sang Y, Ballew SH, Appel LJ, Arima H, Chadban SJ, Cirillo M, Djurdjev O. **Decline in estimated glomerular filtration rate and subsequent risk of end-stage renal disease and mortality**. *JAMA* (2014.0) **311** 2518-2531. DOI: 10.1001/jama.2014.6634 25. Greene T, Teng C-C, Inker LA, Redd A, Ying J, Woodward M, Coresh J, Levey AS. **Utility and validity of estimated GFR–based surrogate time-to-event end points in CKD: a simulation study**. *Am J Kidney Dis* (2014.0) **64** 867-879. DOI: 10.1053/j.ajkd.2014.08.019 26. Inker LA, Heerspink HJL, Mondal H, Schmid CH, Tighiouart H, Noubary F, Coresh J, Greene T, Levey AS. **GFR decline as an alternative end point to kidney failure in clinical trials: a meta-analysis of treatment effects from 37 randomized trials**. *Am J Kidney Dis* (2014.0) **64** 848-859. DOI: 10.1053/j.ajkd.2014.08.017 27. Levey AS, Inker LA, Matsushita K, Greene T, Willis K, Lewis E, De Zeeuw D, Cheung AK, Coresh J. **GFR decline as an end point for clinical trials in CKD: a scientific workshop sponsored by the National Kidney Foundation and the US Food and Drug Administration**. *Am J Kidney Dis* (2014.0) **64** 821-835. DOI: 10.1053/j.ajkd.2014.07.030 28. Badve SV, Palmer SC, Hawley CM, Pascoe EM, Strippoli GF, Johnson DW. **Glomerular filtration rate decline as a surrogate end point in kidney disease progression trials**. *Nephrol Dial Transplant* (2016.0) **31** 1425-1436. DOI: 10.1093/ndt/gfv269 29. Chang WX, Asakawa S, Toyoki D, Nemoto Y, Morimoto C, Tamura Y, Ota T, Shibata S, Fujigaki Y, Shen ZY. **Predictors and the subsequent risk of end-stage renal disease–usefulness of 30% decline in estimated GFR over 2 years**. *PLoS ONE* (2015.0) **10** e0132927. DOI: 10.1371/journal.pone.0132927 30. Lee CL, Chen CH, Wu MJ, Tsai SF. **The variability of glycated hemoglobin is associated with renal function decline in patients with type 2 diabetes**. *Ther Adv Chronic Dis* (2020.0) **11** 2040622319898370. DOI: 10.1177/2040622319898370 31. Low S, Lim SC, Yeoh LY, Liu YL, Liu JJ, Fun S, Su C, Zhang X, Subramaniam T, Sum CF. **Effect of long-term glycemic variability on estimated glomerular filtration rate decline among patients with type 2 diabetes mellitus: Insights from the Diabetic Nephropathy Cohort in Singapore**. *J Diabetes* (2017.0) **9** 908-919. DOI: 10.1111/1753-0407.12512 32. Takenouchi A, Tsuboi A, Terazawa-Watanabe M, Kurata M, Fukuo K, Kazumi T. **Direct association of visit-to-visit HbA1c variation with annual decline in estimated glomerular filtration rate in patients with type 2 diabetes**. *J Diabetes Metab Disord* (2015.0) **14** 69. DOI: 10.1186/s40200-015-0201-y 33. Azizi F, Zadeh-Vakili A, Takyar M. **Review of rationale, design, and initial findings: Tehran lipid and glucose study**. *Int J Endocrinol Metab* (2018.0) **16** e84777. PMID: 30584442 34. Azizi F, Ghanbarian A, Momenan AA, Hadaegh F, Mirmiran P, Hedayati M, Mehrabi Y, Zahedi-Asl S. **Prevention of non-communicable disease in a population in nutrition transition: Tehran lipid and glucose study phase II**. *Trials* (2009.0) **10** 5. DOI: 10.1186/1745-6215-10-5 35. Bild DE, Bluemke DA, Burke GL, Detrano R, Diez Roux AV, Folsom AR, Greenland P, JacobsJr DR, Kronmal R, Liu K. **Multi-ethnic study of atherosclerosis: objectives and design**. *Am J Epidemiol* (2002.0) **156** 871-881. DOI: 10.1093/aje/kwf113 36. Momenan AA, Delshad M, Sarbazi N, RezaeiGhaleh N, Ghanbarian A, Azizi F. **Reliability and validity of the Modifiable Activity Questionnaire (MAQ) in an Iranian urban adult population**. *Arch Iran Med* (2012.0) **15** 279-282. PMID: 22519376 37. Jeon CY, Lokken RP, Hu FB, van Dam RM. **Physical activity of moderate intensity and risk of type 2 diabetes: a systematic review**. *Diabetes Care* (2007.0) **30** 744-752. DOI: 10.2337/dc06-1842 38. Tohidi M, Hasheminia M, Mohebi R, Khalili D, Hosseinpanah F, Yazdani B, Nasiri AA, Azizi F, Hadaegh F. **Incidence of chronic kidney disease and its risk factors, results of over 10 year follow up in an Iranian cohort**. *PLoS ONE* (2012.0) **7** e45304-e45304. DOI: 10.1371/journal.pone.0045304 39. Mena L, Pintos S, Queipo NV, Aizpúrua JA, Maestre G, Sulbarán T. **A reliable index for the prognostic significance of blood pressure variability**. *J Hypertens* (2005.0) **23** 505-511. DOI: 10.1097/01.hjh.0000160205.81652.5a 40. Kazancioğlu R. **Risk factors for chronic kidney disease: an update**. *Kidney Int Suppl (2011)* (2013.0) **3** 368-371. DOI: 10.1038/kisup.2013.79 41. Habte-Asres HH, Wheeler DC, Forbes A. **The association between glycaemic variability and progression of chronic kidney disease: a systematic review**. *SN Comprehensive Clin Med* (2022.0) **4** 102. DOI: 10.1007/s42399-022-01182-5 42. Habte-Asres HH, Wheeler DC, Forbes A. **The association between glycaemic variability and progression of chronic kidney disease: a systematic review**. *SN Comprehensive Clin Med* (2022.0) **4** 1-14. DOI: 10.1007/s42399-022-01182-5 43. Lin C-C, Chen C-C, Chen F-N, Li C-I, Liu C-S, Lin W-Y, Yang S-Y, Lee C-C, Li T-C. **Risks of diabetic nephropathy with variation in hemoglobin A1c and fasting plasma glucose**. *Am J Med* (2013.0) **126** e1011-1017. DOI: 10.1016/j.amjmed.2013.04.015 44. Yang Y-F, Li T-C, Li C-I, Liu C-S, Lin W-Y, Yang S-Y, Chiang J-H, Huang C-C, Sung F-C, Lin C-C. **Visit-to-visit glucose variability predicts the development of end-stage renal disease in type 2 diabetes: 10-year follow-up of Taiwan diabetes study**. *Medicine* (2015.0) **94** e1804. DOI: 10.1097/MD.0000000000001804 45. Oshima M, Toyama T, Haneda M, Furuichi K, Babazono T, Yokoyama H, Iseki K, Araki S, Ninomiya T, Hara S. **Estimated glomerular filtration rate decline and risk of end-stage renal disease in type 2 diabetes**. *PLoS ONE* (2018.0) **13** e0201535. DOI: 10.1371/journal.pone.0201535 46. Inker LA, Chaudhari J. **GFR slope as a surrogate endpoint for CKD progression in clinical trials**. *Curr Opin Nephrol Hypertens* (2020.0) **29** 581-590. DOI: 10.1097/MNH.0000000000000647 47. Peralta CA, Katz R, DeBoer I, Ix J, Sarnak M, Kramer H, Siscovick D, Shea S, Szklo M, Shlipak M. **Racial and ethnic differences in kidney function decline among persons without chronic kidney disease**. *J Am Soc Nephrol* (2011.0) **22** 1327-1334. DOI: 10.1681/ASN.2010090960 48. Patzer RE, McClellan WM. **Influence of race, ethnicity and socioeconomic status on kidney disease**. *Nat Rev Nephrol* (2012.0) **8** 533-541. DOI: 10.1038/nrneph.2012.117 49. Dreyer G, Hull S, Mathur R, Chesser A, Yaqoob M. **Progression of chronic kidney disease in a multi-ethnic community cohort of patients with diabetes mellitus**. *Diabet Med* (2013.0) **30** 956-963. DOI: 10.1111/dme.12197 50. Ceriello A. **The possible role of postprandial hyperglycaemia in the pathogenesis of diabetic complications**. *Diabetologia* (2003.0) **46** M9-M16. DOI: 10.1007/s00125-002-0931-5 51. Monnier L, Colette C, Rabasa-Lhoret R, Lapinski H, Caubel C, Avignon A, Boniface H. **Morning hyperglycemic excursions: a constant failure in the metabolic control of non–insulin-using patients with type 2 diabetes**. *Diabetes Care* (2002.0) **25** 737-741. DOI: 10.2337/diacare.25.4.737 52. Sparkenbaugh E, Pawlinski R. **Interplay between coagulation and vascular inflammation in sickle cell disease**. *Br J Haematol* (2013.0) **162** 3-14. DOI: 10.1111/bjh.12336 53. Esposito K, Nappo F, Marfella R, Giugliano G, Giugliano F, Ciotola M, Quagliaro L, Ceriello A, Giugliano D. **Inflammatory cytokine concentrations are acutely increased by hyperglycemia in humans: role of oxidative stress**. *Circulation* (2002.0) **106** 2067-2072. DOI: 10.1161/01.CIR.0000034509.14906.AE 54. Ceriello A, Esposito K, Piconi L, Ihnat MA, Thorpe JE, Testa R, Boemi M, Giugliano D. **Oscillating glucose is more deleterious to endothelial function and oxidative stress than mean glucose in normal and type 2 diabetic patients**. *Diabetes* (2008.0) **57** 1349-1354. DOI: 10.2337/db08-0063 55. Monnier L, Mas E, Ginet C, Michel F, Villon L, Cristol J-P, Colette C. **Activation of oxidative stress by acute glucose fluctuations compared with sustained chronic hyperglycemia in patients with type 2 diabetes**. *JAMA* (2006.0) **295** 1681-1687. DOI: 10.1001/jama.295.14.1681 56. Ceriello A, Ihnat M. **‘Glycaemic variability’: a new therapeutic challenge in diabetes and the critical care setting**. *Diabet Med* (2010.0) **27** 862-867. DOI: 10.1111/j.1464-5491.2010.02967.x 57. Brownlee M. **Biochemistry and molecular cell biology of diabetic complications**. *Nature* (2001.0) **414** 813-820. DOI: 10.1038/414813a 58. Brownlee M. **The pathobiology of diabetic complications: a unifying mechanism**. *Diabetes.* (2005.0) **54** 1615-1625. DOI: 10.2337/diabetes.54.6.1615 59. Wood AJ, Churilov L, Perera N, Thomas D, Poon A, MacIsaac RJ, Jerums G, Ekinci EI. **Estimating glomerular filtration rate: performance of the CKD-EPI equation over time in patients with type 2 diabetes**. *J Diabetes Complications* (2016.0) **30** 49-54. DOI: 10.1016/j.jdiacomp.2015.08.025 60. Schrauben SJ, Shou H, Zhang X, Anderson AH, Bonventre JV, Chen J, Coca S, Furth SL, Greenberg JH, Gutierrez OM. **Association of multiple plasma biomarker concentrations with progression of prevalent diabetic kidney disease: findings from the chronic renal insufficiency cohort (CRIC) study**. *J Am Soc Nephrol* (2021.0) **32** 115-126. DOI: 10.1681/ASN.2020040487 61. Yun J-S, Ko S-H, Ko S-H, Song K-H, Ahn Y-B, Yoon K-H, Park Y-M, Ko S-H. **Presence of macroalbuminuria predicts severe hypoglycemia in patients with type 2 diabetes: a 10-year follow-up study**. *Diabetes Care* (2013.0) **36** 1283-1289. DOI: 10.2337/dc12-1408 62. Lakoski SG, Cushman M, Criqui M, Rundek T, Blumenthal RS, D'Agostino RB, Herrington DM. **Gender and C-reactive protein: data from the Multiethnic Study of Atherosclerosis (MESA) cohort**. *Am Heart J* (2006.0) **152** 593-598. DOI: 10.1016/j.ahj.2006.02.015
--- title: Association between exposure to multiple air pollutants, transportation noise and cause-specific mortality in adults in Switzerland authors: - Danielle Vienneau - Massimo Stafoggia - Sophia Rodopoulou - Jie Chen - Richard W. Atkinson - Mariska Bauwelinck - Jochem O. Klompmaker - Bente Oftedal - Zorana J. Andersen - Nicole A. H. Janssen - Rina So - Youn-Hee Lim - Benjamin Flückiger - Regina Ducret-Stich - Martin Röösli - Nicole Probst-Hensch - Nino Künzli - Maciek Strak - Evangelia Samoli - Kees de Hoogh - Bert Brunekreef - Gerard Hoek journal: Environmental Health year: 2023 pmcid: PMC10041702 doi: 10.1186/s12940-023-00983-y license: CC BY 4.0 --- # Association between exposure to multiple air pollutants, transportation noise and cause-specific mortality in adults in Switzerland ## Abstract ### Background Long-term exposure to air pollution and noise is detrimental to health; but studies that evaluated both remain limited. This study explores associations with natural and cause-specific mortality for a range of air pollutants and transportation noise. ### Methods Over 4 million adults in Switzerland were followed from 2000 to 2014. Exposure to PM2.5, PM2.5 components (Cu, Fe, S and Zn), NO2, black carbon (BC) and ozone (O3) from European models, and transportation noise from source-specific Swiss models, were assigned at baseline home addresses. Cox proportional hazards models, adjusted for individual and area-level covariates, were used to evaluate associations with each exposure and death from natural, cardiovascular (CVD) or non-malignant respiratory disease. Analyses included single and two exposure models, and subset analysis to study lower exposure ranges. ### Results During follow-up, 661,534 individuals died of natural causes ($36.6\%$ CVD, $6.6\%$ respiratory). All exposures including the PM2.5 components were associated with natural mortality, with hazard ratios ($95\%$ confidence intervals) of 1.026 (1.015, 1.038) per 5 µg/m3 PM2.5, 1.050 (1.041, 1.059) per 10 µg/m3 NO2, 1.057 (1.048, 1.067) per 0.5 × 10–5/m BC and 1.045 (1.040, 1.049) per 10 dB Lden total transportation noise. NO2, BC, Cu, Fe and noise were consistently associated with CVD and respiratory mortality, whereas PM2.5 was only associated with CVD mortality. Natural mortality associations persisted < 20 µg/m3 for PM2.5 and NO2, < 1.5 10–5/m BC and < 53 dB Lden total transportation noise. The O3 association was inverse for all outcomes. Including noise attenuated all outcome associations, though many remained significant. Across outcomes, noise was robust to adjustment to air pollutants (e.g. natural mortality 1.037 (1.033, 1.042) per 10 dB Lden total transportation noise, after including BC). ### Conclusion Long-term exposure to air pollution and transportation noise in Switzerland contribute to premature mortality. Considering co-exposures revealed the importance of local traffic-related pollutants such as NO2, BC and transportation noise. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12940-023-00983-y. ## Introduction Air pollution is an important contributor to morbidity and mortality, with an estimated 6.7 million deaths due to long-term exposure to ambient particulate air pollution worldwide [1]. Though air pollution levels in Europe and premature deaths attributed to particulate matter < 2.5 µm (PM2.5) have generally declined since the mid 2000’s, the burden of disease remains high [2]. The much reduced recommendations for low limit values in the 2021 WHO air quality guidelines (AQG) [3], and the growing body of evidence from developed nations, clearly signal that even low levels of air pollution are harmful [4–9]. Likewise exposure to noise from transportation sources is also known to be detrimental to health and linked to mortality [10–15], with the most recent health policy instrument being the WHO Environmental Noise Guidelines (ENG) for the European Region [16]. Though studies with similar spatially resolved air pollution and noise exposure data are limited, some suggest the effects of noise are independent [12, 17–19] thus leading to additional health burden from diseases also associated with air pollution. For example, a comparative health risk assessment for Switzerland using data from 2010 indicated the external costs of the transportation noise burden is equal to that of air pollution [20]. The multicenter Effects of Low-Level Air Pollution: A Study in Europe (ELAPSE) study investigated associations of long-term exposure to air pollution with natural and cause-specific mortality in both a pooled analysis of over 300,000 adults from eight population-based cohorts with detailed confounder data [21], and a meta-analysis of seven administrative cohorts—including the Swiss National Cohort (SNC)—for a total of over 28 million adults [22]. In both analyses, PM2.5, nitrogen dioxide (NO2) and black carbon (BC) were clearly associated with increased risk of natural, cardiovascular (CVD) and non-malignant respiratory mortality. These associations persisted in subsets of the population residing in areas with low concentrations; and for CVD, the outcome for which noise was considered, the air pollution associations were generally robust to adjustment for transportation noise. In six of the administrative cohorts including the SNC [23], and the pooled cohorts in ELAPSE [24], further investigation of eight PM2.5 components (including copper (Cu), iron (Fe), sulphur (S) and zinc (Zn)) showed associations with natural mortality for most components that often attenuated after considering PM2.5 mass. All air pollution concentration estimates were derived from harmonized, Europe-wide models [25, 26]. Though transportation noise was considered in some administrative cohorts of ELAPSE as a further adjustment for CVD, it was given less attention because the exposure models were cohort specific and thus heterogeneous. Since publication of the WHO ENG [16], newer studies on other outcomes have suggested noise may also be related to respiratory disease and natural mortality [11, 14, 17]. The often high correlations between exposures deriving from similar sources make it difficult to discern which exposure, or combination thereof, drive the mortality associations within any given jurisdiction. The spatial variation and levels of some exposures, such as PM2.5 components and BC, further depend on the local context in which other co-exposures related to traffic, such as transportation noise, may also play a role. Furthermore, air pollutant concentration maps are often estimated via land-use regression approaches, potentially increasing their correlations when similar predictors are used for different pollutants. The complexity of the total exposure environment for disentangling health effects imposes clear challenges to the downstream regulation and public health decision making. Insight into the country-specific mortality associations is thus crucial for the authorities engaged in setting guidelines and standards for air quality and noise, given that the Environmental Law requires protection of all people from emissions that harm health or impair wellbeing [27]. Extending the analyses in ELAPSE to more thoroughly study transportation noise as a co-exposure, the aim of this study was to investigate the associations between long-term exposure to air pollution and total transportation noise with natural, CVD and respiratory mortality in the SNC. To assess the independent effects of air pollution and transportation noise, particular emphasis was on specification of two exposure models. An additional aim was to gain insight into the associations at very low levels via subset analyses. Exposures included PM2.5, NO2, BC and warm season O3 (subsequently referred to as O3 warm), several PM2.5 components (Cu, Fe, S and Zn) and transportation noise. ## Study population The Swiss National Cohort (SNC) is an administrative cohort that links the decennial national census (for 1990 and 2000) and Registry Based Census (from 2010 onward) with births, mortality and emigration [28, 29]. With compulsory participation in the census, virtually all residents in Switzerland are represented, i.e. $98.6\%$ in the 04 December 2000 census [30]. The SNC was approved by the Ethics Committees of the Cantons of Zurich and Bern. Treated as a closed cohort, this study used data from 04 December 2000 (i.e. baseline date) to 31 December 2014 (i.e. end of follow-up). In total, 7.28 million observations of all ages were available at baseline. With the focus on mortality outcomes in adults, the analysis included 4.19 million individuals after excluding the following observations: individuals below 30 years of age ($$n = 2$.6$ million) and those with missing residential coordinates or designated as living in an institution ($$n = 0$.4$ million). A further 0.1 million with missing individual characteristics, specifically education and socio-economic position (SEP), were excluded to ensure a study population with complete data for use in all analyses (Supplement Table S1). ## Outcome definition The outcomes under investigation were primary causes of death from natural causes (International Classification of Disease version 10 [ICD-10]: A00 – R99), CVD (ICD-10: I10 – I70), and non-malignant respiratory disease (ICD-10: J00 – J99). ## Exposure assessment Annual average ambient PM2.5, NO2 and BC concentrations, in addition to O3 warm (based on maximum 8-h running means, during April to September), were available from the European 100 × 100 m hybrid land use regression models for 2010 developed within ELAPSE [26]. Exposure to PM2.5 components including the four investigated here (Cu, Fe and Zn as indicators of non-tailpipe emissions, and S representing long-range transported inorganic aerosols) were also developed within ELAPSE [25]; models derived with supervised linear regression were used. Five-fold hold-out validation was used to evaluate the models, with R2: 0.66 for PM2.5; 0.58 for NO2; 0.51 for BC; 0.60 for O3 warm; 0.48 for Cu; 0.48 for Fe; 0.41 for Zn and 0.79 for S [25, 26]. These models for 2010 were a priori defined as the main exposure models in ELAPSE, and were assigned to participant baseline addresses. Transportation noise exposure for Switzerland derived from the SiRENE project (Short and Long Term Effects of Transportation Noise Exposure) [31, 32], in which noise was modelled by source at each dwelling façade (i.e. by floor of residence) at decennial census years. Models for road traffic, railways and aircraft were respectively based on the sonRoad (with StL-86 propagation model), sonRail (with SEMIBEL propagation model) and FLULA2. The Lden metric (i.e. weighted energetic average of Leq,day (07:00–19:00), Leq,evening (19:00–23:00) and Leq,night (23:00–07:00) with a respective penalty of 5 and 10 dB applied to the evening and night) was computed for each noise source. The energetic sum of these three sources was then determined to derive total transportation noise for each dwelling [33], based on the noise level at the maximum exposed façade. This was the a priori main noise exposure; road traffic noise was included for comparison. Noise exposure from the year 2001 was used to align with the cohort baseline. Noise exposure is stable over time, with moderate to high correlations (r 0.67–0.96 depending on source) previously reported for the SNC [15]. ## Covariates Individual-level characteristics included sex (female/male), marital status (single, married, widowed, divorced), education level (compulsory education or less, upper secondary level education, tertiary level education), mother tongue (German and Rhaeto-Romansch, French, Italian, other language), and nationality (Swiss, non-Swiss). Area-level SEP variables were also developed to provide broader area context not captured by the individual covariates. These were calculated by aggregating the relevant individual variables, including the Swiss-SEP index of socio-economic position (i.e. calculated for small local areas of 50 nearest neighbours as described in Panczak et al. [ 34]). The following were calculated at “neighbourhood” ($$n = 3$$,175 postcode areas) and “regional” ($$n = 26$$ Swiss cantons) level: composite score (mean of the Swiss-SEP index); unemployment rate (% working age population [20 to 65 years] unemployed); low education rate (% adults with compulsory or less education); and high education rate (% adults with tertiary education or higher). All covariates were available for the baseline year. ## Statistical analysis Associations were analysed using the Cox proportional hazards model, with age as the underlying time scale. Models were stratified by sex, and clustered by neighbourhood to properly adjust the standard errors of the estimates for the correlations of subjects residing in the same neighbourhood [22, 35]. Participants were followed until the event, emigration, death by another cause or end of follow-up, which ever came first. The adjustment strategy followed the ELAPSE study protocol. Model 1 included only age (time axis), sex (as strata) and neighbourhood (as cluster). Model 2 added individual level variables applicable to the Swiss context, specifically: education level, occupational status, marital status, country of origin, and mother tongue. Model 3, the a priori main model, added the four types of area-level SEP variables described above, at both the neighbourhood and regional level. The two exposure models were based on Model 3, and included a further adjustment for one of the following main exposures, in turn: PM2.5, NO2, BC, O3 warm or total transportation noise. Hazard ratios were computed including linear terms for the exposure(s) and were expressed per standard units (PM2.5 per 5 µg/m3, NO2 per 10 µg/m3, BC per 0.5 × 10–5/m, O3 warm per 10 µg/m3; PM2.5 Cu per 5 ng/m3, PM2.5 Fe per 100 ng/m3, PM2.5 S per 200 ng/m3, PM2.5 Zn per 10 ng/m3; and total transportation noise or road traffic noise exposure per 10 dB Lden). Previous analyses showed linear to supralinear associations in this cohort [15, 22], thus linear exposure terms were considered justified. To look deeper into the associations at very low levels of the main exposures, subsets were defined by removing participants residing in areas above pre-specified values including guideline limit values. Thus using main Model 3, associations below the following levels were investigated: PM2.5 below 25 (EU limit value), 20, 15 and 12 μg/m3; NO2 below 40 (EU limit value; former 2005 WHO air quality guideline value), 30 and 20 μg/m3; BC below 3.0, 2.5, 2.0 and 1.5 10–5/m and O3 warm below 120 and 100 μg/m3. There were an insufficient number of exposed individuals to investigate associations for PM2.5 and NO2 below 5 and 10 μg/m3 (2021 WHO air quality guideline values). Associations below 60, 55 and 53 dB Lden total transportation noise, with the latter being the WHO ENG level for road traffic noise [16], were also investigated. Additional analyses based on two exposure models included investigating effect modification by sex as well as a non-movers analysis. All analyses were conducted in R v3.4.0 using common scripts developed in ELAPSE. ## Study description In total 4,188,175 adults over the age of 30 years on 04 December 2000 with complete exposure and covariate data were included in the cohort (Table S1). The cohort included slightly more women than men ($52.0\%$), with a high proportion of Swiss nationals ($83.1\%$). The majority had a mother tongue of German or Rhaeto-Romansch ($65.1\%$), were employed ($61.4\%$), and married ($69.3\%$) at baseline. During follow up (mean of 12.7 years, and 53,344,296 person years), 661,534 deaths due to natural causes were recorded, of which $36.6\%$ and $6.6\%$ were due to CVD and respiratory mortality, respectively. The mean age of death was over 70 years for all investigated causes. Compared to the study population, a notably higher proportion of deaths were in those who were widowed or had low education (Table 1).Table 1Characteristics of the study population and deaths during follow-upVariableStudy population (SNC)Deaths during follow-upNatural causeCardiovascularRespiratoryN participants total4,293,521673,946246,84044,489N participants with complete data4,188,175661,534241,98543,612N person years53,344,296Individual level covariates Age (mean +–Sd)52.7 (15.2)72.1 (12.3)75.7 (10.9)74.6 (10.6) Women (%)52.050.452.944.7 Country origin (% Swiss)83.192.494.192.8 *Marital status* (%) Single14.09.69.29.7 Married69.355.751.154.0 Divorced8.77.86.48.3 Widowed8.126.833.327.9 Education (%) Low (compulsory education or less)24.540.444.643.8 Medium (upper secondary level)52.746.544.045.0 High (tertiary level)22.713.211.311.2 *Occupational status* (%) Employed/self-employed61.418.111.911.5 Unemployed2.21.10.70.9 Homemaker14.68.34.87.3 Retired21.872.482.680.4 Mother tongue (%) German and Rhaeto-Romansch65.169.873.765.6 French19.620.918.524.6 Italian7.46.96.17.8 Other8.02.31.72.0Area level covariates (mean +–Sd) Neighborhood socio-economic position (SEP) score63.0 (7.3)62.7 (7.3)62.6 (7.3)62.1 (7.4) Neighborhood unemployment rate3.5 (1.5)3.5 (1.6)3.5 (1.5)3.6 (1.6) Neighborhood low education rate28.4 (7.3)28.9 (7.4)28.9 (7.4)29.4 (7.5) Neighborhood high education rate19.8 (7.5)19.6 (7.5)19.2 (7.3)19.4 (7.6)Exposures (mean +–Sd) PM2.5 (µg/m3)15.9 (2.4)16.0 (2.4)15.9 (2.4)15.9 (2.6) NO2 (µg/m3)23.7 (7.4)24.0 (7.6)23.6 (7.5)24.0 (7.9) BC (10–5/m)1.67 (0.35)1.69 (0.36)1.67 (0.35)1.69 (0.40) O3 warm (µg/m3)94.8 (5.9)94. 7 (6.1)94.9 (6.0)94.8 (6.3) PM2.5 Cu (ng/m3)5.2 (2.7)5.3 (2.7)5.1 (2.7)5.2 (2.9) PM2.5 Fe (ng/m3)108.4 (46.9)110.3 (48.0)107.9 (46.7)110.1 (49.7) PM2.5 S (ng/m3)646.6 (85.0)646.9 (87.8)644.1 (87.4)644.0 (92.3) PM2.5 Zn (ng/m3)20.8 (18.5)21.4 (19.0)21.0 (19.1)21.0 (18.9) Total transportation noise (dB Lden)55.9 (8.2)56.4 (8.2)56.3 (8.2)56.6 (8.3) Road traffic noise (dB Lden)54.2 (8.1)54.8 (8.1)54.8 (8.2)55.0 (8.2) Mean (standard deviation, SD) exposures were 15.9 µg/m3 (2.4) for PM2.5, 23.7 µg/m3 (7.4) for NO2, 1.67 10–5/m (0.35) for BC and 94.8 µg/m3 (5.9) for O3 warm. As indicated by the small standard deviation, the contrast for O3 warm was small. The most abundant PM2.5 components were S (646.6 [85.0] ng/m3) followed by Fe (108.4 [46.9] ng/m3) (Table 1, Table S2). Spearman correlations amongst the air pollutant exposures were moderate to high. For the main pollutants the highest correlation was between NO2 and BC (0.91), and both were correlated > 0.7 with PM2.5. O3 warm was negatively correlated with all exposures, the highest being NO2, BC and PM2.5 (< -0.65). Cu and Fe were highly correlated with NO2 and BC (> 0.88). Amongst PM2.5 components, Cu and Fe were almost perfectly correlated (0.97). The mean (SD) total transportation noise and road traffic noise exposures were 55.9 (8.2) and 54.2 (8.1) dB Lden, respectively. Total transportation noise was dominated by road traffic noise, with a correlation of 0.89 in this study population (Table S3). In terms of correlations, noise was low to moderately correlated (within ± 0.40) with each air pollutant including the PM2.5 components. Most exposures also showed a low positive correlation with neighbourhood SEP score, indicating higher exposures in high SEP neighbourhoods (with these in urban areas in Switzerland). Exceptions were O3 warm in the opposite direction (low negative) and noise which were uncorrelated (Table S3). The corresponding mean exposures by quintiles of neighbourhood SEP score showed a consistent pattern of higher air pollution (lower for O3 warm) with higher SEP across the whole population, and within urban and rural populations (Table S4). ## Single and two exposure models For most air pollution exposures the hazard ratios became stronger with increasing covariate adjustment from Model 1 to main Model 3. The opposite patterns were found for O3 warm (i.e. stronger inverse associations with increasing adjustment). A clear pattern in covariate adjustment for total transportation noise was less obvious (Table S5). All exposures were associated with natural mortality in single exposure models, including O3 warm though in the inverse direction (Table 2). The hazard ratios were 1.026 (1.015, 1.038) per 5 µg/m3 PM2.5, 1.050 (1.041, 1.059) per 10 µg/m3 NO2, 1.057 (1.048, 1.067) per 0.5 × 10–5/m BC, 0.946 (0.939, 0.954) per 10 µg/m3 O3 warm and 1.045 (1.040, 1.049) per 10 dB Lden total transportation noise (road traffic noise as exposure gave highly similar results). PM2.5 components (Cu, Fe, S and Zn) were also significantly associated with natural mortality in single exposure models. PM2.5 (mass and the individual components) associations attenuated, often to the null, after adjusting for NO2, BC or O3 warm; whereas including PM2.5 did not change the other single air pollutant exposure associations. O3 warm associations slightly attenuated when noise was introduced. NO2 and BC associations only slightly decreased though remained indicative of an association in models including noise, with slightly more attenuation when adjusting for road traffic vs. total transportation noise. The noise associations were robust to air pollution co-exposure adjustment; for example, the natural mortality association after BC adjustment was 1.037 (1.033, 1.042) per 10 dB Lden total transportation noise. HRs for noise were less affected by air pollution adjustment than air pollution HRs were for noise adjustment. Table 2Single vs. two exposure model hazard ratios ($95\%$ confidence intervals) for associations with mortality by cause (Model 3)OutcomeExposureIncrementSingleAdjusted for PM2.5Adjusted for NO2Adjusted for BCAdjusted for O3 warmAdjusted for total noiseAdjusted for road traffic noiseNatural cause mortalityPM2.55 µg/m31.026 (1.015, 1.038)-0.992 (0.978, 1.005)0.989 0.976, 1.002)1.000 (0.989, 1.012)1.012 (1.001, 1.023)1.013 (1.001, 1.024)NO210 µg/m31.050 (1.041, 1.059)1.053 (1.043, 1.063)-1.017 (1.003, 1.032)1.033 (1.023, 1.043)1.029 (1.021, 1.038)1.026 (1.017, 1.035)BC0.5 × 10–5/m1.057 (1.048, 1.067)1.062 (1.051, 1.072)1.041 (1.026, 1.056)-1.040 (1.029, 1.051)1.035 (1.026, 1.044)1.031 (1.022, 1.041)O3 warm10 µg/m30.946 (0.939, 0.954)0.946 (0.938, 0.955)0.965 (0.955, 0.975)0.969 (0.959, 0.978)-0.962 (0.955, 0.970)0.963 (0.956, 0.971)PM2.5 Cu5 ng/m31.067 (1.054, 1.080)1.064 (1.050, 1.078)1.014 (0.996, 1.033)0.993 (0.973, 1.014)1.037 (1.022,1.051)1.040 (1.028,1.053)1.035 (1.023, 1.048)PM2.5 Fe100 ng/m31.085 (1.070, 1.100)1.082 (1.066, 1.098)1.041 (1.018, 1.064)1.020 (0.994, 1.047)1.056 (1.039,1.072)1.052 (1.038,1.067)1.046 (1.032, 1.061)PM2.5 S200 ng/m31.035 (1.020, 1.051)1.027 (1.009, 1.045)0.989 (0.972, 1.007)0.989 (0.973, 1.005)0.995 (0.980,1.011)1.014 (1.000,1.028)1.012 (0.998, 1.026)PM2.5 Zn10 ng/m31.004 (1.002, 1.006)1.003 (1.001, 1.005)1.001 (0.999, 1.003)1.000 (0.998, 1.002)1.003 (1.001,1.005)1.003 (1.001,1.005)1.003 (1.001, 1.005)Total Noise10 dB Lden1.045 (1.040, 1.049)1.044 (1.039, 1.048)1.038 (1.034, 1.043)1.037 (1.033, 1.042)1.039 (1.035, 1.044)--Road traffic Noise10 dB Lden1.046 (1.042, 1.050)1.045 (1.041, 1.050)1.040 (1.035, 1.044)1.039 (1.034, 1.043)1.041 (1.037, 1.045)--CVD mortalityPM2.55 µg/m31.026 (1.008, 1.044)-1.011 (0.991, 1.032)1.005 (0.986, 1.024)1.006 (0.988, 1.026)1.012 (0.994, 1.031)1.013 (0.995, 1.031)NO210 µg/m31.026 (1.014, 1.039)1.022 (1.008, 1.036)-0.993 (0.971, 1.015)1.008 (0.993, 1.023)1.004 (0.991, 1.017)1.001 (0.988, 1.014)BC0.5 × 10–5/m1.036 (1.022, 1.051)1.034 (1.019, 1.049)1.043 (1.017, 1.070)-1.019 (1.002, 1.037)1.014 (0.999, 1.029)1.010 (0.995, 1.025)O3 warm10 µg/m30.958 (0.946, 0.970)0.960 (0.947, 0.974)0.963 (0.948, 0.979)0.969 (0.954, 0.985)-0.974 (0.961, 0.987)0.975 (0.962, 0.988)PM2.5 Cu5 ng/m31.048 (1.027, 1.069)1.043 (1.021, 1.065)1.042 (1.009, 1.076)1.015 (0.979, 1.052)1.023 (0.998,1.050)1.023 (1.002, 1.044)1.017 (0.996, 1.039)PM2.5 Fe100 ng/m31.045 (1.023, 1.067)1.038 (1.016, 1.061)1.023 (0.987, 1.059)0.981 (0.940, 1.023)1.016 (0.990,1.043)1.012 (0.990, 1.035)1.005 (0.983, 1.028)PM2.5 S200 ng/m31.018 (0.997, 1.040)1.001 (0.976, 1.027)0.994 (0.969, 1.020)0.988 (0.965, 1.011)0.983 (0.958,1.008)0.998 (0.977, 1.020)0.997 (0.975, 1.019)PM2.5 Zn10 ng/m31.002 (0.997, 1.006)1.000 (0.995, 1.006)1.000 (0.995, 1.005)0.999 (0.993, 1.004)1.001 (0.995,1.006)1.001 (0.996, 1.006)1.001 (0.996, 1.006)Total Noise10 dB Lden1.041 (1.035, 1.047)1.040 (1.034, 1.046)1.040 (1.033, 1.047)1.038 (1.031, 1.045)1.037 (1.031, 1.044)--Road traffic Noise10 dB Lden1.043 (1.037, 1.049)1.042 (1.036, 1.048)1.043 (1.036, 1.049)1.041 (1.034, 1.047)1.041 (1.034, 1.047)--Respiratory mortalityPM2.55 µg/m30.981 (0.953, 1.010)-0.933 0.901, 0.966)0.925 (0.894, 0.956)0.949 (0.917, 0.982)0.963 (0.934, 0.992)0.964 (0.935, 0.993)NO210 µg/m31.051 (1.031, 1.072)1.079 (1.053, 1.105)-0.995 (0.956, 1.037)1.036 (1.012, 1.061)1.024 (1.003, 1.046)1.020 (0.998, 1.042)BC0.5 × 10–5/m1.067 (1.043, 1.090)1.101 (1.073, 1.130)1.071 (1.024, 1.121)-1.055 (1.025, 1.085)1.039 (1.016, 1.064)1.034 (1.010, 1.059)O3 warm10 µg/m30.947 (0.924, 0.971)0.930 (0.905, 0.956)0.968 (0.941, 0.997)0.978 (0.948, 1.010)-0.969 (0.945, 0.994)0.970 (0.946, 0.995)PM2.5 Cu5 ng/m31.056 (1.021, 1.091)1.075 (1.037, 1.113)0.981 (0.928, 1.038)0.926 (0.873, 0.983)1.024 (0.985,1.065)1.022 (0.988, 1.057)1.015 (0.981, 1.05)PM2.5 Fe100 ng/m31.092 (1.053, 1.133)1.113 (1.070, 1.158)1.058 (0.990, 1.129)1.000 (0.927, 1.079)1.068 (1.023,1.114)1.052 (1.013, 1.092)1.043 (1.004, 1.084)PM2.5 S200 ng/m31.009 (0.974, 1.045)1.035 (0.992, 1.079)0.951 (0.909, 0.994)0.944 (0.904, 0.985)0.964 (0.926,1.004)0.982 (0.948, 1.018)0.980 (0.946, 1.015)PM2.5 Zn10 ng/m30.999 (0.989, 1.009)1.000 (0.990, 1.010)0.996 (0.986, 1.006)0.994 (0.984, 1.005)0.998 (0.988,1.007)0.998 (0.988, 1.008)0.998 (0.988, 1.008)Total Noise10 dB Lden1.056 (1.043, 1.069)1.059 (1.045, 1.072)1.050 (1.037, 1.064)1.048 (1.034, 1.061)1.051 (1.038, 1.065)--Road traffic Noise10 dB Lden1.058 (1.045, 1.071)1.061 (1.047, 1.074)1.053 (1.039, 1.067)1.050 (1.036, 1.064)1.054 (1.04, 1.067)--Model 3 (main model) adjusted for age, sex, individual level variables (education level, occupational status, marital status, origin, mother tongue), and area-level SES variables (composite score, unemployment rate, low education rate, high education rate). Additionally adjusted for noted exposure“Total noise” refers to total transportation noise, i.e. the energetic sum of road traffic, railway and aircraft noiseDue to the high correlation between the exposures, the following two-pollutant models are difficult to interpret: between BC and NO2; Cu and NO2; Cu and BC; Fe and NO2; Fe and BC In single exposure models, most exposures were also associated with CVD mortality (all except S and Zn). Similar to natural mortality, the associations were robust to adjustment for PM2.5; however, PM2.5 reduced somewhat after adjusting for most co-exposures. Adding noise to the models also reduced the associations to borderline significance for most pollutants, while noise was robust to adjustment for air pollution (Table 2). For respiratory mortality, again most exposures showed associations in single exposure models (all except PM2.5, S and Zn). For this outcome only, the inverse association for O3 was reduced to unity when BC was included in the model. A robust association between noise and respiratory mortality was also found that persisted on co-exposure adjustment (i.e. 1.056 [1.043, 1.069] per 10 dB Lden total transportation noise vs. 1.048 [1.034, 1.061] per 10 dB Lden total transportation noise with BC adjustment) (Table 2). Overall, the high correlations between NO2 and BC, as well as between NO2 / BC and Cu / Fe, made the two exposure models including these combinations difficult to interpret. ## Subset analysis on single exposure models In subset analysis (Table 3), associations reported in the main Models 3 persisted below 20 µg/m3 PM2.5 for both natural (1.038 [1.026, 1.049] per 5 µg/m3) and CVD (1.039 [1.020, 1.058] per 5 µg/m3) mortality, however with virtually the entire study population ($98.5\%$) residing in such areas. The point estimate was robust, indicative of associations down to the lowest investigated subset of below 12 µg/m3 PM2.5. For NO2, the results differed somewhat by outcome. For natural mortality, the association persisted down to the lowest subset of < 20 µg/m3, that included $31.8\%$ of the study population (1.018 [1.000, 1.036] per 10 µg/m3). The associations for CVD and respiratory mortality remained < 30 µg/m3 (1.038 [1.022, 1.053] per 10 µg/m3 and 1.023 [0.993, 1.053] per 10 µg/m3, respectively). BC associations persisted within the lowest subset of < 1.5 10–5/m for both natural and CVD mortality, and to < 2 10–5/m for respiratory with hazard ratios of: 1.039 (1.013, 1.066), 1.051 (1.011, 1.093) and 1.039 (1.003, 1.076) per 0.5 × 10–5/m BC, respectively. Given the limited spatial variability in Switzerland in the higher range, the subset analysis for O3 warm was not informative compared to the full model. Finally, for total transportation noise, the natural mortality association remained down to < 53 dB with 1.012 (1.002, 1.023) per 10 dB Lden including only $37.3\%$ of the study population, while CVD and respiratory associations clearly persisted < 60 dB Lden (1.028 [1.018,1.038] and 1.030 [1.007, 1.054] per 10 dB Lden, respectively).Table 3Subset analysis: hazard ratios ($95\%$ confidence intervals) for associations with mortality by cause, single exposure models (Model 3)ExposuresubsetN%Natural cause mortalityCVD mortalityRespiratory mortalityPM2.5Full4,188,175100.01.026 (1.015, 1.038)1.026 (1.008, 1.044)0.981 (0.953, 1.010) < 254,184,84299.91.027 (1.017, 1.038)1.028 (1.011, 1.046)0.980 (0.951, 1.009) < 204,127,07798.51.038 (1.026, 1.049)1.039 (1.020, 1.058)0.968 (0.933, 1.003) < 151,128,70126.91.012 (0.992, 1.033)1.016 (0.985, 1.049)0.886 (0.829, 0.946) < 12265,2536.31.024 (0.983, 1.067)1.051 (0.986, 1.120)0.788 (0.690, 0.899)NO2Full4,188,175100.01.050 (1.041, 1.059)1.026 (1.014, 1.039)1.051 (1.031, 1.072) < 404,087,41397.61.057 (1.049, 1.066)1.034 (1.021, 1.047)1.044 (1.020, 1.069) < 303,406,89181.31.059 (1.049, 1.068)1.038 (1.022, 1.053)1.023 (0.993, 1.053) < 201,331,20831.81.018 (1.000, 1.036)0.996 (0.968, 1.024)0.945 (0.889, 1.004)BCFull4,188,175100.01.057 (1.048, 1.067)1.036 (1.022, 1.051)1.067 (1.043, 1.090) < 34,177,68199.71.060 (1.051, 1.070)1.039 (1.024, 1.053)1.068 (1.043, 1.093) < 2.54,094,56597.81.066 (1.055, 1.076)1.045 (1.029, 1.061)1.064 (1.036, 1.092) < 23,494,99783.41.074 (1.062, 1.086)1.054 (1.034, 1.075)1.039 (1.003, 1.076) < 1.51,471,26835.11.039 (1.013, 1.066)1.051 (1.011, 1.093)0.945 (0.867, 1.030)O3 warmFull4,188,175100.00.946 (0.939, 0.954)0.958 (0.946, 0.970)0.947 (0.924, 0.971) < 1204,188,175100.00.946 (0.939, 0.954)0.958 (0.946, 0.970)0.947 (0.924, 0.971) < 1003,411,29781.50.933 (0.921, 0.945)0.963 (0.945, 0.982)0.896 (0.867, 0.926)Total noiseFull4,188,175100.01.045 (1.040, 1.049)1.041 (1.035, 1.047)1.056 (1.043, 1.069) < 602,892,57569.11.031 (1.025, 1.038)1.028 (1.018, 1.038)1.030 (1.007, 1.054) < 551,983,48247.41.018 (1.009, 1.027)1.008 (0.993, 1.022)0.984 (0.953, 1.015) < 531,562,11837.31.012 (1.002, 1.023)0.998 (0.981, 1.015)0.964 (0.929, 1.000)Exposure increments: PM2.5 per 5 µg/m3, NO2 per 10 µg/m3, BC per 0.5 × 10–5/m, O3 warm per 10 µg/m3, Total (transportation) noise per 10 dBModel 3 (main model) adjusted for age, sex, individual level variables (education level, occupational status, marital status, origin, mother tongue), and area-level SES variables (composite score, unemployment rate, low education rate, high education rate) ## Additional analyses on two exposure models Regarding potential effect modification by sex, the association for total transportation noise (adjusted for PM2.5) was stronger in males compared to females. Indications of stronger effects in males compared to females were also found for NO2 and BC, but not PM2.5 or O3 warm in models that were mutually adjusted for noise (Table S6). Separately, stronger associations were found in non-movers compared to the full cohort again for NO2 and BC (adjusted for noise) and for total transportation noise (adjusted for PM2.5) (Table S7). ## Main findings Single exposure models showed almost all air pollutants and noise exposures were positively associated with natural, cardiovascular and respiratory mortality outcomes. Many associations persisted at or below guideline limits for air pollution, as well as for natural mortality in relation to noise. Most associations were robust to adjustment for PM2.5 in two exposure models, however the opposite was not true. Specifically, associations for NO2, BC and O3 warm and both natural and respiratory mortality, as well as for BC and O3 warm for CVD mortality, largely remained after adjustment for co-exposures including transportation noise (total or road traffic noise only). Transportation noise was universally robust to adjustment for air pollution. As an outcome not considered in the WHO Environmental Noise Guidelines [16], the finding of an association between noise and respiratory mortality, independent of air pollution, was particularly novel but should be interpreted with caution given the lack of data on individual health behaviours. The influence of co-exposure adjustment on the PM2.5 component associations with natural and CVD mortality was quite consistent, though it is important to acknowledge the high correlations when interpreting these findings. The single exposure associations for Cu and Fe did not change after adjustment for PM2.5, while they attenuated after NO2 or BC adjustment. The attenuation was stronger for BC, which represents incomplete combustion. Further, the observation of BC being the more influential pollutant than PM2.5 supports the notion that the local air pollution mixture – and traffic related air pollution – may be more important than the broader regional mixture in Switzerland. Another interesting finding, and in line with Vodonos et al. [ 36], is that the associations of the air pollutants almost universally strengthened with increasing covariate adjustment. This applied not only to the pollutants that showed increased risk, but also to the inverse O3 warm association. The exception was Zn that remained stable. The expectation in environmental epidemiology is that the effect estimates typically attenuate with better covariate adjustment. Here, however, better adjustment achieved by including several measures of area-level SEP at two spatial scales led to stronger HR. In Switzerland, built-up areas with higher levels of traffic related pollution are also those with typically higher socio-economic position thus explaining the negative confounding. This finding also challenges the idea that administrative cohorts over-estimate associations due to insufficient adjustment, though indeed lifestyle factors are not available in the SNC to more directly contest this assumption (see Strengths & limitations below). ## Comparison to previous literature on air pollution The SNC was one of the administrative cohorts included in ELAPSE. Compared to the estimates from the meta-analysis including all cohorts, the Swiss effect estimates for natural mortality were slightly weaker for PM2.5 (1.026 [1.015, 1.038] Swiss vs. 1.053 [1.021, 1.085] combined per 5 µg/m3), slightly stronger for BC (1.057 [1.048, 1.067] vs. 1.039 [1.018, 1.059] per 0.5 × 10–5/m), and comparable for NO2 (1.050 [1.041, 1.059] vs. 1.044 [1.019, 1.069] per 10 µg/m3), and O3 warm (0.946 [0.939, 0.954] vs. 0.953 [0.929, 0.979] per 10 µg/m3) [22]. The associations reported in the pooled analysis of the eight detailed cohorts included in the other part of ELAPSE were stronger than in Switzerland [21]. Contrary to findings from North America [37–39], the inverse association for O3 warm was unexpected. Given that ozone is typically reduced by primary tail pipe emissions, one would expect the inverse association to disappear after adjusting for traffic pollutants. This was the behaviour observed in the meta-analysis of the seven administrative cohorts in ELAPSE; the single exposure O3 associations were inverse, though after co-pollutant adjustment the associations largely disappeared [22]. While HRs for Switzerland were marginally closer to unity after adjustment for co-exposures, the inverse relationship with ozone remained in two exposure models. We have no clear explanation for the robustness of the negative association in the Swiss cohort. This could relate to the moderately high negative correlations with the other exposures, perhaps made higher than previous studies given the 100 × 100 m spatial resolution. However, additional analyses in ELAPSE using O3 modelled at a coarser resolution likewise produced inverse associations for Switzerland [40]. The subset analysis for Switzerland revealed associations at lower levels of exposures for all air pollution-outcome pairs except O3 warm that had little exposure contrast. Generally studies from other regions have shown that the air pollution association with mortality persists in the low exposure range. These low-level cohort studies have tended to focus on PM2.5 exposure. The Canadian studies, specifically three waves of the Canadian Census Health and Environment Cohort (CanCHEC) and the Canadian Community Health Survey (CCHS), observed supralinear associations between PM2.5 and mortality starting at very low levels [4, 9]. In the US Medicare cohort, associations were also found with PM2.5, NO2, and summer O3 below the air quality standard levels [41]. Similar to these findings for Switzerland, the effect of PM2.5 on mortality in the Canadian studies attenuated when NO2 was considered [4, 9]. Findings in the Sydney, Australia ’45 and Up Study also suggested both PM2.5 and NO2 were related to premature mortality [7]. None of these studies, except for ELAPSE, considered co-adjustment with transportation noise. ## Role of transportation noise It is not yet common practice to adjust for noise in the wider literature on health effects of traffic related air pollution. The several studies on mortality, however, are available to shed more light on the issue of potential confounding in different study areas and populations in Europe. The Danish Diet, Cancer and Health Cohort reported indicative associations of CVD and all-cause mortality with several air pollutants, after substantial attenuation with noise adjustment [42]. Others, however, found little effect of noise adjustment [43–46]. Within ELAPSE, adjustment for traffic noise hardly changed the association between air pollutants and CVD mortality for most of the included administrative cohorts; associations slightly strengthened for the Dutch cohort and slightly attenuated for the others [22]. The impact of noise adjustment across these cohorts in ELAPSE likely relates to the urban structure and subsequent degree of correlations. Also, unlike air pollution where exposures were from harmonized Europe-wide models, differences in the quality, specification or resolution of the country-specific noise models may play a role. High quality country-specific noise models are expected to better capture small-scale variation in exposures [47]. In Switzerland, the correlations between air pollution and noise exposures were not high. Further the exposure models were all comparably high resolution which should minimize the possibility of effect transfer, i.e. when the effect from the less well measured exposure is transferred to (or mopped up by) the better measured one [42, 48, 49]. Previously using the same data for Switzerland, Vienneau et al. [ 47] demonstrated the sensitivity of noise exposure assessment illustrating that use of fine scale noise maps rather than estimates directly at the household façade underestimates the associations. Several cohorts in ELAPSE, in addition to Switzerland, used state-of-the-art emissions and propagation models (described in [22]) which is the gold standard for noise exposure modelling. Detailed in Karipidis et al. [ 31], the Swiss noise exposure data were available at the façade of the residential location including floor of residence. The models have been shown to perform well in a validation study using measurements at the window (agreement between measured and modelled: mean + 0.5 dB(A), standard deviation 4.0 dB(A)) [50]. Previous Swiss studies on exposure to transportation noise sources and CVD mortality found that associations were robust to air pollution adjustment [12]. This current study, with more detailed air pollution modelling and a broader range of air pollutants strengthens the evidence that transportation noise is an independent risk factor for natural, CVD and respiratory mortality. It should be noted, however, that the subset analysis suggests the air pollution associations are largely linear while noise may have a threshold (i.e., for CVD and respiratory mortality). Thus, the impact of noise adjustment on the air pollution associations for some outcomes may differ between low and high exposure groups. Another concern with noise models, similar to models for air pollution, is that they reflect ambient conditions rather than indoor exposures which may contribute to exposure misclassification. Foraster et al. [ 51] showed that modelling indoor noise at the bedroom substantially reduced correlations with air pollution and led to more consistent associations with blood pressure in a cross-sectional study in Gerona, Spain. Similar correction factors have been investigated by Locher et al. [ 52] in Switzerland, though lack of national data on noise attenuation factors limits their application in the SNC. In this study, however, any exposure misclassification is expected to be non-differential. To date, the strongest evidence for health effects of transportation noise relates to CVD outcomes [13, 15, 53–55]. In addition to CVD mortality, the findings reported here also show that transportation noise was associated with natural and respiratory mortality. The few similar studies also concur with an association for natural mortality, though are less consistent for respiratory mortality. In the Danish Nurse Cohort (DNC), Liu et al. [ 17] reported a HR of 1.15 (1.06, 1.25) per 10 dB Lden road traffic noise in association with chronic obstructive pulmonary disease. In the same cohort, road traffic noise was associated with all-cause mortality (1.09 [1.03, 1.15] per 10 dB 23-year mean Lden), and suggestive for respiratory mortality (1.16 [0.97, 1.40]) [11]. In both DNC studies, the associations similarly persisted after adjusting for air pollution. The Danish Diet, Cancer and Health Cohort study also found associations with road traffic noise and both CVD (1.13 [1.06, 1.19] 10-year mean Lden at the most exposed façade) and all-cause mortality (1.08 [1.05, 1.11], but not with respiratory mortality (1.02 [0.96, 1.09]) [14]. In the Swiss SAPALDIA study, transportation noise was found to be associated with exacerbation of asthma symptoms in asthmatics, but not incidence of asthma [33]. Noise has further been suggested as a risk factor for diabetes and neurodegenerative diseases [56–58]. Hence, there is an increased rationale for assessing mortality from natural causes in relation to noise, in addition to the more traditional cardiovascular causes. Interestingly, associations with noise below 53 dB (the WHO guideline value for road traffic noise) remained significant for natural mortality but not CVD mortality. Note this was based on total transportation noise, though given the near identical associations for road traffic noise we expect the same finding. That the association for natural mortality persists indicates a cause other than CVD is strongly related to noise at low levels. ## Biological mechanisms The underlying mechanisms by which long term air pollution exposure is related to CVD is thought to be oxidative stress and systematic inflammation. The same mechanisms likely play a role in respiratory disease [59, 60]. For noise, research into mechanisms has focused on elucidating the cardio-metabolic pathways, stemming from direct (auditory) sleep disturbance and indirect (non-auditory) disturbances such as annoyance. Noise can trigger a physiological stress response through activation of the Hypothalamic–Pituitary–Adrenal axis. This includes the releases of stress hormones, leading to increased blood pressure, vascular dysfunction, inflammation and oxidative stress [13, 61–63]. Being in a stressed state may also disrupt night-time recovery of the immune system, contributing to inflammation and oxidative stress also in the respiratory tract [63]. There is growing evidence that air pollution accelerates aging, and that multiple organs can be affected [60, 64]. In addition to reduced longevity, aging-related outcomes that have been associated with air pollution range from reduced lung function, increased chronic disease, frailty, to cognitive decline [65–69]. Sleep, which can be disrupted by noise, also changes with aging, with poor sleep linked to many adverse health outcomes [70]. Regarding the cause-specific mortalities investigated here, whether air pollution or noise first affects an individual’s heart or lungs may relate to comorbidities. Self-selection may also be important as suggested in the subset analysis where the effects generally attenuated in those living in less polluted areas. This was particularly noticeable for respiratory mortality, where an inverse association for PM2.5 was found in the lowest subsets. It is plausible that those with severe respiratory disease intentionally move to less polluted areas. ## Strengths & limitations The strengths of this analysis included the large study population followed for up to 14 years. The cohort had near complete coverage given that the census in *Switzerland is* compulsory, minimizing selection bias. Both individual and area-level covariates at baseline could be derived from the same source; the 04 December 2000 census. High spatial resolution air pollution and noise exposures were linked to this population at their place of residence. Given the importance of sleep as one mechanism for noise effects, it may be reasonable to specifically investigate night-time noise exposure in addition to Lden that covers the full 24-h day with an added penalty for evening and night hours. The correlations between Lden and Leq,night in the study population, however, are near identical ($r = 0.99$ total transportation noise, and 1.00 for road traffic noise). Further, *Lden is* the main metric used in studies on chronic health effects, including in the WHO noise guidelines (Leq,night is only used in relation to sleep disturbance) [16]. Individual data on behavioural risk factors were not available in the SNC, thus direct adjustment for such factors could not be performed. However, including individual behaviours has been shown to only modestly attenuate the mortality associations for PM2.5 in some studies that could investigate this in detail [4, 71]. Within ELAPSE, cohort-specific sensitivity analyses – including for the SNC – showed natural mortality associations were robust to indirect adjustment for smoking and body-mass index (Table S13 in Stafoggia et al. [ 22]). Whether this also holds for other mortality outcomes or exposures requires further study. Another possible limitation is that the air pollution exposures related to year 2010 [25, 26] which may introduce some exposure misclassification. For natural mortality, additional sensitivity analysis within ELAPSE included back extrapolating these 2010 exposures to the baseline year 2000 (Table S14 in Stafoggia et al. [ 22]). Accounting for residential history (with $34\%$ participants known to have moved during follow up, the results for Switzerland based on back extrapolated exposures were found to be robust. Back extrapolation, however, was applied in each ELAPSE study area at NUTS-2 regions (Nomenclature of territorial units for statistics, harmonised hierarchical system for Europe, which for Switzerland equals national scale making this a temporal correction only. It is also possible that the true association is diluted due to exposure misclassification from assigning exposure only at baseline. While address history is available in the SNC, as indicated above, the intervals are inconsistent. It is collected via the census at 2000 then annually from 2010 onward. For the gap between 2000 and 2010, we rely on the change in geocode/building ID plus a census question related to community of residence at 2006 to identify movers and broadly assign the timing of a move (i.e. before vs after 2006). Any information about moving for those who die between 2000 and 2010, however, is lost. The sensitivity analysis restricting to non-movers indeed showed stronger associations for NO2, BC and total transportation noise compared to the full cohort. This suggests exposure misclassificaton may be particularly relevant for the more local/traffic-related exposures. Finally, investigating mortality endpoints does not enable differentiating effects on disease etiology from the effects on disease progression. ## Policy implications While the current air quality and environmental noise guidelines focus on regulating single exposures that are pivotal for health protection, more attention is needed regarding the complex multipollutant mixture [72]. This includes gaining insight into the area-specific composition of PM as well as attempting to disentangle associations with noise. As demonstrated in this study, however, the PM components were highly correlated with other exposures lending to results that could not be easily interpreted. The 2021 WHO AQG have set limit values for PM2.5, NO2 and O3 in the peak warm season which, for the annual averaging period, are 5 µg/m3, 10 µg/m3 and 60 µg/m3 [3]. These limits are notably lower than the majority of exposures in Switzerland according to this study. The AQG also include good practice statements, aimed to draw attention to other pollutants conveying risk but with less supporting evidence to develop a limit value such as BC. Though at present there is also no Swiss limit value for BC, this pollutant is of particular concern. As emphasized in a 2013 PM report by the Federal Commission for Air Hygiene (FCAH), BC is considered carcinogenic. Thus, the concentrations should be as low as possible to limit the consequences to not more than 1 case per 1 million [73], that approximately corresponds to an annual mean concentration of 0.1 µg/m3 soot (elemental carbon). FACH called for a reduction strategy to reach $20\%$ of the 2013 levels within 10 years, given still far higher levels of 2–3 µg/m3 observed in many urban locations. ELAPSE and this analysis in the SNC underscore the call for continued and rigorous reductions of BC to protect public health. It also strongly suggests that additional outcomes beyond CVD should be considered when evaluating the burden of disease from transportation noise. ## Conclusion In this study using a large administrative cohort, long-term exposure to air pollution and transportation noise were associated with mortality in Switzerland. Traffic-related air pollutants were more strongly associated with natural, CVD and respiratory mortality than PM2.5 mass and components. Noise was not only associated with CVD, but also with respiratory and natural cause mortality. Associations with noise remained for natural cause mortality below the WHO guideline value for road traffic noise of 53 dB Lden. Considering co-exposures revealed the importance of local traffic-related pollutants such as NO2 and BC as well as transportation noise in relation to mortality. ## Supplementary Information Additional file 1: Table S1. Study population selection. Table S2. Exposure distributions. Table S3. Spearman correlations between exposures (top), and between exposures and neighbourhood socio-economic position (SEP) score (bottom). Table S4. Mean exposure by quintiles of neighbourhood socio-economic position. Table S5. Hazard ratios ($95\%$ confidence intervals) for single exposure associations between air pollution and noise exposures and mortality by cause, with increasing level of covariate adjustment in single exposure models. Table S6. Effect modification by sex, hazard ratios ($95\%$ confidence intervals) in two exposure models for associations with natural cause mortality (Model 3). Table S7. Non-mover analysis, hazard ratios ($95\%$ confidence intervals) in two exposure models for associations with natural cause mortality (Model 3). ## References 1. 1.HEIState of Global Air 2020. Special report2020BostonHealth Effects Institute. *State of Global Air 2020. Special report* (2020.0) 2. 2.EEA. Air quality in Europe 2021. 2021. 3. 3.WHOWHO global air quality guidelines: particulate matter (PM2.5 and PM10), ozone, nitrogen dioxide, sulfur dioxide and carbon monoxide2021GenevaWorld Health Organization. *WHO global air quality guidelines: particulate matter (PM2.5 and PM10), ozone, nitrogen dioxide, sulfur dioxide and carbon monoxide* (2021.0) 4. Christidis T, Erickson AC, Pappin AJ, Crouse DL, Pinault LL, Weichenthal SA, Brook JR, van Donkelaar A, Hystad P, Martin RV, Tjepkema M, Burnett RT, Brauer M. **Low concentrations of fine particle air pollution and mortality in the Canadian Community Health Survey cohort**. *Environ Health* (2019.0) **18** 84. DOI: 10.1186/s12940-019-0518-y 5. Fischer PH, Marra M, Ameling CB, Velders GJM, Hoogerbrugge R, de Vries W, Wesseling J, Janssen NAH, Houthuijs D. **Particulate air pollution from different sources and mortality in 75 million adults - the Dutch Environmental Longitudinal Study (DUELS)**. *Sci Total Environ* (2020.0) **705** 135778. DOI: 10.1016/j.scitotenv.2019.135778 6. Hales S, Atkinson J, Metcalfe J, Kuschel G, Woodward A. **Long term exposure to air pollution, mortality and morbidity in New Zealand: cohort study**. *Sci Total Environ* (2021.0) **801** 149660. DOI: 10.1016/j.scitotenv.2021.149660 7. Hanigan IC, Rolfe MI, Knibbs LD, Salimi F, Cowie CT, Heyworth J, Marks GB, Guo Y, Cope M, Bauman A, Jalaludin B, Morgan GG. **All-cause mortality and long-term exposure to low level air pollution in the ‘45 and up study’ cohort, Sydney, Australia, 2006–2015**. *Environ Int* (2019.0) **126** 762-770. DOI: 10.1016/j.envint.2019.02.044 8. Hvidtfeldt UA, Geels C, Sørensen M, Ketzel M, Khan J, Tjønneland A, Christensen JH, Brandt J, Raaschou-Nielsen O. **Long-term residential exposure to PM2.5 constituents and mortality in a Danish cohort**. *Environ Int* (2019.0) **133** 105268. DOI: 10.1016/j.envint.2019.105268 9. Pappin AJ, Christidis T, Pinault LL, Crouse DL, Brook JR, Erickson A, Hystad P, Li C, Martin RV, Meng J, Weichenthal S, van Donkelaar A, Tjepkema M, Brauer M, Burnett RT. **Examining the shape of the association between low levels of fine particulate matter and mortality across three cycles of the Canadian Census Health and Environment Cohort**. *Environ Health Perspect* (2019.0) **127** 107008. DOI: 10.1289/EHP5204 10. Beelen R, Hoek G, Houthuijs D, van den Brandt PA, Goldbohm RA, Fischer P, Schouten LJ, Armstrong B, Brunekreef B. **The joint association of air pollution and noise from road traffic with cardiovascular mortality in a cohort study**. *Occup Environ Med* (2009.0) **66** 243-250. DOI: 10.1136/oem.2008.042358 11. Cole-Hunter T, So R, Amini H, Backalarz C, Brandt J, Bräuner EV, Hertel O, Jensen SS, Jørgensen JT, Ketzel M, Laursen JE, Lim YH, Loft S, Mehta A, Mortensen LH, Simonsen MK, Sisgaard T, Westendorp R, Andersen ZJ. **Long-term exposure to road traffic noise and all-cause and cause-specific mortality: a Danish Nurse Cohort study**. *Sci Total Environ* (2022.0) **820** 153057. DOI: 10.1016/j.scitotenv.2022.153057 12. Héritier H, Vienneau D, Foraster M, Eze IC, Schaffner E, de Hoogh K, Thiesse L, Rudzik F, Habermacher M, Köpfli M. **A systematic analysis of mutual effects of transportation noise and air pollution exposure on myocardial infarction mortality: a nationwide cohort study in Switzerland**. *Eur Heart J* (2019.0) **40** 598-603. DOI: 10.1093/eurheartj/ehy650 13. Münzel T, Sørensen M, Daiber A. **Transportation noise pollution and cardiovascular disease**. *Nat Rev Cardiol* (2021.0) **18** 619-36. DOI: 10.1038/s41569-021-00532-5 14. Thacher JD, Hvidtfeldt UA, Poulsen AH, Raaschou-Nielsen O, Ketzel M, Brandt J, Jensen SS, Overvad K, Tjønneland A, Münzel T, Sørensen M. **Long-term residential road traffic noise and mortality in a Danish cohort**. *Environ Res* (2020.0) **187** 109633. DOI: 10.1016/j.envres.2020.109633 15. Vienneau D, Saucy A, Schäffer B, Flückiger B, Tangermann L, Stafoggia M, Wunderli JM, Röösli M. **Transportation noise exposure and cardiovascular mortality: 15-years of follow-up in a nationwide prospective cohort in Switzerland**. *Environ Int* (2022.0) **158** 106974. DOI: 10.1016/j.envint.2021.106974 16. 16.WHOEnvironmental noise guidelines for the European region2018CopenhagenWHO Regional Office for Europe. *Environmental noise guidelines for the European region* (2018.0) 17. Liu S, Lim YH, Pedersen M, Jørgensen JT, Amini H, Cole-Hunter T, Mehta AJ, So R, Mortensen LH, Westendorp RGJ, Loft S, Bräuner EV, Ketzel M, Hertel O, Brandt J, Jensen SS, Christensen JH, Sigsgaard T, Geels C, Frohn LM, Brboric M, Radonic J, Turk Sekulic M, Bønnelykke K, Backalarz C, Simonsen MK, Andersen ZJ. **Long-term air pollution and road traffic noise exposure and COPD: the Danish Nurse Cohort**. *Eur Respir J* (2021.0) **58** 2004594. DOI: 10.1183/13993003.04594-2020 18. Roswall N, Raaschou-Nielsen O, Ketzel M, Gammelmark A, Overvad K, Olsen A, Sorensen M. **Long-term residential road traffic noise and NO2 exposure in relation to risk of incident myocardial infarction - a Danish cohort study**. *Environ Res* (2017.0) **156** 80-86. DOI: 10.1016/j.envres.2017.03.019 19. Tétreault L-F, Perron S, Smargiassi A. **Cardiovascular health, traffic-related air pollution and noise: are associations mutually confounded? A systematic review**. *Int J Public Health* (2013.0) **58** 649-666. DOI: 10.1007/s00038-013-0489-7 20. Vienneau D, Perez L, Schindler C, Lieb C, Sommer H, Probst-Hensch N, Künzli N, Röösli M. **Years of life lost and morbidity cases attributable to transportation noise and air pollution: a comparative health risk assessment for Switzerland in 2010**. *Int J Hyg Environ Health* (2015.0) **218** 514-521. DOI: 10.1016/j.ijheh.2015.05.003 21. Strak M, Weinmayr G, Rodopoulou S, Chen J, de Hoogh K, Andersen ZJ, Atkinson R, Bauwelinck M, Bekkevold T, Bellander T, Boutron-Ruault M-C, Brandt J, Cesaroni G, Concin H, Fecht D, Forastiere F, Gulliver J, Hertel O, Hoffmann B, Hvidtfeldt UA, Janssen NAH, Jöckel K-H, Jørgensen JT, Ketzel M, Klompmaker JO, Lager A, Leander K, Liu S, Ljungman P, Magnusson PKE, Mehta AJ, Nagel G, Oftedal B, Pershagen G, Peters A, Raaschou-Nielsen O, Renzi M, Rizzuto D, van der Schouw YT, Schramm S, Severi G, Sigsgaard T, Sørensen M, Stafoggia M, Tjønneland A, Verschuren WMM, Vienneau D, Wolf K, Katsouyanni K, Brunekreef B, Hoek G, Samoli E. **Long term exposure to low level air pollution and mortality in eight European cohorts within the ELAPSE project: pooled analysis**. *BMJ* (2021.0) **374** n1904. DOI: 10.1136/bmj.n1904 22. Stafoggia M, Oftedal B, Chen J, Rodopoulou S, Renzi M, Atkinson RW, Bauwelinck M, Klompmaker JO, Mehta A, Vienneau D, Andersen ZJ, Bellander T, Brandt J, Cesaroni G, de Hoogh K, Fecht D, Gulliver J, Hertel O, Hoffmann B, Hvidtfeldt UA, Jöckel K-H, Jørgensen JT, Katsouyanni K, Ketzel M, Kristoffersen DT, Lager A, Leander K, Liu S, Ljungman PLS, Nagel G, Pershagen G, Peters A, Raaschou-Nielsen O, Rizzuto D, Schramm S, Schwarze PE, Severi G, Sigsgaard T, Strak M, van der Schouw YT, Verschuren M, Weinmayr G, Wolf K, Zitt E, Samoli E, Forastiere F, Brunekreef B, Hoek G, Janssen NAH. **Long-term exposure to low ambient air pollution concentrations and mortality among 28 million people: results from seven large European cohorts within the ELAPSE project**. *Lancet Planet Health* (2022.0) **6** e9-e18. DOI: 10.1016/S2542-5196(21)00277-1 23. Rodopoulou S, Stafoggia M, Chen J, de Hoogh K, Bauwelinck M, Mehta AJ, Klompmaker JO, Oftedal B, Vienneau D, Janssen NAH, Strak M, Andersen ZJ, Renzi M, Cesaroni G, Nordheim CF, Bekkevold T, Atkinson R, Forastiere F, Katsouyanni K, Brunekreef B, Samoli E, Hoek G. **Long-term exposure to fine particle elemental components and mortality in Europe: results from six European administrative cohorts within the ELAPSE project**. *Sci Total Environ* (2022.0) **809** 152205. DOI: 10.1016/j.scitotenv.2021.152205 24. Chen J, Rodopoulou S, de Hoogh K, Strak M, Andersen ZJ, Atkinson R, Bauwelinck M, Bellander T, Brandt J, Cesaroni G, Concin H, Fecht D, Forastiere F, Gulliver J, Hertel O, Hoffmann B, Hvidtfeldt UA, Janssen NAH, Jöckel KH, Jørgensen J, Katsouyanni K, Ketzel M, Klompmaker JO, Lager A, Leander K, Liu S, Ljungman P, MacDonald CJ, Magnusson PKE, Mehta A, Nagel G, Oftedal B, Pershagen G, Peters A, Raaschou-Nielsen O, Renzi M, Rizzuto D, Samoli E, van der Schouw YT, Schramm S, Schwarze P, Sigsgaard T, Sørensen M, Stafoggia M, Tjønneland A, Vienneau D, Weinmayr G, Wolf K, Brunekreef B, Hoek G. **Long-term exposure to fine particle elemental components and natural and cause-specific mortality-a pooled analysis of eight European cohorts within the ELAPSE project**. *Environ Health Perspect* (2021.0) **129** 47009. DOI: 10.1289/EHP8368 25. Chen J, de Hoogh K, Gulliver J, Hoffmann B, Hertel O, Ketzel M, Weinmayr G, Bauwelinck M, van Donkelaar A, Hvidtfeldt UA, Atkinson R, Janssen NAH, Martin RV, Samoli E, Andersen ZJ, Oftedal BM, Stafoggia M, Bellander T, Strak M, Wolf K, Vienneau D, Brunekreef B, Hoek G. **Development of Europe-wide models for particle elemental composition using supervised linear regression and random forest**. *Environ Sci Technol* (2020.0) **54** 15698-15709. DOI: 10.1021/acs.est.0c06595 26. de Hoogh K, Chen J, Gulliver J, Hoffmann B, Hertel O, Ketzel M, Bauwelinck M, van Donkelaar A, Hvidtfeldt UA, Katsouyanni K, Klompmaker J, Martin RV, Samoli E, Schwartz PE, Stafoggia M, Bellander T, Strak M, Wolf K, Vienneau D, Brunekreef B, Hoek G. **Spatial PM2.5, NO2, O3 and BC models for Western Europe – evaluation of spatiotemporal stability**. *Environ Int* (2018.0) **120** 81-92. DOI: 10.1016/j.envint.2018.07.036 27. 27.FOEN. Swiss Environmental Law. A brief guide. 2022. p. 44. 28. Bopp M, Spoerri A, Zwahlen M, Gutzwiller F, Paccaud F, Braun-Fahrländer C, Rougemont A, Egger M. **Cohort profile: the Swiss National Cohort—a longitudinal study of 6.8 million people**. *Int J Epidemiol* (2009.0) **38** 379-384. DOI: 10.1093/ije/dyn042 29. Spoerri A, Zwahlen M, Egger M, Bopp M. **The Swiss National Cohort: a unique database for national and international researchers**. *Int J Public Health* (2010.0) **55** 239-242. DOI: 10.1007/s00038-010-0160-5 30. Renaud A. *Coverage estimation for the Swiss population census 2000: estimation methodology and results, Swiss statistics methodology report* (2004.0) 147 31. Karipidis I, Vienneau D, Habermacher M, Köpfli M, Brink M, Probst-Hensch N, Röösli M, Jean-Marc W. **Reconstruction of historical noise exposure data for environmental epidemiology in Switzerland within the SiRENE project**. *Noise Mapp* (2014.0) **1** 3-14 32. Röösli M, Wunderli J-M, Brink M, Cajochen C, Probst-Hensch N. **Bruit du trafic, mortalité cardiovasculaire, diabète, troubles du sommeil et nuisance - L’étude SiRENE**. *Forum Med Suisse* (2019.0) **19** 77-82 33. Eze IC, Foraster M, Schaffner E, Vienneau D, Héritier H, Pieren R, Thiesse L, Rudzik F, Rothe T, Pons M, Bettschart R, Schindler C, Cajochen C, Wunderli J-M, Brink M, Röösli M, Probst-Hensch N. **Transportation noise exposure, noise annoyance and respiratory health in adults: a repeated-measures study**. *Environ Int* (2018.0) **121** 741-750. DOI: 10.1016/j.envint.2018.10.006 34. Panczak R, Galobardes B, Voorpostel M, Spoerri A, Zwahlen M, Egger M. **A Swiss neighbourhood index of socioeconomic position: development and association with mortality**. *J Epidemiol Comm Health* (2012.0) **66** 1129-1136. DOI: 10.1136/jech-2011-200699 35. Samoli E, Rodopoulou S, Hvidtfeldt UA, Wolf K, Stafoggia M, Brunekreef B, Strak M, Chen J, Andersen ZJ, Atkinson R, Bauwelinck M, Bellander T, Brandt J, Cesaroni G, Forastiere F, Fecht D, Gulliver J, Hertel O, Hoffmann B, de Hoogh K, Janssen NAH, Ketzel M, Klompmaker JO, Liu S, Ljungman P, Nagel G, Oftedal B, Pershagen G, Peters A, Raaschou-Nielsen O, Renzi M, Kristoffersen DT, Severi G, Sigsgaard T, Vienneau D, Weinmayr G, Hoek G, Katsouyanni K. **Modeling multi-level survival data in multi-center epidemiological cohort studies: applications from the ELAPSE project**. *Environ Int* (2021.0) **147** 106371. DOI: 10.1016/j.envint.2020.106371 36. Vodonos A, Awad YA, Schwartz J. **The concentration-response between long-term PM2.5 exposure and mortality; a meta-regression approach**. *Environ Res* (2018.0) **166** 677-689. DOI: 10.1016/j.envres.2018.06.021 37. Crouse DL, Peters PA, Hystad P, Brook JR, van Donkelaar A, Martin RV, Villeneuve PJ, Jerrett M, Goldberg MS, Pope CA, Brauer M, Brook RD, Robichaud A, Menard R, Burnett RT. **Ambient PM2.5, O(3), and NO(2) exposures and associations with mortality over 16 years of follow-up in the Canadian Census Health and Environment Cohort (CanCHEC)**. *Environ Health Perspect* (2015.0) **123** 1180-1186. DOI: 10.1289/ehp.1409276 38. Di Q, Wang Y, Zanobetti A, Wang Y, Koutrakis P, Choirat C, Dominici F, Schwartz JD. **Air pollution and mortality in the medicare population**. *N Engl J Med* (2017.0) **376** 2513-2522. DOI: 10.1056/NEJMoa1702747 39. Turner MC, Jerrett M, Pope CA, Krewski D, Gapstur SM, Diver WR, Beckerman BS, Marshall JD, Su J, Crouse DL, Burnett RT. **Long-term ozone exposure and mortality in a large prospective study**. *Am J Respir Crit Care Med* (2016.0) **193** 1134-1142. DOI: 10.1164/rccm.201508-1633OC 40. 40.Brunekreef B, Strak S, Chen J, Andersen Z, Atkinson R, Bauwelinck M, et al. Mortality and morbidity effects of long-term exposure to low-level PM25, BC, NO2, and O3: an analysis of European cohorts in the ELAPSE project. In: Institute HE, editor. Boston: Health Effects Institute; 2021. 41. Yazdi MD, Wang Y, Di Q, Requia WJ, Wei Y, Shi L, Sabath MB, Dominici F, Coull B, Evans JS, Koutrakis P, Schwartz JD. **Long-term effect of exposure to lower concentrations of air pollution on mortality among US Medicare participants and vulnerable subgroups: a doubly-robust approach**. *Lancet Planet Health* (2021.0) **5** e689-e697. DOI: 10.1016/S2542-5196(21)00204-7 42. Hvidtfeldt UA, Sørensen M, Geels C, Ketzel M, Khan J, Tjønneland A, Overvad K, Brandt J, Raaschou-Nielsen O. **Long-term residential exposure to PM2.5, PM10, black carbon, NO2, and ozone and mortality in a Danish cohort**. *Environ Int* (2019.0) **123** 265-272. DOI: 10.1016/j.envint.2018.12.010 43. Nieuwenhuijsen MJ, Gascon M, Martinez D, Ponjoan A, Blanch J, Garcia-Gil MDM, Ramos R, Foraster M, Mueller N, Espinosa A, Cirach M, Khreis H, Dadvand P, Basagaña X. **Air pollution, noise, blue space, and green space and premature mortality in Barcelona: a mega cohort**. *Int J Environ Res Public Health* (2018.0) **15** 2405. DOI: 10.3390/ijerph15112405 44. Raaschou-Nielsen O, Andersen ZJ, Jensen SS, Ketzel M, Sørensen M, Hansen J, Loft S, Tjønneland A, Overvad K. **Traffic air pollution and mortality from cardiovascular disease and all causes: a Danish cohort study**. *Environ Health* (2012.0) **11** 60. DOI: 10.1186/1476-069X-11-60 45. So R, Jørgensen JT, Lim Y-H, Mehta AJ, Amini H, Mortensen LH, Westendorp R, Ketzel M, Hertel O, Brandt J, Christensen JH, Geels C, Frohn LM, Sisgaard T, Bräuner EV, Jensen SS, Backalarz C, Simonsen MK, Loft S, Cole-Hunter T, Andersen ZJ. **Long-term exposure to low levels of air pollution and mortality adjusting for road traffic noise: a Danish Nurse Cohort study**. *Environ Int* (2020.0) **143** 105983. DOI: 10.1016/j.envint.2020.105983 46. Tonne C, Halonen JI, Beevers SD, Dajnak D, Gulliver J, Kelly FJ, Wilkinson P, Anderson HR. **Long-term traffic air and noise pollution in relation to mortality and hospital readmission among myocardial infarction survivors**. *Int J Hyg Environ Health* (2016.0) **219** 72-78. DOI: 10.1016/j.ijheh.2015.09.003 47. Vienneau D, Héritier H, Foraster M, Eze IC, Schaffner E, Thiesse L, Rudzik F, Habermacher M, Köpfli M, Pieren R, Brink M, Cajochen C, Wunderli JM, Probst-Hensch N, Röösli M. **Façades, floors and maps – Influence of exposure measurement error on the association between transportation noise and myocardial infarction**. *Environ Int* (2019.0) **123** 399-406. DOI: 10.1016/j.envint.2018.12.015 48. Huangfu P, Atkinson R. **Long-term exposure to NO(2) and O(3) and all-cause and respiratory mortality: a systematic review and meta-analysis**. *Environ Int* (2020.0) **144** 105998. DOI: 10.1016/j.envint.2020.105998 49. Münzel T, Sorensen M, Gori T, Schmidt FP, Rao X, Brook J, Chen LC, Brook RD, Rajagopalan S. **Environmental stressors and cardio-metabolic disease: part I-epidemiologic evidence supporting a role for noise and air pollution and effects of mitigation strategies**. *Eur Heart J* (2017.0) **38** 550-556. PMID: 27460892 50. Schlatter F, Piquerez A, Habermacher M, Ragettli MS, Röösli M, Brink M, Cajochen C, Probst-Hensch N, Foraster M, Wunderli JM. **Validation of large scale noise exposure modelling by long-term measurements**. *Noise Mapp* (2017.0) **4** 75. DOI: 10.1515/noise-2017-0006 51. Foraster M, Kunzli N, Aguilera I, Rivera M, Agis D, Vila J, Bouso L, Deltell A, Marrugat J, Ramos R, Sunyer J, Elosua R, Basagana X. **High blood pressure and long-term exposure to indoor noise and air pollution from road traffic**. *Environ Health Perspect* (2014.0) **122** 1193-1200. DOI: 10.1289/ehp.1307156 52. Locher B, Piquerez A, Habermacher M, Ragettli M, Röösli M, Brink M, Cajochen C, Vienneau D, Foraster M, Müller U, Wunderli J. **Differences between outdoor and indoor sound levels for open, tilted, and closed windows**. *Int J Environ Res Public Health* (2018.0) **15** e149. DOI: 10.3390/ijerph15010149 53. Dzhambov AM, Dimitrova DD. **Exposure-response relationship between traffic noise and the risk of stroke: a systematic review with meta-analysis**. *Arh Hig Rada Toksikol* (2016.0) **67** 136-151. PMID: 27331301 54. Thacher JD, Poulsen AH, Raaschou-Nielsen O, Hvidtfeldt UA, Brandt J, Christensen JH, Khan J, Levin G, Münzel T, Sørensen M. **Exposure to transportation noise and risk for cardiovascular disease in a nationwide cohort study from Denmark**. *Environ Res* (2022.0) **211** 113106. DOI: 10.1016/j.envres.2022.113106 55. van Kempen E, Casas M, Pershagen G, Foraster M. **WHO environmental noise guidelines for the European region: a systematic review on environmental noise and cardiovascular and metabolic effects: a summary**. *Int J Environ Res Public Health* (2018.0) **15** 379. DOI: 10.3390/ijerph15020379 56. Clark C, Crumpler C, Notley AH. **Evidence for environmental noise effects on health for the United Kingdom policy context: a systematic review of the effects of environmental noise on mental health, wellbeing, quality of life, cancer, dementia, birth, reproductive outcomes, and cognition**. *Int J Environ Res Public Health* (2020.0) **17** 393. DOI: 10.3390/ijerph17020393 57. Hegewald J, Schubert M, Freiberg A, Romero Starke K, Augustin F, Riedel-Heller SG, Zeeb H, Seidler A. **Traffic noise and mental health: a systematic review and meta-analysis**. *Int J Environ Res Public Health* (2020.0) **17** 6175. DOI: 10.3390/ijerph17176175 58. ZareSakhvidi MJ, ZareSakhvidi F, Mehrparvar AH, Foraster M, Dadvand P. **Association between noise exposure and diabetes: a systematic review and meta-analysis**. *Environ Res* (2018.0) **166** 647-657. DOI: 10.1016/j.envres.2018.05.011 59. Albano GD, Gagliardo RP, Montalbano AM, Profita M. **Overview of the mechanisms of oxidative stress: impact in inflammation of the airway diseases**. *Antioxidants (Basel)* (2022.0) **11** 2237. DOI: 10.3390/antiox11112237 60. Cohen G, Gerber Y. **Air pollution and successful aging: recent evidence and new perspectives**. *Curr Environ Health Rep* (2017.0) **4** 1-11. DOI: 10.1007/s40572-017-0127-2 61. Babisch W. **The noise/stress concept, risk assessment and research needs**. *Noise Health* (2002.0) **4** 1-11 62. Basner M, Babisch W, Davis A, Brink M, Clark C, Janssen S, Stansfeld S. **Auditory and non-auditory effects of noise on health**. *Lancet* (2014.0) **383** 1325-1332. DOI: 10.1016/S0140-6736(13)61613-X 63. Recio A, Linares C, Banegas JR, Diaz J. **Road traffic noise effects on cardiovascular, respiratory, and metabolic health: An integrative model of biological mechanisms**. *Environ Res* (2016.0) **146** 359-370. DOI: 10.1016/j.envres.2015.12.036 64. Peters A, Nawrot TS, Baccarelli AA. **Hallmarks of environmental insults**. *Cell* (2021.0) **184** 1455-1468. DOI: 10.1016/j.cell.2021.01.043 65. Adam M, Schikowski T, Carsin AE, Cai Y, Jacquemin B, Sanchez M, Vierkötter A, Marcon A, Keidel D, Sugiri D, Al Kanani Z, Nadif R, Siroux V, Hardy R, Kuh D, Rochat T, Bridevaux PO, Eeftens M, Tsai MY, Villani S, Phuleria HC, Birk M, Cyrys J, Cirach M, de Nazelle A, Nieuwenhuijsen MJ, Forsberg B, de Hoogh K, Declerq C, Bono R, Piccioni P, Quass U, Heinrich J, Jarvis D, Pin I, Beelen R, Hoek G, Brunekreef B, Schindler C, Sunyer J, Krämer U, Kauffmann F, Hansell AL, Künzli N, Probst-Hensch N. **Adult lung function and long-term air pollution exposure. ESCAPE: a multicentre cohort study and meta-analysis**. *Eur Respir J* (2015.0) **45** 38-50. DOI: 10.1183/09031936.00130014 66. Delgado-Saborit JM, Guercio V, Gowers AM, Shaddick G, Fox NC, Love S. **A critical review of the epidemiological evidence of effects of air pollution on dementia, cognitive function and cognitive decline in adult population**. *Sci Total Environ* (2021.0) **757** 143734. DOI: 10.1016/j.scitotenv.2020.143734 67. Rice MB, Ljungman PL, Wilker EH, Dorans KS, Gold DR, Schwartz J, Koutrakis P, Washko GR, O’Connor GT, Mittleman MA. **Long-term exposure to traffic emissions and fine particulate matter and lung function decline in the Framingham heart study**. *Am J Respir Crit Care Med* (2015.0) **191** 656-664. DOI: 10.1164/rccm.201410-1875OC 68. ZareSakhvidi MJ, Lafontaine A, Yang J, Lequy E, Artaud F, Canonico M, Ozguler A, Vienneau D, Zins M, Jacquemin B. **Association between outdoor air pollution exposure and handgrip strength: findings from the French CONSTANCES study**. *Environ Health Perspect* (2022.0) **130** 57701. DOI: 10.1289/EHP10464 69. ZareSakhvidi MJ, Yang J, Lequy E, Chen J, de Hoogh K, Letellier N, Mortamais M, Ozguler A, Vienneau D, Zins M, Goldberg M, Berr C, Jacquemin B. **Outdoor air pollution exposure and cognitive performance: findings from the enrolment phase of the CONSTANCES cohort**. *Lancet Planet Health* (2022.0) **6** e219-e229. DOI: 10.1016/S2542-5196(22)00001-8 70. Li J, Vitiello MV, Gooneratne NS. **Sleep in normal aging**. *Sleep Med Clin* (2018.0) **13** 1-11. DOI: 10.1016/j.jsmc.2017.09.001 71. Guo C, Yu T, Bo Y, Lin C, Chang L-Y, Wong MCS, Yu Z, Lau AKH, Tam T, Lao XQ. **Long-term exposure to fine particulate matter and mortality: a longitudinal cohort study of 400,459 adults**. *Epidemiology* (2022.0) **33** 309-317. DOI: 10.1097/EDE.0000000000001464 72. Oakes M, Baxter L, Long TC. **Evaluating the application of multipollutant exposure metrics in air pollution health studies**. *Environ Int* (2014.0) **69** 90-99. DOI: 10.1016/j.envint.2014.03.030 73. 73.ELK. Feinstaub in der Schweiz 2013; Statusbericht der Eidgenössischen Kommission für Lufthygiene (ELK), 63 S, Bern. 2013.
--- title: Impact of BMI and smoking in adolescence and at the start of pregnancy on birth weight authors: - Rebecka Bramsved - Staffan Mårild - Maria Bygdell - Jenny M. Kindblom - Ingela Lindh journal: BMC Pregnancy and Childbirth year: 2023 pmcid: PMC10041706 doi: 10.1186/s12884-023-05529-1 license: CC BY 4.0 --- # Impact of BMI and smoking in adolescence and at the start of pregnancy on birth weight ## Abstract ### Background Birth weight is an indicator of intra-uterine conditions but also a determinant for future health. The importance of preconception health for a healthy birth weight has been emphasized, but evidence is lacking on how modifiable factors in adolescence, such as body mass index (BMI) and smoking, affect future pregnancy outcome. We evaluated associations between BMI and smoking in adolescence and at the start of pregnancy and birth weight of the first-born child. ### Methods This longitudinal study included 1256 mothers, born 1962–1992, and their first-born children, born between 1982–2016. Self-reported questionnaire information on weight, height and smoking at age 19 was cross-linked with national register data obtained at the start of pregnancy and with the birth weights of the children. Univariable and multivariable linear regressions were performed to determine the impact of maternal factors at 19 years of age and at the start of the pregnancy respectively, and the importance of BMI status at these points of time for the birth weight of the first child. ### Results BMI and smoking at the start of the pregnancy displayed strong associations with birth weight in a multivariable analysis, BMI with a positive association of 14.9 g per BMI unit ($95\%$ CI 6.0; 23.8 $$p \leq 0.001$$) and smoking with a negative association of 180.5 g ($95\%$ CI -275.7; -85.4) $$p \leq 0.0002$$). Smoking and BMI at 19 years of age did not show this association. Maternal birth weight showed significant associations in models at both time-points. Becoming overweight between age 19 and the start of the pregnancy was associated with a significantly higher birth weight (144.6 ($95\%$ CI 70.7;218.5) $$p \leq 0.0002$$) compared to mothers with normal weight at both time points. ### Conclusions Our findings indicate that the time period between adolescence and first pregnancy could be a window of opportunity for targeted health promotion to prevent intergenerational transmission of obesity. ## Background The global disease burden of overweight and obesity is now one of the greatest threats to public health worldwide [1] and has proven difficult to treat, despite great efforts and vast costs. Especially worrying is the increasing prevalence of obesity among women of reproductive age [2]. Women entering pregnancy with obesity have a higher risk of a wide range of adverse pregnancy and delivery complications. The effects of maternal obesity further extends to the next generation, creating an intergenerational cycle of obesity, with birthweight as a mediator [3, 4]. Life-style interventions in overweight women during pregnancy have not been successful in reducing the frequency of macrosomia [5, 6] and a meta-analysis concluded that weight loss prior to pregnancy is probably needed to achieve optimum pregnancy outcomes [7]. Smoking during pregnancy is another well-known modifiable risk factor that has been linked to reduced birth weight in several large studies [8, 9]. Smoking not only affects birth weight, but also linear growth and head circumference [10]. The negative effect of smoking extends into childhood and has been shown to increase the risk of childhood overweight [11]. As for overweight, interventions to curb the negative effects of smoking during pregnancy have not shown to be sufficient when initiated after conception [12]. Life course epidemiology is a useful technique capable of examining preconceptional factors and their effects on maternal and child health as it considers the timing and duration of exposures and their potential latent or cumulative effects. Adolescence could in this perspective be regarded as a sensitive period for future pregnancy outcome, since unhealthy life-style behaviors such as smoking and poor diet often originate during the teenage years [13]. There is, however, very little knowledge regarding the association between BMI and smoking in adolescence and subsequent pregnancy outcome. The longitudinal study of 19-year-old women in Gothenburg offers a unique possibility to address this knowledge gap. The aim of this study was to evaluate the association between these modifiable maternal factors, BMI and smoking in the period between adolescence and young adulthood with the birth weight of the first-born child. We hypothesized that the adolescent period is a window of opportunity for improving future pregnancy outcome and the health of the next generation. ## Study population and data sources This study originates from the prospective longitudinal study of 19-year-old women in Gothenburg, Sweden, initiated in 1981 [14]. The participants were randomly identified from the population register and were invited at age 19 in 1981 (born 1962), in 1991 (born 1972), in 2001 (born 1982) and in 2011 (born 1992) as previously described [15]. Questionnaires regarding weight, height, smoking and reproductive health were sent out by mail at the study start and subsequently every fifth year. Validation of the questionnaires was performed and is described in detail in a previous publication [16]. No changes were made in the questionnaires between study start and later on, except for the addition of newly available contraceptives among the answer options. For the purpose of the present study, all women who had given birth to at least one child after the age of 19 years at the time of data retrieval were eligible. The first-born children were born between 1982 and 2016. To avoid confounding due to prematurity, children born before 37 weeks of gestation were excluded. One child per twin pair was included. A flow chart of the inclusion process is shown in Fig. 1.Fig. 1Flowchart of inclusion and exclusion criteria for the study population *Auxiliary data* to the information from the questionnaires regarding the women and their first-born children were retrieved by linkage to Swedish national registers, using the unique personal identification number allocated to every Swedish citizen [17]. The Medical Birth Register (MBR) at the National Board of Health and Welfare [18] was initiated in 1973 for surveillance of prenatal, obstetric and neonatal care, as well as for research. High quality data on all deliveries are registered, as well as information from the start of pregnancy, usually at about 8 weeks of pregnancy. Self-reported information on smoking, alcohol consumption, occupation and family situation is obtained, and a midwife measures weight [19]. For women born in 1962 and 1972, birth weights were retrieved from archived child and school health records since the Swedish MBR had not been initiated at the time these women were born. Information on education, income and country of birth was retrieved from the Longitudinal Integrated Database for Health Insurance and Labor Market Studies (LISA) [20] and Total Population Register [21] at Statistics Sweden for the year before the birth of the first child. ## Exposure variables The main exposure variables were maternal birth weight together with weight status at 19 years of age and at the start of the pregnancy and smoking status both at 19 years age and at the start of the pregnancy. Self-reported information at 19 years of age and register derived information from the first visit to the antenatal clinic (defined as start of pregnancy) on weight and height were used to calculate BMI (weight in kg/height in meter2) and categorized into underweight (BMI < 18.5), normal weight (BMI 18.5–24.9), overweight (BMI 25–29.9) or obesity (BMI > = 30) [22]. Women who reported smoking at least 1 cigarette daily were categorized as smokers. Smoking at start of pregnancy was defined as reporting daily smoking at either 3 months before start of pregnancy or at the first visit to the maternal health clinic at 8 weeks of pregnancy. Maternal age at start of pregnancy was analyzed as a continuous variable. Education was categorized as low (0–12 years) or high (> 12 years). To assess income in the study population, the individual disposable income defined by Statistics Sweden was retrieved for the year before the birth of the first child. Income was index-related to the year 2015 and categorized as above (“high”) or below (“low”) the median income for the study population. Birth country of the parents of the women participating in the study was obtained from LISA at Statistics Sweden and categorized as Swedish born if both parents were born in Sweden. Children were categorized as small for gestational age (SGA) if having a birth weight below minus two standard deviations for their gestational age, and large for gestational age (LGA) if birth weight was above two standard deviations for their gestational age. ## Statistical analyses Descriptive data are presented with mean and standard deviation and categorical variables were presented with number and percentage. In order to test changes between 19 years of age and the start of the first pregnancy Wilcoxon sign rank test was used for continuous variables and Sign test for dichotomous and ordered categorical variables. In order to evaluate predictors of birth weight, univariable linear regressions were performed with the child’s birth weight as a dependent variable and maternal birth weight, mothers’ parents’ country of birth, maternal weight, height, BMI and smoking at 19 years as well as maternal age, weight, BMI, smoking, education and income at the start of the first pregnancy as predictors. BMI and smoking at 19 years of age and at the start of the first pregnancy were analyzed in a multivariable regression analysis, adjusted for maternal birth weight and maternal age and year of birth, and in the next step further adjusted for gestational age and sex of the child. Regression analysis was also performed based on groups of change in BMI status between 19 years of age and the start of pregnancy. This analysis was adjusted for maternal age at the first pregnancy, maternal birth year, maternal smoking at the start of pregnancy as well as gestational week and sex of the child. Beta ($95\%$ CI), standardized beta ($95\%$ CI) (inn the univariable analyses), p-value, and R2 are presented from the regression analyses. The standardized beta was calculated as beta*sx/sy, where sx and sy are the (estimated) standard deviations of x and y, respectively. Interaction testing for sex of the child and exposure variables were performed with a significance level set to 0.2. All analyses were performed using SAS 9.4 SAS Institute Inc., Cary NC, USA. All testing was made using a with alpha level 0.05 (interactions 0.20) with two-sided t-tests. ## Characteristics of the study population We included 1256 women-child pairs in the study (Fig. 1). Table 1 depicts the maternal characteristics at birth, at 19 years of age and at the start of pregnancy. The mean age at the start of the first pregnancy was 27.6 years (SD 4.6), min 20 years, max 40 years of age. Between 19 years of age and the start of the pregnancy, the mean maternal weight increased from 59.9 kg (SD 9.1) to 65.1 kg (SD 11.0). The prevalence of overweight and obesity increased accordingly during this time period, from $6.3\%$ to $20.0\%$ for overweight and from $1.4\%$ to $5.3\%$ for obesity. The number of smokers decreased, from $35.8\%$ at 19 years of age to $14.7\%$ at the start of the first pregnancy ($p \leq 0.0001$ for all comparisons). Regarding diagnoses that potentially influence birth weight, 6 women ($0.5\%$) had hypertensive disorders and 9 women ($0.7\%$) had diabetes in any form. No exclusions or adjustments were performed for these diagnoses. Table 2 shows the characteristics of the children at birth. Table 1Characteristics of the mothers ($$n = 1256$$) in the study cohortVariablesMean (SD)/percentagenBirth weight (kg)3.46 (0.52)1047Birth length (cm)50.1 (2.3)706Age at start of first pregnancy27.6 (4.6)1256Disposable Incomea167 200 [85 500]1256 Low (< median 146 800)$49.9\%$627 High (> median 146 800)$50.1\%$629Education levelb1227 Low (0–12 years)$58.6\%$719 High (> 12 years)$41.4\%$508Parental Country of Birth1223 Sweden$72.5\%$887 Other$27.5\%$33619 yearsPregnancy startWeight (kg) mean (SD)59.9 (9.1)65.1 (11.0)12271030Height (cm) mean (SD)167.1 (6.0)1245BMI (kg/m2) mean (SD)21.4 (2.8)23.3 (3.6)12251030 Underweight (< 18.5)$7.9\%$$4.8\%$9749 Normal (18.5–24.9)$84.4\%$$69.9\%$1043720 Overweight (25–29.9)$6.3\%$$20.0\%$77206 Obese (> 30)$1.4\%$$5.3\%$1755Smoking status12331181 Smoker$35.8\%$$14.7\%$441174a At the start of the first pregnancy. The sum of incomes and benefits minus taxes and negativetransfers, per year, in Swedish crowns. Index-related to the year of 2015bAt the start of the first pregnancyTable 2Characteristics of the children ($$n = 1256$$) in the study cohortVariablesMean (SD)/percentagenBirth weight (kg)3.51 (0.48)1256Birth length (cm)50.4 (2.2)1246Boys$52.3\%$657Girls$47.7\%$599Twinsa$1\%$12Large for gestational age (LGA)$1.8\%$22Small for gestional age (SGA)$3.4\%$42a One child per twin pair included ## Associations between maternal factors and infant birth weight Since birth weight differs by sex of the child, we tested whether there was an interaction between sex and the exposure factors in the association with birth weight. There was no significant interaction with sex of the child ($p \leq 0.2$) and hence results are presented for girls and boys together. Univariable linear regressions, shown in Fig. 2, demonstrated significant positive associations of infant birth weight with maternal BMI at both 19 years of age and at the start of the pregnancy. Smoking at 19 years was not significantly associated with birth weight, while smoking at the start of pregnancy was associated with a reduction of birth weight by 163 g ($95\%$ CI -238.9; -86.4, $p \leq 0$,001). Of the socioeconomic factors investigated, maternal income displayed a significant association with infant birth weight, with a reduction in birth weight of 54 g for income below the median ($95\%$ CI -106.8; -1.1 $$p \leq 0.046$$). The highest R2 was seen for maternal birth weight, height at 19 years age and weight at the start of pregnancy, each explaining $6\%$ of the variance in infant birth weight (R2 0.06 $p \leq 0.0001$).Fig. 2Associations between maternal factors and birth weight of first-born child. Results of univariable analyses presented as standardized beta (95 % CI) and beta (95 % CI)The Beta-column show the effect on birth-weight (in gram) for each unit change given for maternal features in the left-hand columnaSmoking categorized as smoker or non-smokerbDisposable income for the year before birth of the first child, income below median of the group 146 800 Swedish Crowns, index related to the year of 2015c0-12 years of educationdBoth parents of the study participant born in Sweden We evaluated the predictive value of maternal BMI and smoking at 19 years of age and at start of pregnancy in multivariable models (Table 3). The models evaluated maternal BMI, smoking and maternal birth weight effect on birth weight; Model 1A and 1B at 19 years of age and Model 2A and 2B at the start of pregnancy. All models were adjusted for maternal age at the start of the first pregnancy and birth year of mother, and models B additionally for gestational age and sex of the child. Neither maternal BMI nor smoking at 19 years of age were independently associated with birth weight of the first-born child (Table 3, Model 1A). In contrast, maternal BMI and smoking at the start of the pregnancy were both associated with birth weight (Table 3, Model 2A), BMI with a positive association of 14.9 g per BMI unit ($95\%$ CI 6.0; 26.0 $$p \leq 0.001$$) and smoking with a negative association of 180.5 g ($95\%$ CI -275.7; -85.4 $$p \leq 0.0002$$). R2 of this model was 0.08. Maternal birth weight was positively associated with birth weight in both models. Adjusting for gestational age and sex of the child increased the degree of explanation in the models but did not change the association with the examined exposures (Table 3, Model 1B and 2B).Table 3Prediction of birth weight of first-born child by BMI and smoking at 19 years of age and at the start of the first pregnancyModel 1At 19 years of age, $$n = 1006$$Model 2At start of first pregnancy, $$n = 846$$AaR2 = 0.06BbR2 = 0.24AaR2 = 0.08BbR2 = 0.24Beta ($95\%$ CI)p valueBeta ($95\%$ CI)p valueBeta ($95\%$ CI)p valueBeta ($95\%$ CI)p valueMaternal Birth Weight (100 g)22.1 (16.6; 27.5) <.000119.0 (14.0; 24.0) <.000120.1 (14.1;26.0) <.000119.5 (14.1;24.9) <.0001BMI (kg/m2)7.0 (-4.1; 18.0)0.223.2 (-6.8; 13.3)0.5314.9 (6.0; 23.8)0.001012.1 (4.0; 20.3)0.0033Smoking-29.9 (-89.8;30.1)0.33-30.3 (-84.6; 24.0)0.27-180.5 (-275.7; -85.4)0.0002-154.1 (-241.1;-67.2)0.0005a Adjusted for maternal age and birth yearb Adjusted for maternal age, birth year, gestational age and sex of child ## Impact of changes in BMI status between 19 years of age and the start of the first pregnancy on the birth weight of the first-born child In multivariable regression models grouped according to change in BMI status between19 years of age and the start of the first pregnancy (shown in Fig. 3), shifting from normal weight at 19 years of age to overweight or obesity at the start of the pregnancy ($$n = 176$$) was associated with significantly higher birth weight of the first-born child compared to that of women with normal weight at both time points ($$n = 626$$), Beta 144.6 ($95\%$ CI 70.7;218.5) $$p \leq 0.0002.$$Fig. 3Impact of changes in BMI status from 19 years of age to the start of the first pregnancy on birth weight, results from a multivariable regression model ($$n = 982$$)*Beta (95 %CI) adjusted for maternal birth year, age at start of pregnancy and smoking, gestational week and sex of childThe group with normal weight (BMI 18.0-24.9 kg/m2) at both 19 years of age and at start of pregnancy is used as referencenw – normal weight, ow/ob – overweight or obesity, uw – underweight, y – years Children born to women who were overweight or obese at both age 19 and at the start of the pregnancy ($$n = 71$$), did not differ in birth weight from the children of mothers in the normal weight reference group. The groups with 54, 19, 26 and 10 individuals respectively were small, with wide confidence intervals, and thus no certain conclusions can be made for these BMI groups. ## Discussion A healthy birth weight is important for the prevention of intergenerational transmission of obesity [23]. Studies indicate that interventions during pregnancy are insufficient in their aim to normalize birth weight, highlighting the importance of preconceptional health [13]. In the present study, which included 1256 women-child pairs, we evaluated the influence of the women’s BMI and smoking in both late adolescence, at 19 years of age, and at the start of the first pregnancy on the birth weight of her first born child. To our knowledge, this is the first study evaluating BMI and smoking in late adolescence in relation to future pregnancy outcomes. We found that BMI and smoking at 19 years of age and at the start of the pregnancy differed regarding association with birth weight. The impact on birth weight of these variables in adolescence was negligible, while both factors at the start of the pregnancy were significantly associated. We hypothesized that the changes that occurred in maternal BMI or smoking from age 19 to the start of the first pregnancy may cause birth weight deviations associated with an increased risk of future disease. Both BMI and smoking changed significantly between these two time points. Mean BMI and prevalence of overweight and obesity increased, and we found that developing overweight after the age of 19 was associated with higher birth weight of the first-born child compared to children of women that maintained a normal weight. The rate of smoking, on the other hand, decreased significantly. Initiating smoking during the adolescent years has widespread negative long-term health effects [24], and smoking cessation remains a public health priority. Our results show that there is no persistent negative effect on birth weight related to smoking in late adolescence if terminated before the start of pregnancy. These results indicate that there are great health benefits to be gained by intensifying efforts to promote smoking cessation in young women in time to avoid affecting future pregnancy, since previous studies have shown that cessation of smoking in early pregnancy is not enough to prevent the negative impact on fetal growth [12]. Maternal obesity when entering pregnancy has been shown to influence not only birth weight but also the distribution and amount of fat mass in relationship to lean mass in the neonate [25]. Maternal BMI seems to be independently associated with increased adiposity in the new-born [26] and an additive effect of high birth weight and obesity at the start of pregnancy has been observed for the risk of having a LGA child [27]. However, the importance of whether maternal overweight is present since childhood or develops after adolescence for the birth weight of their first-born child is not known. No difference was seen for birth weight of the first-born child for women who were overweight or obese at both age 19 and at the start of the pregnancy compared with women with a normal weight at both age 19 and at the start of pregnancy, in contrast to developing overweight or obesity after adolescence. We can only speculate on the reason behind this difference. A recent weight gain in a woman entering pregnancy could be associated with higher levels of IGF-1 and insulin than in a woman with stable overweight. These levels might affect body composition and the birth weight of the child and increase the risk of unhealthy weight development in childhood. However, these pathophysiological theories need to be confirmed in future studies. Our findings support the conclusions by Hanson et al., that lifestyle interventions in pregnant women do not limit gestational weight gain to the degree required to have a meaningful impact on pregnancy outcomes. Focus for interventions should be in the preconception period [28]. Our findings suggest a window of opportunity to promote healthy birth weight development for the first child between late adolescence and pregnancy. Socioeconomic factors, such as family income and education, are well known factors affecting the risk for development of childhood obesity. In a previous study from our group, we found parental income to be positively associated with birth weight [29]. In the present study, income was weakly associated with birth weight, while the other socioeconomic factors investigated did not show any association with birth weight. These findings suggest that socioeconomic status mainly affects childhood BMI development through later lifestyle related factors. ## Strengths and limitations The main strengths of this study are the availability of longitudinal information for each woman on weight, height and smoking status at 19 years of age. This setting allows us to evaluate weight change and changes in smoking status during the time interval between late adolescence and at the start of pregnancy. The national Swedish registers are reliable and nearly complete [19], which allowed us to examine the effects of numerous exposures, such as smoking, maternal age, country of birth, income and education on birth weight. Information regarding these factors has been requested in previous studies on the subject of maternal overweight and birth weight of her child [4]. The limitations include a relatively low prevalence of overweight and obesity among mothers both at 19 years of age and at the start of pregnancy, which might be due to selection bias in the questionnaire study. The prevalence of birth weights outside normal range was also low, which prevented analyses of risk of LGA or SGA in the child. ## Conclusions Obesity is the leading cause of poor health worldwide today [1, 30]. The increasing prevalence of pre-conceptional maternal obesity is especially worrying knowing that it has a negative effect on the health of the next generation [2]. There is increasing evidence suggesting that interventions during pregnancy are not sufficient to reduce the negative effect of maternal overweight [13] and attention must be directed towards better nutrition in adolescence [31]. In this study, we had the rare opportunity to study BMI and smoking habits in women both at 19 years of age and at the start of their first pregnancy. We demonstrate that smoking at age 19 is not independently associated with birth weight of the first-born child, high-lighting late adolescence as highly important for promotion of smoking cessation when striving for a healthy birth weight in the next generation. Our results further indicate that weight gain during the period between late adolescence and the first pregnancy might be of importance for infant birth weight. ## References 1. Afshin A, Reitsma MB, Murray CJL. **Health effects of overweight and obesity in 195 countries**. *N Engl J Med* (2017.0) **377** 1496-1497. PMID: 29020584 2. Poston L, Caleyachetty R, Cnattingius S, Corvalan C, Uauy R, Herring S. **Preconceptional and maternal obesity: epidemiology and health consequences**. *Lancet Diabetes Endocrinol* (2016.0) **4** 1025-1036. DOI: 10.1016/S2213-8587(16)30217-0 3. Heslehurst N, Vieira R, Akhter Z, Bailey H, Slack E, Ngongalah L. **The association between maternal body mass index and child obesity: a systematic review and meta-analysis**. *PLoS Med* (2019.0) **16** e1002817. DOI: 10.1371/journal.pmed.1002817 4. Yu Z, Han S, Zhu J, Sun X, Ji C, Guo X. **Pre-pregnancy body mass index in relation to infant birth weight and offspring overweight/obesity: a systematic review and meta-analysis**. *PLoS ONE* (2013.0) **8** e61627. DOI: 10.1371/journal.pone.0061627 5. Dodd JM, Turnbull D, McPhee AJ, Deussen AR, Grivell RM, Yelland LN. **Antenatal lifestyle advice for women who are overweight or obese: LIMIT randomised trial**. *BMJ* (2014.0) **10** g1285. DOI: 10.1136/bmj.g1285 6. Poston L, Bell R, Croker H, Flynn AC, Godfrey KM, Goff L. **Effect of a behavioural intervention in obese pregnant women (the UPBEAT study): a multicentre, randomised controlled trial**. *Lancet Diabetes Endocrinol* (2015.0) **3** 767-777. DOI: 10.1016/S2213-8587(15)00227-2 7. Farpour-Lambert NJ, Ells LJ, Martinez de B, Scott C. **Obesity and weight gain in pregnancy and postpartum: an evidence review of lifestyle interventions to inform maternal and child health policies**. *Front Endocrinol (Lausanne)* (2018.0) **9** 546. DOI: 10.3389/fendo.2018.00546 8. Salihu HM, Wilson RE. **Epidemiology of prenatal smoking and perinatal outcomes**. *Early Hum Dev* (2007.0) **83** 713-720. DOI: 10.1016/j.earlhumdev.2007.08.002 9. Ko TJ, Tsai LY, Chu LC, Yeh SJ, Leung C, Chen CY. **Parental smoking during pregnancy and its association with low birth weight, small for gestational age, and preterm birth offspring: a birth cohort study**. *Pediatr Neonatol* (2014.0) **55** 20-27. DOI: 10.1016/j.pedneo.2013.05.005 10. Rumrich I, Vahakangas K, Viluksela M, Gissler M, de Ruyter H, Hanninen O. **Effects of maternal smoking on body size and proportions at birth: a register-based cohort study of 1.4 million births**. *BMJ Open* (2020.0) **10** e033465. DOI: 10.1136/bmjopen-2019-033465 11. Oken E, Levitan EB, Gillman MW. **Maternal smoking during pregnancy and child overweight: systematic review and meta-analysis**. *Int J Obes (Lond)* (2008.0) **32** 201-210. DOI: 10.1038/sj.ijo.0803760 12. Lewandowska M, Więckowska B, Sztorc L, Sajdak S. **Smoking and smoking cessation in the risk for fetal growth restriction and low birth weight and additive effect of maternal obesity**. *J Clin Med* (2020.0) **9** 3504. DOI: 10.3390/jcm9113504 13. Stephenson J, Heslehurst N, Hall J, Schoenaker D, Hutchinson J, Cade JE. **Before the beginning: nutrition and lifestyle in the preconception period and its importance for future health**. *Lancet* (2018.0) **391** 1830-1841. DOI: 10.1016/S0140-6736(18)30311-8 14. Andersch B, Milsom I. **Contraception and pregnancy among young women in an urban Swedish population**. *Contraception* (1982.0) **26** 211-219. DOI: 10.1016/0010-7824(82)90070-1 15. Lindh I, Ellstrom AA, Blohm F, Milsom I. **A longitudinal study of contraception and pregnancies in the same women followed for a quarter of a century**. *Hum Reprod* (2010.0) **25** 1415-1422. DOI: 10.1093/humrep/deq095 16. Lindh I, Ellstrom AA, Milsom I. **The effect of combined oral contraceptives and age on dysmenorrhoea: an epidemiological study**. *Hum Reprod* (2012.0) **27** 676-682. DOI: 10.1093/humrep/der417 17. Ludvigsson JF, Otterblad-Olausson P, Pettersson BU, Ekbom A. **The Swedish personal identity number: possibilities and pitfalls in healthcare and medical research**. *Eur J Epidemiol* (2009.0) **24** 659-667. DOI: 10.1007/s10654-009-9350-y 18. 18.Socialstyrelsen. The Swedish Medical Birth Register Stockholm 2018. Tillgänglig från. https://www.socialstyrelsen.se/en/statistics-and-data/registers/register-information/the-swedish-medical-birth-register/. hämtad 2 Sept 2020. 19. 19.Källén B, Källen K. The Swedish Medical Birth Register - A summary of content and quality. Centre for Epidemiology. Swedish National Board of Health and Welfare; 2003. p. 3. Available at https://www.socialstyrelsen.se/globalassets/sharepoint-dokument/artikelkatalog/ovrigt/2003-112-3_20031123.pdf. 20. Ludvigsson JF, Svedberg P, Olen O, Bruze G, Neovius M. **The longitudinal integrated database for health insurance and labour market studies (LISA) and its use in medical research**. *Eur J Epidemiol* (2019.0) **34** 423-437. DOI: 10.1007/s10654-019-00511-8 21. Ludvigsson JF, Almqvist C, Bonamy AK, Ljung R, Michaelsson K, Neovius M. **Registers of the Swedish total population and their use in medical research**. *Eur J Epidemiol* (2016.0) **31** 125-136. DOI: 10.1007/s10654-016-0117-y 22. 22.WHOObesity: preventing and managing the global epidemic2000GenevaWorld Health Organization. *Obesity: preventing and managing the global epidemic* (2000.0) 23. Adamo KB, Ferraro ZM, Brett KE. **Can we modify the intrauterine environment to halt the intergenerational cycle of obesity?**. *Int J Environ Res Public Health* (2012.0) **9** 1263-1307. DOI: 10.3390/ijerph9041263 24. Amialchuk A, Sapci O. **The long-term health effects of initiating smoking in adolescence: evidence from a national longitudinal survey**. *Health Econ* (2022.0) **31** 597-613. DOI: 10.1002/hec.4469 25. Sewell MF, Huston-Presley L, Super DM, Catalano P. **Increased neonatal fat mass, not lean body mass, is associated with maternal obesity**. *Am J Obstet Gynecol* (2006.0) **195** 1100-1103. DOI: 10.1016/j.ajog.2006.06.014 26. Longmore DK, Barr ELM, Lee IL, Barzi F, Kirkwood M, Whitbread C. **Maternal body mass index, excess gestational weight gain, and diabetes are positively associated with neonatal adiposity in the Pregnancy and Neonatal Diabetes Outcomes in Remote Australia (PANDORA) study**. *Pediatr Obes* (2019.0) **14** e12490. DOI: 10.1111/ijpo.12490 27. Cnattingius S, Villamor E, Lagerros YT, Wikstrom AK, Granath F. **High birth weight and obesity–a vicious circle across generations**. *Int J Obes (Lond)* (2012.0) **36** 1320-1324. DOI: 10.1038/ijo.2011.248 28. Hanson M, Barker M, Dodd JM, Kumanyika S, Norris S, Steegers E. **Interventions to prevent maternal obesity before conception, during pregnancy, and post partum**. *Lancet Diabetes Endocrinol* (2017.0) **5** 65-76. DOI: 10.1016/S2213-8587(16)30108-5 29. Bramsved R, Regber S, Novak D, Mehlig K, Lissner L, Marild S. **Parental education and family income affect birthweight, early longitudinal growth and body mass index development differently**. *Acta Paediatr* (2018.0) **107** 1946-1952. DOI: 10.1111/apa.14215 30. Swinburn BA, Kraak VI, Allender S, Atkins VJ, Baker PI, Bogard JR. **The global syndemic of obesity, undernutrition, and climate change: the lancet commission report**. *Lancet* (2019.0) **393** 791-846. DOI: 10.1016/S0140-6736(18)32822-8 31. **The hidden crisis of adolescent nutrition**. *Lancet Child Adolesc Health.* (2022.0) **6** 1. DOI: 10.1016/S2352-4642(21)00381-3
--- title: Multifunctional porous poly (L-lactic acid) nanofiber membranes with enhanced anti-inflammation, angiogenesis and antibacterial properties for diabetic wound healing authors: - Hao Yu - Yijia Li - Yining Pan - Hongning Wang - Wei Wang - Xiaobin Ren - Hang Yuan - Ziru Lv - Yijia Zuo - Zhirong Liu - Wei Lin - Qingqing Yao journal: Journal of Nanobiotechnology year: 2023 pmcid: PMC10041712 doi: 10.1186/s12951-023-01847-w license: CC BY 4.0 --- # Multifunctional porous poly (L-lactic acid) nanofiber membranes with enhanced anti-inflammation, angiogenesis and antibacterial properties for diabetic wound healing ## Abstract With increased diabetes incidence, diabetic wound healing is one of the most common diabetes complications and is characterized by easy infection, chronic inflammation, and reduced vascularization. To address these issues, biomaterials with multifunctional antibacterial, immunomodulatory, and angiogenic properties must be developed to improve overall diabetic wound healing for patients. In our study, we prepared porous poly (L-lactic acid) (PLA) nanofiber membranes using electrospinning and solvent evaporation methods. Then, sulfated chitosan (SCS) combined with polydopamine-gentamicin (PDA-GS) was stepwise modified onto porous PLA nanofiber membrane surfaces. Controlled GS release was facilitated via dopamine self-polymerization to prevent early stage infection. PDA was also applied to PLA nanofiber membranes to suppress inflammation. In vitro cell tests results showed that PLA/SCS/PDA-GS nanofiber membranes immuomodulated macrophage toward the M2 phenotype and increased endogenous vascular endothelial growth factor secretion to induce vascularization. Moreover, SCS-contained PLA nanofiber membranes also showed good potential in enhancing macrophage trans-differentiation to fibroblasts, thereby improving wound healing processes. Furthermore, our in vitro antibacterial studies against *Staphylococcus aureus* indicated the effective antibacterial properties of the PLA/SCS/PDA-GS nanofiber membranes. In summary, our novel porous PLA/SCS/PDA-GS nanofiber membranes possessing enhanced antibacterial, anti-inflammatory, and angiogenic properties demonstrate promising potential in diabetic wound healing processes. ## Background According to 2019 International Diabetes Federation data, an estimated 463 million adults had diabetes, and the numbers are expected to increase to over 700 million by 2045 [1]. With increased diabetes mellitus rates, the incidence of diabetic wounds, one of the most common diabetic complications, has been challenging in clinical settings owing to long treatment duration, high healthcare costs, high recurrence rates, and mortality [2, 3]. Generally, normal wound healing processes are characterized by typical overlap stages comprising hemostasis, inflammation, proliferation, and tissue remodeling [4, 5]. However, diabetic wound healing is hindered by chronic inflammation, ease of infection, and reduced neovascularization [6, 7]. Long-term inflammatory diabetic conditions generate reactive oxygen species (ROS) and inflammatory cytokine overexpression (such as interleukin-6 (IL-6), TNF-alpha (TNF-α) and IL-1beta (IL-1β)) [8]. Also, increased M1 macrophage polarization causes chronic inflammation and excessive inflammation hinders the transition from inflammation to proliferation stages. Hence, immunomodulating macrophage polarization toward M2 phenotypes (anti-inflammatory) to facilitate transition from inflammation to proliferation stages could effectively promote diabetic wound healing [9, 10]. In recent years, polydopamine (PDA) has attracted considerable attention in the wound healing field owing to its good anti-inflammatory activity, ROS scavenging ability, good tissue adhesion, and excellent cell affinity properties [8, 11, 12]. PDA contains biomaterials with rich reductive functional groups, such as catechol and amine groups, which help scavenge radical species [11, 13]. For example, Ma et al. showed that PDA-decorated microneedles inhibited ROS-induced inflammation, further promoting M2 macrophage polarization, suppressing wound inflammation, and facilitating wound healing [14]. Zhang et al. reported that PDA-treated titanium significantly reduced M1 macrophages by activating nuclear factor-κappaB signaling [15]. Additionally, due to plentiful phenol groups, PDA strongly adheres to tissues, including human skin [11]. Hence, PDA is an easy-to-modify substrate and provides anti-inflammatory microenvironments for wound healing. Angiogenesis is a key process that accelerates diabetic wound healing by supplying oxygen and nutrients [16]. Typically, growth factors, such as vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (bFGF), have been used to induce vascularization in diabetic wounds in clinical settings. However, growth factor-based therapy success is significantly hindered by high costs, high-dose requirements, relatively short half-life, and serious side effects [7]. Thus, new strategies need to be investigated. Recently, several studies suggested that M2 macrophages facilitated vascularization as they secreted angiogenic factors, such as platelet-derived growth factor (PDGF) and VEGF [17, 18]. Yu et al. reported that sulfated chitosan (SCS) had high affinity for VEGF, promoted VEGF binding to the VEGF receptor 2 (VEGFR2) and stimulated endothelial cell proliferation, migration, tubule networking, and promoted VEGF-mediated angiogenesis [19]. Studies also indicated that SCS induced macrophages toward the M2 phenotype via the interleukin-4 (IL-4) mediated Stat6 signaling pathway. Moreover, M2 macrophages enhanced macrophage trans-differentiation into fibroblasts [20, 21]. Therefore, SCS may suppress inflammation at early stages and promote neovascularization [20, 21]. Diabetic wounds are prone to bacterial infections because of constant exposure to external environments, which impedes wound treatment [22–24]. Therefore, strategies incorporating multifunctional scaffolds with antimicrobial, immunomodulatory, and angiogenic activities for diabetic wound healing must be promoted [25–27]. Although various wound dressing materials have been used for wound healing, nanofiber scaffolds that mimic both extracellular matrix (ECM) composition and structure have shown promising applications in skin regeneration [28, 29]. The electrospinning method is capable to produce nanofibers with diameters similar to those of natural ECM, and large specific electrospun nanofiber surface areas, together with high electrospun nanofibrous scaffold porosity, facilitate cell adhesion, proliferation, migration and differentiation [30–32]. Electrospun nanofibers with porous structures can also be used as drug carriers for drug release, thus inducing particular cell behaviors. Moreover, nanofiber porous structures increase the specific surface areas and also create opportunities for liquid and gas exchange to facilitate microenvironments for wound healing [1, 33]. Recently, polylactic acid (PLA), an FDA approved biopolymer, has attracted great attention in biomedical applications due to its good biocompatibility, biodegradability, relatively high mechanical properties and easy to process [34]. In this study, porous PLA nanofibers were prepared via electrospinning method by varying solvent ratios, PLA concentrations, and electric field intensity parameters. SCS was applied to these nanofibers to improve the hydrophilicity of PLA nanofiber membranes. SCS also drove macrophage polarization toward the M2 phenotype, thus suppressing inflammation and promoting neovascularization. PDA and the antibiotic gentamicin sulfate (GS) were decorated onto PLA/SCS nanofiber membranes via a PDA-assisted assembly strategy to obtain effective anti-inflammatory and antibacterial properties. PDA also facilitated strong fibrous membrane adherence to wounded tissues via electrostatic interactions and covalent bonds. In summary, porous PLA nanofiber membranes with angiogenic, anti-inflammatory, and antibacterial properties with possible applications in diabetic wound healing were developed. ## Materials PLA and GS were purchased from Sigma-Aldrich. Chitosan (CS, $95\%$ deacetylated powder) and dopamine hydrochloride were provided by Macklin Biochemical Co., Ltd. Chlorosulfonic acid, dichloromethane (DCM) and N, N-dimethylformamide (DMF) were purchased from Aladdin Reagent Co. (Shanghai, China). ## SCS fabrication and characterization CS was dissolved in formamide at 50 ℃ with stirring to generate a 2wt% homogeneous CS solution. Then, a chlorosulfonic acid/DMF (volume ratio = 4:1) mixed solution was dropped into the CS solution and reacted at 50 ℃ for 2 h. This solution against deionized water (DI) for 96 h and replaced the DI water twice a day. Finally, the resultant solution was stored at − 80 ℃ overnight, and then freeze dried for 72 h. To determine CS and SCS chemical composition, Fourier-transform infrared spectroscopy (FTIR, Nicolet, USA) was performed. Samples were prepared by mixing 1 mg of powder with 200 mg of KBr powder and by pressing into pellets. The FTIR spectra were collected in the 4000–400 cm−1 range in the transmission mode with a resolution of 4 and scans of 32. To confirm the CS sulfonation, proton nuclear magnetic resonance (1H NMR) measurements were performed. CS and SCS powders were dried in the oven overnight at 80 ℃, then dissolved in deuterium oxide (D2O) at the concentration of 1 mg/mL, and 1H NMR data were collected at 37 °C and analyzed using a MestReNova software. ## In vitro angiogenesis assays and immune response of SCS The influence of SCS on the angiogenic and immune responses of macropahges, RT-PCR was performed as previously reported [34]. For the SCS, Raw 264.7 cells (5 × 104) were seeded on the 24-well plate and cultured overnight, after that SCS/culture medium was added. Total RNA was extracted from macrophages using the GeneJET™ RNA Purification Kit and reverse transcribed into cDNA using the High Capacity cDNA Reverse Transcript kit according to the manufacturer’s instructions. Angiogenic marker (hypoxia-inducible factor 1-alpha (HIF-1α) and VEGF and inflammatory gene marker expression (tumor necrosis factor-α (TNF-α), IL-1β, IL-6, IL-4, IL-10, and Arg-1) were quantified using RT-PCR. β-actin was the housekeeping gene. The primer sequences are listed in Table 1.Table 1RT-PCR primer sequencesGene nameForward primerReverse primerHIF-1αVEGFACCTTCATCGGAAACTCCAAAGGTCCTCTCCTTACCCCACCTCCTCTGTTAGGCTGGGAAAAGTTAGGCTCACACACACAGCCAAGTCTCCTIL-6TGTGTTTTCCTCCTTGCCTCTGATTGCTGCCTAATGTCCCCTTGAATIL-1βTGTGTTTTCCTCCTTGCCTCTGATTGCTGCCTAATGTCCCCTTGAATTNF-αCTTGTTGCCTCCTCTTTTGCTTACTTTATTTCTCTCAATGACCCGTAGIL-4GCGTGCTTGCTGGTTCTGTCCTGGGCTCCCTCTCArg-1GGCAACCTGTGTCCTTTCTCCTCCCAGCTTGTCTACTTCAGTCATGIL-10GGAAGACAATAACTGCACCCACTCAACCCAAGTAACCCTTAAAGTCC ## Preparation and characterization of PLA nanofiber membranes PLA was dissolved in DCM/DMF mixtures (volume ratios of $\frac{9}{1}$, $\frac{8}{2}$, and $\frac{7}{3}$) at 10 wt% and the solution transferred to a 5 mL syringe for electrospinning. The parameters were as follows: flow rate = 1 mL/h, collecting distance = 20 cm, and voltage = 20 kV. Porous PLA fiber membrane morphologies were observed using scanning electron microscopy (SEM, Prox, Phenom) at 10 kV accelerating voltage. Samples were sputter-coated with gold for 40 s. Fiber diameters and distributions were measured using SEM images by ImageJ (win64, National Institutes of Health, USA), and more than 100 fibers were randomly were analyzed per sample. ## Preparation and characterization of PLA-based nanofiber membranes To prepare PLA/SCS nanofiber membranes, PLA nanofiber membranes were immersed in SCS solution (10 mg/mL) for 1 h and dried overnight in a fume hood. To prepare PLA/SCS/PDA nanofiber membranes, PLA/SCS nanofiber membranes were immersed in Tris-buffer (pH 8), dopamine hydrochloride (10 mg/mL) was then added into the solution. After that, the mixed solutions were reacted overnight with stirring. To prepare PLA/SCS/PDA-GS nanofiber membranes, the prepared PLA/SCS nanofiber membranes were mixed with the Tris-buffer, dopamine hydrochloride and GS were added into the solution and reacted overnight with stirring. ## Wettability tests Porous PLA, PLA/SCS, PLA/SCS/PDA, and PLA/SCS/PDA-GS nanofiber membranes with a dimension of 2 × 2 mm were subjected to the contact angle test. The wettability of the prepared samples was determined using a water contact angle instrument (OCA25, Dataphysics, Germany). Three parallel tests were measured on each nanofiber membrane groups. ## Swelling tests Equilibrated swelling ratios (ESRs) of porous PLA, PLA/SCS, PLA/SCS/PDA, and PLA/SCS/PDA-GS nanofiber membranes were measured in phosphate buffered saline (PBS) solution. The nanofiber membranes were first weighed (dry weight (M0)) and then transferred into PBS at 37 ℃. The nanofiber membranes were removed carefully at different time points and weighed after removing the excess solution (MW) until the swelling equilibrium was reached. ESRs were calculated as follows:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\text{ESRs }} = \, \left({{\text{M}}_{{\text{W}}} - {\text{M}}_{0} } \right)/{\text{M}}_{0} \times { 1}00\%$$\end{document}ESRs=MW-M0/M0×$100\%$where M0 and Mw = membrane weights under dry and swollen conditions, respectively. Three repeated measurements were performed. ## In vitro GS release PLA/SCS/PDA-GS nanofiber membranes were used in the in vitro GS release study. The prepared nanofiber membranes were immersed in Dulbecco's phosphate-buffered saline (DPBS) medium at 37 ℃ and agitated at 90 rpm. At different time points, half the supernatant was collected and the same volume of fresh DPBS was added. After 7 days of release, the released GS amounts were determined using UV-Vis spectrophotometry (ZF-20D, Shanghai, China), as described elsewhere [35]. ## In vitro cell culture Mouse macrophages cell line (Raw 264.7) and human umbilical vein endothelial cells (HUVECs) were purchased from the Procell Life Science&Technology Company (Wuhan, China). Raw 246.7 cells and HUVECs were cultured with DMEM high-glucose. All medium supplemented with $10\%$ FBS and $1\%$ penicillin/streptomycin (P/S, Beyotime). Moreover, all cells were cultured in a humid incubator at 37 °C and a CO2 level of $5\%$. ## In vitro cytotoxicity assays The morphologies of Raw264.7 cells on porous PLA, PLA/SCS, PLA/SCS/PDA, and PLA/SCS/PDA-GS nanofiber membranes were determined using the Calcein-AM and propidium iodide (PI) stains, which labeled live and dead cells, respectively. At each incubation time point, the culture medium was removed and the cells were washed with three times in DPBS. Calcein-AM and PI in DPBS were then added to the 24-well plates and incubated for 15 min at room temperature. Fluorescence images were recorded using a laser scanning microscope (DFC7000 GT DMi8, Leica, Germany). Nanofiber membrane cytotoxicity was quantitatively analyzed using CCK8 assays (Dojindo Molecular Technologies, Inc.) according to the manufacturer’s instruction. Briefly, porous PLA, PLA/SCS, PLA/SCS/PDA, and PLA/SCS/PDA-GS nanofiber membranes were placed in 24-well plates, 5 × 104 Raw264.7 cells were then seeded onto the prepared nanofiber membranes. After 1 and 3 days of culture, the culture medium was removed, $10\%$ CCK8/culture medium was added, the cells were incubated for 2 h, and the absorbance was measured at 450 nm on a microplate reader (Infinite M200, Tecan, USA). PLA nanofiber membranes acted as controls, and the cell viability was expressed as percentages relative to the control group. ## Real-time polymerase chain reaction (RT-PCR) To determine the influence of the PLA-based nanofiber membrane on macrophage angiogenic and immune responses, RT-PCR was performed. For PLA-based samples, Raw 264.7 cells (5 × 104) were seeded on PLA, PLA/SCS, PLA/SCS/PDA, and PLA/SCS/PDA-GS nanofiber membranes and cultured for 1 and 3 days. *And* gene expression data were acquired. ## In vitro antibacterial culture The antibacterial activities of porous PLA, PLA/SCS, PLA/SCS/PDA, and PLA/SCS/PDA-GS nanofiber membranes were studied using the Gram-positive bacteria *Staphylococcus aureus* (S. aureus) (ATCC 6538). Same volume of Staphylococcus aureus-containing suspensions were spread on the agar surface for inoculation. Zones of inhibition (ZOI), shake-flask culture, and bacterial live/dead staining tests were carried out to determine the anti-adhesive and bactericidal functions of nanofiber membranes toward S. aureus. ## ZOI assays S. aureus suspension (1.0 × 106 CFU/mL) were uniformly spread onto nutrient agar, afterward, the prepared PLA, PLA/SCS, PLA/SCS/PDA, and PLA/SCS/PDA-GS nanofiber membranes were placed on the Petri dishes and incubated at 37 ºC for 24 h. ZOIs were measured using a perpendicular caliper. Three parallel samples were measured to assess the antimicrobial activity of the nanofiber membrane. ## Bacteria killing tests Bacteria killing properties were studied by immersing the nanofiber memberanes into bacterial solutions in shake-flask tests. After immersing for 15 min, bacterial solutions had homogeneous coated the membranes, which were placed onto solid agar and incubated, after which bacterial numbers were counted. The survival rate was defined as the percentage of bacteria relative to the initial total number in the suspension. ## Live/dead bacterial staining S. aureus suspensions (1.0 × 105 CFU/mL) were seeded onto the PLA, PLA/SCS, PLA/SCS/PDA, and PLA/SCS/PDA-GS nanofiber membranes and co-cultured for 24 h. Samples were then stained using a live/dead BacLight bacterial viability kit (L-7012, Invitrogen) according to the manufacturer’s instructions. Bacterial morphology was observed by a fluorescence microscopy (Zeiss, Germany). ## Statistical analysis The One-way ANOVA with Tukey’s correction (compare among groups) and Student’s t-tests (between two groups) were used to calculate statistical significance. A value of $p \leq 0.05$ was considered to be statistically significant. All the data are presented as the mean ± standard deviation (SD). ## SCS synthesis and characterization To analyze CS sulfation, FTIR and 1H NMR were measured (Fig. 1a, b). The peak around 3422 cm−1 was belonged to the overlapping of the stretching vibration of O–H and the stretching vibration of N–H groups of chitosan. The peak at 895 cm−1 was the stretching vibration peak of C-O. And the peak around 1600 cm−1 in chitosan was attributed to the bending vibration peak of protonated amino group. When compared with CS, the FTIR spectra of SCS showed the new bands at approximately 810 cm−1 and 1230 cm−1, which were attributed to C-O-S and O = S = O group stretching vibrations, respectively. Moreover, peaks at 3.3 ppm and 4.6 ppm in SCS 1H NMR spectra of SCS confirmed that the sulfonation had occurred positions in C2 and C6 of CS. These data suggest successful CS sulfation. Fig. 1a CS and SCS FTIR spectra; b SCS 1H NMR spectrum; c HIF-1α and d VEGF gene expression in Raw.264.7 cells in SCS/culture medium for 24 h; relative gene expression of e, g TNF-α, IL-1β, IL-6 and f, h IL-4, IL-10, Arg-1 of Raw 264.7 cells in SCS/culture medium for 24 h and 48 h of culture, respectively *Angiogenesis is* a complex and coordinated process involving multiple-factors [36, 37]. It remains a challenge to induce vascularization in engineered tissues by delivering just one growth factor (e.g., VEGF or PDGF-BB). As a key upstream transcription factor, HIF-1α plays important roles in bone tissue engineering via VEGF [38, 39] and stromal cell-derived factor 1 (SDF-1) generation to upregulate the VEGF-induced vascularization and SDF-1 induced progenitor cells recruitment [40, 41]. HIF-1α and VEGF gene expression levels in Raw 264.7 cells were significantly increased after 24 h of SCS treatment (Fig. 1c, d). The pro-angiogenic potential of SCS in human umbilical vein endothelial cells (HUVECs) was also evaluated using capillary tube formation assays (Fig. 2). The exogenous VEGF- (6 ng/mL) treated group was used as a positive control. Capillary-like networks were formed on VEGF- and SCS-treated HUVECs after 3 h of treatment, whereas few capillary tubes were observed in the control group (***$p \leq 0.001$). All branch points, loops per field, and total capillary tube length per field were increased in VEGF and SCS-treated groups in the first 6 h. However, at 9 h, capillary-like networks no longer increased in the VEGF treated group, but more capillary-like networks were formed in the SCS treated group. Thus, SCS alone promoted capillary tube formation without exogenous VEGF addition, with higher enhanced capabilities when compared with the VEGF-treated group. Moreover, SCS effects were was longer than VEGF effects, which could be attributed to the short half-life of VEGF. Therefore, SCS alone promoted angiogenesis in HUVECs and induced pro-angiogenic factor expression in macrophages without exogenous VEGF addition. When compared with VEGF, SCS was easily available, cheap, had low toxicity, and was stable. In the future, SCS may be used as a VEGF substitute in clinical settings. Fig. 2Representative bright light images showing the capillary networks after 3 h, 6 h, 9 h of (a, e, i) Control, (b, f, j) VEGF and (c, g, k) SCS treatment, respectively; quantitative analysis of sprouted HUVECs on (d) number of branch points per field, (h) the number of loops per field and (l) total capillary tube length per field To study the influence of SCS on the macrophage polarization, pro-inflammatory and anti-inflammatory gene expression was investigated using RT-PCR. After 1 day of SCS treatment, M1 phenotype-related genes (IL-1β and IL-6) were decreased, whereas the M2 phenotype-related gene IL-4 was increased. Interestingly, after SCS incubation for 3 days, both M1 (TNF-α, IL-1β, and IL-6) and M2 phenotype-related genes (IL-4, IL-10, and Arg-1) were dramatically increased. As Shen et al. reported, some macrophages participate in proliferation processes during diabetic wound healing, and M1/M2 double-positive macrophages can benefit macrophage trans-differentiation to fibroblasts [21]. Therefore, SCS appears to regulate macrophages toward an anti-inflammatory phenotype and potentially promote macrophage trans-differentiation into fibroblasts. ## Synthesis and characterization of PLA-based nanofiber membranes Our preliminary data showed that the mixture solvent ratios play the leading role in PLA nanofiber porous structures of PLA nanofibers when compared with the PLA concentrations, electric field intensity, and liquid flow rates (data not shown). Hence, the effects of solvent ratios on porous PLA nanofiber morphologies and corresponding diameter distribution were investigated. In the present work, a mixture of solvents with ratios of 7:3, 8:2 and 9:1 was performed. The SEM images showed that when DCM: DMF ratio varied from 7:3 to 9:1, PLA nanofiber porous structures were formed and nanofiber diameters increased stepwise (Fig. 3). Moreover, PLA nanofiber diameters were more uniform in the DCM: DMF 8:2 ratio when compared with 9:1 ratio. Upon organic solvent evaporation, pores were formed on PLA nanofiber surfaces, and owing to evaporation rate differences between DCM and DMF, increased DCM ratios enhanced the pore forming rates. Thus, the DCM: DMF ratios of 8:2 was selected as the optimal parameter to prepare electrospun porous PLA nanofiber membranes and used for subsequent studies. Fig. 3SEM images showing porous PLA nanofibers prepared with different mixed solvent ratios a DCM/DMF = $\frac{7}{3}$, b DCM/DMF = $\frac{8}{2}$ and c DCM/DMF = $\frac{9}{1}$ (Scale bar = 1 μm); Porous PLA nanofiber diameters prepared using different solvent ratios d DCM/DMF = $\frac{7}{3}$, e DCM/DMF = $\frac{8}{2}$ and (f) DCM/DMF = $\frac{9}{1}$ Morphological changes in PLA nanofiber membranes after SCS, PDA, and GS modification were measured using SEM. After SCS modification, relatively smooth surfaces and increased PLA nanofiber diameters were observed, and the porous PLA nanofiber structures had disappeared (Fig. 4a, b). The diameter of PLA nanofiber was increased from 590 ± 120 nm to PLA/SCS (800 ± 300 nm), PLA/SCS/PDA (830 ± 170 nm) and PLA/SCS/PDA-GS (970 ± 230 nm). With self-aggregated PDA on the PLA/SCS nanofiber membrane surface, granular-like morphology was formed (Fig. 4c). Also, PLA/SCS nanofiber roughness and diameters were increased with the addition of GS into dopamine before the self-aggregation of PDA (Fig. 4(d)). PLA hydrophilicity was improved after SCS coating, and the PLA nanofiber membrane contact angle was decreased from 110° to 39° (Fig. 4e). No significant differences were observed between PLA/SCS and PLA/SCS/PDA nanofiber membranes, which indicates that PDA modifications had not altered the hydrophobicity of PLA/SCS nanofiber membranes. However, the contact angle of PLA/SCS/PDA-GS nanofiber membranes was increased due to GS hydrophobicity properties (Fig. 4f). The swell behaviors of PLA, PLA/SCS, PLA/SCS/PDA, and PLA/SCS/PDA-GS nanofiber membranes were examined in PBS (Fig. 4g). All the nanofibers membranes reached equilibrium after 4 h, and the swelling ratios were increased from PLA (ca. $200\%$) to PLA/SCS (ca. $440\%$), PLA/SCS/PDA (ca. $1180\%$), and PLA/SCS/PDA-GS (ca. $1380\%$). GS release profiles from PLA/SCS/PDA-GS nanofiber membranes are shown in Fig. 4h. An initial burst release was observed in the first 12 h, which was followed by a sustained release over 7 days. Thus, the efficient initial GS release from PLA/SCS/PDA-GS nanofiber membranes was beneficial to prevent early-stage wound healing infection. Fig. 4SEM images showing a porous PLA, b PLA/SCS, c PLA/SCS/PDA, and d PLA/SCS/PDA-GS nanofiber membranes, e Contact angle images of PLA, PLA/SCS, PLA/SCS/PDA and PLA/SCS/PDA-GS membranes; f Quantitative contact angle data of PLA, PLA/SCS, PLA/SCS/PDA and PLA/SCS/PDA-GS membranes; g swelling behaviors of PLA, PLA/SCS, PLA/SCS/PDA and PLA/SCS/PDA-GS membranes immersed in PBS solution for 18 h; h GS release profiles from PLA/SCS/PDA-GS membranes in PBS solution. Data are expressed as mean ± SD ($$n = 3$$) ## In vitro cell viability and pro-angiogenic marker gene expression The Live/dead Raw 264.7 cell staining of all nanofiber membranes is shown in Fiugre 5a1-a8. Raw 264.7 cells attached to and spread well on the porous nanofiber membranes, with cells proliferating well on day 3 when compared with day 1. Nanofiber membrane cytotoxicity levels are shown in Fig. 5b. Consistent with live/dead staining data, no obvious cell toxicities were detected on all four nanofiber membranes after 1 and 3 days of culture. Interestingly, after both 1 and 3 days of culture, the PLA/SCS group exhibited higher cell viability when compared with the PLA group, whereas no significant differences were obseverd among PLA, PLA/SCS/PDA and PLA/SCS/PDA-GS groups. Thus, all porous PLA-based nanofibers were suitable for cell growth, and SCS may be beneficial for cell proliferation. Fig. 5Fluorescence images of Raw 264.7 cells cultured on (a1, a5) PLA, (a2, a6) PLA/SCS, (a3, a7) PLA/SCS/PDA and (a4, a8) PLA/SCS/PDA-GS nanofiber membranes after 1 and 3 days of culture, respectively; b cell viabilities on PLA, PLA/SCS, PLA/SCS/PDA and PLA/SCS/PDA-GS nanofiber membranes after 1 and 3 days of culture; c HIF-1α and d VEGF gene expressions in Raw.264.7 cells after cultured in PLA, PLA/SCS, PLA/SCS/PDA and PLA/SCS/PDA-GS nanofiber memebranes for 24 h We previously data showed that SCS alone induced the pro-angiogenic factor (HIF-1α and VEGF) expression in Raw 264.7 cells (Fig. 1c, d). Our in vitro gene expression data demonstrated that the PLA/SCS group expressed higher HIF-1α and VEGF levels when compared with the PLA group after 1 day of culture. Thus, after decorating the porous nanofiber membranes with SCS, the pro-angiogenic function of SCS was maintained. However, both PLA/SCS/PDA and PLA/SCS/PDA-GS groups did not increase HIF-1α and VEGF expression. These might have be due to a PDA layer forming on the PLA/SCS nanofiber membrane surface, and also more time might have been required for SCS to release and induce pro-angiogenic gene expression. Hyperglycemia in diabetes mellitus will cause endothelia dysfunction and damage the activity of VEGF, ultimately lead to impairment of neovascularization [42, 43]. In this work, SCS-containing PLA nanofiber membranes showed ability to increase endogenous VEGF expression and promote angiogenesis in HUVECs, and these will promote diabetic wound healing. ## Macrophage immune responses to porous PLA-based nanofiber membranes To investigate the influence of porous PLA-based nanofiber membranes on the polarization of Raw 264.7 cells, RT-PCR was carried out. Lipopolysaccharide (LPS, 500 ng/mL), a potent activator of the inflammatory response, was used to stimulate macrophage polarization toward the proinflammatory M1 phenotype. After 24 h coculture (Fig. 6), LPS markedly increased pro-inflammatory gene expression (TNF-α, IL-1β and IL-6) (***$p \leq 0.001$), with PLA/SCS, PLA/SCS/PDA, and PLA/SCS/PDA-GS groups showing significantly reduced the TNF-α, IL-1β and IL-6 expression, especially PDA-containing groups. Notably, PLA/SCS and PLA/SCS/PDA groups showed elevated anti-inflammatory gene expression (IL-4, IL-10 and Arg-1). Forty-eight hours after the LPS stimulation (Fig. 7), similar results were observed for pro-inflammatory gene expression (TNF-α, IL-1β and IL-6) in all nanofiber groups. PLA/SCS and PLA/SCS/PDA groups showed increased anti-inflammatory gene expression, whereas GS-containing groups dramatically promoted anti-inflammatory gene expression. Fig. 6The relative gene expression levels of TNF-α, IL-1β, IL-6, IL-4, IL-10, and Arg-1 of Raw 264.7 cells in PLA, PLA/SCS, PLA/SCS/PDA and PLA/SCS/PDA-GS membranes after 24 h of cultureFig. 7The relative gene expression levels of pro-inflammatory genes (TNF-α, IL-1β and IL-6) and anti-inflammatory genes (IL-4, IL-10 and Arg-1) of Raw 264.7 cells in PLA, PLA/SCS, PLA/SCS/PDA and PLA/SCS/PDA-GS membranes after 48 h of culture Long-term inflammatory environment in diabetic wounds impaired macrophage phenotype transition from inflammatory (M1) to anti-inflammatory (M2) status, which will also impede the trans-differentiation of macrophages into fibroblasts. Moreover, M2 macrophages also play a pivotal role in revascularization owing to their ability to release angiogenic factors, such as VEGF and PDGF. Our cell tests data demonstrated that PLA/SCS/PDA-GS nanofiber membranes were able to regulate the macrophage inflammatory responses, forming new vessels, thus promoting diabetic wound healing. ## Antibacterial activity of porous PLA-based nanofiber membranes Bacterial infection is a critical barrier to diabetic wound healing; therefore, antimicrobial activity is essential in diabetic wound dressings to prevent bacterial infection. Here, the S. aureus was used to investigate the in vitro antibacterial properties of the PLA-based nanofiber membrane. The ZOI assays showed no antibactericidal effects against S. aureus in PLA, PLA/SCS, and PLA/SCS/PDA groups, but for the PLA/SCS/PDA-GS nanofiber membranes, an 8 mm ZOI against S. aureus was observed (Fig. 8a, b). Additionally, the quantitative bactericidal effects from membranes were using through shake-flask culture method (Fig. 8c, d). Consistent with ZOI data, S. aureus had high survival rates when incubated with PLA, PLA/SCS, and PLA/SCS/PDA nanofiber membranes, but few S. aureus were observed in the PLA/SCS/PDA-GS treated group after co-cultivation for 15 min. It is worth noting that dopamine deposition had no obvious effects on S. aureus killing. Live/dead staining assays showed that bacteria grew well in PLA, PLA/SCS, and PLA/SCS/PDA groups, whereas GS-loaded nanofiber membranes (PLA/SCS/PDA-GS) showed efficient bacteria killing functions (99.23 ± $0.6\%$) (Fig. 8e1-e12). All the above results indicated that PLA, PLA/SCS and PLA/SCS/PDA nanofiber membranes had no significant anti-bactericidal abilities against S. aureus, whereas GS loading onto nanofiber membranes contributed to effective antibacterial activity. Fig. 8a Bacteriostatic circles and b radius of bacteriostatic circle of S. aureus after treatment with PLA, PLA/SCS, PLA/SCS/PDA and PLA/SCS/PDA-GS nanofiber membranes; c Photographs and d relative bacterial survival rates of S. aureus after coculture with PLA, PLA/SCS, PLA/SCS/PDA and PLA/SCS/PDA-GS nanofiber membranes for 0 min and 15 min respectively; Live/dead staining of S. aureus incubated with (e1, e5, e9) PLA, (e2, e6, e10) PLA/SCS, (e3, e7, e11) PLA/SCS/PDA and (e4, e8, e12) PLA/SCS/PDA-GS nanofiber membranes for 15 min (SYTO9 and PI stained live bacteria and dead bacteria, respectively) ## Conclusions In summary, in the present work, a multifunctional nanofiber membrane was developed via electrospinning combined with surface modification methods. Importantly, the coating of SCS, PDA and GS changed the porous structure of PLA nanofiber, increased the PLA fiber diameter and improved the hydrophilicity property of PLA nanofiber membrane. Additionally, in vitro data confirmed anti-inflammatory properties of PLA/SCS/PDA-GS nanofiber membranes toward macrophages (Raw 264.7 cells) and increased endogenous VEGF secretion to induce vascularization. Moreover, our in vitro antibacterial data demonstrated the efficient antibacterial ability of the GS-loaded nanofiber membrane toward S. aureus. Therefore, our novel nanofiber membranes with antibacterial, anti-inflammatory, and angiogenic properties have promising potential in diabetic wound regeneration applications. ## References 1. Li WC, Jiang LB, Wu SJ, Yang SW, Ren L, Cheng B, Xia J. **A shape-programmable hierarchical fibrous membrane composite system to promote wound healing in diabetic patients**. *Small* (2022) **18** e2107544. DOI: 10.1002/smll.202107544 2. Liu P, Xiong Y, Chen L, Lin CC, Yang YY, Lin Z, Yu YS, Mi BB, Liu GH, Xiao XF, Feng Q. **Angiogenesis-based diabetic skin reconstruction through multifunctional hydrogel with sustained releasing of M2 Macrophage-derived exosome**. *Chem Eng J* (2022) **431** 132413. DOI: 10.1016/j.cej.2021.132413 3. Li G, Wang QX, Feng JW, Wang JL, Wang YQ, Huang XT, Shao TT, Deng XF, Cao YM, Zhou MM, Zhao C. **Recent insights into the role of defensins in diabetic wound healing**. *Biomed Pharmacother* (2022) **155** 113694. DOI: 10.1016/j.biopha.2022.113694 4. Augustine R, Rehman SRU, Ahmed R, Zahid AA, Sharifi M, Falahati M, Hasan A. **Electrospun chitosan membranes containing bioactive and therapeutic agents for enhanced wound healing**. *Int J Biol Macromol* (2020) **156** 153-170. DOI: 10.1016/j.ijbiomac.2020.03.207 5. Dong RN, Guo BL. **Smart wound dressings for wound healing**. *Nano Today* (2021) **41** 101290. DOI: 10.1016/j.nantod.2021.101290 6. Tu ZL, Chen M, Wang M, Shao ZX, Jiang XQ, Wang KY, Yao Z, Yang SW, Zhang XX, Gao WY. **Engineering bioactive M2 macrophage-polarized anti-inflammatory, antioxidant, and antibacterial scaffolds for rapid angiogenesis and diabetic wound repair**. *Adv Funct Mater* (2021) **31** 2100924. DOI: 10.1002/adfm.202100924 7. Wang W, Wang CG, Chen M, Xi YW, Cheng W, Mao C, Xu TZ, Zhang XX, Lin C, Gao WY, Guo Y, Lei B. **Efficient angiogenesis-based diabetic wound healing/skin reconstruction through bioactive antibacterial adhesive ultra-violet shielding nanodressing with exosome release**. *ACS Nano* (2019) **13** 10279-10293. DOI: 10.1021/acsnano.9b03656 8. Li YZ, Yang L, Hou Y, Zhang ZZ, Chen M, Wang MX, Liu J, Wang J, Zhao ZH, Xie CM, Lu X. **Polydopamine-mediated graphene oxide and nanohydroxyapatite-incorporated conductive scaffold with an immunomodulatory ability accelerates periodontal bone regeneration in diabetes**. *Bioact Mater* (2022) **18** 213-227. DOI: 10.1016/j.bioactmat.2022.03.021 9. Xu Z, Deng B, Wang XW, Yu J, Xu ZB, Liu PG, Liu CH, Cai Y, Wang F, Zong RL, Chen ZL, Xing H, Chen G. **Nanofiber-mediated sequential photothermal antibacteria and macrophage polarization for healing MRSA-infected diabetic wounds**. *J Nanobiotechnol* (2021) **19** 404. DOI: 10.1186/s12951-021-01152-4 10. Yuan Y, Fan DD, Shen SH, Ma XX. **An M2 macrophage-polarized anti-inflammatory hydrogel combined with mild heat stimulation for regulating chronic inflammation and impaired angiogenesis of diabetic wounds**. *Chem Eng J* (2022) **433** 133589. DOI: 10.1016/j.cej.2021.133859 11. Zheng DT, Huang CX, Zhu XH, Huang HH, Xu CL. **Performance of polydopamine complex and mechanisms in wound healing**. *Int J Mol Sci* (2021) **22** 10563. DOI: 10.3390/ijms221910563 12. Li T, Ma HS, Ma HZ, Ma ZJ, Qiang L, Yang ZZ, Yang XX, Zhou XJ, Dai KR, Wang JW. **Mussel-inspired nanostructures potentiate the immunomodulatory properties and angiogenesis of mesenchymal stem cells**. *ACS Appl Mater Inter* (2019) **11** 17134-17146. DOI: 10.1021/acsami.8b22017 13. Bao XF, Zhao JH, Sun J, Hu M, Yang XR. **Polydopamine nanoparticles as efficient scavengers for reactive oxygen species in periodontal disease**. *ACS Nano* (2018) **12** 8882-8892. DOI: 10.1021/acsnano.8b04022 14. Ma WJ, Zhang XX, Liu YX, Fan L, Gan JJ, Liu WL, Zhao YJ, Sun LY. **Polydopamine decorated microneedles with Fe-MSC-derived nanovesicles encapsulation for wound healing**. *Adv Sci* (2022) **9** e2103317. DOI: 10.1002/advs.202103317 15. Zhang H, Wu L, Wang GC, Liu PZ, Qin S, Xu KH, Tong D, Ding H, Tang H, Ji F. **Macrophage polarization, inflammatory signaling, and NF-kappaB activation in response to chemically modified titanium surfaces**. *Colloids Surf B Biointerfaces* (2018) **166** 269-276. DOI: 10.1016/j.colsurfb.2018.03.029 16. Guan Y, Niu H, Liu ZT, Dang Y, Shen J, Zayed M, Ma L, Guan JJ. **Sustained oxygenation accelerates diabetic wound healing by promoting epithelialization and angiogenesis and decreasing inflammation**. *Sci Adv* (2021) **7** ahj0153. DOI: 10.1126/sciadv.abj0153 17. Spiller KL, Anfang RR, Spiller KJ, Ng J, Nakazawa KR, Daulton JW, Vunjak-Noyakavic G. **The role of macrophage phenotype in vascularization of tissue engineering scaffolds**. *Biomaterials* (2014) **35** 4477-4488. DOI: 10.1016/j.biomaterials.2014.02.012 18. Wang YQ, Yao DY, Li LH, Qian ZY, He W, Ding R, Liu HF, Fan YB. **Effect of electrospun silk fibroin-silk sericin films on macrophage polarization and vascularization**. *ACS Biomate Sci Eng* (2020) **6** 3502-3512. DOI: 10.1021/acsbiomaterials.0c00175 19. Yu YM, Chen R, Sun Y, Pan YZ, Tang W, Zhang S, Cao LY, Yuan Y, Wang J, Liu CS. **Manipulation of VEGF-induced angiogenesis by 2-N, 6-O-sulfated chitosan**. *Acta Biomater* (2018) **71** 510-521. DOI: 10.1016/j.actbio.2018.02.031 20. Yu YM, Dai K, Gao ZH, Tang W, Shen T, Yuan Y, Wang J, Liu CS. **Sulfated polysaccharide directs therapeutic angiogenesis via endogenous VEGF secretion of macrophages**. *Sci Adv* (2021) **7** 8217. DOI: 10.1126/sciadv.abd8217 21. Shen T, Dai K, Yu YM, Wang J, Liu CS. **Sulfated chitosan rescues dysfunctional macrophages and accelerates wound healing in diabetic mice**. *Acta Biomater* (2020) **117** 192-203. DOI: 10.1016/j.actbio.2020.09.035 22. Chen H, Cheng RY, Zhao X, Zhang YH, Tam A, Yan YF, Shen HK, Zhang YS, Qi J, Feng YH, Liu L, Pan GQ, Cui WG, Deng LF. **An injectable self-healing coordinative hydrogel with antibacterial and angiogenic properties for diabetic skin wound repair**. *NPG Asia Mater* (2019) **11** 3. DOI: 10.1038/s41427-018-0103-9 23. Chen J, Liu YJ, Cheng GP, Guo JH, Du S, Qiu JM, Wang C, Li CC, Yang XF, Chen TK, Chen ZB. **Tailored hydrogel delivering niobium carbide boosts ROS-scavenging and antimicrobial activities for diabetic wound healing**. *Small* (2022) **18** 2201300. DOI: 10.1002/smll.202201300 24. Shi MQ, Du Z, Qi YC, Li WL, Hu HQ, Lin XH, Wang SJ, Tang Z, Zhou M. **Wound microenvironment-responsive glucose consumption and hydrogen peroxide generation synergistic with azithromycin for diabetic wounds healing**. *Theranostics* (2022) **12** 2658-2673. DOI: 10.7150/thno.64244 25. Anand S, Rajinikanth PS, Arya DK, Pandey P, Gupta RK, Sankhwar R, Chidambarm K. **Multifunctional biomimetic nanofibrous scaffold loaded with asiaticoside for rapid diabetic wound healing**. *Pharmaceutics* (2022) **14** 273. DOI: 10.3390/pharmaceutics14020273 26. Xi YW, Ge J, Wang M, Chen M, Niu W, Cheng W, Xue YM, Lin C, Lei B. **Bioactive antiinflammatory antibacterial antioxidative silicon-based nanofibrous dressing enables cutaneous tumor photothermo-chemo therapy and infection-induced wound healing**. *ACS Nano* (2020) **14** 2904-2916. DOI: 10.1021/acsnano.9b07173 27. Qian YN, Zheng YJ, Jin J, Wu X, Xu KJ, Dai ML, Niu Q, Zheng H, He XJ, Shen JL. **Immunoregulation in diabetic wound repair with a photoenhanced glycyrrhizic acid hydrogel scaffold**. *Adv Mater* (2022) **34** 2200521. DOI: 10.1002/adma.202200521 28. Asadi N, Mehdipour A, Ghorbani M, Mesgari-Abbasi M, Akbarzadeh A, Davaran S. **A novel multifunctional bilayer scaffold based on chitosan nanofiber/alginate methacrylate hydrogel for full-thickness wound healing**. *Int J Biol Macromol* (2021) **193** 734-747. DOI: 10.1016/j.ijbiomac.2021.10.180 29. Golipour H, Ezzatzadeh E, Sadeghianmaryan A. **Investigation of co-electrospun gelatin: TiO2/polycaprolactone:silk fibroin scaffolds for wound healing applications**. *J Appl Polym Sci* (2022) **139** e52505. DOI: 10.1002/app.52505 30. Li T, Sun MC, Wu SH. **State-of-the-art review of electrospun gelatin-based nanofiber dressings for wound healing applications**. *Nanomaterials-basel* (2022) **12** 784. DOI: 10.3390/nano12050784 31. Rad ZP, Mokhtari J, Abbasi M. **Biopolymer based three-dimensional biomimetic micro/nanofibers scaffolds with porous structures via tailored charge repulsions for skin tissue regeneration**. *Polym Adv Technol* (2021) **32** 3535-3548. DOI: 10.1002/pat.5364 32. Yin J, Xu L, Ahmed A. **Batch preparation and characterization of electrospun porous polylactic acid-based nanofiber membranes for antibacterial wound dressing**. *Adv Fiber Mater* (2022) **4** 832-844. DOI: 10.1007/s42765-022-00141-y 33. Chandika P, Oh GW, Heo SY, Kim SC, Kim TH, Kim MS, Jung WK. **Electrospun porous bilayer nano-fibrous fish collagen/PCL bio-composite scaffolds with covalently cross-linked chitooligosaccharides for full-thickness wound-healing applications**. *Mater Sci Eng C* (2021) **121** 111871. DOI: 10.1016/j.msec.2021.111871 34. Yao QQ, Cosme JGL, Xu T, Miszuk JM, Picciani PHS, Fong H, Sun HL. **Three dimensional electrospun PCL/PLA blend nanofibrous scaffolds with significantly improved stem cells osteogenic differentiation and cranial bone formation**. *Biomaterials* (2017) **115** 115-127. DOI: 10.1016/j.biomaterials.2016.11.018 35. Yao QQ, Liu Y, Pan YN, Li YJ, Xu LM, Zhong YM, Wang W, Zuo JY, Yu H, Lv ZR, Chen H, Zhang LS, Wang BL, Yong HY, Meng YC. **Long-term induction of endogenous BMPs growth factor from antibacterial dual network hydrogels for fast large bone defect repair**. *J Colloid Interf Sci* (2022) **607** 1500-1515. DOI: 10.1016/j.jcis.2021.09.089 36. Huang JH, Han QX, Cai M, Zhu J, Li L, Yu LF, Wang Z, Fan GT, Zhu Y, Lu JW, Zhou GX. **Effect of angiogenesis in bone tissue engineering**. *Ann Biomed Eng* (2022) **50** 898-913. DOI: 10.1007/s10439-022-02970-9 37. Recoraro AR, Hosfield BD, Li HG, Shelley WC, Markel TA. **Angiogenesis: a cellular response to traumatic injury**. *Shock* (2021) **55** 301-310. DOI: 10.1097/SHK.0000000000001643 38. Hu C, Long LY, Cao J, Zhang SM, Wang YB. **Dual-crosslinked mussel-inspired smart hydrogels with enhanced antibacterial and angiogenic properties for chronic infected diabetic wound treatment via pH-responsive quick cargo release**. *Chem Eng J* (2021) **411** 128564. DOI: 10.1016/j.cej.2021.128564 39. Zimna A, Kurpisz M. **Hypoxia-inducible factor-1 in physiological and pathophysiological angiogenesis: applications and therapies**. *Biomed Res Int* (2015) **2015** 549412. DOI: 10.1155/2015/549412 40. Zhao Y, Zhang M, Lu GL, Huang BX, Wang DW, Shao Y, Lu MJ. **Hypoxic preconditioning enhances cellular viability and pro-angiogenic paracrine activity: the roles of VEGF-A and SDF-1α in rat adipose stem cells**. *Front Cell Dev Biol* (2020) **8** 580131. DOI: 10.3389/fcell.2020.580131 41. Long G, Liu DY, He X, Shen YY, Zhao YN, Hou XL, Chen B, OuYang W, Dai JW, Li X. **A dual functional collagen scaffold coordinates angiogenesis and inflammation for diabetic wound healing**. *Biomater Sci* (2020) **8** 6337-6349. DOI: 10.1039/D0BM00999G 42. Kolluru GK, Bir SC, Kevil CG. **Endothelial dysfunction and diabetes: effects on angiogenesis, vascular remodeling, and wound healing**. *Int J Vasc Med* (2012) **2012** 918267. PMID: 22611498 43. Maruhashi T, Higashi Y. **Pathophysiological association between diabetes mellitus and endothelia dysfunction**. *Antioxidants* (2021) **10** 1306. DOI: 10.3390/antiox10081306
--- title: A low BUN/creatinine ratio predicts histologically confirmed acute interstitial nephritis authors: - López Giacoman Salvador - González Fuentes Carolina - Robles Dávila Jesús - Soto Abraham María Virgilia - Román Acosta Susana - Chávez Íñiguez Jonathan - Salas Pacheco José Luis - Ronco Claudio journal: BMC Nephrology year: 2023 pmcid: PMC10041724 doi: 10.1186/s12882-023-03118-0 license: CC BY 4.0 --- # A low BUN/creatinine ratio predicts histologically confirmed acute interstitial nephritis ## Abstract ### Introduction In hospitalized patients with acute renal injury (AKI), acute tubulointerstitial nephritis (AIN) constitutes one of the leading etiologies. The objective of this study was to identify clinical and biochemical variables in patients with AKI associated with kidney biopsy-confirmed AIN. ### Methods For our prospective study, we recruited hospitalized patients aged 18 years and older who were diagnosed with AKI based on biochemical criteria. Prior to enrollment, each patient was assessed with a complete metabolic panel and a kidney biopsy. ### Results The study consisted of 42 patients (with a mean age of 45 years) and equal numbers of male and female patients. Diabetes and hypertension were the main comorbidities. Nineteen patients had histological findings consistent with AIN. There was a correlation between histology and the BUN/creatinine ratio (BCR) (r = -0.57, $$p \leq 0.001$$). The optimal Youden point for classifying AIN via a receiver operating characteristic (ROC) curve analysis was a BCR ≤ 12 (AUC = 0.73, $$p \leq 0.024$$). Additionally, in diagnosing AIN, BCR had a sensitivity of $76\%$, a specificity of $81\%$, a positive predictive value of $81\%$, a negative predictive value of $76\%$, and OR of 14 ($95\%$ CI = 2.6 to 75.7, $$p \leq 0.021$$). In the multivariable analysis, BCR was the sole variable associated with AIN. ### Conclusion A BCR ≤ 12 identifies AIN in patients with AKI. This study is the first to prospectively assess the relationship between renal biopsy results and BCR. ## Introduction Acute interstitial nephritis (AIN) accounts for 13–$27\%$ of acute kidney injury (AKI) cases in hospitalized patients and is diagnosed in $13\%$ of kidney biopsies performed for AKI [1–3]. AIN histopathology is characterized by interstitial inflammation and tubulitis. Interstitial infiltrates predominantly contain lymphocytes and monocytes, but plasma cells, neutrophils, and histiocytes may also be present [1, 2]. AIN has become increasingly relevant in acute and critical care settings over the last few years [2]. Additionally, AIN could be a potential origin of chronic kidney disease from unidentified etiology [4]. The classic triad described for AIN presentation includes fever, dermatosis, and eosinophilia. Eosinophilia is only present in $10\%$ of patients [5], thus resulting in poor diagnostic performance [6, 7]. Previous studies have evaluated novel biomarkers of AIN by analyzing markers of inflammation, interstitial edema, cellular damage, and tubular lesions; however, their clinical utility remains unknown [8]. Despite the many examined biomarkers, the gold standard continues to be percutaneous renal biopsy [9, 10]. It is widely accepted that the blood-ureic nitrogen to creatinine ratio (BCR) decreases in renal tubular lesions, but prospectively obtained clinical-histological evidence relating low BCR to AIN is lacking [6, 11]. Therefore, this study aimed to determine classic clinical and biochemical predictors (specifically BCR) associated with histopathologically-confirmed AIN in patients with AKI. ## Methods From June 2018 to June 2019, we prospectively recruited hospitalized patients of both sexes aged 18 years and older and diagnosed them with AKI according to the creatinine criteria established by the Kidney Disease: Improving Global Outcomes guidelines [12]. The study exclusion criteria included pregnancy, acute coronary syndrome, advanced liver disease, rhabdomyolysis, previous glucocorticoid treatment, and sepsis. The comprehensive clinical evaluation recorded prescribed and over-the-counter medications, as well as the consumption of herbal medication. An exhaustive physical examination evaluated cutaneous rash and other signs of systemic diseases. Peripheral blood samples were obtained for a complete blood count and a basic metabolic panel. Moreover, a biochemical urine analysis and spot proteinuria analysis were also performed. This study was approved by the Institutional Review Board and was conducted in adherence to the Helsinki Declaration. The protocol followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. Informed consent was obtained from all of the patients. For determining the variables, the nephrotic syndrome was defined as proteinuria > 3.5 g per day, serum albumin < 2.5 g/dL, and clinical evidence of peripheral edema and hyperlipidemia [13]. The nephritic syndrome was defined as oliguria, hematuria with red blood cell casts, subnephrotic proteinuria, and hypertension [14]. The degrees of interstitial fibrosis and arteriosclerosis were reported in accordance with the standardized grading proposed by Sethi et al [15]. ## Percutaneous kidney biopsy An experienced nephrologist performed all of the kidney biopsies. Prior to the procedure, all the patients were normotensive, and kidney size and anatomy were evaluated via ultrasound. Following the localization of the left kidney inferior pole, the puncture site was marked, the skin was prepared with chlorhexidine, and $1\%$ lidocaine was used for local anesthesia. The biopsy procedure was guided by real-time ultrasound and was performed by using an automatic biopsy gun (Magnum, Bard Medical) with a disposable, 18-gauge core needle. Two kidney core biopsies were obtained from each patient. After the procedure, the patients were closely monitored for 24 h. ## Tissue processing and analysis All of the tissues were uniformly processed in the same nephropathology laboratory and analyzed by a single nephropathologist. Of the two core biopsies that were obtained per patient, one specimen was fixed in $10\%$ formaldehyde, embedded in paraffin, and stained with hematoxylin and eosin, PAS, Masson’s trichrome, and Jone’s methenamine silver. The second sample per patient was placed in *Zeus media* tissue fixative, after which it was washed, rehydrated, and frozen at –24 °C for cryosectioning and direct immunofluorescence staining. Sections were stained with antibodies against IgG, IgA, IgM, Ciq, C3c, C4c, fibrinogen, albumin, kappa, and lambda. Histologically, AIN was defined as interstitial edema and interstitial infiltrate consisting primarily of mononuclear or polymorphonuclear leukocytes. Tubulitis was defined as the invasion of the tubular basement membrane by inflammatory cells [13]. ## Statistical analysis In accordance with their respective distributions, continuous variables were expressed by the mean ± standard deviation or the median with interquartile range. The normality of the distribution was evaluated by using the Kolmogorov‒Smirnov test. The chi-square test was performed to evaluate the association between the qualitative variables. Differences between groups were assessed with the Student’s t-test or the Mann‒Whitney U test (according to the distribution). A multivariate analysis was performed with binary logistic regression. The variables with a p-value < 0.1 in the bivariate analysis were selected for inclusion as the explanatory variables within the multivariable model. The dependent variable was dichotomous and expressed as the presence or absence of histologically confirmed AIN. The independent variables that were included in the model were BUN, serum creatinine, BCR, and blood glucose concentration. The optimal BCR value for discriminating AIN was identified by using a ROC analysis. The Youden point was selected to maximize the sensitivity over the specificity. The bilateral p-value for statistical significance was established at < 0.05. The statistical analysis was performed by using R (version 4.04) in the R-Studio. ## Results Five hundred thirty patients diagnosed with AKI were assessed, of whom 42 ($7\%$) patients satisfied the inclusion criteria and underwent a percutaneous renal biopsy (Fig. 1). Table 1 describes the demographic and biochemical characteristics of the patient cohort. The mean age was 45 years, and half of the patients were female. The main comorbidities were hypertension and diabetes ($33\%$ and $17\%$, respectively). In addition, the biopsy indications were unexplained AKI ($66\%$), nephrotic syndrome ($21\%$), and nephritic syndrome ($12\%$). There were no complications from the biopsy procedures. The glomerular disease was reported in twenty-three patients ($54\%$), the primary reported glomerular disease was focal and segmental glomerulosclerosis (FSGS) in $16.6\%$ of patients, followed by membranous nephropathy at $11.9\%$, Ig A nephropathy at $9.5\%$, and amyloidosis at $9.5\%$. Among the study participants, AIN was observed in 19 ($45\%$) patients. Nodular arteriolopathy was reported in $28\%$ of patients, followed by arteriosclerosis in $17\%$ of the patients. Half of the samples had interstitial fibrosis, $47\%$, of which presented with third-degree fibrosis. Fig. 1Flowchart showing the study patient distribution. Forty-two individuals required percutaneous kidney biopsy. In nearly half of AIN patients, the etiology was acute kidney injury. In patients with AIN, the BCR was significantly lower. AIN = acute interstitial nephritis, AKI = acute kidney injury, BCR = BUN/creatinine ratio, FSGS = focal and segmentary glomerulosclerosisTable 1Demographic characteristicsVariableAll $$n = 42$$AINNo AINpn = 19n = 23Gender male2111100.3 female21813 Age (years)44 (± 19)45 (25–53)40 (29–53)0.3 Hypertension14 ($33\%$)7 ($37\%$)7 ($30\%$)0.6 Diabetes7 ($16\%$)4 ($21\%$)3 ($13\%$)0.48 BCR15.8 (± 9.3)11.2 (± 3.8)21.4 (± 8.7)0.023 Creatinine (mg/dl)5.05 (± 5.2)6.6 (3–9.7)3.5 (0.71–4)0.007 BUN (mg/dl)51.3 (± 36.1)62 (38–88)40 (13–68)0.037 Glucose (mg/dl)119 (± 102)99 (77–116)137 (131–140)0.02 Hemoglobin (g/dl)12.09 (± 3.16)11.6 (± 3.1)12.5 (± 3.2)0.78 WBC (cells × 103)7.4 (± 2.75)8.0 (± 2.9)6.8 (± 2.5)0.37 Sodium (mmol/l)135 (± 9.9)132 (± 11.5)137 (± 7.8)0.34 Potassium (mmol/l)4.4 (± 0.66)4.4 (± 0.78)4.5 (± 0.5)0.6Phosphorus (mg/dl)5.4 (± 2.55)6 (3–8.2)4.6 (3.4–4.3)0.14 Chlorine (mmol/l)100 (± 10.6)100 (± 13)100 (± 7.5)0.28 Magnesium (mg/dl)2.1 (± 0.39)2.1 (± 0.4)2.1 (± 0.3)0.36 BCR BUN to creatinine ratio, BUN Blood ureic nitrogen, WBC White blood cellsTable 1 Demographic characteristics and biochemical variables at clinical diagnosis. Patient groups are divided according to the presence of AIN. Quantitative data are presented as the mean, maxima, and minima, whereas qualitative data contain a mean and percentage. The results were considered to be statistically significant when $p \leq 0.05$ The antecedent of recent drug consumption was present in $80\%$ of patients with AIN and $60\%$ in the group with no AIN. The most frequently used drugs were antihypertensives, such as losartan and telmisartan, as well as nonsteroidal anti-inflammatories (such as diclofenac and ketorolac) and analgesics, including tramadol. Concomitantly, two patients with AIN consumed herbal drugs for gastrointestinal disorders. Table 2 shows the complete details of the drugs that were ingested. Table 2Patient-reported drug therapy and herbalismDrugsHerbalismAntihypertensivesLosartan, telmisartan, furosemide,nifedipine, proponolol, valsartan,amlodipine, chlortalidone, andprazosineEysenhardtia polystachyaRuta graveolesnAntimicrobials/Antivirals Ciprofloxacin, metronidazole, and tenofovirNSAIDs/Analgesics Paracetamol, diclofenac, ketorolac, Pregabaline, and tramadolAntidiabeticInsulin, metformin, and glibenclamideAntiacidsPantoprazole, omeprazole, and ranitidineOthersMetoclopramide, tretinoine, allopurinol, and levotiroxine The bivariate analysis showed that patients with AIN had greater creatinine (6.6 mg/dL vs. 3.5 mg/dL) and BUN (62 mg/dL vs. 40 mg/dL) concentrations. The BCR was significantly lower in the AIN group (11.2 ± 3.8 vs. 21.4 ± 8.7, $$p \leq 0.001$$) (Fig. 2). The remaining clinical and biochemical variables were equally distributed between the groups, except for blood glucose concentrations, which were higher in the group with no AIN. The results of the multivariate analysis, BCR had an inverse correlation with the histological finding of AIN (Spearman rho = -0.57, $$p \leq 0.001$$). Table 3 shows the results of the multivariate analysis, with the variables suspected of being risk factors for AIN. Furthermore, the ROC analysis (Fig. 3) showed that the optimal Youden point for identifying AIN was at a BCR lower than 12 (sensitivity = $81\%$, specificity = $83\%$, and AUC = 0.73; $$p \leq 0.024$$).Fig. 2Violin plot showing the distribution of the BUN/creatinine ratio (BCR) in patients with acute kidney injury. In the AIN population, the median BCR is grouped at approximately 12Table 3Results of logistic regression analysisEstimateStd ErrorT valuepIntercept1.2029610.1894536.3500.001BCR0.0255300.0084933.0060.004BUN-0.0073250.004038-1.8140.07Urea0.0047940.0026191.8310.07Creatinine-0.0522930.025448-2.0550.06Glucose0.00020.0041.20.4Fig. 3The ROC plot shows the optimal Youden point discriminating AIN from non-AIN patients with AKI as having a BCR value of 12. BCR = BUN to creatinine ratio The coexistence of AIN and acute tubular injury was observed in 12 patients, with a mean BCR of 12.1 ± 3. Nine patients had isolated acute tubular injury (without AIN), with a mean BCR of 17.5 ± 7. This difference was significant ($$p \leq 0.001$$). However, we did not identify an association between AIN and acute tubular injury ($$p \leq 0.2$$). In patients with glomerular disease, low BCR was observed only $13\%$; whereas in that with AIN, it was found in $84\%$ (X2 = 21, $$p \leq 0.001$$). BCR < 12 had good diagnostic performance for AIN (sensitivity = $76\%$, specificity = $81\%$, positive predictive value = $81\%$, and negative predictive value = $76\%$). Furthermore, BCR < 12 increased the probability of observing AIN in kidney histology (OR = 14.1, $95\%$ CI = 2.6–75, $$p \leq 0.0021$$). In the multivariable analysis, BCR < 12 was the sole biomarker associated with histopathology-confirmed AIN. Individually, serum creatinine and BUN were not associated with AIN. ## Discussion Our retrospective study demonstrated that low BCR (BCR < 12) adequately predicts the diagnosis of histological AIN in patients with AKI. The diagnostic utility of low BCR was previously described in obstructive uropathy, in which low BCR correlated with elevated urea in hospitalized patients [1, 16]. Additionally, BCR < 10 identified AIN in Mesoamerican nephropathy [4]. To our knowledge, there has been no other report on the diagnostic performance of BCR in a prospective study involving patients with AKI and AIN. Creatinine is a product of muscle metabolism. Creatine is broken down into creatinine by a nonenzymatic mechanism. It is filtered in the glomerulus and secreted in the proximal tubule. At steady-state, the daily urinary creatinine that is excreted is equal to that produced, and this quantity is directly related to muscle mass [17]. In contrast, urea is excreted at the same rate as creatinine in the glomerulus but is reabsorbed into the tubules [18–20]. In the hemodynamic etiologies of AKI (prerenal), the activation by antidiuretic hormone causes preferential urea reabsorption, thus resulting in a BCR > 20 [20]. AIN is characterized by tubular dysfunction and an acute decrease in glomerular filtration rate. Edema, interstitial inflammation, and tubulitis are the main histopathological features of AIN [21]. In AIN, the abnormal reabsorption of water and sodium leads to increased urea excretion and decreased reabsorption [22]. The low BCR in patients with AIN can be explained by the combinatorial effects of the changes in urea excretion, filtration, and reabsorption with the reduction in creatinine secretion as the glomerular filtration rate declines. The use of biomarkers in the diagnosis of AIN has been controversial in the literature. Novel biomarkers have focused on subclinical phases or prognostic factors in AKI. Neutrophil gelatinase-associated lipocalin (NGAL) is one of the most studied biomarkers. NGAL expression is related to stress or damage to the loop of Henle and the collecting duct; values greater than 104 ng/ml correlate with kidney damage [23, 24]. Another well-studied, novel biomarker is kidney injury molecule 1 (KIM-1), which is a transmembrane glycoprotein that is upregulated in response to ischemia‒reperfusion kidney injury [25]. Calprotectin is a heterodimer that is synthesized in epithelial cells in the collecting duct in response to damage or inflammation. Calprotectin is proposed to be able to distinguish intrinsic renal injury from prerenal etiology; however, its secretion is known to be increased in pyuria and some systemic inflammatory diseases (such as rheumatoid arthritis and inflammatory bowel disease) [24, 26] thus limiting the specificity of this novel biomarker. As a novel biomarker of AIN, BCR has multiple advantages, including low cost, ease of measurement, and the ability for longitudinal tracking of AIN. Creatinine and BUN are routinely measured in basic biochemical panels, and the establishment of a relationship between BCR and AIN would facilitate the diagnosis and early treatment of these patients prior to the analysis of renal histology. A kidney biopsy is not systematically performed in patients with AIN. Drug-induced AIN can be diagnosed in patients who have a recent history of drug initiation, followed by subsequent creatinine elevation, urinalysis with white cells, white cell casts, eosinophiluria (in some cases), and symptom improvement after the cessation of the offending drug [27, 28]. However, histological confirmation is necessary for patients receiving drugs that are not known to precipitate AIN, with these patients lacking improvements in response to glucocorticoid treatment and having the absence of white cells in urinalysis. In addition, low BCR could help in the differential diagnosis of patients who are not improving with glucocorticoid treatment before the performance of a kidney biopsy. In this study, the BCR was calculated upon hospital admission of the AKI patient. The BCR can also vary over the course of the disease. Of note, BUN and creatinine can increase due to other etiologies, and it may be difficult to differentiate etiologies in patients with multiple factors contributing to AKI. We acknowledge that this study was limited by a modest sample size and the study being performed at a single academic center. The study was also limited to hospitalized patients who consented to their participation, and the implications for outpatient AKI require further validation. Future experiments could involve further validation of the utility of the BCR in larger patient cohorts. These limitations could be overcome in future research by involving more recruiting centers and by increasing the etiologies of AIN. The first etiology of AIN was drug-induced; however, this information was obtained from anecdotal accounts from patients. This implied bias can be overcome by using databases with drugs prescribed to patients. Our prospective study is the first of its kind to indicate a statistically significant correlation between BCR and histopathologically confirmed diagnosis of AIN. In addition, this prospective study was specifically designed to identify BCR as a predictor of AIN, which decreases the risk of bias. ## Conclusion In hospitalized patients with AKI, the presence of BCR ≤ 12 is a robust parameter that suggests the diagnosis of AIN. The BCR is obtained from a basic metabolic panel, and its low cost allows for longitudinal quantification. Additionally, if urinalysis and clinical course are atypical for AIN, a low BCR could provide additional support and allow for the avoidance of kidney biopsy in some patients. ## References 1. Perazella MA. **Clinical Approach to Diagnosing Acute and Chronic Tubulointerstitial Disease**. *Adv Chronic Kidney Dis* (2017.0) **24** 57-63. DOI: 10.1053/j.ackd.2016.08.003 2. Goicoechea M, Rivera F, López-Gómez JM. **Spanish Registry of Glomerulonephritis. Increased prevalence of acute tubulointerstitial nephritis**. *Nephrol Dial Transplant* (2013.0) **28** 112-5. DOI: 10.1093/ndt/gfs143 3. Haas M, Spargo BH, Wit EJ, Meehan SM. **Etiologies and outcome of acute renal insufficiency in older adults: a renal biopsy study of 259 cases**. *Am J Kidney Dis* (2000.0) **35** 433. DOI: 10.1016/S0272-6386(00)70196-X 4. Fischer RSB, Vangala C, Truong L, Mandayam S, Chavarria D, Granera Llanes OM. **Early detection of acute tubulointerstitial nephritis in the genesis of Mesoamerican nephropathy**. *Kidney Int* (2018.0) **93** 681-690. DOI: 10.1016/j.kint.2017.09.012 5. Praga M, González E. **Acute interstitial nephritis**. *Kidney Int* (2010.0) **77** 956. DOI: 10.1038/ki.2010.89 6. Corwin HL, Korbet SM, Schwartz MM. **Clinical correlates of eosinophiluria**. *Arch Intern Med* (1985.0) **145** 1097. DOI: 10.1001/archinte.1985.00360060165025 7. Muriithi AK, Nasr SH, Leung N. **Utility of urine eosinophils in the diagnosis of acute interstitial nephritis.**. *lin J Am Soc Nephrol* (2013.0) **8** 1857. DOI: 10.2215/CJN.01330213 8. Martinez Valenzuela L, Draibe J, Fulladosa X, Torras J. **New biomarkers in acute tubulointerstitial nephritis: a novel approach to a classic condition**. *Int J Mol Sci* (2020.0) **21** 4690. DOI: 10.3390/ijms21134690 9. Robertson H, Kirby JA. **Post-transplant renal tubulitis: the recruitment, differentiation and persistence of intra-epithelial T cells**. *Am J Transplant* (2003.0) **3** 3-10. DOI: 10.1034/j.1600-6143.2003.30102.x 10. Berney-Meyer L, Hung N, Slatter T, Schollum JB, Kitching AR, Walker RJ. **Omeprazole-induced acute interstitial nephritis: a possible Th1-Th17-mediated injury?**. *Nephrology (Carlton)* (2014.0) **19** 359-365. DOI: 10.1111/nep.12226 11. Rosen S, Stillman IE. **Acute tubular necrosis is a syndrome of physiologic and pathologic dissociation**. *J Am Soc Nephrol* (2008.0) **19** 871-875. DOI: 10.1681/ASN.2007080913 12. 12.kidney disease: Improving Global Outcomes (KDIGO) Acute Kidney Injury Work Group. KDIGO Clinical Practice Guideline for Acute Kidney Injury. Kidney International Supplements. 2012 Mar;2(1):1. 13. Hull RP, Goldsmith DJA. **Nephrotic syndrome in adults**. *BMJ* (2008.0) **336** 1185-1189. DOI: 10.1136/bmj.39576.709711.80 14. Khanna R. **Clinical Presentation & Management of Glomerular Diseases: Hematuria**. *Nephritic & Nephrotic Syndrome Mo Med* (2011.0) **108** 33-36. PMID: 21462608 15. 15.Sethi S, D’Agati VD, Nast CC, Fogo AB, De Vriese AS, Markowitz GS, et al. A proposal for standardized grading of chronic changes in native kidney biopsy specimens. Kidney International. 1 april 2017;91(4):787–9. 16. Raza MN, Hadid M, Keen CE, Bingham C, Salmon AH. **Acute tubulointerstitial nephritis, treatment with steroid and impact on renal outcomes**. *Nephrology* (2012.0) **17** 748-753. DOI: 10.1111/j.1440-1797.2012.01648.x 17. Shemesh O, Golbetz H, Kriss JP, Myers BD. **Limitations of creatinine as a filtration marker in glomerulopathic patients**. *Kidney Int* (1985.0) **28** 830. DOI: 10.1038/ki.1985.205 18. Levey AS. **Measurement of renal function in chronic renal disease**. *Kidney Int* (1990.0) **38** 167. DOI: 10.1038/ki.1990.182 19. Dossetor JB. **Creatininemia versus uremia. The relative significance of blood urea nitrogen and serum creatinine concentrations in azotemia**. *Ann Intern Med* (1966.0) **65** 1287. DOI: 10.7326/0003-4819-65-6-1287 20. Miller TR, Anderson RJ, Linas SL, Henrich WL, Berns AS, Gabow PA, Schrier RW. **Urinary diagnostic indices in acute renal failure: a prospective study**. *Ann Intern Med* (1978.0) **89** 47. DOI: 10.7326/0003-4819-89-1-47 21. Su T, Gu Y, Sun P, Tang J, Wang S, Liu G. **Etiology and renal outcomes of acute tubulointerstitial nephritis: a single-center prospective cohort study in China**. *Nephrol Dial Transplant* (2018.0) **33** 1180-1188. DOI: 10.1093/ndt/gfx247 22. Praga M, Sevillano A, Auñón P, González E. **Changes in the aetiology, clinical presentation and management of acute interstitial nephritis, an increasingly common cause of acute kidney injury**. *Nephrol Dial Transplant* (2015.0) **30** 1472-1479. DOI: 10.1093/ndt/gfu326 23. Shang W, Wang Z. **The Update of NGAL in Acute Kidney Injury**. *Curr Protein Pept Sci* (2017.0) **18** 1211-1217. DOI: 10.2174/1389203717666160909125004 24. Zhao WT, Huang JW, Sun PP, Su T, Tang JW, Wang SX, Liu G, Yang L. **Diagnostic roles of urinary kidney injury molecule 1 and soluble C5b–9 in acute tubulointerstitial nephritis**. *Am J Physiol Renal Physiol* (2019.0) **317** F584-F592. DOI: 10.1152/ajprenal.00176.2019 25. 25.drug-associated chronic tubulointerstitial nephritisAm J Med Sci201334628328810.1097/MAJ.0b013e318271f91023147380. *Am J Med Sci* (2013.0) **346** 283-288. DOI: 10.1097/MAJ.0b013e318271f910 26. Heller F, Frischmann S, Grünbaum M, Zidek W, Westhoff TH. **Urinary calprotectin and the distinction between prerenal and intrinsic acute kidney injury**. *Clin J Am Soc Nephrol* (2011.0) **6** 2347-2355. DOI: 10.2215/CJN.02490311 27. Baker RJ, Pusey CD. **The changing profile of acute tubulointerstitial nephritis**. *Nephrol Dial Transplant* (2004.0) **19** 8. DOI: 10.1093/ndt/gfg464 28. 28.Eknoyan G, Raghavan R. Acute tubulointerstitial nephritis. In: Coffman T, Falk R, Molitoris B, Neilson E, Schrier RW (eds). Schrier’s Diseases of the Kidney. 9th edition. Philadelphia, PA: Lippincott, Williams & Wilkins; 2013. p. 994–1017.
--- title: Complement C3 mediates podocyte injury through TLR4/NFΚB-P65 signaling during ischemia–reperfusion acute kidney injury and post-injury fibrosis authors: - Yi Chen - Liyu Lin - Siyi Rao - Xuan Tao - Jiong Cui - Jianxin Wan journal: European Journal of Medical Research year: 2023 pmcid: PMC10041728 doi: 10.1186/s40001-023-01054-1 license: CC BY 4.0 --- # Complement C3 mediates podocyte injury through TLR4/NFΚB-P65 signaling during ischemia–reperfusion acute kidney injury and post-injury fibrosis ## Abstract ### Background The aim of this study was to explore the mechanism of complement C3a mediating podocyte injury during ischemia–reperfusion acute kidney injury (IR-AKI) and post-injury fibrosis. ### Methods Renal artery clamping was used to establish IR-AKI and post-injury fibrosis model. HE and Masson staining were performed to observe renal fibrosis. The protein abundance levels were measured along with inflammatory markers, renal complement C3. Podocytes were treated with C3a with or without Toll-like receptor 4(TLR4) inhibitor. The effects of TLR4 up-regulation by TLR4 plasmids were examined. ### Results C3−/− resulted in amelioration of renal dysfunction by reducing podocyte damage and renal fibrosis. Immunoblot with renal tissue homogenates from IR-AKI mice revealed that C3−/− decreased TLR4/Nuclear Factor-κB (NFκB)-P65. ### Conclusion Our results indicate that modulating C3/TLR4/NFκB-P65 signaling pathway is a novel therapeutic target for the IR-AKI and post-injury fibrosis. ## Introduction Chronic kidney disease (CKD) has increased significantly; there were over 697 million patients worldwide in 2019 [1]. The prevalence rate of CKD in Chinese adults is up to $10.8\%$ [2]. The incidence rate of acute kidney injury (AKI) inpatients in *Chinese* general hospitals is 5–$7\%$; approximately $50\%$ of surviving patients suffer from permanent renal dysfunction, poor prognosis, and heavy medical burden [3–5]. As far back as 2009, the international nephrology community has clearly indicated that AKI is the direct cause of CKD [6]; however, for many years, the treatment of AKI has been symptomatic only, preventing AKI from progressing to CKD has become the focal point of the international nephrology community. The clear pathophysiological mechanisms related to chronicity after AKI are microvascular endothelial cell damage, inflammation, and abnormal renal tubular epithelial cells activation. However, there have been few studies on podocytes. Perspicacious studies found that patients with AKI often display persistent proteinuria and albuminuria after AKI, which is closely related to subsequent CKD [7]. Podocyte damage leads to proteinuria, which is an indicator of most glomerular diseases and is related to the progression of kidney disease. Research by Hu et al. suggested that podocytes involve in the occurrence and progression of AKI [8]. The role of complement in IRI has received increasing attention. However, the mechanism is not clear. Complement deficiency can inhibit elevated TLR4 [9]. Inhibition of TLR4 attenuated Heme-induced complement deposit on endothelial cells. A central role of P-sel is to tag endothelium as a target for complement activation in vivo and provide the missing link between TLR4 and complement activation [10]. Recent research found that less secretion of C3 appears to inhibit the activation of the High-Mobility Group Box 1 (HMGB1)-TLR4-p65 pathway signal pathway and the production of transforming growth factor-β (TGF-β1), thereby alleviating renal fibrosis in unilateral ureteral obstruction (UUO) mice [11]. However, whether C3 and TLR4 interact in CKD after AKI is still unclear. In previous research, we established unilateral renal ischemia IR-AKI model under different ischemia times and found that the mice in the 20 min ischemia group were mainly characterized by mild AKI lesions, without chronic renal manifestation after AKI. Mice in 40 min ischemia group showed severe renal tubule and renal interstitial damage, which was irreversible. However, ischemia 30 min and contralateral kidney excision after 8 days could establish a post-AKI fibrosis kidney model. Electron microscopy was confirmed in mouse podocytes during AKI injury, proteinuria, and subsequent chronic kidney fibrosis, thus confirming that podocyte injury is an important cause of AKI and post-AKI renal fibrosis [12]. Our research found that complement C3 exacerbates renal interstitial fibrosis by facilitating the M1 macrophage polarization in a mouse model of unilateral ureteral obstruction [13]. Complement C3 activation also generates the anaphylatoxins C3a, which have potent proinflammatory effects. Therefore, we hypothesize that complement C3-mediated podocyte injury is involved in the post-injury fibrosis of the kidney after AKI. This study used wild-type C57BL/6 mice and C3 knockout mice to establish an IR-AKI and post-injury fibrosis model. Additionally, by in vitro treatment of mouse podocytes cultured under ischemic and hypoxic conditions, the mechanism of complement C3 promoting IR-AKI and post-injury fibrosis was investigated through examination of the TLR4/NFκB-P65 signaling pathway. Our findings aim to provide new outlooks on the prevention and treatment of post-AKI kidney fibrosis. ## Animals care All mice were raised in specific pathogen-free (SPF) barrier facility at The Laboratory Animal Center of Fujian Medical University. All procedures were approved by The Animal Welfare and Ethics Committee of Fujian Medical University (Approval Number: 2017-062): Ninety 16-week-old C57BL/6 mice (weight: 25–28 g, Shanghai SLAC Laboratory Animal Co., Ltd (production license number: SCXK (Shanghai) 2012-0002)) and twenty C3−/− mice (C3-deficient mice (strain B6.129S4-C3tm1Crr) of C57BL/6 genetic background from Jackson Laboratory (Bar Harbor, ME), 5 mice per cage (365 mm × 207 mm × 140 mm; model GK, Suzhou Feng's Laboratory Animal Equipment Co. Ltd, China) with bedding material (cellulose pellet) housed at a room temperature of 22 ± 2 ℃, the humidity of 55 ± $5\%$, and 12-h light/dark cycle. Mice had unrestricted get food and water. ## Establishment of mouse ischemia–reperfusion acute kidney injury and post-injury fibrosis model. Establishment of mouse IR-AKI and post-injury fibrosis model was according to Yi Chen method [12]. Fifty C57BL/6 mice were intraperitoneally anesthetized with $3\%$ pentobarbital sodium (1.0–1.5 mL/kg). The left renal artery was clamped with a micro-arterial clip for 30 min to restore blood flow. Sham group mice underwent the same surgical procedure without clamping the left renal artery. One week after surgery, the right kidney was excised. Afterward, the mice ($$n = 10$$) were euthanized at 28 days, blood and kidney tissues were also obtained for subsequent analyses. Euthanasia is performed as follows: pentobarbital sodium (60 mg/kg) is injected intraperitoneally and the animal is executed by cervical dislocation after 10 min. C3−/− (C57BL/6 background) mice AKI model were received the same method. ## Determination of urine protein and renal function Urinary protein, serum creatinine (Scr), and urea nitrogen (BUN) were examined with biochemical analyzer. ## Hematoxylin–Eosin (HE) and Masson’s staining The left kidney was excised, $10\%$ formalin fixed, paraffin embedded, and 3 μm serial sectioned. Subsequently, sections were dewaxed, hydrated, hematoxylin–eosin (HE) stained, and Masson stained. ## Renal tissue pathological damage score: Mouse kidney tissue specimens were stained by HE staining and Masson’s staining, and renal pathology scores were evaluated with light microscope as described previously [12]. For each animal, 10 randomly selected nonoverlapping interstitial fields and glomerulus were analyzed, and their average was used as data from one animal sample. And the average value was taken for statistical analysis. An independent pathologist was responsible for pathological scoring according to the following criteria:Glomerular mesangial hyperplasia: no, mild, moderate, and severe mesangial hyperplasia were scoring as 0, 1, 2, and 3 points, respectively. Degree of glomerulosclerosis: glomerulosclerosis rate ≤ $0\%$, < $25\%$, $25\%$ − $50\%$, and > $50\%$, were scoring as 0, 1, 2, and 3 points, respectively. Renal tubular interstitial score (RTIS), namely, [1] renal tubular degeneration and necrosis, [2] renal tubular atrophy, [3] interstitial inflammatory cell infiltration, and [4] interstitial fibrosis. According to the extent and severity of the lesions, 0, 1, 2, and 3 were scored as none, < $25\%$, 25−$50\%$, and > $50\%$, respectively. ## Evaluation of the glomerular and podocyte morphological changes by transmission electron microscopy (TEM) Kidney cortex tissue was obtained, cut out into a 1 mm3 tissue block, and placed in an electron microscope fixative solution. The tissue was then alcohol-acetone dehydrated, embedded, and sliced sections were observed for podocyte ultrastructure and glomerulosclerosis by electron microscope. ## Immunohistochemistry Paraffin sections were sliced and baked, dewaxed and hydrated, and incubated in $3\%$ (hydrogen peroxide solution) H2O2 for 10 min. Sections were incubated in primary antibody (50 μL/piece). Primary antibodies concentrations were Nephrin (Cat.ab2163, Abcam, 1:200), CD2AP (Cat.5478, CST, 1:100), Synaptopodin (Cat.ab22449, Abcam, 1:200), transient receptor potential channel 6 (TRPC6) (Cat.ab233413, Abcam, 1:100), C3 (Cat.ab200999, Abcam, 1:200), TLR4 (Cat.ab13867, Abcam, 1:200), and NFκB-P65 (Cat.3033, CST, 1:200), at 37℃ for 1 h, and rinse with phosphate-buffered saline (PBS). Then they were incubated in secondary antibody (China Zhongshan Golden Bridge) at 37 ℃ for 30 min, followed by rinse with PBS for 10 min × 3. Then after 1–2-min DAB incubation and sections were subsequently hematoxylin counterstained, dehydrated, and sealed. PBS replaced the primary antibody in all negative controls. Sections were evaluated under the microscope at a field of view of 200 times magnification. Five fields were randomly selected from each slice and optical density (IOD) value analysis was performed with the image analysis system, Motic Images Advanced. The protein expression level was represented by IOD value. ## Western blot analysis Kidney cortex tissues were lysate with RIPA lysate. 1 × sodium dodecyl sulfate (SDS) lysis buffer was used to extract total proteins of glomerular podocytes. The protein content was determined according to the instructions of a Braford protein quantification kit. Protein extract was separated with $10\%$ SDS-PAGE, transferred to polyvinylidene fluoride (PVDF) membrane, and incubated in $5\%$ free-fat milk at room temperature for 30 min. The concentrations of antibodies were Nephrin (Cat.ab2163, Abcam, 1:400), CD2AP (Cat.5478, CST, 1:1000), Synaptopodin (Cat.ab22449, Abcam, 1:2000), TRPC6 (Cat.ab233413, Abcam, 1:1000), TLR4 (Cat.ab13867, Abcam, 1:1000), NFκB-P65 (Cat.3033, CST, 1:1000), and incubation at 37℃ for 1.5 h. The membrane was rinsed with PBS for 10 min × 3, incubated in secondary antibody (1: 2000) at 37 ℃ for 1 h, rinsed with PBS for 10 min × 3, exposed, developed, and analyzed with White/Ultraviolet Transilluminator system software. The target protein/β-actin (optical density, OD) ratio was used as the indicator of the target protein level. ## Podocyte culture HSMPs (heat-sensitive mouse podocytes) were donated by Professor Ding Guohua of the School of Medicine of Wuhan University. HSMPs culture protocol has been previously published [14]. In vitro HSMPs were propagated under undifferentiated conditions (33 ℃, $5\%$ CO2, 1640 medium containing 100 U/mL interferon (IFN-γ Sigma, USA) and $10\%$ Fetal Bovine Serum, FBS) for 3–4 days. HSMPs were transferred to differentiation conditions (37 ℃, $5\%$ CO2, 1640 culture medium without interferon, and $10\%$ FBS). HSMPs morphology was observed under a microscope after 14 days. Differentiated podocytes were used experiment. For interruption of the TLR4, podocytes were treated with TAK242 (5 μM), a small-molecule-specific inhibitor of TLR4 signaling for 12 h. ## Podocyte culture ischemia and hypoxia conditioning Differentiated podocytes were seeded on a 24-well culture plate. When the cells grew to 70–$80\%$ confluence, they were replaced in serum-free 1640 medium (1 mL/well) and incubated for 24 h [15]. Buffer liquid in ischemia and hypoxia groups were replaced with NaHCO3 4.5 mM, Na2HPO4 0.8 mM, NaH2PO4 0.2 mM, NaCl 106 mM, KCl 5.4 mM, CaCl2 1.2 mM, MgCl2 0.8 mM, MES 20 Mm, sugar free, and PH 6.6 and placed in a three-gas incubator ($0.5\%$ O2 + $5\%$ CO2 + $94.5\%$ N2, 37 ℃) for 24 h. ## Podocyte nuclear protein extraction Podocytes were harvested with trypsin-Ethylenediaminetetraacetic acid (EDTA) and centrifuged at 500 × g for 5 min. Nuclear and cytoplasmic extracts were performed according to the NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Corporation) protocol. Extracts were stored at − 80℃ until use. ## F-actin staining of phalloidin podocyte cytoskeleton: Cells were mounted on a slide, rinsed 3 times with PBS, and fixed with $4\%$ paraformaldehyde for 15 min, followed by rinse with PBS for 3 min × 3. Phalloidin staining solution was incubated at 37 ℃ for 1 h and rinsed with PBS for 3 min × 3. DAPI was incubated for 15 min. The specimens were stained and washed with PBS for 5 min × 4. The cytoskeletal characteristics of podocytes were observed under a Confocal Laser Scanning Microscope and images were collected. ## Transfection of podocytes with the TLR4 plasmid vector The transfection of TLR4 plasmid was executed according to the instructions of the liposome transfection kit (Liposome transfection kit X-treme GENE Transfection Solution (Invitrogen)). Upon completion, a new serum-free medium was replaced for the intervention test 24 h later. ## Statistical processing All data are shown as mean ± Standard Deviation (SD) (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{x}$$\end{document}x¯± S). One-way ANOVA was performed for comparison between multiple groups, and LSD test for pairwise comparison between multiple groups; $P \leq 0.05$ indicated statistical significance. ## Changes in C3 and renal function during AKI To study the role of the C3 in the development of renal in AKI, we detected the level of C3, urinary protein excretion (UPE), BUN, and Scr in ischemia–reperfusion injury (IRI) mice. Compared with Day 0, the expression of C3 increased at the early stage of renal injury. On Day 2, the level of C3 increased ($P \leq 0.01$), peaking on Day 14 and then slowly declined ($P \leq 0.01$) (Fig. 1).Fig. 1Changes in C3 during post-injury fibrosis. a Western blot gel electrophoresis film exposure image and quantitative analysis of the result, expressed as the ratio of target protein/β-actin (OD). b The levels of complement C3 in renal tissue were measured by ELISA method, $$n = 10$.$ ** $P \leq 0.01$vs 0d; ##$P \leq 0.01$ vs 7d Compared with the sham group, serum Scr and BUN levels in the AKI-CKD group on Day 1 were higher ($P \leq 0.01$). Compared with Day 1, the Scr and BUN levels of Day 2 were decreased ($P \leq 0.01$). Scr and BUN increased consistently through the period between Day 3 and Day 28 ($P \leq 0.01$). Compared with the sham group, the urine protein level in the AKI-CKD group was increased consistently through the period between Day 3 ($P \leq 0.01$) and Day 28 ($P \leq 0.01$) (Table 1).Table 1Changes in urinary protein, BUN, and Scr in each group (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{x}$$\end{document}x¯±s, $$n = 10$$)GroupD0D1D2D3D7D14D28UPE (mg/d)Sham1.39 ± 0.191.43 ± 0.171.45 ± 0.181.44 ± 0.191.432 ± 0.161.41 ± 0.171.52 ± 0.27AKI-CKD1.41 ± 0.151.91 ± 0.331.96 ± 0.272.22 ± 0.35**##3.81 ± 1.23**##12.05 ± 2.38**##17.79 ± 2.86**##BUN(mmol/L)Sham11.21 ± 1.2511.01 ± 1.0311.88 ± 0.6911.92 ± 1.2712.51 ± 1.4212.15 ± 1.4111.97 ± 1.55AKI-CKD11.34 ± 1.3729.97 ± 2.12**##16.53 ± 1.24**##22.79 ± 313**##25.45 ± 2.99**##29.68 ± 1.45**##48.06 ± 6.72**##Scr(umol/L)Sham7.08 ± 0.827.15 ± 0.727.19 ± 0.917.07 ± 0.817.12 ± 0.836.98 ± 0.697.25 ± 0.94AKI-CKD7.43 ± 0.7318.11 ± 1.68**##11.75 ± 1.27**##13.97 ± 1.79**##15.68 ± 1.35**##24.01 ± 1.29**##35.61 ± 3.08**##UPE urinary protein excretion, BUN serum urea nitrogen, *Scr serum* creatinine*$P \leq 0.05$ vs Sham, **$P \leq 0.01$ vs Sham, #$P \leq 0.05$ vs D0, ##$P \leq 0.01$vs D0 ## The effect of C3 on renal function and renal pathology during post-injury fibrosis To determine whether C3 mediates renal fibrosis after IRI, we assessed renal pathology in C3−/− mice compared with age-matched wild-type (WT) mice. Compared with the WT-sham group, the UPE, BUN, and Scr levels of the WT- AKI-CKD group and the C3−/−-AKI-CKD group were higher ($P \leq 0.01$). Compared with the WT -AKI- CKD group, the UPE, BUN, and Scr levels of the C3−/−-AKI-CKD group were statistically lower ($P \leq 0.01$) (Table 2). HE staining showed that the glomerular and renal tubular structures were intact and clear in the WT-sham group and C3−/−-sham group. In the WT-AKI-CKD group, mesangial cells hyperplasia and interstitial inflammatory cells were observed. Renal tubules were atrophy and interstitial structures were blurred. Meanwhile, in the C3−/−-AKI-CKD group, only a small amount of mesangial cells hyperplasia and fewer interstitial inflammatory cells were seen. Tubular and interstitial structures were relatively intact (Fig. 2a). Compared with the Sham group, the AKI-CKD group renal tissue pathological damage grade and score were raised ($P \leq 0.01$). And compared with the WT-AKI-CKD group, the C3−/−-AKI-CKD group had a lower grade and score of renal histopathological damage. However, there was no statistical difference between the WT-sham group and the C3−/−-sham group ($P \leq 0.05$) (Table 2).Table 2Comparison of renal pathological changes and renal function damage of mice in each group (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{x}$$\end{document}x¯ ± SD, $$n = 10$$)WTC3−/−ShamAKI-CKDShamAKI-CKDMesangial hyperplasia score0 ± 01.83 ± 0.45**0 ± 01.15 ± 0.25**#Glomerulosclerosis score0.11 ± 0.281.36 ± 0.33**0 ± 00.76 ± 0.08**#Tubular interstitial score0.33 ± 0.056.55 ± 0.81**0 ± 03.12 ± 0.46**#Glomerular collagen volume fraction (%)0 ± 01.83 ± 0.45**0 ± 01.15 ± 0.25**#Renal tubular collagen volume fraction (%)0 ± 00.138 ± 0.021**0 ± 00.016 ± 0.002**#BUN(mmol/L)11.55 ± 1.1730.20 ± 1.63**10.77 ± 1.1919.36 ± 1.98**##Scr(umol/L)7.12 ± 1.0221.58 ± 2.29**6.94 ± 0.8215.78 ± 0.96**##UPE (mg/d)1.48 ± 0.1210.73 ± 1.08**1.52 ± 0.095.80 ± 1.04**##UPE urinary protein excretion, BUN serum urea nitrogen, *Scr serum* creatinine**$P \leq 0.01$vs WT-sham or C3−/− sham, #$P \leq 0.05$vs WT-AKI-CKD, ##$P \leq 0.01$vs WT-AKI-CKDFig. 2The effect of complement C3 on renal pathology during post-injury fibrosis. a Hematoxylin and eosin (HE) stain and Masson’s trichrome of kidney tissue. Scale bar of glomerular = 25 μm; Scale bar of renal tubular = 50 μm. b Immunohistochemical detection of the expression of Synaptopodin, Nephrin, CD2AP, TRPC6 (Scale bar = 25 μm) in mice of each group. c Compare the levels of podocyte functional protein (IOD) of kidney tissues in each group, $$n = 10$.$ d Western blot gel electrophoresis film exposure image e *Quantitative analysis* of the result, expressed as the ratio of target protein/β-actin (OD), $$n = 10$.$ f Morphological changes of glomerular podocytes in each group of mice (TEM, × 3500): kidney cortical tissue was cut to 1mm3 tissue for electron microscopy, $$n = 10$.$ ** $P \leq 0.01$vs WT-sham or C3−/−-sham, ## $P \leq 0.01$vs WT-AKI-CKD Masson’s staining showed no blue stained tissues in the WT-sham group or the C3−/−-sham group. There were lots of blue staining in the WT-AKI-CKD group, whereas only a small amount of blue staining was observed in the C3−/−-AKI-CKD group (Fig. 2a). Compared with the WT-sham group, glomerular, renal tubular collagen volume fractions in the WT-AKI-CKD group were significantly increased ($P \leq 0.01$). There were substantial decline in the C3−/−-AKI-CKD group than in the WT-AKI-CKD group ($P \leq 0.01$) (Table 2). Synaptopodin, Nephrin, and CD2AP immunohistochemical results showed a large number of brown-yellow granules in the glomeruli of the WT-sham group and C3−/−-sham group, while less in the WT-AKI-CKD group, and a moderate amount in the C3−/−-AKI-CKD group. The results of TRPC6 immunohistochemistry showed that only a small amount of brown-yellow granules was present in the glomeruli of the WT-sham group and the C3−/−-sham group, on the other hand, a large amount in the WT-AKI-CKD group and a moderate amount in the C3−/−-AKI-CKD group (Fig. 2b). Compared with the WT -AKI- CKD group, the IOD values of Synaptopodin, Nephrin, and CD2AP in the C3−/−-AKI- CKD group all increased significantly ($P \leq 0.$ 01), while the TRPC6 IOD value decreased considerably ($P \leq 0.01$) (Fig. 2c). Immunohistochemistry and Western blot analysis showed that, compared with the WT-sham group, the levels of Synaptopodin, Nephrin, and CD2AP in WT-AKI-CKD and C3−/−-AKI-CKD groups were obviously decreased ($P \leq 0.01$), while TRPC6 increased considerably ($P \leq 0.01$). Compared with WT -AKI- CKD group, the levels of Synaptopodin, Nephrin, and CD2AP were significantly increased ($P \leq 0.01$) and the level of TRPC6 protein considerably decreased in the C3−/−-AKI- CKD group ($P \leq 0.01$) (Fig. 2d, e). TEM showed that a complete and clear podocyte structure with no foot process fusion was observed in the WT-sham and the C3−/−-sham group. There was extensive podocyte fusion, and podocyte segment dissection was observed in the WT-AKI-CKD group, whereas only a small amount of podocyte fusion was detected, and no podocyte segment dissection was seen in C3−/−-AKI-CKD group (Fig. 2f). ## C3 deficiency inhibited TLR4/NFκB-P65 signaling pathway activation post-IRI. To verify the C3 promoting inflammation and therapeutic efficacy through the TLR4/NFκB-P65 pathway in vivo, we carried out quantitative and positional measurements. Immunohistochemistry and Western blot analysis showed that, compared with the WT-sham group, the levels of C3, TLR4, NFκB-P65 in the WT-AKI-CKD group were significantly increased ($P \leq 0.01$). Compared with the WT -AKI- CKD group, C3, TLR4, NFκB-P65 in the C3−/−-AKI-CKD group considerably decreased ($P \leq 0.01$) (Fig. 3a–d).Fig. 3C3 deficiency inhibited TLR4 / NFκB-P65 signaling pathway activation during post-injury fibrosis. a Immunohistochemical detection of glomerular complement C3, TLR4, and NFκB-P65 protein expression (Scale bar = 25 μm) in each group of mice. b Renal tissue immunohistochemical staining optical density analysis, $$n = 10$.$ c Western blot gel electrophoresis film exposure map, d Western blot gel electrophoresis exposure film grayscale analysis, with the target protein/β-actin (OD) ratio to express the target protein expression level, $$n = 10$.$ ** $P \leq 0.01$vs WT-sham or C3−/−-sham, ## $P \leq 0.01$vs WT-AKI -CKD ## The effect of C3a on podocyte cytoskeleton and signaling pathways under hypoxic-ischemic conditions To discern whether C3a promoted podocyte phenotypic alteration in hypoxic-ischemic, we exposed podocytes to C3a. Phalloidin staining showed that in the control group, cells were elongated or star shaped, had a small number of protrusions, and displayed regularly arranged cytoplasmic muscle filaments. Cell in the hypoxic-ischemic group demonstrated an irregularly shaped cell body, rounded cells, and disordered cytoplasmic muscle filament arrangement. In the hypoxic-ischemia + C3a group (C3a Merck Millipore, USA), the cell body was also irregularly shaped in a rounded manner and the arrangement of cytoplasmic muscle filaments was either disordered or unclear (Fig. 4a).Fig. 4The effect of C3a on podocyte cytoskeleton and signaling pathways under hypoxic-ischemic conditions. a Podocyte cytoskeleton stained with phalloidin, objective lens (40 ×). b Immunoblotting film exposure map, c Immunoblotting exposure film optical density analysis, $$n = 5$.$ * $P \leq 0.05$vs Control (C3a: 0 h) group. ** $P \leq 0.01$ vs Control (C3a: 0 h) group C3a promoted TLR4 expression and NFκB-P65 nuclear translocation in glomerular podocytes in a time-dependent manner under hypoxic-ischemia conditions. The level of TLR4 was significantly increased after 60 min with C3a (0.1 μM) under hypoxic-ischemia conditions ($P \leq 0.01$), and reaching its peak after 2 h. Meanwhile, cytoplasmic NFκB-P65 was significantly decreased ($P \leq 0.01$) and nucleus NFκB-P65 was markedly increased (Fig. 4b, c). ## Inhibition of TLR4 on C3a-induced podocyte damage and function under hypoxic-ischemic conditions To determine whether TLR4-mediated C3a activation specifically contributed to exacerbated podocytes in hypoxic-ischemic, we used TAK242 to inhibit TLR4 expression. Compared with the Control group, the levels of TLR4 and nuclear NFκB-P65 in the C3a group were significantly increased ($P \leq 0.01$) and the TLR4 receptor inhibitor TAK242 (5 μM) inhibited TLR4 expression and NFκB-P65 nuclear translocation, and compared with the C3a group, TLR4 expression and nuclear NFκB-P65 levels of the C3a + TAK242 group were significantly decreased ($P \leq 0.01$) (Fig. 5a, b).Fig. 5Inhibition of TLR4 on C3a-induced podocyte damage and function under hypoxic-ischemic conditions. a Immunoblotting film exposure map, b Immunoblotting exposure film optical density analysis, $$n = 3$.$ c Immunoblotting film exposure map, d Immunoblotting exposure film optical density analysis. e Inhibition of the effect of TLR4 on the C3a-induced podocyte cytoskeleton under hypoxic-ischemic conditions: stained with phalloidin, objective lens (40 ×). ** $P \leq 0.01$ vs Control, ## $P \leq 0.01$ vs C3a (0.1 μM) Compared with the control group, the levels of Synaptopodin, Nephrin, and CD2AP in the C3a group were markedly reduced ($P \leq 0.01$), but TLR4 and TRPC6 substantially increased ($P \leq 0.01$). Compared with the control group, the levels of TLR4, Synaptopodin, Nephrin, CD2AP, and TRPC6 in the TAK242 group were no statistical difference ($P \leq 0.05$). Compared with the C3a group, TLR4 and TRPC6 in the C3 + TAK242 group showed significantly reduced ($P \leq 0.01$), while Synaptopodin, Nephrin, and CD2AP were significantly elevated ($P \leq 0.01$) (Fig. 5c, d). Phalloidin staining showed that in the control group, cells were elongated or star shaped with a few protrusions, and the cytoplasmic myofilaments were regularly arranged. In the C3a group, the cell body was irregular, with a rounded shape and disordered arrangement of cytoplasmic myofilament. In the TAK242 group, cells were elongated or star shaped, with a small number of protrusions and regularly arranged cytoplasmic muscle filaments. The cells were round or star shaped in the C3a+TAK242 group, with a regularly engineered cytoplasmic muscle filament (Fig. 5e). ## Effect of TLR4 up-regulation on C3a-induced podocyte function and hypoxic-ischemic damage. To further confirm the effect of TLR4 in hypoxic-ischemic, we overexpress the TLR4 in the podocyte. The TLR4 protein expression level was significantly increased in podocytes with TLR4 plasmid transfection ($P \leq 0.01$), and there was no statistically significant difference ($P \leq 0.05$) between empty plasmid group and liposome group and control group (Fig. 6a, b).Fig. 6Effect of up-regulating TLR4 on C3a-induced podocyte damage and function under hypoxic-ischemic conditions. a Immunoblotting film exposure map; b Immunoblotting exposure film optical density analysis, $$n = 3$.$ c Immunoblotting film exposure map; d Immunoblotting film optical density analysis, $$n = 3$.$ e Immunoblotting film exposure image. f Immunoblotting exposure film optical density analysis, $$n = 4$.$ g Podocyte cytoskeleton stained with phalloidin, objective lens (40 ×). ** $P \leq 0.01$ vs Control group; ## $P \leq 0.01$ vs C3a + Vehicle group Compared with the vehicle group, the level of Cyto-NFκB-P65 protein in the C3a + TLR4 plasmid group (C3a 0.1 μM treatment) was significantly declined, the level of Nucleus-NFκB-P65 protein was considerably increased ($P \leq 0.01$) (Fig. 6c, d). The level of TRPC6 protein in the C3a+TLR4 plasmid group significantly decreased. On the other hand, Synaptopodin, Nephrin, and CD2AP protein expression levels were substantially increased ($P \leq 0.01$) (Fig. 6e, f). Finally, phalloidin staining showed that in the Control group, cells were elongated or star shaped and had a small number of protrusions with regularly arranged cytoplasmic myofilaments. In the C3a group or C3a+Vehicle group, the cell body was irregularly shaped, with rounded cells and disordered or unclear cytoplasmic myofilaments. In the C3a+TLR4 plasmid group, the cell nucleus was markedly smaller, with both the membrane protrusion and the cytoplasmic myofilament disappearing (Fig. 6g). ## Discussion At present, in the world, CKD's social and economic burden ranks among the top diseases [16, 17]. The prevalence of CKD in China has been increasing over the years [4, 17]. In recent years, epidemiological and experimental studies have shown that AKI is closely related to CKD, being one of the important factors attributed to the increased incidence of CKD [18–20]. This study found that complement C3 and the inflammation regulatory protein TLR4/NFκB-P65 all increased considerably after renal ischemia–reperfusion AKI and post-injury fibrosis. Complement C3 gene knockdown can mitigate the changes above. In vitro experiments showed that the inhibition of TLR4 can reduce C3a-induced podocyte NFκB-P65 levels under hypoxic-ischemic conditions, stabilizing the podocyte cytoskeleton. Moreover, up-regulating TLR4 levels have the opposite effect. We pointed out that complement C3a can mediate podocyte injury through TLR4/NFκB-P65 signaling pathway, participating in post-AKI fibrosis. Recent study found that podocytes through the Wnt/β-Catenin pathway involve in AKI [8]. Our previous studies have found that podocyte damage may be the main cause of CKD after AKI [12]. Complement activation is involved in the pathogenesis of ischemia reperfusion injury (IRI), which is an inevitable process during AKI [21]. In investigating the causes of podocyte damage, we noticed that complement C3 is an important factor of damage. The Malmö Diet and Cancer cohort study found that in the general population, complement C3 was related to the incidence of first hospitalization of CKD [22]. The complement system activated, forming a membrane attack complex (MAC), resulting in podocyte apoptosis and damage, cytoskeletal changes, abnormal pore membrane protein, filtration barrier damage, and proteinuria [23, 24]. Recent studies have demonstrated that the pathogenic activation of complement by the glomerular subepithelial immune complex is a critical step for proteinuria. And the activation of complement is a key trigger for podocyte loss and activation of the mesangial epithelial cells of the glomeruli, which then lead to glomerulosclerosis [25, 26]. Infiltration of inflammatory cells and activation of inflammatory factors in kidney tissue after IRI have been shown to accelerate the progression of CKD. Studies by Chao Hu et al. [ 21] showed that complement activation induces renal IRI, and the complement inhibitor complement receptor of the immunoglobulin superfamily (CRIg) /FH can improve renal IRI by activating PI3K/AKT signaling. Targeting of a human complement inhibitor, CR1, provided effective protection against cardiac IRI [27]. Complements C3a and C5a could lead to endothelial cell transdifferentiation after IRI, and inhibition of complements C3a and C5a can significantly reduce renal fibrosis [28, 29]. C3a induces mitochondrial dysfunction in podocytes, and inhibition of C3aR significantly limited podocyte loss and enhanced podocyte density [30]. These data, together with the evidence that injury to podocytes is a major cause of glomerulosclerosis, support our speculation that there is a detrimental link between complement C3 and podocyte injury in the post-injury fibrosis of the kidney after AKI. Our study found that podocyte injury and extensive foot process fusion occurred in the WT-AKI-CKD group, with a decrease in the expression of the podocyte functional proteins and an increase in the expression of TRPC6, progressively increasing proteinuria and eventually chronic renal fibrosis and renal dysfunction. Compared with the WT-AKI-CKD group, in the C3−/−-AKI-CKD group, TEM shows improved podocyte fusion, with significantly increased the level of podocyte functional protein and decreased the level of TRPC6, consequently improving proteinuria and renal function. Pathology evaluation further indicated that mesangial hyperplasia, glomerular sclerosis, and renal tubule interstitial fibrosis declined, demonstrating significant improvement in kidney chronic fibrosis. Primary cultured human podocytes and conditionally immortalized mouse podocytes can synthesize and secrete complements C1q, C1r, C2, C3, C7, complement factor H (CFH), CD59, C4bp, CD46, Protein S, complement receptor 2 (CR2), C1qR, C3aR, C5aR, and Crry under physiological conditions. The synthesis increases under the stimulation of inflammatory factor interferon γ, which is affected by various cytokines such as angiotensin II (Ang II), IL6, and TGF-β [31]. In this research, we found that complement C3 was activated during post-AKI renal fibrosis of mice as complement C3 knockout mice can improve renal tissue inflammation and podocyte damage during AKI, reducing post-AKI renal fibrosis. We speculate that the activation of complement C3 is thereby closely related to podocyte injury and post-AKI renal fibrosis. TLR4 is an immunoinflammatory grade trigger protein involved in the signal transduction of the ganglion reaction [32, 33], and widely expressed on the surface of various immune cells, as well as in glomerular podocytes, mesangial cells, and renal tubular epithelial cells [34, 35]. TLR4 and the complement system are widely involved in the occurrence and progression of various diseases, such as renal IRI, AKI and diabetic nephropathy [36–38], which can activate the transcription and expression of NFκB-mediated complement factor, thereby initiating the inflammatory cascade [39, 40]. In vitro experiments, we have observed that under hypoxic-ischemic conditions the differentiation of glomerular podocytes can be induced. Complement C3a promotes TGF-β1 synthesis and the TLR4/ NFκB-P65 protein expression in glomerular podocytes. We also observed that C3a promotes TLR4 expression and NFκB-P65 nuclear translocation in glomerular podocytes under hypoxic-ischemic conditions. Up-regulating the level of TLR4 can increase the levels of NFκB-P65 in C3a-induced podocyte nucleus under hypoxic-ischemic conditions, inducing the synthesis of the glomerular podocyte transdifferentiation proteins Synaptopodin, Nephrin, and CD2AP and reducing the expression of TRPC6, resulting in obvious damage to the podocyte cytoskeleton. TLR4 inhibitors can inhibit NFKB-P65 nuclear translocation, reducing the protein expression level of Synaptopodin, Nephrin, and CD2AP, inducing TRPC6 protein synthesis, and stabilizing the podocyte cytoskeleton. Since podocytes express both the C3a receptor and TLR4, we speculate that complement C3a may activate the TLR4 protein of glomerular podocytes through the TLR4/NFκB-P65 signaling pathway, inducing NFκB-P65 nuclear translocation, participating in the inflammation and transdifferentiation of glomerular podocytes and resulting in the progression of post-AKI renal fibrosis. ## Conclusion In summary, complement C3-mediated inflammatory response-induced glomerular podocyte injury is closely related to the post-injury fibrosis after renal ischemia–reperfusion AKI. These effects may be achieved by regulating podocyte TLR4/NFκB-P65 signaling pathway. ## References 1. **Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019**. *Lancet* (2020) **2020** 1204-1222. DOI: 10.1016/S0140-6736(20)30925-9 2. Zhang L, Wang F, Wang L, Wang W, Liu B, Liu J, Chen M, He Q, Liao Y, Yu X, Chen N, Zhang JE, Hu Z, Liu F, Hong D, Ma L, Liu H, Zhou X, Chen J, Pan L, Chen W, Wang W, Li X, Wang H. **Prevalence of chronic kidney disease in China: a cross-sectional survey**. *Lancet* (2012) **379** 815-822. DOI: 10.1016/S0140-67 3. Fang Y, Teng J, Ding X. **Acute kidney injury in China**. *Hemodial Int* (2015) **19** 2015. DOI: 10.1111/hdi.12193 4. Zhang L, Zhao MH, Zuo L, Wang Y, Yu F, Zhang H, Wang H. **China Kidney Disease Network (CK-NET) 2015 Annual Data Report**. *Kidney Int Suppl* (2019) **9** e1-e81. DOI: 10.1016/j.kisu.2018.11.001 5. Zhou M, Wang H, Zeng X, Yin P, Zhu J, Chen W, Li X, Wang L, Wang L, Liu Y, Liu J, Zhang M, Qi J, Yu S, Afshin A, Gakidou E, Glenn S, Krish VS, Miller-Petrie MK, Mountjoy-Venning WC, Mullany EC, Redford SB, Liu H, Naghavi M, Hay SI, Wang L, Murray CJL, Liang X. **Mortality, morbidity, and risk factors in China and its provinces, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017**. *Lancet* (2019) **394** 1145-1158. DOI: 10.1016/S0140-6736(19)30427-1 6. Coca SG, Yusuf B, Shlipak MG, Garg AX, Parikh CR. **Long-term risk of mortality and other adverse outcomes after acute kidney injury: a systematic review and meta-analysis**. *Am J Kidney Dis* (2009) **53** 961-973. DOI: 10.1053/j.ajkd.2008.11.034 7. Horne KL, Packington R, Monaghan J, Reilly T, Selby NM. **Three-year outcomes after acute kidney injury: results of a prospective parallel group cohort study**. *BMJ Open* (2017) **7** e015316. DOI: 10.1136/bmjopen-2016-015316 8. Hu X, Zhou W, Wu S, Wang R, Luan Z, Geng X, Xu N, Zhang Z, Ruan Z, Wang Z, Li F, Yu C, Ren H. **Tacrolimus alleviates LPS-induced AKI by inhibiting TLR4/MyD88/NF-κB signalling in mice**. *J Cell Mol Med* (2022) **26** 507-514. DOI: 10.1111/jcmm.17108 9. Markarian M, Krattli RP, Baddour JD, Alikhani L, Giedzinski E, Usmani MT, Agrawal A, Baulch JE, Tenner AJ, Acharya MM. **Glia-selective deletion of complement C1q prevents radiation-induced cognitive deficits and neuroinflammation**. *Cancer Res* (2021) **81** 1732-1744. DOI: 10.1158/0008-5472.CAN-20-2565 10. Merle NS, Paule R, Leon J, Daugan M, Robe-Rybkine T, Poillerat V, Torset C, Frémeaux-Bacchi V, Dimitrov JD, Roumenina LT. **P-selectin drives complement attack on endothelium during intravascular hemolysis in TLR-4/heme-dependent manner**. *Proc Natl Acad Sci U S A* (2019) **116** 6280-6285. DOI: 10.1073/pnas.1814797116 11. Wang K, Wei H, Zhan J, Liang X, Zhang C, Liu Y, Xu G. **GSPE alleviates renal fibrosis by inhibiting the activation of C3/ HMGB1/TGF-β1 pathway**. *Chem Biol Interact* (2020) **316** 108926. DOI: 10.1016/j.cbi.2019.108926 12. Chen Y, Lin L, Tao X, Song Y, Cui J, Wan J. **The role of podocyte damage in the etiology of ischemia-reperfusion acute kidney injury and post-injury fibrosis**. *BMC Nephrol* (2019) **20** 106. DOI: 10.1186/s12882-019-1298-x 13. Cui J, Wu X, Song Y, Chen Y, Wan J. **Complement C3 exacerbates renal interstitial fibrosis by facilitating the M1 macrophage phenotype in a mouse model of unilateral ureteral obstruction**. *Am J Physiol Renal Physiol* (2019) **317** F1171-f1182. DOI: 10.1152/ajprenal.00165 14. Shankland SJ, Pippin JW, Reiser J, Mundel P. **Podocytes in culture: past, present, and future**. *Kidney int* (2007) **72** 26-36. DOI: 10.1038/sj.ki.5002291 15. Sauvant C, Schneider R, Holzinger H, Renker S, Wanner C, Gekle M. **Implementation of an in vitro model system for investigation of reperfusion damage after renal ischemia**. *Cell Physiol Biochem* (2009) **24** 567-576. DOI: 10.1159/000257513 16. Webster AC, Nagler EV, Morton RL, Masson P. **Chronic Kidney Disease**. *Lancet* (2017) **389** 1238-1252. DOI: 10.1016/S0140-6736(16)32064-5 17. Bowe B, Xie Y, Li T, Mokdad AH, Xian H, Yan Y, Maddukuri G, Al-Aly Z. **Changes in the US Burden of Chronic Kidney Disease From 2002 to 2016: An Analysis of the Global Burden of Disease Study**. *JAMA Netw Open* (2018) **1** e184412. DOI: 10.1001/jamanetworkopen 18. Hapca S, Siddiqui MK, Kwan RSY, Lim M, Matthew S, Doney ASF, Pearson ER, Palmer CNA, Bell S. **The Relationship between AKI and CKD in Patients with Type 2 Diabetes: An Observational Cohort Study**. *J Am Soc Nephrol* (2021) **32** 138-150. DOI: 10.1681/ASN.2020030323 19. Chawla LS, Eggers PW, Star RA, Kimmel PL. **Acute kidney injury and chronic kidney disease as interconnected syndromes**. *N Engl J Med* (2014) **371** 58-66. DOI: 10.1056/NEJMra1214243 20. Venkatachalam MA, Weinberg JM, Kriz W, Bidani AK. **Failed Tubule Recovery, AKI-CKD Transition, and Kidney Disease Progression**. *J Am Soc Nephrol* (2015) **26** 1765-1776. DOI: 10.1681/ASN.2015010006 21. Hu C, Li L, Ding P, Li L, Ge X, Zheng L, Wang X, Wang J, Zhang W, Wang N, Gu H, Zhong F, Xu M, Rong R, Zhu T, Hu W. **Complement Inhibitor CRIg/FH Ameliorates Renal Ischemia Reperfusion Injury via Activation of PI3K/AKT Signaling**. *J Immunol* (2018) **201** 3717-3730. DOI: 10.4049/jimmunol.1800987 22. Bao X, Borné Y, Muhammad IF, Schulz CA, Persson M, Orho-Melander M, Niu K, Christensson A, Engström G. **Complement C3 and incident hospitalization due to chronic kidney disease: a population-based cohort study**. *BMC Nephrol* (2019) **21** 61. DOI: 10.1186/s12882-019-1248-7 23. Danobeitia JS, Djamali A, Fernandez LA. **The role of complement in the pathogenesis of renal ischemia-reperfusion injury and fibrosis**. *Fibrogenesis Tissue Repair* (2014) **7** 16. DOI: 10.1186/1755-1536-7-16 24. Yu SM, Nissaisorakarn P, Husain I, Jim B. **Proteinuric Kidney Diseases: A Podocyte's Slit Diaphragm and Cytoskeleton Approach**. *Front Med (Lausanne)* (2018) **5** 221. DOI: 10.3389/fmed.2018.00221 25. Morigi M, Locatelli M, Rota C, Buelli S, Corna D, Rizzo P, Abbate M, Conti D, Perico L, Longaretti L, Benigni A, Zoja C, Remuzzi G. **A previously unrecognized role of C3a in proteinuric progressive nephropathy**. *Sci Rep* (2016) **6** 28445. DOI: 10.1038/srep28445 26. Luo W, Olaru F, Miner JH, Beck LH, van der Vlag J, Thurman JM, Borza DB. **Alternative Pathway Is Essential for Glomerular Complement Activation and Proteinuria in a Mouse Model of Membranous Nephropathy**. *Front Immunol* (2018) **9** 1433. DOI: 10.3389/fimmu.2018.01433 27. Zheng C, Sleiman MM, Yang X, He S, Atkinson C, Tomlinson S. **Increasing the efficacy and safety of a human complement inhibitor for treating post-transplant cardiac ischemia reperfusion injury by targeting to a graft-specific neoepitope**. *J Heart Lung Transplant* (2021) **40** 1112-1121. DOI: 10.1016/j.healun.2021.07.004 28. Curci C, Castellano G, Stasi A, Divella C, Loverre A, Gigante M, Simone S, Cariello M, Montinaro V, Lucarelli G, Ditonno P, Battaglia M, Crovace A, Staffieri F, Oortwijn B, van Amersfoort E, Gesualdo L, Grandaliano G. **Endothelial-to-mesenchymal transition and renal fibrosis in ischaemia/reperfusion injury are mediated by complement anaphylatoxins and Akt pathway**. *Nephrol Dial Transplant* (2014) **29** 799-808. DOI: 10.1093/ndt/gft516 29. Durigutto P, Sblattero D, Biffi S, De Maso L, Garrovo C, Baj G, Colombo F, Fischetti F, Di Naro AF, Tedesco F, Macor P. **Targeted Delivery of Neutralizing Anti-C5 Antibody to Renal Endothelium Prevents Complement-Dependent Tissue Damage**. *Front Immunol* (2017) **8** 1093. DOI: 10.3389/fimmu.2017.01093 30. Morigi M, Perico L, Corna D, Locatelli M, Cassis P, Carminati CE, Bolognini S, Zoja C, Remuzzi G, Benigni A, Buelli S. **C3a receptor blockade protects podocytes from injury in diabetic nephropathy**. *JCI Insight* (2020) **5** e131849. DOI: 10.1172/jci.insight.131849 31. Li X, Ding F, Zhang X, Li B, Ding J. **The Expression Profile of Complement Components in Podocytes**. *Int J Mol Sci* (2016) **17** 471. DOI: 10.3390/ijms17040471 32. Molteni M, Gemma S, Rossetti C. **The Role of Toll-Like Receptor 4 in Infectious and Noninfectious Inflammation**. *Mediators Inflamm* (2016) **2016** 6978936. DOI: 10.1155/2016/6978936 33. Yang Y, Lv J, Jiang S, Ma Z, Wang D, Hu W, Deng C, Fan C, Di S, Sun Y, Yi W. **The emerging role of Toll-like receptor 4 in myocardial inflammation**. *Cell Death Dis* (2016) **7** e2234. DOI: 10.1038/cddis.2016.140 34. Mudaliar H, Pollock C, Panchapakesan U. **Role of Toll-like receptors in diabetic nephropathy**. *Clin Sci (Lond)* (2014) **126** 685-694. DOI: 10.1042/CS20130267 35. McCoy KL. **Interaction between Cannabinoid System and Toll-Like Receptors Controls Inflammation**. *Mediators of inflammation. Mediators Inflamm* (2016) **2016** 5831315. DOI: 10.1155/2016/5831315 36. Gluba A, Banach M, Hannam S, Mikhailidis DP, Sakowicz A, Rysz J. **The role of Toll-like receptors in renal diseases**. *Nat Rev Nephrol* (2010) **6** 224-235. DOI: 10.1038/nrneph.2010.16 37. Damman J, Daha MR, van Son WJ, Leuvenink HG, Ploeg RJ, Seelen MA. **Crosstalk between complement and Toll-like receptor activation in relation to donor brain death and renal ischemia-reperfusion injury**. *Am J Transplant* (2011) **11** 660-669. DOI: 10.1111/j.1600-6143.2011.03475.x 38. Zhao H, Perez JS, Lu K, George AJ, Ma D. **Role of Toll-like receptor-4 in renal graft ischemia-reperfusion injury**. *Am J Physiol Renal Physiol* (2014) **306** F801-F811. DOI: 10.1152/ajprenal.00469.2013 39. Rao J, Yue S, Fu Y, Zhu J, Wang X, Busuttil RW, Kupiec-Weglinski JW, Lu L, Zhai Y. **ATF6 mediates a pro-inflammatory synergy between ER stress and TLR activation in the pathogenesis of liver ischemia-reperfusion injury**. *Am J Transplant* (2014) **14** 1552-1561. DOI: 10.1111/ajt.12711 40. Zuniga MC, Raghuraman G, Hitchner E, Weyand C, Robinson W, Zhou W. **PKC-epsilon and TLR4 synergistically regulate resistin-mediated inflammation in human macrophages**. *Atherosclerosis* (2017) **259** 51-59. DOI: 10.1016/j.atherosclerosis.2017.02.021
--- title: FUT2 inhibits the EMT and metastasis of colorectal cancer by increasing LRP1 fucosylation authors: - Lingnan He - Zijun Guo - Weijun Wang - Shuxin Tian - Rong Lin journal: 'Cell Communication and Signaling : CCS' year: 2023 pmcid: PMC10041739 doi: 10.1186/s12964-023-01060-0 license: CC BY 4.0 --- # FUT2 inhibits the EMT and metastasis of colorectal cancer by increasing LRP1 fucosylation ## Abstract ### Background Fucosyltransferase 2(FUT2) and its induced α-1,2 fucosylation is associated with cancer metastasis. However, the role of FUT2 in colorectal cancer (CRC) metastasis remains unclear. ### Methods The expression levels and clinical analyses of FUT2 were assessed in CRC samples. Migration and invasion assays, EMT detection, nude mice peritoneal dissemination models and intestinal specific FUT2 knockout mice (FUT2△IEC mice) were used to investigate the effect of FUT2 on metastasis in colorectal cancer. Quantitative proteomics study of glycosylated protein, UEA enrichment, Co-immunoprecipitation identified the mediator of the invasive-inhibiting effects of FUT2. ### Results FUT2 is downregulated in CRC tissues and is positively correlated with the survival of CRC patients. FUT2 is an inhibitor of colorectal cancer metastasis which, when overexpressed, suppresses invasion and tumor dissemination in vitro and in vivo. FUT2 knock-out mice (FUT2△IEC mice) develop AMO and DSS-induced tumors and promote EMT in colorectal cancers. FUT2-induced α-1,2 fucosylation impacts the ability of low-density lipoprotein receptor-related protein 1(LRP1) to suppress colorectal cancer invasion. ### Conclusions Our study demonstrated that FUT2 induces α-1,2 fucosylation and inhibits EMT and metastasis of colorectal cancer through LRP1 fucosylation, suggesting that FUT2 may serve as a therapeutic target for colorectal cancer. Video Abstract ### Supplementary Information The online version contains supplementary material available at 10.1186/s12964-023-01060-0. ## Background Colorectal cancer (CRC) is the third most common cancer and the second most common cause of cancer-related death worldwide [1]. Despite the recent approval of advanced therapies (i.e., surgery, chemotherapy and radiotherapy), metastatic CRC still occurs in more than $50\%$ of patients who undergo resection [2]. Therefore, the identification of determinants of CRC metastasis is a key step toward effectively controlling tumor progression. Emerging evidence indicates that changes in glycosylation play a crucial functional role in tumor progression and metastasis [3]. Altered glycosylation is implicated in processes related to angiogenic signaling and endothelial cell adhesion [4, 5], which not only directly impact cell growth [6] and survival but also facilitate tumor-induced immunomodulation [7] and eventual metastasis [8, 9]. Glycosylation is a specific posttranslational modification that is mainly controlled by glycosyltransferases and glycosidases that orchestrate the addition of defined glycan structures to glycoproteins and/or lipids [10]. Glycosylation results in several functional changes to glycoproteins that confer unique features that are characteristic of cancer cells and the tumor microenvironment [11]. Aberrantly glycosylated proteins affect different steps of the metastasis process, including epithelial-mesenchymal transformation (EMT), migration, invasion/infiltration, and tumor cell extravasation [12]. One of the most common glycan modifications on proteins or lipids is the attachment of fucoses via the action of various glycosyltransferases [13]. Recent reports have shown that fucosylation is associated with cancer progression [14] and metastasis [15] in colorectal cancer. Aberrant fucosylation was shown to be associated with prometastatic and metastasis-suppressing functions in cancer [16, 17]. However, there is currently limited understanding of which enzymes and related fucosylation modifications are important for the progression and metastasis of colorectal cancer. Fucosyltransferase 2 (FUT2) is a key enzyme that catalyzes the transfer of fucose to the terminal galactose of type 1 or type 2 disaccharide via α1,2-linkage [18]. Aberrant α-1,2 fucosylation is a hallmark of multiple types of cancers [19]. Recent reports have revealed important functions of FUT2 in cancers. For instance, a lectin microarray of melanoma identified FUT2 as an anti-metastatic factor, and silencing FUT2 promoted the invasion of melanoma cells [20]. Nevertheless, FUT2 overexpression increased cell migration and invasion in vitro and metastasis of breast cancer in vivo [21]. The role and molecular mechanisms of FUT2 in colorectal cancer remain largely unclear. Thus, identifying the role and target proteins of FUT2 as well as gaining insight into their biological functions in colorectal cancer are worthwhile. ## Cell lines and cell culture The human colorectal cancer cell lines HCT116 and SW480 were purchased from American Type Culture Collection (ATCC). All CRC cell lines were authenticated by short tandem repeat analysis and were negative for mycoplasma. The cells were cultured with DMEM (Gibco, San Francisco, CA, USA) supplemented with $10\%$ fetal bovine serum (Gibco) and 100 U/mL penicillin and streptomycin (Servicebio, Wuhan, China) at 37 °C in a humidified incubator in $5\%$ CO2. ## Animal studies and clinical specimens BALB/c nude mice and Wild-type C57BL/6 male mice were purchased from Beijing Huafukang Biological Co., Ltd. and housed under standard pathogen-free conditions in the Experimental Animal Center of Tongji Medical College, Huazhong University of Science & Technology. Pvillin-Cre recombinase transgenic C57BL/6 mice (Pvillin-Cre TG mice) and FUT2flox/flox C57BL/6 mice (purchased from GemPharmatech Co. Ltd) were crossed and generated mice with FUT2 gene specifically deleted in intestinal epithelial cell (Pvillin-Cre + FUT2flox/flox mice, abbreviated as FUT2△IEC). 29 human samples were obtained from the Endoscopy Center of Union Hospital in 2020. All the samples were obtained with the patients’ informed consent, and the samples were processed histologically. ## Azoxymethane (AOM)- and dextran sulfate sodium (DSS)-induced colorectal cancer On Day 0, WT and FUT2△IEC mice were intraperitoneally (IP) injected with 10 mg/kg AOM working solution (1 mg/ml in isotonic saline, diluted from a 10 mg/ml stock solution in dH2O that was stored at − 20 deg C). On Day 7, mice were provided a continuous supply of $2.5\%$ (2.5 g/100 ml) DSS solution in their drinking water for seven days. On Day 14, mice were again given standard drinking water for two weeks. The abovementioned DSS and water steps were repeated on Days 28 and 49 to provide a second and third cycle of DSS administration. Mice were sacrificed on Day 70, and samples were prepared for assessment. ## Virus construction and infection Lentivirus for FUT2 overexpression and shRNA sequences targeting LRP1 were designed and constructed by Obio Technology Co. Ltd., Shanghai, China. OeRNA and shRNA lentivirus was added to the culture medium of GC cells with HitransG A (GeneChem, Shanghai, China), and puromycin was used to screen stable clones. The sequence targeting LRP1 was as follows: CATGCTGGACCTCTCCAATAA. ## RNA Extraction and real-time qPCR Total RNA was extracted with RNAiso Plus reagent (Code No.: 9109, TAKARA, Japan) and reverse transcribed into cDNA using a verse Transcription system (Code No.: RR036A, TAKARA, Japan). Quantitative real-time polymerase chain reaction (qRT‒PCR) was performed using TB Green Premix Ex Taq (Tli RNaseH Plus) (Code No.: RR420A, TAKARA, Japan). The 2-(ΔΔCt) method was used to calculate the relative abundance of RNA after normalization to GAPDH. For HCT116 and SW480 cells, cells were suspended and added to the 6-well plates (1.2 × 106 cells/plate). After 24 h, cells grew to $80\%$ fusion and the total RNA was extracted. The primers that were used for PCR are listed in Additional file 1: Table S1. ## Histology, immunohistochemistry (IHC) and immunofluorescence (IF) staining Tissues were embedded and sectioned. After deparaffinization and rehydration, the sections were boiled for 30 min at 100 °C for antigen retrieval, treated with $3\%$ hydrogen peroxide for 4 min and blocked by incubation with $10\%$ goat serum for 30 min. Subsequent staining with primary antibodies was performed at room temperature for 60 min. After incubation with horseradish peroxidase (HRP)-conjugated secondary antibodies, the sections were washed in distilled water, treated with DAB, counterstained with hematoxylin, and dehydrated. ## Cell migration and invasion assay Cell migration and invasion were measured using 24-well inserts (membrane pore size, 8 μm; Corning Life Sciences, MA, USA). For the migration assay, SW480 and HCT116 cells were suspended in serum-free medium (5 × 104 cells/insert) and added to the upper chamber of the 24-well insert. Medium supplemented with $10\%$ serum was added to the lower chamber. After incubation for 18 h, the cells that migrated were fixed in $4\%$ paraformaldehyde and stained with $1\%$ crystal violet for 10 min. For the invasion assay, chamber membranes were coated with Matrigel (BD Bioscience, San Jose, CA, USA). ## Quantitative proteomic analysis of tandem mass tags (TMT) Colon tissues of mice within both groups(WT and FUT2△IEC mice administered with AMO and DSS) were lysed into peptides. Peptides were labeled with the TMT reagent label and mixed into one sample. Then the mixed sample was separated by liquid chromatography, and the distillate was divided into 12 samples in series. An aliquot of each sample was loaded onto an analytical HPLC column using the auto sampler of an EASY-nLC 1000 HPLC. The results were searched against a decoy database and proteins. ## Immunoprecipitation Immunoprecipitation was performed via a protein A immunoprecipitation kit (Roche) according to the manufacturer’s protocol. SW480 and HCT116 cells were suspended and added to the 10 cm dish. After 24 h, cells grew to $80\%$ fusion and were lysed (lysis buffer: 50 mM Tris, 150 mM NaCl, 2 mM EDTA, protease inhibitor cocktail (Complete Mini, Roche), $0.5\%$ Triton X-100) and incubated for 1 h on ice. The cell lysates containing the protein were incubated overnight with primary antibody-conjugated beads at 4 °C overnight. The beads were eluted with lysis buffer, and the proteins were resuspended in SDS sample buffer and resolved by Western blotting. ## Ulex europaeus agglutinin-I (UEA-I) enrichment assays SW480 and HCT116 cells were transfected with a FUT2 overexpressing or control vector. Cells were suspended and added to the 10 cm dish. After 24 h, cells grew to $80\%$ fusion and were were lysed with lysis buffer (50 mM Tris, 150 mM NaCl, 2 mM EDTA, protease inhibitor cocktail) containing $1\%$ Nonidet P-40. The lysates (1000 μg) were mixed with 100 μl of biotin-conjugated UEA-I (0.1 µg/ml) and incubated with rotation at 4 °C overnight. Then, agarose coupled with streptavidin was added and incubated continuously for 4 h. The samples were extracted and separated by SDS‒PAGE and transferred to PVDF membranes. The membranes were incubated with LRP1, LAMB1 and Integrin β1 antibodies. ## Western blotting HCT116 and SW480 cells were suspended and added to the 6-well plates (1.2 × 106 cells/plate). After 24 h, cells grew to $80\%$ fusion. Proteins were extracted from tissues and CRC cells in ice-cold RIPA lysis buffer containing phosphatase and protease inhibitor mixes (Beyotime Biotechnology, China) at 4 °C for 20 min. The proteins were separated by SDS‒PAGE and transferred to nitrocellulose membranes. The membranes were blocked using $10\%$ milk. The membranes were then incubated overnight with primary antibodies at 4 °C, followed by washing three times in TBST before the addition of secondary antibodies and incubation for 1 h at room temperature. The membranes were further exposed to ECL with Millipore immobilon western chemilum HRP substrate. ## In vivo metastasis assay HCT116 cells transduced with LV-FUT2 or LV-CON lentiviruses were harvested and resuspended at a concentration of 5 × 106 cells per 100 μl and maintained on ice until injection. The cell suspensions were mixed with 100 μl Matrigel (Corning, BD) before injection. Two hundred microliters of cell/Matrigel (1:1) suspensions were then subcutaneously or peritoneally injected into four-week-old male BALB/c nude mice. Mice were sacrificed after 30 days, and peritoneal metastatic nodules were counted. ## Statistical analysis All the experiments were repeated at least three times, and the results are expressed as the means ± SDs. All the statistical analyses were performed using GraphPad Prism software. Student’s t test or one-way analysis of variance (ANOVA) was used to analyze the data, and the chi-square test was used to analyze differences in other variables, as appropriate. A P value of < 0.05 was considered statistically significant for all datasets. (* $p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$). ## FUT2 expression is downregulated in tumor tissues and inversely associated with the prognosis of CRC patients Although glycosylation has been studied in colorectal cancer, these studies have focused on the N- and O-glycosylation of proteins [13]. It has been proposed that fucosylation deficiency leads to colitis and adenocarcinoma in mice [22], while it remains unclear whether α-1,2 fucosylation (FUT1, FUT2) is relevant to colorectal cancer. To examine the relationship between FUT2 and colorectal cancer (CRC), we performed bioinformatics analysis of FUT2 expression profiles in the TCGA database and found that FUT2 was significantly downregulated in CRC (Fig. 1A). Furthermore, we observed significantly decreased FUT2 mRNA expression levels in 29 CRC tissues relative to those in normal colorectal tissues (Fig. 1B). FUT2 protein expression was examined, and significantly lower expression levels of FUT2 were observed in CRC tissues than in corresponding normal tissues (Fig. 1C–D). Moreover, we observed a loss of UEA-I, specific for α-1,2 fucosylation, which is used as a global marker for fucose in CRC tissues (Fig. 1E). Gene Expression Profiling Interactive Analysis (GEPIA) showed that lower FUT2 expression was correlated with lower survival rates (Fig. 1F).Fig. 1FUT2 and altered protein fucosylation are linked with prognosis of CRC patients. A FUT2 mRNA in CRC tissues and normal tissues in TCGA dataset. B FUT2 expression level in 29 CRC tissues and adjacent non-tumor tissues detected by qRT-PCR. C FUT2 expression level in CRC tissues and adjacent non-tumor tissues detected by western blotting. D Quantification of C. **$p \leq 0.01.$ E Representative images of UEA-I lectin fluorescence of CRC tissues and adjacent non-tumor tissues. F The overall survival analysis was plotted using GEPIA for patients with CRC ## FUT2 overexpression inhibits CRC cell EMT, migration and invasion in vitro Downregulation of α-1,2 fucosylation (FUT1, FUT2) and the resulting alteration in protein fucosylation were identified as features of metastatic behavior [20]. Keiichiro et al. demonstrated that FUT2 was significantly suppressed in colorectal cancer HT29 and DLD-1 cells undergoing epithelial-mesenchymal transition (EMT) [23]. EMT has been noted as a critical phenotypic alteration in metastatic cancer cells [24]. To further investigate the role of FUT2 in CRC metastasis, we overexpressed FUT2 in the CRC cell lines HCT116 and SW480, both of which express low endogenous levels of FUT2 (Additional file 2: Fig. S1). The stable overexpression of FUT2 was assessed by real-time qPCR and Western blotting (Fig. 2A–B).Fig. 2FUT2 overexpression inhibits CRC cell EMT and invasiveness in vitro. A FUT2 levels in SW480 and HCT116 cells after overexpression of FUT2 were assessed by qRT-PCR and western blot (B). C Quantification of B. *$p \leq 0.05.$ ** $p \leq 0.01.$ D β-catenin and vimentin levels in SW480 and HCT116 cells after overexpression of FUT2 were assessed by western blot. E Quantification of D. **$p \leq 0.01.$ F Transwell migration and invasion by SW480 and HCT116 cells transduced with FUT2 or Vector. G–H Quantification of F. *$p \leq 0.05$ Further analyses of the FUT2-overexpressing cells indicated a change in EMT. There was a clear increase in the surface expression of β-catenin and a decrease in the expression of the mesenchymal marker vimentin in FUT2-overexpressing cells (Fig. 2C–E). Moreover, FUT2 overexpression attenuated the migratory and invasive capacity of HCT116 and SW480 cells (Fig. 2F–H). Together, these observations suggest that FUT2 is a regulator of the EMT program and metastatic capacity in colorectal cancer cells. ## FUT2 overexpression decreases CRC EMT and metastasis in vivo Next, we investigated whether FUT2 overexpression impairs aggressive behavior in vivo using a xenograft model of metastasis. HCT116 cells stably transduced with LV-FUT2 or LV-NC were transplanted intraperitoneally into nude mice, and tumor growth with the occurrence of peritoneal metastases was monitored. Similar to our in vitro observations, analyses of the peritoneal metastases revealed that the tumors derived from FUT2-overexpressing cells exhibited impaired aggression, as evidenced by fewer peritoneal metastatic nodules (Fig. 3A–B). The observation of upregulated FUT2 transcripts by qRT‒PCR confirmed the effective overexpression of FUT2 by both LVRNAs in tumors (Fig. 3C). IHC staining of sections of the xenograft tumors derived from FUT2-overexpressing cells revealed decreased expression of N-cadherin and vimentin as well as increased expression of E-cadherin and β-catenin (Fig. 3D–E). Mice injected with FUT2-overexpressing HCT116 cells exhibited significantly reduced metastases in the lungs and livers (Fig. 3F–H). Together, these data demonstrate that FUT2 overexpression inhibits the metastatic capacity of CRC in vivo. Fig. 3FUT2 overexpression decreases in vivo CRC EMT and metastasis. A Representative images of peritoneal metastases of nude mice injected with HCT116 cells transduced with LV-FUT2 or LV-NC. B Quantification of A. **$p \leq 0.01.$ C FUT2 expression in tumors was assessed by western bloting. D Representative images of IHC staining with N-cadherin, E-cadherin, β-catenin and vimentin antibody in xenograft tumor sections from FUT2 overexpression cells. E Quantification of D. *$p \leq 0.05.$ ** $p \leq 0.01.$ F H&E-stained images of mouse lungs and livers at end-point. G–H Quantification of F. **$p \leq 0.01.$ *** $p \leq 0.001$ ## FUT2-knockout mice develop AMO- and DSS-induced tumors Intestinal-specific FUT2-knockout mice (FUT2△IEC mice) were generated by hybridizing FUT2-knockout mice with Villin-CRE transgenic mice. qRT‒PCR analysis of colon lysates from WT and fut2 − / − mice confirmed the reduction or absence of FUT2 expression (Fig. 4A). UEA-I lectin fluorescence staining confirmed the decrease in α-1,2 fucosylation in FUT2△IEC mice (Fig. 4B). We then examined whether FUT2 knockout led to the development of DSS- and AMO-induced tumors. Mice within both groups (WT and FUT2△IEC mice) were subdivided into two groups and treated with or without AMO and DSS. FUT2△IEC mice developed significantly increased established colon tumors in comparison to WT mice 90 days after AMO and DSS treatment (Fig. 4C–E). To better study the role of FUT2 in colorectal cancer, colons were analyzed for the expression of EMT markers. IHC staining and Western blotting confirmed higher EMT in FUT2△IEC mice treated with AMO and DSS than in WT mice treated with AMO and DSS, as evidenced by increased vimentin expression and decreased β-catenin expression (Fig. 4F–G). In summary, our results demonstrate that FUT2-knockout mice develop more AMO- and DSS-induced tumors and exhibit greater colorectal cancer EMT.Fig. 4FUT2 knock-out mice develop AMO and DSS-induced tumors. A FUT2 levels in colon lysates of WT mice and FUT2 knock-out mice were detected by qRT-PCR. B Representative images of UEA-I lectin fluorescence staining revealed α-1,2 fucosylation in WT mice and Fut2△IEC mice. C Representative images of H&E-stained colon sections after AMO and DSS administered. D–E Quantification of C. *$p \leq 0.05.$ ** $p \leq 0.01.$ F Representative images of IHC staining with N-cadherin and E-cadherin in colon lysates in colon sections in WT mice and FUT2 knock-out mice after AMO and DSS administered. G FUT2 knock-out increased N-cadherin and vimentin, as shown by Western blotting ## 1.2-fucosylated glycoproteins reveal regulators of EMT and metastasis FUT2 can generate α-1,2 fucosylated structures both at the termini of N-acetyllactosamine and on galactose linked with N-acetylgalactosamine, which are epitopes recognized by UEA-I and TJA-II/SNA-II lectins [20, 25]. To identify α-1,2 fucosylated proteins that could mediate the effects of FUT2 dysregulation on colorectal cancer, we performed a quantitative proteomics study of total protein and N-glycosylated TMT markers from colon tissues of mice in both groups (WT and FUT2△IEC mice). A total of 324 candidate downregulated α-1,2 fucosylated proteins were identified in the proteomics study of N-glycosylation in FUT2△IEC mice. Enrichment (Gene Ontology; GO) analysis of genes was performed and revealed enrichment in biological processes relevant to metastasis, such as cell migration and angiogenesis. The proteins that bound to UEA-I and were related to migration include laminin β1 (LAMB1), laminin β2 and laminin γ1, integrin-αV, integrin β1, signal-regulatory protein alpha (Sirpa) and Pro-low-density lipoprotein receptor-related protein 1 (LRP1) (Additional file 3: Table S2 and Fig. 5A–B). The glycosylation state of these proteins in colorectal cancer cells was further verified by UEA-I lectin enrichment followed by Western blotting. The results demonstrated increased integrin β1, LAMB1 and LRP1 levels in FUT2-overexpressing SW480 and HCT116 cells compared to LV-CON cells, consistent with lower α-1,2 fucosylation on those proteins in FUT2△IEC mice, while there was no difference in the expression of these proteins in the input samples (Fig. 5C–D). Additionally, immunoprecipitation (IP) of LRP1 followed by UEA-I blotting showed increased UEA-I binding to LRP1 proteins in FUT2-overexpressing CRC cells (Fig. 5E–F). After SGN-2FF treatment, the observed differences disappeared because SGN-2FF removed all forms of fucosylation (Fig. 5G). Overall, our results suggest that LRP1 was α-1,2 fucosylated, which was consistent with our quantitative proteomics study. Fig. 51.2- fucosylated glycoproteins reveals regulators of EMT and metastasis. A GO enrichment analysis of N-fucosylated proteins in WT and Fut2△IEC mice based on biological processes. B GO enrichment analysis (category of functional categories) of N-fucosylated proteins in WT and Fut2△IEC mice. C Validation of 1,2 fucosylation in proteins bound with UEA-I related to migration in SW480 cells. D Validation of 1,2 fucosylation in proteins bound with UEA-I related to migration in HCT116 cells. E IP of LRP1 from cell lysates of SW480 cells transduced with LV-FUT2 or LV-NC. F IP of LRP1 from cell lysates of HCT116 cells transduced with LV-FUT2 or LV-NC. G LRP1 immunoprecipitation from cell lysates of HCT116 cells transfected with LV-FUT2 or LV-NC. Anti-LRP1 immunoprecipitates were treated with or without SGN-2FF and blotted with UEA-I or α-LRP1. ## LRP1 is a major mediator of the suppressive role of FUT2 in metastasis LRP1 is a highly glycosylated cysteine-rich protein. Numerous studies have suggested a role for LRP1 in the regulation of cell invasion and migration in several cancers [26, 27], and it may suppress CRC progression [28]. Furthermore, bioinformatics analysis of LRP1 expression profiles in the TCGA database indicated that LRP1 was expressed at lower levels in CRC than in noncancerous colon tissues (Fig. 6A). Therefore, it is a possible candidate for the effects of FUT2 in suppressing CRC invasion. We next investigated whether LRP1 silencing accelerated cell invasion and migration in HCT116 cells overexpressing FUT2. Our data suggested that LRP1 exhibits increased ɑ-1,2 fucosylation in FUT2-overexpressing HCT116 cells (Fig. 6B). Furthermore, FUT2 overexpression increased ɑ-1,2 fucosylation and neutralized the pro-invasive effects of LRP1 silencing (Fig. 6C–E). In addition, silencing LRP1 counteracted the inhibitory effects of FUT2 overexpression on EMT (Fig. 6F–G). Our data suggest that decreased ɑ-1,2 fucosylation is critical for the effects of LRP1 in suppressing the invasive phenotype of CRC cells. Fig. 6LRP1 is a major mediator of the suppressing metastatic role by FUT2. A Downregulation of LRP1 mRNA in CRC tissues than normal tissues in TCGA dataset. B LRP1 IP on cell lysates of HCT116 cells stably overexpressing FUT2 or control vector and transfected with NC or LRP1 shRNA. Anti-LRP1 immunoprecipitates were blotted with UEA-I or α-LRP1. C Transwell migration and invasion assays by HCT116 cells stably overexpressing FUT2 or control vector and transfected with NC or LRP1 shRNA. D–E Quantification of C. F β-catenin and vimentin levels in HCT116 cells stably overexpressing FUT2 or control vector and transfected with NC or LRP1 shRNA were assessed by western blot. G Quantification of F ## Discussion Modifications of glycosylation at the surface of tumor cells and tissue induce unique features that are characteristic of cancer cells and the tumor microenvironment [13]. Several studies have revealed that aberrant glycosylation plays important roles in tumor progression and malignant transformation [29]. One of the most common modifications of glycans is the attachment of fucoses via the action of various fucosyltransferases [13]. Nakayama et al. described a novel metastatic pathway that is dependent on the loss of fucosylation in colorectal cancer [15]. Recent studies indicate that carcinogenesis in a subset of colorectal cancer in mice occurs due to a molecular mechanism driven by fucosylation deficiency [22]. There is currently limited understanding of which enzymes and related fucosylation modifications are important in the progression and metastasis of colorectal cancer. The present study demonstrates how ɑ-1,2 fucosylation, mediated by FUT2, impacts CRC progression and metastasis. FUT2, an enzyme governing epithelial α-1,2 fucosylation, is associated with various human disorders [30]. FUT2 mediates the addition of L-fucose via an α-1–2 linkage to the terminal β-D-galactose residues of mucosal glycans, including type 1 or 2 N-acetyllactosamine [31]. Recent reports have shown that FUT2 is preferentially expressed in epithelial cells (ECs) of the gastrointestinal tract [32] and that α-1,2 fucose on enterocytes is specifically regulated by FUT2 [33]. Several studies have revealed that defects in epithelial FUT2 predispose individuals to several diseases [30], including inflammatory bowel disease (IBD, especially Crohn’s disease) [34], acute and chronic inflammatory disorders such as type I diabetes [35], and chronic pancreatitis [36]. However, the role of changes in fucosylation induced by FUT2 in colorectal cancer remains unclear. Our results revealed a loss of α-1,2 fucosylation (UEA-I) in human colorectal cancer tissues. These data are consistent with a previous study showing that fucosylation deficiency in mice leads to colitis and adenocarcinoma [22]. We observed that FUT2 knockout resulted in increased DSS-induced colon tumors and EMT in FUT2△IEC mice, indicating the inhibitory effects of FUT2 and α-1,2 fucosylation on the progression and metastasis of colorectal cancer. Lau et al. identified a role of α-1,2 fucose in suppressing the metastatic potential of melanoma and correlated it with higher survival rates [37]. Our results are consistent with this finding and highlight the inhibitory effects of FUT2 on EMT in colorectal cancer. Next, we focused functional and mechanistic studies of FUT2. FUT2 was significantly suppressed during the EGF- or bFGF-triggered EMT of colorectal cancer cells [23]. However, the role of this enzyme and the α-1,2 fucosylation it causes in colorectal cancer is still unknown. FUT2 overexpression in the colorectal cancer cell lines HCT116 and SW480 attenuated their migratory and invasive capacities. This is in contrast to results in breast cancer cell lines [21] and lung adenocarcinoma cells [38], where FUT2 enhances cell migration and invasion. Overexpressing FUT2 reduced EMT in colorectal cancer cells, suggesting that α-1,2 fucosylation may inhibit EMT. Similar results were obtained in vivo: FUT2 overexpression reduced metastatic dissemination of colorectal cancer cells to the peritoneum and inhibited EMT in vivo. Together with these results, our present study demonstrates that loss of FUT2 and α-1,2 fucosylation may promote EMT and metastasis in colorectal cancer. One of the most important aspects of our study is related to the involvement of LRP1 in the regulation of EMT and metastasis by FUT2. We identified LRP1 as a mediator of the invasive-inhibiting effects of FUT2 using a proteomics study and immunoprecipitation. LRP1 is a cell surface receptor involved in invasion and neovascularization, processes that drive tumor progression and metastasis [39]. Previous studies have shown that LRP1 expression is significantly lower in colon adenocarcinoma cells than in colon mucosa and stromal cells [40] and is associated with worse colorectal cancer outcomes [28]. Consistent with these previous studies, we found that LRP1 was downregulated in CRC tissues compared to nontumor tissues. Overexpressing FUT2 reduced the EMT and invasion of colorectal cancer cells. However, these invasive-inhibiting effects were mostly abrogated by silencing LRP1. Our results demonstrated that α-1,2 fucosylation was crucial for the inhibition of EMT and invasion by LRP1 in colorectal cancer cells. Taken together, our results highlight the inhibitory potential of FUT2 in colorectal cancer. In addition, we provide a number of solid arguments for a regulatory role of FUT2 and the α-1,2 fucosylation it causes in the function of LRP1. These data may not only enrich our knowledge of fucosylation but also suggest a strategy for therapeutic intervention in colorectal cancer. ## Conclusions Our study demonstrated that FUT2 induces α-1,2 fucosylation and inhibits the EMT and metastasis of colorectal cancer through LRP1 fucosylation, suggesting that FUT2 may serve as a therapeutic target for colorectal cancer (Additional file 4). ## Supplementary Information Additional file 1. Table S1. Primers for qRT-PCR.Additional file 2. Figure S1. Expression of FUT2 gene in CRC cell lines. Additional file 3. Table S2. Mass spectrometric analysis of N-glycosylated TMT proteins in migration category of GO analysis from colon tissues of mice in WT and FUT2△IEC mice. Additional file 4. Uncropped images of western blot analysis. ## References 1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A. **Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries**. *CA Cancer J Clin* (2021) **71** 209-49. DOI: 10.3322/caac.21660 2. Ihn MH, Kim DW, Cho S, Oh HK, Jheon S, Kim K. **Curative resection for metachronous pulmonary metastases from colorectal cancer: analysis of survival rates and prognostic factors**. *Cancer Res Treat* (2017) **49** 104-115. DOI: 10.4143/crt.2015.367 3. Chandler KB, Costello CE, Rahimi N. **Glycosylation in the tumor microenvironment: implications for tumor angiogenesis and metastasis**. *Cells* (2019) **8** 544. DOI: 10.3390/cells8060544 4. Stowell SR, Ju T, Cummings RD. **Protein glycosylation in cancer**. *Annu Rev Pathol* (2015) **10** 473-510. DOI: 10.1146/annurev-pathol-012414-040438 5. Nguyen AT, Chia J, Ros M, Hui KM, Saltel F, Bard F. **Organelle specific O-glycosylation drives MMP14 activation, tumor growth, and metastasis**. *Cancer Cell* (2017) **32** 639-653 e6. DOI: 10.1016/j.ccell.2017.10.001 6. Cheng C, Ru P, Geng F, Liu J, Yoo JY, Wu X. **Glucose-mediated N-glycosylation of SCAP is essential for SREBP-1 activation and tumor growth**. *Cancer Cell* (2015) **28** 569-581. DOI: 10.1016/j.ccell.2015.09.021 7. Laubli H, Borsig L. **Altered cell adhesion and glycosylation promote cancer immune suppression and metastasis**. *Front Immunol* (2019) **10** 2120. DOI: 10.3389/fimmu.2019.02120 8. Hauselmann I, Borsig L. **Altered tumor-cell glycosylation promotes metastasis**. *Front Oncol* (2014) **4** 28. DOI: 10.3389/fonc.2014.00028 9. Jassam SA, Maherally Z, Ashkan K, Pilkington GJ, Fillmore HL. **Fucosyltransferase 4 and 7 mediates adhesion of non-small cell lung cancer cells to brain-derived endothelial cells and results in modification of the blood-brain-barrier: in vitro investigation of CD15 and CD15s in lung-to-brain metastasis**. *J Neurooncol* (2019) **143** 405-415. DOI: 10.1007/s11060-019-03188-x 10. Thomas D, Rathinavel AK, Radhakrishnan P. **Altered glycosylation in cancer: a promising target for biomarkers and therapeutics**. *Biochim Biophys Acta* (2021) **1875** 188464 11. Silsirivanit A. **Glycosylation markers in cancer**. *Adv Clin Chem* (2019) **89** 189-213. DOI: 10.1016/bs.acc.2018.12.005 12. Oliveira-Ferrer L, Legler K, Milde-Langosch K. **Role of protein glycosylation in cancer metastasis**. *Semin Cancer Biol* (2017) **44** 141-152. DOI: 10.1016/j.semcancer.2017.03.002 13. Holst S, Wuhrer M, Rombouts Y. **Glycosylation characteristics of colorectal cancer**. *Adv Cancer Res* (2015) **126** 203-256. DOI: 10.1016/bs.acr.2014.11.004 14. Blanas A, Zaal A, van der Haar Avila I, Kempers M, Kruijssen L, de Kok M. **FUT9-driven programming of colon cancer cells towards a stem cell-like state**. *Cancers* (2020) **12** 2580. DOI: 10.3390/cancers12092580 15. Nakayama K, Moriwaki K, Imai T, Shinzaki S, Kamada Y, Murata K. **Mutation of GDP-mannose-4,6-dehydratase in colorectal cancer metastasis**. *PLoS ONE* (2013) **8** e70298. DOI: 10.1371/journal.pone.0070298 16. Scupakova K, Adelaja OT, Balluff B, Ayyappan V, Tressler CM, Jenkinson NM. **Clinical importance of high-mannose, fucosylated, and complex N-glycans in breast cancer metastasis**. *JCI Insight* (2021). DOI: 10.1172/jci.insight.146945 17. Li J, Hsu HC, Mountz JD, Allen JG. **Unmasking fucosylation: from cell adhesion to immune system regulation and diseases**. *Cell Chem Biol* (2018) **25** 499-512. DOI: 10.1016/j.chembiol.2018.02.005 18. Goupille C, Marionneau S, Bureau V, Hallouin F, Meichenin M, Rocher J. **alpha1,2Fucosyltransferase increases resistance to apoptosis of rat colon carcinoma cells**. *Glycobiology* (2000) **10** 375-382. DOI: 10.1093/glycob/10.4.375 19. Tan KP, Ho MY, Cho HC, Yu J, Hung JT, Yu AL. **Fucosylation of LAMP-1 and LAMP-2 by FUT1 correlates with lysosomal positioning and autophagic flux of breast cancer cells**. *Cell Death Dis* (2016) **7** e2347. DOI: 10.1038/cddis.2016.243 20. Agrawal P, Fontanals-Cirera B, Sokolova E, Jacob S, Vaiana CA, Argibay D. **A systems biology approach identifies FUT8 as a driver of melanoma metastasis**. *Cancer cell* (2017) **31** 804-819 e7. DOI: 10.1016/j.ccell.2017.05.007 21. Lai TY, Chen IJ, Lin RJ, Liao GS, Yeo HL, Ho CL. **Fucosyltransferase 1 and 2 play pivotal roles in breast cancer cells**. *Cell Death Discovery* (2019) **5** 74. DOI: 10.1038/s41420-019-0145-y 22. Wang Y, Huang D, Chen KY, Cui M, Wang W, Huang X. **Fucosylation deficiency in mice leads to colitis and adenocarcinoma**. *Gastroenterology* (2017) **152** 193-205 e10. DOI: 10.1053/j.gastro.2016.09.004 23. Sakuma K, Aoki M, Kannagi R. **Transcription factors c-Myc and CDX2 mediate E-selectin ligand expression in colon cancer cells undergoing EGF/bFGF-induced epithelial-mesenchymal transition**. *Proc Natl Acad Sci USA* (2012) **109** 7776-7781. DOI: 10.1073/pnas.1111135109 24. Pastushenko I, Blanpain C. **EMT transition states during tumor progression and metastasis**. *Trends Cell Biol* (2019) **29** 212-226. DOI: 10.1016/j.tcb.2018.12.001 25. Saku A, Hirose K, Ito T, Iwata A, Sato T, Kaji H. **Fucosyltransferase 2 induces lung epithelial fucosylation and exacerbates house dust mite-induced airway inflammation**. *J Aller Clin Immunol* (2019) **144** 698-709. DOI: 10.1016/j.jaci.2019.05.010 26. Pencheva N, Tran H, Buss C, Huh D, Drobnjak M, Busam K. **Convergent multi-miRNA targeting of ApoE drives LRP1/LRP8-dependent melanoma metastasis and angiogenesis**. *Cell* (2012) **151** 1068-1082. DOI: 10.1016/j.cell.2012.10.028 27. Dedieu S, Langlois B, Devy J, Sid B, Henriet P, Sartelet H. **LRP-1 silencing prevents malignant cell invasion despite increased pericellular proteolytic activities**. *Mol Cell Biol* (2008) **28** 2980-2995. DOI: 10.1128/MCB.02238-07 28. Boulagnon-Rombi C, Schneider C, Leandri C, Jeanne A, Grybek V, Bressenot AM. **LRP1 expression in colon cancer predicts clinical outcome**. *Oncotarget* (2018) **9** 8849-8869. DOI: 10.18632/oncotarget.24225 29. Mereiter S, Balmana M, Campos D, Gomes J, Reis CA. **Glycosylation in the era of cancer-targeted therapy: where are we heading?**. *Cancer Cell* (2019) **36** 6-16. DOI: 10.1016/j.ccell.2019.06.006 30. Goto Y, Uematsu S, Kiyono H. **Epithelial glycosylation in gut homeostasis and inflammation**. *Nat Immunol* (2016) **17** 1244-1251. DOI: 10.1038/ni.3587 31. Rouquier S, Lowe JB, Kelly RJ, Fertitta AL, Lennon GG, Giorgi D. **Molecular cloning of a human genomic region containing the H blood group alpha(1,2)fucosyltransferase gene and two H locus-related DNA restriction fragments Isolation of a candidate for the human Secretor blood group locus**. *J Biol Chem* (1995) **270** 4632-4639. DOI: 10.1074/jbc.270.9.4632 32. Domino SE, Zhang L, Lowe JB. **Molecular cloning, genomic mapping, and expression of two secretor blood group alpha (1,2)fucosyltransferase genes differentially regulated in mouse uterine epithelium and gastrointestinal tract**. *J Biol Chem* (2001) **276** 23748-23756. DOI: 10.1074/jbc.M100735200 33. Goto Y, Obata T, Kunisawa J, Sato S, Ivanov II, Lamichhane A. **Innate lymphoid cells regulate intestinal epithelial cell glycosylation**. *Science* (2014) **345** 1254009. DOI: 10.1126/science.1254009 34. McGovern DP, Jones MR, Taylor KD, Marciante K, Yan X, Dubinsky M. **Fucosyltransferase 2 (FUT2) non-secretor status is associated with Crohn's disease**. *Hum Mol Genet* (2010) **19** 3468-3476. DOI: 10.1093/hmg/ddq248 35. Smyth DJ, Cooper JD, Howson JM, Clarke P, Downes K, Mistry T. **FUT2 nonsecretor status links type 1 diabetes susceptibility and resistance to infection**. *Diabetes* (2011) **60** 3081-3084. DOI: 10.2337/db11-0638 36. Aghdassi AA, Weiss FU, Mayerle J, Lerch MM, Simon P. **Genetic susceptibility factors for alcohol-induced chronic pancreatitis**. *Pancreatol Off J Int Assoc Pancreatol* (2015) **15** S23-31. DOI: 10.1016/j.pan.2015.05.476 37. Lau E, Feng Y, Claps G, Fukuda MN, Perlina A, Donn D. **The transcription factor ATF2 promotes melanoma metastasis by suppressing protein fucosylation**. *Sci Signal* (2015) **8** ra124. DOI: 10.1126/scisignal.aac6479 38. Deng G, Chen L, Zhang Y, Fan S, Li W, Lu J. **Fucosyltransferase 2 induced epithelial-mesenchymal transition via TGF-beta/Smad signaling pathway in lung adenocarcinaoma**. *Exp Cell Res* (2018) **370** 613-622. DOI: 10.1016/j.yexcr.2018.07.026 39. Boye K, Pujol N, Ica DA, Chen YP, Daubon T, Lee YZ. **The role of CXCR3/LRP1 cross-talk in the invasion of primary brain tumors**. *Nat Commun* (2017) **8** 1571. DOI: 10.1038/s41467-017-01686-y 40. Obermeyer K, Krueger S, Peters B, Falkenberg B, Roessner A, Rocken C. **The expression of low density lipoprotein receptor-related protein in colorectal carcinoma**. *Oncol Rep* (2007) **17** 361-367. PMID: 17203175
--- title: Daurisoline attenuates H2O2-induced chondrocyte autophagy by activating the PI3K/Akt/mTOR signaling pathway authors: - Yang Zhang - Wenguang Liu - Zhonghao Liu - Yi Liu journal: Journal of Orthopaedic Surgery and Research year: 2023 pmcid: PMC10041752 doi: 10.1186/s13018-023-03717-5 license: CC BY 4.0 --- # Daurisoline attenuates H2O2-induced chondrocyte autophagy by activating the PI3K/Akt/mTOR signaling pathway ## Abstract ### Background Osteoarthritis (OA) is a chronic degenerative joint disease characterized by cartilage degeneration and intra-articular inflammation. Daurisoline (DAS) is an isoquinoline alkaloid isolated from Rhizoma Menispermi, whose antitumor and anti-inflammatory pharmacological effects have been demonstrated, but the effects of DAS on OA have rarely been researched. In this study, we aimed to explore the potential role of DAS in OA and its partial mechanism. ### Materials and methods The cytotoxicity of H2O2 and DAS toward chondrocytes was detected by the Cell Counting Kit-8 assay. Safranin O staining was used to detect chondrocyte phenotype changes. Cell apoptosis was measured by both flow cytometry and quantitative analysis of the protein levels of the apoptosis-related factors Bax, Bcl-2 and cleaved caspase-3 by western blot. Western blotting and immunofluorescence were used to assess the expression of the autophagy-related proteins LC3, Beclin-1 and p62. In addition, key signal pathway targets and matrix-degrading indicators were measured by western blot. ### Results Our results indicated that H2O2 induced human chondrocyte apoptosis and activated autophagy in a dose-dependent manner. DAS treatment dose-dependently reversed the expression of apoptosis-related proteins (Bax, Bcl-2 and cleaved caspase3) and the apoptosis rate induced by H2O2. Western blot and immunofluorescence analyses showed that DAS decreased the H2O2-induced upregulation of the autophagy marker Beclin-1 and the LC3 II/LC3 I ratio and upregulated the p62 protein level. Mechanistically, DAS inhibited autophagy through the activation of the classical PI3K/AKT/mTOR signaling pathway and protected chondrocytes from apoptosis. In addition, DAS alleviated the H2O2-induced degradation of type II collagen and the high expression of matrix metalloproteinase 3 (MMP3) and MMP13. ### Conclusion Our research demonstrated that DAS alleviated chondrocyte autophagy caused by H2O2 through activation of the PI3K/AKT/mTOR signaling pathway and protected chondrocytes from apoptosis and matrix degradation. In conclusion, these findings suggest that DAS may serve as a promising therapeutic strategy for OA. ## Introduction With the acceleration of the global population aging trend, OA has become nearly universal among the middle and elderly worldwide [1, 2]. As a type of chronic degenerative joint disease, OA is characterized by progressive articular cartilage degeneration associated with multiple factors, such as metabolic disorders of the cartilage matrix, chondrocyte apoptosis, subchondral bone remodeling, and intra-articular inflammation [3–5]. The aggravation of the disease leads to total cartilage loss and destruction and ultimately causes limited movement, joint dysfunction, and even disability [6, 7]. Although some medications and physical therapy can alleviate OA symptoms and slow disease progression by improving cartilage metabolism and repairing cartilage, no effective treatment can reverse the progression of the disease [8, 9]. Therefore, it is imperative to research and develop more effective treatments. Autophagy, which is closely related to apoptosis, is an intracellular metabolic pathway of self-degradation that is highly conserved. Autophagy provides extra energy for cells to combat antagonistic conditions, maintain intracellular homeostasis, and regulate the physiological function of cells by eliminating abnormal organelles and unnecessary proteins [10–12]. Autophagy is activated under oxidative stress, but excessive oxidative stress that exceeds the tolerance of cells will cause dysfunctional autophagy [13]. Unrestrained autophagy may lead to autophagic injury and cell apoptosis and even aggravate disease progression [14–16]. Autophagy is involved in physiological processes such as apoptosis regulation and energy metabolism in chondrocytes. In the initial phase of OA, autophagy acts as an adaptable reaction to prevent cell damage. However, excessive autophagy can be activated along with apoptosis as a substitute for cell death in the later period of OA [17]. Autophagic cell death promotes chondrocyte apoptosis and further exacerbates cartilage degeneration. Therefore, chondrocyte therapy based on the regulation of autophagy might be an essential strategy for OA treatment. Plant-derived traditional Chinese medicine has attracted increasing attention by virtue of its efficacy and safety [18, 19]. Daurisoline (DAS) is an isoquinoline extracted from the Chinese herbal medicine Rhizoma Menispermi. Earlier research has shown the potential pharmacological effects of DAS for treating certain diseases, such as its modulation of focal ischemia/reperfusion injury, platelet aggregation, and arrhythmia [20, 21]. Furthermore, DAS has potential autophagy inhibition and exhibits anti-inflammatory and antitumor activity [22–24]. The potential bioactivity of DAS on OA, especially for autophagy, and its underlying mechanisms remain uncertain. This study established a chondrocyte damage model by treating chondrocytes with H2O2, aiming to explore the role of DAS in antioxidative autophagy and elucidate some of its mechanism of action. Furthermore, the effect of DAS on articular cartilage degeneration was investigated for the first time in vitro and in vivo, hoping to provide a new therapeutic strategy for OA. ## Patients and ethics statement In this study, 20 patients with OA who underwent knee arthroplasty surgery at the Second Hospital of Shandong University from May 2020 to June 2021 were selected for primary human chondrocyte extraction. OA was diagnosed according to the American College of Rheumatology (ACR) knee OA criteria, and it was classified based on the Kellgren-Lawrence scoring system. The number of patients with each grade was 12 with grade 2 and 8 with grade 3. Our study was approved by the Ethics Committee of the Second Hospital of Shandong University (approval number: KYLL-2020LKJOA-0026). In addition, all patients signed the informed consent form. ## Primary human chondrocyte extraction and culture Chondrocytes were isolated from cartilage tissue samples taken from all 20 patients with OA included in the study. We only selected macroscopic healthy specimens around the joint wear surface for cell extraction [25, 26]. Briefly, the cartilage tissues were washed with PBS 3 times to remove residual blood. Next, we cut the cartilage tissue into pieces of approximately 1 mm3 with ophthalmic scissors, placed they into a centrifuge tube, and added PBS to wash they three times. The cartilage tissue was digested with $0.2\%$ type II collagenase overnight at 37 °C. The cells were separated from the digested cartilage tissue with a 200-mesh filter. After centrifugation of the cell suspension with the supernatant removed, chondrocytes were seeded in DMEM/F12 medium containing $10\%$ fetal bovine serum (HyClone, USA) and incubated at 37 °C in an incubator with $5\%$ CO2. After a 3-day incubation, the cell growth status was investigated, and the medium was replaced. The culture medium was changed every 3 days. The cells were passaged when they had grown to $80\%$ to $90\%$ confluency. The chondrocytes were used in experiments after at least three passages. ## Cell viability analysis The cytotoxicity of H2O2 and DAS on chondrocytes was detected by the Cell Counting Kit-8 (CCK-8; Beyotime, Beijing, China) assay. Primary human chondrocytes (3000 cells per well) were seeded in a 96-well plate for 24 h for preincubation. First, H2O2 (Aladdin, Shanghai, China) and DAS (Yuanye Bio-technology Co, Shanghai, China, cat. no. B20095) were used to stimulate cells, and then 10 μl of CCK-8 dye was added to each well, followed by incubation in a 96-well plate at 37 °C protected from light for 2–3 h. The absorbance value (OD value) of each well at a wavelength of 450 nm was calculated by a microplate reader (Bio-Rad, USA). ## Safranin O staining Safranin O staining was used to detect chondrocyte phenotype changes. Briefly, chondrocytes were seeded in 6-well plates, prestimulated with H2O2 (200 μM) for 4 h, and then treated with DAS (2.5 and 5 μM) for 24 h. Next, the cells were fixed with $4\%$ paraformaldehyde at room temperature for 30 min. After being washed with PBS three times, the cells were stained with safranin O solution (Solarbio, Beijing, China) at a concentration of $0.1\%$ for 5–10 min. Finally, the cells were washed with PBS three times and then imaged by light microscopy (Leica, Germany, magnification ×100). The statistical processing was done with ImageJ, and values were normalized to the control group. ## Chondrocyte treatment Chondrocytes were seeded in 96-well or 6-well plates and treated with different concentrations of H2O2 (0, 10, 50, 100, 200, 300, 400 and 500 μM) and DAS (0, 1, 2.5, 5, 10, 15 and 20 μM) for 4 h and 24 h to detect cell viability and chondrocyte phenotypic changes. Based on these results, for experiments the cells were pretreated with DAS (2.5 and 5 μM) for 24 h and stimulated with H2O2 (200 μM) for 4 h. In addition, cells were collected for a western blot assay to determine the levels of autophagy markers and apoptosis-related factors, and immunofluorescence was performed to test Beclin-1. To detect the involved signal pathways, we divided the cells into 4 groups that underwent different treatments: control group, H2O2, H2O2 + DAS and H2O2 + IGF-1 as a PI3K/Akt/mTOR pathway activator to stimulate chondrocytes. We detected changes in related factors by western blot. ## Protein extraction and western blot After appropriate treatment, cells were lysed for total protein extraction by RIPA Lysis Buffer (Beyotime) containing protease and phosphatase inhibitors, and the protein concentration was measured using a BCA assay kit (Beyotime, Beijing, China). Then, 20 μg of whole proteins from each group was loaded into $10\%$ or $12\%$ sodium dodecyl sulfate‒polyacrylamide gel electrophoresis (SDS-PAGE) gels for separation and transferred to polyvinylidene difluoride (PVDF) membranes. The membranes were blocked with TBST supplemented with $5\%$ nonfat dried milk and incubated with primary antibodies at 4 °C overnight. Primary antibodies against LC3B (1:1000, CST), p62 (1:2000, Abcam), Beclin-1 (1:1000, CST), BAX (1:1000, Abcam), Bcl-2 (1:2000, Abcam), caspase-3 (1:1000, CST), PI3K (1:500, CST), p-PI3K (1:500, CST), AKT (1:500, CST), p-AKT (1:500, CST), mTOR (1:500, CST), p-mTOR (1:500, CST), Collagen II (1:500, Abcam), MMP-3 (1:1000, CST) and MMP-13 (1:1000, CST) were used in this study. After washing the membranes three times, the membranes were incubated with the relevant secondary antibodies for 1 h at room temperature. The protein bands were evaluated by an enhanced ECL kit (Millipore, USA) and a Chemiluminescence Imaging System (Tanon-4800, Shanghai, China), and the data were analyzed by ImageJ software (National Institutes of Health, USA). ## Flow cytometry Apoptosis of chondrocytes was evaluated by an annexin V-fluorescein isothiocyanate (FITC)/PI kit (Bestbio, China). Briefly, chondrocytes were seeded into 6-well plates and treated for 6 h with serum-free DMEM to ensure synchronization. Next, chondrocytes were treated accordingly with the indicated concentrations of H2O2 and DAS. After treating the cells with $0.25\%$ trypsin, the cells were collected, rinsed twice with PBS, centrifuged, and resuspended in 100 μl binding buffer. The cell suspension was then thoroughly mixed with 5 μl of annexin V-FITC staining solution and 5 μl propidium iodide (PI) and kept in the dark at room temperature for 15 min, followed by the addition of 400 μl of 1× binding buffer. Cell apoptosis was detected by flow cytometry (BD Accuri™ C6 Plus) at a wavelength of 488 nm. The data were analyzed using FlowJo V10.8.1 software (Tree Star). ## Immunofluorescence Chondrocytes were seeded into 20-mm confocal Petri dishes and starved with serum-free DMEM for 24 h after their confluency reached $80\%$. We then randomly divided chondrocytes into H2O2, H2O2 + DAS, and control groups. After washing three times with PBS, the cells were fixed for 15 min with $4\%$ paraformaldehyde and then permeabilized with PBS containing $0.3\%$ Triton X-100 for 15 min, followed by a PBS wash. At room temperature, the cells were blocked with $5\%$ BSA for 1 h. Afterward, the samples were incubated with the rabbit anti-human Beclin-1 antibody (Abcam, Britain; 1:200) at 4 °C overnight. Next, the samples were washed with PBS 3 times and incubated with fluorescein isothiocyanate-conjugated secondary antibody for 1 h at room temperature in the dark. Finally, we stained the cell nuclei with DAPI in darkness for 5 min and sealed the slides with an anti-fluorescent quenching agent. Immunofluorescence analysis was performed with a Leica DMi8 fluorescence microscope. The intensity of fluorescence was measured using ImageJ and normalized to the DAPI signal. ## Statistical analysis Experimental data were analyzed using GraphPad Prism 8.0 (GraphPad Software Inc., San Diego, California, USA), and the measurements are expressed as mean ± standard deviation (SD). The two groups were compared using Student’s t test. One-way ANOVA was used for intergroup comparisons of > 3 groups. At least three independent replicate experiments were done to obtain all data. Differences were statistically significant at $P \leq 0.05$ and are marked with *$P \leq 0.05$, **$P \leq 0.01$, and ***$P \leq 0.001.$ ## H2O2 activates apoptosis and autophagy in human chondrocytes The H2O2-induced chondrocyte damage model is considered a classical in vitro model of OA, which we made with reference to the experimental methods of previous studies [27, 28]. To assess the toxicity of H2O2 on chondrocytes, cells were stimulated with incremental doses of H2O2 for 4 h. The CCK-8 results indicated that chondrocytes were exposed to H2O2 for 4 h to inhibit activity in a dose-dependent fashion, and doses above 200 μM H2O2 had a significant inhibitory effect (Fig. 1A). Therefore, we selected 200 µM H2O2 for subsequent experiments. To search for the potential mechanism of reduced cell activity, we examined chondrocyte apoptosis and autophagy. To explore the apoptosis response of H2O2-stimulated human chondrocytes, we selected different doses of H2O2 (0, 10, 50, 100, 200 and 300 µM) to stimulate cells. After treatment for 24 h, the protein expression levels of Bax, cleaved-caspase3 and Bcl-2 were detected by western blot. As shown in Fig. 1B–E, H2O2 increased the expression of the proapoptotic proteins Bax and cleaved caspase3 and decreased the expression of the antiapoptotic protein Bcl-2 in a concentration-dependent manner. We also examined the effect of H2O2 on chondrocyte autophagy. The results showed that H2O2 enhanced the expression of the autophagy markers LC3-II/LC3-I and Beclin-1 in a concentration-dependent manner (Fig. 1F–H). The apoptosis and autophagy reactions of chondrocytes showed the same enhancement trend after H2O2 treatment. We speculate that overactivated autophagy exacerbated the apoptosis of chondrocytes. Fig. 1H2O2 induces apoptosis and activates autophagy in human chondrocytes. A Cell cytotoxicity detected by CCK-8 assay. Chondrocytes were treated with 0, 10, 50, 100, 200, 300, 400 and 500 μM H2O2 for 4 h. B–E western blot analysis and quantitative correlation analysis of Bax, cleaved caspase3 and Bcl-2 in chondrocytes. F–H western blot analysis and correlation quantitative analysis of LC3 and Beclin-1 in cells. Values are mean ± SD. * $p \leq 0.05$, **$p \leq 0.01$, and ***$p \leq 0.001$ versus the control group ## Effect of DAS on human chondrocyte viability and maintenance of cell phenotype The chemical properties of DAS are shown in Fig. 2A. To assess the effect of DAS on chondrocytes, cells were treated with various doses of DAS for 24 h. According to Fig. 2B, DAS showed no significant toxicity on cells at doses of 0.5, 1, 2.5 and 5 μM. We further investigated the influence of two experimental concentrations of DAS (2.5 and 5 μM) on chondrocyte viability at different treatment times (12, 24, 48 and 72 h). As shown in Fig. 2C, DAS with a treatment time ≤ 48 h had no obvious toxic effect on chondrocytes, but when the treatment time was longer than 48 h, the cell activity was significantly reduced. When chondrocytes were pretreated with DAS at different concentrations (0.5, 1, 2.5 and 5 μM) for 24 h and then cultivated with or without H2O2 (200 μM) for 4 h, cell viability was significantly restored by the pretreatment, and the most protective concentration was 5 μM (Fig. 2D). Based on these data, DAS was added at concentrations of 2.5 and 5 μM was for further experiments. Next, morphological analysis of chondrocytes after different treatments was performed by phase-contrast microscopy. According to Fig. 2E, the number of cell colonies decreased after H2O2 treatment compared to the control value, while DAS effectively prevented H2O2-induced cell depletion. The effect of DAS on the chondrocyte phenotype was also tested by Safranin O staining, which yielded similar results. The application of DAS increased the secretion of glycosaminoglycan (GAG) by chondrocytes and prevented matrix degradation (Fig. 2F–G).Fig. 2Protective effects of DAS on chondrocytes induced by H2O2. A The chemical structures of DAS. B Effects of DAS on chondrocyte activity. After being cultured with DAS (0.5, 1, 2.5, 5, 10, 15 and 20 μM) for 24 h, cell proliferation was determined by CCK8 assay. C Effects of DAS on chondrocyte activity. After being cultured with DAS (2.5 and 5 μM) for different treatment times (12, 24, 48 and 72 h), cell proliferation was determined by CCK8 assay. D Chondrocytes were pretreated with DAS (0.5, 1, 2.5, and 5 μM) for 24 h in the presence and absence of H2O2 (200 μM), and cell activity was measured by CCK8 assay. The values are mean ± SD. # $p \leq 0.05$ versus control group, *$p \leq 0.05$, **$p \leq 0.01$, and ***$p \leq 0.001$ versus control group. E *Morphological analysis* of chondrocytes after different treatments (magnification ×100, scale bar = 100 μm). F Chondrocyte phenotype, glycosaminoglycan production and matrix degradation were evaluated by Safranin O staining (magnification ×100, scale bar = 100 μm). G *Statistical analysis* of Safranin O staining ## DAS alleviates H2O2-induced human chondrocyte apoptosis To assess the effect of DAS on human chondrocyte apoptosis, we measured the expression of apoptosis-related factors by western blot. Chondrocytes were pretreated with DAS for 24 h and then incubated with or without H2O2 (200 μM) for 4 h. As shown in Fig. 3A–D, DAS (2.5 or 5 μM) inhibited the H2O2-induced upregulation of pro-apoptotic Bax and cleaved-caspase3 proteins, whereas it increased the expression of anti-apoptotic Bcl-2. Notably, treatment with 5 μM DAS resulted in a more marked recovery of apoptosis-related protein expression than the dose of 2.5 μM. Subsequently, flow cytometry was used to measure the proportion of apoptotic cells, and similar results were obtained. According to Fig. 3E–F, the apoptosis rate in the H2O2 group ($33.62\%$) was significantly greater than that in the control group ($15.39\%$), but the apoptosis rate decreased to $25.82\%$ after DAS treatment. The above results showed that DAS protected chondrocytes from H2O2-induced apoptosis. Fig. 3Daurisoline (DAS) protects chondrocytes from H2O2-induced apoptosis. A–D Western blot was performed to quantitatively analyze the expression of Bax, Bcl-2 and cleaved caspase-3. The values are mean ± SD. # $p \leq 0.05$ versus the control group. * $p \leq 0.05$, **$p \leq 0.01$, and ***$p \leq 0.001$ versus the control group. E Apoptosis was quantified by flow cytometry. ( F) Relevant quantitative analysis of flow cytometry ## DAS downregulates the level of autophagy in chondrocytes To clarify whether autophagy participates in the protective effect of DAS against H2O2-induced chondrocyte apoptosis, we detected the expression of autophagy markers. As shown in Fig. 4A–D, under H2O2 treatment, the autophagy markers LC3 and Beclin-1 increased, and the protein level of p62 was significantly reduced. Under DAS pretreatment (2.5 and 5 μM), the expression of LC3-II/LC3-I and Beclin-1 were significantly downregulated, while the p62 protein level was upregulated. To further confirm the role of DAS in chondrocyte autophagy, immunofluorescence was performed to detect the expression of Beclin-1. As shown in Fig. 5A, B, H2O2 (200 μM, 4 h) significantly upregulated the expression of Beclin-1 compared to that in the control group, while treatment with DAS (5 μM, 24 h) significantly antagonized the increase in H2O2-induced Beclin-1 expression. These results indicated that DAS inhibited H2O2-induced excessive autophagy. Fig. 4DAS inhibits H2O2-induced chondrocyte autophagy. A–D Western blot and correlation quantitative analysis of Beclin-1, LC3 and p62 in chondrocytes. The values are mean ± SD. # $p \leq 0.05$ versus the control group. * $p \leq 0.05$, **$p \leq 0.01$, and ***$p \leq 0.001$ versus the control groupFig. 5Beclin-1 expression as assayed by immunofluorescence. A Beclin-1 expression in chondrocytes was determined by immunofluorescence. Blue, DAPI; green, Beclin-1. Magnification ×400. Scale bar, 50 μm. B Fluorescence intensity quantification of Beclin1 expression ## DAS activates the PI3K/AKT/mTOR signaling pathway The PI3K/AKT/mTOR signaling pathway has been reported to play a crucial role in chondrocyte apoptosis and autophagy. Therefore, to better understand the mechanism through which DAS prevents H2O2-induced chondrocyte apoptosis and autophagy damage, we investigated the involvement of the PI3K/AKT/mTOR signaling pathway by western blot. As shown in Fig. 6A–D, compared to the control group, p-PI3K, p-AKT and p-mTOR levels were reduced in chondrocytes treated with H2O2, while their expression was upregulated in chondrocytes treated with DAS (2.5 and 5 μM). These data suggest that the activation of PI3K/Akt/mTOR signaling may be related to the mechanism by which DAS inhibits autophagy injury. Fig. 6DAS activates the PI3K/AKT/mTOR signaling pathway. A The p-AKT, T-AKT, p-PI3K, T-PI3K, p-mTOR and T-mTOR expression levels in H2O2-stimulated chondrocytes with or without DAS were assayed by western blot. B–D Relevant quantitative analysis of the blots shown in A. The data represent the mean ± SD. # $p \leq 0.05$ versus the control group. * $p \leq 0.05$, **$p \leq 0.01$, and ***$p \leq 0.001$ versus the control group ## DAS inhibits autophagy marker expression through the PI3K/AKT/mTOR pathway and protects chondrocytes from apoptosis and matrix degradation To further verify whether DAS inhibited anomalous chondrocyte autophagy via the PI3K/AKT/mTOR signaling pathway, we treated chondrocytes with insulin-like growth 1 as a positive control (IGF-1, a PI3K activator). The cells were divided into the following 4 groups: control, H2O2, H2O2 + DAS and H2O2 + IGF-1. Then, the expression of relevant proteins was analyzed by western blot. As shown in Fig. 7A–D, H2O2 significantly inhibited phosphorylation of the PI3K/AKT/mTOR signaling pathway, while the separate addition of DAS and IGF-1 upregulated the phosphorylation levels of H2O2-inhibited PI3K, AKT, and mTOR to varying degrees. The results of autophagy marker expression showed that DAS and IGF-1 blocked autophagy. Western blot analysis showed that DAS and IGF-1 inhibited the autophagy-related proteins LC3 and Beclin-1 and promoted the expression of p62 (Fig. 7E–H). In addition, we found that DAS and IGF-1 reduced the stimulating effect of H2O2 on the expression of cleaved caspase-3 and Bax protein and enhanced the expression of the anti-apoptotic factor Bcl-2, as shown in Fig. 7I–L.Fig. 7DAS inhibits autophagy markers and apoptosis-related factors through the PI3K/AKT/mTOR signaling pathway. A–D Western blot analysis of the protein levels of p-AKT, T-AKT, p-PI3K, T-PI3K, p-mTOR and T-mTOR and the quantification of associated proteins in the blots shown. E–H western blot and quantitative correlation analysis of Beclin-1, LC3 and p62 in chondrocytes. I–L Western blot was performed to quantitatively analyze the expression of Bax, Bcl-2 and cleaved caspase-3. The values represent the mean ± SD. # $p \leq 0.05$ versus the control group. * $p \leq 0.05$, **$p \leq 0.01$, and ***$p \leq 0.001$ versus the control group As shown in Fig. 2E, we have preliminarily demonstrated the role of DAS in cartilage matrix metabolism. To further explore the effect of DAS on human chondrocytes, western blots were performed to detect the expression of the matrix component protein (collagen II) and matrix-degrading indicators (MMP-3 and MMP-13). The results showed that DAS significantly reversed the H2O2-induced downregulation of collagen II, inhibited the upregulation of MMP3 and MMP13, and blocked cartilage stromal decomposition (Fig. 8A–D). Therefore, DAS could protect human chondrocytes from stromal decomposition by restoring the H2O2-induced deletion of matrix component proteins and downregulating the expression of matrix-degrading proteins. In summary, DAS has a similar effect as IGF-1 and may block excessive autophagy and apoptosis of human chondrocytes by upregulating the activation of PI3K/AKT/mTOR signaling and alleviating cartilage matrix degradation. Fig. 8Effects of DAS on H2O2-induced matrix degradation of human chondrocytes. A–D Western blot and quantitative correlation analysis of collagen II, MMP-3 and MMP-13 in chondrocytes. The values represent the mean ± SD. # $p \leq 0.05$ versus the control group. * $p \leq 0.05$, **$p \leq 0.01$, and ***$p \leq 0.001$ versus the control group ## Discussion Osteoarthritis is a chronic degenerative disease characterized by chronic intra-articular inflammation and articular cartilage degeneration, often resulting in chronic pain and joint disability, and is more prevalent in the elderly population [6, 29, 30]. While drugs such as nonsteroidal anti-inflammatory drugs (NSAIDs) can relieve chronic pain and swelling caused by OA, the deterioration of OA is not effectively reversed. In addition, the long-term application of such drugs can cause a range of side effects [31]. Therefore, finding safe and effective drugs to treat OA is imperative. In recent years, traditional Chinese medicine (TCM) has been recognized worldwide for its natural advantages [32]. Daurisoline is an isoquinoline alkaloid mainly extracted from the Chinese herbal medicine Rhizoma Menispermi, which exhibits a wide range of pharmacological effects in the treatment of cardiovascular and cerebrovascular diseases, including anti-inflammatory and antitumor effects [20, 22, 23]. In this study, we investigated the effect and molecular mechanisms of DAS in H2O2-induced oxidative stress injury of chondrocytes and found that DAS alleviated H2O2-induced autophagy, apoptosis and cartilage matrix degradation in chondrocytes. The potential mechanism was associated with the activation of the PI3K/AKT/mTOR pathway in human OA chondrocytes. Oxidative stress is one of the leading causes of OA pathogenesis. It plays a major pathological driving role in ECM degradation, chondrocyte apoptosis and the inflammatory response [33–35]. It affects chondrocyte homeostasis through multiple pathways, such as mitochondrial dysfunction, telomerase shortening and DNA damage. In this study, we used H2O2 to induce oxidative stress injury in human OA chondrocytes to mimic the OA cell model. As expected, H2O2 markedly reduced chondrocyte viability and GAG secretion and significantly activated cell apoptosis. Moreover, the administration of DAS promoted the recovery of cell viability and phenotype and alleviated H2O2-induced chondrocyte apoptosis. These data indicate that DAS has a protective effect against H2O2-induced chondrocyte injury. Autophagy is a critical biological process of self-digestion. It is believed to play a dual role in the pathogenesis of OA [36]. Many studies have shown that autophagy is intimately involved in the protection of cartilage lesions, and restoration of autophagy alleviates chondrocyte apoptosis in OA [37, 38]. On the other hand, when the degradation capacity of lysosomes is insufficient to eliminate the autophagosome contents, abnormal autophagy promotes the activation of the apoptotic cascade, and autophagy damage induces aging-related cell death and aggravates the progression of OA in patients [39, 40]. Therefore, autophagy in cartilage appears to be subtly regulated, and whether it is beneficial or harmful to the progression of OA depends on the balance between the amount of substrate and the capacity of the autophagy mechanism, and maintaining an appropriate level of autophagy under stress is crucial. In this study, we found that H2O2-induced autophagy markers LC3 and Beclin-1 were significantly increased in human chondrocytes, accompanied by increased apoptotic signals, such as increased Bax and decreased Bcl-2, suggesting that cartilage injury was caused by excessive autophagy-induced apoptosis. By detecting the protein expression of LC3II/LC3I, Beclin-1 and p62, we found that DAS strongly inhibited autophagy. In addition, immunofluorescence was used to detect Beclin-1 to determine whether DAS can inhibit chondrocyte autophagy. When autophagy was blocked by DAS, the expression of apoptosis-related proteins (Bax and cleaved caspase 3) was downregulated, the level of anti-apoptotic Bcl-2 was increased, and the percentage of apoptotic cells was also decreased. These findings suggest that DAS mitigated oxidative stress-induced apoptosis signaling by inhibiting excessive autophagy in chondrocytes, making this process a potential therapeutic target for OA. Autophagy is controlled by the major negative regulator mTOR and its upstream regulator PI3K/Akt [41]. In addition to mediating autophagy, the PI3K/Akt/mTOR signaling pathway affects various physiological and pathological processes, such as cell proliferation, differentiation, inflammation, apoptosis, and cancer [42]. As the PI3K/Akt/mTOR pathway has been studied more, results have confirmed that the PI3K/AKT/mTOR pathway plays a role in apoptosis and autophagy of chondrocytes [43]. Considering the above reasons, we speculated that the regulatory effect of DAS on autophagy in osteoarthritis chondrocytes might be realized through the PI3K/AKT/mTOR signaling pathway. In our study, DAS significantly activated H2O2-induced inhibition of the PI3K/AKT/mTOR signaling pathway. Furthermore, IGF-1 was employed to activate the PI3K/Akt/mTOR signaling pathway as a positive control. Our results suggested that both DAS and IGF-1 treatment reversed H2O2-induced autophagy, apoptosis and cartilage matrix degradation. Therefore, these data indicate that DAS can reverse H2O2-induced adverse outcomes through the PI3K/Akt/mTOR signaling pathway, establishing the relationship between DAS and the PI3K/Akt/mTOR signaling pathway. Activation of mTOR is known to inhibit autophagosome formation, and inhibition of mTOR pathway signaling leads to cell death related to apoptosis and autophagy [44, 45]. Therefore, we suspect that the inhibition of apoptosis induced by excessive autophagy is synergistic with the antiapoptotic effect of PI3K/AKT/mTOR signaling pathway activation, which makes the antiapoptotic effect of DAS more significant. In summary, these results suggest that the potential mechanism by which DAS alleviates autophagy and apoptosis in chondrocytes is related to the activation of the PI3K/AKT/mTOR pathway. Whether PI3K/AKT/mTOR is a direct target of DAS remains unknown. Further studies are needed to elucidate the exact mechanism by which DAS regulates the PI3K/AKT/mTOR signaling pathway. In addition, the data obtained from in vitro experiments may differ from the results of in vivo experiments. Therefore, the curative effect of DAS on OA needs to be further studied. ## Conclusion This study preliminarily demonstrated that DAS inhibits chondrocyte apoptosis and improves inflammatory responses and matrix degradation in vitro by inhibiting H2O2-induced excessive autophagy. These effects may be mediated by activation of the PI3K/AKT/mTOR pathway. Our findings indicate the great potential of DAS in the treatment of OA. Future studies should focus on the efficacy and more comprehensive mechanisms of DAS in vivo, to determine whether there is strong evidence supporting the application of DAS in clinical trials. ## References 1. Bortoluzzi A, Furini F, Scirè CA. **Osteoarthritis and its management—epidemiology, nutritional aspects and environmental factors**. *Autoimmun Rev* (2018) **17** 1097-1104. DOI: 10.1016/j.autrev.2018.06.002 2. Coryell PR, Diekman BO, Loeser RF. **Mechanisms and therapeutic implications of cellular senescence in osteoarthritis**. *Nat Rev Rheumatol* (2021) **17** 47-57. DOI: 10.1038/s41584-020-00533-7 3. Al Faqeh H, Nor Hamdan BM, Chen HC, Aminuddin BS, Ruszymah BH. **The potential of intra-articular injection of chondrogenic-induced bone marrow stem cells to retard the progression of osteoarthritis in a sheep model**. *Exp Gerontol* (2012) **47** 458-464. DOI: 10.1016/j.exger.2012.03.018 4. Guilak F, Nims RJ, Dicks A, Wu CL, Meulenbelt I. **Osteoarthritis as a disease of the cartilage pericellular matrix**. *Matrix Biol* (2018) **71–72** 40-50. DOI: 10.1016/j.matbio.2018.05.008 5. Burr DB, Gallant MA. **Bone remodelling in osteoarthritis**. *Nat Rev Rheumatol* (2012) **8** 665-673. DOI: 10.1038/nrrheum.2012.130 6. Chen D, Shen J, Zhao W, Wang T, Han L, Hamilton JL, Im HJ. **Osteoarthritis: toward a comprehensive understanding of pathological mechanism**. *Bone Res* (2017) **5** 16044. DOI: 10.1038/boneres.2016.44 7. Hunter DJ, Bierma-Zeinstra S. **Osteoarthritis**. *Lancet* (2019) **393** 1745-1759. DOI: 10.1016/S0140-6736(19)30417-9 8. Cooper C, Chapurlat R, Al-Daghri N, Herrero-Beaumont G, Bruyère O, Rannou F, Roth R, Uebelhart D, Reginster JY. **Safety of oral non-selective non-steroidal anti-inflammatory drugs in osteoarthritis: What does the literature say?**. *Drugs Aging* (2019) **36** 15-24. DOI: 10.1007/s40266-019-00660-1 9. Lemes IR, Pinto RZ. **Critically appraised paper: physiotherapy improved pain and functional disability more than glucocorticoid injection in people with knee osteoarthritis [commentary]**. *J Physiother* (2021) **67** 64. DOI: 10.1016/j.jphys.2020.11.009 10. Mizushima N, Levine B, Cuervo AM, Klionsky DJ. **Autophagy fights disease through cellular self-digestion**. *Nature* (2008) **451** 1069-1075. DOI: 10.1038/nature06639 11. Rubinsztein DC, Mariño G, Kroemer G. **Autophagy and aging**. *Cell* (2011) **146** 682-695. DOI: 10.1016/j.cell.2011.07.030 12. Pierrefite-Carle V, Santucci-Darmanin S, Breuil V, Camuzard O, Carle GF. **Autophagy in bone: self-eating to stay in balance**. *Ageing Res Rev* (2015) **24** 206-217. DOI: 10.1016/j.arr.2015.08.004 13. Filomeni G, De Zio D, Cecconi F. **Oxidative stress and autophagy: the clash between damage and metabolic needs**. *Cell Death Differ* (2015) **22** 377-388. DOI: 10.1038/cdd.2014.150 14. Zhang M, Sui W, Xing Y, Cheng J, Cheng C, Xue F, Zhang J, Wang X, Zhang C, Hao P, Zhang Y. **Angiotensin IV attenuates diabetic cardiomyopathy via suppressing FoxO1-induced excessive autophagy, apoptosis and fibrosis**. *Theranostics* (2021) **11** 8624-8639. DOI: 10.7150/thno.48561 15. Mattiolo P, Yuste VJ, Boix J, Ribas J. **Autophagy exacerbates caspase-dependent apoptotic cell death after short times of starvation**. *Biochem Pharmacol* (2015) **98** 573-586. DOI: 10.1016/j.bcp.2015.09.021 16. Levine B, Kroemer G. **Autophagy in the pathogenesis of disease**. *Cell* (2008) **132** 27-42. DOI: 10.1016/j.cell.2007.12.018 17. Musumeci G, Castrogiovanni P, Trovato FM, Weinberg AM, Al-Wasiyah MK, Alqahtani MH, Mobasheri A. **Biomarkers of chondrocyte apoptosis and autophagy in osteoarthritis**. *Int J Mol Sci* (2015) **16** 20560-20575. DOI: 10.3390/ijms160920560 18. Liu JH, Zhao N, Zhang GJ, Yu SS, Wu LJ, Qu J, Ma SG, Chen XG, Zhang TQ, Bai J, Chen H, Fang ZF, Zhao F, Tang WB. **Bioactive quassinoids from the seeds of**. *J Nat Prod* (2012) **75** 683-688. DOI: 10.1021/np200920c 19. Zeng L, Wang W, Rong XF, Zhong Y, Jia P, Zhou GQ, Li RH. **Chondroprotective effects and multi-target mechanisms of Icariin in IL-1 beta-induced human SW 1353 chondrosarcoma cells and a rat osteoarthritis model**. *Int Immunopharmacol* (2014) **18** 175-181. DOI: 10.1016/j.intimp.2013.11.021 20. Yang XY, Jiang SQ, Zhang L, Liu QN, Gong PL. **Inhibitory effect of dauricine on inflammatory process following focal cerebral ischemia/reperfusion in rats**. *Am J Chin Med* (2007) **35** 477-486. DOI: 10.1142/S0192415X07004990 21. Lingenhöhl K, Small DL, Monette R, Buchan AM, Morley P, Allegrini PR, Fröst LW, Sauer D, Schmutz M, Knöpfel T. **Exploration of P-type Ca**. *Neuropharmacology* (1997) **36** 107-113. DOI: 10.1016/S0028-3908(96)00158-X 22. Zhou J, Wu H, Hou J, Wang J, Wang J, Li M, Yao X, Gao J, Zhang Q. **Daurisoline alleviated experimental colitis in vivo and in vitro: Involvement of NF-κB and Wnt/β-Catenin pathway**. *Int Immunopharmacol* (2022) **108** 108714. DOI: 10.1016/j.intimp.2022.108714 23. Huang XH, Yan X, Zhang QH, Hong P, Zhang WX, Liu YP, Xu WW, Li B, He QY. **Direct targeting of HSP90 with daurisoline destabilizes β-catenin to suppress lung cancer tumorigenesis**. *Cancer Lett* (2020) **489** 66-78. DOI: 10.1016/j.canlet.2020.05.024 24. Zhang X, Zhang JG, Mu W, Zhou HM, Liu GL, Li Q. **The role of daurisoline treatment in hepatocellular carcinoma: Inhibiting vasculogenic mimicry formation and enhancing sensitivity to sorafenib**. *Phytomedicine* (2021) **92** 153740. DOI: 10.1016/j.phymed.2021.153740 25. Wang FS, Kuo CW, Ko JY, Chen YS, Wang SY, Ke HJ. **Irisin mitigates oxidative stress, chondrocyte dysfunction and osteoarthritis development through regulating mitochondrial integrity and autophagy**. *Antioxidants* (2020) **9** 66. DOI: 10.3390/antiox9090810 26. Ansari MY, Khan NM, Haqqi TM. **A standardized extract of**. *Biomed Pharmacother* (2017) **96** 198-207. DOI: 10.1016/j.biopha.2017.09.140 27. D'Adamo S, Cetrullo S, Guidotti S, Silvestri Y, Minguzzi M, Santi S. **Spermidine rescues the deregulated autophagic response to oxidative stress of osteoarthritic chondrocytes**. *Free Radic Biol Med* (2020) **153** 159-172. DOI: 10.1016/j.freeradbiomed.2020.03.029 28. Lim HD, Kim YS, Ko SH, Yoon IJ, Cho SG, Chun YH, Choi BJ, Kim EC. **Cytoprotective and anti-inflammatory effects of melatonin in hydrogen peroxide-stimulated CHON-001 human chondrocyte cell line and rabbit model of osteoarthritis via the SIRT1 pathway**. *J Pineal Res* (2012) **53** 225-237. DOI: 10.1111/j.1600-079X.2012.00991.x 29. Zheng L, Zhang Z, Sheng P, Mobasheri A. **The role of metabolism in chondrocyte dysfunction and the progression of osteoarthritis**. *Ageing Res Rev* (2021) **66** 101249. DOI: 10.1016/j.arr.2020.101249 30. van den Bosch MHJ. **Osteoarthritis year in review 2020: biology**. *Osteoarthr Cartil* (2021) **29** 143-150. DOI: 10.1016/j.joca.2020.10.006 31. Bannuru RR, Osani MC, Vaysbrot EE, Arden NK, Bennell K, Bierma-Zeinstra SMA, Kraus VB, Lohmander LS, Abbott JH, Bhandari M, Blanco FJ, Espinosa R, Haugen IK, Lin J, Mandl LA, Moilanen E, Nakamura N, Snyder-Mackler L, Trojian T, Underwood M, McAlindon TE. **OARSI guidelines for the non-surgical management of knee, hip, and polyarticular osteoarthritis**. *Osteoarthr Cartil* (2019) **27** 1578-1589. DOI: 10.1016/j.joca.2019.06.011 32. Tian Z, Zhang X, Sun M. **Phytochemicals mediate autophagy against osteoarthritis by maintaining cartilage homeostasis**. *Front Pharmacol* (2021) **12** 795058. DOI: 10.3389/fphar.2021.795058 33. Loeser RF, Collins JA, Diekman BO. **Ageing and the pathogenesis of osteoarthritis**. *Nat Rev Rheumatol* (2016) **12** 412-420. DOI: 10.1038/nrrheum.2016.65 34. Ansari MY, Ahmad N, Haqqi TM. **Oxidative stress and inflammation in osteoarthritis pathogenesis: role of polyphenols**. *Biomed Pharmacother* (2020) **129** 110452. DOI: 10.1016/j.biopha.2020.110452 35. Jallali N, Ridha H, Thrasivoulou C, Underwood C, Butler PE, Cowen T. **Vulnerability to ROS-induced cell death in ageing articular cartilage: the role of antioxidant enzyme activity**. *Osteoarthr Cartil* (2005) **13** 614-622. DOI: 10.1016/j.joca.2005.02.011 36. Arias C, Salazar LA. **Autophagy and polyphenols in osteoarthritis: a focus on epigenetic regulation**. *Int J Mol Sci* (2021) **23** 66. DOI: 10.3390/ijms23010421 37. Zhang X, Li Q, Sun Y, Chen L, Wang J, Liang L. **Chondroitin sulfate from sturgeon bone protects rat chondrocytes from hydrogen peroxide-induced damage by maintaining cellular homeostasis through enhanced autophagy**. *Int J Biol Macromol* (2020) **164** 2761-2768. DOI: 10.1016/j.ijbiomac.2020.07.313 38. Kurakazu I, Akasaki Y, Tsushima H, Sueishi T, Toya M, Kuwahara M, Uchida T, Lotz MK, Nakashima Y. **TGFβ1 signaling protects chondrocytes against oxidative stress via FOXO1-autophagy axis**. *Osteoarthr Cartil* (2021) **29** 1600-1613. DOI: 10.1016/j.joca.2021.07.015 39. Almonte-Becerril M, Navarro-Garcia F, Gonzalez-Robles A, Vega-Lopez MA, Lavalle C, Kouri JB. **Cell death of chondrocytes is a combination between apoptosis and autophagy during the pathogenesis of Osteoarthritis within an experimental model**. *Apoptosis* (2010) **15** 631-638. DOI: 10.1007/s10495-010-0458-z 40. Hwang HS, Yang CM, Park SJ, Kim HA. **Monosodium urate crystal-induced chondrocyte death via autophagic process**. *Int J Mol Sci* (2015) **16** 29265-29277. DOI: 10.3390/ijms161226164 41. Yang Z, Klionsky DJ. **Mammalian autophagy: core molecular machinery and signaling regulation**. *Curr Opin Cell Biol* (2010) **22** 124-131. DOI: 10.1016/j.ceb.2009.11.014 42. Ersahin T, Tuncbag N, Cetin-Atalay R. **The PI3K/AKT/mTOR interactive pathway**. *Mol Biosyst* (2015) **11** 1946-1954. DOI: 10.1039/C5MB00101C 43. Sun K, Luo J, Guo J, Yao X, Jing X, Guo F. **The PI3K/AKT/mTOR signaling pathway in osteoarthritis: a narrative review**. *Osteoarthr Cartil* (2020) **28** 400-409. DOI: 10.1016/j.joca.2020.02.027 44. Kim YC, Guan KL. **mTOR: a pharmacologic target for autophagy regulation**. *J Clin Invest* (2015) **125** 25-32. DOI: 10.1172/JCI73939 45. Will M, Qin AC, Toy W, Yao Z, Rodrik-Outmezguine V, Schneider C, Huang X, Monian P, Jiang X, de Stanchina E, Baselga J, Liu N, Chandarlapaty S, Rosen N. **Rapid induction of apoptosis by PI3K inhibitors is dependent upon their transient inhibition of RAS-ERK signaling**. *Cancer Discov* (2014) **4** 334-347. DOI: 10.1158/2159-8290.CD-13-0611
--- title: The characteristics of extrachromosomal circular DNA in patients with end-stage renal disease authors: - Yue Peng - Yixi Li - Wei Zhang - Yu ShangGuan - Ting Xie - Kang Wang - Jing Qiu - Wenjun Pu - Biying Hu - Xinzhou Zhang - Lianghong Yin - Donge Tang - Yong Dai journal: European Journal of Medical Research year: 2023 pmcid: PMC10041755 doi: 10.1186/s40001-023-01064-z license: CC BY 4.0 --- # The characteristics of extrachromosomal circular DNA in patients with end-stage renal disease ## Abstract ### Background End-stage renal disease (ESRD) is the final stage of chronic kidney disease (CKD). In addition to the structurally intact chromosome genomic DNA, there is a double-stranded circular DNA called extrachromosomal circular DNA (eccDNA), which is thought to be involved in the epigenetic regulation of human disease. However, the features of eccDNA in ESRD patients are barely known. In this study, we identified eccDNA from ESRD patients and healthy people, as well as revealed the characteristics of eccDNA in patients with ESRD. ### Methods Using the high-throughput Circle-Sequencing technique, we examined the eccDNA in peripheral blood mononuclear cells (PBMCs) from healthy people (NC) ($$n = 12$$) and ESRD patients ($$n = 16$$). We analyzed the length distribution, genome elements, and motifs feature of eccDNA in ESRD patients. Then, after identifying the specific eccDNA in ESRD patients, we explored the potential functions of the target genes of the specific eccDNA. Finally, we investigated the probable hub eccDNA using algorithms. ### Results In total, 14,431 and 11,324 eccDNAs were found in the ESRD and NC groups, respectively, with sizes ranging from 0.01 kb to 60 kb at most. Additionally, the ESRD group had a greater distribution of eccDNA on chromosomes 4, 11, 13, and 20. In two groups, we also discovered several motifs of specific eccDNAs. Furthermore, we identified 13,715 specific eccDNAs in the ESRD group and 10,585 specific eccDNAs in the NC group, both of which were largely annotated as mRNA catalog. Pathway studies using Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) showed that the specific eccDNA in ESRD was markedly enriched in cell junction and communication pathways. Furthermore, we identified potentially 20 hub eccDNA-targeting genes from all ESRD-specific eccDNA-targeting genes. Also, we found that 39 eccDNA-targeting genes were associated with ESRD, and some of these eccDNAs may be related to the pathogenesis of ESRD. ### Conclusions Our findings revealed the characteristics of eccDNA in ESRD patients and discovered potentially hub and ESRD-relevant eccDNA-targeting genes, suggesting a novel probable mechanism of ESRD. ### Supplementary Information The online version contains supplementary material available at 10.1186/s40001-023-01064-z. ## Background End-stage renal disease (ESRD) is the final stage of chronic kidney disease (CKD), characterized by a glomerular filtration rate (GFR) of less than 15 ml/min/1.73m2, as well as structural and functional deterioration for at least 3 months [1]. According to the recent studies, diabetes and hypertension are the primary causes of ESRD in developed countries [2], whereas the transition from glomerulonephritis to metabolic kidney disorders has become the leading cause in low- and middle-income nations [3, 4]. The prevalence of ESRD has risen considerably as diabetes and hypertension have become more prevalent [5, 6]. Currently, kidney replacement therapy (KRT), which includes hemodialysis, peritoneal dialysis, and kidney transplantation, is the most common treatment for ESRD. Global KRT usage is predicted to reach 5.4 million by 2030, with Asia seeing the most growth [7]. Cardiovascular illnesses, uremia, volume overload, malnutrition, uncontrolled hypertension, cancer, severe infections, and dialysis-related complications are all major causes of death in people with ESRD [8–10]. ESRD has become a global concern since it would significantly diminish life quality and increase financial cost on families and society. Yasuo Hotta and Alix Bassel discovered extrachromosomal circular DNA (eccDNA) in pig sperm and wheat nuclei in 1964 [11]. EccDNA is a double-stranded circular DNA that exists in addition to the structurally complete chromosome genomic DNA. Following that, eccDNAs were discovered in human tumor cells [12] and practically all organisms [13, 14]. EccDNAs have been classified as micro DNA (100–400 bp) [15], small polydispersed circular DNAs (spcDNAs, 100 bp–10 kb) [16], mitochondrial DNAs (mtDNAs, 16 kb) [17], and double-minutes (DMs, 100 kb–3 Mb) [17, 18] based on their length and features. Previous research has linked eccDNA to genomic rearrangement, cell apoptosis, episome enlargement, translocation, and amplification [19–22]. Several studies have proclaimed that eccDNA potentially play an important role in disorders involving epigenetic modulation, such as cancer [23]. EccDNA can cause gene deletion, mutation, duplication, or amplification, resulting in genetic heterogeneity and adaptive evolution between cells. Recent research by lv [24] examined the physical characteristics of eccDNA, such as length, GC content, and motif signature, etc. and miRNA in urine from CKD patients, and found that the eccDNA count in the CKD group was higher than it was in the healthy group. Although preliminary study on the properties of eccDNA in CKD patients revealed certain differences between the patients and healthy individuals, there is still a dearth of knowledge regarding the role and probable mechanism of eccDNA in CKD, particularly in ESRD patients, which we intended to focus on. In this study, we used a high-throughput approach, namely Circle-sequencing, to collect and identify eccDNAs from peripheral blood mononuclear cells (PBMCs) of ESRD patients ($$n = 16$$) and healthy people ($$n = 12$$). We detected 14,431 and 11,324 eccDNAs in the ESRD and NC groups, respectively, with a large range of eccDNA sizes. Furthermore, we identified that the specific eccDNA in ESRD was markedly enriched in cell junction and communication pathways. Out of all the eccDNA-targeting genes specific to ESRD, we discovered potentially 20 hub genes. Additionally, we found that eccDNA-targeting genes, including CCL2, CCR2, MYH9, and IL10, were critical in the development of ESRD, indicating that these eccDNAs had novel biological roles for ESRD patients. ## Participants From January 2020 to January 2021, 16 ESRD patients and 12 healthy individuals participated in this study at Shenzhen People’s Hospital (Shenzhen, China). This study’s ESRD diagnostic criteria were based on the CKD diagnostic and grading guideline of the 2012 Kidney Disease: Improving Global Outcomes (KDIGO) and Kidney Disease Outcomes Quality Initiative (KDOQI) guideline [25], with the following inclusion criteria: [1] age ≥ 18 years; [2] abnormalities of kidney structure present for > 3 months, including: abnormalities detected by imaging or histology, electrolyte and other abnormalities due to tubular disorders; [3] abnormalities of kidney function (GFR < 15 mL/min/1.73m2), present for > 3 months; [4] maintenance hemodialysis three times a week for at least 6 months; [5] without acute or chronic infection; [6] without malignant tumor; [7] without no serious complications of important organs, including cardiovascular, hepatic, pulmonary, or brain. The individuals selected in the normal control (NC) group were from the Physical Examination Center of Shenzhen People's Hospital (Shenzhen, China), which had no relation in birth with ESRD patients, with the inclusion and exclusion criteria as follows: [1] age ≥ 18 years; [2] no clinical or laboratory evidence for renal diseases; [3] without acute or chronic infection; [4] without malignant tumor; [5] without critical basic diseases, including urinary, cardiovascular, hepatic, pulmonary, or brain related issues. The ESRD group was studied using demographic profiles, such as age, gender, hemodialysis duration, hypertension (HTN), diabetic nephropathy (DN), chronic glomerulonephritis (CGN), and lupus nephropathy (LN), as well as laboratory data such as albumin, hemoglobin (Hb), neutrophil percentage, estimated glomerular filtration rate (eGFR), serum creatinine (Scr), blood urea nitrogen (BUN), and parathyroid hormone (PTH) as independent variables. Shenzhen People's Hospital Ethics Committee approved the study, which was carried out in accordance with the Declaration of Helsinki. All participants signed their informed consent. ## PBMC extraction A heparinized vacuum container was used to collect each 10 mL blood sample from the ESRD and NC groups. PBMCs were isolated from blood samples using Ficoll-Hypaque Solution (GE Healthcare, Marlborough, MA) by density gradient centrifugation at 1200 rpm for 3 min at room temperature, and lysed using TRIzol reagent (Invitrogen; Thermo Fisher Scientific, Inc.) after standing at 4 °C for 20 min [26, 27]. The suspension was then transferred to a new tube and centrifuged again (2000 rpm for 15 min). Finally, the supernatant was immediately collected, and the PBMCs were kept in a refrigerator at − 80 °C. ## Total DNA isolation To rupture plasma cell membranes and eliminate DNase and RNase, total DNA was isolated from PBMCs in lysis buffer containing L1 suspension solution (A&A Biotechnology) supplemented with Proteinase K [28]. Samples were incubated in the aforementioned lysate for 16 to 24 h at 50 °C and 700 rpm, until the suspension was homogeneous and acellular mass, before being chilled to room temperature. DNA samples were purified and enhanced using the Plasmid Mini AX Kit’s instructions for column chromatography (A&A Biotechnology). ## Linear chromosomal DNA digestion The purified DNA was treated with cutting endonuclease to facilitate specific digestion of linear DNA and part of mtDNA by exonuclease. The samples were incubated for 16 h at 37 °C, then the endonuclease was heat inactivated for 5 min at 80 °C. Exonuclease (Plasmid-Safe ATP-Dependent DNase kit, Epicentre) was used to digest linear single-stranded and double-stranded DNA. After removal of the above DNA, samples from the exonuclease-treated solution were confirmed to eliminate chromosomal linear DNA and mtDNA by quantitative polymerase chain reaction (qPCR) using gene COX5B and Human mt separately. The exonuclease solution was then heat inactivated for 30 min at 70 °C and the digested eccDNA was purified by magnetic beads (Agencourt AMPure XP beads). ## eccDNA amplification Enriched and purified eccDNA was subjected to rolling circle amplification with the REPLI-g Mini Kit (Qiagen, 150023), and the amplified product was purified with AMPure XP beads (Beckman, A63880). ## High-throughput sequencing With a focused ultrasonicator (biorupter), the rolling circle amplification products were ultrasonically sheared to roughly 200–300 bp. The NEBNext Ultra DNA Library Prep Kit for Illumina was used to recover the fragmented products and build libraries. Finally, the libraries were sent to an Illumina Novaseq 6000 system for sequencing. Paired-end 150 sequencing approach (PE150). ## Raw sequencing data analysis After the library check was qualified, raw sequencing data was generated from the Illumina NovaSeq 6000 system. Trimmomatic software [29] filtered raw data to remove adaptor and low-quality reads, and default reference values were utilized for the key parameters. The clean reads produced from the raw reads by eliminating the adaptor sequences were then aligned to reference genome sequences (hg38_gencode) using the BWA tool after raw reads and clean reads quality testing. The reads aligned to the gene were used for subsequent identification and analysis of circular DNA. The circular DNA was detected by the Circle-Map [30, 31], and the split was used for screening eccDNA. Each eccDNA contained at least 1 split read (split ≥ 1), which was chosen to be further analyzed. ## Gene annotation and functional analyses *The* gene annotation of eccDNA was based on BEDTools [32]. The HOMER's findMotifsGenome.pl tool was used for motif analysis. Gene Ontology (GO) analysis of the specifically expressed genes on eccDNAs was derived from Database for Annotation, Visualization, and Integrated Discover (DAVID, https://david.ncifcrf.gov/tools.jsp). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway was utilized to describe specifically expressed gene pathways on the basis of the DAVID database as well. Fisher exact test was performed to detect the overlap between the GO/KEGG annotation lists. $P \leq 0.05$ was considered to be statistically significant of GO terms/KEGG pathways. When the eccDNAs were classified into distinct types of RNAs, including miRNA, lncRNA, mRNA, pseudogene, and other types of genes (others), if multiple sorts were annotated, eccDNAs were classified as multipleType. For the sake of obtaining known and predicted functional associations between specific expression genes on eccDNAs, STRING database (https://cn.string-db.org/) was used. Furthermore, the Molecular Complex Detection (MCODE) in Cytoscape was used to inspect the modules of the gene–gene interaction (GGI) network (degree cutoff = 2, node score cutoff = 0.2, k-core = 2, and max. Depth = 100). The cytoHubba plugin in Cytoscape was used to identify the top 20 hub genes in network GGI by MCC method. The known genes associated with ESRD were searched in the Phenopedia database (https://phgkb.cdc.gov/PHGKB/startPagePhenoPedia.action) and Disgenet database (https://www.disgenet.org). The Metascape database (https://metascape.org) was conducted for pathway analyses of ESRD-related genes. ## Statistical analysis Categorical variables were expressed as frequency and percentage (%), and compared by the χ2 test between groups. Normally distributed numerical variables were described as mean ± SD and compared by the Student’s t test between groups, while non-normally distributed numerical variables were presented as median [P25, P75]. ## Basic information of participants The clinical and demographic data of ESRD patients and healthy volunteers was shown in Table 1. The control group had the matched gender with ESRD group ($P \leq 0.05$), while the age of the controls meeting the grouping conditions was younger. By comparison, each laboratory profile of patients with ESRD was visibly aberrant, while hemoglobin, neutrophil percentage, serum creatinine, and blood urea nitrogen of the NC group were in the normal range and had a significant difference from patients with ESRD ($P \leq 0.05$).Table 1The demographic and laboratory data in ESRD and NC groupsVariablesESRD ($$n = 16$$)NC ($$n = 12$$)P valueAge, years (± SD)45 ± 1033 ± 130.007Male, n (%)8 (50.00)4 (33.33)0.3778HTN, n (%)14 (87.50)//DN, n (%)4 (25.00)//CGN, n (%)10 (62.50)//LN, n (%)2 (12.50)//Duration of hemodialysis, years (M [P25, P75])4 (2.75, 5.00)//Hb, g/dL (± SD)93.63 ± 16.31137.00 ± 14.72 < 0.0001Neutrophil percentage,% (± SD)66.29 ± 7.4656.22 ± 10.160.0076Scr, µmol/L (± SD)1027.76 ± 367.3571.97 ± 13.75 < 0.0001BUN, mmol/L (± SD)21.81 ± 7.623.95 ± 0.73 < 0.0001eGFR, mL/min/1.73m2 (± SD)4.93 ± 2.99//Albumin, g/dL (± SD)39.28 ± 6.57//PTH, pg/mL (± SD)381.57 ± 300.49//ESRD, end-stage renal disease; NC, normal control; HTN, hypertension; DN, diabetes nephropathy; CGN, chronic glomerulonephritis; LN, lupus nephritis; Hb, hemoglobin; Scr, serum creatinine; BUN, blood urea nitrogen; eGFR, estimated glomerular filtration rate; PTH, parathyroid hormone ## The quality control of circle sequencing After sequencing, we yielded 159,339,104, and 133,069,578 raw reads from the NC and ESRD groups, respectively. Then, using Trimmomatic to remove adaptor and low-quality sequences, we obtained 156,391,184 and 130,528,434 clean reads from the NC and ESRD samples, respectively. Following that, we found that the quality values Q20 and Q30 in two groups were both greater than $90\%$ as analyzed through FastQC, which suggested strong sequencing accuracy. Subsequently, we discovered that the unique mapped rate was both over $80\%$, indicating good sequencing quality. The data of quality control were listed in Table 2.Table 2The quality control data of circle-sequencingSampleRaw readsClean readsQ20 (%)Q30 (%)Unique mapped rate (%)NC159,339,104156,391,18498.3294.6188.45ESRD133,069,578130,528,43498.3594.6884.53 ## The length ranges and genomic distribution of eccDNA We used the Circle-Map software under the screening condition with split reads ≥ 1 to detect eccDNA from similar pair areas of the human genome. As a result, we identified more than 10,000 distinct eccDNAs in the ESRD ($$n = 14$$,431) and NC groups ($$n = 11$$,324), respectively, and eccDNA ranged from 0.01 KB to 60 KB at most (Fig. 1A, B). In the NC group, there were nine eccDNAs greater than 1 MB, with the longest being 218 Mb. There were 15 eccDNAs larger than 1 MB in the ESRD sample, with the largest being 88 MB. The fraction of eccDNA in ESRD was higher than that in NC in most length ranges, but in the 0.3 kb to 1 kb range, NC's proportion was higher than ESRD's (Fig. 1C). As was previously reported, the full range of circular DNAs was included in our study [23]. We found that ESRD patients' eccDNAs were less distributed than the NC group's on chromosomes 2, 17, and 19 and sex chromosome Y, but that healthy individuals' eccDNAs were less disseminated than ESRD samples on chromosomes 4, 11, 13, and 20 (Fig. 1D, E).Fig. 1EccDNA sizes, distribution, repetitive elements, and catalogs. The distribution of eccDNA sizes (< 100 kb) in the NC (A) and ESRD (B) groups. Plots of the size distributions (< 100 kb) of ESRD- and NC-derived eccDNA (C). Distribution of eccDNA on the chromosome genome in NC (D) and ESRD (E) groups. Repetitive regions from general mapped reads for ESRD- and NC-derived eccDNA samples (F) *In previous* studies, it has been known that eccDNA consists of all types of elements, including repetitive genomic elements, and the repetitive regions exist in proportions to the genome [33]. In our study, we also found that a large number of reads mapped to long interspersed nuclear elements (LINE), short interspersed nuclear Alu (SINE-Alu), long terminal repeat (LTR), satellite, CpG island, telomere, centromere, mitochondria (MT), ribosomal RNA (rRNA), transfer RNA (tRNA), small nuclear RNA (snRNA), single-cell RNA (scRNA), and signal recognition particle RNA (srpRNA), implying the eccDNAs derived from all types of repetitive sequences on genome (Fig. 1F). With the University of California Santa Cruz database (UCSC, http://genome.ucsc.edu), we discovered that the eccDNA reads in LINE ($18.2\%$ of ESRD, $18.5\%$ of NC), SINE-Alu ($23.7\%$ of ESRD, $26.1\%$ of NC), and LTR ($11.2\%$ of ESRD, $9.6\%$ of NC) were obviously enriched, but reads in telomere were fewer. In the ESRD group, reads in MT ($2.7\%$) and tRNA ($0.1\%$) regions were apparently more abundant than in the NC group ($1.3\%$ of MT, $0.07\%$ of tRNA). On the contrary, reads in rRNA regions in ESRD ($0.1\%$) were fewer than in NC ($0.2\%$). ## The identified motifs on both sides of eccDNA junctions We probed recurring motifs that appeared on both sides of the eccDNA junction site and their corresponding transcription factors (TFs) to explore the effect of TFs on eccDNA functions. After identifying the junction sites of eccDNA, we retrieved the DNA sequences that were enlarged to 200 bp in the upstream and downstream directions of eccDNA coordinate start and end, respectively. The start and end junction locations are thought to be the consequence of chromosomal DNA sequences being cut off to generate eccDNA [34]. According to 37,874 and 39,462 genome DNA sequences in the NC and ESRD groups, separately, findMotifsGenome.pl tool of Homer software was used to identify the known motifs and relevant TFs. As a result, we discovered 976 motifs in total, 20 and 38 of which were substantially connected to the TFs (P value ≤ 0.01) in the ESRD and NC groups, respectively. The 20 motifs in ESRD patients’ eccDNA were collectively listed in Table 3. The enrichment scores of the top 10 TF binding motifs were also recorded with a comparison of the NC group (Fig. 2A, B).Table 3Identification of motif patterns in ESRDMotif sequencesP valueCTCCCTGGGAGGCCN1e−5ACAGGAAGTG1e−4CRGCTGBNGNSNNSAGATAA1e−3YAGATCTRAW1e−3TAATCCCN1e−2BTBRAGTGSN1e−2TTTTTTTTTT1e−2RATWCCGTTA1e−2AGTTTCAKTTTC1e−2NGCGTGGGCGGR1e−2AAAAAAAAAA1e−2AVCCGGAAGT1e−2AAACMATTAN1e−2RSCACTYRAG1e−2RGGATTAR1e−2CDCCGCCGTC1e−2ATGCATAATTCA1e−2BCWGATTCAATCAAN1e−2AAAGRGGAAGTG1e−2TWCCHWATWDGGAAA1e−2N, degenerate bases of A/T/C/G; R, degenerate bases of A/G; B, degenerate bases of T/C/G; S, degenerate bases of C/G; Y, degenerate bases of T/C; W, degenerate bases of A/TFig.2Identification of motifs and transcription factors. Enrichment scores of top 10 predicted transcription factors binding motifs in ESRD group and compared to NC group (A). The top 10 transcription factors related motif sequences (B) ## Functional investigations of target genes of ESRD-specific eccDNAs To define the specific eccDNA in ESRD patients which may have an effect on the occurrence of ESRD, we analyzed the eccDNAs detected in ESRD patients and healthy people using Venn analysis. To avoid influencing the results, 1 Mb of eccDNA was left out of the analysis (1 Mb in length would interfere with the intersection analysis). The result revealed 13,715 specific eccDNAs in the ESRD group and 10,585 in the NC groups, respectively (Fig. 3A). After that, to probe the probable functions of these peculiar eccDNAs in the development of ESRD, we performed enrichment analysis of the target genes of the eccDNA using BEDtools. Then, we classified the eccDNAs into different modules based on the catalogs of their targeting genes. Eventually, there was no significant difference in quantity between specific eccDNAs in ESRD patients and the NC group among the diverse modules. Most part of the specifically expressed eccDNA was categorized as mRNA (Fig. 3B).Fig. 3Specifically expressed eccDNA number, catalogs, and GO/KEGG functional enrichment bubble plots in patients with ESRD compared. Venn diagram shows the number of commonly and specifically expressed eccDNA in the ESRD group and NC group (A). The catalog distribution of specifically expressed eccDNA in NC and ESRD groups (B). The GO classifications of specifically expressed eccDNA in patients with ESRD in the biological processes (C), the cellular component (D), and molecular function (E). The specifically eccDNA in patients with ESRD in the KEGG functional enrichment (F) Furthermore, we conducted GO and KEGG analysis of these specific eccDNAs’ targeting genes. Compared with the NC group, genes on eccDNAs in ESRD patients were particularly enriched in the biological processes, namely regulation of cell shape, positive regulation of chondrocyte differentiation, and Golgi organization (Fig. 3C), the cellular components, such as neuronal cell body, axon, and extrinsic component of cytoplasmic side of plasma membrane (Fig. 3D), and the molecular functions, which were glutamate receptor activity, Ras guanyl-nucleotide exchange factor activity, and motor activity (Fig. 3E). Meanwhile, the KEGG analysis highlighted three pathways, including phosphatidylinositol signaling system, neurotrophin signaling pathway, and inositol phosphate metabolism (Fig. 3F). According to the aforementioned findings, eccDNAs might be involved in the pathways, such as chondrocyte differentiation, Golgi function, and phosphatidylinositol signaling, to promote the development and progression of ESRD. ## The potentially hub eccDNA in ESRD To identify the eccDNA that might be crucial in the development of ESRD, we performed computer calculations of the target genes corresponding to these ESRD-specific eccDNA. We built a gene–gene interaction network model of the genes on distinct eccDNAs in the ESRD group using STRING database (Fig. 4A). Subsequently, we uncovered the closely-tied group within the network using the MCODE plugins in Cytoscape, and two clusters were found (Fig. 4B). Moreover, we analyzed the hub nudes among all the genes using cytoHubba and then identified 20 top hub genes. Among the hub genes, GRIN2A, NCAM1, GRIK1, RPTOR, PRKAG2, and EFCAB1 exhibited the highest scores, that was, the greatest gene-to-gene linkages (Fig. 4C).Fig. 4The GGI network between the specifically expressed eccDNA in ESRD. Overview of the GGI network (A). Module analysis by MCODE (B). The hub genes network (C) After that, we looked for the target genes of ESRD-specific eccDNA that have been reported to link to ESRD, suggesting that their responding eccDNAs might also contribute to ESRD. We searched for the published functional genes of kidney diseases using the phenopedia and DisGeNet databases, and the result showed that a total of 68 genes of eccDNA in ESRD patients were found in the genes known for kidney diseases. The venn diagram exhibited the distribution of 68 genes in various diseases, including 39 genes in ESRD, 51 genes in DN, 14 genes in CGN, and 12 genes in LN (Fig. 5A, Additional file 1: Table S1). Subsequently, we investigated the functions of the 39 ESRD-related genes using Metascape database and found that the principal biological processes that these genes participated in were response to stimulus, regulation of biological process, and negative response of biological process (Fig. 5B). Besides, the pathways analysis highlighted pathways such as cellular response to lipid, cytokine signaling in immune system, and regulation of epithelial cell proliferation (Fig. 5C).Fig. 5The function of ESRD-related eccDNA-targeting gene. Venn diagram shows the number of overlap target genes of eccDNA and known kidney-related genes in diverse kidney diseases (A). The GO biological processes of ESRD-related overlap genes enriched in Metascape (B). Pathway and process network analysis of ESRD-related overlap genes (C) ## Discussion Multiple variables have a role in the development of ESRD because it can be exacerbated by the evolution of a range of kidney illnesses. For starters, renal and systemic inflammation are significant causes of ESRD in patients with kidney diseases [35]. Furthermore, multiple studies have revealed that TFs, DNA, and chromosome damage, as well as metabolites in urine and blood, are all linked to the development of ESRD and can be manifested to be potential diagnostic and prognostic biomarkers [36–38]. However, there are currently no daily clinical examinational biomarkers for ESRD. With advances in technology of sequencing, quantities of latest studies have discovered that eccDNA physically excised from the chromosome is involved in a wide range of biological processes, including cell–cell communication, aging, intercellular genetic heterogeneity, regulating innate immunity, transcribed into noncoding RNAs, and participates in the cancer physiological processes [39–41]. However, the biochemical function of eccDNA in ESRD patients is inadequately defined. As far as we know, this study is the first to illustrate the expression and function of eccDNA in ESRD patients. This study extends the knowledge of the characteristics of eccDNA for ESRD patients. Using the circle-seq technology, we were able to identify eccDNA in PBMCs from ESRD patients and healthy people, as well as reveal the characteristics of eccDNA in PBMCs from patients with ESRD, such as the number of eccDNA, length distribution, genome distribution, motif, and function of genes on eccDNA. Finally, we identified 20 hub genes and 39 ESRD-related genes of ESRD-specific eccDNA-targeting genes. EccDNA is featured with motifs siding the start and end. Motifs are DNA sequences that provide binding sites for a type of protein called TFs, which govern the activation or repression of gene expression by recognizing motifs found at regulatory regions to regulate downstream chromatin processes [42]. The major ESRD transcription factor GLIS3 in this study, which belongs to the Krüppel-like zinc finger protein family, is mostly expressed in the kidney, thyroid, and pancreas. GLIS3 has been indicated to act a significant part in preserving the natural structure and function of the kidney as part of transcription regulatory networks, and GLIS3 mutant develops polycystic kidney disease [43]. However, whether GLIS3 leads to other kidney diseases even ESRD remains room for further research. Furthermore, the transcription factor Egr2 has been demonstrated to play a role in neutrophil degranulation and immunological activation in ESRD patients on nocturnal hemodialysis [44]. Pitx1, IRF3, and ETS1 are also TFs involved in the pathogenesis of kidney diseases [45–47]. Hence, we conjecture that eccDNA develops ESRD as a result of the TFs mentioned above. In this work, we discovered 20 top eccDNA hub genes in which NCAM1, NFATC1, PRKCB, LEF1, PRKAG2, and GRM8 were strongly linked to a variety of kidney disorders. For example, gene NFATC1 has been associated with LN [48], gene PRKCB1 has been linked to the progression from DN to ESRD [49], transcription factor LEF1 encoded by gene LEF1 engaged in the Wnt signaling pathway is linked to CKD [50], and gene PRKAG2 is a fresh locus for CKD [51]. In addition, we searched databases and discovered that some ESRD-specific targeting-genes were known to be related to several kidney diseases involved in this study (Additional file 1: Table S1). Obviously, the majority of ESRD-related genes were linked to DN, CGN, or LN as well, which was consistent with the progression of primary diseases into ESRD. Among them, we noticed that CCL2, CCR2, MYH9, and IL10 were present in all four diseases. According to the reports, monocyte chemoattractant protein 1 (MCP-1) encoded by CCL2 was a biomarker in kidney diseases, suggesting kidney damage and inflammation, and CCL2 itself increased in macrophages and was related to renal fibrosis in a renal atrophy model [52]. The genotype frequency of polymorphisms in CCR2 and IL10 showed a great difference between ESRD and controls, especially IL10, which demonstrated their susceptibility to ESRD [53]. MYH9 mutation might disorder renal epithelial transport pathways and further result in kidney diseases [54]. The above ESRD-related genes and functional analysis in Fig. 5C were displayed that inflammation and renal epithelial cell dysfunction were essential mechanisms in ESRD and implied that eccDNA in these genes played a significant part in the progression of ESRD which was the subject of further research. According to GO analysis, “regulation of cell shape” is the most enriched biological process among the genes predicted by specifically expressed eccDNA in the ESRD group as compared to the NC group. Planar cell polarity (PCP) refers to the coordinated orientation of cells in the tissue plane. Protein encoded by PCP genes and PCP signaling pathway regulate cell shape and behavior, as well as kidney development and diseases, such as polycystic kidney disease and Congenital Anomalies of the Kidney and Urinary Tract (CAKUT) [55]. In addition, actomyosin, a prominent cellular target of the PCP signaling pathway, not only regulates cell shape and motility, but it can also be cleaved by activated caspase-3, causing muscle atrophy in patients with CKD [56]. These findings suggest that eccDNA hub genes may have a role in the pathophysiology of ESRD via regulating cell shape. Furthermore, in comparison to the NC group, "glutamate receptor activity" is the most enriched molecular function of specifically expressed eccDNA in the ESRD group. Ionotropic receptors, such as NMDA receptors, AMPA receptors, and KA receptors, along with metabotropic L-Glu receptors (mGluRs), are the two types of glutamate receptors. The toxicity of overactivated NMDA receptors on renal cells has been demonstrated [57]. Ca2+ influx and oxidative stress are caused by sustained NMDA receptors activation, which can contribute to glomerulosclerosis. Ca2+ influx pathways TRPC6 that amplifies Ca2+ excess activated by NMDA receptors regulates Rac1 of the Rho protein family to modulate signal transduction that influences a great many aspects of cell behavior, including cytoskeletal dynamics in podocytes [58]. The abnormality of actin cytoskeleton then breaches the barrier of proteinuria and finally gives rise to CGN, such as focal segmental glomerulosclerosis (FSGS) [59], which is the most common cause of ESRD and has the most cases in this study, implying that the aforesaid hub genes may be involved in the pathogenic process of ESRD. The primary pathway of ESRD patients with a comparison of healthy individuals in the present study is “phosphatidylinositol signaling system” among the top 20 enrichment KEGG pathways. Phosphatidylinositol signaling pathway is involved in a variety of biological activities, such as cell proliferation, cell differentiation, apoptosis, and membrane trafficking [60]. Among this signaling system, abnormal activation of the phosphoinositide 3-kinase gamma (PI3Kγ) signaling pathway has been shown to play an important role in the regulation of profibrotic phenotypes. In kidney disease, blocking PI3Kγ signaling pathway in Ang-II-induced kidney damage could alleviate renal injury and fibrosis, and thus improve renal functions, as well as be investigated as a fresh therapeutic method for the treatment of renal fibrosis, renal hypertension, and CKD [61]. Furthermore, protein-energy wasting characterized by muscle wasting is obviously manifested in ESRD patients, increasing the morbidity and mortality [62]. A decrease in PI3K activity in skeletal muscle has been shown to aggravate caspase-3 activity and enhance protein degradation, leading to and speeding up wasting through hemodialysis [63]. This latently indicated that eccDNAs from hub genes could be utilized as biomarkers of diagnosis and progression for ESRD. We look forward to further experimental testimonies to validate these findings. Despite a bit of advances made by relying on genomic and bioinformatics analysis, it is vital to be acknowledged that there are still limitations to current study. Above important, although the discovery of expression of hub genes on eccDNA in the PBMCs of ESRD patients has been observed by sequencing, there is a paucity of further validation. Additionally, their expression and mechanisms have not yet been authenticated by functional experiments in ESRD. In the next part, fewer cases were included and failed to be gathered by a single primary disease. ## Conclusions Our genomics study revealed the characteristics and specific expression profiles of eccDNA in the PBMCs from ESRD, enabling us to explore the expression and preliminary functions of eccDNA-targeting genes in the pathogenesis of ESRD. Further research is awaited to analyze and prove the significance of eccDNA to the mechanisms of ESRD. ## Supplementary Information Additional file 1: Table S1. The known disease-related genes from specific eccDNA-targeting genes in ESRD group. ## References 1. Johnson CA, Levey AS, Coresh J, Levin A, Eknoyan JG. **K/DOQI clinical practice guidelines for chronic kidney disease: evaluation, classification, and stratification**. *Am J Kidney Dis* (2002) **39** S1-266. PMID: 11904577 2. Wang J, Zhang L, Tang SC, Kashihara N, Kim YS, Togtokh A. **Disease burden and challenges of chronic kidney disease in North and East Asia**. *Kidney Int* (2018) **94** 22-25. DOI: 10.1016/j.kint.2017.12.022 3. Zhang L, Wang J, Yang CW, Tang SC, Kashihara N, Kim YS. **International Society of Nephrology Global Kidney Health Atlas: structures, organization and services for the management of kidney failure in North and East Asia**. *Kidney Int Suppl.* (2011) **11** e77-e85. DOI: 10.1016/j.kisu.2021.01.011 4. Zhang L, Long J, Jiang W, Shi Y, He X, Zhou Z. **Trends in chronic kidney disease in China**. *N Engl J Med* (2016) **375** 905-906. DOI: 10.1056/NEJMc1602469 5. Malekmakan L, Tadayon T, Roozbeh J, Sayadi M. **End-stage renal disease in the Middle East: a systematic review and meta-analysis**. *Iran J Kidney Dis* (2018) **12** 195-203. PMID: 30087213 6. Malekmakan L, Malekmakan A, Daneshian A, Pakfetrat M, Roosbeh J. **Hypertension and diabetes remain the main causes of chronic renal failure in Fars Province, Iran 2013**. *Saudi J Kidney Dis Transpl* (2016) **27** 423-424. DOI: 10.4103/1319-2442.178594 7. Liyanage T, Ninomiya T, Jha V, Neal B, Patrice HM, Okpechi I. **Worldwide access to treatment for end-stage kidney disease: a systematic review**. *Lancet* (2015) **385** 1975-1982. DOI: 10.1016/S0140-6736(14)61601-9 8. Arogundade FA, Omotoso BA, Adelakun A, Bamikefa T, Ezeugonwa R, Omosule B. **Burden of end-stage renal disease in sub-Saharan Africa**. *Clin Nephrol* (2020) **93** 3-7. DOI: 10.5414/CNP92S101 9. Herzog CA, Asinger RW, Berger AK, Charytan DM, Díez J, Hart RG. **Cardiovascular disease in chronic kidney disease. A clinical update from kidney disease: improving global outcomes (KDIGO)**. *Kidney Int* (2011) **80** 572-586. DOI: 10.1038/ki.2011.223 10. Wong G, Staplin N, Emberson J, Baigent C, Turner R, Chalmers J. **Chronic kidney disease and the risk of cancer: an individual patient data meta-analysis of 32,057 participants from six prospective studies**. *BMC Cancer* (2016). DOI: 10.1186/s12885-016-2532-6 11. Hotta Y, Bassel A. **Molecular size and circularity of DNA in cells of mammals and higher plants**. *Proc Natl Acad Sci USA* (1965) **53** 356-362. DOI: 10.1073/pnas.53.2.356 12. Cox D, Yuncken C, Spriggs AI. **Minute chromatin bodies in malignant tumours of childhood**. *Lancet* (1965) **1** 55-58. DOI: 10.1016/S0140-6736(65)90131-5 13. Møller HD, Mohiyuddin M, Prada-Luengo I, Sailani MR, Halling JF, Plomgaard P. **Circular DNA elements of chromosomal origin are common in healthy human somatic tissue**. *Nat Commun* (2018) **9** 1069. DOI: 10.1038/s41467-018-03369-8 14. Molin WT, Yaguchi A, Blenner M, Saski CA. **Autonomous replication sequences from the**. *BMC Res Notes* (2020) **13** 330. DOI: 10.1186/s13104-020-05169-0 15. Paulsen T, Shibata Y, Kumar P, Dillon L, Dutta A. **Small extrachromosomal circular DNAs, microDNA, produce short regulatory RNAs that suppress gene expression independent of canonical promoters**. *Nucleic Acids Res* (2019) **47** 4586-4596. DOI: 10.1093/nar/gkz155 16. Cohen S, Regev A, Lavi S. **Small polydispersed circular DNA (spcDNA) in human cells: association with genomic instability**. *Oncogene* (1997) **14** 977-985. DOI: 10.1038/sj.onc.1200917 17. Taanman JW. **The mitochondrial genome: structure, transcription, translation and replication**. *Biochim Biophys Acta* (1999) **1410** 103-123. DOI: 10.1016/S0005-2728(98)00161-3 18. Storlazzi CT, Lonoce A, Guastadisegni MC, Trombetta D, D'Addabbo P, Daniele G. **Gene amplification as double minutes or homogeneously staining regions in solid tumors: origin and structure**. *Genome Res* (2010) **20** 1198-1206. DOI: 10.1101/gr.106252.110 19. Barr FG, Nauta LE, Davis RJ, Schäfer BW, Nycum LM, Biegel JA. **In vivo amplification of the PAX3-FKHR and PAX7-FKHR fusion genes in alveolar rhabdomyosarcoma**. *Hum Mol Genet* (1996) **5** 15-21. DOI: 10.1093/hmg/5.1.15 20. Carroll SM, DeRose ML, Gaudray P, Moore CM, Needham-Vandevanter DR, Von Hoff DD. **Double minute chromosomes can be produced from precursors derived from a chromosomal deletion**. *Mol Cell Biol* (1988) **8** 1525-1533. PMID: 2898098 21. Zhang CZ, Spektor A, Cornils H, Francis JM, Jackson EK, Liu S. **Chromothripsis from DNA damage in micronuclei**. *Nature* (2015) **522** 179-184. DOI: 10.1038/nature14493 22. Mehanna P, Gagné V, Lajoie M, Spinella JF, St-Onge P, Sinnett D. **Characterization of the microDNA through the response to chemotherapeutics in lymphoblastoid cell lines**. *PLoS ONE* (2017) **12** e0184365. DOI: 10.1371/journal.pone.0184365 23. Wang M, Chen X, Yu F, Ding H, Zhang Y, Wang K. **Extrachromosomal circular DNAs: origin, formation and emerging function in cancer**. *Int J Biol Sci* (2021) **17** 1010-1025. DOI: 10.7150/ijbs.54614 24. Lv W, Pan X, Han P, Wang Z, Feng W, Xing X. **Circle-Seq reveals genomic and disease-specific hallmarks in urinary cell-free extrachromosomal circular DNAs**. *Clin Transl Med* (2022) **12** e817. DOI: 10.1002/ctm2.817 25. Inker LA, Astor BC, Fox CH, Isakova T, Lash JP, Peralta CA. **KDOQI US commentary on the 2012 KDIGO clinical practice guideline for the evaluation and management of CKD**. *Am J Kidney Dis* (2014) **63** 713-735. DOI: 10.1053/j.ajkd.2014.01.416 26. Panda SK, Ravindran B. **Isolation of human PBMCs**. *Bio Protoc* (2013) **3** e323. DOI: 10.21769/BioProtoc.323 27. Panda SK, Kumar S, Tupperwar NC, Vaidya T, George A, Rath S. **Chitohexaose activates macrophages by alternate pathway through TLR4 and blocks endotoxemia**. *PLoS Pathog* (2012) **8** e1002717. DOI: 10.1371/journal.ppat.1002717 28. Moller HD. **Circle-Seq: isolation and sequencing of chromosome-derived circular DNA elements in cells**. *Methods Mol Biol* (2020). DOI: 10.1007/978-1-0716-0323-9_15 29. Bolger AM, Lohse M, Usadel B. **Trimmomatic: a flexible trimmer for Illumina sequence data**. *Bioinformatics* (2014) **30** 2114-2120. DOI: 10.1093/bioinformatics/btu170 30. Prada-Luengo I, Krogh A, Maretty L, Regenberg B. **Sensitive detection of circular DNAs at single-nucleotide resolution using guided realignment of partially aligned reads**. *BMC Bioinform* (2019) **20** 663. DOI: 10.1186/s12859-019-3160-3 31. Møller HD, Bojsen RK, Tachibana C, Parsons L, Botstein D, Regenberg B. **Genome-wide purification of extrachromosomal circular DNA from eukaryotic cells**. *J Vis Exp* (2016) **110** e54239 32. Quinlan AR, Hall IM. **BEDTools: a flexible suite of utilities for comparing genomic features**. *Bioinformatics* (2010) **26** 841-842. DOI: 10.1093/bioinformatics/btq033 33. Møller HD, Ramos-Madrigal J, Prada-Luengo I, Gilbert MTP, Regenberg B. **Near-random distribution of chromosome-derived circular DNA in the condensed genome of pigeons and the larger, more repeat-rich human genome**. *Genome Biol Evol* (2020) **12** 3762-3777. DOI: 10.1093/gbe/evz281 34. Sin STK, Jiang P, Deng J, Ji L, Cheng SH, Dutta A. **Identification and characterization of extrachromosomal circular DNA in maternal plasma**. *Proc Natl Acad Sci USA* (2020) **117** 1658-1665. DOI: 10.1073/pnas.1914949117 35. Gusev E, Solomatina L, Zhuravleva Y, Sarapultsev A. **The pathogenesis of end-stage renal disease from the standpoint of the theory of general pathological processes of inflammation**. *Int J Mol Sci.* (2021) **22** 11453. DOI: 10.3390/ijms222111453 36. Choi BH, Kang KS, Kwak MK. **Effect of redox modulating NRF2 activators on chronic kidney disease**. *Molecules* (2014) **19** 12727-12759. DOI: 10.3390/molecules190812727 37. Titan SM, Venturini G, Padilha K, Goulart AC, Lotufo PA, Bensenor IJ. **Metabolomics biomarkers and the risk of overall mortality and ESRD in CKD: results from the Progredir Cohort**. *PLoS ONE* (2019) **14** e0213764. DOI: 10.1371/journal.pone.0213764 38. Gandhi G, Mehta T, Contractor P, Tung G. **Genotoxic damage in end-stage renal disease**. *Mutat Res Genet Toxicol Environ Mutagen.* (2018) **835** 1-10. DOI: 10.1016/j.mrgentox.2018.08.005 39. Paulsen T, Kumar P, Koseoglu MM, Dutta A. **Discoveries of extrachromosomal circles of DNA in normal and tumor cells**. *Trends Genet* (2018) **34** 270-278. DOI: 10.1016/j.tig.2017.12.010 40. Wu S, Turner KM, Nguyen N, Raviram R, Erb M, Santini J. **Circular ecDNA promotes accessible chromatin and high oncogene expression**. *Nature* (2019) **575** 699-703. DOI: 10.1038/s41586-019-1763-5 41. Cai M, Zhang H, Hou L, Gao W, Song Y, Cui X. **Inhibiting homologous recombination decreases extrachromosomal amplification but has no effect on intrachromosomal amplification in methotrexate-resistant colon cancer cells**. *Int J Cancer* (2019) **144** 1037-1048. DOI: 10.1002/ijc.31781 42. de Mendoza A, Sebé-Pedrós A. **Origin and evolution of eukaryotic transcription factors**. *Curr Opin Genet Dev.* (2019) **58** 25-32. DOI: 10.1016/j.gde.2019.07.010 43. Jetten AM. **GLIS1-3 transcription factors: critical roles in the regulation of multiple physiological processes and diseases**. *Cell Mol Life Sci* (2018) **75** 3473-3494. DOI: 10.1007/s00018-018-2841-9 44. Dai H, Zhou J, Zhu B. **Gene co-expression network analysis identifies the hub genes associated with immune functions for nocturnal hemodialysis in patients with end-stage renal disease**. *Medicine* (2018) **97** e12018. DOI: 10.1097/MD.0000000000012018 45. Geng XD, Wang WW, Feng Z, Liu R, Cheng XL, Shen WJ. **Identification of key genes and pathways in diabetic nephropathy by bioinformatics analysis**. *J Diabetes Investig* (2019) **10** 972-984. DOI: 10.1111/jdi.12986 46. Li W, Tan Y, Gao F, Xiang M. **Overexpression of TRIM3 protects against LPS-induced acute kidney injury via repressing IRF3 pathway and NLRP3 inflammasome**. *Int Urol Nephrol.* (2022). DOI: 10.1007/s11255-021-03017-z 47. Huang YS, Lu KC, Chao TK, Chen JS, Chen A, Guo CY. **Role of melatonin receptor 1A and pituitary homeobox-1 coexpression in protecting tubular epithelial cells in membranous nephropathy**. *J Pineal Res* (2018) **65** e12482. DOI: 10.1111/jpi.12482 48. Lanata CM, Nititham J, Taylor KE, Chung SA, Torgerson DG, Seldin MF. **Genetic contributions to lupus nephritis in a multi-ethnic cohort of systemic lupus erythematous patients**. *PLoS ONE* (2018) **13** e0199003. DOI: 10.1371/journal.pone.0199003 49. Ma RC, Tam CH, Wang Y, Luk AO, Hu C, Yang X. **Genetic variants of the protein kinase C-beta 1 gene and development of end-stage renal disease in patients with type 2 diabetes**. *JAMA* (2010) **304** 881-889. DOI: 10.1001/jama.2010.1191 50. Zhou L, Li Y, Hao S, Zhou D, Tan RJ, Nie J. **Multiple genes of the renin-angiotensin system are novel targets of Wnt/β-catenin signaling**. *J Am Soc Nephrol* (2015) **26** 107-120. DOI: 10.1681/ASN.2014010085 51. Köttgen A, Pattaro C, Böger CA, Fuchsberger C, Olden M, Glazer NL. **New loci associated with kidney function and chronic kidney disease**. *Nat Genet* (2010) **42** 376-384. DOI: 10.1038/ng.568 52. Puthumana J, Thiessen-Philbrook H, Xu L, Coca SG, Garg AX, Himmelfarb J. **Biomarkers of inflammation and repair in kidney disease progression**. *J Clin Invest.* (2021). DOI: 10.1172/JCI139927 53. Manchanda PK, Singh R, Mittal RD. **Cytokine (IL-10 -1082 and -819) and chemokine receptor (CCR2 and CCR5) gene polymorphism in North Indian patients with end-stage renal disease**. *DNA Cell Biol* (2009) **28** 177-183. DOI: 10.1089/dna.2008.0822 54. Otterpohl KL, Busselman BW, Ratnayake I, Hart RG, Hart KR, Evans CM. **Conditional Myh9 and Myh10 inactivation in adult mouse renal epithelium results in progressive kidney disease**. *JCI Insight.* (2020) **5** e138530. DOI: 10.1172/jci.insight.138530 55. Torban E, Sokol SY. **Planar cell polarity pathway in kidney development, function and disease**. *Nat Rev Nephrol* (2021) **17** 369-385. DOI: 10.1038/s41581-021-00395-6 56. Du J, Hu Z, Mitch WE. **Molecular mechanisms activating muscle protein degradation in chronic kidney disease and other catabolic conditions**. *Eur J Clin Invest* (2005) **35** 157-163. DOI: 10.1111/j.1365-2362.2005.01473.x 57. Kim EY, Anderson M, Dryer SE. **Sustained activation of N-methyl-D-aspartate receptors in podoctyes leads to oxidative stress, mobilization of transient receptor potential canonical 6 channels, nuclear factor of activated T cells activation, and apoptotic cell death**. *Mol Pharmacol* (2012) **82** 728-737. DOI: 10.1124/mol.112.079376 58. Tian D, Jacobo SM, Billing D, Rozkalne A, Gage SD, Anagnostou T. **Antagonistic regulation of actin dynamics and cell motility by TRPC5 and TRPC6 channels**. *Sci Signal.* (2010) **3** ra77. DOI: 10.1126/scisignal.2001200 59. Greka A, Mundel P. **Calcium regulates podocyte actin dynamics**. *Semin Nephrol* (2012) **32** 319-326. DOI: 10.1016/j.semnephrol.2012.06.003 60. Poli A, Billi AM, Mongiorgi S, Ratti S, McCubrey JA, Suh PG. **Nuclear phosphatidylinositol signaling: focus on phosphatidylinositol phosphate kinases and phospholipases C**. *J Cell Physiol* (2016) **231** 1645-1655. DOI: 10.1002/jcp.25273 61. Yu X, Xia Y, Zeng L, Zhang X, Chen L, Yan S. **A blockade of PI3Kγ signaling effectively mitigates angiotensin II-induced renal injury and fibrosis in a mouse model**. *Sci Rep* (2018) **8** 10988. DOI: 10.1038/s41598-018-29417-3 62. Kang SS, Chang JW, Park Y. **Nutritional status predicts 10-year mortality in patients with end-stage renal disease on hemodialysis**. *Nutrients* (2017) **9** 399. DOI: 10.3390/nu9040399 63. Garibotto G, Bonanni A, Verzola D. **Effect of kidney failure and hemodialysis on protein and amino acid metabolism**. *Curr Opin Clin Nutr Metab Care* (2012) **15** 78-84. DOI: 10.1097/MCO.0b013e32834d9df6
--- title: 'What are the current diabetic foot assessment methods in private podiatry practices in Flanders, Belgium: an exploratory mixed method study' authors: - Irene Vansteenland - Rachel Forss journal: Journal of Foot and Ankle Research year: 2023 pmcid: PMC10041772 doi: 10.1186/s13047-023-00615-1 license: CC BY 4.0 --- # What are the current diabetic foot assessment methods in private podiatry practices in Flanders, Belgium: an exploratory mixed method study ## Abstract ### Background Diabetic foot assessments detect patients at risk for developing a diabetes-related foot ulceration and can significantly reduce the risk of amputation. In order to organize this assessment effectively, diabetic foot assessment guidelines are required according to the International Working Group of the Diabetic Foot. However, these international guidelines have not been adapted into a national guideline for podiatrists in Flanders, Belgium. This study aims to identify the methods and guidelines currently used to assess the diabetic foot in private podiatry practices in Flanders, Belgium and to explore the podiatrists’ opinions on developing a national diabetic foot assessment guideline. ### Methods This exploratory mixed method study was composed of an anonymous online survey comprising of open- and closed-ended questions followed by 1:1 online semi-structured interviews. Participants were recruited via e-mail and a closed private Facebook group of podiatry alumni. Data was analyzed using SPSS statistics and thematic analysis described by Braun and Clarke. ### Results This study showed that the vascular assessment of the diabetic foot exists solely of a medical history and palpation of the pedal pulses. Non-invasive tests such as doppler, toe brachial pressure index or ankle brachial pressure index are seldom used. Only $66\%$ reported to use a guideline for the diabetic foot assessment. There was a variety of reported guidelines and risk stratification systems in use in private podiatry practices in Flanders, Belgium. ### Conclusion Non-invasive tests such as the doppler, ankle brachial pressure index or toe brachial pressure index are rarely used for the vascular assessment of the diabetic foot. Diabetic foot assessment guidelines and risk stratification systems to identify patients at risk for developing a diabetic foot ulcer were not frequently used. International guidelines of the International Working Group of the Diabetic Foot have not yet been implemented in private podiatry practices in Flanders, Belgium. This exploratory research has provided useful information for future research studies. ## Background The prevalence of diabetes mellitus (DM) is increasing at an alarming rate. In 2018, the prevalence of diabetes in the Belgian population has increased to $6.1\%$ as a result of the population ageing and an increase in overweight or obesity [1]. Diabetes-related foot ulceration (DFU) is one of the most prevalent and serious complications of DM [2]. The annual incidence of DFUs in *Belgium is* $2\%$ and the lifetime risk of developing a DFU has been estimated between 19 and $34\%$ [3, 4]. Moreover, recent data suggest that between 25 to $40\%$ of patients with a history of a DFU experience a recurrence within 1 year after the ulcer has healed [3, 4]. DFUs are the most common cause of non-traumatic amputation in Western countries, with $85\%$ of all lower limb amputations reported as being preceded by a DFU [5]. Early detection of the patients at risk for developing a DFU, through a diabetic foot assessment, can significantly reduce the risk of amputation [6, 7]. The study of Lavery et al. showed that implementation of a diabetic foot assessment with complementary preventive and acute care services, according to patient’s risk factor, reduces the incidence of amputations by $47\%$ [7]. Peer reviews of the diabetic foot services in the South-West region of the UK indicated that introducing regular foot examinations and offering advice or referral to preventive and acute services decreases the incidence of DFU related major amputations [8]. Podiatrists have an essential role in performing these foot assessments and providing preventive or acute services. The study of Blanchette et al. [ 9] reported that multidisciplinary teams with podiatry services lead to a significant reduction in lower extremity amputations. Moreover, providing podiatry services for patients with diabetes before the onset of a DFU reduces hospital admissions [10]. In Belgium, the initial diabetic foot assessment is carried out by the general practitioner (GP). When patients are at low or moderate risk of developing a DFU, they are referred to podiatrists, working in the private sector, for annual diabetic foot assessments and prevention services [11]. In order to organize these assessments and services effectively, guidelines are required according to the International Working Group of the Diabetic Foot (IWGDF) [12]. The IWGDF has developed an evidence-based international diabetic foot assessment guideline for all health care professionals [13, 14]. This practical guideline is aimed at the global community of health care professionals involved in the diabetic foot care [14]. The working group recommend that those guidelines may have to be adapted based on local circumstances taking into account accessibility to health care resources and various cultural factors [14]. However, to date, these international diabetic foot assessment guidelines have not been adapted into a national guideline for podiatrists working in the private sector in Flanders, Belgium. Therefore, the primary objective of this research was to examine which methods and guidelines are currently used to assess the diabetic foot in private podiatry practices in Flanders, Belgium. The second objective of this study was to explore the podiatrists’ perceptions on developing a national guideline for the diabetic foot assessment in Flanders, Belgium. ## Methods To our knowledge, this is the first research investigating the diabetic foot assessment methods in private podiatry practices in Flanders, Belgium. Therefore, an exploratory mixed method research was conducted. The quantitative phase of this research involved the collection of data by using an anonymous online survey to determine which methods and guidelines are currently used to assess the diabetic foot. A sequential qualitative phase followed the quantitative phase to clarify the results of the survey an to explore podiatrists’ perceptions on developing a national diabetic foot assessment guideline. This phase consisted of online 1:1 interviews. The target population of this study were podiatrists registered with the Belgian National Institute for Health and Disability Insurance (NIHDI) and working in the private sector in Flanders, Belgium. The School of Health Sciences Research Ethics Panel, University of Brighton approved the study on the 11th of March 2021. ## Study design An anonymous online survey comprising of 6 open- and 8 closed-ended questions was drawn up to generate quantitative data. These questions were generated following a literature review. The IT services of the Jisc UK digital, data and technology agency were used to host the online survey. The front page of the survey included information on storage of data, purpose of the study and data protection. Participants had to give their informed consent in order to gain access to this survey. The survey questions were distributed on 2 screen pages and were translated in English by a translator. Participants could choose to write the answers to the open-ended questions in their preferred language to encourage participation. They could review and change their answers before finishing the survey. All questions had to be completed in order to finish the survey. A pilot of this survey was completed by the researcher, supervisor and an experienced podiatrist, prior to sending out the survey links. From May until July 2021, invitations to participate in the online survey were sent via e-mail to 362 podiatrists working in the private sector in Flanders, Belgium. These e-mail addresses were retrieved from an internet search and were publicly available. The survey was also promoted on the private Facebook group of all podiatry alumni of the Artevelde University in Flanders, Belgium. Reminder e-mails and Facebook posts were sent two weeks after the initial invitation or post to improve the response rate. In order to retrieve qualitative data on podiatrist’s perceptions on developing a national guideline, survey respondents were asked in the survey to indicate if they were interested in participating in a one-on-one semi-structured interview. Only 9 out of the 50 participants indicated they would like to participate in these interviews. Invitations for these interviews were sent via e-mail in July 2021. After the invitation e-mails were sent out, only 4 volunteers were interested to take part in the next step of this research. The researcher conducted the interviews in Dutch. These interviews were held online and recorded via Microsoft Teams. The interviews recordings were deleted after transcription. The interviews lasted approximately 1 h and were carried out once per participant, repeat interviews were not conducted. The researcher developed an interview guide which outlined the structure of the interview and was used as a tool to check if every question was answered by the interviewees. This guide was reviewed by the supervisor prior to the first interview. The interviews were coded using thematic inductive analysis described by Braun and Clarke [15]. The 3 core themes developed from this analysis are presented in Table 1 and the identified subthemes are also illustrated. Participating in the survey and interviews was voluntary. There were no incentives or rewards offered to participate in this research. Table 1Themes thematic analysisThemesSubthemesGuidance documents or guidelines for the diabetic foot assessment-Use of guidelines in private podiatry practices-Retrieving guidelines-Communication of the latest international guidelinesDiabetic foot assessment-Use of a Doppler-Inconsistencies in the interpretation of the diabetic foot risk stratification systemNeed for a change-Lack of referral pathway from the general practitioner to the private podiatrist-Reimbursement for podiatry consultations in Belgium *Inclusion criteria* for this research were registered podiatrists working part-time or full-time in the private sector in Flanders, Belgium. Therefore, podiatrists working solely in multidisciplinary diabetic foot clinics in the public sector (MDFCs) were excluded from the study, as these clinics are bound to guidelines for clinical audits. Other health care professionals such as diabetes nurses, physiotherapists and GPs were also excluded from the study. The survey was only available in English and Dutch. As a result, the link to the survey was not sent out to podiatrists working in the private sector in the French part of Belgium, Wallonie. These podiatrists were therefore also excluded from the study. ## Data analysis Data analysis was conducted solely by the researcher. Data from the survey were kept within the “JISC online survey” tool and analyzed using Excel (Microsoft 365) and SPSS Statistics version 26 for Windows (IBM Corporation). Categorical data were compared using a Fishers exact test. A p-value of < 0.05 was considered statistically significant. Data from free text responses to the survey questions and online interviews were coded using the inductive thematic analysis described by Braun and Clarke [15]. Personal details of the interviewees were removed from the interview transcripts. Thematic analysis of the qualitative data involved familiarization and analysis of data, developing core themes around podiatrists’ experiences and opinions and reviewing these themes before reporting the results. The core themes and most interesting quotes were translated in English after the thematic analysis. All interview transcripts were returned to the participants for comments and/or corrections. As a result, member checking along with the methodological triangulation, using survey and interviews for data collection, helped to mitigate bias in coding. ## Survey Out of the 362 podiatrists contacted in Flanders, Belgium, a total of 50 participated in the survey ($14\%$ response rate). Demographic data from the respondents are presented in Table 2.Table 2Demographics of respondentsDuration of employment as a podiatrist a:($$n = 50$$)•0–5 years$50\%$ [25]•5–10 years$16\%$ [8]• > 10 years$34\%$ [17]Practice office setting of podiatrists:($$n = 50$$)•Private practice$40\%$ [20]•Part-time hospital/ Part-time private (multidisciplinary) practice$32\%$ [16]•Multidisciplinary private practice$28\%$ [14]Number of diabetic patients treated in an average week a:($$n = 50$$)•Less than 5 patients$38\%$ [19]•Between 5 and 10 patients$24\%$ [12]•More than 10 patients per week$38\%$ [19]aVariables were not mutually exclusive. Results should be interpreted with caution ## Diabetic foot assessment methods The survey investigated which screening assessment methods were routinely performed to identify diabetic neuropathy and periperal arterial disease (PAD). These results are presented in Fig. 1 and 2. The results of the survey indicate that the most reported screening tests for diabetic neuropathy are history of symptoms of neuropathy ($82\%$) and the 10 g monofilament test ($98\%$) (see Fig. 1). In addition to history and 10 g monofilament test, at least one other test was used by $72\%$ of the respondents such as the 128 Hz tuning fork or the Ipswich Touch test [16].Fig. 1Screening tests for diabetic neuropathy. Detailed summary of tests used to assess diabetic neuropathy in private podiatry practices in Flanders, Belgium. $$n = 50$$ podiatristsFig. 2Screening tests for peripheral arterial disease. Detailed summary of tests used to perform the vascular assessment of the diabetic foot in private podiatry practices in Flanders, Belgium. $$n = 50$$ podiatrists For the assessment of PAD, a minimum combination of medical history, palpation of the pedal pulses and capillary refill time was used by $68\%$ of the respondents. An audible handheld doppler, ankle brachial pressure index (ABPI) and toe brachial pressure index (TBPI) were seldom used ($18\%$, $4\%$ and $6\%$ respectively). However, when podiatrists were asked which tests they are not currently performing but would like to, $22\%$ reported they would like to use an audible handheld doppler and $6\%$ would like to perform an ABPI. The survey investigated the use of complementary tests for the diabetic foot assessment. Inspection of foot deformities, footwear (shoes and socks) and skin/nails/wounds was reported by $6\%$, $8\%$ and $18\%$ of the podiatrists respectively. Clinical tests to detect joint mobility was the most reported complementary test ($28\%$). ## Diabetic foot assessment guidelines or guidance documents and risk stratification systems Sixty-six percent of all podiatrists use guidance documents or guidelines for the assessment of the diabetic foot. The Fisher’s exact test was used to determine the relationship between years of experience and the use of guidelines or guidance document for diabetic foot assessments. Although recently graduated podiatrists are introduced to the latest guidelines of the IWGDF, the test showed that they are not more likely to use guidelines compared to experienced podiatrists (fisher exact test. $$p \leq 0$$,837). The findings of this survey indicate that a minimum of 9 different guidance documents or guidelines are used by private podiatrists working in Flanders, Belgium. Most of the podiatrists ($24\%$) develop their own guidance documents or guidelines for the diabetic foot assessment. $21\%$ of the respondents did not specify the name of the document used for this assessment while others reported published risk guidance classification documents such as the guidance documents available on patient management software designed for podiatrists ($12\%$), the Sims classification ($9\%$) [17] or the perfusion, extent, depth, infection and sensation classification system ($9\%$) [18] (PEDIS), the IWGDF guidelines ($6\%$) [14], documents retrieved from MDFC ($9\%$), DN4 and VAS-score questionnaire ($6\%$) and documents retrieved from the podiatry undergraduate course ($3\%$). Risk stratification systems are used to determine the risk of developing a DFU and are often used to determine the frequency of podiatry visits. However, only $66\%$ of the respondents use a diabetic foot risk stratification system. The most popular system ($42\%$) is the one provided by the NIHDI [19] to determine the reimbursement of podiatry consultations for patients with diabetes. The “SIMS classification” [17] is the second most described risk stratification system in use ($36\%$). ## Frequency of diabetic foot assessments Figure 3 shows how podiatrists determine the frequency of diabetic foot assessments. It highlights the two most commonly cited reasons for assessment frequency being the risk classification of the patient ($44\%$) and depending on the reimbursement arrangements for podiatry consultations ($16\%$). Other reported reasons were the frequency of scheduled appointments ($6\%$), presence of a foot wound ($4\%$) or patients request for a diabetic foot assessment ($2\%$). Some podiatrists perform the diabetic foot assessment only during annual ($6\%$) or bi annual ($12\%$) podiatry consultations. Fig. 3Frequency of diabetic foot assessments. How often and when do podiatrists perform a diabetic foot assessment in private podiatry practices in Flanders, Belgium. $$n = 50$$ podiatrists ## Guidance documents or guidelines for the diabetic foot assessment The majority of the participants interviewed reported the use of a variety of guidelines for the diabetic foot assessment in private practices, which was also apparent in the survey results. Furthermore, these guidelines and guidance documents were retrieved in various ways.“ I attend the International Symposium on the Diabetic Foot every year”“ We retrieved guidelines or guidance documents through conferences, guidelines from other countries and following Prof. Dr. Armstrong on social media”“ I use the guidance document that I received during my undergraduate internship” When asked about possibly introducing a national diabetic foot assessment guideline, the participants suggested that it would be impossible to introduce a national guideline immediately. They reported that the process of developing and introducing a national guideline should start with increasing the Belgian private podiatrists’ awareness of the IWGDF guidelines. “Communicating the latest international standards on diabetic foot assessment, through BVP-ABP (the podiatric medical association in Belgium), would be a step in the right direction. Although, we do not know if podiatrists will read this information.” ## Diabetic foot assessment It was apparent in the survey results that podiatrists use a minimum combination of medical history, palpation of the pedal pulses and capillary refill for the vascular assessment. The interviews explored why highly recommended methods such as a Doppler or ABPI are not integrated in the diabetic foot assessment in private podiatry practices in Belgium. The majority of participants agreed with the statement that it is not worth investing in the costly equipment for ABPI or Doppler. “You have to buy additional equipment and I am wondering what is the cost–benefit analysis?”“*It is* quite expensive and you do not need it that often. I would rather invest in other equipment that I need for daily use”. This view was echoed by another participant who stated that besides this costly investment, the current pricing of podiatry consultations and non-existent referral pathways for patients with diabetes hold podiatrists back to invest in the equipment needed. “The pricing of the podiatry consultations has not been correctly determined. I really would like to invest in this equipment but if I only need to use it once a year due to the lack of referral of patients with diabetes, it is not worth the investment. Another common view amongst interviewees was the inconsistencies in the interpretation of the diabetic foot risk factor between different private podiatrists. It was suggested that although there is no national guideline on diabetic foot assessment, podiatrists use the same assessment methods. However, when the results of this assessment must be interpreted there could be inconsistencies in how these are used to stratify the patients risk factor for developing a DFU.“Patient with a medical history of revascularization, presenting with good palpable pedal pulses remains a high risk foot in my opinion. However, colleagues could interpret this differently. They could reason that because the pedal pulses are palpable the patient would fall back to the low risk category.” ## Need for change A recurrent theme in the interviews was a sense amongst interviewees that some issues in assessing the diabetic foot highlight the need for change in the diabetic foot care. Firstly, there is no established referral pathway from the GP to the private podiatrist in Belgium. “There are few GPs referring patients with diabetes for a diabetic foot assessment. I have been trying for years to change this, without any result. They only refer patients when DFUs occur.” Secondly, the lack of referral could be attributed to the fact that GPs are not aware of the professional capabilities of podiatrists. “I think GPs have no idea of what the podiatrists’ professional capabilities really are. They suppose we solely perform diabetic foot care and do not realize we perform a thorough diabetic foot assessment prior to the foot care. What does the GP expect from the podiatrist and vice versa? I think this needs further discussion.” Lastly, podiatry consultations are only reimbursed twice a year, which is not sufficient for patients with a moderate to high risk of developing a DFU.“A lot of high risk patients need frequent podiatry consultations. However, they also have a lot of other medical expenses, so they stick to the two reimbursed podiatry consultations a year because they can’t afford it. If we could treat these patients every month, this would reduce the occurrence of diabetic foot problems remarkably. ”“Patients with diabetes have a lot of medical expenses. Podiatry consultations cost between 30 to 35 euros. If you have to pay this out of your own pocket every time, these medical expenses will increase and some patients will eventually cut out these expenses.” ## Discussion To our knowledge, this is the first exploratory mixed method study that evaluated the podiatrists’ experiences and methods in assessing the diabetic foot in the private sector in Flanders, Belgium. The major findings of this research were firstly the lack of use of non-invasive tests for the vascular assessment of the diabetic foot. Secondly, only $66\%$ of the respondents use guidelines to assess the DF. Moreover, this research has shown that the IWGDF guidelines have not yet been implemented in the private podiatry practices in Flanders, Belgium. Lastly, one third of the respondents do not use a risk stratification system to identify patients at risk for developing a DFU. The most important finding was the limitation of non-invasive tests used for the vascular assessment of the diabetic foot. PAD is an independent risk factor for subsequent DFU [20]. Studies have shown that it is present in up to $50\%$ of patients presenting with a DFU [21, 22]. Moreover, previous research has established that patients with diabetes with PAD were five times more likely to have undergone a lower-extremity amputation (LEA) and had higher mortality compared to non-diabetes patients [23–25]. The UK NICE guidelines recommend to assess the vascular status as an important predictor of ulceration in the diabetic foot [26]. Considering this evidence, it seems that identification of PAD in patients with diabetes is key in minimizing the risk of LEA. The results of the survey and interviews showed that the vascular assessment in private podiatry practices in Flanders, Belgium solely exists of a medical history and palpation of the pedal pulses. These results are consistent with previous research, which evaluated the vascular assessment techniques of podiatrists in the UK [27]. Nevertheless, the IWGDF guidelines suggest that the presence of palpable foot pulses cannot be used in isolation to reliably exclude PAD [28]. Pedal pulse examination has a poor sensitivity and is not independently sufficient to conclusively diagnose PAD [29, 30]. Therefore, guidelines recommend a more objective evaluation with a Doppler, ABPI or TBPI for identifying PAD [28–30]. Research has shown that non-invasive testing such as Doppler and TBPI are more accurate and viable screening tests to identify PAD among patients with diabetes [31–33]. TBPI is often preferred for diagnostic testing because research has shown that this test provides a more accurate diagnosis in patients with diabetes with carotid atherosclerosis compared to ABPI [32, 34, 35]. The results of the interviews of this study indicated that the lack of podiatry consultation reimbursement and referral pathways hinders private podiatrists in Flanders, Belgium to invest in the equipment needed for noninvasive testing. Lack of equipment has been reported as a frequent barrier to performing a vascular assessment in previous studies performed in the UK and Australia [36, 37]. This shows that in order to improve the quality of diabetic foot assessments, podiatrists should get the opportunity to invest in proper equipment. The second finding of this research was that the IWGDF guidelines have not yet been implemented in private podiatry practices in Flanders, Belgium. Since 1999, the IWGDF has developed international clinical practice guidelines for the prevention and management of the diabetic foot [14]. These guidelines are systematically developed statements to assist health care professionals’ decisions, to standardize the diabetic foot care and improve the quality of health care [38, 39]. The IWGDF advises that those guidelines may have to be adapted based on local circumstances taking into account accessibility to health care resources and various cultural factors [14]. When nations consider to develop a national diabetic foot assessment guideline, it is advised to adopt a similar methodology to that used by the IWGDF [40, 41]. Moreover, the guidelines must be as specific as possible to reduce ambiguity and confusion among clinicians managing patients with DFUs [40]. Several studies have shown that developing a national diabetic foot assessment guideline, based on the international recommendations, not only increases the frequency of diabetic foot assessments [42, 43] but also reduces the incidence of diabetes-related LEA [19, 44]. However, in Belgium, the international recommendations have not yet been implemented into a national diabetic foot assessment guideline. This could be the reason why only $66\%$ of all private podiatrists in Flanders, Belgium reported to use guidance documents or guidelines for the assessment of the diabetic foot. Solely $6\%$ of these podiatrists are using the IWGDF guidelines. It also raises a question to what are the other $34\%$ using? Although, with only a response rate of $14\%$, we do not have a true representative sample of private podiatrists in Flanders, Belgium and these results must therefore be interpreted with caution. Another possible explanation for the lack of implementation of diabetic foot assessment guidelines in private podiatry practices in Flanders, Belgium could be the variety of available guidelines or guidance documents published by various organizations and experts in the field [45, 46]. This could create confusion among podiatrists as to which guidelines should be implemented in clinical practice and explains why 9 different guidelines or guidance documents were identified as being used in private podiatry practices in Flanders, Belgium. Moreover, studies have shown that there is a high variability in the recommended methods for the diabetic foot assessment and a lack of consistency regarding the levels of evidence and grades of these recommended methods between different guidelines [45, 47, 48]. As a result, the variation of guidelines used in private podiatry practice in Flanders, Belgium could lead to differences in interpretation of the diabetic foot risk stratification system between podiatrists and affect the quality of diabetic foot care ultimately received by the patient. The concerns regarding the inconsistent interpretation of the diabetic foot risk stratification system among the private podiatrists were widespread in the interviews. This concern could be explained by the lack of implementation of these systems in podiatry practice, which was apparent in the survey results. One third of the respondents do not use a risk stratification system to identify patients at risk for developing a DFU. Diabetic foot risk stratification systems are designed to determine the appropriate management and assessment frequency of the diabetic foot [46]. Studies of Boyko et al. [ 49] and Leese et al. [ 50] showed that risk stratification systems based on the IWGDF guidelines have an excellent ability in accurately quantifying and defining an individual’s risk of developing a DFU. Moreover, it provides a more accurate prediction of the foot ulcer risk than the individual results considered in isolation [49]. The survey results indicated that one-third of the podiatrists are not using any risk stratification system. As a result, it could be assumed that these podiatrists determine the patients risk factor based on individual predictors potentially resulting in inconsistent diabetic foot risk scores. Podiatrists that did use a risk stratification system, most frequently rely on the system provided by the Belgian NIHDI [51] (Table 3) which is based on the on the Coleman’s risk stratification [52, 53]. Although this system is provided by the Belgian Institute, it is important to note that this risk stratification has never been adapted to the latest research or recommendations of the IWGDF. Moreover, there are no studies that have validated this risk stratification system, which could explain why there could be inconsistencies in the diabetic foot risk stratification interpretation between podiatrists. Therefore, adopting a new risk stratification system in compliance with the latest international recommendations could decrease the inconsistencies in the interpretation of the diabetic foot risk stratification between podiatrists. This would ultimately improve the diabetic foot care for patients at risk for developing a DFU.Table 3Risk stratification system NIDHI & IWGDFRisk GroupRisk classification Belgian NIDHI [40]IWGDF guidelines [9]0-no LOPS or PAD1LOPSLOPS or PAD2A)Moderate foot deformities such as prominence of metatarsal heads, hyperkeratosis and/or flexible hammer- or claw toes and/or moderate hallux abducto valgus (< 30°)B)Severe foot deformitiesLOPS + PADLOPS + foot deformityPAD + foot deformity3PADHistory of DFUAmputationCharcotLOPS or PAD and one or more of the following:•History of foot ulcer•LEA (minor or major)•End-stage renal diseaseLOPS Loss of protective sensation, PAD Peripheral arterial disease, DFU Diabetic foot ulcer, LEA Lower extremity amputation ## Limitations The present study has presented a number of limitations and the results should be interpreted cautiously. Firstly, the trustworthiness of this study was subject to certain limitations. The transferability of this research was affected by the small sample size. Invitations for the survey and interviews were solely sent out to private podiatrists in Flanders, Belgium. Moreover, only a small proportion of private podiatrists responded ($14\%$) resulting in a poor external validity. Moreover, this study was conducted as an exploratory mixed method research. Therefore, there was no pilot study or peer review performed prior to this research to validate the survey and interview questions. This resulted in a poor reliability and dependability. In order to generalize the results nationally, this study should be repeated including podiatrists working in the private sector in Wallonie and measures should be undertaken to improve the response rate and to validate the survey and interview questions. Secondly, the survey limited the researcher in her ability to further explore the content of the different risk stratification systems used in private podiatry practice. The second most reported risk stratification used was the “SIMS risk stratification system”. This system was developed in 1988 [17] and introduced in the neighboring country the Netherlands. According to diabetic foot assessment guidelines in the Netherlands, their risk stratification system has been adapted in 2006 to the current IWGDF guidelines. However, the term “SIMS” was kept as podiatrists kept associating the term with diabetic foot risk stratification [54]. As a result it is possible that Belgian private podiatrists reporting the use of the “SIMS risk stratification system” actually use the current IWGDF guidelines, not the original reported system of 1988, which could have influenced the results related to this survey question. Lastly, the audit reports from the MFCS in secondary care in Belgium provide the only data available on national diabetic foot care. The lack of organization of services for the diabetic foot in the private sector results in a gap of knowledge on the current diabetic foot assessment methods and foot care. Furthermore, it makes it impossible to analyze how the current practice could influence the results of the audit report from the MFCS. ## Conclusion The findings of this study demonstrated that podiatrists in the private sector in Flanders, Belgium rarely use non-invasive tests such as the audible handheld doppler, ABPI or TBPI for the vascular assessment of the diabetic foot. Diabetic foot assessment guidelines and risk stratification system to identify patients at risk for developing a DFU are not frequently used in private podiatry practices. Furthermore, the international guidelines of the IWGDF have not yet been implemented in these practices, which highlights the need for the development of a uniform national diabetic foot assessment guideline. These findings are important targets for further investigation. In order to generalize these results, future research targeting registered podiatrists working in the private and public sectors in *Belgium is* needed. Conducting a pilot study or peer review to validate the questionnaire could improve the reliability of these future studies. ## Authors information Forss *Rachel is* a senior lecturer in Podiatry at the University of Brighton, UK. Vansteenland *Irene is* a Msc Podiatry candidate at the University of Brighton, UK. She is the owner of the podiatry practice “Podoconsult Vansteenland Irene” in Lievegem, Belgium and works as a private podiatrist in “Dermatologie Gent” and the Multidisciplinary Diabetic Foot Clinic of AZ-Sint Jan Brugge in Belgium. ## References 1. 1.Steeds meer belgen met diabetes [press release]. 2020. 2. Everett E, Mathioudakis N. **Update on management of diabetic foot ulcers**. *Ann N Y Acad Sci* (2018.0) **1411** 153-165. DOI: 10.1111/nyas.13569 3. Armstrong DG, Boulton AJM, Bus SA. **Diabetic foot ulcers and their recurrence**. *N Engl J Med* (2017.0) **376** 2367-2375. DOI: 10.1056/NEJMra1615439 4. 4.Doggen, 2018, Initiatief voor kwaliteitsbevordering en epidemiologie bij multidisciplinaire diabetes voetklinieken (IKED-voet)`; resultaten van de 5de gegevensverzameling. 5. Holt RIG, Cockram CS, Flyvbjerg A, Goldstein BJ. *Textbook of diabetes* (2017.0) 6. 6.Apelqvist J, Larsson J. What is the most effective way to reduce incidence of amputation in the diabetic foot? Diabetes Metabolism Research and Reviews. 2000;16;(4;S1):S75-S83. 7. Lavery LA, Hunt NA, Lafontaine J, Baxter CL, Ndip A, Boulton AJM. **Diabetic Foot Prevention: a neglected opportunity in high-risk patients**. *Diabetes Care* (2010.0) **33** 1460-1462. DOI: 10.2337/dc10-0310 8. Paisey RB, Abbott A, Levenson R, Harrington A, Browne D, Moore J. **Diabetes-related major lower limb amputation incidence is strongly related to diabetic foot service provision and improves with enhancement of services: peer review of the South-West of England**. *Diabet Med* (2018.0) **35** 53-62. DOI: 10.1111/dme.13512 9. Blanchette V, Brousseau-Foley M, Cloutier L. **Effect of contact with podiatry in a team approach context on diabetic foot ulcer and lower extremity amputation: systematic review and meta-analysis**. *J Foot Ankle Res* (2020.0) **13** 15. DOI: 10.1186/s13047-020-0380-8 10. Gibson TB, Driver VR, Wrobel JS, Christina JR, Bagalman E, DeFrancis R. **Podiatrist care and outcomes for patients with diabetes and foot ulcer: podiatrist care and diabetic foot ulcer**. *Int Wound J* (2014.0) **11** 641-648. DOI: 10.1111/iwj.12021 11. 11.Koeck P, Bastiaens H, Benhalima K, Cloetens H, Feyen L, Sunaert P, et al. Diabetes Mellitus type 2. In: Medica D, editor.: Domus Medica; 2015. 12. Schaper NC, Van Netten JJ, Apelqvist J, Lipsky BA, Bakker K. **Prevention and management of foot problems in diabetes: a summary guidance for daily practice 2015, based on the IWGDF guidance documents**. *Diabetes Res Clin Pract* (2017.0) **124** 84-92. DOI: 10.1016/j.diabres.2016.12.007 13. Bus SA, Lavery LA, Monteiro-Soares M, Rasmussen A, Raspovic A, Sacco ICN. **Guidelines on the prevention of foot ulcers in persons with diabetes (IWGDF 2019 update)**. *Diabetes/metabol Res Rev.* (2020.0) **36** e3269-n/a 14. Schaper NC, van Netten JJ, Apelqvist J, Bus SA, Hinchliffe RJ, Lipsky BA. **Practical Guidelines on the prevention and management of diabetic foot disease (IWGDF 2019 update)**. *Diabetes/metabol Res Rev.* (2020.0) **36** e3266-n/a. DOI: 10.1002/dmrr.3266 15. Braun V, Clarke V. **Using thematic analysis in psychology**. *Qual Res Psychol* (2006.0) **3** 77-101. DOI: 10.1191/1478088706qp063oa 16. Rayman G, Vas PR, Baker N, Taylor CG, Gooday C, Alder AI. **The Ipswich Touch Test: a simple and novel method to identify inpatients with diabetes at risk of foot ulceration**. *Diabetes Care* (2011.0) **34** 1517-1518. DOI: 10.2337/dc11-0156 17. Sims JDS, Cavanagh PR, Ulbrecht JS. **Risk factors in the diabetic foot**. *Recognit Manage Phys Ther* (1988.0) **68** 1887 18. 18.Schaper NC. Diabetic foot ulcer classification system for research purposes: a progress report on criteria for including patients in research studies. Diabetes/Metabol Res Rev. 2004;20(4;S1;):S90–5. 19. Anichini R, Zecchini F, Cerretini I, Meucci G, Fusilli D, Alviggi L. **Improvement of diabetic foot care after the Implementation of the International Consensus on the Diabetic Foot (ICDF): Results of a 5-year prospective study**. *Diabetes Res Clin Pract* (2006.0) **75** 153-158. DOI: 10.1016/j.diabres.2006.05.014 20. Brownrigg JRW, Apelqvist J, Bakker K, Schaper NC, Hinchliffe RJ. **Evidence-based management of PAD & the Diabetic Foot**. *Eur J Vasc Endovasc Surg* (2013.0) **45** 673-681. DOI: 10.1016/j.ejvs.2013.02.014 21. Prompers L, Huijberts M, Apelqvist J, Jude E, Piaggesi A, Bakker K. **High prevalence of ischaemia, infection and serious comorbidity in patients with diabetic foot disease in Europe. Baseline results from the Eurodiale study**. *Diabetologia.* (2007.0) **50** 18-25. DOI: 10.1007/s00125-006-0491-1 22. Morbach S, Furchert H, Rümenapf G, Abbas ZG, Bharara M, Armstrong DG. **Long-term prognosis of diabetic foot patients and their limbs: amputation and death over the course of a decade**. *Diabetes care.* (2012.0) **35** 2021-7. DOI: 10.2337/dc12-0200 23. Moulik PK, Mtonga R, Gill GV. **Amputation and mortality in new-onset diabetic foot ulcers stratified by etiology**. *Diabetes Care* (2003.0) **26** 491-494. DOI: 10.2337/diacare.26.2.491 24. Jude EB, Oyibo SO, Chalmers N, Boulton AJM. **Peripheral arterial disease in diabetic and nondiabetic patients**. *Diabetes Care* (2001.0) **24** 1433. DOI: 10.2337/diacare.24.8.1433 25. Spreen MI, Gremmels H, Teraa M, Sprengers RW, Verhaar MC, Statius van Eps RG. **Diabetes is associated with decreased limb survival in patients with critical limb ischemia: pooled data from two randomized controlled trials**. *Diabetes care.* (2016.0) **39** 2058. DOI: 10.2337/dc16-0850 26. 26.NICE. Diabetic foot problems: prevention and management of foot problems in people with diabetes. Excellence NIfC, editor. 2015. p. 1–35. 27. Normahani P, Mustafa C, Standfield NJ. **Management of peripheral arterial disease in diabetes: a national survey of podiatry practice in the United Kingdom**. *J Foot Ankle Res* (2018.0) **11** 29. DOI: 10.1186/s13047-018-0270-5 28. Hinchliffe RJ, Forsythe RO, Apelqvist J, Boyko EJ, Fitridge R, Hong JP. **Guidelines on diagnosis, prognosis, and management of peripheral artery disease in patients with foot ulcers and diabetes (IWGDF 2019 update)**. *Diabetes/metabol Res Rev.* (2020.0) **36** e3276-n/a. DOI: 10.1002/dmrr.3276 29. Collins TC, Suarez-Almazor M, Peterson NJ. **An absent pulse is not sensitive for the early detection of peripheral arterial disease**. *Fam Med* (2006.0) **38** 38-42. PMID: 16378257 30. Khan NA, Rahim SA, Anand SS, Simel DL, Panju A. **Does the clinical examination predict lower extremity peripheral arterial disease?**. *JAMA, J Am Med Assoc* (2006.0) **295** 536. DOI: 10.1001/jama.295.5.536 31. Normahani P, Poushpas S, Alaa M, Bravis V, Sounderajah V, Aslam M, Jaffer U. **Diagnostic accuracy of point-of-care tests used to detect arterial disease in diabetes: TEsting for Arterial Disease in Diabetes (TrEAD) Study**. *Ann Surg* (2022.0) **276** e605-e612. DOI: 10.1097/SLA.0000000000004545 32. 32.Vriens B, D’Abate F, Ozdemir BA, Fenner C, Maynard W, Budge J, Carradice D, Hinchliffe RJ. Clinical examination and non-invasive screening tests in the diagnosis of peripheral artery disease in people with diabetes-related foot ulceration. Diabet Med. 2018;35(7):895–902. 10.1111/dme. 33. Tehan PE, Bray A, Chuter VH. **Non-invasive vascular assessment in the foot with diabetes: sensitivity and specificity of the ankle brachial index, toe brachial index and continuous wave Doppler for detecting peripheral arterial disease**. *J Diabetes Complications.* (2016.0) **30** 155-60. DOI: 10.1016/j.jdiacomp.2015.07.019 34. Darban Hosseini Amirkhiz G, Babaei MR, Madani NH, Khamseh ME. **Toe-brachial index is beyond a peripheral issue in patients with type 2 diabetes**. *PLoS ONE* (2021.0) **16** e0253138. DOI: 10.1371/journal.pone.0253138 35. Singh AK, Bahirani P. **Is toe brachial index a better tool than ankle brachial index for predicting outcomes in diabetic foot patients?**. *Int J Anatomy Radiol Surg.* (2021.0) **10** SO39-SO44 36. Tehan PE, Fox M, Stewart S. **Lower limb vascular assessment techniques of podiatrists in the United Kingdom: a national survey**. *J Foot Ankle Res* (2019.0) **12** 31. DOI: 10.1186/s13047-019-0341-2 37. Tehan PE, Chuter VH. **Vascular assessment techniques of podiatrists in Australia and New Zealand: a web-based survey**. *J Foot Ankle Res* (2015.0) **8** 71. DOI: 10.1186/s13047-015-0130-5 38. Pooler A. *An introduction to evidence-based practice in nursing & healthcare* (2014.0) 39. Hewitt-Taylor J. *Dawsonera* (2006.0) 40. Parker CN, Van Netten JJ, Parker TJ, Jia L, Corcoran H, Garrett M, Kwok CF, Nather A, Que MT, Srisawasdi G, Wraight P, Lazzarini PA. **Differences between national and international guidelines for the management of diabetic foot disease**. *Diabetes Metab Res Rev.* (2019.0) **35** e3101. DOI: 10.1002/dmrr.3101 41. Blanchette V, Patry J, Brousseau-Foley M. **Adequacy between canadian clinical guidelines and recommendations compared with international guidelines for the management of diabetic foot ulcers**. *Can J Diabetes* (2021.0) **45** 761-767.e12. DOI: 10.1016/j.jcjd.2021.03.004 42. 42.Chan CB, Dmytruk K, Labbie M, O’Connell P. Organizational changes in diabetic foot care practices for patients at low and moderate risk after implementing a comprehensive foot care program in Alberta, Canada. J Foot Ankle Res. 2020;13(1):26. 43. Leese GP, Stang D, Pearson DW. **Scottish Diabetes Foot Action G A national approach to diabetes foot risk stratification and foot care**. *Scottish Med J.* (2011.0) **56** 151-5. DOI: 10.1258/smj.2011.011113 44. Ortegon M, Redekop K, Niessen LW. **Cost-effectiveness of prevention and treatment of the diabetic foot: a Markov analysis**. *Diabetes Care* (2004.0) **27** 901-907. DOI: 10.2337/diacare.27.4.901 45. Formosa C, Gatt A, Chockalingam N. **A critical evaluation of existing diabetic foot screening guidelines**. *Rev Diabet Stud* (2016.0) **13** 158-186. DOI: 10.1900/RDS.2016.13.158 46. Boulton AJM, Armstrong DG, Albert SF, Frykberg RG, Hellman R, Kirkman MS. **Comprehensive foot examination and risk assessment: a report of the task force of the foot care interest group of the American Diabetes Association, with endorsement by the American Association of Clinical Endocrinologists**. *Diabetes Care* (2008.0) **31** 1679-1685. DOI: 10.2337/dc08-9021 47. Pérez-Panero AJ, Ruiz-Muñoz M, Cuesta-Vargas AI, Gónzalez-Sánchez M. **Prevention, assessment, diagnosis and management of diabetic foot based on clinical practice guidelines: a systematic review**. *Medicine (Baltimore).* (2019.0) **98** e16877-e. DOI: 10.1097/MD.0000000000016877 48. Parker CN, Van Netten JJ, Parker TJ, Jia L, Corcoran H, Garrett M. **Differences between national and international guidelines for the management of diabetic foot disease**. *Diabetes Metab Res Rev* (2019.0) **35** e3101. DOI: 10.1002/dmrr.3101 49. 49.Boyko EJ, Ahroni JH, Cohen V, Nelson KM, Heagerty PJ. Prediction of diabetic foot ulcer occurrence using commonly available clinical information : the seattle diabetic foot study. Diabetes Care. 2006;29(6):1202–7. 50. Leese GP, Reid F, Green V, McAlpine R, Cunningham S, Emslie-Smith AM. **Stratification of foot ulcer risk in patients with diabetes: a population-based study**. *Int J Clin Pract* (2006.0) **60** 541-545. DOI: 10.1111/j.1368-5031.2006.00899.x 51. 51.RIZIV. Diabetes: tegemoetkoming in de kosten van diëtetiek- en podologieverstrekkingen 2017. Available from: https://www.riziv.fgov.be/nl/themas/kost-terugbetaling/ziekten/endocriene-metabole/Paginas/diabetes-tegemoetkoming-kosten-verstrekkingen-dietetiek-podologie.aspx. 52. 52.Janisse DJ, Coleman W. Pedorthic care of the diabetic foot: correlation with risk category. Bowker JHMD, Pfeifer MAMDCDEF, editors. 7th ed. 2008. p. 529–46. 53. Meijer JWG, Links TP, Smit AJ, Groothoff JW, Eisma WH. **Evaluation of a screening and prevention programme for diabetic foot complications**. *Prosthet Orthot Int* (2001.0) **25** 132-138. DOI: 10.1080/03093640108726586 54. 54.Diabetische voet: Federatie Medische Specialisten; 2017 [Uitvoering voetcontrole - Richtlijn - Richtlijnendatabase].
--- title: 'Correlation between nonalcoholic fatty liver disease and left ventricular diastolic dysfunction in non-obese adults: a cross-sectional study' authors: - Fangyuan Cong - Luying Zhu - Lihua Deng - Qian Xue - Jingtong Wang journal: BMC Gastroenterology year: 2023 pmcid: PMC10041784 doi: 10.1186/s12876-023-02708-4 license: CC BY 4.0 --- # Correlation between nonalcoholic fatty liver disease and left ventricular diastolic dysfunction in non-obese adults: a cross-sectional study ## Abstract ### Background and aims Non-alcoholic fatty liver disease (NAFLD) is associated with a greater risk of developing cardiovascular disease and have adverse impacts on the cardiac structure and function. Little is known about the effect of non-obese NAFLD upon cardiac function. We aimed to compare the echocardiographic parameters of left ventricle (LV) between non-obese NAFLD group and control group, and explore the correlation of non-obese NAFLD with LV diastolic dysfunction. ### Methods and results In this cross-sectional study, 316 non-obese inpatients were enrolled, including 72 participants with NAFLD (non-obese NAFLD group) and 244 participants without NAFLD (control group). LV structural and functional indices of two groups were comparatively analyzed. LV diastolic disfunction was diagnosed and graded using the ratio of the peak velocity of the early filling (E) wave to the atrial contraction (A) wave and E value. Compared with control group, the non-obese NAFLD group had the lower E/A〔(0.80 ± 0.22) vs (0.88 ± 0.35), $t = 2.528$, $$p \leq 0.012$$〕and the smaller LV end-diastolic diameter〔(4.51 ± 0.42)cm vs (4.64 ± 0.43)cm, $t = 2.182$, $$p \leq 0.030$$〕. And the non-obese NAFLD group had a higher prevalence of E/A < 1 than control group ($83.3\%$ vs $68.9\%$, X2 = 5.802, $$p \leq 0.016$$) while two groups had similar proportions of LV diastolic dysfunction ($58.3\%$ vs $53.7\%$, X2 = 0.484, $$p \leq 0.487$$). Multivariate logistic regression analysis showed that non-obese NAFLD was associated with an increase in E/A < 1 (OR = 6.562, $95\%$CI 2.014, 21.373, $$p \leq 0.002$$). ### Conclusions Non-obese NAFLD was associated with decrease of E/A, while more research will be necessary to evaluate risk of non-obese NAFLD for LV diastolic dysfunction in future. ## Introduction In recent years, nonalcoholic fatty liver disease (NAFLD) has become the most common chronic liver disease globally. The disease is associated with many metabolic risk factors such as obesity, diabetes mellitus (DM), insulin resistance, dyslipidemia [1]. In the general population, the prevalence of NAFLD is estimated to be $25\%$ in the world [2] and it is about $30\%$ in China [3]. Previous studies have shown that NAFLD is closely related to obesity which will increase the prevalence of NAFLD [4], while it may also affect normal-weight individuals, a condition termed as non-obese NAFLD [5], which includes individuals with BMI < 30 kg/m2 in the Caucasian population and BMI < 25 kg/m2 in the Asian population [6, 7]. The global incidence of NAFLD in the non-obese population was about 25 per 1000 person-years and around $40\%$ of the global NAFLD population were classified as non-obese [5, 8]. In Asia, about $30\%$ of NAFLD population were non-obese [9]. The prevalence of NAFLD is about 7–$19\%$ [10, 11] and 8–$20\%$ among people with BMI < 25 kg/m2 in Asia and China respectively [12, 13], which is increasing with years [14]. Non-obese NAFLD is similar to obese NAFLD in pathophysiological mechanisms, such as hepatic lipid accumulation [15], insulin resistance [13], metabolic dysfunction of visceral fat [16], genetic susceptibility [17]. Many studies have shown that NAFLD has an adverse impact on the cardiac structure and function [18–20], which may associate with myocardial glucose uptake, myocardial fat infiltration, inflammation, oxidative stress [21, 22]. Studies have shown that an incrementally increased risk for left ventricular (LV) diastolic dysfunction according to fibrosis grade was prominent in the non-obese population [23]. However, less studies are about the effect of non-obese NAFLD on LV structure and function, so the purpose of this study is to assess the correlation of non-obese NAFLD with LV diastolic dysfunction by comparing echocardiographic parameters of LV between non-obese NAFLD group and control group. ## Subjects The subjects of this cross-sectional study were inpatients from the Department of Geriatrics, Peking University People's Hospital from January 2018 to December 2020. All inpatients were hospitalized for physical examination. The inclusion criteria were: [1] age ≥ 40 years old; [2] BMI < 25 kg/m2; [3] the imaging examination of liver (abdominal ultrasound or CT) and echocardiography were performed during hospitalization; [4] complete demographic, laboratory and imaging information. The exclusion criteria refer to the Guidelines of Prevention and Treatment for Nonalcoholic Fatty Liver Disease (2018 Updated Edition) [24] as follows: [1] excessive alcohol intake (> 30 g/d in men and > 20 g/d in women); [2] detected positive serum markers of hepatitis B and C; [3] secondary causes of fatty liver including viral hepatitis, drug-induced liver disease, autoimmune liver disease, hepatolenticular degeneration, total parenteral nutrition, inflammatory bowel disease, celiac disease, Cushing's syndrome, β-lipoprotein deficiency, lipid atrophy diabetes mellitus, Mauriac syndrome; [4] end-stage liver diseases including hepatic fibrosis, liver cirrhosis, liver cancer and liver failure; [5] basic heart diseases including coronary heart disease, congenital heart disease, valvular heart disease, pulmonary heart disease, hypertrophic cardiomyopathy, cardiac surgery, aortic dissection, heart failure; [6] chronic renal failure or malignancies; [7] pregnancy. Patients who met the inclusion criteria were divided into non-obese NAFLD group and control group according to the results of imaging examination of liver. The imaging diagnostic criteria of NAFLD refer to the Guidelines for Management of Nonalcoholic Fatty Liver Disease (2010 Revision) [25]. ## Clinical and laboratory evaluations Complete blood count, blood biochemistry, blood glucose metabolism and other indicators of patients were collected retrospectively. Alanine aminotransferase (ALT), aspartate aminotransferase (AST), γ-glutamyl transpeptidase (γ-GT), serum albumin (Alb), serum creatinine (Cr), serum uric acid (UA), fasting blood glucose (FBG), total cholesterol (TC), triglyceride (TG), high-density lipoprotein cholesterol (HDL-C) and low-density lipoprotein cholesterol (LDL-C) were detected by automatic biochemical analyzer AU5832. Hemoglobin (Hb) and platelet count (PLT) were measured by blood cell analyzer DxH800, HbA1c was measured by glycosylated hemoglobin analyzer Primus9210, and estimated glomerular filtration rate (eGFR) was obtained by CKD-EPI method [26]. Height, body weight, heart rates (HR), systolic blood pressure (SBP) and diastolic blood pressure (DBP) were measured using a digital scale. The body mass index (BMI) was calculated as weight (kg)/height (m2). The body surface area (BSA, male) was calculated as 0.0057 × height (cm) + 0.0121 × weight (kg) + 0.0082, and BSA (female) was calculated as 0.0073 × height (cm) + 0.0127 × weight (kg)-0.2106 [27]. Obesity is defined as BMI ≥ 25 kg/m2 and non-obesity is defined as BMI < 25 kg/m2 [7]. We collected current comorbidities, including hypertension, DM and obstructive sleep apnea–hypopnea syndrome (OSAHS), and medication history, including antihypertensive, lipid-lowering and hypoglycemic drugs. Smoking history and family history of heart diseases were also collected. Smokers were defined as individuals who had a continuous or cumulative smoking time ≥ 1 year. ## Echocardiographic evaluations The results of echocardiography were collected retrospectively. Echocardiography was performed by Acusonsc-2000 Full Digital Color Doppler Ultrasonic Instrument, which was completed and reviewed by two sonographers (at least one of them is associate chief physician or chief physician). In this study, the results of echocardiography were used to evaluate the cardiac structure and function, and the echocardiographic parameters included interventricular septal thickness at diastole (IVSd), left ventricular posterior wall thickness at diastole (LVPWd), left ventricular mass (LVM), left ventricular mass index (LVMI), ejection fraction (EF), left ventricular end-systolic diameter (LVESD), left ventricular end-diastolic diameter (LVEDD), the peak velocity of the early filling (E) wave, the peak velocity of the atrial contraction (A) wave and E/A. E/A ratio is usually > 1 in healthy adults while it often decreases with aging and is affected by HR. In this study, LV diastolic function was assessed using E/A ratio and E value. Normal LV diastolic function was defined as 0.8 < E/A < 2. LV diastolic dysfunction was then graded as 1 (mild), 2 (moderate), and 3 (severe). Grade 1 diastolic dysfunction, which is also called delayed relaxation filling pattern, was defined as 1) E/A ≤ 0.8 and 2) E ≤ 50 cm/s. Grade 2 diastolic dysfunction, which is called pseudonormal filling pattern, was defined as 1) E/A ≤ 0.8 and 2) E > 50 cm/s. Grade 3, which is called restrictive filling pattern, was defined as E/A ≥ 2 [28]. ## Statistical analysis Statistical analyses were conducted using SPSS 26.0 software package of IBM. The measurement data conforming to the normal distribution was expressed as the mean ± standard deviation (‾x ± s), and the independent sample t-test or variance analysis was used to compare the continuous variables between two groups. The categorical variables were analyzed by x2 test. Univariate logistic regression was used to find out confounding factors, and two-class logistic regression (backward: LR) was used to analyze the relationship between non-obese NAFLD and LV diastolic dysfunction. p value of < 0.05 was considered statistically significant. ## General data A total of 316 subjects met the inclusion criteria for the study and were finally included in this analysis, which including 118 males and 198 females, with an average age of (69 ± 12) years, (72 ± 13) years for males and (67 ± 12) years for females respectively. According to the imaging results, 72 subjects ($22.8\%$) were diagnosed as non-obese NAFLD and 244 subjects ($77.2\%$) belonged to control group. *The* general data of the subjects with and without non-obese NAFLD is provided in Table 1. BMI and BSA were significantly higher in subjects with non-obese NAFLD compared with control group ($p \leq 0.01$), but there was no significant statistical difference in sex composition, age, HR, SBP and DBP between the two groups. Table 1Comparison of general data between non-obese NAFLD group and control groupNon-obese NAFLDgroup($$n = 72$$)Control group($$n = 244$$)t/x2P-valueMales〔n(%)〕22(30.6)96(39.3)1.8350.176Age(years)67.57 ± 11.5969.62 ± 12.881.2870.200BMI(kg/m2)22.82 ± 1.7421.85 ± 2.26-3.896 < 0.001BSA(m2)1.74 ± 0.131.69 ± 0.13-2.6300.009HR73.36 ± 8.2473.98 ± 7.440.6000.549SBP128.13 ± 13.84128.94 ± 16.370.3590.720DBP75.94 ± 10.2373.65 ± 9.34-1.6550.099 ## Biochemical and glucose metabolism Table 2 describes the biochemical and glucose metabolic characteristics of the study cohort according to the presence of non-obese NAFLD. Subjects with non-obese NAFLD had higher ALT, γ-GT, TG, Alb, UA, Hb, FBG and HbA1c, and lower HDL-C than subjects in the control group ($p \leq 0.05$). There was no significant difference in other indices. Table 2Comparison of biochemistry and glucose metabolism between non-obese NAFLD group and control groupNon-obese NAFLDgroup($$n = 72$$)Control group($$n = 244$$)t/x2P-valueALT(U/L)20.69 ± 9.6316.29 ± 9.90-3.3310.001AST(U/L)20.19 ± 4.8620.06 ± 6.96-0.1550.877γ-GT(U/L)28.31 ± 22.7723.01 ± 18.14-2.0420.042TC(mmol/L)4.79 ± 1.174.51 ± 1.00-1.8640.065TG(mmol/L)1.88 ± 1.231.19 ± 0.58-4.668 < 0.001HDL-C(mmol/L)1.17 ± 0.291.30 ± 0.342.9900.003LDL-C(mmol/L)2.95 ± 0.852.73 ± 0.98-1.6800.094Alb(g/L)40.74 ± 3.8338.82 ± 4.28-3.4120.001UA(umol/L)354.85 ± 81.05304.18 ± 77.59-3.897 < 0.001Cr(umol/L)66.31 ± 17.3971.58 ± 22.511.8340.068eGFR(ml/min*1.73m2)86.76 ± 15.7183.50 ± 17.21-1.4390.151Hb(g/L)135.54 ± 12.51127.23 ± 15.31-4.211 < 0.001PLT(× 109/L)214.13 ± 44.86203.52 ± 67.82-1.5510.123FBG(mmol/L)6.02 ± 1.855.15 ± 1.13-3.757 < 0.001HbA1c(%)6.73 ± 1.386.01 ± 0.93-4.043 < 0.001 ## Complications and medications The subjects' complications and medications are compared in Table 3. Subjects with non-obese NAFLD had higher prevalence rates of DM and longer course of DM than subjects in the control group ($p \leq 0.05$). The proportion of patients taking hypoglycemic drugs in the non-obese NAFLD group was significantly higher than that in the control group ($p \leq 0.01$).Table 3Comparison of complications and medications between non-obese NAFLD group and control groupNon-obese NAFLDgroup($$n = 72$$)Control group($$n = 244$$)t/x2P-valueHypertension〔n(%)〕35(50.0)102(42.1)1.3590.244Course of hypertension(years)6.85 ± 10.826.48 ± 11.17-0.2480.804DM〔n(%)〕30(42.9)52(21.6)12.653 < 0.001Course of DM(years)4.96 ± 7.662.52 ± 5.89-2.4610.016OSAHS〔n(%)〕0[0]2(0.9)0.0001.000Smoking history〔n(%)〕8(11.1)32(13.1)0.2020.653Family history of heart disease〔n(%)〕7(9.7)8(3.3)3.7790.052History of taking lipid-lowering drugs〔n(%)〕23(31.9)67(27.5)0.5490.459History of taking antihypertensive drugs〔n(%)〕32(44.4)95(38.9)0.7020.402History of taking hypoglycemic drugs〔n(%)〕30(41.7)51(20.9)12.575 < 0.001 ## LV structure and function LV structure and function of two groups were assessed by subjects' echocardiographic parameters, which are compared in Table 4. Subjects with non-obese NAFLD had more unfavorable echocardiographic parameters, including a lower E/A and a lower LVEDD, than the control group ($p \leq 0.05$). There was no significant difference in other indices. Table 4Comparison of echocardiographic characteristics between non-obese NAFLD group and control groupNon-obese NAFLDgroup($$n = 72$$)Control group($$n = 244$$)tP-valueIVSd(cm)0.91 ± 0.150.88 ± 0.13-1.6280.107LVPWd(cm)0.84 ± 0.120.87 ± 0.240.2960.397LVM(g)130.14 ± 28.95132.94 ± 34.910.6840.495LVMI(g/m2)74.82 ± 15.2678.44 ± 19.681.6440.102EF(%)68.49 ± 5.2169.64 ± 5.551.5630.119LVESD(cm)2.78 ± 0.312.81 ± 0.330.7160.474LVEDD(cm)4.51 ± 0.424.64 ± 0.432.1820.030E(cm/s)68.34 ± 15.9472.26 ± 18.051.6640.097A(cm/s)88.16 ± 16.8786.05 ± 20.66-0.7890.431E/A0.80 ± 0.220.88 ± 0.352.5280.012 In the non-obese NAFLD group, 42 ($58.3\%$) subjects had LV diastolic disfunction, including 8 ($11.1\%$) subjects presented with grade 1 and 34 ($47.2\%$) subjects presented with grade 2. In the control group, 131 ($53.7\%$) subjects had LV diastolic disfunction, including 29 ($11.9\%$) subjects presented with grade 1, 98 ($40.2\%$) subjects presented with grade 2 and 4 ($1.6\%$) subjects presented with grade 3. The prevalence rates of LV diastolic dysfunction in two groups were similar ($58.3\%$ vs $53.7\%$, X2 = 0.484, $$p \leq 0.487$$). E/A ratio is usually > 1 in healthy adults despite the lack of diagnostic significance, then we also analyzed the proportions of E/A < 1. The proportion of patients with E/A < 1 in the non-obese NAFLD group was significantly higher than the control group ($p \leq 0.05$). The specific data are shown in Table 5.Table 5Comparison of LV diastolic dysfunction between non-obese NAFLD group and control groupNon-obese NAFLDgroup($$n = 72$$)Control group($$n = 244$$)X2P-valueLV diastolic disfunction 〔n(%)〕42(58.3)131(53.7)0.4840.487Grade 1〔n(%)〕8(11.1)29(11.9)0.0320.858Grade 2〔n(%)〕34(47.2)98(40.2)1.1390.286Grade 3〔n(%)〕0[0]4(1.6)0.2440.622E/A < 1〔n(%)〕60(83.3)168(68.9)5.8020.016 ## Non-obese NAFLD and LV diastolic dysfunction Although there is no statistical difference in LV diastolic dysfunction proportions of two groups, the non-obese NAFLD group had a lower E/A and a higher proportion of E/A < 1, maybe suggesting worse LV diastolic function, so we analyzed the association between non-obese NAFLD and LV diastolic dysfunction using logistic regression analysis. According to whether 0.8 < E/A < 2, 173 subjects ($54.7\%$) were divided into LV diastolic dysfunction group (abnormal group) and 143 subjects ($45.3\%$) belonged to normal group. There were 42 ($24.3\%$) and 30 ($21.0\%$) non-obese NAFLD patients in the abnormal group and the normal group respectively. With the LV diastolic dysfunction as dependent variable, univariate logistic regression analysis showed that age, BMI, hypertension, course of hypertension, course of DM, history of taking antihypertensive drugs, HR, SBP, HbA1c, Alb and eGFR were associated with LV diastolic dysfunction ($p \leq 0.05$). In the univariate model, subjects with non-obese NAFLD had a 1.2-fold increased risk for LV diastolic dysfunction with no statistical significance (OR = 1.208, $95\%$CI 0.710, 2.055, $$p \leq 0.487$$, Table 6).Table 6Univariate analysis of the risk of left ventricular diastolic dysfunctionβSEWaldOR$95\%$CIP-valueLowerUpperMale0.3520.2362.2281.4220.8962.2570.136Age0.0970.01262.941.1011.0751.128 < 0.001BMI0.1450.0547.3311.1561.0411.2840.007BSA-1.0760.9021.4220.3410.0581.9980.233Hypertension0.9660.23816.4982.6281.6494.190 < 0.001Course of hypertension0.0470.01313.1891.0481.0221.075 < 0.001DM0.4020.2632.3261.4940.8922.5040.127Course of DM0.0410.0194.5201.0421.0031.0820.034OSAHS-0.2231.4190.0250.8000.05012.9190.875NAFLD0.1890.2710.4831.2080.7102.0550.487History of taking lipid-lowering drugs0.2990.2531.3971.3490.8212.2150.237History of taking antihypertensive drugs0.9000.23914.1272.4591.5383.932 < 0.001History of taking hypoglycemic drugs0.4540.2652.9451.5750.9382.6450.086Smoking history-0.2180.3390.4150.8040.4141.5610.519Family history of heart disease-0.0600.5300.0130.9420.3332.6640.942HR0.0470.0168.9931.0481.0161.0810.003SBP0.0200.0085.5551.0201.0031.0370.018DBP0.0050.0130.1701.0050.9801.0310.680FBG0.1140.0871.7181.1210.9451.3290.190HbA1c0.3600.1317.5901.4331.1091.8510.006BNP0.0030.0022.0491.0030.9991.0070.152ALT-0.0060.0110.2820.9940.9721.0160.596AST0.0210.0181.2811.0210.9851.0580.258γ-GT0.0060.0060.8511.0060.9941.0180.356TC-0.2090.1113.5420.8120.6531.0090.060TG-0.0480.1370.1220.9530.7301.2460.727HDL-C-0.3370.3390.9880.7140.3671.3880.320LDL-C-0.0880.1200.5390.9150.7231.1590.463Alb-0.0820.0316.9140.9210.8670.9790.009UA0.0010.0020.6751.0010.9981.0050.411Cr0.0100.0062.7741.0100.9981.0210.096eGFR-0.0360.00821.3320.9650.9500.979 < 0.001Hb-0.0050.0080.4410.9950.9801.0100.507PLT0.0020.0021.2301.0020.9981.0060.267 Further stepwise multivariate logistic regression analysis, including the above-mentioned significant variables, NAFLD and DM as a known risk factor, showed that non-obese NAFLD was associated with an increase in LV diastolic dysfunction with no statistical significance (OR = 1.206, $95\%$CI 0.566, 2.569, $$p \leq 0.628$$, Table 7). LV diastolic dysfunction also related to age, BMI and HR ($p \leq 0.001$).Table 7Multivariate logistic regression analysis of the risk of left ventricular diastolic dysfunctionBSEWaldOR$95\%$CI P-valueLowerUpperAge0.1180.01749.5591.1251.0891.163 < 0.001BMI0.3130.08613.2811.3681.1561.619 < 0.001HR0.0770.02113.0151.0801.0361.126 < 0.001NAFLD0.1870.3860.2351.2060.5662.5690.628constant-20.5433.23840.244 < 0.001-- < 0.001 ## Non-obese NAFLD and decrease of E/A As the non-obese NAFLD group had a higher proportion of E/A < 1 than the control group, then according to whether E/A < 1, 228 subjects ($72.2\%$) were divided into decreased E/A group and 88 subjects ($27.8\%$) belonged to normal E/A group. There were 60 ($26.3\%$) and 12 ($13.6\%$) non-obese NAFLD patients in the decreased E/A group and the normal E/A group respectively. With the E/A < 1 as dependent variable, univariate logistic regression analysis showed that NAFLD, gender, age, BMI, hypertension, course of hypertension, course of DM, history of taking lipid-lowering and antihypertensive drugs, SBP, HbA1c, UA, Cr and eGFR were associated with E/A < 1 ($p \leq 0.05$). In the univariate model, subjects with non-obese NAFLD had a 2.3-fold increased risk for E/A < 1 (OR = 2.262, $95\%$CI 1.150, 4.449, $$p \leq 0.018$$, Table 8).Table 8Univariate analysis of the risk of E/A < 1βSEWaldOR$95\%$CIP-valueLowerUpperMale0.6260.2755.2031.8711.0923.2040.023Age0.0910.01348.6681.0951.0671.123 < 0.001BMI0.1550.0567.6971.1681.0471.3030.006BSA-0.1340.9940.0180.8740.1256.1340.892Hypertension1.0250.27813.6442.7881.6184.804 < 0.001Course of hypertension0.0480.0168.3411.0491.0151.0830.004DM0.5760.3143.3721.7790.9623.2900.066Course of DM0.0630.0265.7181.0651.0111.1220.017OSAHS-1.0191.4210.5140.3610.0225.8450.361NAFLD0.8160.3455.5932.2621.1504.4490.018History of taking lipid-lowering drugs0.9720.3249.0232.6431.4034.9830.003History of taking antihypertensive drugs1.1010.28714.7213.0071.7135.276 < 0.001History of taking hypoglycemic drugs0.5850.3133.4921.7940.9723.3130.062Smoking history-0.2550.3640.4910.7750.3801.5810.483Family history of heart disease0.0630.5980.0111.0650.3303.4360.917HR0.0260.0172.3601.0260.9931.0600.124SBP0.0300.0108.9531.0301.0101.0500.003DBP0.0140.0150.8571.0140.9851.0430.355FBG0.1020.1011.0351.1080.9101.3490.309HbA1c0.3620.1615.0781.4371.0481.9690.024BNP0.0010.0020.3561.0010.9971.0060.551ALT-0.0030.0120.0620.9960.9731.0210.457AST0.0060.0200.0930.9990.9681.0460.940γ-GT-0.0070.0061.1290.9980.9821.0060.309TC-0.1860.1222.2950.8310.6531.0560.130TG0.1720.1790.9161.1870.8351.6870.338HDL-C-0.8460.3755.0991.0180.2060.8940.672LDL-C-0.0520.1290.1590.9500.7371.2240.690Alb-0.0270.0310.7680.9730.9161.0340.381UA0.0050.0025.3261.0051.0011.0090.021Cr0.0140.0073.9021.0141.0001.0290.048eGFR-0.0440.01020.8380.9570.9390.975 < 0.001Hb0.0020.0080.0381.0020.9861.0180.845PLT0.0000.0020.0611.0000.9961.0030.805 Further stepwise multivariate logistic regression analysis, including the above-mentioned significant variables and DM as a known risk factor, showed that non-obese NAFLD was associated with an increase risk in E/A < 1 (OR = 6.562, $95\%$CI 2.014, 21.373, $$p \leq 0.002$$, Table 9).Table 9Multivariate logistic regression analysis of the risk of E/A < 1BSEWaldOR$95\%$CI P-valueLowerUpperNAFLD1.8810.6039.7496.5622.01421.3730.002Age0.1040.02123.8101.1091.0641.157 < 0.001BMI0.2830.1086.8121.3261.0731.6400.009constant-13.0333.18616.735 < 0.001-- < 0.001 ## Discussion In this study, non-obese NAFLD was associated with decrease of E/A, independent of well-identified cardiovascular risk factors, while it was not an independent risk factor of LV diastolic dysfunction. Previously, several studies have suggested that NAFLD was an independent risk factor affecting cardiac structure and function, but there are few studies on the correlation between non-obese NAFLD and LV function or structure in Chinese adults. Therefore, this paper studied the changes of LV structure and function in hospitalized non-obese NAFLD patients, and discussed the correlation between non-obese NAFLD and LV diastolic dysfunction, aiming to provide scientific evidence for clinicians to pay attention to the cardiac structure and function of non-obese NAFLD patients. Our study showed that compared with the control group, non-obese NAFLD patients had higher BMI, BSA, levels of liver enzymes, blood lipids, proportion of DM, and worse glucose metabolism, which were consistent with previous reports. Although BMI of a non-obese NAFLD patient is within the normal range, it is still higher than that of a healthy adult and the visceral fat index is also high [13, 29]. Obesity is related to higher all-cause mortality [30], so both obese and non-obese NAFLD patients can benefit from losing weight [31]. For non-obese NAFLD patients, a 5–$10\%$ reduction in body weight through lifestyle intervention may be a reasonable target and can also benefit them a lot [32]. There are also many drugs as good candidates to cure NAFLD/NASH, and consequently to reduce the burden of cardiovascular diseases (CVD) [33]. In this study, LV structure and function were mainly evaluated by echocardiographic parameters, among which E/A ratio and E value were important indices to evaluate LV diastolic function. In the early stage of LV diastolic dysfunction, the E value declines due to the decrease of the maximum mitral blood flow velocity in the early LV diastole, which leads to E/A < 1 [34]. With the further decline of LV diastolic function, the reduction in diastolic LV filling results in an atrial residuum, which increases left atrial (LA) pressure, causing an increase of E value. Finally, the increased load imposed on the LA as a result of a poorly compliant LV may lead to decreased atrial contractile reserve, LA enlargement, and a decrease of A value. The combined effects of a rise in E value and a decrease in A value result in an increase of E/A [28]. We compared the LV structure and function indices between non-obese NAFLD group and control group, and the results showed that non-obese NAFLD patients had lower E/A while there was no significant statistical difference in proportions of LV diastolic dysfunction. The reason why we considered the result may be that decreased E/A or declined function in non-obese NAFLD patients was at an extremely early stage so that had not yet reached the diagnostic standard of LV diastolic dysfunction. Then in addition to diastolic dysfunction, we also analyzed E/A < 1 as a dependent variable. The results of this study showed that in non-obese people, subjects with NAFLD had a 6.6-fold increased risk for decrease of E/A while non-obese NAFLD was not an independent risk factor of LV diastolic dysfunction. Maybe the above results were still controversial because there are few similar studies about non-obese NAFLD at present. A previous research showed that non-obese NAFLD patients have worse echocardiographic measurements, including IVSd, LVEDD, LVPWd, EF, LVM and E velocity [35]. Most studies about NAFLD, whether the patient was obese or not, showed that NAFLD patients had worse cardiac structure and function [21, 36–41], even related to the degree of liver fibrosis [42]. And latest research revealed that LVMI increased progressively and LV diastolic function worsened with increasing number of steatosis scores in NAFLD patients. It also showed that liver steatosis, as identified by use of biochemical scores, predicted LV hypertrophy and diastolic dysfunction independently of blood pressure and obesity, and this association was independent of the HOMA-index for LV diastolic dysfunction [43]. Then there was a research showing that metabolic dysfunction-associated fatty liver disease patients had increased interventricular septum thickness, left ventricular posterior wall thickness, LVM and LVMI, and more patients with MAFLD had LV diastolic dysfunction compared to the normal group ($60.8\%$ vs $24.6\%$, $p \leq 0.001$). On the contrary, some studies believed that correlation between NAFLD and cardiac function remained to be confirmed because of confounding factors such as obesity [44, 45]. A previous Chinese cohort study revealed that no significant association was observed between non-obese NAFLD and incident coronary artery disease after adjusting other traditional cardiovascular risk factors [46]. Surprisingly, non-obese NAFLD patients have smaller LVEDD, which is usually found in late stage of LV diastolic dysfunction. We consider this result probably because non-obese NAFLD patients had more complications or some medication history leading to abnormal ventricular structure. Moreover, despite the exclusion of all patients with known heart disease, there may be patients with asymptomatic coronary heart disease even if all echocardiographic results showed no abnormal wall motion. Both multivariate logistic regression analyses showed that age and BMI were independent risk factors of LV diastolic dysfunction or E/A decrease, which was consistent with previous consensus that age and BMI are independent risk factors of CVD. Non-obese NAFLD is similar to obese NAFLD in pathophysiology and non-obese NAFLD individuals seem to have an intermediate metabolic phenotype between healthy individuals and obese NAFLD patients. A previous study based on liver biopsy showed that compared with obese NAFLD patients, non-obese NAFLD patients had lighter degree of hepatocytic steatosis, lobular inflammation and advanced liver fibrosis, and lower prevalence of NASH ($54.1\%$ vs $71.2\%$, $p \leq 0.001$) [47], and it also believed that liver fibrosis in non-obese NAFLD patients was obviously related to metabolic disorders. Another meta-analysis including 493 non-obese NAFLD patients and 2209 obese NAFLD patients compared the liver histological features between the two groups, which also showed that pathological changes of non-obese NAFLD were mild [48]. Insulin resistance is universal in NAFLD patients whether they are obese or not [12, 49, 50]. Non-obese NAFLD patients generally had higher prevalence of DM and glucose intolerance than healthy subjects, while there was no statistical difference between non-obese and obese NAFLD patients [51]. The changes of intestinal microbiota are also associated with the progress of NAFLD and liver fibrosis [52, 53]. A previous report about gut microbiota composition showed that Eubacterium abundance was significantly decreased in non-obese NAFLD patients compared with that in obese NAFLD patients and healthy subjects, then the results demonstrated a negative correlation between Eubacterium and hepatic fibrosis and that the decrease in the abundance of Eubacterium producing butyric acid may play an important role in the development of non-obese NAFLD [54]. It was found that a variety of gene sites are related to the risk, disease severity, hepatic steatosis and advanced fibrosis of NAFLD, including PNPLA3, TM6SF2, GCKR, MBOAT7, APOC3, HSD17B13, etc. [ 55–60]. Among them, PNPLA3 is one of the earliest genes related to NAFLD in genome-wide association studies. the PNPLA3 rs738409 GG genotype was found in 13–$19\%$ of the general population in Asian, which is higher than that in other regions [12]. PNPLA3 not only plays a role in increasing the susceptibility to NAFLD, but also is related to abdominal visceral fat accumulation [61], and this gene has been proved to be one of the risk factors for NAFLD in non-obese people [62]. TM6SF2 has a protective effect on cardiovascular system, but it participates in hepatic steatosis and increases the susceptibility to NASH and hepatic fibrosis [63]. Compared with obese NAFLD patients, TM6SF2 E167K mutation is more common in non-obese NAFLD patients [64]. A new study found that loss of immunity-related GTPase GM4951 leads to nonalcoholic fatty liver disease without obesity [65]. At present, it is considered that NAFLD is not only related to systemic insulin resistance, but also related to endothelial dysfunction, oxidative stress, plaque formation, vascular tone change, systemic inflammatory response, metabolic disorders of blood lipid and so on [66–68]. Previous studies have shown that compared with healthy people, non-obese NAFLD patients also have a higher risk of coronary heart disease [69], and there is no statistical difference between non-obese NAFLD patients and obese NAFLD patients in the risk of CVD and malignant tumors, and they all have a higher risk of all-cause mortality [65]. The main causes of death of non-obese NAFLD patients are malignant tumors and CVD [70]. Despite the fact that NAFLD is usually associated with obesity, it has also been noted that the prevalence of NAFLD is increasing in non-obese individuals. Non-obese NAFLD is similar to the obese NAFLD in pathophysiological mechanism and influence on other related diseases. Compared with the healthy individuals, the non-obese NAFLD patients have a higher risk of liver cirrhosis, hypertension, DM, coronary heart disease and other diseases, as well as the risk of all-cause death, which needs to be confirmed by more studies in the future. The research is meaningful for clinicians and patients because it can remind clinicians to pay more attention to cardiac structure and function of non-obese NAFLD patients, and early intervention on non-obese NAFLD to delay its progress may be helpful to prevent myocardial dysfunction. However, this study has several limitations. First, the cross-sectional design of this study was difficult to explore the causal relationship between NAFLD and LV diastolic dysfunction. Second, imaging examination were used to diagnose NAFLD and we were unable to obtain liver histological samples, the gold standard for the diagnosis of NAFLD. Third, data on visceral adiposity, such as waist circumference, was lacking. Forth, despite the exclusion of all patients with known heart disease, the further research may need the coronary angiography as the diagnostic gold standard to exclude patients with asymptomatic coronary heart disease. Fifth,we used E/A and E value to evaluate LV diastolic dysfunction, while only this approach is difficult to identify patients with Grade 2 diastolic dysfunction (pseudonormal filling pattern) when E/A is in the normal range [28]. Therefore, more accurate methods will be needed in the future. Sixth, the cohort in this research was a selected population, so may not be representative of the general population. In addition, this study only included subjects of East Asian ethnicity, so the conclusions may not be generalizable to other ethnic groups. Further studies are needed to validate our results. In conclusion, non-obese NAFLD was associated with decrease of E/A, while more research will be necessary to evaluate risk of non-obese NAFLD for LV diastolic dysfunction in future. ## References 1. Chalasani N, Younossi Z, Lavine JE. **The diagnosis and management of non-alcoholic fatty liver disease: practice Guideline by the American Association for the Study of Liver Diseases, American College of Gastroenterology, and the American Gastroenterological Association**. *Hepatology* (2012.0) **55** 2005-2023. DOI: 10.1002/hep.25762 2. Younossi ZM. **Non-alcoholic fatty liver disease - A global public health perspective**. *J Hepatol* (2019.0) **70** 531-544. DOI: 10.1016/j.jhep.2018.10.033 3. Zhou F, Zhou J, Wang W. **Unexpected Rapid Increase in the Burden of NAFLD in China From 2008 to 2018: A Systematic Review and Meta-Analysis**. *Hepatology* (2019.0) **70** 1119-1133. DOI: 10.1002/hep.30702 4. Estes C, Anstee QM, Arias-Loste MT. **Modeling NAFLD disease burden in China, France, Germany, Italy, Japan, Spain, United Kingdom, and United States for the period 2016–2030**. *J Hepatol* (2018.0) **69** 896-904. DOI: 10.1016/j.jhep.2018.05.036 5. Ye Q, Zou B, Yeo YH. **Global prevalence, incidence, and outcomes of non-obese or lean non-alcoholic fatty liver disease: a systematic review and meta-analysis**. *Lancet Gastroenterol Hepatol* (2020.0) **5** 739-752. DOI: 10.1016/s2468-1253(20)30077-7 6. Weir CB, Jan A. **BMI Classification Percentile And Cut Off Points**. *StatPearls* (2022.0) 7. 7.Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies. Lancet. 2004,363:157–163. 10.1016/s0140-6736(03)15268-3 8. Eslam M, Fan JG, Mendez-Sanchez N. **Non-alcoholic fatty liver disease in non-obese individuals: the impact of metabolic health**. *Lancet Gastroenterol Hepatol* (2020.0) **5** 713-715. DOI: 10.1016/s2468-1253(20)30090-x 9. Kam L, Huang DQ, Teng MLP. **Clinical profiles of Asians with NAFLD: A systematic review and meta-analysis**. *Dig Dis* (2021.0). DOI: 10.1159/000521662 10. Niriella MA, Kasturiratne A, Pathmeswaran A. **Lean non-alcoholic fatty liver disease (lean NAFLD): characteristics, metabolic outcomes and risk factors from a 7-year prospective, community cohort study from Sri Lanka**. *Hepatol Int* (2019.0) **13** 314-322. DOI: 10.1007/s12072-018-9916-4 11. Singh SP, Kar SK, Panigrahi MK. **Profile of patients with incidentally detected nonalcoholic fatty liver disease (IDNAFLD) in coastal eastern India**. *Trop Gastroenterol* (2013.0) **34** 144-152. DOI: 10.7869/tg.118 12. Fan JG, Kim SU, Wong VW. **New trends on obesity and NAFLD in Asia**. *J Hepatol* (2017.0) **67** 862-873. DOI: 10.1016/j.jhep.2017.06.003 13. Feng RN, Du SS, Wang C. **Lean-non-alcoholic fatty liver disease increases risk for metabolic disorders in a normal weight Chinese population**. *World J Gastroenterol* (2014.0) **20** 17932-17940. DOI: 10.3748/wjg.v20.i47.17932 14. Nakatsuka T, Tateishi R, Koike K. **Changing clinical management of NAFLD in Asia**. *Liver Int* (2021.0). DOI: 10.1111/liv.15046 15. Kuchay MS, Martínez-Montoro JI, Choudhary NS. **Non-Alcoholic Fatty Liver Disease in Lean and Non-Obese Individuals: Current and Future Challenges**. *Biomedicines.* (2021.0) **9** 1346. DOI: 10.3390/biomedicines9101346 16. Ha Y, Seo N, Shim JH. **Intimate association of visceral obesity with non-alcoholic fatty liver disease in healthy Asians: A case-control study**. *J Gastroenterol Hepatol* (2015.0) **30** 1666-1672. DOI: 10.1111/jgh.12996 17. Zou ZY, Wong VW, Fan JG. **Epidemiology of nonalcoholic fatty liver disease in non-obese populations: Meta-analytic assessment of its prevalence, genetic, metabolic, and histological profiles**. *J Dig Dis* (2020.0) **21** 372-384. DOI: 10.1111/1751-2980.12871 18. Hallsworth K, Hollingsworth KG, Thoma C. **Cardiac structure and function are altered in adults with non-alcoholic fatty liver disease**. *J Hepatol* (2013.0) **58** 757-762. DOI: 10.1016/j.jhep.2012.11.015 19. Sunbul M, Kivrak T, Durmus E. **Nonalcoholic Steatohepatitis Score is an Independent Predictor of Right Ventricular Dysfunction in Patients with Nonalcoholic Fatty Liver Disease**. *Cardiovasc Ther* (2015.0) **33** 294-299. DOI: 10.1111/1755-5922.12145 20. Li X, Heiskanen JS, Ma H. **Fatty liver index and left ventricular mass: prospective associations from two independent cohorts**. *J Hypertens* (2021.0) **39** 961-969. DOI: 10.1097/hjh.0000000000002716 21. Lee YH, Kim KJ, Yoo ME. **Association of non-alcoholic steatohepatitis with subclinical myocardial dysfunction in non-cirrhotic patients**. *J Hepatol* (2018.0) **68** 764-772. DOI: 10.1016/j.jhep.2017.11.023 22. Lee M, Kim KJ, Chung TH. **Nonalcoholic fatty liver disease, diastolic dysfunction, and impaired myocardial glucose uptake in patients with type 2 diabetes**. *Diabetes Obes Metab* (2021.0) **23** 1041-1051. DOI: 10.1111/dom.14310 23. Chung GE, Lee JH, Lee H. **Nonalcoholic fatty liver disease and advanced fibrosis are associated with left ventricular diastolic dysfunction**. *Atherosclerosis* (2018.0) **272** 137-144. DOI: 10.1016/j.atherosclerosis.2018.03.027 24. 24.Guidelines of prevention and treatment for nonalcoholic fatty liver disease:a 2018 update. J Pract Hepatol. 2018,21:177–186.doi:10.3969/j.issn.1672-5069.2018.02.007 25. 25.Guidelines for management of nonalcoholic fatty liver disease:an updated and revised edition. Chin J Hepatol. 2010,18:163–166.doi:10.3760/cma.j.issn.1007-3418.2010.03.002 26. Levey AS, Stevens LA, Schmid CH. **A new equation to estimate glomerular filtration rate**. *Ann Intern Med* (2009.0) **150** 604-612. DOI: 10.7326/0003-4819-150-9-200905050-00006 27. Hu YM, Wu XL, Hu ZH. **Study of formula for calculating body surface areas of the Chinese adults**. *Acta Physiol Sin.* (1999.0) **51** 45-48. DOI: 10.3321/j.issn:0371-0874.1999.01.008 28. Silbiger JJ. **Pathophysiology and Echocardiographic Diagnosis of Left Ventricular Diastolic Dysfunction**. *J Am Soc Echocardiogr* (2019.0) **32** 216-232.e212. DOI: 10.1016/j.echo.2018.11.011 29. Shida T, Oshida N, Suzuki H. **Clinical and anthropometric characteristics of non-obese non-alcoholic fatty liver disease subjects in Japan**. *Hepatol Res* (2020.0) **50** 1032-1046. DOI: 10.1111/hepr.13543 30. Flegal KM, Kit BK, Orpana H. **Association of all-cause mortality with overweight and obesity using standard body mass index categories: a systematic review and meta-analysis**. *JAMA* (2013.0) **309** 71-82. DOI: 10.1001/jama.2012.113905 31. Li Y, Chen Y, Tian X. **Comparison of Clinical Characteristics Between Obese and Non-Obese Patients with Nonalcoholic Fatty Liver Disease (NAFLD)**. *Diabetes Metab Syndr Obes* (2021.0) **14** 2029-2039. DOI: 10.2147/dmso.S304634 32. Wong VW, Wong GL, Chan RS. **Beneficial effects of lifestyle intervention in non-obese patients with non-alcoholic fatty liver disease**. *J Hepatol* (2018.0) **69** 1349-1356. DOI: 10.1016/j.jhep.2018.08.011 33. Negi CK, Babica P, Bajard L. **Insights into the molecular targets and emerging pharmacotherapeutic interventions for nonalcoholic fatty liver disease**. *Metabolism.* (2022.0) **126** 154925. DOI: 10.1016/j.metabol.2021.154925 34. Mitter SS, Shah SJ, Thomas JD. **A Test in Context: E/A and E/e' to Assess Diastolic Dysfunction and LV Filling Pressure**. *J Am Coll Cardiol* (2017.0) **69** 1451-1464. DOI: 10.1016/j.jacc.2016.12.037 35. Aksu E, Sokmen A, Ispiroglu M. **Early cardiac electrical and structural changes in patients with non-obese non-alcoholic fatty liver disease**. *Kardiologiia* (2021.0) **61** 51-58. DOI: 10.18087/cardio.2021.5.n1416 36. Catena C, Brosolo G, Da Porto A. **Association of non-alcoholic fatty liver disease with left ventricular changes in treatment-naive patients with uncomplicated hypertension**. *Front Cardiovasc Med* (2022.0) **9** 1030968. DOI: 10.3389/fcvm.2022.1030968 37. Borges-Canha M, Neves JS, Libânio D. **Association between nonalcoholic fatty liver disease and cardiac function and structure-a meta-analysis**. *Endocrine* (2019.0) **66** 467-476. DOI: 10.1007/s12020-019-02070-0 38. Wijarnpreecha K, Lou S, Panjawatanan P. **Association between diastolic cardiac dysfunction and nonalcoholic fatty liver disease: A systematic review and meta-analysis**. *Dig Liver Dis* (2018.0) **50** 1166-1175. DOI: 10.1016/j.dld.2018.09.004 39. Lai YH, Su CH, Hung TC. **Association of Non-Alcoholic Fatty Liver Disease and Hepatic Fibrosis with Epicardial Adipose Tissue Volume and Atrial Deformation Mechanics in a Large Asian Population Free from Clinical Heart Failure**. *Diagnostics (Basel).* (2022.0) **12** 916. DOI: 10.3390/diagnostics12040916 40. Yong JN, Ng CH, Lee CW. **Non-alcoholic fatty liver disease association with structural heart, systolic and diastolic dysfunction: a meta-analysis**. *Hepatol Int* (2022.0) **16** 269-281. DOI: 10.1007/s12072-022-10319-6 41. Jung JY, Park SK, Ryoo JH. **Effect of non-alcoholic fatty liver disease on left ventricular diastolic function and geometry in the Korean general population**. *Hepatol Res* (2017.0) **47** 522-532. DOI: 10.1111/hepr.12770 42. Petta S, Argano C, Colomba D. **Epicardial fat, cardiac geometry and cardiac function in patients with non-alcoholic fatty liver disease: association with the severity of liver disease**. *J Hepatol* (2015.0) **62** 928-933. DOI: 10.1016/j.jhep.2014.11.030 43. Zhang Z, Xun Y, Rong S. **Loss of immunity-related GTPase GM4951 leads to nonalcoholic fatty liver disease without obesity**. *Nat Commun* (2022.0) **13** 4136. DOI: 10.1038/s41467-022-31812-4 44. Chiu LS, Pedley A, Massaro JM. **The association of non-alcoholic fatty liver disease and cardiac structure and function-Framingham Heart Study**. *Liver Int* (2020.0) **40** 2445-2454. DOI: 10.1111/liv.14600 45. VanWagner LB, Wilcox JE, Ning H. **Longitudinal Association of Non-Alcoholic Fatty Liver Disease With Changes in Myocardial Structure and Function: The CARDIA Study**. *J Am Heart Assoc.* (2020.0) **9** e014279. DOI: 10.1161/jaha.119.014279 46. Dai W, Zhang Z, Zhao S. **The Risk of Type 2 Diabetes and Coronary Artery Disease in Non-obese Patients With Non-alcoholic Fatty Liver Disease: A Cohort Study**. *Front Cardiovasc Med.* (2021.0) **8** 680664. DOI: 10.3389/fcvm.2021.680664 47. Younes R, Govaere O, Petta S. **Caucasian lean subjects with non-alcoholic fatty liver disease share long-term prognosis of non-lean: time for reappraisal of BMI-driven approach?**. *Gut* (2022.0) **71** 382-390. DOI: 10.1136/gutjnl-2020-322564 48. Sookoian S, Pirola CJ. **Systematic review with meta-analysis: the significance of histological disease severity in lean patients with nonalcoholic fatty liver disease**. *Aliment Pharmacol Ther* (2018.0) **47** 16-25. DOI: 10.1111/apt.14401 49. Watt MJ, Miotto PM, De Nardo W. **The Liver as an Endocrine Organ-Linking NAFLD and Insulin Resistance**. *Endocr Rev* (2019.0) **40** 1367-1393. DOI: 10.1210/er.2019-00034 50. Portillo-Sanchez P, Bril F, Maximos M. **High Prevalence of Nonalcoholic Fatty Liver Disease in Patients With Type 2 Diabetes Mellitus and Normal Plasma Aminotransferase Levels**. *J Clin Endocrinol Metab* (2015.0) **100** 2231-2238. DOI: 10.1210/jc.2015-1966 51. Feldman A, Eder SK, Felder TK. **Clinical and Metabolic Characterization of Lean Caucasian Subjects With Non-alcoholic Fatty Liver**. *Am J Gastroenterol* (2017.0) **112** 102-110. DOI: 10.1038/ajg.2016.318 52. Boursier J, Mueller O, Barret M. **The severity of nonalcoholic fatty liver disease is associated with gut dysbiosis and shift in the metabolic function of the gut microbiota**. *Hepatology* (2016.0) **63** 764-775. DOI: 10.1002/hep.28356 53. Loomba R, Seguritan V, Li W. **Gut Microbiome-Based Metagenomic Signature for Non-invasive Detection of Advanced Fibrosis in Human Nonalcoholic Fatty Liver Disease**. *Cell Metab* (2017.0) **25** 1054-1062.e1055. DOI: 10.1016/j.cmet.2017.04.001 54. Iwaki M, Kessoku T, Ozaki A. **Gut microbiota composition associated with hepatic fibrosis in non-obese patients with non-alcoholic fatty liver disease**. *J Gastroenterol Hepatol* (2021.0) **36** 2275-2284. DOI: 10.1111/jgh.15487 55. Kalia HS, Gaglio PJ. **The Prevalence and Pathobiology of Nonalcoholic Fatty Liver Disease in Patients of Different Races or Ethnicities**. *Clin Liver Dis* (2016.0) **20** 215-224. DOI: 10.1016/j.cld.2015.10.005 56. Hernaez R, McLean J, Lazo M. **Association between variants in or near PNPLA3, GCKR, and PPP1R3B with ultrasound-defined steatosis based on data from the third National Health and Nutrition Examination Survey**. *Clin Gastroenterol Hepatol* (2013.0) **11** 1183-1190.e1182. DOI: 10.1016/j.cgh.2013.02.011 57. Sookoian S, Castaño GO, Scian R. **Genetic variation in transmembrane 6 superfamily member 2 and the risk of nonalcoholic fatty liver disease and histological disease severity**. *Hepatology* (2015.0) **61** 515-525. DOI: 10.1002/hep.27556 58. Mancina RM, Dongiovanni P, Petta S. **The MBOAT7-TMC4 Variant rs641738 Increases Risk of Nonalcoholic Fatty Liver Disease in Individuals of European Descent**. *Gastroenterology* (2016.0) **150** 1219-1230.e1216. DOI: 10.1053/j.gastro.2016.01.032 59. Petersen KF, Dufour S, Hariri A. **Apolipoprotein C3 gene variants in nonalcoholic fatty liver disease**. *N Engl J Med* (2010.0) **362** 1082-1089. DOI: 10.1056/NEJMoa0907295 60. Abul-Husn NS, Cheng X, Li AH. **A Protein-Truncating HSD17B13 Variant and Protection from Chronic Liver Disease**. *N Engl J Med* (2018.0) **378** 1096-1106. DOI: 10.1056/NEJMoa1712191 61. Graff M, North KE, Franceschini N. **PNPLA3 gene-by-visceral adipose tissue volume interaction and the pathogenesis of fatty liver disease: the NHLBI family heart study**. *Int J Obes (Lond)* (2013.0) **37** 432-438. DOI: 10.1038/ijo.2012.65 62. Wei JL, Leung JC, Loong TC. **Prevalence and Severity of Nonalcoholic Fatty Liver Disease in Non-Obese Patients: A Population Study Using Proton-Magnetic Resonance Spectroscopy**. *Am J Gastroenterol.* (2015.0) **110** 1306-1314. DOI: 10.1038/ajg.2015.235 63. Dongiovanni P, Petta S, Maglio C. **Transmembrane 6 superfamily member 2 gene variant disentangles nonalcoholic steatohepatitis from cardiovascular disease**. *Hepatology* (2015.0) **61** 506-514. DOI: 10.1002/hep.27490 64. Fracanzani AL, Petta S, Lombardi R. **Liver and Cardiovascular Damage in Patients With Lean Nonalcoholic Fatty Liver Disease, and Association With Visceral Obesity**. *Clin Gastroenterol Hepatol* (2017.0) **15** 1604-1611.e1601. DOI: 10.1016/j.cgh.2017.04.045 65. Iwaki M, Kessoku T, Tanaka K. **Comparison of long-term prognosis between non-obese and obese patients with non-alcoholic fatty liver disease**. *JGH Open* (2022.0) **6** 696-703. DOI: 10.1002/jgh3.12808 66. Francque SM, van der Graaff D, Kwanten WJ. **Non-alcoholic fatty liver disease and cardiovascular risk: Pathophysiological mechanisms and implications**. *J Hepatol* (2016.0) **65** 425-443. DOI: 10.1016/j.jhep.2016.04.005 67. Stahl EP, Dhindsa DS, Lee SK. **Nonalcoholic Fatty Liver Disease and the Heart: JACC State-of-the-Art Review**. *J Am Coll Cardiol* (2019.0) **73** 948-963. DOI: 10.1016/j.jacc.2018.11.050 68. Targher G, Byrne CD, Tilg H. **NAFLD and increased risk of cardiovascular disease: clinical associations, pathophysiological mechanisms and pharmacological implications**. *Gut* (2020.0) **69** 1691-1705. DOI: 10.1136/gutjnl-2020-320622 69. Chang Y, Ryu S, Sung KC. **Alcoholic and non-alcoholic fatty liver disease and associations with coronary artery calcification: evidence from the Kangbuk Samsung Health Study**. *Gut* (2019.0) **68** 1667-1675. DOI: 10.1136/gutjnl-2018-317666 70. Ahmed OT, Gidener T, Mara KC. **Natural History of Nonalcoholic Fatty Liver Disease With Normal Body Mass Index: A Population-Based Study**. *Clin Gastroenterol Hepatol* (2021.0). DOI: 10.1016/j.cgh.2021.07.016
--- title: The risk factors for Group B Streptococcus colonization during pregnancy and influences of intrapartum antibiotic prophylaxis on maternal and neonatal outcomes authors: - Xiaoli Chen - Sijia Cao - Xiaochun Fu - Yan Ni - Bixuan Huang - Jiayin Wu - Ling Chen - Shuying Huang - Jiali Cao - Weiwei Yu - Huiming Ye journal: BMC Pregnancy and Childbirth year: 2023 pmcid: PMC10041798 doi: 10.1186/s12884-023-05478-9 license: CC BY 4.0 --- # The risk factors for Group B Streptococcus colonization during pregnancy and influences of intrapartum antibiotic prophylaxis on maternal and neonatal outcomes ## Abstract ### Background Group B Streptococcus (GBS), also referred as Streptococcus agalactiae, is one of the leading causes of life-threatening invasive diseases such as bacteremia, meningitis, pneumonia and urinary tract infection in pregnant women and neonates. Rates of GBS colonization vary by regions, but large-sample studies on maternal GBS status are limited in southern China. As a result, the prevalence of GBS among pregnant women and its associated risk factors and the efficacy of intrapartum antibiotic prophylaxis (IAP) intervention in preventing adverse pregnancy and neonatal outcomes remain poorly understood in southern China. ### Methods To fill this gap, we retrospectively analyzed demographic and obstetrical data of pregnant women who have undergone GBS screening and delivered between 2016 and 2018 in Xiamen, China. A total of 43,822 pregnant women were enrolled and only a few GBS-positive women did not receive IAP administration. Possible risk factors for GBS colonization were assayed by univariate and multivariate logistic regression analysis. Generalized linear regression model was applicated to analyze whether IAP is one of the impact factors of the hospital length of stay of the target women. ### Results The overall GBS colonization rate was $13.47\%$ ($\frac{5902}{43}$,822). Although women > 35 years old ($$P \leq 0.0363$$) and women with diabetes mellitus (DM, $$P \leq 0.001$$) had a higher prevalence of GBS colonization, the interaction between ages and GBS colonization was not statistically significant in Logistic *Regression analysis* (adjusted OR = 1.0014; $95\%$ CI, 0.9950, 1.0077). The rate of multiple births was significantly dropped in GBS-positive group than that of GBS-negative group ($$P \leq 0.0145$$), with no significant difference in the rate of fetal reduction ($$P \leq 0.3304$$). Additionally, the modes of delivery and the incidences of abortion, premature delivery, premature rupture of membranes, abnormal amniotic fluid and puerperal infection were not significantly different between the two groups. The hospitalization stays of the subjects were not influenced by GBS infection. As for neonatal outcomes, the cases of fetal death in maternal GBS-positive group did not statistically differ from that in maternal GBS-negative group. ### Conclusion Our data identified that pregnant women with DM are at high risk of GBS infection and IAP is highly effective in prevention of adverse pregnancy and neonatal outcomes. This stressed the necessity of universal screening of maternal GBS status and IAP administration to the target population in China, and women with DM should be considered as priorities. ## Background Group B Streptococcus (GBS), also known as Streptococcus agalactiae, is a Gram-positive bacterium which asymptomatically colonizes in women rectovaginal areas and could result in adult and neonatal invasive diseases under certain conditions [1]. GBS infection can lead to invasive diseases such as bacteremia and skin/soft tissue infection in nonpregnant adults, the burden of which has been increased significantly during the past few years [2–4]. The GBS carriage of pregnant women can be chronic, intermittent or transient and is implicated in urinary tract infection, premature rupture of membranes, and preterm birth [5–7]. According to a literature, the prevalence of GBS invasive diseases in pregnant women was twice as much as that in nonpregnant women [8]. GBS could be transmitted vertically from colonized mothers to their offspring through genital tract at or just before delivery, which may cause early-onset invasive neonatal GBS disease (EOD) that occurs < 7 days of life, often manifesting as bacteremia and pneumonia [9, 10]. The incidence of EOD among infants born to women with GBS colonization was 29 times higher than that of infants born to women without GBS colonization [11]. Invasive neonatal GBS disease that appears from 7 to 90 days of life is referred as late-onset disease (LOD), the common manifestations of which are bacteremia, urinary tract infection, and meningitis [12, 13]. Newborns with LOD are exposed to GBS by horizontal transmission. Although the risk factors of LOD are not as well understood as EOD, it was suggested that babies usually acquired the same serotype of GBS as their mothers’ colonized strains, and GBS-positive breast milk was implicated in heavy neonatal infection that GBS could be isolated from their throat, ear and rectum at least once [14]. It was reported by a meta-analysis that estimates of maternal GBS colonization in the world vary by regions, with rates ranging from 11 to $35\%$ [15]. Annually, GBS infection causes high morbidity and mortality worldwide. The incidence rate of systemic invasive GBS diseases in pregnant women is 0.38 per 1000 pregnancies with case fatality rate of $0.2\%$ [8]. The invasive GBS disease rate in newborns is 0.49 per 1000 live births [9]. The Centers for Disease Control and Prevention (CDC) recommends pregnant women at 35–37 weeks of gestation should be screened for GBS carriage through culture-based strategy or risk-based approach. And culture-positive women or women with any risk factors for EOD should receive intrapartum antibiotic prophylaxis (IAP) [16]. According to the recommendation provided by CDC, GBS-colonized parturient women were offered IAP at the time of labor onset or rupture of membranes until their delivery, and penicillin, ampicillin and cefazolin were currently the agents of choice for IAP. In China, universal prenatal screening for GBS carriage has not been carried out. Given the regional variations in GBS infection, it is necessary to develop a proper strategy to test GBS colonization status in women at late pregnancy. Since there are no available GBS vaccines, it is essential to evaluate the efficacy of IAP in preventing GBS-related adverse outcomes at different regions [17, 18].We previously reported the combination of GBS chromogenic media with GBS carrot agar and β-γ detection agar enhanced the detection rate of GBS in vaginal and rectal swabs significantly [19], which was also used in this research to test GBS status among pregnant women. Wenjing Ji et al. found the incidence of invasive neonatal GBS diseases was 0.31 cases/1,000 live births and the case-fatality rate was $2.3\%$ in China, suggesting enhanced surveillance and preventive strategies should be carried out in China [20]. Moreover, Yao Zhu’s group showed that IAP was effective in reducing GBS-EOD and recommended universal screening of maternal GBS and subsequent IAP intervention in China [11]. In this study, we aim to determine the prevalence of GBS among pregnant women in southern China, to find out the high risk-group of invasive GBS diseases and to calculate the efficacy of IAP treatment, which may aid in the development of intervention programs. ## Population and Study Design This study was carried out at Women and Children’s Hospital, School of Medicine, Xiamen University, China. Pregnant women who attended the hospital for prenatal tests including GBS screening and delivered from 2016 to 2018 were eligible for inclusion. Data on the subjects and their babies were obtained from their medical records and reviewed retrospectively. According to the algorithms provided by the CDC, women admitted with signs and symptoms of preterm labor and women with rupture of membranes at < 37 weeks and 0 days’ gestation should be collected vaginal-rectal swabs for GBS culture and started GBS prophylaxis once they entered true labor [16]. Therefore, a minority of the participants in this study with threatened preterm labor, risks of preterm delivery such as multiple births or premature rupture of membranes might have the test before 35 weeks of gestation. The antenatal screening results would guide GBS management of pregnant women at the time of labor. Since a negative GBS screen is considered valid for 5 weeks [16], the negative participants who did not give birth in 5 weeks would undergo repeat GBS screening before their delivery. This research was approved by the Human Research Ethics Committee of Women and Children’s Hospital of Xiamen University (KY-2020-103) and was conducted strictly in accordance with the approval. ## GBS colonization determination GBS was tested as previously described [19]. The vaginal and rectal samples of each patient were collected using GBS TranSwab (Creative Lifesciences, China), which were subsequently incubated at 37 ℃ in $5\%$ CO2 for 18 to 24 h within 2 h of collection. The detection of color change or red-orange pigment after enrichment was specific for the presence of β-hemolytic GBS strains. The negative samples (no color change or red-orange spots) were further inoculated onto GBS Carrot Agar and β-γ Detection Agar (Creative Lifesciences, China) and were cultivated for another 24 h in order to generate β-hemolysis in γ-hemolytic strains. The red-orange pigmented β-hemolytic colonies were indicative of nonhemolytic GBS strains, and the nonpigmented β-hemolytic colonies were finally detected by CAMP test or matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) to confirm the presence of nonpigmented or CAMP test-negative GBS strains. Once the pregnant women were confirmed GBS culture-positive, they would be provided with IAP. ## IAP administration Pregnant women with threatened preterm labor, risks of preterm delivery such as multiple births or premature rupture of membranes could be collected vaginal-rectal swabs for GBS culture before 35 weeks of gestation. And all other pregnant women were screened for GBS at 35–37 weeks of gestation. At the time of labor or rupture of membranes, IAP should be given to women who tested positive for GBS colonization. If GBS culture results were unknown at the time of delivery onset, women at < 37 weeks and 0 days’ gestation, had a duration of membrane rupture ≥ 18 h, or had a temperature of ≥ 100.4º F (≥ 38.0ºC) were also treated with IAP. Once the GBS status were available prior to delivery and were negative, the GBS prophylaxis would be discontinued, otherwise IAP administration would be continued until their delivery. However, there were 90 culture-confirmed GBS-positive pregnant women were not administrated with IAP in time or more than 4 h. IAP agents and dosage were applied to patients according to the guidelines released by the CDC [16]. An initial dose of 4.8 million U of penicillin were given to GBS culture-positive women at the time of labor onset via intravenous injection, followed by 2.4 million U of penicillin at an interval of 4 h. Pregnant women with premature rupture of membranes were intravenously injected with an initial dose of 2 g ampicillin, followed by 1 g ampicillin at an interval of 4 h. Penicillin-allergic women could be administrated with cefazolin or clindamycin. ## Data abstraction and statistical analysis The vaginal and rectal samples of each patient were collected followed consisted standard. The laboratory has passed ISO15189 quality management system certification. The medical record apartment has passed the five-level evaluation of the application level of the electronic medical record system of the National Health Commission of China. *The* general information and obstetrical data of the target population were entered into a standard Excel form and reviewed twice by the investigators, then we used R v4.0.5 software for statistical analysis. Normal distribution data were presented as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\stackrel{-}{x}$$\end{document}± s and skewness distribution data as M (P25, P75). Enumeration data were presented as absolute numbers. Z test or Mann-Whitney U test was employed for the comparison of measurement data between groups of women with GBS colonization vs. women without GBS colonization. Chi-square test or Fisher’s exact test was used for the comparison of enumeration data between groups of women with GBS colonization vs. women without GBS colonization. Generalized linear regression model was applicated to analyze the impact factors of the hospital length of stay of the target women. Logistic regression analysis model was used to assay the risk factors for GBS colonization and fetal distress. $P \leq 0.05$ (*) was considered statistically significant. ## Risk factors for GBS colonization among pregnant women A total of 43,822 women were included in this study, with ages ranging from 16 to 56 years old. The demographic data of the investigated population were summarized in Table 1. The overall mean age of the 43,822 pregnant women was 30.67 ± 4.42 years old. Of them, about $13.46\%$ ($\frac{5902}{43}$,822) were identified as GBS carriers. The correlation between age group and GBS colonization was statistically different ($$P \leq 0.0363$$). In detail, the proportion of GBS-positive women was higher in ≥ 35 years old group. The disparities in characteristics including ABO blood group and Rh blood group were statistically insignificant. Ethnic backgrounds or native places did not differ significantly between these two groups. Table 1Demographic characteristics of pregnant womenVariablesGBS colonizationP valuePositive($$n = 5902$$)Negative($$n = 37$$,920)Age (years)0.0363*≤ 2014 ($0.24\%$)138 ($0.36\%$)[20, 35]4686 ($79.40\%$)30,501 ($80.44\%$)≥ 351202 ($20.37\%$)7281 ($19.20\%$)ABO Blood Group0.5703A1666 ($28.23\%$)10,773 ($28.41\%$)B1431 ($24.25\%$)9135 ($24.09\%$)O2438 ($41.31\%$)15,481 ($40.83\%$)AB367 ($6.22\%$)2531 ($6.67\%$)Rh Blood Group0.9309Negative32 ($0.54\%$)209 ($0.55\%$)Positive5870 ($99.46\%$)37,711 ($99.45\%$)Ethnic Group0.1774Han5877 ($99.58\%$)37,715 ($99.46\%$)She2 ($0.03\%$)31 ($0.08\%$)Miao3 ($0.05\%$)25 ($0.07\%$)Hui1 ($0.02\%$)25 ($0.07\%$)Mongolian1 ($0.02\%$)23 ($0.06\%$)Other18 ($0.30\%$)101 ($0.27\%$)Native Place0.1644Fujian4924 ($83.43\%$)31,047 ($81.88\%$)Jiangxi183 ($3.10\%$)1334 ($3.52\%$)Sichuan87 ($1.47\%$)689 ($1.82\%$)Henan76 ($1.29\%$)592 ($1.56\%$)Anhui77 ($1.30\%$)473 ($1.25\%$)Other555 ($9.40\%$)3785 ($9.98\%$) The obstetrical data and underlying diseases of the study population were shown in Table 2. Women with threatened preterm labor, risks of preterm delivery such as multiple births or premature rupture of membranes might have the test before 35 gestational weeks. No significant difference was observed between GBS-positive women and GBS-negative women in terms of parity ($$P \leq 0.1688$$), gravidity ($$P \leq 0.4070$$) and gestational weeks ($$P \leq 0.6769$$). The proportion of pregnant women with diseases including anemia ($$P \leq 0.3946$$), eclampsia ($$p \leq 0.8879$$), cholestasis ($$P \leq 0.7179$$) and thyroid dysfunction ($$P \leq 0.8058$$) showed no significant difference between the two groups. Whereas the proportion of pregnant women with diabetes were significantly higher in GBS carriers than noncarriers ($$P \leq 0.0010$$). Table 2Obstetric characteristics of pregnant womenCharacteristicsGBS colonizationP valuePositive($$n = 5902$$)Negative($$n = 37$$,920)Parity (times)0.168812097 ($35.53\%$)13,236 ($34.91\%$)21901 ($32.21\%$)11,968 ($31.56\%$)31119 ($18.96\%$)7294 ($19.24\%$)≥ 4785 ($13.30\%$)5422 ($14.30\%$)Gravidity (times)0.40700254 ($4.30\%$)1773 ($4.68\%$)12835 ($48.03\%$)18,330 ($48.34\%$)22547 ($43.15\%$)16,237 ($42.82\%$)3252 ($4.27\%$)1474 ($3.89\%$)≥ 414 ($0.24\%$)106 ($0.28\%$)Gestational Age (weeks)0.1571≤ 28212 ($3.59\%$)1356 ($3.58\%$)(28, 32]102 ($1.73\%$)830 ($2.19\%$)(32, 34]118 ($2.00\%$)805 ($2.12\%$)(34, 37]366 ($6.20\%$)2465 ($6.50\%$)> 375104 ($86.48\%$)32,464 ($85.61\%$)Eclampsia0.8879Yes110 ($1.86\%$)693 ($1.83\%$)No5792 ($98.14\%$)37,227 ($98.17\%$)Cholestasis0.7179Yes114 ($1.93\%$)763 ($2.01\%$)No5788 ($98.07\%$)37,157 ($97.99\%$)Thyroid Dysfunction0.8058Yes501 ($8.49\%$)3179 ($8.38\%$)No5401 ($91.51\%$)34,741 ($91.62\%$)Diabetic Mellitus0.0010**Yes1233 ($20.89\%$)7230 ($19.07\%$)No4669 ($79.11\%$)30,690 ($80.93\%$) As indicated in Table 3, variables including ages and DM were further included in a multiple logistic regression analysis. After adjustment, only DM was significantly associated with GBS colonization (OR = 1.1175, $95\%$ CI = 1.0423, 1.1973), while the relation between ages and GBS colonization was no longer statistically significant (OR = 1.0014, $95\%$ CI = 0.9950, 1.0077). Table 3Logistic regression analysis of the risk factors for maternal GBS-colonizationFactorsβ (SD)$95\%$ CI of βOR$95\%$ CI of ORP valueAges0.0014 (0.0032)(-0.0050, 0.0077)1.0014(0.9950, 1.0077)0.6724Diabetes Mellitus0.1111 (0.0354)(0.0414, 0.1801)1.1175(1.0423, 1.1973)0.0016** Taken together, our data suggested that diabetic mellitus might be one of the risk factors for GBS colonization during pregnancy. ## Pregnancy outcomes of the Study Population by GBS colonization status To evaluate the influence of IAP on pregnancy outcomes, we analyzed the rates of abortion and multiple-pregnancy between GBS carriers and negative controls and results were listed in Table 4. It seemed that the incidence of abortion (including spontaneous abortion and induced abortion) was similar in these two groups ($$P \leq 0.2084$$). In contrast, the proportion of multiple-pregnancy in GBS carriers was dropped significantly than that in GBS-negative group ($$P \leq 0.0145$$), with no significant difference in the rate of multifetal pregnancy reductions ($$P \leq 0.3304$$). Table 4Comparison of pregnancy outcomes between GBS carriers and noncarriersParametersGBS colonizationP valuePositiven = 5902Negativen = 37,920Abortion0.2084Yes518 ($8.78\%$)3525 ($9.30\%$)No5384 ($91.22\%$)34,395 ($90.70\%$)Multiple-pregnancy0.0145*Yes170 ($2.88\%$)1332 ($3.51\%$)No5732 ($97.12\%$)36,588 ($96.49\%$)Multiple-pregnancy with fetal reduction0.3304Yes5 ($2.94\%$)66 ($4.95\%$)No165 ($97.06\%$)1266 ($95.05\%$) We further excluded 4019 women who did not deliver (ended their pregnancy with abortion or stillbirths). As shown in Table 5, the prevalence of GBS among pregnant women at different gestational ages was similar ($$P \leq 0.6769$$). The proportion of cesarean section in GBS-positive group did not differ from that in GBS-negative group significantly ($$P \leq 0.5573$$). In addition, there was no significant difference in terms of the rates of premature delivery ($$P \leq 0.2077$$), premature rupture of membranes ($$P \leq 0.6769$$) or abnormal amniotic fluid ($$P \leq 0.6634$$). Only one case of puerperal infection was found in GBS culture-negative pregnant women. Table 5The association between GBS colonization and adverse pregnancy outcomesFactorsGBS colonizationP valuePositiven = 5382Negativen = 34,421Modes of Delivery0.5573Vaginal Birth3368 ($62.58\%$)21,687 ($63.01\%$)Cesarean Section2014 ($34.12\%$)12,734 ($36.99\%$)Premature Delivery0.2077Yes457 ($8.49\%$)3108 ($9.03\%$)No4925 ($91.51\%$)31,313 ($90.97\%$)Premature Rupture of Membranes0.1282Yes1164 ($21.63\%$)7129 ($20.71\%$)No4218 ($78.37\%$)27,292 ($79.29\%$)Abnormal amniotic fluid0.6634Yes180 ($3.34\%$)1195 ($3.47\%$)No5202 ($96.66\%$)33,226 ($96.53\%$)Puerperal Infection1Yes01No538234,420 From these results, we concluded that though the multiple births rate was reduced among GBS-positive pregnant women, IAP intervention was important in ameliorating the adverse pregnancy outcomes including premature delivery, premature rupture of membranes, abnormal amniotic fluid and puerperal infection. ## Factors affecting the hospitalization stays of pregnant women analyzed by generalized Linear Regression Analysis GBS related invasive diseases remains a heavy burden of public health system and the therapy of which costs a lot. Herein we used hospital length of stay as an indicator of disease severity and disease burden as Meredith Deutscher et al. suggested [21]. In order to evaluate the severity of GBS infection after IAP application, we assayed the effect of GBS carriage on the hospital length of stays of pregnant women by generalized linear regression analysis (Table 6). The data indicated that the hospitalization days have no association with GBS colonization status, ages of the pregnant women or occurrence of puerperal infection, but were positively related to cesarean section, gravidity, abortion, premature delivery, diabetes, eclampsia, anemia, thyroid dysfunction, cholestasis as well as multiple births, and were negatively related to parity and premature rupture of membranes. Table 6Generalized Linear Regression Analysis of the Hospital Length of Stays of Pregnant WomenFactorsβ (SD)$95\%$CIP valueGBS Colonization0.049 (0.031)(-0.012, 0.109)0.1135Ages0.005 (0.003)(-0.001, 0.010)0.0796Cesarean Section1.868 (0.023)(1.823, 1.914)< 0.001***Gravidity0.062 (0.012)(0.038, 0.084)< 0.001***Parity-0.477 (0.025)(-0.526, -0.428)< 0.001***Abortion1.318 (0.288)(0.753, 1.883)< 0.001***Premature Delivery1.993 (0.043)(1.908, 2.078)< 0.001***Diabetes Mellitus0.190 (0.027)(0.137, 0.244)< 0.001***Premature Rupture of Membranes-0.197 (0.027)(-0.249, -0.145)< 0.001***Eclampsia0.416 (0.076)(0.267, 0.565)< 0.001***Puerperal Infection2.730 (2.091)(-1.367, 6.828)0.1916Anemia0.335 (0.024)(0.288, 0.381)< 0.001***Thyroid Dysfunction0.119 (0.038)(0.045, 0.193)0.0017**Cholestasis0.983 (0.080)(0.457, 0.735)< 0.001***Multiple-pregnancy0.596 (0.071)(0.648, 0.921)< 0.001*** These results revealed that after IAP therapy, the hospitalization stays of pregnant women were not affected by GBS infection. ## Outcomes of newborns by maternal GBS colonization status Of the 40,905 fetuses carried by the study population, 81 were stillbirths. As shown in Table 7, the incidence of stillbirth between maternal GBS-colonized group and noncolonized group did not differ significantly ($$P \leq 0.8975$$). In order to figure out the effect of IAP application to GBS carriers on infants, we examined the characteristics of the remaining 40,824 neonates and summarized in Table 8. No significant difference was observed in fetal gender, weight, height or the rate of nuchal cord or cord torsion ($$P \leq 0.6284$$) between maternal GBS-positive group and maternal GBS-negative group. Although the APGAR score was comparable in these two groups, the incidence of fetal distress in maternal GBS-colonized group was significantly declined than that in noncolonized group ($$P \leq 0.0385$$), which was out of our expectation. Table 7The incidence of stillbirth in maternal GBS-positive group and maternal GBS-negative groupMaternal GBS ColonizationP valuePositiven = 5502Negativen = 35,403StillbirthYes10 ($0.18\%$)71 ($0.20\%$)0.8975No5492 ($99.82\%$)35,332 ($99.80\%$) Table 8The basic information of live birth infantsCharacteristicsMaternal GBS ColonizationP valuePositiven = 5492Negativen = 35,332Fetal Gender0.1213Male2923 ($53.22\%$)19,204 ($54.35\%$)Female2569 ($46.78\%$)16,128 ($45.65\%$)Weight (grams)3195.96 ± 479.793184.97 ± 497.770.1163Height (centimeters)49.58 ± 1.9749.56 ± 1.940.4375Nuchal Cord or Cord Torsion0.6284Yes1848 ($33.65\%$)12,010 ($33.99\%$)No3644 ($66.35\%$)23,322 ($66.01\%$)Fetal Distress0.0385*Yes302 ($5.50\%$)2201 ($6.23\%$)No5190 ($94.50\%$)33,131 ($93.77\%$)APGAR19.87 ± 0.559.87 ± 0.590.3487APGAR59.93 ± 0.369.92 ± 0.390.7426APGAR109.94 ± 0.319.94 ± 0.350.6973 This finding prompted us to re-analyze our data with mothers taken as the study subjects. As summarized in Table 9, the differences between pregnant women with fetal distress and pregnant women without distress were statistically significant in ages ($P \leq 0.001$), gestational ages ($P \leq 0.001$), the rate of multiple pregnancies ($P \leq 0.001$), the incidence of nuchal cord or cord torsion ($P \leq 0.001$) as well as abnormal amniotic fluid ($P \leq 0.001$). However, no significant difference was found between these two groups in terms of the incidences of cholestasis ($$P \leq 0.5967$$), diabetic mellitus ($$P \leq 0.0915$$) and GBS colonization ($$P \leq 0.4841$$). We further assayed the potential impact factors of fetal distress in a logistic regression model. The results in Table 10 indicated that among the factors, ages, gestational ages and the occurrence of nuchal cord or cord torsion and abnormal amniotic fluid were significant. The risk of fetal distress ascended with decreasing ages (OR = 0.9673, $95\%$ CI=-0.046, -0.0204), and descended with increasing gestational ages (OR = 1.0997, $95\%$ CI = 1.0712, 1.1312) and the occurrence of nuchal cord or cord torsion (OR = 2.1794, $95\%$ CI = 1.9587, 2.4252) and abnormal amniotic fluid (OR = 3.8839, $95\%$ CI = 3.2448, 4.6204). All other factors including multiple-pregnancy, cholestasis, diabetes mellitus and GBS colonization were found to be insignificant. Table 9Maternal characteristics by fetal distressCharacteristicsFetal DistressP valueYesn = 1436Non = 42,386Age (Years)29.91 ± 4.2530.63 ± 4.42< 0.001***Gestational age (weeks)39.71 (38.57, 40.43)39.14 (38.14, 40.00)< 0.001***Multiple births< 0.001***Yes22 ($1.53\%$)1480 ($3.49\%$)No1414 ($98.47\%$)40,906 ($96.51\%$)Cholestasis0.5967Yes32 ($2.23\%$)845 ($2.00\%$)No1404 ($97.77\%$)41,541 ($98.01\%$)Diabetic Mellitus0.0915Yes252 ($17.55\%$)8211 ($19.37\%$)No1184 ($82.45\%$)34,175 ($80.63\%$)Nuchal Cord or Cord Torsion< 0.001***Yes736 ($51.25\%$)13,032 ($30.75\%$)No700 ($48.75\%$)29,354 ($69.25\%$)Abnormal amniotic fluid< 0.001***Yes158 ($11.00\%$)1307 ($3.08\%$)No1278 ($89.00\%$)41,079 ($96.92\%$)GBS colonization0.4841Yes184 ($12.81\%$)5718 ($13.49\%$)No1252 ($87.19\%$)36,668 ($86.51\%$) Table 10Logistic regression analysis of the risk factors for fetal distressFactorsβ (SD)$95\%$ CI of βOR$95\%$ CI of ORP valueAges-0.0332 (0.0066)(-0.046, -0.0204)0.9673(0.9549, 0.9798)< 0.001***Gestational ages0.0950 (0.0139)(0.0687, 0.1233)1.0997(1.0712, 1.1312)< 0.001***Multiple-pregnancy-0.4253 (0.2221)(-0.8896, -0.0148)0.6535(0.4108, 0.9853)0.0555Cholestasis0.3176 (0.1850)(-0.0644, 0.6630)1.3738(0.9376, 1.9407)0.0859Diabetes Mellitus0.0187 (0.0726)(-0.1255, 0.1592)1.0188(0.8820, 1.1726)0.7972Nuchal Cord or Cord Torsion0.7790 (0.0545)(0.6723, 0.8859)2.1794(1.9587, 2.4252)< 0.001***Abnormal amniotic fluid1.3568 (0.0901)(1.1771, 1.5305)3.8839(3.2448, 4.6204)< 0.001***GBS colonization status0.0512 (0.0808)(-0.1044, 0.2125)1.0525(0.9008, 1.2368)0.5263 Overall, our research suggested that IAP treatment was highly effective in preventing adverse neonatal outcomes. Although the fetal distress rate of babies born to GBS carriers was reduced in comparison with non-carriers, the logistic regression analysis showed the relation between fetal distress and GBS colonization was insignificant. ## Discussion The current study suggested that the prevalence of GBS colonization was $13.47\%$ ($\frac{5902}{43}$,822) in Xiamen, China. We demonstrated that in ≥ 35 years old pregnant women, the proportion of GBS carriers was significantly higher than non-carriers. Whereas the association between ages and GBS colonization was not statistically significant. In the contrast, DM is one of the risk factors for GBS colonization. After IAP usage, pregnancy and neonatal outcomes and hospitalization of stays were not affected by GBS status. The regional difference of GBS colonization rate has been verified by lots of researches. Others reported $4.9\%$ in Shenzhen, China [22], $17\%$ in Karachi, Pakistan [23], $6.5\%$ in Kocaeli, Turkey [24], 4.2 to $28.4\%$ in Brazil [25] and $21\%$ in the Hague, The Netherlands [26]. The disparities in sampling sites, detection methods and target populations might attribute to the regional variations. It stressed the necessity to promote universal GBS screening at different regions. Consisted with our finding, Tsering Chomu Dechen et al. showed that the relation between age group and GBS culture positivity was statistically insignificant [27]. While an early study has indicated the statistically significant association between increasing ages and lower GBS culture-positive rate [28], which was contrary to our conclusion. Since some studies have reported the correlation of GBS detection and decreased α-diversity and Lactobacillus species, we reckoned the impaired vaginal microecology in elder pregnant women might account for the increasing GBS detection rate [29, 30]. If an association between maternal age and GBS colonization exists in the southern of China, it might be swamped by other variables in our research. Several studies have claimed that GBS were highly invasive in diabetic patients [4, 31], which supported our observation. There were also some studies declaring the significant association between increasing parity, gravidity and gestational ages and higher GBS colonization rate [22, 32, 33]. The increased susceptibility to GBS of these women might result from their decreased immunity and ability to eliminate the organism. Many studies have showed the significantly higher burden of GBS invasive diseases in black race than nonblack women [7, 34]. Furthermore, a study focused on a multicultural pregnant population from the Netherlands, showing that compared to European women, African women were at higher risk for GBS carriage while Asian women at lower risk [26]. However, the reason for these differences remains elusive. To our knowledge this is the first report to identify the rate of multiple births was significantly dropped in GBS-positive group than that in GBS-negative group, with no significant difference in the rate of multifetal pregnancy reductions. But further studies are needed to examine the underlying mechanism. Once they were confirmed GBS culture-positive, the pregnant women would be offered IAP based on the guideline developed by the CDC. Our data revealed that after IAP usage, the pregnancy outcomes were not significantly influenced by GBS infection. And the hospital length of stay of pregnant women was not correlated to GBS status, suggesting IAP is effective in preventing severe GBS invasive diseases and reducing the disease burden to GBS-positive pregnant women and the public health system. An earlier study focused on pregnant women who were not on IAP intervention at the third trimester showed that GBS infection was significantly associated with premature delivery and premature rupture of membranes [27]. And the finding that GBS colonization rate was significantly reduced in penicillin G-treated group than that in untreated-group also supports our conclusion [35]. As to neonatal outcomes, we did not directly investigate the correlation between GBS carriage and invasive neonatal GBS diseases due to the lack of the relevant data, being one of the main flaws of this study. However, we found the fatality rate, gender, weight, height, APGAR scores and the rates of nuchal cord or cord torsion were not affected by maternal GBS colonization. While the incidence of fetal distress was significantly declined by maternal GBS colonization, which might result from some cofounders since no significant association between GBS colonization and fetal distress was observed in logistic regression analysis. In contrast, elevated risk of fetal distress was significantly correlated with decreasing ages, increasing gestational weeks as well as the occurrence of nuchal cord or cord torsion and abnormal amniotic fluid. There was a report demonstrated that IAP resulted in a $50\%$ decrease in the occurrence of GBS-associated neonatal sepsis [36]. Another research clarified that standard IAP was a protective factor for GBS-EOD by logistic regression analysis [11]. Additionally, the incidence of EOD-GBS was reduced from 0.7 cases/1,000 live births in 1997 to 0.21–0.25 cases/1,000 live births in 2014 and 2015 [1]. These studies, in concert with our research, suggested that IAP usage was highly effective in preventing newborns from adverse outcomes including GBS invasive diseases. According to the guideline released by the CDC, the strategies for GBS screening and IAP intervention based on risk factors or antenatal universal GBS-culture were available [16]. Whereas the comparison of risk factors versus intrapartum culture screening indicated that the latter was more precise in detecting GBS and was optimal in guiding IAP therapy [18]. Herein we described the prevalence of GBS in pregnant women at late pregnancy was $13.47\%$ based on the culture method. Additionally, we found that women with diabetics were more susceptible to GBS colonization. Though IAP usage was effective in preventing pregnant women and newborns from adverse outcomes, there were still 90 culture-confirmed GBS-positive pregnant women in this study were not administrated with IAP in time or more than 4 h. It underlies the urgent need for rapid tests to detect GBS status. Laboratory detection of GBS takes a lot of time to culture GBS, ranging from 24 to 72 h. To solve this problem, we are now researching a new fluorescent immunochromatographic GBS antigen detection kit, aiming to shorten the testing time to 4–6 h. We hope this detection kit could be applied to clinical practice in future and help to improve the efficacy of IAP treatment. The high prevalence of GBS colonization and the high efficacy of IAP protection has underlined the need for GBS universal screening and IAP application for the target population in China. ## References 1. 1.Raabe VN, Shane AL. Group B Streptococcus (Streptococcus agalactiae).Microbiol Spectr2019, 7(2). 2. Tazi A, Morand PC, Reglier-Poupet H, Dmytruk N, Billoet A, Antona D, Trieu-Cuot P, Poyart C. **Invasive group B streptococcal infections in adults, France (2007–2010)**. *Clin Microbiol Infect* (2011.0) **17** 1587-9. DOI: 10.1111/j.1469-0691.2011.03628.x 3. Chaiwarith R, Jullaket W, Bunchoo M, Nuntachit N, Sirisanthana T, Supparatpinyo K. **Streptococcus agalactiae in adults at Chiang Mai University Hospital: a retrospective study**. *BMC Infect Dis* (2011.0) **11** 149. DOI: 10.1186/1471-2334-11-149 4. Skoff TH, Farley MM, Petit S, Craig AS, Schaffner W, Gershman K, Harrison LH, Lynfield R, Mohle-Boetani J, Zansky S. **Increasing burden of invasive group B streptococcal disease in nonpregnant adults, 1990–2007**. *Clin Infect Dis* (2009.0) **49** 85-92. DOI: 10.1086/599369 5. Balachandran L, Jacob L, Al Awadhi R, Yahya LO, Catroon KM, Soundararajan LP, Wani S, Alabadla S, Hussein YA. **Urinary tract infection in pregnancy and its Effects on maternal and perinatal outcome: a retrospective study**. *Cureus* (2022.0) **14** e21500. PMID: 35223276 6. Huang J, Zheng L, Su Y, Wang F, Kong H, Chang Y, Xin H. **Effects of group B streptococcus infection on vaginal micro-ecology and pregnancy outcomes of pregnant women in late pregnancy**. *Eur J Obstet Gynecol Reprod Biol* (2021.0) **267** 274-9. DOI: 10.1016/j.ejogrb.2021.11.419 7. Stapleton RD, Kahn JM, Evans LE, Critchlow CW, Gardella CM. **Risk factors for group B streptococcal genitourinary tract colonization in pregnant women**. *Obstet Gynecol* (2005.0) **106** 1246-52. DOI: 10.1097/01.AOG.0000187893.52488.4b 8. Hall J, Adams NH, Bartlett L, Seale AC, Lamagni T, Bianchi-Jassir F, Lawn JE, Baker CJ, Cutland C, Heath PT. **Maternal Disease with Group B Streptococcus and serotype distribution Worldwide: systematic review and Meta-analyses**. *Clin Infect Dis* (2017.0) **65** 112-S124. DOI: 10.1093/cid/cix660 9. Madrid L, Seale AC, Kohli-Lynch M, Edmond KM, Lawn JE, Heath PT, Madhi SA, Baker CJ, Bartlett L, Cutland C. **Infant Group B Streptococcal Disease Incidence and Serotypes Worldwide: systematic review and Meta-analyses**. *Clin Infect Dis* (2017.0) **65** 160-S172. DOI: 10.1093/cid/cix656 10. 10.Schrag SJ, Farley MM, Petit S, Reingold A, Weston EJ, Pondo T, Hudson Jain J, Lynfield R. Epidemiology of Invasive Early-Onset Neonatal Sepsis, 2005 to 2014.Pediatrics2016, 138(6). 11. Zhu Y, Huang J, Lin XZ, Chen C. **Group B Streptococcus colonization in late pregnancy and invasive infection in neonates in China: a Population-Based 3-Year study**. *Neonatology* (2019.0) **115** 301-9. DOI: 10.1159/000494133 12. Berardi A, Rossi C, Lugli L, Creti R, Bacchi Reggiani ML, Lanari M, Memo L, Pedna MF, Venturelli C, Perrone E. **Group B streptococcus late-onset disease: 2003–2010**. *Pediatrics* (2013.0) **131** e361-368. DOI: 10.1542/peds.2012-1231 13. 13.Creti R, Imperi M, Berardi A, Lindh E, Alfarone G, Pataracchia M, Recchia S. The Italian Network On N, Infant Gbs I: Invasive Group B Streptococcal Disease in Neonates and Infants, Italy, Years 2015–2019.Microorganisms2021, 9(12). 14. Berardi A, Rossi C, Creti R, China M, Gherardi G, Venturelli C, Rumpianesi F, Ferrari F. **Group B streptococcal colonization in 160 mother-baby pairs: a prospective cohort study**. *J Pediatr* (2013.0) **163** 1099-1104e1091. DOI: 10.1016/j.jpeds.2013.05.064 15. Russell NJ, Seale AC, O’Driscoll M, O’Sullivan C, Bianchi-Jassir F, Gonzalez-Guarin J, Lawn JE, Baker CJ, Bartlett L, Cutland C. **Maternal colonization with Group B Streptococcus and serotype distribution Worldwide: systematic review and Meta-analyses**. *Clin Infect Dis* (2017.0) **65** 100-S111. DOI: 10.1093/cid/cix658 16. Verani JR, McGee L, Schrag SJ. **Division of Bacterial Diseases NCfI, respiratory Diseases CfDC, Prevention: Prevention of perinatal group B streptococcal disease–revised guidelines from CDC, 2010**. *MMWR Recomm Rep* (2010.0) **59** 1-36. PMID: 21088663 17. Li S, Huang J, Chen Z, Guo D, Yao Z, Ye X. **Antibiotic Prevention for maternal Group B Streptococcal colonization on neonatal GBS-Related adverse outcomes: a Meta-analysis**. *Front Microbiol* (2017.0) **8** 374. PMID: 28367139 18. Johansen NR, Kjaerbye-Thygesen A, Jonsson S, Westh H, Nilas L, Rorbye C. **Prevalence and treatment of group B streptococcus colonization based on risk factors versus intrapartum culture screening**. *Eur J Obstet Gynecol Reprod Biol* (2019.0) **240** 178-81. DOI: 10.1016/j.ejogrb.2019.06.037 19. Chen XLWJ, Chen L, Huang QL, Ye HM. **Application of three screening approaches for group B streptococcus in the third trimester of pregnancy**. *Chin J Clin Lab Sci* (2019.0) **37** 290-2 20. Ji W, Liu H, Madhi SA, Cunnington M, Zhang Z, Dangor Z, Zhou H, Mu X, Jin Z, Wang A. **Clinical and molecular epidemiology of Invasive Group B Streptococcus Disease among Infants, China**. *Emerg Infect Dis* (2019.0) **25** 2021-30. DOI: 10.3201/eid2511.181647 21. Deutscher M, Lewis M, Zell ER, Taylor TH, Van Beneden C, Schrag S. **Active bacterial core surveillance T: incidence and severity of invasive Streptococcus pneumoniae, group a Streptococcus, and group B Streptococcus infections among pregnant and postpartum women**. *Clin Infect Dis* (2011.0) **53** 114-23. DOI: 10.1093/cid/cir325 22. Chen Z, Wen G, Cao X, Li S, Wang X, Yao Z, Wu C, Ye X. **Group B streptococcus colonisation and associated risk factors among pregnant women: a hospital-based study and implications for primary care**. *Int J Clin Pract* (2019.0) **73** e13276. DOI: 10.1111/ijcp.13276 23. Najmi N, Jehan I, Sikandar R, Zuberi NF. **Maternal genital tract colonisation by group-B streptococcus: a hospital based study**. *J Pak Med Assoc* (2013.0) **63** 1103-7. PMID: 24601185 24. Yucesoy G, Caliskan E, Karadenizli A, Corakci A, Yucesoy I, Huseyinoglu N, Babaoglu K. **Maternal colonisation with group B streptococcus and effectiveness of a culture-based protocol to prevent early-onset neonatal sepsis**. *Int J Clin Pract* (2004.0) **58** 735-9. DOI: 10.1111/j.1368-5031.2004.00025.x 25. do Nascimento CS, Dos Santos NFB, Ferreira RCC, Taddei CR. **Streptococcus agalactiae in pregnant women in Brazil: prevalence, serotypes, and antibiotic resistance**. *Braz J Microbiol* (2019.0) **50** 943-52. DOI: 10.1007/s42770-019-00129-8 26. Valkenburg-van den Berg AW, Sprij AJ, Oostvogel PM, Mutsaers JA, Renes WB, Rosendaal FR, Joep Dorr P. **Prevalence of colonisation with group B streptococci in pregnant women of a multi-ethnic population in the Netherlands**. *Eur J Obstet Gynecol Reprod Biol* (2006.0) **124** 178-83. DOI: 10.1016/j.ejogrb.2005.06.007 27. Dechen TC, Sumit K, Ranabir P. **Correlates of Vaginal colonization with Group B Streptococci among pregnant women**. *J Glob Infect Dis* (2010.0) **2** 236-41. DOI: 10.4103/0974-777X.68536 28. Anthony BF, Okada DM, Hobel CJ. **Epidemiology of group B Streptococcus: longitudinal observations during pregnancy**. *J Infect Dis* (1978.0) **137** 524-30. DOI: 10.1093/infdis/137.5.524 29. Altoparlak U, Kadanali A, Kadanali S. **Genital flora in pregnancy and its association with group B streptococcal colonization**. *Int J Gynaecol Obstet* (2004.0) **87** 245-6. DOI: 10.1016/j.ijgo.2004.08.006 30. Brzychczy-Wloch M, Pabian W, Majewska E, Zuk MG, Kielbik J, Gosiewski T, Bulanda MG. **Dynamics of colonization with group B streptococci in relation to normal flora in women during subsequent trimesters of pregnancy**. *New Microbiol* (2014.0) **37** 307-19. PMID: 25180845 31. Jackson LA, Hilsdon R, Farley MM, Harrison LH, Reingold AL, Plikaytis BD, Wenger JD, Schuchat A. **Risk factors for group B streptococcal disease in adults**. *Ann Intern Med* (1995.0) **123** 415-20. DOI: 10.7326/0003-4819-123-6-199509150-00003 32. Kadanali A, Altoparlak U, Kadanali S. **Maternal carriage and neonatal colonisation of group B streptococcus in eastern Turkey: prevalence, risk factors and antimicrobial resistance**. *Int J Clin Pract* (2005.0) **59** 437-40. DOI: 10.1111/j.1368-5031.2005.00395.x 33. Yow MD, Leeds LJ, Thompson PK, Mason EO, Clark DJ, Beachler CW. **The natural history of group B streptococcal colonization in the pregnant woman and her offspring. I. colonization studies**. *Am J Obstet Gynecol* (1980.0) **137** 34-8. DOI: 10.1016/0002-9378(80)90382-8 34. Phares CR, Lynfield R, Farley MM, Mohle-Boetani J, Harrison LH, Petit S, Craig AS, Schaffner W, Zansky SM, Gershman K. **Epidemiology of invasive group B streptococcal disease in the United States, 1999–2005**. *JAMA* (2008.0) **299** 2056-65. DOI: 10.1001/jama.299.17.2056 35. Lewin EB, Amstey MS. **Natural history of group B streptococcus colonization and its therapy during pregnancy**. *Am J Obstet Gynecol* (1981.0) **139** 512-5. DOI: 10.1016/0002-9378(81)90509-3 36. Schrag SJ, Zywicki S, Farley MM, Reingold AL, Harrison LH, Lefkowitz LB, Hadler JL, Danila R, Cieslak PR, Schuchat A. **Group B streptococcal disease in the era of intrapartum antibiotic prophylaxis**. *N Engl J Med* (2000.0) **342** 15-20. DOI: 10.1056/NEJM200001063420103
--- title: 'Adult asthma associated with roadway density and housing in rural Appalachia: the Mountain Air Project (MAP)' authors: - W. Jay Christian - John Flunker - Beverly May - Susan Westneat - Wayne T. Sanderson - Nancy Schoenberg - Steven R. Browning journal: Environmental Health year: 2023 pmcid: PMC10041800 doi: 10.1186/s12940-023-00984-x license: CC BY 4.0 --- # Adult asthma associated with roadway density and housing in rural Appalachia: the Mountain Air Project (MAP) ## Abstract ### Background Appalachian *Kentucky is* a rural area with a high prevalence of asthma among adults. The relative contribution of environmental exposures in the etiology of adult asthma in these populations has been understudied. ### Objective This manuscript describes the aims, study design, methods, and characteristics of participants for the Mountain Air Project (MAP), and focuses on associations between small area environmental exposures, including roadways and mining operations, and lifetime and current asthma in adults. ### Methods A cohort of residents, aged 21 and older, in two Kentucky counties, was enrolled in a community-based, cross-sectional study. Stratified cluster sampling was used to select small geographic areas denoted as 14-digit USGS hydrologic units (HUCs). Households were enumerated within selected HUCs. Community health workers collected in-person interviews. The proximity of nearby active and inactive coal mining operations, density of oil and gas operations, and density of roadways were characterized for all HUCs. Poisson regression analyses were used to estimate adjusted prevalence ratios. ### Results From 1,459 eligible households contacted, 1,190 individuals were recruited, and 972 persons completed the interviews. The prevalence of lifetime asthma was $22.8\%$; current asthma was $16.3\%$. Adjusting for covariates, roadway density was positively associated with current asthma in the second (aPR = 1.61; $95\%$ CI 1.04–2.48) and third tertiles (aPR = 2.00; $95\%$ CI 1.32–3.03). Increased risk of current asthma was associated with residence in public, multi-unit housing (aPR = 2.01; $95\%$ CI 1.27–3.18) compared to a residence in a single-family home. There were no notable associations between proximity to coal mining and oil and gas operations and asthma prevalence. ### Conclusions This study suggests that residents in rural areas with higher roadway density and those residing in public housing units may be at increased risk for current asthma after accounting for other known risk factors. Confirming the role of traffic-related particulates in producing high asthma risk among adults in this study contributes to the understanding of the multiple environmental exposures that influence respiratory health in the Appalachia region. ## Introduction There are many potential environmental exposures in rural mining communities that raise concerns for respiratory health outcomes. Contributions from both indoor and outdoor environments are potentially significant contributors to respiratory disease. Surface coal-mining operations, including mountaintop removal (MTR) mines, emit atmospheric particulate matter (PM) to surrounding areas. The air quality impacts from these forms of mining have been a community concern since soil, mineral dust, emissions from diesel equipment and blasting, and wind-driven re-suspension of PM may all contribute to ambient PM [1, 2]. Traffic-related pollutants, which differ in composition from mining-related pollutants, may play a role in respiratory disease, but these pollutants have been understudied in rural areas [3]. Further, PM exposures may be influenced by the characteristics of the homes, personal activity patterns, and local topography [4]. Finally, cigarette smoking and the resulting environmental tobacco smoking exposures (ETS) in the homes of rural Appalachian residents have been well-documented [5]. Understanding the impacts of these multiple exposures on asthma and other respiratory outcomes is critical to designing appropriate interventions. Epidemiologic studies have demonstrated associations between PM and respiratory symptoms and new-onset asthma, asthma hospitalizations, emergency department visits, and deaths [6–9]. Penttinen et al. [ 2006] reported fine particles are associated with respiratory morbidity in adults with asthma, with the strongest associations between ultrafine and fine particles and decreased lung function [10]. The sources of ambient PM in mining communities could potentially include coal haul roads, blasting operations, meteorological conditions (carried by the wind) or some combination of multiple exposure sources. Additionally, explosives used in the mining process to remove coal overburden contain ammonium nitrate and diesel fuel and release CO2, CO, NO, SO2, and ammonia during combustion [2, 11]. Studies examining traffic-related air pollutants in adults have shown positive associations with both asthma prevalence and current asthma [12–14]. Residents of Central Appalachia experience the nation’s highest rates of serious respiratory disease [15, 16]. In addition to high rates of adult smoking [17], regional epidemiologic studies suggest associations between adverse respiratory health outcomes and residence in coal mining areas [18–20]. In an ecological analysis, Hendryx and Ahern found that counties in West Virginia with annual coal production of 4 million tons or greater had higher adjusted odds of chronic obstructive pulmonary disease (COPD), lung disease, and other health outcomes in comparison to persons residing in counties with lower coal production levels [19]. Cross-sectional data from 892 adults indicated current asthma prevalence in a coal mining community in Appalachian Kentucky was $18.2\%$; COPD was estimated at $25.9\%$ [18]. The adjusted prevalence ratio for those who self-reported current asthma was 1.68 ($95\%$ CI: 1.11–2.54) and for COPD was 2.47 ($95\%$ CI: 1.62–3.74) in communities with mountaintop removal coal mining in comparison to referent communities without mining [18]. These rates exceed national prevalence estimates for the general population by over two and four-fold, respectively. The body of research by Hendryx et al. has attracted substantial interest from researchers as well as residents of mining communities. In addition to reported associations between communities with potential mining-related exposures, especially surface and mountaintop mining operations, he and allied researchers have found associations between residence in “primary mining communities” and cardiovascular disease [21, 22], birth defects [23], cancers [24, 25], and general health status [19, 20]. These results call into question the specificity of these associations resulting from discrete environmental pollutants, which may be associated with mining. The separation of the confounding and mediating effects of socioeconomic status has been debated and proposed as alternative explanations for these findings [26]. Most of these studies have been ecologic in design or cross-sectional investigations premised on convenience samples, with ecologic measures of exposure. Studies have been limited by an absence of individual-level, quantitative exposure data of dust levels and other environmental exposures (e.g., motor vehicles, ambient pollution) or small area exposure estimates. In addition, these studies have generally lacked adequate control for critical behavioral risk factors (e.g., smoking, diet, physical activity) and social determinants (e.g., lower socioeconomic status, occupation, age, gender) of respiratory diseases like asthma. While potential environmental exposures from mining operations, roadways, and other emissions sources have attracted concern, lifestyle factors including the high prevalence of cigarette smoking, exposure to environmental tobacco smoke (ETS), housing conditions including dampness and mold exposures, lack of physical activity, and obesity rates are also likely to impact adult asthma prevalence. For example, roughly one-third of adults in this region are current smokers, and an estimated $65\%$ of adults have ever lived with someone who smokes [17, 27]. Few previous studies have closely considered the factors related to housing condition and housing type (single-family, mobile home, apartment), which may influence adult asthma symptoms and episodes [28]. The Mountain Air Project (MAP) is a community-based participatory research project which was designed to investigate the extent, nature, and source of respiratory health inequities in two counties in southeastern Kentucky through a community-engaged assessment of environmental and individual-level exposures. In addition, the study was undertaken to develop an environmental health action strategy to address community-identified exposures of concern for respiratory diseases, including asthma and COPD, and implement and evaluate a community-based intervention [29]. A study goal was to enhance the collection of environmental exposure data for small geographic units and obtain individual-level data on risk factors for several respiratory disease outcomes. The specific aims of this paper are to describe the sampling and design characteristics of the Mountain Air Project (MAP) and the baseline characteristics of the cohort. Further, we present the results of the associations of several small area metrics of exposures to traffic, mining, and oil and gas operations with lifetime and current asthma while adjusting for other asthma risk factors. While the MAP study focuses on several respiratory conditions, this paper presents results only for lifetime and current asthma. ## Methods A cross-sectional epidemiologic survey was designed and implemented as a component of the Mountain Air Project (MAP) to meet the multiple objectives described above. For the survey, sampling and interviewing of residents would occur on-site in the community with face-to-face interviews. The survey was developed to include a broad range of community-identified exposures of concern, including those from mining, traffic and roadways, and oil and gas operations. In an effort to address some of the limitations of previous ecologic studies [26], the survey also included home environmental exposures and lifestyle risk factors by obtaining individual-level exposure data related to asthma prevalence. ## Study area and population Our study focused on two Appalachian Kentucky counties (Letcher and Harlan) with long histories of intensive resource extraction and economic disadvantage. This area has the nation’s highest burden of respiratory disease [30], as well as environmental justice concerns stemming from airborne contaminant exposures. The study area was selected based on the presence of extractive industries, marked disparities in respiratory disease, community concerns regarding exposure from coal mining and other extractive industries, and existing infrastructure for mobilizing the project from previous community-based health research [31–33]. In addition, community stakeholders suggested using "hollows" (or hydrological units) as the most relevant geographic unit for defining “neighborhoods” for the cross-sectional study and for ambient air sampling. This approach was discussed in a previous publication [29]. The study was approved by the University of Kentucky Medical Institutional Review Board, and written informed consent was obtained from all participants. ## Eligibility Inclusion criteria included non-institutionalized, English-speaking adults age 21 or older, of any race or ethnicity, residing within an eligible household in either Letcher or Harlan counties. Eligible households consisted of single-family residences, apartment housing, or mobile homes. One adult participant was recruited per household. If an adult in the household reported having asthma, COPD, coal worker’s pneumoconiosis (black lung), or other respiratory health condition, he or she was encouraged to serve as the participant for that household. If the person with respiratory disease declined to participate and another adult household member without a respiratory condition was eligible, then that person was recruited for the study. ## Geographic site selection and characterization We used a stratified cluster sampling technique to select small geographic areas in Harlan and Letcher counties to be the sampling units. We defined candidate ‘hollows’ using GIS map layers representing the boundaries of 14-digit hydrologic unit codes (HUCs). These are the smallest hydrologic units available and often coincide with residential development patterns in the study region since streets and homes are typically ordered in a linear fashion along narrow valleys. We obtained the GIS data for these HUCs from the Kentucky Geological Survey. We imported the HUC boundary polygons into ArcGIS 10.3 [34] and characterized the HUCs by their relationship to several other layers that indicate potential sources of airborne particulates and pollution. These sources of pollutants included the surface boundaries for (a) active underground and surface coal mining sites, and (b) inactive underground and surface coal mining sites, from the Kentucky Mine Mapping Information System; (c) all streets and roads, and (d) roads officially designated as ‘coal haul routes’, from the Kentucky Transportation Cabinet, and (e) the point locations of active oil and gas wells within the HUC, from the Kentucky Geologic Survey. All data sets were the most recent available as of July 2015, just before participant recruitment. Figure 1 displays the distribution of these small area metrics in the study area. From these map layers, we calculated the following metrics to characterize each HUC in ArcGIS:1) abandoned mining (surface or underground), as a percent of HUC total surface area, 2) active mining (surface or underground), as a percent of HUC total surface area, 3) road miles per square mile, 4) coal haul route miles per square mile, and 5) oil and gas wells per square mile. We then summed these ordinal values to create an index of overall environmental risk to respiratory health. The index was divided into tertiles of high, medium, and low presumptive exposure levels to the five sources of airborne particulates above. Ten hollows per each index level were then randomly selected for a total of 30 HUCs included in the study. Fig. 1Geographic Distribution of Study HUCs (“Hollows”) and Small Area Exposure Metrics (14 digit HUC) for Harlan and Letcher Counties, Kentucky Homes in each HUC were enumerated on the ground by field staff. Within each selected HUC, homes were sampled by dividing the total number of homes by an appropriate factor to yield at least ten homes per HUC for the study. Eligible homes were selected using a random starting point and a systematic sample of every nth home. Each selected home had its GPS coordinates recorded using a QStarz GPS data logger. The GPS data were linked with the household survey data and the environmental sampling data so that we were able to develop maps and integrate other datasets with the final epidemiologic files. HUCs with insufficient residences to yield at least ten eligible households were eliminated from the sample, and randomly selected replacement HUCs were provided to the field staff. Ten replacement HUCs were identified through a two-step process that included random selection followed by a rooftop survey using Google satellite imagery. Therefore, although 40 HUCs (hollows) were selected in the sampling process, 30 HUCs with sufficient populations were used in data collection, which was halted when the sample approached 1,000 persons. ## Enrollment of study subjects and survey data collection Community health workers (CHWs), most with previous experience in community-based research and familiarity with the local community, were hired for recruiting participants and survey data collection. Details of the field operations for the MAP study are described elsewhere [29]. One CHW was responsible for recruiting, obtaining informed consent, and using GPS to locate each home. Household contact forms were used to collect demographic information and respiratory health status for each member of the household. Data from these forms were entered into a database for tracking response and participation rates. Following the initial agreement to participate and the setting of appropriate times to interview, four CHWs were assigned to participating households to administer questionnaires and collect spirometry. Community Health Workers (CHWs) were trained in collection of lung function data using a portable hand-held ndd Easy on-PC Spirometry System (ndd Medical Technologies Inc. Andover, MA, USA). After the CHW determined the procedure was not contraindicated by a stroke or myocardial infarction in the past 30 days, the technique was explained and the participant seated. A minimum of three trials was attempted by the participant, with coaching by the CHW. The team pulmonologist reviewed each spirogram for reliability and repeatability using American Thoracic Society/European Respiratory Society (ATS/ERS) grading criteria. Further details regarding the procedures used in the collection of the spirometry are provided in May et al. [ 29]. CHWs used REDCap survey software on Ipads for all data collection. REDCap (Research Electronic Data Capture) is a secure, web-based software platform designed to support data capture for research studies, providing 1) an intuitive interface for validated data capture; 2) audit trails for tracking data manipulation and export procedures; 3) automated export procedures for seamless data downloads to statistical packages, and 4) procedures for data integration and interoperability with external sources [35]. The REDCap database, stored and backed up on servers, was exported to SAS datasets. Participants received $40 for survey completion. The survey, which took approximately 40 min to administer, focused on established and potential risk factors for respiratory health outcomes and obtained data on current and past symptoms of respiratory health over the past 2 and 12 months before the survey. Questions were drawn primarily from established questionnaires, including the Multi-Ethic Study of Atherosclerosis (MESA) spirometry questionnaire, which provides questions similar to the National Health Interview Survey (NHIS) questions for adult asthma, and the Seattle Healthy Homes I baseline questionnaire [36–40]. Family history of respiratory disease, allergies, chronic conditions, and eczema was obtained by self-report on the questionnaire. In addition, we assessed self-reported chemical and biological triggers, focusing on environmental tobacco smoke (ETS), pesticides, VOCs, dust mites, molds, rodent and cockroach feces, and animal dander; home heating (wood, coal, gas, space heaters), home cooking (electric, wood, gas, oil), indoor smoking, pets, molds, and dampness were also assessed. Detailed information was obtained on demographic and health behavior and lifestyle factors (education, marital status, employment status, job titles, work in mining and related occupations, dietary intake, alcohol consumption, and tobacco use). At the time of the interview, pulmonary function tests were also administered to all participants who did not report a stroke or myocardial infarction in the past 30 days. At a later date, a convenience subsample of 70 participants received indoor air quality and exposure assessment to quantify fine particulates and record in-home and outdoor exposure sources. The details of pulmonary function testing and in-home air quality assessment are described in more detail elsewhere [29, 41]. ## Asthma Outcomes We examined two asthma outcomes: lifetime (“ever”) and current asthma. Lifetime asthma was coded as those responding affirmatively to the question, “Have you ever had asthma?”. Among those who reported lifetime asthma, current asthma was coded as ‘yes’ if participants reported in the past 12 months: 1) at least one asthma attack, or 2) a routine medical visit for asthma, or 3) were unable to work due to asthma, or 4) talked with a doctor or other health professional about asthma, or 5) took asthma medication, or 6) experienced any symptoms of asthma. Lifetime and current asthma variables in the analysis are based on the responses to the self-reported questions and subject to the known limitations of data without subsequent medical validation. ## Covariates Demographic variables in the descriptive analysis included age as a three-level variable (21–34, 35–64, and 65 years and older); marital status as married/partnered or not; level of education as high school graduate (or GED) or higher; annual household income below $25,000 annually, $25,000–50,000, or greater than $50,000. Health risk behaviors examined included exposure to secondary tobacco smoke before age 16 and in adulthood; history of having smoked greater than 100 cigarettes in lifetime (never vs. ever smoker); current smoker; former smoker greater than ten years. Body mass index (BMI) was calculated as weight in pounds/height in inches2 multiplied by 703 and categorized as underweight < 18.5, normal 18.5–24.9, overweight 25–29.9, and obese 30 or greater. Housing type was categorized as a single-family home, mobile home, or multi-unit housing. Self-reported seasonal allergies were recorded as an indicator of atopy. ## Statistical Analysis Frequency distributions of the demographic characteristics of the sample of respondents were calculated using Stata 12 [42]. There were relatively few missing values, so the final analysis simply omitted individuals with missing values for key variables. An exception was household income, where $24.4\%$ of values were missing. These missing values were retained as a separate category of household income. The primary outcome variables, lifetime and current asthma were highly prevalent (> $10\%$) in our sample. Consequently, we used Poisson models with robust variance estimators to calculate prevalence ratios (PRs) adjusted for multiple covariates [43]. Models were adjusted for individual-level covariates (age, gender, seasonal allergies, BMI, educational attainment, type of dwelling, living with a current smoker, smoking status) that alone comprised base models, and each included a single HUC-level variable describing the intensity of a potential source of airborne particulates in the environment. Other than the base models, this resulted in ten regression analyses, one for each of the five potential sources of airborne particulates—roadways (miles per square mile), coal haul routes (miles per square mile), oil/gas wells (per square mile), abandoned mines (percent of HUC area), and active mines (percent of HUC area)—for both current and lifetime asthma. The base models were developed through consideration of significant bivariate associations observed in this study and others reported in the literature. Main effects models were fit and checked for multicollinearity and violations of other model assumptions. ## Response rates From November 2015 to July 2017, a total of 4,291 dwellings were enumerated within 40 HUCs in the study area. From 1,459 eligible households contacted, 1,190 individual participants ($82\%$) were recruited into the study. Of those, 218 participants did not complete the survey due to refusal, loss to follow up, or death. Of the 972 individuals recruited who completed the survey, 872 provided valid spirograms (data not presented in this paper). ## Demographic characteristics of the sample The sample (Table 1) was primarily composed of females ($59\%$), participants aged 35 to 64 years ($61\%$), and those with a high school education or above ($74\%$). Participant age ranged from 21 to 96 years with a median age of 55 years. Annual household income was reported by $76\%$ of participants. Of these, $46\%$ reported income less than $25,000 for the household. While most persons resided in single-family homes ($65\%$), nearly a third lived in mobile homes, and $5\%$ lived in multi-unit housing. Only one-fifth of our sample were employed full time, with $23\%$ reporting being retired and $19\%$ disabled. Table 1Characteristics study participants and distributions of cases and prevalence ratios (PR) for ever asthma and current asthma in the Mountain Air Project study (2016–2017)TotalEver diagnosed w/asthma ($$n = 963$$)*Current asthma* ($$n = 963$$)n%cases%PR$95\%$ CIcases%PR$95\%$ CIAll participants97210021822.6…15716.3…County of Residence ($$n = 972$$) Harlan58960.612020.6Ref––-8113.9Ref––- Letcher38339.49825.61.240.98–1.577619.91.431.07–1.90Season of Interview ($$n = 972$$) Winter19219.84925.81.320.90–1.923518.41.661.00–2.74 Spring39640.79223.61.200.86–1.697118.21.641.04–2.58 Summer19520.14020.61.050.71–1.573015.51.390.83–2.34 Fall18919.43719.6Ref––-2111.1Ref––-Age group ($$n = 972$$) 18–3414014.34532.42.691.77–4.112719.42.061.22–3.47 35–6459460.914524.52.041.40–2.9710818.31.941.26–2.99 65 + 23824.82812.0Ref––-229.4Ref––-Gender ($$n = 972$$) Female57158.714225.11.321.03–1.6910518.61.421.05–1.93 Male40141.37619.1Ref––-5213.1Ref––-Educational attainment ($$n = 971$$) < High School25426.15823.21.070.82–1.404819.21.330.96–1.83 High school or some college59961.612921.7Ref––-8614.5Ref––- College11812.13025.41.170.83–1.652218.61.290.84–1.97 Missing10.11100.01100.0Annual household income ($$n = 735$$) < $25 K45146.411124.91.210.85–1.738218.41.220.79–1.88 $25-$50 K13814.23223.51.150.74–1.782014.70.980.56–1.71 $50 K + 14615.03020.5Ref––-2215.1Ref––- Missing23724.44519.10.930.62–1.413314.00.930.57–1.53Type of dwelling ($$n = 972$$) Single family house63465.213421.3Ref––-9615.2Ref––- Mobile home28829.66422.51.060.81–1.384315.10.990.71–1.38 Multi-unit housing505.12040.81.921.33–2.781836.72.411.60–3.64Body mass index (BMI) (kg/m2) ($$n = 925$$) Underweight (< 18.5)192.0526.31.350.61–2.96421.01.530.61–3.84 Normal (18.5–24.9)21722.34520.91.070.75–1.523214.91.080.70–1.67 Overweight (25.0–29.9)27628.45419.6Ref––-3813.8Ref––- Obese (30.0 +)41342.510225.11.280.96–1.727518.41.340.93–1.92 Missing474.81226.1817.4Season allergies ($$n = 969$$) Yes54656.215528.71.911.47–2.4911320.92.001.44–2.76 No42343.56315.0Ref––-4410.5Ref––- Missing30.300.000.0Smoking status ($$n = 969$$) Current32032.97323.01.020.78–1.345517.31.120.80–1.56 Former23023.74921.90.980.72–1.323615.81.020.70–1.49 Never41943.19122.5Ref––-6415.5Ref––- Missing30.3266.7266.7Lives with current smoker ($$n = 972$$) Yes62664.414523.31.090.85–1.4010617.11.140.84–1.56 No34635.67321.4Ref––-5114.9Ref––-Current employment ($$n = 972$$) Homemaker19420.05126.62.051.35–3.113920.31.911.18–3.08 Full-time21522.15123.81.841.21–2.803415.91.490.91–2.45 Part-time / Full-time student / Unemployed15916.43321.01.621.02–2.572012.71.200.68–2.10 Retired21922.52813.0Ref––-2310.7Ref––- Disabled18519.05529.92.311.53–3.484122.32.091.31–3.35Ever had dusty job, including mining ($$n = 952$$) Yes35036.06418.4Ref––-4613.3Ref––- No62264.015425.01.361.04–1.7611118.01.360.99–1.87PR prevalence ratio Cigarette smoking remains highly prevalent in this region, with $33\%$ of the sample reporting current smoking while $24\%$ were classified as former smokers; $65\%$ of the sample reported living in the home of a current smoker as an adult. Reflecting regional health characteristics, $29\%$ of the sample were overweight, and $42\%$ were obese. More than one-third had been employed in a dusty job, including mining, during their working career. Among men, however, $80\%$ had been employed in mining, compared to only $5\%$ among women. The overall rate of persons reporting they had ever been diagnosed with asthma was $22.8\%$, and the prevalence of current asthma was $16.3\%$. Lifetime and current asthma were most prevalent in women and among those in the youngest age group (21–34 years). Both lifetime asthma (PR = 1.92; $95\%$ CI: 1.33–2.78) and current asthma (PR = 2.41; $95\%$ CI: 1.60–3.64) were roughly two-fold higher in residents of multi-unit housing compared to those in single-family homes. Those reporting seasonal allergies also had a higher prevalence of both lifetime (PR = 1.91) and current asthma (PR = 2.00) in comparison to those not reporting allergies. Homemakers, as well as those who were disabled, had a higher prevalence of lifetime asthma and current asthma compared to the retired, who had the lowest prevalence. Those who had ever worked in dusty occupations reported lower rates of lifetime and current asthma. An unadjusted analysis of the associations between our HUC-based GIS metrics and lifetime and current asthma is displayed in Table 2 for the 963 participants with complete data. There was a significant positive dose–response relationship between roadway density (road miles/square mile) and lifetime and current asthma prevalence. For current asthma, the prevalence ratio is more than two-fold higher (PR = 2.14; $95\%$ CI: 1.43–3.22) at the highest tertile of roadway density. The rates of both lifetime and current asthma appeared to increase with greater density of oil and gas wells in the hollow, but the associations were not significant. The prevalence rates of lifetime and current asthma were lowest at the highest tertiles of the intensity of active mining (measured as a percent of the hollow area) in this unadjusted analysis. Table 2Prevalence of ever asthma and current asthma by geographic exposure metrics among study participants in the Mountain Air Project study (2016–2017)Ever diagnosed w/asthma ($$n = 963$$)Current asthma($$n = 963$$)n%PR$95\%$ CIn%PR$95\%$ CIRoad miles/sq. mi Tertile 1 (0.19–1.89)4916.21.0–-299.61.0–- Tertile 2 (2.18–2.76)7523.01.421.03–1.965918.11.881.24–2.86 Tertile 3 (2.96–6.00)9428.11.731.27–2.356920.62.141.43–3.22Coal haul miles/sq. mi Zero6621.91.0–-5016.61.0–- Below median (0.00–0.62)10822.41.020.78–1.348116.81.010.73–1.40 Above median (0.72–1.27)4424.41.110.80–1.562614.40.870.56–1.35Oil/gas wells/sq. mi Zero5020.81.0–-3213.31.0–- Below median (0.34–3.51)13222.51.090.81–1.459616.71.260.87–1.82 Above median (4.38–8.44)3626.51.280.88–1.852719.91.500.94–2.39Abandoned mining, %HUC area Tertile 1 (0.00–8.04)6521.21.0–-4514.71.0–- Tertile 2 (9.72–44.84)7722.11.040.78–1.405315.21.040.72–1.49 Tertile 3 (46.40–196.45)7624.61.160.87–1.555919.11.300.91–1.85Active mining, %HUC area Zero11325.31.0–-8418.81.0–- Below median (0.00–8.02)7023.00.910.70–1.185016.40.870.63–1.20 Above median (9.68–48.31)3516.50.650.46–0.922310.90.580.37–0.89PR prevalence ratio Adjusted models—the base models—using the robust Poisson regressions are shown in Table 3. The adjusted prevalence ratios (aPR) indicate significantly higher lifetime and current asthma prevalence in the youngest age group (18–34 years) compared to those over age 65 and significantly higher prevalence of current asthma in women (aPR = 1.40, $$p \leq 0.04$$). Seasonal allergies and residing in a multi-unit housing complex (in comparison to single-family housing) remained significant risk factors for both lifetime and current asthma in the base models. Current asthma was $50\%$ (aPR = 1.50; $95\%$ CI: 1.09–2.07) more prevalent among those with less than a high school education in the model. Table 3Base model for lifetime and current asthmaLifetime asthma ($$n = 910$$)*Current asthma* ($$n = 910$$)aPRP-value$95\%$ CIaPRP-value$95\%$ CIAge group 18–342.83 < 0.0011.77–4.512.040.011.15–3.62 35–642.10 < 0.0011.42–3.121.900.011.21–2.97 65 + Ref––-Ref––-Gender Female1.210.140.94–1.561.400.041.02–1.93 MaleRef––-Ref––-Seasonal allergies Yes1.86 < 0.0011.42–2.431.91 < 0.0011.37–2.67 NoRef––-Ref––-Body mass index (BMI) (kg/m2) Underweight (< 18.5)1.250.550.60–2.571.370.470.59–3.22 Normal (18.5–24.9)1.040.830.73–1.481.000.990.64–1.56 Overweight (25.0–29.9)Ref––-Ref––- Obese (30.0 +)1.170.300.87–1.571.220.270.85–1.76Educational attainment < High School1.190.210.90–1.581.500.011.09–2.07 High school or some collegeRef––-Ref––- College1.340.110.94–1.931.470.100.93–2.34Type of dwelling Single family houseRef––-Ref––- Mobile home1.020.870.77–1.360.980.890.69–1.38 Multi-unit housing1.710.011.13–2.572.010.0031.27–3.18Lives with current smoker Yes1.230.220.89–1.701.200.390.79–1.83 NoRef––-Ref––-*Smoking status* CurrentRef––-Ref––- Former1.260.180.90–1.771.160.460.78–1.73 Never1.280.160.91–1.811.120.600.73–1.72PR prevalence ratio Table 4 provides prevalence ratio estimates for each of the environmental exposure metrics and lifetime and current asthma adjusted for all variables in the base models (i.e., Table 3). Roadway density (road miles /square mile) was positively associated with current asthma when comparing the second tertile (aPR = 1.61; $95\%$ CI:1.04–2.48) and third tertile (aPR = 2.00; $95\%$ CI:1.32–3.03) to the lowest first tertile of road miles/square mile. For lifetime asthma, roadway density was positively associated with asthma prevalence, with a significant $56\%$ increased prevalence at the third tertile. For current asthma, there was some elevation in prevalence (aPR = 1.70) among those living in areas with oil and gas well density above the median. Adjusted prevalence ratios for lifetime and current asthma were lowest at the above-median levels of active mining density within the HUC, with statistically significant protective ratios for lifetime (aPR = 0.64) and current (aPR = 0.54) asthma. Table 4Adjusted prevalence ratios for lifetime and current asthma by environmental exposureLifetime asthma ($$n = 910$$)*Current asthma* ($$n = 910$$)aPRP-value$95\%$ CIaPRP-value$95\%$ CIRoad miles/sq. mi Tertile 1 (0.19–1.89)Ref––-Ref––- Tertile 2 (2.18–2.76)1.220.240.88–1.711.610.031.04–2.48 Tertile 3 (2.96–6.00)1.560.011.14–2.142.000.0011.32–3.03Coal haul miles/sq. mi ZeroRef––-Ref––- Below median (0.00–0.62)0.870.340.66–1.150.840.300.60–1.17 Above median (0.72–1.27)1.030.880.73–1.440.790.300.50–1.24Oil/gas wells/sq. mi ZeroRef––-Ref––- Below median (0.34–3.51)1.040.810.76–1.431.260.280.83–1.91 Above median (4.38–8.44)1.300.190.88–1.921.700.041.03–2.80Abandoned mining, %HUC area Tertile 1 (0.00–8.04)Ref––-Ref––- Tertile 2 (9.72–44.84)1.010.960.74–1.370.940.760.64–1.38 Tertile 3 (46.40–196.45)1.180.280.88–1.581.310.140.91–1.89Active mining, %HUC area ZeroRef––-Ref––- Below median (0.00–8.02)0.870.330.66–1.150.780.160.56–1.10 Above median (9.68–48.31)0.640.020.45–0.920.540.010.35–0.85aPRadjusted prevalence ratioPrevalence ratios adjusted for variables in base model, but not other environmental exposures ## Discussion The prevalence estimates of lifetime ($22.8\%$) and current asthma ($16.3\%$) in the MAP sample affirm the high prevalence documented in previous work among rural Appalachian populations [15]. Our prevalence estimates are subject to a small upward bias for asthma (see below) since our methods of enrolling persons intentionally favored the selection of persons with any respiratory health outcome (COPD, asthma, other lung disease) into the study. While this approach increased the likelihood of recruiting those with respiratory health conditions beyond a random sample of eligible adults in the household, the approach provided a more robust analysis of the associations of the environmental exposures with asthma. Current smoking, exposure to environmental tobacco smoke, and obesity rates are significantly higher in this sample than in the US population and may account, in part, for high asthma prevalence in this geographic area. The paper by Mabila et al. with the NHI survey affirms the strong association of respiratory disease, including asthma, in men working in the dusty trades industries [44]. In the MAP cohort, $69\%$ of men had worked in coal mining occupations, including above and underground, as well as MTR. Consequently, the ability to separate the occupational component from environmental exposure is challenging, especially in men. Only 15 women in our study reported working in mining or other dusty trades. The asthma prevalence in this study is higher in the younger age and in the lower exposure groups among men in the mining industry. Another bias not often mentioned is the healthy smoker effect, whereby those with better lung function may be more likely to be smokers than persons initially with asthma [45]. The longitudinal data examining the association between smoking and asthma are definitive; associations between smoking and asthma are less strong in cross-sectional data. ## Roadway density The positive associations of lifetime and current asthma with increasing roadway density, measured at the level of the HUC in the adjusted analysis, is consistent with other literature that examined asthma in relation to traffic density, particularly in urban areas [3, 4]. In our analyses, the association with roadway density was the strongest and most consistent finding, following adjustment for other known risk factors. Our finding of a strong association between roadway density and risk for lifetime and current asthma in a rural mining community is notable and requires further confirmation. The majority of studies of traffic-related air pollutants (TRAP) and asthma have been undertaken in urban environments and focused primarily on childhood asthma. Porebski et al. [ 2014] has examined the relationship between current asthma symptoms in children in Poland and distance to major roadways, with symptom prevalence being greatest for those living less than 200 m from the roadway [46]. A study among 6,399 adults from the Framingham cohort found that living close to a major roadway (less than 400 m) was associated with an increased prevalence of adult asthma with an adjusted odds ratio of 1.35 ($95\%$ CI: 1.06–1.72) for those living 200-less than 400 m away in comparison to those living greater than 400 m from a roadway. A study by Rice in among participants of the Framingham Offspring and Third Generation studies showed the association between traffic-related air pollution and changes in lung function decline over relatively short distances and with relatively low levels of air pollutants [47]. Lindgren et al. reported significant associations between asthma exacerbations and residential proximity to traffic in a cohort of adults and children in Minnesota [48]. Generally, the higher levels of TRAP, which are associated with roadway proximity or density, reflect exposure to particulates, nitrogen oxide, diesel, carbon monoxide, sulfur oxide, and other volatile organic hydrocarbons [49]. There are multiple components of traffic-related emissions that are likely involved in influencing asthma, which is itself a heterogeneous condition. In addition to nitrogen oxides, sulfur oxides, and other volatile organic hydrocarbons, traffic emissions are composed of fine, ultrafine, and nanoparticles which have recently been shown to be among the most hazardous particles [48]. Traffic and proximity to roadways may be associated with other exposures, including noise. Metrics such as traffic density and distance to roadways may better measure the cumulative effects and various components of TRAP exposures rather than considering any one component individually. Gowers et al. provide a review of potential mechanisms including oxidative stress, airway remodeling, and inflammation and sensitization by which TRAP may induce new cases of asthma as well as exacerbate the symptoms of asthma [50] Our study is unique in highlighting the effect of roadway density in a rural area in eastern Kentucky on adult asthma. There are few EPA air pollution monitoring stations in these areas, and consequently, there are limited data for PM levels in these rural areas compared to more intensively monitored urban areas. Geographic characteristics of these rural areas may trap pollutants, perhaps during cold inversions events, and enhance PM levels [41]. The type of housing (see below) may also be a factor influencing the levels of indoor pollutants, in addition to lifestyle factors such as the amount of time rural residents spend at home and spend in outdoor locations, such as porches, driveways, and yards. ## Public housing In multivariable modeling, residence in multi-unit housing was a significant predictor of both lifetime asthma and current asthma. Apartments included in the sample were almost exclusively public or subsidized housing. This finding is consistent with previous studies focused on children living in large public housing facilities in urban areas [36, 38, 51]. Our findings suggest that residence in public housing may also be an independent contributor to differences in asthma prevalence in rural adults. In an extensive review of the literature, Mendell et al. found sufficient evidence for an association between dampness and mold and asthma development, current asthma, and ever asthma [52]. Walkthrough data from the homes and anecdotal observation from our study indicated that dampness was prevalent in many of the homes [41]. Their suggestion was that indoor dampness and mold prevention are likely to reduce the risk of asthma even without consideration of the specific microbiologic agents involved. ## Mining and other GIS metrics While we found a high prevalence of current asthma ($16.3\%$) among adults in the MAP study, consistent with a previous cross-sectional study by Hendryx et al. [ 2013], no significant positive associations between current or lifetime asthma prevalence and increasing levels of active coal mining within the HUCs were found in multivariable analysis. Rather, we observed a significant protective effect associated with the highest tertile of active mining. Roadway density—a metric more closely associated with current and lifetime asthma in adjusted analyses—was significantly negatively correlated with active mining, however, which suggests an explanation for this counterintuitive finding. Furthermore, during the period of data collection, few active surface mines remained in the study area, yielding only 35 participants with asthma who resided within above-median active mining surface areas. It is important to consider the average duration of residence in the interpretation of the results. Long-time residents may have had asthma due to dusty conditions in the past. The average duration of residence for persons in the entire study was 17.2 years. If mining-related dusts are one potential causal factor related to asthma incidence, or if they trigger asthma episodes, then one may have expected higher prevalence ratios with closer proximity to these mining sites. Further, we do not have data on particulate levels in the past, which may have been much higher, from the local coal mining operations [41]. Finally, the mining operations in the area, especially the surface and mountaintop removal mines, may have only been operational for a limited time in the past, and dust-related exposures due to blasting, coal transport, and active mining operations may have been of relatively limited duration. While our small area metrics of potential exposures at the level of the HUC to mining dust or other mining-related contaminants are limited, they likely represent an improvement to earlier studies that simply characterized whole counties by a mining status classification [18, 19, 21]. Our HUC-based metrics characterize potential exposures at a neighborhood level with a median HUC size of less than 1.8 square miles. There is the potential for exposure misclassification with our strategy, which assigned exposure values at the level of the HUC for roadway density and coal haul routes (road miles/square mile), active and abandoned mining sites (percent of HUC total surface area), and oil and gas wells (wells/square mile). Unknown variables such as the location of dust-producing activities within the permitted mine area, the operational status of facilities, wind directions, length of time in operation (open and closing dates), size of the mining facility, type of mining, and adherence to regulations are all factors which impact PM levels within small areas. However, recent work by our group does indicate a strong concordance between the levels of PM2.5 measured at residences and the HUC estimates of the tertiles of roadway density. In the case of roadway density, our study created a substantial range in exposure variability, which may have allowed for the detection of an effect, whereas our data on mining activity was more limited, with a smaller proportion of HUCs having "high mining exposure." It would be a reasonable expectation that our exposure classification using HUCs was nondifferential by adult self-reported asthma status. ## Strengths and limitations This community-based study was undertaken in two counties in Kentucky which provided geographic diversity and exposure to both active and historic coal mining operations and documented a high prevalence of respiratory health outcomes. The enumeration and recruitment occurred through direct door-to-door contact. In rural community studies of adults, using community-based interviewers is essential to securing participation and gaining a high level of completeness for the questionnaires. Community members were engaged and supportive of a study to address these issues. The overall response rate for eligible households in this study was $82\%$, and we obtained relatively complete data collection from the face-to-face surveys due to well-trained community-based interviewers and diligent effort [29]. We further note that 28 of 40 HUCs enumerated had recruitment rates of $80\%$ or greater and five HUCs had recruitment rates exceeding $90\%$ which reduced issues with potential selection bias across the small geographic areas (HUCs) [29]. Our sampling approach prioritized enrolling those who reported respiratory illness. Such an approach may have introduced a selection bias, potentially inflating the prevalence estimates above what may have occurred in a random sample. This would occur in the situation where there were two or more adults in the home, and at least one adult did not have a respiratory health outcome and the other did and agreed to participate. This would lead to a small increase in the prevalence of respiratory disease in the sample relative to the eligible population. For current asthma, we estimate a $3\%$ increase in asthma prevalence in our enrolled sample in comparison to the fully enumerated sample (all household members) in the study. Although the prevalence of the outcome may be overestimated, the magnitude of the effect estimates for the predictors of asthma outcomes in the Poisson models, where the effect estimates are adjusted prevalence ratios, should not be affected by the sampling approach that we used. While our sample provided good representation across education and income levels which are comparable to Census estimates, we do note that we had greater participation from women compared to men and those persons who were more likely to be home at the time of contact with our enumerators when comparing the enrolled sample to the fully enumerated sample frame in the study [29]. This cross-sectional study has the standard limitations of being conducted during a single time period and obtaining most of the health outcome and covariate data by self-report on the interviewer-administered questionnaire. Computer-assisted data entry using iPads limited the amount of missing data from the questionnaire. The enrollment of adults from eligible households has the potential for selection bias in that recruitment required a person was at the household during one of the multiple attempts to make contact. For example, persons who may have worked in the evenings or weekends or spent little time at home may not have been available for inclusion. There was no sampling frame for all occupied households or a phone listing in these communities, such as a 911 emergency listing for use in the study. Details of the enumeration and survey approach are provided elsewhere [29]. The population of adults residing in these two counties has been relatively stable; we report a duration of residence of 17 years among study participants in their current homes. However, since 2010 there has been substantial outmigration of segments of the population, with an estimated $9\%$ abandoned dwellings estimated by our field staff during the enumeration survey. This result has been supported by a recent Census projection of about a $10\%$ loss to the population in this area. Those who stay and who participated in the survey have been there for a relatively long time, but a segment of the younger population no longer resides in these communities, and they are communities with an older structure to their age pyramids. A series of reports by Hendryx et al. and others have attracted substantial attention for documenting associations, generally at the county level in ecologic analysis, between living in areas with active surface or mountaintop removal mining and a range of health outcomes, including mortality and morbidity from respiratory diseases such as asthma and COPD [18–21, 23–25]. There has been an ongoing debate regarding whether the associations between mining related exposures, typically residence within a county with high coal production or residence near mining locations, and respiratory disease outcomes are likely causal or due to confounding or other biases. There are biologically plausible reasons to infer that exposures from mining including mineral dusts, metals, diesel, air pollutants (NO, SO2, CO) may be biologically related to respiratory impairments. In addition, the two counties in our study are located within the Central Appalachian region – well known as some of the highest areas of unemployment, poverty, lower educational attainment, and poor health status. The limitations of ecologic studies for individual level inferences are well known and, in particular, the limitations of the series of studies, primarily among Appalachian mining communities in West Virginia and Kentucky, have been discussed in detail by Borak et al. [ 26]. *In* general, the inferences made at the ecologic level are not comparable to those made at the individual level. Among the several critiques directed at the previous studies, the limitations with the exposure assessment, typically county level proxies of mining activity, the lack of individual level and complete covariate adjustment, concerns regarding missing variables, and selection biases are the most concerning. For most of these economically distressed counties in Appalachia, there is a substantial overlay of lifestyle risk factors (poverty, unemployment, lower educational levels, poor health status, smoking, and obesity) and potential exposures from coal mining, proximity to roadways, and other sources of contaminants that may impact respiratory function. For successful statistical control of these variables, it is required that there is a differential distribution of each across the units of comparison, whether counties or individuals, to detect effects. The ecologic study design is insufficient to address the question of the individual level health outcomes from a complex mixture or exposure and lifestyle variables. The MAP study provides small area estimates of exposures, at the level of the hydrologic unit classification, to surface and underground mining, traffic density, coal haul routes, and oil and gas operations. This allowed for a range of these exposures within the county and assigned at the individual level, along with individual level covariates for current smoking, environmental tobacco smoke, occupational exposures, and other lifestyle risk factors. The variation of these exposures within participants in our study contrasts with the assignment of only a county-level exposure variable in most previous research. Our efforts to enhance the sampling of persons throughout the two counties by different geographic regions and in relation to their proximity to diverse sources of exposure improves on studies which have employed convenience samples. ## Conclusion Our data suggest that proximity to roadways and conditions of public housing, in particular, increase the asthma risk among adults. Current exposures to dust from mining operations may have influenced respiratory outcomes in the past, but we found no evidence of an association with the small area metrics that we used for the environmental exposures related to mining and oil and gas operations. From the public health perspective, a focus on continuing efforts to reduce exposure to ETS in homes may have the largest impact on the reduction of adult asthma. Well-established risk factors for asthma incidence and episodes, such as ETS, current smoking, and dampness, should be the focus of intervention efforts in these communities. Future research with the MAP data will more closely examine asthma episodes, the heterogeneity of the disease, use of medication, and exposure to PM 2.5. ## References 1. Epstein PR, Buonocore JJ, Eckerle K, Hendryx M, Stout Iii BM, Heinberg R, Clapp RW, May B, Reinhart NL, Ahern MM. **Full cost accounting for the life cycle of coal**. *Ann N Y Acad Sci* (2011.0) **1219** 73-98. DOI: 10.1111/j.1749-6632.2010.05890.x 2. Kurth LM, McCawley M, Hendryx M, Lusk S. **Atmospheric particulate matter size distribution and concentration in West Virginia coal mining and non-mining areas**. *J Expo Sci Environ Epidemiol* (2014.0) **24** 405-411. DOI: 10.1038/jes.2014.2 3. Khreis H, Kelly C, Tate J, Parslow R, Lucas K, Nieuwenhuijsen M. **Exposure to traffic-related air pollution and risk of development of childhood asthma: A systematic review and meta-analysis**. *Environ Int* (2017.0) **100** 1-31. DOI: 10.1016/j.envint.2016.11.012 4. Salam MT, Islam T, Gilliland FD. **Recent evidence for adverse effects of residential proximity to traffic sources on asthma**. *Curr Opin Pulm Med* (2008.0) **14** 3-8. DOI: 10.1097/MCP.0b013e3282f1987a 5. Cardarelli K, Westneat S, Dunfee M, May B, Schoenberg N, Browning S. **Persistent disparities in smoking among rural Appalachians: evidence from the Mountain Air Project**. *BMC Public Health* (2021.0) **21** 270. DOI: 10.1186/s12889-021-10334-6 6. 6.Fan J, Li S, Fan C, Bai Z, Yang K: The impact of PM2.5 on asthma emergency department visits: a systematic review and meta-analysis. Environ Sci Pollut Res Int 2016, 23(1):843–850. 7. 7.Jung CR, Young LH, Hsu HT, Lin MY, Chen YC, Hwang BF, Tsai PJ: PM2.5 components and outpatient visits for asthma: A time-stratified case-crossover study in a suburban area. Environ Pollut 2017, 231(Pt 1):1085–1092. 8. 8.Loftus C, Yost M, Sampson P, Arias G, Torres E, Vasquez VB, Bhatti P, Karr C: Regional PM2.5 and asthma morbidity in an agricultural community: a panel study. Environ Res 2015, 136:505–512. 9. 9.Mirabelli MC, Vaidyanathan A, Flanders WD, Qin X, Garbe P: Outdoor PM2.5, Ambient Air Temperature, and Asthma Symptoms in the Past 14 Days among Adults with Active Asthma. Environ Health Perspect 2016, 124(12):1882–1890. 10. Penttinen P, Vallius M, Tiittanen P, Ruuskanen J, Pekkanen J. **Source-specific fine particles in urban air and respiratory function among adult asthmatics**. *Inhal Toxicol* (2006.0) **18** 191-198. DOI: 10.1080/08958370500434230 11. 11.Ayers H, Boyer J, Feeney P, McGrath L, Owen S, Rodgers K, Wasson M. Appalachian Coalfield Delegation Position Paper on Sustainable Energy, United Nations Commission on Sustainable Development 15th Session. New York. 2010. https://www.datacenter.org/wp-content/uploads/csdpositionpaper.pdf. 12. 12.Cai Y, Zijlema WL, Doiron D, Blangiardo M, Burton PR, Fortier I, Gaye A, Gulliver J, de Hoogh K, Hveem K et al: Ambient air pollution, traffic noise and adult asthma prevalence: a BioSHaRE approach. Eur Respir J 2017, 49(1):1502127. 13. 13.Kunzli N, Bridevaux PO, Liu LJ, Garcia-Esteban R, Schindler C, Gerbase MW, Sunyer J, Keidel D, Rochat T, Swiss Cohort Study on Air P et al: Traffic-related air pollution correlates with adult-onset asthma among never-smokers. Thorax 2009, 64(8):664–670. 14. Lindgren A, Bjork J, Stroh E, Jakobsson K. **Adult asthma and traffic exposure at residential address, workplace address, and self-reported daily time outdoor in traffic: A two-stage case-control study**. *BMC Public Health* (2010.0) **10** 716. DOI: 10.1186/1471-2458-10-716 15. Barry AC, Mannino DM, Hopenhayn C, Bush H. **Exposure to indoor biomass fuel pollutants and asthma prevalence in Southeastern Kentucky: results from the Burden of Lung Disease (BOLD) study**. *J Asthma* (2010.0) **47** 735-741. DOI: 10.3109/02770903.2010.485661 16. Methvin JN, Mannino DM, Casey BR. **COPD prevalence in southeastern Kentucky: the burden of lung disease study**. *Chest* (2009.0) **135** 102-107. DOI: 10.1378/chest.08-1315 17. Christian WJ, Walker CJ, Huang B, Hahn EJ. **Effect of Local Smoke-Free Ordinances on Smoking Prevalence in Kentucky, 2002–2009**. *South Med J* (2019.0) **112** 369-375. DOI: 10.14423/SMJ.0000000000001000 18. 18.Hendryx M: Personal and Family Health in Rural Areas of Kentucky with and without Mountaintop Coal Mining. J Rural Health. 2013 29 Suppl 1:s79. 19. Hendryx M, Ahern MM. **Relations between health indicators and residential proximity to coal mining in West Virginia**. *Am J Public Health* (2008.0) **98** 669-671. DOI: 10.2105/AJPH.2007.113472 20. Hendryx M, Ahern MM. **Mortality in Appalachian coal mining regions: the value of statistical life lost**. *Public Health Rep* (2009.0) **124** 541-550. DOI: 10.1177/003335490912400411 21. Hendryx M. **Mortality from heart, respiratory, and kidney disease in coal mining areas of Appalachia**. *Int Arch Occup Environ Health* (2009.0) **82** 243-249. DOI: 10.1007/s00420-008-0328-y 22. Hendryx M, Zullig KJ. **Higher coronary heart disease and heart attack morbidity in Appalachian coal mining regions**. *Prev Med* (2009.0) **49** 355-359. DOI: 10.1016/j.ypmed.2009.09.011 23. Ahern MM, Hendryx M, Conley J, Fedorko E, Ducatman A, Zullig KJ. **The association between mountaintop mining and birth defects among live births in central Appalachia, 1996–2003**. *Environ Res* (2011.0) **111** 838-846. DOI: 10.1016/j.envres.2011.05.019 24. 24.Hendryx M, O’Donnell K, Horn K. Lung cancer mortality is elevated in coal-mining areas of Appalachia. Lung Cancer. 2008;62(1):1–7. 25. Hendryx M, Wolfe L, Luo J, Webb B. **Self-reported cancer rates in two rural areas of West Virginia with and without mountaintop coal mining**. *J Community Health* (2012.0) **37** 320-327. DOI: 10.1007/s10900-011-9448-5 26. Borak J, Salipante-Zaidel C, Slade MD, Fields CA. **Mortality disparities in Appalachia: reassessment of major risk factors**. *J Occup Environ Med* (2012.0) **54** 146-156. DOI: 10.1097/JOM.0b013e318246f395 27. Schoenberg NE, Huang B, Seshadri S, Tucker TC. **Trends in cigarette smoking and obesity in Appalachian Kentucky**. *South Med J* (2015.0) **108** 170-177. DOI: 10.14423/SMJ.0000000000000245 28. Gan WQ, Sanderson WT, Browning SR, Mannino DM. **Different types of housing and respiratory health outcomes**. *Prev Med Rep* (2017.0) **7** 124-129. DOI: 10.1016/j.pmedr.2017.05.018 29. May BA, Cardarelli KM, Silver R, Christian WJ, Schoenberg NE, Browning SR. **Hollows as Sampling Units for Community-Based Participatory Research in Appalachia: The Mountain Air Project**. *Prog Community Health Partnersh* (2019.0) **13** 401-410. DOI: 10.1353/cpr.2019.0065 30. 30.Villarroel MB, Debra QuickStats: Percentage of Adults Aged 18–64 Years with Current Asthma, by State — National Health Interview Survey, 2014–2016. In. Edited by Rep MMMW, 67:590: CDC 2018. 31. 31.Schoenberg NE, Hatcher J, Dignan MB. Appalachian women’s perceptions of their community’s health threats. J Rural Health. 2008;24(1):75–83. 32. Schoenberg NE, Hatcher J, Dignan MB, Shelton B, Wright S, Dollarhide KF. **Faith Moves Mountains: an Appalachian cervical cancer prevention program**. *Am J Health Behav* (2009.0) **33** 627-638. DOI: 10.5993/AJHB.33.6.1 33. Schoenberg NE, Howell BM, Fields N. **Community strategies to address cancer disparities in Appalachian Kentucky**. *Fam Community Health* (2012.0) **35** 31-43. DOI: 10.1097/FCH.0b013e3182385d2c 34. 34.ESRI: ArcGis 10.3. In. Redlands, California. 35. Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. **Research electronic data capture (REDCap)–a metadata-driven methodology and workflow process for providing translational research informatics support**. *J Biomed Inform* (2009.0) **42** 377-381. DOI: 10.1016/j.jbi.2008.08.010 36. Krieger J, Takaro TK, Song L, Beaudet N, Edwards K. **A randomized controlled trial of asthma self-management support comparing clinic-based nurses and in-home community health workers: the Seattle-King County Healthy Homes II Project**. *Arch Pediatr Adolesc Med* (2009.0) **163** 141-149. DOI: 10.1001/archpediatrics.2008.532 37. Krieger JK, Takaro TK, Allen C, Song L, Weaver M, Chai S, Dickey P. **The Seattle-King County healthy homes project: implementation of a comprehensive approach to improving indoor environmental quality for low-income children with asthma**. *Environ Health Perspect* (2002.0) **110** 311-322. DOI: 10.1289/ehp.02110s2311 38. Krieger JW, Takaro TK, Song L, Weaver M. **The Seattle-King County Healthy Homes Project: a randomized, controlled trial of a community health worker intervention to decrease exposure to indoor asthma triggers**. *Am J Public Health* (2005.0) **95** 652-659. DOI: 10.2105/AJPH.2004.042994 39. 39.Braun-Fahrlander C, Gassner M, Grize L, Minder CE, Varonier HS, Vuille JC, Wuthrich B, Sennhauser FH: Comparison of responses to an asthma symptom questionnaire (ISAAC core questions) completed by adolescents and their parents. SCARPOL-Team. Swiss Study on Childhood Allergy and Respiratory Symptoms with respect to Air Pollution. Pediatr Pulmonol 1998, 25(3):159–166. 40. Valle SO, Kuschnir FC, Sole D, Silva MA, Silva RI, Da Cunha AJ. **Validity and reproducibility of the asthma core International Study of Asthma and Allergies in Childhood (ISAAC) written questionnaire obtained by telephone survey**. *J Asthma* (2012.0) **49** 390-394. DOI: 10.3109/02770903.2012.669440 41. 41.Flunker J: "Environmental Respiratory Exposures and Pulmonary Function Among Residents of Rural Appalachia, Kentucky" https://uknowledge.uky.edu /epb_etds/28. Dissertation Lexington , Kentucky University of Kentucky 2021. 42. 42.StataCorp: Stata Statistical Software: Release 12. In. College Station: StataCorp LP; 2011. 43. Barros AJ, Hirakata VN. **Alternatives for logistic regression in cross-sectional studies: an empirical comparison of models that directly estimate the prevalence ratio**. *BMC Med Res Methodol* (2003.0) **3** 21. DOI: 10.1186/1471-2288-3-21 44. 44.Mabila SL, Almberg KS, Friedman L, Cohen R: High exposure mining occupations are associated with obstructive lung disease, National Health Interview Survey (NHIS), 2006–2015. Am J Ind Med. 2018;61:715–24. 45. Weiss ST, Utell MJ, Samet JM. **Environmental tobacco smoke exposure and asthma in adults**. *Environ Health Perspect* (1999.0) **107** 891-895. DOI: 10.1289/ehp.99107s6891 46. Porebski G, Wozniak M, Czarnobilska E. **Residential proximity to major roadways is associated with increased prevalence of allergic respiratory symptoms in children**. *Ann Agric Environ Med* (2014.0) **21** 760-766. DOI: 10.5604/12321966.1129929 47. 47.Rice MB, Ljungman PL, Wilker EH, Dorans KS, Gold DR, Schwartz J, Koutrakis P, Washko GR, O’Connor GT, Mittleman MA. Long-term exposure to traffic emissions and fine particulate matter and lung function decline in the Framingham heart study. Am J Respir Crit Care Med. 2015;191(6):656–64. 48. Lindgren P, Johnson J, Williams A, Yawn B, Pratt GC. **Asthma exacerbations and traffic: examining relationships using link-based traffic metrics and a comprehensive patient database**. *Environ Health* (2016.0) **15** 102. DOI: 10.1186/s12940-016-0184-2 49. 49.Liu SV, Chen FL, Xue J: Evaluation of traffic density parameters as an indicator of vehicle emission-related near-road air pollution: a case study with NEXUS measurement data on black carbon. Int J Environ Res Public Health 2017, 14(12):1581. 50. Gowers AM, Cullinan P, Ayres JG, Anderson HR, Strachan DP, Holgate ST, Mills IC, Maynard RL. **Does outdoor air pollution induce new cases of asthma? Biological plausibility and evidence; a review**. *Respirology* (2012.0) **17** 887-898. DOI: 10.1111/j.1440-1843.2012.02195.x 51. Krieger JW, Song L, Takaro TK, Stout J. **Asthma and the home environment of low-income urban children: preliminary findings from the Seattle-King County healthy homes project**. *J Urban Health* (2000.0) **77** 50-67. DOI: 10.1007/BF02350962 52. Mendell MJ, Mirer AG, Cheung K, Tong M, Douwes J. **Respiratory and allergic health effects of dampness, mold, and dampness-related agents: a review of the epidemiologic evidence**. *Environ Health Perspect* (2011.0) **119** 748-756. DOI: 10.1289/ehp.1002410
--- title: 'Evaluation of overweight control applications with cognitive‐behavioral therapy approach: A systematic review' authors: - Negin Ebrahimi - Niloofar Mohammadzadeh - Seyed Mohammad Ayyoubzadeh journal: Health Science Reports year: 2023 pmcid: PMC10041866 doi: 10.1002/hsr2.1157 license: CC BY 4.0 --- # Evaluation of overweight control applications with cognitive‐behavioral therapy approach: A systematic review ## Abstract ### Background and Aims Overweight and obesity lead to the development of physical diseases. Cognitive factors play a vital role in controlling one's weight. Currently, cognitive‐behavioral therapy (CBT) interventions are recognized as a subcategory of lifestyle modification programs that can be implemented to control weight and modify eating patterns as well as physical activity. Nowadays, smartphone‐based applications are utilized to implement behavioral interventions. The main purpose of this study is to evaluate the quality of CBT‐based smartphone applications available on Google Play and the App Store in the field of overweight control. ### Methods Smartphone‐based utility applications available on Google Play and App Store were identified in March 2021. Weight control smartphone applications were obtained based on inclusion and exclusion criteria. The app name, platform, version, number of downloads, password protection, affiliations, and features of retrieved apps were tabulated. The Mobile Application Rating Scale was utilized to evaluate the quality of the identified apps. ### Results Seventeen CBT‐based weight control smartphone apps were retrieved. The average engagement, functionality, aesthetics, and information quality scores were 3.65, 3.92, 3.80, and 3.91, respectively. Also, the average score in an aspect containing the usefulness of the app, frequency of using the application, cost, and user satisfaction was 3.5. ### Conclusion Future applications related to this field can be improved by providing a personalization program according to the needs of users and the possibility of online chatting with the therapist. Further improvements can be achieved by improving the areas of engagement, aesthetics, and subjective quality as well as having appropriate privacy policies. ## INTRODUCTION Being overweight is one of the most significant factors that lead to developing physical diseases such as cardiovascular disease, diabetes, joint damage, lumbago, certain types of cancer, hypertension, fatty liver, and an increased mortality rate. 1, 2 The World Health Organization has classified the prevalence of being overweight as a global epidemic due to the rapidly rising rate of overweight and obesity throughout the globe. 3 Cognitive factors play a key role in controlling one's weight. Altering these maladaptive cognitions using implementing control of overeating and satiety is highly effective in the success achieved through dieting. 4 Currently, cognitive‐behavioral therapy (CBT) interventions are recognized as a subcategory of lifestyle modification programs that can be implemented to control weight and modify eating patterns as well as physical activity. 5 This approach ultimately assists people in altering their eating habits, not in the short term but throughout their entire lifetime, by changing their manner of thinking. 6 Nowadays, smartphone‐based applications are utilized to implement behavioral interventions. 7 The use of such applications to help control weight is on the rise. However, the quality and quantity of the effectiveness as well as efficiency of these applications have not been entirely determined. 8 The results of a systematic review by Antoun et al. indicated that mobile‐based interventions aimed at losing weight were more effective on target populations than the control groups by providing a suitable diet, physical activity program, and behavioral indicators that determine weight management. 9 However, based on a review study, none of these applications have addressed behavioral aspects and evaluated the dysfunctional thoughts that prevent accomplishing the permanent weight loss objective. 10 Meanwhile, studies have indicated that smartphone devices may be used to boost self‐efficacy by activating specific behavior and altering dysfunctional thoughts in other fields of health. Companies such as Apple and Google provide users with numerous health applications through the App Store and Google Play platforms, respectively. However, only a few of these apps have been approved by the Food and Drug Administration. 11 Furthermore, there are currently no regulations for developing smartphone‐based health applications. Thus, evaluating health apps is of the utmost importance. Numerous tools are available for evaluating applications, including the QUIZ tools, the Nielsen, Norman, and Shneiderman models, as well as the Mobile Application Rating Scale (MARS). One of the tools utilized to evaluate smartphone applications, particularly in the field of health applications, is the MARS scale. 12 This scale consists of 23 items that evaluate the quality of smartphone applications in four qualitative aspects (engagement, functionality, aesthetics, and information quality) and one subjective aspect. 13 In addition to evaluating applications from a quantitative perspective, this scale also qualitatively evaluates them. This scale has been put into use to evaluate smartphone apps in the field of diabetes, 14 pregnancy, 15 epilepsy, 16 sleep self‐care control, 17 health behavior change, 18 COVID‐19 management, 19 and food allergies. 20 The main purpose of this study is to evaluate the quality of smartphone applications available on the Google Play and App Store in the field of overweight control based on CBT by applying the MARS scale. ## Search strategy Smartphone‐based utility applications that are available on Google Play and App Store were identified in March 2021 using the following expressions and keywords: “Cognitive Behavioral Therapy,” “Obesity,” “Weight,” and “Eating Disorder.” *The data* were then extracted and analyzed in April and May 2021. ## App selection The app selection process is described in Figure 1. The inclusion criteria for evaluating the applications in this study included: [1] those being relevant to CBT or weight control, or a combination of both, [2] apps that have a minimum rating of 4 by users, [3] apps that use the English language, and [4] those that are free to download and install. The exclusion criteria in this study included: [1] the apps that were only based on diet, [2] apps that were only based on physical activity, [3] apps that were only based on CBT, and [4] apps that were are not accessible. **Figure 1:** *Flowchart indicating the screening process.* ## Data extraction All identified apps were registered in an initial list to count the total number of apps and the number of duplicates. *The* general characteristics of the included apps were extracted from the information in the app stores, while the main app features were verified by the authors by utilizing the app. ## MARS app quality assessment The MARS can be utilized to evaluate the quality of the identified smartphone apps. This scale enables the quality assessment of smartphone health apps with their engagement, functionality, aesthetics, information quality, and subjective features. The MARS scale is a 23‐item tool featuring a 19‐item qualitative part divided into four aspects (engagement, functionality, aesthetics, and information quality) and, finally, one subjective aspect (Table 1). The MARS scale has a reliability of 90 (α = 90) and a validity of 79 (interclass correlation coefficient = 79. 12 **Table 1** | 1. Engagement “Entertainment, Interest, Customization, Interactivity, Target Group” | | --- | | 2. Functionality “Functionality, Ease of Use, Navigation, Gestural Design” | | 3. Aesthetics “Layout, Graphics, Visual Appeal” | | 4. Information quality “Accuracy of App Description, Goals, Quality of Information, Quantity of Information, Visual Information, Credibility, Evidence‐Based” | | 5. Subjective quality Would you recommend this app to people who might benefit from it? Would you pay for this app? How many times do you think you would use this app in the next 12 months if it were relevant to you? What is your overall star rating for the app? | ## Data analysis According to the MARS scale, all the information is rated based on a 5‐point Likert scale (1 = inadequate, 2 = poor, 3 = acceptable, 4 = good, and 5 = excellent). For the quality assessment of the apps, three experts of the research team, including one health information management, one medical informatics, and one software computer engineer answered 23 questions of the MARS scale. Their research areas were e‐health and mobile app development. Additionally, they had at least 2 years of relevant work experience. The higher the score, the better the quality of the smartphone application. Microsoft Excel 2010 was used in this step. ## RESULTS In the first step, all relevant applications were identified according to the keywords, and 246 apps were screened according to the study inclusion criteria. The identified applications were then screened based on the exclusion criteria. Later, the remaining 22 smartphone applications were downloaded by research members. Seventeen CBT‐based weight control smartphone apps were examined according to the search strategy and their characteristics (Table 2). All of the identified 17 applications were then rated by research members using the MARS scale and on a 5‐point Likert scale (Table 3). Among the apps, two apps named Noom and Peace with Food were available for free download and installation. However, gaining full access to these apps' features required in‐app payment. The evaluation of apps using the MARS scale has been as follows: ## Engagement The scores for this aspect were obtained in five criteria (entertainment, interest, customization, engagement, and target setting) with an average of 3.65. These scores were within the range of 1.2–4.8 out of 5. The Mindful application received the highest score in terms of engagement. This smartphone app helps people find their inner motivation, follow their diet consciously, create and maintain a healthy lifestyle, resist food cravings, and overcome laziness and other issues related to losing weight. ## Functionality The scores regarding this aspect of the study were obtained in four criteria with an average of 3.92. The minimum and maximum scores were 1.7 and 5 out of 5. The Mindful application received the highest score and is designed to achieve the target weight by following a safe and efficient approach. This smartphone app not only includes weight loss features but also offers scientifically designed plans and informed eating guides. ## Aesthetics The scores for this aspect were obtained in three criteria with an average of 3.80. These scores ranged from 2 to 5. The CBT Companion and Mindful smartphone apps received the highest scores in aesthetics due to their user interface quality and visual attractiveness. ## Information quality The scores for this aspect were obtained in seven criteria with an average of 3.91. These scores ranged from 2.1 to 4.9 range. The Eat Right Now and Mindful smartphone apps ultimately received the top scores regarding this aspect of the study. Eat Right *Now is* an educational application that enables users to change their eating habits. Furthermore, this app provides step‐by‐step tutorials which assist the user in controlling their food cravings. The application provides its tutorials through audio and video playlists, target‐setting tools, and daily reminders. ## Subjective quality The scores regarding this aspect were obtained in four criteria (usefulness of the app, frequency of using the application, cost, and user satisfaction) with an average of 3.5. These scores had a minimum and maximum of 1.8 and 4.8, respectively. The Eat Right Now and Mindful smartphone apps received the top scores regarding this aspect of the study. ## DISCUSSION This study used the MARS scale to examine the qualitative evaluation of available smartphone CBT‐based applications in the field of overweight control. Relevant applications were downloaded and reviewed according to the inclusion and exclusion criteria of the study. They were then scored according to the MARS scale. The smartphone apps reviewed in this study focused primarily on sending users daily reminders, recording daily food intake, and registering everyday emotions. Users should be sufficiently aware of the amount of food they consume during the day to be able to control their weight. Thus, users can be guided by sending reminders throughout the day. Additionally, the daily emotions of each person have a profound effect on controlling their weight. Thus, registering such feelings on the app and being provided with exercises to relieve as well as eliminate negative emotions can prevent users from overeating due to such nervous tensions. The results of the current study are in line with previous studies indicating that frequent recording of daily food intake is the key to overweight treatment. 21 In addition, according to the results of this research, sending motivational messages about the importance of self‐monitoring in addition to incentivizing patients to use the app, may enhance adherence in overweight management. 22 The results of the reports suggest that daily or weekly feedback and encouragement by sending motivational messages can also promote dietary intake and self‐efficacy. 23, 24 In addition to the mentioned cases, psychological training plays a significant role in controlling people's weight. Users can alter their eating behaviors for the better by using the tutorials provided by these apps. According to the results reported by Wadden et al., combining a weight loss program with a weight loss counseling program can be a powerful combination of tools. 25 The results revealed all of the apps related to overweight control based on CBT were commercial apps and therefore there is a lack of science‐based apps in the apps market. Thus, there is a need to develop more scientific‐based apps by academic institutions. One of the smartphone applications that scored the highest on the MARS scale is the Mindful app. This application can guide users to eat consciously and provides users with 4‐week exercises which help them relax and focus more on their diet. In the study, it was identified that reducing the manual entry of diet and physical activity information as well as turning the apps into semiautomatic ones can enhance their engagement, attractiveness, and aesthetics and ultimately increase the chance of them being frequently used by users. The current study's findings indicated that the Functionality, Information Quality, and Aesthetics of apps should not be considered the only essential aspects when designing such applications. The engagement and Subjective Quality of the apps should also be taken into account. Furthermore, the positive role of behavioral skills in weight control was observed in only 2 out of the 23 applications. Thus, the findings revealed, as mentioned by Bardus et al., 26 the apps with higher quality include the following features: semiautomatic tracking (self‐monitoring), having a forum, and sharing their content on social media (for instance, providing social support), and employing notifications (such as commands/behavioral cues). Comparing the overall MARS score and the number of downloads indicated that apps with a score of less than 4 generally had less than 50K number of downloads. However, one app titled “Rise Up” had a low overall score of MARS (2.8) and a high number of downloads (+100K). The use of specialized tools such as the MARS scale to evaluate the quality of smartphone applications is practical and efficient in this regard. 12, 27 Qualitative indicators such as reliability, comprehensiveness, accuracy, and control of data sharing should also be taken into account regarding this matter. To adequately meet the demands of users, the applications available for download should be user‐friendly, simple, and attractive. The main features that affect the scores of a smartphone app in this category include sending motivational and informative reminders to users. Also, providing psychological training and solutions could help eliminate negative thoughts and establish correct eating behaviors in users. The possibility of recording food intake and emotions daily as well as setting forth approaches would encourage users to use the application again. ## Limitations The most important limitation of this study was that we could not include any apps that did not have a free trial version in the evaluation. Additionally, we may have missed some apps that did not include any of our search criteria related to “CBT” and “Diet” in their titles or descriptions. Furthermore, the app market is constantly changing, with old apps being updated or removed from the app store and new apps being added. As such, the review studies such as this one will need to be regularly updated to keep up with the rapidly changing digital healthcare landscape. ## CONCLUSIONS Future applications related to this field can be improved by providing a personalization program according to the needs of users and the possibility of online chatting with the therapist. Further improvements can be achieved by ameliorating the areas of engagement, aesthetics, and subjective quality as well as having appropriate privacy policies. ## AUTHOR CONTRIBUTIONS Negin Ebrahimi: Investigation; writing—original draft. Niloofar Mohammadzadeh: Conceptualization; methodology; supervision. Seyed Mohammad Ayyoubzadeh: Writing—review and editing. ## CONFLICT OF INTEREST STATEMENT The authors declare no conflict of interest. ## TRANSPARENCY STATEMENT The lead author Niloofar Mohammadzadeh affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained. ## DATA AVAILABILITY STATEMENT The data that support the findings of this study are available from the corresponding author upon reasonable request. ## References 1. Celik O, Yildiz BO. **Obesity and physical exercise**. *Minerva Endocrinol* (2021) **46** 131-144 2. Lin X, Li H. **Obesity: epidemiology, pathophysiology, and therapeutics**. *Front Endocrinol* (2021) **12** 3. 3 World Health Organization . Obesity: Preventing and Managing the Global Epidemic. Report of a World Health Organization Consultation. WHO Obesity Technical Report Series; 2000.. *Obesity: Preventing and Managing the Global Epidemic. Report of a World Health Organization Consultation* (2000) 4. Hilbert A, Petroff D, Herpertz S. **Meta‐analysis on the long‐term effectiveness of psychological and medical treatments for binge‐eating disorder**. *Int J Eat Disord* (2020) **53** 1353-1376. PMID: 32583527 5. Jacob A, Moullec G, Lavoie KL. **Impact of cognitive‐behavioral interventions on weight loss and psychological outcomes: a meta‐analysis**. *Health Psychol* (2018) **37** 417-432. PMID: 29698017 6. Madjd A, Taylor MA, Delavari A, Malekzadeh R, Macdonald IA, Farshchi HR. **Effects of cognitive behavioral therapy on weight maintenance after successful weight loss in women: a randomized clinical trial**. *Eur J Clin Nutr* (2020) **74** 436-444. PMID: 31462695 7. Stawarz K, Preist C, Tallon D, Wiles N, Coyle D. **User experience of cognitive behavioral therapy apps for depression: an analysis of app functionality and user reviews**. *J Med Internet Res* (2018) **20**. PMID: 29875087 8. Islam MM, Poly TN, Walther BA, Li YC. **Use of mobile phone app interventions to promote weight loss: meta‐analysis**. *JMIR Mhealth Uhealth* (2020) **8**. PMID: 32706724 9. Antoun J, Itani H, Alarab N, Elsehmawy A. **The effectiveness of combining nonmobile interventions with the use of smartphone apps with various features for weight loss: systematic review and meta‐analysis**. *JMIR Mhealth Uhealth* (2022) **10**. PMID: 35394443 10. Paramastri R, Pratama SA, Ho DKN. **use of mobile applications to improve nutrition behaviour: a systematic review**. *Comput Methods Programs Biomed* (2020) **192**. PMID: 32234632 11. Shan R, Sarkar S, Martin SS. **Digital health technology and mobile devices for the management of diabetes mellitus: state of the art**. *Diabetologia* (2019) **62** 877-887. PMID: 30963188 12. Terhorst Y, Philippi P, Sander LB. **Validation of the Mobile Application Rating Scale (MARS)**. *PLoS One* (2020) **15**. PMID: 33137123 13. Davalbhakta S, Advani S, Kumar S. **A systematic review of smartphone applications available for corona virus disease 2019 (COVID19) and the assessment of their quality using the Mobile Application Rating Scale (MARS)**. *J Med Syst* (2020) **44** 164. PMID: 32779002 14. Chavez S, Fedele D, Guo Y. **Mobile apps for the management of diabetes**. *Diabetes Care* (2017) **40** e145-e146. PMID: 28774944 15. Brown HM, Bucher T, Collins CE, Rollo ME. **A review of pregnancy iPhone apps assessing their quality, inclusion of behaviour change techniques, and nutrition information**. *Matern Child Nutr* (2019) **15**. PMID: 30569549 16. Mohammadzadeh N, Khenarinezhad S, Ghazanfarisavadkoohi E, Safari MS, Pahlevanynejad S. **Evaluation of M‐Health applications use in epilepsy: a systematic review**. *Iran J Publ Health* (2021) **50** 459-469 17. Choi YK, Demiris G, Lin SY. **Smartphone applications to support sleep self‐management: review and evaluation**. *J Clin Sleep Med* (2018) **14** 1783-1790. PMID: 30353814 18. McKay FH, Cheng C, Wright A, Shill J, Stephens H, Uccellini M. **Evaluating mobile phone applications for health behaviour change: a systematic review**. *J Telemed Telecare* (2018) **24** 22-30. PMID: 27760883 19. Salehinejad S, Niakan Kalhori SR, Hajesmaeel Gohari S, Bahaadinbeigy K, Fatehi F. **A review and content analysis of national apps for COVID‐19 management using Mobile Application Rating Scale (MARS)**. *Inform Health Soc Care* (2021) **46** 42-55. PMID: 33164594 20. Mandracchia F, Llauradó E, Tarro L, Valls RM, Solà R. **Mobile phone apps for food allergies or intolerances in app stores: systematic search and quality assessment using the Mobile App Rating Scale (MARS)**. *JMIR Mhealth Uhealth* (2020) **8**. PMID: 32936078 21. Wadden TA, Tronieri JS, Butryn ML. **Lifestyle modification approaches for the treatment of obesity in adults**. *Am Psychol* (2020) **75** 235-251. PMID: 32052997 22. Patrick K, Raab F, Adams MA. **A text message‐based intervention for weight loss: randomized controlled trial**. *J Med Internet Res* (2009) **11** 23. Burke LE, Styn MA, Sereika SM. **Using mHealth technology to enhance Self‐Monitoring for weight loss**. *Am J Prev Med* (2012) **43** 20-26. PMID: 22704741 24. Carter MC, Burley VJ, Nykjaer C, Cade JE. **adherence to a smartphone application for weight loss compared to website and paper diary: pilot randomized controlled trial**. *J Med Internet Res* (2013) **15** 25. Wadden TA, Volger S, Sarwer DB. **A two‐year randomized trial of obesity treatment in primary care practice**. *N Engl J Med* (2011) **365** 1969-1979. PMID: 22082239 26. Bardus M, van Beurden SB, Smith JR, Abraham C. **A review and content analysis of engagement, functionality, aesthetics, information quality, and change techniques in the most popular commercial apps for weight management**. *Int J Behav Nutr Phys Act* (2016) **13** 35. PMID: 26964880 27. Kalhori SRN, Hemmat M, Noori T, Heydarian S, Katigari MR, Kalhori SRN. **Quality evaluation of English mobile applications for gestational diabetes: app review using Mobile Application Rating Scale (MARS)**. *Curr Diabetes Rev* (2021) **17** 161-168. PMID: 32619173
--- title: '“Not Today, Diabetes”: Using Blog Analysis to Understand Emotional Interactions and Support Among People With Type 1 Diabetes' authors: - Heather L. Stuckey - Sean M. Oser - Erin L. Miller - Tamara K. Oser - Mark Peyrot - Aditi Sharma journal: Frontiers in Clinical Diabetes and Healthcare year: 2021 pmcid: PMC10041872 doi: 10.3389/fcdhc.2020.613569 license: CC BY 4.0 --- # “Not Today, Diabetes”: Using Blog Analysis to Understand Emotional Interactions and Support Among People With Type 1 Diabetes ## Abstract The goal of this study is to understand how internet blogs are used by people with type 1 diabetes (T1D) to provide or exchange social support. A stratified, clustered proportionate probability sample of entries from 10 Internet blogs focusing on T1D was obtained. A random sample of 100 days generated 200 blogger posts and 1,606 commenter responses. Entries were coded using qualitative analysis software and analyzed thematically. Blogs were used as a dynamic, interactional form of emotional support from others who understood diabetes from personal experience; and as a source of sharing lived user experience of having diabetes, more often than as a way of communicating medical knowledge or facts about diabetes. Blog participation contributed to a sense of belonging for participants in the “Diabetes Online Community” where there was a shared culture. In conclusion, blogs provide unobtrusive access to the experiences of people with T1D that are driven by their interests rather than those of qualitative research interviewers or healthcare providers. In addition to permitting analysis of the way that participants use blogs to address their own personal wants and needs, blog data can serve as an inexpensive and unobtrusive method for studying topics of interests to researchers and healthcare providers. ## Introduction Few health conditions require as much self-management as type 1 diabetes (T1D) [1, 2]. Due to the complexity of the condition and its management, people with T1D (PWT1D) require various types of support for self-management and social/emotional support (3–7). Traditional sources of support have been healthcare providers (HCPs), support groups, family members and friends. The Internet has emerged as a wealth of ongoing, interactive information and social support for diabetes (8–14). Illness blogs (online journals about a content area) allow for study of the experience of illness in a naturalistic and longitudinal manner, often with greater detail than data relying only on participant recall. Participants produce online illness blogs to share their own illness narratives and connect with others going through similar processes [15]. Social media has enabled PWT1D to find and maintain connections with peers with T1D for self-management support. “ Diabetes Online Community” (DOC) is a widely used term that encompasses all of the people who engage in various online activities related to living with diabetes across a collection of web-based platforms including community forums, blogs and social media sites such as Facebook and Twitter [16]. The theory of social constructivism suggests that as people share background knowledge and participate in collaborative activities, they negotiate meaning and build knowledge, not as individuals, but as a community [17]. An online community provides opportunity for expression to make sense of the condition through sharing, negotiating, and building knowledge. As such, social media can increase the sense of connectedness [18], and the use of social media has been increasing as an area of research interest related to T1D self-management and support (10, 19–22). One form of social media is blogs, which provide insights, information, and comments on a specific topic area. In recent years, blogging activity has increased dramatically—from just $4\%$ of social media usage in 2008 to $47\%$ in 2018 with 3.196 billion social media users worldwide (23–25). Unlike other social networking forums, such as Facebook and Twitter, blogs provide sustained and focused dialogue with peers, remain accessible and open to all who seek or subscribe to them, are socially constructed, and are often moderated by peers who filter inaccurate or harmful comments and commercial promotions [26]. Moderated discussions have been found to result in increased social communication, improved participation, and increased trust among participants [27]. Blogs are one way that the people of the DOC create and share culture together in a virtual community, comprised of three roles: (a) “Bloggers” journal their experiences for others; (b) “Commenters” read and actively comment on others’ posts, and they may or may not also be bloggers themselves; and (c) “Lurkers” read others’ blogs without commenting [28]. With approximately $85\%$ of the estimated 1.25M Americans with T1D being adults (and 3M people worldwide), and with 40,000 new T1D diagnoses annually [29], it is likely that blogs will continue to be a significant source of information and self-management support for PWT1D [30]. Popular T1D blogs, such as SixUntilMe, Scott’sDiabetes, and DiabetesMine, are viewed approximately 90,000 times per month, with 45,000 of these views by unique individuals. Blogging activity has positive association with health outcomes and is valuable as a means of providing support and promoting self-management [14, 16, 31]. One venue for dissemination of insights from and benefits of T1D blogs is through citations and referrals by healthcare providers (HCPs). However, many HCPs are not engaged with the same types of social media as their patients [30] and therefore are not presently serving as collaborators on T1D blogs to inform and provide guidance [32]. The Association of Diabetes Care & Education Specialists (ADCES) has endorsed referral to the DOC and has developed a handout to provide to PWT1D to refer them to the DOC (Warshaw & Edelman, 2019), however only one in 3 diabetes educators ($34.7\%$) recommends the DOC to their patients [33]. Recently, the ADCES further endorsed referral to online peer support communities such as the DOC in their 2019 Perspectives in Practice [34]. Although much has been written about the complexities of managing T1D, some HCPs do not understand the day-to-day realities of living with such a complex condition [35]. Studies show discordance between PWT1D and HCP perceptions of different barriers to diabetes care and their importance [36, 37]. Whereas HCPs perceived informational barriers as most important and believed that people with diabetes (PWD) wanted more education, PWD wanted more psychosocial support and identified psychological barriers as most important [6]. PWD may be able to get some of this support from their HCPs, but PWD may know more about certain aspects of life with diabetes and its self-management than do their HCPs [37], and they may turn to peers for such support and to tackle these daily life self-management issues (38–40). In this paper we examine the use of blogs to understand how the DOC provide and exchange support to one another. ## Materials and Methods The Internet ethnography (“netnography”) [41] methods used in this study were approved by the Penn State College of Medicine Institutional Review Board. ## Recruitment and Sampling Strategy Through snowball sampling [42], ten online blogs written by adults with T1D were selected for analysis (list included in Table 1). Sampling began with the most visited and visible adult T1D blog, Six Until Me (www.sixuntilme.com). Identified blog authors gave permission to retrospectively analyze data publicly available on their blogs posted between June 1, 2012 and July 31, 2014. **Table 1** | Blog Name | Posts | Number of Total Comments | Number of Unique Commenters | | --- | --- | --- | --- | | Bittersweet-Diabetes | 21 | 154 | 71 | | Diabetesaliciousness | 28 | 102 | 95 | | InDpendence | 15 | 43 | 21 | | Ninjabetic | 14 | 60 | 46 | | The Perfect D | 17 | 110 | 53 | | Scott’s Diabetes | 12 | 96 | 59 | | Six Until Me | 44 | 654 | 255 | | Strangely Diabetic | 11 | 0 | 0 | | Sweetly Voiced | 9 | 89 | 54 | | Texting My Pancreas | 29 | 298 | 141 | | Six Until Me (2019) | 59 | 472 | 274 | | Diabetesaliciousness (2019) | 22 | 34 | 10 | | Total | 281 | 2112 | 1079 | Blog posts including comments were selected for inclusion based on a stratified, clustered, proportionate probability sampling strategy in which calendar days were randomly selected. Any blog posts on the randomly selected date were included in the data set, representing a cluster of content. This ensured a probability sample of all content from the ten blog sites, as blogs published more frequently had a higher likelihood of having published on any given date, and therefore a higher proportion of content included in the sample. ## Data Analysis Data were imported into a qualitative data management program (Nvivo 10). The primary coders (ELM, TKO) reviewed the data and made memos of initial codes to form a coding scheme (primary codes and sub-codes), which was revised by the study team. To assess inter-coder reliability, the primary coders coded $10\%$ of the data (kappa = 0.97). Coding proceeded past saturation for an additional $20\%$ of the dataset to confirm saturation. After coding, data were processed by “sifting”, which involved identifying and selecting the most essential data from the coded items [43]. The first stage of sifting sorted the data (codes) into two categories: one that related to the research question of emotional and practical support shared among T1D blog participants; and one that did not. The resulting blog data from the first stage were ‘sifted,’ into two primary analytic themes where blogs were used as a: (a) dynamic, interactional form of emotional support from others who understood diabetes from personal experience; and (b) source of sharing lived user experience of having diabetes, rather than as a source of communicating medical knowledge or facts about diabetes. Therefore, the thematic analysis describes how these blogs were used by PWT1D to exchange support. In order to ensure that the data from 2014 was relevant to the present time, another round of retrospective analysis was conducted between January 2019 and December 2019. Two active blog sites were chosen for the analysis: Six Until Me and Diabetesaliciousness. The year 2019 was chosen as it is the most recent complete year. All blog posts from the two sites were analyzed. ## Results There were approximately 2,470 blog posts within the initial 24-month sampling frame. The study team analyzed data for the first 50 randomly selected dates, and continually assessed for saturation as coding proceeded. After coding the initial 140 blog posts from the first 50 dates, saturation had not been achieved; therefore, additional randomly selected dates were analyzed. Twenty-seven [27] blog posts were excluded from analysis as irrelevant to the research question (e.g., not discussing diabetes, consisting only of a link to another site without original content, etc.). A total of 200 blog posts including 8 guest posts were included for analysis in the final sample of the 2014 data. These included variable numbers of comments, averaging 8.2 comments per blog post (range 0–38). Considering the age of the data, an additional 80 blog entries including 2 guest posts were included from 2019. Findings remain constant among the two samples. Table 2 lists the primary codes and sub-codes used in the thematic analysis, along with the number of occurrences and an illustrative quote. **Table 2** | Primary Codes | Secondary Codes, # of Occurrences, and Illustrative Quotes (in addition to quotes in the primary manuscript) | | --- | --- | | Shared ways of coping (570) | Humor and sarcasm (130) If we can laugh at something, we can own it, and that includes diabetes. I often find myself in a totally silly and very entertaining conversation that leaves me giggling long after I’ve shut down the computer. | | | T1D to T1D emotional support from DOC (363) Just by reading a blog post … the DOC automatically lifts my spirits, makes me smile and makes diabetes in all dimensions less challenging. And those some days become better days - And that’s so awesome! On the days when diabetes gets me down - It’s the [DOC] that pulls me up - Through blog posts, Instagram pictures, tweets, Facebook messages, texts or phone calls - And I am very grateful indeed~ | | | T1D to T1D informational support from DOC (41) It’s times like these that I’m so glad to have found the DOC. Because posts from blogs and Facebook began to flash through my mind. Someone used ketone sticks to test soda for sugar … Someone else tested soda on their meter. I finally learned that I use 10% less insulin in the first two weeks of my menstrual cycle than I do in the second two weeks. Now I have two different basal rate patterns that I switch every two weeks. It has made a big difference for whether or not I am low or high all the time. | | | T1D to T1D instrumental support from DOC (36) The DOC … real people with real day to day D issues and that was the only key I was lacking in the clinical supports I already had! The DOC has been a huge resource for me. I downloaded a copy of the Diabetes Health Magazine, it has coupons for glucose tabs and other supplies, and some recipes. It’s definitely worth checking out. | | | Affirmations (598) Thank you for your honesty, Karen. This was probably the most touching post I’ve read today. Hugs to you!! I am glad you punched through that writer’s block; this was a great post!! rage on, sometimes it’s all we can do. | | Shared user barriers to technology (222) | Impact of wearing technology and carrying it with you (184) The pump was snatched out of my hand last night when the infusion tubing snagged on a moving ceiling fan, was whipped around at high speed and thrown violently into the closet. I would bet that nobody in the quality department at Medtronic envisioned that scenario! I once had a sensor on my bum, right behind the back pocket. They kept saying “ma’am, I’ll ask you one more time- please remove everything from your pockets”…the joys of diabetes. | | | Intrusion of device alarms (17) Between waking up to treat lows, waking up to correct highs, the many beeps of the CGM that have disrupted my sleep, and even waking up to wonder if I’m high or low … it’s a lot of sleep I’ve been missing out on. I’ve taken my CGM off (for about 4 wks now) just to catch up on sleep lost from all the beep beep beeping. | | | Lack of usability (21) Among my top-5 nagging annoyances with diabetes is the air-bubble-problem. The one issue I did have was with the touch screen - I had to touch the screen three to five times before it would register the contact. | ## Theme 1. Blogs Were Used as a Dynamic, Interactional Form of Emotional Support From Others Who Understood Diabetes in the DOC PWT1D used words like “our” or “ours”, instead of “my” or “mine”, even though they were talking about their own personal experiences. For example, if one has to have a condition like diabetes that “tries to control us and our happiness but fails”, then it is best to give it a “smack in the face” by being healthy and taking care of it. The expression of personal experience was found in community: Advocacy regarding healthcare access and awareness regarding charities supporting T1D were constant features in the blogs: Emotional support and connection was an important part of the blog content. Participants affirmed each other frequently ($$n = 598$$), thanking an individual for sharing a response, or to give further encouragement. Instead of directing comments at an individual, bloggers and commenters in the DOC posted comments for the community-at-large to see. Participants believed that their medical team was helpful during medical appointments with diabetes, one even calling them “lifesavers”, but they did not lean on the medical team for emotional support. HCPs start by “asking me how my diabetes was … and he wanted that answered with an A1c result”. So PWT1D looked to the DOC for support. PWT1D used blogs to feel like part of a community that understood diabetes from the inside out. Sometimes they used “inside” humor or sarcasm that people who did not have diabetes, or did not live with someone who had diabetes, would not find as humorous. If we can laugh at something, we can own it, and that includes diabetes. One example is the often unpleasant taste of glucose tablets to treat hypoglycemia: But hey, we want to make sure the whole DOC can voice their opinions on glucose tablet flavors. He put together a clever and funny little survey that we hope everyone will take a few seconds to answer. It contains all of the brilliant, silly and downright gross suggestions, along with spaces for your own tab flavor creations. Are you game?? A PWT1D talked about meeting with other people who had diabetes, even though “diabetes is no party”, then said, “that’s a good enough reason to put on a funny hat and celebrate”. One person accidentally was in the wrong place at the wrong time, and a commenter said, “I am still laughing at the mental image of an insulin pump being whipped around by a ceiling fan! *That is* a new one!” to which the response was, “Good thing I wasn’t still connected to the tubing too, right? That could have been messy”. One person called the interactions with tubing in the restroom the “tubing tango”. When a stranger was looking at one PWT1D’s technology while they were waiting in line for coffee, the PWT1D “grabbed my coffee from the counter. I smiled. And I leaned in to whisper, I am not the droid you’re looking for”. Some of the humor was directed at food: *As a* person with Type 1 diabetes, there is not a food on this planet that I am not permitted to consume. ( There are many that you couldn’t pay me to try.) I didn’t feel even a little bit bad for ordering a glass of wine on my 9 am flight. Nope. A hypoglycemia tip was to avoid drinking juice when wearing a white dress shirt … “am I the only one who only buys apple juice boxes because of the lighter stain value?” Some issues were more serious, such as sarcasm directed toward things which cannot be immediately changed: Do they have a working model? Um. No. Have they figured out the anti-rejection issue? Um. No … But they do have a nifty bunch of ideas and a spiffy bunch of animations and pictures of people looking through microscopes. “ I’m sorry, which type of diabetes is the simple type?” I blurted out. “ Because I think we would all like to sign up for that one”. I got a few laughs, but seriously … A treatment plan may be simple from a clinician’s standpoint, but I guarantee you, to the patient, their disease is challenging and frustrating, whether they are asked to make lifestyle changes and take a pill or whether they are asked to struggle with variable basal rates and complicated medical devices. A quote from one blogger describes how humor can be a source of emotional support: It’s really, really hard to allow ourselves to be vulnerable with each other but I find that there is a certain empowerment that comes with allowing ourselves to let go of a little fear. Add in humor, and you’ve hit my sweet spot of emotional support. In summary, PWT1D are looking for emotional support and ways to feel understood. Each person has different needs, but as one person commented, “I’d be lost without my DOC”. Not everyone is interested in on-line activities; some want social settings, and some want presentations. One participant advised: It’s okay to move on and find a better fit—or start your ideal group yourself. Poke around online until you find the connections you are looking for. We all need different things, and over time the things we need often change and grow. ## Theme 2: Blogs Were Used as a Source of Sharing Lived User Experience of Having Diabetes, More Often Than as a Source of Communicating Medical Knowledge or Facts About Diabetes Blogs were a platform to communicate shared lived experience, typically not related to improving HbA1c or discussing complications. Blogs were not often used to seek or offer medical advice, such as reducing HbA1c, complications, or hyper/hypo-glycemia. One area discussed frequently was advances in technology, which were helpful in managing diabetes, but practically could be described as problematic. Many people in the DOC talked about how wearing diabetes technology impacted their lives and made them feel different. By wearing technology on the body, diabetes became more easily visible. It made the awareness of diabetes ever more present by being attached constantly to a device that is keeping the PWT1D alive. For some, the impact of wearing technology was a reason they delayed going on an insulin pump for the inconvenience of “having something attached to me all the time”. The benefits of wearing the pump generally outweighed the inconvenience, but it did take time getting used to wearing a device and responding to unsolicited questions. For many PWT1D, wearing a device meant that they received more attention (wanted and unwanted) from strangers, asking about their pumps or continuous glucose monitors (CGMs). This exposure could go either way, with PWT1D becoming more “open and comfortable” about their diabetes at times and for some, but other times and/or for other people, becoming “embarrassed” by having to wear a device. Another lived experience described was the impact of the alarms that come with pump and/or CGM user. For example, the alarms could draw unwanted attention to the pump or CGM user, such as “Mama? Your [CGM] has beeps!” Additionally, alarms often interrupted coveted sleep. Those without diabetes—the “other”—are able to sleep throughout the night without alarms, but those who have alarms lose sleep over them. This was an “insider” view that other people who did not wear technology could not fully understand: I was exhausted. I muffled the [CGM] under my pillow so the kid wouldn’t wake up. Can we count the dreams/nightmares about D in that number, too? I’ve taken my CGM off (for about 4 wks now) just to catch up on sleep lost from all the beep beep beeping. Although wearing a pump or CGM generated curiosity or attention from strangers, the familiar sound of a device from another person helped form an “instant connection when I see someone on the subway with their tubing sticking out”, as one blogger wrote. She continued, “I know we have something intimate and intense in common”. There is also the physical impact of wearing a pump, where the “weight and bulk” of the device can be inconvenient: Not everyone in the DOC sample of bloggers and commenters was a CGM/pump user and would not be able to comment from the role of an “insider”. One responded: I also wish that the community, as a whole, wouldn’t assign pumps as “necessary” for diabetes control. Pumps are a tool that I’m grateful we have available to us, but not using an insulin pump doesn’t equal out to “not trying hard enough”. Injections work really well for some people. Your diabetes may vary. Practical informational support came from lived experience that supplemented advice received in a medical office. There were accounts of how people solved problems in everyday life related to diabetes: *And this* is where a reader came in with a suggestion that saved my skin. She wrote, “You need to spray steroid nasal spray on the site after the alcohol or IV prep and before you insert”. She also attached a photo of a rash she received from a CGM and it looked just like mine. I travel frequently and always go through the metal detector with pump in one pocket, [CGM] receiver in another, and transmitter in my abdomen. After giving a disclaimer that one should seek medical advice, one commenter commiserated by talking about how he/she used a complicated formula to take insulin for sushi, which was a “problem food” for them. There was a comment about someone who put his/her CGM in a glass at night so they could avoid sleeping through the alarm. PWT1D talked about how to get through insurance hassles and how to override denial claims. Others talked about how they bolused or changed their insulin for exercise, but were quick to say that this was their personal experience with a dosing strategy, and it may not work for everyone. Participants also shared about developments of new clinical trials and their participation in them. This sparked discussion in the comments about various ongoing clinical trials and also brought hope to people. ## Discussion The primary finding of this study is that T1D blogs provide an exchange of emotional support in dealing with the heavy, unrelenting workload of living with and managing T1D. Blogs empower PWT1D to help each other and PWT1D are able to learn from the people who “get it”. It is interesting that the blogs did not mention much about how to improve HbA1c or prevent complications of diabetes. One possible reason for this is that these topics are discussed during visits with HCPs, and PWT1D can ask for clinical advice during those visits, leaving less need to pursue such topics with online peers. PWT1D peer support is often emotional in focus, and HCPs may often not be able to easily provide this, even if they want to, because they do not typically have access to the lived experience of having and managing diabetes. HCPs also may have less opportunity to focus on emotional support during appointments that are usually more clinical. Even when PWT1D peer support is informational, it is about lived experience, i.e. “tips and tricks” for managing the diabetes treatment regimen [44], rather than medical information about treatment procedures or efficacy. There may be some lessons for the healthcare community based on the value that blog-reading PWT1D seem to place on peer support, especially in light of how some note that their peers understand and “get it” better than their HCPs generally do. PWT1D may appreciate if this were simply acknowledged (e.g., that I, as your HCP may not fully understand your day-to-day efforts as well as a peer). They might also appreciate an HCP’s recommendation about where to find online peer support that is trustworthy, as well as sites that might be better to avoid. Even if HCPs cannot provide the same type of support as their peers, HCPs can help them navigate the world of online peer support. The findings of this study are supported by other research, both using social media and conventional data. Consider the topic of wearable technology. Prior to a recent publication [45], there was little published research that addressed the intrusiveness of wearing visible technology. Whereas that study was conducted via survey, the finding of similar results in the current naturalistic blog study of information obtained without prompts strengthens the finding itself, as well as the use of blogs as a source of data. More recently, another blog-based study found that there is a growing trend toward being “out and proud”, including the #showmeyourpump hashtag [44]. This is a testimony to the impact of social media in creating culture, specifically a culture that empowers and normalizes PWT1D. In addition, it indicates that monitoring of blog content allows researchers to stay abreast of rapidly developing culture around living with diabetes. ## Strengths and Limitations One strength of using blogs as a data source is that it is the ultimate naturalistic inquiry; it focuses on topics that are of importance to the participant, and it is unobtrusive in that the research does not influence the nature of the data. Moreover, there is no recall bias. There are no geographic boundaries in that persons can participate in the blog no matter where they live. However, a digital divide may exist in that some people (e.g. people who are older, people who are less educated) are less likely to access the Internet. Also, it is more likely that blog contributors are motivated people with diabetes; thus our results cannot be generalized to PWT1D who do not blog. In addition, it appeared as though many of the PWT1D bloggers were technologically familiar with insulin pumps and CGMs, which is not generalizable to low-end users of technology. We are able to capture data only from bloggers and commenters, and cannot know the experience among “lurkers” (participants who read the blogs but do not actively comment) by analyzing blogs and comments alone; doing so would require interviewing and/or surveying them [12]. Further research could explore the experience of lurkers through individual interviews, but that was not the purpose of this study. Also, it was not possible to have specific inclusion/exclusion criteria, or even to know the demographics, much less the identity, of participants; thus, it is not possible to examine how different types of people participate differently in blogging. The initial research was conducted in 2014. To ensure relevance, we conducted additional analysis in 2019. However, many blogs had stopped activity and some may have moved to other engaging online spaces such as Instagram. ## Conclusion This study represents a significant reversal of the usual paradigm of discovery and innovation by taking a belief held by a community of PWT1D (that blogs provide them with much-needed support), investigating it, and then bringing it to the healthcare community. More often, concepts come from within the scientific community, are disseminated to the broader clinical community, and then are eventually brought to PWT1D. Our research methodology contributes to the current movement in healthcare to become more patient-centered, in recognition of the primary role people have in managing their chronic conditions. In fact, blogs themselves may facilitate the empowerment and activation of PWT1D from passive to active participant, and from individual to community member [44]. Although not implemented in this paper, another possible use of diabetes blog data is to examine self-management barriers and facilitators. Blogs provide a unique window into patient-driven—rather than HCP-driven—concerns. As such, blogs may offer distinct advantages and economies not only for participants, but also for researchers and the healthcare system. The flexibility in timeliness (not just at scheduled appointment or group meeting times), convenience, and the ability of blogs to reach many individuals at the same time may improve outcomes and reduce costs. Researchers may also save time and money by using existing data sources as a foundation. These dual uses of blog data suggest that blogs are a growing part of health research and care. ## Data Availability Statement The data presented in the study are included in the article; further inquiries can be directed to the corresponding author. ## Ethics Statement The Internet ethnography (“netnography”) [41] methods used in this study were approved by the Penn State College of Medicine Institutional Review Board. Identified blog authors gave permission to retrospectively analyze data publicly available on their blogs posted between June 1, 2012 and July 31, 2014. ## Author Contributions HS was responsible for the draft of the first manuscript, ensuring the codebook, and inter-rater reliability was completed accurately and rigorously. SO reviewed and edited the drafts, discussed the coding and coding procedures, and provided clinical support for unhelpful or unsafe practices mentioned in the blogs. EM is the project manager for the study, whose role includes all administrative tasks as well as being the primary coder for the “first phase” of the data, and reviewed the manuscript for accuracy of the methodology. TO reviewed and edited the drafts, was the secondary coder of the data and ensured rigor of the data, and also provided clinical support. MFP contributed to the design of the stratified, clustered proportionate probability sample, and provided input into the organization of this manuscript. AS coded and analyzed the data from the “second phase” of the blogs and reviewed the final manuscript. All authors contributed to the article and approved the submitted version. ## Funding Research reported in this publication was supported by the National Institute of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health under Award Number DP3DK104054. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. ## Conflict of Interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer MMF declared a past co-authorship with several of the authors MP, HS to the handling editor. ## References 1. Coffen RD. **The 600-step program for type 1 diabetes self-management in youth: the magnitude of the self-management task**. *Postgrad Med* (2009) **121**. DOI: 10.3810/pgm.2009.09.2059 2. Greenhalgh T, Collard A, Campbell-Richards D, Vijayaraghavan S, Malik F, Morris J. **Storylines of self-management: narratives of people with diabetes from a multiethnic inner city population**. *J Health Serv Res Policy* (2011) **16** 37-43. DOI: 10.1258/jhsrp.2010.009160 3. Hempler NF, Joensen LE, Willaing I. **Relationship between social network, social support and health behaviour in people with type 1 and type 2 diabetes: cross-sectional studies**. *BMC Public Health* (2016) **16** 198. DOI: 10.1186/s12889-016-2819-1 4. Hill K, Ward P, Gleadle J. **“I kind of gave up on it after a while, became too hard, closed my eyes, didn’t want to know about it”—adults with type 1 diabetes mellitus describe defeat in the context of low social support**. *Heal Expect* (2019) **22**. DOI: 10.1111/hex.12850 5. Lee AA, Piette JD, Heisler M, Janevic MR, Rosland A-M. **Diabetes self-management and glycemic control: The role of autonomy support from informal health supporters**. *Heal Psychol* (2019) **38** 122. DOI: 10.1037/hea0000710 6. Nicolucci A, Kovacs Burns K, Holt RIG, Comaschi M, Hermanns N, Ishii H. **Diabetes Attitudes, Wishes and Needs second study (DAWN2**. *Diabetes Med* (2013) **30**. DOI: 10.1111/dme.12245 7. Stuckey HL, Mullan-Jensen CB, Reach G, Burns KK, Piana N, Vallis M. **Personal accounts of the negative and adaptive psychosocial experiences of people with diabetes in the second Diabetes Attitudes, Wishes and Needs (DAWN2) study**. *Diabetes Care* (2014) **37**. DOI: 10.2337/dc13-2536 8. Brady E, Segar J, Sanders C. **Accessing support and empowerment online: The experiences of individuals with diabetes**. *Heal Expect* (2017) **20**. DOI: 10.1111/hex.12552 9. Oser TK, Oser SM, McGinley EL, Stuckey HL. **A novel approach to identifying barriers and facilitators in raising a child with type 1 diabetes: qualitative analysis of caregiver blogs**. *JMIR Diabetes* (2017) **2** e27. DOI: 10.2196/diabetes.8966 10. Litchman ML, Rothwell E, Edelman LS. **The diabetes online community: older adults supporting self-care through peer health**. *Patient Educ Couns* (2018) **101**. DOI: 10.1016/j.pec.2017.08.023 11. Litchman ML, Walker HR, Ng AH, Wawrzynski SE, Oser SM, Greenwood DA. **State of the science: a scoping review and gap analysis of diabetes online communities**. *J Diabetes Sci Technol* (2019) **13**. DOI: 10.1177/1932296819831042 12. Oser TK, Oser SM, Parascando JA, Grisolano LA, Krishna KB, Hale DE. **Challenges and Successes in Raising a Child With Type 1 Diabetes and Autism Spectrum Disorder: Mixed Methods Study**. *J Med Internet Res* (2020) **22** e17184. DOI: 10.2196/17184 13. Oser TK, Oser SM, Parascando JA, Hessler-Jones D, Sciamanna CN, Sparling K. **Social Media in the Diabetes Community: a Novel Way to Assess Psychosocial Needs in People with Diabetes and Their Caregivers**. *Curr Diabetes Rep* (2020) **20** 1-10. DOI: 10.1007/s11892-020-1294-3 14. Oser SM, Stuckey HL, Parascando JA, McGinley EL, Berg A, Oser TK. **Glycated hemoglobin differences among blog-reading adults with type 1 diabetes compared with those who do not read blogs: cross-sectional study**. *JMIR Diabetes* (2019) **4** e13634. DOI: 10.2196/13634 15. Keim-Malpass J, Steeves RH, Kennedy C. **Internet ethnography: A review of methodological considerations for studying online illness blogs**. *Int J Nurs Stud* (2014) **51**. DOI: 10.1016/j.ijnurstu.2014.06.003 16. Hilliard EM, Sparling MK, Hitchcock J, Oser KT, Hood KK. **The emerging diabetes online community**. *Curr Diabetes Rev* (2015) **11**. DOI: 10.2174/1573399811666150421123448 17. Vygotsky LS. *Mind in society: The development of higher psychological processes* (1980). DOI: 10.2307/j.ctvjf9vz4 18. Dolan P. **Patients want to use social media tools to manage health care**. *Am Med News* (2012) **55** 19. Cooper A, Kar P. **A new dawn: the role of social media in diabetes education**. *J Diabetes Nurs* (2014) **18** 68-71 20. Farrell H. *Impact and evidence of social media use amongst people with diabetes* (2014) 21. Greene JA, Choudhry NK, Kilabuk E, Shrank WH. **Online social networking by patients with diabetes: a qualitative evaluation of communication with Facebook**. *J Gen Intern Med* (2011) **26**. DOI: 10.1007/s11606-010-1526-3 22. Shaffer-Hudkins E, Johnson N, Melton S, Wingert A. **Social media use among individuals with diabetes**. *Int J Commun Heal* (2014) **4** 38-43. DOI: 10.2196/diabetes.8603 23. Saric M. *Blogging statistics and trends you should know in 2020 [Internet]* (2020) 24. Sarasohn-Kahn J. *The wisdom of patients: Health care meets online social media* (2008) 25. Chaffey D. *Global social media research summary 2020 [Internet]* (2020) 26. McMahon KL. **Power and pitfalls of social media in diabetes care**. *Diabetes Spectr* (2013) **26**. DOI: 10.2337/diaspect.26.4.232 27. Das A, Faxvaag A. **What influences patient participation in an online forum for weight loss surgery? A qualitative case study**. *Interact J Med Res* (2014) **3** e4. DOI: 10.2196/ijmr.2847 28. Bair AL. *Blogging for dummies* (2019) 29. 29 JDFR. Type 1 Diabetes Facts [Internet]. (2020). Available at: https://www.jdrf.org/t1d-resources/about/facts/.. *Type 1 Diabetes Facts [Internet]* (2020) 30. Antheunis ML, Tates K, Nieboer TE. **Patients’ and health professionals’ use of social media in health care: motives, barriers and expectations**. *Patient Educ Couns* (2013) **92**. DOI: 10.1016/j.pec.2013.06.020 31. Patel R, Chang T, Greysen SR, Chopra V. **Social media use in chronic disease: a systematic review and novel taxonomy**. *Am J Med* (2015) **128**. DOI: 10.1016/j.amjmed.2015.06.015 32. Magnezi R, Bergman YS, Grosberg D. **Online activity and participation in treatment affects the perceived efficacy of social health networks among patients with chronic illness**. *J Med Internet Res* (2014) **16** e12. DOI: 10.2196/jmir.2630 33. Rinker J, Dickinson JK, Litchman ML, Williams AS, Kolb LE, Cox C. **The 2017 diabetes educator and the diabetes self-management education national practice survey**. *Diabetes Educ* (2018) **44**. DOI: 10.1177/0145721718765446 34. Warshaw H, Hodgson L, Heyman M, Oser TK, Walker HR, Deroze P. **The Role and Value of Ongoing and Peer Support in Diabetes Care and Education**. *Diabetes Educ* (2019) **45**. DOI: 10.1177/0145721719882007 35. Stuckey HL, Vallis M, Burns KK, Mullan-Jensen CB, Reading JM, Kalra S. **“I do my best to listen to patients”: qualitative insights into DAWN2 (diabetes psychosocial care from the perspective of health care professionals in the second diabetes attitudes, wishes and needs study)**. *Clin Ther* (2015) **37**. DOI: 10.1016/j.clinthera.2015.06.010 36. Peyrot M, Rubin RR, Siminerio LM. **Physician and nurse use of psychosocial strategies in diabetes care: results of the cross-national Diabetes Attitudes, Wishes and Needs (DAWN) study**. *Diabetes Care* (2006) **29**. DOI: 10.2337/dc05-2444 37. Snow R, Humphrey C, Sandall J. **What happens when patients know more than their doctors? Experiences of health interactions after diabetes patient education: a qualitative patient-led study**. *BMJ Open* (2013) **3** 4. DOI: 10.1136/bmjopen-2013-003583 38. Aziz Z, Riddell MA, Absetz P, Brand M, Oldenburg B, Investigators AP for PDP. **Peer support to improve diabetes care: an implementation evaluation of the Australasian Peers for Progress Diabetes Program**. *BMC Public Health* (2018) **18** 262. DOI: 10.1186/s12889-018-5148-8 39. Piette JD, Resnicow K, Choi H, Heisler M. **A diabetes peer support intervention that improved glycemic control: mediators and moderators of intervention effectiveness**. *Chronic Illn* (2013) **9**. DOI: 10.1177/1742395313476522 40. Riddell MA, Renwick C, Wolfe R, Colgan S, Dunbar J, Hagger V. **Cluster randomized controlled trial of a peer support program for people with diabetes: study protocol for the Australasian peers for progress study**. *BMC Public Health* (2012) **12** 843. DOI: 10.1186/1471-2458-12-843 41. Kozinets RV. **On netnography: Initial reflections on consumer research investigations of cyberculture**. *ACR North Am Adv* (1998) **25** 42. Johnson TP. *Snowball sampling: introduction* (2014). DOI: 10.1002/9781118445112.stat05720 43. Chowdhury MF. **Coding, sorting and sifting of qualitative data analysis: Debates and discussion**. *Qual Quant* (2015) **49**. DOI: 10.1007/s11135-014-0039-2 44. Tenderich A, Tenderich B, Barton T, Richards SE. **What are PWDs (people with diabetes) doing online? A netnographic analysis**. *J Diabetes Sci Technol* (2019) **13**. DOI: 10.1177/1932296818813192 45. Hood M, Wilson R, Corsica J, Bradley L, Chirinos D, Vivo A. **What do we know about mobile applications for diabetes self-management? A review of reviews**. *J Behav Med* (2016) **39**. DOI: 10.1007/s10865-016-9765-3
--- title: Validation of sociocultural attitudes towards appearance questionnaire and its associations with body-related outcomes and eating disorders among Chinese adolescents authors: - Houyi Huang - Zhongting Liu - Haoran Xiong - Fabian Herold - Jin Kuang - Erle Chen - Alyx Taylor - Albert Yeung - Jing Sun - Md M. Hossain - Arthur Kramer - Tianyou Guo - Liye Zou journal: Frontiers in Psychiatry year: 2023 pmcid: PMC10041934 doi: 10.3389/fpsyt.2023.1088769 license: CC BY 4.0 --- # Validation of sociocultural attitudes towards appearance questionnaire and its associations with body-related outcomes and eating disorders among Chinese adolescents ## Abstract ### Introduction The Sociocultural Attitudes Towards Appearance Questionnaire-4 Revised (SATAQ-4R) has been widely used in Western countries to link body appearance that is related to eating disorders and body dissatisfaction being commonly reported by adolescents. However, a comprehensive psychometric validation of the SATAQ-4R in Chinese adolescent samples is still lacking. To this end, the aim of the current study was to validate the gender-appropriate SATAQ-4R in a sample of Chinese adolescents, following by an investigation of its associations with body-related outcomes and eating disorder symptoms. ### Methods Two gender-specific studies were conducted to examine the psychometric properties of the SATAQ-4R-Female and SATAQ-4R-Male respectively among adolescent girls (Study1, $$n = 344$$, with 73 participants at retest) and boys (Study2, $$n = 335$$, with 64 participants at retest). Confirmatory factor analysis was employed to examine the factor structure and their test-retest reliability, the internal consistency and convergent validity were subsequently examined. ### Results For the SATAQ-4R-Females, the seven-factor model has a reasonable fit, with Chi-square =1112.769 ($p \leq 0.001$), CFI = 0.91, RMSEA = 0.071, SRMR = 0.067. For the SATAR-4R-Males, an acceptable seven-factor model with Chi-square = 982.92 ($p \leq 0.001$), CFI = 0.91, RMSEA = 0.08, SRMR= 0.06 was observed. With respect to test-retest reliability, the internal consistency for 7 subscales was rated as good (Cronbach’s alpha =0.74 to 0.95) among female adolescents, likewise the internal consistency of the seven subscales was also rated as good (Cronbach’s alpha =0.70 to 0.96) among male participants. Good convergent validity was observed, reflected by associations of the subscales of the gender-specific SATAQ-4R with muscularity-related attitude, body image-acceptance, body appearance, perceived stress level, symptoms of eating disorder and self-esteem. ### Discussion For women and men, the original 7-factor structure was validated among Chinese adolescents, internal reliability coefficients for the seven subscale scores were good and test-retest reliability was acceptable. Our results also confirmed the convergent validity of the two different gender-appropriate scales. ## Introduction Body dissatisfaction refers to persistent negative emotions and thoughts about his or her own body weight and physical appearance (1–4). Such body-related misconception is relatively common among Chinese adolescents, as a study reported that $72.7\%$ of Chinese middle-school students perceived themselves as fat, even though only $5.3\%$ of them were objectively overweight [5]. Another study observed that $47.2\%$ and $56.5\%$ of Chinese boys and girls, respectively, reported body dissatisfaction and were over-concerned about their body size [6]. Of note, the occurrence of body dissatisfaction has been linked to a higher risk of developing disordered eating patterns [7, 8], which is perhaps caused by excessive preoccupations with respect to the individual physical appearance [9]. In addition, Chinese adolescents who reported a higher level of shape- or weight-related concerns or dissatisfaction were linked to lower self-esteem and higher levels of depressive symptoms [10, 11], which, in turn, resulted in the increased prevalence of suicidal thought [12]. As a result, body dissatisfaction does not only negatively influence adolescent well-being [13], but also impacts health and social systems due to the collective burden of eating disorders and other comorbidities associated with body dissatisfaction [14]. Based on a wide range of negative consequences of body dissatisfaction among the affected individuals and communities, there is an urgent need to examine factors that contribute to shape- or body-related perceptions (i.e., body dissatisfaction). To assess the influence of sociocultural attitudes on the perception of adolescents’ physical appearance, the Sociocultural Attitudes Towards Appearance Questionnaire (SATAQ) was developed and has been widely used to investigate relationships between eating disorders and body image concerns. Since its first publication in 1997, the SATAQ has gone through several revisions to account for the shift in aesthetic standards and appearance ideals. Of note, the SATAQ-4 is the first version that includes both males and females [15]. The psychometric properties of the SATAQ have been confirmed in multiple countries, including eastern countries such as Japan whose culture, at least partly, resembles Chinese culture. Despite SATAQ-4 providing valid measures of the internalization of appearance ideals and appearance-associated pressures derived from peers, family, and media, there are some conceptual limitations (e.g., items about athleticism represented not only muscularity but also some physical competence such as coordination and agility, resulting in ambiguity in subscale score interpretation) that required further adjustment to provide a more precise and accurate assessment. Therefore, a newly adapted SATAQ-4R was developed and its advantages have been documented in a previous study [16]. The SATAQ-4R has been applied in a study [17], focusing on the validation of the Chinese version of the Acceptance of Cosmetic Surgery Scale (ACSS), in which six factors for male adults (Internalization of Muscularity, Internalization of Thinness, Internalization of General Attractiveness, Pressure from Family, Pressure from Peers and Significant Others, and Pressure from Media) and four factors for female adults (Internalization of Muscularity, Internalization of Thinness and General Attractiveness, Pressure from media) were supported. However, given the main focus of the above-mentioned study, a comprehensive psychometric validation of the SATAQ-4R in Chinese samples (e.g., adolescents) is still lacking. As the SATAQ-4R has not been appropriately validated in Chinese adolescents, further studies are needed since adolescence is a unique stage in individual development. To summarize, given the high prevalence of body dissatisfaction (i.e., misconception about physical appearance) and its significant impact on Chinese adolescents’ physical and mental health, effective tools to measure theoretical constructs that can explain factors (e.g., internalization of ideal appearance and sociocultural pressure) contributing to the development of body dissatisfaction are urgently required. Considering that there are gender-related differences in the field of body dissatisfaction, the SATAQ-4R which considers and accounts for these gender-related differences, is a more suitable instrument than previous versions of the questionnaire. Thus, the SATAQ-4R can be helpful for further research aiming to better understand the complex relationships between, for instance, disordered eating and body dissatisfaction. Thus, in this study, we translated and validated the Chinese version of the SATAQ-4R in a sample of Chinese adolescent girls and boys. The aim of the current study was to validate the SATAQ-4R in a sample of Chinese adolescents. Since the seven-factor model has been confirmed in different cultures [18], the original seven-factor model was examined in the current study. Based on previous studies, we expected the internalization of a thin ideal, internalization of muscularity ideal, and the internalization of general appearance would have a medium to large correlation with the body image acceptance, drive for muscularity, and body-esteem, respectively. In addition, we expected the pressure subscales would demonstrate a medium to large correlation with the perceived stress scale. Moreover, we expected that all the SATAQ-4R subscales demonstrate a medium to large correlation with the eating disorder symptoms and a small to medium correlation with self-esteem. In addition, as the SATAQ-4R is an instrument that is based on sociocultural theory and was previously created and validated mostly within western culture, we expected a lower fit for Chinese adolescents due to cultural differences. ## Study participants The participants of the current study were recruited from 3 high schools located in South China and data collection occurred from July 18th to August 5th, 2022. Of note, before data collection, a meeting with school counselors was conducted to ensure that students: [1] were able to understand the questionnaires; [2] had no psychiatric disorder like an eating disorder; [3] were not physically disabled or suffered from other types of physical illness that negatively affect his or her perception about his or her physical condition or body image. Given the nature of SATAQ-4R, female and male participants were asked to complete two respective scales with a different number of items (see Subsection of “Measures”), which generated gender-based samples for data analysis. For the pretest, the female sample consisted of 344 Chinese high school girls aged 13–18 years (15.9 ± 0.49 years; 20.35 ± 3.47 kg/m2), while the male sample consisted of 335 Chinese high school boys aged 15–18 years (16.04 ± 0.5 years; 21.02 ± 3.45 kg/m2). For a 1-week re-test, 73 female participants (15.88 ± 0.43 years; 20.21 ± 2.76 kg/m2) and 64 male participants (15.98 ± 0.45 years; 20.85 ± 3.83 kg/m2) volunteered to complete the SATAQ-4R, respectively. Demographic information is presented in Table 1. ## Procedure In the context of this study, following the standard procedures in our previous validation studies (19–21), the SATAQ-4R was translated into Chinese and back-translated into English. Prior to this translation, the authors of the original SATAQ-4R had granted their permission to do so. Specifically, two native Chinese speakers translated the SATAQ-4R items from English into the Chinese language. This step was followed by a meeting with the leading author (L.Y.Z) and 3 graduate students who were fluent in English, which generated a Chinese-language scale. To ensure the accuracy of the translated scale, a back translation was conducted by another Chinese-English bilingual scholar who was not involved in the study. The back-translated English version and the original version were carefully compared to create a prefinal Chinese version, which captured the complete meaning of the original English scale clearly and accurately. Ten high-school students were then invited to evaluate the readability of the translated scale and some minor feedback was received, which was used to develop the final version of the Chinese-language SATAQ-4R. As a part of the research project investigating validity and reliability of exercise-related instruments among Chinese samples, this study was approved by the Ethics Committee of Shenzhen University in China (NO. PN-2022-00026). Participants were informed that all data were kept confidential, and their responses were recorded anonymously. Demographic items were presented for completion first, followed by six self-reported scales including the SATAQ-4R, Drive for Muscularity Scale (DMS), Perceived Stress Scale (PSS), Eating Disorder Diagnostic Scale (EDDS), Body Image Acceptance and Action Questionnaire −5 (BIAAQ-5), Body Esteem Scale-Appearance (BES-appearance) and the Rosenberg Self-Esteem Scale (RSES). In addition, some participants voluntarily continued to the re-test of the SATAQ-4R about 1 week later. All data were collected via an online survey platform (Wenjuanxing). Three polygraph questions were inserted in the scales to rule out those participants who did not carefully read the items. ## Measures Two gender-appropriate SATAQ-4Rs were used to assess the internationalization of appearance ideals and appearance-related sociocultural pressures from peers, families, media, and significant others [16]. The SATAQ-4R-Female has 31 items within 7 sub-scales: [1] Internalization: thin/low body fat (four items, e.g., “I want my body to look very thin”); [2] Internalization: muscular (five items, e.g., “*It is* important for me to look muscular”); [3] Internalization: general attractiveness (six items, e.g., “I think a lot about my appearance”); [4] Pressures: family (four items, e.g., “I feel pressure from family members to look thinner”); [5] Pressures: peers (four items, e.g., “My peers encourage me to get thinner”); [6] Pressures: significant others (four items, e.g., “I feel pressure from significant others to improve my appearance”); and [7] Pressures: media (four items, e.g., “I feel pressure from the media to look in better shape”). Each item has 5 options of 1 (definitely disagree), 2 (mostly disagree), 3 (neither agree nor disagree), 4 (mostly agree), and 5 (definitely agree). Score on each sub-scale can be obtained as the sum of all items divided by the number of items, with higher mean scores indicating greater internalization or pressures. The SATAQ-4R-Male contains 28 items within 7 sub-scales/identical naming (see Supplementary Table 1). The Drive for Muscularity Scale (DMS) was used to assess participants’ desire for muscularity and engagement in behaviors and attitudes to achieve a muscular physique [22]. This scale consists of 15 items with two factors (e.g., “I wish I were more muscular,” “I think that my legs are not muscular enough”), with each item being rated on a 6-point response scale (0 = never to 6 = always). The score on each factor was calculated by the sum of all items divided by the number of all items, and a higher mean score indicates a greater level of motivation to acquire a muscular appearance. The DMS has been validated within Chinese female adults, demonstrating good internal consistency and validity [23]. The Chinese-speaking DMS was used in this study, Cronbach’s alpha in the current sample was high with 0.83. The Body Image Acceptance and Action Questionnaire −5 (BIAAQ-5) was used to assess participants’ flexible responses to body-related thoughts and feelings [24]. This scale contains 5 items (e.g., “I shut down when I feel bad about my body shape or weight,” “Feeling fat causes problems in my life”) with each response ranging from 1 (never true) to 7 (always true). All items were reverse-coded. The score was calculated by the sum of all items, with higher sum scores indicating a higher level of body image flexibility. The BIAAQ-5 was validated among Chinese undergraduates, with good internal consistency and validity [25]. The Chinese speaking BIAAQ-5 was used in this study, Cronbach’s alpha in the current sample was very high with 0.88. One sub-scale (namely BES-appearance) of the Body Esteem Scale (BES) was used to assess an individual’s general feeling about one’s appearance [26] This sub-scale contains 10 positively or negatively worded items (e.g., “I’m pretty happy about the way I look,” “There are lots of things I’d change about my looks if I could”). The respondents indicate their degree of agreement with the questions on a 5-point Likert scale ranging from 0 (never) to 4 (always). The score was calculated by the sum of all items. The BES-appearance was validated among Chinese adolescents, with good internal consistency and validity [27]. The Chinese-speaking BES-appearance was used in this study, Cronbach’s alpha of the BES-appearance in the current sample was high with 0.77. The Perceived Stress Scale (PSS) was used to assess an individual’s stress level during last month [28]. The PSS contains 10 items (e.g., “In the last month, how often have you felt nervous and stressed?,” “ In the last month, how often have you felt that things were going your way?”). The respondents indicate their degree of agreement on a 5-point Likert scale ranging from 0 (never) to 4 (very often). The score is calculated by the sum of all items. PSS was validated among Chinese adolescents, with good internal consistency and validity [29]. The Chinese-speaking PSS was used in this study, Cronbach’s alpha in the current sample was high with 0.83. The Eating Disorder Diagnostic Scale (EDDS) is a brief self-reported scale for capturing eating disorder symptoms [30]. The EDDS has 22 items, most of which are measured on a seven-point scale (0 = not at all to 6 = extremely), based on the past 3 months (e.g., “Over the past 3 months, have you felt fat?”). The EDDS also includes some Yes/No questions (e.g., “During these episodes of overeating and loss of control did you eat much more rapidly than normal? YES/NO”) and frequency assessments (e.g., “How many times per week on average over the past 3 months have you made yourself vomit to prevent weight gain or counteract the effects of eating?”). It yields a symptom-related composite score criteria for eating disorders. The EDDS had good reliability and validity in a large sample of male and female Hong Kong pupils [31]. The Chinese-speaking EDDS was used in the current study, Cronbach’s alpha for the EDDS-5 was high of 0.87. The Rosenberg Self-Esteem Scale (RSES) was used to assess global self-esteem and general feelings of self-worth. All 10 items are rated on a 4-point scale (1 = strongly disagree to 4 = strongly agree), with five items stating positive feeling and five items stating negative feeling. Score on items for negative feeling were firstly reversed and then add up to. The score was calculated by the sum of all items, and a higher sum score indicates a greater self-esteem. The Chinese-speaking RSES was translated and validated with good construct validity, internal consistency, and test–retest reliability [32]. The Chinese-speaking RSES was used in this study, Cronbach’s alpha in the current sample was very high with 0.89. ## Statistical analysis Data analysis was performed using Amos 28 (IBM SPSS, Chicago, United States). Given that the gender-appropriate scales were used across different countries and have been previously validated in a sample of Chinese adults, the female sample ($$n = 346$$) and the male sample ($$n = 344$$) were separately analyzed using the Confirmatory factor analysis (CFA) to evaluate internal consistency and convergent validity while maximum likelihood estimation was selected to test the seven-factor model presented in the original study [16]. Model fit indices were used to determine if this seven-factor model was acceptable among Chinese adolescents in terms of gender: [1] Chi-square test $p \leq 0.05$; [2] RMSEA = 0.05 to 0.08 (reasonable fit); [3] CFI = 0.90 to 0.95 (reasonable fit); [4] SRMR<0.1. With regard to each sub-scale of the two gender-appropriate SATAQ-4R, its acceptable internal consistency was assessed using Cronbach’s alpha >0.70 [33]. while convergent validity was tested using Pearson product–moment correlations, with r of 0.1 (small), 0.3 (medium), and 0.5 (large) [34]; Specifically, correlations between sub-scales of the gender-appropriate SATAQ-4R and other variables (i.e., drive for muscularity, self-esteem, perceived stress, general appearance feeling and eating disorder symptoms) were examined for convergent validity. The test–retest reliability for all sub-scales of the SATAQ-4R were assessed using the intraclass correlation coefficients (ICC) in terms of gender-based sample. Its cutoff was rated as follows: poor: < 0.40; fair: ≥ 0.40 and ≤ 0.59, good: ≥ 0.60 and ≤ 0.74, and excellent: ≥ 0.75 and ≤ 1.00 [34, 35]. ## Confirmatory factor analyses For the SATAQ-4R-Females, descriptive statistics were calculated for all items. For each item, skew was lower than 2 and kurtosis was lower than 7, suggesting normal distribution of all items. All items loaded onto their original factor, ranging from 0.37 (Item 11) to 0.97 (Item 27; See Supplementary Table 1). The seven-factor model has a reasonable fit, with Chi-square = 1112.769 ($p \leq 0.001$), CFI = 0.91, RMSEA = 0.071, SRMR = 0.067 (Figure 1). For the SATAR-4R-Males, all items were loaded strongly on their intended factor, ranging from 0.54 to 0.98 (see Supplementary Table 2), generating an acceptable seven-factor model with Chi-square = 982.92 ($p \leq 0.001$), CFI = 0.91, RMSEA = 0.08, SRMR = 0.06 (Figure 2). **Figure 1:** *Confirmatory factor analysis for female adolescents.* **Figure 2:** *Confirmatory factor analysis for male adolescents.* ## Means, internal consistency, and intercorrelation between sub-scales For the SATAQ-4R-Female, the mean score (ranging from 2.41 to 3.69) for sub-scales indicated a moderate level of ideal internalization and sociocultural pressure. The internal consistency for 7 subscales is rated as good, as indicated by Cronbach’s alpha =0.74 to 0.95 (see Supplementary Table 1). In addition, intercorrelations between every two sub-scales were significantly positive, except for the correlation between muscularity and general attractiveness ($r = 0.07$, $p \leq 0.05$) in terms of internalization (Table 2). For the SATAQ-4R-Male, the mean score for each sub-scale indicated that adolescent boys experienced a moderate level of appearance ideal internalization (ranging from 2.47 to 3.32) and low to moderate level of sociocultural pressures (ranging from 2.37 to 2.76). The internal consistency of the seven sub scales is rated as good (Cronbach’s alpha =0.70 to 0.96; see Supplementary Table 2). There were three non-significant correlations (i) between thin/low body fat-Internalization and general attractiveness-Internalization; (ii) general attractiveness-Internalization and family-Pressures and (iii) general attractiveness-Internalization and media-Pressures (See Table 3). **Table 3** | Unnamed: 0 | Mean (SD) | 1 | 2 | 3 | 4 | 5 | 6 | | --- | --- | --- | --- | --- | --- | --- | --- | | Internalization: Thin/Low Body Fat | 2.47 (0.94) | | | | | | | | Internalization: Muscular | 3.15 (0.86) | 0.12* | | | | | | | Internalization: General attractiveness | 3.32 (0.87) | 0.08 | 0.12* | | | | | | Pressures: Family | 2.37 (0.77) | 0.29** | 0.22** | 0.03 | | | | | Pressures: Peers | 2.76 (0.93) | 0.25** | 0.48** | 0.25** | 0.52** | | | | Pressures: significant others | 2.59 (0.96) | 0.25** | 0.36** | 0.16** | 0.50** | 0.62** | | | Pressures: Media | 2.29 (0.96) | 0.24** | 0.25** | 0.08 | 0.51** | 0.50** | 0.64** | ## Convergent validity For the SATAQ-4R-Female, except for muscularity, other sub-scales were: [1] positively (lightly to moderately) associated with eating disorder symptoms; [2] negatively (lightly to moderately) associated with body image flexibility; [3] negatively (lightly) associated with self-esteem. The Internalization: Muscular subscale demonstrated large associations with the variable drive for muscularity, and small associations with eating disorder symptomatology, but was not significantly associated with body image flexibility and self-esteem. The Pressures subscales presented small to medium positive associations with measures of perceived stress (see Table 4). For the SATAQ-4R-Male, the Internalization: Thin/Low Body Fat subscale exhibited a small positive correlation with body image flexibility. The Internalization: Muscular subscale exhibited large negative associations with drive for muscularity (assessed via DMS). The Internalization: General Attractiveness presented medium associations with body image flexibility. The Pressure: Peers and Pressure: Significant Others subscales presented small associations with measures of perceived stress (see Table 4). **Table 4** | Unnamed: 0 | Internalization | Internalization.1 | Internalization.2 | Internalization.3 | Internalization.4 | Internalization.5 | Pressures | Pressures.1 | Pressures.2 | Pressures.3 | Pressures.4 | Pressures.5 | Pressures.6 | Pressures.7 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | Thin/ Low Body Fat | Thin/ Low Body Fat | Muscular | Muscular | General attractiveness | General attractiveness | Family | Family | Peers | Peers | Significant others | Significant others | Media | Media | | | F | M | F | M | F | M | F | M | F | M | F | M | F | M | | DFM | −0.19** | −0.03 n.s | −0.56** | −0.67** | −0.11 n.s | −0.24** | −0.21** | −0.21** | −0.11* | −0.51** | −0.14** | −0.42** | −0.07 n.s | −0.29** | | BIAAQ | 0.53** | 0.29** | 0.10 n.s | 0.19** | 0.41** | 0.17** | 0.43** | 0.30**. | 0.51** | 0.35** | 0.49** | 0.35** | 0.52** | 0.38** | | BES-APP | .-0.35** | −0.19** | −0.06 | −0.05 n.s | −0.31** | −0.08 n.s | .-0.26** | −0.24** | −0.36** | −0.31** | 0.28** | −0.26** | −0.30** | −0.21** | | PSS | 0.29** | 0.09 n.s | 0.07 n.s | 0.26** | 0.34** | 0.05 n.s | 0.18** | 0.05 n.s | 0.29** | 0.17** | 0.26** | 0.24** | 0.30** | 0.10 n.s | | EDDS | 0.43** | 0.19** | 0.15** | 0.08 n.s | 0.33** | 0.13* | 0.31** | 0.16** | 0.33** | 0.22** | 0.34** | 0.25** | 0.33** | 0.31** | | RESE | −0.22** | −0.17** | −0.06 | 0.09 n.s | −0.24** | −0.08 n.s | −0.15** | −0.10 n.s | −0.23** | −0.15 | −0.22** | −0.13* | −0.20** | −0.04 | Consistent with our hypotheses (male participants), all the pressure-related subscales and Thin/low body fat-Internalization subscale demonstrated small to medium positive correlation with eating disordered symptoms and medium negative correlations with body image acceptances/flexibility, which is in line with findings in samples of American and Italian men samples [16, 18]. This agrees with the tripartite model suggesting that perceived sociocultural pressures and internalization of thinness contribute to negative eating and body image outcomes. With PSS, only two pressure-related subscales, peer-Pressures, and significant others-Pressures, demonstrated medium to large positive associations with perceived stress levels. Taken together, it appears that elevated perceived stress levels, increased eating disorder symptoms, and decreased body image acceptance may be caused by the fact that Asian people including Chinese are more concerned about (good) physical appearance (through very strict eating habits) as they are linked to career success in the modern society– naturally developing pressures from peers and significant others. In addition, only significant others-Pressures exhibited a medium negative association with self-esteem, which may be attributed to the unique cultural context of *China a* great emphasis on collectivism [43]. Such Chinese cultural collectivism linked to perceived stress could lead to reduced self-esteem. The female version of the SATAQ-4R has established good convergent validity among the sample of Chinese adolescent girls. As we have mentioned above, muscularity is more relevant to males than females. Chinese females have a lower motivation for muscularity due to different beauty standards and sociocultural impact in the Chinese culture, so that it seems reasonable to suggest that the absence of statistically significant correlations between internalization of muscularity and body image flexibility, body esteem, perceived stress, or self-esteem is related to the former (i.e., beauty ideal in China). However, a small correlation was observed between internalization of muscularity and eating disorder symptoms. Furthermore, muscularity-Internalization demonstrated a large positive association with a drive for muscularity, but it was not associated with eating disorder symptoms and self-esteem. Such results are different from previous studies focusing on American and Italian males [16, 18]. Unlike Western society which has an inveterate link between muscularity and masculinity, Chinese adolescent boys seem to have less exposure to exaggerated muscular body ideals and may not attach muscularity to individual popularity on campus [44], therefore, whether being muscular or not seems not to influence their eating behavior or self-esteem (i.e., based on the absence of correlations in the empirical data of the current study). General attractiveness-Internalization demonstrated a small positive association with eating disorder symptoms, but it failed to correlate with general appearance feelings. It is likely that the items of BES-appearance were perceived to assess facial attractiveness as opposed to overall body figure in the general-attractiveness-Internalization based on their respective contexts. Therefore, these two scales might evaluate two distinctive constructs due to their contextual differences, which might be the reason why they did not reach convergence. The validated version of SATAQ-4R would be valuable for advancing psycho-social health and well-being among Chinese adolescents emphasizing gender-specific challenges associated with body dissatisfaction. Given a high burden of body image disturbances and associated mental health crises in this population [45, 46], the findings of this study will allow researchers to use robust measurement approaches to assess and address such psycho-social problems. Also, future research evaluating the relevance of the current scale and other psychometric instruments may benefit from the findings of the current study in the given population context. The use of validated instruments is critical for evaluating and treating mental and behavioral health problems, which necessitates further research promoting context-specific estimations of psycho-social health problems for maximizing health and social outcomes across different populations. ## Discussion The current study evaluated the reliability and validity of the gender appropriate SATAQ-4R within the sample of Chinese adolescent girls and boys. Specifically, each of the subscales demonstrated good internal consistency and acceptable test–retest reliability. In addition, our results also confirmed its convergent validity of the two different gender-appropriate scales. Also, our study has clarified the structure of the SATAQ-4R among age the group between 15 and 17, which was not specifically validated in previous research. In the following, we will discuss our findings with further details. ## Factor structure Generally, this study provided evidence that the gender-specific SATAQ-4Rs are valid and reliable instruments that can be applied in samples of Chinese female and male adolescents. With respect to the factor structure, the results of the present study support the findings of the original study with US college students with a mean age of 20 years [16] and other culturally adapted scales among Italian adults with a mean age of 20 years [18] and among female college students in Turkey [36] in terms of gender-specific SATAQ-4R. On the other hand, other previous validation studies had also generated different factor structures. For example, another validation study demonstrated two 5-factor scales among Brazilian children aged 7–11 years [37] while a 6-factor structure for 225 male participants and a 4-factor structure for 308 male participants were observed among Chinese adults aged 18 to 68 years – of note, this non-validation study only reported Cronbach’s a and EFA of the SATAQ-4R (but no CFA-related fit indices and test–retest reliability) [17]. Taking the above-presented findings into account, it seems reasonable to hypothesize that age is a factor that may alter the factor structure of the SATAQ-4R. This assumption is supported by the fact that high-school students in China (mean age of 16 years) being studied in the current trial and western adults (mean age of 20 years) show a comparable factor structure [16], whereas in samples of Brazilian children and Chinese adults [38] (mean age of 9 years and 30 years, respectively) age-related effects may contribute to the differences being observed concerning the factor structure (6-factor and 4-factor structure, respectively). To verify or refute the assumption that the factor structure of the SATAQ-4R varies as a function of age, more high-quality studies following the standard validation procedures should be conducted among various age groups and health conditions. ## Internal consistency Overall, the results emerging from the female sample demonstrated a good internal consistency of the SATAQ-4R scale. Of note, muscularity-Internalization did not show a strong inter-correlation with other subscales of SATAQ-4R, ranging from 0.12 (media-Pressure) to 0.26 (significant others-Pressure). This finding is in line with the previous studies conducted in a different sociocultural context (e.g., western societies such as the United States) or within Asian culture [16, 18, 37, 39]. In addition, the correlation between muscularity-Internalization and general attractiveness-Internalization was not observed among female adolescents. Such a finding was expected as previous research has shown that muscularity is not the sociocultural feminine standard of attractiveness in the Chinese culture [40]. Moreover, there is evidence in the literature suggesting that Chinese females have a lower motivation to exercise and a lower level of internalization for muscularity comparison to females in western societies [41, 42]. The sociocultural ideal of female beauty that is commonly mentioned in the Chinese culture includes attributes such as being tall and curvy [40]. Taken together, muscularity does not play an important role in the sociocultural expectation of the ideal body image of Chinese females which might explain the non-significant correlation between these two subscales. Male-related results in the present study demonstrated acceptable internal consistency. In contrast to studies conducted within the Western culture, no correlation was observed between the subscales Internalization: general attractiveness and internalization of thin ideal, pressure from family, or pressure from media. However, this result is in line with the previous research conducted within Chinese adults [17], suggesting that males in China are less likely to be encouraged to evaluate their appearance by their family or on social media compared to females. ## Study limitations This study also has some limitations that need to be acknowledged and considered when interpreting our findings. The test–retest reliability assessment in the current study is based on rather small samples which, in turn, might somewhat limit the generalizability of these findings, although our results are encouraging as they suggest a good test–retest reliability of both versions of the SATAQ-4R (i.e., female and male versions). However, to address the latter limitation, future studies should seek to evaluate the test–retest reliability of both versions of the SATAQ-4R in larger samples with different age groups. ## Conclusion There is evidence in the literature suggesting that mental health conditions including eating disorders generally emerge among Chinese adolescents and young adults. Considering the social and health burden resulting from these mental health problems (mainly depressive disorders and eating disorders), the examination of sociocultural factors that influence the physical and mental health of Chinese adolescents is urgently needed. In this context, the findings of the current studies which provide evidence that the SATAQ-4R-Male and SATAQ-4R-Female have good psychometric properties to assess appearance-ideal internalization and appearance pressures in Chinese adolescent girls and boys, pave the way for future investigations on the appearance-ideal internalization among Chinese adolescents. The SATAQ-4R may also be used in future studies to test various theoretical models of the etiology of disordered eating, such as the tripartite influence model, within Chinese adolescents. Additionally, eating disorder treatments mainly aim to reduce the internalization of appearance ideals and address negative appearance-related pressures from sociocultural groups. Therefore, clinicians and researchers may consider the inclusion of the current measure (i.e., translated and validated version of the SATAQ-4R) to identify mechanisms of change in eating disorder prevention with patients or clinical research trials. ## Data availability statement The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. ## Ethics statement The studies involving human participants were reviewed and approved by the Ethics Committee of Shenzhen University in China (No. PN-2022-00026). As a part of the research project investigating validity and reliability of exercise-related instruments among Chinese samples. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin. ## Author contributions HH, ZL, HX, and LZ: conceptualization and formal analysis. HH, ZL, and HX: data curation, investigation, methodology, software, and writing – original draft. LZ: project administration, resources, and validation. JK and LZ: supervision. HH, ZL, HX, FH, JK, EC, AT, AY, JS, MH, AK, TG, and LZ: writing – review and editing. All authors contributed to the article and approved the submitted version. ## Funding This study was partially supported by Start-up Research Grant of Shenzhen University [20200807163056003] and Start-Up Research Grant (Peacock Plan: 20191105534C), the Natural Science Foundation of Guangdong Province of China (Grant no. 2018A030307002), the National Nature Science Foundation of China (Grant no. 31871115), and funding from the Shenzhen Humanities & Social Sciences Key Research Bases of the Center for Mental Health, Shenzhen University. ## Conflict of interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ## Publisher’s note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. ## Supplementary material The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpsyt.2023.1088769/full#supplementary-material ## References 1. Lucena-Santos P, Carvalho SA, da Silva OM, Pinto-Gouveia J. **Body-image acceptance and action questionnaire: its deleterious influence on binge eating and psychometric validation**. *Int J Clin Health Psychol* (2017) **17** 151-60. DOI: 10.1016/j.ijchp.2017.03.001 2. Quittkat HL, Hartmann AS, Düsing R, Buhlmann U, Vocks S. **Body dissatisfaction, importance of appearance, and body appreciation in men and women over the lifespan**. *Front Psychol* (2019) **10** 17. DOI: 10.1186/s41155-020-00149-6 3. Zou L, Yang P, Herold F, Liu W, Szabo A, Taylor A. **The contribution of BMI, body image inflexibility, and generalized anxiety to symptoms of eating disorders and exercise dependence in exercisers**. *Int J Ment Health Promot* (2022) **24** 811-23. DOI: 10.32604/ijmhp.2022.024862 4. Lucena-Santos P, Carvalho SA, Oliveira MDS, Pinto-Gouveia J. **Body-image acceptance and action questionnaire: its deleterious influence on binge eating and psychometric validation**. *Int J Clinical Health Psychol: IJCHP* (2017) **17** 151-60. DOI: 10.1016/j.ijchp.2017.03.001 5. Ro Y, Hyun W. **Comparative study on body shape satisfaction and body weight control between Korean and Chinese female high school students**. *Nutrition Res Prac* (2012) **6** 334-9 6. Li Y, Hu X, Ma W, Wu J, Ma G. **Body image perceptions among Chinese children and adolescents**. *Body Image* (2005) **2** 91-103. DOI: 10.1016/j.bodyim.2005.04.001 7. Keski-Rahkonen A, Mustelin L. **Epidemiology of eating disorders in Europe: prevalence, incidence, comorbidity, course, consequences, and risk factors**. *Current Opinion Psychiatry* (2016) **29** 340-5. DOI: 10.1097/YCO.0000000000000278 8. Vaquero-Cristóbal R, Alacid F, Muyor JM, López-Miñarro PÁ. **Body image; literature review**. *Nutr Hosp* (2013) **28** 27-35. DOI: 10.4162/nrp.2012.6.4.334 9. Jackson T, Chen H. **Predicting changes in eating disorder symptoms among Chinese adolescents: A 9-month prospective study**. *J Psychosom Res* (2008) **64** 87-95. DOI: 10.1016/j.jpsychores.2007.08.015 10. Chen G, Guo G, Gong J, Xiao S. **The association between body dissatisfaction and depression: an examination of the moderating effects of gender, age, and weight status in a sample of Chinese adolescents**. *J Psychol Couns Sch* (2015) **25** 245-60. DOI: 10.1007/s11031-014-9450-1 11. Li Y-P, Ma G, Schouten EG, Hu X, Cui Z, Wang D. **Report on childhood obesity in China (5) body weight, body dissatisfaction and depression symptoms of Chinese children aged 9–10 years**. *Biomed Environ Sci* (2007) **20** 11. PMID: 17458136 12. Hesketh T, Ding Q, Jenkins R. **Suicide ideation in Chinese adolescents**. *Soc Psychiatry Psychiatr Epidemiol* (2002) **37** 230-5. DOI: 10.1007/s00127-002-0536-9 13. Griffiths S, Hay P, Mitchison D, Mond JM, McLean SA, Rodgers B. **Sex differences in the relationships between body dissatisfaction, quality of life and psychological distress**. *Aust N Z J Public Health* (2016) **40** 518-22. DOI: 10.1177/0269881116675513 14. Samnaliev M, Noh HL, Sonneville KR, Austin SB. **The economic burden of eating disorders and related mental health comorbidities: An exploratory analysis using the US Medical Expenditures Panel Survey**. *Prev Med Rep* (2015) **2** 32-4. DOI: 10.1016/j.pmedr.2014.12.002 15. Schaefer LM, Burke NL, Thompson JK, Dedrick RF, Heinberg LJ, Calogero RM. **Development and validation of the sociocultural attitudes towards appearance questionnaire-4 (SATAQ-4)**. *Psychol Assess* (2015) **27** 54. DOI: 10.1037/a0037917 16. Schaefer LM, Harriger JA, Heinberg LJ, Soderberg T, Kevin Thompson J. **Development and validation of the sociocultural attitudes towards appearance questionnaire-4-revised (SATAQ-4R)**. *Int J Eat Disord* (2017) **50** 104-17. DOI: 10.1002/eat.22590 17. Wu Y, Alleva JM, Mulkens S. **Factor analysis and psychometric properties of the Chinese translation of the acceptance of cosmetic surgery scale**. *Body Image* (2020) **33** 244-56. DOI: 10.1016/j.bodyim.2020.03.009 18. Stefanile C, Nerini A, Matera C, Schaefer LM, Thompson JK. **Validation of an Italian version of the sociocultural attitudes towards appearance questionnaire-4-revised (SATAQ-4R) on non-clinical Italian women and men**. *Body Image* (2019) **31** 48-58. DOI: 10.1016/j.bodyim.2019.08.005 19. Kuang J, Zhong J, Yang P, Bai X, Liang Y, Cheval B. **Psychometric evaluation of the inventory of dimensions of emerging adulthood (IDEA) in China**. *Int J Clin Health Psychol* (2023) **23** 100331. DOI: 10.1016/j.ijchp.2022.100331 20. Wang T, Cheval B, Maltagliati S, Zenko Z, Herold F, Ludyga S. *Validation of the Chinese version of the affective exercise experiences questionnaire (AFFEXX-C)* (2022a) 21. Wang T, Kuang J, Herold F, Taylor A, Ludyga S, Zhang Z. **Validity and reliability of the preference for and tolerance of the intensity of exercise questionnaire among Chinese college students**. *Int J Ment Health Promot* (2022b) **25** 1-12. DOI: 10.32604/ijmhp.2022.022504 22. McCreary DR, Sasse DK, Saucier DM, Dorsch KD. **Measuring the drive for muscularity: factorial validity of the drive for muscularity scale in men and women**. *Psychol Men Masculinity* (2004) **5** 49. DOI: 10.1037/1524-9220.5.1.49 23. He J, Murray S, Compte EJ, Song J, Nagata JM. **The muscularity-oriented eating test, drive for muscularity scale, and muscle dysmorphic disorder inventory among Chinese men: Confirmatory factor analyses**. *Int J Environ Res Public Health* (2021a) **18** 11690. DOI: 10.3390/ijerph182111690 24. Basarkod G, Sahdra B, Ciarrochi J. **Body image–acceptance and action questionnaire–5: An abbreviation using genetic algorithms**. *Behav Ther* (2018) **49** 388-402. DOI: 10.1016/j.beth.2017.09.006 25. He J, Cai Z, Chen X, Lu T, Fan X. **Validation of the Chinese version of the body image acceptance and action questionnaire and the mediating role of body image flexibility in the relationship between body dissatisfaction and psychological distress**. *Behav Ther* (2021b) **52** 539-51. DOI: 10.1016/j.beth.2020.07.003 26. Mendelson BK, Mendelson MJ, White DR. **Body-esteem scale for adolescents and adults**. *J Pers Assess* (2001) **76** 90-106. DOI: 10.1207/S15327752JPA7601_6 27. Yang X, Huang B, Wong KM. **Prevalence and socio-demographic, anthropometric, and cognitive correlates of internet gaming disorder among children in China**. *Child Youth Serv Rev* (2021) **122** 105893. DOI: 10.3390/ijerph17020579 28. Cohen S, Kamarck T, Mermelstein R. **A Global Measure of Perceived Stress**. *Journal of Health and Social Behavior* (1983) **24** 385-396. DOI: 10.2307/2136404 29. Liu X, Zhao Y, Li J, Dai J, Wang X, Wang S. **Factor structure of the 10-item perceived stress scale and measurement invariance across genders among Chinese adolescents**. *Front Psychol* (2020) **11** 537. DOI: 10.3389/fpsyg.2020.00537 30. Stice E, Telch CF, Rizvi SL. **Development and validation of the Eating Disorder Diagnostic Scale: a brief self-report measure of anorexia, bulimia, and binge-eating disorder**. *Psychol Assess* (2000) **12** 123. PMID: 10887758 31. Lee SW, Stewart SM, Striegel-Moore RH, Lee S, Sy H, Lee PW. **Validation of the eating disorder diagnostic scale for use with Hong Kong adolescents**. *Int J Eat Disord* (2007) **40** 569-74. DOI: 10.1002/eat.20413 32. Chen F, Bi C, Han M. **The reliability and validity of the Chinese version of the revised-positive version of Rosenberg self-esteem scale**. *Adv Psychol* (2015) **05** 531-5. DOI: 10.12677/AP.2015.59068 33. Blunch N. *Introduction to structural equation modeling using IBM SPSS statistics and AMOS* (2012) 34. Cohen J. *Statistical power analysis for the behavioral sciences* (2013) 35. Zhu W. *J Sport Health Sci.* (2016) **5** 77-79. DOI: 10.1016/j.jshs.2016.01.019 36. Cihan B, Bozo Ö, Schaefer LM, Thompson JK. **Psychometric properties of the sociocultural attitudes towards appearance Questionnaire-4-Revised (SATAQ-4R) in turkish women**. *Eat Behav* (2016) **21** 168-71. DOI: 10.1016/j.eatbeh.2016.03.003 37. Neves CM, Meireles JFF, FFR M, ACS A, MEC F. **Translation, adaptation and psychometric properties of SATAQ-4R for Brazilian Children**. *Psicologia: Reflexão e Crítica* (2020) **33** 12. DOI: 10.1186/s41155-020-00149-6 38. Mehta CM, Arnett JJ, Palmer CG, Nelson LJ. **Established adulthood: A new conception of ages 30 to 45**. *Am Psychol* (2020) **75** 431. PMID: 32378940 39. Yamamiya Y, Shimai S, Schaefer L, Thompson J, Shroff H, Sharma R. **Psychometric properties and validation of the Sociocultural Attitudes Towards Appearance Questionnaire-4 (SATAQ-4) with a sample of Japanese adolescent girls**. *Body image* (2016) **19** 89-97. DOI: 10.1016/j.bodyim.2016.08.006 40. Leung F, Lam S, Sze S. **Cultural expectations of thinness in Chinese women**. *Eat Disord* (2001) **9** 339-50. DOI: 10.1080/106402601753454903 41. Tang C, Barnhart WR, Zhang B, He J. **Psychometric properties of the Chinese version of the female muscularity scale among Chinese young women**. *Eating Weight Disord-Stud Anorexia, Bulimia Obesity* (2022) **27** 3187-96. DOI: 10.1007/s40519-022-01448-w 42. Yan Z, Cardinal BJ. **Perception of physical activity participation of Chinese female graduate students: A case study**. *Res Q Exerc Sport* (2013) **84** 384-96. DOI: 10.1080/02701367.2013.813895 43. Chiu RK, Kosinski FA. **Chinese cultural collectivism and work-related stress: Implications for employment counselors**. *J Employ Couns* (1995) **32** 98-110. DOI: 10.1002/j.2161-1920.1995.tb00980.x 44. Jung J, Forbes GB, Chan P. **Global body and muscle satisfaction among college men in the United States and Hong Kong-China**. *Sex Roles* (2010) **63** 104-17. DOI: 10.1007/s11199-010-9760-z 45. Zhang Y, Liu B, Sun L. **Association between subjective body image, body mass index and psychological symptoms in Chinese adolescents: a Nationwide cross-sectional study**. *Health* (2021) **9** 1299. DOI: 10.3390/healthcare9101299 46. Tie B, Tang C, Zhu C, He J. **Body dissatisfaction and non-suicidal self-injury among Chinese young adults: a moderated mediation analysis**. *Eating Weight Disord-Stud Anorexia, Bulimia Obesity* (2022) **27** 2051-2062. DOI: 10.1037/amp0000600
--- title: 'Toward reliable calcification detection: calibration of uncertainty in object detection from coronary optical coherence tomography images' authors: - Hongshan Liu - Xueshen Li - Abdul Latif Bamba - Xiaoyu Song - Brigitta C. Brott - Silvio H. Litovsky - Yu Gan journal: Journal of Biomedical Optics year: 2023 pmcid: PMC10042069 doi: 10.1117/1.JBO.28.3.036008 license: CC BY 4.0 --- # Toward reliable calcification detection: calibration of uncertainty in object detection from coronary optical coherence tomography images ## Abstract. ### Significance Optical coherence tomography (OCT) has become increasingly essential in assisting the treatment of coronary artery disease (CAD). However, unidentified calcified regions within a narrowed artery could impair the outcome of the treatment. Fast and objective identification is paramount to automatically procuring accurate readings on calcifications within the artery. ### Aim We aim to rapidly identify calcification in coronary OCT images using a bounding box and reduce the prediction bias in automated prediction models. ### Approach We first adopt a deep learning-based object detection model to rapidly draw the calcified region from coronary OCT images using a bounding box. We measure the uncertainty of predictions based on the expected calibration errors, thus assessing the certainty level of detection results. To calibrate confidence scores of predictions, we implement dependent logistic calibration using each detection result’s confidence and center coordinates. ### Results We implemented an object detection module to draw the boundary of the calcified region at a rate of 140 frames per second. With the calibrated confidence score of each prediction, we lower the uncertainty of predictions in calcification detection and eliminate the estimation bias from various object detection methods. The calibrated confidence of prediction results in a confidence error of ∼0.13, suggesting that the confidence calibration on calcification detection could provide a more trustworthy result. ### Conclusions Given the rapid detection and effective calibration of the proposed work, we expect that it can assist in clinical evaluation of treating the CAD during the imaging-guided procedure. ## Introduction Optical coherence tomography (OCT) can acquire high-resolution cross-sectional images of coronary arteries. The high-quality and detailed information from coronary OCT images facilitates the treatment of coronary artery disease (CAD). CAD causes 1 of every 5 deaths in Europe1 and 1 of every 6 deaths in the United States,2 and it remains one of the leading causes of morbidity and mortality in developed countries.3 *Coronary atherosclerosis* is caused by the gradual buildup of plaque resulting from the depositing of calcium, lipids, and macrophages from the luminal blood into the arterial intima. Coronary atherosclerosis compounds and augments the risks of heart attack and heart failure. When treated improperly or left unattended, coronary atherosclerosis blocks the pathways to the heart’s main arteries, known as the coronary arteries. The potential effects of plaque in CAD include chest pain, shortness of breath, heart failure, myocardial infarction, and sudden death. A typical treatment for CAD is percutaneous coronary intervention (PCI), which is a nonsurgical procedure used to treat the narrowing of the heart’s coronary arteries. Unidentified calcified tissues within a narrowing artery often negatively impact the benefits of treatment. Approximately 700,000 PCIs are performed every year in the United States, and calcifications have been found in $17\%$ to $35\%$ of patients undergoing the procedure,4–6 highlighting a need to precisely locate the existence and extent of calcifications. Most PCI procedures involve using stents to open up obstructed coronary arteries.7 During the PCI procedure, a catheter with a tiny, folded balloon on its tip is inserted into the blood vessels until it arrives at the site where the plaque buildup is causing a blockage. At that point, the balloon is inflated to compress the plaque against the artery walls, therefore widening the passageway and restoring blood flow to the heart. After that, the balloon is deflated and removed. A stent implantation is performed in the plaque buildup area to keep the artery open after removing the balloon.8 Excess coronary calcification is highly related to the suboptimal deployment of the stent in the coronary during the PCI.9 Major calcifications are of great concern for two reasons.10 Calcifications can lead to stent underexpansion and strut malapposition. Malapposition of stent struts (e.g., an empty space between the strut and the adjacent vessel wall) might preclude healthy endothelial tissue growth. Even though stent deployment is generally effective in the short term, stent efficacy can be reduced and the risk can be increased by adverse clinical events, such as in-stent restenosis and thrombosis in the medium- and long-term.11–17 Coronary imaging guidance during PCI is one of the key determinants of treatment outcomes. Imaging is integral to every stage of PCI, such as assessment of lesion severity, preprocedural planning, optimization, and management of immediate complications.18,19 OCT has significant advantages for characterizing coronary calcification that typically has a signal-poor area with sharply delineated borders.20 A typical OCT system can achieve a high axial resolution at the micron level and a penetration depth of up to 2 mm, indicating superior imaging capability.21,22 The detection of calcified regions within coronary OCT images is critical for intervention.23 On account of this, developing an object detection algorithm that is capable of detecting calcification in OCT images is essential. Deep learning has been increasingly explored in analyzing the diseased tissue in coronary OCT images.24 In existing research works,25–30 extensive studies have been conducted to automatically identify plaque in coronary OCT images. A weighted majority voting from different convolutional neural networks (CNN)26 was used to solve the multiclass classification problem of pathological formations in coronary artery tissues. A deep convolutional architecture named SegNet segmented calcification in coronary OCT images.10 A two-step deep learning approach27 characterized plaques in coronary arteries in OCT images by first localizing the major calcification lesions using the CNN model and then applying the deep learning model (SegNet) to provide pixel-wise classifications of calcified plaques. A modified deep convolutional segmentation model UNet28 was used to identify calcification in coronary OCT images. The segmentation module in MASK-RCNN was employed to identify the erosion region.31 Currently, the most popular way to perform automated analysis on OCT images is deep learning-based segmentation, which makes the pixel-wise classification and outputs the detailed shape and location of the tissue of interest. Demonstrably, the segmentation architecture results in large computational costs due to the burden of pixel-wise classification. By virtue of this, a more efficient way of enacting automated analysis of coronary images is through the use of object detection, which outputs the bounding box of the tissue region rather than the pixel-wise classification of the tissue region, to efficiently identify the diseased region in coronary images. Although existing works also focus more on increasing the accuracy of deep learning models, the quality of predictions can be negatively impacted by overconfident deep learning models.32 The problem of overconfidence can be produced by deep learning models in the form of providing high confidence scores for predictions.33–35 *In* general, recalibration methods of the well-trained model, such as Platt scaling,36 histogram binning,37 and temperature scaling,38 can improve the calibration of the overconfident prediction results. In addition, model ensemble methods39,40 can also reduce overconfidence by aggregating the prediction results over multiple models. However, there are limited studies on correcting overconfident predictions in coronary OCT images. In OCT-related CAD treatment, overconfidence could be dangerous as confidence is often learned as the likelihood that the prediction is correct. Therefore, in safety-critical and risk-sensitive applications in clinical diagnosis, it is crucial to quantify and calibrate the uncertainty of predictions. In this work, we aim to achieve reliable calcification detection for patients with CAD to boost the efficiency of clinical diagnosis. We summarize our contributions as follows. ## Methods The workflow is shown in Fig. 1. The steps are as follows: [1] the coronary OCT data are first processed by a data augmentation module to create motion-blurred and horizontally flipped copies of each original OCT image. [ 2] The coronary OCT data after augmentation are trained by deep learning object detection models, and the detection results on test data are output. [ 3] Detections containing bounding box coordinates and confidence scores are processed through dependent logistic calibration, and a calibrated confidence score is output for each predicted bounding box. **Fig. 1:** *Flowchart of the proposed work. Scale bar: 500  μm.* ## Data Collection Samples are imaged by the spectral domain OCT system (Thorlabs Ganymede, Newton, New Jersey, United States) with an axial resolution of 3 μm and a lateral resolution of 4 μm in air. Autopsy specimens of human heart vessels are collected and imaged through the same protocol given in Refs. 41 and 42. All images are acquired in the laboratory at the University of Alabama. ## Data Augmentation Various data augmentation techniques have been proposed to improve the performance of deep learning models.43 During imaging, the quality of OCT images may be impacted due to degradation caused by motion blur,44,45 which can be caused by sample and device movement.46–48 A motion blur filter is used to simulate this effect of real-world conditions.49 Other common augmentation strategies, such as flipping, cropping, scaling, Gaussian noise, rotation, and shears, are routinely performed.50 Noticeably, we do not prefer vertical flipping or rotation in OCT images because the light propagates in a fixed direction, and applying such methods will change the nature of OCT images. Therefore, in this work, two copies of each OCT image are created by applying a motion-blurring filter and flipping horizontally for training the deep learning model. ## Object Detection Object detection creates bounding box regions that identify an object’s position, size, and class within an image. We opt to use You-Only-Look-Once v5 (YOLO)51 to rapidly identify the bounding box and tissue types within an OCT image. Because of its lightweight and feature-reuse properties, the YOLO architecture is powerful at realizing fast and accurate detection. As the conceptual schematic shown in Fig. 2, to better predict objects of different sizes, YOLO enhances the detection performance by utilizing different scales of feature maps that are generated by applying filters to the input image or the feature map output of the prior layers. **Fig. 2:** *Schematic of YOLO object detector and calibration.* The predictions have the outputs in two branches: confidence scores (confidence) and bounding boxes (xcenter, ycenter, width, and height). In the confidence score branch, the confidence score indicates a certain level that the prediction is true. In the bounding boxes branch, the values of the center coordinate, together with the width and height of the bounding box, depict the location of the predicted bounding box. To evaluate the performance of calcification detection, three metrics, precision, recall, and f1-score, are calculated, as given in the following eqautions: precision=TPTP+FP,[4]recall=TPTP+FN,[5]f1-score=2×precision×recallprecision+recall,[6]where the true positive (TP) means the model correctly predicts the region with calcification, the false negative (FN) is the wrong prediction for the region that has calcification, and the false positive (FP) is the wrong prediction for the region with no calcification. Precision indicates the number of correct predictions among all detections. Recall measures the fraction of correct predictions among ground truths. The f1-score is a measure of overall model performance determined by combining precision and recall. Qualitatively, in Fig. 3, YOLO predicts all calcification in this coronary OCT image. The SSD and Faster RCNN fail to detect the calcification region in relatively lower contrast. The low recall of the SSD and Faster RCNN in Fig. 4 reveals higher FN predictions, which agrees with the observation in Fig. 3. **Fig. 3:** *Example of (a) ground-truth label, (b) corresponding histology, and object detection results from (c) Faster RCNN, (d) SSD, and (e) YOLO. Scale bar: 500  μm.* **Fig. 4:** *Object detection results of deep learning models with a threshold of 0.4 in precision, recall, and f-1 score. The gray bars are the results of Faster RCNN, the blue bars are the results of the SSD, and the green bars are the results of YOLO.* In addition, as shown in Table 1, the processing speed of YOLO is 140 frames per second (fps), showing that YOLO has great capability for real-time detection, which is especially desirable in the circumstance of processing a large volume of OCT images. The runtimes of the SSD and Faster RCNN are 68 and 35 fps, respectively. The runtime of OCT segmentation of DeepRetina58 and CNN-S59 is ∼5 fps, which indicates a larger computational burden than detection. **Table 1** | Unnamed: 0 | Faster RCNN | SSD | YOLO | | --- | --- | --- | --- | | Runtime (fps) | 35 | 68 | 140 | ## Uncertainty Measurement and Confidence Calibration The common use of calibration is for the classification task, in which only the confidence score is utilized for a given image. In object detection, one additional piece of information that can be included for calibration is the location and scale of the bounding box. Therefore, in the object detection task, the criterion of a calibrated model is defined as the precision of a prediction given the confidence, class category, and bounding box information,52 as in the following equation: P($z = 1$|p=^conf,y=^y,r=^r)=conf,∀ conf∈[0,1], y∈Y, r∈[0,1]K,[1]where $z = 1$ indicates that the prediction is correct, conf denotes the confidence of prediction, y is the predicted class in the set of all classes denoted by Y, and r is the bounding box information with k dimensions. The expected calibration error (ECE) is used to measure the uncertainty of the prediction of the deep learning model. The ECE of object detection is calculated by binning the confidence p^ into M equally spaced bins. Samples with different confidence scores fall into corresponding bin m. Bm is the number of samples in a bin, and N is the number of total samples. The *Prec is* the precision that represents the correct predictions among all predictions as defined in Eq. [ 4], and the conf denotes the average confidence score of the predictions. The ECE is given by ECE=∑$m = 1$M|Bm|N|Prec(m)−conf(m)|.[2] For confidence calibration, in this work, we take two additional bounding box pieces of information, the center-x and center-y positions, along with the confidence score to calibrate the prediction results using the dependent logistic calibration,52 with the multivariate probability density function being used to model the log-likelihood ratio (lr) of the combined input(confidence,bounding box). Taking the correlations between the confidence and bounding box into consideration, the calibration map is defined as g and is given as g(input)≈11+e−lr(input),lr(s)=12[(s−TΣ−−1s−)−(s+TΣ+−1s+)]+c.[3] For the variables, s+=s−μ+ and s−=s−μ−, and c=log|Σ−Σ+|, with μ+ and μ− as the mean vectors and Σ− and Σ+ as the covariance matrices for the incorrect and correct predictions, respectively. As shown in the calibrated predictions block in Fig. 2, a new confidence score for each prediction is obtained by mapping the input to the calibration map g. The ECEs of the prediction results before and after calibration are calculated to test the effect of calibration on model uncertainty. We evaluate the effectiveness of the calibration of predictions for the deep learning models. In Fig. 5, the adjustment of confidence scores is observed in the calibrated predictions. In Figs. 5(c) and 5(d), the predictions from Faster RCNN and the SSD in the red box show that the confidence score is slightly adjusted. In Fig. 5(e), the overconfident predictions shown in the yellow box reduces the confidence score from $46\%$ to $18\%$ after calibration, whereas the other confidence scores of predicted boxes are slightly adjusted. **Fig. 5:** *Example result: (a) ground-truth label, (b) corresponding histology, and confidence calibration results of (c) Faster RCNN, (d) SSD, and (e) YOLO. Scale bar: 500  μm.* For quantitative evaluation, we use the ECE to measure the uncertainty using the mean value of all test targets. In Table 2, before calibration, YOLO has a lower level of uncertainty in ECE, indicating that YOLO produces more reliable predictions. For the three deep learning models, the calibration errors are lowered to the same level around ∼0.14 after calibration, which shows the effectiveness of the calibration process that helps rectify the overconfident predictions. **Table 2** | Unnamed: 0 | Faster RCNN | SSD | YOLO | | --- | --- | --- | --- | | ECE | 0.429 | 0.731 | 0.233 | | Calibrated ECE | 0.151 | 0.134 | 0.146 | | Before/after calibration | 0.278 | 0.585 | 0.099 | ## Experimental Setup For model development, we use 943 OCT images from 14 OCT specimen segments for a threefold cross validation. The OCT images were acquired from specimens that contain calcification regions, which include essential information for CAD treatments. Within each OCT volume, B-scans were sampled at an interval of 20 B-scans. Each B-scan has a size of 1024 × 1500 pixels, corresponding to a space of 1.98×3 mm2. In the confidence calibration stage, $60\%$ predictions are used to fit the calibration model, with the remaining $40\%$ predictions to be tested. The ground-truth annotations used for training and testing were made under the guidance of the pathologists. The YOLO was built in Python 3.8, PyTorch 1.10, CUDA 11.1, and NVIDIA RTX 6000, and a pretrained weight53 was used in this work, with a batch size of 8 and a learning rate of 0.001 using the Adam optimizer with a weight decay of 0.01. Two other popular object detection deep learning models were implemented to show the effectiveness of the calibration. A single-shot multibox detector (SSD)54 and faster region-based convolutional neural networks (Faster RCNN)55 were built in Python 3.8, PyTorch 1.10, CUDA 11.1, and NVIDIA RTX 6000. The training process of the SSD was started by loading the pretrained weight,56 with a batch size of 8 and a learning rate of 0.001 using the stochastic gradient descent optimizer with a momentum of 0.9. Faster RCNN used a pretrained weight57 and was trained with a batch size of 8 and a learning rate of 0.0001 using the Adam optimizer with a weight decay of 0.001. ## Discussion and Conclusion In this work, we reported calcification detection in coronary OCT images using deep learning models with uncertainty measurements and confidence calibration to reduce the bias in deep learning models. Although tissue detection and segmentation in OCT images have been studied, to our best knowledge, this work is the first to implement uncertainty measurement and confidence calibration for deep learning-based calcification detection in coronary OCT images. We investigated the calcification detection performance of deep learning object detection models and evaluated the reliability of predictions by detection accuracy and uncertainty measures. With an exceptional runtime of 140 fps, YOLO had the potential to become the real-time detector for predicting calcification in coronary OCT images. This work also implemented confidence calibration by integrating the bounding box information with the confidence score. The quantitative and qualitative results showed the effectiveness of the calibration, indicating its practical value in safe-critical and risk-sensitive applications, for example, the calcification detection in coronary OCT images during PCI. In the future, we will implement other calibration methods on the predicted confidence score and seek to ensemble multiple models to produce more robust and reliable predictions for calcification detection in OCT images. Furthermore, by providing additional information critical to diagnosis, the calibrated confidence and uncertainty measures can be used in future clinical practice. ## Data Availability Statement The datasets generated and analyzed in this work are available from the corresponding author upon reasonable request. ## References 1. Nichols M.. **Cardiovascular disease in Europe: epidemiological update**. *Eur. Heart J.* (2013) **34** 3028-3034. DOI: 10.1093/eurheartj/eht356 2. Go A. S.. **Heart disease and stroke statistics—2014 update: a report from the American Heart Association**. *Circulation* (2014) **129** e28-e292. DOI: 10.1161/01.cir.0000442015.53336.12 3. Okrainec K., Banerjee D. K., Eisenberg M. J.. **Coronary artery disease in the developing world**. *Am. Heart J.* (2004) **148** 7-15. DOI: 10.1016/j.ahj.2003.11.027 4. Moussa I.. **Impact of coronary culprit lesion calcium in patients undergoing paclitaxel-eluting stent implantation (a TAXUS-IV sub study)**. *Am. J. Cardiol.* (2005) **96** 1242-1247. DOI: 10.1016/j.amjcard.2005.06.064 5. Farag M.. **Treatment of calcified coronary artery lesions**. *Expert Rev. Cardiovasc. Ther.* (2016) **14** 683-690. DOI: 10.1586/14779072.2016.1159513 6. Kawaguchi R.. **Impact of lesion calcification on clinical and angiographic outcome after sirolimus-eluting stent implantation in real-world patients**. *Cardiovasc. Revasc. Med.* (2008) **9** 2-8. DOI: 10.1016/j.carrev.2007.07.004 7. Kim L. K.. **Rate of percutaneous coronary intervention for the management of acute coronary syndromes and stable coronary artery disease in the United States (2007 to 2011)**. *Am. J. Cardiol.* (2014) **114** 1003-1010. DOI: 10.1016/j.amjcard.2014.07.013 8. Migliavacca F.. **Mechanical behavior of coronary stents investigated through the finite element method**. *J. Biomech.* (2002) **35** 803-811. DOI: 10.1016/S0021-9290(02)00033-7 9. Sinitsyn V.. **Relationships between coronary calcification detected at electron beam computed tomography and percutaneous transluminal coronary angioplasty results in coronary artery disease patients**. *Eur. Radiol.* (2003) **13** 62-67. DOI: 10.1007/s00330-002-1399-x 10. Gharaibeh Y.. **Coronary calcification segmentation in intravascular OCT images using deep learning: application to calcification scoring**. *J. Med. Imaging* (2019) **6** 045002. DOI: 10.1117/1.JMI.6.4.045002 11. Edelman E. R., Rogers C.. **Pathobiologic responses to stenting**. *Am. J. Cardiol.* (1998) **81** 4E-6E. DOI: 10.1016/S0002-9149(98)00189-1 12. Dangas G. D.. **In-stent restenosis in the drug-eluting stent era**. *J. Am. Coll. Cardiol.* (2010) **56** 1897-1907. DOI: 10.1016/j.jacc.2010.07.028 13. Fujii K.. **Stent underexpansion and residual reference segment stenosis are related to stent thrombosis after sirolimus-eluting stent implantation: an intravascular ultrasound study**. *J. Am. Coll. Cardiol.* (2005) **45** 995-998. DOI: 10.1016/j.jacc.2004.12.066 14. Attizzani G. F.. **Mechanisms, pathophysiology, and clinical aspects of incomplete stent apposition**. *J. Am. Coll. Cardiol.* (2014) **63** 1355-1367. DOI: 10.1016/j.jacc.2014.01.019 15. Doi H.. **Impact of post-intervention minimal stent area on 9-month follow-up patency of paclitaxel-eluting stents: an integrated intravascular ultrasound analysis from the TAXUS IV, V, and VI and TAXUS ATLAS workhorse, long lesion, and direct stent trials**. *JACC: Cardiovasc. Interv.* (2009) **2** 1269-1275. DOI: 10.1016/j.jcin.2009.10.005 16. Hong M.-K.. **Intravascular ultrasound predictors of angiographic restenosis after sirolimus-eluting stent implantation**. *Eur. Heart J.* (2006) **27** 1305-1310. DOI: 10.1093/eurheartj/ehi882 17. Uren N. G.. **Predictors and outcomes of stent thrombosis. An intravascular ultrasound registry**. *Eur. Heart J.* (2002) **23** 124-132. DOI: 10.1053/euhj.2001.2707 18. Mehrotra S., Mishra S., Paramasivam G.. **Imaging during percutaneous coronary intervention for optimizing outcomes**. *Indian Heart J.* (2018) **70** S456-S465. DOI: 10.1016/j.ihj.2018.08.012 19. Richards G., Johnson T.. **A vision of percutaneous coronary revascularisation in 2021: how to take advantage of intra-coronary imaging to perform more effective PCI**. *JRSM Cardiovasc. Dis.* (2021) **10** 204800402110499. DOI: 10.1177/20480040211049978 20. Fujino A.. **A new optical coherence tomography-based calcium scoring system to predict stent underexpansion**. *EuroIntervention* (2018) **13** 2182-2189. DOI: 10.4244/EIJ-D-17-00962 21. Prati F.. **Expert review document part 2: methodology, terminology and clinical applications of optical coherence tomography for the assessment of interventional procedures**. *Eur. Heart J.* (2012) **33** 2513-2520. DOI: 10.1093/eurheartj/ehs095 22. Bezerra H. G.. **Intracoronary optical coherence tomography: a comprehensive review: clinical and research applications**. *JACC: Cardiovasc. Interv.* (2009) **2** 1035-1046. DOI: 10.1016/j.jcin.2009.06.019 23. Wu W.. **Automatic detection of coronary artery stenosis by convolutional neural network with temporal constraint**. *Comput. Biol. Med.* (2020) **118** 103657. DOI: 10.1016/j.compbiomed.2020.103657 24. Tong Y.. **Application of machine learning in ophthalmic imaging modalities**. *Eye Vision* (2020) **7** 1-15. DOI: 10.1186/s40662-020-00183-6 25. Athanasiou L. S.. **A deep learning approach to classify atherosclerosis using intracoronary optical coherence tomography**. *Proc. SPIE* (2019) **10950** 109500N. DOI: 10.1117/12.2513078 26. Abdolmanafi A.. **Characterization of coronary artery pathological formations from OCT imaging using deep learning**. *Biomed. Opt. Express* (2018) **9** 4936-4960. DOI: 10.1364/BOE.9.004936 27. Lee J.. **Segmentation of coronary calcified plaque in intravascular OCT images using a two-step deep learning approach**. *IEEE Access* (2020) **8** 225581-225593. DOI: 10.1109/ACCESS.2020.3045285 28. Avital Y.. **Identification of coronary calcifications in optical coherence tomography imaging using deep learning**. *Sci. Rep.* (2021) **11** 11269. DOI: 10.1038/s41598-021-90525-8 29. Shibutani H.. **Automated classification of coronary atherosclerotic plaque in optical frequency domain imaging based on deep learning**. *Atherosclerosis* (2021) **328** 100-105. DOI: 10.1016/j.atherosclerosis.2021.06.003 30. Lee J.. **Automated plaque characterization using deep learning on coronary intravascular optical coherence tomographic images**. *Biomed. Opt. Express* (2019) **10** 6497-6515. DOI: 10.1364/BOE.10.006497 31. Sun H.. **In vivo detection of plaque erosion by intravascular optical coherence tomography using artificial intelligence**. *Biomed. Opt. Express* (2022) **13** 3922-3938. DOI: 10.1364/BOE.459623 32. Nguyen A., Yosinski J., Clune J.. **Deep neural networks are easily fooled: High confidence predictions for unrecognizable images**. (2015) 427-436 33. Wei H.. **Mitigating neural network overconfidence with logit normalization**. (2022) 23631-23644 34. Mishra A. A.. **Uncertainty quantification for deep learning in particle accelerator applications**. *Phys. Rev. Accel. Beams* (2021) **24** 114601. DOI: 10.1103/PhysRevAccelBeams.24.114601 35. Nguyen A., Yosinski J., Clune J.. **Deep neural networks are easily fooled: high confidence predictions for unrecognizable images**. (2015) 427-436. DOI: 10.48550/arXiv.1412.1897 36. Platt J.. **Probabilistic outputs for support vector machines and comparisons to regularized likelihood methods**. *Adv. Large Margin Classifiers* (1999) **10** 61-74 37. Zadrozny B., Elkan C.. **Obtaining calibrated probability estimates from decision trees and naive Bayesian classifiers**. (2001) 609-616 38. Guo C.. **On calibration of modern neural networks**. (2017) 1321-1330 39. Wyatt M.. **Using ensemble methods to improve the robustness of deep learning for image classification in marine environments**. *Methods Ecol. Evol.* (2022) **13** 1317-1328 40. Ju C., Bibaut A., van der Laan M.. **The relative performance of ensemble methods with deep convolutional neural networks for image classification**. *J. Appl. Stat.* (2018) **45** 2800-2818. PMID: 31631918 41. Cao S.. **Super-resolution technology to simultaneously improve optical & digital resolution of optical coherence tomography via deep learning**. (2020) 1879-1882. DOI: 10.1109/EMBC44109.2020.9175777 42. Li X.. **Multi-scale reconstruction of undersampled spectral-spatial OCT data for coronary imaging using deep learning**. *IEEE Trans. Biomed. Eng.* (2022) **69** 3667-3677. DOI: 10.1109/TBME.2022.3175670 43. Hussain Z.. **Differential data augmentation techniques for medical imaging classification tasks**. (2017) **2017** 979 44. Kraus M. F.. **Motion correction in optical coherence tomography volumes on a per A-scan basis using orthogonal scan patterns**. *Biomed. Opt. Express* (2012) **3** 1182-1199. DOI: 10.1364/BOE.3.001182 45. Lian J.. **Deblurring retinal optical coherence tomography via a convolutional neural network with anisotropic and double convolution layer**. *IET Comput. Vision* (2018) **12** 900-907. DOI: 10.1049/iet-cvi.2018.0016 46. Tearney G.. **Rapid acquisition of in vivo biological images by use of optical coherence tomography**. *Opt. Lett.* (1996) **21** 1408-1410. DOI: 10.1364/OL.21.001408 47. Cui Y.. **Imaging artifacts and segmentation errors with wide-field swept-source optical coherence tomography angiography in diabetic retinopathy**. *Transl. Vis. Sci. Technol.* (2019) **8** 18. DOI: 10.1167/tvst.8.6.18 48. Schwarzhans F.. **Generating large field of view en-face projection images from intra-acquisition motion compensated volumetric optical coherence tomography data**. *Biomed. Opt. Express* (2020) **11** 6881-6904. DOI: 10.1364/BOE.404738 49. Abayomi-Alli O. O.. **Cassava disease recognition from low-quality images using enhanced data augmentation model and deep learning**. *Expert Syst.* (2021) **38** e12746. DOI: 10.1111/exsy.12746 50. Sharif Razavian A.. **CNN features off-the-shelf: an astounding baseline for recognition**. (2014) 806-813. DOI: 10.1109/CVPRW.2014.131 51. Jocher G.. **YOLOv5 by ultralytics**. (2020) 52. Kuppers F.. **Multivariate confidence calibration for object detection**. (2020) 326-327. DOI: 10.1109/CVPRW50498.2020.00171 53. Jocher G.. **YOLO5s pre-trained weight using coco dataset**. (2020) 54. Liu W.. **SSD: single shot multibox detector**. (2016) 21-37 55. Ren S.. **Faster R-CNN: towards real-time object detection with region proposal networks**. *Adv. Neural Inf. Process. Syst.* (2015) **28** 91-99 56. Yu H.. **TensorFlow model garden**. (2020) 57. **TorchVision: PyTorch’s computer vision library**. (2016) 58. Li Q.. **DeepRetina: layer segmentation of retina in OCT images using deep learning**. *Transl. Vision Sci. Technol.* (2020) **9** 61 59. Shah A.. **Multiple surface segmentation using convolution neural nets: application to retinal layer segmentation in OCT images**. *Biomed. Opt. Express* (2018) **9** 4509-4526. DOI: 10.1364/BOE.9.004509
--- title: Simple prediction of COVID-19 convalescent plasma units with high levels of neutralization antibodies authors: - Katerina Jazbec - Mojca Jež - Klemen Žiberna - Polonca Mali - Živa Ramšak - Urška Rahne Potokar - Zdravko Kvrzić - Maja Černilec - Melita Gracar - Marjana Šprohar - Petra Jovanovič - Sonja Vuletić - Primož Rožman journal: Virology Journal year: 2023 pmcid: PMC10042109 doi: 10.1186/s12985-023-02007-0 license: CC BY 4.0 --- # Simple prediction of COVID-19 convalescent plasma units with high levels of neutralization antibodies ## Abstract ### Background Hyperimmune convalescent COVID-19 plasma (CCP) containing anti-SARS-CoV-2 neutralizing antibodies (NAbs) was proposed as a therapeutic option for patients early in the new coronavirus disease pandemic. The efficacy of this therapy depends on the quantity of neutralizing antibodies (NAbs) in the CCP units, with titers ≥ 1:160 being recommended. The standard neutralizing tests (NTs) used for determining appropriate CCP donors are technically demanding and expensive and take several days. We explored whether they could be replaced by high-throughput serology tests and a set of available clinical data. ### Methods Our study included 1302 CCP donors after PCR-confirmed COVID-19 infection. To predict donors with high NAb titers, we built four [4] multiple logistic regression models evaluating the relationships of demographic data, COVID-19 symptoms, results of various serological testing, the period between disease and donation, and COVID-19 vaccination status. ### Results The analysis of the four models showed that the chemiluminescent microparticle assay (CMIA) for the quantitative determination of IgG Abs to the RBD of the S1 subunit of the SARS-CoV-2 spike protein was enough to predict the CCP units with a high NAb titer. CCP donors with respective results > 850 BAU/ml SARS-CoV-2 IgG had a high probability of attaining sufficient NAb titers. Including additional variables such as donor demographics, clinical symptoms, or time of donation into a particular predictive model did not significantly increase its sensitivity and specificity. ### Conclusion A simple quantitative serological determination of anti-SARS-CoV-2 antibodies alone is satisfactory for recruiting CCP donors with high titer NAbs. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12985-023-02007-0. ## Introduction The hyperimmune convalescent COVID-19 plasma (CCP) with anti-SARS-CoV-2 antibodies became an appealing source of therapy early in the COVID-19 pandemic. Millions of CCP units were collected in the following two years and became available for controlled prospective clinical studies. The promising prospect was that the antibodies present in CCP units could neutralize the SARS-CoV-2 virus contributing to its clearance from the patient [1, 2]. During COVID-19 infection, the immune system generates various classes of anti-SARS-CoV-2 antibodies (Abs) that are polyspecific and polyclonal. The early IgM and IgA Abs peak between day 16 to day 30 after symptom onset and disappear 1–2 months later, whereas the IgG Abs persist much longer. In most infected patients, anti-SARS-CoV-2 IgG seropositivity lasts for at least six months after infection [3, 4], which renders them potential donors of CCP. Several initial studies on low numbers of patients reported a beneficial effect of CCP treatment with no serious side effects [5–7]. The initial studies had some drawbacks, such as a lack of suitable control groups, variability of therapeutic approaches, concomitant pathology, and poor recording of adverse reactions. Probably the most significant failure was the lack of uniform timing and dosing of CCP therapy. However, the results gave specific common knowledge that only high titer plasma should be used, that it should be used within 72 h of the symptom appearance, and that only selected groups of patients are legible for CCP therapy, such as mildly affected patients, and especially the ones with immune deficiencies [8, 9]. However, several parallel studies haven’t yielded such booming therapeutic results [10, 11]. At least two meta-analyses did not provide evidence of a reduction in mortality or any benefit for other clinical outcomes [12, 13]. Another subset of patients who received CCP within 72 h of symptoms onset did not show significant improvement [14], even when two CCP units with high antibody titer were administered and the patients had some benefit. International trials such as RECOVERY and REMAP-CAP demonstrated a low likelihood of improvement in organ support-free days and mortality [15, 16]. Also, later on in 2022, several groups failed to report the mortality benefit of CCP treatment in patients with mild disease [17] as well as with severe disease [18–22], which led to the most recent AABB expert panel’s clinical practice guidelines recommending CCP therapy only for outpatients with COVID-19 who are at high risk for disease progression, for hospitalized patients with moderate or severe disease, and for the patients with COVID-19 who do not have SARS-CoV-2 antibodies at admission or with preexisting immunosuppression. Besides, the CCP should not be used prophylactically for uninfected persons with close contact exposure and should be transfused with high neutralizing titers to infected patients early after symptom onset [23]. Therefore, it was somehow surprising that in 2022, the benefit of early therapy was again confirmed in several larger cohorts. Sullivan et al. reported that early administration of high titer CCP reduced hospitalizations by more than $50\%$ [24]. Similarly, Sanz et al. reported that transfusion of CCP caused a significant reduction in the 30-day mortality rate, suggesting that CCP can still be helpful in selected patients and calling for further studies before withdrawing CCP from the COVID-19 therapeutic armamentarium [25]. Franchini et al. even urge their colleagues to review the available CCP efficacy data and incorporate its use in the treatment of the vulnerable population [26]. Many other studies reported favorable responses to CCP treatment, too [27–30]. During 2020 and 2021, we collected approximately 4000 hyperimmune CCP units from voluntary blood donors. The entering criterion was a recovery after PCR-positive COVID-19 disease. CCP donors are often selected based upon their neutralizing antibody (NAb) count, which is assessed by a plaque reduction neutralization test (NT) that needs a feasible isolate, replication-competent cell lines, and qualified staff [31]. Our chosen threshold for effective CCP units was the NAb titer ≥ 1:160, based on the references stating that people with NAb titer ≥ 1:160 can be protected from the SARS-CoV-2 infection [32, 33]. Since the standard NTs are technically demanding, expensive, and require biosafety level 3 containment, we looked for a simple prediction method for acquiring CCP units with corresponding therapeutic efficiency based on the demographic, clinical, and serological status of the convalescent donors. We intended to identify the key features that would predict the high NT values. These results would be applicable not only for determining appropriate hyperimmune CCP but also to prevent and control COVID-19 infection and optimize vaccine doses [34]. ## The study design For this retrospective cohort study, we analyzed 1302 CCP units collected from convalescent donors that donated plasma between June 2020 and August 2021 and divided them into a group with low NT values (< 160) and a group with high NT values (≥ 160). With univariate analysis, we determined the demographic and laboratory parameters that differed in each group and were therefore considered informative. We created four different models, which included different potentially informative parameters such as serology assays, demographic data, and the post-COVID-19 period of collection. ## Ethics The study was approved by the National Medical Ethics Committee of the Republic of Slovenia (0120–$\frac{241}{2020}$-11, from 7.12.2020; 0120–$\frac{241}{2020}$/14, from 17.5.2021; 0120–$\frac{241}{2020}$-8, from 18.6.2020). ## Convalescent plasma donors All CCP donors met the criteria for normal blood or plasma donation. Only the first donations from 1302 CCP donors (400 females and 902 males)aged between 18 and 65 years (mean 43.9 ± 0.3 years) with a history of polymerase chain reaction (PCR) confirmed SARS-CoV-2 infection in nasopharyngeal swabs were included. Out of these, $86\%$ of participants reported mild symptoms, $14\%$ reported moderate to severe symptoms, with only 6 participants requiring hospital treatment. CCP units were donated 63.0 days [IQR 43.0-121.0 days] after confirmed SARS-CoV-2 infection. Since the vaccination in our region started after January 2021, 235 convalescent donors have also received the vaccination before the first CPP collection. Basic donor information is presented in Table 2. ## Serological testing Two semi-quantitative and one quantitative serological test were used to detect anti-SARS-CoV-2 antibodies: (i) Wantai SARS-CoV-2 Ab ELISA, an enzyme-linked immunosorbent assay for the qualitative detection of total IgG and IgM antibodies to the RBD of SARS-CoV-2 spike protein that was performed on 285 samples; (ii) Abbott SARS-CoV-2 IgG assay, a chemiluminescent microparticle assay (CMIA) for the qualitative detection of IgG Abs to the nucleocapsid protein that was performed on 569 samples; and (iii) Abbott SARS-CoV-2 IgG II Quant, second-generation CMIA for the quantitative determination of IgG Abs to the RBD of the S1 subunit of the SARS-CoV-2 spike protein, including the neutralizing Abs that was performed on 844 samples. All tests were performed according to the manufacturers’ instructions. ## Neutralizing antibody testing For neutralizing antibody testing, a standard live SARS-CoV-2 microneutralization assay was used [35]. Briefly, two-fold serial dilutions of CCP from 1:10 to 1:1280 were prepared and mixed with a viral solution containing 100 TCID50 (TCID50–$50\%$ tissue culture infective dose) and incubated. Virus lineage B.1.1 (G614) was used. 10,000 Vero E6 cells per well of 96-well plate were preseeded 24 h before the experiment. After incubation, the virus/CCP mixture was added to Vero E6 cells into a 96-well plate and incubated for five days at 37 °C. The assay readout was the cytopathic effect, and the assay cut-off titer was < 1:20. Neutralization test was performed at the Institute of Microbiology and Immunology, Faculty of Medicine, University of Ljubljana. ## Statistics The analysis was performed on the COVID-19 dataset from September 2020 to August 2021. A subset table was created using only the first measurement for every donor (some donors have donated plasma multiple times) and only using data where NT measurement was obtained. This selection yielded a total of 1302 samples used for this analysis. Data in the analysis are presented as the mean and standard error of the mean for normally distributed data, median and interquartile range (IQR) for non-parametric data, and as the count and the percentage for binary data. Statistical comparison between the two groups was performed using a Student’s t-test for normally distributed data, a Mann-Whitney U test for non-parametric data, and Fisher exact test for binary data. Four different multiple logistic regression models were created to analyze the relative importance of selected features in predicting high neutralization test values. The dataset was split into the train set ($70\%$ of the data) that was used to fit the model. The test (validation) set ($30\%$) was used to assess the model’s prediction performance. A receiver operating characteristics (ROC) curve was used to compare the sensitivity and specificity of different models. The analysis was performed using Python 3.8, NumPy 1.19, Pandas 1.1, SciPy 1.7, SciKit Learn 0.23, and Statsmodels 0.12. ## Results We compared the groups with low NAb titers and high NAb titers. The first important difference between the groups was the COVID-19 vaccination. More than $44\%$ of samples in the high NT titer were from donors who were vaccinated against COVID-19 and also recovered from the COVID-19 infection (compared to $1.7\%$ in the low NT titer group) (Table S1). Amongst the samples from the vaccinated donors (235 of 1302 total samples, Table S2), $94.5\%$ of vaccinated ones were in the high NT titer group compared to $5.5\%$ in the low NT titer group (Table 1). Table 1Unvaccinated donor’s characteristics in low and high NT titer groupsLow NT titer (< 1:160)High NT titer (≥ 1:160)p-value Demographic parameters Gender - female263 ($34.3\%$) ($$n = 766$$)84 ($27.9\%$) ($$n = 301$$)0.050Age (years)42.4 ± 0.4 ($$n = 766$$)47.4 ± 0.6 ($$n = 301$$)< 0.001Body weight (kg)85.2 ± 0.7 ($$n = 637$$)90.3 ± 1.1 ($$n = 199$$)< 0.001Height (cm)175.8 ± 0.3 ($$n = 634$$)176.0 ± 0.6 ($$n = 196$$)0.759Body mass index (kg/m2)27.4 ± 0.2 ($$n = 633$$)29.1 ± 0.3 ($$n = 196$$)< 0.001 Blood groups and total IgG Blood group 0252 ($32.9\%$) ($$n = 766$$)106 ($35.2\%$) ($$n = 301$$)0.472Blood group A314 ($41.0\%$) ($$n = 766$$)127 ($42.2\%$) ($$n = 301$$)0.730Blood group B134 ($17.5\%$) ($$n = 766$$)48 ($15.9\%$) ($$n = 301$$)0.588Blood group AB66 ($8.6\%$) ($$n = 766$$)20 ($6.6\%$) ($$n = 301$$)0.319Rh(D) factor569 ($81.8\%$) ($$n = 696$$)204 ($80.0\%$) ($$n = 255$$)0.574Total IgG (AU/ml)10.4 [9.1–11.9] ($$n = 732$$)10.6 [9.4–11.9] ($$n = 280$$)0.143 Serological testing Wantai semi-quantitative SARS-CoV-2 Ab test (index S/C)*17.9 [11.2–19.3] ($$n = 233$$)19.6 [18.6–20.5] ($$n = 52$$)< 0.001Abbott semi-quantitative SARS-CoV-2 Ab test (index S/C)*4.32 [2.62-6.0] ($$n = 455$$)6.94 [5.62–7.88] ($$n = 114$$)< 0.001Abbott quantitative SARS-CoV-2 Ab test (BAU/ml)187 [141–282] ($$n = 385$$)498 [310–1019] ($$n = 224$$)< 0.001Neutralization test (titer)31.2 [30.3–32.1] ($$n = 766$$)300.0 [288.2-312.4] ($$n = 301$$)< 0.001 Timeline Days after start of COVID-19 symptoms53.0 [41.0-79.2] ($$n = 712$$)52.5 [40.0-79.8] ($$n = 254$$)0.8220–60 days after start of COVID-19 symptoms418 ($59.2\%$) ($$n = 706$$)150 ($61.2\%$) ($$n = 245$$)0.59760–120 days after start of COVID-19 symptoms221 ($31.3\%$) ($$n = 706$$)76 ($31.0\%$) ($$n = 245$$)1.000120–180 days after start of COVID-19 symptoms67 ($9.5\%$) ($$n = 706$$)19 ($7.8\%$) ($$n = 245$$)0.518 Symptoms Hospitalization4 ($1.0\%$) ($$n = 335$$)1 ($1.0\%$) ($$n = 122$$)1.000Fever191 ($58.0\%$) ($$n = 329$$)88 ($71.0\%$) ($$n = 124$$)0.013Maximum body temperature (°C)38.0 [37.5–38.5] ($$n = 461$$)38.5 [37.8–39.0] ($$n = 189$$)0.001Number of days with fever2.0 [1.0–4.0] ($$n = 559$$)4.0 [2.0–8.0] ($$n = 218$$)< 0.001Cough106 ($32.0\%$) ($$n = 329$$)60 ($48.0\%$) ($$n = 124$$)0.002Anosmia209 ($63.0\%$) ($$n = 330$$)58 ($47.0\%$) ($$n = 124$$)0.002Myalgia146 ($44.0\%$) ($$n = 330$$)60 ($48.0\%$) ($$n = 124$$)0.460Dyspnea34 ($10.0\%$) ($$n = 329$$)25 ($20.0\%$) ($$n = 123$$)0.007Fatigue168 ($51.0\%$) ($$n = 329$$)63 ($51.0\%$) ($$n = 124$$)1.000Headache108 ($33.0\%$) ($$n = 327$$)36 ($29.0\%$) ($$n = 123$$)0.497Notes: *All data* was not available for every plasma donor. The N represents the total number of samples for which the data was available for a particular parameter*Index S/C - signal/cut-off index In order to depict which other parameters (besides vaccination status) could also be different between the low and high NAb titer groups, we separately analyzed the samples from unvaccinated donors (1067 of 1302 total samples). Table 1 presents how basic demographic features, blood types, SARS-CoV-2 antibody testing, and COVID-19 symptoms differ between unvaccinated donors in the low and high NAb titer groups. The unvaccinated donors in the high NAb titer group were older ($p \leq 0.001$), and had greater body weight and body mass index ($p \leq 0.001$) compared to donors in the low NAb titer group. There were no differences between the blood group types, the total IgG value, and the duration between the start of COVID-19 and the date of plasma collection. Comparing the clinical presentation of COVID-19 between the groups, we note that the higher proportion of donors in the high NAb titer group had fever ($$p \leq 0.01$$), higher maximum temperature ($$p \leq 0.001$$), and longer symptom duration ($p \leq 0.001$). Furthermore, a higher proportion of donors in the high NAb titer group also had cough ($$p \leq 0.002$$), anosmia ($$p \leq 0.002$$), and dyspnea ($$p \leq 0.007$$). There were no differences in the number of hospitalizations, frequency of myalgia, and fatigue between the donors in the two groups. Unsurprisingly, the most important features differentiating the low and high NAb titer group samples are serological measurements of SARS-CoV-2 antibodies ($p \leq 0.001$). However, there are differences between different commercial tests. The difference in medians between low NT and high NT titer groups using the Wantai semi-quantitative S/C index was 1.7 or $9\%$. The difference in medians between the groups using Abbott semi-quantitative S/C index was 2.6 or $61\%$. And, finally, the difference in medians between the groups using Abbot quantitative test was 311 BAU/mL or $166\%$. Four different multiple logistic regression models were created (Table 3) to assess which parameter or combination of parameters has the most significant predictive power for choosing CCP donors with high SARS-CoV-2 NAb titer (≥ 160). The first multiple logistic regression model (Model 1) contained seven demographic and clinical variables and had relatively poor performance, with 0.75 ROC AUC (area under the ROC curve), 0.65 sensitivity, and 0.74 specificity at the optimal cut-off threshold (Fig. 1). Fig. 1The receiver operating characteristic (ROC) curve of four different multiple logistic regression models (defined in Table 3) predicting a high NT titer group. Model 1 was created with demographic and clinical variables, Model 2 consists of Abbott quantitative SARS-CoV-2 Ab test only, the Model 3 uses Abbott quantitative SARS-CoV-2 Ab test variable and vaccination status, and the Model 4 includes all variables The second model (Model 2), consisting of the Abbott SARS-CoV-2 IgG II Quant test only, performed better than the previous model and produced a 0.83 ROC AUC with 0.80 sensitivity and 0.86 specificity at the optimal cut-off threshold. The variable Abbott SARS-CoV-2 IgG II Quant test was highly significant ($p \leq 0.001$) in this model. An increase of 100 BAU/mL results in a 1.5x increased probability of having a high NT titer. Adding additional variables to the Abbott SARS-CoV-2 IgG II Quant test resulted in minor improvements in predicting high NT titers over the previous model. Model 3 incorporated the Abbott SARS-CoV-2 IgG II Quant test and vaccination variables and produced a 0.92 ROC AUC with 0.80 sensitivity and 0.86 specificity at the optimal threshold. Both variables were statistically significant ($p \leq 0.001$ and $$p \leq 0.02$$, respectively). The odds ratio (OR) for 100 BAU/mL of variable Abbott SARS-CoV-2 IgG II Quant test was 1.44, and the OR for being vaccinated was 2.64. Finally, Model 4 included all previously selected variables and resulted in 0.91 ROC AUC with 0.73 sensitivity and 0.89 specificity at the optimal threshold. Only the Abbott SARS-CoV-2 IgG II Quant test variable and vaccination status were statistically significant ($p \leq 0.001$ and $$p \leq 0.004$$ respectively) in this model with OR of 1.38 and 5.66, respectively. Even though the univariate analysis depicted differences in many variables when comparing the low NAb titer group with the high NAb titer group (Table 2), their inclusion into the predictive model only minimally improved its predictive power. Besides, the results of the simplest model, i.e., Model 2, are also very informative. The CCP donors with SARS-CoV-2 IgG values above 850 BAU/ml had an $80\%$ probability of having high NT (Fig. 2). Altogether, the simplest model (Model 2) seems to be sufficient for a good prediction of high Nab titers in CCP donors, although it only relied on one variable, i.e., the quantitative serological chemiluminescent microparticle assay (CMIA) test. Table 2Number of vaccinated and unvaccinated donors with low or high neutralizing antibody titersLow NAb titer (< 1:160)High NAb titer (≥ 1:160)Unvaccinated766 ($71.8\%$)301 ($28.2\%$)Vaccinated13 ($5.5\%$)222 ($94.5\%$)Notes: Vaccinated donors were more likely to have a high NT titer than unvaccinated CCP donors ($p \leq 0.001$) Table 3Different multiple logistic regression models predict whether a donor will yield a high NT titerLog OR ($95\%$ CI)OR ($95\%$ CI)p-value Model 1 N training = 491, N evaluation = 211Gender (female)0.14 [-0.49-0.78]1.15 [0.61–2.17]0.66Age (years)0.02 [0.0-0.04]1.02 [1.00-1.04]0.07Body weight (kg)-0.15 [-0.32-0.02]0.86 [0.72–1.02]0.09Height (cm)0.14 [-0.03-0.32]1.16 [0.97–1.38]0.11Body mass index (kg/m2)0.46 [-0.07-1]1.59 [0.93–2.71]0.09Days after start of COVID-19 symptoms0.02 [0.01–0.02]1.02 [1.01–1.02]< 0.001COVID-19 symptoms duration (days)0.01 [-0.02-0.03]1.01 [0.98–1.03]0.53 Model 2 (univariate logistic regression) N training = 590, N evaluation = 254Abbott quantitative SARS-CoV-2 Ab test (100 BAU/ml)0.41 [0.33–0.50]1.51 [1.38–1.64]< 0.001 Model 3 N training = 590, N evaluation = 254Abbott quantitative SARS-CoV-2 Ab test (100 BAU/ml)0.37 [0.28–0.46]1.44 [1.32–1.58]< 0.001Vaccination0.97 [0.14–1.8]2.64 [1.15–6.04]0.02 Model 4 N training = 420, N evaluation = 180Gender (female)-0.53 [-1.43-0.36]0.59 [0.24–1.44]0.24Age (years)0.00 [-0.02-0.03]1.00 [0.98–1.03]0.77Body weight (kg)-0.23 [-0.47-0.02]0.80 [0.62–1.02]0.07Height (cm)0.19 [-0.06-0.44]1.21 [0.94–1.56]0.14Body mass index (kg/m2)0.7 [-0.05-1.45]2.01 [0.95–4.27]0.07Days after start of COVID-19 symptoms0.00 [-0.01-0.01]1.00 [0.99–1.01]0.66COVID-19 symptoms duration (days)0.00 [-0.03-0.04]1.00 [0.97–1.04]0.88Abbott quantitative SARS-CoV-2 Ab test (100 BAU/ml)0.32 [0.22–0.43]1.38 [1.25–1.53]< 0.001Vaccination1.73 [0.56–2.9]5.66 [1.76–18.2]0.004Notes: Model 1 was trained using data from 491 donors and its performance was evaluated on 211 donors. Model 2 was trained using data from 590 donors and its performance was evaluated on 254 donors. Model 3 was trained using data from 590 donors and its performance was evaluated on 254 donors. Model 4 was trained using data from 420 donors and its performance was evaluated on 180 donors Fig. 2The modeled relationship between the Abbot quantitative SARS-CoV-2 Ab test and high NAb titer probability. The probability of a high NT titer is obtained from Model 2 (as defined in Table 3) ## Discussion Several authors tried to find the best recruitment strategy and predictors of high antibody levels needed for improving the supply of high-quality CCP from the donors. Prudente et al. stated that among 102 individuals, the ones with a longer time interval between symptom onset and sample collection, who had been hospitalized and were above 35 years old, presented with stronger antibody response [36]. Similarly, Mehew et al. found in 29,585 CCP donors that older male donors who had been hospitalized with COVID-19 were most likely to harbor high levels of antibodies [37]. Yang et al. suggest that SARS-CoV-2 viral specific antibody response profiles are distinct in different age groups [38]. Vinkenoog et al. found that in 2,082 convalescent donors six symptoms (dry cough, fatigue, diarrhea, fever, dyspnea, and muscle weakness) predicted higher IgG concentrations [39]. Our data also shows higher neutralization antibodies in unvaccinated plasma donors (see Table 2), if they were older, had greater body weight and body mass index, and if they had higher body temperature during the infection, higher number of days with fever, cough, anosmia, and dyspnea. While obesity is a well-established independent risk factor for developing severe COVID-19, the effect of obesity on neutralizing antibody is not entirely clear with reports of positive as well as negative correlation between body weight/BMI and SARS-CoV-2 antibodies [40, 41]. The aim of our analysis was not to investigate this relationship in great detail, but it would be interesting to explore in greater detail the role of body weight on the immune response, in particular in younger patients. We checked whether serological titer alone could represent an excellent predictive factor. Similar to previous studies [42–44], we found higher NAb titers in older male patients with higher BMI, longer-lasting fever, and higher body temperature. Our search for an optimal prediction model showed that the most crucial predictor of a suitable CCP donor was the result of the serological Abbott Quant test. In contrast, donor demographics, clinical signs, or the time of donation were not that important, and adding these variables into our logistic regression model only minimally improved its predictive power. Even a model containing the two most significant individual predictors of high NAb titers (serological titer and vaccination status) did not improve the predictive power beyond a simple model containing only the serological titer as the variable. This is due to the fact that a vast majority of the vaccinated subjects also had high serological titer values as well as high NAb values, therefore the vaccination information is already contained in the serological titer itself. Moreover, as long as the donor had a titer of anti-SARS-CoV-2 IgG above 850 BAU/ml, the probability of a high NAb titer was high (probability of 0.8). This finding suggests that measuring only SARS-CoV-2 IgG antibody concentration is sufficient to predict whether a CCP donor will have a high NT titer. This also leads to the conclusion that the laborious NT that is currently considered the gold standard can be supplemented by surrogate serological quantitative assays, which was also proved by other authors [45–52]. On the other hand, several other authors claim that commercialized serological tests, including those targeting the RBD, cannot substitute for NT assay functionality [53, 54]. In our case, we used three different serological methods (Wantai semi-quantitative SARS-CoV-2 Ab test, Abbott semi-quantitative SARS-CoV-2 Ab test, and Abbott quantitative SARS-CoV-2 Ab test) to measure anti-SARS-CoV-2 binding antibodies. The methods exhibited various abilities to separate the samples into the low and high NAb titer groups. The median was significantly different between high and low NAb titer groups for all three tests. For the Wantai test, the difference between groups was only around $10\%$, and the overlap in the values measured in the low and high NAb titer groups was significant. The difference between groups was approximately $60\%$ for the semi-quantitative Abbott test. Both tests were not very useful for predicting high or low NAb titer. The quantitative Abbott test was much more helpful since the difference between groups was around $500\%$. Also, it is automated, easy to perform, and represents a helpful tool for providing CCP units with clinically relevant antibody titers. In the Emergency Use Authorization (EUA), from December 2021, the FDA states that the use of CCP should be limited to units with high titers of anti-SARS-CoV-2 antibodies and that the testing criteria used in the qualification of CCP should be revised to better assure high neutralization titers in CCP. In the case of Abbott quantitative tests, previous qualifying values of ≥ 840 AU/mL are increased to ≥ 1280 AU/mL in the revised qualifying result (i.e. ≥ 119 BAU/ml to ≥ 181 BAU/ml, respectively) [55]. In our particular case, the median titers in the low and high titer groups were 192 BAU/mL, and 1123 BAU/mL, respectively (see Table S1). The main advantage of our study is its large, representative sample size covering the whole country. However, the main limitation of our study is that the SARS-CoV-2 variant Abs specificities were limited to early variants only. The samples were collected when alpha and beta were prevalent variants. The second limitation of the study is the lack of other clinical data, such as extensive clinical parameters describing the pulmonary, immune, biochemical, and prothrombotic status during the COVID-19 infection, which might improve the prediction in similar models. Since our donors were non-hospitalized and only suffered from mild disease, such data could not be captured. There remain several unsolved questions about whether, in the light of COVID-19 vaccination, the rise of new SARS-CoV-2 virus variants of concern (VOC), and upcoming monoclonal antibody therapies, CCP therapy is losing its potential value as a bridge therapy of the pandemic. We agree that CCP remains an option for the treatment of COVID-19 patients with impaired humoral immunity and that it could play its role as an affordable and easily accessible therapeutic and prophylactic option, especially in middle- and low-income countries [24, 25]. Besides, the national blood transfusion services may continue collecting VOC-specific CCP for the deliberate production of hyperimmune SARS-CoV-2 gamma globulins with distinct variant specificities. Some SARS-CoV-2 VOC may be less susceptible to neutralization by CCP, vaccine-elicited plasma/sera, or SARS-CoV-2 monoclonal antibodies than the earlier SARS-CoV-2 strains. In a population with a newer viral variant, locally collected CCP units will also contain neutralizing antibodies against the local variant and could be used therapeutically. ## Conclusion The quantitative serological determination of anti-SARS-CoV-2 antibodies proved to be a sufficient predictor of high NAb titers, and adding additional demographic parameters did not improve the sensitivity and specificity of our model. Abbot Quant test for detecting anti-SARS-CoV-2 antibodies proved to be highly sensitive and specific for detecting the early SARS-CoV-2 variants and proved to be a surrogate for NT in collecting CCP units with clinically relevant antibody titers. Its value will have to be tested again in the coming months when the new waves of SARS-CoV-2 variants will emerge. ## Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 ## References 1. Du Z, Zhu F, Guo F, Yang B, Wang T. **Detection of antibodies against SARS-CoV-2 in patients with COVID-19**. *J Med Virol* (2020.0) **92** 1735-8. DOI: 10.1002/jmv.25820 2. Hou H, Wang T, Zhang B, Luo Y, Mao L, Wang F. **Detection of IgM and IgG antibodies in patients with coronavirus disease 2019**. *Clin Transl Immunology* (2020.0) **9** e01136. DOI: 10.1002/cti2.1136 3. Munitz A, Edry-Botzer L, Itan M, Tur-Kaspa R, Dicker D, Marcoviciu D. **Rapid seroconversion and persistent functional IgG antibodies in severe COVID-19 patients correlates with an IL-12p70 and IL-33 signature**. *Sci Rep* (2021.0) **11** 3461. DOI: 10.1038/s41598-021-83019-0 4. Jiang XL, Wang GL, Zhao XN, Yan FH, Yao L, Kou ZQ. **Lasting antibody and T cell responses to SARS-CoV-2 in COVID-19 patients three months after infection**. *Nat Commun* (2021.0) **12** 897. DOI: 10.1038/s41467-021-21155-x 5. Duan K, Liu B, Li C, Zhang H, Yu T, Qu J. **Effectiveness of convalescent plasma therapy in severe COVID-19 patients**. *Proc Natl Acad Sci U S A* (2020.0) **117** 9490-6. DOI: 10.1073/pnas.2004168117 6. 6.Shen C, Wang Z, Zhao F, Yang Y, Li J, Yuan J et al. Treatment of 5 Critically Ill Patients With COVID-19 With Convalescent Plasma.JAMA. 2020. 7. 7.Ye M, Fu D, Ren Y, Wang F, Wang D, Zhang F et al. Treatment with convalescent plasma for COVID-19 patients in Wuhan, China.J Med Virol. 2020. 8. Libster R, Perez Marc G, Wappner D, Coviello S, Bianchi A, Braem V. **Early high-titer plasma therapy to prevent severe Covid-19 in older adults**. *N Engl J Med* (2021.0) **384** 610-8. DOI: 10.1056/NEJMoa2033700 9. Tobian AAR, Shaz BH. **Earlier the better: convalescent plasma**. *Blood* (2020.0) **136** 652-4. DOI: 10.1182/blood.2020007638 10. Agarwal A, Mukherjee A, Kumar G, Chatterjee P, Bhatnagar T, Malhotra P. **Convalescent plasma in the management of moderate covid-19 in adults in India: open label phase II multicentre randomised controlled trial (PLACID Trial)**. *BMJ* (2020.0) **371** m3939. DOI: 10.1136/bmj.m3939 11. 11.Avendano-Sola C, Ramos-Martinez A, Munez-Rubio E, Ruiz-Antoran B, Malo de Molina R, Torres F et al. A multicenter randomized open-label clinical trial for convalescent plasma in patients hospitalized with COVID-19 pneumonia.J Clin Invest. 2021;131(20). 12. Janiaud P, Axfors C, Schmitt AM, Gloy V, Ebrahimi F, Hepprich M. **Association of Convalescent plasma treatment with clinical outcomes in patients with COVID-19: a systematic review and Meta-analysis**. *JAMA* (2021.0) **325** 1185-95. DOI: 10.1001/jama.2021.2747 13. Chai KL, Valk SJ, Piechotta V, Kimber C, Monsef I, Doree C. **Convalescent plasma or hyperimmune immunoglobulin for people with COVID-19: a living systematic review**. *Cochrane Database Syst Rev* (2020.0) **10** CD013600. PMID: 33044747 14. Simonovich VA, Burgos Pratx LD, Scibona P, Beruto MV, Vallone MG, Vazquez C. **A Randomized Trial of Convalescent plasma in Covid-19 severe pneumonia**. *N Engl J Med* (2021.0) **384** 619-29. DOI: 10.1056/NEJMoa2031304 15. Axfors C, Janiaud P, Schmitt AM, Van’t Hooft J, Smith ER, Haber NA. **Association between convalescent plasma treatment and mortality in COVID-19: a collaborative systematic review and meta-analysis of randomized clinical trials**. *BMC Infect Dis* (2021.0) **21** 1170. DOI: 10.1186/s12879-021-06829-7 16. **Convalescent plasma in patients admitted to hospital with COVID-19 (RECOVERY): a randomised controlled, open-label, platform trial**. *Lancet* (2021.0) **397** 2049-59. DOI: 10.1016/S0140-6736(21)00897-7 17. Snow TAC, Saleem N, Ambler G, Nastouli E, McCoy LE, Singer M. **Convalescent plasma for COVID-19: a meta-analysis, trial sequential analysis, and meta-regression**. *Br J Anaesth* (2021.0) **127** 834-44. DOI: 10.1016/j.bja.2021.07.033 18. Troxel AB, Petkova E, Goldfeld K, Liu M, Tarpey T, Wu Y. **Association of Convalescent plasma treatment with clinical status in patients hospitalized with COVID-19: a Meta-analysis**. *JAMA Netw Open* (2022.0) **5** e2147331. DOI: 10.1001/jamanetworkopen.2021.47331 19. Song ATW, Rocha V, Mendrone-Júnior A, Calado RT, De Santis GC, Benites BD. **Treatment of severe COVID-19 patients with either low- or high-volume of convalescent plasma versus standard of care: a multicenter bayesian randomized open-label clinical trial (COOP-COVID-19-MCTI)**. *Lancet Reg Health Am* (2022.0) **10** 100216. PMID: 35308034 20. De Santis GC, Oliveira LC, Garibaldi PMM, Almado CEL, Croda J, Arcanjo GGA. **High-dose convalescent plasma for treatment of severe COVID-19**. *Emerg Infect Dis* (2022.0) **28** 548-55. DOI: 10.3201/eid2803.212299 21. Alemany A, Millat-Martinez P, Corbacho-Monne M, Malchair P, Ouchi D, Ruiz-Comellas A. **High-titre methylene blue-treated convalescent plasma as an early treatment for outpatients with COVID-19: a randomised, placebo-controlled trial**. *Lancet Respir Med* (2022.0) **10** 278-88. DOI: 10.1016/S2213-2600(21)00545-2 22. Goossens H, Derde L, Horby P, Bonten M. **The european clinical research response to optimise treatment of patients with COVID-19: lessons learned, future perspective, and recommendations**. *Lancet Infect Dis* (2022.0) **22** e153-e8. DOI: 10.1016/S1473-3099(21)00705-2 23. 23.Estcourt LJ, Cohn CS, Pagano MB, Iannizzi C, Kreuzberger N, Skoetz N et al. Clinical Practice Guidelines From the Association for the Advancement of Blood and Biotherapies (AABB): COVID-19 Convalescent Plasma.Ann Intern Med. 2022. 24. 24.Sullivan DJ, Gebo KA, Shoham S, Bloch EM, Lau B, Shenoy AG et al. Early Outpatient Treatment for Covid-19 with Convalescent Plasma. The New England journal of medicine. 2022. 25. 25.Sanz C, Nomdedeu M, Pereira A, Sauleda S, Alonso R, Bes M et al. Efficacy of early transfusion of convalescent plasma with high-titer SARS-CoV-2 neutralizing antibodies in hospitalized patients with COVID-19.Transfusion. 2022. 26. 26.Franchini M, Casadevall A, Joyner MJ, Focosi D. WHO Is Recommending against the Use of COVID-19 Convalescent Plasma in Immunocompromised Patients?Life (Basel). 2023;13(1). 27. Fodor E, Müller V, Iványi Z, Berki T, Kuten Pella O, Hornyák I. **Early transfusion of Convalescent plasma improves the clinical outcome in severe SARS-CoV2 infection**. *Infect Dis Ther* (2022.0) **11** 293-304. DOI: 10.1007/s40121-021-00514-7 28. Lang-Meli J, Fuchs J, Mathé P, Ho HE, Kern L, Jaki L. **Case Series: convalescent plasma therapy for patients with COVID-19 and primary antibody Deficiency**. *J Clin Immunol* (2022.0) **42** 253-65. DOI: 10.1007/s10875-021-01193-2 29. 29.Franchini M, Glingani C, Donno G, Lucchini G, Beccaria M, Amato M et al. Convalescent Plasma for Hospitalized COVID-19 Patients: A Single-Center Experience.Life (Basel). 2022;12(3). 30. Bajpai M, Maheshwari A, Dogra V, Kumar S, Gupta E, Kale P. **Efficacy of convalescent plasma therapy in the patient with COVID-19: a randomised control trial (COPLA-II trial)**. *BMJ Open* (2022.0) **12** e055189. DOI: 10.1136/bmjopen-2021-055189 31. Chavda VP, Bezbaruah R, Dolia S, Shah N, Verma S, Savale S. **Convalescent plasma (hyperimmune immunoglobulin) for COVID-19 management: an update**. *Process Biochem* (2023.0) **127** 66-81. DOI: 10.1016/j.procbio.2023.01.018 32. Peterhoff D, Gluck V, Vogel M, Schuster P, Schutz A, Neubert P. **A highly specific and sensitive serological assay detects SARS-CoV-2 antibody levels in COVID-19 patients that correlate with neutralization**. *Infection* (2021.0) **49** 75-82. DOI: 10.1007/s15010-020-01503-7 33. 33.Matusali G, Colavita F, Lapa D, Meschi S, Bordi L, Piselli P et al. SARS-CoV-2 Serum Neutralization Assay: A Traditional Tool for a Brand-New Virus.Viruses. 2021;13(4). 34. Bonanni P, Cantón R, Gill D, Halfon P, Liebert UG, Crespo KAN. **The role of Serology Testing to strengthen vaccination initiatives and policies for COVID-19 in Europe**. *COVID* (2021.0) **1** 20-38. DOI: 10.3390/covid1010004 35. 35.Nguyen D, Simmonds P, Steenhuis M, Wouters E, Desmecht D, Garigliany M et al. SARS-CoV-2 neutralising antibody testing in Europe: towards harmonisation of neutralising antibody titres for better use of convalescent plasma and comparability of trial data.Euro Surveill. 2021;26(27). 36. Prudente TP, Castro RG, Candido MA, Rodrigues RL, de Souza LM, Roberti M. **Antibody response against SARS-CoV-2 in convalescent plasma donors: can we predict subjects’ eligibility?**. *Hematol Transfus Cell Ther* (2022.0) **44** 1-6. DOI: 10.1016/j.htct.2021.07.008 37. 37.Mehew J, Johnson R, Roberts D, Griffiths A, Harvala H. Convalescent plasma for COVID-19: Donor demographic factors associated high neutralising antibody titres.Transfus Med. 2022. 38. Yang HS, Costa V, Racine-Brzostek SE, Acker KP, Yee J, Chen Z. **Association of Age with SARS-CoV-2 antibody response**. *JAMA Netw Open* (2021.0) **4** e214302. DOI: 10.1001/jamanetworkopen.2021.4302 39. Vinkenoog M, Steenhuis M, Brinke AT, van Hasselt JGC, Janssen MP, van Leeuwen M. **Associations between symptoms, Donor characteristics and IgG antibody response in 2082 COVID-19 convalescent plasma donors**. *Front Immunol* (2022.0) **13** 821721. DOI: 10.3389/fimmu.2022.821721 40. Ou X, Jiang J, Lin B, Liu Q, Lin W, Chen G. **Antibody responses to COVID-19 vaccination in people with obesity: a systematic review and meta-analysis**. *Influenza Other Respir Viruses* (2023.0) **17** e13078. DOI: 10.1111/irv.13078 41. Soffer S, Glicksberg BS, Zimlichman E, Efros O, Levin MA, Freeman R. **The association between obesity and peak antibody titer response in COVID-19 infection**. *Obes (Silver Spring)* (2021.0) **29** 1547-53. DOI: 10.1002/oby.23208 42. Wardhani SO, Fajar JK, Nurarifah N, Hermanto DH, Fatonah S, Djajalaksana S. **The predictors of high titer of anti-SARS-CoV-2 antibody of convalescent plasma donors**. *Clin Epidemiol Glob Health* (2021.0) **11** 100763. DOI: 10.1016/j.cegh.2021.100763 43. Gniadek TJ, Thiede JM, Matchett WE, Gress AR, Pape KA, Fiege JK. **SARS-CoV-2 neutralization and serology testing of COVID-19 convalescent plasma from donors with nonsevere disease**. *Transfusion* (2021.0) **61** 17-23. DOI: 10.1111/trf.16101 44. Li L, Tong X, Chen H, He R, Lv Q, Yang R. **Characteristics and serological patterns of COVID-19 convalescent plasma donors: optimal donors and timing of donation**. *Transfusion* (2020.0) **60** 1765-72. DOI: 10.1111/trf.15918 45. Annen K, Morrison TE, DomBourian MG, McCarthy MK, Huey L, Merkel PA. **Presence and short-term persistence of SARS-CoV-2 neutralizing antibodies in COVID-19 convalescent plasma donors**. *Transfusion* (2021.0) **61** 1148-59. DOI: 10.1111/trf.16261 46. Tang MS, Case JB, Franks CE, Chen RE, Anderson NW, Henderson JP. **Association between SARS-CoV-2 neutralizing antibodies and commercial serological assays**. *Clin Chem* (2020.0) **66** 1538-47. DOI: 10.1093/clinchem/hvaa211 47. von Rhein C, Scholz T, Henss L, Kronstein-Wiedemann R, Schwarz T, Rodionov RN. **Comparison of potency assays to assess SARS-CoV-2 neutralizing antibody capacity in COVID-19 convalescent plasma**. *J Virol Methods* (2021.0) **288** 114031. DOI: 10.1016/j.jviromet.2020.114031 48. Goodhue Meyer E, Simmons G, Grebe E, Gannett M, Franz S, Darst O. **Selecting COVID-19 convalescent plasma for neutralizing antibody potency using a high-capacity SARS-CoV-2 antibody assay**. *Transfusion* (2021.0) **61** 1160-70. DOI: 10.1111/trf.16321 49. Harvala H, Robb ML, Watkins N, Ijaz S, Dicks S, Patel M. **Convalescent plasma therapy for the treatment of patients with COVID-19: Assessment of methods available for antibody detection and their correlation with neutralising antibody levels**. *Transfus Med* (2021.0) **31** 167-75. DOI: 10.1111/tme.12746 50. Lamikanra A, Nguyen D, Simmonds P, Williams S, Bentley EM, Rowe C. **Comparability of six different immunoassays measuring SARS-CoV-2 antibodies with neutralizing antibody levels in convalescent plasma: from utility to prediction**. *Transfusion* (2021.0) **61** 2837-43. DOI: 10.1111/trf.16600 51. 51.Walker GJ, Naing Z, Ospina Stella A, Yeang M, Caguicla J, Ramachandran V et al. SARS Coronavirus-2 Microneutralisation and Commercial Serological Assays Correlated Closely for Some but Not All Enzyme Immunoassays.Viruses. 2021;13(2). 52. 52.Luchsinger LL, Ransegnola BP, Jin DK, Muecksch F, Weisblum Y, Bao W et al. Serological Assays Estimate Highly Variable SARS-CoV-2 Neutralizing Antibody Activity in Recovered COVID-19 Patients.J Clin Microbiol. 2020;58(12). 53. Bal A, Pozzetto B, Trabaud MA, Escuret V, Rabilloud M, Langlois-Jacques C. **Evaluation of high-throughput SARS-CoV-2 serological assays in a longitudinal cohort of patients with mild COVID-19: clinical sensitivity, specificity, and Association with Virus Neutralization Test**. *Clin Chem* (2021.0) **67** 742-52. DOI: 10.1093/clinchem/hvaa336 54. Perkmann T, Perkmann-Nagele N, Koller T, Mucher P, Radakovics A, Marculescu R. **Anti-spike protein assays to Determine SARS-CoV-2 antibody levels: a Head-to-Head comparison of five quantitative assays**. *Microbiol Spectr* (2021.0) **9** e0024721. DOI: 10.1128/Spectrum.00247-21 55. 55.Villa HC. COVID-19 Convalescent Plasma Clinical Memorandum December 27, 2021: FDA; 2021 Available from: https://www.fda.gov/media/155159/download.
--- title: Bone marrow stromal cell‐derived exosomes improve oxidative stress and pyroptosis in doxorubicin‐induced myocardial injury in vitro by regulating the transcription of GSDMD through the PI3K‐AKT‐Foxo1 pathway authors: - Hong Zeng - Yong Yang - Fangfang Tou - Yuliang Zhan - Songtao Liu - Pengtao Zou - Yanmei Chen - Liang Shao journal: Immunity, Inflammation and Disease year: 2023 pmcid: PMC10042126 doi: 10.1002/iid3.810 license: CC BY 4.0 --- # Bone marrow stromal cell‐derived exosomes improve oxidative stress and pyroptosis in doxorubicin‐induced myocardial injury in vitro by regulating the transcription of GSDMD through the PI3K‐AKT‐Foxo1 pathway ## Abstract Bone marrow stromal cells ‐derived exosomes phosphorylate forkhead box O1 (FoxO1) by activating the phosphatidylinositol 3 kinase (PI3K)‐protein kinase B (AKT) signaling pathway to repress gasdermin D (GSDMD) transcription, thus ameliorating doxorubicin (DOX)‐induced oxidative stress and pyroptosis in cardiomyocytes. ### Objectives Doxorubicin (DOX) can contribute to severe myocardial injury, and bone marrow stromal cells (BMSC)‐exosomes (Exos) improves acute myocardial infarction. Hence, this research investigated whether BMSC‐Exos alleviated DOX‐induced myocardial injury. ### Methods BMSC‐derived Exos were isolated and identified, and the optimal concentration of DOX was confirmed. H9C2 cells were treated with DOX and BMSC‐Exos or in combination with the protein kinase B (AKT) inhibitor. Reactive oxygen species (ROS) and JC‐1 were detected to assess oxidative stress (OS) and mitochondrial membrane damage, respectively. In addition, the expression of pyroptosis‐related molecules was measured. The expression of phosphatidylinositol 3 kinase (PI3K)‐AKT pathway‐related proteins and the phosphorylation and acetylation of forkhead box O1 (Foxo1) in the cell nucleus and cytoplasm were tested. Last, interactions between Foxo1 and gasdermin D (GSDMD) were assessed. ### Results BMSC‐Exo treatment increased viability and mitochondrial membrane potential and reduced lactic dehydrogenase release and ROS levels in DOX‐treated H9C2 cells. Furthermore, the addition of BMSC‐Exos suppressed DOX‐induced activation and upregulation of NLRP3 and apoptosis‐associated speck‐like protein containing A CARD (ASC) and in vitro cleavage of caspase‐1, GSDMD, interleukin (IL)‐1β, and IL‐18 proteins. Additionally, BMSC‐Exo treatment enhanced the expression of phosphorylated (p)‐PI3K, p‐AKT, and p‐mTOR in DOX‐treated H9C2 cells and the levels of phosphorylated Foxo1 in the cytoplasm of DOX‐treated H9C2 cells. Foxo1 was enriched in the promoter region of GSDMD. Moreover, the AKT inhibitor API‐2 annulled the effects of BMSC‐Exos on OS, pyroptosis, and Foxo1 phosphorylation in DOX‐treated H9C2 cells. ### Conclusions BMSC‐Exos phosphorylated Foxo1 and inactivated Foxo1 transcription via the PI3K‐AKT pathway to diminish GSDMD expression, thus restraining DOX‐induced pyroptosis and OS of myocardial cells. ## INTRODUCTION Isolated from streptomyces peucetius, doxorubicin (DOX) is an effective antibiotic in treatment of various tumors. 1 However, a large dosage use of DOX results in hard‐to‐treat injuries in the myocardium, which may bring about heart dysfunction, severe heart failure, and even death, especially among children and adolescents. 2 DOX was known to be responsible for myocardial cell death in terms of multiple cellular processes such as autophagy, ferroptosis, necroptosis, and apoptosis. 3 Oxidative stress (OS) and pyroptosis are two crucial mechanisms of DOX‐induced myocardial injury. 4, 5, 6 Nevertheless, the currently accepted therapeutic methods for DOX‐induced myocardial injury are limited and none of them obtain completely satisfactory efficacy. 7 To advance the treatment of DOX‐induced myocardial injury, it is necessary to probe the molecular mechanism underlying OS and pyroptosis in the myocardium. Exosomes (Exos) are 30–150 nm extracellular vesicles produced by cells and play crucial roles in cell‐to‐cell communication, normal life activities, and disease diagnosis and treatment. 8, 9 Exos have been documented to confer protection against OS and pyroptosis in diseases. 10, 11 Bone marrow stromal cells (BMSCs) are a heterogeneous mixture of variable subpopulations with functional and molecular properties 12 and Exos derived from these cells were known to manipulate the function and processes of different cells to be implicated in the treatment of diseases. 13, 14 It was reported that BMSC‐derived Exos participate in the alleviation of OS in ischemic stroke and function in reducing inflammasome‐associated pyroptosis. 15, 16 More importantly, previous research stated that BMSC‐Exos improve acute myocardial infarction. 17 As reported, the phosphatidylinositol 3 kinase (PI3K)‐protein kinase B (AKT) pathway participates in the repressive effects of BMSC‐Exos on cell apoptosis in myocardial ischemia–reperfusion injury. 18 The PI3K‐AKT was a vital pathway regulating cellular processes such as cell survival, division, and differentiation. 19 Of note, the PI3K‐AKT pathway was displayed to protect against DOX‐induced cardiotoxicity. 20 Meanwhile, the PI3K‐AKT pathway can phosphorylate forkhead box O1 (Foxo1), one of the downstream molecules of this pathway. 21, 22 Foxo1 is an important molecule regulating transcription and antioxidative enzymes. 23 Additionally, an earlier study uncovered that Foxo1 activation participates in alleviating DOX‐induced cardiac dysfunction. 24 Besides, gasdermin D (GSDMD), a pyroptosis execution protein, was revealed to mediate pyroptosis in spinal cord injury and can be activated by Foxo1. 25, 26 It was illustrated that GSDMD was capable of binding to DOX and mediating pyroptosis of myocardial cells. 27 Based on all these studies, we proposed a hypothesis that BMSC‐Exos modulate GSDMD levels through the PI3K‐AKT‐Foxo1 pathway to affect OS and pyroptosis in DOX‐induced myocardial injury. Thereafter, we conducted a set of experiments to confirm this hypothesis, thus providing novel promising therapeutic targets for DOX cardiotoxicity. ## Cell culture and identification Rat BMSCs and H9C2 cells were obtained respectively from the previous study in our lab and American Type Culture Collection (ATCC) and cultured in a Dulbecco's Modified Eagle Medium (DMEM; Thermo Fisher Scientific) which contained $10\%$ fetal bovine serum (FBS), streptomycin (100 g/mL), and penicillin (100 units/mL) under a condition of 37°C and $5\%$ CO2. BMSCs at the third passage and with $90\%$ confluence were seeded in six‐well plates at 1 × 105 cells/mL. Cells were induced for differentiation separately with adipogenic differentiation and osteogenic differentiation mediums. After 14 days, cells were stained with Oil Red O staining (Sigma‐Aldrich) for detection of adipogenic differentiation of BMSCs. Alizarin Red (Sigma‐Aldrich) was then utilized for staining to measure the osteogenic differentiation of BMSCs 21 days later. BMSCs underwent flow cytometry analysis and probed for 30 min with anti‐CD29‐fluoresceinisothiocyanate (FITC), anti‐CD90‐FITC, anti‐CD44‐FITC, anti‐CD45‐FITC, and anti‐CD34‐FITC. Subsequently, goat anti‐rabbit immunoglobulin G (IgG, ab6717, 1:1000; Abcam) secondary antibodies coupled with FITC fluorochrome were added and cultured for 30 min. A flow cytometer was employed for measurement. ## Isolation and identification of Exos Exos were isolated from BMSC supernatant and incubated for 36 h in a serum‐free medium to avoid the influence of FBS on Exos. After removal of cell debris and related apoptotic debris with differential centrifugation, concentrated cell supernatants were filtered and further characterized by density gradient centrifugation with iodixanol (OptiPrep™; Axis‐Shield). 28 Electron microscope observation was also conducted: Exo resuspensions were precipitated with ultracentrifugation and subjected to 1‐h fixing ($2\%$ paraformaldehyde and $2.5\%$ glutaraldehyde) at 4°C, three washes with phosphate‐buffered saline (PBS, 15 min/time), 1.5 h fixing with $1\%$ osmic acid, and three rinses with PBS (15 min each time) in sequence. Next, after dehydration with graded ethanol and overnight soaking and embedding in epoxy resin, samples underwent 24‐h aggregation at 35°C, 45°C, and 60°C. Samples were then cut into ultrathin sections and treated with lead–uranium double staining, followed by observations under a transmission electron microscope (JEM‐1011; JEOL) at 80 kV voltage of acceleration. Side‐mounted Camera‐Megaview III (Soft Imaging System) was used for photography. Each experiment was repeated three times. Exos were also identified with western blot: Alix, CD63, Golgi Matrix Protein 130 (GM130), cytochrome c, and calnexin expression was tested in BMSCs and BMSC‐derived Exos. ## Nanoparticle tracking analysis (NTA) to observe the concentration and size distribution of Exos NTA was utilized to measure the size and concentration of Exos: Exo samples were resuspended in PBS and diluted 500 times with Milli‐Q water. Afterward, diluted Exo samples were injected into the sample room of NanoSight LM10 (Malvern Panalytical Ltd.) with a sterilized syringe to ensure no air bubbles till the sample room was full. NanoSight LM10 was equipped with a 640‐nm laser and a fluoroelastomer Oring (Viton™; DuPont). NanoSight version 2.3 software (Malvern Panalytical Ltd.) was applied for video analysis, and the motion trails of particles were recorded with the gain value of 6.0 and the threshold value of 11 for all samples. Figures of concentration and size distribution of the diluted samples were output and the Exo concentration of the original liquid was figured out according to the dilution ratio. The experiment was repeated three times. ## Laser scanning microscope (LSM) to examine the uptake of Exos by H9C2 cells BMSCs were seeded in the 24‐well plates at 1 × 105 cells/well, followed by culture in a 37°C incubator containing $5\%$ CO2. Subsequent to 24 h, BMSCs were washed three times with PBS and then cultured for 24 h in a low glucose‐DMEM. Subsequently, Exos were isolated from transfected BMSCs and then labeled with a PKH‐67 fluorochrome kit (MINI67‐1KT; Sigma‐Aldrich). Next, Exos were quantified with a bicinchoninic acid (BCA) kit (Pierce). Thereafter, 1 mL Diluent C solution was added to 200 μg Exos and 4 μL PKH67 fluorochromes, respectively. These two solutions were gently and evenly mixed for 5 min, after which the mixture underwent 2‐h centrifugation at 4°C and 100,000g. The supernatants were then discarded and rinsed twice with PBS. Last, after 2 h of centrifugation at 100,000g (4°C), the labeled Exos were collected. H9C2 cells (5 × 104 cells/well) were spread in the 24‐well plates, incubated overnight, and then cultured for 4 h with PKH67‐labeled Exos at 37°C with $5\%$ CO2, with an equal volume of PBS as the blank control. H9C2 cells were then fixed in methanol and arranged into PBS and Exo groups. LSM 780 (Carl Zeiss) was applied to evaluate the internalization of BMSC‐Exos by H9C2 cells. ## Western blot H9C2 cells were lysed in enhanced radio‐immunoprecipitation assay cell lysis buffer (BOSTER) with protease inhibitor, followed by measurement of protein concentration with the BCA quantitation kit (BOSTER). Proteins were separated with $10\%$ sodium dodecyl sulfate‐polyacrylamide gel electrophoresis, after which they were transferred to polyvinylidene fluoride membranes and sealed for 2 h in $5\%$ bovine serum albumin (BSA) at room temperature to block nonspecific binding. The membranes were supplemented respectively with diluted primary antibodies against Alix (ab275377, 1:1000; Abcam), CD63 (PA5‐92370, 1:1000; Thermo Fisher Scientific), GM130 (PA5‐95727, 1:2000; Thermo Fisher Scientific), cytochrome c (ab133504, 1:1000; Abcam), calnexin (ab133615, 1:1000; Abcam), nucleotide‐binding oligomerization domain, leucine rich repeat and pyrin domain containing 3 (NLRP3, ab263899, 1:1200; Abcam), apoptosis‐associated speck‐like protein containing A CARD (ASC, 307560, 1:1000; Abcam), cleaved caspase‐1 (89332S, 1:1000; Cell Signaling Technologies [CST]), interleukin (IL)‐18 (67775S, 1:1000; CST), cleaved IL‐1β (63124S, 1:1000; CST), active N‐terminal fragment of GSDMS (NT‐GSDMD, 10137S, 1:1000; CST), phosphorylated‐phosphatidylinositol 3‐kinase (p‐PI3K, ab278545, 1:1000; Abcam), p‐protein kinase B (p‐AKT, 4060T, 1:1000; CST), total‐AKT (4685S, 1:1000; CST), p‐mechanistic target of rapamycin kinase (p‐mTOR, ab109268, 1:1000; Abcam), p‐Foxo1 (ab259337, 1:1000; Abcam), acetyl‐Foxo1 (PA5‐104560, 1:1000; Thermo Fisher Scientific), and β‐actin (ab8226, 1:2500; Abcam) for overnight probing at 4°C. Subsequent to washing, the membrane was probed for 1 h with horseradish peroxidase‐labeled secondary antibodies at room temperature. Electrogenerated chemiluminescence (ECL) working solution (Millipore) was utilized to incubate with the membrane for 1 min, after which the redundant ECL reagents were removed and the membrane was sealed with plastic wrap and then exposed for 5–10 min with X‐ray film in a dark box, followed by developing and fixing. ImageJ analysis software (National Institutes of Health [NIH]) was applied to quantify the gray scales of protein bands in western blot figures. β‐actin was regarded as the internal reference. Each experiment was repeated thrice. ## Cell counting kit (CCK)‐8 assay CCK‐8 (#A311‐02‐A; Vazyme Bio‐technology Co., Ltd.) was adopted for cell viability detection in simple and efficient manners as instructed in the manuals of the manufacturer. Briefly speaking, after 24‐h DOX treatment (0.5–20 μM), H9C2 cells were treated for 1 h with CCK‐8 solution. A microplate reader (Imark‐22353; Bio‐Rad) was used to measure the optical density (OD) value at 450 nm. Cell viability = OD value of each group/OD value of the control group × $100\%$. The experiment was repeated independently thrice. ## Lactic dehydrogenase (LDH) release test DOX (0.5–20 μM) was utilized to stimulate H9C2 cells, after which LDH levels in cell supernatants were tested with an LDH cytotoxicity analysis kit (#C0016; Beyotime) per the protocols of the manufacturer. To assess the LDH levels in cell supernatants, the supernatants of H9C2 cells were transferred to a 96‐well plate and each well was added with 60 μL testing solution. Subsequent to 30‐min incubation avoiding light at room temperature, the OD values were determined. The ratio of the OD value to that in the control group was considered the change of multiple. Each experiment was repeated thrice independently. ## Determination of intracellular reactive oxygen species (ROS) After grouping, the 20,70‐dichlorofluorescein diacetate (DCFH‐DA) method was employed for assessment of intracellular ROS levels with a ROS detection kit (#S0033S; Beyotime). In a word, H9C2 cells were stained for 30 min in the dark with DCFH‐DA. Next, a fluorescence microscope was used to monitor cells, and ImageJ software (NIH) to examine the fluorescence intensity of ROS. ## JC‐1 mitoscreen assay Mitochondrial damage was tested with a JC‐1 detection kit (#C2006; Beyotime) in light of instructions from the manufacturer. Cells were incubated with JC solutions and then analyzed with the fluorescence microscope. ImageJ software was applied to quantify the green and red fluorescence intensities in three random fields of view, with the mean value obtained. JC‐1 fluorescence intensity = green fluorescence intensity (mean)/red fluorescence intensity (mean). Three fields of view were randomly selected for each group and the experiment was repeated thrice independently. ## Immunofluorescence After centrifugation, cells were supplemented in fixing agents and permeabilization solution (MultiSciences). H9C2 cells were subject to incubation with $5\%$ BSA, co‐incubation with primary antibodies (NLRP3 and caspase‐1), and overnight probing with secondary antibodies at 4°C. H9C2 cells were then stained for 10 min with 4ʹ,6‐diamidino‐2‐phenylindole (#S0033S; Beyotime). Antifluorescence quenching sealing tablets (#S2100; Solarbio) were utilized for sealing, and a fluorescence microscope (DFM‐90C; Shanghai Cai Kang Optical Instrument Co., Ltd.) was adopted to capture representative images. The green fluorescence intensity in three random fields of view of each group was quantified with ImageJ software and the mean value was figured out. For each experiment, three replicates were conducted. ## Enzyme‐linked immunosorbent assay (ELISA) Contents of IL‐18 (ab213909) and IL‐1β (ab255730) in cell supernatants were detected as directed in the instructions of the manufacturer. ## Dual‐luciferase reporter gene assay The binding site of Foxo1 to the promoter of GSDMD was predicted through the Jaspar database (https://jaspar.genereg.net/). According to the predictive results, wild type sequence (wt‐GSDMD) and mutation sequence (mut‐GSDMD) of the binding site were designed, synthesized, and then respectively inserted into luciferase reporter gene vectors (pGL3‐Promoter), followed by respective co‐transfection with OE‐NC or OE‐Foxo1 into 293T cells. After 48‐h culture, Firefly luciferase activity and Renilla luciferase activity (transfected with phRL‐TK vectors) were measured with luciferase activity detection kits. Renilla luciferase activity was regarded as the internal reference and the ratio of Firefly luciferase activity to Renilla luciferase activity was the relative activity of luciferases. The experiment was repeated three times. ## Chromatin immunoprecipitation (ChIP) assay Subsequent to $4\%$ methanol (final concentration: $1\%$) treatment, H9C2 cells underwent ultrasonication, after which anti‐Foxo1 (ab179450, 1:30; Abcam) and Foxo1‐GSDMD promoter were supplemented for mutual binding. After complete binding through overnight incubation at 4°C, Protein A Agarose/SaLmon Sperm DNA was added to bind to and precipitate Foxo1 antibody‐Foxo1‐GSDMD promoter complexes. The complexes were washed to remove nonspecific bindings and eluted to obtain the enriched Foxo1‐GSDMD promoter complexes. Subsequent to 5‐min centrifugation at 12,000g, the supernatant was discarded, and nonspecific complexes were washed. Samples were subjected to overnight de‐crosslinking at 65°C. Phenol/chloroform was used to extract, purify, and recycle DNA fragments. The enriched fragments of GSDMD promoters were purified, followed by polymerase chain reaction, with IgG (ab109489, 1:100; Abcam) as the negative control. The experiment was repeated thrice. ## Statistical analysis Data were statistically analyzed with GraphPad prism8 software and presented as mean ± standard deviation. After normality analysis of data, the two‐tailed Student's t test was adopted for comparisons between two groups and one‐way analysis of variance for those among multiple groups (except for the experimental groups tested by two‐way analysis of variance as per special instructions). Tukey's test was applied for post hoc multiple comparisons. Differences were statistically significant at $p \leq .05.$ ## The isolation and identification of BMSCs BMSCs were isolated, cultured, and observed under microscopes. The results showed that BMSCs were spindle‐shaped and that the isolated BMSCs exhibited the abilities of osteogenesis, adipogenesis, and chondrogenic differentiation (Figure 1A,B). Flow cytometry detections showed that CD29, CD90, and CD44 had relatively strong positive signals while CD34 and CD45 appeared to be negative (Figure 1C), indicating the successful isolation of BMSCs. Exos were isolated from the BMSC medium and observed to be round or oval membranous vesicles under the transmission electron microscope, and vesicles were surrounded by membranous structures at the periphery, with a low‐density component in the center of the vesicle (Figure 1D). NTA analyses manifested that the particle sizes of major isolated vesicles ranged from 50 to 200 nm (Figure 1E). Western blot results displayed that surface marker proteins Alix and CD63 were expressed but GM130, cytochrome c, and calnexin were not expressed in the isolated BMSC‐Exos (Figure 1F). These results validated that the vesicles isolated from BMSCs were Exos. The uptake condition of Exos by H9C2 cells was assessed under LSM with PBS as the control. No signals of green fluorescence were found in the PBS group of H9C2 cells and PHK67‐labeled Exo fluorescence was observed in the Exo group (Figure 1G), which implicated that BMSC‐Exos were internalized by H9C2 cells. **Figure 1:** *The extraction and identification of BMSCs. (A) Ordinary optical microscope observations of BMSC morphology (×100). (B) Ordinary optical microscope analyses of adipogenic and osteogenic differentiation in BMSCs in vitro (×200). (C) Flow cytometer measurement of surface antigens (positive: CD29, CD90, and CD44; negative: CD45 and CD34). (D) Transmission electron microscope observations of BMSC‐Exos. (E) NTA analyses of BMSC‐Exo size. (F) Western blot examination of the expression of Exo surface marker proteins Alix, CD63, GM130, cytochrome c, and calnexin. (G) Laser scanning microscope evaluation of Exo uptake in H9C2 cells. BMSC, bone marrow stromal cells; Exo, exosome; NTA, nanoparticle tracking analysis.* ## BMSC‐Exos alleviated OS and mitochondrial damage in DOX‐induced H9C2 cells To ascertain whether BMSC‐derived Exos ameliorate DOX‐induced H9C2 cell injury, CCK‐8 and LDH release assays were carried out. The 24‐h treatment of different concentrations of DOX (0–20 μM) dose‐dependently inhibited the viability of H9C2 cells (Figure 2A) and dose‐dependently increased LDH release (Figure 2B). Given these results and previous clinical research, 29 1 μM DOX was selected to perform in vitro experiments for the stable construction of the in vitro cardiotoxicity model. **Figure 2:** *BMSC‐Exos improves OS and mitochondrial damage of DOX‐treated H9C2 cells. (A) CCK‐8 assay to test viability of H9C2 cells 24 h after treatments of different concentrations of DOX (0–20 μM). (B) LDH release of H9C2 cells 24 h after treatment of DOX at above concentrations. (C) CCK‐8 assay to measure changes of cell viability after treatment of 1 μM DOX. (D) LDH release of each group. (E, F) DCFH‐DA staining to test ROS production of cells. (G, H) JC‐1 immunofluorescence to measure the changes of mitochondrial membrane potentials of cells. ***p < .001 versus the control group or the 0 μM group; ## p < .01; ### p < .001 versus the DOX + BMSC group. One‐way analysis of variance was adopted to confirm p values, with Tukey's test for post hoc multiple comparisons. Each experiment was independently repeated thrice. BMSC, bone marrow stromal cells; CCK, cell counting kit; DCFH‐DA, 20,70‐dichlorofluorescein diacetate; DOX, doxorubicin; Exo, exosome; LDH, lactic dehydrogenase; OS, oxidative stress; ROS, reactive oxygen species.* Next, BMSC‐derived Exos and DOX were simultaneously used to treat H9C2 cells. The results of CCK‐8 and LDH release assays indicated that the cell viability was memorably lowered and LDH release was increased in H9C2 cells by DOX treatment, while BMSC‐Exos evidently enhanced cell viability and reduced LDH release in DOX‐treated cells (Figure 2C,D). DCFH‐DA probe examination of ROS contents displayed that single treatment of DOX signally augmented ROS contents in H9C2 cells, whereas treatment of BMSC‐Exos markedly diminished ROS contents in DOX‐treated H9C2 cells. Meanwhile, treatment of BMSCs did not change ROS levels in DOX‐treated H9C2 cells (Figure 2E,F). Additionally, immunofluorescence was adopted to measure the mitochondrial membrane potential of different groups to evaluate the mitochondrial damage. JC‐1 is a fluorescent lipophilic carbonyl blue dye used for the measurement of mitochondrial membrane potential. When the mitochondrial membrane potential is relatively high, JC‐1 converges in matrix to form J‐aggregates, generating red fluorescence (excitation wavelength [Ex]/emission wavelength [Em] = $\frac{585}{590}$ nm). When the potential is relatively low, JC‐1 fails to converge in matrix and then exists in the form of monomer, thus producing green fluorescence (Ex/Em = $\frac{510}{527}$ nm). In this way, the reduction in mitochondrial membrane potentials can be easily observed through the change of JC‐1 from red fluorescence to green fluorescence. Results reflected that BMSC‐Exos distinctly reversed the DOX‐induced decrease in mitochondrial membrane potentials of H9C2 cells (the ratio of green fluorescence to red fluorescence decreased) (Figure 2G,H). These results indicated that BMSC‐Exos reduced DOX‐induced OS and mitochondrial damage in H9C2 cells. ## BMSC‐Exos improved DOX‐induced H9C2 cell pyroptosis Whether BMSCs‐derived Exos inhibited pyroptosis in H9C2 cells was further explored in our research. Western blot results displayed that the treatment of DOX markedly induced the activation of NLRP3 and ASC and the cleavage of caspase‐1, GSDMD, IL‐1β, and IL‐18 proteins, which was suppressed by the further addition of BMSC‐Exos (Figure 3A,B). Meanwhile, ELISA detections of IL‐18 and IL‐1β contents in cell supernatants revealed coincident results (Figure 3C,D). Immunofluorescence demonstrated that BMSC‐Exos diminished DOX‐caused elevations in positive rates of NLRP3 and caspase‐1 (Figure 3E,F). The above data indicated that BMSC‐Exos ameliorated DOX‐induced pyroptosis of H9C2 cells. **Figure 3:** *BMSC‐Exos ameliorates DOX‐induced H9C2 cell pyroptosis. (A, B) Western blot detection of the protein expression of NLRP3, ASC, cleaved caspase‐1, cleaved IL‐1β, cleaved IL‐18, and NT‐GSDMD in cells. (C, D) ELISA to test IL‐18 and IL‐1β levels in cell supernatants. (E, F) Immunofluorescence to examine the positive rates of NLRP3 and caspase‐1. ***p < .001 versus the control group; ## p < .01; ### p < .001 versus the DOX + BMSC group. One‐way analysis of variance was adopted to confirm p values, with Tukey's test for post hoc multiple comparisons. Each experiment was independently repeated thrice. ASC, apoptosis‐associated speck‐like protein containing A CARD; BMSC, bone marrow stromal cells; DOX, doxorubicin; ELISA, enzyme‐linked immunosorbent assay; Exo, exosome; IL, interleukin; NLRP3, nucleotide‐binding oligomerization domain, leucine rich repeat and pyrin domain containing 3; NT‐GSDMD, active N‐terminal fragment of gasdermin D.* ## BMSC‐Exos reduced the transcription of GSDMD via the PI3K‐AKT‐Foxo1 axis in DOX‐treated H9C2 cells Next, we explored the mechanism of BMSC‐Exos in DOX‐induced myocardial injury. Western blot results manifested that the treatment of BMSC‐Exos conspicuously increased the levels of p‐PI3K, p‐AKT, and p‐mTOR in DOX‐treated H9C2 cells (Figure 4A). In addition, western blot data uncovered insignificant differences in total protein expression of Foxo1 in both cytoplasm and nucleus (Figure 4B), and acetylated Foxo1 expression exhibited a similar condition (Figure 4C). However, BMSC‐Exos treatment evidently enhanced the levels of phosphorylated Foxo1 in the cytoplasm of DOX‐treated H9C2 cells, accompanied by no significant difference in the levels of phosphorylated Foxo1 in the nucleus (Figure 4C). These findings illustrated BMSC‐Exos might phosphorylate Foxo1 through the PI3K‐AKT pathway to promote the translocation of Foxo1 from the nucleus to the cytoplasm and inactivate its transcription, thus suppressing Foxo1‐modulated downstream gene expression and further regulating OS and pyroptosis in H9C2 cells. In addition, it was found via a public database that Foxo1, as a transcription factor, regulated the transcription of GSDMD (Figure 4D). The dual‐luciferase reporter gene assay unveiled that OE‐Foxo1 treatment prominently elevated the luciferase activity of 293T cells transfected with wt‐GSDMD promoter sequence‐inserted reporter vectors but did not alter the luciferase activity of 293T cells transfected with mut‐GSDMD promoter sequence‐inserted reporter vectors (Figure 4E). ChIP assay results showed that Foxo1 remarkably enriched in the promoter region of GSDMD (Figure 4F). Conclusively, BMSC‐Exos repressed GSDMD transcription through the PI3K‐AKT‐Foxo1 axis in DOX‐treated H9C2 cells. **Figure 4:** *BMSC‐Exos regulates GSDMD transcription through the PI3K‐AKT‐Foxo1 pathway in DOX‐induced H9C2 cells. (A) Western blot to examine the expression of p‐PI3K, p‐AKT, total‐AKT, and p‐mTOR in cells. *p < .05; **p < .01; ***p < .001 versus the DOX + BMSC group. (B) Western blot to measure the protein expression levels of total‐Foxo1 in cell nucleus and cytoplasm. (C) Western blot to determine the expression of acetylated Foxo1, phosphorylated Foxo1, and total‐Foxo1,  ***p < .001 versus the DOX + BMSC group. (D) Public database to analyze the binding site sequence of Foxo1 to GSDMD. (E) Dual‐luciferase reporter gene assay to validate binding of Foxo1 to GSDMD promoter,  ***p < .001 versus the OE‐NC group. (F) ChIP assay to test the enrichment level of Foxo1 in the promoter region of GSDMD, ***p < .001 versus the anti‐IgG group. The two‐tailed test was employed to confirm p values. Except for special statements, one‐way analysis of variance was adopted to confirm p values, with Tukey's test for post hoc multiple comparisons. Each experiment was independently repeated thrice. AKT, protein kinase B; BMSC, bone marrow stromal cells; ChIP, chromatin immunoprecipitation; DOX, doxorubicin; Exo, exosome; GSDMD, gasdermin D; IgG, immunoglobulin G; mTOR, mechanistic target of rapamycin kinase; OS, oxidative stress; p, phosphorylated; PI3K, phosphatidylinositol 3‐kinase.* ## Inhibition of the PI3K‐AKT pathway by API‐2 nullified the alleviatory impact of BMSC‐Exos on DOX‐induced OS and pyroptosis in H9C2 cells To further clarify whether BMSC‐Exos regulated GSDMD transcription via the PI3K‐AKT‐Foxo1 to improve OS and pyroptosis in DOX‐induced myocardial injury, the selective inhibitor of AKT, API‐2, was chosen to treat H9C2 cells after Exo treatment. DCFH‐DA staining exhibited that the ROS level was enormously augmented in the DOX + BMSC‐Exo + API‐2 group versus the DOX + BMSC‐Exo group (Figure 5A,B). JC‐1 immunofluorescence demonstrated that in contrast to the DOX + BMSC‐Exo group, the potential difference of mitochondrial membrane was observably lowered in the DOX + BMSC‐Exo + API‐2 group (the ratio of green fluorescence to red fluorescence was enhanced) (Figure 5C,D). Meanwhile, western blot results revealed that the activation of NLRP3 and ASC and in vitro cleavage of caspase‐1, GSDMD, IL‐1β, and IL‐18 proteins distinctly increased in the DOX + BMSC‐Exo + API‐2 group in comparison with the DOX + BMSC‐Exo group (Figure 5E,F). Western blot showed that API‐2 markedly decreased the phosphorylation levels of Foxo1 in the cytoplasm of H9C2 cells (Figure 5G,H). In summary, API‐2 blocked the improving effect of BMSC‐Exos on DOX‐induced OS and pyroptosis in H9C2 cells. **Figure 5:** *Inactivation of the PI3K‐AKT pathway abrogates the relieving impacts of BMSC‐Exos on DOX‐induced OS and pyroptosis in H9C2 cells. (A, B) DCFH‐DA staining to measure ROS levels in cells. (C, D) JC‐1 immunofluorescence to monitor the changes of potential difference of mitochondrial membrane in cells. (E, F) Western blot to test the expression of NLRP3, ASC, cleaved caspase‐1, NT‐GSDMD, cleaved IL‐1β, and cleaved IL‐18 in cells. (G, H) Western blot to measure the phosphorylation levels and total protein levels of Foxo1 in cell nucleus and cytoplasm; ***p < .001 versus the control group; ## p < .01; ### p < .001 versus the DOX group; & p < .05; && p < .01; &&& p < .001 versus the DOX + BMSC‐Exo group. One‐way analysis of variance was adopted to confirm p values, with Tukey's test for post hoc multiple comparisons. Each experiment was independently repeated thrice. AKT, protein kinase B; ASC, apoptosis‐associated speck‐like protein containing A CARD; BMSC, bone marrow stromal cells; DCFH‐DA, 20,70‐dichlorofluorescein diacetate; DOX, doxorubicin; Exo, exosome; Foxo1, forkhead box O1; GSDMD, gasdermin D; IL, interleukin; NLRP3, nucleotide‐binding oligomerization domain, leucine rich repeat and pyrin domain containing 3; OS, oxidative stress; PI3K, phosphatidylinositol 3‐kinase; ROS, reactive oxygen species.* ## DISCUSSION DOX causes myocardial injury, to which OS and pyroptosis were closely linked. 30, 31 Recent research revealed that MSC‐derived Exos protect against myocardial infarction and myocardial ischemia–reperfusion injury, so increasing attention is paid to their potential therapeutic values in treating myocardial problems. 32, 33 To probe the specific role of BMSC‐Exos in DOX cardiotoxicity, this study was conducted to ascertain the function of BMSC‐Exos in OS and pyroptosis during DOX‐induced myocardial injury and the related mechanism. It was elucidated from our data that BMSC‐Exos ameliorated OS and pyroptosis in DOX‐induced myocardial injury via the PI3K‐AKT‐Foxo1‐GSDMD axis. BMSC‐*Exos is* implicated in the suppression of DOX‐induced cardiotoxicity. 34 Therefore, to delve into the specific role of BMSC‐Exos in DOX cardiotoxicity, Exos were isolated from rat BMSCs, and the following experiments were conducted. LDH is an enzyme that plays a part in myocardial ischemia–reperfusion injury. 35, 36 ROS is a product in the pro‐/antioxidant balance and contributes to harmful OS when excessively generated, the clearance of which is crucial for the improvement of myocardial injury. 37, 38 JC‐1 is a cationic fluorescent dye to assess the mitochondrial membrane potential that reflects the cytotoxicity when lost. 39, 40 Therefore, these three factors were selected for the detection of OS and mitochondrial damage, along with the measurement of cell viability. The results elucidated that BMSC‐Exo treatment elevated viability and red‐to‐green fluorescence ratio of JC‐1 and downregulated ROS levels in DOX‐treated H9C2 cells. Intriguingly, a former study showed accordant trends that addition of BMSC‐Exos increased proliferation and mitochondrial potential and decreased ROS levels in hypoxia/reoxygenation‐treated myocardial cells. 41 Meanwhile, NLRP3 is an inflammasome participating in cell pyroptosis. 42 ASC is an adaptor protein required for driving pyroptosis. 43 Caspase‐1 is an inflammatory mediated enzyme, which cleaves and activates inflammatory factors. 44 Along with the inflammatory factors IL‐1β and IL‐18, caspase‐1 contributes to pyroptosis. 45, 46, 47 Therefore, the aforementioned factors were all tested in our experiments to evaluate pyroptosis during DOX‐induced myocardial injury. Earlier research discovered that BMSC‐Exos reduced the level of IL‐1β in H2O2‐treated H9c2 cells, 48 which was concurrent with our results. Additionally, we also observed that BMSC‐Exos suppressed the activation of NLRP3 and ASC and the expression of caspase‐1, GSDMD, and IL‐18. Zeng et al. also concluded that BMSC‐Exos reduced pyroptosis in cerebral ischemia–reperfusion injury. 49 It has been documented that the PI3K‐AKT pathway plays a vital role in protecting against myocardial injury. 50 Furthermore, DOX blocked the PI3K‐AKT‐mTOR pathway to induce myocardial injury. 51 In addition, the PI3K‐AKT pathway activation is involved in the alleviation of OS‐induced apoptosis and ischemia/reperfusion‐induced pyroptosis. 52, 53 Of note, BMSC‐Exos has been validated to activate the PI3K‐AKT pathway to reduce myocardial ischemia–reperfusion injury. 54 Concordantly, our data also uncovered that BMSC‐Exo treatment increased the levels of p‐PI3K and p‐AKT in cells and p‐Foxo1 in the cytoplasm of cells. Previous documentation concluded that posttranslational modifications of Foxo1 are related to myocardial ischemia 55 and that the knockdown of Foxo1 attenuates DOX‐induced myocardial injury. 56 In addition, Foxo1 can be phosphorylated by AKT, resulting in the translocation of Foxo1 from the nucleus to the cytoplasm. 57 Additionally, a previous study revealed that Foxo1 can bind to GSDMD promoter during myocardial damage. 58 Of note, our observations also verified the binding relationship between Foxo1 and GSDMD. More importantly, GSDMD is an important pyroptosis mediator and DOX directly binds to GSDMD to promote pyroptosis of myocardial cells. 27 In addition, GSDMD axis is deeply associated with OS and pyroptosis of myocardial cells in myocardial infarction. 59 These discussions confirmed the linkages among PI3K‐AKT, Foxo1, and GSDMD in DOX‐induced myocardial injury and implicated their involvement in the alleviatory effects of BMSC‐Exo on DOX‐induced myocardial injury. Therefore, the inhibitor of AKT, API‐2, 60 was further used in our research for assessing the effect of this pathway on OS and pyroptosis in DOX‐induced myocardial injury. Our data unraveled that API‐2 exaggerated OS and pyroptosis mitigated by BMSC‐Exos in DOX‐induced myocardial injury. To be specific, API‐2 treatment elevated ROS generation, lowered the mitochondrial membrane potential, and enhanced the protein levels of cleavage‐caspase‐1, GSDMD, IL‐1β, and IL‐18 in H9C2 treated with DOX + BMSC‐Exos, accompanied by the diminished phosphorylated levels of Foxo1 in the cytoplasm. In a word, it can be implicated from our results that BMSC‐Exos improved OS and pyroptosis in DOX‐induced myocardial injury by downregulating GSDMD via the PI3K‐AKT‐Foxo1 pathway. However, we only probed this mechanism in the cellular model. Further related animal and clinical studies are warranted to be carried out to further validate our findings. Nevertheless, our experiment results might assist in providing data for promoting investigations on the therapeutic potential of BMSC‐Exos in myocardial injury. ## AUTHOR CONTRIBUTIONS SL, ZH, YY and TFF conceived the ideas; designed the experiments. ZH, YY, TFF, ZYL and LST performed the experiments. ZH, YY, TFF, ZPT and CYM analysed the data. ZYL and LST provided critical materials. ZPT and CYM wrote the manuscript. SL supervised the study. All the authors have read and approved the final version for publication. ## CONFLICT OF INTEREST STATEMENT The authors declare no conflict of interest. ## DATA AVAILABILITY STATEMENT The data sets used or analyzed during the current study are available from the corresponding author on reasonable request. ## References 1. Yarmohammadi F, Rezaee R, Karimi G. **Natural compounds against doxorubicin‐induced cardiotoxicity: a review on the involvement of Nrf2/ARE signaling pathway**. *Phytother Res* (2021) **35** 1163-1175. PMID: 32985744 2. Zhu JG, Jin K, Ren Y. **Ulinastatin reduces myocardial injury induced by doxorubicin in SD rats**. *Eur Rev Med Pharmacol Sci* (2020) **24** 10769-10778. PMID: 33155237 3. Christidi E, Brunham LR. **Regulated cell death pathways in doxorubicin‐induced cardiotoxicity**. *Cell Death Dis* (2021) **12** 339. PMID: 33795647 4. Songbo M, Lang H, Xinyong C, Bin X, Ping Z, Liang S. **Oxidative stress injury in doxorubicin‐induced cardiotoxicity**. *Toxicol Lett* (2019) **307** 41-48. PMID: 30817977 5. Tian L, Cao W, Yue R. **Pretreatment with Tilianin improves mitochondrial energy metabolism and oxidative stress in rats with myocardial ischemia/reperfusion injury via AMPK/SIRT1/PGC‐1 alpha signaling pathway**. *J Pharmacol Sci* (2019) **139** 352-360. PMID: 30910451 6. Zhang L, Jiang YH, Fan C. **MCC950 attenuates doxorubicin‐induced myocardial injury in vivo and in vitro by inhibiting NLRP3‐mediated pyroptosis**. *Biomed Pharmacother* (2021) **143**. PMID: 34474337 7. Kouzi SA, Uddin MN. **Aerobic exercise training as a potential cardioprotective strategy to attenuate doxorubicin‐induced cardiotoxicity**. *J Pharm Pharm Sci* (2016) **19** 399-410. PMID: 27806245 8. Yang D, Zhang W, Zhang H. **Progress, opportunity, and perspective on exosome isolation—efforts for efficient exosome‐based theranostics**. *Theranostics* (2020) **10** 3684-3707. PMID: 32206116 9. Zhang Y, Bi J, Huang J, Tang Y, Du S, Li P. **Exosome: areview of its classification, isolation techniques, storage, diagnostic and targeted therapy applications**. *Int J Nanomedicine* (2020) **15** 6917-6934. PMID: 33061359 10. Wang T, Jian Z, Baskys A. **MSC‐derived exosomes protect against oxidative stress‐induced skin injury via adaptive regulation of the NRF2 defense system**. *Biomaterials* (2020) **257**. PMID: 32791387 11. Yan B, Zhang Y, Liang C. **Stem cell‐derived exosomes prevent pyroptosis and repair ischemic muscle injury through a novel exosome/circHIPK3/FOXO3a pathway**. *Theranostics* (2020) **10** 6728-6742. PMID: 32550900 12. Sivaraj KK, Jeong HW, Dharmalingam B. **Regional specialization and fate specification of bone stromal cells in skeletal development**. *Cell Rep* (2021) **36**. PMID: 34260921 13. Shi Y, Kang X, Wang Y. **Exosomes derived from bone marrow stromal cells (BMSCs) enhance tendon‐bone healing by regulating macrophage polarization**. *Med Sci Monit* (2020) **26**. PMID: 32369458 14. Tang S, Tang T, Gao G, Wei Q, Sun K, Huang W. **Bone marrow mesenchymal stem cell‐derived exosomes inhibit chondrocyte apoptosis and the expression of MMPs by regulating Drp1‐mediated mitophagy**. *Acta Histochem* (2021) **123**. PMID: 34656826 15. Yang H, Chen J. **Bone marrow mesenchymal stem cell‐derived exosomes carrying long noncoding RNA ZFAS1 alleviate oxidative stress and inflammation in ischemic stroke by inhibiting microRNA‐15a‐5p**. *Metab Brain Dis* (2022) **37** 2545-2557. PMID: 35907132 16. Zhao Y, Chen Y, Wang Z. **Bone marrow mesenchymal stem cell exosome attenuates inflammasome‐related pyroptosis via delivering circ_003564 to improve the recovery of spinal cord injury**. *Mol Neurobiol* (2022) **59** 6771-6789. PMID: 36038697 17. Xiao M, Zeng W, Wang J. **Exosomes protect against acute myocardial infarction in rats by regulating the renin‐angiotensin system**. *Stem Cells Dev* (2021) **30** 622-631. PMID: 33765842 18. Sun XH, Wang X, Zhang Y, Hui J. **Exosomes of bone‐marrow stromal cells inhibit cardiomyocyte apoptosis under ischemic and hypoxic conditions via miR‐486‐5p targeting the PTEN/PI3K/AKT signaling pathway**. *Thromb Res* (2019) **177** 23-32. PMID: 30844685 19. Jafari M, Ghadami E, Dadkhah T, Akhavan‐Niaki H. **PI3k/AKT signaling pathway: erythropoiesis and beyond**. *J Cell Physiol* (2019) **234** 2373-2385. PMID: 30192008 20. Sun L, Wang H, Xu D, Yu S, Zhang L, Li X. **Lapatinib induces mitochondrial dysfunction to enhance oxidative stress and ferroptosis in doxorubicin‐induced cardiomyocytes via inhibition of PI3K/AKT signaling pathway**. *Bioengineered* (2022) **13** 48-60. PMID: 34898356 21. Li F, Xie W, Fang Y. **HnRNP‐F promotes the proliferation of bladder cancer cells mediated by PI3K/AKT/FOXO1**. *J Cancer* (2021) **12** 281-291. PMID: 33391425 22. Yan Y, Huang H. **Interplay among PI3K/AKT, PTEN/FOXO and AR signaling in prostate cancer**. *Adv Exp Med Biol* (2019) **1210** 319-331. PMID: 31900915 23. Xing Y, Li A, Yang Y, Li X, Zhang L, Guo H. **The regulation of FOXO1 and its role in disease progression**. *Life Sci* (2018) **193** 124-131. PMID: 29158051 24. Yan F, Liu R, Zhuang X, Li R, Shi H, Gao X. **Salidroside attenuates doxorubicin‐induced cardiac dysfunction partially through activation of QKI/FoxO1 pathway**. *J Cardiovasc Transl Res* (2021) **14** 355-364. PMID: 32671648 25. Wang K, Sun Q, Zhong X. **Structural mechanism for GSDMD targeting by autoprocessed caspases in pyroptosis**. *Cell* (2020) **180** 941-955. PMID: 32109412 26. Xu S, Wang J, Zhong J. **CD73 alleviates GSDMD‐mediated microglia pyroptosis in spinal cord injury through PI3K/AKT/Foxo1 signaling**. *Clin Transl Med* (2021) **11**. PMID: 33463071 27. Ye B, Shi X, Xu J. **Gasdermin D mediates doxorubicin‐induced cardiomyocyte pyroptosis and cardiotoxicity via directly binding to doxorubicin and changes in mitochondrial damage**. *Transl Res* (2022) **248** 36-50. PMID: 35545198 28. Willis GR, Fernandez‐Gonzalez A, Anastas J. **Mesenchymal stromal cell exosomes ameliorate experimental bronchopulmonary dysplasia and restore lung function through macrophage immunomodulation**. *Am J Respir Crit Care Med* (2018) **197** 104-116. PMID: 28853608 29. Priya LB, Baskaran R, Huang CY, Padma VV. **Neferine ameliorates cardiomyoblast apoptosis induced by doxorubicin: possible role in modulating NADPH oxidase/ROS‐mediated NFκB redox signaling cascade**. *Sci Rep* (2017) **7**. PMID: 28947826 30. Abdelghffar EA, Obaid WA, Elgamal AM, Daoud R, Sobeh M, El Raey MA. **Pea (**. *Biomed Pharmacother* (2021) **143**. PMID: 34649330 31. Zheng X, Zhong T, Ma Y. **Bnip3 mediates doxorubicin‐induced cardiomyocyte pyroptosis via caspase‐3/GSDME**. *Life Sci* (2020) **242**. PMID: 31862454 32. Zhang J, Lu Y, Mao Y. **IFN‐γ enhances the efficacy of mesenchymal stromal cell‐derived exosomes via miR‐21 in myocardial infarction rats**. *Stem Cell Res Ther* (2022) **13** 333. PMID: 35870960 33. Zhao J, Li X, Hu J. **Mesenchymal stromal cell‐derived exosomes attenuate myocardial ischaemia‐reperfusion injury through miR‐182‐regulated macrophage polarization**. *Cardiovasc Res* (2019) **115** 1205-1216. PMID: 30753344 34. Tian C, Yang Y, Li B. **Doxorubicin‐induced cardiotoxicity may be alleviated by bone marrow mesenchymal stem cell‐derived exosomal lncRNA via inhibiting inflammation**. *J Inflamm Res* (2022) **15** 4467-4486. PMID: 35966005 35. Chen W, Wei L, Zhang Y. **Involvement of the two l‐lactate dehydrogenase in development and pathogenicity in fusarium graminearum**. *Curr Genet* (2019) **65** 591-605. PMID: 30474697 36. Ding M, Wang Y, Sun D. **Punicalagin pretreatment attenuates myocardial ischemia‐reperfusion injury via activation of AMPK**. *Am J Chin Med* (2017) **45** 53-66. PMID: 28081629 37. Daenen K, Andries A, Mekahli D, Van Schepdael A, Jouret F, Bammens B. **Oxidative stress in chronic kidney disease**. *Pediatr Nephrol* (2019) **34** 975-991. PMID: 30105414 38. Zheng Y, Shi B, Ma M, Wu X, Lin X. **The novel relationship between Sirt3 and autophagy in myocardial ischemia‐reperfusion**. *J Cell Physiol* (2019) **234** 5488-5495. PMID: 30485429 39. Elefantova K, Lakatos B, Kubickova J, Sulova Z, Breier A. **Detection of the mitochondrial membrane potential by the cationic dye JC‐1 in L1210 cells with massive overexpression of the plasma membrane ABCB1 drug transporter**. *Int J Mol Sci* (2018) **19** 1985. PMID: 29986516 40. Gharbaran R, Shi C, Onwumere O, Redenti S. **Plumbagin induces cytotoxicity via loss of mitochondrial membrane potential and caspase activation in metastatic retinoblastoma**. *Anticancer Res* (2021) **41** 4725-4732. PMID: 34593421 41. Zhang JK, Zhang Z, Guo ZA. **The BMSC‐derived exosomal lncRNA Mir9‐3hg suppresses cardiomyocyte ferroptosis in ischemia‐reperfusion mice via the Pum2/PRDX6 axis**. *Nutr Metab Cardiovasc Dis* (2022) **32** 515-527. PMID: 34953631 42. Shen S, He F, Cheng C, Xu B, Sheng J. **Uric acid aggravates myocardial ischemia‐reperfusion injury via ROS/NLRP3 pyroptosis pathway**. *Biomed Pharmacother* (2021) **133**. PMID: 33232925 43. Man SM, Kanneganti TD. **Regulation of inflammasome activation**. *Immunol Rev* (2015) **265** 6-21. PMID: 25879280 44. Molla MD, Ayelign B, Dessie G, Geto Z, Admasu TD. **Caspase‐1 as a regulatory molecule of lipid metabolism**. *Lipids Health Dis* (2020) **19** 34. PMID: 32143623 45. Libby P. **Interleukin‐1 beta as a target for atherosclerosis therapy**. *J Am Coll Cardiol* (2017) **70** 2278-2289. PMID: 29073957 46. Ni XQ, Hu ZY. **Remifentanil improves myocardial ischemia‐reperfusion injury in rats through inhibiting IL‐18 signaling pathway**. *Eur Rev Med Pharmacol Sci* (2020) **24** 3915-3922. PMID: 32329867 47. Rauf A, Shah M, Yellon DM, Davidson SM. **Role of caspase 1 in ischemia/reperfusion injury of the myocardium**. *J Cardiovasc Pharmacol* (2019) **74** 194-200. PMID: 31356550 48. Liu HY, Yu LF, Zhou TG. **Lipopolysaccharide‐stimulated bone marrow mesenchymal stem cells‐derived exosomes inhibit H2O2‐induced cardiomyocyte inflammation and oxidative stress via regulating miR‐181a‐5p/ATF2 axis**. *Eur Rev Med Pharmacol Sci* (2020) **24** 10069-10077. PMID: 33090414 49. Zeng Q, Zhou Y, Liang D. **Exosomes secreted from bone marrow mesenchymal stem cells attenuate oxygen‐glucose deprivation/reoxygenation‐induced pyroptosis in PC12 cells by promoting AMPK‐dependent autophagic flux**. *Front Cell Neurosci* (2020) **14** 182. PMID: 32765221 50. Ghafouri‐Fard S, Khanbabapour Sasi A, Hussen BM. **Interplay between PI3K/AKT pathway and heart disorders**. *Mol Biol Rep* (2022) **49** 9767-9781. PMID: 35499687 51. Yu W, Qin X, Zhang Y. **Curcumin suppresses doxorubicin‐induced cardiomyocyte pyroptosis via a PI3K/Akt/mTOR‐dependent manner**. *Cardiovasc Diagn Ther* (2020) **10** 752-769. PMID: 32968631 52. Diao MY, Zhu Y, Yang J. **Hypothermia protects neurons against ischemia/reperfusion‐induced pyroptosis via m6A‐mediated activation of PTEN and the PI3K/Akt/GSK‐3β signaling pathway**. *Brain Res Bull* (2020) **159** 25-31. PMID: 32200003 53. Gong Y, Luo S, Fan P, Zhu H, Li Y, Huang W. **Growth hormone activates PI3K/Akt signaling and inhibits ROS accumulation and apoptosis in granulosa cells of patients with polycystic ovary syndrome**. *Reprod Biol Endocrinol* (2020) **18** 121. PMID: 33287836 54. Zhang L, Wei Q, Liu X. **Exosomal microRNA‐98‐5p from hypoxic bone marrow mesenchymal stem cells inhibits myocardial ischemia‐reperfusion injury by reducing TLR4 and activating the PI3K/Akt signaling pathway**. *Int Immunopharmacol* (2021) **101**. PMID: 34715573 55. Kappel BA, Stöhr R, De Angelis L, Mavilio M, Menghini R, Federici M. **Posttranslational modulation of FoxO1 contributes to cardiac remodeling in post‐ischemic heart failure**. *Atherosclerosis* (2016) **249** 148-156. PMID: 27105158 56. Xia P, Chen J, Liu Y, Fletcher M, Jensen BC, Cheng Z. **Doxorubicin induces cardiomyocyte apoptosis and atrophy through cyclin‐dependent kinase 2‐mediated activation of forkhead box O1**. *J Biol Chem* (2020) **295** 4265-4276. PMID: 32075913 57. Zhang W, Hou C, Du L. **Protective action of pomegranate peel polyphenols in type 2 diabetic rats via the translocation of Nrf2 and FoxO1 regulated by the PI3K/Akt pathway**. *Food Funct* (2021) **12** 11408-11419. PMID: 34673854 58. Zhang B, Sun C, Liu Y, Bai F, Tu T, Liu Q. **Exosomal miR‐27b‐3p derived from hypoxic cardiac microvascular endothelial cells alleviates rat myocardial ischemia/reperfusion injury through inhibiting oxidative stress‐induced pyroptosis via Foxo1/GSDMD signaling**. *Oxid Med Cell Longevity* (2022) **2022** 1-19 59. Lei Q, Yi T, Chen C. **NF‐κB‐Gasdermin D (GSDMD) axis couples oxidative stress and NACHT, LRR and PYD domains‐containing protein 3 (NLRP3) inflammasome‐mediated cardiomyocyte pyroptosis following myocardial infarction**. *Med Sci Monit* (2018) **24** 6044-6052. PMID: 30161099 60. Chen J, Lian X, Du J. **Inhibition of phosphorylated Ser473‐Akt from translocating into the nucleus contributes to 2‐cell arrest and defective zygotic genome activation in mouse preimplantation embryogenesis**. *Dev Growth Differ* (2016) **58** 280-292. PMID: 27004788
--- title: Short‐Term Effects of Carbon Monoxide on Morbidity of Chronic Obstructive Pulmonary Disease With Comorbidities in Beijing authors: - Zhiwei Li - Feng Lu - Mengmeng Liu - Moning Guo - Lixin Tao - Tianqi Wang - Mengyang Liu - Xiuhua Guo - Xiangtong Liu journal: GeoHealth year: 2023 pmcid: PMC10042128 doi: 10.1029/2022GH000734 license: CC BY 4.0 --- # Short‐Term Effects of Carbon Monoxide on Morbidity of Chronic Obstructive Pulmonary Disease With Comorbidities in Beijing ## Abstract The association between CO and chronic obstructive pulmonary disease (COPD) has been widely reported; however, the association among patients with type 2 diabetes mellitus (T2DM) or hypertension has remained largely unknown in China. Over‐dispersed generalized additive model was adopted to quantity the associations between CO and COPD with T2DM or hypertension. Based on principal diagnosis, COPD cases were identified according to the International Classification of Diseases (J44), and a history of T2DM and hypertension was coded as E12 and I10‐15, O10‐15, P29, respectively. A total of 459,258 COPD cases were recorded from 2014 to 2019. Each interquartile range uptick in CO at lag 03 corresponded to $0.21\%$ ($95\%$CI: $0.08\%$–$0.34\%$), $0.39\%$ ($95\%$CI: $0.13\%$–$0.65\%$), $0.29\%$ ($95\%$CI: $0.13\%$–$0.45\%$) and $0.27\%$ ($95\%$CI: $0.12\%$–$0.43\%$) increment in admissions for COPD, COPD with T2DM, COPD with hypertension and COPD with both T2DM and hypertension, respectively. The effects of CO on COPD with T2DM ($Z = 0.77$, $$P \leq 0.444$$), COPD with hypertension ($Z = 0.19$, $$P \leq 0.234$$) and COPD with T2DM and hypertension ($Z = 0.61$, $$P \leq 0.543$$) were insignificantly higher than that on COPD. Stratification analysis showed that females were more vulnerable than males except for T2DM group (COPD: $Z = 3.49$, $P \leq 0.001$; COPD with T2DM: $Z = 0.176$, $$P \leq 0.079$$; COPD with hypertension: $Z = 2.48$, $$P \leq 0.013$$; COPD with both T2DM and hypertension: $Z = 2.44$, $$P \leq 0.014$$); No statistically significant difference could be found between age groups (COPD: $Z = 1.63$, $$P \leq 0.104$$; COPD with T2DM: $Z = 0.23$, $$P \leq 0.821$$; COPD with hypertension: $Z = 0.53$, $$P \leq 0.595$$; COPD with both T2DM and hypertension: $Z = 0.71$, $$P \leq 0.476$$); Higher effects appeared in cold seasons than warm seasons on COPD ($Z = 0.320$, $P \leq 0.001$). This study demonstrated an increased risk of COPD with comorbidities related to CO exposure in Beijing. We further provided important information on lag patterns, susceptible subgroups, and sensitive seasons, as well as the characteristics of the exposure‐response curves. ## Key Points Exposure to CO was associated with an increased risk of chronic obstructive pulmonary disease (COPD) with comorbiditiesFemales were more vulnerable to CO exposure among COPD with comorbiditiesHigher effect appeared in cold season than warm season among COPD group ## Introduction Acute exacerbation of chronic obstructive pulmonary disease (COPD), defined as “an acute worsening of respiratory symptoms that result in additional therapy,” is the main cause of high hospitalization rates and mortality among patients with respiratory diseases (Whittaker Brown & Braman, 2020). COPD is usually linked with several chronic diseases, such as type 2 diabetes mellitus (T2DM), cardiovascular disease, especially hypertension, which are the most common comorbidities (Wielscher et al., 2021). T2DM is present in $22\%$–$40\%$ of patients hospitalized with COPD (Rambaran et al., 2019). The prevalence of T2DM is higher in patients with acute exacerbation of chronic obstructive pulmonary disease (AECOPD) than in the general population. In addition, COPD patients with T2DM are associated with hypertension (Lin et al., 2021). Comorbidities can lead to disease progression and therefore play an important role in the prognosis of COPD (Vogelmeier et al., 2017). In recent years, air pollution, as an emerging risk factor for morbidity and mortality of COPD, has attracted much attention (Renzi et al., 2022). The adverse effects of particular matters on COPD (Lee et al., 2021; Pothirat et al., 2021), T2DM (Song et al., 2018) and hypertension (Qin et al., 2021) patients have been widely reported. Carbon monoxide (CO) is a minor component of air pollution, but as a pollutant that is not easily perceived by the human senses, it can have significant health effects. High concentration of CO exposure will affect the function of heart and nervous system and bring adverse effects on human health, which should be paid enough attention. Short‐term exposure to CO has been positively associated with COPD‐related emergency department visits, hospital admissions, and mortality (Boehm et al., 2021; Chen et al., 2020; Du et al., 2021; Qu et al., 2019; Wang et al., 2021). However, the effect of CO on COPD comorbidities has not been comprehensively reported. In this study, we examined the association between short‐term exposure to CO and morbidity of COPD with comorbid T2DM or hypertension in Beijing from 2014 to 2019. In addition, we further explored whether COPD patients with diabetes and/or hypertension were more vulnerable to adverse effects due to CO exposure. ## Data Collection Beijing, the capital city of the People's Republic of China, is situated at the northern tip of the North China Plain, at 39°56′N and 116°20′E region. The total area of the city is about 16,410.54 square kilometers, including 16 districts, with a population of 21.89 million by the end of 2020. Admission data were obtained from the Information Center of Beijing Municipal Health Commission (http://www.phic.org.cn/). The data mainly included the time of admission, gender, age, main diagnosis and secondary diagnosis of COPD patients admitted to 258 public hospitals in 16 districts in Beijing from 1 January 2014 to 31 December 2019. The study population was divided into four groups based on whether hospitalized patients with COPD were simultaneously diagnosed with T2DM and/or hypertension. According to the 10th revision of the International Classification of Diseases (ICD‐10), COPD was coded as J44.0‐J44.1. T2DM was coded as E12, which was defined as fasting blood glucose ≥7.1 mmol/L, and/or current treatment of diabetes with antidiabetic medication before admission. Hypertension (ICD‐10: I10‐15, O10‐15, P29) was defined as systolic blood pressure ≥140 mmHg and/or diastolic blood pressure ≥90 mmHg and/or taking antihypertensive agents. Daily hospital admissions were further categorized by gender, age (≤60; >60 years of age). This study only included hospitalized cases of COPD in Beijing residents. The data recording system in the study area has been proven to be of high validity (Aklilu et al., 2020; M. Liu et al., 2021; X. Liu et al., 2021). Hourly concentrations of ambient CO and fine particulate matter (PM2.5), inhalable particulate matter (PM10), nitrogen dioxide (NO2), sulfur dioxide (SO2), ozone (O3) from 35 stations in 16 districts in Beijing were retrieved from the Beijing Environmental Protection Bureau during the same period. In this study, we averaged air pollution data as daily levels. Given that meteorological parameters may alter the associations between air pollution exposure and COPD, hourly data of temperature (°C) and relative humidity (%) of 18 meteorological monitoring stations in Beijing were obtained from online China Meteorological Data Sharing Service System online over the study period. ## Ethical Clearance The study was approved by the Institutional Review Board of Capital Medical University (No. IRB00009511). Informed consent was not specifically required since personal identifiers were not collected. ## Statistical Analysis As the daily admissions generally followed a Poisson distribution, the effects of CO on COPD hospital admissions were investigated using an over‐dispersed generalized additive model, Model 1 as follows: log[E(Yt)]=intercept+βZt+s(time,7/year)+s(temp,3)+s(RH,3)+DOWt+Holidayt where E(Yt) represents the number of hospital admissions for COPD on dayt; *Zt is* the CO concentration on day t; β represents the log‐relative risk (RR) of COPD admissions associated with an interquartile range (IQR) increase in CO concentration; s() indicates a natural spline function to filter out long‐term trends and seasonal patterns in daily COPD admissions; temp is the daily mean temperature (℃); RH is the relative humidity (%);DOWt is the day of the week; and DOWt and Holidayt were included as categorical variables. Considering that a nonlinear relationship has been shown between temperature, relative humidity and hospitalizations for COPD in the previous studies (Tian, Xiang et al., 2018; Wang et al., 2021), a natural cubic spline with 3 degrees of freedom (df) was used for each weather condition variable. We created a binary variable for season, with 0 for warm season (from May to October) and 1 for cold season (November to April). Then we added a product term between pollutant concentrations and season into the core model to test the possible interaction between air pollution and season (Guo et al., 2010). Model 2 as follows: log[E(Yt)]=intercept+β1Zt+β2season+β3Zt+s(time,7)+s(temp,3)+s(RH,3)+DOWt+Holidayt β1 signifies the main effect of the pollutant in the warm season; (β1+β3) was the pollutant effect in the cold season; β2 is a vector for coefficients of the season, and β3 is a vector for coefficient of the interactive term between pollutant and season. Sensitivity analysis was conducted to check the stability of the model. First, we checked the df values of the time variable from 4 to 10 per year. Moreover, we also conducted bi‐pollutant model by including PM2.5, PM10, NO2, SO2, and O3 one at a time into the model and changed the number of df values of the time variable from 4 to 10 per year to assess the robustness of the effect estimate. Given the uncertainty in determining the best lag days for estimation, we used multiple lag structures, including single‐day lags from 0 to 3 and moving average exposures from multiple days. A Z value was calculated to test the statistical significance of subgroup differences as follows: Z=β1−β2/SE12+SE22, where β1 and β2 were the effect estimates for the two categories (e.g., males and females) and SE1 and SE2 were the corresponding standard errors (Altman & Bland, 2003). Then, a P value could be obtained from the standard normal distribution based on the Z value. Spearman's correlation coefficients were calculated to assess the degrees of correlation between air pollutants and meteorological variables. The statistical tests were two‐sided, and associations with $P \leq 0.05$ were considered statistically significant. The effects are described as the percentage change and $95\%$ confidence interval (CI) in daily counts of admissions for COPD per IQR increase in air pollutants. All statistical models were run in R software (version 4.0.2) using the mgcv package. ## Descriptive Analyses Table 1 summarized the descriptive statistics of daily admissions in COPD, stratified by gender, age and season. In total, 459,258 admissions for COPD were included in this study during the period from 2014 to 2019, of which 121,330 patients with comorbid both T2DM and hypertension (accounting for $26.42\%$). The majority of patients were male ($67.34\%$) and over 60 years old ($93.09\%$). The annual average values of daily mean concentrations of air pollutants were 0.99 mg/m3 for CO, 65.94 μg/m3 for PM2.5, 98.52 μg/m3 for PM10, 42.91 μg/m3 for NO2, 9.24 μg/m3 for SO2 and 60.77 μg/m3 for O3. The annual average values were 12.41°C for temperature and $53.52\%$ for relative humidity (Table S1 in Supporting Information S1). **Table 1** | Unnamed: 0 | COPD | COPD.1 | COPD with hypertension | COPD with hypertension.1 | COPD with T2DM | COPD with T2DM.1 | COPD with both T2DM and hypertension | COPD with both T2DM and hypertension.1 | Total | Total.1 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | No. | (%) | No. | (%) | No. | (%) | No. | (%) | No. | (%) | | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | | Male | 128546 | 69.66 | 27320 | 63.77 | 72916 | 65.96 | 80491 | 66.34 | 309273 | 67.34 | | Female | 55987 | 30.34 | 15524 | 36.23 | 37635 | 34.04 | 40839 | 33.66 | 149985 | 32.66 | | Age (years) | Age (years) | Age (years) | Age (years) | Age (years) | Age (years) | Age (years) | Age (years) | Age (years) | Age (years) | Age (years) | | ≤60 | 15436 | 8.36 | 2826 | 6.60 | 6154 | 5.57 | 7302 | 6.02 | 31718 | 6.906 | | >60 | 169097 | 91.64 | 40018 | 93.40 | 104397 | 94.43 | 114028 | 93.98 | 427540 | 93.09 | | Season | Season | Season | Season | Season | Season | Season | Season | Season | Season | Season | | Warm | 83596 | 45.30 | 19378 | 45.23 | 49487 | 44.76 | 54463 | 44.89 | 206924 | 45.06 | | Cold | 100937 | 54.70 | 23466 | 54.77 | 61064 | 55.24 | 66867 | 55.11 | 252334 | 54.94 | | Total | 184533 | 40.18 | 42844 | 9.33 | 110551 | 24.07 | 121330 | 26.42 | 459258 | 100 | Figure S1 in Supporting Information S1 showed pairwise Spearman correlation coefficients between air pollutants and weather conditions. Air pollutants except O3 were strongly correlated with each other (Spearman's correlation coefficients distributed from 0.59 to 0.89) and were moderately correlated with weather conditions (−0.46–0.76). ## Association Between Pollutants and COPD Admissions Figure 1 summarized the effects of CO on hospital admissions for COPD. Significant associations were found between CO and daily admissions for COPD comorbidities in Beijing. Delayed effects of CO were significantly associated with admissions on the COPD comorbidities at lag03 with the largest effect. For each IQR uptick in CO, the corresponding percentage changes were $0.21\%$ ($95\%$CI: $0.08\%$–$0.34\%$) for COPD, $0.39\%$ ($95\%$CI: $0.13\%$–$0.65\%$) for COPD with T2DM, $0.29\%$ ($95\%$CI: $0.13\%$–$0.45\%$) for COPD with hypertension and $0.27\%$ ($95\%$CI: $0.12\%$–$0.43\%$) for COPD with both T2DM and hypertension, respectively, at lag03. Specifically, the effects of CO on COPD with T2DM, COPD with hypertension, COPD with both T2DM and hypertension were higher than those in COPD group at different lag structures. However, no significant difference was found between T2DM or hypertension groups and those without comorbidities ($Z = 1.19$, $$P \leq 0.234$$; $Z = 0.77$, $$P \leq 0.444$$; $Z = 0.61$, $$P \leq 0.543$$). **Figure 1:** *Percentage changes with 95% confidence interval in chronic obstructive pulmonary disease (COPD) admissions associated with per IQR increase in CO concentrations for different lag structures in single pollutant generalized additive model. Note: COPD: acute exacerbation of COPD; T2DM: type 2 diabetes mellitus; CO: carbon monoxide.* ## Stratification Analyses The associations between a IQR increase pollutants in the lag03 and the risk of COPD admissions by gender, age and season groups are presented in Table 2. CO were significantly associated with hospital admissions for COPD among female, age >60 years patients. There was significant difference between gender group in the effects of CO on the COPD comorbidities in addition to T2D group (COPD:$Z = 3.49$, $P \leq 0.001$; COPD with T2DM: $Z = 0.176$, $$P \leq 0.079$$; COPD with hypertension: $Z = 2.48$, $$P \leq 0.013$$; COPD with both T2DM and hypertension: $Z = 2.44$, $$P \leq 0.014$$). No statistically significant difference could be found between age groups (COPD: $Z = 1.63$, $$P \leq 0.104$$; COPD with T2DM: $Z = 0.23$, $$P \leq 0.821$$; COPD with hypertension: $Z = 0.53$, $$P \leq 0.595$$; COPD with both T2DM and hypertension: $Z = 0.71$, $$P \leq 0.476$$). The effects on admissions of COPD comorbidities were statistically significant except for COPD with T2DM group in warm season. Effect of CO on COPD was significantly higher in warm season than cold season at lag03 ($Z = 3.20$, $$P \leq 0.001$$). We found that there was no significant difference between the seasonal groups in the COPD comorbidities group (COPD with T2DM: $Z = 0.17$, $$P \leq 0.866$$; COPD with hypertension: $Z = 1.10$, $$P \leq 0.272$$; COPD with both T2DM and hypertension: $Z = 1.25$, $$P \leq 0.211$$). **Table 2** | Group | Gender | Gender.1 | Gender.2 | Age (years) | Age (years).1 | Age (years).2 | Season | Season.1 | Season.2 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | Group | Male | Female | P | ≤60 | >60 | P | Warm | Cold | P | | COPD | 0.06(−0.09, 0.21) | 0.54(0.32, 0.77) | <0.001 | −0.14(−0.58, 0.30) | 0.24(0.11, 0.38) | 0.104 | 0.73(0.38, 1.09) | 0.15(0.11, 0.20) | 0.001 | | COPD with T2DM | 0.21(−0.12, 0.54) | 0.69(0.27, 1.12) | 0.079 | 0.28(−0.71, 1.28) | 0.40(0.13, 0.67) | 0.821 | 0.32(−0.41, 1.06) | 0.39(0.16, 0.62) | 0.866 | | COPD with hypertension | 0.14(−0.06, 0.34) | 0.57(0.30, 0.84) | 0.013 | 0.11(−0.56, 0.79) | 0.30(0.14, 0.47) | 0.595 | 0.54(0.08, 1.00) | 0.26(0.08, 0.45) | 0.272 | | COPD with both T2DM and hypertension | 0.15(−0.02, 0.31) | 0.48(0.26, 0.71) | 0.018 | 0.05(−0.57, 0.68) | 0.29(0.13, 0.45) | 0.476 | 0.54(0.10, 0.98) | 0.24(0.08, 0.40) | 0.211 | ## Exposure Response Curves Exposure response curves between pollutants and daily COPD comorbidity admissions were plotted in Figure 2. Four‐day moving average (lag 03) concentrations of CO appeared to have positive effect on admissions, especially for pollutants with very high concentrations. The exposure response curves of COPD, COPD with hypertension group and COPD with T2DM and hypertension were relatively flat at the concentration of 1–3 mg/m3 and increased sharply after exceeding 3 mg/m3. For COPD with T2DM group, the exposure response curves appeared to be generally linear, with an increasing trend observed. **Figure 2:** *Exposure‐response curves between CO (lag03) and chronic obstructive pulmonary disease (COPD) comorbidity admissions in Beijing, 2014–2019, after adjusted for day of week and public holidays, temperature, and relative humidity. Note: COPD: acute exacerbation of COPD; T2DM: type 2 diabetes mellitus; CO: carbon monoxide.* ## Sensitivity Analysis After adjusting for other pollutants (Figure 3), the impact of CO on COPD admission has changed, but most of the results with statistical significance in the single pollutant model are still statistically significant in the double pollutant model. Figure 4 showed the cumulative lag effects of pollutants on the admissions of COPD complications for different df of time trend. The results showed that when the df varied in the range of 4–10, the effects of pollutants on COPD comorbidities were consistent, that is, CO were positively correlated. It can be seen that different df of time trend have little effect on the study of the relationship between CO and the acute effect of COPD comorbidities, and the model results of this study are relatively robust. Compared to the original model results, the results were not statistically altered when ultraviolet radiation was included as an additional covariate in the model (P values for all Z test were greater than 0.05). **Figure 3:** *Percentage changes with 95% confidence interval in chronic obstructive pulmonary disease (COPD) admissions associated with per IQR increase in pollutants concentrations at lag03 in two‐pollutant generalized additive model. Note: COPD: acute exacerbation of COPD; T2DM: type 2 diabetes mellitus; PM2.5: particles with an aerodynamic diameter ≤2.5 μm; PM10: particles with an aerodynamic diameter ≤10 μm; NO2: nitrogen dioxide; SO2: sulfur dioxide; O3: carbon monoxide; CO: carbon monoxide.* **Figure 4:** *Cumulative effect (lag03) of pollutants on chronic obstructive pulmonary disease (COPD) and comorbidities for different degrees of freedom (df) of time trend. Note: COPD: acute exacerbation of COPD; T2DM: type 2 diabetes mellitus; CO: carbon monoxide.* ## Discussion In this study, we found significant associations between short‐term exposures to CO and hospital admissions for COPD. Our results indicated that the effects of CO on COPD with T2DM, COPD with hypertension, COPD with both T2DM and hypertension were higher than those observed in COPD group at different lag days, but these differences were not statistically significant. Additionally, our findings suggest that females and individuals exposed during warm seasons may be more vulnerable to daily CO exposure. Although prior studies have reported on the association between air pollution and COPD admissions (Boehm et al., 2021; Chen et al., 2020; Du et al., 2021; Mohebbichamkhorami et al., 2020; Sun et al., 2018, 2019), little is known about the effects of CO on the admissions of COPD comorbidities. A study conducted in Qingdao, China, on the association between air pollution and COPD showed no association between CO and COPD (Yang et al., 2021). A study conducted in Hong Kong found a possible negative association between CO and COPD (Tian et al., 2014). Another review on the health effects of gaseous pollutants, published in 2022, found only a positive association between CO and cardiovascular diseases and Parkinson's disease (Chen et al., 2022) and did not focus on the effect of CO on COPD. More evidence is needed to establish a solid association between CO and COPD comorbidities admissions. The results are consistent with previous findings that a study conducted in Beijing (Liang et al., 2019) with consistent data sources showed that in single‐pollutant models at lag0, the RR of hospitalization for COPD per IQR increase in pollutant was 1.024 ($95\%$ CI 1.018–1.029) for CO. Our study did not find that COPD patients with T2DM and/or hypertension were more vulnerable to adverse effects from CO exposure. This may be related to the particularity of the effect of CO on human body. CO, which mainly comes from traffic pollution, is a highly toxic pollutant to the blood and nervous system. CO could enter human blood through the respiratory system and combines with hemoglobin in blood, myoglobin in muscle and respiratory enzyme containing divalent iron to form a reversible conjugate. It not only reduces the ability of blood cells to carry oxygen, but also inhibits and delays the dissociation and release of oxyhemoglobin, resulting in the necrosis of body tissues due to hypoxia. A clinical follow‐up study of 45 patients with COPD in Beijing, China showed that an increase in the quartile range of the moving average of traffic pollution exposure was associated with a significant decrease in large and small airway function at lag 7 days (Wang et al., 2022). T2DM is a common comorbidity in patients with COPD and is probably associated with increased systemic inflammation and poor outcomes. Studies have shown that COPD patients hospitalized for exacerbation are at high risk for impaired glucose metabolism (Mekov et al., 2016). COPD with diabetes was observed to be positively correlated with hypertension, suggesting that patients both with COPD and T2D were more likely to suffer from hypertension than patients with COPD (Lin et al., 2021). However, Co protects pancreas β Cells are protected from apoptosis induced by cytokines and hydrogen deficiency and promote apoptosis β Cell regeneration. CO is of great value in the treatment of T2DM and the prevention of prediabetes from developing into diabetes (Bahadoran et al., 2022). After adjusting for other pollutants, we found that CO still had harmful effects on COPD patients with T2DM and/or hypertension. This may be related to the co‐exposure of CO with other pollutants. Although it is still unclear about the pathogenic mechanisms of co‐exposure to COPD, some pathways may explain it. A study from Italy showed that CO decreases forced expiratory volume in 1s (FEV1) and forced vital capacity (FVC) (Canova et al., 2010). Other gaseous pollutants, such as SO2, have also been shown to significantly reduce FEV1 (Ghozikali et al., 2015). Particulate pollutants inhaled by humans can be deposited in the lungs, especially in the alveoli, also decreasing FEV1 and FVC (Kyung & Jeong, 2020). In addition, particulate pollutants promote oxidative stress and induce cellular inflammation, reducing the body's immunity (Alemayehu et al., 2020). All of these can make people more susceptible to COPD. Identification of the potentially susceptible populations is crucial to public health in developing more targeted intervention strategies. We found a larger association of air pollution in female patients, which is consistent with previous studies (Tian, Li et al., 2018). Different airways sizes, airway reactivity, lung structural and deposition of particles in the lungs between females and males may partly explain the gender differences. Another explanation is that non‐smokers may be more sensitive to air pollution than smokers, since in China, the smoking rate is much lower in females than males (Xu et al., 2016). In addition, we found similar results as previous studies that older people are more susceptible to air pollution (Guan et al., 2016; Liang et al., 2019). Due to poorer lung function and the weaker immune system, elderly patients are more likely to suffer from air pollution. These findings suggested that elderly COPD patients were most susceptible and should be intensively protected from exposure to outdoor air pollution. Furthermore, we found a stronger association during the warm season. One possible explanation is that people tend to be more active outdoors in the warm season compared to the cold season, which leads to longer exposure to CO (Zhuang et al., 2021). However, more studies are needed to confirm the seasonal patterns in the associations of air pollution on COPD and its comorbidity occurrence. Several limitations should be noted in our study. First, we used fixed‐site monitor measurements as a proxy for personal exposure, which may result in exposure errors and an underestimation of the associations between ambient air pollution and diseases. However, measuring every participant's exposure directly is not feasible in such a large epidemiologic study. Second, the generalizability of our results might be limited, as the study collected data from only one highly‐polluted city. Third, since the date of diagnosis is not available, the accuracy of the occurrence of COPD, T2D and hypertension cannot be determined. Fourth, owing to the limited availability of data, we were unable to eliminate the influence of planned admissions. ## Conclusions Short‐term CO exposure were associated with increased admission of COPD and its comorbidity. Our study indicates that the prevention and control for COPD should be given more attention on people with T2DM or hypertension, suggesting that more efforts may be required to mitigate air pollution in Beijing, China. ## Conflict of Interest The authors declare no conflicts of interest relevant to this study. ## Data Availability Statement Air pollutant data and meteorological data are available at https://quotsoft.net/air/ and http://data.cma.cn, but users need to register for free on this website to access the data. Due to the Information Center of Beijing Municipal Health Commission's data policy, admission data for patients with COPD and COPD comorbidities are not available to the public. Detailed steps for downloading air pollutant and meteorological data can be found in the Supporting Information S1 “Data Download Tutorial” section. ## References 1. Aklilu D., Wang T., Amsalu E., Feng W., Li Z., Li X.. **Short‐term effects of extreme temperatures on cause specific cardiovascular admissions in Beijing, China**. *Environmental Research* (2020) **186**. DOI: 10.1016/j.envres.2020.109455 2. Alemayehu Y. A., Asfaw S. L., Terfie T. A.. **Exposure to urban particulate matter and its association with human health risks**. *Environmental Science and Pollution Research International* (2020) **27** 27491-27506. DOI: 10.1007/s11356-020-09132-1 3. Altman D. G., Bland J. M.. **Interaction revisited: The difference between two estimates**. *BMJ* (2003) **326**. DOI: 10.1136/bmj.326.7382.219 4. Bahadoran Z., Mirmiran P., Kashfi K., Ghasemi A.. **Carbon monoxide and β‐cell function: Implications for type 2 diabetes mellitus**. *Biochemical Pharmacology* (2022) **201**. DOI: 10.1016/j.bcp.2022.115048 5. Boehm A., Aichner M., Sonnweber T., Tancevski I., Fischer T., Sahanic S.. **COPD exacerbations are related to poor air quality in Innsbruck: A retrospective pilot study**. *Heart & Lung* (2021) **50** 499-503. DOI: 10.1016/j.hrtlng.2021.02.012 6. Canova C., Torresan S., Simonato L., Scapellato M. L., Tessari R., Visentin A.. **Carbon monoxide pollution is associated with decreased lung function in asthmatic adults**. *European Respiratory Journal* (2010) **35** 266-272. DOI: 10.1183/09031936.00043709 7. Chen C., Liu X., Wang X., Qu W., Li W., Dong L.. **Effect of air pollution on hospitalization for acute exacerbation of chronic obstructive pulmonary disease, stroke, and myocardial infarction**. *Environmental Science and Pollution Research International* (2020) **27** 3384-3400. DOI: 10.1007/s11356-019-07236-x 8. Chen Z., Liu N., Tang H., Gao X., Zhang Y., Kan H.. **Health effects of exposure to sulfur dioxide, nitrogen dioxide, ozone, and carbon monoxide between 1980 and 2019: A systematic review and meta‐analysis**. *Indoor Air* (2022) **32**. DOI: 10.1111/ina.13170 9. Du W., Zhang W., Hu H., Zhang M., He Y., Li Z.. **Associations between ambient air pollution and hospitalizations for acute exacerbation of chronic obstructive pulmonary disease in Jinhua, 2019**. *Chemosphere* (2021) **267**. DOI: 10.1016/j.chemosphere.2020.128905 10. Ghozikali M. G., Mosaferi M., Safari G. H., Jaafari J.. **Effect of exposure to O₃, NO₂, and SO₂ on chronic obstructive pulmonary disease hospitalizations in Tabriz, Iran**. *Environmental Science and Pollution Research International* (2015) **22** 2817-2823. DOI: 10.1007/s11356-014-3512-5 11. Guan W. J., Zheng X. Y., Chung K. F., Zhong N. S.. **Impact of air pollution on the burden of chronic respiratory diseases in China: Time for urgent action**. *Lancet* (2016) **388** 1939-1951. DOI: 10.1016/s0140-6736(16)31597-5 12. Guo Y., Barnett A. G., Zhang Y., Tong S., Yu W., Pan X.. **The short‐term effect of air pollution on cardiovascular mortality in Tianjin, China: Comparison of time series and case‐crossover analyses**. *Science of the Total Environment* (2010) **409** 300-306. DOI: 10.1016/j.scitotenv.2010.10.013 13. Kyung S. Y., Jeong S. H.. **Particulate‐matter related respiratory diseases**. *Tuberculosis and Respiratory Diseases* (2020) **83** 116-121. DOI: 10.4046/trd.2019.0025 14. Lee Y. G., Lee P. H., Choi S. M., An M. H., Jang A. S.. **Effects of air pollutants on airway diseases**. *International Journal of Environmental Research and Public Health* (2021) **18**. DOI: 10.3390/ijerph18189905 15. Liang L., Cai Y., Barratt B., Lyu B., Chan Q., Hansell A. L.. **Associations between daily air quality and hospitalisations for acute exacerbation of chronic obstructive pulmonary disease in Beijing, 2013–17: An ecological analysis**. *The Lancet Planetary Health* (2019) **3** e270-e279. DOI: 10.1016/s2542-5196(19)30085-3 16. Lin L., Shi J., Kang J., Wang Q.. **Analysis of prevalence and prognosis of type 2 diabetes mellitus in patients with acute exacerbation of COPD**. *BMC Pulmonary Medicine* (2021) **21**. DOI: 10.1186/s12890-020-01371-9 17. Liu M., Li Z., Lu F., Guo M., Tao L., Liu M.. **Acute effect of particulate matter pollution on hospital admissions for cause‐specific respiratory diseases among patients with and without type 2 diabetes in Beijing, China, from 2014 to 2020**. *Ecotoxicology and Environmental Safety* (2021) **226**. DOI: 10.1016/j.ecoenv.2021.112794 18. Liu X., Li Z., Guo M., Zhang J., Tao L., Xu X.. **Acute effect of particulate matter pollution on hospital admissions for stroke among patients with type 2 diabetes in Beijing, China, from 2014 to 2018**. *Ecotoxicology and Environmental Safety* (2021) **217**. DOI: 10.1016/j.ecoenv.2021.112201 19. Mekov E. V., Slavova Y. G., Genova M. P., Tsakova A. D., Kostadinov D. T., Minchev D. D.. **Diabetes mellitus type 2 in hospitalized COPD patients: Impact on quality of life and lung function**. *Folia Medica* (2016) **58** 36-41. DOI: 10.1515/folmed-2016-0005 20. Mohebbichamkhorami M., Arbabi M., Mirzaei M., Ahmadi A., Hassanvand M. S., Rouhi H.. **Ambient air particulate matter (PM(10)) satellite monitoring and respiratory health effects assessment**. *Journal of Environmental Health Science & Engineering* (2020) **18** 1247-1258. DOI: 10.1007/s40201-020-00542-4 21. Pothirat C., Chaiwong W., Liwsrisakun C., Bumroongkit C., Deesomchok A., Theerakittikul T.. **The short‐term associations of particular matters on non‐accidental mortality and causes of death in Chiang Mai, Thailand: A time series analysis study between 2016‐2018**. *International Journal of Environmental Health Research* (2021) **31** 538-547. DOI: 10.1080/09603123.2019.1673883 22. Qin P., Luo X., Zeng Y., Zhang Y., Li Y., Wu Y.. **Long‐term association of ambient air pollution and hypertension in adults and in children: A systematic review and meta‐analysis**. *Science of the Total Environment* (2021) **796**. DOI: 10.1016/j.scitotenv.2021.148620 23. Qu F., Liu F., Zhang H., Chao L., Guan J., Li R.. **Comparison of air pollutant‐related hospitalization burden from AECOPD in Shijiazhuang, China, between heating and non‐heating season**. *Environmental Science and Pollution Research International* (2019) **26** 31225-31233. DOI: 10.1007/s11356-019-06242-3 24. Rambaran K., Bhagan B., Ali A., Ali F., Toolsie S., Lobin R.. **High prevalence of diabetes mellitus in a Cohort of patients with chronic obstructive pulmonary disease in Trinidad, West Indies**. *Turkish Thoracic Journal* (2019) **20** 12-17. DOI: 10.5152/turkthoracj.2018.18036 25. Renzi M., Scortichini M., Forastiere F., De' Donato F., Michelozzi P., Davoli M.. **A nationwide study of air pollution from particulate matter and daily hospitalizations for respiratory diseases in Italy**. *Science of the Total Environment* (2022) **807**. DOI: 10.1016/j.scitotenv.2021.151034 26. Song J., Liu Y., Zheng L., Gui L., Zhao X., Xu D., Wu W.. **Acute effects of air pollution on type II diabetes mellitus hospitalization in Shijiazhuang, China**. *Environmental Science and Pollution Research International* (2018) **25** 30151-30159. DOI: 10.1007/s11356-018-3016-9 27. Sun Q., Liu C., Chen R., Wang C., Li J., Sun J.. **Association of fine particulate matter on acute exacerbation of chronic obstructive pulmonary disease in Yancheng, China**. *Science of the Total Environment* (2019) **650** 1665-1670. DOI: 10.1016/j.scitotenv.2018.09.278 28. Sun X. W., Chen P. L., Ren L., Lin Y. N., Zhou J. P., Ni L., Li Q. Y.. **The cumulative effect of air pollutants on the acute exacerbation of COPD in Shanghai, China**. *Science of the Total Environment* (2018) **622–623** 875-881. DOI: 10.1016/j.scitotenv.2017.12.042 29. Tian L., Ho K. F., Wang T., Qiu H., Pun V. C., Chan C. S.. **Ambient carbon monoxide and the risk of hospitalization due to chronic obstructive pulmonary disease**. *American Journal of Epidemiology* (2014) **180** 1159-1167. DOI: 10.1093/aje/kwu248 30. Tian Y., Li G., Yi Z., Liu J., Yang B.. **A low‐frequency MEMS piezoelectric energy harvester with a rectangular hole based on bulk PZT film**. *Journal of Physics and Chemistry of Solids* (2018) **117** 21-27. DOI: 10.1016/j.jpcs.2018.02.024 31. Tian Y., Xiang X., Juan J., Song J., Cao Y., Huang C.. **Short‐term effects of ambient fine particulate matter pollution on hospital visits for chronic obstructive pulmonary disease in Beijing, China**. *Environmental Health* (2018) **17**. DOI: 10.1186/s12940-018-0369-y 32. Vogelmeier C. F., Criner G. J., Martinez F. J., Anzueto A., Barnes P. J., Bourbeau J.. **Global strategy for the diagnosis, Management, and prevention of chronic obstructive lung disease 2017 report: GOLD executive summary**. *European Respiratory Journal* (2017) **49**. DOI: 10.1183/13993003.00214-2017 33. Wang T., Xu H., Zhu Y., Sun X., Chen J., Liu B.. **Traffic‐related air pollution associated pulmonary pathophysiologic changes and cardiac injury in elderly patients with COPD**. *Journal of Hazardous Materials* (2022) **424**. DOI: 10.1016/j.jhazmat.2021.127463 34. Wang Z., Zhou Y., Zhang Y., Huang X., Duan X., Chen D.. **Association of change in air quality with hospital admission for acute exacerbation of chronic obstructive pulmonary disease in Guangdong, China: A province‐wide ecological study**. *Ecotoxicology and Environmental Safety* (2021) **208**. DOI: 10.1016/j.ecoenv.2020.111590 35. Whittaker Brown S. A., Braman S.. **Recent advances in the Management of acute exacerbations of chronic obstructive pulmonary disease**. *Medical Clinics of North America* (2020) **104** 615-630. DOI: 10.1016/j.mcna.2020.02.003 36. Wielscher M., Amaral A. F. S., van der Plaat D., Wain L. V., Sebert S., Mosen‐Ansorena D.. **Genetic correlation and causal relationships between cardio‐metabolic traits and lung function impairment**. *Genome Medicine* (2021) **13**. DOI: 10.1186/s13073-021-00914-x 37. Xu Q., Li X., Wang S., Wang C., Huang F., Gao Q.. **Fine particulate air pollution and hospital emergency room visits for respiratory disease in urban areas in Beijing, China, in 2013**. *PLoS One* (2016) **11**. DOI: 10.1371/journal.pone.0153099 38. Yang J., Liu M., Cheng Q., Yang L., Sun X., Kan H.. **Investigating the impact of air pollution on AMI and COPD hospital admissions in the coastal city of Qingdao, China**. *Frontiers of Environmental Science & Engineering* (2021) **16**. DOI: 10.1007/s11783-021-1490-7 39. Zhuang J., Bai H., Sun J., Zhang T., Li J., Chen Y.. **The association between fine particulate matter and acute lower respiratory infections in Yancheng City, China**. *Environmental Science and Pollution Research International* (2021) **28** 61723-61731. DOI: 10.1007/s11356-021-15102-y
--- title: Widely targeted metabolome profiling of different plateau raspberries and berry parts provides innovative insight into their antioxidant activities authors: - Xiaoli Ren - Shulin Wang - Jinying Wang - Dan Xu - Ying Ye - Yangbo Song journal: Frontiers in Plant Science year: 2023 pmcid: PMC10042140 doi: 10.3389/fpls.2023.1143439 license: CC BY 4.0 --- # Widely targeted metabolome profiling of different plateau raspberries and berry parts provides innovative insight into their antioxidant activities ## Abstract Raspberries are highly nutritious and have powerful antioxidant properties, making them functional berries with positive effects on physiological functioning. However, there is limited information available on the diversity and variability of metabolites in raspberry and its parts, especially in plateau raspberries. To address this, commercial raspberries and their pulp and seeds from two plateaus in China were subjected to LC-MS/MS-based metabolomics analysis and evaluated for antioxidant activity using four assays. A metabolite-metabolite correlation network was established based on antioxidant activity and correlation analysis. The results showed that 1661 metabolites were identified and classified into 12 categories, with significant variations in composition between the whole berry and its parts from different plateaus. Flavonoids, amino acids and their derivatives, and phenolic acids were found to be up-regulated in Qinghai’s raspberry compared to Yunnan’s raspberry. The main differently regulated pathways were related to flavonoid, amino acid, and anthocyanin biosynthesis. The antioxidant activity of Qinghai’s raspberry was stronger than Yunnan’s raspberry, and the order of antioxidant capacity was seed > pulp > berry. The highest FRAP (420.31 µM TE/g DW) values was found in the seed of Qinghai’s raspberry. Overall, these findings suggest that the environment in which the berries grow can affect their chemical composition, and comprehensive exploitation and cultivation of whole raspberry and its parts from different plateaus can lead to new opportunities for phytochemical compositions and antioxidant activity. ## Introduction Berries, including raspberries, are an excellent source of natural antioxidants and are an essential part of a healthy diet. Red raspberry (*Rubus ideaus* L.) is an aggregate fruit and is a genus of suspension berries (Rubus) in the Rosaceae family. Different cultivars and varieties of raspberries are grown worldwide, primarily in Europe, North America, China, Russia, and Japan (Sójka et al., 2016). It is grown for the fresh fruit market and primarily commercial processing into individually quick frozen fruit, juice (Jiang et al., 2022), pulp (Stüwe et al., 2022), dried fruit (Noratto et al., 2017), wine (Cao et al., 2022), and other products. Raspberries have been widely studied by the pharmaceutical, cosmetical, agricultural, and food industries (Kausar and Akhtar, 2016; Gomes et al., 2017; Kirina et al., 2020). Many in vitro and in vivo investigations on human health have demonstrated that raspberry has antibacterial (Goodman et al., 2020), anti-inflammatory, antioxidant (Wang et al., 2019), antiaging (Kobori et al., 2021), and anticancer (Seeram et al., 2006; Bader Ul Ain et al., 2022). Raspberries are known for their various biological activities, which can be primarily attributed to the presence of a diverse range of phytochemicals, such as flavonoids, phenolic acids, tannins, water-soluble vitamins, amino acids, and lignans (Seeram et al., 2006; Nile and Park, 2014; Huang et al., 2022). Many studies analyzed the bioactivity and antioxidant capacity of different parts of plants. Afonso et al. [ 2020] studied bioactive substances and antioxidants in sweet cherry fruit stems and seed kernels. Lenucci et al. [ 2022] focused on the characterization of several bioactive molecule classes simultaneously performed in different fruit fractions of two mango cultivars. Xi et al. [ 2017] examined five fruit segments for phenolic chemicals and antioxidant capacity. Nevertheless, prior studies on raspberries has often focused on the whole berry or on specific tissues such as the pulp or seed. Additionally, some studies have only investigated a limited number of metabolites, rather than looking at the global profile of compounds present in raspberries and its parts. A whole raspberry comprises around 100 drupelets, each with a juicy pulp and a single central seed (Iannetta et al., 2000). Usually, red raspberries are eaten in whole fresh berries, and different parts of raspberry, such as pulp and seed, show various phytochemicals composition with distinctive antioxidant activities. Anthocyanins, flavanols, vitamins, superoxide dismutase, and phenolic acids are present in raspberry pulp, which may be advantageous to health (Badin et al., 2023), which are associated with different organoleptic attributes of color, aroma, and taste (Zhao et al., 2022). Although many research articles have been published on the valorization of byproducts from the agroindustry (Marcillo-Parra et al., 2021), the seed of some fruit remains under investigation. Raspberries’ seeds, unlike pulp, contain $10\%$ oil with a unique chemical composition, making them a viable fatty raw material (Kosmala et al., 2015). Raspberry seeds are rich sources of polyunsaturated fatty acids and antioxidants, namely, polyphenols, flavonoids, and ellagitannins, that may improve the antioxidative status of a consumer (Gođevac et al., 2009; Kosmala et al., 2015). As far as we know, oxidative stress leads to various diseases (Gao et al., 2019). The health effects of raspberry, including its seeds, can prevent cancer by inhibiting cell proliferation, inducing autophagy, and inducing apoptosis (Bilawal et al., 2021). Plant metabolomics can provide a comprehensive analysis of metabolites, a novel technology to analyze the quality and quantity of all metabolites in plants and organs (Li et al., 2021; Segla Koffi Dossou et al., 2022). To date, widely targeted metabolome, based on UPLC-ESI-triple quadrupole-linear ion trap (QTRAP)-MS/MS with a multiple reaction monitoring (MRM) mode, combines the merits of targeted and nontargeted metabolomics. With widely targeted metabolomics analysis, thousands of metabolites can be quickly detected and accurately quantified to assess the metabolome underlying the phenotype of organisms and investigate metabolite variability among different organs, varieties, and species (Scalbert et al., 2011). This approach has been widely used in fruit metabolite analysis, including peach (Gedük and Atsız, 2022), kiwifruit (Li et al., 2023), cherry (Yang et al., 2021), and blueberry (Zheng et al., 2020). The antioxidant capabilities and bioactive chemicals composition of these berries, pulp, and seeds are not fully understood, despite the nutritional and commercial significance of raspberries. Moreover, studies have shown that raspberry polyphenols have antioxidant properties (Lebedev et al., 2022). Commercial raspberries, their pulp, and seeds from two plateaus (Yunnan and Qinghai, China) were used to confirm the diversity and variability of metabolites in whole raspberries and berry components and to reveal crucial metabolites and pathways causing variations in antioxidant activity. Several studies have documented that the raspberries’ composition and bioactivity are affected by some factors, such as the edaphoclimatic conditions of the growing sites, among others (Kafkas et al., 2008; de Souza et al., 2014; Vara et al., 2020). So, the purpose of this research was to investigate the antioxidant properties of different parts of red raspberries grown in the Qinghai-Xizang Plateau and Yunnan-Guizhou Plateau. We used UPLC-MS/MS-based widely targeted metabolome profiling to compare the metabolites present in the berry, pulp, and seeds of the raspberries and combined these findings with biochemical indicators to identify the key metabolites responsible for the antioxidant properties of red raspberries. This study provides new information on the phytochemical composition of different regions and parts of red raspberries, which could offer valuable insights for the wider cultivation and exploitation of red raspberries. ## Berry materials and sample preparation Qinghai raspberries belong to species Autumn Bliss, which were harvested at commercial maturity from the red raspberry base of Datong County (36°80’ N, 101°63’ E, altitude: 2280 m), Qinghai, Qinghai-Xizang Plateau. For Yunnan raspberries belong to species Heritage, or the classical ‘Driscoll’s’, one of the finest fresh berry producers across the globe, which were collected from an orchard in Zhanyi Area (27°49’ N, 103°80’ E, altitude: 2000 m), Yunnan, Yunnan-Guizhou Plateau. The average annual temperature, annual sunshine hours, and rainfall of Datong County were 4.9°C, 2553 h, and 523 mm, respectively, and those of the Zhanyi Area were 17.4°C, 2098 h, and 1002 mm. The chosen berries were uniformly colored and free of mechanical damage. Fresh berries were transported to the lab in a –18°C cold chamber. For the preparation of pulp, the defrosted berry was extracted on a low-speed juicer (model number JYL-C93T, Joyoung Company Limited, China), and the seeds and pulp were collected separately. According to locations and tissues (berry, pulp, and seeds), the red raspberry’s sample from Qinghai-Xizang Plateau was marked as Q-R, Q-RP, and Q-RS, and the red raspberry’s sample from Yunnan-Guizhou Plateau was noted as Y-R, Y-RP, and Y-RS. Samples were stored at –80°C until analysis. The samples are freeze-dried with a vacuum freeze-dryer (Scientz-100F). The freeze-dried sample was crushed using a mixer mill (MM 400, Retsch, Germany) with a zirconia bead for 1.5 min at 30 Hz. Dissolve 50 mg of lyophilized powder with 1.2 mL $70\%$ methanol solution, vortex 30 sec per 30 min for 6 rounds. Following centrifugation at 12000 rpm for 3 min, the supernatant was collected and filtered by a microporous membrane filter (0.22 μm pore size) and stored in the sample injection bottle for UPLC-MS/MS analysis. ## Chemicals Methanol, acetonitrile, and ethanol were all chromatographic purity purchased from Merck (Darmstadt, Hesse, Germany). Formic acid was purchased from Aladdin (Shanghai, China). Standard compounds such as gallic acid and (+)-catechin were obtained from Yuanye company (Shanghai, China). The remaining chemical reagents were purchased from Sinopharm Co. Ltd. (Shanghai, China). ## Widely-targeted metabolomics analysis After sample preparation, the sample was performed at Metware (Wuhan, China) for widely-targeted metabolomics measurement. The operation process was strictly following the operation flow of the UPLC and MASS spectrometry. The column, Agilent SB-C18 (1.8 μm, 2.1 mm * 100 mm), was used, and the mobile phase consisted of solvent A (pure water with $0.1\%$ formic acid) and solvent B (acetonitrile with $0.1\%$ formic acid). Samples were measured using a gradient program with $95\%$ A and $5\%$ B as starting conditions. Within 9 min, a linear gradient to $5\%$ A, $95\%$ B was programmed, and a composition of $5\%$ A, $95\%$ B was held for 1 min. Subsequently, a composition of $95\%$ A, $5\%$ B was adjusted within 1.1 min and kept for 2.9 min. The flow rate was 0.35 mL/min, the column temperature was 40°C, and the injection volume was 4 μL. Data filtration, alignment, and calculation were carried out using Analyst 1.6.1 software (AB SCIEX Pet. Ltd, Framingham, Massachusetts, USA). The ESI source operation parameters were as follows: [1] source temperature 550°C; [2] ion spray voltage (IS) 5500 V (positive ion mode)/– 4500 V (negative ion mode); ion source gas I (GSI), gas II(GSII), and curtain gas (CUR) were set at 50, 60, and 25 psi, respectively; the collision-activated dissociation (CAD) was high. Triple quadrupole (QQQ) scans were acquired as multiple reaction monitoring (MRM) experiments with optimized declustering potential (DP) and collision energy (CE) for each individual MRM transition, and a specific set of the transitions were monitored during each period based on the eluted metabolites (Zhu et al., 2013). ## DPPH method For DPPH assay, the procedure followed the method of Afonso et al. [ 2020] with some modifications. Briefly, 15 mL of the extract and 15 mL of 0.1 mM DPPH solution were added to a test tube and incubated for 30 min in dark. Two hundred microliters of the final mixture were added to a microplate for absorbance determination at 517 nm. Assays were performed after appropriate dilution for samples with high antioxidant activity. $80\%$ methanol was used as control, and Trolox was used as standard. The results were expressed as μg Trolox equivalent per g of fresh weight (μg TE/g FW). ## FRAP method The ferric ion reducing antioxidant power (FRAP) assay was also performed according to the method reported by Li et al. [ 2012]. with some modifications. Briefly, 40 μL of extraction solution diluted with distilled water (1:20, v/v) was mixed with 200 μL of freshly prepared ferric-tripyridyl-triazine agent (Fe3+-TPTZ) in a 96-well microplate. Fe3+-TPTZ was prepared by mixing 20 mM FeCl3·6H2O, 10 mM TPTZ in 40 mM HCl, and 300 mM acetate buffer in a 1:1:10 (v/v/v) ratio. The plates were maintained in the dark at 25°C for 30 min, and their absorbances at 593 nm were measured., and the results were expressed as µM Trolox equivalent per g of dry weight (μM TE/g DW) as the mean ± standard deviation (SD) of three replicates. ## ABTS method The ABTS radical scavenging activity was evaluated in a 96-well microplate using the method of Re et al. [ 1999]. An ABTS radical solution was prepared by mixing 7 mM of ABTS at pH 7.4 (5 mM NaH2PO4, 5 mM Na2HPO4, and 154 mM NaCl) with 2.5 mM potassium persulfate and storing the mixture in the dark at room temperature for 16 h. The mixture was then diluted with ethanol to give an absorbance of 0.70 ± 0.02 units at 734 nm. In each microplate well, 15 µL of the extract was mixed with 285 µL of the freshly prepared ABTS solution and incubated at room temperature in the dark for 10 min. A standard calibration curve was constructed for Trolox at 0, 80, 160, 240, 320, and 400 μmol/L concentrations. Absorbance values were measured at 734 nm, and the results were expressed as µM Trolox equivalent per g of dry weight (μM TE/g DW) as the mean ± standard deviation (SD) of three replicates. ## ORAC ORAC is based on a hydrogen atom transfer (HAT) process, with the oxidation of a fluorescent probe by peroxyl radicals. The procedure for the ORAC assay was performed on plasma according to the instructions supplied with the Oxygen Radical Antioxidant Capacity (ORAC) assay kit from CELL BIOLABS, INC (San Diego, USA). The kit included a 96-well microtiter plate with clean bottom black plate, fluorescein probe 100 ×, free radical initiator, antioxidant standard (Trolox™), and assay diluent (4 ×). The samples were dissolved in a ratio of 1:100 (Rolnik et al., 2021). ## Data analysis The antioxidant ability’s results were expressed as the mean ± standard deviation (SD) using SPSS 17.0 (Chicago, IL, USA). One-way analysis of variance (ANOVA) and Tukey’s test were applied to establish the significance of the differences among samples ($p \leq 0.05$). Origin Pro software (2021b, OriginLab Inc.) was used for image processing. The metabolomic data were processed by multivariate statistical analysis methods, including principal component analysis (PCA), hierarchical cluster analysis (HCA), and orthogonal partial least squares discriminant analysis (OPLS-DA). PCA was first performed on all samples (including QC samples) to determine the overall metabolite differences as well as the variation degree among the red raspberries samples. The HCA (Hierarchical Cluster Analysis) results for samples and metabolites are presented as heat maps and are performed by the R package Complex Heatmap. Orthogonal partial least squares-discriminant analysis (OPLS-DA) was generated using the Metabo Analyst R package. Identified metabolites were annotated using Kyoto Encyclopedia of Genes and Genomes (KEGG) compound databases (http://www.kegg.jp/kegg/compound/) and Metware database, which were then mapped to KEGG Pathway database (http://www.kegg.jp/kegg/pathway.html), followed by enrichment and topological analysis of the pathways where differential metabolites were present. Key pathways were further screened based on the number of differential metabolites (Xiao et al., 2021). ## Widely targeted metabolomics analysis In order to get a complete picture of the differences in the phytochemical composition of red raspberries, widely targeted metabolomics of raspberries and different parts (berry, pulp, and seeds) was performed using UPLC-ESI-MS/MS. Totally 1661 metabolites were identified in Supplementary Table S1. These metabolites were classified into 12 categories, among which 283 phenolic acids, 135 lipids, 337 flavonoids, 89 organic acids, 171 amino acids and derivatives, 70 nucleotides and derivatives, 166 terpenoids, 85 lignans and coumarins, 79 alkaloids, 61 tannins, 161 other, and 24 quinones. Other classes mainly contain sugars, vitamins, aldehydes, ketones, lactones, and others. We found the highest relative content of flavonoids and phenolic acids, with $20.29\%$ and $17.04\%$, respectively (Figure 1A). Analyst 1.6.3 processed mass spectrum data. Supplemental Figures S1, S2 showed the mixed sample’s TIC and MRM metabolite detection multi-peak plots. **Figure 1:** *Overview analysis of widely-targeted metabolomics of different plateau raspberries. Classifications of these metabolites identified based on the mass-to-charge ratio of the compounds in Metware database (A); PCA analysis (B); hierarchical cluster analysis (C).* ## Metabolite variations between Qinghai raspberry and Yunnan raspberry Principal component analysis (PCA) is a multivariate method commonly used to summarize data variances, reveal group differences, and measure sample variability within a group (Wang et al., 2018). PC1 contributed $47.49\%$ and PC2 $25.38\%$. Two primary components contributed $72.87\%$ (Figure 1B). Separation of samples from other plateau occurred in the second principal component, and PCA classified Qinghai and Yunnan raspberry samples differently. There were significant differences between seed and pulp or berry in Qinghai raspberry metabolites, indicating that the relative quantification of the metabolites was significantly different among different plateaus and parts. The clustering heatmap of the metabolites also clearly showed the similarity between the biological replicates and the differences among the red raspberry from two plateaus (Figure 1C). Furthermore, the OPLS-DA model validation Q2 indicates the predictive power, R2Y and R2X indicate the explanation rate of the Y matrix and the X matrix, respectively. And R2Y and Q2 scores were above 0.9, which means the model was appropriate, $p \leq 0.005$ model was excellent. We observed high predictability (Q2) and strong goodness of fit (R2X and R2Y). For instance, the Q2 values between Y-R vs Q-R, Y-RP vs Q-RP, Y-RS vs Q-RS, Q-R vs Q-RP, Q-R vs Q-RS, Q-RP vs Q-RS, Y-R vs Y-RP, Y-R vs Y-RS, and Y-RP vs Y-RS was 0.989, 0.996, 0.997, 0.95, 0.999, 1, 0.964, 0.929, and 0.954 indicating the metabolite profiles of parts and plateaus were distinctly different (Figure 2). OPLS-DA score plots showed that the same parts of red raspberries in different plateaus and different parts of raspberries were well-separated in pairs, suggesting significant differences in metabolic phenotypes of the two kinds of raspberries (Figure 3). Respectively, according to VIP ≥ 1 screened from OPLS-DA results and a fold change ≥ 2 or ≤ 0.5. The screening results are presented in volcano plots (Figure 4). **Figure 2:** *OPLS-DA Validation Chart, Y-R vs Q-R, Y-RP vs Q-RP, Y-RS vs Q-RS, Q-R vs Q-RP, Q-R vs Q-RS, Q-RP vs Q-RS, Y-R vs Y-RP, Y-R vs Y-RS, and Y-RP vs Y-RS (A-I). The horizontal coordinate indicates the model R2Y, Q2 values, and the vertical coordinate is the frequency of the model classification effect in 200 random permutation experiments.* **Figure 3:** *The OPLS-DA score plots of. OPLS-DA model plots for the comparison groups, Y-R vs Q-R, Y-RP vs Q-RP, Y-RS vs Q-RS, Q-R vs Q-RP, Q-R vs Q-RS, Q-RP vs Q-RS, Y-R vs Y-RP, Y-R vs Y-RS, and Y-RP vs Y-RS (A–I).* **Figure 4:** *Volcano plots showing the differences in the expression levels of metabolites in red raspberry, red spots indicate up-regulated differentially expressed metabolites; blue spots indicate down-regulated differentially expressed metabolites, and grey spots indicate detected metabolites, the differences were not significant at p < 0.05, Y-R vs Q-R, Y-RP vs Q-RP, Y-RS vs Q-RS, Q-R vs Q-RP, Q-R vs Q-RS, Q-RP vs Q-RS, Y-R vs Y-RP, Y-R vs Y-RS, and Y-RP vs Y-RS (A–I).* For raspberry from two plateaus, Y-R vs Q-R contained 586 significantly different metabolites (452 up-regulated, 134 down-regulated). Of these, $27.21\%$ of flavonoids, $16.15\%$ of phenolic acids, $11.28\%$ of terpenoids, and $7.74\%$ of amino acids and their derivatives were up-regulated in Q-R. The key differentially regulated metabolites included $49.78\%$ of polyphenols, up-regulated in Qinghai raspberry (Figures 4A, 5A). Y-RP vs Q-RP showed 760 substantially different metabolites (618 up-regulated and 142 down-regulated). Of these, $24.27\%$ of flavonoids, $16.50\%$ of phenolic acids, $11.49\%$ of terpenoids, and $10.52\%$ of lipids were up-regulated in Q-RP (Figures 4B, 5B). Between Y-RS and Q-RS, 1028 substantially different metabolites (395 up-regulated and 633 down-regulated) were found. Among these, the Q-RS showed an up-regulation of $26.58\%$ flavonoids, $18.98\%$ phenolic acids, $13.92\%$ terpenoids, and $9.62\%$ amino acids and their derivatives (Figures 4C, 5C). **Figure 5:** *Number of different types of differential metabolites for the comparison group, Y-R vs Q-R, Y-RP vs Q-RP, Y-RS vs Q-RS, Q-R vs Q-RP, Q-R vs Q-RS, Q-RP vs Q-RS, Y-R vs Y-RP, Y-R vs Y-RS, and Y-RP vs Y-RS (A-I). Yellow column indicates metabolites that were significantly up-regulated, blue column indicates metabolites that were significantly down-regulated. Significantly regulated metabolites between groups were determined by VIP ≥ 1 and FC ≥ 2 or ≤ 0.5.* The analysis of different parts of raspberry showed significant differences in metabolites between various comparisons. For instance, Q-R vs Q-RP showed 119 differential metabolites, out of which 107 were up-regulated, and 12 were down-regulated. Similarly, Q-R vs Q-RS, Q-RP vs Q-RS, Y-R vs Y-RP, Y-R vs Y-RS, and Y-RP vs Y-RS showed 1251, 1250, 331, 320, and 420 differential metabolites, respectively (Figures 4D–I). Flavonoids, phenolic acids, amino acids and derivatives, and terpenoids were the major categories of differential metabolites observed in these comparisons, comprising over $50\%$ of the total differential metabolites (Figure 5D–I). Differential accumulated metabolite (DAM) mainly consisted of phenolic acids, flavonoids, and amino acids and their derivatives, and about half of the DAMs was phenolic acids and flavonoids. These metabolites are important secondary metabolites in many plants. They contribute to the antioxidant activity of plants (Ramalingam et al., 2021). Thus, the difference in DAMs in pairwise comparisons suggests that the functional activity of the raspberries from the two plateaus may differ. ## KEGG annotation and enrichment analysis of differential metabolites To further conduct the major pathways of DAMs in samples, the KEGG enrichment analysis of each cluster to obtain detailed information about the metabolic pathways was shown in bubble plots (Figure 6). The DAMs were enriched into 88, 88, and 97 pathways in Y-R vs Q-R, Y-RP vs Q-RP, and Y-RS vs Q-RS, respectively (Supplementary Table S2). Y-R and Q-R occurred in flavonoid biosynthesis, flavone, and flavonol biosynthesis. The primary enrichment of differential metabolites between Y-RP and Q-RP occurred in purine metabolism and flavonoid biosynthesis. In addition, tthe differential metabolites between the Y-RS and Q-RS were involved in anthocyanin biosynthesis, biosynthesis of amino acids. These results indicate that environmental and climate factors have great changes in metabolites in the anthocyanin, flavonoid, flavone, and flavonol biosynthesis pathways. **Figure 6:** *KEGG enrichment of differential metabolites between the comparison groups, Y-R vs Q-R, Y-RP vs Q-RP, Y-RS vs Q-RS, Q-R vs Q-RP, Q-R vs Q-RS, Q-RP vs Q-RS, Y-R vs Y-RP, Y-R vs Y-RS, and Y-RP vs Y-RS (A–I). Each bubble in the plot represents a metabolic pathway whose abscissa and bubble size jointly indicate the magnitude of the impact factors of the pathway. A larger bubble size indicates a larger impact factor. The bubble colors represent the p-values of the enrichment analysis, with darker colors showing a higher degree of enrichment. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).* The DAMs for part’s groups (Q-R vs Q-RP, Q-R vs Q-RS, Q-RP vs Q-RS, Y-R vs Y-RP, Y-R vs Y-RS, and Y-RP vs Y-RS) were involved in 38, 100, 100, 52, 77, and 80 pathways (Supplementary Table S2). Figures 6D-I displayed the top 20 enriched pathways for the DAMs. Multiple primary metabolism pathways and secondary metabolic pathways (purine metabolism, phenylpropanoid biosynthesis, biosynthesis of various plant secondary metabolites, cysteine and methionine metabolism, flavonoid biosynthesis, anthocyanin biosynthesis, biosynthesis of unsaturated fatty acids, alpha-linolenic acid metabolism, linoleic acid metabolism, valine, leucine, and isoleucine biosynthesis) enriched by these DAMs. In these comparison groups, it is worth noting that these pathways are related to the biosynthesis of flavonoids, phenolic acids, fatty acids, and amino acids and their derivatives. The flavonoid and anthocyanin biosynthesis pathways, for example, are located downstream of the phenylpropanoid biosynthetic pathway (Qiu et al., 2020). In these comparison groups, some metabolic pathways overlapped, such as purine metabolism. The process of purine metabolism serves as a fundamental stage in the synthesis of nucleic acids and is intricately linked to the primary and secondary metabolic pathways of plants, which are also closely associated with the polyphenol metabolic pathway. The metabolic pathways appear to be significantly linked to the potent antioxidant properties observed in raspberry seeds. ## Key significantly differential metabolites In order to identify the crucial metabolites present in raspberries grown on various plateaus, a Venn diagram was created to compare the differential metabolites between Y-R and Q-R, Y-RP and Q-RP, and Y-RS and Q-RS (Figure 7A). We discovered that the majority of these metabolites, which were among the 264 that overlapped between Y-R and Q-R, Y-RP and Q-RP, and Y-RS and Q-RS, were up-regulated in Qinghai raspberries. ( Supplementary Table S3). There were $43.56\%$ and $7.58\%$ respectively of polyphenolic compounds and amino acids and their derivatives. Also, the data were merged into the appropriate maps based on the KEGG annotation, and the metabolic pathways of the most pertinent overlapping differential metabolites (Figure 8). We took the up-regulated metabolites in metabolic pathways related to antioxidant properties as the standard and screened out three differential metabolites related to antioxidant quality in the overlapping substances. The anthocyanin biosynthesis, flavonoid biosynthesis, flavone and flavonol biosynthesis, and biosynthesis of amino acids were enriched among the different comparison groups. Some metabolites of raspberries grown in QZP, including Chlorogenic acid (3-O-Caffeoylquinic acid)*, 5-O-p-Coumaroylquinic acid*, Luteolin-7-O-glucuronide-5-O-rhamnoside, Quercetin-3-O-sambubioside*, Isosalipurposide (Phlorizin Chalcone), Hesperetin-7-O-glucoside*, Quercetin-3-O-rhamnoside(Quercitrin), Dihydromyricetin (Ampelopsin), 3,5,7-Trihydroxyflavanone (Pinobanksin), 3-O-Acetylpinobanksin, Apigenin-6-C-glucoside (Isovitexin), Naringenin chalcone; 2’,4,4’,6’-Tetrahydroxychalcone, Butin; 7,3’,4’-Trihydroxyflavanone, Cyanidin-3-O-glucoside (Kuromanin), Naringenin (5,7,4’-Trihydroxyflavanone), L-Glutamine, and L-Lysine were up-regulated (Figures 9, 10). In addition, we identified distinct metabolites in amino acids and their derivatives, polyphenols, and fatty acids in berry, pulp, and seed. Using Venn diagrams to compare each group, we found only 41 metabolites were common among Q-R vs Q-RP, Q-R vs Q-RS, and Q-RP vs Q-RS, while only 20 metabolites were shared among Y-R vs Y-RP, Y-R vs Y-RS, and Y-RP vs Y-RS (Figures 7B, C). These findings suggest that the metabolites responsible for the variations between berry, pulp, and seed were significantly distinct. ( Supplementary Table S3). **Figure 7:** *Plotted Venn diagram of the different metabolites in the above comparison groups from raspberries, Y-R vs Q-R, Y-RP vs Q-RP, and Y-RS vs Q-RS Venn diagram, (A); Q-R vs Q-RP, Q-R vs Q-RS, and Q-RP vs Q-RS, (B); Y-R vs Y-RP, Y-R vs Y-RS, and Y-RP vs Y-RS, (C), were plotted using Venn diagram package.* **Figure 8:** *Overview of the probable regulation of some key metabolites mapped to metabolic pathways in pairwise comparisons of the raspberry from different plateaus. The red color small rectangle indicates that the metabolite content is significantly up-regulated; the blue rectangle indicates that the metabolite content is significantly down-regulated.* **Figure 9:** *Variation of two selected bioactive phenolic acid in two plateaus. 5-O-p-Coumaroylquinic acid* (A); Chlorogenic acid (3-O-Caffeoylquinic acid)* (B). Bars indicate the s.d. of three replicates. “ns” means not significant, *** p < 0.005.* **Figure 10:** *Difference in the relative content of thirteen selected bioactive flavonoids and two amino acids and their derivatives in two plateaus. 3,5,7-Trihydroxyflavanone (Pinobanksin) (A), 3-O-Acetylpinobanksin (B), Quercetin-3-O-rhamnoside(Quercitrin) (C), Apigenin-6-C-glucoside (Isovitexin) (D), Butin; 7,3',4'-Trihydroxyflavanone (E), Quercetin-3-O-sambubioside* (F), Cyanidin-3-O-glucoside (Kuromanin) (G), Dihydromyricetin (Ampelopsin) (H), Hesperetin-7-O-glucoside* (I), Isosalipurposide (Phlorizin Chalcone) (J), L-Glutamine (K), L-Lysine (L), Naringenin (5,7,4'-Trihydroxyflavanone) (M), Luteolin-7-O-glucuronide-5-O-rhamnoside (N), Naringenin chalcone; 2',4,4',6'-Tetrahydroxychalcone (O). Bars indicate the s.d. of three replicates. “ns” means not significant, * p < 0.05, ** p < 0.01, *** p < 0.005.* ## Antioxidant activity analysis of the raspberry from two plateaus The most frequently employed techniques to evaluate the antioxidant capacity of foods include FRAP (ferric reducing activity of plasma), ABTS (2,2-azinobis(3-ethylbenzthiazoline-6-sulfonic acid)), DPPH (1,1-diphenyl-2-picrylhydrazyl), and ORAC (oxygen radical absorbance capacity). Furthermore, they carved two groups: a hydrogen atom transfer-based assay and an electron transfer-based assay (Burton-Freeman et al., 2016). Compared to common fruits like grapes, apples, and citrus, raspberries exhibit higher levels of antioxidant activity (Fu et al., 2011). As shown in Table 1, the antioxidant capacity of red raspberries in different plateaus was significantly different. For the DPPH scavenging performance test and FRAP total antioxidant test, Q-RS and Q-R showed higher antioxidant capacity than Y-RS and Y-R. For the ABTS radical scavenging test, Q-R and Q-RP showed higher antioxidant capacity than Y-R and Y-RP, which exhibited a higher ABTS radical scavenging capacity for Q-R (49.57 ± 1.26 μM TE/g DW), which was 10-fold higher than Y-R (4.68 ± 0.45 μM TE/g DW). The ORAC values of Q-RP and Q-RS were higher than those of Y-RP and Y-RS. In addition, the average ORAC values raised in the direction: seed < pulp < berry, the rise of DPPH, ABTS, and FRAP values were berry < pulp < seed, and the DPPH average values of berry, pulp, and seed were above 10-fold compared to ABTS values. These results may be due to the different reaction mechanisms in ABTS, DPPH, FRAP, and ORAC assays. These results indicated that there were significant differences in antioxidant activity between the same parts in red raspberries from different plateaus, with berry, pulp, and seeds of Qinghai red raspberries having higher antioxidant capacity compared to Yunnan red raspberries. In addition, the antioxidant capacity of the seed of Qinghai raspberry was the maximum (420.31 µM TE/g DW, respectively) of all, over 20 times higher than the least value (Yunnan berry). The order of the antioxidant capacity was seed > pulp > berry. **Table 1** | Index | Y-R | Y-RP | Y-RS | Q-R | Q-RP | Q-RS | | --- | --- | --- | --- | --- | --- | --- | | DPPH value (μg TE/g FW) | 191.31 ± 7.57d | 5345.70 ± 76.03c | 5072.01 ± 50.23b | 211.91 ± 11.77d | 5022.00 ± 23.50a | 5218.34 ± 78.16c | | FRAP value (μΜ TE/g DW) | 25.34 ± 7.10e | 121.94 ± 6.01c | 402.86 ± 16.11b | 27.06 ± 5.94e | 100.95 ± 5.08d | 420.31 ± 8.65a | | ORAC (%) | 71.43 ± 1.07a | 45.53 ± 0.04d | 44.36 ± 0.06e | 69.23 ± 1.11b | 50.74 ± 0.06c | 44.44 ± 0.03de | | ABTS value (μΜ TE/g DW) | 4.68 ± 0.45f | 151.05 ± 1.71d | 351.72 ± 5.71a | 49.57 ± 1.26e | 172.07 ± 5.99c | 326.72 ± 10.09b | ## Antioxidants and related metabolites correlation analysis To reveal potential correlations between the metabolite profiles of red raspberries and their parts and antioxidant activities (Table 1), we compared the relative content of each class of metabolites. As shown in Figure 11, Y-R, Y-RP, Y-RS, Q-R, Q-RP, and Q-RS significantly differed in the relative contents of flavonoids, phenolic acids, amino acids and derivatives, lipids, and terpenoids. Unlike Y-R, Y-RP, and Y-RS, the accumulation of phenolic acids and flavonoids was higher in Q-R, Q-RP, and Q-RS. Additionally, in comparison to Y-RS, Q-RS has larger concentrations of flavonoids, amino acids and their derivatives. **Figure 11:** *The relative content of each class of metabolites of the raspberry from different plateaus and parts. Comparison of the relative content of each category of metabolites in Y-RS, Y-RP, Y-R, Q-RS, Q-RP, and Q-R.* ## Phenolic compounds composition in raspberry from two plateaus and their relation to antioxidant activity In the current study, phenolic compound composition and concentration were evaluated in order to identify possible metabolites responsible for variations in antioxidant activity in raspberries and their parts from distinct plateaus. ## Flavonoids A total of flavonoid metabolites were identified (28 flavanols, 13 flavanonols, 75 flavones, 28 anthocyanidins, 125 flavonols, 34 flavanones, 15 chalcones, 1 dihydroisoflavones, 7 isoflavones, 11 other flavonoids, of which 8 chalcones (4,4'-dihydroxy-2'-methoxychalcone, sieboldin, Phloretin-4'-O-glucoside (Trilobatin), Isosalipurposide (Phlorizin Chalcone), 3,4,2',4',6'-Pentahydroxychalcone-4'-O-glucoside, Phloretin-4'-O-(6''-Caffeoyl)glucoside, Phloretin-4'-O-(6''-p-Coumaroyl)glucoside, 2,4,2',4'-tetrahydroxy-3'-prenylchalcone), a dihydroflavone (Pinocembrin-7-O-(2''-O-arabinosyl)glucoside), (Diosmetin (5,7,3'-Trihydroxy-4'-methoxyflavone), Kaempferol-3-O-glucoside-7-O-rhamnoside*) were the most abundant flavonoids. Compared to the Y-R, Kaempferol-3-O-sambubioside and 6,7,8-tetrahydroxy-5-methoxyflavone were found to be specific to Q-R. Compared to the Q-R, 3,5,4'-Trihydroxy-7-methoxyflavone (Rhamnocitrin) was found to be specific to Y-R. Compared with Y-RP, there are flavonoids (Kaempferol-3-O-sambubioside, Kaempferol-3-O-(6''-Malonyl)glucoside-7-O-Glucoside) that are specific to Q-RP, 3,5,4’-trihydroxy-7-(rhamnocitrin) and 3,5,4'-Trihydroxy-7-methoxyflavone (Rhamnocitrin) that are specific to Y-RP. 3,5,4’-Trihydroxy-7-methoxyflavone (Rhamnocitrin) and Eriodictyol-7-O-glucoside* are specific to Y-RP, and 82 flavonoids are specific to Q-RS compared to Y-RS. Notably, the present study focused on the berry, pulp, and seeds of raspberries from two plateaus. Among them, compared to Y-R, flavonoids in core difference metabolites. For instance, Luteolin-7-O-glucuronide-5-O-rhamnoside, Quercetin-3-O-sambubioside*, Isosalipurposide (Phlorizin Chalcone), Quercetin-3-O-rhamnoside(Quercitrin), Dihydromyricetin (Ampelopsin), 3,5,7-Trihydroxyflavanone (Pinobanksin), 3-O-Acetylpinobanksin, Apigenin-6-C-glucoside (Isovitexin), Naringenin chalcone; 2’,4,4’,6’-Tetrahydroxychalcone, Butin; 7,3’,4’-Trihydroxyflavanone, Naringenin (5,7,4’-Trihydroxyflavanone) (log2FC = 6.37, 3.53, 3.31, 1.43, 1.77, 1.45, 4.32, 2.77, 1.52, 1.51, 1.46) were found at higher levels in Q-R (Figure 12A). Compared to Y-RP, the relative contents of Isosalipurposide (Phlorizin Chalcone), Hesperetin-7-O-glucoside*, Dihydromyricetin (Ampelopsin), 3,5,7-Trihydroxyflavanone (Pinobanksin), and Naringenin chalcone; 2’,4,4’,6’-Tetrahydroxychalcone (log2FC = 3.45, 2.22, 1.11, 1.49, 1.34) were significantly higher in Q-RP (Figure 12B). Compared to Q-RS, the Cyanidin-3-O-glucoside (Kuromanin) (log2FC = 6.22) contents were higher than that in Y-RS (Figure 12C). Based on these findings, it can be inferred that certain substances in the pathway of flavonoid biosynthesis may be influenced by environmental factors. As a result, the variations in the types and quantities of flavonoids present in the raspberries from the two plateaus could potentially result in varying levels of antioxidant activity. **Figure 12:** *Variance multiplier bar chart, the horizontal coordinate is the value of the difference multiplier of the difference metabolite taken as logarithm with a base of 2. The vertical coordinate is the difference metabolite. Red represents up-regulation of metabolite content, blue represents down-regulation of metabolite content. Y-R vs Q-R, Y-RP vs Q-RP, Y-RS vs Q-RS, Q-R vs Q-RP, Q-R vs Q-RS, Q-RP vs Q-RS, Y-R vs Y-RP, Y-R vs Y-RS, and Y-RP vs Y-RS (A–I).* ## Phenolic acids Phenolic acids possess antioxidant properties and are typically found in plant cell walls, where they are closely associated with polysaccharides (Río Segade et al., 2019). Phenolic acids in core difference metabolites, levels of chlorogenic acid (3-O-caffeoylquinic acid)* (log2FC = 2.78) and 5-O-p-coumaroylquinic acid* (log2FC = 1.85) were significantly higher in Q-RP than those in Y-RP. Therefore, the composition and amount of phenolic acids in raspberries from different regions may influence their antioxidant capacity. This could be due to variations in growing conditions, such as climate, soil type, and altitude, which can affect the plant’s ability to produce and accumulate phenolic compounds. ## Amino acids and their derivatives compounds composition in raspberry from two plateaus and their relation to antioxidant activity Plants are the only organisms capable of synthesizing the essential amino acids (leucine, isoleucine, methionine, phenylalanine, arginine, histidine, tryptophan, valine, threonine, and lysine) (Kumar et al., 2017). The amino acid and derivative compositions in Y-RS, Y-RP, Y-R, Q-RP, and Q-R detected were very comparable, only with Q-RS containing 24 unique amino acids and derivatives. Notably, among the top 20 differential up-regulated or down-regulated metabolites (Y-RS and Q-RS), 7 compounds were amino acids and their derivatives (Figure 12C). In Qinghai berries, L-Lysine and L-Glutamine were mainly present in the seeds, and the more antioxidant properties of the seeds might be related to the amino acid content. To gain a better understanding of the antioxidative metabolite composition, antioxidant activity was measured. Therefore, Spearman’s rank correlation tests were used to test for correlation analysis of the amino acids and their derivatives DAMs with ORAC, DPPH, FRAP, and ABTS. We can find from Figure 13 that different amino acids and their derivatives have different correlations with antioxidant capacity, with L-Lysine and L-Glutamine showing the most significant correlation with antioxidant properties. DPPH, FRAP, and ABTS were clustered, indicating that these three antioxidant assays had positive correlations and showed the lowest correlation with ORAC assay. **Figure 13:** *Heat map of the correlation between amino acids and their derivatives in core metabolites and antioxidants. Different letters above columns and the color of the column indicate the correlation, with the larger and the redder the color and the greater the correlation coefficient.* ## Metabolite-metabolite correlation 289 pairwise correlation values were obtained for 17 different metabolites in red raspberry from two plateaus, of which flavonoids and phenolic acids were determined to be significant (r > 0.80, $p \leq 0.05$) (Figure 14). Furthermore, two amino acids (L-Glutamine and L-Lysine) and Cyanidin-3-O-glucoside (Kuromanin) were found to be significantly correlated. In antioxidant assays, they demonstrated synergism. The presence of amino acid metabolites in significant metabolite-metabolite correlations suggested that amino acids positively effect the antioxidant activity. **Figure 14:** *Map of significant seed metabolite-metabolite correlations. Metabolites are represented by circles, and the same color indicates metabolites in the same metabolic function group. Correlations are indicated by connected lines. Positive correlations are red and negative correlations are blue. The thickness of the line represents the magnitude of the absolute value of Pearson’s correlation coefficient r. The thicker the line, the larger the |r|.* ## Discussion Because of their highly functional ingredients content and antioxidant activities, raspberries are becoming increasingly popular. However, differences exist in the metabolites of raspberries grown in different environments. Therefore, the present study was designed to provide a widely targeted metabolomic analysis of the relationship between metabolites and antioxidants in raspberries and parts. ## Differential metabolites between raspberries from two plateaus Flavonoids, phenolic acids, amino acids and derivatives, and lipids were the most significant differential metabolites between the two plateaus. About half of the 264 key significantly differential metabolites were flavonoids and phenolic acids. The analysis of metabolic pathways for the different metabolites in Y-R vs Q-R, Y-RP vs Q-RP, and Y-RS vs Q-RS groups showed a significant involvement of flavonoid biosynthesis, anthocyanin biosynthesis, biosynthesis of amino acids, and flavone and flavonol biosynthesis metabolic pathways. This suggests that there are notable differences between the two plateaus in terms of weather, rainfall, altitude, and temperature. Previous studies found during red raspberry development, in stage 2 with stage 3 (25 DAFB) d after full bloom (DAFB), which mainly involves KEGG pathways related to flavonoid biosynthesis and phenylpropanoid biosynthesis (Huang et al., 2022). This previous study is consistent with the main pathway in the present study and suggests that antioxidant-related substances in raspberries are produced during the second stage of raspberry growth. Interestingly, linoleic acid metabolism was significantly up-regulated ($p \leq 0.05$) in the comparison of in Y-RP vs Q-RP group, indicating that the γ-Linolenic Acid* is the main differential metabolite. The study showed that the two plateaus growing environment causes drastic changes in polyphenols metabolites. Recent research by Grand View Research, Inc. has suggested that the worldwide polyphenols market is arriving at USD 2.9 billion by 2030 (Sarv et al., 2021). Polyphenols, as antioxidants from many berries and particularly raspberries. The mechanism of their antioxidant activity can be characterized by the direct scavenging or quenching of oxygen radicals or excited oxygen species and the inhibition of oxidative enzymes that produce these reactive oxygen species (Baby et al., 2018). As food supplements can enhance our body’s antioxidant defense system, reduce life-threatening diseases caused by oxidative stress, and greatly reduce the risk of cancer (Alarcón-Flores et al., 2013; Yang et al., 2018). The main determinants of the total antioxidant capacity of fruits such as berries are particularly rich in flavonoids (Panche et al., 2016). The flavonoid synthesis pathway involves the condensation of phenylpropanoid derivatives with malonyl-CoA, as reported by Ono et al. [ 2006]. In addition, transcription factors play a role in regulating certain aspects of this pathway. According to Mudge et al. [ 2016], berries are known to have elevated concentrations of Quercetin-4’-O-glucoside, Quercetin-3, 4’-O-diglucoside, and Quercetin-3-O-rutinoside. Our study supports these previous findings, as we found a substantial amount of Quercetin-4’-O-glucoside in raspberries. In order to investigate the key metabolites related to antioxidant capacity in raspberries from two different plateaus, a total of 264 overlapping differential metabolites were identified. Out of these, three differential metabolites associated with antioxidant activity were selected by using the up-regulated metabolites between different groups (Y-R vs Q-R, Y-RP vs Q-RP, and Y-RS vs Q-RS) as a criterion for screening. We found the main up-regulated flavonoids in Qinghai raspberry’s berry, pulp, and seed included Quercetin-3-O-rhamnoside (Quercitrin), Cyanidin-3-O-glucoside (Kuromanin), Naringenin (5,7,4’-Trihydroxyflavanone) (Figures 10C, G, and M). According to the research, Cyanidin-3-O-glucoside is responsible for the antioxidant properties of blackberries (Huang et al., 2022). Quercetin-3-O-rhamnoside could be a major contributor to antioxidant activity (Nie et al., 2020). Naringenin (5,7,4'-trihydroxyflavanone) is a naturally occurring bioactive flavanone, which is documented to have bioactive effects on human health, such as antidiabetic, immunomodulatory, anticancer, anti-inflammatory (Stec et al., 2020). These molecules responded particularly to increased antioxidant activity and, hence, can serve as indicators for selecting Qinghai berries. The molecules known as α-amino acids consist of an α-carbon atom attached to an amino group (NH2), a carboxyl group (COOH), a hydrogen atom (H), and a side chain (R), where the NH2 group is connected to the α-carbon. Due to structural differences in the side chains, these amino acids’ antioxidation mechanisms and capacities vary (Xu et al., 2017). Other studies have reported that lysine with strong antioxidative capacity in 20 amino acids. Researchers found that tolerant citrus germplasm possesses a large number of amino acids with high antioxidant potential, such as lysine, tyrosine, phenylalanine, tryptophan, and asparagine (Killiny and Hijaz, 2016). The reported case illustrates that amino acids are associated with plant resistance to several abiotic and biotic stresses. In this study, one of the reasons for the higher antioxidant capacity of Qinghai raspberry may be due to the synthesis of a large number of amino acids with antioxidant capacity under biotic stress. Y-RS compared to Q-RS, we found the highest relative content of L-Lysine and L-Glutamine among amino acids and their derivatives in Qinghai raspberry seeds. L-*Lysine is* known to have a positive impact on antioxidant status by primarily boosting the effectiveness of the GSH and peroxidase systems, which are crucial components of the body’s antioxidant defense mechanism. By doing so, L-Lysine acts to eliminate free radicals and guard against the harmful effects of oxidative damage caused by these radicals (Al-Malki, 2015). In addition, oxidative stress and neuronal apoptosis were decreased by L-Glutamine (Luo et al., 2019), and *It is* a precursor for the generation of a variety of secondary defensive metabolites. Finally, amino acid metabolic pathways stimulated plant defense to increase resistance. Transgenic rice overexpressing M. oryzae Systemic Defense Trigger 1 showed this pattern. ( Duan et al., 2021). The correlation between amino acids and their derivatives in the core metabolites and antioxidants was analyzed, and L-Lysine and L-Glutamine showed a significant positive correlation with antioxidants. The strongest correlations were found between DPPH, ABTS, and FRAP assays, especially ABTS and FRAP. Then, ORAC had the lowest correlations. Unlike the others, the ORAC assay determined the kinetic action of antioxidants, which might explain the discrepancy. The DPPH, ABTS, FRAP, and ORAC assays gave comparable results for the antioxidant activity tested in extracts from guava fruit. The highest correlations were found between DPPH, ABTS, and FRAP assays, especially between ABTS and FRAP assays, a result previously reported by Thaipong et al. [ 2006]. Another study by Dudonné et al. [ 2009] that evaluated the antioxidant properties of 30 industrial plant extracts using DPPH, ABTS, FRAP, SOD, and ORAC assays also obtained consistent results. Correlation analysis of different metabolites is helpful to find the relationship between metabolites and discover potential key metabolic regulatory (Fukushima et al., 2011). The red raspberry metabolic network displayed a highly concerted interplay of amino acids, providing evidence for the essential conserved roles of amino acids in plant’s metabolism. These are highly positive correlations because the correlated metabolites were either participating in the same metabolic biosynthesis of amino acids reaction as a substrate-product or were controlled by the same regulatory enzymes’ activities. Also, L-Lysine and L-Glutamine were found to be highly positively correlated in our study. The information on these significant metabolite correlations could help better understand red raspberry’s metabolic regulatory network and potentially discover novel metabolic pathways. Furthermore, the synergistic effect among natural antioxidants is one of the mechanisms by which amino acids exert their antioxidant activity, also known as the synergistic antioxidant effect with other antioxidants (Zhang et al., 2022). Due to the increase in the content of amino acids and phenolic compounds and their derivatives. Their derivative content induces the activity of enzymes related to amino acids and phenolic compounds, and the expression of genes related to amino acid and phenolic metabolism contributes to the improvement of antioxidant capacity (Wang et al., 2021). The above research indicated that amino acids and polyphenols have synergistic effects in terms of antioxidant activity. In our correlation analysis, L-Lysine, L-Glutamine, and Cyanidin-3-O-glucoside (Kuromanin) were found to have strong correlations demonstrating their potential to be used as a potential to increase antioxidant capacity. Taken together, besides some specific flavonoids and phenolic acids, the reason for the strong antioxidant capacity of raspberry seeds and Qinghai raspberry may be related to the fact that L-Lysine and L-Glutamine content and synergistic effect with polyphenols. Plant antioxidant defense mechanism has attracted many researchers’ attention (Mazur et al., 2014; Bhattacharjee, 2019). Some metabolites as metabolic antioxidant defense systems are key components of plant growth during adaptation to biotic and abiotic stress conditions, such as fatty acids, amino acids, phenolic acids, flavonoids, and anthocyanins (colored pigments) (Wink, 2008). For most plants, external factors (light, temperature, precipitation) can significantly affect their ability to synthesize secondary metabolites (Yang et al., 2018). Other studies have found that cultivar, climate, and growing conditions all have an impact on the quality and chemical composition of ten red raspberry genotypes (Mazur et al., 2014). Yang et al. [ 2020] also found that the most of the berries harvested from the dried temperate continental climate plateaus present higher antioxidant activity than the ones harvested from the continental climate (Georgescu et al., 2022). Recent research on blueberry and chokeberry revealed that the metabolite composition also depends on geo-climatic conditions, especially latitude. Berries harvested in geo-climatic zones of different latitudes have relatively high amino acid contents (Sim et al., 2017). In many cases, UV irradiation induces the processing of secondary metabolites, such as UV irradiation significantly increases flavonoid content (Chaves et al., 1997; Umek et al., 1999; Afshar et al., 2022), enhancing the total accumulation of anthocyanins (Gläßgen et al., 1998; Araguirang and Richter, 2022). This phenomenon, known as the defensive effect, is attributed to the ability of ROS scavenging as well as defending plants from excessive sunlight to growth enhancement (Dehghanian et al., 2022). These studies suggest that the environment induces the metabolites of raspberries grown in different environments. Climatic conditions are a major influencing factor. In our study, we found that the antioxidant properties of Qinghai raspberries were higher than those of Yunnan raspberries, which may be due to Qinghai being located in QZP with low temperature, high altitude, and long hours of sunlight of climatic conditions. These lead to increased production of ROS in plant cells to achieve antioxidant defense and increased content of metabolites with antioxidant properties. Primarily, flavonoid, phenolic acids, and amino acids and their derivatives were responses to abiotic stresses such as ultraviolet radiation, drought resistant, and cold resistance. Considering that Qinghai raspberry has strong antioxidant properties and the seeds are rich in secondary metabolites, the transcriptome and metabolite profile can be jointly analyzed to find how related genes regulate metabolites to have different expressions in Qinghai raspberry, thus promoting the diversity and variability of nutrients and bioactive compounds in Qinghai raspberry. ## Differential metabolites among raspberries in different parts The current study found that working with Qinghai red raspberry and Yunnan raspberry, the major part of the weight of fresh raspberries included pulp, therefore, it is expected that the berry and pulp could have comparable composition. In recent years, researchers investigated the correlation analysis and observed positive correlations between the antioxidant assays and polyphenolic groups of raspberries (Basu and Maier, 2016; Carmichael et al., 2021). In this study, the antioxidant activity was found in berry and pulp samples and may be related to the high concentration of 4 anthocyanidins (Pelargonidin-3-O-glucoside, Peonidin-3-O-glucoside, Cyanidin-3-O-sambubioside [Cyanidin-3-O-(2’’-O-xylosyl) glucoside], Cyanidin-3-O-(6’’-O-p-Coumaroyl) glucoside). The KEGG pathway enrichment analysis revealed that the differential metabolites between Q-R and Q-RP were mostly involved in phenylpropanoid biosynthesis, linoleic acid metabolism, biosynthesis of various plant secondary metabolites, and purine metabolism. By clustering all the differential metabolites in these pathways, in the Q-RP, phenolic acids (p-Coumaryl alcohol, Coniferaldehyde, p-Coumaraldehyde, Coniferyl alcohol), and amino acids and derivatives (L-Phenylalanine) are overrepresented in the phenylpropanoid biosynthesiss (Supplementary Figure S3A). Such as lignans and Coumarins (3,4-Dihydrocoumarin, Pinoresinol), phenolic acids (Gallic acid, Coniferyl alcohol) and amino acids and derivatives (L-Phenylalanine, L-Methionine) were clustered in the biosynthesis of various plant secondary metabolites as well as their derivatives were detected at a high level (Supplementary Figure S3B). Free fatty acids (13-KODE; (9Z,11E)-13-Oxooctadeca-9,11-dienoic acid, 12,13-Epoxy-9-Octadecenoic Acid, 7S,8S-DiHODE; (9Z,12Z)-(7S,8S)-Dihydroxyoctadeca-9,12-dienoic acid, 9[10]-EpOME;(9R,10S)-(12Z)-9,10-Epoxyoctadecenoic acid) which were highly detected in the Q-RP might attribute the antioxidant activity. In this study, linoleic acid metabolism and phenylpropanoid biosynthesis were significantly enriched in Y-R and Y-RP ($P \leq 0.05$). Caffeic acid, cinnamic acid, ferulic acid, and sinapic acid were up-regulated in Y-RP (Supplementary Figure S3D). 12 free fatty acids were significantly up-regulated in Y-R (Supplementary Figure S3C, E). This also showed that the antioxidant activity in berry and pulp causes differences in flavonoid and unsaturated fatty acids metabolites. Compared to the berry and pulp, seeds hold higher concentrations of flavones (Hispidulin-7-O-glucoside (Homoplantaginin)), free fatty acids (9[10]-EpOME;(9R,10S)-(12Z)-9,10-Epoxyoctadecenoic acid, 9,10,13-Trihydroxy-11-Octadecenoic Acid), phenolic acids (Raspberry ketone glucoside, Ethyl ferulate, 4’-Hydroxypropiophenone, Vanillin acetate), flavanones (hesperetin-7-O-rutinoside (hesperidin)*, hesperetin-7-O-neohesperidoside(Neohesperidin)*), tannin (pterocaryaninB), amino acids and derivatives (L-Lysine, L-Glutamine), making the seed fraction a considerable source of natural antioxidants. These findings prove that the seed part obtained from raspberry can supply a valuable source of functional ingredients with antioxidant properties for functional food and pharmaceutical purposes. Polyunsaturated fatty acids (PUFAs), including γ-Linolenic Acid* play multiple key roles in host defense and immunity, including anti-inflammation and antioxidative activity (Michalak et al., 2016). In addition, fatty acids in Q-RS mainly include Elaidic Acid, LysoPC 18:3*, LysoPC 18:3(2n isomer)*, α-Linolenic Acid*, and LysoPC 18:1*. In another study, Caidan et al. [ 2013] also reported different profiles consisting of palmitic, linolenic, linoleic, and stearic acids. Compared with the seeds, berry, and pulp of Qinghai raspberry, 325 metabolites are unique to Q-RS, including 67 phenolic acids, 52 terpenoids, 14 lipids (8 free fatty acids), 24 amino acids and their derivatives, 82 flavonoids, and 4 tannins (Supplementary Table S4). More than half of the total metabolites are relatively higher in the pulp. Therefore, the higher antioxidant capacity of Qinghai raspberry seeds is inseparable from their unique metabolites. As shown in Figure 12, the results of the top 20 metabolites in terms of the fold of difference in each group comparison. As presented in Figure 12E and F, Comparison of Q-RP and Q-RS, Q-R and Q-RS, Cyanidin-3-O-(6’’-O-p-coumaroyl) glucoside-5-O-glucoside, Cyanidin-3,5-O-diglucoside (Cyanin), Quercetin3-O-galactoside (log2FC = 21.99, 21.13, 20.45) are specific to Q-R and Q-RP. Tannins, as metabolic antioxidants, allows plants to respond to adverse environmental conditions (Casacchia et al., 2019). Procyanidin B2 and Butyl isobutyl phthalate (log2FC = 20.68, 20.73) were only detected in Qinghai raspberries’ seed. Besides, Procyanidin B2 is described as an important and unique anthocyanin in the red raspberry seed. Notably, in the present study, among the top 20 differential up-regulated or down-regulated metabolites (Q-RP and Q-RS, Q-R and Q-RS). Trans-4-Hydroxy-L-proline, DL-Leucine, and DL-Methionine (log2FC = 21.43, 21.06, 20.44) were unique amino acids and their derivatives to seeds in Qinghai raspberries. The difference in composition and content of flavonoids, phenolic acid, amino acids and their derivatives, and fatty acids might lead to strong antioxidant activities in the seeds from Qinghai raspberry. The FAO reported that the world population is projected to grow by $34\%$, from 6.8 billion today to 9.1 billion in 2050. The total production of fruits (650,684 tons) was adequate for the year 2017 (Mishra et al., 2022). However, it is not easy to evaluate the availability and cost of fruits on account of these variable population estimates. Therefore, greater production will be needed, particularly raspberries with a high antioxidant capacity, by the years 2025 and 2050, as the consumer trust that an appropriate diet reduces illness with decreased costs of pharmaceuticals is also considered to promote the demand for raspberries. Atkinson et al. [ 2013] proved that an inadequate winter chill induced the declining yield of perennial fruit species in Europe and the Americas. Moreover, the surveys of the FAO in 2017 have reported that raspberry productivity was 2.19 tons per ha in Serbia (Europe, an annual average temperature: 5-15°C), while this was only 0.64 tons per ha in Mexico (Americas, an annual average temperature: 16-28°C). Therefore, although *Qinghai is* located in Qinghai-Xizang plateau, its climate and geographical conditions are appropriate for planting raspberries, and the quality of raspberry grown in this area are better than in Yunnan. In addition, to meet the demand for functional substances and prevent human diseases in the context of global population growth. Therefore, we suggest that Qinghai raspberries with high antioxidant substances contents are widely grown in Qinghai. As a significant part of phytochemicals remain in the raspberry seed fraction, particularly Qinghai raspberry, it can be a potential source of functional ingredients to increase the utilization of raspberry. ## Conclusions In this study, LC-MS-based metabolomics and biochemical indicators were used to investigate the antioxidant properties of raspberries and their parts from two different plateaus. The results showed that there were metabolic differences between Qinghai raspberry and Yunnan raspberry, with different pathways being affected in each. The antioxidant capacity of the raspberry was found to be primarily related to the content and types of flavonoids, free fatty acids, and phenolic acids in the berry and pulp. In addition, the unique composition and content of flavonoids [Cyanidin-3-O-(6’’-O-p-coumaroyl) glucoside-5-O-glucoside, Cyanidin-3,5-O-diglucoside (Cyanin), and Quercetin3-O-galactoside], tannin (Procyanidin B2), phenolic acids (Butyl isobutyl phthalate), amino acids (L-Lysine and L-Glutamine) and fatty acids found in the seeds of Qinghai raspberry were shown to contribute to their strong antioxidant activities. The study suggests that the red raspberry seeds fraction could be a valuable source of functional ingredients for increasing the utilization of red raspberry. Overall, the research provides new insights into the chemical composition of red raspberry and its parts, and highlights the potential health benefits associated with its consumption. ## Data availability statement The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding authors. ## Author contributions XR conceived the project, data analysis, and edited the manuscript. YS designed the research, reviewed and edited the manuscript, made strict revisions to the grammar of the manuscript and performed the funding acquisition. SW, JW, and DX revised the paper. YY performed the research and discussed the results. All authors contributed to the article and approved the submitted version. ## Conflict of interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ## Publisher’s note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. ## Supplementary material The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2023.1143439/full#supplementary-material ## References 1. Afonso S., Oliveira I. V., Meyer A. S., Aires A., Saavedra M. J., Gonçalves B.. **Phenolic profile and bioactive potential of stems and seed kernels of sweet cherry fruit**. *Antioxidants.* (2020) **9**. DOI: 10.3390/antiox9121295 2. Afshar M., Najafian S., Radi M.. **Seasonal variation on the major bioactive compounds: total phenolic and flavonoids contents, and antioxidant activity of rosemary from shiraz**. *Nat. Prod. Res.* (2022) **36** 4287-4292. DOI: 10.1080/14786419.2021.1978998 3. Alarcón-Flores M. I., Romero-González R., Vidal J. L. M., Frenich A. G.. **Multiclass determination of phytochemicals in vegetables and fruits by ultra high performance liquid chromatography coupled to tandem mass spectrometry**. *Food Chem.* (2013) **141** 1120-1129. DOI: 10.1016/j.foodchem.2013.03.100 4. Al-Malki A. L.. **Suppression of acute pancreatitis by l-lysine in mice**. *BMC Complement Altern. Med.* (2015) **15** 193. DOI: 10.1186/s12906-015-0729-x 5. Araguirang G. E., Richter A. S.. **Activation of anthocyanin biosynthesis in high light - what is the initial signal**. *New Phytol.* (2022) **236** 2037-2043. DOI: 10.1111/nph.18488 6. Atkinson C. J., Brennan R. M., Jones H. G.. **Declining chilling and its impact on temperate perennial crops**. *Environ. Exp. Bot.* (2013) **91** 48-62. DOI: 10.1016/j.envexpbot.2013.02.004 7. Baby B., Antony P., Vijayan R.. **Antioxidant and anticancer properties of berries**. *Crit. Rev. Food Sci. Nutr.* (2018) **58** 2491-2507. DOI: 10.1080/10408398.2017.1329198 8. Bader Ul Ain H., Tufail T., Javed M., Tufail T., Arshad M. U., Hussain M.. **Phytochemical profile and pro-healthy properties of berries**. *Int. J. Food Pro.* (2022) **25** 1714-1735. DOI: 10.1080/10942912.2022.2096062 9. Badin E. E., Augusto P. E. D., Quevedo-Leon R. A., Ibarz A., Ribotta P. D., Lespinard A. R.. **Raspberry pulp pasteurization: Computational fluid dynamics modeling and experimental validation of color and bioactive compound retention**. *J. Food Process Engineering.* (2023) **46** e14168. DOI: 10.1111/jfpe.14168 10. Basu P., Maier C.. *Phcog Res.* (2016) **8** 258. DOI: 10.4103/0974-8490.188875 11. Bhattacharjee S.. *Reactive oxygen species in plant biology* (2019) 107-125 p 12. Bilawal A., Ishfaq M., Gantumur M.-A., Qayum A., Shi R., Fazilani S. A.. **A review of the bioactive ingredients of berries and their applications in curing diseases**. *Food Biosci.* (2021) **44**. DOI: 10.1016/j.fbio.2021.101407 13. Burton-Freeman B. M., Sandhu A. K., Edirisinghe I.. **Red raspberries and their bioactive polyphenols: cardiometabolic and neuronal health links**. *Adv. Nutr.* (2016) **7** 44-65. DOI: 10.3945/an.115.009639 14. Caidan R., Cairang L., Liubin , Yourui S.. **Simultaneous analysis of fatty acids in rubus niveus thunb. fruits by HPLC-MS/MS**. *Asian J. Chem.* (2013) **25** 1866-1870. DOI: 10.14233/ajchem.2013.13204 15. Cao Z., Li Y., Yu C., Li S., Zhang X., Tian Y.. **Effect of high hydrostatic pressure on the quality of red raspberry wine**. *J. Food Process. Preserv.* (2022) **46** e16030. DOI: 10.1111/jfpp.16030 16. Carmichael E., Najjar R., Meister M., Feresin R.. **Comparing the polyphenolic and flavonoid content, antioxidant and radical scavenging capacity of widely consumed berries**. *Curr. Developments Nutr.* (2021) **5** 301-301. DOI: 10.1093/cdn/nzab037_011 17. Casacchia T., Occhiuzzi M. A., Grande F., Rizzuti B., Granieri M. C., Rocca C.. **A pilot study on the nutraceutical properties of the citrus hybrid tacle**. *J. Funct. Foods.* (2019) **52** 370-381. DOI: 10.1016/j.jff.2018.11.030 18. Chaves N., Escudero J. C., Gutierrez-Merino C.. **Role of ecological variables in the seasonal variation of flavonoid content of cistus ladanifer exudate**. *J. Chem. Ecol.* (1997) **23** 579-603. DOI: 10.1023/B:JOEC.0000006398.79306.09 19. Dehghanian Z., Habibi K., Dehghanian M., Aliyar S., Asgari Lajayer B., Astatkie T.. **Reinforcing the bulwark: unravelling the efficient applications of plant phenolics and tannins against environmental stresses**. *Heliyon.* (2022) **8**. DOI: 10.1016/j.heliyon.2022.e09094 20. de Souza V. R., Pereira P. A. P., da Silva T. L. T., de Oliveira Lima L. C., Pio R., Queiroz F.. **Determination of the bioactive compounds, antioxidant activity and chemical composition of Brazilian blackberry, red raspberry, strawberry, blueberry and sweet cherry fruits**. *Food Chem.* (2014) **156** 362-368. DOI: 10.1016/j.foodchem.2014.01.125 21. Duan G., Li C., Liu Y., Ma X., Luo Q., Yang J.. **Magnaporthe oryzae systemic defense trigger 1 (MoSDT1)-mediated metabolites regulate defense response in rice**. *BMC Plant Biol.* (2021) **21** 40. DOI: 10.1186/s12870-020-02821-6 22. Dudonné S., Vitrac X., Coutière P., Woillez M., Mérillon J.-M.. **Comparative study of antioxidant properties and total phenolic content of 30 plant extracts of industrial interest using DPPH, ABTS, FRAP, SOD, and ORAC assays**. *J. Agric. Food Chem.* (2009) **57** 1768-1774. DOI: 10.1021/jf803011r 23. Fu L., Xu B.-T., Xu X.-R., Gan R.-Y., Zhang Y., Xia E.-Q.. **Antioxidant capacities and total phenolic contents of 62 fruits**. *Food Chem.* (2011) **129** 345-350. DOI: 10.1016/j.foodchem.2011.04.079 24. Fukushima A., Kusano M., Redestig H., Arita M., Saito K.. **Metabolomic correlation-network modules in arabidopsis based on a graph-clustering approach**. *BMC Syst. Biol.* (2011) **5**. DOI: 10.1186/1752-0509-5-1 25. Gao Y., Fang L., Wang X., Lan R., Wang M., Du G.. **Antioxidant activity evaluation of dietary flavonoid hyperoside using saccharomyces cerevisiae as a model**. *Molecules.* (2019) **24**. DOI: 10.3390/molecules24040788 26. Gedük A.Ş., Atsız S.. **LC-MS/MS phenolic composition of peach (**. *S. Afr. J. Bot.* (2022) **147** 636-645. DOI: 10.1016/j.sajb.2022.02.026 27. Georgescu C., Frum A., Virchea L.-I., Sumacheva A., Shamtsyan M., Gligor F.-G.. **Geographic variability of berry phytochemicals with antioxidant and antimicrobial properties**. *Molecules.* (2022) **27**. DOI: 10.3390/molecules27154986 28. Gläßgen W. E., Rose A., Madlung J., Koch W., Gleitz J., Seitz H. U.. **Regulation of enzymes involved in anthocyanin biosynthesis in carrot cell cultures in response to treatment with ultraviolet light and fungal elicitors**. *Planta* (1998) **204** 490-498. DOI: 10.1007/s004250050283 29. Gođevac D., Tešević V., Vajs V., Milosavljević S., Stanković M.. **Antioxidant properties of raspberry seed extracts on micronucleus distribution in peripheral blood lymphocytes**. *Food. Chem. Toxicol.* (2009) **47** 2853-2859. DOI: 10.1016/j.fct.2009.09.006 30. Gomes M., de S. C., das G. M., Guimarães A. C. G., Guerreiro A. C., Gago C. M. L.. **Effect of edible coatings with essential oils on the quality of red raspberries over shelf-life**. *J. Sci. Food Agric.* (2017) **97** 929-938. DOI: 10.1002/jsfa.7817 31. Goodman C., Lyon K., Scotto A., Roe M. M., Moghimpour F., Sebrell T. A.. **A high-throughput metabolic microarray assay reveals antibacterial effects of black and red raspberries and blackberries against helicobacter pylori infection**. *Antibiotics.* (2020) **10** 845. DOI: 10.1101/2020.12.14.422663 32. Huang X., Wu Y., Zhang S., Yang H., Wu W., Lyu L.. **Variation in bioactive compounds and antioxidant activity of rubus fruits at different developmental stages**. *Foods.* (2022) **11**. DOI: 10.3390/foods11081169 33. Iannetta P. P. M., Wyman M., Neelam A., Jones C., Taylor M. A., Davies H. V.. **A causal role for ethylene and endo-beta-1,4-glucanase in the abscission of red-raspberry (**. *Physiol. Plant* (2000) **110** 535-543. DOI: 10.1111/j.1399-3054.2000.1100417.x 34. Jiang Y., Luo T., Tang Y., Chen S., Ni H., Chen Q.. **Isolation of a novel characterized**. *Food Sci. Hum. Wellness.* (2022) **11** 1018-1027. DOI: 10.1016/j.fshw.2022.03.029 35. Kafkas E., Özgen M., Özoğul Y., Türemiş N.. **Phytochemical and fatty acid profile of selected red raspberry cultivars: a comparative study**. *J. Food Quality.* (2008) **31** 67-78. DOI: 10.1111/j.1745-4557.2007.00184.x 36. Kausar R., Akhtar N.. **Evaluation of depegmenting and antierythemic effects of cosmetic emulgels containing raspberry fruit extract on human cheek skin**. *Int. J. Pharm. Pharm. Sci.* (2016) **9** 236. DOI: 10.22159/ijpps.2017v9i1.15123 37. Killiny N., Hijaz F.. **Amino acids implicated in plant defense are higher in Candidatus Liberibacter asiaticus-tolerant citrus varieties**. *Plant signaling & behavior* (2016) **11**. DOI: 10.1007/s004250050283 38. Kirina I. B., Belosokhov F. G., Titova L. V., Suraykina I. A., Pulpitow V. F.. **Biochemical assessment of berry crops as a source of production of functional food products**. *IOP Conf. Ser.: Earth Environ. Sci.* (2020) **548**. DOI: 10.1088/1755-1315/548/8/082068 39. Kobori R., Yakami S., Kawasaki T., Saito A.. **Changes in the polyphenol content of red raspberry fruits during ripening**. *Horticulturae.* (2021) **7**. DOI: 10.3390/horticulturae7120569 40. Kosmala M., Zduńczyk Z., Juśkiewicz J., Jurgoński A., Karlińska E., Macierzyński J.. **Chemical composition of defatted strawberry and raspberry seeds and the effect of these dietary ingredients on polyphenol metabolites, intestinal function, and selected serum parameters in rats**. *J. Agric. Food Chem.* (2015) **63** 2989-2996. DOI: 10.1021/acs.jafc.5b00648 41. Kumar V., Sharma A., Kaur R., Thukral A. K., Bhardwaj R., Ahmad P.. **Differential distribution of amino acids in plants**. *Amino Acids* (2017) **49** 821-869. DOI: 10.1007/s00726-017-2401-x 42. Lebedev V. G., Lebedeva T. N., Vidyagina E. O., Sorokopudov V. N., Popova A. A., Shestibratov K. A.. **Relationship between phenolic compounds and antioxidant activity in berries and leaves of raspberry genotypes and their genotyping by SSR markers**. *Antioxidants.* (2022) **11**. DOI: 10.3390/antiox11101961 43. Lenucci M. S., Tornese R., Mita G., Durante M.. **Bioactive compounds and antioxidant activities in different fractions of mango fruits (**. *Antioxidants.* (2022) **11**. DOI: 10.3390/antiox11030484 44. Li H., Deng Z., Wu T., Liu R., Loewen S., Tsao R.. **Microwave-assisted extraction of phenolics with maximal antioxidant activities in tomatoes**. *Food Chem.* (2012) **130** 928-936. DOI: 10.1016/j.foodchem.2011.08.019 45. Li W., Wen L., Chen Z., Zhang Z., Pang X., Deng Z.. **Study on metabolic variation in whole grains of four proso millet varieties reveals metabolites important for antioxidant properties and quality traits**. *Food Chem.* (2021) **357**. DOI: 10.1016/j.foodchem.2021.129791 46. Li Z., Yang S., Wang X., Liao Q., Zhang W., Liu J.. **Widely targeted metabolomics analysis reveals the effect of exogenous auxin on postharvest resistance to botrytis cinerea in kiwifruit (**. *Postharvest Biol. Technol.* (2023) **195**. DOI: 10.1016/j.postharvbio.2022.112129 47. Luo L.-L., Li Y.-F., Shan H.-M., Wang L.-P., Yuan F., Ma Y.-Y.. **L-glutamine protects mouse brain from ischemic injury**. *CNS Neurosci. Ther.* (2019) **25** 1030-1041. DOI: 10.1111/cns.13184 48. Marcillo-Parra V., Tupuna-Yerovi D. S., González Z., Ruales J.. **Encapsulation of bioactive compounds from fruit and vegetable by-products for food application - a review**. *Trends Food Sci. Tech.* (2021) **116** 11-23. DOI: 10.1016/j.tifs.2021.07.009 49. Mazur S. P., Nes A., Wold A.-B., Remberg S. F., Aaby K.. **Quality and chemical composition of ten red raspberry (**. *Food Chem.* (2014) **160** 233-240. DOI: 10.1016/j.foodchem.2014.02.174 50. Michalak A., Mosińska P., Fichna J.. **Polyunsaturated fatty acids and their derivatives: therapeutic value for inflammatory, functional gastrointestinal disorders, and colorectal cancer**. *Front. Pharmacol.* (2016) **7**. DOI: 10.3389/fphar.2016.00459 51. Mishra R., Tripathi A. D., Singh R. B., Tomar R. S., Wilson D. W., Smail M. M.. *Estimates of functional food and nutraceutical availability in the world, with reference to food peroxidation and food safety* (2022) 23-42 52. Mudge E., Applequist W. L., Finley J., Lister P., Townesmith A. K., Walker K. M.. **Variation of select flavonols and chlorogenic acid content of elderberry collected throughout the Eastern united states**. *J. Food Compos. Anal.* (2016) **47** 52-59. DOI: 10.1016/j.jfca.2015.12.003 53. Nie X.-R., Li H.-Y., Wei S.-Y., Han Q.-H., Zhao L., Zhang Q.. **Changes of phenolic compounds, antioxidant capacities, and inhibitory effects on digestive enzymes of kiwifruits (**. *Food Measure.* (2020) **14** 1765-1774. DOI: 10.1007/s11694-020-00424-1 54. Nile S. H., Park S. W.. **Edible berries: Bioactive components and their effect on human health**. *Nutrition.* (2014) **30** 134-144. DOI: 10.1016/j.nut.2013.04.007 55. Noratto G. D., Chew B. P., Atienza L. M.. **Red raspberry (**. *Food Chem.* (2017) **227** 305-314. DOI: 10.1016/j.foodchem.2017.01.097 56. Ono E., Hatayama M., Isono Y., Sato T., Watanabe R., Yonekura-Sakakibara K.. **Localization of a flavonoid biosynthetic polyphenol oxidase in vacuoles**. *Plant J.* (2006) **45** 133-143. DOI: 10.1111/j.1365-313X.2005.02625.x 57. Panche A. N., Diwan A. D., Chandra S. R.. **Flavonoids: an overview**. *J. Nutr. Sci.* (2016) **5** e47. DOI: 10.1017/jns.2016.41 58. Qiu W., Su W., Cai Z., Dong L., Li C., Xin M.. **Combined analysis of transcriptome and metabolome reveals the potential mechanism of coloration and fruit quality in yellow and purple**. *J. Agric. Food Chem.* (2020) **68** 12096-12106. DOI: 10.1021/acs.jafc.0c03619 59. Ramalingam A. P., Mohanavel W., Premnath A., Muthurajan R., Prasad P. V. V., Perumal R.. **Large-Scale non-targeted metabolomics reveals antioxidant, nutraceutical and therapeutic potentials of sorghum**. *Antioxidants.* (2021) **10**. DOI: 10.3390/antiox10101511 60. Re R., Pellegrini N., Proteggente A., Pannala A., Yang M., Rice-Evans C.. **Antioxidant activity applying an improved ABTS radical cation decolorization assay**. *Free Radical Bio. Med.* (1999) **26** 1231-1237. DOI: 10.1016/S0891-5849(98)00315-3 61. Río Segade S., Paissoni M. A., Giacosa S., Bautista-Ortín A. B., Gómez-Plaza E., Gerbi V.. **Winegrapes dehydration under ozone-enriched atmosphere: Influence on berry skin phenols release, cell wall composition and mechanical properties**. *Food Chem.* (2019) **271** 673-684. DOI: 10.1016/j.foodchem.2018.07.218 62. Rolnik A., Soluch A., Kowalska I., Olas B.. **Antioxidant and hemostatic properties of preparations from asteraceae family and their chemical composition - comparative studies**. *Biomed. Pharmacother.* (2021) **142**. DOI: 10.1016/j.biopha.2021.111982 63. Sarv V., Venskutonis P. R., Rätsep R., Aluvee A., Kazernavičiūtė R., Bhat R.. **Antioxidants characterization of the fruit, juice, and pomace of sweet rowanberry (**. *Antioxidants.* (2021) **10**. DOI: 10.3390/antiox10111779 64. Scalbert A., Andres-Lacueva C., Arita M., Kroon P., Manach C., Urpi-Sarda M.. **Databases on food phytochemicals and their health-promoting effects**. *J. Agric. Food Chem.* (2011) **59** 4331-4348. DOI: 10.1021/jf200591d 65. Seeram N. P., Adams L. S., Zhang Y., Lee R., Sand D., Scheuller H. S.. **Blackberry, black raspberry, blueberry, cranberry, red raspberry, and strawberry extracts inhibit growth and stimulate apoptosis of human cancer cells**. *J. Agric. Food Chem.* (2006) **54** 9329-9339. DOI: 10.1021/jf061750g 66. Segla Koffi Dossou S., Xu F., You J., Zhou R., Li D., Wang L.. **Widely targeted metabolome profiling of different colored sesame (**. *Food Res. Int.* (2022) **151**. DOI: 10.1016/j.foodres.2021.110850 67. Sim I., Suh D. H., Singh D., Do S.-G., Moon K. H., Lee J. H.. **Unraveling metabolic variation for blueberry and chokeberry cultivars harvested from different geo-climatic regions in Korea**. *J. Agric. Food Chem.* (2017) **65** 9031-9040. DOI: 10.1021/acs.jafc.7b04065 68. Sójka M., Macierzyński J., Zaweracz W., Buczek M.. **Transfer and mass balance of ellagitannins, anthocyanins, flavan-3-ols, and flavonols during the processing of red raspberries (**. *J. Agric. Food Chem.* (2016) **64** 5549-5563. DOI: 10.1021/acs.jafc.6b01590 69. Stec K., Kozłowska J., Wróblewska-Kurdyk A., Kordan B., Anioł M., Gabryś B.. **Effect of naringenin and its derivatives on the probing behavior of**. *Molecules.* (2020) **25**. DOI: 10.3390/molecules25143185 70. Stüwe K., Robl J., Turab S. A., Sternai P., Stuart F. M.. **Feedbacks between sea-floor spreading, trade winds and precipitation in the southern red Sea**. *Nat. Commun.* (2022) **13** 5405. DOI: 10.1038/s41467-022-32293-1 71. Thaipong K., Boonprakob U., Crosby K., Cisneros-Zevallos L., Hawkins Byrne D.. **Comparison of ABTS, DPPH, FRAP, and ORAC assays for estimating antioxidant activity from guava fruit extracts**. *J. Food Compos Anal.* (2006) **19** 669-675. DOI: 10.1016/j.jfca.2006.01.003 72. Umek A., Kreft S., Kartnig T., Heydel B.. **Quantitative phytochemical analyses of six**. *Planta Med.* (1999) **65** 388-390. DOI: 10.1055/s-2006-960798 73. Vara A. L., Pinela J., Dias M. I., Petrović J., Nogueira A., Soković M.. **Compositional features of the “Kweli” red raspberry and its antioxidant and antimicrobial activities**. *Foods.* (2020) **9**. DOI: 10.3390/foods9111522 74. Wang P.-W., Cheng Y.-C., Hung Y.-C., Lee C.-H., Fang J.-Y., Li W.-T.. **Red raspberry extract protects the skin against UVB-induced damage with antioxidative and anti-inflammatory properties**. *Oxid. Med. Cell Longev.* (2019) **2019** 1-14. DOI: 10.1155/2019/9529676 75. Wang A., Li R., Ren L., Gao X., Zhang Y., Ma Z.. **A comparative metabolomics study of flavonoids in sweet potato with different flesh colors (**. *Food Chem.* (2018) **260** 124-134. DOI: 10.1016/j.foodchem.2018.03.125 76. Wang L., Wang Y., Hou Y., Zhu X., Zheng Y., Jin P.. **Physiological and metabolomic analyses of hot water treatment on amino acids and phenolic metabolisms in peach cold tolerance**. *Postharvest Biol. Technol.* (2021) **179**. DOI: 10.1016/j.postharvbio.2021.111593 77. Wink M.. **Plant secondary metabolism: diversity, function and its evolution**. *Nat. Prod. Commun.* (2008) **3**. DOI: 10.1177/1934578X0800300801 78. Xi W., Lu J., Qun J., Jiao B.. **Characterization of phenolic profile and antioxidant capacity of different fruit part from lemon (**. *J. Food Sci. Technol.* (2017) **54** 1108-1118. DOI: 10.1007/s13197-017-2544-5 79. Xiao J., Gu C., He S., Zhu D., Huang Y., Zhou Q.. **Widely targeted metabolomics analysis reveals new biomarkers and mechanistic insights on chestnut (**. *Food Res. Int.* (2021) **141**. DOI: 10.1016/j.foodres.2021.110128 80. Xu N., Chen G., Liu H.. **Antioxidative categorization of twenty amino acids based on experimental evaluation**. *Molecules.* (2017) **22**. DOI: 10.3390/molecules22122066 81. Yang J., Cui J., Chen J., Yao J., Hao Y., Fan Y.. **Evaluation of physicochemical properties in three raspberries (**. *Sci. Hortic.* (2020) **263**. DOI: 10.1016/j.scienta.2019.109146 82. Yang H., Tian C., Ji S., Ni F., Fan X., Yang Y.. **Integrative analyses of metabolome and transcriptome reveals metabolomic variations and candidate genes involved in sweet cherry (**. *PloS One* (2021) **16**. DOI: 10.1371/journal.pone.0260004 83. Yang L., Wen K.-S., Ruan X., Zhao Y.-X., Wei F., Wang Q.. **Response of plant secondary metabolites to environmental factors**. *Molecules.* (2018) **23**. DOI: 10.3390/molecules23040762 84. Zhang S., Liu Z., Li X., Abubaker M. A., Liu X., Li Z.. **Comparative study of three raspberry cultivar (**. *Appl. Sci.* (2022) **12**. DOI: 10.3390/app12030990 85. Zhao W., Wang L., Chuai P., Wang R., Su Y., Yang R.. **The effects of cold pectinase maceration and pulp juice fermentation on physicochemical properties of raspberry wine**. *J. Food Process. Preserv.* (2022) **46**. DOI: 10.1111/jfpp.16646 86. Zheng J., Wu Z., Yang N., Zhou K., Hu W., Ou S.. **Widely targeted UHPLC-MS/MS metabolomic analysis on the chemical variation in blueberry-filled pastries during processing**. *Front. Nutr.* (2020) **7**. DOI: 10.3389/fnut.2020.569172 87. Zhu Z.-J., Schultz A. W., Wang J., Johnson C. H., Yannone S. M., Patti G. J.. **Liquid chromatography quadrupole time-of-flight mass spectrometry characterization of metabolites guided by the METLIN database**. *Nat. Protoc.* (2013) **8** 451-460. DOI: 10.1038/nprot.2013.004
--- title: 'Exploring Korean adolescent stress on social media: a semantic network analysis' authors: - JongHwi Song - JunRyul Yang - SooYeun Yoo - KyungIn Cheon - SangKyun Yun - YunHee Shin journal: PeerJ year: 2023 pmcid: PMC10042152 doi: 10.7717/peerj.15076 license: CC BY 4.0 --- # Exploring Korean adolescent stress on social media: a semantic network analysis ## Abstract ### Background Considering that adolescents spend considerable time on the Internet and social media and experience high levels of stress, it is difficult to find a study that investigates adolescent stress through a big data-based network analysis of social media. Hence, this study was designed to provide basic data to establish desirable stress coping strategies for adolescents based on a big data-based network analysis of social media for Korean adolescent stress. The purpose of this study was to [1] identify social media words that express stress in adolescents and [2] investigate the associations between those words and their types. ### Methods To analyse adolescent stress, we used social media data collected from online news and blog websites and performed semantic network analysis to understand the relationships among keywords extracted in the collected data. ### Results The top five words used by Korean adolescents were counselling, school, suicide, depression, and activity in online news, and diet, exercise, eat, health, and obesity in blogs. As the top keywords of the blog are mainly related to diet and obesity, it reflects adolescents’ high degree of interest in their bodies; the body is also a primary source of adolescent stress. In addition, blogs contained more content about the causes and symptoms of stress than online news, which focused more on stress resolution and coping. This highlights the trend that social blogging is a new channel for sharing personal information. ### Conclusions The results of this study are valuable as they were derived through a social big data analysis of data obtained from online news and blogs, providing a wide range of implications related to adolescent stress. Hence this study can contribute basic data for the stress management of adolescents and their mental health management in the future. ## Introduction Desirable mental health is central to adolescent development into healthy adulthood, so multifaceted efforts are urgently needed to properly manage excessive adolescent stress. The 2021 perception rate of stress ($38.8\%$) among Korean adolescents is $32.3\%$ for male students and $45.6\%$ for female students, and it tends to increase as they progress to higher grades (Korea Disease Control and Prevention Agency, 2022). Additionally, $26.8\%$ of Korean adolescents have experienced depression, $12.7\%$ thought of suicide, and $2.2\%$ have attempted suicide (Korea Disease Control and Prevention Agency, 2022). Likewise, approximately $37\%$ of high school students in the United States have experienced periods of persistent feelings of sadness or hopelessness during the past year, and nearly half of all female students in Korea have experienced persistent feelings of sadness or hopelessness in 2019 (CDC, 2021). Adolescents cannot escape from stress in the fierce competition of modern society. Overstressed adolescents often express their stress through delinquency or violence, causing social problems and, in severe cases, leading to suicidal thoughts (Park, Kang & Lee, 2017). Adolescents who cannot find an appropriate way to cope with stress levels that exceed their personal resources and can threaten their well-being often attempt suicide as a way of escaping reality (Kang & Shin, 2015). Suicidal ideation in adolescents was positively correlated with stress experienced in daily life. Although many previous studies on adolescent stress have been reported, such as factors affecting adolescent stress (Jang, 2022), related problems (Oh & Kweon, 2019), coping methods, and interventions to reduce stress (Lim, 2021), it remains a serious health and social problem. Korean adolescents spent an average of 285.2 and 397.1 min per day on weekdays and weekends, respectively, on their smartphones (Korea Disease Control and Prevention Agency, 2022). Simultaneously, social media use has markedly increased among adolescents. In the US, the proportion of young people between the ages of 13 and 17 years who have a smartphone has reached $89\%$, more than doubling over a 6-year period, and $70\%$ of teenagers use social media multiple times in a day, up from one-third of all teenagers in 2012 (Rideout & Robb, 2018). Vernon, Modecki & Barber [2018] analysis of Australian longitudinal data found that $86\%$ of students owned smartphones in Grade 8, increasing to $93\%$ by Grade 11, with increased use of social media communication; additionally, most adolescents rely on smartphones to obtain health information (Chau, Burgermaster & Mamykina, 2018). Recently, with the rapid spread of smartphones, smart TVs, and mobile Internet and social media, the available data has increased exponentially, leading to an era of big data whereby data are used in various fields, particularly in healthcare (Song, 2013). Thus, researchers collect social media messages to expand their knowledge and analyse the meaning of the data using social media analytics (Song & Ryu, 2015). Big data in social media is not just a technology to collect, process, and analyse massive amounts of data. The meaning that can be created from such data is more valuable. The core of big data technology is to analyse the pouring information and provide valuable new information and services (Choi, 2015). The information on stress on social media can be extremely useful for adolescents, given the amount of time they spend on social media platforms. Network analysis is a useful way to derive the characteristics of network types and characterise topics of interest in relation to each other (Kim & Kim, 2016). Semantic networks were used to infer the subjects used in the texts. Semantic Network Analysis (SNA) describes the relationships between related concepts through word co-occurrence analysis. By evaluating the networks that appear, the SNA can highlight the most prominent information in the body of the text (Featherstone et al., 2020). In addition, such analysis and visualisation help to easily understand the knowledge structure and implicit meaning of the phenomenon of interest (Yoon, 2013). Therefore, by analysing and categorising the connectivity of big data-based collection, analysis, and processing, the characteristics and structure of the contents related to stress in adolescents can be identified. ## Previous studies Previous studies that attempted big data-based SNA on adolescents analysed research trends related to childhood and adolescent cancer survivors in South Korea using word co-occurrence network analysis (Kang et al., 2021) and the knowledge structure of students with severe and multiple disabilities (Song, 2018). Another study analysed the perception of sports and physical activity in Korean adolescents through big data analysis over the last 10 years, collating data from Naver, Daum, and Google, which are the most widely used search engines (Park et al., 2020) in Korea, using TEXTOM 4.0. Yet another study collected data from search engines widely used in Korea to identify social media words that express adolescents’ dietary behaviours and identify the associations and types of such words and behaviours. It used text-mining techniques and SNA for related big data collected from the Internet on adolescents’ dietary behaviours (Song et al., 2022). A study on physical activity and exercise in school-age youth was conducted to provide a solution by analysing a large number of scientific articles through text mining (Pans et al., 2021). In the belief that social media plays an important role in adolescents’ life, a study describing the big data approach to social media has also been presented by analysing an ad hoc dataset from the eating disorder forum of a social media website (Moessner et al., 2018). Considering the long time spent on internet use and the high levels of stress among adolescents, we had difficulty finding a study that investigated adolescent stress using big data-based network analysis of social media in the process of reviewing previous studies. Therefore, this study was designed to provide basic data to establish desirable stress coping strategies for adolescents, based on big data-based network analysis of social media for Korean adolescent stress. This study aimed to [1] identify social media words that express stress in adolescents and [2] investigate the associations between those words and their types. ## Materials & Methods In this study, we used social media big data to analyse adolescents’ awareness of stress. The data collection and analysis processes used in this study are shown in Fig. 1 and this is the same method as described in the previous study (Song et al., 2022). First, we collected data on adolescent stress by crawling online news and blog websites. Then, we extracted keywords using natural language processing (NLP) and performed pre-processing of the extracted keywords. Next, a SNA was performed to understand the relationships among the extracted keywords. For this study, we implemented two Python programs using suitable libraries instead of using a non-free web-based big data analysis solution such as TEXTOM (TheIMC, 2018); one program collects data by web crawling and the other performs the pre-processing of collected data and SNA, except CONCOR analysis. The UCINET package was used for CONCOR analysis and data visualisation. **Figure 1:** *The study process.* ## Data collection We collected relevant data from online news and blogs on Naver [2022] and Daum [2022], the largest search engines in Korea. In August 2022, using the search keyword ‘adolescent stress’, we collected 654 news articles from Naver news and 1,654 blog posts from Naver and Daum blogs using a web crawling programme implemented in Python. Although the period for data collection was not specified, more than $90\%$ of the data are from the last 10 years. We overcame the anti-crawling functions of some websites using the Selenium library, which automates web browser interactions in the programme. Duplicate content frequently occurs in online news because several sources provide news articles with the same. Cosine similarity, defined as the cosine of the angle between two word vectors, is widely used for similarity measurement between documents (Rahutomo, Kitasuka & Aritsugi, 2012). We eliminated duplicate content by using cosine similarity. ## Pre-processing We refined the collected data by selecting only nouns, verbs, and adjectives by the unigram method using KoNLPy, an open-source Python library for Korean NLP (Park & Cho, 2014), and excluding stop-words which are commonly used words or unimportant words. Next, we selected the top 30 keywords based on The Term Frequency–Inverse Document Frequency (TF-IDF) values (Boom et al., 2015; Leskovec, Rajaraman & Ullman, 2011), the opinions of a counselling teacher, and a network analysis expert. The TF-IDF is the formal measure of how the occurrences of a given word are concentrated into relatively few documents. TF-IDF is calculated as tf(d,t) idf(d,t), where term frequency tf(d,t) represents the number of appearances of a specific word t in a specific document d. Inverse document frequency idf(d,t) is a factor which diminishes the weight of terms that occur very frequently in the document set and increases the weight of terms that occur rarely; it is represented by log(N/(df(t)+1)), where df(t) is the number of documents in which a specific word t appears. Because there is no quantitative criterion for the number of selected keywords, we have selected top 30 words as in many similar studies (Choi et al., 2022; Jung, 2022; Ye et al., 2022; Park, Lee & Hong, 2023). For the top 30 keywords, we constructed a frequency table; thereafter, a document term matrix (DTM) (Anandarajan, Hill & Nolan, 2019) was generated to represent the frequency of words per document. DTM can quantify the relationships between words and documents. Subsequently, a co-occurrence matrix (COM) was constructed to represent the relationship between the simultaneous appearances of words in all documents. To simplify the network analysis of the COM, the constructed COM was transformed into a binary matrix using the median of all its elements as a threshold value. If an element was higher than the threshold value, it was changed to 1; otherwise, it was changed to 0. ## Semantic network analysis and visualisation SNA was performed to understand the relationships among the top 30 words related to adolescent stress. Network centralities were calculated to identify important words, and a CONvergence analysis of an iterative CORrelation (CONCOR) analysis was performed to identify groups of words with the same relationship pattern. We calculated four network centralities of COM: [1] degree centrality—the number of nodes a particular node (Xie, 2005) is connected to; [2] betweenness centrality—a measure of the mediation role of a node in a network; [3] closeness centrality—the inverse of the mean distance to all other nodes, which indicates how close a node is to all other nodes; and [4] eigenvector centrality—a measure of the influence of a node in a network (Tabassum et al., 2018) using the NetworkX (Hagberg, Swart & Chult, 2008) Python library. CONCOR repeatedly partitions nodes into subsets based on structural equivalence and analyses Pearson’s correlations to search for groups with certain levels of similarity. It forms clusters, including nodes with similarities to each other (Breiger, Boorman & Arabie, 1975). CONCOR analysis was performed using the UCINET 6.0 software package (Borgatti, Everett & Freeman, 2002) for the analysis of social networks, and the clustering results were visualised using NetDraw. ## The frequencies of keywords related to adolescent stress The frequencies of the top 30 words for adolescent stress in online news and blogs are shown in Table 1. The top five keywords were ‘counselling’, ‘school’, ‘suicide’, ‘depression’, and ‘activity’ in online news, and ‘diet’, ‘exercise’, ‘eat’, ‘health’, and ‘obesity’ in blogs. **Table 1** | rank | Word (news) | Freq | Word (blog) | Freq.1 | | --- | --- | --- | --- | --- | | 1 | counseling | 1143 | diet | 6553 | | 2 | school | 682 | exercise | 3863 | | 3 | suicide | 654 | eat | 2649 | | 4 | depression | 629 | health | 2626 | | 5 | activity | 619 | obesity | 2479 | | 6 | health | 585 | study | 2007 | | 7 | education | 509 | counseling | 1659 | | 8 | problem | 499 | treatment | 1463 | | 9 | parent | 454 | sleep | 1434 | | 10 | self-harm | 440 | problem | 1382 | | 11 | mental health | 423 | parent | 1379 | | 12 | treatment | 415 | skin | 1354 | | 13 | family | 403 | weight | 1351 | | 14 | study | 401 | help | 1316 | | 15 | experience | 398 | intake | 1241 | | 16 | participation | 348 | boxing | 1232 | | 17 | smoking | 335 | acne | 1113 | | 18 | friend | 331 | friend | 1099 | | 19 | game | 329 | oneself | 1077 | | 20 | oneself | 305 | person | 1076 | | 21 | person | 276 | activity | 1052 | | 22 | mind | 260 | mind | 915 | | 23 | relieve | 258 | body | 906 | | 24 | rest | 242 | rest | 880 | | 25 | online | 241 | school | 855 | | 26 | body | 230 | hair loss | 844 | | 27 | anxiety | 229 | control | 840 | | 28 | relationship | 226 | worry | 801 | | 29 | worry | 213 | relieve | 618 | | 30 | career | 188 | depression | 589 | ## Analysis of centralities of keywords related to adolescent stress Table 2 shows the centralities of keywords created using the keyword COM for online news. As the keyword ‘counselling’ was the most connected with three centralities, it had the highest degree of centrality, followed by ‘school’, ‘suicide’, ‘problem’, ‘self-harm’, and ‘depression’; the highest closeness centrality, followed by ‘school’, ‘suicide’, ‘problem’, ‘self-harm’, and ‘depression’; and the highest eigenvector centrality, followed by ‘school’, ‘suicide’, ‘problem’, ‘self-harm’, and ‘parent’. The keyword ‘suicide’ had the highest betweenness centrality, followed by ‘problem’, ‘counselling’, ‘school’, ‘activity’, and ‘depression’. **Table 2** | Rank | Keyword | Cd | Keyword.1 | Cb | Keyword.2 | Cc | Keyword.3 | Ce | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | 1 | counseling | 0.897 | suicide | 0.061 | counseling | 0.906 | counseling | 0.26 | | 2 | school | 0.897 | problem | 0.061 | school | 0.906 | school | 0.26 | | 3 | suicide | 0.897 | counseling | 0.058 | suicide | 0.906 | suicide | 0.259 | | 4 | problem | 0.862 | school | 0.058 | problem | 0.879 | problem | 0.253 | | 5 | self-harm | 0.828 | activity | 0.043 | self-harm | 0.853 | self-harm | 0.251 | | 6 | depression | 0.759 | depression | 0.04 | depression | 0.806 | parent | 0.237 | | 7 | activity | 0.759 | self-harm | 0.036 | activity | 0.806 | family | 0.23 | | 8 | parent | 0.759 | health | 0.034 | parent | 0.806 | activity | 0.23 | | 9 | health | 0.724 | parent | 0.03 | health | 0.784 | depression | 0.23 | | 10 | mental health | 0.724 | mental health | 0.029 | mental health | 0.784 | mental health | 0.228 | | 11 | family | 0.724 | education | 0.023 | family | 0.784 | health | 0.225 | | 12 | education | 0.69 | family | 0.022 | education | 0.763 | education | 0.214 | | 13 | oneself | 0.552 | treatment | 0.011 | oneself | 0.69 | oneself | 0.197 | | 14 | experience | 0.517 | experience | 0.007 | experience | 0.674 | experience | 0.185 | | 15 | treatment | 0.483 | oneself | 0.003 | treatment | 0.659 | study | 0.176 | | 16 | study | 0.483 | study | 0.003 | study | 0.659 | friend | 0.174 | | 17 | friend | 0.483 | friend | 0.003 | friend | 0.659 | treatment | 0.161 | | 18 | person | 0.414 | person | 0.0 | person | 0.63 | relationship | 0.156 | | 19 | relationship | 0.414 | relationship | 0.0 | relationship | 0.63 | person | 0.154 | | 20 | mind | 0.31 | body | 0.0 | mind | 0.592 | mind | 0.121 | | 21 | anxiety | 0.31 | game | 0.0 | anxiety | 0.592 | anxiety | 0.118 | | 22 | participation | 0.276 | participation | 0.0 | participation | 0.58 | relieve | 0.11 | | 23 | relieve | 0.276 | smoking | 0.0 | relieve | 0.58 | participation | 0.108 | | 24 | game | 0.241 | mind | 0.0 | game | 0.569 | online | 0.096 | | 25 | rest | 0.241 | relieve | 0.0 | rest | 0.569 | rest | 0.094 | | 26 | online | 0.241 | rest | 0.0 | online | 0.558 | career | 0.092 | | 27 | career | 0.241 | online | 0.0 | career | 0.558 | game | 0.09 | | 28 | smoking | 0.172 | anxiety | 0.0 | body | 0.537 | worry | 0.071 | | 29 | body | 0.172 | worry | 0.0 | worry | 0.537 | smoking | 0.063 | | 30 | worry | 0.172 | career | 0.0 | smoking | 0.527 | body | 0.06 | Table 3 shows the centralities of keywords which were made by using the keyword COM for blogs. The keyword ‘diet’ had the highest degree of centrality, followed by ‘exercise’, ‘health’, ‘study’, ‘treatment’, and ‘problem’; the highest closeness centrality, followed by ‘exercise’, ‘health’, ‘study’, ‘treatment’, and ‘problem’. The keyword ‘treatment’ had the highest betweenness centrality, followed by ‘exercise’, ‘diet’, ‘study’, ‘health’, and ‘sleep’. The keyword ‘health’ had the highest eigenvector centrality, followed by ‘diet’, ‘exercise’, ‘problem’, ‘study’, and ‘parent’. **Table 3** | Rank | Keyword | Cd | Keyword.1 | Cb | Keyword.2 | Cc | Keyword.3 | Ce | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | 1 | diet | 0.862 | treatment | 0.068 | diet | 0.879 | health | 0.254 | | 2 | exercise | 0.862 | exercise | 0.068 | exercise | 0.879 | diet | 0.254 | | 3 | health | 0.862 | diet | 0.061 | health | 0.879 | exercise | 0.253 | | 4 | study | 0.828 | study | 0.058 | study | 0.853 | problem | 0.247 | | 5 | treatment | 0.793 | health | 0.057 | treatment | 0.829 | study | 0.245 | | 6 | problem | 0.793 | sleep | 0.034 | problem | 0.829 | parent | 0.24 | | 7 | counseling | 0.759 | counseling | 0.032 | counseling | 0.806 | counseling | 0.234 | | 8 | parent | 0.759 | help | 0.027 | parent | 0.806 | eat | 0.234 | | 9 | eat | 0.724 | problem | 0.026 | eat | 0.784 | treatment | 0.233 | | 10 | sleep | 0.69 | parent | 0.021 | sleep | 0.763 | obesity | 0.216 | | 11 | obesity | 0.655 | eat | 0.015 | obesity | 0.744 | sleep | 0.215 | | 12 | help | 0.655 | friend | 0.014 | help | 0.744 | help | 0.209 | | 13 | oneself | 0.586 | activity | 0.012 | oneself | 0.707 | person | 0.204 | | 14 | person | 0.586 | obesity | 0.011 | person | 0.707 | oneself | 0.201 | | 15 | activity | 0.586 | oneself | 0.007 | activity | 0.707 | activity | 0.198 | | 16 | friend | 0.552 | person | 0.006 | friend | 0.69 | friend | 0.186 | | 17 | mind | 0.448 | skin | 0.005 | mind | 0.644 | mind | 0.165 | | 18 | body | 0.448 | body | 0.003 | body | 0.644 | body | 0.157 | | 19 | weight | 0.414 | worry | 0.002 | weight | 0.63 | school | 0.143 | | 20 | school | 0.379 | weight | 0.002 | school | 0.617 | control | 0.142 | | 21 | control | 0.379 | control | 0.001 | control | 0.617 | weight | 0.14 | | 22 | rest | 0.345 | intake | 0.0 | rest | 0.604 | rest | 0.133 | | 23 | intake | 0.31 | mind | 0.0 | intake | 0.58 | intake | 0.107 | | 24 | worry | 0.276 | school | 0.0 | worry | 0.58 | worry | 0.1 | | 25 | skin | 0.241 | boxing | 0.0 | skin | 0.558 | depression | 0.089 | | 26 | depression | 0.241 | acne | 0.0 | depression | 0.547 | skin | 0.073 | | 27 | acne | 0.138 | rest | 0.0 | acne | 0.518 | relieve | 0.052 | | 28 | relieve | 0.138 | hair loss | 0.0 | hair loss | 0.518 | hair loss | 0.041 | | 29 | boxing | 0.103 | relieve | 0.0 | relieve | 0.518 | acne | 0.041 | | 30 | hair loss | 0.103 | depression | 0.0 | boxing | 0.5 | boxing | 0.039 | ## Semantic network of clusters through CONCOR analysis related to adolescent stress Network groupings and visualisation of adolescent stress are shown in Figs. 2 and 3, respectively. Figure 2 shows the CONCOR analysis of the online news network of adolescent stress consisting of five clusters. We represented the cluster consisting of words 1, 2, … in [word 1, word 2, …]. The cluster [body, smoking, person, anxiety, mind, treatment, rest] could be considered ‘a pattern that occurs when adolescents are under stress’, as the cluster reflects having an anxious mind, searching for someone to be with, smoking, resting one’s body, or receiving treatment. The cluster [problem, parent, health, mental health, depression, experience, oneself] can be interpreted as ‘the causes and consequences of adolescent stress’, as adolescents’ experiences with problems between themselves and their parents affect their health and mental health, especially leading to depression. The cluster [worry, game, activity, relationship, friend] could be regarded as ‘a way for adolescents to relieve stress’, that is, to find a friend to relieve stress, talk about their worry, and play games or activities. The cluster [relieve, study, online, career, participation] could be interpreted as ‘to relieve the stress of studying, they participate in events such as online career experiences’. The cluster [suicide, self-harm, school, family, education, counselling] could be considered ‘education and counselling about self-harm or suicide due to stress is required at school and in the family’. **Figure 2:** *CONCOR analysis of news network of the adolescent stress.* **Figure 3:** *CONCOR analysis of blog network of the adolescent stress.* Figure 3 shows the CONCOR analysis of the blog network of adolescent stress, comprising seven clusters. The clusters [acne, skin, weight, intake] and [eat, obesity, health, diet, exercise, problem, parent] can be considered ‘sources of stress’; the cluster [school, rest, friend, boxing, relief] as in ‘coping with stress such as spending time with friends at school, resting, or boxing’; the cluster [study, counselling, activity] could be interpreted as ‘coping with the stress by studying, counselling, or activities’. The cluster [person, mind, depression, oneself] could be interpreted as ‘relationships between themselves and others causing stress and depression’, and the cluster [hair loss, control, body, worry] could be considered ‘a phenomenon that can occur when adolescents are under stress’ which were reflected in hair loss, (emotional) control, body (imbalance), and worry. The cluster [treatment, help, sleep] could be seen as ‘strategies adolescents can adopt to relieve stress’. ## Discussion Since the mental health of adolescents is extremely important for their future growth into healthy adults, various efforts are urgently needed to appropriately manage their current levels of stress. Therefore, in this study, we analysed online social big data on adolescent stress using text mining techniques. As a result of text mining of keywords related to adolescent stress, the top five words with high frequency were ‘counselling’, ‘school’, ‘suicide’, ‘depression’, and ‘activity’ in online news, and ‘diet’, ‘exercise’, ‘eat’, ‘health’, and ‘obesity’ in blogs. The tendency of the top five words appeared such that the words of the blogs appeared somewhat lighter than those of the online news; this reflects the fact that the blogs were lighter, more personal, and informal, and these features distinguished them from the news (Tereszkiewicz, 2014). The fact that the top keywords of the blog were mainly related to diet and obesity reflects the high interest of adolescents in their bodies (Yun, 2018), which was also confirmed as a source of immense stress among adolescents. Analysis of online news revealed that words that refer to resolving stress or behaviours caused by stress—such as counselling, school, self-harm, problem, and suicide—have high centrality. In contrast, in the centrality analysis of blogs, although treatment was highlighted, high centrality words reflecting the cause of stress—such as diet, study, and health—were also indicated. The result of ‘counselling’ having the highest connection with other keywords in network centrality analysis of the online news may indicate a lot of resolution in consideration of the social issues of adolescent stress due to the nature of news which has a formal character (Jeong & Kim, 2010). As the characteristics of blogs are considered temporary, personal, and informal (Thorsen & Jackson, 2018), it was confirmed that adolescents frequently shared personal causes of stress through individual blogs. Furthermore, since social media can influence adolescents’ self-views and interpersonal relationships through social comparisons and negative interactions, social media content often promotes self-harm and suicidal thoughts among adolescents (Abi-Jaoude, Naylor & Pignatiello, 2020). Corroborating the descriptions in this last cited study, another study reported that adolescents with depression and/or suicidality often use more social media and report worsening mood and suicide risk (John et al., 2018). Nonetheless, researchers also found that lower levels of social media use (overall and messaging) are associated with a greater likelihood of having suicidal ideation with plan over the next 30 days (Hamilton et al., 2021). Therefore, it is relevant to consider the importance of social media as an additional context for the topic of adolescent suicide and to educate adolescents to avoid placing indiscriminate trust on social media. Through CONCOR analysis, five clusters were identified in online news and seven in blogs. The causes and symptoms of stress and coping strategies were confirmed in both online news and blogs. However, online news contained multiple coping strategies for relieving stress, whereas blogs focused more on the causes of stress. ‘ Study’ as a cause of stress was included in both online news and blogs. Diet, obesity, acne, skin, hair loss, and various other causes of stress, such as school, and family, have been included in blogs and highlighted in previous studies. Contrarily, online news focused more on coping strategies for relieving adolescent stress. The importance of counselling for adolescent concerns, education, and counselling in schools and families was also confirmed. Korean adolescents are under a lot of stress due to their studies, and with excessive academic stress, they may experience mental health-related problems, such as depression (Kang, 2022). Because modern society tends to judge and evaluate people based on their appearance, it is common for adolescents to rate themselves based on their appearance. Acne occurrence is associated with stress and depression; therefore, acne treatment and skin care are considered necessary for improving mental health (Shin & Kim, 2019). Our analyses indicate that blogs contain more content about the causes and symptoms of stress than online news, reflecting the trend of social networks of casual and informal youth blogging as a new channel for sharing personal information (Li et al., 2016). In one study, researchers designed and implemented a microblogging platform to detect and relieve stress in teenagers; the authors mentioned the potential for stress detection because stressed individuals view microblogs as a channel for emotional release and interaction (Zhao et al., 2016). Young people who experience illnesses, including stress, tend to blog about them, and such blogs often have many followers. By means of blogging, young people living with an illness may succeed in having a social life and uphold and even extend their self-knowledge and self-esteem. The content of a blog can foster familiarity between the author and readers. Blogs can provide the authors’ unique experience-based knowledge and reflection to readers who read published articles. Therefore, blogging, especially on specific issues involving stress, should be continuously explored and recognised as a valuable source for such content in the future (Nesby & Salamonsen, 2016). This finding suggests that differences in blog authors’ subjective thoughts and direct experiences are used as the main basis for blogs, whereas online news focuses on delivering objective information and explanations based on the values of fairness and responsibility (Jeong & Kim, 2010). Nowadays, Internet usage is unavoidable for the younger generations. The online world is the primary source of information and quick communication; therefore, education about the correct use of the internet should be made reasonable at the earliest (Prievara & Piko, 2016). Finally, the results suggest that when establishing a stress coping strategy for adolescents, first, the information currently present in social media can be utilised by stakeholders; this is in consideration that blogs focused more on the causes of stress and online news contain relatively large amounts of information on stress coping. Second, since elements related to appearance and academics are often cited as causes of stress for Korean adolescents, they should be incorporated into stress coping interventions aimed at this population. Third, social media could be given greater importance within the context of adolescent suicide. ## Conclusions To contribute to a strategy for preventing and managing adolescent stress, we analysed social media data using text-mining techniques and derived the words and word associations from online news and blog content. We collected data from the two largest portals for Koreans, including Korean adolescents, using the search term ‘Adolescent stress’ and related words. In this study, we collected only data that is in Korean from Korean portal sites. Although information about adolescent stress is available on websites worldwide, only data in Korean language were collected to ensure consistency with keyword selection. Despite these limitations, the results of this study are valuable as they were derived through social big data analysis of data obtained from online news and blogs. The findings provide a wide range of implications related to adolescent stress; hence this study can contribute as basic data for the stress management of adolescents and their mental health management in the future. ## References 1. Abi-Jaoude E, Naylor KT, Pignatiello A. **Smartphones, social media use and youth mental health**. *Canadian Medical Association Journal* (2020) **192** E136-E141. DOI: 10.1503/cmaj.190434 2. Anandarajan M, Hill C, Nolan T, Anandarajan M, Hill C, Nolan T. **Term-document representation**. *Practical text analytics* (2019) 61-73 3. Boom CD, Canneyt SV, Bohez S, Demeester T, Dhoedt B. **Learning semantic similarity for very short texts**. (2015) 1229-1234 4. Borgatti SP, Everett MG, Freeman LC. *Ucinet for windows: software for social network analysis* (2002) 5. Breiger RL, Boorman SA, Arabie P. **An algorithm for clustering relational data with applications to social network analysis and comparison with multidimensional scaling**. *Journal of Mathematical Psychology* (1975) **12** 328-383. DOI: 10.1016/0022-2496(75)90028-0 6. **Youth risk behavior survey data summary & trend report 2009-2019**. (2021) 7. Chau MM, Burgermaster M, Mamykina L. **The use of social media in nutrition interventions for adolescents and young adults—a systematic review**. *International Journal of Medical Informatics* (2018) **120** 77-91. DOI: 10.1016/j.ijmedinf.2018.10.001 8. Choi CM, Lee JH, Machado J, Kim GW. **Big-data-based text mining and social network analysis of landscape response to future environmental change**. *Land* (2022) **11** 2183. DOI: 10.3390/land11122183 9. Choi JY. **Utilization value of medical big data created in operation of medical information system**. *Journal of the Korea Institute of Electronic Communication Sciences* (2015) **10** 1403-1410. DOI: 10.13067/JKIECS.2015.10.12.1403 10. **Daum portal**. (2022) 11. Featherstone JD, Ruiz JB, Barnett GA, Millam BJ. **Exploring childhood vaccination themes and public opinions on Twitter: a semantic network analysis**. *Telematics and Informatics* (2020) **54** 101474. DOI: 10.1016/j.tele.2020.101474 12. Hagberg A, Swart P, Chult DS. **Exploring network structure, dynamics, and function using networkx**. (2008) p 11 13. Hamilton JL, Biernesser C, Moreno MA, Porta G, Hamilton E, Johnson K, Poling KD, Sakolsky D, Brent DA, Goldstein TG. **Social media use and prospective suicidal thoughts and behaviors among adolescents at high risk for suicide**. *Suicide & Life-threatening Behavior* (2021) **51** 1203-1212. DOI: 10.1111/sltb.12801 14. Jang JS. **Influencing factors according to the stress perception in Korean adolescents**. *Korean Public Health Research* (2022) **48** 69-78. DOI: 10.22900/kphr.2022.48.3.005 15. Jeong IK, Kim YR. **Understanding online news types: a Comparative study of professional online news and blog news**. *Korean Journal of Journalism & Communication Studies* (2010) **54** 146-166 16. John A, Glendenning AC, Marchant A, Montgomery P, Stewart A, Wood S, Lloyd K, Hawton K. **Self-harm, suicidal behaviours, and cyberbullying in children and young people: systematic review**. *Journal of Medical Internet Research* (2018) **20**. DOI: 10.2196/jmir.9044 17. Jung S. **A study on changes in the food service industry about keyword before and after COVID-19 using big data**. *International Journal of Internet, Broadcasting and Communication* (2022) **14** 85-90. DOI: 10.7236/IJIBC.2022.14.3.85 18. Kang K-A, Han SJ, Chun J, Kim H-Y. **Research trends related to childhood and adolescent cancer survivors in South Korea using word co-occurrence network analysis**. *Child Health Nursing Research* (2021) **27** 201-210. DOI: 10.4094/chnr.2021.27.3.201 19. Kang JH, Shin TS. **The effects of adolescents‘ stress on suicidal ideation: focusing on the moderating and mediating effects of depression and social support**. *Korean Journal of Youth Studies* (2015) **22** 26-50 20. Kang SI. **The effects of academic and friendship stress on high-school students’ depression: mediating and moderating effects of self-esteem**. *Korean Journal of Social Welfare Education* (2022) **58** 31-55 21. Kim SH, Kim HJ. **Students social network analysis according to physical factors**. *Journal of Learner-Centered Curriculum and Instruction* (2016) **16** 439-456 22. **The statistics on the 16th Korea youth risk behavior web-based survey in 2021**. (2022) 23. Leskovec J, Rajaraman A, Ullman JD, Leskovec J, Rajaraman A, Ullman JD. **Data mining**. *Mining of Massive Datasets* (2011) 1-17 24. Li Q, Xue Y, Zhao L, Jia J, Feng L. **Analyzing and identifying teens’ stressful periods and stressor events from a microblog**. *IEEE Journal of Biomedical and Health Informatics* (2016) **21** 1434-1448. DOI: 10.1109/JBHI.2016.2586519 25. Lim YT. **Influence of behavioral activation system on coping with music among high school students: the mediating effect of music use in daily life**. *Korean Journal of Youth Studies* (2021) **28** 273-291. DOI: 10.21509/KJYS.2021.10.28.10.273 26. Moessner M, Feldhege J, Wolf M, Bauer S. **Analyzing big data in social media: text and network analyses of an eating disorder forum**. *International Journal of Eating Disorders* (2018) **51** 656-667. DOI: 10.1002/eat.22878 27. **Naver portal**. (2022) 28. Nesby L, Salamonsen A. **Youth blogging and serious illness**. *Medical Humanities* (2016) **42** 46-51. DOI: 10.1136/medhum-2015-010723 29. Oh DK, Kweon SY. **Causal relations among academic stress, career stress, anxiety, impulsivity, depression, suicide idea of adolescent**. *Journal of Future Oriented Youth Society* (2019) **16** 29-45 30. Pans M, Madera J, González L-M, Pellicer-Chenoll M. **Physical activity and exercise: text mining analysis**. *International Journal of Environmental Research and Public Health* (2021) **18** 9642. DOI: 10.3390/ijerph18189642 31. Park EL, Cho S. **KoNLPy: Korean natural language processing in Python**. (2014) 133-136 32. Park JH, Kang UG, Lee YH. **Big data decision analysis of stress on adolescent mental health**. *Journal of the Korea Society of Computer and Information* (2017) **22** 89-96. DOI: 10.9708/jksci.2017.22.11.089 33. Park JY, Lee J, Hong B. **Keyword Network Analysis of Infusion Nursing from Posts on the Q & A Board in the Intravenous Nurses Café**. *Healthcare Informatics Research* (2023) **29** 75-83. DOI: 10.4258/hir.2023.29.1.75 34. Park SU, Ahn HY, Kim DK, So WY. **Big data analysis of sports and physical activities among Korean adolescents**. *International Journal of Environmental Research and Public Health* (2020) **17** 5577. DOI: 10.3390/ijerph17155577 35. Prievara DK, Piko B. **Being online without a purpose—study of background variables of problematic internet use**. *Psychiatria Hungarica* (2016) **31** 146-156. PMID: 27244870 36. Rahutomo F, Kitasuka T, Aritsugi M. **Semantic cosine similarity**. (2012) **4** 1 37. Rideout V, Robb MB. *Social media, social life: teens reveal their experiences* (2018) 38. Shin DH, Kim SH. **Influence of acne skin on teenager’s stress, depression, and self-esteem—focusing on middle school students**. *International Journal of Advancements in Computing Technology* (2019) **9** 264-271. DOI: 10.22156/CS4SMB.2019.9.10.264 39. Song JH, Yoo SY, Yang JR, Yun SK, Shin YH. **Network analysis based on big data in social media of Korean adolescents’ diet behaviors**. *PLOS ONE* (2022) **17**. DOI: 10.1371/journal.pone.0273570 40. Song SM. **Analysis of knowledge structure of study on students with severe and multiple disabilities using semantic nNetwork analysis**. *Korean Journal of Physical, Multiple, & Health Disabilities* (2018) **61** 75-99. DOI: 10.20971/kcpmd.2018.61.4.75 41. Song TM. **Trend and applicaton of big data on health and welfare in Korea**. *Science & Technology Policy* (2013) **192** 56-73 42. Song TM, Ryu S. **Big data analysis framework for healthcare and social sectors in Korea**. *Healthcare Informatics Research* (2015) **21** 3-9. DOI: 10.4258/hir.2015.21.1.3 43. Tabassum S, Pereira FSF, Fernandes S, Gama J. **Social network analysis: an overview**. *WIREs Data Mining and Knowledge Discovery* (2018) **8**. DOI: 10.1002/widm.1256 44. Tereszkiewicz A. **I’m not sure what that means yet, but we’ll soon find out–the discourse of newspaper live blogs**. *Studia Linguistica Universitatis Iagellonicae Cracoviensis* (2014) **131** 299-319. DOI: 10.4467/20834624SL.14.018.2326 45. **TEXTOM**. (2018) 46. Thorsen E, Jackson D. **Seven characteristics defining online news formats**. *Digital Journalism* (2018) **6** 847-868. DOI: 10.1080/21670811.2018.1468722 47. Vernon L, Modecki KL, Barber BL. **Mobile phones in the bedroom: trajectories of sleep habits and subsequent adolescent psychosocial development**. *Child Development* (2018) **89** 66-77. DOI: 10.1111/cdev.12836 48. Xie Z. **Centrality measures in text mining: prediction of noun phrases that appear in abstracts**. *Association for Computational Linguistics Annual Meeting Conference Proceedings* (2005) **43** 103-108 49. Ye J, Zhang J, Gao L, Zhou Y, Liu Z, Han J. **Neo-Chinese style furniture design based on semantic analysis and connection**. *KSII Transactions on Internet & Information Systems* (2022) **16** 2704-2719. DOI: 10.3837/tiis.2022.08.013 50. Yoon HK. **Research on the application methods of big data within the cultural industry**. *Global Cultural Contents* (2013) **10** 157-179 51. Yun HJ. **Effect of body image distortion on mental health in adolescents**. *Journal of the Korea Socity of Health Informatics and Statistics* (2018) **43** 191-199. DOI: 10.21032/jhis.2018.43.3.191 52. Zhao L, Li Q, Xue Y, Jia J, Feng L. **A systematic exploration of the micro-blog feature space for teens stress detection**. *Health Information Science and Systems* (2016) **4** 3. DOI: 10.1186/s13755-016-0016-3
--- title: 'A new model predicts hepatocellular carcinoma in patients with HBV-related decompensated liver cirrhosis and long-term antiviral therapy: a prospective study' authors: - Hao-dan Mao - Shu-qin Zheng - Su-hua Yang - Ze-yu Huang - Yuan Xue - Min Zhou journal: PeerJ year: 2023 pmcid: PMC10042153 doi: 10.7717/peerj.15014 license: CC BY 4.0 --- # A new model predicts hepatocellular carcinoma in patients with HBV-related decompensated liver cirrhosis and long-term antiviral therapy: a prospective study ## Abstract ### Background We aimed to evaluate the prediction values of non-invasive models for hepatocellular carcinoma (HCC) development in patients with HBV-related liver cirrhosis (LC) and long-term NA treatment. ### Methods Patients with compensated or decompensated cirrhosis (DC), who achieved long-term virological response, were enrolled. DC and its stages were defined by the complications including ascites, encephalopathy, variceal bleeding, or renal failure. Prediction accuracy of several risk scores, including ALBI, CAMD, PAGE-B, mPAGE-B and aMAP, was compared. ### Results The median follow-up duration was 37 (28–66) months. Among the 229 patients, 9 ($9.57\%$) patients in the compensated LC group and 39 ($28.89\%$) patients in the DC group developed HCC. The incidence of HCC was higher in the DC group (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\cal X$\end{document}X2 = 12.478, $P \leq 0.01$). The AUROC of ALBI, aMAP, CAMD, PAGE-B and mPAGE-B scores were 0.512, 0.667, 0.638, 0.663, 0.679, respectively. There was no significant difference in AUROC between CAMD, aMAP, PAGE-B and mPAGE-B (all $P \leq 0.05$). Univariable analysis showed that age, DC status and platelet were associated with HCC development, and multivariable analysis showed that age and DC status (both $P \leq 0.01$) were independent risk factors for HCC development, then Model (Age_DC) was developed and its AUROC was 0.718. Another model, Model (Age_DC_PLT_TBil) consisting of age, DC stage, PLT, TBil was also developed, and its AUROC was larger than that of Model (Age_DC) (0.760 vs. 0.718). Moreover, AUROC of Model (Age_DC_PLT_TBil) was larger than the other five models (all $P \leq 0.05$). With an optimal cut-off value of 0.236, Model (Age_DC_PLT_TBil) achieved $70.83\%$ sensitivity, $76.24\%$ specificity. ### Conclusion There is a lack of non-invasive risk scores for HCC development in HBV-related DC, and a new model consisting of age, DC stage, PLT, TBil may be an alternative. ## Introduction Liver cancer is one of the most common cancers worldwide, of which over $90\%$ of the cases were hepatocellular carcinoma (HCC) (Su et al., 2022). The outcomes vary based on the severity of the underlying chronic liver disease and the tumor stage. HCC diagnosed at an intermediate or advanced stage, is related to a poor prognosis (Foerster et al., 2022). Despite long-term antiviral therapy, the risk of HCC remains high among patients with chronic HBV infection (Kim et al., 2020; Zhu et al., 2020). It is striking that HCC developed in $31.8\%$ of the patients with decompensated cirrhosis (DC) during a 5-year follow-up (Zhu et al., 2020). Considering that the incidence of HCC supports that increasing and early diagnosis improves the patients’ prognosis, HCC surveillance is recommended, particularly in patients with liver cirrhosis (LC). At the present time, several non-invasive models for HCC prediction have been reported, including risk estimation for HCC in chronic hepatitis B (CHB) score (REACH-B) (Yang et al., 2011), cirrhosis, age, male, and diabetes score (CAMD) (Hsu et al., 2018), platelets, age and gender score (PAGE-B) (Papatheodoridis et al., 2016), the modified PAGE-B score (mPAGE-B) (Kim et al., 2018), age, male, albumin-bilirubin, and platelets (aMAP) (Fan et al., 2020), albumin-bilirubin (ALBI) score (Casadei Gardini et al., 2019). PAGE-B, for which AUROC was higher than that of REACH-B, can easily identify high-risk cases of HCC in patients treated with nucleot(s)ide analogs (NAs) (Kirino et al., 2020). For patients with compensated LC, aMAP showed better predictive performance than CAMD, mPAGE-B and PAGE-B, according to the Harrell’s c-index (Gui et al., 2021). It should be noted that patients with DC were excluded from the above studies (Hsu et al., 2018; Papatheodoridis et al., 2016; Kim et al., 2018; Fan et al., 2020; Kirino et al., 2020; Lee et al., 2019). To date, there is limited information about the non-invasive models for HCC surveillance in patients with DC. Herein, we investigated the predictive value of the non-invasive prediction models in patients with HBV-related LC and long-term NA treatment. Moreover, a new model consisting of age, DC stage, platelet, and total bilirubin (TBil), was developed and it can identify patients with a high risk of HCC. ## Patients and primary endpoint From May 2010 to December 2021, 307 patients with LC admitted to the Third People’s hospital of Changzhou, were recruited, and were prospectively followed up. HBV-related compensated LC and DC were diagnosed according to Chinese guidelines for prevention and treatment of CHB (2019 version) (Jia et al., 2020). Patients with nodules in the hepatic parenchyma found in histological or ultrasonographic examination, or gastroesophageal varices detected by endoscopic evaluation, were diagnosed with compensated LC. DC and its stages were defined by the complications including ascites, encephalopathy, variceal bleeding, or hepatorenal syndrome (D’Amico, Garcia-Tsao & Pagliaro, 2006; D’Amico et al., 2018). HCC was diagnosed according to the imaging techniques, alpha-fetoprotein and/or histological findings (Health Commission of the People’s Republic of China, 2020). Patients co-infected with other hepatitis virus or suffered from malignant tumor, were excluded. Patients who had positive HBV DNA at the end of follow-up, developed HCC within 6 months during follow-up, or lost to follow-up were also excluded. All the patients received NAs treatment, including Lamivudine, Adefovir, Telbivudine, Entecavir or Tenofovir at admission, and routinely underwent clinical examination, laboratory tests and ultrasonography every 3 to 6 months. Demographic and clinical data, including age, sex, alanine transaminase (ALT), aspartate transaminase (AST), total bilirubin (TBil), gamma-glutamyl transpeptidase (GGT), albumin, international standard ratio (INR), HBV serologic markers, serum HBV DNA, blood cell count, and complications were collected at the first time of admission. The endpoints of the study were HCC development or death. The study was non-interventional and not harmful to the patients, and written consents were obtained from all the participants. The protocol was approved by the Ethics Committee of the Third People’s Hospital of Changzhou according to the Declaration of Helsinki, 2013 (No. CZSY2018-0601). ## Score systems Risk scores including CAMD (Hsu et al., 2018), PAGE-B (Papatheodoridis et al., 2016), mPAGE-B (Kim et al., 2018), aMAP (Fan et al., 2020), ALBI score (Casadei Gardini et al., 2019) were calculated as described before. ## Statistical analysis All data were analyzed using SPSS version 25.0 (Armonk, NY, USA). Continuous variables were presented as median (interquartile range, IQR), and were analyzed using the Mann-Whitney U tests. Categorical values were presented as frequencies, and were compared using the chi-square test. Correlation analysis was evaluated using the Spearman correlation test. Independent risk factors for HCC development were identified using univariate and multivariate logistic regression analysis. Accuracy of the scoring systems was compared according to the area under the receiver operating characteristic curve (AUROC), which was calculated using MedCalc version 15.2.2 software for Windows (Medcalc Software, Mariakerke, Belgium). The cutoff value was identified using MedCalc software, and then Kaplan-*Meier analysis* was performed using GraphPad Prism version 5.0 for Windows (GraphPad Software, San Diego, CA, USA). P value < 0.05 was considered statistically significant. ## Characteristics of patients Among the 307 patients, 78 patients were excluded, including 58 patients lost to follow-up, three patients developed HCC during 6 months of follow-up, 13 patients had positive HBV DNA at the last visit, one patient coinfected with hepatitis C virus, two patients suffered from lung cancer, and one patient suffered from kidney tumor. Data from 229 patients were analyzed, including 94 individuals with compensated LC and 135 individuals with DC (Fig. 1). Of the 135 patients with DC, 12 patients had hepatic encephalopathy, six patients had variceal bleeding, 121 patients had ascites, and one patient had renal failure. A total of 28 patients were stratified as late decompensated state based on the refractory ascites, recurrent encephalopathy, and renal failure. **Figure 1:** *Screening of patients with liver cirrhosis.LC, liver cirrhosis; DC, decompensated cirrhosis; HCC, hepatocellular carcinoma.* The median follow-up duration was 37 (28–66) months. A total of 9 ($9.57\%$) patients in the compensated LC group and 39 ($28.89\%$) patients in the DC group developed HCC. The incidence of HCC was higher in the DC group (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\cal X$\end{document}X2 = 12.478, $P \leq 0.01$). The characteristics of the 48 patients who developed HCC are shown in Table 1. Patients who developed HCC were older than those without HCC ($P \leq 0.01$). CAMD, PAGE-B, mPAGE-B and aMAP scores were higher, while the platelet count was lower in patients with HCC ($P \leq 0.01$). **Table 1** | Variables | Patients did not developed HCC (n = 181) | Patients developed HCC (n = 48) | Z or \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\cal X$\end{document}X2 | P value | | --- | --- | --- | --- | --- | | Age (years) | 50.0 (45.0–60.0) | 57.5 (51.5–64.0) | −3.850 | <0.01 | | Male, n (%) | 119 (65.8) | 34 (70.8) | 0.443 | 0.51 | | DC, n (%) | 96 (53.0) | 39 (81.3) | 12.478 | <0.01 | | Diabetes, n (%) | 52 (28.7) | 17 (35.4) | 0.806 | 0.37 | | ALT, U/L | 33.0 (21.0–65.0) | 34.0 (23.0–56.5) | −0.168 | 0.87 | | AST, U/L | 36.0 (25.0–54.0) | 42.5 (31.3–73.0) | −1.308 | 0.19 | | GGT, U/L | 50.0 (25.5–79.0) | 64.5 (26.5–107.8) | −1.578 | 0.11 | | TBil, µmol/L | 22.5 (15.4–34.7) | 20.0 (14.6–31.8) | 0.566 | 0.57 | | Albumin, g/L | 39.1 (32.8–43.3) | 38.0 (31.0–42.2) | 0.787 | 0.43 | | Creatinine, µmol/L | 77.7 (68.4–92.0) | 77.9 (66.6–92.6) | 0.464 | 0.64 | | PLT, E+09/L | 96.0 (56.5–141.0) | 70.5 (53.5–101.5) | 2.684 | <0.01 | | Serum sodium, mmol/L | 141.0 (139.4–142.2) | 141.0 (139.4–142.3) | −0.117 | 0.91 | | INR | 1.1 (1.1–1.3) | 1.2 (1.1–1.3) | −0.826 | 0.41 | | ALBI | −2.5 (−2.8 to −1.8) | −2.3 (−2.8 to −1.6) | −0.533 | 0.59 | | CAMD | 14.0 (13.0–17.0) | 16.0 (14.3–18.0) | −4.061 | <0.01 | | PAGE_B | 16.0 (14.0–19.5) | 19.0 (17.3–21.0) | −3.939 | <0.01 | | mPAGE_B | 12.0 (11.0–15.0) | 15.0 (13.3–17.0) | −4.412 | <0.01 | | aMAP | 60.0 (54.5–65.8) | 66.5 (62.6–69.0) | −4.399 | <0.01 | | Model (Age_ DC_PLT_TBil) | 0.1 (0.1–0.2) | 0.3 (0.2–0.4) | −5.538 | <0.01 | | MELD score | 10.0 (8.0–13.0) | 9.0 (8.0–12.8) | 0.244 | 0.81 | | Duration of follow-up, months | 38.0 (30.5–66.5) | 30.0 (19.0–60.0) | 2.189 | 0.03 | ## Independent risk factors for HCC development As shown in Table 2, the univariate logistic analysis showed that age, DC status and platelet were associated with HCC development. Multivariable analysis showed that age and DC status were independent risk factors for HCC development (both $P \leq 0.01$). Then a prediction models were developed: Model (Age_DC) ($y = 1$) = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}${{\rm exp}\,\left({ -\ 3.658\ +\ 0.049\ *\ {\rm Age}\ -\ 1.107\ *\ {\rm DC}\left(1 \right)} \right)} \over {1\ +\ {\rm exp}\,\left({ -\ 3.658 \,+\, 0.049\ *\ {\rm Age}\ -\ 1.107\ *\ {\rm DC}\left(1 \right)} \right)}$\end{document}exp(− 3.658 + 0.049 ∗ Age − 1.107 ∗ DC[1])1 + exp(− 3.658+0.049 ∗ Age − 1.107 ∗ DC[1]). Considering that TBil and PLT were important indexes of several non-invasive models, another model incorporating TBil and PLT into Model (Age_DC), Model (Age_DC_PLT_TBil) ($y = 1$) = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}${{\rm exp}\,\left({ -\ 2.406\ +\ 0.051\ *\ {\rm Age}\ -\ 1.090\ *\ {\rm DC}\left(1 \right)\ -\ 0.022\ *\ {\rm TBil}\ -\ 0.008\ *\ {\rm PLT}} \right)} \over {1\ +\ {\rm exp}\,\left({ -\ 2.406\ +\ 0.051\ *\ {\rm Age}\ -\ 1.090{\rm DC}\left(1 \right)\ -\ 0.022{\rm TBil}\ -\ 0.008\ *\ {\rm PLT}} \right)}$\end{document}exp(− 2.406 + 0.051 ∗ Age − 1.090 ∗ DC[1] − 0.022 ∗ TBil − 0.008 ∗ PLT)1 + exp(− 2.406 + 0.051 ∗ Age − 1.090DC[1] − 0.022TBil − 0.008 ∗ PLT), was developed. The AUROC of the two models were compared, and Model (Age_DC_PLT_TBil) was more potent than Model (Age_DC) (AUROC: 0.760 and 0.718). For patients with compensated LC, AUROC of Model (Age_DC_PLT_TBil) was larger than Model (Age_DC) (AUROC: 0.778 and 0.684). For patients with DC, AUROC of Model (Age_DC_PLT_TBil) was also larger than Model (Age_DC) (AUROC: 0.691 and 0.629). Patients who developed HCC had significantly higher Model (Age_DC_PLT_TBil) scores than those patients who did not develop HCC ($Z = 5.538$, $P \leq 0.01$). **Table 2** | Baseline variables | Univariate | Univariate.1 | Univariate.2 | Multivariate | Multivariate.1 | Multivariate.2 | | --- | --- | --- | --- | --- | --- | --- | | Baseline variables | Odds ratio | 95% CI | P | Odds ratio | 95% CI | P | | Age | 1.062 | [1.027–1.098] | <0.01 | 1.050 | [1.015–1.086] | <0.01 | | Male | 1.265 | [0.632–2.533] | 0.51 | | | | | DC | 0.261 | [0.119–0.569] | <0.01 | 0.331 | [0.148–0.739] | <0.01 | | ALT | 0.998 | [0.994–1.001] | 0.24 | | | | | AST | 0.999 | [0.995–1.002] | 0.44 | | | | | Albumin | 0.979 | [0.935–1.026] | 0.38 | | | | | TBil | 0.998 | [0.980–1.017] | 0.87 | | | | | INR | 1.540 | [0.408–5.812] | 0.52 | | | | | Creatinine | 0.997 | [0.985–1.009] | 0.61 | | | | | Serum sodium | 0.994 | [0.874–1.130] | 0.93 | | | | | PLT | 0.988 | [0.979–0.997] | 0.01 | | | | | WBC | 0.922 | [0.787–1.081] | 0.32 | | | | | Diabetes | 0.735 | [0.375–1.441] | 0.37 | | | | ## Comparison of ALBI, aMAP, CAMD, PAGE-B, mPAGE-B and Model (Age_DC_PLT_TBil) As shown in Fig. 2, the AUROC of ALBI, aMAP, CAMD, PAGE-B, mPAGE-B and Model (Age_DC_PLT_TBil) scores were 0.512, 0.667, 0.638, 0.663, 0.679 and 0.760, respectively. There was no significant difference in AUROC between CAMD, aMAP, PAGE-B and mPAGE-B (all $P \leq 0.05$). Moreover, AUROC of Model (Age_DC_PLT_TBil) was larger than the other five models (all $P \leq 0.05$). Moreover, the AUROC of Child-Pugh score in predicting HCC development was just 0.529, which was comparable with ALBI. **Figure 2:** *Comparison of AUROC between ALBI, CAMD, aMAP, PAGE-B and mPAGE-B and Model (Age_DC_PLT_TBil).(A) Total population; (B) compensated liver cirrhosis; (C) decompensated cirrhosis.* For patients with compensated LC, AUROC of ALBI, aMAP, CAMD, PAGE-B, mPAGE-B and Model (Age_DC_PLT_TBil) were 0.548, 0.641, 0.662, 0.665, 0.637 and 0.778, respectively. Model (Age_DC_PLT_TBil) had larger AUROC than mPAGE-B ($P \leq 0.01$) and ALBI ($$P \leq 0.02$$). For patients with DC, AUROC of ALBI, aMAP, CAMD, PAGE-B, mPAGE-B and Model (Age_DC_PLT_TBil) were 0.574, 0.623, 0.607, 0.642, 0.638 and 0.691, respectively. AUROC of Model (Age_DC_PLT_TBil) seemed to be larger than those of ALBI ($$P \leq 0.05$$) and CAMD ($$P \leq 0.07$$). ## Correlation between ALBI, aMAP, CAMD, PAGE-B, mPAGE-B and Model (Age_DC_PLT_TBil) Model (Age_DC_PLT_TBil) positively correlated with CAMD, PAGE_B, mPAGE_B and aMAP (all $P \leq 0.01$), but not with ALBI ($$P \leq 0.99$$ and 0.10) in patients with compensated LC or DC. Moreover, for patients with compensated LC or DC, ALBI correlated with mPAGE_B and aMAP (all $P \leq 0.05$), but not with other three models (Fig. 3). **Figure 3:** *Correlation analysis between ALBI, CAMD, aMAP, PAGE-B and mPAGE-B and Model (Age_DC_PLT_TBil).(A) Total population; (B) compensated liver cirrhosis; (C) decompensated cirrhosis.* ## Risk stratification for cumulative incidence of HCC With an optimal cut-off value of 0.236, Model (Age_DC_PLT_TBil) achieved $70.83\%$ sensitivity, $76.24\%$ specificity. Then the patients were divided into two groups: the low-risk group (Model (Age_DC_PLT_TBil) <0.236) and high-risk group (Model (Age_DC_PLT_TBil) ≥0.236). Patients with DC in high-risk group had a poor prognosis ($P \leq 0.01$) (Fig. 4). **Figure 4:** *Kaplan-Meier analysis of Model (Age_DC_PLT_TBil).(A) Total population; (B) compensated liver cirrhosis; (C) decompensated cirrhosis.* ## Discussion Data from the present study showed that approximately one third ($28.89\%$) patients in the DC group developed HCC during a median duration of 37-month follow-up. Although sustained virological response was achieved by NAs treatment, patients with DC remain at high risk of HCC (Zhu et al., 2020). Therefore, HCC surveillance should be paid more attention in clinical practice. Moreover, Child-Pugh score which is a common non-invasive scoring system in evaluating the severity and outcomes of LC, has a poor predictive accuracy according to the AUROC (0.529). To date, for patients with DC, these is no available models for HCC prediction (Lee et al., 2022; Yu et al., 2022). Several non-invasive models have been investigated in patients with CHB or compensated LC (Hsu et al., 2018; Papatheodoridis et al., 2016; Kim et al., 2018; Fan et al., 2020; Kirino et al., 2020; Lee et al., 2019), or chronic hepatitis C (Casadei Gardini et al., 2019). The present study focuses on the predictive accuracy of these models in patients with DC. Comparison of AUROC shows that ALBI was not suitable for HCC prediction in HBV-related LC. In addition, aMAP, CAMD, PAGE-B and mPAGE-B did not seem to be potent yet (all AUROC <0.7). Based on a logistic regression analysis, Model (Age_DC_PLT_TBil) consisting of age, DC stage, platelets and TBil, was developed and it showed superiority in stratifying patients at high risk. Compared with aMAP (Gui et al., 2021) or transient elastography (Lee et al., 2021), this new model with potential clinical utility, is more easily to calculate, and it does not increase the economic burden to the patients. For patients at high-risk group, surveillance strategy including alpha-fetoprotein and imaging examinations are recommended to be performed routinely every 3 to 6 months (Sohn et al., 2022). As shown in Fig. 4B, Model (Age_DC_PLT_TBil) shows non-optimal stratification accuracy for compensated LC. The main reason is that patients with Model (Age_DC_PLT_TBil) ≥0.236 are very few in the compensated LC group (just one patient), and the incidence of HCC is relatively lower. Normal PLT counts and TBil in patients with compensated LC and long-term antiviral therapy, result in lower Model (Age_DC_PLT_TBil) score. Moreover, baseline GGT seemed to be higher in patients who developed HCC, but the difference was not significant, then univariate logistic analysis showed that GGT was not related to HCC development. Recently, Huang et al. [ 2022] reported that GGT 6 months after initiating NAs strongly predicted HCC development in CHB patients, but not in LC patients. It would be interesting to analyze on-treatment GGT in patients with DC. Accumulated evidences are needed to validate its accuracy in studies with large sample-size. It should be noted that HBV genotype, viral loading at baseline, mutation of HBV, as well as the kind of NAs, may influence the accuracy of Model (Age_DC_PLT_TBil). Recent data regarding the HBV genotype C2 in Korea, revealed that risk of HCC steadily persisted despite long-term antiviral treatment (Kim et al., 2020). It is still controversial that *Tenofovir is* superior to Entecavir in reducing the incidence of HCC (Kim et al., 2019; Ha et al., 2020; Chon et al., 2021; Lee et al., 2020). Moreover, HCC inhibition by different NAs remains largely unknown in patients with DC. Regardless of the kind of NAs, persistent HBV DNA positivity is speculated to be more harmful in patients undergoing long-term NAs treatment, especially for DC. Positive HBV DNA accompanying with abnormal TBil and decreased PLT counts due to hypersplenism, may present as high Model (Age_DC_PLT_TBil) score in patients with DC. There are several limitations to the present study. First, in the present single-center study, the sample size is small and the follow-up duration is short, so multi-center study with large sample size is of importance for future research. Second, obesity and cirrhosis-associated immune dysfunction in patients with end-stage liver diseases may affect HCC development (Shin, Jun & Yi, 2022), and it should be interesting to evaluate the several models in patients with HBV-related LC and nonalcohol fatty liver diseases. Third, the family history of HCC, which has been confirmed to be a susceptible risk factor for HCC, was not collected in the present study. In conclusion, it lacks of non-invasive risk scores for HCC development in HBV-related DC, and a new model consisting of Age, DC stage, PLT, TBil may be an alternative. ## References 1. Casadei Gardini A, Foschi FG, Conti F, Petracci E, Vukotic R, Marisi G, Buonfiglioli F, Vitale G, Ravaioli F, Gitto S, Verucchi G, Lenzi M, Bolondi L, Mazzella G, Brillanti S, Andreone P. **Immune inflammation indicators and ALBI score to predict liver cancer in HCV-patients treated with direct-acting antivirals**. *Digestive and Liver Disease* (2019) **51** 681-688. DOI: 10.1016/j.dld.2018.09.016 2. Chon HY, Ahn SH, Kim YJ, Yoon JH, Lee JH, Sinn DH, Kim SU. **Efficacy of entecavir, tenofovir disoproxil fumarate, and tenofovir alafenamide in treatment-naive hepatitis B patients**. *Hepatology International* (2021) **15** 1328-1336. DOI: 10.1007/s12072-021-10262-y 3. D’Amico G, Garcia-Tsao G, Pagliaro L. **Natural history and prognostic indicators of survival in cirrhosis: a systematic review of 118 studies**. *Journal of Hepatology* (2006) **44** 217-231. DOI: 10.1016/j.jhep.2005.10.013 4. D’Amico G, Morabito A, D’Amico M, Pasta L, Malizia G, Rebora P, Valsecchi MG. **New concepts on the clinical course and stratification of compensated and decompensated cirrhosis**. *Hepatology International* (2018) **12** 34-43. DOI: 10.1007/s12072-017-9808-z 5. Fan R, Papatheodoridis G, Sun J, Innes H, Toyoda H, Xie Q, Mo S, Sypsa V, Guha IN, Kumada T, Niu J, Dalekos G, Yasuda S, Barnes E, Lian J, Suri V, Idilman R, Barclay ST, Dou X, Berg T, Hayes PC, Flaherty JF, Zhou Y, Zhang Z, Buti M, Hutchinson SJ, Guo Y, Calleja JL, Lin L, Zhao L, Chen Y, Janssen HLA, Zhu C, Shi L, Tang X, Gaggar A, Wei L, Jia J, Irving WL, Johnson PJ, Lampertico P, Hou J. **aMAP risk score predicts hepatocellular carcinoma development in patients with chronic hepatitis**. *Journal of Hepatology* (2020) **73** 1368-1378. DOI: 10.1016/j.jhep.2020.07.025 6. Foerster F, Gairing SJ, Ilyas SI, Galle PR. **Emerging immunotherapy for hepatocellular carcinoma: a guide for hepatologists**. *Hepatology* (2022) **75** 1604-1626. DOI: 10.1002/hep.32447 7. Gui H, Huang Y, Zhao G, Chen L, Cai W, Wang H, Guo Q, Xie Q. **External validation of aMAP hepatocellular carcinoma risk score in patients with chronic hepatitis B-related cirrhosis receiving ETV or TDF therapy**. *Frontiers in Medicine* (2021) **8** 677920. DOI: 10.3389/fmed.2021.677920 8. Ha I, Chung JW, Jang ES, Jeong SH, Kim JW. **Comparison of the on-treatment risks for hepatocellular carcinoma between entecavir and tenofovir: a propensity score matching analysis**. *Journal of Gastroenterology and Hepatology* (2020) **35** 1774-1781. DOI: 10.1111/jgh.15031 9. **Guidelines for diagnosis and treatment of primary liver cancer in China (2019 edition)**. *Zhonghua Gan Zang Bing Za Zhi* (2020) **28** 112-128. DOI: 10.3760/cma.j.issn.1007-3418.2020.02.004 10. Hsu YC, Yip TCF, Ho HJ, Wong VWS, Huang YT, El-Serag HB, Lee TY, Wu MS, Lin JT, Wong GLH, Wu CY. **Development of a scoring system to predict hepatocellular carcinoma in Asians on antivirals for chronic hepatitis B**. *Journal of Hepatology* (2018) **69** 278-285. DOI: 10.1016/j.jhep.2018.02.032 11. Huang CF, Jang TY, Jun DW, Ahn SB, An J, Enomoto M, Takahashi H, Ogawa E, Yoon E, Jeong SW, Shim JJ, Jeong JY, Kim SE, Oh H, Kim HS, Cho YK, Kozuka R, Inoue K, Cheung KS, Mak LY, Huang JF, Dai CY, Yuen MF, Nguyen MH, Yu ML. **On-treatment gamma-glutamyl transferase predicts the development of hepatocellular carcinoma in chronic hepatitis B patients**. *Liver International* (2022) **42** 59-68. DOI: 10.1111/liv.15085 12. Jia JD, Hou JL, Wei L, Zhuang H. **Highlights of the guidelines of prevention and treatment for chronic hepatitis B (2019 version)**. *Zhonghua Gan Zang Bing Za Zhi* (2020) **28** 21-23. DOI: 10.3760/cma.j.issn.1007-3418.2020.01.006 13. Kim JH, Kim YD, Lee M, Jun BG, Kim TS, Suk KT, Kang SH, Kim MY, Cheon GJ, Kim DJ, Baik SK, Choi DH. **Modified PAGE-B score predicts the risk of hepatocellular carcinoma in Asians with chronic hepatitis B on antiviral therapy**. *Journal of Hepatology* (2018) **69** 1066-1073. DOI: 10.1016/j.jhep.2018.07.018 14. Kim SU, Seo YS, Lee HA, Kim MN, Lee YR, Lee HW, Park JY, Kim DY, Ahn SH, Han KH, Hwang SG, Rim KS, Um SH, Tak WY, Kweon YO, Kim BK, Park SY. **A multicenter study of entecavir vs. tenofovir on prognosis of treatment-naive chronic hepatitis B in South Korea**. *Journal of Hepatology* (2019) **71** 456-464. DOI: 10.1016/j.jhep.2019.03.028 15. Kim SU, Seo YS, Lee HA, Kim MN, Lee EJ, Shin HJ, Lee YR, Lee HW, Park JY, Kim DY, Ahn SH, Han KH, Um SH, Tak WY, Kweon YO, Kim BK, Park SY. **Hepatocellular carcinoma risk steadily persists over time despite long-term antiviral therapy for hepatitis B: a multicenter study**. *Cancer Epidemiology, Biomarkers & Prevention* (2020) **29** 832-837. DOI: 10.1158/1055-9965.EPI-19-0614 16. Kirino S, Tamaki N, Kaneko S, Kurosaki M, Inada K, Yamashita K, Osawa L, Hayakawa Y, Sekiguchi S, Watakabe K, Okada M, Wang W, Shimizu T, Higuchi M, Takaura K, Maeyashiki C, Yasui Y, Nakanishi H, Tsuchiya K, Itakura J, Takahashi Y, Izumi N. **Validation of hepatocellular carcinoma risk scores in Japanese chronic hepatitis B cohort receiving nucleot(s)ide analog**. *Journal of Gastroenterology and Hepatology* (2020) **35** 1595-1601. DOI: 10.1111/jgh.14990 17. Lee HW, Kim SU, Park JY, Kim DY, Ahn SH, K-Hyub H, Kim BK. **External validation of the modified PAGE-B score in Asian chronic hepatitis B patients receiving antiviral therapy**. *Liver International* (2019) **39** 1624-1630. DOI: 10.1111/liv.14129 18. Lee SW, Kwon JH, Lee HL, Yoo SH, Nam HC, Sung PS, Nam SW, Bae SH, Choi JY, Yoon SK, Han NI, Jang JW. **Comparison of tenofovir and entecavir on the risk of hepatocellular carcinoma and mortality in treatment-naive patients with chronic hepatitis B in Korea: a large-scale, propensity score analysis**. *Gut* (2020) **69** 1301-1308. DOI: 10.1136/gutjnl-2019-318947 19. Lee JS, Lee HW, Lim TS, Shin HJ, Lee HW, Kim SU, Park JY, Kim DY, Ahn SH, Kim BK. **Novel liver stiffness-based nomogram for predicting hepatocellular carcinoma risk in patients with chronic hepatitis B virus infection initiating antiviral therapy**. *Cancers* (2021) **13** 5892. DOI: 10.3390/cancers13235892 20. Lee JS, Lim TS, Lee HW, Kim SU, Park JY, Kim DY, Ahn SH, Lee HW, Lee JI, Kim JK, Min IK, Kim BK. **Suboptimal performance of hepatocellular carcinoma prediction models in patients with hepatitis B virus-related cirrhosis**. *Diagnostics* (2022) **13** 3. DOI: 10.3390/diagnostics13010003 21. Papatheodoridis G, Dalekos G, Sypsa V, Yurdaydin C, Buti M, Goulis J, Calleja JL, Chi H, Manolakopoulos S, Mangia G, Gatselis N, Keskin O, Savvidou S, de la Revilla J, Hansen BE, Vlachogiannakos I, Galanis K, Idilman R, Colombo M, Esteban R, Janssen HLA, Lampertico P. **PAGE-B predicts the risk of developing hepatocellular carcinoma in Caucasians with chronic hepatitis B on 5-year antiviral therapy**. *Journal of Hepatology* (2016) **64** 800-806. DOI: 10.1016/j.jhep.2015.11.035 22. Shin HS, Jun BG, Yi SW. **Impact of diabetes, obesity, and dyslipidemia on the risk of hepatocellular carcinoma in patients with chronic liver diseases**. *Clinical and Molecular Hepatology* (2022) **28** 773-789. DOI: 10.3350/cmh.2021.0383 23. Sohn W, Kang D, Kang M, Guallar E, Cho J, Paik YH. **Impact of nationwide hepatocellular carcinoma surveillance on the prognosis in patients with chronic liver disease**. *Clinical and Molecular Hepatology* (2022) **28** 851-863. DOI: 10.3350/cmh.2022.0037 24. Su GL, Altayar O, O’Shea R, Shah R, Estfan B, Wenzell C, Sultan S, Falck-Ytter Y. **AGA clinical practice guideline on systemic therapy for hepatocellular carcinoma**. *Gastroenterology* (2022) **162** 920-934. DOI: 10.1053/j.gastro.2021.12.276 25. Yang HI, Yuen MF, Chan HLY, Han KH, Chen PJ, Kim DY, Ahn SH, Chen CJ, Wong VWS, Seto WK. **Risk estimation for hepatocellular carcinoma in chronic hepatitis B (REACH-B): development and validation of a predictive score**. *The Lancet Oncology* (2011) **12** 568-574. DOI: 10.1016/S1470-2045(11)70077-8 26. Yu JH, Cho SG, Jin YJ, Lee JW. **The best predictive model for hepatocellular carcinoma in patients with chronic hepatitis B infection**. *Clinical and Molecular Hepatology* (2022) **28** 351-361. DOI: 10.3350/cmh.2021.0281 27. Zhu DM, Xie J, Ye CY, Qian MY, Xue Y. **Risk of hepatocellular carcinoma remains high in patients with HBV-related decompensated cirrhosis and long-term antiviral therapy**. *Canadian Journal of Gastroenterology and Hepatology* (2020) **2020** 8871024. DOI: 10.1155/2020/8871024
--- title: N6-methyladenosine reader YTHDF1 regulates the proliferation and migration of airway smooth muscle cells through m6A/cyclin D1 in asthma authors: - Juan Wang - Lei Wang - Xingfeng Tian - Lingping Luo journal: PeerJ year: 2023 pmcid: PMC10042154 doi: 10.7717/peerj.14951 license: CC BY 4.0 --- # N6-methyladenosine reader YTHDF1 regulates the proliferation and migration of airway smooth muscle cells through m6A/cyclin D1 in asthma ## Abstract Asthma is a chronic inflammatory respiratory disease, which is involved in multiple pathologic molecular mechanisms and presents a huge challenge to clinic nursing. Emerging evidence suggests that N6-methyladenosine (m6A) plays critical roles in respiratory system disease. Thus, present work tried to investigate the functions of m6A reader YTHDF 1 in asthma. The results indicated that YTHDF1 significantly upregulated in platelet-derived growth factor (PDGF) induced airway smooth muscle cells (ASMCs). Functionally, overexpression of YTHDF1 promoted the proliferation and migration of ASMCs, while YTHDF1 knockdown repressed the proliferation and migration. Mechanistically, there was a m6A modification site on cyclin D1 RNA (CCND1 genome) and YTHDF1 combined with cyclin D1 mRNA, thereby enhancing its mRNA stability via m6A-dependent manner. Collectively, these findings reveal a novel axis of YTHDF1/m6A/cyclin D1 in asthma’s airway remodeling, which may provide novel therapeutic strategy for asthma. ## Introduction Asthma is well known as a common respiratory disorder, hyper-responsiveness accompanied by airway remodeling or chronic bronchial airway inflammatory (Ferry, De Castro & Bragg, 2020; Kutlu & Unal, 2019). In the pathophysiological process of asthma, the airway structures are involved, including airway cells, cell components and inflammatory substances (Lee, Lee & Hong, 2020; MacDonald & Barrett, 2019). The pathogenesis of asthma is predominantly attributed by immune cells’ excessive immune response and substantial amounts of inflammatory cytokines production, e.g., interleukin $\frac{4}{5}$/13. These cytokines mediate mucus and immunoglobulin E overproduction, airway hyperresponsiveness and eosinophilic infiltration (Sánchez et al., 2019). Thus, the most effective way to deal with asthma is to discover its precise pathogenesis. N6-methyladenosine (m6A) methylation is a post-transcriptional RNA modification epigenetically occurred on mRNAs, which regulates gene expression and affects the RNA fate (Liu et al., 2021; Sweaad et al., 2021; Tuncel & Kalkan, 2019). Presently, m6A methylation gains more and more attention about its function and mechanism. In the asthma pathophysiological process, there is only primary probation and initiatory findings (Wang et al., 2021; Wu et al., 2021). For example, Teng et al. [ 2021] found that dysregulated or hypermethylated m6A peaks in 329 mRNAs and 150 hypomethylated m6A peaks in 143 mRNAs in asthmatic mice. In addition, Dai et al. [ 2021] found that 5 candidate m6A regulators (FMR1, KIAA1429, WTAP, YTHDF2, ZC3HAV1) are in close contact with the risk of childhood asthma. Therefore, these literatures inspire us that m6A may participate in the asthma pathology. In clinical nursing, the particularity of asthma brings great challenge to nursing work. Some environmental factors can trigger or stimulate asthma, such as pollen, excessively cold climate, strenuous exercise or pets (Sonney et al., 2022). Asthma can cause recurrent episodes of wheezing, shortness of breath, chest tightness, and/or coughing (Yamada et al., 2022). In addition, these emergencies often occur at night/morning. Therefore, this special situation requires us to pay close attention to the patient’s changes during the course of nursing. Here, our research found that, in the cellular asthma model induced by PDGF-BB, the m6A modification significantly varied and the m6A regulator key-enzymes also altered. To investigate the potential roel of m6A in asthma, we focused on a critical m6A reader YTHDF1. Results indicated that YTHDF up-regulated in the PDGF-BB induced ASM cells. Functionally, YTHDF1 posituvely promoted the proliferation and migration of ASM cells. Interestingly, an important element cyclin D1 (CCND1) acted as the downstream target of YTHDF1 via m6A-depedent manner. Mechanistically, YTHDF1 significantly combined with cyclin D1 mRNA, thereby enhancing its mRNA stability through m6A-depedent pattern, which may provide novel therapeutic strategy for asthma. ## Asthmatic cellular model As previously described (Ba et al., 2018), the primary cultured human ASM cells were obtained from 2nd–4th generation mainstem bronchi of patients undergoing lung resection surgery in accordance with procedures. Written informed consent was obtained from every human participant. The assay had been approved by the Ethics Committee of Shanxi Medical University (No. SXMU201905047). In brief, the mainstem bronchi segments were cut into pieces and ASMCs were isolated from it. After digestion, ASMCs were placed in DMEM medium containing $10\%$ fetal bovine serum (Gibco, NY, USA) in humidified incubator at $5\%$ CO2 37 °C with. Cells from passages 4–7 were used for following assays, ASMCs were treated with PDGF-BB (25 ng/ml) to mimic the asthma. ## Transfection For the transfection of YTHDF1, ASMCs were transfected with sequences following the manufacturer’s instructions. For silencing of YTHDF1, the shRNA sequences of YTHDF1 were synthesized by OBiOc (Shanghai, China) and the vectors containing shRNAs were inserted into PLKO.1. The transfection of plasmids was performed using the Lipofectamine 3000 kit (Invitrogen) according to the manufacturer’s instructions. For the overexpression of YTHDF1, the full-length cDNA sequences of human YTHDF1 (gene ID: 54915) were cloned into a pLentiEF1a-EGFP-Puro-CMV-MCS-3Flag lentivirus vector. The transfection efficiency was evaluated with qRT-PCR or western blot. ## Reverse transcription quantitative polymerase chain reaction (RT- qPCR) Total RNA was extracted according to the instruction of HiScript II 1st Strand cDNA Synthesis Kit (Vazyme, Nanjing, China). Then, cDNA was synthesized using an PrimeScript RT reagent kit (Takara, Dalian, China). Real-time PCR was performed on the 7900 Real-time PCR System using Taqman RNA assay kit (Thermo Fisher Scientific, Rockford, IL, USA). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) acted as a control. The primer sequences were listed in Table S1. The results of transcript levels were analyzed by the 2−ΔΔCt method. ## Western blot analysis Total protein in ASMC cells was extracted using radio immunoprecipitation assay (RIPA) lysis buffer with phenylmethanesulfonyl fluoride (PMSF) (Solarbio, Beijing, No. R0010). After incubation on ice and then concentration, the protein concentration was measured by bicinchoninic acid (BCA) kit and adjusted by deionized water. Protein samples were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis ($10\%$ SDS-PAGE) and transferred to polyvinylidene fluoride membranes (PVDF) (Millipore, Billerica, MA, USA). The transferred PVDF membranes (No. ISEQ00010; Millipore, Billerica, MA, USA) was added with Tris-buffered saline tween (TBST) containing $5\%$ dried dskimmed milk. Then, PVDF membranes were incubated with the primary antibodies (anti-Cyclin D1, ab16663; Abcam; anti-YTHDF1, 1:1000, ab220162; Abcam). Beta-actin (1:1,000; Abcam) was used as an internal control. After washing with phosphate buffer five times containing Tween-20 (PBST), PVDF membranes were incubated for 1 h at room temperature. Finally, pierce ECL western blotting substrate was employed to develop the protein bands and quantification was conducted by Image Lab software (Bio-Rad, Hercules, CA, USA). ## Proliferation assays and cycle analysis The proliferation of ASMCs was detected using CCK-8 assays using Cell counting kit-8 (8 µl of CCK-8; Dojindo, Kumamoto, Japan) with 100 µl serum free medium and incubated for 90 min. In brief, the transfected ASMCs (5 ×103 cells/well) was inoculated at into 96-well plates at 24, 48, and 72 h of culture at 37 °C overnight. Eventually, the absorbance at 450 nm was detected by a microplate reader (Thermo Fisher Scientific, Waltham, MA, USA). For the cycle analysis, Coulter EPICS XL flow cytometer (Beckman Coulter, Inc., Fullerton, CA, USA) was performed by flow cytometry on flow cytometry with Modifit software (BD Biosciences). ## Transwell migration assays and wound healing assay The migration of ASMCs was detected by transwell assays. In brief, the transfected ASMCs (1 ×105 cells/well) were suspended in serum-free medium upper transwell chamber (pore size: 8 µm; Corning, Inc., Corning, NY, USA). In the bottom chamber, medium was supplemented with 600 µl of $10\%$ FBS. After being incubated for 24 h at 37 °C, the cells in the top compartment were wiped off by cotton swabs and the migrated cells were fixed for 20 min with $4\%$ paraformaldehyde and stained for 10 min with $0.1\%$ crystal violet staining solution (Sigma-Aldrich, Louis, MO, USA). Lastly, images were taken under the optical microscope (Olympus, Tokyo, Japan). For the wound healing assay, the monolayer was manually scraped by sterile pipette tip. After being 24 h incubation at 37 °C, the Images of wound closure were evaluated by inverted microscope (Olympus, Tokyo, Japan). The migration rate was calculated by the formula: migration rate = migration distance/original distance. ## RNA immunoprecipitation (RIP)-PCR The interaction within RNA binding proteins and mRNA was identified using RIP-PCR. In brief, the RIP experiment was carried out by EZ-Magna RIP Kit (Millipore) according to the manufacturer’s protocol. ASMCs were lysed in complete RIP lysis buffer, and the cell extract was incubated with protein A/G agarose beads conjugated by anti-YTHDF1 (ab220162, 1:30; Abcam) or control IgG (ab172730; Abcam) for 2 h at 4 °C. After being washed, beads were incubated with Proteinase K to remove protein in complex. Lastly, the purified RNAs were subjected to qRT-PCR analysis. ## RNA stability assay To detect the cyclin D1 mRNA stability, the ASMCs were treated with actinomycin (Act D, 2 µg/mL) treatment for 0, 3 and 6 h. The relative remaining RNA level was detected by qRT-PCR and the half-life of cyclin D1 mRNA was examined by transcript levels at indicated time points relative to those before Act D treatment. ## Luciferase reporter assay The wild-type or mutant Cyclin D1 3′-UTR was synthesized and inserted into pmirGLO reporter vector (Promega, Madison, WI, USA), and then co-transfected with WT- Cyclin D1 or Mut- Cyclin D1 pcDNA 3.0 expressing plasmid into cells using Lipofectamine 3000. Cells were transfected with pGL3-Luc (1 µg) as a control for transfection efficiency (Promega), according to the manufacturer’s instructions. The luciferase activity was tested using a luciferase reporter commercial kit (Promega, Madison, WI, USA). ## Statistical analysis Experiments were repeated three times and data was presented as means ± standard deviation (SD) in this study. Variables between was compared by student’s t-test and Statistical analyses were performed using the SPSS 20.0 software (SPSS, Inc., Chicago, IL, USA) and GraphPad Prism 8.0 software (GraphPad, San Diego, CA, USA). A two-sided p-value of less than 0.05 was considered statistically significant. ## YTHDF1 was up-regulated in the PDGF-induced ASMCs To mimic the cellular asthma model, PDGF-induced ASMCs and blank control cells were constructed. Firstly, we found that YTHDF1 mRNA was up-regulated in the PDGF-induced ASMCs (Fig. 1A). Besides, the YTHDF1 protein level also up-regulated in the PDGF-induced ASMCs (Fig. 1B). Moreover, we found cyclin D1 (CCND1), an essential cell cycle control gene closely correlated to the development of asthma (Thun et al., 2013), was up-regulated in the PDGF-induced ASMCs (Fig. 1C). Besides, the cyclin D1 protein level also up-regulated in the PDGF-induced ASMCs (Fig. 1D). Overall, these findings revealed that YTHDF1 was up-regulated in the PDGF-induced ASMCs. **Figure 1:** *YTHDF1 was up-regulated in the PDGF-induced ASMCs.(A) RT-qPCR analysis was performed to detect the YTHDF1 mRNA level in the PDGF-induced ASMCs and control group. (B) Western blot analysis was carried out to measure the YTHDF1 protein level in the PDGF-induced ASMCs. (C) The cyclin D1 mRNA level was detected using RT-qPCR analysis in the PDGF-induced ASMCs. (D) The cyclin D1 protein level was detected using western blot analysis in PDGF-induced ASMCs. **p < 0.01.* ## YTHDF1 positively regulated the proliferation and migration of ASMCs The bio-functional roles of YTHDF1 were explored in the PDGF-induced ASMCs with YTHDF1 knockdown and overexpression. The transfection efficiency was detected using RT-PCR (Fig. 2A) and western blot (Fig. 2B). Proliferation analysis by CCK-8 assay unveiled that YTHDF1 silencing repressed the proliferative ability of ASMCs, while the YTHDF1 enforced overexpression up-regulated the proliferative ability (Fig. 2C). Furthermore, the migrative ability of ASMCs was determined by transwell assay, and results illustrated that YTHDF1 silencing repressed the migrative ability of ASMCs, while the YTHDF1 enforced overexpression up-regulated the migrative ability (Figs. 2D, 2E). Overall, these findings revealed that YTHDF1 positively regulated the proliferation and migration of ASMCs. **Figure 2:** *YTHDF1 positively regulated the proliferation and migration of ASMCs.(A) The transfection efficiency for YTHDF1 was detected using RT-PCR and (B) western blot. Results reflected the YTHDF1 mRNA and YTHDF1 protein after YTHDF1 silencing (sh-YTHDF1-1#/2#, sh-NC) and YTHDF1 enforced overexpression (YTHDF1, vector). (C) Proliferative ability of ASMCs was detected by CCK-8 assay after YTHDF1 silencing (sh-YTHDF1-1#/2#, sh-NC) and YTHDF1 enforced overexpression (YTHDF1, vector). (D, E) The migrative ability of ASMCs was determined by transwell assay. *p < 0.05; **p < 0.01.* ## YTHDF1 positively facilitated the cell cycle progression and migration To detect the role of YTHDF1 on PDGF-induced ASMCs, flow cytometry cell cycle analysis was performed. The results showed that YTHDF1 knockdown induced the cycle arrest at G1/S phase, while YTHDF1 overexpression promoted the cycle progression of ASMCs (Fig. 3A). Furthermore, the migrative ability of ASMCs was determined by wound healing assay, and results illustrated that YTHDF1 knockdown repressed the migrative ability of ASMCs, while the YTHDF1 overexpression up-regulated the migrative ability (Fig. 3B). Therefore, these data showed that YTHDF1 positively facilitated the cell cycle progression and migration. **Figure 3:** *YTHDF1 positively facilitated the cell cycle progression and migration.(A) Flow cytometry cell cycle analysis was performed in PDGF-induced ASMCs with YTHDF1 knockdown (sh-YTHDF1) or control (sh-NC). (B) Wound healing assay was performed to determine the the migrative ability of ASMCs with YTHDF1 knockdown (sh-YTHDF1) or control (sh-NC). *p < 0.05; **p < 0.01.* ## Cyclin D1 acted as the target of YTHDF1 To discovery the potential downstream target of YTHDF1 in asthma, we utilized the predictive tool (SRAMP, http://www.cuilab.cn/sramp) to analyze the m6A site in these targets (Fig. 4A). Moreover, we found that the exact site of m6A modification (ATGGAC) on the Cyclin D1 gene (Fig. 4B). The m6A motif on the Cyclin D1 mRNA was predicted (https://rna.sysu.edu.cn/rmbase/) (Fig. 4C). The molecular interaction within Cyclin D1 and YTHDF1 was determined by RIP-PCR, and results indicated that YTHDF1 closely combined with Cyclin D1 (Fig. 4D). Moreover, further RIP-PCR analysis found that YTHDF1 silencing repressed the combination within Cyclin D1 and YTHDF1, and YTHDF1 overexpression enhanced the combination (Fig. 4E). Taken together, these findings revealed that Cyclin D1 acted as the target of YTHDF1. **Figure 4:** *Cyclin D1 acted as the target of YTHDF1.(A) The predictive tool (SRAMP, http://www.cuilab.cn/sramp) was utilized to analyze the m6A site in these potential downstream targets of YTHDF1 in asthma. (B) The exact site of m6A modification (ATGGAC) on the Cyclin D1 gene. (C) The m6A motif on the Cyclin D1 mRNA was predicted (https://rna.sysu.edu.cn/rmbase/). (D) RIP-PCR was performed to determine the molecular interaction within Cyclin D1 and YTHDF1 in PDGF-induced ASMCs (p = 0.004). (E) RIP-PCR was performed in ASMCs transfected with YTHDF1 silencing (sh-YTHDF1-1#/2#, sh-NC) and YTHDF1 enforced overexpression (YTHDF1, vector) (p = 0.038). *p < 0.05; **p < 0.01.* ## YTHDF1 enhanced the RNA stability of cyclin D1 mRNA via m6A-dependent manner Previous researches had revealed that YTHDF1 could recognize the m6A site on mRNA and then enhance its stability (Chen et al., 2021; Xu et al., 2021). In our study, we found that YTHDF1 might also target cyclin D1 mRNA to enhance its stability. Firstly, we detected the cyclin D1 mRNA level in ASMCs with YTHDF1 silencing or overexpression, and results showed that YTHDF1 silencing or overexpression didn’t significantly regulate the cyclin D1 mRNA (Figs. 5A, 5B). RNA stability analysis revealed that YTHDF1 silencing reduced the half life time (t$\frac{1}{2}$) of cyclin D1 mRNA upon Act D treatment, while YTHDF1 overexpression up-regulated the half life time (t$\frac{1}{2}$) of cyclin D1 mRNA (Figs. 5C, 5D). Luciferase reporter assay using wild-type (WT) cyclin D1 3′UTR or mutant (Mut) was performed and results indicated that YTHDF1 accelerated the luciferase activity within cyclin D1 wild type (Figs. 5E, 5F). In conclusion, we found that YTHDF1 enhanced the RNA stability of cyclin D1 mRNA via m6A-dependent manner (Fig. 6). **Figure 5:** *YTHDF1 enhanced the RNA stability of cyclin D1 mRNA.(A, B) RT-PCR analysis was performed to detect the cyclin D1 mRNA level in ASMCs with YTHDF1 silencing or overexpression. (C, D) RNA stability analysis was performed to revealed the half life time (t1/2) of cyclin D1 mRNA in ASMCs upon Act D treatment. (E, F) cyclin D1 mRNA 3′ UTR containing m6A modification site was cloned into luciferase reporter vectors, including mutation (Mut) of m6A consensus sequence and mutant by replacing adenosine with cytosine. The luciferase activity within cyclin D1 wild type sequence and YTHDF1 overexpression or control was detected. Relative luciferase activity was computed by the ratio of Firefly and Renilla luciferase values.* **Figure 6:** *YTHDF1 enhanced the RNA stability of cyclin D1 mRNA to positively regulate asthma.* ## Discussion Currently, the pathophysiology of asthma is complex, including airway remodeling and inflammatory cells invasion and so on. Airway remodeling refers to a series of structural changes in airway structure in patients with asthma, including epithelial injury, increased basement membrane thickness, airway smooth muscle thickening, goblet cell metaplasia, and airway vascular and lymphatic proliferation (Xu et al., 2022). This study unveiled a novel finding that m6A reader YTHDF1 play acritical role in asthma airway remodeling, involving ASMCs proliferation and migration abilities. N6-methyladenosine (m6A) is the most abundant modification in mRNA, which is regulated by m6A methyltransferases, demethylases and readers. In the respiratory diseases, there are more and more literature reveal the important functions of m6A via variety of evidence (Xu et al., 2022). For instance, m6A writer METTL3 is up-regulated in PM2.5 exposured mice lung injury and METTL3 up-regulated the m6A modification of Interleukin 24 (IL24) through via METTL3/YTHDF1-coupled epitranscriptomal regulation (He et al., 2022). In lung ischemia/reperfusion injury, the m6A reader YTHDF3 or IGF2BP2 knockdown attenuates the hypoxia/reoxygenation-mediated inhibitory effects in BEAS-2B cells, as well as the hypoxia/reoxygenation-induced cell apoptosis (Xiao et al., 2022). Thus, these findings show the important function of m6A modification in respiratory diseases. Here, present research indicated that a novel m6A reader YTHDF1 also significantly up-regulated in the asthma cellular model. In the PDGF-induced ASMCs, we found that m6A reader YTHDF1 up-regulated and the functional assays suggested that YTHDF1 overexpression promoted the proliferation and migration of ASMCs. Thus, our assays’ data revealed the potential function of m6A reader YTHDF1 in asthma. Moreover, we found that a important element cyclin D1 (CCND1) up-regulated in the asthma cellular model (PDGF-induced ASMCs). Mechanistically, there were remarkable m6A modified site on cyclin D1 mRNA. Then, RIP-PCR assays was performed and results indicated that YTHDF1 significantly combine with cyclin D1 mRNA, thereby enhancing its mRNA stability through m6A-depedent manner. Collectively, these findings reveal a YTHDF1/cyclin D1 axis in asthma. As regarding the role of cyclin D1 (CCND1), an essential cell cycle control gene, convictive literature has revealed that CCND1 is closely correlated to the development of asthma (Thun et al., 2013). Besides, Li et al. [ 2021] reported that cell cycle regulation may play a role in asthma initiation and development, and CCND1 rs9344 genotype serves as an early detection marker for asthma. Overall, we could conclude that cyclin D1 (CCND1) significantly participate in the asthma. Asthma is a chronic inflammation; however, this inflammation is not primarily caused by bacterial infections. Asthma attacks are mostly related to exposure to allergens, cold air, physical and chemical stimulation, emotional changes, respiratory tract infection and exercise (Zhou et al., 2021). Thus, the complex pathogenesis put forward higher request to clinical nursing. For the clinical nursing of asthma, there are many situations that require protection, such as the allergies. To eliminate allergens, we need to clean the house dust mite thoroughly. Asthma can cause recurrent episodes of wheezing, shortness of breath, chest tightness, and/or coughing. In addition, these emergencies often occur at night/morning. Therefore, this special situation requires us to pay close attention to the patient’s changes during the course of nursing. In conclusion, our research utilized the PDGF-induced ASMCs to investigate the function and mechanism of YTHDF1 in asthma. These findings revealed the regulation of YTHDF1 on ASMCs’ proliferation and migration. Mechanistically, YTHDF1 significantly combined with cyclin D1 mRNA, thereby enhancing its mRNA stability through m6A-depedent manner (Fig. 6). Overall, this study may provide new insight for asthma. ## References 1. Ba M, Rawat S, Lao R, Grous M, Salmon M, Halayko AJ, Gerthoffer WT, Singer CA. **Differential regulation of cytokine and chemokine expression by MK2 and MK3 in airway smooth muscle cells**. *Pulmonary Pharmacology & Therapeutics* (2018) **53** 12-19. DOI: 10.1016/j.pupt.2018.09.004 2. Chen Z, Zhong X, Xia M, Zhong J. **The roles and mechanisms of the m6A reader protein YTHDF1 in tumor biology and human diseases**. *Molecular Therapy - Nucleic Acids* (2021) **26** 1270-1279. DOI: 10.1016/j.omtn.2021.10.023 3. Dai B, Sun F, Cai X, Li C, Liu H, Shang Y. **Significance of RNA N6-methyladenosine regulators in the diagnosis and subtype classification of childhood asthma using the gene expression omnibus database**. *Frontiers in Genetics* (2021) **12**. DOI: 10.3389/fgene.2021.634162 4. Ferry B, De Castro A, Bragg S. **Common prescription medications used in athletes**. *Primary Care: Clinics in Office Practice* (2020) **47** 49-64. DOI: 10.1016/j.pop.2019.10.003 5. He X, Zhang L, Liu S, Wang J, Liu Y, Xiong A, Jiang M, Luo L, Ying X, Li G. **Methyltransferase-like 3 leads to lung injury by up-regulation of interleukin 24 through N6-methyladenosine-dependent mRNA stability and translation efficiency in mice exposed to fine particulate matter 2.5**. *Environmental Pollution* (2022) **308**. DOI: 10.1016/j.envpol.2022.119607 6. Kutlu A, Unal D. **Mammalian meat allergy accompanied by venom allergy: a review of 12 cases**. *Iranian Journal of Allergy, Asthma and Immunology* (2019) **18** 584-588. DOI: 10.18502/ijaai.v18i5.1928 7. Lee E, Lee SY, Hong SJ. **The past, present, and future of humidifier disinfectant-associated interstitial lung diseases in children**. *Clinical and Experimental Pediatrics* (2020) **63** 251-258. DOI: 10.3345/cep.2019.01326 8. Li CH, Chiu KL, Hsia TC, Shen TC, Chen LH, Yu CC, Mong MC, Chang WS, Tsai CW, Bau DT. **Significant association of cyclin D1 promoter genotypes with asthma susceptibility in Taiwan**. *In Vivo* (2021) **35** 2041-2046. DOI: 10.21873/invivo.12473 9. Liu K, Gao Y, Gan K, Wu Y, Xu B, Zhang L, Chen M. **Prognostic roles of N6-methyladenosine METTL3 in different cancers: a system review and meta-analysis**. *Cancer Control* (2021) **28**. DOI: 10.1177/1073274821997455 10. MacDonald ES, Barrett JG. **The potential of mesenchymal stem cells to treat systemic inflammation in horses**. *Frontiers in Veterinary Science* (2019) **6** 507. DOI: 10.3389/fvets.2019.00507 11. Sánchez J, Sánchez MR, Macías-Weinmann A, Barreto B, Ensina LF, Uriarte-Obando SA, Castro-Almarales RL, Adorni R, Lázaro M, Callero-Viera A, Ale I, Álvarez L. **Systematic review about 10 interventions in dermatitis, a document from the Latin American Society of Allergy, Asthma, and Immunology**. *Revista Alergia México* (2019) **66** 426-455. DOI: 10.29262/ram.v66i4.636 12. Sonney J, Ward T, Thompson HJ, Kientz JA, Segrin C. **Improving Asthma Care Together (IMPACT) mobile health intervention for school-age children with asthma and their parents: a pilot randomised controlled trial study protocol**. *BMJ Open* (2022) **12**. DOI: 10.1136/bmjopen-2021-059791 13. Sweaad WK, Stefanizzi FM, Chamorro-Jorganes A, Devaux Y, Emanueli C. **Relevance of N6-methyladenosine regulators for transcriptome: implications for development and the cardiovascular system**. *Journal of Molecular and Cellular Cardiology* (2021) **160** 56-70. DOI: 10.1016/j.yjmcc.2021.05.006 14. Teng F, Tang W, Wuniqiemu T, Qin J, Zhou Y, Huang X, Wang S, Zhu X, Tang Z, Yi L, Wei Y, Dong J. **N (6)-methyladenosine methylomic landscape of lung tissues in murine acute allergic asthma**. *Frontiers in Immunology* (2021) **12**. DOI: 10.3389/fimmu.2021.740571 15. Thun GA, Imboden M, Berger W, Rochat T, Probst-Hensch NM. **The association of a variant in the cell cycle control gene CCND1 and obesity on the development of asthma in the Swiss SAPALDIA study**. *Journal of Asthma* (2013) **50** 147-154. DOI: 10.3109/02770903.2012.757776 16. Tuncel G, Kalkan R. **Importance of m N(6)-methyladenosine (m(6)A) RNA modification in cancer**. *Medical Oncology* (2019) **36** 36. DOI: 10.1007/s12032-019-1260-6 17. Wang Y, Su X, Zhao M, Xu M, Chen Y, Li Z, Zhuang W. **Importance of N(6)-methyladenosine RNA modification in lung cancer (Review)**. *Molecular and Clinical Oncology* (2021) **14** 128. DOI: 10.3892/mco.2021.2290 18. Wu S, Li XF, Wu YY, Yin SQ, Huang C, Li J. **N(6) -methyladenosine and rheumatoid arthritis: a comprehensive review**. *Frontiers in Immunology* (2021) **12**. DOI: 10.3389/fimmu.2021.731842 19. Xiao K, Liu P, Yan P, Liu Y, Song L, Liu Y, Xie L. **N6-methyladenosine reader YTH N6-methyladenosine RNA binding protein 3 or insulin like growth factor 2 mRNA binding protein 2 knockdown protects human bronchial epithelial cells from hypoxia/reoxygenation injury by inactivating p38 MAPK, AKT, ERK1/2, and NF-**. *Bioengineered* (2022) **13** 11973-11986. DOI: 10.1080/21655979.2021.1999550 20. Xu L, Zhou L, Yan C, Li L. **Emerging role of N6-methyladenosine RNA methylation in lung diseases**. *Experimental Biology and Medicine* (2022). DOI: 10.1177/15353702221128564 21. Xu Y, Zhang W, Shen F, Yang X, Liu H, Dai S, Sun X, Huang J, Guo Q. **YTH domain proteins: a family of m(6)A readers in cancer progression**. *Frontiers in Oncology* (2021) **11**. DOI: 10.3389/fonc.2021.629560 22. Yamada J, Lam Shin Cheung J, Gagne M, Spiegel-Feld C, Aaron SD, FitzGerald JM, Gershon AS, Gupta S. **Barriers and enablers to objective testing for asthma and COPD in primary care: a systematic review using the theoretical domains framework**. *Chest* (2022) **161** 888-905. DOI: 10.1016/j.chest.2021.10.030 23. Zhou X, Zhang Y, Liu L, Feng X, Zhang H. **Therapeutic effect of acupuncture combined montelukast sodium on cough variant asthma in children: a protocol for systematic review and meta-analysis**. *Medicine* (2021) **100**. DOI: 10.1097/md.0000000000028048
--- title: Baseline gene expression in subcutaneous adipose tissue predicts diet-induced weight loss in individuals with obesity authors: - Ali Oghabian - Birgitta W. van der Kolk - Pekka Marttinen - Armand Valsesia - Dominique Langin - W. H. Saris - Arne Astrup - Ellen E. Blaak - Kirsi H. Pietiläinen journal: PeerJ year: 2023 pmcid: PMC10042157 doi: 10.7717/peerj.15100 license: CC BY 4.0 --- # Baseline gene expression in subcutaneous adipose tissue predicts diet-induced weight loss in individuals with obesity ## Abstract ### Background Weight loss effectively reduces cardiometabolic health risks among people with overweight and obesity, but inter-individual variability in weight loss maintenance is large. Here we studied whether baseline gene expression in subcutaneous adipose tissue predicts diet-induced weight loss success. ### Methods Within the 8-month multicenter dietary intervention study DiOGenes, we classified a low weight-losers (low-WL) group and a high-WL group based on median weight loss percentage ($9.9\%$) from 281 individuals. Using RNA sequencing, we identified the significantly differentially expressed genes between high-WL and low-WL at baseline and their enriched pathways. We used this information together with support vector machines with linear kernel to build classifier models that predict the weight loss classes. ### Results Prediction models based on a selection of genes that are associated with the discovered pathways ‘lipid metabolism’ (max AUC = 0.74, $95\%$ CI [0.62–0.86]) and ‘response to virus’ (max AUC = 0.72, $95\%$ CI [0.61–0.83]) predicted the weight-loss classes high-WL/low-WL significantly better than models based on randomly selected genes ($P \leq 0.01$). The performance of the models based on ‘response to virus’ genes is highly dependent on those genes that are also associated with lipid metabolism. Incorporation of baseline clinical factors into these models did not noticeably enhance the model performance in most of the runs. This study demonstrates that baseline adipose tissue gene expression data, together with supervised machine learning, facilitates the characterization of the determinants of successful weight loss. ## Introduction Weight loss resulting from a low-calorie diet (LCD) effectively improves obesity and reduces the health risks among people with overweight and obesity (Bray et al., 2016). A moderate body weight reduction of 5–$10\%$ induced through dieting significantly reduces disease risk and improves the metabolic profile (Gaal, Wauters & Leeuw, 1997). However, inter-individual variability in weight loss in response to lifestyle interventions is large and long-term weight maintenance remains rather poor (Anderson et al., 2001). Identifying markers that can predict the responsiveness to various weight loss interventions is significant for clinical care. In recent years, several studies have applied machine learning methods in the obesity field (for extensive review see Safaei et al., 2021) and a few studies have attempted to predict weight variations in humans by applying machine learning methods. These machine learning approaches help us better understand the biological complexity underlying inter-individual variability in weight loss. The prediction studies have mostly focused on energy intake and expenditure (Thomas et al., 2011, 2014), macronutrient balance (Chow & Hall, 2008), anthropometrics (Finkler, Heymsfield & St-Onge, 2012), and glycemic and insulinemic statuses (Ritz et al., 2019; Valsesia et al., 2020), blood DNA methylation (Keller et al., 2020) as predictor variables. Another attractive factor to predict weight loss is based on adipose tissue function (Goossens, 2017) as adipose tissue is a main regulator for post-dieting physiology (MacLean et al., 2015), responds dynamically to caloric intake (van Baak & Mariman, 2019), and weight loss leads to a desired reduction in adiposity. Indeed, a few studies have utilized supervised machine learning on subcutaneous adipose tissue gene expression data to predict weight-loss response to dietary interventions (Mutch et al., 2011; Armenise et al., 2017; Valsesia et al., 2020). But these studies mostly focused on glycemic control responders following weight loss (Valsesia et al., 2020) or they used the adipose tissue gene expression response to caloric restriction for the prediction models (Mutch et al., 2011; Armenise et al., 2017). Overall, these studies emphasize the importance of adipose tissue metabolism in weight loss interventions and its potential in weight loss predictions. The current work aims to identify whether baseline gene expression in abdominal subcutaneous adipose tissue can predict the weight loss status following a 2-month low-calorie diet (LCD) and a subsequent 6-months weight maintenance period. We reanalyzed retrospectively baseline RNA-seq data from individuals from the previously published Diet Obesity and Genes (DiOGenes) study (Larsen et al., 2010). This study provides important knowledge on baseline adipose tissue metabolism that may prove beneficial in the long term for the development of personalized strategies to improve successful weight maintenance. ## Study design DiOGenes is a multicenter, randomized, dietary intervention study (trial number NCT00390637), described in detail elsewhere (Larsen et al., 2010). Briefly, 938 healthy adults with overweight or obesity (BMI 27–45 kg/m2) without concomitant medications entered the first phase of the study: a 2-month weight-loss period using a complete meal replacement low calorie diet (LCD) providing 800 kcal/day (Modifast; Nutrition et Santé France, Revel, France). Among the 781 participants who completed the LCD period, 773 achieved >$8\%$ weight loss and those individuals were randomly assigned to one of five different weight maintenance diets for the next 6 months, differing in macronutrient compositions (a 2 × 2 factorial combination diet of low/high protein and a low/high glycemic index or a control diet). A total of 548 ($71\%$) participants completed the whole 8-month intervention. Here, we examined data from 281 out of these 548 participants, for whom abdominal subcutaneous adipose tissue RNA sequencing data at baseline was available. Weight development and metabolic health were assessed at baseline, 2, and 8 months. The ethics committee of each center/country approved the protocol. The committees included [1] Medical ethical commission from Maastricht University, NL [2] Copenhagen ethical research commission, DK [3] Bedfordshire local Research Ethics Committee, Luton and Dunstable Hospital NHS Trust, UK [4] Ethics Committee of the Faculty Hospital, Prague University, CZ [5] Ethical Commission by NMTI, Sofia, BG [6] Ethical Commission University Potsdam, D [7] Ethical Commission Medical University, Navarra, SP [8] Scientific council *Heraklion* general university hospital, Heraklion, GR and [9] Commission Cantonale d’ éthique de la recherche sur l’ être humain, Canton de Vaud, CH. Furthermore, the protocol was in accordance with the Declaration of Helsinki. All study participants provided written consent. ## Subcutaneous adipose tissue biopsy and transcriptomic analyses Abdominal subcutaneous adipose tissue needle biopsies were collected 6–8-cm laterally from the umbilicus under local anesthesia. Thereafter, biopsy specimens were snap-frozen in liquid nitrogen and stored at −80 °C until further analysis. Total RNA was extracted and analyzed from subcutaneous adipose tissue biopsy specimens as described elsewhere (Armenise et al., 2017; van der Kolk et al., 2019). For each sample, gene expression was then examined using 100-nucleotide-long paired-end RNA sequencing with an Illumina HiSeq 2000 of libraries prepared using the Illumina TruSeq kit according to the manufacturer’s instructions. We used GenomicAlignments to retrieve the number of reads mapping onto 53,343 genes (GRCh37 assembly). Only reads with both ends mapping onto coordinates of a single gene were considered (Armenise et al., 2017; van der Kolk et al., 2019). The data is accessible through Gene Expression Omnibus under the accession no. GSE95640. ## Construction, training and testing of weight loss prediction models A diagram illustrating our method is shown in Figs. 1A and 1B. The RNAseq data from the studied individuals at baseline were split into training and testing subsamples. To prevent overfitting of the prediction models, we used a five-fold cross-validation with 100 resampling (i.e., 100 times our samples were randomly divided to $80\%$ training and $20\%$ testing subsets). This machine learning setup was used for all the prediction models that are reported in this article. Within each training and testing subset, participants were divided into high-weight losers (WL) and low-WL groups based on the cutoff of median weight loss percentage in the subset. **Figure 1:** *Data analysis workflow.(A) Abstract illustration of the analysis performed during a single run (out of 100). Note that for each analysis (i.e., feature/gene extraction, and construction, training and testing of the prediction models) the necessary normalization and adjustment for sex, age and study center effects are carried out. (B) The “feature/gene extraction” precedes all other analysis (i.e., prediction model construction, training and testing) and is shown in further details. In each run the corresponding training data is analyzed. The outputs during the 100 runs are considered to extract the significant GO categories and genes that are used for construction of prediction models. Furthermore, the dimensions PC1-PCn collectively explain 70% of the gene expression variance.* We used support vector machines (SVM) with linear kernel (and cost value of 1) to build classifier prediction models. Prior to the training/testing of the models, we applied the ComBatSeq() function supported by the SVA R library (Johnson, Li & Rabinovic, 2006; Zhang, Parmigiani & Johnson, 2020) on the count matrix to adjust for biases and effects introduced by sex, age and study center. We then applied Log2 Count Per Million (LCPM) scaling while adjusting the library sizes by normalization factors that were calculated using the Trimmed Mean of M values (TMM) (Robinson, McCarthy & Smyth, 2010). The normalizations were carried out independently for the $80\%$ training and the remaining $20\%$ testing sub-samples. Furthermore, the training and testing of the prediction models were carried out on their corresponding sub-samples. We finally used the ROCR R-package to measure the area under the receiver operating characteristic (ROC) curve, i.e., AUC. The $95\%$ confidence interval (CI) of the AUC of the most optimal models were calculated using the DeLong method supported by the pROC R package (DeLong, DeLong & Clarke-Pearson, 1988). To ensure that the performances of the models are not affected by unaccounted for biases within the expression data, the performances of all prediction models (based on the studied gene-lists, see below) were compared to those that featured randomly selected genes as their input parameters. In further detail, we created 100 lists of randomly picked genes (each list featuring ten genes) over the whole genome, from the genes with >5 mapped reads across all samples. Next, we built 100 prediction models based on the randomly selected genes. For each prediction model a P-value was measured. The P-value is the fraction of the 100 prediction models that were based on the randomly picked genes that resulted in higher median AUCs compared to the prediction model (i.e., based on a studied gene-list). We considered $P \leq 0.05$ as significant. ## Selection of genes for the prediction models We studied differentially expressed genes (DEGs), comparing high-WL to low-WL, by applying DESeq2 on the training samples in each cross-validation run (Love, Huber & Anders, 2014). We adjusted the models for sex, age and study center and we considered the nominal $P \leq 0.01$ as significant. We then extracted the biological process Gene Ontology (GO) categories enriched in the significant DEGs, using the enrichGO() function supported by the clusterProfiler R/Bioconductor package (with the default parameter settings of $P \leq 0.05$ and q < 0.2 cutoffs) (Yu et al., 2012). This function applies hypergeometric distribution to calculate P-values based on the number of selected genes associated with the GO category, the overall number of genes associated with the GO category, and the overall number of genes in background (Boyle et al., 2004). The Benjamini Hochberg method was used to correct for multiple testing (Benjamini & Hochberg, 1995) and we considered FDR <0.05 significant. Furthermore, we constructed enrichment maps based on the similarities of the ontologies using the enrichplot R package and the redundant GO categories were filtered based on a ≥$70\%$ similarity cutoff. We applied the Jaccard similarity coefficient method supported by the pairwise_termsim() function of the enrichplot R package to construct the similarity matrix of the GO categories (Yu & Hu, 2020). The ‘components’ within the enrichment maps, i.e., biological process GO categories that for each pair at least one path exists, were finally extracted. We used the vertices within these components, i.e., classes of interrelated biological process GO categories, to extract the suitable genes for construction of the prediction models. The biological process GO categories discovered from comparing the high-WL vs. low-WL include fatty acid metabolic pathways (referred to as Lipid), mitotic/meiotic pathways (referred to as Mitosis), and ‘response to virus’ pathways (referred to as Virus). Expression data of the genes that are associated with a pathway (i.e., Lipid, Mitosis or Virus) were extracted and principle component analysis (PCA) was performed. From each list of genes associated with a pathway, ten genes that most contributed to the PC dimensions that collectively explained ≥$70\%$ of the variance within the expression data was chosen. Since we used five-fold cross-validation with 100 resampling, the training data varied across the different runs and the discovered genes from a specific pathway across the different runs somewhat varied too. Therefore, from each studied Lipid/Mitosis/*Virus* gene list, the ten most frequently featured genes during the 100 runs were selected as input parameters to the corresponding prediction models. Finally, to further analyze the performance of the prediction models based on lipid metabolism and ‘response to virus’ genes, we also constructed models based on combined gene lists (named as Lipid + Virus), and prediction models based on genes featured in one but absent from the other (i.e., Lipid—Virus and Virus—Lipid). ## Incorporating clinical factors into the prediction models We studied the effects of anthropometric and clinical factors (Table S1 and S2) by building a prediction model solely based on these factors as well as incorporating them into the input parameters of the Lipid and Virus models. The values of the clinical factors were adjusted for sex and age using the non-parametric method supported by ComBat() function of the sva R/Bioconductor package (Johnson, Li & Rabinovic, 2006; Zhang, Parmigiani & Johnson, 2020). The highly skewed clinical data (i.e., skewness > 1 or skewness < −1) were loge scaled. For each Lipid and Virus model that incorporated a clinical factor, a P-value was calculated by using paired t-test, by comparing their performances to the original Lipid/Virus prediction models that lack the clinical factor information. We considered $P \leq 0.05$ as significant. ## Expression heatmaps We normalized the gene expression within all samples across all three time points using variance stabilizing transformation (VST) supported by DESeq2, whilst adjusting for sex, age and study center in the design model (Anders & Huber, 2010). ## Baseline characteristics and overall clinical outcome Baseline characteristics of the participants are described in Table 1. For descriptive reasons, we created this table based the median weight loss percentage achieved at 8-months ($9.9\%$) from 281 participants. Sixty-five percent of the participants were females, and the participants were on average 42-years-old. At baseline, the BMI of included participants was 34.6 ± 0.2 kg/m2. The high-WL and low-WL groups were comparable at baseline in term of weight and insulin sensitivity (Table 1). Upon LCD at the 2-month time point, both groups had achieved significant weight loss (>$8\%$ initial body weight); yet the high-WL group had significantly higher weight loss (mean difference $2.3\%$, $P \leq 0.001$). By design, following 8 months, the high-WL group lost significantly more of their initial body weight 15.4 ± $4.6\%$ compared to the low-WL group who lost 5.8 ± $3.4\%$ ($P \leq 0.001$). **Table 1** | Unnamed: 0 | High-WLbaseline | 2 Months | 8 Months | Low-WLbaseline | 2 Months.1 | 8 Months.1 | P-valuebaseline | | --- | --- | --- | --- | --- | --- | --- | --- | | Female sex (%) | 64.5 | | | 65.7 | | | 0.836 | | Age (y) | 42.2 ± 6.7 | | | 42.5 ± 5.7 | | | 0.685 | | Weight loss (%) | | 12.3 ± 2.9 | 15.4 ± 4.6 | | 10.0 ± 1.9* | 5.8 ± 3.4# | | | Weight (kg) | 100.4 ± 18.3 | 87.9 ± 15.9 | 84.8 ± 15.8 | 97.7 ± 16.7 | 88.0 ± 15.1 | 92.0 ± 15.7 | 0.211 | | BMI (kg/m2) | 34.6 ± 4.8 | 30.3 ± 4.3 | 29.2 ± 4.2 | 34.0 ± 4.4 | 30.6 ± 4.1 | 32.0 ± 4.1 | 0.291 | | Cholesterol (mmol/L) | 4.8 ± 0.9 | 4.2 ± 0.8 | 4.8 ± 0.8 | 4.9 ± 1.1 | 4.3 ± 1.0 | 5.0 ± 1.0 | 0.339 | | LDL (mmol/L) | 3.0 ± 0.8 | 2.5 ± 0.7 | 2.9 ± 0.8 | 3.1 ± 0.9 | 2.6 ± 0.8 | 3.0 ± 0.9 | 0.479 | | HDL (mmol/L) | 1.2 ± 0.3 | 1.2 ± 0.3 | 1.4 ± 0.3 | 1.3 ± 0.3 | 1.2 ± 0.3 | 1.4 ± 0.3 | 0.582 | | TAG (mmol/L) | 1.4 ± 0.6 | 1.1 ± 0.5 | 1.1 ± 0.5 | 1.4 ± 0.7 | 1.1 ± 0.4 | 1.4 ± 0.6 | 0.682 | | NEFA (µmol/L) | 680 ± 294 | 747 ± 231 | 574 ± 229 | 640 ± 281 | 685 ± 218 | 564 ± 227 | 0.295 | | Glucose (mmol/L) | 5.1 ± 0.6 | 4.8 ± 0.5 | 4.9 ± 0.5 | 5.1 ± 0.6 | 4.9 ± 0.5 | 5.1 ± 0.5 | 0.734 | | Insulin (mU/L) | 12.6 ± 18.1 | 9.2 ± 15.1 | 9.9 ± 14.3 | 10.9 ± 6.5 | 7.8 ± 3.9 | 9.1 ± 4.5 | 0.312 | | HOMA-IR | 3.4 ± 5.0 | 2.2 ± 3.9 | 2.5 ± 4.0 | 2.9 ± 1.9 | 1.9 ± 1.0 | 2.3 ± 1.3 | 0.358 | ## Prediction models based on biological processes associated with weight-loss To create a predictive model for diet-induced weight loss success based on global subcutaneous adipose tissue gene expression and biological processes, we extracted the significantly DEGs in the training subsamples used in each cross-validation run by comparing the high-WL to the low-WL subsamples at baseline. Subsequently, we created GO enrichment maps that illustrate the relationships of the discovered significantly enriched biological process GOs based on their pairwise gene overlaps. Throughout the 100 runs, the most frequently discovered biological process GO categories were associated with fatty acid metabolic pathways (i.e., Lipid, discovered in 42 runs), mitosis/meiosis pathways (Mitosis, discovered in 36 runs), and ‘response to virus’ pathways (Virus, discovered in 27 runs) (Table 2). In 34 runs, however, no significant GOs were discovered. The largest nonoverlapping Lipid, Mitosis, and Virus associated biological process classes discovered across the 100 cross validation runs are illustrated in Fig. 2A. Next, we extracted ten genes from each of the three discovered pathways (Fig. 2B and Table 3). We used the baseline expression of these genes as input parameters to construct weight-loss prediction models. The Lipid model (median AUC = 0.58, $P \leq 0.01$) (Fig. 3A) and Virus model (median AUC = 0.59, $P \leq 0.01$) (Fig. 3C) predicted the 8-months weight loss of the individuals in our study significantly better than 100 models that feature randomly selected genes as their input parameters or the Mitosis model (median AUC = 0.50, $$P \leq 0.21$$) (Fig. 3B). The maximum AUC achieved by the Lipid models was 0.74 with $95\%$ CI [0.62–0.86], whilst for the Virus models was 0.72 with $95\%$ CI [0.61–0.83] and for the Mitosis models was 0.66 with $95\%$ CI [0.54–0.78]. Notably, the genes in the Virus model were highly in common with the genes in the Lipid models (Fig. 2B) indicating that the discovered ‘response to virus’ pathways have genes with heterogenous functions that are also associated with the lipid metabolism pathways. Therefore, we further examined prediction models that feature combinations of the Lipid and *Virus* genes. Prediction models based on combined ‘response to virus’ and lipid metabolism gene-lists (Lipid + Virus) did not perform significantly better than Virus or Lipid models (Figs. S1A, S1B). Furthermore, excluding the genes that are annotated with ‘response to virus’ from the Lipid models (Lipid–Virus) did not affect the performance of the Lipid metabolism significantly (Fig. S1C and Table S3). However, prediction models that featured genes associated with ‘response to virus’ but not annotated with lipid metabolism (Virus–Lipid) performed drastically different compared to the Virus models (Fig. S1D). The performance of these models was no longer significantly better than those models that feature randomly selected genes (median AUC = 0.51, max AUC = 0.67, $95\%$ CI [0.51–0.75], $$P \leq 0.16$$) (Fig. S2A and Table S3). This shows that the performance of the Virus models is highly dependent on the genes that they have in common with the Lipid models. In contrast, the performance of the Lipid models are not dependent on the genes that they have in common with the Virus models (Fig. S2B and Table S3). To unravel the conditions under which the Lipid and Virus prediction models best performed, and to explain their bimodal AUC distributions (Figs. 3A, 3C), we compared the training subsamples in the runs that resulted the AUCs higher than 0.60 vs. those in the runs that resulted lower AUCs. Note that the overall weight loss during the 8-months correlated highly with the weight loss during the 6-months weight maintenance phase ($r = 0.90$, $P \leq 2.2$e−16, Pearson correlation test) (Fig. 3D), but our results show that the models perform worst when the training dataset include more samples from individuals who gained weight during the 6-months weight maintenance phase despite being in the high-WL group. In detail, 30 high-WL individuals gained weight during the 6-months weight maintenance of which seven samples regained more than $4\%$ (i.e., 5th percentile) weight from the LCD-induced weight loss (Fig. 3D). We realized that runs with AUCs higher than 0.6 included significantly less samples from those seven individuals ($P \leq 0.05$, Jonckheere–Terpstra test) in their training sets compared to other runs. This was true for both, Lipid ($$P \leq 0.048$$, Jonckheere–Terpstra test) and Virus ($$P \leq 0.034$$, Jonckheere–Terpstra test) prediction models. These results indicate that training the models on data from high-WL individuals who regained weight considerably can lead to less efficient prediction models. ## Incorporating clinical factors to improve the prediction models Whilst most clinical parameters did not display strong separation between the High-WL and Low-WL groups (Table 1), we still hypothesized that the combination of easily accessible clinical parameters could predict weight loss. We constructed a prediction model based on 30 anthropometric and clinical factors, but this model did not predict the weight loss status of the samples significantly better than the models that feature randomly selected genes (median AUC = 0.51, max AUC = 0.75, $95\%$ CI [0.60–0.90], $$P \leq 0.12$$) (Fig. S3). We then hypothesized that adding these 30 anthropometric and clinical factors, one by one, to the input parameters of Lipid and Virus prediction models may yield to improved predictions. The following factors improved the performance of the Lipid models: systolic blood pressure (median AUC = 0.59, max AUC = 0.74, $95\%$ CI [0.63–0.86], $$P \leq 0.037$$) and fibrinogen (median AUC = 0.59, max AUC = 0.77, $95\%$ CI [0.67–0.88], $$P \leq 0.049$$) (Table S1). The performance of the Virus models was improved by adding systolic blood pressure (median AUC = 0.61, max AUC = 0.76, $95\%$ CI [0.65–0.86], $$P \leq 0.018$$), diastolic blood pressure (median AUC = 0.59, max AUC = 0.74, $95\%$ CI [0.65–0.86], $$P \leq 0.037$$) or waist-to-hip ratio (median AUC = 0.59, max AUC = 0.74, $95\%$ CI [0.67–0.88], $$P \leq 0.037$$) (Table S2). Notably, in most of the runs, no significant changes of performance were observed upon adding anthropometric and clinical factors (ΔAUC < 0.1) (Figs. 4A–4E). However, in one to six out of 100 runs where the Lipid and Virus models had performed poorly (AUC < 0.5), adding these parameters significantly improved the prediction model performances. In contrast, in a few runs, these parameters worsened the prediction performances where the Lipid and Virus models solely based on genes had performed well (AUC > 0.5). Overall, these results indicate that the effect of incorporating anthropometric and clinical factors into the weight-loss prediction model is negligible and should be carried out with caution. **Figure 4:** *The effect of incorporating anthropometric and clinical factors into the lipid and virus models.(A and B) Scatter plots illustrating AUC measurements for weight loss prediction models that as input parameters combine the Lipid genes models and (A) systolic blood pressure and (B) fibrinogen. (C–E) Scatter plots illustrating the AUC measurements for weight loss prediction models that as input parameters combine Virus models and (C) waist-to-hip ratio, (D) systolic blood pressure, and (E) diastolic blood pressure. The Y axis corresponds to the AUC measurements for the combined models whilst the X axis corresponds to the AUCs for the models solely based on lipid metabolism or ‘response to virus’ gene expression. Dots or triangles indicate individual runs. The black dots represent those runs where the AUC performance differences between the models that the Y axis and X axis represent were less than 0.1. The threshold cutoffs (|Y−X| = 0.1) are shown with dashed lines. The solid line represents Y = X. The red triangles show the runs in which the difference of the hybrid and original models are significant different (|Y−X| > 0.1). Sys BP, Systolic blood pressure; Fibr, fibrinogen; WHR, waist-to-hip ratio; Sys BP, systolic blood pressure.* ## Discussion We aimed to study whether baseline gene expression in subcutaneous adipose tissue can predict the weight loss achieved following a caloric restriction and weight maintenance period. By using supervised machine learning, we show that lipid metabolism and ‘response to virus’ related gene expression in subcutaneous adipose tissue at baseline can predict the weight-loss status (i.e., high-WL or low-WL) of individuals with obesity. Adding clinical variables as input parameters did not necessarily improve the Lipid or Virus prediction models. In this study, we show that the combined expression of ten lipid metabolism related genes (Lipid models) can predict the weight-loss status following long-term (i.e., 8 months) intervention up to a maximum AUC of 0.74. These models featured the genes including GPAM, DGAT2 and ELOVL6 which are involved in triacylglycerol synthesis and elongation (Morgan-Bathke et al., 2015; Matsuzaka, 2021). The high-WL group showed relatively lower baseline expression across all these lipid metabolism genes, which indicate that individuals with a more active triacylglycerol formation in subcutaneous adipose tissue at baseline may have a worse chance of achieving long-term weight loss. This is consistent with our previous work whereby alterations in lipid metabolism genes were observed in differentiating glycemic responders from non-responders (Valsesia et al., 2020). Here, we extend these findings by showing that the baseline expression of lipid metabolism genes can also predict the weight loss status of individuals with obesity. However, additional studies are needed to fully elucidate the underlying mechanisms of adipose tissue lipid turnover in weight loss success. The prediction models composed of the Virus pathways genes were also capable of predicting the weight loss groups with a performance as high as AUC of 0.72. We recognized that the group of genes is broad and heterogenous in the Virus models with half of these genes overlapping with the Lipid metabolism pathway genes (i.e., ACSL1, CYP3A4, CYP3A5, CYP1A1 and PLA2G4C). In fact, the performance of Virus models was highly dependent on the overlapping genes. *These* genes are related to metabolizing xenobiotics and endogenous molecules including steroid hormones and poly-unsaturated fatty acids (Bishop-Bailey et al., 2014). Indeed, obesity is known to dysregulate the CYP epoxygenase pathway and evoke a marked suppression of adipose-derived epoxyeicosatrienoic acids levels, thereby affecting adipogenesis (Zha et al., 2014). Individuals with less active adipogenesis genes at baseline may have a better chance of achieving long-term weight loss. Therefore, an important role for adipose tissue lipid metabolism in weight loss success is emphasized by these results. We tried to improve the performance by incorporating multiple anthropometric and clinical factors into the Lipid and Virus prediction models. Although the incorporation of a few parameters (i.e., systolic and diastolic blood pressure, fibrinogen and/or waist-to-hip ratio) significantly improved the performances of Lipid and Virus models in a few runs, overall, they did not affect the performance of the Lipid and Virus models in most runs. Furthermore, the prediction models solely based on the anthropometric and clinical factors did not predict significantly better than models based on randomly selected genes, which is in line with our previous results (Valsesia et al., 2020). Therefore, the performances of the models were not noticeably enhanced, and this once again illustrates the difficulty in predicting weight maintenance based on initial phenotype (Varkevisser et al., 2019). Our results are strengthened by the fact that the DiOGenes study is one of the largest studies of its kind. In order to prevent any overfitting of the prediction models to the studied data, we applied a five-fold cross-validation with 100 resampling. Even though the individuals are deeply phenotyped, the number of included individuals in this study may still be limited for an effective data integration and prediction analysis. Furthermore, we attempted to understand the limits of our prediction models and the dependency of the performance of our discovered prediction models on the training data. The Lipid and Virus models performed the best when their training data included less samples from high-WL individuals who gained weight considerably (>$4\%$) during the weight-maintenance phase, even though during the LCD they had lost weight to an extent that their overall weight loss during the whole 8-months intervention was higher than the median weight loss. This suggests that perhaps some of the individuals who were labelled as high-WL should have been labelled as low-WL or NA. These individuals may have been detected as High-WL due to the limitations of the data rather than their weight loss. As their weight is rising during the weight maintenance phase, we can speculate that they may have been labelled as low-WL if the weight of the studied individuals were continued to be measured for a longer period than 8 months. Furthermore, with our data it is almost impossible to use a more accurate method to label high-WL and low-WL samples (e.g., considering the rate of the weight regain of the studied individuals during the weight maintenance period) as only weight measurements at three time points are available. Another limitation of the current study is that we have classified the participants solely on changes in body weight during the whole dietary intervention and have attempted to predict the weight loss outcome using gene expression in subcutaneous adipose tissue and baseline clinical information. Furthermore, we did not consider the different macronutrient contents of the diets in our analyses since the differences in SAT gene expression was previously reported to be associated with weight variations rather than the dietary macronutrient content (Márquez-Quiñones et al., 2010). However, we fully acknowledge that the regulation of body weight is multifactorial and can be influenced by many other factors including lifestyle/behavioral and environmental factors (Varkevisser et al., 2019), for which detailed data is lacking in our study. ## Conclusions In conclusion, our study demonstrates the importance of phenotyping on molecular level to identify possible mechanisms involved in the dynamics of weight loss success. We show that prediction models based on adipose tissue gene expression related to lipid metabolism, and ‘response to virus’ (with genes that are also highly associated with lipid metabolism) can predict the weight loss outcome following a dietary intervention. Furthermore, the performance of the prediction models based on ‘response to virus’ genes were highly dependent on those genes that are also annotated with lipid metabolism. Therefore, our results confirm the importance of lipid metabolism in subcutaneous adipose tissue in the long-term weight loss of individuals with obesity. Finally, our methodologies and findings can also benefit clinicians and scientists by providing a better understanding of the large inter-individual variability in response to dietary interventions and methods to tackle challenges they introduce. ## References 1. Anders S, Huber W. **Differential expression analysis for sequence count data**. *Genome Biology* (2010) **11** R106. DOI: 10.1186/gb-2010-11-10-r106 2. Anderson JW, Konz EC, Frederich RC, Wood CL. **Long-term weight-loss maintenance: a meta-analysis of US studies**. *American Journal of Clinical Nutrition* (2001) **74** 579-584. DOI: 10.1093/ajcn/74.5.579 3. Armenise C, Lefebvre G, Carayol J, Bonnel S, Bolton J, Cara AD, Gheldof N, Descombes P, Langin D, Saris WH, Astrup A, Hager J, Viguerie N, Valsesia A. **Transcriptome profiling from adipose tissue during a low-calorie diet reveals predictors of weight and glycemic outcomes in obese, nondiabetic subjects**. *The American Journal of Clinical Nutrition* (2017) **106** ajcn156216. DOI: 10.3945/ajcn.117.156216 4. Benjamini Y, Hochberg Y. **Controlling the false discovery rate—a practical and powerful approach to multiple testing**. *Journal of the Royal Statistical Society Series B-Statistical Methodology* (1995) **57** 289-300. DOI: 10.1111/j.2517-6161.1995.tb02031.x 5. Bishop-Bailey D, Thomson S, Askari A, Faulkner A, Wheeler-Jones C. **Lipid-metabolizing CYPs in the regulation and dysregulation of metabolism**. *Annual Review of Nutrition* (2014) **34** 261-279. DOI: 10.1146/annurev-nutr-071813-105747 6. Boyle EI, Weng S, Gollub J, Jin H, Botstein D, Cherry JM, Sherlock G. **GO: : TermFinder—open source software for accessing Gene Ontology information and finding significantly enriched Gene Ontology terms associated with a list of genes**. *Bioinformatics* (2004) **20** 3710-3715. DOI: 10.1093/bioinformatics/bth456 7. Bray GA, Frühbeck G, Ryan DH, Wilding JPH. **Management of obesity**. *The Lancet* (2016) **387** 1947-1956. DOI: 10.1016/S0140-6736(16)00271-3 8. Chow CC, Hall KD. **The dynamics of human body weight change**. *PLOS Computational Biology* (2008) **4**. DOI: 10.1371/journal.pcbi.1000045 9. DeLong ER, DeLong DM, Clarke-Pearson DL. **Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach**. *Biometrics* (1988) **44** 837-845. DOI: 10.2307/2531595 10. Finkler E, Heymsfield SB, St-Onge M-P. **Rate of weight loss can be predicted by patient characteristics and intervention strategies**. *Journal of the Academy of Nutrition and Dietetics* (2012) **112** 75-80. DOI: 10.1016/j.jada.2011.08.034 11. Gaal LFV, Wauters MA, Leeuw IHD. **The beneficial effects of modest weight loss on cardiovascular risk factors**. *International Journal of Obesity* (1997) **21 Suppl 1** S5-S9. PMID: 9130034 12. Goossens GH. **The metabolic phenotype in obesity: fat mass, body fat distribution, and adipose tissue function**. *Obesity Facts* (2017) **10** 207-215. DOI: 10.1159/000471488 13. Johnson WE, Li C, Rabinovic A. **Adjusting batch effects in microarray expression data using empirical Bayes methods**. *Biostatistics* (2006) **8** 118-127. DOI: 10.1093/biostatistics/kxj037 14. Keller M, Meir AY, Bernhart SH, Gepner Y, Shelef I, Schwarzfuchs D, Tsaban G, Zelicha H, Hopp L, Müller L, Rohde K, Böttcher Y, Stadler PF, Stumvoll M, Blüher M, Kovacs P, Shai I. **DNA methylation signature in blood mirrors successful weight-loss during lifestyle interventions: the CENTRAL trial**. *Genome Medicine* (2020) **12** 97. DOI: 10.1186/s13073-020-00794-7 15. Larsen TM, Dalskov S-M, van Baak M, Jebb SA, Papadaki A, Pfeiffer AFH, Martinez JA, Handjieva-Darlenska T, Kunesova M, Pihlsgård M, Stender S, Holst C, Saris WHM, Astrup A. **Diets with high or low protein content and glycemic index for weight-loss maintenance**. *New England Journal of Medicine* (2010) **363** 2102-2113. DOI: 10.1056/NEJMoa1007137 16. Love MI, Huber W, Anders S. **Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2**. *Genome Biology* (2014) **15** 21-31. DOI: 10.1186/s13059-014-0550-8 17. MacLean PS, Higgins JA, Giles ED, Sherk VD, Jackman MR. **The role for adipose tissue in weight regain after weight loss**. *Obesity Reviews* (2015) **16** 45-54. DOI: 10.1111/obr.12255 18. Márquez-Quiñones A, Mutch DM, Debard C, Wang P, Combes M, Roussel B, Holst C, Martinez JA, Handjieva-Darlenska T, Kalouskova P, Jebb S, Babalis D, Pfeiffer AFH, Larsen TM, Astrup A, Saris WHM, Mariman E, Clement K, Vidal H, Langin D, Viguerie N, Project D. **Adipose tissue transcriptome reflects variations between subjects with continued weight loss and subjects regaining weight 6 mo after caloric restriction independent of energy intake**. *The American Journal of Clinical Nutrition* (2010) **92** 975-984. DOI: 10.3945/ajcn.2010.29808 19. Matsuzaka T. **Role of fatty acid elongase Elovl6 in the regulation of energy metabolism and pathophysiological significance in diabetes**. *Diabetology International* (2021) **12** 68-73. DOI: 10.1007/s13340-020-00481-3 20. Morgan-Bathke M, Chen L, Oberschneider E, Harteneck D, Jensen MD. **More insights into a human adipose tissue GPAT activity assay**. *Adipocyte* (2015) **5** 93-96. DOI: 10.1080/21623945.2015.1068977 21. Mutch DM, Pers TH, Temanni MR, Pelloux V, Márquez-Quiñones A, Holst C, Martinez JA, Babalis D, van Baak MA, Handjieva-Darlenska T, Walker CG, Astrup A, Saris WH, Langin D, Viguerie N, Zucker J-D, Clement K. **A distinct adipose tissue gene expression response to caloric restriction predicts 6-mo weight maintenance in obese subjects**. *American Journal of Clinical Nutrition* (2011) **94** 1399-1409. DOI: 10.3945/ajcn.110.006858 22. Ritz C, Astrup A, Larsen TM, Hjorth MF. **Weight loss at your fingertips: personalized nutrition with fasting glucose and insulin using a novel statistical approach**. *European Journal of Clinical Nutrition* (2019) **73** 1529-1535. DOI: 10.1038/s41430-019-0423-z 23. Robinson MD, McCarthy DJ, Smyth GK. **edgeR: a bioconductor package for differential expression analysis of digital gene expression data**. *Bioinformatics* (2010) **26** 139-140. DOI: 10.1093/bioinformatics/btp616 24. Safaei M, Sundararajan EA, Driss M, Boulila W, Shapi’i A. **A systematic literature review on obesity: understanding the causes & consequences of obesity and reviewing various machine learning approaches used to predict obesity**. *Computers in Biology and Medicine* (2021) **136** 104754. DOI: 10.1016/j.compbiomed.2021.104754 25. Thomas DM, Gonzalez MC, Pereira AZ, Redman LM, Heymsfield SB. **Time to correctly predict the amount of weight loss with dieting**. *Journal of the Academy of Nutrition and Dietetics* (2014) **114** 857-861. DOI: 10.1016/j.jand.2014.02.003 26. Thomas DM, Martin CK, Heymsfield S, Redman LM, Schoeller DA, Levine JA. **A simple model predicting individual weight change in humans**. *Journal of Biological Dynamics* (2011) **5** 579-599. DOI: 10.1080/17513758.2010.508541 27. Valsesia A, Chakrabarti A, Hager J, Langin D, Saris WHM, Astrup A, Blaak EE, Viguerie N, Masoodi M. **Integrative phenotyping of glycemic responders upon clinical weight loss using multi-omics**. *Scientific Reports* (2020) **10** 9236-9314. DOI: 10.1038/s41598-020-65936-8 28. van Baak MA, Mariman ECM. **Mechanisms of weight regain after weight loss—the role of adipose tissue**. *Nature Reviews Endocrinology* (2019) **15** 1-14. DOI: 10.1038/s41574-018-0148-4 29. van der Kolk BW, Kalafati M, Adriaens M, van Greevenbroek MMJ, Vogelzangs N, Saris WHM, Astrup A, Valsesia A, Langin D, van der Kallen CJH, Eussen SJPM, Schalkwijk CG, Stehouwer CDA, Goossens GH, Arts ICW, Jocken JWE, Evelo CT, Blaak EE. **Subcutaneous adipose tissue and systemic inflammation are associated with peripheral but not hepatic insulin resistance in humans**. *Diabetes* (2019) **68** 2247-2258. DOI: 10.2337/db19-0560 30. Varkevisser RDM, Stralen MM, Kroeze W, Ket JCF, Steenhuis IHM. **Determinants of weight loss maintenance: a systematic review**. *Obesity Reviews* (2019) **20** 171-211. DOI: 10.1111/obr.12772 31. Yu G, Hu E. **Enrichplot: visualization of functional enrichment result. Bioconductor version: Release (3.11)**. (2020) 32. Yu G, Wang L-G, Han Y, He Q-Y. **clusterProfiler: an R package for comparing biological themes among gene clusters**. *OMICS: A Journal of Integrative Biology* (2012) **16** 284-287. DOI: 10.1089/omi.2011.0118 33. Zha W, Edin ML, Vendrov KC, Schuck RN, Lih FB, Jat JL, Bradbury JA, DeGraff LM, Hua K, Tomer KB, Falck JR, Zeldin DC, Lee CR. **Functional characterization of cytochrome P450-derived epoxyeicosatrienoic acids in adipogenesis and obesity**. *Journal of Lipid Research* (2014) **55** 2124-2136. DOI: 10.1194/jlr.M053199 34. Zhang Y, Parmigiani G, Johnson WE. **ComBat-seq: batch effect adjustment for RNA-seq count data**. *NAR Genomics and Bioinformatics* (2020) **2** lqaa078. DOI: 10.1093/nargab/lqaa078
--- title: 'Hormonal drugs: Influence on growth, biofilm formation, and adherence of selected gut microbiota' authors: - Zainab K. Hammouda - Reham Wasfi - Nourtan F. Abdeltawab journal: Frontiers in Cellular and Infection Microbiology year: 2023 pmcid: PMC10042233 doi: 10.3389/fcimb.2023.1147585 license: CC BY 4.0 --- # Hormonal drugs: Influence on growth, biofilm formation, and adherence of selected gut microbiota ## Abstract Many studies have reported the influence of hormonal drugs on gut microbiota composition. However, the underlying mechanism of this interaction is still under study. Therefore, this study aimed to evaluate the possible in vitro changes in selected members of gut bacteria exposed to oral hormonal drugs used for years. Selected members of gut bacteria were Bifidobacterium longum, Limosilactobacillus reuteri, Bacteroides fragilis, and *Escherichia coli* representing the four main phyla in the gut. Selected hormonal drugs used for a long time were estradiol, progesterone, and thyroxine. The effect of intestinal concentrations of these drugs on the selected bacterial growth, biofilm formation, and adherence to Caco-2/HT-29 cell line was assessed. Short-chain fatty acids (SCFAs) have been included in host functions including the gut, immune and nervous functions; thus, the drug’s effects on their production were assayed using High- Performance Liquid Chromatography. Sex steroids significantly increased the growth of all tested bacteria except B. longum, similarly, thyroxine increased the growth of tested Gram-negative bacteria however reducing that of tested Gram-positive bacteria. The effect of drugs on biofilm formation and bacterial adherence to cell lines cocultures was variable. Progesterone decreased the biofilm formation of tested Gram-positive bacteria, it nevertheless increased L. reuteri adherence to Caco-2/HT-29 cell line cell lines coculture. By contrast, progesterone increased biofilm formation by Gram-negative bacteria and increased adherence of B. fragilis to the cell lines coculture. Moreover, thyroxine and estradiol exhibited antibiofilm activity against L. reuteri, while thyroxine increased the ability of E. coli to form a biofilm. Moreover, hormones affected bacterial adherence to cell lines independently of their effect on hydrophobicity suggesting other specific binding factors might contribute to this effect. Tested drugs affected SCFAs production variably, mostly independent of their effect on bacterial growth. In conclusion, our results showed that the microbiota signature associated with some hormonal drug consumption could be the result of the direct effect of these drugs on bacterial growth, and adherence to enterocytes besides the effect of these drugs on the host tissue targets. Additionally, these drugs affect the production of SCFAs which could contribute to some of the side effects of these drugs. ## Introduction Gut bacteria play an important role in maintaining human health through metabolic, protective, and trophic mechanisms; however, altering their composition has been linked to the development of certain diseases such as irritable bowel disease (IBD), colorectal cancer, obesity, diabetes, autism, etc. ( Prakash et al., 2011; DeGruttola et al., 2016). Among the important functions played by the gut bacteria is the production of short-chain fatty acids (SCFAs) as the end products of the metabolism of complex carbohydrates. SCFAs have several effects on maintaining the health and integrity of the human body besides being an energy source for colonocytes, having a strong anti-inflammatory effect, maintaining the integrity of the blood-brain barrier (BBB), and reducing the oxidative stress in the colon. The most abundant SCFAs produced in the colon are acetate, butyrate, and propionate. ( Prakash et al., 2011; Valdes-Varela et al., 2016; Silva et al., 2020). Different types of microorganisms, including bacteria, yeast, and viruses, make up the gut microbiota. Bacteroidetes and Firmicutes account for $90\%$ of the intestinal bacteria, and Proteobacteria, Actinobacteria, and Verrucomicrobia comprise the other major phyla that constitute the gut (Scotti et al., 2017; Rinninella et al., 2019). Microbiota composition and functions are exposed to dysbiosis by host factors and environmental pressure (Thursby and Juge, 2017). Among these factors contributing to possible dysbiosis is drug intake (Nicholson et al., 2012; Hasan and Yang, 2019). Research in population-based cohorts has revealed a connection between various non-antibiotic drug classes and specific gut microbiome signatures highlighting that this interaction might contribute to the development of diseases in addition to the change in drug metabolism by gut microbiota (Vich Vila et al., 2020). Among the drugs that have drawn the attention of researchers to study their effect on gut microbiota are hormonal drugs because they are usually used for a long time ranging from 3 months to a lifetime (Benvenga et al., 2017; Clabaut et al., 2021; Gong et al., 2021). However, the direct effect of these drugs on the growth, adherence, and production of SCFAs by gut bacteria was not studied in vitro except for a few studies that utilized concentrations higher than the estimated intestinal concentrations. Therefore, this study aimed to study the impact of hormonal drugs at the intestinal concentration on the growth, adherence, and production of SCFAs and lactic acid by selected members of the gut microbiota to explain the in vivo changes in gut microbiota upon consumption of hormonal drugs. ## Bacterial strains and culturing condition Four bacterial strains were used including *Bacteroides fragilis* (ATCC 25285), *Bifidobacterium longum* (ATCC 15707), *Escherichia coli* (ATCC 8739), and Limosilactobacillus reuteri (ATCC 23272) representing the 4 main phyla Bacteroidetes, Actinobacteria, Proteobacteria, and Firmicutes, respectively, which comprise the gut microbiota. All strains were cultured in their recommended media: De Man, Rogosa & Sharpe (MRS) broth and lactobacillus selective (LBS) agar for L. reuteri; Brain-heart infusion supplemented (BHIS) broth and neomycin anaerobic blood agar (NABA) for Bacteroides fragilis; Reinforced clostridial media (RCM) broth and MRS agar supplemented with $0.05\%$ cysteine for Bifidobacterium longum; Luria Bertani (LB) broth and MacConkey agar for E. coli. All the previous cultures were incubated under anaerobic conditions (BBL anaerobic jar, Gas pack Anaerobic system, Franklin, New Jersey, USA) at 37°C. Strains were preserved in glycerol stock at −20°C. ## Preparation of drug stock solutions Steroid sex hormones of ethinyl estradiol and progesterone were obtained from Qinhuangdao Zizhu Pharmaceutical (Hebei, China) and Zhejiang Shenzhou Pharmaceutical Co (Zhejiang, China), respectively, while L-thyroxine hormone was supplied by Azico Biophore Ltd (Pradesh, India). Stock solution (100x intestinal concentration) of each drug compound was prepared by dissolving drugs using the least amount of Dimethyl Sulfoxide (DMSO) and diluted with sterile distilled water. Drug stocks were stored at -20°C. The intestinal concentrations used were previously calculated by Maier and colleagues [2018] and the used final concentrations were 0.562, 211.99, and 0.0481µM for ethinyl estradiol, progesterone and L-thyroxine, respectively (Maier et al., 2018). ## Screening the effect of hormonal drugs on bacterial growth This assay was carried out according to Maier et al. [ 2018] with some modifications. The bacterial strains were cultured on their specific medium, and then one isolated colony was allowed to grow overnight on modified Gifu anaerobic (mGAM) broth at 37°C under anaerobic conditions and that was repeated twice to ensure that the culture was robust and consistent. The bacterial overnight culture was adjusted to reach starting OD600 of 0.02. Two controls were used simultaneously along with the test for this experiment. Control [1] bacteria in medium, control [2] bacteria in medium containing DMSO at a final concentration as in diluted drugs. The effect of DMSO at final concentrations on the growth of bacterial strains was monitored using controls [1] and [2]. The drug stocks were thawed, and an aliquot was diluted to reach 2x intestinal concentrations. In a 96-well flat bottom plate (Greiner Bio-one®, Germany), 50 µl of the adjusted bacterial suspension (OD600 = 0.02) were added to 50 µl of the drug concentration (2x). Plates were incubated at 37°C under anaerobic conditions. For the viable count of bacterial growth at zero and 24 hours, tenfold serial dilutions [10-1-10-6] were performed. An aliquot (10 µl) was spotted on NABA, MRS supplemented with cystine (MRS-C), MacConkey agar, and LBS for the enumeration of B. fragilis, B. longum, E. coli, and L. reuteri, respectively. The plates were incubated at 37°C under anaerobic conditions for 48 hours. Following incubation, colonies were counted and used to calculate the number of colonies per ml (CFU/ml) (Wang et al., 2017; Maier et al., 2018). ## Screening the effect of hormonal drugs on auto-aggregation and cell surface hydrophobicity The tendency of identical bacterial cells for self-adherence (auto-aggregation) and cell surface hydrophobicity (CSH) are two independent traits that influence bacterial adhesion ability to surfaces (Rahman et al., 2008). ## Preparation of bacterial inoculum L. reuteri, B. fragilis, B. longum, and E. coli were grown in a 5 ml broth of MRS, BHIS, RCM, and LB, respectively, with different drugs at their intestinal concentrations. Tubes were incubated under anaerobic conditions for 18 hours at 37°C. Positive control was run simultaneously where bacteria were cultured in media with a DMSO concentration equivalent to that used in dissolving the drugs. Bacterial pellets were harvested by centrifugation at 9500 rpm for 10 min at 18°C followed by washing twice with ice-cold phosphate buffer saline to be used in the auto-aggregation and cell surface hydrophobicity (Botes et al., 2008). ## Auto-aggregation Bacterial cells were suspended in saline and adjusted to OD600 of 0.3. One milliliter of the adjusted bacterial suspension was transferred into a sterile eppendorf tube, then centrifuged at 2000 rpm for 2 min (Botes et al., 2008). The supernatant’s optical density (OD600) was measured immediately (A0) and after one hour (A60)(Botes et al., 2008). Auto-aggregation percentage was calculated using the following equation: ## Cell surface hydrophobicity The assay of microbial adhesion to hydrocarbons was used to characterize microbial hydrophobicity according to Lather and his colleagues [2016] with some modifications. Bacterial cells were suspended in saline and adjusted to OD600 = 0.5. In a glass tube, 0.8 ml volume of xylene was added to 4.8 ml of the adjusted bacterial suspension. The mixture was shaken vigorously for 1 min and allowed to separate at room temperature for 60 min. Bacterial cells were distributed between aqueous and organic phases according to bacterial hydrophobicity. The aqueous phase was removed with caution using a micropipette and measured by spectrophotometer at wavelength 500 nm (A) (Lather et al., 2016). Percentage hydrophobicity was calculated using the following equation: A0 is the absorbance before the addition of xylene while A is the absorbance in the aqueous phase after the addition of xylene. ## Screening the drug activity on biofilm formation The effect of drugs on biofilm formation was measured by crystal violet assay. A selective medium supplemented with $1\%$ glucose was used for the incubation of each bacterium at 37°C for 24 hours under anaerobic conditions. Peptone yeast glucose (PYG), LB, BHIS, and RCM were used for the assessment of L. reuteri, E. coli, B. fragilis, and B. longum, respectively. The cell density of bacterial suspension was adjusted to OD600 = 1, followed by dilution 1:100 using the selective fresh medium for each bacterium. In a 96-well flat bottom plate (Greiner Bio-one®, Germany), 100 µl of the bacterial cell suspensions were inoculated with 100 µl of drugs (2x) in each well, incubated for 48 hours at 37°C under anaerobic conditions. After incubation, the bacterial growth was measured using a microtiter plate reader (STAT FAX 2200, Awareness Technology, Florida, USA) at a wavelength of 630 nm. Adhered cells were washed and then stained with $0.1\%$ crystal violet for 30 min followed by washing and solubilization. The colored solution was measured at wavelength 545 nm. The readings were used to calculate the biofilm formation index (Kwasny and Opperman, 2010; Coffey and Anderson, 2014; Woo et al., 2017; Jang et al., 2020). Five technical replicates were used for each bacterial strain to compensate for variability and three biological replicates were performed. A positive control with bacteria in addition to DMSO was used simultaneously along with the test. OD 545 is colorimetric absorbance of stained bacteria, OD 630 is absorbance of bacterial growth, and OD control is absorbance of negative control. ## Screening the effect of selected drugs on bacterial adherence to cell lines Co-culture cells of Caco-2/HT29 (90:10) was used to simulate intestinal tissue which was prepared according to the method described by Kleiveland [2015]. ## Cytotoxicity assay of tested drug on cell line The effect of drugs and DMSO on cell line viability was measured using 3, -4,5 dimethyithiazol-2,5 diphenyl tetrazolium bromide (MTT) assay. Co-culture cells of Caco-2/HT29 were seeded in 96 well microtiter flat bottom plate (Greiner Bio-one®, Germany) using Roswell Park Memorial Institute (RPMI) medium supplemented with $2\%$ Fetal bovine serum (FBS) and incubated overnight in $5\%$ CO2 at 37°C. Following incubation, 100 µl of each drug in their working dilution was added to co-culture cells (three technical replicates for each concentration). A control was run simultaneously: a negative control with DMSO concentrations equivalent to that in drug solution. The plates were sealed and incubated overnight in $10\%$ CO2 at 37°C. MTT was dissolved in fresh medium at concentration $0.05\%$, added to each well, and incubated for 2 hours under the same conditions. After incubation, the medium was aspirated, and 100 µl of DMSO was used for solubilization. Color was measured at wavelength 545 nm and readings were used to calculate percentage viability (Mueller et al., 2004). ## Adherence to cell line assay The bacterial strains were cultured on their specific media for 20 hours under anaerobic conditions at 37°C. Bacterial cells were centrifuged at 6000 rpm for 5 min at 4°C, the pellet was washed twice with PBS and the cell density for each bacterium was adjusted to 1x108 CFU/ml using PBS. In a 24 well flat bottom plate (Greiner Bio-one®, Germany), Caco-2/HT-29 co-culture was grown and maintained using RPMI medium except for E. coli where DMEM medium was used. On plates seeded with cell line coculture, 100 µl of adjusted bacterial suspension and 100 µl of drugs (2x intestinal concentration) were added. Thus, final concentration of the drugs in each well will be equivalent to intestinal concentration (1x). Plates were sealed and incubated at 37°C for 2 hours in $5\%$ CO2. Following incubation, the medium was removed, and wells were washed using 200 µl of fresh medium to remove non-adherent cells. The cells were then lysed by adding 100 µl of $0.1\%$ Triton X-100 for 10 min at room temperature, then the reaction was stopped by addition of 900 µl of fresh medium. Viable bacterial cells were enumerated by spreading 10µl of diluted bacterial culture (drop plate technique) on their selective medium prior to incubation with cell line (CFU of initial inoculum) and after incubation for 2 hours (CFU of adhered cells)(Letourneau et al., 2011; Gagnon et al., 2013; Reddy and Austin, 2017). The count was recorded and used to calculate the percentage of adhered bacteria. ## Imaging of cell adherence to CaCo-2/HT29 co-culture using scanning electron microscope The same steps of adherence assay were followed for preparation of samples in a 12 well flat bottom plate (Greiner Bio-one®, Germany) for imaging the cell adherence using scanning electron microscope. After incubation for 2 hours, the plate was washed with phosphate buffer saline (PBS) twice, and $5\%$ glutaraldehyde (prepared in 0.1M sodium cacodylate) was added for 2 hours for fixation. The wells were dehydrated by passing them through graded ethanol (25, 50, 70, 80, and $90\%$) for 10 min in each concentration at room temperature. The last concentration used for dehydration was $100\%$ for 15 min. The wells were coated with gold and examined using scanning electron microscope (Quanta 250 FEG, West Bengal, India) with a magnification power of 5000x and 10000x (Heckman et al., 2007; Ude et al., 2019). ## Measurement of the change in Short Chain Fatty Acids and lactic acid production by bacteria under the effect of hormonal drugs using high performance liquid chromatography (HPLC) Strict anaerobic bacteria are known for their ability to produce short-chain fatty acids by the saccharolytic fermentation of complex polysaccharides (Nogal et al., 2021), therefore the effect of hormonal drugs on production of SCFA by B. fragilis and B. longum were studied. Analytical samples were prepared by inoculating colonies of B. fragilis and B. longum in BHIS broth and MRS broth, respectively for 48 hours at 37°C under anaerobic conditions. The bacterial OD was adjusted to 0.01 at 600 nm and incubated with the intestinal concentration of the three hormonal drugs at 37°C for 16 hours under anaerobic conditions. The suspension was then centrifuged at 9500 rpm for 15 min at 4°C and the supernatant was collected. The analytical sample was injected into the HPLC system (Smart line, Knauer, Germany) using the autosampler after its conditions was set. The HPLC system was equipped with Rezex™ column (Phenomenex, California, USA) for organic acid analysis. The flow rate was set at 0.6 ml/min, UV detector set at 214 nm, column oven temperature kept constant at 65°C, and the mobile phase was 0.005M H2SO4. To create the standard curve, a standard solution containing lactic, acetic, propionic, and butyric acids was prepared at concentrations of 1, 10, 100, 500, and 1000 ppm. The SCFA quantities were determined using the standard curves’ appropriate linear regression equations (R2 ≥ 0.99). The response factor is a measurement of the analyte’s relative spectral response to its external standard at the specified retention time, followed by calculation of SCFAs in ppm. Positive control was prepared by growing bacteria in media containing DMSO in concentrations equivalent to their final concentration in drug solution. The data was integrated by clarity chrom software (DataApex, Praha, Czechia) ## Statistical analysis GraphPad Prism 9.1.1 (GraphPad Software Inc., CA, USA) was used for statistical analysis. Multiple unpaired t-tests and multiple comparisons using the Holm-Šídák method were used to compare auto-aggregation, hydrophobicity, and formation of biofilm in the presence and absence of drugs. For statistical analysis of the viable count of the microorganisms in the screening of the antibacterial activity of drugs and adherence assay to cell lines coculture, the Mann-Whitney t-test was used. An unpaired t-test was used for the statistical analysis of the effect of drug on the viability of the intestinal cell lines. The readings were considered significant at $p \leq 0.05.$ ## Alteration in bacterial growth under the effect of hormonal drugs B. longum was the most affected bacteria in presence of the three tested hormonal drugs showing reduction in viable count by 3, 2, and 4 logs in presence of progesterone, estradiol, thyroxine, respectively (Figure 1C). The growth of Gram-negative bacteria was enhanced by hormonal treatment showing increase by 1 to 2 logs by B. fragilis (Figure 1A) and E. coli (Figure 1B). The effect on the growth of L. reuteri was variable where the steroid hormones increased its growth by 2 to 3 logs while thyroxine reduced its growth by one log (Figure 1D). **Figure 1:** *Growth of bacteria under the effect of hormonal drugs. Effect of hormonal drugs on the growth of: Gram-negative bacteria (A) Bacteroid fragilis, (B) Escherichia coli, and Gram-positive bacteria (C) Bifidobacterium longum, (D) Limosilactobacillus reuteri represented as viable count (CFU/ml). Tested hormonal drugs include progesterone, ethinyl estradiol, and L-thyroxine in their intestinal concentrations, 211.99, 0.562, and 0.0481µM, respectively. Control represents the growth of bacteria in addition to DMSO. Mann-Whitney t-test was used to statistically compare the effect on bacterial growth. Significance level of **(P<0.001), ***(P<0.0001). ns, non significant.* ## Progesterone changed the auto-aggregation of tested bacteria Progesterone was the only hormonal drug under test that showed an effect on bacterial auto-aggregation (Figures 2A–D) as it caused a significant increase ($$P \leq 0.0008$$, α=0.05) in the auto-aggregation of E.coli from $5.5\%$ in control to $10\%$ in presence of drug (Figure 2C). On the other hand, progesterone remarkably reduced ($$P \leq 0.007$$, α=0.05) the auto-aggregation of L. reuteri cells to $5\%$ compared to the control (Figure 2D). **Figure 2:** *Auto-aggregation of bacteria with hormonal drugs. Change in percentage auto- aggregation of (A) B. fragilis; (B) E. coli; (C) B. longum; (D) L. reuteri after incubation with hormonal drugs at 37°C for 60 min. Values were expressed as the mean of the percentage of three experiments with error bars (SE). Tested hormonal drugs include progesterone, ethinyl estradiol, and L-thyroxine in their intestinal concentrations, 211.99, 0.562, and 0.0481µM, respectively. Control represents the bacteria in addition to DMSO. Multiple unpaired t-tests along with Holm-Šídák for multiple corrections were used to statistically compare the effect of different drugs on bacterial auto-aggregation. * Significant difference (p<0.05).* ## Hormonal drugs change cell surface hydrophobicity of tested bacteria The hydrophobicity of B. fragilis increased significantly ($P \leq 0.01$) in the presence of the three hormonal drugs (Figure 3A). Estradiol reduced significantly ($P \leq 0.01$) the hydrophobicity of B. longum and L. reuteri while progesterone reduction to hydrophobicity was limited to L. reuteri (Figures 3C, D). E. coli did not show significant change in hydrophobicity in presence of the three drugs (Figure 3B). **Figure 3:** *Hydrophobicity of bacteria under the effect of hormonal drugs. Change in percentage hydrophobicity of (A) B. fragilis; (B) E. coli; (C) B. longum; (D) L. reuteri with hormonal drugs after incubation at 37 °C for 60 min. Values were expressed as the mean of the percentage of three experiments with error bars (SE). Tested hormonal drugs include progesterone, ethinyl estradiol, and L-thyroxine in their intestinal concentrations, 211.99, 0.562, and 0.0481µM, respectively. Control represents the bacteria in addition to DMSO. Multiple unpaired t-test along with Holm-Šídák for multiple corrections was used to statistically compare the effect of different drugs on bacterial hydrophobicity. * Significant difference (p<0.05).* ## Hormonal drugs changed biofilm formation ability of tested bacteria B. longum showed the highest biofilm index among tested isolates. Hormonal drugs increased biofilm formation by Gram-negative bacteria (Figures 4A, B) but reduced the ability of Gram-positive bacteria to form biofilm (Figures 4C, D). **Figure 4:** *Biofilm formation ability of selected bacteria under the effect of hormonal drugs. Change in biofilm formation index by (A) B. fragilis; (B) E. coli (C); B. longum; (D) L. reuteri with hormonal drugs. Values were expressed as the mean of the percentage of three experiments with error bars (SE). Tested hormonal drugs include progesterone, ethinyl estradiol, and L-thyroxine in their intestinal concentrations, 211.99, 0.562, and 0.0481µM, respectively. Control represents the bacteria grown in culture with DMSO. Multiple unpaired t-test along with Holm-Šídák for multiple corrections was used to statistically compare the effect of different drugs on bacterial biofilm. * Significant difference (p<0.05).* ## Effect of tested drugs on bacterial adherence to Caco-2/HT-29 coculture cell lines An unpaired t-test was used to compare the difference in cell viability in presence of drugs compared with the untreated coculture cell lines and no significant effect was observed on coculture viability when treated with the three drugs in their intestinal concentrations (data not shown). The effect of hormonal drugs on adherence of tested bacteria was variable. No growth of B. fragilis was observed on NABA after 2 hours of incubation in both control and drug-treated samples except with progesterone which showed relatively low adherence of $0.13\%$ (Figure 5A). Progesterone reduced the adherence of E. coli and B. longum while increasing the adherence of L. reuteri (Figures 5B–D). Thyroxine remarkably increased the percentage of adhered E. coli and B. longum to cell lines coculture while reducing adherence of L. reuteri when compared to control. Ethinyl estradiol reduced the adherence of E. coli and increased the adherence of B. longum and L. reuteri. **Figure 5:** *Bacterial adherence to Caco-2/HT-29 co-culture under the effect of hormonal drugs. Change in the percentage of adhered bacterial cells (A) B. fragilis; (B) E. coli; (C) B. longum; (D) L. reuteri under the effect of hormonal drugs: progesterone, ethinyl estradiol, and thyroxine in their intestinal concentrations 211.99, 0.562, and 0.0481µM, respectively. Values were expressed as the mean of the percentage of three experiments with error bars (SE). Positive control with bacteria and DMSO at different concentrations equivalent to that used to dissolve drug ( control (1): DMSO 1% and Control (2):DMSO 0.001%).* ## Scanning electron microscope The influence of hormonal drugs on bacterial adherence was confirmed by SEM images. As shown in Figure 6A, untreated B. fragilis showed no visible bacterial attachment to the extracellular matrix, while exposure to progesterone increased adherence to the cell line which appeared as few longitudinal rods attaching to the cell line co-culture in Figure 6B. Ethinyl estradiol increased the number of B. longum bacteria adhering to cell line coculture (Figure 6D) when compared to the control (Figure 6C). The effect of progesterone on the adherence of L. reuteri in the presence (Figure 6F) and absence of the drug (Figure 6E) where a visible slight increase in the number of L. reuteri was observed in treated cells. **Figure 6:** *Scanning electron micrograph showing the effect of hormonal drugs on adherence of bacteria to Caco-2/HT-29 co-culture. Effect of progesterone on B. fragilis adherence (A) Control untreated bacterial cells (B) Drug-treated bacterial cells (Magnification power 10000x); B longum (C) Control untreated bacterial cells (D) Drug-treated bacterial cells (Magnification power 5000x); L. reuteri (E) Control with bacterial cells (F) Drug-treated cells bacterial cells (Magnification power 10000x). The adherence assay was carried out using RPMI medium in 5% CO2 at 37°C, and imaging was performed using a scanning electron microscope (Quanta 250 FEG, West Bengal, India).* ## Alteration in SCFAs and lactic acid production by B. fragilis and B. longum in presence of hormonal drugs The reference chromatograms obtained from the standard solution revealed that SCFAs and lactic acid were detected at different retention times: 13.100, 14.876, 17.967, and 20 min; for lactic, acetic, propionic and butyric acids, respectively. The amount of butyric acid produced wasn’t detected in treated and untreated samples under the test conditions. The response factor was calculated for each SCFA, which represented the measurement of the analyte’s relative spectral response to its external standard. The highest acid produced by B. fragilis was acetic acid followed by lactic acid then propionic acid. The interpretation of chromatogram (Supplementary Figure 1 and Supplementary Table 1) showed a reduction in the amount of both lactic acid and propionic acid and an increase in the amount of acetic acid produced by B. fragilis in presence of progesterone (Table 1). The concentration of lactic acid and acetic acid produced by B. fragilis was reduced after being treated with ethinyl estradiol, conversely an increase in the amount of propionic acid was observed under the effect of the same drug. Both lactic acid and propionic acid produced by B. fragilis increased in amount when the bacteria were treated with thyroxine hormone while acetic acid levels were reduced under the same conditions. **Table 1** | Drugs | Compound | Concentration (ppm) | Concentration (ppm).1 | | --- | --- | --- | --- | | Drugs | Compound | Control* | Sample | | Progesterone | Lactic acid | 1738.136 | 411.882 ↓ | | Progesterone | Acetic acid | 5661.760 | 5809.819 ↑ | | Progesterone | Propionic acid | 409.485 | 253.075 ↓ | | Ethinyl estradiol | Lactic acid | 986.53 | 376.539 ↓ | | Ethinyl estradiol | Acetic acid | 5138.918 | 4933.959 ↓ | | Ethinyl estradiol | Propionic acid | 347.249 | 497.194 ↑ | | Thyroxine | Lactic acid | 986.53 | 1218.416 ↑ | | Thyroxine | Acetic acid | 5138.918 | 4512.039 ↑ | | Thyroxine | Propionic acid | 347.249 | 403.225 ↓ | Propionic and butyric acid weren’t detected in B. longum control or drug-treated samples. The chromatogram (Supplementary Figure 2 and Supplementary Table 2) showed a reduction in the amount of both lactic acid and acetic acid produced by B. longum when treated with progesterone. Bifidobacterium longum produced higher levels of both lactic acid and acetic acid after being treated with ethinyl estradiol. Both lactic acid and butyric acid produced by B. longum increased in amount when the bacteria were treated with thyroxine hormone (Table 2). **Table 2** | Drugs | Compound | Concentration (ppm) | Concentration (ppm).1 | | --- | --- | --- | --- | | Drugs | Compound | Control* | Sample | | Progesterone | Lactic acid | 315.68 | 116.640 ↓ | | Progesterone | Acetic acid | 7994.02 | 7296.440 ↓ | | Ethinyl estradiol | Lactic acid | 289.266 | 331.122 ↑ | | Ethinyl estradiol | Acetic acid | 8133.807 | 8405.94 ↑ | | Thyroxine | Lactic acid | 986.53 | 309.830 ↓ | | Thyroxine | Acetic acid | 5138.918 | 8322.263 ↑ | ## Discussion The human gastrointestinal (GI) tract is a niche to a complex and dynamic community of bacteria known as the gut microbiota, which has a significant impact on the host during health and disease (Thursby and Juge, 2017). The composition and function of the gut microbiome are also influenced by different factors including the use of medications (Wen and Duffy, 2017). In our study, the steroid hormones such as ethinyl estradiol and progesterone significantly increased the growth of B. fragilis, E. coli, and L. reuteri whereas decreasing the growth of B. longum. Conversely, previous studies reported the antibacterial effect of steroid derivatives by preventing the normal development of the cell membrane integrity and permeability. Thus, it is thought that the reason for the antibacterial effect of steroid bile acids is due to their binding to phospholipids in bacterial membranes resulting in membrane destruction and ultimately cell death (DoĞAn et al., 2017; Bustos et al., 2018; Vollaro et al., 2020; Crowley et al., 2022). The resistance of the selected gut bacteria to the deleterious effect of steroids could be due to the presence of bile salt hydrolase enzyme in these genera which protect them from the damaging effect of steroid bile acids (Staley et al., 2017). Another explanation for the resistance of L. reuteri to bile salts was the protective effect of this bacteria against steroids as bile acids arising from precipitation of the deconjugated bile salts and physical binding of bile salts by a bacterium, so rendering the detrimental bile salts accessible (De Boever et al., 2000). The increase in growth of B. fragilis could be explained by the study carried out by Kornman and Loesche [1982] using labeled C14 steroid hormones who proved the uptake of these hormones by *Bacteroides bacteria* and explained their ability to substitute vitamin K compounds, an essential growth factor, with progesterone and ethinyl estradiol resulting in an increased growth curve (Kornman and Loesche, 1982). Both steroid hormones had a significant effect in reducing the growth of B. longum in this study supported by findings of previous studies on Gram-positive bacteria, which demonstrated that different steroids reduce in vitro growth and increase cell leakage (Souza et al., 2021). B. longum growth was reduced under the same treatment which could be attributed to their lower resistance to bile acids compared to other Bifidobacterium species (Ibrahim and Bezkorovainy, 1993; Clark and Martin, 1994; Dunne et al., 2001). Many cross sectional studies for the influence of sex steroids on gut bacteria have proved the existence of correlation between sex hormone levels and microbiota composition despite the inevitable interfering factors, including genetics and environment (Santos-Marcos et al., 2018; Shin et al., 2019; Zhao et al., 2019; He et al., 2021). Researchers reported that lower levels of estradiol in postmenopausal women and men is accompanied by increase in Bacteroidetes sp. and depletion of Lactobacillus sp. Compared to women with higher levels of sex hormones (Santos-Marcos et al., 2018; Zhao et al., 2019). Likewise, the direct effect of the estradiol hormone on the growth of L. reuteri has been proved in our study. Treatment of sex steroid deficient mice with *Lactobacillus rhamnosus* could avoid bone loss which indicates the importance of this bacterium in preserving bone density (Li et al., 2016). The effect of sex steroids on gut bacteria in vivo could not be compared by their effect in vitro because, in some cases, hormonally related microbial shifts result from endogenous steroid-induced tissue and immunological changes rather than from steroids’ direct effect on bacteria (Lester and Hechter, 1958; Feraco et al., 2016). While studying the effect of L-thyroxine on the viability of gut microbiota, results revealed a significant increase in the growth of Gram-negative B. fragilis and E. coli and a significant reduction in the growth of Gram-positive B. longum and L. reuteri. The results withstood the previous findings by Garber and Lupowitz-Donenfeld [1973] that L-thyroxine had an inhibitory effect on the growth of Gram-positive bacteria however, it has no significant effect on Gram-negative bacteria. The inhibition of Gram-positive bacteria by L-thyroxine was reduced by cations such as Mn2+, Fe2+, and Ca2+ which support the hypothesis that L-thyroxine chelating effect is one of the factors contributing to its inhibitory effect (Garber and Lupowitz-Donenfeld, 1973; Benvenga et al., 2017). Metagenomic analysis of gut microbiome in hyperthyroidism patients showed as a significant decline in Bifidobacterium and Lactobacillus (Zhou et al., 2014). Similarly, reduction in these two genera was detected in our study under the influence of L-thyroxine which also explain the need for probiotic supplement in hypothyroidism patients treated with L-thyroxine to keep bone density and optimizing thyroid function (Knezevic et al., 2020). The process of bacterial adhesion to different surfaces is a complex process that involves contact between bacterial membranes and interacting surfaces. Specific and nonspecific binding are the two essentially distinct strategies that cause bacterial adhesion (Piette and Idziak, 1992). Electrostatic or hydrophobic interactions have a major role in the non-specific binding and considerably affect adhesion strength (Piette and Idziak, 1992). Two factors were assessed in this study to understand the influence of their changes in altering bacterial adherence in the presence of hormones. The auto-aggregation and cell surface hydrophobicity assays were carried out to evaluate the effect of the drugs on non-specific binding of these bacteria to different surfaces. Previous study showed that biofilm formation is correlated with hydrophobicity (Krasowska and Sigler, 2014) however this correlation was detected in L. reuteri under the effect of tested drugs where reduction in hydrophobicity was accompanied by the reduction in biofilm formation ability of this bacterium. Conversely, hydrophobicity was not correlated to bacteria adherence to Caco-2/HT-29 co-culture. Similarly, some previous studies showed that the correlation between hydrophobicity and adhesion to hydrophobic mucosal cells was strain specific (Kos et al., 2003; Muñoz-Provencio et al., 2009). This has suggested that nonspecific binding factors such as hydrophobicity is not an accurate measure of adhesive potential to enterocytes (Van Tassell and Miller, 2011) and interactions between microbes and hosts depend greatly on the structure of the cell surface rather than nonspecific binding (Nishiyama et al., 2021). Many studies have shown that, components of a protein nature such as mucus adhesins in L. reuteri (Vélez et al., 2007; Sánchez et al., 2008) and B. longum (Izquierdo et al., 2008) as well as capsular polymer in B. fragilis (Nakano et al., 2008; Reis et al., 2014) are primarily important for bacterial adherence to intestinal mucin types and/or epithelial cells beside saccharide moieties and lipoteichoic acid. A previous study linked hormonal drugs’ impact on bacterial growth with its impact on bacterial biofilm (Fteita et al., 2014). Similarly, in our results, a change in bacterial growth under the effect of the drug was associated with a similar change in biofilm formation except L. reuteri showing a reduction in biofilm in presence of sex steroids despite the increase in growth and this could be explained by the increased production of biosurfactant by Lactobacillus species in the presence of sex steroids thus decreasing biofilm formation (Clabaut et al., 2021). Additionally, the effect of progesterone on quorum sensing has been reported for some bacteria which could affect phenotypes that depend on cell-to-cell communication (Cadavid et al., 2018). The gut microbiota makes use of SCFAs as a source of carbon by cross-feeding, but SCFAs can be harmful to some gut bacterial species when present in high concentrations (Feng et al., 2018). The amount of SCFAs produced by the anaerobic bacteria B. fragilis and B. longum was measured using HPLC. Surprisingly we did not detect butyric acid after incubation of B. longum for 48 hours despite being known as butyric acid producer (Aguirre et al., 2016). Additionally, propionate was not detected in B. longum culture and similar result was obtained by Rios-Covian et al. [ 2017]. The effect of the tested drug compounds in this study on SCFAs production was variable and mostly independent of their effect on bacterial growth, in contrast to earlier studies’ results that a change in abundance of gut bacteria owing to drug usage was correlated to a similar effect on SCFAs production (Hojo et al., 2018). Notably, numerous studies reveal that Bacteroidetes are the main producers of propionate in the human gut (Salonen et al., 2014; Aguirre et al., 2016) and this SCFA is known for its ability to suppress inflammation and the increase in its level is associated with cognitive decline in elderly (Kawasoe et al., 2022; Neuffer et al., 2022) while low level of propionic acid and acetate was linked to autism(Tetel et al., 2018). Progesterone and L-thryoxine reduced the production of propionic acid by B. fragilis. A key organic acid in the fermentation of prebiotics is lactic acid, which is generated in the GIT by the bacteria Lactobacilli and Bifidobacteria. Lactic acid does not significantly accumulate in the intestinal lumen, and it is further metabolized by cross-feeding species, particularly with the butyrate-producing bacteria, to acetate or butyrate, or propionate. ## Conclusion Oral hormonal drugs can affect the growth, biofilm formation and adherence of gut bacteria at intestinal concentrations in vitro which can explain specific microbiome signatures associated with long term use of these drugs. Steroid hormones increase the growth of B. fragilis, E. coli, and L. reuteri while reducing the growth B. longum. However, thyroxine increased the growth of Gram-negative bacteria and reduced the growth of Gram-positive one. The effect of these drugs on biofilm formation by selected bacteria was linked to their effect on growth except L. reuteri where steroid hormones reduced biofilm formation by this bacterium despite increasing bacterial growth. The effect of hormonal drugs on bacterial adherence to Caco-2/HT-28 coculture was not solely dependent on the change in hydrophobicity but other specific binding factors might contribute to their effect. The effect of the tested drug compounds on SCFAs and lactic acid production was variable and mostly independent of their effect on bacterial growth. ## Data availability statement The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding author. ## Author contributions NA and RW conceptualized the work. ZH performed the experiments. NA, RW, and ZH analyzed the results. All authors contributed to the article and approved the submitted version. ## Conflict of interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ## Publisher’s note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. ## Supplementary material The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcimb.2023.1147585/full#supplementary-material ## References 1. Aguirre M., Eck A., Koenen M. E., Savelkoul P. H., Budding A. E., Venema K.. **Diet drives quick changes in the metabolic activity and composition of human gut microbiota in a validated**. *Res. Microbiol.* (2016) **167** 114-125. DOI: 10.1016/j.resmic.2015.09.006 2. Benvenga S., Di Bari F., Vita R.. **Undertreated hypothyroidism due to calcium or iron supplementation corrected by oral liquid levothyroxine**. *Endocrine* (2017) **56** 138-145. DOI: 10.1007/s12020-017-1244-2 3. Botes M., Loos B., van Reenen C. A., Dicks L. M.. **Adhesion of the probiotic strains enterococcus mundtii ST4SA and lactobacillus plantarum 423 to caco-2 cells under conditions simulating the intestinal tract, and in the presence of antibiotics and anti-inflammatory medicaments**. *Arch. Microbiol.* (2008) **190** 573-584. DOI: 10.1007/s00203-008-0408-0 4. Bustos A. Y., Font de Valdez G., Fadda S., Taranto M. P.. **New insights into bacterial bile resistance mechanisms: the role of bile salt hydrolase and its impact on human health**. *Food Res. Int.* (2018) **112** 250-262. DOI: 10.1016/j.foodres.2018.06.035 5. Cadavid E., Robledo S. M., Quiñones W., Echeverri F.. **Induction of biofilm formation in klebsiella pneumoniae ATCC 13884 by several drugs: the possible role of quorum sensing modulation**. *Antibiotics* (2018) **7** 103. DOI: 10.3390/antibiotics7040103 6. Clabaut M., Suet A., Racine P.-J., Tahrioui A., Verdon J., Barreau M.. **Effect of 17β-estradiol on a human vaginal lactobacillus crispatus strain**. *Sci. Rep.* (2021) **11** 7133. DOI: 10.1038/s41598-021-86628-x 7. Clark P. A., Martin J. H.. **Selection of bifidobacteria for use as dietary adjuncts in cultured dairy foods. III. tolerance to simulated bile concentrations of human small intestines**. *Cultured dairy products J.* (1994) **29** 18. DOI: 10.12938/bifidus.23.93 8. Coffey B. M., Anderson G. G., Filloux A., Ramos J.-L.. **Biofilm formation in the 96-well microtiter plate**. *Pseudomonas methods and protocols* (2014) 631-641 9. Crowley J., Withana M., Deplazes E.. **The interaction of steroids with phospholipid bilayers and membranes**. *Biophys. Rev.* (2022) **14** 163-179. DOI: 10.1007/s12551-021-00918-2 10. De Boever P., Wouters R., Verschaeve L., Berckmans P., Schoeters G., Verstraete W.. **Protective effect of the bile salt hydrolase-active lactobacillus reuteri against bile salt cytotoxicity**. *Appl. Microbiol. Biotechnol.* (2000) **53** 709-714. DOI: 10.1007/s002530000330 11. DeGruttola A. K., Low D., Mizoguchi A., Mizoguchi E.. **Current understanding of dysbiosis in disease in human and animal models**. *Inflammatory bowel Dis.* (2016) **22** 1137-1150. DOI: 10.1097/MIB.0000000000000750 12. DoĞAn A., Otlu S., Çelebi Ö., Aksu Kiliçle P., Gülmez Sağlam A., DoĞAn A.. **An investigation of antibacterial effects of steroids**. *Turkish J. Of Veterinary And Anim. Sci.* (2017) **41** 302-305. DOI: 10.3906/vet-1510-24 13. Dunne C., O'Mahony L., Murphy L., Thornton G., Morrissey D., O'Halloran S.. *Am. J. Clin. Nutr.* (2001) **73** 386s-392s. DOI: 10.1093/ajcn/73.2.386s 14. Feng W., Ao H., Peng C.. **Gut microbiota, short-chain fatty acids, and herbal medicines**. *Front. Pharmacol.* (2018) **9**. DOI: 10.3389/fphar.2018.01354 15. Feraco D., Blaha M., Khan S., Green J. M., Plotkin B. J.. **Host environmental signals and effects on biofilm formation**. *Microbial pathogenesis* (2016) **99** 253-263. DOI: 10.1016/j.micpath.2016.08.015 16. Fteita D., Könönen E., Söderling E., Gürsoy U. K.. **Effect of estradiol on planktonic growth, coaggregation, and biofilm formation of the prevotella intermedia group bacteria**. *Anaerobe* (2014) **27** 7-13. DOI: 10.1016/j.anaerobe.2014.02.003 17. Gagnon M., Zihler Berner A., Chervet N., Chassard C., Lacroix C.. **Comparison of the caco-2, HT-29 and the mucus-secreting HT29-MTX intestinal cell models to investigate salmonella adhesion and invasion**. *J. Microbiol. Methods* (2013) **94** 274-279. DOI: 10.1016/j.mimet.2013.06.027 18. Garber N., Lupowitz-Donenfeld B.. **The effect of thyroxine and triiodothyronine on bacterial growth**. *Canadian Journal of Microbiology* (1973) **19** 1401-1405. DOI: 10.1139/m73-226%M4203514 19. Gong B., Wang C., Meng F., Wang H., Song B., Yang Y.. **Association between gut microbiota and autoimmune thyroid disease: A systematic review and meta-analysis**. *Front. Endocrinol.* (2021) **12**. DOI: 10.3389/fendo.2021.774362 20. Hasan N., Yang H.. **Factors affecting the composition of the gut microbiota, and its modulation**. *PeerJ* (2019) **7** e7502-e7502. DOI: 10.7717/peerj.7502 21. He S., Li H., Yu Z., Zhang F., Liang S., Liu H.. **The gut microbiome and sex hormone-related diseases**. *Frontiers in Microbiology* (2021) **12**. DOI: 10.3389/fmicb.2021.711137 22. Heckman C., Kanagasundaram S., Cayer M., Paige J.. **Preparation of cultured cells for scanning electron microscope**. *Protocol Exchange* (2007). DOI: 10.1038/nprot.2007.504 23. Hojo M., Asahara T., Nagahara A., Takeda T., Matsumoto K., Ueyama H.. **Gut microbiota composition before and after use of proton pump inhibitors**. *Digestive Dis. Sci.* (2018) **63** 2940-2949. DOI: 10.1007/s10620-018-5122-4 24. Ibrahim S. A., Bezkorovainy A.. **Survival of bifidobacteria in the presence of bile salt**. *Journal of the Science of Food and Agriculture* (1993) **62** 351-354. DOI: 10.1002/jsfa.2740620407 25. Izquierdo E., Medina M., Ennahar S., Marchioni E., Sanz Y.. **Resistance to simulated gastrointestinal conditions and adhesion to mucus as probiotic criteria for bifidobacterium longum strains**. *Curr. Microbiol.* (2008) **56** 613-618. DOI: 10.1007/s00284-008-9135-7 26. Jang H.-I., Rhee K.-J., Eom Y.-B.. **Antibacterial and antibiofilm effects of α-humulene against bacteroides fragilis**. *Can. J. Microbiol.* (2020) **66** 389-399. DOI: 10.1139/cjm-2020-0004 27. Kawasoe J., Uchida Y., Kawamoto H., Miyauchi T., Watanabe T., Saga K.. **Propionic acid, induced in gut by an inulin diet, suppresses inflammation and ameliorates liver ischemia and reperfusion injury in mice**. *Frontiers in Immunol.* (2022) **13**. DOI: 10.3389/fimmu.2022.862503 28. Kleiveland C. R., Verhoeckx K., Cotter P., López-Expósito I.. **Co-cultivation of Caco-2 and HT-29MTX**. *The Impact of Food Bioactives on Health:* (2015). DOI: 10.1007/978-3-319-16104-4_13 29. Knezevic J., Starchl C., Tmava Berisha A., Amrein K.. **Thyroid-Gut-Axis: How does the microbiota influence thyroid function**. *Nutrients* (2020) **12** 1769. DOI: 10.3390/nu12061769 30. Kornman K. S., Loesche W. J.. **Effects of estradiol and progesterone on bacteroides melaninogenicus and bacteroides gingivalis**. *Infection Immun.* (1982) **35** 256-263. DOI: 10.1128/iai.35.1.256-263.1982 31. Kos B., Susković J., Vuković S., Simpraga M., Frece J., Matosić S.. **Adhesion and aggregation ability of probiotic strain lactobacillus acidophilus M92**. *J. Appl. Microbiol.* (2003) **94** 981-987. DOI: 10.1046/j.1365-2672.2003.01915.x 32. Krasowska A., Sigler K.. **How microorganisms use hydrophobicity and what does this mean for human needs**. *Front. Cell. Infection Microbiol.* (2014) **4**. DOI: 10.3389/fcimb.2014.00112 33. Kwasny S. M., Opperman T. J.. **Static biofilm cultures of gram-positive pathogens grown in a microtiter format used for anti-biofilm drug discovery**. *Curr. Protoc. Pharmacol.* (2010) **50** 13A.18.11-13A.18.23. DOI: 10.1002/0471141755.ph13a08s50 34. Lather P., Mohanty A. K., Jha P., Garsa A. K.. **Contribution of cell surface hydrophobicity in the resistance of staphylococcus aureus against antimicrobial agents**. *Biochem. Res. Int.* (2016) **2016**. DOI: 10.1155/2016/1091290 35. Lester G., Hechter O.. **Effect of deoxycorticosterone on the growth of microorganisms**. *J. Bacteriology* (1958) **76** 365-367. DOI: 10.1128/jb.76.4.365-367.1958 36. Letourneau J., Levesque C., Berthiaume F., Jacques M., Mourez M.. *J. Vis. Exp.* (2011) **51**. DOI: 10.3791/2783 37. Li J.-Y., Chassaing B., Tyagi A. M., Vaccaro C., Luo T., Adams J.. **Sex steroid deficiency–associated bone loss is microbiota dependent and prevented by probiotics**. *J. Clin. Invest.* (2016) **126** 2049-2063. DOI: 10.1172/JCI86062 38. Maier L., Pruteanu M., Kuhn M., Zeller G., Telzerow A., Anderson E. E.. **Extensive impact of non-antibiotic drugs on human gut bacteria**. *Nature* (2018) **555** 623-628. DOI: 10.1038/nature25979 39. Mueller H., Kassack M. U., Wiese M.. **Comparison of the usefulness of the MTT, ATP, and calcein assays to predict the potency of cytotoxic agents in various human cancer cell lines**. *J. biomolecular screening* (2004) **9** 506-515. DOI: 10.1177/1087057104265386 40. Muñoz-Provencio D., Llopis M., Antolín M., de Torres I., Guarner F., Pérez-Martínez G.. **Adhesion properties of lactobacillus casei strains to resected intestinal fragments and components of the extracellular matrix**. *Arch. Microbiol.* (2009) **191** 153-161. DOI: 10.1007/s00203-008-0436-9 41. Nakano V., Piazza R. M. F., Cianciarullo A., Bueris V., Santos M., Menezes M.. **Adherence and invasion of bacteroidales isolated from the human intestinal tract**. *Clin. Microbiol. infection* (2008) **14** 955-963. DOI: 10.1111/j.1469-0691.2008.02069.x 42. Neuffer J., González-Domínguez R., Lefèvre-Arbogast S., Low D. Y., Driollet B., Helmer C.. **Exploration of the gut-brain axis through metabolomics identifies serum propionic acid associated with higher cognitive decline in older persons**. *Nutrients* (2022) **14** 4688. DOI: 10.3390/nu14214688 43. Nicholson J. K., Holmes E., Kinross J., Burcelin R., Gibson G., Jia W.. **Host-gut microbiota metabolic interactions**. *Science* (2012) **336** 1262. DOI: 10.1126/science.1223813 44. Nishiyama K., Yokoi T., Sugiyama M., Osawa R., Mukai T., Okada N.. **Roles of the cell surface architecture of bacteroides and bifidobacterium in the gut colonization**. *Front. Microbiol.* (2021) **12**. DOI: 10.3389/fmicb.2021.754819 45. Nogal A., Valdes A. M., Menni C.. **The role of short-chain fatty acids in the interplay between gut microbiota and diet in cardio-metabolic health**. *Gut Microbes* (2021) **13** 1-24. DOI: 10.1080/19490976.2021.1897212 46. Piette J. P., Idziak E. S.. **A model study of factors involved in adhesion of pseudomonas fluorescens to meat**. *. Appl. Environ. Microbiol* (1992) **58** 2783-2791. DOI: 10.1128/aem.58.9.2783-2791.1992 47. Prakash S., Rodes L., Coussa-Charley M., Tomaro-Duchesneau C.. **Gut microbiota: next frontier in understanding human health and development of biotherapeutics**. *. Biologics: Targets Ther* (2011) **5** 71-86. DOI: 10.2147/BTT.S19099 48. Rahman M., Kim W.-S., Kumura H., Shimazaki K.-I.. **Autoaggregation and surface hydrophobicity of bifidobacteria**. *World J. Microbiol. Biotechnol* (2008) **24** 1593-1598. DOI: 10.1007/s11274-007-9650-x 49. Reddy S., Austin F.. **Adhesion and invasion assay procedure using caco-2 cells for listeria monocytogenes**. *Bio Protoc.* (2017) **7** e2267. DOI: 10.21769/BioProtoc.2267 50. Reis A. C., Silva J. O., Laranjeira B. J., Pinheiro A. Q., Carvalho C. B.. **Virulence factors and biofilm production by isolates of bacteroides fragilis recovered from dog intestinal tracts**. *Braz. J. Microbiol.* (2014) **45** 647-650. DOI: 10.1590/s1517-83822014000200037 51. Rinninella E., Raoul P., Cintoni M., Franceschi F., Miggiano G. A. D., Gasbarrini A.. **What is the healthy gut microbiota composition? a changing ecosystem across age, environment, diet, and diseases**. *Microorganisms* (2019) **7**. DOI: 10.3390/microorganisms7010014 52. Rios-Covian D., Salazar N., Gueimonde M., de Los Reyes-Gavilan C. G.. **Shaping the metabolism of intestinal bacteroides population through diet to improve human health**. *Front. Microbiol.* (2017) **8**. DOI: 10.3389/fmicb.2017.00376 53. Salonen A., Lahti L., Salojärvi J., Holtrop G., Korpela K., Duncan S. H.. **Impact of diet and individual variation on intestinal microbiota composition and fermentation products in obese men**. *ISME J.* (2014) **8** 2218-2230. DOI: 10.1038/ismej.2014.63 54. Sánchez E., Nadal I., Donat E., Ribes-Koninckx C., Calabuig M., Sanz Y.. **Reduced diversity and increased virulence-gene carriage in intestinal enterobacteria of coeliac children**. *BMC Gastroenterol.* (2008) **8**. DOI: 10.1186/1471-230X-8-50 55. Santos-Marcos J. A., Rangel-Zuñiga O. A., Jimenez-Lucena R., Quintana-Navarro G. M., Garcia-Carpintero S., Malagon M. M.. **Influence of gender and menopausal status on gut microbiota**. *Maturitas* (2018) **116** 43-53. DOI: 10.1016/j.maturitas.2018.07.008 56. Scotti E., Boué S., Sasso G. L., Zanetti F., Belcastro V., Poussin C.. **Exploring the microbiome in health and disease: Implications for toxicology**. *Toxicol. Res. Appl.* (2017) **1** 2397847317741884. DOI: 10.1177/2397847317741884 57. Shin J.-H., Park Y.-H., Sim M., Kim S.-A., Joung H., Shin D.-M.. **Serum level of sex steroid hormone is associated with diversity and profiles of human gut microbiome**. *Res. Microbiol.* (2019) **170** 192-201. DOI: 10.1016/j.resmic.2019.03.003 58. Silva Y. P., Bernardi A., Frozza R. L.. **The role of short-chain fatty acids from gut microbiota in gut-brain communication**. *Front. Endocrinol.* (2020) **11**. DOI: 10.3389/fendo.2020.00025 59. Souza C., Barbosa C. D., Coelho H. I., Santos Júnior M. N., Barbosa E. N., Queiroz E. C.. **Effects of 17β-estradiol on Monocyte/Macrophage response to staphylococcus aureus: An in vitro study**. *Front. Cell. infection Microbiol.* (2021) **11**. DOI: 10.3389/fcimb.2021.701391 60. Staley C., Weingarden A. R., Khoruts A., Sadowsky M. J.. **Interaction of gut microbiota with bile acid metabolism and its influence on disease states**. *Appl. Microbiol. Biotechnol.* (2017) **101** 47-64. DOI: 10.1007/s00253-016-8006-6 61. Tetel M. J., de Vries G. J., Melcangi R. C., Panzica G., O'Mahony S. M.. **Steroids, stress and the gut microbiome-brain axis**. *J. Neuroendocrinol* (2018) **30**. DOI: 10.1111/jne.12548 62. Thursby E., Juge N.. **Introduction to the human gut microbiota**. *Biochem. J.* (2017) **474** 1823-1836. DOI: 10.1042/bcj20160510 63. Ude V. C., Brown D. M., Stone V., Johnston H. J.. **Using 3D gastrointestinal tract**. *J. Nanobiotechnology* (2019) **17** 70. DOI: 10.1186/s12951-019-0503-1 64. Valdes-Varela L., Hernandez-Barranco A. M., Ruas-Madiedo P., Gueimonde M.. **Effect of bifidobacterium upon clostridium difficile growth and toxicity when Co-cultured in different prebiotic substrates**. *Front. Microbiol.* (2016) **7**. DOI: 10.3389/fmicb.2016.00738 65. Van Tassell M. L., Miller M. J.. **Lactobacillus adhesion to mucus**. *Nutrients* (2011) **3** 613-636. DOI: 10.3390/nu3050613 66. Vélez M. P., De Keersmaecker S. C., Vanderleyden J.. **Adherence factors of lactobacillus in the human gastrointestinal tract**. *FEMS Microbiol. Lett.* (2007) **276** 140-148. DOI: 10.1111/j.1574-6968.2007.00908.x 67. Vich Vila A., Collij V., Sanna S., Sinha T., Imhann F., Bourgonje A. R.. **Impact of commonly used drugs on the composition and metabolic function of the gut microbiota**. *Nat. Commun.* (2020) **11** 362. DOI: 10.1038/s41467-019-14177-z 68. Vollaro A., Esposito A., Antonaki E., Iula V. D., D'Alonzo D., Guaragna A.. **Steroid derivatives as potential antimicrobial agents against staphylococcus aureus planktonic cells**. *Microorganisms* (2020) **8** 468. DOI: 10.3390/microorganisms8040468 69. Wang J., Woo M., Yan C.. **Spot plating assay for the determination of survival and plating efficiency of escherichia coli in sub-MIC levels of antibiotics**. *JEMI Methods* (2017) **1** 26-29 70. Wen L., Duffy A.. **Factors influencing the gut microbiota, inflammation, and type 2 diabetes**. *J. Nutr.* (2017) **147** 1468s-1475s. DOI: 10.3945/jn.116.240754 71. Woo S.-G., Lee S.-M., Lee S.-Y., Lim K.-H., Ha E.-J., Kim S.-H.. **The effectiveness of anti-biofilm and anti-virulence properties of dihydrocelastrol and dihydrocelastryl diacetate in fighting against methicillin-resistant staphylococcus aureus**. *Arch. Microbiol.* (2017) **199** 1151-1163. DOI: 10.1007/s00203-017-1386-x 72. Zhao H., Chen J., Li X., Sun Q., Qin P., Wang Q.. **Compositional and functional features of the female premenopausal and postmenopausal gut microbiota**. *FEBS letters* (2019) **593** 2655-2664. DOI: 10.1002/1873-3468.13527 73. Zhou L., Li X., Ahmed A., Wu D., Liu L., Qiu J.. **Gut microbe analysis between hyperthyroid and healthy individuals**. *Curr. Microbiol.* (2014) **69** 675-680. DOI: 10.1007/s00284-014-0640-6
--- title: Digital spatial profiling of human parathyroid tumors reveals cellular and molecular alterations linked to vitamin D deficiency authors: - Chia-Ling Tu - Wenhan Chang - Julie A Sosa - James Koh journal: PNAS Nexus year: 2023 pmcid: PMC10042281 doi: 10.1093/pnasnexus/pgad073 license: CC BY 4.0 --- # Digital spatial profiling of human parathyroid tumors reveals cellular and molecular alterations linked to vitamin D deficiency ## Abstract Primary hyperparathyroidism (PHPT) is a common endocrine neoplastic disorder characterized by disrupted calcium homeostasis secondary to inappropriately elevated parathyroid hormone (PTH) secretion. Low levels of serum 25-hydroxyvitamin D (25OHD) are significantly more prevalent in PHPT patients than in the general population (1–3), but the basis for this association remains unclear. We employed a spatially defined in situ whole-transcriptomics and selective proteomics profiling approach to compare gene expression patterns and cellular composition in parathyroid adenomas from vitamin D-deficient or vitamin D-replete PHPT patients. A cross-sectional panel of eucalcemic cadaveric donor parathyroid glands was examined in parallel as normal tissue controls. Here, we report that parathyroid tumors from vitamin D-deficient PHPT patients (Def-Ts) are intrinsically different from those of vitamin D-replete patients (Rep-Ts) of similar age and preoperative clinical presentation. The parathyroid oxyphil cell content is markedly higher in Def-Ts ($47.8\%$) relative to Rep-Ts ($17.8\%$) and normal donor glands ($7.7\%$). Vitamin D deficiency is associated with increased expression of electron transport chain and oxidative phosphorylation pathway components. Parathyroid oxyphil cells, while morphologically distinct, are comparable to chief cells at the transcriptional level, and vitamin D deficiency affects the transcriptional profiles of both cell types in a similar manner. These data suggest that oxyphil cells are derived from chief cells and imply that their increased abundance may be induced by low vitamin D status. Gene set enrichment analysis reveals that pathways altered in Def-Ts are distinct from Rep-Ts, suggesting alternative tumor etiologies in these groups. Increased oxyphil content may thus be a morphological indicator of tumor-predisposing cellular stress. ## Introduction Low levels of serum 25-hydroxyvitamin D (25OHD) are commonly associated with primary hyperparathyroidism (PHPT) (1–3). Although parathyroid hormone (PTH) levels are known to be increased in PHPT patients with vitamin D deficiency (<20 ng/ml), the underlying pathophysiological basis for this relationship remains poorly understood (4–6). Downstream systemic effects driven by elevated PTH may enhance renal conversion of 25OHD to [1,25(OH)2D] with subsequent suppression of 25OHD production in the skin and liver [2, 7], but the possibility that chronic 25OHD deficiency can itself initiate changes in parathyroid tissue that predispose to adenoma development and PHPT remains an open question. Although treatment with vitamin D analogues can inhibit PTH transcription and cellular proliferation in cultured bovine parathyroid cells [8, 9], genetic ablation of the vitamin D receptor (VDR) in the parathyroid glands of transgenic mice did not induce gland hyperplasia and only modestly increased serum PTH [10], leaving the potential direct effects of 25OHD as a driver of PHPT biology largely uncertain. To investigate the potential downstream actions of 25OHD on parathyroid tissue, we performed a comparative molecular analysis of parathyroid adenomas derived from PHPT patients with preoperative vitamin D deficiency relative to tumors from vitamin D-replete PHPT patients and normal organ donor parathyroid glands. The parathyroid glands of healthy adults are comprised predominantly of chief cells, but in conditions such as secondary hyperparathyroidism (SHPT) caused by chronic kidney disease, the relative abundance of a second parathyroid cell type known as oxyphils can increase dramatically [11, 12]. Parathyroid oxyphils are characterized by high mitochondrial content and appear more abundant in older individuals [13], but the origin and physiological functions of these cells are unknown. Previously, we and others have shown that similar to chief cells, parathyroid oxyphil cells can respond to extracellular calcium stimulation via calcium sensing receptor (CaSR)-dependent signal transduction [14], express multiple parathyroid-specific factors including GCM2 and PTH [11, 15], and contain key regulatory factors such as the vitamin D-receptor, Klotho, and mitochondrial components involved in cellular respiration [16]. Because oxyphil-dominant parathyroid tumors appear to drive biochemically more severe disease presentation in PHPT (17–21) and oxyphilic hyperplasia is associated with the loss of calcimimetic responsiveness in SHPT [12], we hypothesized that chronic 25OHD deficiency could produce changes in parathyroid oxyphil content and gene expression that reduce calcium sensitivity, increase PTH hypersecretion, and promote adenoma development in PHPT. To test this idea, we employed a novel, spatially indexed approach to isolate and capture oxyphil and chief cells separately from parathyroid tissue sections for subsequent transcriptomic and proteomic comparative analysis. Here, we report that parathyroid tumors from PHPT patients presenting with preoperative vitamin D deficiency (Def-Ts) are molecularly distinct from vitamin D-replete patient tumors (Rep-Ts). The oxyphil content is markedly increased in Def-Ts, and both chief and oxyphil cells from these tumors share a discrete transcriptional signature enriched for genes involved in oxidative phosphorylation, cellular respiration, and proteasomal catabolism. In contrast, Rep-Ts are more heterogeneous and are characterized by upregulation of the ras and myc signaling pathways, suggestive of an oncogene-driven etiology. In either vitamin D context, chief and oxyphil cells are highly similar at the transcriptome level, supporting the notion that oxyphil cells are derivatives of chief cells rather than arising through an independent lineage. Interestingly, PTH transcript abundance is equivalent between normal tissue, Def-Ts, and Rep-Ts, indicating that aberrant tumor-specific PTH gene expression is not a driver of hormonal hypersecretion in PHPT. Consistent with this finding, proteomic analysis revealed that the ratio of CaSR to the type B γ-aminobutyric acid receptor (GABBR) is altered in favor of biochemically silent CaSR:GABBR heteromers [22] in Def-Ts, suggesting that CaSR antagonism could contribute to PTH hypersecretion uncoupled from calcium sensing in these tumors. Finally, a comparison between PHPT adenomas and normal parathyroid tissue independent of vitamin D status revealed that genes involved in cytoskeletal structure and tissue remodeling were significantly upregulated in the tumors, suggesting that the ability to modify cellular structure and the physical tumor microenvironment is a common denominator, acquired phenotype in PHPT adenomas. These data suggest that PHPT adenomas share certain features related to cellular structure and tissue remodeling but that vitamin D status strongly influences the intrinsic gene expression patterns and biochemical behavior of the tumors. Collectively, these results indicate that oxyphil differentiation in Def-Ts may be an indicator of metabolic stress and that vitamin D deficiency can induce gene expression changes that uncouple calcium sensing from PTH secretion. ## Parathyroid adenomas from vitamin D-deficient PHPT patients display increased oxyphil cell abundance To determine whether the preoperative vitamin D status of PHPT patients was associated with specific changes in the cellular composition and transcriptional profile of their parathyroid adenomas, we assembled a cohort of study subjects and normal donor controls for comparative analysis (Table 1). The PHPT patients were drawn from a preexisting registry of female study subjects who underwent parathyroidectomy at our institution from 2018 to the present. Nine patients whose preoperative vitamin D levels met the Institute of Medicine definition [23] of vitamin D deficiency (25OHD ≤ 20 ng/ml) were selected, along with 34 patients who presented with a range of preoperative 25OHD levels above the vitamin D-replete threshold (25OHD ≥ 30 ng/ml). Preoperative vitamin D levels for these patients were confirmed by repeat reads of blood collected intraoperatively from each study participant. Control parathyroid glands from 12 eucalcemic organ donors were evaluated in parallel as normal reference benchmarks. The vitamin D-deficient and vitamin D-replete groups were similar in age ($$P \leq 0.1814$$), preoperative serum calcium ($$P \leq 0.9580$$), and resected gland mass ($$P \leq 0.9274$$), and all had single gland adenomas. Preoperative PTH differed significantly as has been previously reported [24], with a median of 188 pg/ml in the vitamin D-deficient group compared to 112.5 pg/ml in the vitamin D-replete group ($$P \leq 0.0019$$ by two-tailed t-test). The age of the normal donor controls was significantly younger than either of the PHPT cohorts ($$P \leq 0.001$$ by ANOVA). **Table 1.** | Parameter | Normal donor (n = 12) | Vitamin D-deficient PHPT (n = 9) | Vitamin D-replete PHPT (n = 34) | | --- | --- | --- | --- | | Age (yr) | 38 (30,74) | 63 (53,71) | 69 (59,74) | | Sex | 8 females/4 males | 9 females | 34 females | | Pre-op. PTH (pg/ml) | | 188 (120.5, 240)a | 112.5 (87.5, 141.8)a | | Pre-op. serum Ca++ (mg/dl) | | 10.8 (10.45, 11.10) | 10.7 (10.4, 11.1) | | Gland weight (mg) | | 485 (264, 686) | 486 (204, 695) | | Pre-op. vitamin D (ng/ml) | | 17 (10, 18) | 38 (33, 46.5) | To investigate the relative abundance of chief and oxyphil cells in the adenomas from each vitamin D group, formalin-fixed, paraffin-embedded sections were prepared and examined by immunofluorescence. To exclude vascular components, adipocytes, and other nonparathyroid cell types, the sections were stained for the calcium sensing receptor (CaSR), since elevated abundance of this protein is a recognized hallmark of parathyroid cells. The sections were costained for TOMM20 (translocase of outer mitochondrial membrane 20), a mitochondrial marker that is highly enriched in parathyroid oxyphil cells. To maintain consistent staining conditions for comparative purposes, each glass slide included tissue sections from a normal parathyroid gland (Fig. 1A, NL), an adenoma from a vitamin D-replete patient (Fig. 1A, REP), and an adenoma from a vitamin D-deficient patient (Fig. 1A, DEF). The entire specimen area of each slide was scanned with a 20× objective in a GeoMx Digital Spatial Profiler (DSP). The proportions of CaSR-positive cells expressing high (oxyphil cells) versus low (chief cells) levels of the TOMM20 marker were then quantified by ImageJ and expressed as a fraction of the total cell number determined by SYTO13-positive nuclear counts of the CaSR-positive population. **Fig. 1.:** *Parathyroid adenomas from PHPT patients with preoperative vitamin D deficiency contain a higher proportion of oxyphil cells. (A) The relative abundance of oxyphil cells was evaluated by immunofluorescence in tissue sections from normal donor parathyroid tissue (NL) and parathyroid adenomas from vitamin D-deficient (DEF) or vitamin D-replete (REP) PHPT patients. CaSR is stained in red to identify parathyroid cells; the mitochondrial biogenesis protein TOMM20, a marker of mitochondria-rich oxyphil cells, is stained in green. Scale bar = 0.25 mm. (B) The relative abundance of chief (blue) or oxyphil cells (red) were quantitated in a panel of normal (n = 12) or adenoma tissue sections from vitamin D-deficient (n = 9) or vitamin D-replete (n = 34) PHPT patients. Percentages are based upon cell counts from full-section low-power fields for each tumor.* As expected, normal parathyroid tissue contained the lowest proportion of oxyphil cells ($7.7\%$), likely due in part to the younger age of the donor group, as the oxyphil content is known to increase with age [25]. In the age-matched PHPT patient cohorts, Def-Ts had significantly greater oxyphil abundance (Fig. 1B), accounting for nearly half ($47.8\%$) of the cellular content of the tumors on average. In contrast, oxyphil abundance was much lower in Rep-Ts ($13.9\%$). The oxyphil cells in Def-Ts tended to occur in large, contiguous areas as opposed to the more scattered pockets or isolated cells observed in normal tissue or in Rep-Ts. ## Transcriptional spatial profiling reveals distinct gene signatures in parathyroid adenomas from PHPT patients with preoperative vitamin D deficiency We then sought to determine whether the gene expression profiles of Def-Ts and Rep-Ts were similar or distinct. Because the oxyphil content of these tumor groups was clearly different, we employed a digital spatial profiling (DSP) approach to capture chief and oxyphil cells separately for direct transcriptional comparisons between cell types from each tissue specimen. Multiple regions of interest (ROIs) encompassing both cell types were selected from each formalin-fixed, paraffin-embedded tissue section, and CaSR+/TOMM20-high (oxyphil cells) and CaSR+/TOMM20-low (chief cells) were marked for selective capture (Fig. 2A). A GeoMx DSP instrument (NanoString Technologies) was used to interrogate gene expression within the chief or oxyphil cells separately captured from each ROI. **Fig. 2.:** *Transcriptional spatial profiling reveals discrete gene signatures in parathyroid adenomas from PHPT patients with preoperative vitamin D deficiency. (A) Tissue sections were stained with Syto13 (blue) to visualize nuclei and anti-TOMM20 (green) to identify oxyphil cells. The upper panels are from a tumor from a vitamin D-replete patient. The lower panels are from a vitamin D-deficient patient. Regions of interests (ROIs) were selected (white circles) that included both chief (TOMM2−) and oxyphil (TOMM20+) cells. Chief and oxyphil cells were captured separately from each other in each ROI using the indicated masks (teal = oxyphil cells; purple = chief cells). Scale bar = 0.25 mm. (B) Distribution of total Q3 normalized, log2 transformed counts per gene by the tissue group (left) or by cell type (right). Upper and lower box boundaries represent the 75th and 25th percentiles of each data group. The white horizontal line indicates the median, and error bars indicate the standard deviation. (C) Principal component analysis visualized in a three-dimensional tSNE plot. Green symbols = normal parathyroid tissue. Blue symbols = tumors from vitamin D-replete patients. Orange symbols = tumors from vitamin D-deficient patients. Circles = chief cells. Squares = oxyphil cells.* Utilizing a human whole transcriptome oligonucleotide library and next-generation sequencing, we performed a quantitative analysis of over 18,000 unique transcripts in the targeted cells captured from each ROI. Quality control metrics including target gene saturation, total transcript counts, and mean counts per transcript were evaluated for each tissue source (normal, Def-Ts, or Rep-Ts) and for each cell type (chief vs oxyphil) (Table S1, Fig. S1). No significant differences in these metrics were observed between input groups. For example, the mean counts per transcript did not vary when comparing either tissue source or cell type (Fig. 2B). The raw sequencing counts were normalized through the Q3 (third quartile of all targets above the limit of quantitation) method [26], using the top $25\%$ of expressors to normalize across all ROIs and captured cell subsets. The normalized data were then subjected to principal component analysis to visualize gene expression effects associated with the cell type or vitamin D status. In a three-dimensional t-distributed stochastic neighbor embedding (tSNE) plot, both chief and oxyphil cells from Def-Ts formed a discrete cluster widely separated from normal parathyroid tissue and from Rep-Ts (Fig. 2C). Two-dimensional projections from UMAP and PCA modeling generated similar results (Fig. S2). Compared to cells from Def-Ts, chief and oxyphil cells from Rep-Ts are grouped more loosely and, in some cases, appeared to be closely related to normal tissue. Notably, both chief and oxyphil cells segregated by tissue source rather than by cell type. In one specific case, the chief cell and oxyphil cell inputs from a single Def-T appeared as outliers grouping more closely to vitamin D-replete tumor tissue; this patient's preoperative vitamin D level was the highest in the deficient group (25OHD = 18.2 ng/ml). ## Unsupervised hierarchical cluster analysis identifies gene signatures associated with vitamin D status in parathyroid tumors To visualize transcriptome profiles potentially associated with vitamin D status, the 12,762 unique transcripts detected were rank ordered by the degree of differential expression, and unsupervised hierarchical cluster analysis was performed using the ComplexHeatmap R/Bioconductor package. Genes encoded on the X and Y chromosomes were excluded to remove sex as a differentiating variable, as four of the normal tissue donors were male. Consistent with the PCA results, the interrogated samples segregated by vitamin D status, with normal parathyroid tissue clustering separately from both tumor groups (Fig. 3A). The cell type did not emerge as a primary organizing variable, as chief cells and oxyphil cells both clustered within each tissue group rather than segregating independently. A total of six transcriptional profile clusters emerged: two from normal parathyroid tissue (clusters 1 and 2), three within Rep-Ts (clusters 3, 4, and 5), and two (clusters 5 and 6) within the Def-Ts. The chief and oxyphil cells from the one outlier vitamin D-deficient patient sample noted in the PCA plot sorted into cluster 5, a profile shared with eight Rep-T specimens. Cluster 6 represented the predominant signature of the Def-Ts, encompassing tumors from eight of the nine vitamin D-deficient patients. The two clusters from normal tissue each contained both chief and oxyphil cells. The tumors from the Rep-Ts were the most heterogeneous, with three different profile clusters identified. Strikingly, the three clusters visualized among Rep-Ts correlated with the mean preoperative vitamin D levels of the patients in each group (Fig. 3B), demonstrating a potential dose-dependent relationship between tumor transcriptional profile patterns and vitamin D status. The differences in the mean preoperative 25OHD levels of the patients in each Rep-T transcriptional cluster were highly significant ($P \leq 0.0001$ by ANOVA). Cluster 6, the Def-T group which by definition had the lowest mean 25OHD level, segregates as a distinct transcriptional profile and does not appear to be closely related to any of the three Rep-T profiles. **Fig. 3.:** *Digital spatial profiling reveals distinct transcriptional pathway changes in parathyroid tumors from vitamin D-replete and vitamin D-deficient PHPT patients. (A) Q3 normalized counts were compared across normal parathyroid tissue and parathyroid adenomas from vitamin D-deficient or vitamin D-replete PHPT patients and analyzed by unsupervised two-way hierarchical clustering. The DSP heatmap shows the top 150 differentially expressed genes with signature pattern relationships shown on the left of the heatmap. (B) Preoperative vitamin D levels show dosage effects (P < 0.0001 by ANOVA) correlating with gene expression clusters of PHPT patients. Serum 25OHD mean and standard deviations for each cluster group are shown, with each dot representing an individual patient value. The dotted line demarcates 20 ng/ml, the established threshold for vitamin D deficiency.* Pathway analysis of the DSP data revealed that genes associated with oxidative phosphorylation, mitochondrial electron transport chain function, and the citrate TCA cycle were significantly upregulated in both chief and oxyphil cells from the Def-Ts relative to normal tissue. Surprisingly, oxyphil cells from normal tissue and from Rep-Ts did not display upregulation of these same pathways despite the high mitochondrial content of these cells. Relative to Def-Ts, Rep-Ts were more heterogeneous and demonstrated a greater degree of overlap with normal tissue expression patterns. While vitamin D deficiency is associated with a distinct pattern of gene expression, the correlation between Rep-T clusters and preoperative vitamin D levels in the absence of a readily apparent unifying profile suggests that additional variables beyond vitamin D status contribute to transcriptomic heterogeneity in PHPT adenomas. To evaluate potential functional differences between the groups, gene set enrichment analysis was performed [27]. Annotated databases from the curated Gene Ontology/Biological Processes (GOBP), Kyoto Encyclopedia of Genes and Genomes (KEGG), and GSEA Molecular Signatures Database (MSigDB) repositories were queried for association with the transcriptome profiles revealed in the hierarchical cluster analysis. Four phenotype-associated signatures of differentially expressed genes were identified (Table 2). Rep-Ts were preferentially associated with activation of Myc target genes (MSig DB: M5926, correlation coefficient 0.811), genes regulated in response to amyloid beta (GOBP: 1904645, correlation coefficient 0.747), and genes activated by K-ras signaling (MSigDB: M5953, correlation coefficient 0.632). In contrast, Def-Ts were strongly associated with genes involved in oxidative phosphorylation (MSigDB: M5936, correlation coefficient 0.942) and to a lesser extent genes linked to proteasomal catabolism (GOBP: 0010499, correlation coefficient 0.587) and the TCA cycle (KEGG: M3985, correlation coefficient 0.546). Genes differentially expressed in normal parathyroid tissue were strongly associated with downregulation of K-ras signaling (MSigDB: M5956, correlation coefficient 0.942) and regulation of calcium transport (GOBP: 0051924, correlation coefficient 0.545). A fourth signature shared by both normal tissue and Rep-Ts was most closely associated with genes involved in fat-soluble vitamin metabolic processes (GOBP:0006775, correlation coefficient 0.570). **Table 2.** | Signature | Sample group | Gene set | Gene set ID | Correlation coefficient | | --- | --- | --- | --- | --- | | S1 | 25OHD-replete PHPT | Myc targets | MSigDB: M5926 | 0.811 | | | | Response to amyloid beta | GOBP: 1904645 | 0.747 | | | | K-ras signaling UP | MSigDB: M5953 | 0.632 | | S2 | Normal parathyroid | K-ras signaling DOWN | MSigDB: M5956 | 0.942 | | | | Regulation of Ca++ transport | GOBP:0051924 | 0.545 | | S3 | 25OHD-deficient PHPT | Oxidative phosphorylation | MSigDB:M5936 | 0.982 | | | | Proteasomal catabolism | GOBP: 0010499 | 0.587 | | | | TCA cycle | KEGG:M3985 | 0.546 | | S4 | Normal, replete | Fat soluble vitamin metabolic processes | GOBP: 0006775 | 0.57 | ## Chief and oxyphil cells share highly similar transcriptional profiles, while vitamin D status is associated with gene expression changes that affect both cell types To identify individual genes with the greatest degree of differential expression between cell types and vitamin D status groups, we performed a series of pairwise comparisons employing the DESeq Wald test, edgeR quasi-likelihood F test, and limma.zoom, each implemented as Bioconductor modules in R (28–30). When comparing chief cells to oxyphils, either collectively or considering the two cell types in each vitamin D status group separately, no genes met a differentially expressed gene (DEG) threshold of log2(fold change) > ±1 and q < 1e−6 (Fig. 4A). In contrast, when comparing Def-Ts to Rep-Ts, 26 DEGs were detected that met the same differential expression criteria (Fig. 4B). The DEGs, their fold change in the Def-Ts relative to the Rep-Ts, and the corresponding q-values are listed in Table 3. EPB41L3 (erythrocyte membrane protein band 4.1-like 3), a cytoskeleton protein–membrane anchor with suspected tumor suppressor properties [31, 32], showed the greatest fold change, with a mean fold increase of almost 16-fold in Def-Ts. *This* gene has been found to be upregulated in benign meningiomas, while its loss by mutational inactivation or gene silencing has been associated with enhanced invasiveness and malignant transformation in gastric and colorectal cancers [32]. Expression of MAPK8IP1 (mitogen-activated protein kinase 8 interacting protein 1), a key regulatory protein that opposes MAPK8-mediated activation of downstream transcription factors and colocalizes with amyloid deposits in the neurofibrillary tangles of Alzheimer's disease patients [33], is significantly diminished in Def-Ts. Notably, expression of GABBR1 (gamma aminobutyric acid type B receptor 1), which our group has recently shown to oppose CaSR-mediated calcium sensing in the parathyroid [34], is significantly elevated in Def-Ts. *Additional* genes found to be significantly upregulated in these tumors act in signaling pathways controlling apoptosis (TNFSF10, tumor necrosis superfamily member 10) [35, 36], cellular stress response (SGK1, serum glucocorticoid regulated kinase 1) [37], inositol-derived second messenger production (IPMK, inositol polyphosphate multikinase) [38], and energetic metabolism (AMPD3, adenosine monophosphate deaminase 3) [39]. Despite the higher preoperative PTH observed in vitamin D-deficient PHPT patients, PTH transcript abundance was not significantly different between tissue sources ($$P \leq 0.2862$$ by one-way ANOVA) or when comparing chief to oxyphil cells ($$P \leq 0.7284$$ by unpaired t-test). These data suggest that alterations in secretory or sensing mechanisms, rather than aberrant overexpression of PTH transcription, are the primary drivers of PTH hypersecretion in PHPT. **Fig. 4.:** *(A) Chief and oxyphil cells share highly similar gene expression patterns; (B) vitamin D status is associated with gene expression changes that affect both cell types. Volcano plots depict the log2(fold change) in individual gene expression between the indicated comparison groups on the x-axis and log10(adjusted statistical significance) on the y-axis. Genes highlighted in blue had a log2(fold change) value of greater than 1 and a false discovery rate (FDR) significance threshold of <1e−06.* TABLE_PLACEHOLDER:Table 3. ## Increased GABBR protein abundance in tumors from vitamin D-deficient PHPT patients favors the formation of biochemically inactive CaSR/GABBR heteromers The modest but statistically significant increase in GABBR1 transcript abundance in Def-Ts suggested that the stoichiometric balance between active, calcium-responsive CaSR:CaSR homomers relative to signaling-attenuated CaSR:GABBR heteromers [22] might be shifted in favor of the inactive complex in Def-Ts. To assess this notion, we utilized the DSP platform to determine the protein abundance of CaSR, GABBR1, and GABBR2 in normal parathyroid tissue sections and in Def-Ts or Rep-Ts. The protein abundance of both GABBR variants was higher in Def-Ts, while CaSR abundance in these tumors was consistently lower (Fig. 5A). These changes resulted in a significantly reduced ratio of CaSR to GABBR proteins in both chief and oxyphil cells from vitamin D-deficient patients, while Rep-Ts were indistinguishable from normal tissue in this assay (Fig. 5B). **Fig. 5.:** *Inactive CASR/GABBR heteromer formation is favored in tumors from vitamin D-deficient PHPT patients. (A) Protein abundance of CASR, GABBR1, and GABBR2 was determined in situ from parathyroid tissue sections using a NanoString DSP instrument and a custom conjugated antibody panel. The relative abundance of the individual proteins and the ratio of CASR to the GABBA receptor proteins were determined in sections from vitamin D-deficient patient tumors (DEF), vitamin D-replate patient tumors (REP), or normal donor parathyroid tissue (NORMAL). (B) The ratio of CaSR protein to GABBR proteins was determined in the indicated cell types and tissue sources. Green bars = normal tissue. Orange bars = tumors from vitamin D-deficient patients. Blue bars = tumors from vitamin D-replete patients. Hashmarked bars = chief cells. Plain bars = oxyphil cells. Values shown are the means ± standard deviation for each sample group. n = 6 (normals), n = 13 (replete), and n = 6 (deficient).* ## Parathyroid adenomas as a group express higher levels of genes involved in tissue remodeling and cytoskeletal function When we compared all the parathyroid adenomas in our cohort as a group to normal parathyroid tissue using the same DEG criteria described above, twelve genes emerged as differentially expressed (Fig. 6A). *Ten* genes were upregulated in the tumors relative to normal tissue, while two genes were downregulated relative to normal tissue (Fig. 6B). The top three (ranked by q-value) and six of the 10 upregulated genes have roles in cellular structure and tissue architecture. COL6A6 (collagen type VI alpha 6 chain) is a component of the basal lamina of epithelial cells and plays a central role in maintaining extracellular matrix structure and function [40]. PLAT (tissue-type plasminogen activator) is a secreted serine protease whose enzymatic action is essential for cell migration and tissue remodeling [41]. AFAP1L2 (actin filament-associated protein 1-like 2) is an adaptor protein whose elevated expression is associated with the epithelial–mesenchymal transition, cellular migration, and tissue repair (42–44). EPB41L3, cited earlier, is a cytoskeletal protein anchor, and ALCAM (activated leukocyte cell adhesion molecule) has been shown to play an important role in invasive cellular behavior, mesenchymal stromal cell activity, and extracellular vesicular trafficking (45–47). TOX2 (TOX high-mobility group box family 2) is a transcriptional coactivator that modulates multiple pathways including tissue remodeling and tumor microenvironment functions [48]. One of the two downregulated genes, IGFBP5 (insulin-like growth factor binding protein 5), is a key regulator of osteogenic differentiation, and agents that antagonize its expression have been shown to promote osteoporosis [49, 50]. **Fig. 6.:** *Genes differentially expressed between parathyroid adenomas and normal parathyroid tissue. (A) Volcano plot of differentially expressed genes, depicted as described in Fig. 4. (B) Differentially expressed genes listed in descending order by q-value, with the log2(fold change) values representing expression in tumor tissue relative to normal tissue.* ## Discussion The current paradigm for explaining the well-established association between low vitamin D levels and PHPT posits that 25OHD hypovitaminosis is a consequence of constitutively elevated PTH. Multiple mechanisms to support this idea have been proposed, including PTH-mediated suppression of 25OHD synthesis, shortened serum half-life, and restricted bioavailability (2, 51–53). Here, we have explored an alternative viewpoint by investigating whether vitamin D deficiency can exert downstream effects on parathyroid tissue, testing the hypothesis that low vitamin D status could act as a potential driver of PHPT development and PTH hypersecretion. Our data reveal that parathyroid adenomas from PHPT patients with preoperative vitamin D deficiency (Def-Ts) are intrinsically different from tumors from vitamin D-replete PHPT patients (Rep-Ts) with respect to cellular content and transcriptional profile. The cellular composition of Def-Ts reflects a striking increase in the relative abundance of parathyroid oxyphil cells compared to age-matched Rep-Ts. At the transcriptional level, genes involved in cellular respiration are preferentially upregulated in Def-Ts, while Rep-Ts are more heterogeneous, showing enhanced expression of genes in ras and myc oncogene-activated pathways and beta-amyloid protein signaling. These differences suggest that the respective etiologies of Def-Ts and Rep-Ts may be distinct, with the former being driven by adaptive metabolic responses and the latter by oncogenic signal transduction pathway activation. The increased oxyphilic content of Def-*Ts is* reminiscent of reports characterizing the hyperplastic parathyroid glands of patients with chronic kidney disease who have developed secondary hyperparathyroidism (SHPT) (11–13, 16). In the most recent of these studies, Mao and coworkers compared chief and oxyphil cell nodules from the parathyroid glands of uremic SHPT patients and found that the oxyphil cells were enriched for mitochondrial proteins; expressed lower levels of proliferation-associated genes and regulatory factors such as the vitamin D receptor, Klotho, and CaSR; and secreted higher levels of PTH [16]. This cellular phenotype is similar to Def-Ts, where mitochondrial genes are upregulated in the absence of a proliferative signature, and higher PTH secretion is observed relative to Rep-T patients. However, several important distinctions between the findings of the Mao study and the data that we report here suggest that calcimimetic-resistant SHPT and vitamin D deficiency-associated PHPT arise through different underlying molecular mechanisms. In Def-Ts, both oxyphil and chief cells display upregulation of oxidative phosphorylation, electron transport chain, and TCA cycle components, suggesting that both cell types have mobilized a response to increased energetic demand. This signature is not apparent in oxyphils from Rep-Ts or normal parathyroid tissue, indicating that the increased mitochondrial content of oxyphil cells alone does not explain the higher transcript levels of the cellular respiration genes. The absence of a purely mitochondrial signature in all oxyphils regardless of tissue source argues against compensatory mitochondrial biogenesis caused by mitochondrial mutations [25] in our vitamin D-deficient cohort. In contrast to the divergent chief vs oxyphil signatures observed in the hyperplastic glands of SHPT patients [16], the high degree of transcriptional similarity between chief and oxyphil cells within individual parathyroid glands in our study suggests that vitamin D deficiency exerts the same effect on both cell types in parathyroid adenomas. The fact that oxyphil and chief cells share expression of definitive parathyroid markers and retain highly similar transcriptomic profiles that respond similarly to vitamin D deficiency is consistent with oxyphils being a derivative of chief cells. Because the greater oxyphil content of Def-Ts does not appear to be accompanied by increased cellular proliferation relative to Rep-Ts, it is most likely that these cells arise through postmitotic differentiation of preexisting chief cells rather than through expansion of an independent cellular lineage. The increased oxyphil content in Def-Ts could indicate that this phenotypic differentiation is induced in response to vitamin D deficiency. Parathyroid adenomas are benign, relatively indolent neoplastic lesions characterized by a low mitotic index [54, 55], but activating mutations in proliferation-inducing oncogenes such as cyclin D1 have been shown to occur in 20–$40\%$ of sporadic PHPT tumors [56]. Our data are consistent with Rep-Ts arising through growth-promoting mechanisms, as genes associated with increased ras signaling and myc activation are preferentially expressed in these tumors but not in Def-Ts. Conversely, genes associated with downregulation of ras signaling are enriched in normal tissue, suggesting that deregulation of this pathway is a key event in the etiology of Rep-Ts. Supporting this idea, cyclin D1 transcription is modestly elevated in Rep-Ts compared to Def-Ts (log2(fold change) = 0.325; $$P \leq 0.045$$). Rep-Ts and normal parathyroid tissue retain the expression of genes involved in fat-soluble vitamin metabolic processes, while Def-Ts lose this signature, consistent with the concept that Rep-Ts may be driven more by proliferative changes than by metabolic disruption. Rep-Ts also retain sensitivity to vitamin D, displaying three distinct transcriptional profile clusters that correlate with preoperative vitamin D levels. This heterogeneity suggests that Rep-Ts preserve the capacity to respond to vitamin D levels and implies that the etiology of these tumors is not dependent upon the absence or diminution of 25OHD-mediated signaling. In contrast to the oncogene-activated signature of Rep-Ts, Def-Ts are characterized by differential expression of genes involved in pathways linked to opposition of cellular invasiveness, MAPK signaling, calcium sensing, and cellular stress response. EPB41L3, a protein–membrane anchor, is highly overexpressed in Def-Ts and has been found to exhibit tumor suppressor properties in multiple other tissues, with increased expression in early-stage benign tumors and loss or inactivation upon malignant transformation and the onset of invasive cellular behavior [57]. It is possible that EPB41L3 upregulation in the Def-T subset of parathyroid adenomas reflects a self-limiting protective mechanism similar to that seen with other tumor suppressor gene pathways in early-stage neoplasms [58]. Loss of this gene could be investigated as a potential marker of malignancy in parathyroid tumors and may yield a useful new indicator for the histopathological diagnosis of parathyroid carcinomas. *Two* genes associated with amyloid protein-dependent signaling perturbations in Alzheimer's disease (AD) may play an important role in Def-Ts and could reveal an intriguing connection linking amyloidosis in aging individuals to disruptions in parathyroid function. MAPK8IP1, which opposes MAPK8-mediated signal transduction, is downregulated in Def-Ts and has been found to colocalize with amyloid deposits in AD neurofibrillary tangles [33]. GABBR1, upregulated in Def-Ts, can be biochemically activated by amyloid-derived peptides [59]; our group has recently shown that GABBR1 can oppose CaSR signaling in parathyroid tissue by forming CaSR:GABBR1 heteromers that displace calcium-responsive CaSR:CaSR homomers [34]. These observations suggest that the reduced expression of MAPK8IP1 and increased levels of GABBR1 in vitamin D-deficient patients could render them more susceptible to increased amyloid protein levels, with enhanced amyloid-initiated aggregation and inactivation of MAPK8IP1 and increased opposition to CaSR signaling through amyloid-liganded GABBR1 activation both contributing to PTH hypersecretion. Interestingly, Rep-Ts also exhibit enrichment for genes associated with beta-amyloid signaling, suggesting that amyloid peptides may exert effects on both classes of parathyroid tumors. Future studies defining the role of amyloid peptides in influencing calcium sensing and PTH secretion in the parathyroid gland will provide a clearer understanding of this previously unrecognized relationship. Collectively, the parathyroid adenomas in our overall PHPT cohort express elevated levels of genes involved in tissue remodeling, cellular structure, and tumor microenvironment interactions when compared to normal parathyroid tissue. Consistent with the low mitotic index of parathyroid adenomas, the tumors do not share a dominant proliferative signature, but as a group, they appear to mobilize genes that promote morphological processes such as the epithelial–mesenchymal transition (EMT). The enhanced activity of genes in these pathways could enable features common to all parathyroid adenomas such as higher cell density and other tumor-specific structural changes including disruption of stromal boundaries, epithelial cell polarity, and basement membrane attachment. IGFBP5, a gene downregulated in both Def-Ts and Rep-Ts, is an important stimulator of osteogenic differentiation [50]. Because microRNA-mediated silencing of IGFBP5 has been shown to promote osteoporosis [49], it is possible that attenuated expression of this gene in parathyroid adenomas could contribute to bone mineral density loss in PHPT in addition to the direct osteoclastic effects of PTH. Transcriptomic studies of parathyroid tumors to date have largely employed candidate gene analysis or aggregate evaluation of bulk tumor input. As our primary focus was comparing cellular subsets within Def-Ts and Rep-Ts, the principal gene expression differences that we identified are reflective of vitamin D status in age- and sex-matched PHPT patients. Because our current study utilized a spatially indexed approach for whole transcriptome profiling of specific cellular subsets within parathyroid tumors and normal tissue, the results reported here are not directly comparable to previously published work. Nonetheless, certain informative commonalities and distinctions are apparent. Consistent with earlier studies [60, 61], we found that PTH transcript abundance was not elevated in parathyroid adenomas relative to normal tissue, suggesting that the hypersecretory behavior of these tumors is not dependent upon increased PTH gene expression. While we did not observe significantly reduced PTH mRNA levels in parathyroid tumor cells compared to normal tissue, our experimental design utilized a cross-sectional panel of independent normal donor glands as a reference standard rather than tumor-adjacent histologically normal parathyroid cells that may be influenced by the adjoining tumor [60] or parathyroid glands obtained from thyroid carcinoma patients undergoing thyroidectomy [61]. Variability between these respective control groups could contribute to the detection of alternative sets of differentially expressed genes when compared to parathyroid adenomas. Using a fold change threshold of ≥2 and a false discovery cutoff of <0.001, Chai et al [61] identified 247 DEGs (45 up-regulated, 202 down-regulated) with enrichment in KEGG pathways associated with protein processing in the ER, protein export, RNA transport, glycosylphosphotidylinositol-anchor biosynthesis, and pyrimidine metabolism. In our study, we utilized a similar fold change threshold but employed a much more stringent false discovery criterion of 1 × 10−6 in order to limit potentially artifactual differences that could arise from closely related interrogative comparisons (i.e. chief cells and oxyphil cells from a single tumor, or PHPT tumors of different vitamin D status, instead of tumor vs normal comparisons where a wider degree of divergence would be anticipated). This increased stringency, coupled with our directed comparisons between the Rep-T and Def-T subsets of PHPT adenomas, allowed us to identify vitamin D-correlated signatures that may not have emerged in aggregate tumor versus normal tissue analysis. Nonetheless, our findings are consistent with Chai et al. in revealing that parathyroid tumors in general do not exhibit a predominant proliferative gene expression profile. These results support the concept that the PHPT disease process is driven more by parathyroid tumor changes in metabolic behavior, alterations in the relationship with the extracellular environment, and protein processing in parathyroid adenomas rather than primarily by mitotically activating oncogenic events. Intriguingly, tumor-specific alterations in the beta amyloid signaling pathway were identified in the Chai et al. report as well as the current study. Further investigation of this novel relationship could yield important new insights into the potential mechanistic function of the beta-amyloid peptide in PHPT. There are several important limitations to the current study that will warrant further investigation. The cohort of vitamin D-deficient patients is small, since preoperative vitamin D deficiency is frequently restored by supplementation prior to surgery. Future studies of tumors from PHPT patients who initially presented with vitamin D deficiency at the time of diagnosis but were repleted prior to parathyroidectomy could reveal whether the Def-T profile can be dynamically switched by vitamin D therapy. It could prove useful to search for additional factors associated with vitamin D deficiency in PHPT patients to determine whether there are unrecognized variables or comorbidities contributing to their low vitamin D status and clinical presentation. Inclusion of the larger cohort of PHPT patients presenting with vitamin D insufficiency (20 ng/ml < 25OHD < 30 ng/ml) could provide a transitional state for identifying dose-dependent dynamic shifts in gene expression as confirmation of the vitamin D dependency of differentially expressed candidate genes or gene pathways identified in the current study. In future work, we will test the proposed vitamin D-dependent stratification of gene expression patterns in parathyroid cells by incorporating tumors from vitamin D-insufficient patients into our profiling studies for comparison to the vitamin D-deficient and vitamin D-replete groups. Our study drew upon a preexisting registry of female PHPT patients accrued through an ongoing project at our institution. While this experimental design allowed us to remove sex as a factor in our comparisons, future studies evaluating sex as a contributing variable could provide new key insights into the biology of PHPT. Specifically, it will be highly informative to determine whether the increased incidence of PHPT among older women is primarily a consequence of demographic factors or, alternatively, is indicative of important underlying biological differences, including the metabolism and actions of vitamin D in parathyroid tissue in females. In summary, our study demonstrates that vitamin D deficiency is associated with cellular and transcriptomic changes in parathyroid tissue that could contribute to tumor development in PHPT. Parathyroid adenomas from PHPT patients with preoperative vitamin D deficiency are molecularly distinct from tumors from PHPT patients who are vitamin D replete. Gene set enrichment comparative analysis of Def-Ts and Rep-Ts suggests the possibility of alternative etiologies for these tumors and supports the notion that oxyphil cells in either context are lineal descendants of chief cells. The differential expression in parathyroid tumors of genes associated with beta-amyloid signaling reveals a potential connection between the increased amyloid burden in aging adults [62, 63] and the heightened incidence of PHPT among older individuals [64]. Collectively, these findings suggest that PHPT in vitamin D-deficient patients may be a distinct subset of the disease with an alternative etiology, tumor composition, and cellular behavior. Further experiments assessing the reversibility of the tumor phenotype associated with vitamin D deficiency are warranted to determine whether vitamin D supplementation could potentially mitigate the clinical phenotype in PHPT patients presenting with low vitamin D status. ## Normal donor parathyroid tissue Normal human parathyroid tissue was obtained through our institution's solid organ transplant service from an unselected sequential series of eucalcemic donors, using a fully authorized tissue procurement protocol for the recovery of viable, intact parathyroid glands. The vitamin D levels of the donors were not determined. Dissected glands were immediately fixed in $4\%$ paraformaldehyde (PFA), embedded, and sectioned as previously described [65]. ## Parathyroid adenoma collection Parathyroid adenoma specimens were obtained under an IRB-approved protocol (IRB protocol number 19-27072) from patients undergoing surgery for primary hyperparathyroidism at our high-volume endocrine surgery center. Clinical, demographic, and pathological patient data were collected from the medical record and anonymized by the study clinical research coordinator in compliance with IRB requirements. The tumor samples were fixed, embedded, and sectioned using standard methods [65]. Briefly, parathyroid tissue was fixed in $4\%$ paraformaldehyde (PFA) in 0.1 M PBS (pH 7.6) overnight at room temperature. After fixation, the tissue was rinsed with ddH2O and the PFA was replaced with $70\%$ ethanol for storage. The tissue was embedded in paraffin and 5-micron sections were prepared for analysis. ## Image analysis for oxyphil quantitation Immunofluorescence images of complete tissue sections from each specimen were exported as single-channel TIFF files from the GeoMx Digital Spatial Profiler instrument (NanoString Technologies) and analyzed using ImageJ. The total cell number was quantitated from the nuclear stain SYTO13 channel using the Analyze Particles module of ImageJ. Oxyphils were quantitated from the anti-TOMM20-AlexaFluor594 channel using the 3D Object Counter module of ImageJ. Cell counts were limited to CaSR-positive cells (marked by the anti-CaSR-AlexaFluor647 channel) in each section to exclude vascular elements and other nonparathyroid components. TOMM20 was detected using an AlexaFluor594-conjugated mouse monoclonal antibody (catalog number sc-17764, Santa Cruz) at a concentration of 2 μg/ml. CaSR was detected using an AlexaFluor647-conjugated mouse monoclonal antibody, clone 3H8E9 [34] at a concentration of 5 μg/ml. ## Digital spatial profiling with the GeoMx Human Whole Transcriptome Atlas Five μm-thick FFPE sections were prepared from normal donor human parathyroid glands or from human parathyroid adenomas resected from patients undergoing parathyroidectomy for primary hyperparathyroidism. The sections were then processed using the GeoMx DS-NGS RNA FFPE slide prep protocol (NanoString Technologies). The slides were first deparaffinized and subjected to heat-inducible antigen retrieval procedures (15 min at 100°C with 1× Tris-EDTA buffer pH 9) and proteinase K digestion (1 μg/ml, 15 min at 37°C). The treated slides were hybridized to the Human Whole Transcriptome Atlas probe set (1: 12.5 dilution, 16 h at 37°C) and slides were washed twice in fresh 2× saline sodium citrate buffer (SSC). Prior to imaging on the GeoMx Digital Spatial Profiler (DSP) instrument, parathyroid tissue morphology was visualized using fluorescent-labeled antibodies (anti-CaSR and anti-Tomm20) and nuclei were visualized with 500 nM Syto13, a fluorescent DNA stain. Entire slides were imaged at 20× magnification, and 8 to 16 regions of interest (ROI) were selected per sample. ROIs were chosen based on morphology markers (CaSR+/Tomm20+/Syto13+ and CaSR+/Tomm20-/Syto13+ for chief cell-enriched and oxyphil-enriched compartment, respectively). CaSR and TOMM20 positive cells were defined as those with immunofluorescence signal intensity in the top $40\%$ of the signal range for each section stained with anti-CaSR-AF647 or anti-TOMM20-AF594, respectively. Negative cells for each marker were defined as those with immunofluorescence intensity in the bottom $40\%$ of the signal range for each section. The GeoMx instrument was then exposed ROIs to 385 nm UV light, releasing the indexing oligos and collecting them with a microcapillary. Indexing oligos were then deposited in a 96-well plate for subsequent processing. The indexing oligos were dried down overnight and resuspended in 10 μl of diethyl pyrocarbonate (DEPC)-treated water. Sequencing libraries were generated by PCR from the photo-released indexing oligos and ROI-specific Illumina adapter sequences, and unique i5 and i7 sample indices were added. Each polymerase chain reaction (PCR) used 4 μl of indexing oligos, 4 μl of indexing primer mix, and 2 μl of NanoString 5X PCR Master Mix. Thermocycling conditions were 37°C for 30 min, 50°C for 10 min, and 95°C for 3 min; 18 cycles of 95°C for 15 s, 65°C for 1 min, and 68°C for 30 s; and 68°C for 5 min. PCR reactions were pooled and purified twice using AMPure XP beads (Beckman Coulter, A63881) according to manufacturer's protocol. Pooled libraries were sequenced at 2 × 27 base pairs and with the dual-indexing workflow on an Illumina NextSeq500 sequencer. ## Analysis of human GeoMx DSP data Gene expression counts were determined using the GeoMx Human Whole Transcriptome Atlas-Human RNA for Illumina Systems (GMX-RNA-NGS-HuWTA-4) RNA probe set. This panel profiles the whole transcriptome by targeting 18,000+ unique transcripts from human protein encoding genes plus ERCC negative controls. The panel excludes uninformative high-abundance RNAs such as ribosomal subunits and includes RNA probes designed for Illumina NGS readout with the Seq Code library prep. Raw Illumina counts were Q3 normalized using the GeoMx software and standardized QC threshold settings as recommended by the manufacturer. The data were then log2 transformed prior to downstream analysis. Principal component analysis was computed through the irlba package in R, using the top 1,000 most differentially expressed genes (DEGs) out of 12,762 unique transcripts detected. T-distributed stochastic neighbor embedding (t-SNE) calculations were performed using the Rtsne R package, reducing the top 1,000 DEGs to 50 PCA dimensions before computing the t-SNE embedding. The perplexity was heuristically set to $25\%$ of the sample size. Uniform Manifold Approximation and Projection (UMAP) was computed using the uwot package in R, with the same input and heuristic settings. Heat maps were generated using the ComplexHeatmap R/Bioconductor package on scaled log-expression values using Euclidean distance and Ward linkage. The standard deviation was used to rank the genes, with the top 150 genes with the greatest degree of differential expression (largest standard deviation between groups) incorporated into the heatmap. Statistical testing of differential gene set enrichment was performed using Fisher's exact test (fGSEA), Camera, and GSVA/limma [66]. The maximum q-value of the three methods was taken as the aggregate q-value, which corresponds to taking the intersection of significant genes from all three tests. Gene sets polled were from public databases including Gene Ontology [67, 68], the Kyoto Encyclopedia of Genes and Genomes (KEGG) [69], and MSigDB [70]. Enrichment scores were calculated using the GSEA (v14) algorithm (https://www.gsea-msigdb.org/gsea/index.jsp) as previously described [27]. Differential expression analysis was performed by comparing groups (tumors from vitamin D-deficient vs vitamin D-replete patients; chief vs oxyphil cells; and normal tissue vs all tumors), with the false discovery rate (FDR) threshold set to 1 × 10−6 and the log2(fold change) threshold set to ±1. R-based analysis was performed in the Omics Playground platform [66], implemented in R using the open-source Shiny Server web application framework. The source code for the platform was cloned from a publicly available GitHub repository (https://github.com/bigomics/omicsplayground.git). ## GeoMx DSP protein nCounter quantitation Five µm-thick FFPE sections were prepared using the GeoMx DSP Protein slide prep protocol (NanoString Technologies). Briefly, the slides were first deparaffinized and subjected to standard heat-inducible antigen retrieval procedures (15 min at ∼95°C in 1× pH6 citrate buffer in a pressure cooker). The slides were then coincubated with fluorescent-conjugated morphology marker antibodies (as described above in the GeoMx DSP transcriptome methods), together with photocleavable oligonucleotide-labeled primary antibodies (profiling antibodies, see below), followed by incubation with 500 nM Syto13 nuclear stain. The stained slides were then loaded into the GeoMx DSP instrument and were scanned at 20× magnification to produce a digital fluorescence image of the entirety of the tissue sections on each slide. Circular regions of interest (660 µM2) were selected to capture roughly equivalent numbers of chief and oxyphil cells as described above. To obtain cell type-specific protein measurements, we utilized generated molecularly defined compartments within each ROI using the TOMM20 and CaSR morphology markers described above. Chief cells and oxyphil cells were marked and oligonucleotide tags corresponding to bound antibodies within each cell compartment were released by UV (385 nm) photocleavage. The released oligonucleotides were recovered and dispensed into 96-well plates. The indexing oligonucleotides were dried down and resuspended in 7 μl of diethyl pyrocarbonate (DEPC)-treated water, hybridized to 4-color, 6-spot optical barcodes, and digitally counted using the nCounter system (NanoString Technologies). GeoMx software was used to normalize the digital counts using internal spike-in controls (ERCCs) and a housekeeping gene panel as previously described [71]. The profiling antibodies were from a custom GPCR module that includes internal GeoMX DSP controls and antibodies to detect CaSR (MAb 1C12D7), metabotropic gamma-aminobutyric acid (GABA) receptors GABBR1 (Abcam, ab264069, RabMAb EPR22954-47), and GABBR2 (Abcam, ab230136, RabMAb EP2411). ## Ethics Approval Statement Our full study protocol was approved by the Human Research Protection Program Institutional Review Board (IRB) at UCSF. Fully informed consent was obtained from all study participants as required by the IRB protocol. ## Supplementary Material Supplementary material is available at PNAS Nexus online. ## Funding This work was supported by funding from the following agencies: the California Institute of Regenerative Medicine, grant #CLIN-2-11437 (JK), the National Institutes of Health, grants 1R01 CA228399-05A1 (JAS and JK) and 1R21 AG070721-02A1 (WC, JAS, and JK), and a research grant from the University of Pennsylvania Orphan Disease Center in Partnership with the Hypopara Research Foundation, grant 21-001-10 (JAS and JK). ## Author Contributions C.-L. T. performed experiments and generated and compiled data. W.C. designed and performed experiments, analyzed study results, and interpreted data. J.A.S. supervised human subject data and specimen collection, designed the overall study, and analyzed data. J.K. designed the research plan, performed the analysis of the proteomic, transcriptomic, and imaging data, and wrote the paper. All authors participated in reviewing and editing the paper. ## Data Availability The data supporting this work include private health information from the medical records of the study participants. As mandated under the IRB protocol authorizing the study, these data cannot be released publicly. Supporting data in the fully anonymized form will be made available upon request from the corresponding author after publication. ## References 1. Boudou P, Ibrahim F, Cormier C, Sarfati E, Souberbielle JC. **A very high incidence of low 25 hydroxy-vitamin D serum concentration in a French population of patients with primary hyperparathyroidism**. *J Endocrinol Invest* (2006) **29** 511-515. PMID: 16840828 2. Walker MD, Bilezikian JP. **Vitamin D and primary hyperparathyroidism: more insights into a complex relationship**. *Endocrine* (2017) **55** 3-5. PMID: 27858283 3. Moosgaard B. **Vitamin D status, seasonal variations, parathyroid adenoma weight and bone mineral density in primary hyperparathyroidism**. *Clin Endocrinol (Oxf)* (2005) **63** 506-513. PMID: 16268801 4. Moosgaard B. **Vitamin D metabolites and skeletal consequences in primary hyperparathyroidism**. *Clin Endocrinol (Oxf)* (2008) **68** 707-715. PMID: 17980013 5. Harkness L, Cromer B. **Low levels of 25-hydroxy vitamin D are associated with elevated parathyroid hormone in healthy adolescent females**. *Osteoporos Int* (2005) **16** 109-113. PMID: 15175848 6. Raposo L, Martins S, Ferreira D, Guimaraes JT, Santos AC. **Vitamin D, parathyroid hormone and metabolic syndrome - the PORMETS study**. *BMC Endocr Disord* (2017) **17** 71. PMID: 29149839 7. Clements MR. **The role of 1,25-dihydroxyvitamin D in the mechanism of acquired vitamin D deficiency**. *Clin Endocrinol (Oxf)* (1992) **37** 17-27. PMID: 1424188 8. Silver J, Russell J, Sherwood LM. **Regulation by vitamin D metabolites of messenger ribonucleic acid for preproparathyroid hormone in isolated bovine parathyroid cells**. *Proc Natl Acad Sci U S A* (1985) **82** 4270-4273. PMID: 3858880 9. Kremer R, Bolivar I, Goltzman D, Hendy GN. **Influence of calcium and 1,25-dihydroxycholecalciferol on proliferation and proto-oncogene expression in primary cultures of bovine parathyroid cells**. *Endocrinology* (1989) **125** 935-941. PMID: 2502380 10. Meir T. **Deletion of the vitamin D receptor specifically in the parathyroid demonstrates a limited role for the receptor in parathyroid physiology**. *Am J Physiol Renal Physiol* (2009) **297** F1192-F1198. PMID: 19692484 11. Ritter CS, Haughey BH, Miller B, Brown AJ. **Differential gene expression by oxyphil and chief cells of human parathyroid glands**. *J Clin Endocrinol Metab* (2012) **97** E1499-E1505. PMID: 22585091 12. Li S. **Comparative proteomic analysis of chief and oxyphil cell nodules in refractory uremic hyperparathyroidism by iTRAQ coupled LC-MS/MS**. *J Proteomics* (2018) **179** 42-52. PMID: 29526777 13. Christie AC. **The parathyroid oxyphil cells**. *J Clin Pathol* (1967) **20** 591-602. PMID: 4880406 14. Shi Y, Hogue J, Dixit D, Koh J, Olson JA. **Functional and genetic studies of isolated cells from parathyroid tumors reveal the complex pathogenesis of parathyroid neoplasia**. *Proc Natl Acad Sci U S A* (2014) **111** 3092-3097. PMID: 24510902 15. Koh J. **Transcriptional profiling reveals distinct classes of parathyroid tumors in PHPT**. *Endocr Relat Cancer.* (2018) **25** 407-420. PMID: 29475894 16. Mao J. **Integrated transcriptomic and proteomic analyses for the characterization of parathyroid oxyphil cells in uremic patients**. *Amino Acids* (2022) **54** 749-763. PMID: 35348903 17. Bedetti CD, Dekker A, Watson CG. **Functioning oxyphil cell adenoma of the parathyroid gland: a clinicopathologic study of ten patients with hyperparathyroidism**. *Hum Pathol* (1984) **15** 1121-1126. PMID: 6389314 18. Erbil Y. **The positive effect of adenoma weight and oxyphil cell content on preoperative localization with 99mTc-sestamibi scanning for primary hyperparathyroidism**. *Am J Surg* (2008) **195** 34-39. PMID: 18082541 19. Fleischer J, Becker C, Hamele-Bena D, Breen TL, Silverberg SJ. **Oxyphil parathyroid adenoma: a malignant presentation of a benign disease**. *J Clin Endocrinol Metab* (2004) **89** 5948-5951. PMID: 15579742 20. Howson P. **Oxyphil cell parathyroid adenomas causing primary hyperparathyroidism: a clinico-pathological correlation**. *Endocr Pathol* (2015) **26** 250-254. PMID: 26091632 21. Metgudmath RB, Metgudmath VV, Malur PR, Das AT, Metgudmath AR. **Functioning oxyphil parathyroid adenoma: a case report**. *J Clin Diagn Res* (2014) **8** QD07-08 22. Chang W. **PTH hypersecretion triggered by a GABA(B1) and Ca(2+)-sensing receptor heterocomplex in hyperparathyroidism**. *Nat Metab* (2020) **2** 243-255. PMID: 32694772 23. Ross AC. **The 2011 report on dietary reference intakes for calcium and vitamin D from the Institute of Medicine: what clinicians need to know**. *J Clin Endocrinol Metab* (2011) **96** 53-58. PMID: 21118827 24. Walker MD. **Vitamin D in primary hyperparathyroidism: effects on clinical, biochemical, and densitometric presentation**. *J Clin Endocrinol Metab* (2015) **100** 3443-3451. PMID: 26079779 25. Costa-Guda J, Tokura T, Roth SI, Arnold A. **Mitochondrial DNA mutations in oxyphilic and chief cell parathyroid adenomas**. *BMC Endocr Disord* (2007) **7** 8. PMID: 17916247 26. Bhattacharya A. **An approach for normalization and quality control for NanoString RNA expression data**. *Brief Bioinform* (2021) **22** bbaa163. PMID: 32789507 27. Mootha VK. **PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes**. *Nat Genet* (2003) **34** 267-273. PMID: 12808457 28. Liu S. **Three differential expression analysis methods for RNA sequencing: limma, EdgeR, DESeq2**. *J Vis Exp.* (2021). DOI: 10.3791/62528 29. Maza E. **In papyro comparison of TMM (edgeR), RLE (DESeq2), and MRN normalization methods for a simple two-conditions-without-replicates RNA-Seq experimental design**. *Front Genet* (2016) **7** 164. PMID: 27695478 30. Varet H, Brillet-Gueguen L, Coppee JY, Dillies MA. **SARTools: a DESeq2- and EdgeR-based R pipeline for comprehensive differential analysis of RNA-Seq data**. *PLoS One* (2016) **11** e0157022. PMID: 27280887 31. Son HJ, Choi EJ, Yoo NJ, Lee SH. **Mutation and expression of a candidate tumor suppressor gene EPB41L3 in gastric and colorectal cancers**. *Pathol Oncol Res* (2020) **26** 2003-2005. PMID: 31828581 32. Giuliano AR. **Methylation of HPV 16 and EPB41L3 in oral gargles: associations with oropharyngeal cancer detection and tumor characteristics**. *Int J Cancer* (2020) **146** 1018-1030. PMID: 31304592 33. Beeler N, Riederer BM, Waeber G, Abderrahmani A. **Role of the JNK-interacting protein 1/islet brain 1 in cell degeneration in Alzheimer disease and diabetes**. *Brain Res Bull* (2009) **80** 274-281. PMID: 19616077 34. Chang W. **PTH hypersecretion triggered by a GABAB1 and Ca2+ -sensing receptor heterocomplex in hyperparathyroidism**. *Nat Metab* (2020) **2** 243-255. PMID: 32694772 35. Hofle J. **Engagement of TRAIL triggers degranulation and IFN**. *EMBO Rep* (2022) **23** e54133 36. Airiau K. **TRAIL triggers CRAC-dependent calcium influx and apoptosis through the recruitment of autophagy proteins to death-inducing signaling complex**. *Cells* (2021) **11** 57. PMID: 35011619 37. Ghani MJ. **SGK1, autophagy and cancer: an overview**. *Mol Biol Rep* (2022) **49** 675-685. PMID: 34669124 38. Kim E, Ahn H, Kim MG, Lee H, Kim S. **The expanding significance of inositol polyphosphate multikinase as a signaling hub**. *Mol Cells* (2017) **40** 315-321. PMID: 28554203 39. Hsu CM. **Down-regulation of AMPD3 is associated with poor survival in head and neck squamous cell carcinoma**. *In Vivo* (2022) **36** 704-712. PMID: 35241525 40. Groulx JF. **Collagen VI is a basement membrane component that regulates epithelial cell-fibronectin interactions**. *Matrix Biol* (2011) **30** 195-206. PMID: 21406227 41. Saito J, Ishikawa Y, Yokoyama U. **Role of tissue-type plasminogen activator in remodeling of the ductus arteriosus**. *Circ Rep* (2020) **2** 211-217. PMID: 33693232 42. Chen B. **XB130 is overexpressed in prostate cancer and involved in cell growth and invasion**. *Oncotarget* (2016) **7** 59377-59387. PMID: 27509056 43. Shen J. **XB130 enhances invasion and migration of human colorectal cancer cells by promoting epithelialmesenchymal transition**. *Mol Med Rep* (2017) **16** 5592-5598. PMID: 28849225 44. Poosekeaw P. **Adaptor protein XB130 regulates the aggressiveness of cholangiocarcinoma**. *PLoS One* (2021) **16** e0259075. PMID: 34780466 45. Kalantari E. **Significant co-expression of putative cancer stem cell markers, EpCAM and CD166, correlates with tumor stage and invasive behavior in colorectal cancer**. *World J Surg Oncol* (2022) **20** 15. PMID: 35016698 46. Brinkhof B, Zhang B, Cui Z, Ye H, Wang H. **ALCAM (CD166) as a gene expression marker for human mesenchymal stromal cell characterisation**. *Gene* (2020) **763S** 100031. PMID: 34493362 47. Cardenes B. **ALCAM/CD166 is involved in the binding and uptake of cancer-derived extracellular vesicles**. *Int J Mol Sci* (2022) **23** 5753. PMID: 35628559 48. Tessema M. **Differential epigenetic regulation of TOX subfamily high mobility group box genes in lung and breast cancers**. *PLoS One* (2012) **7** e34850. PMID: 22496870 49. Zhou JG. **MicroRNA-1286 inhibits osteogenic differentiation of mesenchymal stem cells to promote the progression of osteoporosis via regulating FZD4 expression**. *Eur Rev Med Pharmacol Sci* (2020) **24** 1-10. PMID: 31957812 50. Zhang ZM, Min L, Jiang DL, Han ZY, Wang LH. **Insulin-like growth factor binding protein 5: an important regulator of early osteogenic differentiation of hMSCs**. *Folia Biol (Praha)* (2021) **67** 118-125. PMID: 35151245 51. Bilezikian JP. **Primary hyperparathyroidism**. *J Clin Endocrinol Metab* (2018) **103** 3993-4004. PMID: 30060226 52. Viccica G, Cetani F, Vignali E, Miccoli M, Marcocci C. **Impact of vitamin D deficiency on the clinical and biochemical phenotype in women with sporadic primary hyperparathyroidism**. *Endocrine* (2017) **55** 256-265. PMID: 27033542 53. Battista C. **Vitamin D status in primary hyperparathyroidism: effect of genetic background**. *Endocrine* (2017) **55** 266-272. PMID: 27154872 54. Ghandur-Mnaymneh L, Kimura N. **The parathyroid adenoma. A histopathologic definition with a study of 172 cases of primary hyperparathyroidism**. *Am J Pathol* (1984) **115** 70-83. PMID: 6711681 55. Wieneke JA, Smith A. **Parathyroid adenoma**. *Head Neck Pathol* (2008) **2** 305-308. PMID: 20614300 56. Arnold A. **Molecular pathogenesis of primary hyperparathyroidism**. *J Bone Miner Res* (2002) **17** N30-N36. PMID: 12412775 57. Zeng R. **EPB41L3 is a potential tumor suppressor gene and prognostic indicator in esophageal squamous cell carcinoma**. *Int J Oncol* (2018) **52** 1443-1454. PMID: 29568917 58. Zhu H. **Oncogene-induced senescence: from biology to therapy**. *Mech Ageing Dev* (2020) **187** 111229. PMID: 32171687 59. Rice HC. **Secreted amyloid-beta precursor protein functions as a GABABR1a ligand to modulate synaptic transmission**. *Science* (2019) **363** eaao4827. PMID: 30630900 60. Haglund F. **Diffuse parathyroid hormone expression in parathyroid tumors argues against important functional tumor subclones**. *Eur J Endocrinol* (2016) **174** 583-590. PMID: 26865585 61. Chai YJ. **Comparative gene expression profiles in parathyroid adenoma and normal parathyroid tissue**. *J Clin Med* (2019) **8** 297. PMID: 30832348 62. Rodrigue KM. **β-Amyloid burden in healthy aging: regional distribution and cognitive consequences**. *Neurology* (2012) **78** 387-395. PMID: 22302550 63. Tasaki M. **Age-related amyloidosis outside the brain: a state-of-the-art review**. *Ageing Res Rev* (2021) **70** 101388. PMID: 34116224 64. Minisola S. **Epidemiology, pathophysiology, and genetics of primary hyperparathyroidism**. *J Bone Miner Res* (2022) **37** 2315-2329. DOI: 10.1002/jbmr.4665 65. Kelly YM. **Effects of multi-stage procurement on the viability and function of human donor parathyroid glands**. *J Surg Res* (2022) **276** 404-415. PMID: 35468367 66. Akhmedov M, Martinelli A, Geiger R, Kwee I. **Omics playground: a comprehensive self-service platform for visualization, analytics and exploration of Big Omics Data**. *NAR Genom Bioinform* (2020) **2** lqz019. PMID: 33575569 67. Ashburner M. **Gene Ontology: tool for the unification of biology. The Gene Ontology Consortium**. *Nat Genet* (2000) **25** 25-29. PMID: 10802651 68. **The Gene Ontology resource: enriching a GOld mine**. *Nucleic Acids Res* (2021) **49** D325-D334. PMID: 33290552 69. Kanehisa M, Goto S. **KEGG: Kyoto Encyclopedia of Genes and Genomes**. *Nucleic Acids Res* (2000) **28** 27-30. PMID: 10592173 70. Liberzon A. **The Molecular Signatures Database (MSigDB) hallmark gene set collection**. *Cell Syst* (2015) **1** 417-425. PMID: 26771021 71. Moutafi MK. **Spatially resolved proteomic profiling identifies tumor cell CD44 as a biomarker associated with sensitivity to PD-1 axis blockade in advanced non-small-cell lung cancer**. *J Immunother Cancer* (2022) **10** e004757. PMID: 36002182
--- title: 'Patterns of sedentary behavior in adults: A cross-sectional study' authors: - Gustavo O. Silva - Paolo M. Cunha - Max D. Oliveira - Diego G. D. Christofaro - William R. Tebar - Aline M. Gerage - Hélcio Kanegusuku - Marilia A. Correia - Raphael M. Ritti-Dias journal: Frontiers in Cardiovascular Medicine year: 2023 pmcid: PMC10042287 doi: 10.3389/fcvm.2023.1116499 license: CC BY 4.0 --- # Patterns of sedentary behavior in adults: A cross-sectional study ## Abstract ### Introduction Sedentary behavior (SB) has been associated with adverse health outcomes, however, it is not completely clear whether total time in SB during the day or prolonged uninterrupted SB are interrelated. The aim of the current study was to describe the different patterns of SB of adults, their relationships, and associated factors. ### Methods The sample included 184 adults aged ranging from 18 to 59 years old. SB was objectively measured by an accelerometer and the following SB pattern parameters were obtained: total time in sedentary bouts, mean time of sedentary bouts, and total time in sedentary breaks. Demographic data (age and sex), anthropometry [weight, height, body mass index (BMI)], blood pressure (BP), medical history (self-reported comorbid conditions), and cardiac autonomic modulation, were assessed to identify factors associated with SB. Multiple linear regressions were used to analyze the relationship between SB parameters and the associated factors. ### Results The parameters of SB indicated 2.4 (0.9) h/day for total time in sedentary bouts, 36.4 (7.9) min for the mean time of sedentary bouts, and 9.1 (1.9) h/day for the total time in sedentary breaks. Multiple adjusted regression indicated that age was the only factor associated with SB patterns ($p \leq 0.05$) after adjustment for confounding variables (sex, age, BMI, dyslipidemia, systolic and diastolic BP). Young adults (18–39 years old) spent more time in sedentary bouts and less time in uninterrupted sedentary bouts compared to middle-aged adults (40–59 years old) (2.58 (0.88) h/day vs. 2.13 (0.90) h/day, respectively; $$p \leq 0.001$$ and 34.5 (5.8) min 18–39 years old vs. 38.8 (9.6) min 40–59 years old; p ≤ 0.001; respectively). The total time in sedentary breaks was similar between age groups ($$p \leq 0.465$$). The total time in sedentary bouts was significantly correlated with the mean time of sedentary bouts (r = −0.58; p ≤ 0.001), and with the total time in sedentary breaks (r = −0.20; $$p \leq 0.006$$). The mean time of sedentary bouts was significantly related to the total time in sedentary breaks (r -= 0.19; $$p \leq 0.007$$). ### Discussion and Conclusion In conclusion, age seems to be a relevant factor associated with sedentary behavior with young adults spending more time in SB and accumulating this behavior in a higher amount of sedentary bouts compared to middle-aged adults. ## Introduction Sedentary behavior (SB) is characterized by any behavior while awake in a sitting, reclining, or lying posture with an energy expenditure of ≤1.5 metabolic equivalents [1], and high time spent in SB has been associated with adverse health outcomes and risk of mortality from all-causes [2, 3], even in adults that meet guidelines for physical activity [4]. In the literature, some studies have shown that adults spend on average 6–8 h a day in sedentary behavior, which can eventually lead to elevated levels of blood pressure [5], a worsening in vascular function [6] and impaired cardiac autonomic modulation [7], which are risk factors for cardiovascular disease morbidity and mortality, and the emergence and progression of atherosclerotic lesions (8–10). The physiology behind the harms of SB has been studied in laboratory-controlled studies employing 3–8 h of uninterrupted sitting. The results indicated that prolonged SB promoted impairments in vascular function and increases in blood pressure [11]. However, whether this pattern of SB occurs in real-life situations is unclear. In contrast, most epidemiological data about the consequences of SB on health have considered the overall time spent on SB during the day, but the patterns of sedentary time can also consider the duration of the sedentary bouts, with studies suggesting that prolonged, uninterrupted sedentary time is also detrimental (12–16). However, it is not completely clear whether total time in SB during the day and prolonged uninterrupted SB are interrelated. In this study, we describe the different patterns of SB in adults, their interrelationships, and associated factors. ## Participants This cross-sectional study is an exploratory analysis of previous work [17]. The sample was comprised of adults aged 18–59 years old, from the city of Santo Anastácio in the southeast of Brazil, as previously described [17]. The sample size of 126 participants was calculated based on previous studies [17, 18] and these factors: (a) the population aged 18 years or over in the city of Santo Anastácio is 16,000; (b) the correlation among sedentary behavior parameters and age of $r = 0.17$; (c) $80\%$ power; and (d) $5\%$ alpha error [19]. To take into consideration possible errors in reading accelerometry data, misuse of equipment, along with adjustment for confounding factors, we recruited a total of 220 subjects to ensure a sufficient number of participants. This study was approved by the Ethical Research Committee from Sao Paulo State University—Unesp, under protocol CAAE: 72191717.9.0000.5402. All participants who agreed to participate signed a Written Informed Consent Form. ## Sedentary behavior parameters Sedentary behavior parameters were measured by the Actigraph GT3X accelerometer (ActiGraph, LLC, Pensacola, FL, United States). The accelerometer was placed on the right side at the waistline and participants were given instructions on how to care for the device and to wear it for the entire day (waking hours), taking off the device only when sleeping and while doing water activities (i.e., hygiene, swimming). Participants were also given instructions to use the accelerometer after receiving it for a minimum of 10 h a day for the following seven days. The 60-second epoch period was considered for this study since it is closest to the pattern of a long-duration activity [20]. Consecutive hours of zero counts and days with less than 10 h of monitoring were not considered for the final analysis [21]. At least five completed days were considered acceptable for data analysis, three weekdays and two weekend days [22]. Sedentary behavior was defined as activities lower than 200 counts per minute (cpm). The wear time percentage calculation was performed by the division of minutes per week in each physical activity intensity by the total device wear time. Sedentary bouts were defined as periods of uninterrupted sedentary behavior, while sedentary breaks were defined as the time spent in interruptions of sedentary bouts with physical activities (Time not spent being sedentary). For the analyses, we considered: - The total time in sedentary bouts: The time spent in sedentary bouts (<200 cpm for at least 10 min) during the day; - Mean time in sedentary bouts: The time spent in sedentary bouts (<200 cpm for at least 10 min) divided by the total number of sedentary bouts. - The total time in sedentary breaks: *The sum* of the time spent in physical activities (≥200 cpm for at least 10 min).Figure 1 shows a schematic view of the sedentary behavior variables considered in the analyses. **Figure 1:** *Schematic view of the sedentary behavior variables considered in the analyses.* ## Factors associated Demographic factors (age, sex, diabetes, hypertension, arthritis, obesity, and dyslipidemia) were analyzed to examine their association with SB patterns. The presence of comorbid conditions was assessed by self-report. Weight was measured with a digital scale and height with a stadiometer. Body mass index (BMI) was calculated by the following formula: body weight/height2. The participants with BMI values up to 24.99 kg/m2 were classified as having normal weight, the ones with values from 25.00 to 29.99 kg/m2 as being overweight, and the ones with a BMI ≥ 30 kg/m2 as obese. Waist circumference was measured at the midpoint between the iliac crest and the last rib with an inextensible tape with a length of 2 m and an accuracy of 0.1 cm. Blood pressure was measured with an automatic monitor (HEM-742, Omro Healthcare, Japan). To do so, the participants spent 5 min in the supine position, with the use of an adequate cuff for the arm circumference. Continuous measures were performed with 1 min of the interval among them, on the right arm, until arriving at a difference below 4 mmHg between two measurements. The value used for analysis was the average of the last two measures, as recommended by the American Society of Cardiology [23]. This same equipment also provides resting heart rate values together with blood pressure, so the same procedures were used to assess resting heart rate. Cardiac autonomic modulation assessment was done using the heart rate variability (HRV) analysis. For this assessment, the participants received the following instruction: not to consume beverages containing alcohol or caffeine and not to practice any type of exercise on the previous 12 h before the HRV measurement, in order to prevent any impact on cardiac autonomic modulation during the measurement [24]. The HRV was recorded for 30 min, with the subjects resting in the supine position, and maintaining normal breathing during the period of the recording. The HRV indexes were calculated using linear methods and analyzed in the time and frequency domains. The following linear indices were calculated: RMSSD and SDNN [25]. The following domains were used for the frequency domain analysis: the low-frequency (LF −0.04 to 0.15 Hz) and high-frequency (HF −0.15 to 0.4 Hz) spectral components in standardized units, representing the relative value of each spectral component in relation to the total power minus the very low-frequency component. All analyses were performed using the Kubios HRV Analysis, version 2.0 (Kupio University, Finland) software, and the Visual Recurrence Analysis, version 4.9 (Eugene Kononov, United States). ## Statistical analysis Continuous variables were presented as mean and standard deviation, while categorical variables were presented as absolute frequency. Student's t-test was used to compare SB (Total time and mean time of sedentary bouts, total time in sedentary breaks) by age group. Multiple linear regression analyses were performed to analyze the relationship between SB, demographic (age, sex, and self-reported comorbid conditions), and cardiometabolic (weight, height, body mass index blood pressure, and cardiac autonomic modulation) factors. Possible confounding variables (Sex, age, BMI, dyslipidemia, systolic and diastolic blood pressure) were tested in the bivariate analyses and all those with a p-value <0.20 were entered simultaneously in the final model. Multicollinearity analysis was performed assuming variance inflation factors less than 5, on which in case of not attending to this criteria the variables would be removed from the analysis model. The significance level was set at $p \leq 0.05.$ ## Results Of the 220 participants recruited, 36 were excluded from the final analysis due to either missing physical activity data or due to significant outlier values due to misuse of the accelerometer device, making a total of 184 participants included in the final analysis. The clinical characteristics, SB, and comorbid conditions of the participants included in the final analysis are described in Table 1. Most participants were women ($$n = 99$$, $54\%$) and overweight. Hypertension was observed in 39 participants ($21\%$). The accelerometer was used for 6.7 (1.2) days for 13.9 (1.8) h/day. The parameters of SB indicated 2.4 (0.9) h/day for the total time in sedentary bouts, 36.4 (7.9) min for the mean time of sedentary bouts, and 9.1 (1.9) h/day for the total time in sedentary breaks. **Table 1** | Variables | Values | | --- | --- | | Sex, women (N, %) | 99, 54 | | Age (years) | 37 (13) | | Body mass index (kg/m2) | 28.4 (5.5) | | Systolic blood pressure (mmHg) | 122 (16) | | Diastolic blood pressure (mmHg) | 75 (11) | | Heart rate (bpm) | 71 (11) | | LF/HF | 2.4 (2.0) | | Sedentary behavior | | | Total time in sedentary bouts (h/day) | 2.38 (0.92) | | Mean time of sedentary bouts (min) | 36.4 (7.9) | | Total time in sedentary breaks (h/day) | 9.07 (1.93) | | Average days of equipment use | 6.7 (1.2) | | Average time of equipment use (h/day) | 13.99 (1.81) | | Comorbidities (N, %) | | | Obesity | 51, 35 | | Hypertension | 39, 21 | | Diabetes | 7, 4 | | Dyslipidemia | 26, 14 | | Arthritis | 15, 8 | Figure 2 shows the correlation between different parameters of SB. The total time in sedentary bouts was significantly correlated with the mean time of sedentary bouts (Figure 2A, r = −0.58; p ≤ 0.001) and with the total time in sedentary breaks (Figure 2B, r = −0.20; $$p \leq 0.006$$). The mean time of sedentary bouts presented a significant correlation with the total time in sedentary breaks (Figure 2C, $r = 0.19$; $$p \leq 0.007$$). **Figure 2:** *Correlation between different parameters of sedentary behavior.* Table 2 shows multiple crude and adjusted linear regression analyses of the factors associated with the SB parameters. In the adjusted models, age was the only factor associated with total time in sedentary bouts (b = −14.9; SE = 5.9; CI $95\%$: −26.7, −3.3; $$p \leq 0.012$$), mean time in sedentary bouts ($b = 0.15$; SE = 0.06; CI $95\%$: 0.03, 0.27; $$p \leq 0.015$$) and total time in sedentary breaks ($b = 16.7$; SE = 8.4; CI $95\%$: 0.1, 33.3; $$p \leq 0.049$$). **Table 2** | Variables | Variables.1 | Crude | Crude.1 | Crude.2 | Adjusted* | Adjusted*.1 | Adjusted*.2 | | --- | --- | --- | --- | --- | --- | --- | --- | | Variables | Variables | b (SE) | CI 95% | p | b (SE) | CI 95% | p | | Total time in SED bouts | Sex | −253.8 (126.3) | −503.1; −4.6 | 0.046 | −176.9 (136.1) | −446.2; 92.4 | 0.196 | | Total time in SED bouts | Age | −13.7 (4.9) | −23.5; −4.1 | 0.006 | −14.9 (5.9) | −26.7; −3.3 | 0.012 | | Total time in SED bouts | Dyslipidemia | −370.9 (179.2) | −724.6; −17.3 | 0.040 | −155.7 (183.4) | −518.6; 207.1 | 0.397 | | Total time in SED bouts | SBP | −6.9 (4.1) | −15.1; 1.3 | 0.097 | −8.8 (6.6) | −22.0; 4.3 | 0.187 | | Total time in SED bouts | DBP | −10.2 (6.3) | −22.8; 2.3 | 0.110 | 13.5 (10.1) | −6.7; 33.7 | 0.186 | | Mean time of SED bouts | Sex | 1.1 (1.2) | −1.3; 3.3 | 0.391 | 0.8 (1.4) | −2.0; 3.5 | 0.593 | | Mean time of SED bouts | Age | 0.14 (0.04) | 0.06; 0.23 | 0.002 | 0.15 (0.06) | 0.03; 0.27 | 0.015 | | Mean time of SED bouts | Dyslipidemia | 3.7 (1.7) | 0.4; 7.0 | 0.027 | 1.0 (1.9) | −2.7; 4.8 | 0.589 | | Mean time of SED bouts | SBP | 0.03 (0.04) | −0.05; 0.11 | 0.371 | 0.04 (0.07) | −0.09; 0.18 | 0.566 | | Mean time of SED bouts | DBP | 0.07 (0.06) | −0.05; 0.19 | 0.247 | −0.06 (0.10) | −0.27; 0.15 | 0.546 | | Total time in SED breaks | Sex | 173.7 (262.9) | −344.6; 692.1 | 0.509 | 184.8 (193.6) | −198.2; 567.8 | 0.342 | | Total time in SED breaks | Age | 9.2 (6.7) | −4.2; 22.6 | 0.177 | 16.7 (8.4) | 0.1; 33.3 | 0.049 | | Total time in SED breaks | Dyslipidemia | 75.5 (369.1) | −652.3; 803.2 | 0.838 | −81.9 (260.9) | −597.8; 434.1 | 0.754 | | Total time in SED breaks | SBP | −8.5 (9.2) | −26.7; 9.7 | 0.359 | −1.8 (9.5) | −20.5; 16.9 | 0.850 | | Total time in SED breaks | DBP | −15.9 (13.9) | −43.6; 11.6 | 0.255 | −7.8 (14.5) | −36.4; 20.8 | 0.591 | Figure 3 presents the total time in sedentary bouts (A), mean time of sedentary bouts (B), and total time in sedentary breaks (C) by age groups. Total time in sedentary bouts was higher in young adults (18–39 years old) [2.58 (0.88) h/day] compared to middle-aged adults (40–59 years old) [2.13 (0.90) h/day] ($$p \leq 0.001$$). The mean time of sedentary bouts was lower in young compared to middle-aged adults [Young: 34.5 (5.8) min vs. Middle-age: 38.8 (9.6)] (p ≤ 0.001). Total time in sedentary breaks was similar between age groups ($$p \leq 0.465$$). **Figure 3:** *Total time in sedentary bouts (A), mean time of sedentary bouts (B) and total time in sedentary breaks (C) by age group. *Significant difference between age groups.* ## Discussion The main results of this study were: (i) Adults spend on average 36.4 (7.9) min in each sedentary bout, totaling 2.4 (0.9) h/day in these bouts, and 9.1 (1.9) h/day in sedentary breaks; (ii) the total time sedentary bouts was significantly correlated with the mean time of sedentary bouts and with the total time in sedentary breaks, while the mean time of sedentary bouts was significantly correlated with the total time in sedentary breaks; (iii) age was the only factor associated with SB patterns after adjustment for confounding variables; (iv) young adults are more sedentary than middle-aged adults and spend this behavior in shorter, more frequent sedentary bouts. The average time that adults spend in sedentary bouts found in the results of the current study is in agreement with a previous study in Americans that found the mean time of sedentary bouts of 30 min [26]. These values are higher than those performed in studies analyzing the consequences of sedentary behavior in health, which ranges from approximately 11–25 min per bouts/day [2, 18]. These observed discrepancies between studies could have been due to different populations studied, methodology used (such as accelerometer type and data reduction), locations, and differences in sample sizes. These findings reveal the importance of placing SB as an aim for interventions in behavior and also identify populations’ risks and associated factors. The total time in sedentary bouts was associated with the mean time of sedentary bouts and the total time in sedentary breaks, while the mean time of sedentary bouts was associated with the total time in sedentary breaks. However, even though the SB parameters were associated among themselves, a stronger association was found only between the total time sedentary behavior and the mean time of sedentary bouts, while the other SB parameters did not show such strong associations. This demonstrates that these SB parameters provide different information about SB, and thus measuring one does not provide good enough information about the other. In other words, these SB parameters seem to reflect different information about SB, thus making it important to investigate several different parameters of SB [26]. Age was the only factor associated with SB patterns after adjustment for confounding variables. We found that middle-aged adults spent less time in SB, but accumulate this behavior in longer sedentary bouts compared to young adults. These results show that young adults are more sedentary than their peers and that their sedentary behavior is accumulated through shorter but more frequent sedentary bouts. In this context, it is known that technological advancements have led to an increasingly sedentary lifestyle [27], which could be even more evident in young adults that are more prone to using technology and thus could explain the higher sedentary behavior time found in this age group, accumulated in shorter but more frequent sedentary bouts. These results demonstrate that when accounting for only the time spent in uninterrupted sedentary bouts instead of the overall time spent in sedentary behavior, only age seems to predict these more specific SB patterns and not other factors. Interestingly, anthropometric and cardiometabolic parameters were not associated with SB patterns. This contrasts with studies indicating a detrimental association between sedentary time and cardiometabolic biomarkers (2, 28–32). For example, a study reported a significant association between total sedentary time and insulin, waist circumference, HDL-cholesterol, C-reactive protein and triglycerides [2]. However, these associations accounted for the total time in sedentary behavior, and did not consider the time in uninterrupted, sedentary bouts, which was the case in our study and probably explains the divergencies found. Additionally, Bellettiere et al. found a linear dose-response relationship of sedentary time with cardiovascular disease events and showed that an additional hour of sedentary time was associated with a $12\%$ increase in multivariable-adjusted risk for cardiovascular disease. Moreover, they also showed that women with higher sedentary time and bout durations presented the greatest cardiovascular disease risk [33]. This highlights the need for more robust studies investigating the association of objectively-measured SB patterns (i.e., total and mean time of uninterrupted sedentary bouts, sedentary breaks) with anthropometric and cardiometabolic parameters, in order to clarify the true influence of SB patterns on cardiometabolic health, especially since there is robust evidence indicating that breaking up uninterrupted sedentary bouts can lead to many health benefits (33–37). This study has limitations that can be highlighted. This is a cross-sectional study that precludes cause-effect inference. In addition, we were unable to verify the relationship between SB and other important blood biomarkers (i.e., C-reactive protein, lipoproteins, and others). The sample size included adults from a single small inner city, which can affect the SB patterns. Lastly, we only studied adults and the results cannot be generalized for the elderly population. In conclusion, age seems to be a relevant factor associated with sedentary behavior with young adults spending more time in SB and accumulating this behavior in a higher amount of sedentary bouts compared to middle-aged adults. These findings are an indication that interventions aimed at reducing sedentary behavior could focus on different approaches for different age groups based on their respective patterns of accumulation of SB and that associations between SB patterns and cardiometabolic parameters should take age into consideration. ## Data availability statement The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. ## Ethics statement The studies involving human participants were reviewed and approved by Ethical Research Committee from Sao Paulo State University. The patients/participants provided their written informed consent to participate in this study. ## Author contributions GS and PC performed the statistical analyses and wrote the manuscript with support from AG, HK, MC and RR-D. MO created one of the figures. DC and WT performed the data collection. All authors contributed to the article and approved the submitted version. ## Conflict of interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ## Publisher's note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. ## References 1. Tremblay MS, Aubert S, Barnes JD, Saunders TJ, Carson V, Latimer-Cheung AE. **Sedentary behavior research network (SBRN) – terminology consensus project process and outcome**. *Int J Behav Nutr Phys Act* (2017.0) **14** 75. DOI: 10.1186/s12966-017-0525-8 2. Healy GN, Matthews CE, Dunstan DW, Winkler EAH, Owen N. **Sedentary time and cardio-metabolic biomarkers in US adults: NHANES 2003-06**. *Eur Heart J* (2011.0) **32** 590-7. DOI: 10.1093/EURHEARTJ/EHQ451 3. Stamatakis E, Gale J, Bauman A, Ekelund U, Hamer M, Ding D. **Sitting time, physical activity, and risk of mortality in adults**. *J Am Coll Cardiol* (2019.0) **73** 2062-72. DOI: 10.1016/J.JACC.2019.02.031 4. Biswas A, Oh PI, Faulkner GE, Bajaj RR, Silver MA, Mitchell MS. **Sedentary time and its association with risk for disease incidence, mortality, and hospitalization in adults**. *Ann Intern Med* (2015.0) **162** 123-32. DOI: 10.7326/M14-1651 5. Dempsey PC, Larsen RN, Dunstan DW, Owen N, Kingwell BA. **Sitting less and moving more: implications for hypertension**. *Hypertension* (2018.0) **72** 1037-76. DOI: 10.1161/HYPERTENSIONAHA.118.11190 6. Paterson C, Fryer S, Zieff G, Stone K, Credeur DP, Barone Gibbs B. **The effects of acute exposure to prolonged sitting, with and without interruption, on vascular function among adults: a meta-analysis**. *Sports Med* (2020.0) **50** 1929-42. DOI: 10.1007/S40279-020-01325-5 7. Hallman DM, Krause N, Jensen MT, Gupta N, Jørgensen MB, Holtermann A. **Objectively measured sitting and standing in workers: cross-sectional relationship with autonomic cardiac modulation**. *Int J Environ Res Public Health* (2019.0) **16**. DOI: 10.3390/IJERPH16040650 8. Lim SS, Vos T, Flaxman AD, Danaei G, Shibuya K, Adair-Rohani H. **A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990-2010: a systematic analysis for the global burden of disease study 2010**. *Lancet* (2012.0) **380** 2224-60. DOI: 10.1016/S0140-6736(12)61766-8 9. Widlansky ME, Gokce N, Keaney JF, Vita JA. **The clinical implications of endothelial dysfunction**. *J Am Coll Cardiol* (2003.0) **42** 1149-60. DOI: 10.1016/S0735-1097(03)00994-X 10. Thayer JF, Yamamoto SS, Brosschot JF. **The relationship of autonomic imbalance, heart rate variability and cardiovascular disease risk factors**. *Int J Cardiol* (2010.0) **141** 122-31. DOI: 10.1016/J.IJCARD.2009.09.543 11. Dunstan DW, Dogra S, Carter SE, Owen N. **Sit less and move more for cardiovascular health: emerging insights and opportunities**. *Nat Rev Cardiol* (2021.0) **18** 637-48. DOI: 10.1038/s41569-021-00547-y 12. Hamilton MT, Hamilton DG, Zderic TW. **Role of low energy expenditure and sitting in obesity, metabolic syndrome, type 2 diabetes, and cardiovascular disease**. *Diabetes* (2007.0) **56** 2655-67. DOI: 10.2337/DB07-0882 13. Dunstan DW, Kingwell BA, Larsen R, Healy GN, Cerin E, Hamilton MT. **Breaking up prolonged sitting reduces postprandial glucose and insulin responses**. *Diabetes Care* (2012.0) **35** 976-83. DOI: 10.2337/DC11-1931 14. Saunders TJ, Larouche R, Colley RC, Tremblay MS. **Acute sedentary behaviour and markers of cardiometabolic risk: a systematic review of intervention studies**. *J Nutr Metab* (2012.0) **2012**. DOI: 10.1155/2012/712435 15. Bergouignan A, Rudwill F, Simon C, Blanc S. **Physical inactivity as the culprit of metabolic inflexibility: evidence from bed-rest studies**. *J Appl Physiol* (2011.0) **111** 1201-10. DOI: 10.1152/JAPPLPHYSIOL.00698.2011/ASSET/IMAGES/LARGE/ZDG0101197960004.JPEG 16. Carson V, Wong SL, Winkler E, Healy GN, Colley RC, Tremblay MS. **Patterns of sedentary time and cardiometabolic risk among Canadian adults**. *Prev Med* (2014.0) **65** 23-7. DOI: 10.1016/J.YPMED.2014.04.005 17. Christofaro DGD, Tebar WR, Vanderlei LCM, Fernandes RA, Mota J, Mielke GI. **Association of cardiac autonomic modulation with different intensities of physical activity in a small Brazilian inner city: a gender analysis**. *Eur J Sport Sci* (2022.0). DOI: 10.1080/17461391.2022.2044913 18. Harris L, McGarty AM, Hilgenkamp T, Mitchell F, Melville CA. **Patterns of objectively measured sedentary behaviour in adults with intellectual disabilities**. *J Appl Res Intellect Disabil* (2019.0) **32** 1428-36. DOI: 10.1111/JAR.12633 19. Miot HA. **Tamanho da amostra em estudos clínicos e experimentais**. *J Vasc Bras* (2011.0) **10** 275-8. DOI: 10.1590/S1677-54492011000400001 20. Trost SG, Pate RR, Freedson PS, Sallis JF, Taylor WC. **Using objective physical activity measures with youth: how many days of monitoring are needed?**. *Med Sci Sports Exerc* (2000.0) **32** 426-31. DOI: 10.1097/00005768-200002000-00025 21. Craig CL, Marshall AL, Sjöström M, Bauman AE, Booth ML, Ainsworth BE. **International physical activity questionnaire: 12-country reliability and validity**. *Med Sci Sports Exerc* (2003.0) **35** 1381-95. DOI: 10.1249/01.MSS.0000078924.61453.FB 22. Troiano RP, Berrigan D, Dodd KW, Mâsse LC, Tilert T, Mcdowell M. **Physical activity in the United States measured by accelerometer**. *Med Sci Sports Exerc* (2008.0) **40** 181-8. DOI: 10.1249/MSS.0B013E31815A51B3 23. Pickering TG, Hall JE, Appel LJ, Falkner BE, Graves JW, Hill MN. **Recommendations for blood pressure measurement in humans: an AHA scientific statement from the council on high blood pressure research professional and public education subcommittee**. *J Clin Hypertens* (2005.0) **7** 102-9. DOI: 10.1111/j.1524-6175.2005.04377.x 24. Catai AM, Pastre CM, de Godoy MF, da Silva E, de Takahashi ACM, Vanderlei LCM. **Heart rate variability: are you using it properly? Standardisation checklist of procedures**. *Braz J Phys Ther* (2020.0) **24** 91-102. DOI: 10.1016/j.bjpt.2019.02.006 25. Vanderlei LCM, Pastre CM, Hoshi RA, de Carvalho TD, de Godoy MF. **Noções básicas de variabilidade da frequência cardíaca e sua aplicabilidade clínica**. *Rev Bras Cir Cardiovasc* (2009.0) **24** 205-17. DOI: 10.1590/S0102-76382009000200018 26. Diaz KM, Howard VJ, Hutto B, Colabianchi N, Vena JE, Blair SN. **Patterns of sedentary behavior in US middle-age and older adults: the REGARDS study**. *Med Sci Sports Exerc* (2016.0) **48** 430. DOI: 10.1249/MSS.0000000000000792 27. Woessner MN, Tacey A, Levinger-Limor A, Parker AG, Levinger P, Levinger I. **The evolution of technology and physical inactivity: the good, the bad, and the way forward**. *Front Public Health* (2021.0) **9** 655491. DOI: 10.3389/FPUBH.2021.655491 28. Ekelund U, Griffin SJ, Wareham NJ. **Physical activity and metabolic risk in individuals with a family history of type 2 diabetes**. *Diabetes Care* (2007.0) **30** 337-42. DOI: 10.2337/DC06-1883 29. Healy GN, Dunstan DW, Salmon J, Cerin E, Shaw JE, Zimmet PZ. **Objectively measured light-intensity physical activity is independently associated with 2-h plasma glucose**. *Diabetes Care* (2007.0) **30** 1384-9. DOI: 10.2337/DC07-0114 30. Healy GN, Wijndaele K, Dunstan DW, Shaw JE, Salmon J, Zimmet PZ. **Objectively measured sedentary time, physical activity, and metabolic risk the Australian diabetes, obesity and lifestyle study (AusDiab)**. *Diabetes Care* (2008.0) **31** 369-71. DOI: 10.2337/DC07-1795 31. Helmerhorst HJF, Wijndaele K, Brage S, Wareham NJ, Ekelund U. **Objectively measured sedentary time may predict insulin resistance independent of moderate- and vigorous-intensity physical activity**. *Diabetes* (2009.0) **58** 1776-9. DOI: 10.2337/DB08-1773 32. 32.Bellettiere J, Lamonte MJ, Evenson KR, Rillamas-Sun E, Kerr J, Lee IM, et al. Sedentary behavior and cardiovascular disease in older women: The Objective Physical Activity and Cardiovascular Health (OPACH) Study. Circulation. (2019) 139:1036–46. 10.1161/CIRCULATIONAHA.118.035312. DOI: 10.1161/CIRCULATIONAHA.118.035312 33. Kruse NT, Hughes WE, Benzo RM, Carr LJ, Casey DP. **Workplace strategies to prevent sitting-induced endothelial dysfunction**. *Med Sci Sports Exerc* (2018.0) **50** 801-8. DOI: 10.1249/MSS.0000000000001484 34. Janssen H, Dunstan DW, Bernhardt J, Walker FR, Patterson A, Callister R. **Breaking up sitting time after stroke (BUST-stroke)**. *Int J Stroke* (2017.0) **12** 425-9. DOI: 10.1177/1747493016676616 35. Larsen RN, Kingwell BA, Sethi P, Cerin E, Owen N, Dunstan DW. **Breaking up prolonged sitting reduces resting blood pressure in overweight/obese adults**. *Nutr Metab Cardiovasc Dis* (2014.0) **24** 976-82. DOI: 10.1016/J.NUMECD.2014.04.011 36. Grace MS, Dempsey PC, Sethi P, Mundra PA, Mellett NA, Weir JM. **Breaking up prolonged sitting alters the postprandial plasma lipidomic profile of adults with type 2 diabetes**. *J Clin Endocrinol Metab* (2017.0) **102** 1991-9. DOI: 10.1210/JC.2016-3926 37. Saunders TJ, McIsaac T, Douillette K, Gaulton N, Hunter S, Rhodes RE. **Sedentary behaviour and health in adults: an overview of systematic reviews**. *Appl Physiol Nutr Metab* (2020.0) **45** S197-217. DOI: 10.1139/APNM-2020-0272
--- title: Analysis of gene expression and use of connectivity mapping to identify drugs for treatment of human glomerulopathies authors: - Chen-Fang Chung - Joan Papillon - José R. Navarro-Betancourt - Julie Guillemette - Ameya Bhope - Amin Emad - Andrey V. Cybulsky journal: Frontiers in Medicine year: 2023 pmcid: PMC10042326 doi: 10.3389/fmed.2023.1122328 license: CC BY 4.0 --- # Analysis of gene expression and use of connectivity mapping to identify drugs for treatment of human glomerulopathies ## Abstract ### Background Human glomerulonephritis (GN)—membranous nephropathy (MN), focal segmental glomerulosclerosis (FSGS) and IgA nephropathy (IgAN), as well as diabetic nephropathy (DN) are leading causes of chronic kidney disease. In these glomerulopathies, distinct stimuli disrupt metabolic pathways in glomerular cells. Other pathways, including the endoplasmic reticulum (ER) unfolded protein response (UPR) and autophagy, are activated in parallel to attenuate cell injury or promote repair. ### Methods We used publicly available datasets to examine gene transcriptional pathways in glomeruli of human GN and DN and to identify drugs. ### Results We demonstrate that there are many common genes upregulated in MN, FSGS, IgAN, and DN. Furthermore, these glomerulopathies were associated with increased expression of ER/UPR and autophagy genes, a significant number of which were shared. Several candidate drugs for treatment of glomerulopathies were identified by relating gene expression signatures of distinct drugs in cell culture with the ER/UPR and autophagy genes upregulated in the glomerulopathies (“connectivity mapping”). Using a glomerular cell culture assay that correlates with glomerular damage in vivo, we showed that one candidate drug – neratinib (an epidermal growth factor receptor inhibitor) is cytoprotective. ### Conclusion The UPR and autophagy are activated in multiple types of glomerular injury. Connectivity mapping identified candidate drugs that shared common signatures with ER/UPR and autophagy genes upregulated in glomerulopathies, and one of these drugs attenuated injury of glomerular cells. The present study opens the possibility for modulating the UPR or autophagy pharmacologically as therapy for GN. ## Introduction Human glomerular diseases, including primary glomerulonephritis (GN)—membranous nephropathy (MN), focal segmental glomerulosclerosis (FSGS) and IgA nephropathy (IgAN), as well as diabetic nephropathy (DN) are leading causes of chronic kidney disease, and have a major impact on health [1]. Current therapies of GN and DN are only partially effective, significantly toxic and lack specificity. Thus, mechanism-based therapies are desirable. Glomerular visceral epithelial cells (GECs, podocytes), mesangial and endothelial cells may all be involved in the pathogenesis of GN and DN. Among these cells, podocytes are vital in maintaining glomerular capillary wall permselectivity [2, 3] and podocyte injury is believed to be key to the pathogenesis of GN and DN. Injury may be initiated by autoantibodies to glomerular components or circulating immune complexes that deposit in glomeruli and lead to the activation of complement (MN and IgAN) [4]. Alternatively, a circulating factor toxic to podocytes induces injury (FSGS) [5]. In DN, hyperglycemia and oxidative stress lead to podocyte and mesangial injury [6]. These distinct stimuli may activate or disturb various metabolic pathways in glomerular cells; e.g. in podocytes this results in disruption of the cytoskeleton, membrane composition and structure, adhesion to the glomerular basement membrane, or the function of organelles (due to ATP depletion). Conversely, other pathways may be activated in parallel in the glomerulus to attenuate cell injury or promote repair. These pathways may include protein kinases, cytokines, endoplasmic reticulum (ER) stress/unfolded protein response (UPR), ubiquitin-proteasome system and autophagy [7]. In the ER, secreted and membrane proteins are covalently modified (e.g., glycosylated) and attain a correctly folded conformation by the action of folding enzymes and chaperones, prior to transport to the secretory pathway [7]. Intact ER function is important for protein homeostasis (“proteostasis”) in podocytes, including production of components of the slit diaphragm, focal adhesion complexes (FACs) and glomerular basement membrane [2, 3]. Protein misfolding causes ER stress and activates a signaling network called the UPR [8, 9]. The UPR is regulated by three transducers in the ER membrane: inositol requiring enzyme-1α (IRE1α), activating transcription factor 6 (ATF6) and protein kinase R-like ER kinase. Actions of the UPR include upregulation of ER chaperones (to enhance protein folding capacity), attenuation of mRNA translation (to reduce the protein load to a damaged ER), and degradation of misfolded proteins, e.g., via linkage to autophagy. There is experimental evidence for upregulation of ER chaperones and dilatation of the podocyte ER in human GN and DN, supporting a role for ER stress in disease pathogenesis (reviewed in [7, 10]). There is also evidence for activation of the UPR in human GN and DN [7, 10]. *In* general, the UPR is beneficial, as it promotes protein folding thereby alleviating injury, although severe/prolonged ER stress can be detrimental and can lead to apoptosis. Autophagy is another cytoprotective process that helps clear misfolded proteins from cells, and it may be linked to the UPR [7, 11]. There is considerable evidence that the UPR and autophagy contribute to the maintenance of podocyte/glomerular homeostasis under basal conditions and that they attenuate disease [7, 10]. For example, mice with podocyte-specific genetic deletion of IRE1α show reduced induction of ER chaperones and autophagy that is associated with podocyte injury as the mice age, and exacerbation of injury in experimental glomerulopathies. “ Preconditioning” of animals to induce the UPR or treatment of animals with chemical chaperones to reduce ER protein misfolding attenuate glomerular disease [7, 12, 13]. Mice with podocyte-specific deletion of the key autophagy mediators ATG5 or ATG7 show age-related podocyte injury and exaggerated injury in experimental glomerulopathies [7, 14]. Together, these observations support mechanistic roles for the UPR and autophagy in glomerular health and disease, particularly in attenuating injury or promoting repair. Furthermore, there is recent evidence that stimulating IRE1α and the UPR pharmacologically may protect cells from injury induced by protein misfolding [15]. While multiple mechanisms mediate the pathogenesis of GN and DN, defining specific pathways that mediate injury, such as ER stress/UPR and autophagy, may be an opportunity for the establishment of more reliable diagnostic testing and development of precise therapies. In this study, we examined gene transcriptional pathways in human GN and DN, and we provide evidence that MN, FSGS, IgAN and DN are all associated with increased expression of ER/UPR and autophagy genes. We then used this knowledge of gene expression in glomerulopathies to search for drugs that activate analogous gene expression pathways in cells, with the intent of selecting candidate drugs for mechanism-based therapeutics. Using this approach, we identified neratinib as a drug that ameliorated injury in cultured GECs. ## Dataset analyses The publicly accessible Nephroseq dataset “JuCKD-Glom” (GSE47183) was used for the expression analysis of glomerular genes, including genes associated with the ER/UPR and autophagy [16, 17]. Nephroseq contains microarray gene expression data (mRNAs) of laser-captured glomeruli from human kidney biopsies. These data are presented as the fold-increase of gene expression in disease above healthy control. Additional analyses were performed using publicly accessible datasets GSE108109 [18], GSE115857 (unpublished) and GSE141295 [19] (Supplementary Table S1). It should be noted that since these datasets have been generated/published independently, we did not pool the data together and we analyzed the datasets separately. The p-values and p-values adjusted for the false discovery rate (e.g., Q-values), which we present in this manuscript, are those calculated and reported within the datasets by their original authors. We employed principal component analysis of gene expression, an algorithm that identifies the maximal variations in the data and reduces the dimensionality to a few components [17, 20]. Pathway overrepresentation and gene ontology (GO) enrichment analyses were performed using the ConsensusPathDB interaction database [21, 22]. This database provides the adjusted p-values (hypergeometric test, corrected for multiple comparisons) that we present in this manuscript [22]. To search for gene signatures induced by chemical perturbagens (drugs), i.e., “connectivity mapping,” as well as ligands and protein kinases, we used the Library of Integrated Network-based Cellular Signatures (LINCS) L1000 dataset [23, 24], which is an extension of the original connectivity mapping (CMAP) dataset (“old CMAP”) [25], and contains 1,319,138 profiles from 42,080 perturbagens (19,811 small molecules, 18,493 shRNAs, 3,462 cDNAs and 314 biologics; treatment vs. vehicle control pairs), corresponding to 25,200 biological entities and 473,647 signatures carried out in 3–77 cells lines. To query this database, we utilized LINCS analytical tools,1 including BD2K-LINCS DCIC and Enrichr. Signatures from the LINCS L1000 data are computed using the moderated Z-score method; the p-values and adjusted p-values presented in this manuscript are those reported directly by the LINCS data portal. One limitation of this dataset is that it includes the HA1E immortalized normal kidney epithelial cell line and various tumor cell lines, but glomerular cells are not included. Relationships between GO terms were produced by using the QuickGO tool [26]. The chord plot was produced by using the R package tool GOPlot [27]. Protein–protein interaction networks were constructed using NetworkAnalyst2 and STRING interactome database with medium confidence score of 600 and experimentally verified criteria [28]. ## Generation of ER/UPR and autophagy gene lists To compile a set of genes associated with ER function and ER stress/UPR, we combined genes listed in the Protein Processing in the ER KEGG pathway (which includes UPR and other ER-related genes) [29], and in the Qiagen human UPR PCR Array (PAHS-089Z, Qiagen) (Supplementary Table S2). Second, we examined genes or genes encoding proteins that were reported to be inducible by X-box binding protein-1 (XBP1; an effector of IRE1α), ATF6 and XBP1 + ATF6 in HEK293T cells [30]. There was a total of 639 genes in this dataset, of which 70 were also present in the KEGG+Qiagen dataset (Supplementary Table S2). We then subjected the non-overlapping 570 XBP1- and ATF6-inducible genes to a GO analysis (Supplementary Table S3), and we selected genes corresponding to ER/UPR pathways. This resulted in an additional 116 genes, which we added to the 203 ER/UPR genes in the KEGG+Qiagen dataset for a total of 319 ER/UPR genes (Supplementary Table S4). Of the 319 ER/UPR genes, 271 ($85\%$) were found in the Nephroseq microarray (48 were not in the microarray). To compile a set of autophagy-related genes, we took a list of 98 genes from the Autophagy KEGG pathway and Qiagen autophagy PCR array (PAXX-084Y, Qiagen), and we subjected these genes to a GO analysis (Supplementary Table S5). Then, we selected autophagy-related pathways and genes corresponding to these GO pathways. This resulted in an additional 449 genes, which we added to the original 98 genes for a total of 547 autophagy genes (Supplementary Table S4). Of the 547 autophagy genes, 391 ($71\%$) were found in the Nephroseq microarray (156 were not in the microarray). Protein–protein interaction networks were produced based on proteins encoded by the 319 ER/UPR and 547 autophagy genes. The large majority of the ER/UPR (Supplementary Figure S1a) and autophagy genes (Supplementary Figure S1b) were mapped to these networks, and the majority of these genes interconnected with each other. There was also a small number of intermediate proteins that connected the selected ER/UPR and autophagy genes (Supplementary Figure S1). As expected, the largest nodes in the ER/UPR network are chaperones and ubiquitin-proteasome system components. ## GEC culture, immunoblotting and quantification of FACs Primary mouse GECs were generated according to previously published methods [12, 31]. These cells contain a floxed IRE1α gene, but in the absence of transduced Cre recombinase, the cells express normal levels of IRE1α, and are phenotypically normal [12]. In IRE1α knockout (KO) cells, Cre recombinase has deleted IRE1α [12]. Studies were done after the cells were cultured at the differentiation temperature (37°C). In experiments, GECs were incubated with drugs added to culture medium. Drugs included adriamycin (doxorubicin), chloroquine, tunicamycin (Sigma-Aldrich Canada, Oakville, ON); geldanamycin, radicicol, NVP-AUY922 (luminespib), neratinib (Cayman Chemical, Ann Arbor, MI); rapamycin (BioShop Canada, Burlington, ON); or IXA6 (N-[(4-chlorophenyl)methyl]-N-[2-(2,3-dihydro-1H-indol-1-yl)-2-oxoethyl]pyridine-3-sulfonamide; Life Chemicals Inc., Niagara-on-the-Lake, ON). Stock solutions of drugs were prepared in DMSO. The protocol for immunoblotting was described previously [12]. Antibodies included rat anti-GRP94 (sc-32,249; Santa Cruz Biotechnology, Santa Cruz, CA), rabbit anti-BiP/GRP78 (ADI-SPA-826F, Enzo Life Sciences, Ann Arbor, MI), mouse anti-HSP70 (3A3, sc-32,239, Santa Cruz), rabbit anti-LC3B (2,775, Cell Signaling Technology, Danvers, MA), rabbit anti-p62/SQSTM1 (23,214, Cell Signaling), rabbit anti-mesencephalic astrocyte-derived neurotrophic factor (MANF/ARMET; PAB13301, Abnova, Walnut, CA), and rabbit anti-actin (A2066, MilliporeSigma, Mississauga, ON). Chemiluminescence signals were detected in a ChemiDoc Touch Imaging System (Bio-Rad; Mississauga, ON). The intensities of bands in each immunoblot were measured in samples derived from the same experiment and were quantified using ImageJ software. The actin signal was used as loading control for normalization of signals. We then calculated the relative intensities of all bands in each immunoblot. We ensured that the intensities of signals were all within a linear range. The lactate dehydrogenase (LDH) release assay was described previously [31]. To visualize FACs, cultured GECs were fixed with $4\%$ paraformaldehyde (37°C), permeabilized with $0.5\%$ Triton X-100 and blocked with $3\%$ BSA. Cells were stained with mouse antibody against the FAC adaptor protein vinculin (Sigma V9131) for 24 h (4°C) plus rhodamine-goat-anti-mouse IgG, as well as fluoresceinated-phalloidin (to stain F-actin and outline the cells), and Hoechst H33342 (to stain nuclei), as described previously [17, 32]. Z-stack images were acquired on a Zeiss Axio Observer inverted fluorescence microscope with visual output connected to an AxioCam MRm monochrome camera (Carl Zeiss AG; Toronto, ON). Immunofluorescence intensities and cell measurements were performed using ImageJ, as described previously [17, 32]. ImageJ allows pre-definition of particle size and the threshold of immunofluorescence intensity to count FACs and estimate cell size. ## Statistics As noted above, the statistical parameters (p-values and adjusted p-values) of gene expression datasets, as well as the ConsensusPathDB interaction and L1000 databases that we present in this manuscript are those calculated and reported within these datasets. To address the overlap of genes among GNs, we calculated Jaccard similarity coefficients and assessed the significance using a hypergeometric test. Experimental data that we generated are presented as mean ± standard deviation (SD) or standard error (SE; as indicated). Comparisons between two groups were done by a two-tailed Student’s t-test. For three or more groups, statistical differences were assessed using one-way analysis of variance. Where significant differences were found, post-hoc analyses were performed using Sidak’s multiple comparisons test. ## Human glomerulopathies reveal many shared upregulated genes The datasets employed and the workflow for this study are presented in Table 1. Initially, we examined all upregulated genes in FSGS ($$n = 25$$), MN ($$n = 21$$) and IgAN ($$n = 27$$) in the JuCKD-Glom (GSE47183) dataset in Nephroseq (Supplementary Table S1) [16]. This dataset contains microarray gene expression data of laser-captured glomeruli from human kidney biopsies. The dataset provides the fold-change in gene expression compared with healthy controls ($$n = 21$$). The mean ages of the patients with the three GNs ranged from 36.4 to 53.7 years, compared with 47.2 years in controls (Supplementary Table S1). The mean estimated glomerular filtration rates of the three GNs were slightly reduced (73.9–74.5 ml/min/1.73m2), compared with control (105.4 ml/min/1.73m2), while blood pressures were in the normal range (Supplementary Table S1). A total of 2,671 genes (out of 11,933 examined) showed increased expression in the three GNs; MN = 1,840, FSGS = 2,064 and IgAN = 1,878 (Figure 1A; Supplementary Table S6). There were 1,046 upregulated genes common to the three GNs, i.e., more than $50\%$ of genes upregulated in each GN were common to the three (Figure 1A). Furthermore, a number of additional genes were common to at least two GNs (Figure 1A). We calculated the Jaccard similarity coefficients corresponding to the overlap of the top 200, 500 and 1,000 upregulated genes in the GNs (Figure 1B). The results suggested a highly significant overlap (hypergeometric test $p \leq 1$ × 10−37 in all cases), confirming that upregulated genes were shared among diseases. To determine if among the upregulated genes in MN, FSGS and IgAN there were genes associated with ER stress and the UPR, as well as autophagy, we performed a GO enrichment analysis (Supplementary Table S7). Examination of all GO terms in each GN revealed a number of GO terms associated with ER components, ER function and the UPR, such as ER chaperones, protein folding, protein transport, stress responses and others (Figure 2; Supplementary Table S8). A smaller number of GO terms were associated with autophagy (Figure 2; Supplementary Table S8). Approximately $50\%$ of 19 selected GO terms were present in all three GNs with a further $20\%$ in two (Figure 3). Together, these results imply that the UPR and autophagy are common responses in multiple types of glomerular injury. **Figure 2:** *Relationships between GO terms selected from genes upregulated in MN, FSGS, and IgAN JuCKD-Glom datasets. Boxes highlighted in yellow denote 19 GO terms represented in the glomerulopathies. “A is a B”: node A is a subtype of node B; “A part of B”: node A is a part of node B; “A regulates B”: node A regulates node B positively or negatively.* **Figure 3:** *Occurrence summary (chord plot) between 19 GO terms selected from genes upregulated in MN, FSGS, and IgAN JuCKD-Glom datasets (highlighted in yellow in Figure 2). The figure highlights pathways common among the three glomerulopathies.* In JuCKD-Glom, there are also genes whose expression is downregulated in GNs (Supplementary Table S9). We performed a GO analysis combining all downregulated genes that were common to at least two of MN, FSGS and IgAN ($$n = 2$$,257). No ER/UPR or autophagy pathways were found and there were only 3 ER-related ontology terms (Supplementary Table S9). Probably, downregulation of UPR and autophagy genes is less biologically relevant, compared with upregulation, and since our focus is on the UPR and autophagy, we do not include further analysis of downregulated genes in the present study. ## ER and autophagy gene expression is increased in human glomerulopathies We compiled a set of 319 genes associated with ER function/stress/UPR (ER/UPR) and a second set of 547 genes associated with autophagy (Materials and methods). Among these genes, 271 and 391, respectively, were present in the Nephroseq microarray. FSGS, MN and IgAN all demonstrated increases in the expression of ER/UPR genes. Among 271 ER/UPR genes, 92 were increased when the three GNs were considered together, and of these, 42 were common to the three diseases (Figures 1C, 4; Supplementary Table S10). Furthermore, a significant number of ER/UPR genes were common to at least two GNs (Figure 1C; Supplementary Table S10). Principal component analysis of changes in ER/UPR gene expression indicates that normal controls and FSGS, MN and IgAN patients can be clearly separated into non-overlapping populations (Figure 5). By analogy, FSGS, MN and IgAN all demonstrated increases in the expression of autophagy genes. Among 391 autophagy genes, 108 were increased when the three GNs were considered together, and of these 38 were common to the three diseases (Figure 1D, 4; Supplementary Table S10). Similar to the ER/UPR genes, a significant number of autophagy genes were common to at least two GNs (Figure 1D; Supplementary Table S10). **Figure 4:** *ER and autophagy genes that are upregulated significantly in glomerulopathies. The figure presents the 93 ER/UPR genes and 108 autophagy genes whose expression was increased in at least one the three glomerulopathies.* **Figure 5:** *Principal component analysis of ER/UPR gene expression in glomerulopathies. Healthy controls are clearly separated from patients with FSGS, MN and IgAN.* The JuCKD-Glom dataset also contains microarray data of laser-captured glomeruli from patients with DN. Although DN is not a primary GN, we nevertheless examined expression of genes in DN, given its clinical importance. The Jaccard similarity coefficients and the highly significant p-values (hypergeometric test $p \leq 1$ × 10−37 in all cases) of the overlap of the top upregulated genes in DN and the GNs indicated that upregulated genes were common not only among FSGS, MN and IgAN, but also among DN and the three GNs (Figure 1B). Moreover, there were 34 ER/UPR genes and 44 autophagy genes upregulated in DN. Interestingly, there was only one ER/UPR gene (Supplementary Table S10) and two autophagy genes (Supplementary Table S10) upregulated in DN, which were not part of the groups of genes upregulated in FSGS, MN and IgAN. *These* genes were added to the 92 ER/UPR and 108 autophagy genes for a total of 93 ER/UPR and 110 autophagy genes, which were used in further studies. Of the 93 upregulated ER/UPR genes and 110 upregulated autophagy genes, 89 were mapped to the ER/UPR protein–protein interaction network, and 106 to the autophagy network, respectively (Supplementary Figure S1). ## Upregulation of ER and autophagy genes is recapitulated in additional GN datasets To determine if upregulation of genes in the JuCKD-Glom GN dataset (GSE47183) was more broadly applicable, we studied gene expression in additional datasets of GN and control subjects, including GSE108109, GSE115857 and GSE141295 (Supplementary Table S1) [18, 19]. We selected these datasets because there were at least 10 human glomerular GN samples per dataset, each dataset provides the fold-change in gene expression compared with healthy controls, and the samples were ascertained to be distinct from those in JuCKD-Glom. When examining upregulation of individual genes (i.e., the 319 ER/UPR genes and 547 autophagy genes), at least as many of these genes were upregulated in GSE108109, GSE115857 and GSE141295 as in JuCKD-Glom, although only a modest to moderate number of these genes overlapped with those observed in JuCKD-Glom (Supplementary Table S1). This may be at least in part due to the use of different microarrays or techniques between the datasets. Importantly, a substantial majority of the ER/UPR and autophagy-related GO biological pathways and gene categories that were evident in JuCKD-Glom overlapped with those identified in GSE108109, GSE115857 and GSE141295, and there were additional biological pathways and gene categories related to the ER/UPR and autophagy identified in these datasets (Supplementary Table S8). Together, these analyses of gene expression support the view that ER/UPR and autophagy pathways are activated in human GNs. ## ER/UPR and autophagy genes are expressed in multiple glomerular cell types To determine if the upregulated ER/UPR ($$n = 93$$) and autophagy genes ($$n = 110$$) in glomerulopathies (Figure 4) were expressed in a specific glomerular cell type, we interrogated single cell RNA sequence datasets. In two older datasets that examined a limited number of glomerular cells [33, 34], $\frac{73}{93}$ ER/UPR genes and $\frac{63}{110}$ autophagy genes were expressed in podocytes, while $\frac{70}{93}$ ER/UPR genes and $\frac{74}{110}$ autophagy genes were expressed in glomerular mesangial cells (Supplementary Table S11). In a more recent and more extensive dataset [35], $\frac{86}{93}$ ER/UPR genes and $\frac{100}{110}$ autophagy genes were expressed in podocytes, and the same genes were expressed in glomerular endothelial cells (Supplementary Table S11). It is therefore reasonable to conclude that at least basal expression of these genes is ubiquitous and is not restricted to a single glomerular cell type. ## Connectivity mapping (CMAP) identifies drugs that target the ER/UPR and autophagy MN, FSGS and IgAN showed common signatures in UPR and autophagy gene expression (Figure 1), as well as biological pathways and gene categories (Supplementary Table S8). We proceeded to use the LINCS L1000 dataset to identify candidate drugs for treatment of glomerulopathies by relating the gene expression signatures of different drugs with the 93 ER/UPR and 110 autophagy genes upregulated in these glomerulopathies (JuCKD-Glom). A considerable number of studies (see Introduction) indicate that the UPR and autophagy are cytoprotective mechanisms in glomerular diseases, and this cytoprotection is believed to be dependent on upregulation of UPR and autophagy genes/proteins [7, 10]. Therefore, in contrast to other studies that have evaluated drugs as inhibitors, we asked which drugs could potentially stimulate the UPR or autophagy, i.e., induce upregulated gene signatures in cultured cell lines that are similar to the upregulated gene signatures in glomerular diseases? This approach is supported by recent studies that have demonstrated cytoprotective effects of drugs that stimulate the IRE1α-XBP1 UPR pathway [15, 36]. Given that the glomerulopathies have a high degree of overlap in gene expression profile, we used the union of the upregulated genes in the connectivity mapping analysis. The union allows use of a greater number of genes in the analysis and to identify candidate drugs with greater statistical power. In addition to drugs, we searched for signatures induced by ligands (growth factors, cytokines) and protein kinase knockdowns with shRNAs. In the LINCS L1000 dataset, there were 52 drugs (“chemical perturbagens”) in 17 cell lines (Supplementary Tables S12, S13), which produced gene signatures that showed a statistically significant similarity (adjusted $p \leq 0.05$) with the 93 upregulated ER/UPR genes in glomerulopathies (Supplementary Table S10). By analogy, there were 227 drugs in 27 cell lines (Supplementary Tables S12, S13), which produced gene signatures showing significant similarity with the 110 upregulated autophagy genes in glomerulopathies (Supplementary Table S10). Among these two groups of drugs, there were six drugs that showed similarities with both the ER/UPR and autophagy gene signatures, and for which descriptions were available in the literature (Supplementary Table S12). We selected these six candidate drugs for further characterization and refer to them as “L1000-ER/UPR/autophagy drugs.” *Within this* group, geldanamycin, NVP-AUY922 (luminespib) and radicicol are Hsp90 inhibitors, although radicicol may have multiple actions. Hsp90 inhibition can lead to the activation of cytosolic and ER stress responses. Neratinib is a highly selective epidermal growth factor receptor (EGFR) and HER2 inhibitor. Celastrol and withaferin-a have multiple reported actions, including activation of cytosolic and ER stress responses [37, 38]. Connectivity mapping of the UPR genes common to MN, FSGS and IgAN demonstrated the presence of the same six L1000-ER/UPR/autophagy drugs (Supplementary Table S13), in keeping with our approach that used the union of the upregulated genes. Comparison of ER/UPR gene signatures induced by the L1000-ER/UPR/autophagy drugs showed that the drugs with effects in the greatest number of cell lines and time points had the greatest number of upregulated genes in total. For example, geldanamycin increased expression of 35 genes, representing $38\%$ of the 93 genes that were upregulated in the glomerulopathies, while neratinib increased only 9 genes or $10\%$ (Supplementary Table S14). There were also considerable similarities in the upregulated genes among the drugs (Supplementary Table S14). *Autophagy* gene signatures were generally smaller compared with the ER/UPR genes (Supplementary Table S14). For example, neratinib increased expression of 11 genes, representing $10\%$ of the 110 genes that were upregulated in the glomerulopathies (Supplementary Table S14). We anticipated that the LINCS L1000 dataset would reveal ligands and kinases with gene signatures that would match the 93 ER/UPR or the 110 autophagy genes. Identification of various ligands or kinases could potentially have been helpful in validating the identified drugs. Surprisingly, there were only a few ligands (growth factors, cytokines) identified with signatures similar to the 93 ER/UPR genes (Supplementary Table S13), and the genes stimulated by these ligands included only HLA genes. Also, there was only a single protein kinase identified (NUAK1), with only four genes overlapping with the 93 ER/UPR genes (Supplementary Tables S13, S14). The original CMAP drug database (“old CMAP”) was largely in keeping with LINCS L1000, as among candidate drugs, there were many Hsp90 inhibitors, as well as another EGFR inhibitor (Supplementary Table S13). This result strengthens the validity of the drugs selected via the L1000 dataset analysis. Importantly, thapsigargin, a drug that releases calcium from the ER and a well-known inducer of the UPR, was identified in the old CMAP drug database (Supplementary Table S13). The identification of thapsigargin, a “gold standard” ER/UPR drug, further supports the validity of the search method. Additional analysis of the thapsigargin gene datasets showed that the drug increased expression of 13 ER/UPR genes or $14\%$ of the 93 genes that were upregulated in the glomerulopathies (Supplementary Table S14). Thus, the number of genes upregulated by the six L1000-ER/UPR/autophagy drugs is in the same range (or even greater) than the number upregulated by thapsigargin. There were no ligands, kinases or additional drugs identified with signatures similar to the 110 autophagy genes (Supplementary Table S13). It should be noted that while neratinib was the single EGFR/HER2 inhibitor identified as matching both ER/UPR and autophagy gene signatures, four other EGFR inhibitor drugs showed significant matches with the autophagy signatures of the glomerulopathies (Supplementary Table S13), strengthening the validity of neratinib as a candidate autophagy-inducing drug. Finally, since the above search involved genes in the JuCKD-Glom dataset, we also searched the LINCS L1000 database for candidate drugs using input of ER/UPR and autophagy genes upregulated in the other GN datasets (GSE108109, GSE115857 and GSE141295; Supplementary Table S1). As stated above, a number of ER/UPR and autophagy genes showed upregulation in GSE108109, GSE115857 and GSE141295, with modest-moderate overlap with JuCKD-Glom. There was considerable similarity between the L1000-ER/UPR/autophagy drugs selected based on JuCKD-Glom and the drugs selected using these other gene inputs. Specifically, geldanamycin, NVP-AUY922, radicicol, celastrol, withaferin-a and neratinib were also identified as candidate drugs based on GSE108109, GSE115857 and GSE141295 (Supplementary Table S13). ## L1000-ER/UPR/autophagy drugs stimulate expression of ER chaperones and/or proteins involved in autophagy Four of the L1000-ER/UPR/autophagy candidate drugs were tested to see if they stimulate the UPR in cultured GECs. We examined the potential of the drugs to upregulate expression of the ER chaperones GRP94 (HSP90B1), BiP (HSPA5) and mesencephalic astrocyte-derived neurotrophic factor (MANF), which were previously shown to be upregulated as part of the UPR in GECs [12]. After incubating GECs with geldanamycin, NVP-AUY922 and radicicol, we observed increases in GRP94, BiP and MANF proteins (most consistent with radicicol), although these increases, as expected, were not as robust as those induced by tunicamycin, a potent inducer of the UPR, used as a positive reference control (Figures 6A,C). As predicted, since these three drugs inhibit Hsp90, this leads to derepression of heat shock factor-1 and an increase in the cytosolic stress protein, Hsp70 (HSPA1A; Figures 6A,C). In an earlier study, we demonstrated that celastrol (another one of the L1000-ER/UPR/autophagy drugs) induces the UPR and Hsp70 in GECs [39]. Neratinib did not increase GRP94, BiP or MANF protein expression; however, genes specifically encoding these three chaperones were not upregulated by neratinib, whereas neratinib upregulated another ER chaperone gene (ERP29), as well as XBP1, a transcription factor for multiple ER chaperones (Supplementary Table S14). **Figure 6:** *Induction of UPR (A,C) and autophagy genes (B,D) by L1000-ER/UPR/autophagy drugs. Cultured GECs were incubated for 24 h with drugs (10 μM concentrations). In panel B, drugs were added together with CQ, as indicated. Lysates were immunoblotted, as indicated. (A,B) Representative immunoblots. (C–E) Densitometric quantification (the expression of each protein is normalized to actin). *p < 0.05, **p < 0.01, +p < 0.001, ++p < 0.0001 vs. untreated. Mean ± SE, 3 experiments performed in duplicate. (E) Autophagosomal flux, as determined by the level of LC3-II in the presence of chloroquine (densitometric quantification of specific lanes in panel B). *p < 0.05, **p < 0.01, vs. chloroquine. Tunicamycin and rapamycin are stimulators of the UPR or autophagy (positive controls). Untr, untreated; Tunic, tunicamycin, Gel, geldanamycin; Rad, radicicol; NVP, NVP-AUY922; Ner, neratinib; Rap, rapamycin; CQ, chloroquine.* Expression of ER chaperones in the UPR is regulated transcriptionally, while autophagy is believed to be primarily a post-translational process; however, more recently it has become evident that transcriptional regulation of autophagy genes and their protein products may contribute to autophagy [7]. Thus, we examined whether radicicol and neratinib can increase expression of proteins involved in autophagy, including LC3 (MAP1LC3B) and p62/SQSTM1 [12] (Supplementary Table S14). Incubation of GEC with neratinib (but not radicicol) increased total LC3 significantly, i.e., LC3-I + LC3-II, as well as p62 (Figures 6B,E), although, as with ER chaperones, the increases were not as robust as those induced by tunicamycin. To monitor autophagic flux, we incubated GECs with chloroquine, which blocks the fusion of autophagosomes with lysosomes and prevents autolysosomal protein degradation, allowing comparison of the rates of autophagosome formation. Neratinib+chloroquine increased LC3-II compared with chloroquine alone, in keeping with stimulation of autophagy (Figures 6B,E). LC3-II tended to be greater with radicicol, but the change did not reach statistical significance. Therefore, while neratinib did not stimulate expression of ER chaperones in GECs, the principal action of this drug was to stimulate autophagy proteins. ## Neratinib reduces GEC injury First, we demonstrated that the L1000-ER/UPR/autophagy drugs were not toxic to GECs, using a LDH assay (Supplementary Table S12). In this assay, we included adriamycin, a drug known to injure podocytes and induce experimental FSGS in rodents in vivo, as well as podocyte ER stress and autophagy [12, 40]. Low dose adriamycin (0.5 μM) did not induce LDH release. In glomerulopathies, podocyte injury is generally not lethal or cytolytic [4]. Cultured GECs form FACs, which mediate adhesion to extracellular matrix [2]. To monitor injury, we used a GEC assay that quantifies FAC density based on vinculin expression. In this assay, dissolution of FACs correlates with injury, i.e., podocyte foot process effacement in primary FSGS in vivo [17, 32]. We selected two drugs with putatively distinct primary targets (neratinib and radicicol) to test if the drugs could protect GECs from injury. We did not test celastrol in this assay, because in our previous study, celastrol proved to be toxic in mice in vivo [39]. GECs were untreated, or were treated with low dose adriamycin (0.5 μM) to induce sublethal injury. In parallel, GECs were treated with neratinib or adriamycin+neratinib, as well as with radicicol or adriamycin+radicicol. Low dose adriamycin reduced the number of FACs (visualized by vinculin immunostaining), consistent with FAC dissolution due to sublethal injury (Figures 7A,B). Neratinib independently did not affect the number of FACs, and importantly, it attenuated the reduction in FACs induced by adriamycin by more than $50\%$ (Figures 7A,B), implying that neratinib protected GECs from injury. Adriamycin and neratinib did not alter vinculin expression (Figure 7C), indicating that changes in the number of FACs were due to vinculin redistribution in cells. In the FAC assay, radicicol (0.5–1 μM) independently decreased FAC number, comparably to the effect of adriamycin shown in Figure 7B; thus, radicicol appeared to be toxic. A lower dose of radicicol (100 nM) did not independently reduce FAC number; however, this dose was not effective in attenuating FAC dissolution by adriamycin. **Figure 7:** *Effect of neratinib on FACs. Cultured GECs were untreated, or were incubated with adriamycin (ADR; 0.5 μM) together with or without neratinib (Ner; 100 nM) for 24 h. (A and B) Cells were then stained with anti-vinculin antibody (red; FACs), phalloidin (green; F-actin cytoskeleton) and H33342 (blue; nuclei). Representative photomicrographs (A) and quantification of FACs (B) are presented. FACs are evident throughout the cells, and F-actin bundles are seen to originate within the FACs. ADR reduces FACs and this effect is attenuated by neratinib. Bar = 25 μm. *p < 0.0001 Untreated vs. ADR and ADR vs. ADR + neratinib. Mean ± SE, 3 experiments with 10–16 measurements per group per experiment. (C) Representative immunoblot of vinculin shows comparable levels among the four conditions. Densitometric quantification of 4 experiments performed in duplicate is shown below the blot (vinculin normalized to actin; mean ± SE). There are no statistically significant differences.* Finally, while neratinib was identified due to its induction of ER/UPR and autophagy gene signatures, it should be noted that the GECs used in this study express low levels of EGFR. This conclusion is based on mass spectrometry spectral counts, which are a semi-quantitative practical measure of protein abundance. The mean spectral counts of EGFR were 4.33 in untreated control GECs and 5.33 in KO GECs (3 measurements per group), reflecting low expression [13]. HER2 spectral counts were not detected. ## IXA6, a drug directed at IRE1α, reduces GEC injury In this set of experiments, we first verified that the UPR protects GEC from injury in the FAC assay. We compared injury in control (IRE1α-replete) GECs with IRE1α KO cells, in which activation of the UPR was shown to be impaired [12]. Control and IRE1α KO GECs were untreated or were incubated with adriamycin (as above) to induce sublethal injury. By analogy to the results in Figure 7, the number of FACs (visualized by vinculin immunostaining) was reduced after adriamycin treatment; however, the reduction in FACs was almost 2-fold greater in IRE1α KO cells compared to IRE1α-replete control (Figures 8A,B). Therefore, it can be concluded that the IRE1α UPR pathway is protective. **Figure 8:** *Effect of the UPR and IXA6 on FACs. (A,B) Control and IRE1α KO GECs were untreated or were incubated with adriamycin (ADR; 0.5 μM) for 24 h. Cells were then stained with anti-vinculin antibody (red; FACs), phalloidin (green; F-actin cytoskeleton) and H33342 (blue; nuclei). Representative photomicrographs (A) and quantification of FACs (B) are presented. The number of FACs was reduced after ADR, and the reduction was greater in IRE1α KO cells compared to control. Bar = 25 μm. *p < 0.05, **p < 0.01, ***p < 0.0001. Mean ± SE, 3 experiments with 10–20 measurements per group per experiment. (C and D) GECs were untreated, or were incubated with ADR together with or without IXA6 (IXA; 10 μM) for 24 h. Cells were then stained as above. IXA6 attenuated the reduction in FACs induced by ADR. **p < 0.01, ***p < 0.0001. Mean ± SE, 4 experiments with 10–21 measurements per group per experiment. (E,F) GECs were incubated with IXA6 (10 μM) for 24 h. Lysates were immunoblotted as indicated. Representative immunoblots (E) and densitometric quantification of MANF normalized to actin (F) are presented **p < 0.01. Mean ± SE, 3 experiments performed in duplicate.* Next, we used control (IRE1α-replete) GECs to examine if IXA6, a drug shown to stimulate the IRE1α RNase [15], could attenuate GEC injury. GECs were untreated, or treated with adriamycin to induce sublethal injury. In parallel, GECs were treated with IXA6 or adriamycin+IXA6. IXA6 independently did not affect the number of FACs, and the drug significantly attenuated the reduction in FACs induced by adriamycin (Figures 8C,D), indicating that IXA6 was cytoprotective. Finally, although the action and specificity of IXA6 toward IRE1α was already characterized extensively [15], we verified that IXA6 activates the IRE1α pathway in GECs. Incubation of control GECs with IXA6 increased expression of MANF, an ER chaperone that we previously showed is activated via the IRE1α pathway in GECs [12]. IXA6 did not affect expression of GRP94, a chaperone that is more dependent on the activation of the ATF6 pathway, nor as expected, expression of Hsp70 (Figures 8E,F). ## Discussion Analysis of gene expression in human GN (MN, FSGS and IgAN) and DN demonstrated that there are many common upregulated glomerular genes, including genes related to the ER/UPR and autophagy. Postulating that activation of the UPR and autophagy are mechanisms that are cytoprotective in glomerular diseases, we employed connectivity mapping to identify drugs that can potentially activate these pathways. These drugs included inhibitors of Hsp90 or EGFR/HER2. Candidate drugs were then shown experimentally to induce the UPR or autophagy in cultured GECs. Finally, in a glomerular cell culture assay that correlates with glomerular damage in vivo, one candidate drug – neratinib (EGFR/HER2 inhibitor) was shown to be cytoprotective. Our analysis of the JuCKD-Glom dataset demonstrated that ~2000 genes were upregulated in glomeruli, and interestingly, more than 1,000 of these genes were common to MN, FSGS and IgAN. GO analysis revealed a number of pathways associated with ER components, ER function and the UPR, and a smaller number of pathways associated with autophagy. The majority of these pathways were present in all three GNs. We then developed sets of genes associated with the ER/UPR and autophagy, and showed that ~$90\%$ of genes in ER/UPR and autophagy protein–protein interaction networks were present in our gene sets. Analysis of the ER/UPR and autophagy genes in MN, FSGS and IgAN showed that these genes were upregulated in all of these GNs, and 35–$45\%$ of these genes were common to the three GNs. One limitation of our analyses is that the JuCKD-Glom dataset contains a relatively small number of patients with GNs. However, analyses of glomerular gene expression in three additional GN datasets confirmed activation of ER/UPR and autophagy pathways. In these additional datasets, a substantial number of ER/UPR and autophagy genes showed increased expression, and while the overlap of specific genes with those observed in JuCKD-Glom was only modest-moderate, a substantial majority of the ER/UPR and autophagy-related GO biological pathways and gene categories overlapped with JuCKD-Glom. DN is likely more heterogeneous in pathogenesis compared with GN, and only a limited number of patients with DN undergo kidney biopsies. Nevertheless, analysis of a small number of DN biopsies in JuCKD-Glom showed increased expression of ER/UPR and autophagy genes, and almost all of these genes were also increased in the three GNs. MN, FSGS and IgAN have distinct pathogenic mechanisms and there is variability in the primary cellular targets of injury in the glomerulus. Despite this, the three GNs share many upregulated genes, and specifically, all show increases in genes associated with activation of ER stress and the UPR, as well as autophagy. Furthermore, single cell RNA sequencing analysis of glomerular cells shows that basal expression of the upregulated ER/UPR and autophagy genes is not restricted to a single glomerular cell type. These results suggest that despite distinct pathogenic mechanisms believed to initiate these GNs, there may be significant commonalities in pathways that are activated to mediate glomerular injury, and common pathways that attenuate injury include the UPR and autophagy. When considering therapeutic approaches to GN, a key question is whether one searches for drugs that might be therapeutic in all of these diseases, or is each disease an entirely separate entity? *Our* gene expression analysis suggested that common pathways, i.e., UPR and autophagy, may potentially be targeted with drugs. We approached this using connectivity mapping, which aims to identify candidate drugs for a disease by relating the gene expression signatures of different drugs with that of the disease. Thus, given the GN-associated ER/UPR and autophagy gene signatures, and the premise that activation of the UPR and autophagy are mechanisms that are cytoprotective in glomerular diseases [7], we used the LINCS L1000 dataset to interrogate which drugs induce gene expression signatures in cultured cell lines that are similar to those in GN. The rationale is that candidate drugs stimulating the UPR or autophagy when given to patients with GN would further stimulate the endogenous UPR or autophagy, or provide for more sustained activation of these processes, and thereby enhance cytoprotective mechanisms, reduce protein misfolding and improve proteostasis [15]. At the same time, such drugs should not trigger overwhelming ER stress that would lead to toxicity. Our search resulted in the identification of multiple drugs that induced ER/UPR or autophagy gene signatures in cell lines. Among these, six drugs appeared to stimulate both ER/UPR and autophagy gene signatures, and for which descriptions were available in the literature (see below). We considered the latter to be an important criterion for selection of candidate drugs for further study, since description in the literature implies that the drug is either already FDA-approved or has been tested in vivo, thereby facilitating potential “repurposing” for eventual treatment of GN. The six candidate drugs were identified using glomerular gene expression signatures from several distinct datasets, and they included Hsp90 inhibitors, a EGFR/HER2 inhibitor and drugs with multiple potential actions. We attempted to substantiate the validity of our drug search by examining for potential growth factors, cytokines or protein kinases that would induce signatures similar to GNs and drugs in cultured cells; however, there were very few candidates, and this approach was not useful. A limitation of the LINCS L1000 dataset is that although it contains immortalized kidney cell lines, there are no primary glomerular cell lines in this dataset, and the large majority of cell lines are immortalized tumor cells. In addition, the L1000 technology measures expression for 978 “landmark” genes directly, while the expression values of the remaining transcriptome is estimated using a computational model [23]. We proceeded to show experimentally that three drugs, which are categorized as Hsp90 inhibitors (geldanamycin, radicicol and NVP-AUY922) increased protein expression of three ER chaperones in cultured GECs, in keeping with the activation of the UPR. Earlier, we showed that celastrol was also able to activate the UPR in vivo [39]. The Hsp90 inhibitors did not, however, stimulate significant increases in proteins involved in autophagy. In contrast, neratinib did not increase expression of the three ER chaperones we analyzed, but neratinib increased expression of LC3 and p62 proteins, in keeping with the ability of this drug to upregulate autophagy genes. Thus, the drugs identified as stimulators of the UPR or autophagy genes using connectivity mapping were, at least in part, validated experimentally. The ultimate goal after discovering cytoprotective drugs by connectivity mapping is to demonstrate their protective effect on cell injury experimentally. For this, we used a validated GEC culture assay that quantifies the density of FACs based on vinculin expression [17, 32]. We induced injury of GECs with a low dose adriamycin, which was reflected as dissolution of FACs, (a surrogate measure of podocyte foot process effacement in vivo). When GECs were treated in parallel with neratinib, the adriamycin-induced dissolution of FACs was attenuated significantly. The Hsp90 inhibitor radicicol, however, proved to be independently toxic in this assay. Thus, drugs that inhibit EGFR, such as neratinib, deserve further consideration for therapy of GN. Using the FAC assay, we also demonstrated that the IRE1α UPR pathway is critically important in protecting cells from adriamycin-induced injury, as deletion of IRE1α exacerbated adriamycin-induced dissolution of FACs. Furthermore, stimulation of the IRE1α RNase with IXA6 attenuated the adriamycin-induced reduction in FACs, supporting the view that stimulation of the UPR/autophagy is protective and a viable option for potential therapy for GN. In the future, our results in cell culture models will require corroboration in vivo. Some of the drugs that we identified in this study by connectivity mapping or drugs with analogous mechanisms of action have been examined in a few animal models of glomerular disease. For example, a Hsp90 inhibitor ameliorated high fat diet-induced renal failure in diabetes [41] and lessened disease in the MRL/lpr mouse model of systemic lupus erythematosus [42]. The EGFR inhibitor erlotinib was associated with a reduction in albuminuria and glomerular injury in diabetic nephropathy and an increase in autophagy in the kidney [43, 44]. An EGFR inhibitor also ameliorated experimental mesangial-proliferative GN [45]. Celastrol improved insulin resistance and attenuated renal injury in db/db diabetic mice [46]; however, in our experience, celastrol proved to be toxic and ineffective in a mouse model of FSGS [39]. Given that we identified neratinib through transcriptional signatures similar to the ER/UPR or autophagy, an intriguing possibility that arises from these results is that the cytoprotective effects of neratinib or other EGFR inhibitor drugs can be attributed to activation of the UPR and/or autophagy rather than blocking EGFR signaling. Actually, autophagy and EGFR signaling may be linked, as EGFR activation was reported to downregulate autophagy via phosphorylation of Beclin-1 [47]. Future studies will be required to address the specific mechanisms involving the cytoprotection of EGFR signaling inhibitors. Before effectively administering drugs in vivo, there is a need to establish non-invasive biomarkers to diagnose ER stress and to monitor drug therapy. Monitoring excretion of certain ER chaperones into the urine may be one approach [48, 49]. Such approaches will allow the ongoing monitoring of mechanistic parameters of drug therapy in addition to drug effectiveness. Our “proof of concept” study is timely, since excellent drug targets for various aspects of podocyte function are emerging, and a number of these drugs are already FDA-approved [7, 50]. By providing additional evidence that ER stress/UPR and autophagy are components of GN, the present study opens the possibility of modulating ER stress/autophagy pharmacologically as therapy for GN. ## Data availability statement The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary material. ## Author contributions AC, C-FC, and AE conceived and designed the research. C-FC, JN-B, AB, AE, and AC analyzed the data. JP and JG performed experiments. AC wrote the manuscript. All authors contributed to the article and approved the submitted version. ## Funding This work was supported by Research Grants from the McGill Initiative in Computational Medicine, Canadian Institutes of Health Research (MOP-133492, PJ9-166216, and PJ9-169678), the Kidney Foundation of Canada, and the Catherine McLaughlin Hakim Chair. ## Conflict of interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ## Publisher’s note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. ## Supplementary material The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmed.2023.1122328/full#supplementary-material ## References 1. **Kidney disease: improving global outcomes (KDIGO) glomerulonephritis work group. KDIGO clinical practice guideline for glomerulonephritis**. *Kidney Int Suppl* (2012) **2** 139-274. DOI: 10.1038/kisup.2012.9 2. Pavenstadt H, Kriz W, Kretzler M. **Cell biology of the glomerular podocyte**. *Physiol Rev* (2003) **83** 253-307. DOI: 10.1152/physrev.00020.2002 3. Greka A, Mundel P. **Cell biology and pathology of podocytes**. *Annu Rev Physiol* (2012) **74** 299-323. DOI: 10.1146/annurev-physiol-020911-153238 4. Cybulsky AV. **Membranous nephropathy**. *Contrib Nephrol* (2011) **169** 107-25. DOI: 10.1159/000313948 5. Rosenberg AZ, Kopp JB. **Focal segmental glomerulosclerosis**. *Clin J Am Soc Nephrol* (2017) **12** 502-17. DOI: 10.2215/CJN.05960616 6. Alicic RZ, Rooney MT, Tuttle KR. **Diabetic kidney disease: challenges, progress, and possibilities**. *Clin J Am Soc Nephrol* (2017) **12** 2032-45. DOI: 10.2215/CJN.11491116 7. Cybulsky AV. **Endoplasmic reticulum stress, the unfolded protein response and autophagy in kidney diseases**. *Nat Rev Nephrol* (2017) **13** 681-96. DOI: 10.1038/nrneph.2017.129 8. Cybulsky AV. **The intersecting roles of endoplasmic reticulum stress, ubiquitin-proteasome system, and autophagy in the pathogenesis of proteinuric kidney disease**. *Kidney Int* (2013) **84** 25-33. DOI: 10.1038/ki.2012.390 9. Wang M, Kaufman RJ. **Protein misfolding in the endoplasmic reticulum as a conduit to human disease**. *Nature* (2016) **529** 326-35. DOI: 10.1038/nature17041 10. Navarro-Betancourt JR, Cybulsky AV. **The IRE1alpha pathway in glomerular diseases: the unfolded protein response and beyond**. *Front Mol Med* (2022) **2** 971247. DOI: 10.3389/fmmed.2022.971247 11. Rashid HO, Yadav RK, Kim HR, Chae HJ. **ER stress: autophagy induction, inhibition and selection**. *Autophagy* (2015) **11** 1956-77. DOI: 10.1080/15548627.2015.1091141 12. Navarro-Betancourt JR, Papillon J, Guillemette J, Iwawaki T, Chung CF, Cybulsky AV. **Role of IRE1alpha in podocyte proteostasis and mitochondrial health**. *Cell Death Discov* (2020) **6** 128. DOI: 10.1038/s41420-020-00361-4 13. Navarro-Betancourt JR, Papillon J, Guillemette J, Chung CF, Iwawaki T, Cybulsky AV. **The unfolded protein response transducer IRE1alpha promotes reticulophagy in podocytes**. *Biochim Biophys Acta Mol basis Dis* (2022) **1868** 166391. DOI: 10.1016/j.bbadis.2022.166391 14. Lenoir O, Jasiek M, Henique C, Guyonnet L, Hartleben B, Bork T. **Endothelial cell and podocyte autophagy synergistically protect from diabetes-induced glomerulosclerosis**. *Autophagy* (2015) **11** 1130-45. DOI: 10.1080/15548627.2015.1049799 15. Grandjean JMD, Madhavan A, Cech L, Seguinot BO, Paxman RJ, Smith E. **Pharmacologic IRE1/XBP1s activation confers targeted ER proteostasis reprogramming**. *Nat Chem Biol* (2020) **16** 1052-61. DOI: 10.1038/s41589-020-0584-z 16. Ju W, Greene CS, Eichinger F, Nair V, Hodgin JB, Bitzer M. **Defining cell-type specificity at the transcriptional level in human disease**. *Genome Res* (2013) **23** 1862-73. DOI: 10.1101/gr.155697.113 17. Chung CF, Kitzler T, Kachurina N, Pessina K, Babayeva S, Bitzan M. **Intrinsic tumor necrosis factor-alpha pathway is activated in a subset of patients with focal segmental glomerulosclerosis**. *PLoS One* (2019) **14** e0216426. DOI: 10.1371/journal.pone.0216426 18. Grayson PC, Eddy S, Taroni JN, Lightfoot YL, Mariani L, Parikh H. **Metabolic pathways and immunometabolism in rare kidney diseases**. *Ann Rheum Dis* (2018) **77** annrheumdis-2017-212935-3. DOI: 10.1136/annrheumdis-2017-212935 19. Park S, Yang SH, Jeong CW, Moon KC, Kim DK, Joo KW. **RNA-Seq profiling of microdissected glomeruli identifies potential biomarkers for human IgA nephropathy**. *Am J Physiol Renal Physiol* (2020) **319** F809-21. DOI: 10.1152/ajprenal.00037.2020 20. Ringner M. **What is principal component analysis?**. *Nat Biotechnol* (2008) **26** 303-4. DOI: 10.1038/nbt0308-303 21. Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM. **Gene ontology: tool for the unification of biology. The Gene Ontology Consortium**. *Nat Genet* (2000) **25** 25-9. DOI: 10.1038/75556 22. Kamburov A, Pentchev K, Galicka H, Wierling C, Lehrach H, Herwig R. **ConsensusPathDB: toward a more complete picture of cell biology**. *Nucleic Acids Res* (2011) **39** D712-7. DOI: 10.1093/nar/gkq1156 23. Subramanian A, Narayan R, Corsello SM, Peck DD, Natoli TE, Lu X. **A next generation connectivity map: L1000 platform and the first 1,000,000 profiles**. *Cell* (2017) **171** 1437-1452.e17. DOI: 10.1016/j.cell.2017.10.049 24. Keenan AB, Jenkins SL, Jagodnik KM, Koplev S, He E, Torre D. **The library of integrated network-based cellular signatures NIH program: system-level cataloging of human cells response to perturbations**. *Cell Syst* (2018) **6** 13-24. DOI: 10.1016/j.cels.2017.11.001 25. Lamb J, Crawford ED, Peck D, Modell JW, Blat IC, Wrobel MJ. **The connectivity map: using gene-expression signatures to connect small molecules, genes, and disease**. *Science* (2006) **313** 1929-35. DOI: 10.1126/science.1132939 26. Binns D, Dimmer E, Huntley R, Barrell D, O'Donovan C, Apweiler R. **QuickGO: a web-based tool for gene ontology searching**. *Bioinformatics* (2009) **25** 3045-6. DOI: 10.1093/bioinformatics/btp536 27. Walter W, Sanchez-Cabo F, Ricote M. **GOplot: an R package for visually combining expression data with functional analysis**. *Bioinformatics* (2015) **31** 2912-4. DOI: 10.1093/bioinformatics/btv300 28. Xia J, Benner MJ, Hancock RE. **NetworkAnalyst--integrative approaches for protein-protein interaction network analysis and visual exploration**. *Nucleic Acids Res* (2014) **42** W167-74. DOI: 10.1093/nar/gku443 29. Kanehisa M, Furumichi M, Tanabe M, Sato Y, Morishima K. **KEGG: new perspectives on genomes, pathways, diseases and drugs**. *Nucleic Acids Res* (2017) **45** D353-61. DOI: 10.1093/nar/gkw1092 30. Shoulders MD, Ryno LM, Genereux JC, Moresco JJ, Tu PG, Wu C. **Stress-independent activation of XBP1s and/or ATF6 reveals three functionally diverse ER proteostasis environments**. *Cell Rep* (2013) **3** 1279-92. DOI: 10.1016/j.celrep.2013.03.024 31. Elimam H, Papillon J, Kaufman DR, Guillemette J, Aoudjit L, Gross RW. **Genetic ablation of calcium-independent phospholipase A2gamma induces glomerular injury in mice**. *J Biol Chem* (2016) **291** 14468-82. DOI: 10.1074/jbc.M115.696781 32. Kachurina N, Chung CF, Benderoff E, Babayeva S, Bitzan M, Goodyer P. **Novel unbiased assay for circulating podocyte-toxic factors associated with recurrent focal segmental glomerulosclerosis**. *Am J Physiol Renal Physiol* (2016) **310** F1148-56. DOI: 10.1152/ajprenal.00349.2015 33. Lu Y, Ye Y, Bao W, Yang Q, Wang J, Liu Z. **Genome-wide identification of genes essential for podocyte cytoskeletons based on single-cell RNA sequencing**. *Kidney Int* (2017) **92** 1119-29. DOI: 10.1016/j.kint.2017.04.022 34. Lu Y, Ye Y, Yang Q, Shi S. **Single-cell RNA-sequence analysis of mouse glomerular mesangial cells uncovers mesangial cell essential genes**. *Kidney Int* (2017) **92** 504-13. DOI: 10.1016/j.kint.2017.01.016 35. Park J, Shrestha R, Qiu C, Kondo A, Huang S, Werth M. **Single-cell transcriptomics of the mouse kidney reveals potential cellular targets of kidney disease**. *Science* (2018) **360** 758-63. DOI: 10.1126/science.aar2131 36. Madhavan A, Kok BP, Rius B, Grandjean JMD, Alabi A, Albert V. **Pharmacologic IRE1/XBP1s activation promotes systemic adaptive remodeling in obesity**. *Nat Commun* (2022) **13** 608. DOI: 10.1038/s41467-022-28271-2 37. Kim S, Chen J, Cheng T, Gindulyte A, He J, He S. **PubChem 2019 update: improved access to chemical data**. *Nucleic Acids Res* (2019) **47** D1102-9. DOI: 10.1093/nar/gky1033 38. Wang Z, Lachmann A, Keenan AB, Ma'ayan A. **L1000FWD: fireworks visualization of drug-induced transcriptomic signatures**. *Bioinformatics* (2018) **34** 2150-2. DOI: 10.1093/bioinformatics/bty060 39. Yee A, Papillon J, Guillemette J, Kaufman DR, Kennedy CRJ, Cybulsky AV. **Proteostasis as a therapeutic target in glomerular injury associated with mutant alpha-actinin-4**. *Am J Physiol Renal Physiol* (2018) **315** F954-66. DOI: 10.1152/ajprenal.00082.2018 40. Elimam H, Papillon J, Guillemette J, Navarro-Betancourt JR, Cybulsky AV. **Genetic ablation of calcium-independent phospholipase A2gamma exacerbates glomerular injury in adriamycin nephrosis in mice**. *Sci Rep* (2019) **9** 16229. DOI: 10.1038/s41598-019-52834-x 41. Zhang HM, Dang H, Kamat A, Yeh CK, Zhang BX. **Geldanamycin derivative ameliorates high fat diet-induced renal failure in diabetes**. *PLoS One* (2012) **7** e32746. DOI: 10.1371/journal.pone.0032746 42. Shimp SK, Chafin CB, Regna NL, Hammond SE, Read MA, Caudell DL. **Heat shock protein 90 inhibition by 17-DMAG lessens disease in the MRL/lpr mouse model of systemic lupus erythematosus**. *Cell Mol Immunol* (2012) **9** 255-66. DOI: 10.1038/cmi.2012.5 43. Zhang MZ, Wang Y, Paueksakon P, Harris RC. **Epidermal growth factor receptor inhibition slows progression of diabetic nephropathy in association with a decrease in endoplasmic reticulum stress and an increase in autophagy**. *Diabetes* (2014) **63** 2063-72. DOI: 10.2337/db13-1279 44. Li Z, Li Y, Overstreet JM, Chung S, Niu A, Fan X. **Inhibition of epidermal growth factor receptor activation is associated with improved diabetic nephropathy and insulin resistance in type 2 diabetes**. *Diabetes* (2018) **67** 1847-57. DOI: 10.2337/db17-1513 45. Rintala JM, Savikko J, Rintala SE, Palin N, Koskinen PK. **Epidermal growth factor receptor inhibition with erlotinib ameliorates anti-thy 1.1-induced experimental glomerulonephritis**. *J Nephrol* (2016) **29** 359-65. DOI: 10.1007/s40620-015-0233-x 46. Kim JE, Lee MH, Nam DH, Song HK, Kang YS, Lee JE. **Celastrol, an NF-kappaB inhibitor, improves insulin resistance and attenuates renal injury in db/db mice**. *PLoS One* (2013) **8** e62068. DOI: 10.1371/journal.pone.0062068 47. Fantin VR, Abraham RT. **Self-eating limits EGFR-dependent tumor growth**. *Cells* (2013) **154** 1184-6. DOI: 10.1016/j.cell.2013.08.040 48. Kim Y, Lee H, Manson SR, Lindahl M, Evans B, Miner JH. **Mesencephalic astrocyte-derived neurotrophic factor as a urine biomarker for endoplasmic reticulum stress-related kidney diseases**. *J Am Soc Nephrol* (2016) **27** 2974-82. DOI: 10.1681/ASN.2014100986 49. Tousson-Abouelazm N, Papillon J, Guillemette J, Cybulsky AV. **Urinary ERdj3 and mesencephalic astrocyte-derived neutrophic factor identify endoplasmic reticulum stress in glomerular disease**. *Lab Investig* (2020) **100** 945-58. DOI: 10.1038/s41374-020-0416-5 50. Park SJ, Kim Y, Yang SM, Henderson MJ, Yang W, Lindahl M. **Discovery of endoplasmic reticulum calcium stabilizers to rescue ER-stressed podocytes in nephrotic syndrome**. *Proc Natl Acad Sci U S A* (2019) **116** 14154-63. DOI: 10.1073/pnas.1813580116
--- title: 'Social determinants of patiromer adherence and abandonment: An observational, retrospective, real-world claims analysis' authors: - Nathan Kleinman - Jennifer Kammerer - Kevin LaGuerre - Charuhas V. Thakar journal: PLOS ONE year: 2023 pmcid: PMC10042334 doi: 10.1371/journal.pone.0281775 license: CC BY 4.0 --- # Social determinants of patiromer adherence and abandonment: An observational, retrospective, real-world claims analysis ## Abstract ### Background Hyperkalemia is a frequent and serious complication in chronic kidney disease (CKD) that can impede continuation of beneficial evidence-based therapies. Recently, novel therapies such as patiromer have been developed to treat chronic hyperkalemia, but their optimal utility hinges on adherence. Social determinants of health (SDOH) are critically important and can impact both medical conditions and treatment prescription adherence. This analysis examines SDOH and their influence on adherence to patiromer or abandonment of prescriptions for hyperkalemia treatment. ### Methods This was an observational, retrospective, real-world claims analysis of adults with patiromer prescriptions and 6- and 12-months pre- and post-index prescription data in Symphony Health’s Dataverse during 2015–2020, and SDOH from census data. Subgroups included patients with heart failure (HF), hyperkalemia-confounding prescriptions, and any CKD stages. Adherence was defined as >$80\%$ of proportion of days covered (PDC) for ≥60 days and ≥6 months, and abandonment as a portion of reversed claims. Quasi-Poisson regression modeled the impact of independent variables on PDC. Abandonment models used logistic regression, controlling for similar factors and initial days’ supply. Statistical significance was $p \leq 0.05.$ ### Results $48\%$ of patients at 60 days and $25\%$ at 6 months had a patiromer PDC >$80\%$. Higher PDC was associated with older age, males, Medicare/Medicaid coverage, nephrologist prescribed, and those receiving renin-angiotensin-aldosterone system inhibitors. Lower PDC correlated with higher out-of-pocket cost, unemployment, poverty, disability, and any CKD stage with comorbid HF. PDC was better in regions with higher education and income. ### Conclusions SDOH (unemployment, poverty, education, income) and health indicators (disability, comorbid CKD, HF) were associated with low PDC. Prescription abandonment was higher in patients with prescribed higher dose, higher out-of-pocket costs, those with disability, or designated White. Key demographic, social, and other factors play a role in drug adherence when treating life-threatening abnormalities such as hyperkalemia and may influence patient outcomes. ## Background Although various acute and chronic conditions impact potassium (K+) homeostasis, hyperkalemia is a potentially life-threatening electrolyte imbalance often problematic in patients with chronic kidney disease (CKD), heart failure (HF), and diabetes mellitus (DM) [1]. Hyperkalemia can occur in up to $34.6\%$ of patients with CKD and $30\%$ of those with HF [2]. The progressive nature of CKD and HF further increases the risk of hyperkalemia as kidney function declines over time. Treatment with renin-angiotensin-aldosterone system inhibitors (RAASi) has revolutionized the treatment of CKD and HF [1, 3–5]. However, onset of persistent hyperkalemia can impede our ability to offer appropriate evidence-based therapies. Until recently, sodium polystyrene sulfonate (SPS) was the only treatment for hyperkalemia, which was typically utilized in emergency circumstances or for short-term use. Over the past few years, novel therapies have been developed and approved for chronic management of hyperkalemia. Patiromer, the first novel approved therapy, is a sodium-free, nonabsorbed, daily-use K+-binding polymer indicated for treating hyperkalemia. Its use has been supported by the pivotal study OPAL-HK and other efficacy data in diverse, medically complex patients [6, 7]. Clinical and real-world evidence suggests patients are more likely to be on RAASi therapy or at optimal RAASi doses while taking a concurrent K+ binder such as patiromer [6–9]. Successful treatment of chronic comorbid conditions, such as CKD or HF, and their complications hinges on medication adherence. Several factors influence patient adherence, including social determinants of health (SDOH). SDOH serve as nonclinical markers of general security, including social support, community connectedness, economic stability, and living environment. The World Health Organization defines SDOH as avoidable inequities in health that arise from circumstances in which people grow, live, work, and age [10], further shaped by political, social, and economic forces. Such factors (eg, income, education, household size/support, environmental safety, proximity to care) may influence health behaviors and correlate with health risks such as medication nonadherence, delays in seeking care, lack of a primary care provider, and toxic stress [10, 11]. Optimal care delivery in chronic conditions such as CKD, HF, or DM is intricately linked with medical therapy, and SDOH. When combined with claims and clinical data, SDOH may enhance complete care planning, address patient-specific barriers by individualizing interventions, and assess the most at-risk groups. The objective of this study is to assess the effects of real-world experience of SDOH on patiromer adherence or abandonment of initial prescription. We aim to achieve our objective by utilizing the strengths of a large national claims-based database, which allows correlation of pharmacy, clinical, and key SDOH parameters. This study was executed in a multi-payer data set that represents a broader real-world population than a single-payer data set. It also incorporated relevant methodologies that reasonably excluded patients who may—in a single-payer data set—switch insurance and be otherwise ineligible for study inclusion. Adherence methodology (PDC) was specifically chosen as the well-accepted approach to HEDIS quality measurement of adherence, whereas prior patiromer studies used “continuous exposure” or modified PDC methodology. In addition, this is the first known assessment of abandonment patterns, which indicate a different barrier to access where a patient has successfully been prescribed patiromer but never started it, for unknown reasons. ## Materials and study design This was an observational, retrospective analysis of data on adults with prescription claims, using a predefined data extract from Symphony Health Analytics claims. Census data supplemented ZIP Code–based regional information at a population level, providing insights on population density, race, per-capita income, education, poverty level, and disability status. This retrospective, administrative claims database analysis was based on historic deidentified patient data and did not involve patients directly; therefore, institutional review board/ethics committee approval was not necessary or applicable. ## Patient population The adherence analysis included patients with a final approved patiromer claim during 2015–2018, and the abandonment analysis included those with a final approved or reversed patiromer claim during 2015–2018. Patients were at least 18 years old 6 months prior to the index date, which was defined as the first patiromer fill date, and all patients had at least 6 months of pre-index and 12 months of post-index (any) prescription activity. Subgroups evaluated included those with HF, hyperkalemia-confounding medications, and distinct stages of CKD (International Classification of Diseases, 10th edition–based CKD stages 1–2 vs CKD stages 3–4 vs CKD stage 5/end-stage renal disease [ESRD]). ## Outcomes and variables The primary adherence outcome was proportion of days covered (PDC). To identify SDOH associations with starting or continuing patiromer, patients were considered adherent if the PDC calculated over 60 days was >$80\%$. To broadly gauge and describe chronicity of patiromer use once started, PDC was also calculated over 6 months. Abandonment was defined as the proportion of reversed initial patiromer claims among all reversed or approved claims with no evidence of a subsequent additional adjudicated paid claim. This was applied to only a first patiromer fill without prior history of patiromer in the look back period and did not include any fill attempts after the first (initial) fill. Thus, we defined an indirect measure of abandonment by representing the frequency of prescriptions never picked up once filled. The following population-descriptive statistics were identified for study patients: gender; age; region (based on the first digit of a patient’s ZIP Code); initial prescription plan type; initial patiromer dosage strength; initial prescription mode of transmission; initial prescriber specialty; pre-index number of daily-use prescription medications, use of other K+ binders, or hyperkalemia diagnosis; and post-index patiromer out-of-pocket cost per day supplied, non-patiromer out-of-pocket cost/month. The patient 3-digit ZIP Code was linked to census-based averages for employment status, race, poverty level, disability, education, per-capita income, and population density. ## Statistical methods Both 60-day and 6-month PDC values were characterized using quasi-Poisson regression models. Quasi-Poisson models were chosen because they account for overdispersion in the days covered count data. Independent variables were controlled for in the regression models to assess and measure their impacts on PDC. These variables included the patient, prescription, provider, plan, and regional census factors listed above, as well as post-index K+-confounding medications and post-index medical claim activity for HF, CKD, and ESRD. Prescription K+-confounding medications were grouped by class and included RAASi, K+ supplements, nonsteroidal anti-inflammatory drugs, immunosuppressants, K+-increasing antibiotics, and others (amiloride, sacubitril/valsartan, and triamterene). Only one regional census variable could be included in the regression model at a time due to collinearity issues. Prescription abandonment was modeled using logistic regression, controlling for factors similar to those in the adherence models in addition to initial dispensed days’ supply and pharmacy type. The regression models were run using RStudio version 1.2.5042 [12]. Statistical significance in all regression models was set at p-value <0.05. This study used observational analysis of an aggregate dataset where the data are anonymized and considered exempt from further informed consent. ## Results In the adherence analysis, initially 33,329 patients had an approved patiromer prescription claim, of which 33,250 met age requirements and 24,405 met all requirements (index, age, claim activity). Average age was 63±13 years, and $41\%$ of the population was female. Patients were well represented across the United States, with the heaviest prescribing in the southeastern, southern, and western areas (Fig 1). **Fig 1:** *Patiromer prescribing, adherence, and abandonment rates by region.n = patiromer patients (% total). Numbers 0–9 indicate the first digit of a region’s ZIP code. Abbreviations: ABR, abandonment rate; PDC2M, 60-day proportion of days covered; PDC6M, 6-month proportion of days covered. Adapted from iStock image. https://www.istockphoto.com/vector/united-states-of-america-map-us-blank-map-template-outline-usa-map-background-gm1301588831-393587962.* In the pre-index period, defined as that before starting patiromer, patients averaged 9.9±5.4 unique daily-use medications and $47.19±$96.30 in out-of-pocket costs per month on all daily-use prescription medications. Twenty-five percent of patients had a pre-index SPS prescription, with an average gap of 53±49 (median 35) days between last SPS fill and first patiromer fill (Table 1). **Table 1** | Patiromer Patients (N = 24,405) | Mean±SD or Count (%) | | --- | --- | | 6-Month Pre-Index | | | Number of unique daily-use medications | 9.9±5.4 | | Average total out-of-pocket per month, any (non-patiromer) prescription, USD | 47.19±96.30 | | Patients with SPS use | 5,991 (25) | | Average days from last SPS prescription fill date to patiromer start | 52.9±48.9 (median 35 days) | | 6-Month Post-Index | | | Average number of fills per patient | 2.9±2.0 | | Average days’ supply per prescription | 34.4±17.3 | | Patients with initial patiromer dose of 8.4 g | 22,453 (92) | | Average number of dose changes | 0.070±0.31 | | Average out-of-pocket per patiromer prescription, USD | 52.48±166.84 | | Average patiromer out-of-pocket cost per day supplied, USD | 1.79±5.22 | | Average total out-of-pocket per month, any non-patiromer prescription, USD | 46.34±152.86 | | Number of unique daily-use medications | 10.8±5.4 | | Average 60-day patiromer PDC, % | 72.9±0.26 | | PDC >80% | 11,812 (48) | | PDC >90% | 10,490 (43) | | Average 6-month patiromer PDC, % | 49.4±0.31 | | PDC >80% | 6,180 (25) | | PDC >90% | 4,060 (17) | In the post-index analyses (defined as after initiating patiromer), most ($92\%$) patients started patiromer at a dose of 8.4 g (with few [0.07±0.30] dose changes). Average number of fills was 2.9±2.0, while average days’ supply was 34.4±17.3 days. Patients showed an increase of about one prescription in mean number of unique daily-use medications from pre-index (9.9±5.4) to post-index (10.8±5.4). Mean out-of-pocket cost for a patiromer prescription was $52.48±$166.84, averaging out to $1.79±$5.22 per day supplied (Table 1), while average out-of-pocket spending on non-patiromer prescriptions remained similar at $46.34±$152.86. Average 60-day PDC was $72.9\%$; $48\%$ of the patients had 60-day PDC >$80\%$. At 6 months, average PDC was lower ($49.4\%$), and $25\%$ were adherent with PDC >$80\%$ (Table 1). However, adherence did not appear to be dose-dependent. A number of independent variables in the adherence regression models had a significant association with PDC, including the region of the country in which the patient lived. At 60 days, adjusted patiromer PDC remained relatively high throughout all regions (range: $69\%$ in region 2 to $74\%$ in regions 1, 4, 8, and 9; Fig 1). At 6 months, adjusted PDC ranged from $46\%$ to $51\%$ across regions (Fig 1). Adjusted mean PDC values for additional significant categorical variables are shown in Table 2. Holding other variables in the model constant, those variables favoring high PDC included older age, male gender, Medicare coverage, initial prescription by nephrologist (vs generalist or cardiologist), and recent hyperkalemia diagnosis or SPS treatment (Tables 2 and 3). For continuous independent factors associated with PDC, Fig 2 illustrates association of adherence with regionally higher per capita income and concentrations of advanced education (bachelor’s or master’s degree). **Fig 2:** *60-day and 6-month regression-adjusted PDC for continuous independent variables.***$p \leq 0.001$; **$p \leq 0.01$; *$p \leq 0.05$; †$$p \leq 0.11$$; ‡$$p \leq 0.17.$$ Abbreviations: PDC, proportion of days covered; USD, United States dollars.* TABLE_PLACEHOLDER:Table 2 TABLE_PLACEHOLDER:Table 3 In the multivariate analysis, the lowest PDC rates were observed in patients in eastern and southern regions (ZIP 0, 2, and 7) and were associated with cash-paid transactions and those with comorbid HF and advanced stages of renal dysfunction (Tables 2 and 3). Patiromer adherence declined where unemployment, poverty, and disability were higher, but adherence was most adversely affected by patiromer out-of-pocket costs per days supplied (Fig 2). Only half of patients on patiromer were also on RAASi (Table 3). Patiromer-specific out-of-pocket costs significantly influenced adherence (PDC), in contrast to either total (non-patiromer) medication costs or pre-index total number of daily use medications (Fig 3). **Fig 3:** *Regression-adjusted abandonment rate for first patiromer prescriptions (all p<0.001).Abbreviations: USD, United States dollars.* In the abandonment analysis, 36,185 patients had an approved or reversed patiromer claim, of which 36,110 met age criteria and 25,762 met all inclusion criteria. The average age was 63±13 years, and $41\%$ were female. Regional distribution was similar to that shown in Fig 1. Table 4 provides regression-adjusted abandonment rates for categorical variables. Abandonment ranged from $26.0\%$ to $34\%$ throughout regions. Holding other variables constant, groups with the lowest abandonment (i.e., the groups most likely to pick up a prescription once filled) were those who were age 55 to <65 years, were in region 7 (south central), were covered by Medicaid or commercial/employer plans, or had a recent hyperkalemia diagnosis (Tables 3 and 4). Unexpectedly, abandonment was less likely in patients with a greater baseline medication burden (Fig 3). **Table 4** | Unnamed: 0 | N | Regression-Adjusted Abandonment Rate (%) | p-Value | | --- | --- | --- | --- | | Demographic variables | | | | | Age <45 years (ref) | 2430.0 | 31.0 | | | 45 ≤ Age <55 years | 3499.0 | 29.0 | 0.11 | | 55 ≤ Age <65 years | 6259.0 | 28.0 | 0.003 | | 65 ≤ Age <75 years | 7806.0 | 31.0 | 0.78 | | Age ≥75 years | 5768.0 | 29.0 | 0.18 | | Female | 10473.0 | 29.0 | 0.06 | | Male (ref) | 15289.0 | 30.0 | | | ZIP Code | | | | | 1st digit = 0 | 1335.0 | 32.0 | 0.09 | | 1st digit = 1 | 2633.0 | 28.0 | 0.45 | | 1st digit = 2 | 2184.0 | 31.0 | 0.07 | | 1st digit = 3 | 4802.0 | 28.0 | 0.64 | | 1st digit = 4 | 2099.0 | 32.0 | 0.04 | | 1st digit = 5 | 659.0 | 31.0 | 0.38 | | 1st digit = 6 | 2061.0 | 34.0 | <0.001 | | 1st digit = 7 | 4679.0 | 26.0 | 0.005 | | 1st digit = 8 | 1618.0 | 32.0 | 0.02 | | 1st digit = 9 (ref) | 3692.0 | 29.0 | | | Days’ supply | | | | | Supply <28 days | 1537.0 | 29.9 | 0.60 | | Supply ≥28 days and ≤30 days (ref) | 20914.0 | 29.3 | | | Supply >30 days | 3311.0 | 29.8 | 0.54 | | Pharmacy type | | | | | Retail (ref) | 17082.0 | 31.0 | | | Mail/specialty | 4536.0 | 32.0 | 0.15 | | Non-retail and non-mail/specialty | 4144.0 | 23.0 | <0.001 | | Pharmacy plan type | | | | | PBM | 1986.0 | 30.0 | 0.104 | | Cash | 768.0 | 28.0 | 0.03 | | Employer/commercial | 2713.0 | 22.0 | <0.001 | | Medicaid | 1596.0 | 23.0 | <0.001 | | Medicare (ref) | 16322.0 | 32.0 | | | Other | 2377.0 | 25.0 | <0.001 | | Prescription strength | | | | | 8.4 g (ref) | 23773.0 | 29.0 | | | 16.8 g | 1826.0 | 34.0 | <0.001 | | 25.2 g | 163.0 | 36.0 | 0.05 | | Initial prescriber classification and specialization | | | | | Internal medicine; nephrology (ref) | 18000.0 | 29.0 | | | Generalists | 3969.0 | 29.0 | 0.47 | | Internal medicine; cardiovascular disease | 415.0 | 32.0 | 0.29 | | Other | 3378.0 | 30.0 | 0.72 | Factors associated with the highest rates of abandonment (groups least likely to pick up a prescription once filled) resided in region 6 (Midwest), had Medicare coverage, and had starting doses above 8.4 g (Table 4). Abandonment was also higher where either patiromer or non-patiromer out-of-pocket costs were higher and in regions with greater concentrations of disabled or White populations (Fig 3). Gender, comorbidities, and concomitant hyperkalemia-confounding medications had little or no influence on abandonment rates (Tables 4 and 5). **Table 5** | Unnamed: 0 | N | % | p-Value | | --- | --- | --- | --- | | Specific 6-Month Pre-Index Factors | | | | | SPS Rx | 6103.0 | 30.0 | 0.42 | | No SPS Rx (ref) | 19659.0 | 29.0 | | | Hyperkalemia Dx | 7232.0 | 28.0 | 0.001 | | No Hyperkalemia Dx (ref) | 18530.0 | 30.0 | | | No HF (ref) | 22387.0 | 29.0 | | | HF (no CKD/ESRD) | 525.0 | 31.0 | 0.39 | | HF and CKD 1–4 | 1557.0 | 30.0 | 0.37 | | HF and (CKD 5 or ESRD) | 1293.0 | 31.0 | 0.32 | | 6-Month Pre-Index Confounding Medication Use | | | | | RAASi Rx | 13692.0 | 30.0 | 0.42 | | No RAASi Rx (ref) | 12070.0 | 29.0 | | | Potassium Supplement Rx | 554.0 | 30.0 | 0.64 | | No Potassium Supplement Rx (ref) | 25208.0 | 29.0 | | | NSAID <28-day Rx | 1088.0 | 28.0 | 0.31 | | No <28-day NSAID Rx (ref) | 24674.0 | 29.0 | | | NSAID ≥28-day Rx | 1432.0 | 28.0 | 0.33 | | No ≥28-day NSAID Rx (ref) | 24330.0 | 29.0 | | ## Discussion Since the release of novel daily-use agents to treat hyperkalemia, this is among the first real-world analyses describing abandonment of or adherence to K+-binder prescriptions. Moreover, the analysis utilizes the strengths of claims data to examine the effect of SDOH on these two critical outcomes to inform opportunities that may improve outcomes in often-complex comorbid conditions. Our analysis examined the impact of SDOH, demographic factors, and comorbidities by looking at K+-binder use in two ways: i) adherence based on PDC; and ii) abandonment or proportion filled at a pharmacy but never picked up by the patient. First, while the mean 60-day PDC was $72.9\%$, 6-month PDC dropped to $49.4\%$; both rates were below the Centers for Medicare and Medicaid Services–defined threshold for long-term adherence with chronic medications [13]. The portion of patients achieving the PDC >$80\%$ threshold at 60 days and 6 months was $48.4\%$ and $25.3\%$, respectively. Although patiromer was studied for chronic daily use, long-term use in practice appears limited, suggesting an important opportunity to identify and address the barriers to not only continuing treatment once started but also those preventing patients with a prescription from starting it. In multivariate analyses, the lowest PDC rates were observed in patients in the eastern and southern regions; associated with cash-paid transactions; and in those with HF particularly with comorbid advanced CKD stages. Patiromer adherence declined where unemployment, poverty, and disability were higher, but was most adversely affected by patiromer out-of-pocket costs per days supplied. Interestingly, neither medication burden prior to use of patiromer nor out-of-pocket costs for other prescriptions influenced PDC. Another observation was that PDC was higher when patiromer was initially prescribed by a specialist, which may reflect the relationship or communication quality. Beyond nonadherence, some medication-specific flags of low literacy that may be tackled include frequently missed appointments, identifying medications by sight rather than label, inability to name a medication, how to use it or its purpose, and asking few questions [14]. For patiromer patients, providers may consider demonstrating how to mix and use it, describing need-to-know points in plain language, and using teach-back methods to ensure understanding [14]. This combined with the findings above regarding 6-month PDC indicate that different patient and physician education or awareness strategies are needed to improve long-term adherence in the treatment of chronic conditions. Abandonment rates, defined as proportion of prescriptions never picked up once filled, ranged between $22\%$ to $34\%$ depending on baseline patient characteristics. Patients with the lowest rates of abandoning a patiromer prescription were those who were prescribed the lowest commercially available dose of patiromer (8.4 g) in age groups 55 to <65, living in region 7 (South Central), had drug coverage under Medicaid or commercial/employer plans, or had a recent hyperkalemia diagnosis. Abandonment was greater with higher out-of-pocket costs or in geographic areas where a higher proportion of the population reported disability. Daily medication complexity, prescription cost, and age-related cognitive changes are often thought to be substantial pressures affecting medication adherence [15]. In this study, patiromer adherence was lowest in the youngest age group and highest in the 65 years and older group, consistent with (but perhaps counterintuitive) other data showing higher adherence in older patients [15]. Regions that had patients with more upper education and higher income were more likely to abandon an original patiromer prescription, but also showed greater adherence rates after starting. We speculate that higher health literacy may favor self-management or ability to achieve lifestyle or dietary changes to counter hyperkalemia before or instead of drug treatment [16]. Our observation suggests that patiromer was more heavily prescribed in regions that also included the OPAL-HK [7] pivotal trial study sites (CA, FL, GA, MO, NY, PA, TX) and, like the OPAL-HK study population, patients were more likely to be older, Medicare-eligible men with substantial polypharmacy indicative of multiple chronic conditions. Out-of-pocket spending for prescription drugs appears to be in line with other reported analyses in a Medicare population [17]. While those in OPAL-HK initiated patiromer dosing as stratified by baseline K+ level, $92\%$ of this real-world population started on the lowest dose of 8.4 g. The current study showed approximately half of patients on patiromer also had a RAASi prescription, which may partially be attributed to gaps in claims data (ie, failure to capture cash-paid transactions, including those filled under low-cost generic programs). Our real-world data also show substantial usage of patiromer in patients with CKD stage 5 or ESRD, while heart failure diagnoses were less frequent ($20\%$) than in OPAL-HK ($42\%$). Our study provides unique insight into prescribing patterns that otherwise may get generalized when compared historically to overlapping geography involved in clinical trials for the same condition. Although this study assessed only prespecified comorbidities (CKD/ESRD, HF), real-world patients in this analysis were taking an average of nearly 10 daily-use medications before starting patiromer, which may serve as a reasonable surrogate of complex health status. Taking more than four or five daily-use medications is often considered polypharmacy when assessing medication-related risks (eg, harmful effects, drug interactions, hospitalizations) [18–20]. One-third of chronically ill patients struggle to afford food, medication, or both [21]. Those with comorbidities are likely to have added financial stress, which accounts for medication nonadherence in as much as $40\%$ of patients with diabetes [21]. These financial obstacles may require essential provider time and development of a well-established, trusting relationship. Our study has limitations. A large observational database of claims-level de-identified data provides associative information. On the other hand, it may uniquely offer insight on payer-agnostic prescription patterns and correlates with real-life socioeconomic indicators over large geographic/national areas. Similarly, the comorbid disease burden is derived from claims and can be specific but not very sensitive. Unlike registries, claims data may be incompletely captured or have gaps. Thus, this study cannot conclusively correlate adherence with patient outcomes. However, pharmacoepidemiologic analyses support broader correlations and highlight the differences between trial results and adoption of such into real-world practices, and they may inform interventional study planning to improve patient outcomes. In summary, specific SDOH (unemployment, poverty, education, income) and health indicators (disability, comorbid CKD and HF) were associated with low PDC for patiromer use. Prescription abandonment was higher in patients prescribed higher initial patiromer doses, those with high out-of-pocket costs, those with disability, or those who were designated White. We successfully demonstrated that key SDOH play important roles in optimizing treatments for chronic illnesses. Considering SDOH barriers when designing interventions to improve drug access and adherence may be important to chronic hyperkalemia management and associated comorbidities. Furthermore, different approaches may be needed to overcome barriers to abandonment, those contributing to non-adherence, and other patient-specific factors. Further research on such individualized approaches is needed, as is further exploration on reasons patients abandon a first patiromer prescription or are non-adherent with chronic use. Lastly, examination of the impact of SDOH on prescription adherence may be extrapolated to other important drug classes that impact health outcomes, particularly emerging novel therapies in CKD. ## References 1. Einhorn LM, Zhan M, Hsu VD, Walker LD, Moen MF, Seliger SL. **The frequency of hyperkalemia and its significance in chronic kidney disease**. *Arch Intern Med* (2009.0) **169** 1156-62. DOI: 10.1001/archinternmed.2009.132 2. Epstein M, Alvarez PJ, Reaven NL, Funk SE, McGaughey KJ, Brenner MS. **Evaluation of clinical outcomes and costs based on prescribed dose level of renin-angiotensin-aldosterone system inhibitors.**. *Am J Manag Care.* (2016.0) **22** S311-24. PMID: 27668789 3. Weir MR, Rolfe M. **Potassium homeostasis and renin-angiotensin-aldosterone system inhibitors**. *Clin J Am Soc Nephrol* (2010.0) **5** 531-48. DOI: 10.2215/CJN.07821109 4. Filippatos G, Anker SD, Agarwal R, Ruilope LM, Rossing P, Bakris GL. **Finerenone reduces risk of incident heart failure in patients with chronic kidney disease and type 2 diabetes: analyses from the FIGARO-DKD trial**. *Circulation* (2022.0) **145** 437-47. DOI: 10.1161/CIRCULATIONAHA.121.057983 5. Pitt B, Filippatos G, Agarwal R, Anker SD, Bakris GL, Rossing P. **Cardiovascular events with finerenone in kidney disease and type 2 diabetes**. *N Engl J Med* (2021.0) **385** 2252-63. DOI: 10.1056/NEJMoa2110956 6. 6VELTASSA® (patiromer) [package insert]. Redwood City, CA: Relypsa, Inc. May, 2018. 7. Weir MR, Bakris GL, Bushinsky DA, Mayo MR, Garza D, Stasiv Y. **Patiromer in patients with kidney disease and hyperkalemia receiving RAAS inhibitors**. *N Engl J Med* (2015.0) **372** 211-21. DOI: 10.1056/NEJMoa1410853 8. Weir MR, Bushinsky DA, Benton WW, Woods SD, Mayo MR, Arthur SP. **Effect of patiromer on hyperkalemia recurrence in older chronic kidney disease patients taking RAAS inhibitors**. *Am J Med* (2018.0) **131** 555-564.e3. DOI: 10.1016/j.amjmed.2017.11.011 9. Desai NR, Rowan CG, Alvarez PJ, Fogli J, Toto RD. **Hyperkalemia treatment modalities: a descriptive observational study focused on medication and healthcare resource utilization.**. *PLoS One* (2020.0) **15** e0226844. DOI: 10.1371/journal.pone.0226844 10. 10World Health Organization. Social Determinants of Health. Available at: https://www.who.int/health-topics/social-determinants-of-health#tab=tab_1. Accessed July 7, 2022. 11. Graham H.. **Social determinants and their unequal distribution: clarifying policy understandings**. *Milbank Q* (2004.0) **82** 101-24. DOI: 10.1111/j.0887-378x.2004.00303.x 12. 12RStudio [Computer Software]. Boston, MA: RStudio, PBC; 2020. Available at: http://www.rstudio.com/. Accessed July 7, 2022.. *RStudio* (2020.0) 13. **Pharmacy Quality Alliance**. (2021.0) 14. 14Agency for Healthcare Research and Quality. Health Literacy Universal Precautions Toolkit 2nd Edition. Available at: https://www.ahrq.gov/sites/default/files/publications/files/healthlittoolkit2_3.pdf. Accessed July 7, 2022. 15. Kennedy J, Tuleu I, Mackay K. **Unfilled prescriptions of Medicare beneficiaries: prevalence, reasons, and types of medicines prescribed.**. *J Manag Care Pharm* (2008.0) **14** 553-60. DOI: 10.18553/jmcp.2008.14.6.553 16. Lindquist LA, Go L, Fleisher J, Jain N, Friesema E, Baker DW. **Relationship of health literacy to intentional and unintentional non-adherence of hospital discharge medications**. *J Gen Intern Med* (2012.0) **27** 173-8. DOI: 10.1007/s11606-011-1886-3 17. 17Cubanski J, Koma W, Damico A, Neuman T. How much do Medicare beneficiaries spend out of pocket on health care? Kaiser Family Foundation. Available at: https://www.kff.org/medicare/issue-brief/how-much-do-medicare-beneficiaries-spend-out-of-pocket-on-health-care/. Accessed July 7, 2022. 18. Masnoon N, Shakib S, Kalisch-Ellett L, Caughey GE. **What is polypharmacy? A systematic review of definitions**. *BMC Geriatr.* (2017.0) **17** 230. DOI: 10.1186/s12877-017-0621-2 19. Gnjidic D, Hilmer SN, Blyth FM, Naganathan V, Waite L, Seibel MJ. **Polypharmacy cutoff and outcomes: five or more medicines were used to identify community-dwelling older men at risk of different adverse outcomes.**. *J Clin Epidemiol* (2012.0) **65** 989-95. DOI: 10.1016/j.jclinepi.2012.02.018 20. Rankin A, Cadogan CA, Patterson SM, Kerse N, Cardwell CR, Bradley MC. **Interventions to improve the appropriate use of polypharmacy for older people.**. *Cochrane Database Syst Rev* (2018.0) **9** CD008165. DOI: 10.1002/14651858.CD008165.pub4 21. Patel MR, Piette JD, Resnicow K, Kowalski-Dobson T, Heisler M. **Social determinants of health, cost-related nonadherence, and cost-reducing behaviors among adults with diabetes: findings from the National Health Interview Survey**. *Med Care* (2016.0) **54** 796-803. DOI: 10.1097/MLR.0000000000000565
--- title: How does Indian news media report smokeless tobacco control? A content analysis of the gutka ban enforcement authors: - Vivek Dsouza - Pratiksha Mohan Kembhavi - Praveen Rao S. - Kumaran P. - Pragati B. Hebbar journal: PLOS Global Public Health year: 2023 pmcid: PMC10042338 doi: 10.1371/journal.pgph.0001724 license: CC BY 4.0 --- # How does Indian news media report smokeless tobacco control? A content analysis of the gutka ban enforcement ## Abstract Smokeless tobacco (SLT) products like gutka and paan masala are a growing public health crisis in India. Despite enacting a ban—the highest form of regulation—little is known about implementation progress. The purpose of this study was to look at how enforcement of gutka ban is covered in Indian news media and if media is a reliable source of data. We conducted a content analysis of online news reports ($$n = 192$$) from 2011 to 2019. News characteristics such as name and type of publication, language, location, slant and beat coverage, visuals, and administrative focus were quantified. Similarly, news contents were inductively coded to examine dominant themes and the implementation landscape. We found that coverage was initially low but increased after 2016. Overall, news reports were in favor of the ban. Five leading English newspapers covered the majority of the ban enforcement reports. Prominent themes like consumption, health hazards, tobacco control responses, impact on livelihoods, and illicit trade were drawn from the textual analysis as the main arguments in relation to the ban. Gutka is largely seen as an issue of crime reflected by the contents, sources, and frequent use of pictures depicting law enforcement. The interconnected distribution channels of the gutka industry hindered enforcement, highlighting the need to study the complexities of regional and local SLT supply chains. ## Introduction India is one of the largest producer and consumer of smokeless tobacco (SLT) products in the world, with twice the prevalence as compared to smoking cigarettes or bidis [1]. SLT—consumed in various forms—causes oral and esophageal cancers and increases the risk of hypertension and cardiovascular diseases [2, 3]. An estimated 200,000 people die each year due to SLT which costs the Indian economy INR 464.2 billion [4, 5]. In addition, long-term SLT use is driven by a myriad of socio-economic, cultural, and behavioral factors, and is more widespread in rural areas with limited healthcare options [6, 7]. Considering the detrimental health impact of SLTs, India enacted a number of international and domestic tobacco control laws to discourage consumption. In 2004, it ratified the WHO Framework Convention on Tobacco Control (WHO-FCTC), and since then, several best practices have been implemented [8]. More recently, state governments in India banned the sale of gutka and pan masala—popular SLT variants—using food laws as a result of multi-sectoral coordination between the Supreme court, policymakers, and NGOs [9, 10]. Despite concerted efforts, it is unclear how much of illicit gutka is widely available in the market and what is required to strengthen SLT control in India [11]. In such a case, the media can provide additional insight regarding the enforcement of the ban. The media is a valuable resource that helps shape people’s knowledge, beliefs, and attitudes by drawing attention to contemporary issues and encouraging critical debates around policy priorities and outcomes. As a health promotion strategy, public health advocates frequently use the media to communicate the risks of tobacco. Studies have shown that media campaigns focusing on pictorial health warnings effectively reduce the likelihood of tobacco use initiation and motivate people to quit [12–15]. Corporations, on the other hand, persuade media firms to omit facts and report uncritically. There is strong evidence of tobacco industries employing a number of tactics to undermine tobacco control legislations. These include corporate social responsibility initiatives (to cultivate a favorable image), aggressive marketing and promotion of new tobacco products (to retain customers and profits), and disseminating industry-sponsored research (to mislead the public), among others [16, 17]. According to a content analysis of US newspapers from 2006 to 2010, the media regularly downplayed SLT-related health hazards by referring to it as a ‘less harmful’ product [18]. This is alarming given the $576.1 million investment to increase the acceptability of SLT in regions where smoking is declining [19]. More importantly, the way non-communicable diseases are framed indicates that large-scale private sector companies are able to persuade the media to overlook the commercial determinants of health and shift the burden of responsibility to the user [20–23]. To safeguard business interests, tobacco industries employ similar tactics to influence how SLT is portrayed in the media by determining what is considered newsworthy and which sources are frequently cited [24]. In light of growing industry-media relations and increasing use of digital media services, how has Indian news media reported the enforcement of the gutka ban? As part of our ongoing study on the implementation of tobacco control policies in India [25], we aim to analyze patterns in the coverage of the ban and examine whether the media is a reliable source of enforcement-related data. The exploratory nature of the inquiry will allow us to better understand the gutka ban enforcement landscape, actors involved, and the facilitators and barriers that exist. ## Materials and methods We conducted a media content analysis using Braun and Clarke’s six-step method: 1) Data familiarization, 2) Generating initial codes, 3) Searching for themes, 4) Reviewing themes, 5) Defining and naming themes, and 6) Producing the final output [26]. ## Search strategy We used Google search engine database to search for news reports on the gutka ban. We used the search term ‘gutka seizure’ as it is one of the most commonly used enforcement strategies in India. A secondary search was conducted for each Indian state and union territory using the syntax [“gutka seizure” AND “Indian state/union territory” AND “state capital” AND “three highly populated cities”]. Since gutka was banned at different time periods in different Indian states, we did not apply restrictions by date. The searches were carried out in Bengaluru, the city capital of Karnataka state in southern India. ## Inclusion and exclusion criteria We included online news reports published till the year 2019. News on enforcement of smoke-free policies were not included. Our search terms were developed with the intention of retrieving English news. However, we came across many Hindi and Marathi news reports during the initial search. We included them based on their high readership [27] and circulation [28], state-specific relevance, and the research team’s time and ability to translate the language. We did not conduct specific or additional searches in Hindi and Marathi. Due to limitations in fully understanding, interpreting, and translating regional languages, we did not include news published in 20 other scheduled languages including Kannada—the state language of Karnataka. News reports with expired web addresses, behind paywalls and published on blogsites were also excluded. In terms of content (subject matter), we included news reports with and without images and excluded audio-visual and verbatim (exact duplicate) text. We included two or more reports on the same enforcement activity only if they offered any additional details about the incident. ## Selection and analysis of news reports PR (having a background in media relations and advocacy) conducted the online searches and compiled the news reports in a Microsoft Word document. Next, VD (trained in interdisciplinary social sciences) and PH (experienced in tobacco control and health policy and systems research) independently assessed the news as per their eligibility; any disagreements were resolved through team discussions. To familiarize with the data, PK (early career public health researcher) and VD read and re-read each eligible news report and took handwritten notes that were later developed into twelve categories. Descriptions for each category were curated and refined with the help of KP (having a background in journalism) and PH using an online resource manual as shown in Table 1 [29]. Descriptive analysis was used to quantify and analyze the news characteristics. **Table 1** | Category | Description | | --- | --- | | Year | Year of publication. | | Name of publication | Name of the publishing house or groups. | | News type | Classification of news media such as newspaper, news agency, news magazine, financial magazine, news portal, news media company. | | News content | News article or News analysis. | | Language | Language in which news is reported. | | By-line | The name of the writer and/or group mentioned in the story. | | Location | State or union territory where the news is reported. | | Slant | | | • Positive | Supportive of the ban | | • Neutral | Neither supportive nor opposing the ban but presents a mixed review | | • Negative | Opposing or against the ban | | Beat | A subject matter that a reporter frequently covers. | | Administrative focus | Level at which enforcement is reported (national, state, district/city). | | Visuals | Pictures used to convey the story | | Enforcement timing | Stage in the gutka supply chain at which enforcement is reported. | For textual analysis, PK, PR, and KP extracted preliminary data in a Microsoft Excel spread sheet. PH and VD reviewed a $10\%$ sample and differences were resolved with the team. Next, the news reports were imported into NVivo software (version 12). Open coding format was applied where PK and VD inductively coded the contents of the news reports while PH reviewed the themes and provided feedback for improvement. Differences were addressed through multiple discussions with PR and KP. ## Results We retrieved 173 news reports during the initial search and 55 reports during the second-level search. After screening and removing duplicates, we evaluated 226 reports by reading their headlines (title), and eight were found to be unrelated, leaving 218 reports. Of these, we assessed the news contents and excluded 26 based on the eligibility criteria, resulting in 192 reports for the review as depicted in Fig 1. **Fig 1:** *Steps involved in identifying, screening, and assessing the eligibility of news reports.* ## Characteristics of the news reports Table 2 describes the news characteristics that met our eligibility criteria. Detailed news characteristics are included in S1 and S2 Tables. Among the news reports studied, news articles ($$n = 152$$) were written using the who, what, when, where, and why format and news analysis pieces ($$n = 40$$) provided additional information regarding the ban enforcement. **Table 2** | Category | Description | Total (N) | % | | --- | --- | --- | --- | | Year of publication | | | | | | 2019 | 79 | 41.1 | | | 2018 | 69 | 35.9 | | | 2017 | 16 | 8.3 | | | 2016 | 4 | 2.1 | | | 2015 | 6 | 3.1 | | | 2014 | 3 | 1.6 | | | 2013 | 7 | 3.6 | | | 2012 | 6 | 3.1 | | | 2011 | 2 | 1.0 | | Name of publication | | | | | | Publications with >10 reports | 116 | 60.4 | | | Times of India | 49 | 25.5 | | | The Hans India | 27 | 14.1 | | | Deccan Chronicle | 14 | 7.3 | | | The Hindu | 13 | 6.8 | | | The New Indian Express | 13 | 6.8 | | | Publications with <10 reports | 76 | 39.6 | | News type | | | | | | Newspaper | 165 | 85.9 | | | News Portal | 16 | 8.3 | | | News Agency | 4 | 2.1 | | | News media Company | 4 | 2.1 | | | News Magazine | 2 | 1.0 | | | Financial Magazine | 1 | 0.5 | | News content | | | | | | News Article | 152 | 79.2 | | | News Analysis | 40 | 20.8 | | Language | | | | | | English | 166 | 86.5 | | | Hindi | 19 | 9.9 | | | Marathi | 7 | 3.6 | | By-line | | | | | | Bureau | 130 | 67.7 | | | Writers’ names | 62 | 32.3 | | Location | | | | | | States with >10 coverage | 156 | 81.3 | | | Tamil Nadu | 52 | 27.1 | | | Telangana | 39 | 20.3 | | | Maharashtra | 35 | 18.2 | | | Andhra Pradesh | 18 | 9.4 | | | Chhattisgarh | 12 | 6.3 | | | States with <10 coverage | 36 | 18.8 | | Slant | | | | | | Positive | 171 | 89.1 | | | Negative | - | - | | | Neutral | 21 | 10.9 | | Beat | | | | | | Crime | 156 | 81.3 | | | City and neighbourhood | 27 | 14.1 | | | City and crime | 6 | 3.1 | | | Health | 2 | 1.0 | | | Others | 1 | 0.5 | | Administrative focus | | | | | | State | 54 | 28.1 | | | District/city | 138 | 71.9 | | Visuals | | | | | | No picture | 55 | 28.6 | | | Representational | 44 | 22.9 | | | Consuming gutka | 17 | 8.9 | | | Handcuffs | 13 | 6.8 | | | Law enforcement | 6 | 3.1 | | | Warning sign | 3 | 1.6 | | | Other tobacco products | 2 | 1.0 | | | Jail | 1 | 0.5 | | | Ganja | 1 | 0.5 | | | City | 1 | 0.5 | | | Enforcement | 67 | 34.9 | | | Enforcement team with seized gutka products | 28 | 14.6 | | | Enforcement team with offenders and seized gutka products | 10 | 5.2 | | | Only seized gutka products | 15 | 7.8 | | | Enforcement activity | 8 | 4.2 | | | Enforcement team with the offenders | 2 | 1.0 | | | Offenders and seized gutka products | 1 | 0.5 | | | Offenders | 1 | 0.5 | | | Seized money | 1 | 0.5 | | | Seized ganja products | 1 | 0.5 | | | Others | 26 | 13.5 | | | Gutka products on display | 16 | 8.3 | | | Pan shop | 6 | 3.1 | | | Gutka manufacturing unit | 2 | 1.0 | | | Image of the victim | 1 | 0.5 | | | Google map | 1 | 0.5 | | Enforcement timing | | | | | | Transporting | 78 | 40.6 | | | Storing | 51 | 26.6 | | | Selling | 31 | 16.1 | | | Manufacturing | 11 | 5.7 | | | Unassigned | 10 | 5.2 | | | Not applicable | 11 | 5.7 | ## Consumption behavior Gutka use experimentation and initiation were highest among young people. Despite the ban, news reports expressed concerns regarding over consumption, dual use, and switching to twin packets (pan masala and flavored tobacco). Increase in prices or scarcity of the products had little or no effect on users as they were willing to pay higher prices for it. ## Health and environmental hazards Addiction among adolescents was the most frequently reported health concern. This was because of the product’s accessibility in the market. Due to the reliance on gutka, companies began selling adulterated products, often laced with toxic substances, endangering both food safety and health. Oral cancers and other tobacco-related diseases were also reported. One report identified plastic waste caused by discarded gutka packets as having a negative impact on the environment. ## Response of tobacco control and activism Four reports mentioned different tobacco control tactics. To raise awareness about gutka-related health hazards, the local police issued public service announcements and held awareness drives for students, as well as for shops, during raids. Civil society acted as watchdogs, inquiring about tobacco product sales and the enforcement status under the Right to Information (RTI) act. They contributed by either supporting or intervening when enforcement was lax. In one case, local activists called for the boycott of gutka by burning the seized products. ## Impact on livelihoods Poverty was a deciding factor in entering the industry as shopkeepers would earn a living by selling gutka, pan masala, and other tobacco products. Despite public support, news reports expressed caution that the ban would negatively impact livelihoods and lead to unintended consequences. Without any alternatives, shopkeepers were forced to engage in illegal activities to compensate for their losses. Revenue losses at the national level were also attributed to the ban. ## Illicit gutka trade News reports expressed concerns regarding the unabated sale of gutka in several retail outlets. One of the reasons was tax evasion through smuggling, in which illegal traders smuggled gutka and other tobacco products from neighboring states and sold in the black market. These products did not have health warnings or manufacturer’s details. Because the smuggled gutka is sold at higher premiums, the money that went into the black market was used for a variety of criminal activities. Five news reports citied the involvement of top government officials and industry executives in a multi-crore gutka scam. ## Stakeholders Individual stakeholders and those representing institutions at the national, sub-national, and local level were reported as being involved in formal policymaking and implementation of the ban as shown in Table 3. **Table 3** | Stakeholder | Role(s) | | --- | --- | | Judiciary (n = 8) | | | • Supreme court | Policymaking, resolving disputes, and enforcing norms | | • High courts | Policymaking, resolving disputes, and enforcing norms | | Government (n=11) | | | • Union and state governments | Policymaking, regulation and enforcement in public health | | • District administration | Policymaking, regulation and enforcement in public health | | • Municipal corporations, nagarpalika | Policymaking, regulation and enforcement in public health | | Central intelligence and investigation institutions (n = 11) | | | • Central Bureau of Investigation (CBI) | Investigating corruption | | • Directorate General of GST Intelligence (DGGI) | Fighting tax evasion | | • Narcotics Control Bureau | Combating drug trafficking | | State departments (n = 10) | | | • Health | Prevention, control, and management of diseases | | • State tobacco control cell | Implementing provisions of COTPA* | | • Commercial tax | Collecting taxes | | • Food civil supplies and consumer protection | Enforcing orders passed under Essential Commodities Act, 1955 | | Regulatory authorities (n = 62) | | | • Food Safety and Standards Authority of India (FSSAI) | Regulating food safety | | • Food and Drug Administration (FDA) | Enforcing legislation, testing samples | | Police (n = 152) | | | • Criminal Investigation Department | Investigating crimes | | • Local police, special operations team, task forces, squads, railway protection force, local crime branch, detectives, rapid response (RR) teams, mukhbir or ‘informant’ | Maintaining law and order, public safety, enforcing laws, detecting and preventing criminal activities | | Academia and NGOs (n = 4) | Tobacco control research and advocacy | ## Quantity and value of seized gutka Between 2011 and 2019, local police in coordination with agencies like the FSSAI, FDA, and DGGI enforced the ban at different locations and at multiple levels of the supply chain including facilities where gutka is produced and consumed. These were plotted on QGIS software, a geographic information system application, using location-based data reported in the news (Figs 2 and 3). The base map (shapefile) of the Indian state boundaries was downloaded from ‘Community Created Maps of India,’ a project run by the {DataMeet} community. During seizure, gutka worth lakhs of rupees in sachets as well as bags containing crores-worth of gutka were recovered. Maharashtra collected the most fines totaling to ~ INR 561.28 crore followed by Telangana (~ INR 11 crore), and Tamil Nadu, Andhra Pradesh and Chhattisgarh (~ INR 4 crore each). Karnataka topped the list with a single-day seizure valued at ~ INR 400 crore. The quantity and value of gutka and other items recovered from different states is illustrated in S3 Table. **Fig 2:** *Gutka seizures at manufacture units, storage facilities, and shops as reported in the media.State Boundaries Maps are provided by Data{Meet} Community Maps Project. It is made available under the Creative Commons Attribution 2.5 India.* **Fig 3:** *Illicit transport of gutka in Karnataka and Gujarat as reported in the media.State Boundaries Maps are provided by Data{Meet} Community Maps Project. It is made available under the Creative Commons Attribution 2.5 India.* ## Motivations and barriers Important motivational factors to the ban enforcement included champions, tip-offs, high-level support (orders), media scrutiny, and cash incentives (rewards). Barriers included ambiguity in the laws, contested viewpoints on gutka as a product, poor coordination, litigation tactics, and lack of resources which hindered enforcement. Table 4 represents a summary of motivations and barriers accompanied by the relevant quotations reported in the media. **Table 4** | Motivations | Quotes | | --- | --- | | Champion’s role in enforcement (n = 88) | “Two persons have been detained so far. The sample of the seized products has been sent for testing. Further investigation is on.” Said the Durg city superintendent of Police, who is leading the campaign against the intoxicating drugs and the prohibited tobacco-based products.” — The New Indian Express, 2019 (032) | | Tip-offs (n = 51) • Credible sources • Informants hired by the police | “K.A. Special team of Superintendent of Police got information that banned gutka was going from Khamgaon to Nagpur in truck number 16C-1780.” — Lokmat, 2019 (096) | | Orders/directives issued by higher authorities (n = 14) • Union/state governments • Food safety department • Revenue department | “On Wednesday, the state government issued directives to all government agencies—police, district collectors, and civic bodies—to implement the order. The directives were issued following a Supreme Court decision to allow the state to take action against offenders under section 328 of the Indian Penal Code (IPC).” — Hindustan Times, 2018 (089) | | Negative image of the police in the media (n = 5) | “After the two seizures came under intense media scrutiny, Pon Manickavel, inspector-general of police Railways, directed the state railway police to take charge of the confiscated goods and to register an FIR.” — The Times of India, 2018 (143) | | Rewards (n = 2) | “Superintendent of Police appreciated Gudihathnoor police and assured of presenting rewards to them.” — The Hans India, 2019 (209) | | Barriers | Quotes | | Legislative and organizational issues • Lack of clarity in the act • Loopholes in the rules • Poor coordination | “The officials also suggest that the central government remove chewable tobacco from the purview of COTPA and bring it under FSSAI as per court order. The grey areas in the acts give freehand to the manufacturers of the stuff like chewable tobacco.” — The Times of India, 2019 (125) | | Lack of resources • Human and financial resources • Lack of infrastructure | “The FDA office near ESIS Hospital premises in Wagle Industrial Estate has already become a virtual godown to all seized goods with no room to move around.” — Moneylife India, 2012 (100) | | Litigation tactics (n = 1) | “Challenging the FDA resolution that imposed a complete prohibition on the transportation of finished products—pan masala and chewing tobacco—through the state even if manufactured and meant to be sold outside, the petitioner stated that they being transporters, only carried the goods which were not banned in many other states.” — The Times of India, 2019 (082) | ## Gutka supply-chain network Four major supply-chain functions were identified based on the sequence of events described in the news. These were procurement of raw tobacco ingredients and finished products, processing (manufacturing and packaging of gutka), logistics (storing, transporting, and supplying), and selling (to vendors, shopkeepers and users). Key industry players (partners or associates, traders, wholesale distributors, retail vendors, merchants, shopkeepers, intermediaries like agents, brokers, and salesmen, supporting industries such as transport companies) were involved in the gutka supply chain—all of whom were connected by direct and indirect networks as shown in Fig 4. **Fig 4:** *Gutka supply-chain linking people, processes, and infrastructure to end users in India.* ## Discussion SLT is a public health threat in low- and middle-income countries (LMICs), with changing consumption patterns and an increase in dual tobacco use [30]. In complex and fragmented systems with organizational, political, and information constraints, immediate action is required to identify and generate context-specific (local) evidence to inform and strengthen SLT control. We attempt to do this through a content analysis of online news sources to determine how gutka ban enforcement is covered in Indian media. Unlike tobacco control laws such as COTPA, which have national and state-level review systems for the prohibition of smoking, there is limited knowledge on the monitoring, reviewing, and supervision of SLT control, partly due to inconsistent enforcement and a lack of publicly available datasets [31]. We used enforcement data generated by the media stories to fill this knowledge gap as the news are accessible and are reported in ‘real-world’ settings. To our knowledge, this is the first media content analysis studying the enforcement component of SLT control in the LMIC context. Earlier content analyses were conducted in high-income countries (HICs) and studied SLT in relation to advertising and promotion among youth and adults in print and social media [32–37]. The main findings of our study showed that gutka ban enforcement is communicated to the public using a variety of arguments and visuals embedded in the actors, processes, and outcomes as shown in Fig 5. **Fig 5:** *How does Indian news media report gutka ban enforcement?* The frequency of news coverage typically increases during major policy developments. In our case, however, there was low enforcement coverage since the enactment of the first ban (March 31, 2012). During the time period studied (2011 to 2019), there was a sharp increase in coverage after 2016. This was most likely due to an increase in the number of states passing the ban, as well as FSSAI’s directive urging all states to strictly comply with the Supreme Court order [38]. In India, public health policies and programmes are designed at the national and sub-national levels while implementation occurs at the district and city level, which partially explains why most of the news ($$n = 138$$) focused on local-level enforcement. However, in terms of circulation, five English newspapers—The Times of India, The Hans India, Deccan Express, The Hindu, and The New Indian Express—having national and regional focus published the reports. This could be indicative of the small number of local news media included in our study. That being said, other factors like the proliferation of regional media outlets, the role of bureaus ($$n = 130$$) and news agencies in gathering and disseminating news remotely, and a lack of local correspondents could equally play a role [39]. Overall, the majority of the reports ($$n = 171$$) were in favor of the ban albeit for different reasons. Most of the news covered issues around health and illicit trade, which were the primary arguments for supporting the ban. A strong anti-tobacco culture (on health grounds) also appeared to have bolstered support. Some reports, although few in number, addressed the ban’s negative impact on the livelihoods of tobacco vendors. Despite health being central to the news, we observed that public health experts were rarely interviewed. On the contrary, most reports considered law enforcement agencies to be reliable and established sources of information, creating an impression that the police—a powerful figure and a symbol of the state—are largely responsible for protecting and promoting public health. Our claim is further supported by the frequent use of images depicting handcuffs, arrests, and prisons, as well as a strong preference for reporting news under the crime beat, which suggests that gutka is seen more as an individual-level criminal offence than a public health concern (which is a result of larger social and structural determinants) [40]. Our findings are similar to an Australian study on reporting practices, which found that when faced with resource and time constraints, writers rely on trusted sources and tend to be generalists, which impacts the quality and newsworthiness of a particular issue [41]. Several health and non-health sectors were involved in gutka ban enforcement. Among the stakeholders identified, police and regulatory authorities were well-represented in the news. We observed that in circumstances where enforcement was difficult, the police leveraged contacts with the CBI, FDA, and education departments to jointly enforce the ban [42]. Likewise, enforcement was dynamic when champions led the inspection drives and there was access to information (tip-offs) and high-level support. Rewards in the form of cash ($$n = 2$$) were also a motivating factor, which relate to two studies arguing that the effect of enforcement is stronger when officials are compensated by the fines collected [43, 44]. Several barriers raised by the news are consistent with global and Indian tobacco control literature, particularly with regard to gutka, where legal loopholes, lack of awareness and enforcement, and litigation by tobacco industries are associated with violation of the ban, sale of twin packets, without mandated health warnings [31, 45, 46]. Only four reports mentioned the role of academia or NGOs, in either individual or institutional capacity, implying they are also involved but receive insufficient coverage. In terms of seizures, we found news media to be a useful resource in documenting the quantity and value of illicit gutka and other tobacco products. However, the data offered limited insight on state-wise implementation due to a range of SLT products in the market, a lack of disaggregated data, different units of measurement applied, and limited reporting capacities. To ensure data representation and validity, formal data collection and reporting systems need to be established and institutionalized so that comparative studies can determine which states are making progress and which are left behind. We observed a growing presence of illegal gutka businesses in western, central, and southern India. Many storage facilities were reported in Gujarat and Tamil Nadu, which have a strong presence of chewing tobacco cultivation [47, 48]. Raw materials and finished products were transported from Gujarat to neighboring states like Maharashtra, Telangana Uttar Pradesh and Madhya Pradesh. However, several gutka manufacturing and storage units were also reported in Andhra Pradesh, dominated by FCV tobacco (used in cigarettes) which we did not expect (Fig 2). As chewing tobacco is not grown in Karnataka, it has likely emerged as a transit state for supplying gutka to neighboring states via border cities like Bidar and Bengaluru (Fig 3). Since tobacco is a profitable venture, our mapping of different supply chain actors demonstrated an integrated network—bound by tacit information sharing practices and access to infrastructure—where people with varied roles enter the chain on shared economic interests. The supply chain is made up of processes that operate independently but cooperatively. Traders were involved in procuring, manufacturing, storing and transporting as well as buying and re-selling gutka to vendors. Traders and vendors reportedly had close links with shopkeepers and sold exclusively to known customers. Shopkeepers frequently interacted with retailers and wholesale traders, while wholesalers supplied gutka to shopkeepers. Additionally, shopkeepers displayed gutka on purpose to lure customers into buying cigarettes and paan (betel leaf). Our observations are in line with results from a study conducted in Nepal, Bangladesh, and Pakistan that emphasizes the complexity and non-linearity of SLT supply chain management systems [49]. It is possible that this non-linearity, in conjunction with a variety of other tobacco products, allowed supply chain actors to adapt to external shocks such as the ban without compromising personal gains. A compliance study conducted in Mumbai found that consumer base and vendor profits were unaffected as alternatives like cigarettes and non-gutka SLT products were freely sold during the gutka ban [50]. Combating the use of SLT is slowly but surely becoming a policy priority in Southeast Asia, with countries such as Nepal, Bhutan, and Bangladesh amending existing tobacco control laws to prohibit SLT in public places, ban manufacturing and supply, and advertising. Simultaneously, multi-pronged strategies such as cessation services, pictorial warnings, taxation, education and awareness, vendor licensing, and litigation measures are being implemented [31, 51, 52]. Our study adds significant value and can assist policymakers, researchers, and practitioners better understand the gutka ban policy measure in India. Further investigation into the illicit supply-chain routes, enforcement timings, and complex social networks can empower tobacco control advocates and implementers in devising evidence-based interventions to ensure that the WHO-FCTC supply reduction measures are effectively implemented. Training and capacity building of science-communicators, journalists, and media firms can enable them to better represent public health interests by documenting and exposing industry interference. Long-term sustained media advocacy can contribute to the denormalization of SLT. High-burden countries can use media content analysis to identify local contexts in which tobacco market operates to fully explain the implementation progress. ## Strengths and limitations Our study has several strengths. We examined news media sourced from Google search engine which has the largest database of reports available in the public domain. The mixed-method features of content analysis was useful in describing coverage patterns as well as interpret news stories [53]. Categories were drawn from the data as we read, re-read and familiarized ourselves with the news. An inductive approach allowed us to examine which arguments were reported, who had more influence, and how various stakeholders responded to the ban enforcement. In addition, a diverse research team of public health professionals, journalists, and social scientists facilitated cross-pollination of ideas, rich discussions, and critique resulting in a multidisciplinary body of work. Our study also encountered several methodological and procedural limitations. Based on the exploratory nature of the study, we used limited search terms. Despite this, the endless results produced by Google were overwhelming. We had insufficient knowledge about how Google search algorithms work to ensure replication of results. To minimize this, we restricted our searches to the first five pages. Manual search activities such as browsing history as well as differences in spelling (gutka vs gutkha), search terms (gutka vs smokeless tobacco) and place-based results mean that searches in different locations would yield different results, making generalizability of findings a challenge. Our analysis focused on English language news and excluded regional languages. As a result, ban enforcement coverage in some high-burden states like Tripura, Manipur, Odisha, Assam, and Arunachal Pradesh were inadequately captured. Furthermore, we did not include audio-visual content in the analysis, which means that their effect on perception and recall were not explored. Future studies need to examine local and regional news and audio-visual content as they are the important primary sources of information for many states in India. Since we worked with a large sample of data, the content analysis exercise was time-consuming and exhaustive, especially during the thematic analysis and interpretation of results. Administering such a method increases the likelihood to make simplistic comparisons as well as errors based on the subjective experiences of researchers [53]. We made an effort to minimize researcher bias by having two co-authors code the data and conducting numerous team discussions to resolve disagreements at each stage of the research process. ## Conclusions Media content analysis of a policy intervention such as the gutka ban has provided rich descriptions of the stakeholders involved, arguments reported in the news, and the channels and processes associated with the banned SLT product. News media is a useful advocacy tool for communicating SLT-related health hazards, documenting potential industry interferences, studying competing interests, and analyzing implementation progress. Policymakers and implementers in charge of tobacco control in India could benefit from investing in news media research and analysis by incorporating the findings into developing targeted strategies. This is possible through collaboration and sustained engagement with media firms. Future research should also prioritize media coverage in vernacular languages and study other tobacco control policy interventions to improve the overall reporting of SLT control in India. ## References 1. 1MoHFW G of I, TISS. Global Adult Tobacco Survey- Second round INDIA 2016–17 Report. 2018. 2. Gupta R, Gupta S, Sharma S, Sinha DN, Mehrotra R. **A systematic review on association between smokeless tobacco & cardiovascular diseases**. *Indian J Med Res* (2018.0) **148** 77. DOI: 10.4103/IJMR.IJMR_2020_17 3. Asthana S, Vohra P, Labani S. **Association of smokeless tobacco with oral cancer: A review of systematic reviews.**. *Tob Prev Cessat* (2019.0) **5**. DOI: 10.18332/TPC/112596 4. Siddiqi K, Husain S, Vidyasagaran A, Readshaw A, Mishu MP, Sheikh A. **Global burden of disease due to smokeless tobacco consumption in adults: An updated analysis of data from 127 countries.**. *BMC Med* (2020.0) **18** 1-22. DOI: 10.1186/S12916-020-01677-9/FIGURES/4 5. John RM, Sinha P, Munish VG, Tullu FT. **Economic Costs of Diseases and Deaths Attributable to Tobacco Use in India, 2017–2018.**. *Nicotine & Tobacco Research.* (2021.0) **23** 294-301. DOI: 10.1093/NTR/NTAA154 6. Solhi M, Fattahi E, Manzari ZS, Gupta PC, Kargar M, Kasmaei P. **The Reasons for Using Smokeless Tobacco: A Review.**. *Iran J Public Health.* (2021.0) **50** 492-501. DOI: 10.18502/IJPH.V50I3.5589 7. Zhao L, Mbulo L, Twentyman E, Palipudi K, King BA. **Disparities in smokeless tobacco use in Bangladesh, India, and Pakistan: Findings from the Global Adult Tobacco Survey, 2014–2017.**. *PLoS One.* (2021.0) **16** e0250144. DOI: 10.1371/JOURNAL.PONE.0250144 8. 8WHO Framework Convention on Tobacco Control DGO. WHO Framework Convention on Tobacco Control. 2003 [cited 18 Nov 2022]. Available: www.who.int/fctc 9. 9WHO Framework Convention on Tobacco Control DGO. Implementation practices in smokeless tobacco control. Geneva; 2022 Oct. 10. Dsouza R, Bhojani U. **Strategic and contested use of food laws to ban smokeless tobacco products in India: a qualitative analysis of litigation.**. *Tob Control.* (2021.0). DOI: 10.1136/tobaccocontrol-2020-056241 11. Arora M, Chugh A, Jain N, Mishu M, Boeckmann M, Dahanayake S. **Global impact of tobacco control policies on smokeless tobacco use: a systematic review protocol**. *BMJ Open* (2020.0) **10** e042860. DOI: 10.1136/BMJOPEN-2020-042860 12. Bala MM, Strzeszynski L, Topor-Madry R. **Mass media interventions for smoking cessation in adults.**. *Cochrane Database of Systematic Reviews.* (2017.0) **2017**. DOI: 10.1002/14651858.CD004704.PUB4/MEDIA/CDSR/CD004704 13. Jepson R, Harris F, Rowa-Dewar N, MacGillivray S, Hastings G, Kearney N. *A Review Of The Effectiveness Of Mass Media Interventions Which Both Encourage Quit Attempts And Reinforce Current And Recent Attempts To Quit Smoking* (2006.0) 14. Munthali GNC, Wu XL, Rizwan M, Daru GR, Shi Y. **Assessment of Tobacco Control Policy Instruments, Status and Effectiveness in Africa: A Systematic Literature Review.**. *Risk Manag Healthc Policy.* (2021.0) **14** 2913-2927. DOI: 10.2147/RMHP.S311551 15. Bafunno D, Catino A, Lamorgese V, del Bene G, Longo V, Montrone M. **Impact of tobacco control interventions on smoking initiation, cessation, and prevalence: a systematic review.**. *J Thorac Dis* (2020.0) **12** 3844-3856. DOI: 10.21037/jtd.2020.02.23 16. Savell E, Gilmore AB, Fooks G. **How Does the Tobacco Industry Attempt to Influence Marketing Regulations? A Systematic Review.**. *PLoS One.* (2014.0) **9** e87389. DOI: 10.1371/JOURNAL.PONE.0087389 17. Reed G, Hendlin Y, Desikan A, MacKinney T, Berman E, Goldman GT. **The disinformation playbook: how industry manipulates the science-policy process—and how to restore scientific integrity.**. *J Public Health Policy* (2021.0) **42** 622-634. DOI: 10.1057/S41271-021-00318-6/TABLES/1 18. Wackowski OA, Lewis MJ, Delnevo CD, Ling PM. **Smokeless Tobacco Risk Comparison and Other Debate Messages in the News.**. *Health Behav Policy Rev.* (2014.0) **1** 183. DOI: 10.14485/HBPR.1.3.2 19. 19Federal Trade Commission. Federal Trade Commission Smokeless Tobacco Report for 2019. Washington DC; 2021. Available: https://www.ftc.gov/smokelesstobaccodata. 20. Capewell S, Lloyd-Williams F. **The role of the food industry in health: lessons from tobacco?**. *Br Med Bull* (2018.0) **125** 131-143. DOI: 10.1093/bmb/ldy002 21. Nixon L, Mejia P, Cheyne A, Wilking C, Dorfman L, Daynard R. **“We’re Part of the Solution”: Evolution of the Food and Beverage Industry’s Framing of Obesity Concerns Between 2000 and 2012.**. *Am J Public Health.* (2015.0) **105** 2228. DOI: 10.2105/AJPH.2015.302819 22. Jeong M, Gilmore JS, Bleakley A, Jordan A. **Local News Media Framing of Obesity in the Context of a Sugar-Sweetened Beverage Reduction Media Campaign**. *J Nutr Educ Behav* (2014.0) **46** 583-588. DOI: 10.1016/j.jneb.2014.04.294 23. Chang A, Schulz PJ, Cheong AW. **Online Newspaper Framing of Non-Communicable Diseases: Comparison of Mainland China, Taiwan, Hong Kong and Macao.**. *International Journal of Environmental Research and Public Health* (2020.0) **17** 5593. DOI: 10.3390/ijerph17155593 24. Weishaar H, Dorfman L, Freudenberg N, Hawkins B, Smith K, Razum O. **Why media representations of corporations matter for public health policy: A scoping review.**. *BMC Public Health* (2016.0) **16** 1-11. DOI: 10.1186/S12889-016-3594-8/TABLES/3 25. Hebbar PB, Dsouza V, Bhojani U, van Schayck OCP, Babu GR, Nagelhout G. **Implementation research for taking tobacco control policies to scale in India: a realist evaluation study protocol**. *BMJ Open* (2021.0) **11** e050859. DOI: 10.1136/bmjopen-2021-050859 26. Braun V, Clarke V. **Using thematic analysis in psychology.**. *Qual Res Psychol* (2006.0) **3** 77-101. DOI: 10.1191/1478088706QP063OA 27. 27MRUC. Indian Readership Survey Q1 2019. 2019 May. 28. 28Audit Bureau of Circulations. Highest Circulated Dailies, Weeklies & Magazines amongst Member Publications (across languages). 29. 29The News Manual—Glossary. [cited 25 Nov 2022]. Available: https://www.thenewsmanual.net/Resources/glossary.html#P 30. Chen DT-H, Girvalaki C, Mechili EA, Millett C, Filippidis FT. **Global Patterns and Prevalence of Dual and Poly-Tobacco Use: A Systematic Review.**. *Nicotine & Tobacco Research.* (2021.0) **23** 1816-1820. DOI: 10.1093/ntr/ntab084 31. Yadav A, Singh PK, Yadav N, Kaushik R, Chandan K, Chandra A. **Smokeless tobacco control in India: policy review and lessons for high-burden countries**. *BMJ Glob Health* (2020.0) **5** e002367. DOI: 10.1136/bmjgh-2020-002367 32. Timberlake DS, Pechmann C, Tran SY, Au V. **A Content Analysis of Camel Snus Advertisements in Print Media.**. *Nicotine & Tobacco Research.* (2011.0) **13** 431-439. DOI: 10.1093/ntr/ntr020 33. Seidenberg AB, Rodgers EJ, Rees VW, Connolly GN. **Youth Access, Creation, and Content of Smokeless Tobacco (“Dip”) Videos in Social Media.**. *JAH* (2012.0) **50** 334-338. DOI: 10.1016/j.jadohealth.2011.09.003 34. Wackowski OA, Jane Lewis M, Delnevo CD, Ling PM. **A Content Analysis of Smokeless Tobacco Coverage in U.S. Newspapers and News Wires.**. *Nicotine & Tobacco Research.* (2013.0) **15** 1289-1296. DOI: 10.1093/ntr/nts332 35. O’Brien EK, Navarro MA, Hoffman L. **Mobile website characteristics of leading tobacco product brands: cigarettes, smokeless tobacco, e-cigarettes, hookah and cigars.**. *Tob Control* (2019.0) **28** 532-539. DOI: 10.1136/TOBACCOCONTROL-2018-054549 36. Moran MB, Heley K, Baldwin K, Xiao C, Lin V, Pierce JP. **Selling tobacco: A comprehensive analysis of the U.S. tobacco advertising landscape.**. *Addictive behaviors.* (2019.0) **96** 100. DOI: 10.1016/j.addbeh.2019.04.024 37. O’brien EK, Hoffman L, Navarro MA, Ganz O. **Social media use by leading US e-cigarette, cigarette, smokeless tobacco, cigar and hookah brands.**. *Tob Control* (2020.0) **29** e87-e97. DOI: 10.1136/tobaccocontrol-2019-055406 38. 38Central Arecanut Marketing Company v. Union of India Transfer Case (Civil) No. 1/2010 (2016). 2016. Available: https://www.tobaccocontrollaws.org/litigation/decisions/central-arecanut-marketing-company-v-union-of-india 39. Chatterjee R.. **India shifts local: People are hungry for local news, but badly paid journalists are tempted by bribes.**. *Index on Censorship* (2019.0) **48** 51-53. DOI: 10.1177/0306422019842099/ASSET/IMAGES/LARGE/10.1177_0306422019842099-FIG1.JPEG 40. Pollak JM, Kubrin CE. **Crime in the News: How Crimes, Offenders and Victims are Portrayed in the Media**. *Journal of Criminal Justice and Popular Culture* (2007.0) **14** 59-83 41. Leask J, Hooker C, King C. **Media coverage of health issues and how to work more effectively with journalists: a qualitative study.**. *BMC Public Health* (2010.0) **10** 535. DOI: 10.1186/1471-2458-10-535 42. Dsouza V, Hebbar P. **Media’s role in advancing multi-sectoral action for smokeless tobacco control: a content analysis of gutka ban-enforcement in India.**. (2022.0). DOI: 10.6084/M9.FIGSHARE.21572223.V1 43. Baumann F, Bienenstock S, Friehe T, Ropaul M. **Fines as enforcers’ rewards or as a transfer to society at large? Evidence on deterrence and enforcement implications.**. *Public Choice.* (2022.0) 1-27. DOI: 10.1007/S11127-022-01000-5/FIGURES/6 44. Hebbar PB, Dsouza V, Bhojani U, Prashanth NS, van Schayck OC, Babu GR. **How do tobacco control policies work in low-income and middle-income countries? A realist synthesis**. *BMJ Glob Health* (2022.0) **7** e008859. DOI: 10.1136/bmjgh-2022-008859 45. Adhikari K, Pednekar MS, Stepanov I, Singh A, Nikam S, Singhavi H. **Observed Circumvention of the Gutka Smokeless Tobacco Ban in Mumbai, India.**. *Tob Regul Sci.* (2020.0) **6** 331. DOI: 10.18001/trs.6.5.3 46. Welding K, Saraf S, Iacobelli M, Smith KC, Puntambekar N, Gupta PC. **Beyond Gutka: Evidence of Illegal Smokeless Tobacco in Rural and Semi-urban Areas in India.**. *Nicotine & Tobacco Research.* (2022.0) **24** 929-932. DOI: 10.1093/ntr/ntab251 47. 47Prasad V. Case study of tobacco cultivation and alternate crops in India. 2007 Feb. 48. 48LIST OF RELEASED / IDENTIFIED VARIETIES OF VARIOUS TOBACCO TYPES FOR CULTIVATION IN INDIA. Central Tobacco Research Institute (CTRI); 49. Siddiqi K, Scammell K, Huque R, Khan A, Baral S, Ali S. **Smokeless Tobacco Supply Chain in South Asia: A Comparative Analysis Using the WHO Framework Convention on Tobacco Control.**. *Nicotine & Tobacco Research.* (2016.0) **18** 424-430. DOI: 10.1093/ntr/ntv067 50. Kumar G, Pednekar MS, Narake S, Dhumal G, Gupta PC. **Feedback from vendors on gutka ban in two States of India**. *Indian J Med Res* (2018.0) **148** 98. DOI: 10.4103/ijmr.IJMR_121_18 51. Huque R, al Azdi Z, Sheikh A, Ahluwalia J, Mishu M, Mehrotra R. **Policy priorities for strengthening smokeless tobacco control in Bangladesh: A mixed-methods analysis.**. *Tob Induc Dis* (2021.0) **19** 1-10. DOI: 10.18332/tid/140826 52. Yadav A, Singh A, Khadka B, Amarasinghe H, Yadav N, Singh R. **Smokeless tobacco control: Litigation & judicial measures from Southeast Asia**. *Indian Journal of Medical Research* (2018.0) **148** 25. DOI: 10.4103/ijmr.IJMR_2063_17 53. Allen M.. *The SAGE Encyclopedia of Communication Research Methods* (2017.0). DOI: 10.4135/9781483381411
--- title: 'Wayfinding artificial intelligence to detect clinically meaningful spots of retinal diseases: Artificial intelligence to help retina specialists in real world practice' authors: - Hideki Shiihara - Shozo Sonoda - Hiroto Terasaki - Kazuki Fujiwara - Ryoh Funatsu - Yousuke Shiba - Yoshiki Kumagai - Naoto Honda - Taiji Sakamoto journal: PLOS ONE year: 2023 pmcid: PMC10042340 doi: 10.1371/journal.pone.0283214 license: CC BY 4.0 --- # Wayfinding artificial intelligence to detect clinically meaningful spots of retinal diseases: Artificial intelligence to help retina specialists in real world practice ## Abstract ### Aim/background To aim of this study is to develop an artificial intelligence (AI) that aids in the thought process by providing retinal clinicians with clinically meaningful or abnormal findings rather than just a final diagnosis, i.e., a “wayfinding AI.” ### Methods Spectral domain optical coherence tomography B-scan images were classified into 189 normal and 111 diseased eyes. These were automatically segmented using a deep-learning based boundary-layer detection model. During segmentation, the AI model calculates the probability of the boundary surface of the layer for each A-scan. If this probability distribution is not biased toward a single point, layer detection is defined as ambiguous. This ambiguity was calculated using entropy, and a value referred to as the ambiguity index was calculated for each OCT image. The ability of the ambiguity index to classify normal and diseased images and the presence or absence of abnormalities in each layer of the retina were evaluated based on the area under the curve (AUC). A heatmap, i.e., an ambiguity-map, of each layer, that changes the color according to the ambiguity index value, was also created. ### Results The ambiguity index of the overall retina of the normal and disease-affected images (mean ± SD) were 1.76 ± 0.10 and 2.06 ± 0.22, respectively, with a significant difference ($p \leq 0.05$). The AUC used to distinguish normal and disease-affected images using the ambiguity index was 0.93, and was 0.588 for the internal limiting membrane boundary, 0.902 for the nerve fiber layer/ganglion cell layer boundary, 0.920 for the inner plexiform layer/inner nuclear layer boundary, 0.882 for the outer plexiform layer/outer nuclear layer boundary, 0.926 for the ellipsoid zone line, and 0.866 for the retinal pigment epithelium/Bruch’s membrane boundary. Three representative cases reveal the usefulness of an ambiguity map. ### Conclusions The present AI algorithm can pinpoint abnormal retinal lesions in OCT images, and its localization is known at a glance when using an ambiguity map. This will help diagnose the processes of clinicians as a wayfinding tool. ## Introduction The advent of deep learning has spurred remarkable developments in artificial intelligence (AI) [1]. AI has also been applied in the classification of various images in the field of ophthalmology, with reports emphasizing the possibility of diagnosis based on images with high accuracy. Some studies have even begun working on clinical AI applications [2–11]. However, a major drawback of AI is its inability to show specific clinical findings based on AI diagnosis from images; this is referred to as black box AI [12]. To date, heatmaps have been added to help understand the region of the fundus image referenced to some extent [13, 14]. However, this only complements AI diagnosis. Recently, Adler-Milstein et al. proposed the concept of Wayfinding AI for next-generation diagnostics [15]. “ Wayfinding” is defined as interpreting the context and providing cues that guide a diagnostician. More specifically, AI immediately presents a final diagnosis. However, for medical professionals and patients, the diagnostic journey begins by removing uncertain information from large numbers of data. Typically, a care plan is developed during this journey. A final diagnosis from the outset can never garner the trust of clinicians, emphasizing the need for wayfinding AI, a next-generation AI system that advises on the process, as opposed to the current AI, which handles the final diagnoses. Optical coherence tomography (OCT) is an instrument for irradiating tomographic images of tissues based on the interferometric property of light and is widely used in numerous medical fields. Particularly in the field of ophthalmology, it is an indispensable tool for the diagnosis of retinal and glaucoma diseases and is widely used globally [16]. In 2019, Seeboeck et al. reported an AI algorithm for detecting abnormalities in OCT-B scan images of the retina [17]. Although this method is excellent for screening a large number of OCT images, it is difficult to use in real-world clinical practice because it is impractical to examine every section of OCT images of the fundus of every patient with AI at busy retinal clinics. Furthermore, the AI’s diagnosis is so called “black and white,” and it is not clear how much trust should be placed in the points that are considered abnormal. This does not necessarily make it suitable for retinal specialists to use AI in daily clinical practice. This study aimed to create a useful AI model for retinal specialists in daily practice. In the diagnosis of ocular OCT, cases that confused humans were found to be equally confusing with AI to the same degree [18]. Most importantly, the greatest benefit of AI is its ability to quantify the level of confidence. By using this function, it is possible to numerically express the strength of the confidence of the diagnosis in each spot. By looking at a heat map of these data on the 2D fundus image, a physician can see the spot of structural abnormality/ambiguity at a glance. Subsequently, a physician can re-examine the indicated area carefully using OCT and other devices. Repeating these processes would be a real thinking process that leads to easy diagnosis correction. This AI-based work is a process called wayfinding, in which AI assists the thinking process to reach the correct diagnosis with a retina specialist. ## Methods This study was approved by the Ethics Committee of Kagoshima University Hospital (Kagoshima, Japan) and was registered with the University Hospital Medical Network (UMIN)Clinical Trials Registry (CTR). The registration title is “UMIN000031747, Research on retinal/choroidal structure analysis by novel image analysis technique and machine learning.” on March 2018. The detailed protocol is available at https://upload.umin.ac.jp/cgi-open-bin/ctr/ctr_view.cgi?recptno=R000036250. Written informed consent was obtained from all participants after an explanation of the procedures to be used and possible complications. All the procedures conformed to the tenets of the Declaration of Helsinki. The subjects underwent OCT at Sonoda Eye Clinic, Kagoshima, Japan, from January 15, 2019, to March 29, 2019, with the data constituting OCT images of 679 eyes. OCT images of the macular region passing through the foveal centralis were analyzed. Images of poor quality due to opacity of the optic media, poor fixation, etc., as well as images that were inverted due to posterior staphyloma were excluded. From 617 eyes that survived the exclusion criteria, 300 were randomly selected for the analysis. ## Imaging protocol Imaging was performed using SD-OCT (RS-3000 Advance2) manufactured by NIDEK (Tokyo, Japan). Optical coherence tomography (OCT) was performed horizontally through the fovea centralis. As the OCT data, a 9-mm scan width, 1024 A-scan images, and 50 additive sheets were used. The image was extracted as a 1024 × 512 pixel Microsoft Windows bitmap image (bmp). ## Segmentation AI All OCT images were segmented using an OCT B-scan automatic image segmentation model utilizing deep learning [19]. For a brief overview, the segmentation model is based on U-Net [20] and consists of an encoder and decoder, a skip connection between the two, and a multiple dilated convolution (MDC) block. The training of the model was based on previous reports [21]. Briefly, layer boundaries were manually annotated from 171 OCT B-scan images and used as training data. Training was conducted using the AMSGrad optimizer (http://www. Satyenkale. Com/papers/amsgrad. Pdf) using cross-entropy loss function. The batch size was set to 5. To increase the number of training data, the annotated b-scan image was divided vertically into 4 segments. For each segmented image, following data augmentation was applied: left-to-right flipping, scaling (0.9 to 1.1x), and ±15° rotation. Implementation was performed using Sony’s NNabla (https://nnabla.org/). The input was an OCT image, and the output was a probability map for each boundary layer (Fig 1). The encoder and decoder performed a 7×1 vertical convolution to extract the features of the horizontal edges (vertical brightness changes). The MDC block expanded the receptive field by combining convolutions with different dilations to capture the positional relationships of a wide range of features. In the output layer of the model, SoftMax was applied in the vertical direction (the direction of the A-scan). This makes it possible to obtain the probability distribution of the position (depth) of each boundary layer from the A-scan. Finally, the position (depth) with the maximum probability distribution in each A-scan was detected as the boundary layer. This model enables the segmentation of the internal limiting membrane (ILM), nerve fiber layer-ganglion cell layer (NFL-GCL), inner plexiform layer, inner nuclear layer (IPL-INL), outer plexiform layer-outer nuclear layer (OPL-ONL), ellipsoid zone (EZ), and retinal pigment epithelium-Bruch’s membrane (RPE-BM) boundaries (S1 Fig). **Fig 1:** *An OCT B-scan automatic image segmentation model was created based on U-Net.The output of this model is the probability distribution of each boundary layer, thus enabling the automatic detection of the boundary for each layer of the retina.* ## Calculating degree of ambiguity The degree of ambiguity of the retinal interface is calculated using the automatic segmentation model described above. In the model, among the 512 points, which is the number of vertical pixels per A-scan, those corresponding to the boundary layer were initially calculated for each A-scan of the OCT image, where the probability distribution of the boundary layer was the output, and the point with the highest probability was detected as the boundary layer. If the probability distribution is not biased toward one point and varies, the detection of the boundary layer is considered uncertain. The variation in this probability distribution was calculated using entropy. Entropy is an index indicating the degree of chaos and irregularity of a state, and is calculated using the following formula: The entropy value (Entropyx,l) at position x of boundary layer l of the A-scan is Entropyx,l=−∑zpx,z,llogpx,z,l Here, px,z,l is the output value of the network at the coordinates (x,z) of boundary layer l. In other words, the larger the entropy, the more scattered the probability distribution, and the more uncertain the layer detection. Entropy was calculated for each boundary layer of the retina in the OCT image, with the average value defined as the ambiguity index (Amb-I). The lower the Amb-I, the smaller the variation in the probability distribution; the higher the Amb-I, the greater the variation in the probability distribution (Fig 2). An example of segmentation using Amb-I is shown in Fig 3. **Fig 2:** *Conceptual diagram of the entropy value of each point in OCT A-scan.The OCT A-scan line in the normal eyes was blue, and in the disease-affected eyes, the scan line was red. In normal eyes, the probability distribution does not vary, so the entropy is small; however, in disease-affected eyes, the variation is large, so the entropy value is large.* **Fig 3:** *Example of segmentation using the ambiguity index.Normal eyes (A) and diabetic retinopathy (B). The boundary line of the layer was drawn using AI in this study. Because each boundary line is represented by a series of dots, there is an entropy value for each dot. The points where the value was high were delineated by a thick dotted line, and the points where the value was small were delineated by a thin solid line. Red; ILM; yellow, IPL/INL; green, OPL/ONL; blue, EZ; purple, RPE/BM.* ## Heatmap/Ambiguity map creation In the proposed method, a heat map for each layer can be created by arranging the entropy calculated for each A-scan in the OCT volume (Fig 4). The created heatmap was smoothened using a Gaussian filter with σ = 1 and normalized such that the minimum value would be 0 and the maximum value would be 1. A jet color map was applied to color the normalized heat map, the result of which is called an ambiguity map. **Fig 4:** *Scheme of ambiguity map productive process.By calculating entropy in the volume scan, it is possible to create a heatmap for each layer of the retina and visualize the part with high entropy.* ## Image labeling Two retinal experts labeled the presence or absence of abnormalities in each layer of the OCT image. Those with an epiretinal membrane (ERM), retinal edema, hard exudate, retinal pigment epithelium (RPE) abnormality, serous retinal detachment (SRD), pigment epithelial detachment (PED), and drusen were defined as abnormalities. The retina was also examined to determine the layer in which these abnormalities were observed. If two examiners had different opinions, abnormalities were established after discussions between the two examiners. ## Statistical analysis All statistical analyses were performed using SPSS Statistics 19 for Windows (SPSS Inc., IBM, Somers, NY, USA). The difference between the mean values of normal and abnormal Amb-I levels was examined using the Mann-Whitney U test. The ability to classify normal and abnormal parts was evaluated using the AUC of the ROC curve. A p-value of 0.05 or less was considered significant. ## Results After consultation with two experts (HS and SS), if the diagnosis was split, a third expert (HT) was included, and the images were classified into 189 normal and 111 disease-affected eyes. Thirty patients had abnormalities at the boundary between the internal limiting membrane (ILM), 44 at the boundary between the nerve fiber layer (NFL) and ganglion cell layer (GCL), 50 at the boundary between the inner plexiform layer (IPL) and inner nuclear layer (INL), 48 at the boundary between the outer plexiform layer (OPL) and outer nuclear layer (ONL), 78 in the ellipsoid zone (EZ), and 48 at the boundary between the retinal pigment epithelium (RPE) and Bruch’s membrane (BM). ## Relationship between the presence or absence of disease and overall ambiguity The overall Amb-I of the normal image was 1.76 ± 0.10, and the Amb-I of the diseased image was 2.06 ± 0.22, showing a significant difference ($p \leq 0.05$, Mann-Whitney U test) (Table 1). Fig 5 shows the ROC curve for the classification of normal and diseased images using the Amb-I. The AUC for this ROC curve was 0.92. We suggest that it should be possible to classify diseased and normal eyes using Amb-I. **Fig 5:** *The ROC curve for the classification of normal and abnormal images by mean Amb-I of each layer of the retina, classified by AUC = 0.92.* TABLE_PLACEHOLDER:Table 1 ## Relationship between the boundary abnormality of each layer and Amb-I Fig 6 shows the ROC curve of the image classification for each layer of the retina. The Amb-I ROC curve for the presence or absence of ILM boundary abnormalities had AUC = 0.588. The Amb-I ROC curve for the presence or absence of NFL/GCL boundary abnormalities had an AUC of 0.902. The Amb-I ROC curve for the presence or absence of IPL/INL boundary abnormalities had AUC of 0.920. The Amb-I ROC curve for the presence or absence of OPL/ONL boundary abnormalities had AUC of 0.882. The Amb-I ROC curve for the presence or absence of EZ had AUC of 0.926. The Amb-I ROC curve for the presence or absence of RPE/BM boundary abnormalities had AUC = 0.866. Representative cases are shown in S2 Fig. **Fig 6:** *Ambiguity index distribution map.Blue: Normal Amb-I distribution map; Red: Abnormal Amb-I distribution map. It was observed that IPL/INL and EZ can be classified with high accuracy.* ## Ambiguity map A Heatmap/Ambiguity-Map was created, and representative cases were presented to reveal the usefulness of this algorithm. Case 1: A 78-year-old male was treated for neovascular AMD for three years. Routine examination of the ocular fundus with OCT and color photography revealed that the physicians did not notice the ILM layer lesion. The Ambiguity-Map image revealed the presence of a high Amb-I spot on the ILM layer in addition to the pigment epithelial layer. After careful examination of the OCT B-scan image, an epiretinal membrane was found (Fig 7). The epiretinal membrane was not noticed until the Ambiguity-Map. This case shows that the Ambiguity-Map can reveal some important lesions that are often overlooked in cases of other evident diseases. Although this ERM does not require treatment at this time, many ERMs worsen over time and require careful observation in the future. It is thanks to this AI that we were able to identify such findings early. Early detection will ensure a better decision on the appropriate time for treatment, and this thought process can be described as "wayfinding". **Fig 7:** *Ocular findings of case 1.(A) Color fundus photograph. (B) Amb-I heatmap image at RPE level. (C) Amb-I heatmap image at ILM level. (D) OCT B-scan image, scan of the white part of the dashed line in (A). Amb-I: ambiguity index, RPE: retinal pigment epithelium, ILM: internal limiting membrane, ERM: epiretinal membrane.* Case 2: A 54-year-old male with simple diabetic retinopathy. The color fundus photograph, B-scan OCT image, and thickness map supported the diagnosis of simple diabetic retinopathy. The Ambiguity map of the OPL/ONL boundary layer shows some light blue islands. Then, looking at the indicated lesion carefully at high magnification, a distortion in the OPL/ONL could be found. This case shows that subtle but important lesions can be found in the Ambiguity map (Fig 8). **Fig 8:** *Ocular findings of case 2.Color fundus photography (A). OCT image of A-scan with dotted red, yellow, and blue lines (B, C, and D). The color map (E) reflects the thickness of the retina in the same case. These findings are consistent with those of simple diabetic retinopathy. However, when the heatmap (F) based on Amb-I shows an image of the OPL/ONL boundary layer, a light-blue island with a large Amb-I is observed in the layer below the macula. Therefore, if that part is magnified (dotted line qualification part of D and F), the presence of a distortion in the OPL/ONL (G) can be observed. From this, it can be determined that distortion begins at the layer structure of the retina, despite simple retinopathy.* Case 3: The patient was a 31-year-old female with acute zonal occult outer retinopathy (AZOOR). At the initial visit, although fundus photography showed no apparent changes, OCT revealed that the normal outer retinal structure was impaired, which was consistent with AZOOR. The area of the impaired retina can be clearly seen as light blue-colored areas in the Ambiguity map image. After three months, the impaired lesions disappeared on fundus photography and OCT. This was confirmed using an Ambiguity map image. This case showed that the chronological change in the lesion became evident in the Ambiguity-Map image (Fig 9). **Fig 9:** *Ocular findings of case 3.Color fundus photographs at the first visit (A) and after 3 months (B). B-scan OCT image at the first visit (C) and after 3 months (D). EZ level Amb-I heatmap image at the first visit (E) and after three months (F). Amb-I: ambiguity index; EZ: ellipsoid zone.* ## Discussion The existing automatic segmentation model of OCT B-scan images using deep learning enables the quantification of the ambiguity of the boundary surface determination of each layer of the retina. By applying quantified data, it is possible to detect abnormal or ambiguous parts with a specific value, Amb-I. Above all, the Ambiguity-Map enabled the detection of abnormal areas instantly and correctly. There have been reports of the use of deep learning to classify OCT normals and abnormalities with high accuracy [22, 23]. However, deep learning has a problem that the analysis process is a black box and it is difficult to clarify the reason for the classification [12] =. This is a major obstacle in the clinical application of AI. In this study, abnormalities were detected by quantifying the analysis process and it was possible to numerically determine why the abnormalities were classified as they were. The advantage of this method is that it is easy to understand the degree of abnormality, because quantified numerical values can be obtained. In addition, because abnormalities can be detected for each layer, it is easy to determine which layer has an abnormality. The indicated areas could then be re-examined carefully using OCT. We believe that repeating the processes of physicians concerning these parts can facilitate correct diagnoses rather than providing the final diagnosis. This is consistent with the concept of wayfinding. Moreover, as the segmentation model was applied, it was not necessary to incorporate a new program for abnormality detection into the machine. Previously, an OCT segmentation model was applied to an automatic diagnostic model [24]. However, in this case, transfer learning using a deep learning model was performed in the diagnostic process, and because the diagnostic process is a black box, it is very different from this model. Seeboeck et al. used uncertainty to detect anomalies in OCT [17]. In their study, uncertainty was used to infer dropouts in order to realize a neural network, and because no explanation could be given as to why the dropout occurred, it is difficult to conclude that their algorithm was an explainable AI. Another difference is the high cost of computational processing compared with that described in the present study, where anomalies were detected using simple entropy. Yim et al. recently developed an AI algorithm for OCT images of suspected AMD that points to the location of findings such as pigment epithelial detachment, which are important for AMD diagnosis [25]. Although this is a breakthrough, it remains unclear how well it can be adapted t like o diseased eyes other than AMD. By contrast, our AI is applicable to all retinal diseases. From a liability perspective, this AI, which points out lesions, is useful because the final diagnosis is still and will always be made by humans. Again, the algorithm developed by Yim et al. requires a large GPU and thus a large computation time, as found in a study by Seebock et al. Our algorithm is simple and can be implemented using current OCT instruments with a short computation time. The ability to detect abnormalities at the ILM interface was not high, according to the AUC obtained in terms of specificity and sensitivity. The lesions on the ILM interface were mainly ERM, but in the OCT image of ERM, the brightness of only the lesion site increased. This was probably because the entropy value often remained low in the segmentation, such that there were many cases where the boundary surface was not disturbed. Furthermore, the inclusion of cases with a narrow range of ERM is also considered to be one of the factors. In contrast, strong abnormalities were detected between the IPL/INL interface and the EZ. Diabetic macular edema is a typical disease that causes IPL/INL abnormalities, while AMD and central serous chorioretinopathy (CSC) are typical diseases that cause EZ abnormalities. However, it is probable that the entropy value was high and the detection power was high because of the significant changes in the OCT image caused by the lesion. In this study, the entropy of each layer of OCT B-scan images in one cross-section passing through the macula was averaged and calculated as Amb-I. Note that by taking the average, small changes that should be detectable may be missed. However, by applying this method to the OCT volume scan and displaying the entropy of each measurement point on a color scale and applying a mapping, the lesion site can be visually comprehended, as shown in Figs 7–9. Distortions in the retinal layer can also be visualized in cases of early diabetic retinopathy (Fig 8). It can be argued that a change in this magnitude is not a type of wayfinding because it does not affect the treatment. However, our definition of “wayfinding” includes not only findings that immediately affect treatment, but also subtle abnormal findings that may affect treatment in the future. It is therefore important to not miss such subtle findings in diabetic retinopathy. To identify unexpected lesions that are often overlooked in regular medical care, for example, when treating age-related macular degeneration, the doctor may concentrate on the macula while often overlooking mild ERM; however, as shown in Fig 7, ERM can be diagnosed at an early stage. In this method, an area around the abnormal area is attached and scanned to obtain an OCT B-scan image. With this software, the average entropy value of each layer was calculated, and it was possible to numerically indicate the layer with an abnormality. In other words, in the heatmap, the area where there is a high possibility of any abnormality is marked and scanned with OCT, making it possible to indicate the cross section of the image that is abnormal. Most importantly, what makes this method useful is that ambiguous areas can be seen at a glance. Numerical results are useful; however, in busy outpatient settings, abnormal areas may be overlooked. The heatmap method, which shows the location of the lesion with the intensity of ambiguity, has a great advantage in busy outpatient clinics, because this information is instantly available. However, owing to the absence of the inner retinal layer, it should be noted that Amb-I is higher in the central fovea. Care should therefore be taken and processing should be implemented from the beginning. In clinical practice, it is necessary to identify abnormal areas within a short timeframe. Because numerous OCT image slices per case need to be carefully observed within a short timeframe, it is difficult to find new lesions (particularly small or subtle ones) that appear in areas outside the main lesion. This occurs because humans tend to look carefully at slices that contain major lesions but not as carefully at slices that do not. However, such small and often-overlooked lesions may be precursors to major changes in the future. Therefore, by creating a heat map/ambiguity map, we made it possible to easily observe the overall areas. This provides a rough distribution of lesions within a short timeframe and is an extremely useful tool for clinicians. In our experience, other than the main lesion, this has significantly reduced the number of abnormal areas missed. It should be noted that this model is unsuitable for the final diagnosis of a specific disease. Although many AI systems provide a final diagnosis, this is not the intent of our developed AI, which is not considered a major drawback. This AI pinpoints “abnormal” images in an OCT image that are otherwise invisible in a fundus image or through a single-section OCT. The AI used for diagnosis is trained on annotated data for each disease, and thus it can make diagnoses that follow the annotations but cannot make diagnoses for other diseases. However, in clinical practice, multiple lesions, borderline lesions, and small early lesions that cannot be diagnosed are hidden, and many diseases that cannot be annotated are important. We believe that our method can compensate for the shortcomings of existing models. Should a person specifically identify an overlooked or unnoticed abnormality, they would consider it and determine the next necessary test for the examination, which would finally lead to a correct diagnosis. Wayfinding AI is yet to be clearly defined. However, the AI developed in this study is in line with the idea of Wayfinding AI, in that reaching the correct diagnosis through trial and error helps human thinking, and this diagnosis process has been adopted for a long time. The proposed AI is a new AI that would be more acceptable to doctors than normal AI, which remains a black box in the diagnostic process. Overdependence on AI may result in unthinkable misdiagnoses, which can be prevented by the proposed AI. ## References 1. LeCun Y, Bengio Y, Hinton G. **Deep learning.**. *Nature* (2015.0) **521** 436-444. DOI: 10.1038/nature14539 2. Gulshan V, Peng L, Coram M, Stumpe MC, Wu D, Narayanaswamy A. **Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs**. *JAMA* (2016.0) **316** 2402-2410. DOI: 10.1001/jama.2016.17216 3. Abràmoff MD, Lou Y, Erginay A, Clarida W, Amelon R, Folk JC. **Improved automated detection of diabetic retinopathy on a publicly available dataset through integration of deep learning**. *Invest Ophthalmol Vis Sci* (2016.0) **57** 5200-5206. DOI: 10.1167/iovs.16-19964 4. Gargeya R, Leng T. **Automated identification of diabetic retinopathy using deep learning**. *Ophthalmology* (2017.0) **124** 962-969. DOI: 10.1016/j.ophtha.2017.02.008 5. Ting DSW, Cheung CY-L, Lim G, Tan GSW, Quang ND, Gan A. **Development and validation of a deep learning system for diabetic retinopathy and related eye diseases using retinal images from multiethnic populations with diabetes**. *JAMA* (2017.0) **318** 2211-2223. DOI: 10.1001/jama.2017.18152 6. Burlina PM, Joshi N, Pekala M, Pacheco KD, Freund DE, Bressler NM. **Automated grading of age-related macular degeneration from color fundus images using deep convolutional neural networks**. *JAMA Ophthalmol* (2017.0) **135** 1170-1176. DOI: 10.1001/jamaophthalmol.2017.3782 7. Burlina PM, Joshi N, Pacheco KD, Freund DE, Kong J, Bressler NM. **Use of deep learning for detailed severity characterization and estimation of 5-year risk among patients with age-related macular degeneration**. *JAMA Ophthalmol* (2018.0) **136** 1359-1366. DOI: 10.1001/jamaophthalmol.2018.4118 8. Li Z, He Y, Keel S, Meng W, Chang RT, He M. **Efficacy of a deep learning system for detecting glaucomatous optic neuropathy based on color fundus photographs**. *Ophthalmology* (2018.0) **125** 1199-1206. DOI: 10.1016/j.ophtha.2018.01.023 9. Schmidt-Erfurth U, Sadeghipour A, Gerendas BS, Waldstein SM, Bogunović H. **Artificial intelligence in retina**. *Prog Retin Eye Res* (2018.0) **67** 1-29. DOI: 10.1016/j.preteyeres.2018.07.004 10. Grassmann F, Mengelkamp J, Brandl C, Harsch S, Zimmermann ME, Linkohr B. **A deep learning algorithm for prediction of age-related eye disease study severity scale for age-related macular degeneration from color fundus photography**. *Ophthalmology* (2018.0) **125** 1410-1420. DOI: 10.1016/j.ophtha.2018.02.037 11. Brown JM, Campbell JP, Beers A, Chang K, Ostmo S, Chan RVP. **Automated diagnosis of plus disease in retinopathy of prematurity using deep convolutional neural networks**. *JAMA Ophthalmol* (2018.0) **136** 803-810. DOI: 10.1001/jamaophthalmol.2018.1934 12. Castelvecchi D.. **Can we open the black box of AI?**. *Nature* (2016.0) **538** 20-23. DOI: 10.1038/538020a 13. Keel S, Wu J, Lee PY, Scheetz J, He M. **Visualizing deep learning models for the detection of referable diabetic retinopathy and glaucoma**. *JAMA Ophthalmol* (2019.0) **137** 288-292. DOI: 10.1001/jamaophthalmol.2018.6035 14. Ting DSW, Pasquale LR, Peng L, Campbell JP, Lee AY, Raman R. **Artificial intelligence and deep learning in ophthalmology**. *Br J Ophthalmol* (2019.0) **103** 167-175. DOI: 10.1136/bjophthalmol-2018-313173 15. Adler-Milstein J, Chen JH, Dhaliwal G. **Next-Generation Artificial Intelligence for Diagnosis: From Predicting Diagnostic Labels to “Wayfinding”.**. *JAMA* (2021.0) **326** 2467-2468. DOI: 10.1001/jama.2021.22396 16. Fujimoto J, Swanson E. **The Development, Commercialization, and Impact of Optical Coherence Tomography**. *Invest Ophthalmol Vis Sci* (2016.0) **57** OCT1-OCT13. DOI: 10.1167/iovs.16-19963 17. Seeböck P, Orlando JI, Schlegl T, Waldstein SM, Bogunović H, Klimscha S. **Exploiting epistemic uncertainty of anatomy segmentation for anomaly detection in retinal OCT**. *IEEE Trans Med Imaging* (2020.0) **39** 87-98. DOI: 10.1109/TMI.2019.2919951 18. Sonoda S, Shiihara H, Terasaki H, Kakiuchi N, Funatsu R, Tomita M. **Artificial intelligence for classifying uncertain images by humans in determining choroidal vascular running pattern and comparisons with automated classification between artificial intelligence**. *PLOS ONE* (2021.0) **16** e0251553. DOI: 10.1371/journal.pone.0251553 19. Hussain S, Guo F, Li W, Shen Z. **DilUnet: A U-net based architecture for blood vessels segmentation**. *Comput Methods Programs Biomed* (2022.0) **218** 106732. DOI: 10.1016/j.cmpb.2022.106732 20. 20Ronneberger O, Fischer P, Brox T, editors. U-net: Convolutional networks for biomedical image segmentation. International Conference on Medical image computing and computer-assisted intervention. Springer, Cham. 2015;234–231. doi: 10.1007/978-3-319-24574-4_28. DOI: 10.1007/978-3-319-24574-4_28 21. He Y, Carass A, Liu Y, Jedynak BM, Solomon SD, Saidha S. **Fully Convolutional Boundary Regression for Retina OCT Segmentation.**. *Med Image Comput Comput Assist Interv.* (2019.0) **11764** 120-128. DOI: 10.1007/978-3-030-32239-7_14 22. Lee CS, Baughman DM, Lee AY. **Deep learning is effective for classifying normal versus age-related macular degeneration OCT images.**. *Ophthalmol Retina* (2017.0) **1** 322-327. DOI: 10.1016/j.oret.2016.12.009 23. De Fauw J, Ledsam JR, Romera-Paredes B, Nikolov S, Tomasev N, Blackwell S. **Clinically applicable deep learning for diagnosis and referral in retinal disease**. *Nat Med* (2018.0) **24** 1342-1350. DOI: 10.1038/s41591-018-0107-6 24. Rim TH, Lee AY, Ting DS, Teo KYC, Yang HS, Hyeonmin K. **Computer-aided detection and abnormality score for the outer retinal layer in optical coherence tomography**. *Br J Ophthalmol* (2021.0) **106** 1301-1307. DOI: 10.1136/bjophthalmol-2020-317817 25. Yim J, Chopra R, Spitz T, Winkens J, Obika A, Kelly C. **Predicting conversion to wet age-related macular degeneration using deep learning**. *Nat Med* (2020.0) **26** 892-899. DOI: 10.1038/s41591-020-0867-7
--- title: Value of the clinical pharmacist interventions in the application of the American College of Cardiology (ACC/AHA) 2018 guideline for cholesterol management authors: - Mohammad M. AlAhmad - Sham ZainAlAbdin - Khozama AlAhmad - Iqbal AlAhmad - Salah AbuRuz journal: PLOS ONE year: 2023 pmcid: PMC10042341 doi: 10.1371/journal.pone.0283369 license: CC BY 4.0 --- # Value of the clinical pharmacist interventions in the application of the American College of Cardiology (ACC/AHA) 2018 guideline for cholesterol management ## Abstract ### Objectives The study aims to examine the extent to which the updated ACC/AHA management of blood cholesterol guideline [2018] is implemented in practice and to assess the value of the clinical pharmacist interventions in improving physicians’ adherence the guidelines recommendations. ### Methods We utilized in this study an interventional before-after design. The study was conducted on 272 adult patients who visited the study site internal medicine clinics and were candidates for statin therapy based on the 2018 ACC/AHA guidelines for cholesterol management. Adherence to guideline recommendations was measured before and after clinical pharmacists’ interventions by calculating the percentage of patients receiving statin therapy as per guideline recommendation, the type and intensity (moderate or high intensity) of statin therapy used, and the need for additional non-statin therapy. ### Results Adherence with guideline recommendations was significantly improved from $60.3\%$ to $92.6\%$ (X2 = 79.1, $$p \leq 0.0001$$) after clinical pharmacist interventions. Among patients who were on statin therapy, the percentage of those who were on proper statin intensity increased significantly from $47.6\%$ to $94.4\%$ (X2 = 72.5, $$p \leq 0.0001$$). The combination of statins with non-statin therapies such as ezetimibe and PCSK9 inhibitors increased from $8.5\%$ to $30.6\%$ (X2 = 95, $p \leq 0.0001$) and from $0.0\%$ to $1.6\%$ (X2 = 6, $$p \leq 0.014$$), respectively. The use of other lipid-lowering agents was diminished from $14.6\%$ to $3.2\%$ (X2 = 19.2, $p \leq 0.0001$). ### Conclusion Collaboration between physicians and clinical pharmacists is a crucial strategy to improve patients’ treatment and hence, achieve better health outcomes among patients suffering from dyslipidemia. ## Introduction Coronary heart disease (CHD) is considered one of the main causes of morbidity and mortality worldwide. One of the major risk factors that contributes to the development and progression of CHD is dyslipidemia [1]. According to the World Health Organization (WHO), an estimated 17.9 million patients died each year because of cardiovascular diseases, of which more than $80.0\%$ of the deaths were due to stroke and myocardial infarction [2]. Ischemic heart diseases and stroke are among the top 3 leading causes of years of life lost (YLL) and mortality according to a global burden disease study [3]. Another recent study in UAE showed that the overall prevalence of dyslipidemia among adults was $72.5\%$, where the total cholesterol and LDL-C levels were high in $42.8\%$ and $38.6\%$ of the participants, respectively [4]. In addition, a study including an expatriate population in the UAE reported a high prevalence of either overweight or obesity ($75.3\%$) as well as known associated risk factors for developing both metabolic syndrome and dyslipidemia [5]. A large number of clinical trials have reported the benefits of lowering cholesterol levels, particularly LDL-C, in reducing the mortality rate among CHD patients. Based on that, the American College of Cardiology (ACC/AHA) published the 2013 blood cholesterol treatment guidelines to reduce atherosclerotic cardiovascular risk in adults. This guideline has been updated several times since then [6]. The latest update of the ACC/AHA management of blood cholesterol guideline [2018] emphasizes on the importance of categorizing patients into the four statins benefit groups and on the importance of statin therapy using evidence-based intensity level (high- or moderate-intensity statins). Furthermore, it highlights the importance of adding PCSK9 inhibitor therapy after receiving the maximum tolerated statin therapy and ezetimibe to achieve LDL-C < 70 mg/dl or non-HDL-C < 100 mg/dl for some patients [7, 8]. The prevalence and treatment rates of dyslipidemia are high in the UAE and worldwide [4], however, it was found that a significant percentage of patients worldwide were not taking appropriate lipid-lowering agents or were taking statins but were not meeting the primary treatment goal [9, 10]. One of the reasons identified was the low adherence to the guideline recommendations. Clinical pharmacists play important role in individualizing patient treatment and improving adherence to guideline’s recommendations. The present study aims to examine the extent to which the updated ACC/AHA management of blood cholesterol guideline [2018] is implemented in practice and to assess the value of the clinical pharmacists’ interventions in improving physicians’ adherence to the guideline’s recommendations. In UAE, the ACC / AHA guidelines are the most commonly followed and recommended guidelines by the health authorities. ## Subjects and settings The study was conducted on adult patients attending an internal medicine clinic at a large hospital in Al Ain City, UAE, from January to April 2019 ($$n = 647$$). Patients’ information had been collected through Hospital Information System (HIS). The study pharmacist evaluated the data for all patients attending the internal medicine clinic on daily basis during the study period to identify eligible patients. Patients aged ≥ 21 years who met the criteria of one of the statin benefit groups requiring high- or moderate-intensity statin therapy according to the 2018 ACC/AHA guidelines were included in this study if they had no of the below exclusion criteria (number of included patients = 272, $42\%$). As per 2018 ACC / AHA guideline recommendations, the following were the statins benefits groups: The main exclusion criteria were: 1) history of statin-induced rhabdomyolysis or myopathy; 2) history of allergic reaction to statins; 3) current active liver disease; 4) creatine kinase levels >3 times the upper limit of normal; 5) Any contraindications to statins use; 6. Patients for whom a lipid profile was not available or who did not have a sufficient data to classify them into statin benefit groups or enough information for calculating the ASCVD risk score at the time the study was conducted were excluded as well. ## Ethical consideration This study was approved by the Hospital Research Ethics Committee (Ref. CR /$\frac{2018}{40}$). All methods were performed in accordance with relevant guidelines and regulations. The study was explained to patients and their consent was obtained before participation. ## Study design and data collection We utilized in this study an interventional before-after design. Demographic and clinical characteristics of the study sample were extracted from the hospital information system. Data collected did not include any personal or sensitive information such as patient’s identity or medical record number. Fig 1 represents the flowchart of the study design. **Fig 1:** *Study design flowchart.* ## Data collection before clinical pharmacists’ intervention Eligibility for statin therapy was evaluated based on 2018 ACC / AHA guideline recommendations as stated above. Adherence to guideline recommendations before clinical pharmacist’s intervention was measured by calculating the percentage of patients identified per each statin benefit group, the percentage of patients receiving statin therapy as per guideline recommendation, the type and intensity (moderate or high intensity) of statin therapy used, and the need for additional non-statin therapy. We calculated only adherence to class 1 and class 2A guideline’s recommendations. ## Clinical pharmacists’ interventions The clinical pharmacists received the medication order for each patient after the patients’ appointment with the physician. The clinical pharmacists evaluated all patients’ data and the appropriateness of their medication order and recommended therapy modifications to meet the 2018 ACC/AHA cholesterol management guideline recommendations. Physicians were automatically notified with the pharmacist’s intervention in the online system and responded accordingly by agreeing, modifying, or rejecting the pharmacist’s recommendation. Physician response and treatment plan changes due to clinical pharmacist interventions were extracted from the Hospital Information System. ## Data collection after clinical pharmacists’ intervention Adherence to guideline recommendations was measured by calculating the percentage of patients identified per each statin benefit group, the percentage of patients receiving statin therapy as per guideline recommendations, the type and intensity (moderate or high intensity) of statin therapy used, and the need for additional non-statin therapy. We calculated only adherence to class 1 and class 2A recommendations. Rejecting a recommendation that is based on a class 2B guideline recommendation was not considered as nonadherent. These class 2B recommendations are usually considered by the guideline as “may be reasonable” and “weak” where their benefit is ≥ risk. ## Study outcomes The following outcomes were examined in this study: ## Statistical analysis All data were entered and analyzed using Statistical Package for the Social Sciences (SPSS) version 22 (IBM SPSS Statistics for Windows, Version 24.0. Armonk, NY: IBM Corp.). Descriptive statistics were used to measure the frequencies and percentages. The chi-square test was used to compare adherence to guidelines before and after clinical pharmacist interventions. A p-value of 0.05 was considered statistically significant, using a $95.0\%$ confidence interval for differences. ## Demographic and clinical characteristics of the study sample The demographic and clinical characteristics of the study sample are shown in Table 1. The mean age of the studied patients was 52.6 ±10.5, and $71.3\%$ ($$n = 194$$) of them were males. Majority of the patients ($95.6\%$, $$n = 260$$) had previous illness or chronic disease. Out of these patients, 178 patients ($65.4\%$) had hypertension, 150 patients ($55.1\%$) had dyslipidemia, and 196 patients ($72.1\%$) had diabetes mellitus. Nevertheless, 20 patients ($7.4\%$) had LDL-C levels less than 70 mg/dl, 236 patients ($86.8\%$) had LDL-C levels between 70–189 mg/dl, and 16 patients ($5.9\%$) had LDL-C levels equal to or more than 190. **Table 1** | Unnamed: 0 | Age Categories | Age Categories.1 | Age Categories.2 | Unnamed: 4 | | --- | --- | --- | --- | --- | | | 21 - <40 | 40–75 | >75 | Total | | Total participants | 36 (13.2%) | 226 (83.1%) | 10 (3.7%) | 272 (100.0%) | | Population characteristics | | | | | | Gender: | | | | | | Male | 30 (11.0%) | 160 (58.8%) | 4 (1.5%) | 194 (71.3%) | | Female | 6 (2.2%) | 66 (24.3%) | 6 (2.2%) | 78 (28.7%) | | Race: | | | | | | White and others | 28 (10.2%) | 176 (64.7%) | 4 (1.5%) | 208 (76.5%) | | Black | 8 (2.9%) | 50 (18.4%) | 6 (2.2%) | 64 (23.5%) | | Smoking | 22 (8.1%) | 78 (28.7%) | 2 (0.75%) | 102 (37.5%) | | Hypertensive patients | 20 (7.4%) | 150 (56.6%) | 8 (2.9%) | 178 (65.4%) | | Total cholesterol | 213±53.1 | 188±45.1 | 198±20.8 | 192±46.5 | | < 200mg/dl (< 5.2 mmol/L) | 14 (5.1%) | 132 (48.5%) | 2 (0.75%) | 148 (54.4%) | | 200 – 239mg/dl (5.2–6.1 mmol/L) | 10 (3.7%) | 58 (21.3%) | 8 (2.9%) | 76 (27.9%) | | ≥ 240mg/dl (≥ 6.2 mmol/L) | 12 (4.4%) | 36 (13.2%) | 0 (0%) | 48 (17.6%) | | HDL | 37±5.6 | 39±8.6 | 46±6.3 | 39±8.3 | | < 40mg/dl (<1.03 mmol/L) | 24 (8.8%) | 132 (48.5%) | 0 (0%) | 156 (57.4%) | | ≥ 40mg/dl (≥ 1.03 mmol/L) | 12 (4.4%) | 94 (34.6%) | 10 (3.7%) | 116 (42.6%) | | LDL | 128±44.2 | 120±37.3 | 135±5.1 | 121±38.1 | | < 70mg/dl (<1.81 mmol/L) | 0 (0.0%) | 20 (7.4%) | 0 (0.0%) | 20 (7.4%) | | 70 – 189mg/dl (1.81–4.88 mmol/L) | 32 (11.8%) | 194 (71.3%) | 10 (3.7%) | 236 (86.8%) | | ≥ 190mg/dl (≥ 4.9 mmol/L) | 4 (1.5%) | 12 (4.4%) | 0 (0%) | 16 (5.9%) | | Dyslipidemia and on statin | 18 (6.6%) | 126 (46.3%) | 6 (2.2%) | 150 (55.1%) | | Diabetes and on diabetic medications | 26 (9.6%) | 166 (61.1%) | 4 (1.5%) | 196 (72.1%) | | eGFR: | 99.3±22.8 | 87.1±22.3 | 69.6±27.8 | 88.3±22.9 | | BMI: | 28.9±4.9 | 30.2±6.7 | 25.7±8.3 | 30±6.4 | | Mean of estimated 10 years risk of CVD*: | 9.1±4.9 | 15.9±10.3 | 30.3±7.4 | 15.5±10.1 | | < 7.5% | 6 (2.2%) | 44 (16.2%) | 0 (0%) | 50 (18.4%) | | ≥ 7.5-<20% | 22 (8%) | 88 (32.4%) | 4 (1.5%) | 110 (40.5%) | | ≥20% | 0 (0%) | 50 (18.4%) | 10 (3.7%) | 60 (22.1%) | | History of CVD | 8 (2.9%) | 44 (16.2%) | 0 (0.0%) | 52 (19.1%) | | ASCVD not at very high risk | 6 (2.2%) | 16 (5.9%) | 0 (0.0%) | 22 (8.1%) | | Very high risk ASCVD | 2 (0.75%) | 28 (10.3%) | 0 (0.0%) | 30 (11.0%) | Additionally, 52 patients ($19.1\%$) had a history of clinical ASCVD. Of these, 38 patients ($14.0\%$) were at very high risk of recurrent CVD. Out of all participants without a history of clinical ASCVD, 50 patients ($18.4\%$) had an estimated 10-year CVD risk less than $7.5\%$, 110 patients ($40.5\%$) had an estimated 10-year CVD risk greater than or equal to $7.5\%$ but less than $20.0\%$, and 60 patients ($22.1\%$) had an estimated 10-year CVD risk equal to or greater than $20.0\%$. ## The adherence with the 2018 ACC/AHA guideline recommendations for the management of cholesterol in adults before clinical pharmacist interventions Adherence with the 2018 ACC/AHA guideline recommendations for the management of cholesterol in adults is shown in Table 2. Based on the inclusion criteria, all the patients who were enrolled ($100.0\%$, $$n = 272$$) were identified as statin benefit groups according to the 2018 ACC/AHA guideline recommendations. Of these, only $60.3\%$ ($$n = 164$$) were initiated on statin therapy. **Table 2** | Statin benefit groups | Total “n” | Use of statins | Use of statins.1 | Statin intensity level | Statin intensity level.1 | Statin intensity level.2 | Statin intensity level.3 | Addition of non-statin therapy to achieve LDL-C goals | Addition of non-statin therapy to achieve LDL-C goals.1 | Addition of non-statin therapy to achieve LDL-C goals.2 | Addition of non-statin therapy to achieve LDL-C goals.3 | Addition of non-statin therapy to achieve LDL-C goals.4 | Addition of non-statin therapy to achieve LDL-C goals.5 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | | Adherence | Non-adherence | Low | Moderate | High | Adherence | Ezetimibe | Ezetimibe | PCSK9 Inhibitors | PCSK9 Inhibitors | Other lipid lowering agents | Other lipid lowering agents | | | | | | | | | | Required | Initiated | Required | Initiated | Initiated | Drug used | | History of ASCVD (Not at very high risk) | 20 | 14 | 6 | 0 | 11 | 3 | 3 | 7 | 2 | 0 | 0 | 0 | | | History of ASCVD (Very high risk⸙) | 28 | 20 | 8 | 0 | 12 | 8 | 8 | 10 | 2 | 0 | 0 | 2 | Fenofibrate 300mg | | DM / (ASCVD risk score <7.5%) | 44 | 26 | 18 | 4 | 14 | 8 | 22 | 10 | 2 | 0 | 0 | 0 | | | DM / (ASCVD risk score ≥7.5<20%) | 74 | 46 | 28 | 6 | 23 | 17 | 17 | 19 | 4 | 0 | 0 | 16 | Fenofibrate 145mg | | DM / (ASCVD risk score ≥20%) | 48 | 32 | 16 | 6 | 12 | 14 | 14 | 14 | 0 | 0 | 0 | 4 | Fenofibrate 145mg | | ASCVD risk score ≥7.5<20% | 30 | 6 | 24 | 0 | 2 | 4 | 6 | 4 | 0 | 0 | 0 | 0 | | | ASCVD risk score ≥20% | 12 | 8 | 4 | 0 | 4 | 4 | 4 | 4 | 0 | 0 | 0 | 0 | | | LDL-C ≥ 190 mg/dl | 16 | 12 | 4 | 0 | 8 | 4 | 4 | 16 | 4 | 6 | 0 | 2 | Gemifibrozil 600mg | | Total | 272 | 164 (60.3%) | 108 (39.7%) | 16 (9.8%) | 86 (52.4%) | 62 (37.8%) | 78 47.6%) | 84 (51.2%) | 14 (8.5%) | 6 (3.7%) | 0 (0%) | 24 (14.6%) | | | Compliance with guideline | Compliance with guideline | 60.3% | 60.3% | 47.6% | 47.6% | 47.6% | 47.6% | 16.7% | 16.7% | 0% | 0% | | | Out of those who were on statin therapy, $9.8\%$ ($$n = 16$$) were on low intensity statin (e.g., simvastatin 10 mg and pitavastatin 1 mg), $52.4\%$ ($$n = 86$$) were on moderate intensity statin (e.g., simvastatin 40 mg, rosuvastatin 10 mg, atorvastatin 20 mg, pitavastatin 2 mg and 4 mg) and $37.8\%$ ($$n = 62$$) were on high intensity statin (e.g., atorvastatin 40 mg and 80 mg and rosuvastatin 20 mg and 40 mg). Adherence to the recommend level of statins intensity was identified in only $47.6\%$ of patients ($$n = 78$$). The addition of non-statin therapies to achieve LDL-C goals was also assessed, and ezetimibe was required for $51.2\%$ ($$n = 84$$) of those who were on statin therapy. While it was initiated for only $8.5\%$ ($$n = 14$$). PCSK9 inhibitors were required for $3.7\%$ ($$n = 6$$) of those who were on statin and ezetimibe therapies. However, such treatment was not initiated in any patient. Other lipid-lowering agents, such as fibric acid derivatives (fenofibrate 145 mg and 300 mg and gemfibrozil 600 mg), were initiated on $14.6\%$ ($$n = 24$$) of those who were on statin therapy. ## The value of the clinical pharmacist’s interventions on applying the 2018 ACC/AHA guideline recommendations The impact of the clinical pharmacist interventions on applying the 2018 ACC/AHA guideline recommendations is shown in Table 3. In patients with LDL-C<70 mg/dl, 18 recommendations were made, ranging from adding moderate- or high-intensity statins for those who were not initiated on statins (need additional therapy–class I and IIa recommendations as per 2018 ACC/AHA guideline definition of recommendation class), changing to moderate- or high-intensity statin agents for those who were on lower-intensity statin agents and stopping other lipid-lowering agents that may not help in achieving LDL-C goals (dose adjustment/stop unnecessary medications—class I and IIa recommendations). However, the physicians’ acceptance of the aforementioned recommendations was only $22.2\%$. **Table 3** | LDL group | Recommendations | Recommendations.1 | Interventions | Interventions.1 | Physicians response | Physicians response.1 | | --- | --- | --- | --- | --- | --- | --- | | | No. | Type | Type | Rational* | % | Reasons for rejecting | | Patients with LDL-C<70mg/dl | 4 | Adding high intensity statin | Need additional therapy | Class I recommendationClass IIa recommendation | 50% | Low LDL-C level, concerns about side effect and the additional cost | | Patients with LDL-C<70mg/dl | 2 | Adding moderate intensity statin | Need additional therapy | Class I recommendation | 100% | | | Patients with LDL-C<70mg/dl | 2 | Change to high intensity statin | Change drug | Class IIa recommendation | 0% | Low LDL-C level, concerns about side effect and the additional cost | | Patients with LDL-C<70mg/dl | 4 | Change to high intensity statin | Dose adjustment | Class I recommendation | 0% | Low LDL-C level, concerns about side effect and the additional cost | | Patients with LDL-C<70mg/dl | 4 | Change to high intensity statin/stop fenofibrate | Dose adjustment/Unnecessary medication | Class IIa recommendation | 0% | Patients can’t tolerate the higher dose | | Patients with LDL-C<70mg/dl | 2 | Change to moderate intensity statin | Dose adjustment | Class I recommendation | 0% | Low LDL-C level, concerns about side effect and the additional cost | | Total = | 18 | | | | 22.2% | | | Patients with LDL-C between 70-189mg/dl | 50 | Adding ezetimibe | Need additional therapy | Class I recommendationClass IIa recommendation | 90% | Low LDL-C level (72), concerns about side effect and the additional cost | | Patients with LDL-C between 70-189mg/dl | 8 | Adding ezetimibe/stop fenofibrate | Need additional therapy/unnecessary medication | Class IIa recommendation | 100% | | | Patients with LDL-C between 70-189mg/dl | 82 | Adding high intensity statin | Need additional therapy | Class I recommendationClass IIa recommendation | 80.5% | Low LDL-C level (70–87), concerns about side effect and the additional cost | | Patients with LDL-C between 70-189mg/dl | 16 | Adding moderate intensity statin | Need additional therapy | Class I recommendation | 87.5% | Low LDL-C level (79), concerns about side effect and the additional cost | | | 6 | Change to high intensity statin | Change drug | Class I recommendation | 100% | | | | 60 | Change to high intensity statin | Dose adjustment | Class I recommendationClass IIa recommendation | 71.7% | 18.3% Patients can’t tolerate the higher dose.10% Low LDL-C level (70–76), concerns about side effect and the additional cost | | | 4 | Change to high intensity statin/stop fenofibrate | Change drug/ unnecessary medication | Class I recommendation | 0% | Patients can’t tolerate the higher dose | | | 24 | Change to high intensity statin/stop fenofibrate | Dose adjustment/unnecessary medication | Class I recommendationClass IIa recommendation | 83.4% | Patients can’t tolerate the higher dose. | | | 12 | Change to high intensity statin/stop fenofibrate/add ezitimibe | Dose adjustment/unnecessary medication/need additional therapy | Class I recommendation | 50% | Patients can’t tolerate the higher dose. | | Total = | 262 | | | | 79.4% | | | Patient with LDL-C ≥190mg/dl | 6 | Adding ezetimibe/adding PCSK9/stop gemfibrozil | Need additional therapies/unnecessary medication | Class I recommendation/very high LDL-C level | 66% | The decision for adding PCSK9 inhibitors has been delayed for the next follow-up | | Patient with LDL-C ≥190mg/dl | 8 | Adding high intensity statin/ Adding ezetimibe | Need additional therapy | Class I recommendation/very high LDL-C level | 100% | | | Patient with LDL-C ≥190mg/dl | 4 | Adding PCSK9 | Need additional therapy | Class I recommendation/very high LDL-C level | 100% | 50% not tolerating moderate or high intensity statin | | | 12 | Change to high intensity statin/adding ezitimibe | Dose adjustment/need additional therapy | Class I recommendation/very high LDL-C level | 100% | | | Total = | 30 | | | | 93.2% | | In patients with LDL-C between 70–189 mg/dl, 262 recommendations were carried out ranged from adding ezetimibe and stopping other ineffective LDL-C lowering agents for those who were in maximum tolerated dose of statin (need additional therapy/stop the unnecessary medication–class I and IIa recommendations), adding moderate or high intensity statin for those with who were not initiated on statin (need additional therapy–class I and IIa recommendations) and changing to high intensity statin dose or drug for those who were on lower intensity statin agents and stopping other ineffective LDL-C lowering agents (dose adjustment/change drug/stop the unnecessary medications—class I and IIa recommendations). The physicians’ acceptance of these recommendations was $79.4\%$. In patients with LDL-C ≥190 mg/dl, 30 recommendations were submitted, ranging from adding ezetimibe, PCSK9 inhibitor and stopping gemfibrozil for those with very high LDL-C results and requiring a more than $25.0\%$ reduction in LDL-C levels despite the use of high-intensity statins (requiring additional therapy–class I recommendation), adding high-intensity statins together with ezetimibe for those who were not on statins (requiring additional therapies–class I recommendation), adding PCSK9 inhibitors for those with high LDL-C levels, although they were on high-intensity statins together with ezetimibes (requiring additional therapy–class I recommendation) and changing to high-intensity statins and adding ezetimibe for those on moderate-intensity statins even though the LDL-C level was more than or equal to 190 mg/dl (dose adjustment/requiring additional therapy–class I recommendation). Interestingly, the physicians’ acceptance of these recommendations was $93.2\%$. Fig 2 summarizes the number of recommendations and type of interventions performed by the clinical pharmacist to achieve the desired outcomes. **Fig 2:** *The total number of recommendations and type of interventions performed by the clinical pharmacist (n = 310).* ## Adherence with the 2018 ACC/AHA guideline after clinical pharmacist’s interventions Adherence with the 2018 ACC/AHA guideline for the management of cholesterol in adults after clinical pharmacist interventions is shown in Table 4. Accordingly, the number of patients who were initiated on statin therapy increased significantly up to $92.6\%$ ($$n = 252$$) after the clinical pharmacist interventions were implemented (X2 (df = 1, $$n = 272$$) = 79.1, $$p \leq 0.0001$$). **Table 4** | Statin benefit groups | Total “n” | Use of statins | Use of statins.1 | Statin intensity level | Statin intensity level.1 | Statin intensity level.2 | Statin intensity level.3 | Adding of non-statin therapy to achieve LDL-C goals | Adding of non-statin therapy to achieve LDL-C goals.1 | Adding of non-statin therapy to achieve LDL-C goals.2 | Adding of non-statin therapy to achieve LDL-C goals.3 | Adding of non-statin therapy to achieve LDL-C goals.4 | Adding of non-statin therapy to achieve LDL-C goals.5 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | | Adherence | Non-adherence | Low | Moderate | High | Adherence | Ezetimibe | Ezetimibe | PCSK9 Inhibitors | PCSK9 Inhibitors | Other lipid lowering agents | Other lipid lowering agents | | | | | | | | | | Required | Initiated | Required | Initiated | Initiated | Drug used | | History of ASCVD (Not at very high risk) | 20 | 18 | 2 | 0 | 4 | 14 | 18 | 7 | 7 | 0 | 0 | 0 | | | History of ASCVD (Very high risk⸙) | 28 | 24 | 4 | 0 | 6 | 18 | 18 | 10 | 8 | 0 | 0 | 2 | Fenofibrate 300mg | | DM / (ASCVD risk score <7.5%) | 44 | 42 | 2 | 2 | 20 | 20 | 40 | 10 | 10 | 0 | 0 | 0 | | | DM / (ASCVD risk score ≥7.5<20%) | 74 | 66 | 8 | 2 | 7 | 57 | 64 | 19 | 18 | 0 | 0 | 2 | Fenofibrate 145mg | | DM / (ASCVD risk score ≥20%) | 48 | 46 | 2 | 2 | 4 | 40 | 42 | 14 | 12 | 0 | 0 | 4 | Fenofibrate 145mg | | ASCVD risk score ≥7.5<20% | 30 | 28 | 2 | 0 | 2 | 26 | 28 | 4 | 2 | 0 | 0 | 0 | | | ASCVD risk score ≥20% | 12 | 12 | 0 | 0 | 0 | 12 | 12 | 4 | 4 | 0 | 0 | 0 | | | LDL-C ≥ 190 mg/dl | 16 | 16 | 0 | 0 | 2 | 14 | 16 | 16 | 16 | 6 | 4 | 0 | | | Total | 272 | 252 (92.6%) | 20 (7.4%) | 6 (2.4%) | 45 (17.9%) | 201 (79.8%) | 238 (94.4%) | 84 (33.3%) | 77 (30.6%) | 6 (2.4%) | 4 (1.6%) | 8 (3.2%) | | | Adherence with guideline | Adherence with guideline | 92.6% | 92.6% | 94.4% | 94.4% | 94.4% | 94.4% | 91.7% | 91.7% | 66.7% | 66.7% | | | Consequently, the number of patients who were on low- or moderate-intensity statins decreased to $2.4\%$ ($$n = 6$$) and $17.9\%$ ($$n = 45$$), respectively. However, the number of patients who were on high-intensity statins potentially increased to $79.8\%$ ($$n = 201$$). Based on that, adherence with the recommendations regarding the level of statin intensity used was significantly improved to $94.4\%$ ($$n = 238$$) after the clinical pharmacist interventions (X2(df = 1, $$n = 252$$) = 72.5, $$p \leq 0.0001$$). The use of ezetimibe as an add-on nonstatin therapy was encouraged and effectively added to the treatment plan to achieve LDL-C goals. The number of patients who were initiated ezetimibe increased significantly to $91.7\%$ ($$n = 77$$) after the clinical pharmacist interventions (X2 (df = 1, $$n = 84$$) = 95, $p \leq 0.0001$). Interestingly, for those who were on statin and ezetimibe therapies and required PCSK9 inhibitors to achieve LDL-C goals, adherence with the recommendations was effectively improved to $66.7\%$ ($$n = 4$$); (X2 (df = 1, $$n = 6$$) = 6, $$p \leq 0.014$$). The use of other lipid-lowering agents, such as fibrates, was markedly reduced to $3.2\%$ ($$n = 8$$) for those who were on statin therapy after the clinical pharmacist interventions (X2(df = 1, $$n = 208$$) = 19.2, $p \leq 0.0001$). Fig 3 shows comparison of adherence with the 2018 ACC/AHA guideline recommendations for the management of cholesterol before and after clinical pharmacist interventions regarding the initiation of statins, the proper use of moderate- or high-intensity statins, evidence-based addition of ezetimibe and PCSK9 inhibitors and minimization of other lipid-lowering agent abuse. **Fig 3:** *Comparison of adherence with the 2018 ACC/AHA guideline recommendations before and after clinical pharmacists’ interventions (n = 272).* ## Discussion Based on this study, adherence with the 2018 ACC/AHA guideline recommendation for the management of cholesterol in adult patients before clinical pharmacist interventions was $60.3\%$ for the initiation of statins therapy and $47.6\%$ for adherence to proper intensity statin therapy. Accordingly, the initiation of statins, particularly high-intensity statins, is prescribed to far fewer patients than recommended. Consequently, the use of non-statin therapies such as ezetimibe and PCSK9 inhibitors was nearly diminished, taking into consideration that several studies highlighted the importance of pharmacist intervention on cholesterol risk management and revealed the treatment gap between research evidence and clinical practice [12–15]. According to our findings, the clinical pharmacist plays a crucial role in the management of cholesterol levels by recommending new therapies, adjusting or increasing drug doses and stopping or changing medications. Furthermore, systematic reviews and meta-analyses of randomized trials conducted by Machado et al. and Santschi et al. emphasized the importance of pharmaceutical care interventions in the management of CVDs [16, 17]. Pharmacist interventions achieved greater reductions in systolic and diastolic blood pressure (BP), total cholesterol (TC), and LDL-C and in the risk of smoking compared with the usual care group [16, 18, 19]. Nevertheless, various clinical trials have illustrated great benefits of statin use, such as pleiotropic effects, which could be beneficial for the treatment and management of several comorbidities [20–22]. In this study, adherence with the 2018 ACC/AHA guideline to achieve the required LDL-C goals was significantly improved after clinical pharmacist interventions and the implementation of the appropriate recommendations. Consistently, Bozovich et al. 2000 and Tahaineh et al. 2011 showed significant improvement in achieving LDL-C goals when clinical pharmacists managed lipid clinics or through clinical pharmacy services under the supervision of cardiologists [23, 24]. The same was achieved by Tsuyuki RT et al. [ 2016] [25]. In the current study, physicians’ acceptance of the clinical pharmacist’s recommendation according to the guidelines was variable based on patients’ LDL-C levels. For instance, physicians” acceptance of clinical pharmacist interventions was high among patients with LDL-C ≥70 mg/dl. Subsequently, this resulted in greater improvement of LDL-C levels and improvement in health outcomes. Likewise, recent studies reported that primary healthcare physicians significantly relied on clinical pharmacists in assessing and improving patients’ adherence to their medications as well as in educating and counseling the patients to avoid clinical malpractice and achieve better health outcomes [26, 27]. Several studies presented the major explanations for statin refractoriness reported by healthcare practitioners, and patients were concerned about adverse events [12, 28–33]. Rosenson, R. S 2016 stated that evaluation of potential adverse events requires validated tools to differentiate between statin-associated adverse events versus nonspecific complaints. Additionally, treatment options for statin-intolerant patients include the use of different statins, often at a lower dose or frequency. To lower LDL cholesterol, lower doses of statins may be combined with ezetimibe or bile acid sequestrants [34]. Newer treatment options for patients with statin-associated muscle symptoms may include proprotein convertase subtilisin kexin 9 (PCSK9) inhibitors [12]. There are many reasons that contribute to non-adherence with the guidelines or rejecting the pharmacists’ recommendations such as the availability of specific drugs and patient’s reluctant for treatment initiation or dose escalation. The results of the current study indicated that many physicians are reluctant to prescribe high intensity statins due the worry about side effects and myopathy. In some cases, physicians stop or change the dose of statins when patients’ report intolerance of such medications or due to the high cost. Another reason for rejecting the pharmacists’ recommendations in this study was low bassline LDL-C level for some patients despite the patient being categorized as a statin benefit group. In clinical trials, statin-associated adverse events showed no differences between participants assigned to statins or placebo [12, 28]. However, it is important to know that these trials select patients with better tolerability and lower risk for myopathy based on their ages, absence of musculoskeletal complaints, normal renal function and less concomitant medications that may alter the pharmacokinetic pathways [12, 29]. One of the solutions to overcome the problem is to switch to the fully human monoclonal antibodies proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibitors (alirocumab and evolocumab) that caused fewer muscle symptoms based on clinical trials and were no more often than when ezetimibe was used [30–32]. However, the cost of such treatment is still one of the main barriers. ## Limitations This study is an interventional before after design. This design is usually used in circumstances where it is not possible to use a control group for ethical or practical issues. Although this design is suitable for the current study, the lack of control group makes this design prone to bias and many confounders. Therefore, the outcome can instead be related to any changes that occurred around the same time as the intervention. Although the clinical pharmacists and physicians who treated the patients in the studied clinics remained the same during the study period, other unknown confounders could have occurred. Another limitation is that it is not an easy task to initiate statin therapy for those are low or intermediate risk for ASCVD. It involves looking for wide range of risk-enhancing factors which could favor the initiation of statin therapy for the low or intermediate risk group, for example the presence of premature CVD in the family. These risk-enhancing factors can be missed during patient assessment or during data extraction. on the other hand, the patient with low or intermediate ASCVD risk is allowed to option for not taking statins for primary ASCVD prevention. Therefore, it is difficult to precisely judge physicians for adherence to guideline for initiating lipid-lowering agents. In addition, the estimate ASCVD risk is more imprecise in some patients when cholesterol levels were used after treatment. ## Conclusions The clinical pharmacist has a key role in improving the management of blood cholesterol by recommending therapies, adjusting doses and stopping or changing medications. Furthermore, adherence with the latest updated guideline recommendations to achieve the desired treatment goals was notably enhanced after the clinical pharmacist interventions and the implementation of the appropriate recommendations. This study illustrates how collaboration between physicians and clinical pharmacists can be crucial strategy to improve patients’ treatment and hence, achieve better health outcomes among patients suffering from dyslipidemia. ## References 1. Tsao CW, Aday AW, Almarzooq ZI, Alonso A, Beaton AZ. **Heart Disease and Stroke Statistics-2022 Update: A Report From the American Heart Association**. *Circulation* (2022.0) **145** e153-e639. DOI: 10.1161/CIR.0000000000001052 2. 2World Health Organization (WHO). Cardiovascular diseases (CVDs). [Internet] Geneva: WHO 2021 [updated 2021 June 11, cited 2022 Sept 13]. Available from: https://www.who.int/news-room/fact-sheets/detail/cardiovascular-diseases-(cvds).. (2021.0) 3. Li Z, Lin L, Wu H, Yan L, Wang H. **Global, Regional, and National Death, and Disability-Adjusted Life-Years (DALYs) for Cardiovascular Disease in 2017 and Trends and Risk Analysis From 1990 to 2017 Using the Global Burden of Disease Study and Implications for Prevention.**. *Front Public Health* **9** 559751. DOI: 10.3389/fpubh.2021.559751 4. Mahmoud I, Sulaiman N. **Dyslipidaemia prevalence and associated risk factors in the United Arab Emirates: a population-based study**. *BMJ Open* (2019.0) **9** e031969. DOI: 10.1136/bmjopen-2019-031969 5. Sulaiman N, Elbadawi S, Hussein A, Abusnana S, Madani A. **Prevalence of overweight and obesity in United Arab Emirates Expatriates: the UAE National Diabetes and Lifestyle Study**. *Diabetol Metab Syndr* (2017.0) **9** 88. DOI: 10.1186/s13098-017-0287-0 6. Stone NJ, Robinson JG, Lichtenstein AH, Bairey Merz CN, Blum CB. **2013 ACC/AHA guideline on the treatment of blood cholesterol to reduce atherosclerotic cardiovascular risk in adults: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines**. *J Am Coll Cardiol* (2014.0) **63** 2889-2934. DOI: 10.1016/j.jacc.2013.11.002 7. Wilson PWF, Polonsky TS, Miedema MD, Khera A, Kosinski AS. **Systematic Review for the 2018 AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA Guideline on the Management of Blood Cholesterol: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines**. *J Am Coll Cardiol* (2019.0) **73** 3210-3227. DOI: 10.1016/j.jacc.2018.11.004 8. Grundy SM, Stone NJ. **2018 American Heart Association/American College of Cardiology/Multisociety Guideline on the Management of Blood Cholesterol-Secondary Prevention**. *JAMA Cardiol* (2019.0) **4** 589-591. DOI: 10.1001/jamacardio.2019.0911 9. Underberg J, Toth PP, Rodriguez F. **LDL-C target attainment in secondary prevention of ASCVD in the United States: barriers, consequences of nonachievement, and strategies to reach goals**. *Postgraduate Medicine.* (2022.0). DOI: 10.1080/00325481.2022.2117498 10. Gitt AK, Parhofer KG, Laufs U, März W, Paar WD. **Hypercholesterolemia diagnosis, treatment patterns and target achievement in patients with acute coronary syndromes in Germany**. *Clinical Research in Cardiology* (2022.0) 1-13. DOI: 10.1007/s00392-022-02108-w 11. Goff DC, Lloyd-Jones DM, Bennett G, Coady S, D’Agostino RB. **2013 ACC/AHA guideline on the assessment of cardiovascular risk: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines**. *Circulation* (2014.0) **129** S49-73. DOI: 10.1161/01.cir.0000437741.48606.98 12. Rosenson RS. **Statin non-adherence: clinical consequences and proposed solutions**. *F1000Res 5.* (2016.0). DOI: 10.12688/f1000research.8215.1 13. Nguyen PV, Biron P. **Underutilization of High-Intensity Statin Therapy After Hospitalization for Coronary Heart Disease: A Cause for Concern, But a Few Words of Caution**. *J Am Coll Cardiol* (2015.0) **65** 2768-2769. DOI: 10.1016/j.jacc.2015.02.083 14. Tsuyuki RT, Johnson JA, Teo KK, Simpson SH, Ackman ML. **A randomized trial of the effect of community pharmacist intervention on cholesterol risk management: the Study of Cardiovascular Risk Intervention by Pharmacists (SCRIP).**. *Arch Intern Med* (2002.0) **162** 1149-1155. DOI: 10.1001/archinte.162.10.1149 15. Yun H, Safford MM, Brown TM, Farkouh ME, Kent S. **Statin use following hospitalization among Medicare beneficiaries with a secondary discharge diagnosis of acute myocardial infarction**. *J Am Heart Assoc* (2015.0) **4**. DOI: 10.1161/JAHA.114.001208 16. Santschi V, Chiolero A, Burnand B, Colosimo AL, Paradis G. **Impact of pharmacist care in the management of cardiovascular disease risk factors: a systematic review and meta-analysis of randomized trials**. *Arch Intern Med* (2011.0) **171** 1441-1453. DOI: 10.1001/archinternmed.2011.399 17. Machado M, Nassor N, Bajcar JM, Guzzo GC, Einarson TR. **Sensitivity of patient outcomes to pharmacist interventions. Part III: systematic review and meta-analysis in hyperlipidemia management**. *Ann Pharmacother* (2008.0) **42** 1195-1207. DOI: 10.1345/aph.1K618 18. George J, McNamara K, Stewart K. **The roles of community pharmacists in cardiovascular disease prevention and management**. *Australas Med J* (2011.0) **4** 266-272. DOI: 10.4066/AMJ.2011.698 19. Dunn SP, Birtcher KK, Beavers CJ, Baker WL, Brouse SD. **The role of the clinical pharmacist in the care of patients with cardiovascular disease**. *J Am Coll Cardiol* (2015.0) **66** 2129-2139. DOI: 10.1016/j.jacc.2015.09.025 20. Bedi O, Dhawan V, Sharma PL, Kumar P. **Pleiotropic effects of statins: new therapeutic targets in drug design**. *Naunyn Schmiedebergs Arch Pharmacol* (2016.0) **389** 695-712. DOI: 10.1007/s00210-016-1252-4 21. Yoshida T, Hayashi M. **Pleiotropic effects of statins on acute kidney injury: involvement of Kruppel-like factor 4**. *Clin Exp Nephrol* (2017.0) **21** 175-181. DOI: 10.1007/s10157-016-1286-4 22. Mohammad S, Nguyen H, Nguyen M, Abdel-Rasoul M, Nguyen V. **Pleiotropic Effects of Statins: Untapped Potential for Statin Pharmacotherapy**. *Curr Vasc Pharmacol* (2019.0) **17** 239-261. DOI: 10.2174/1570161116666180723120608 23. Bozovich M, Rubino CM, Edmunds J. **Effect of a clinical pharmacist-managed lipid clinic on achieving National Cholesterol Education Program low-density lipoprotein goals**. *Pharmacotherapy* (2000.0) **20** 1375-1383. DOI: 10.1592/phco.20.17.1375.34895 24. Tahaineh L, Albsoul-Younes A, Al-Ashqar E, Habeb A. **The role of clinical pharmacist on lipid control in dyslipidemic patients in North of Jordan.**. *Int J Clin Pharm* (2011.0) **33** 229-236. DOI: 10.1007/s11096-011-9479-0 25. Tsuyuki RT, Rosenthal M, Pearson GJ. **A randomized trial of a community-based approach to dyslipidemia management: Pharmacist prescribing to achieve cholesterol targets (RxACT Study).**. *Can Pharm J (Ott)* (2016.0) **149** 283-292. DOI: 10.1177/1715163516662291 26. Doellner JF, Dettloff RW, DeVuyst-Miller S, Wenstrom KL. **Prescriber acceptance rate of pharmacists’ recommendations**. *J Am Pharm Assoc (2003)* (2017.0) **57** S197-S202. DOI: 10.1016/j.japh.2017.03.002 27. Meredith AH, Ramsey D, Schmelz A, Berglund R. **Resident physicians’ perceptions of ambulatory care pharmacy**. *Pharm Pract (Granada)* (2019.0) **17** 1509. DOI: 10.18549/PharmPract.2019.3.1509 28. Ganga HV, Slim HB, Thompson PD. **A systematic review of statin-induced muscle problems in clinical trials**. *Am Heart J* (2014.0) **168** 6-15. DOI: 10.1016/j.ahj.2014.03.019 29. Cohen JD, Brinton EA, Ito MK, Jacobson TA. **Understanding Statin Use in America and Gaps in Patient Education (USAGE): an internet-based survey of 10,138 current and former statin users.**. *J Clin Lipidol* (2012.0) **6** 208-215. DOI: 10.1016/j.jacl.2012.03.003 30. Sullivan D, Olsson AG, Scott R, Kim JB, Xue A. **Effect of a monoclonal antibody to PCSK9 on low-density lipoprotein cholesterol levels in statin-intolerant patients: the GAUSS randomized trial**. *Jama* (2012.0) **308** 2497-2506. DOI: 10.1001/jama.2012.25790 31. Stroes E, Colquhoun D, Sullivan D, Civeira F, Rosenson RS. **Anti-PCSK9 antibody effectively lowers cholesterol in patients with statin intolerance: the GAUSS-2 randomized, placebo-controlled phase 3 clinical trial of evolocumab**. *J Am Coll Cardiol* (2014.0) **63** 2541-2548. DOI: 10.1016/j.jacc.2014.03.019 32. Moriarty PM, Thompson PD, Cannon CP, Guyton JR, Bergeron J. **Efficacy and safety of alirocumab vs ezetimibe in statin-intolerant patients, with a statin rechallenge arm: The ODYSSEY ALTERNATIVE randomized trial**. *J Clin Lipidol* (2015.0) **9** 758-769. DOI: 10.1016/j.jacl.2015.08.006 33. Aburuz S, Al-Bekairy A, Alqahtani AA, Harbi K, Al Nuhait M. **Comparison of the application of treatment Panel III and American College of Cardiology/American heart Association guidelines for blood cholesterol treatment in Saudi Arabia**. *J Saudi Heart Assoc* (2018.0) **30** 349-355. DOI: 10.1016/j.jsha.2018.08.003 34. Cannon CP, Blazing MA, Giugliano RP, McCagg A, White JA. **Ezetimibe Added to Statin Therapy after Acute Coronary Syndromes**. *N Engl J Med* (2015.0) **372** 2387-2397. DOI: 10.1056/NEJMoa1410489
--- title: Increased hepatic interleukin-1, arachidonic acid, and reactive oxygen species mediate the protective potential of peptides shared by gut cysteine peptidases against Schistosoma mansoni infection in mice authors: - Hatem Tallima - Rashika El Ridi journal: PLOS Neglected Tropical Diseases year: 2023 pmcid: PMC10042345 doi: 10.1371/journal.pntd.0011164 license: CC BY 4.0 --- # Increased hepatic interleukin-1, arachidonic acid, and reactive oxygen species mediate the protective potential of peptides shared by gut cysteine peptidases against Schistosoma mansoni infection in mice ## Abstract ### Background Multiple antigen peptide (MAP) construct of peptide with high homology to Schistosoma mansoni cathepsin B1, MAP-1, and to cathepsins of the L family, MAP-2, consistently induced significant ($P \leq 0.05$) reduction in challenge S. mansoni worm burden. It was, however, necessary to modify the vaccine formula to counteract the MAP impact on the parasite egg counts and vitality, and discover the mechanisms underlying the vaccine protective potential. ### Methodology Outbred mice were immunized with MAP-2 in combination with alum and/or MAP-1. Challenge infection was performed three weeks (wks) after the second injection. Blood and liver pieces were obtained on an individual mouse basis, 23 days post-infection (PI), a time of S. mansoni development and feeding in the liver before mating. Serum samples were examined for the levels of circulating antibodies and cytokines. Liver homogenates were used for assessment of liver cytokines, uric acid, arachidonic acid (ARA), and reactive oxygen species (ROS) content. Parasitological parameters were evaluated 7 wks PI. ### Principal findings Immunization of outbred mice with MAP-2 in combination with alum and/or MAP-1 elicited highly significant ($P \leq 0.005$) reduction of around $60\%$ in challenge S. mansoni worm burden and no increase in worm eggs’ loads or vitality, compared to unimmunized or alum pre-treated control mice. Host memory responses to the immunogens are expected to be expressed in the liver stage when worm feeding and cysteine peptidases release start to be active. Serum antibody and cytokine levels were not significantly different between control and vaccinated mouse groups. Highly significant ($P \leq 0.05$ - <0.0001) increase in liver interleukin-1, ARA, and ROS content was recorded in MAP-immunized compared to control mice. ### Conclusion/Significance The findings provided an explanation for the gut cysteine peptidases vaccine-mediated reduction in challenge worm burden and increase in egg counts. ## Author summary Adjuvant-free cysteine peptidases consistently elicited remarkable reduction in challenge schistosome worm burden in outbred rodents, whether used in an enzymatically active or inactive construct. The findings together suggested that peptide sequences shared by these cysteine peptidases may substitute for the whole molecule and form the basis of a safe, cost-effective, chimeric protein vaccine, easy to manufacture and deliver in countries with limited resources. In support, multiple antigen peptide (MAP) construct of two peptides, MAP-1 and MAP-2, showing high homology to helminth gut cysteine peptidases induced $25\%$-$30\%$ reduction in challenge Schistosoma mansoni worm burden in outbred mice. It was, however, necessary to modify the vaccine formula to counteract the effect on the parasite egg counts and vitality. Immunization of mice with MAP-2 in combination with alum and/or MAP-1 elicited reduction of around $60\%$ in challenge S. mansoni worm burden and no increase in worm eggs’ loads or vitality, compared to unimmunized or alum pre-treated control mice. Considerable increase in liver interleukin-1, arachidonic acid, and reactive oxygen species content in MAP-immunized compared to control mice appeared to elucidate the mechanisms underlying the dual impact of the cysteine peptidase-based schistosomiasis vaccine. ## Introduction Adjuvant-free, enzymatically active or inactive cysteine peptidases, notably Schistosoma mansoni cathepsin B1 (SmCB1), and cathepsin L3 (SmCL3), *Schistosoma haematobium* cathepsin L (ShCL), *Fasciola hepatica* cathepsin L1 (FhCL1) and the prototype, papain consistently elicited highly significant ($P \leq 0.005$) reduction in S. mansoni and S. haematobium challenge worm burden in outbred mice and hamsters, respectively [1–10]. These findings indicated that protection might be induced independently of the enzymes proteolytic activity, i.e., host hydrolysis products generated following primary and boost immunization are not essential for worm elimination. The findings together suggested that peptide sequences shared by these cysteine peptidases may substitute for the whole molecule and form the basis of a cost-effective, chimeric protein vaccine, easy to manufacture and deliver in countries with limited resources. The vaccine would be safe, because peptide homology with host corresponding molecules leads to limited antibody generation to the immunogens, but precludes autoimmune responses [1,2]. In support, immunization of outbred mice with adjuvant-free, cysteine peptidases-derived MAP constructs, (MAP-1 and MAP-2) elicited modest, type 2-skewed antibody and cytokine responses conducive to significant ($P \leq 0.05$) reduction of about $30\%$ in challenge S. mansoni worm burden. Yet, MAP-2 immunization was associated with egg loads in liver and small intestine not different from infected control mice. The small intestine circumoval granulomas number and diameter were larger than in MAP-1-immunized mice that showed considerable egg counts in liver and intestine. The mechanism(s) underlying these effects were not explored, but it was necessary to modify the vaccine formulation in an attempt to skew the host responses towards restricting parasite eggs production and vitality. It was recommended to intensify the immune responses to MAP-2, in view of controlling the vitality of the eggs produced by the worms, and only use MAP-1 combined with MAP-2 in order to elicit host responses restricting parasite fecundity [11]. Experiments were, therefore, performed using MAP-1 and MAP-2, alone or in mixture, or combined with alum adjuvant [12,13]. The aims were to increase the challenge worm reduction level, control the caveats of peptide immunization on the egg counts and granulomas formation, and find clues to the mechanism(s) allowing host responses to simple peptides to significantly interfere with challenge worm survival and reproduction. ## Ethics statement All experiments involving animals were conducted according to the ethical policies and procedures approved by the Ethics Committee of the Faculty of Science, Cairo University, Egypt (Approval no. CU/I/F/$\frac{65}{19}$). ## Multiple antigen peptide synthesis Peptides IRDQSRCGSSWAFGAVEAMS, and EQQLVDCSYKYGNDGCQGG, showing highest sharing of amino acid sequences with helminth and murine cathepsins B and L, papain, and major allergens were synthesized as endotoxin-free, tetra-branched multiple antigen peptide (MAP) constructs at Thermo Fisher scientific (Waltham, MA, USA), and designated as MAP-1 and MAP-2, respectively [11]. ## Mice and parasites Outbred, female, six week-old CD1 mice were obtained from the Schistosome Biological Supply Program (SBSP) at Theodore Bilharz Research Institute (TBRI) Giza, Egypt and maintained throughout experimentation at the animal facility of the Zoology Department, Faculty of Science, Cairo University. Cercariae of an Egyptian strain of S. mansoni were obtained from SBSP/TBRI, and used for infection immediately after shedding from *Biomphalaria alexandrina* snails. ## Experimental plan Two independent experiments were performed in parallel, and in each, 5 of 45 mice were retained without immunization or schistosome infection, and considered as naïve, while 40 mice were randomly distributed into three groups and vaccinated intramuscularly, twice with a three weeks (wks)-interval. In Experiment 1, each of 13 or 14 mice were injected with immunogen- and adjuvant-free Dulbecco’s phosphate-buffered saline, pH 7.1 (D-PBS), 25 μg MAP-2 adsorbed on 130 μg alum (Alhydrogel, Aluminum Hydroxide Gel 13 mg/mL, Sigma-Aldrich-Merck, Darmstadt, Germany), or 12.5 μg MAP-1 +12.5 μg MAP-2 in 100 μL D-PBS. In Experiment 2, each of 13 or 14 mice were immunized with immunogen- and alum-free D-PBS (control mice), immunogen-free alum (130 μg/mouse; adjuvant controls), or 15 μg MAP-1 + 10 μg MAP-2 + 130 μg alum adjuvant. Challenge infection was performed three wks after the second injection via percutaneous exposure of each mouse to 100 viable cercariae of S. mansoni, as described [11]. Blood and liver pieces were obtained from 6 to 8 mice per group, 23 days post-infection (PI), a time of schistosome development and feeding in the liver before mating [14], and immediately processed before storing at -20°C until use. Serum samples were examined for the levels of circulating antibodies and cytokines. Liver cell extracts were used for assessment of hepatic cytokines, uric acid, arachidonic acid (ARA), and reactive oxygen species (ROS) content. Parasitological parameters were evaluated in 5 to 7 mice per group 7 wks PI. No attempt was made to assess cytokine or antibody responses to the peptide immunogens at this interval because of the confounding strong reactivities to the parasite egg antigens [1–9,11]. ## Parasitological parameters Worm burden was evaluated by hepatic portal venous system and mesenteric blood vessels perfusion as described in detail previously [11]. After perfusion, the liver and small intestine of each mouse were harvested, and 50 mg pieces processed for histological examination. Parasite egg burden of individual mice was evaluated in 200 mg liver or small intestine following incubation in $4\%$ KOH for 1 h at 40°C as described [1–11]. Percent change in worm and egg burden was evaluated by the formula: % change = [mean number in infected controls − mean number in immunized infected mice / mean number in infected controls] × 100. Liver and small intestine sections from each mouse were stained with haematoxylin and eosin and examined for the number/field and diameter of circumoval granulomas [6–9]. Hepatic egg granulomas numbers are mean ± SD /field of 5–10 fields per section of 5 mice per group. Granuloma diameters (μm) are shown as mean ± SD of all circumoval granulomas in sections of five mice per group. Photographs were acquired by light microscopy (Olympus, Tokyo, Japan). Following our recommendation to improve MAP-1 and MAP-2 protective potential [11], MAP-2 emulsified in alum, or combined with MAP-1 was used to immunize mice against challenge S. mansoni infection. Highly significant ($P \leq 0.005$, Mann-Whitney) reduction in total, male and female worm burden was recorded, varying between 52.1 and $61.6\%$ (Table 1). Whilst MAP-2—alum immunization led to substantial, albeit insignificant, decrease in egg load in liver and small intestine, adding MAP-2 to MAP-1 failed to modify the ability of the latter to elicit considerable increase in number of eggs retrieved in small intestine (Table 1). Thus, MAP-1 + MAP-2 immunization led to increase in fecundity of surviving worms (Fig 1A and 1B). Despite the differences in worm survival and egg loads, the mean number /field (10 x 10) and diameter of liver (Table 1) and small intestine egg granulomas did not differ between the three groups, as assessed by one-way ANOVA. The data confirm that host responses to MAP-1 are associated with increase in number of liver and small intestine eggs that are characterized by limited viability and immunogenicity [11]. **Fig 1:** *Parasite egg parameters.Worm fecundity was evaluated based on liver (A) and small intestine (B) parasite egg counts in individual mice, at wk 7 post infection (PI).* TABLE_PLACEHOLDER:Table 1 Since it was not recommended to use MAP-2 immunogen without adjuvant, MAP-1 and MAP-2 immunogen mixture was combined with alum, and impact on challenge worm parameters compared to untreated and alum-administered controls. Alum administration failed to affect challenge worm burden or egg counts and granulomas number compared to untreated controls (Figs 2 and 3). MAP-1+ MAP-2+ alum immunization elicited a record ($$P \leq 0.0011$$, Mann-Whitney) reduction in total, male and female worm burden of $65.8\%$, higher than for any full-length cysteine peptidase tested, including papain [1–11] (Fig 2). The egg counts in liver and small intestine were not increased and did not differ from infected and alum controls (Fig 2), for the first time with MAP-1 inclusion. The number and diameter of liver egg granulomas were similar in the control (untreated and alum-administered) and immunized mice (S1 Fig). Yet, and even more importantly than the reduction in worm burden, immunization with MAP-1 + MAP-2 + alum was associated with significant decrease in number ($$P \leq 0.0378$$) and diameter ($$P \leq 0.0003$$) of small intestine egg granulomas compared to the control mice (Table 2 and Fig 3). The data indicated that addition of alum or MAP-1 to MAP-2 was associated with impaired parasite egg ability to transit to exit points. **Fig 2:** *Parasite worm and egg counts.Five to 7 mice per group were examined for parasitological parameters 7 wks post challenge infection. P values indicate levels of statistical (Mann-Whitney) differences between immunized and infected control mice.* **Fig 3:** *Small intestine haematoxylin/eosin-stained sections.D-PBS (A), alum (B), MAP-1+ MAP-2+ alum-administered mice, 7 wks post challenge infection with S. mansoni. Typical of 5 mice per group. The arrows point to the egg granulomas. x 100.* TABLE_PLACEHOLDER:Table 2 ## Serum cytokine and antibody assays Quantitative determination of mouse interleukin (IL)-4, IL-5, and IL-17, and interferon-gamma (IFN-γ) (ELISA MAX Set, BioLegend, San Diego, CA, USA) was evaluated in individual mouse sera using capture enzyme-linked immunosorbent assay (ELISA), following the manufacturer’s instructions. The antibody isotypes to a mother cysteine peptidase molecule, recombinant FhCL1 [3], a gift of Professor Dr. John P. Dalton, were determined by indirect ELISA in 1:100-diluted sera, assayed on an individual mouse basis, as described [11]. Absorbances of duplicate wells were evaluated spectroscopically at 405 nm (Multiskan EX, Labsystems, Helsinki, Finland). ## Liver extracts preparation and protein content The weight of a liver piece from each mouse was recorded before homogenization in D-PBS supplemented with $0.1\%$ Triton X-100, and protease inhibitors: leupeptin (4 μg/mL) and 1 mM phenyl methyl sulfonyl fluoride (Merck). The homogenates were incubated on ice with shaking for 30 min, and then centrifuged at 400 x g for 10 min [15]. The supernatants were retrieved in ice-cold reaction tubes. Liver extracts were assessed for protein content spectrophotometrically at 280 and 260 nm, using the formula: protein concentration mg/mL = 1.55 x A280−0.76 x A260 and at 595 nm for the Bio-Rad Protein Assay, and stored at -20°C until use. ## Cytokines analysis Quantitative determination of mouse thymic stromal lymphopoietin (TSLP), IL-25, IL-33, IL-1β, IL-10, IL-13 (R&D Systems, Minneapolis, MN, USA), IL-4, IL-5, IL-17, and IFN-γ (BioLegend), was evaluated in individual mouse liver Triton X-100 extracts (200 μg protein per each of duplicate wells) using capture ELISA, following the manufacturer’s instructions. ## Uric acid assays Liver Triton X-100 extracts were assayed for uric acid content in duplicate 50 μg protein samples per well using in parallel Uric Acid Assay Kit (ab65344, Abcam, Cambridge, UK), and Uric Acid Kit (Chronolab Systems, S.L., Barcelona, Spain), following the manufacturers recommendations and procedures. Additionally, 10 mg liver were thoroughly homogenized and added with 200 μL uric acid assay buffer (100 mM Tris-HCl, pH 7.5), incubated for 30 min on ice and centrifuged at 5,000 x g for 2 min. The supernatant was retrieved in ice-cold reaction tubes and 2.5 and 5 μL/well in duplicates were immediately examined for uric acid content using the Uric Acid Kit (Chronolab). ## Arachidonic acid assays Free ARA content in liver Triton X-100 extracts [16] was evaluated by capture ELISA. Wells were coated with 250 ng unlabelled rabbit polyclonal antibody to ARA (MyBioSource, San Diego, CA, USA, MBS2003715) overnight at 4°C. Following washing in 0.1 M PBS/$0.05\%$ Tween 20 (PBS-T), 200 μg liver protein of naive, control and immunized mice were added in duplicate wells to a total volume of 100 μL PBS-T, and incubated for 2 h at room temperature. The wells were thoroughly washed and added with 150 ng horseradish peroxidase-linked polyclonal antibody to ARA (MyBioSource, MBS2051576) for 1 h at room temperature. The reaction was visualized 30 min after adding 3,3’,5,5’ tetramethylbenzidine substrate (Sigma). Arachidonic acid content was additionally evaluated by immunohistochemistry as described previously [8,9,17], except that liver sections were exposed to $3\%$ hydrogen peroxide (Sigma) to block endogenous peroxidase activity, then incubated with 0 or 0.5 μg horseradish peroxidase-linked polyclonal antibody to ARA (MyBioSource, MBS2051576) overnight at 10°C. The reaction was visualized with Dako Liquid DAB + Substrate Chromogen System (Agilent Dako, Santa Clara, CA, USA). Photographs were acquired by light microscopy. ## Reactive oxygen species assays 2’,7’-dichlorodihydrofluorescein diacetate (DCHF-DA) is a cell-permeable non-fluorescent probe. After mixing with cell homogenates, DCHF-DA is deacetylated by cellular esterases to a non-fluorescent compound which is readily oxidized by ROS into a highly fluorescent compound, 2’, 7’–dichlorofluorescein [18]. Duplicates of 25, 50, 100 and 200 μg liver proteins of individual mice were incubated with 20 μM DCHF-DA (Merck, D6883) at room temperature, in the dark, for 1 h and ROS release estimated by fluorescence spectroscopy with maximum excitation (Ex) and emission (Em) spectra of 485 nm and 535 nm, respectively (Victor X4 Multi-Label Plate Reader, PerkinElmer, Waltham, MA, USA). ## Statistical analysis All values were tested for normality. Students’–t- 2-tailed, Mann-Whitney, and one-way ANOVA with post test were used to analyze the statistical significance of differences between selected values, and considered significant at $P \leq 0.05$ (GraphPad InStat, San Diego, CA, USA). ## Serum immune responses Enough serum was collected from 4 mice and tested on an individual mouse basis for levels of circulating IL-4, IL-5, IL-17 and IFN-γ, and antibody isotype response to a mother cysteine peptidase molecule, on day 23 PI, at time developing worms are still in the liver [14]. Infection with S. mansoni in untreated and alum-administered mice elicited increase in serum IL-4, IL-17, and IFN-γ, compared to naïve mice. Serum cytokine levels in MAP-vaccinated mice were lower than in naïve and control infected mice, except for IL-4 and IL-17 ($P \leq 0.05$) in mice immunized with MAP-1 + MAP-2 (Fig 4A–4D). Serum anti-cysteine peptidase antibodies levels and isotypes were not different among naïve, control and vaccinated mice, except for increase in IgG2a antibodies in MAP-1 + MAP-2 + alum-immunized mice (Fig 4E). **Fig 4:** *Serum cytokine and antibody isotypes levels in mice examined 23 days PI.(A-D), Each column represents mean cytokine levels of 4 mice assessed on an individual basis, and vertical bars the standard error (SE) about the mean. (E), each symbol represents mean absorbance of 4 mice assessed on an individual basis with SE < 5%. Statistical differences were assessed between MAP-vaccinated mice versus naïve and infected controls (Inf). * P < 0.05.* ## Liver cytokines Liver cells of healthy, untreated and uninfected naïve mice released a plethora of type 1, type 2, and type 17 cytokines. Developing, 23 days-old S. mansoni worms released molecules that were not able to modulate the levels of the type 2 cytokines, TSLP, IL-25, IL-4, IL-5, and IL-13, and elicited significant decrease of IL-33 ($P \leq 0.01$). No impact was recorded on the levels of released IL-1, IL-10, and IFN-γ compared to naïve mice, while the most remarkable change concerned significant ($P \leq 0.01$) increase in hepatic cell release of IL-17 (Fig 5). Booster alum administration 3 wks before percutaneous infection with S. mansoni cercariae was associated liver cells production of significantly ($P \leq 0.05$- $P \leq 0.005$) less of each cytokine tested, realizing again balance of cytokine types, only at a lower quantitative level than in naïve animals. Except for IFN-γ, levels of all cytokine tested were significantly ($P \leq 0.05$- $P \leq 0.005$) lower when compared to infected mice (Fig 5). **Fig 5:** *Cytokine levels in liver Triton X-100 extracts on day 23 PI.Each column represents ng cytokine/mg liver proteins of 5–8 individual mice/group, and vertical bards denote the SE about the mean. Asterisks denote significant (P < 0.05- < 0.005) differences between infected and naïve or MAP-immunized and infected mice.* Immunization of mice with MAP-2 + alum was associated with liver cells production of cytokine levels significantly ($P \leq 0.05$) lower than naïve and infected mice, likely because of the impact of alum. Conversely, IL-1 levels were significantly ($P \leq 0.005$) higher than untreated and alum-administered infected mice. Immunization with alum-free MAP-1 and MAP-2 led to highly significant decrease in challenge worm burden that correlated again with significant ($P \leq 0.05$) increase in hepatic cells IL-1. Decrease ($P \leq 0.05$) in liver cells type 1 and type 2 cytokines production, compared to unimmunized infected mice, allowed IL-1 preponderance, perhaps giving a clue for the large increase in production of eggs with limited immunogenicity and viability in this group. MAP-2+MAP-1+ alum immunization induced highly significant ($P \leq 0.005$) decrease in challenge worm recovery and small intestine pathology, and the most extremely significant decrease ($P \leq 0.005$) in cytokines tested, namely all type 2 cytokines, IL-10, and IL-17, with the exception of IL-1, which levels did not differ from unimmunized infected mice (Fig 5). ## Uric acid Liver Triton X-100 extracts were assayed in duplicate wells for uric acid levels using two separate assays. Similar results were obtained and therefore pooled. Twenty three days S. mansoni infection whether preceded or not by alum treatment or MAP immunization elicited no statistically (Anova and Mann Whitney tests) significant changes in host liver uric acid content (S2A Fig). Similar results were obtained using liver uric acid buffer extracts (S2B Fig). ## Arachidonic acid Repeat capture ELISA tests confirmed that the levels of liver ARA readily extracted by Triton X-100 [16] significantly ($$P \leq 0.0007$$) increased at 23 days PI compared to naïve mice, provided that alum was not administered before infection (Fig 6). Immunization with alum adjuvanted MAP-2 and MAPs mixture overcame ($P \leq 0.05$) the alum impact and was associated with ARA content significantly higher ($P \leq 0.05$ - < 0.002) than naïve, but not control infected, mice (Fig 6). Histochemical findings mirrored the capture ELISA results (S3 Fig). **Fig 6:** *Liver free arachidonic acid reactivity in capture ELISA.Each point represents mean of duplicate wells for 5–7 individual naïve, infected (Inf) and MAP-immunized mice, 23 days PI, and horizontal lines depict the median. Values in the different groups significantly (P = 0.0014) differed as assessed by ANOVA. Values of mice immunized with MAP-2 + alum, MAP-1 + MAP-2, and MAP-1 + MAP-2+ alum differed significantly from naïve (P <0.05-<0.005) and from alum/infected (P <0.05) mice.* ## Reactive oxygen species Results illustrated in Fig 7 and Table 3 show the effect of MAP immunogen formulations on liver ROS content on day 23 PI. **Fig 7:** *Reactive oxygen species fluorescence.Each point represents mean reactivity of 25 (blue diamond), 50 (brown squares), 100 (green triangles), and 200 (black circles) μg liver proteins of 6 to 8 mice per group.* TABLE_PLACEHOLDER:Table 3 ## Discussion The results obtained in the present study indicated that our recommendations regarding the use of MAP-1 and MAP-2 were judicious as the novel formulations elicited highly significant ($P \leq 0.005$) of > $60\%$ to up to $68\%$ (with MAP-1 + MAP-2 + alum) challenge worm burden reduction. Decrease in liver and intestine worm egg counts, immunogenicity, and viability was also achieved. Attempts at deciphering the mechanism(s) underlying protection were made at the liver stage, 23 days PI, prior to worm maturity, mating, and migration to the final abode in the capillaries of intestine mesenteries and egg laying for two reasons. First, in the liver, developing worms voraciously feed on host erythrocytes, and increasingly produce the cysteine peptidases necessary for digestion [14,19–21]. Release of cathepsins B and L would activate immune memory responses to the MAP immunogens. Second, like for the lung capillaries, the liver sinusoids represent a danger for the migrating worms because of the ease of extravasation to certain demise [2,14,22–24]. Despite that at this time worms are feeding and regurgitate cathepsins B and L-rich products [20], the induced primary (in untreated and alum-administered infected mice) and memory (mice immunized with cathepsins-related MAP) antibody responses to the target cathepsin molecule were negligible. These data are in full accord with findings in mice reported and discussed previously [11], and in rabbits immunized thrice with SmCB1-derived peptides suspended in complete and incomplete Freund’s adjuvant [25]. Alum was not co-administered with the invading schistosome cercariae and was not expected to potentiate the humoral responses to worm-released molecules. However, increase in serum IgG2a antibodies to FhCL1 in MAP-1+MAP-2+alum-immunized mice could be related to alum co-administration [12,13,26]. Mice of this group showed the highest reduction in challenge worm burden and small intestine egg granulomas number and diameter (Table 2 and Figs 2 and 3), likely because of activation of inflammatory immune cells by antibody/released gut cathepsin complexes, a mechanism of worm attrition discussed previously [1,2,9] (Fig 8). **Fig 8:** *Proposed mechanisms underlying cysteine peptidases-based vaccine impact on challenge Schistosoma mansoni. Icons modified from https://commons.wikimedia.org.* Increase in serum IL-4, IL-17, and IFN-γ levels was noticed in infected control mice, associated with developing worms secretion of SmCB1 [27]. Alum co-administered with the MAP immunogens failed to potentiate their serum cytokine response. Indeed, MAP immunization was associated with serum cytokine levels lower than in infected control mice, except for increase in IL-4 and IL-17 in adjuvant-free MAP-1 + MAP-2-immunized mice, showing the only cases of large egg loads in liver and small intestine. Of interest, the hepatic levels of cytokines, ARA, and ROS significantly ($P \leq 0.05$ - <0.0001) differed among the various mouse groups. The balance of type 1, type 2, and type 17 cytokines released by liver cells of untreated and uninfected naïve mice was remarkable. Liver cytokines of 23 days-infected mice differed from naïve mice in significant decrease in the levels of IL-33, reported to be dispensable for S. mansoni maturation [28]. The most striking finding ($P \leq 0.01$) involved increase in IL-17, likely responsible for the changes in cellular infiltration in the liver as early as 4 wks PI, compared to naïve mice, described by Costain et al. [ 29]. Despite the dramatic increase reported by Costain et al. [ 29] in numbers of hepatic eosinophils and macrophages in 4 wks-infected mice, no changes in ARA or ROS hepatic content was recorded in 23 days-infected mice, compared to naïve controls. The adjuvant used, alum (aluminum hydroxide, 130 μg/injection) failed to impact challenge worms’ survival or fecundity, associated with induction of significant ($P \leq 0.05$ to $P \leq 0.005$) reduction in liver cytokines, including IL-17, ARA, and ROS content when compared to levels recorded in naïve or unimmunized infected mice. It is important to recall that adjuvants, including alum, trigger stromal cells responses at the site of injection [12] and may not impact remote sites. Compared to naïve and untreated or alum-administered infected mice, MAP-immunized mice showed at 23 days PI, significant ($P \leq 0.05$ - <0.0001) increase in content of hepatic IL-1, ARA, and ROS. Interleukin-1β is predominantly produced by myeloid cells but also by B and T lymphocytes [30–33], and is a key mediator in inflammation initiation, maintenance, and amplification [30–36]. We surmise that elevated IL-1 release induces accumulation and activation of neutrophils, eosinophils, and macrophages. Activation of these inflammatory cells promotes release of free ARA from their cell membranes [37,38] that in turn enhances their ROS production [37,39]. Both ARA [for review see 40] and ROS [41–43] activate schistosomes and cells surface membrane-associated neutral sphingomyelinase (nSMase)-2. Hepatic cells are not damaged because of lack or very low levels of surface membrane-associated nSMase-2 [42], while nSMase-mediated hydrolysis of worms surface membrane sphingomyelin is a direct killing hit [9,40]. Increase in hepatic cells ROS levels is counteracted by the persistent high content of the anti-oxidant uric acid [44]. Conversely, the juvenile worms surface membrane lipids and proteins are irreversibly oxidized, further impairing the integrity of their outer membrane shield [45–47]. Thus, IL-1-related increase in hepatic free ARA and ROS content in MAP-immunized mice provides an explanation for the recorded highly significant ($P \leq 0.005$) reduction in challenge worm burden. However, IL-1β promotes triglycerides and cholesterol accumulation in murine and human liver cells [48–50]. Elevated levels of triglycerides, cholesterol, and ARA [48–52] promote the reproductive activities and egg production of the surviving worms, explaining the difficulties in controlling worm egg output in mice immunized with cysteine peptidase in full length [1–10] or peptide [11] constructs (Fig 8). The solution would be to devise peptide-based formulations that induce increase in IL-1, ARA, and ROS targeting both the lung and liver schistosome stages. Lung-stage larvae are exceedingly sensitive to ARA [40] and ROS [45–47] schistosomicidal effects, while it is premature for IL-1-related activation of lipids metabolism to impact the surviving larvae reproductive functions. Support is provided by the observation that the highest reduction in challenge S. mansoni worm burden ($76.5\%$, $$P \leq 0.0006$$), and liver and small intestine egg loads ($61.6\%$, $$P \leq 0.0006$$; and $57.1\%$, $$P \leq 0.0023$$, respectively) was achieved by outbred mouse immunization with SmCB1 + SmCL3 combined with SG3PDH [5], S. mansoni glyceraldehyde phosphate dehydrogenase, a prominent lung-stage larvae excretory-secretory product [1,2,5]. ## Conclusions Peptides common to several gut cysteine peptidases in MAP construct, formulated in mixture and/or combined with the adjuvant, alum elicited highly significant ($P \leq 0.005$) reduction in challenge worm burden. Memory responses to the immunogens are expected to be expressed at the post-lung stage, notably in the liver. The impairment in challenge worm survival in immunized mice was associated with hepatic increase in the levels of the pro-inflammatory IL-1, and the schistosomicidal ARA and ROS. Interleukin-1-related activation of hepatic lipids, ARA, and cholesterol synthesis likely promotes the reproductive functions and egg formation in the surviving liver-stage schistosomes, explaining the difficulty in reducing parasite egg outputs in mice immunized with cysteine peptidases in full-length or peptidic constructs (Fig 8). Reduction of worm egg output is proposed to be achieved via peptide formulations eliciting memory immune responses targeting both the lung- and liver- schistosome developmental stages. ## References 1. El Ridi R, Tallima H. **Vaccine-induced protection against murine schistosomiasis mansoni with larval excretory-secretory antigens and papain or type-2 cytokines**. *J Parasitol* (2013) **99** 194-202. DOI: 10.1645/GE-3186.1 2. El Ridi R, Tallima H. **Novel therapeutic and prevention approaches for schistosomiasis**. *J Adv Res* (2013) **4** 467-478. DOI: 10.1016/j.jare.2012.05.002 3. El Ridi R, Tallima H, Selim S, Donnelly S, Cotton S, Gonzales SB, Dalton JP. **Cysteine peptidases as schistosomiasis vaccines with inbuilt adjuvanticity**. *PloS One* (2014) **9** e85401. DOI: 10.1371/journal.pone.0085401 4. Tallima H, Dalton JP, El Ridi R. **Induction of protective immune responses against**. *Front Immunol* (2015) **6** 130. DOI: 10.3389/fimmu.2015.00130 5. Tallima H, Dvořák J, Kareem S, Abou El Dahab M, Abdel Aziz N, Dalton JP. **Protective immune responses against**. *PLoS Negl Trop Dis* (2017) **11** e0005443. DOI: 10.1371/journal.pntd.0005443 6. Tallima H, Abou El Dahab M, Kareem S, Dalton JP, El Ridi R. **Protection against**. *Vaccine* (2017) **135** 6977-6983. DOI: 10.1016/j.vaccine.2017.10.069 7. Abdel Aziz N, Tallima H, Abou El Dahab M, El Ridi R. **Immunogenicity and protective capacity of**. *Vaccine Res* (2019) **6** 1-8. DOI: 10.29252/vacres.6.2.1 8. Tallima H, Abou El Dahab M, El Ridi R. **Role of T lymphocytes and papain enzymatic activity in the protection induced by the cysteine protease against**. *J Adv Res* (2019) **17** 73-84. DOI: 10.1016/j.jare.2018.12.008 9. Tallima H, Abou El Dahab M, El Ridi R. **Specific antibodies and arachidonic acid mediate the protection induced by the**. *Vaccines* (2020) **8** 682. DOI: 10.3390/vaccines8040682 10. Hassan AS, Perera DJ, Ward BJ, Ndao M. **Therapeutic activity of a**. *Vaccine* (2021) **39** 5580-5588. DOI: 10.1016/j.vaccine.2021.08.031 11. Al-Naseri A, Al-Absi S, Mahana N, Tallima H, El Ridi R. **Protective immune potential of multiple antigenic peptide (MAP) constructs comprising peptides that are shared by several cysteine peptidases against**. *Mol Biochem Parasitol* (2022) **248** 111459. DOI: 10.1016/j.molbiopara.2022.111459 12. Lambrecht BN, Kool M, Willart MA, Hammad H. **Mechanism of action of clinically approved adjuvants**. *Curr Opin Immunol* (2009) **21** 23-29. DOI: 10.1016/j.coi.2009.01.004 13. He P, Zou Y, Hu H. **Advances in aluminum hydroxide-based adjuvant research and its mechanism**. *Hum Vaccin Immunother* (2015) **11** 477-488. DOI: 10.1080/21645515.2014.1004026 14. Nation CS, Da’dara AA, Marchant JK, Skelly JP. **Schistosome migration in the definitive host.**. *PLoS Negl Trop Dis* (2020) **14** e0007951. DOI: 10.1371/journal.pntd.0007951 15. Nono JK, Mpotje T, Mosala P, Aziz NA, Musaigwa F, Hlaka L. **Praziquantel treatment of**. *Front Immunol* (2021) **12** 748387. DOI: 10.3389/fimmu.2021.748387 16. London E, Brown DA. **Insolubility of lipids in triton X-100: physical origin and relationship to sphingolipid/cholesterol membrane domains (rafts)**. *Biochim Biophys Acta* (2000) **1508** 182-195. DOI: 10.1016/s0304-4157(00)00007-1 17. Hanna VS, Gawish A, Abou El-Dahab M, Tallima H, El Ridi R. **Is arachidonic acid an endoschistosomicide?**. *J Adv Res* (2018) **11** 81-89. DOI: 10.1016/j.jare.2018.01.005 18. Chen X, Zhong Z, Xu Z, Chen L, Wang Y. **2’,7 Dichlorodihydrofluorescein as a fluorescent probe for reactive oxygen species measurement: Forty years of application and controversy**. *Free Radic Res* (2010) **44** 587-604. DOI: 10.3109/10715761003709802 19. Caffrey CR, Rheinberg CE, Moné H, Jourdane J, Li YL, Ruppel A. *Parasitol Res* (1997) **83** 37-41. DOI: 10.1007/s004360050204 20. Zerda KS, Dresden MH, Chappell CL. *Exp Parasitol* (1988) **67** 238-246. DOI: 10.1016/0014-4894(88)90071-9 21. Skelly PJ, Da’dara AA, Li XH, Castro-Borges W, Wilson RA. **Schistosome feeding and regurgitation**. *PLoS Pathog* (2014) **10** e1004246. DOI: 10.1371/journal.ppat.1004246 22. Miller P, Wilson RA. **Migration of the schistosomula of**. *Parasitology* (1980) **80** 267-288. DOI: 10.1017/s0031182000000743 23. Delgado VS, McLaren DJ. *Parasitology* (1990) **100** 57-63. DOI: 10.1017/s003118200006011x 24. Wilson RA. **The saga of schistosome migration and attrition**. *Parasitology* (2009) **136** 1581-1592. DOI: 10.1017/S0031182009005708 25. Noya IO, De Noya BA, Ballen DE, Bermúdez H, Bout D, Hoebeke J. **Immunogenicity of synthetic peptides from the Sm31 antigen (cathepsin B) of the**. *Parasite Immunol* (2001) **23** 567-573. DOI: 10.1046/j.1365-3024.2001.00414.x 26. Fan J, Jin S, Gilmartin L, Toth I, Hussein WM, Stephenson RJ. **Advances in infectious disease vaccine adjuvants**. *Vaccines (Basel)* (2022) **10** 1120. DOI: 10.3390/vaccines10071120 27. Soloviova K, Fox EC, Dalton JP, Caffrey CR, Davies SJ. **A secreted schistosome cathepsin B1 cysteine protease and acute schistosome infection induce a transient T helper 17 response**. *PLoS Negl Trop Dis* (2019) **13** e0007070. DOI: 10.1371/journal.pntd.0007070 28. Mukendi JPK, Nakamura R, Uematsu S, Hamano S. **Interleukin (IL)-33 is dispensable for**. *Parasit Vectors* (2021) **14** 70. DOI: 10.1186/s13071-020-04561-w 29. Costain AH, Phythian-Adams AT, Colombo SAP, Marley AK, Owusu C, Cook PC. **Dynamics of host immune response development during**. *Front Immunol* (2022) **13** 906338. DOI: 10.3389/fimmu.2022.906338 30. Demczuk S, Baumberger C, Mach B, Dayer JM. **Expression of human IL 1 alpha and beta messenger RNAs and IL 1 activity in human peripheral blood mononuclear cells**. *J Mol Cell Immunol* (1987) **3** 255-265. PMID: 3509926 31. Dinarello CA. **Interleukin-1 and interleukin-1 antagonism**. *Blood* (1991) **77** 1627-1652. PMID: 1826616 32. Brown S, Boules M, Hamza N, Wang X, Whalen M. **Synthesis of interleukin 1 beta and interleukin 6 in human lymphocytes is stimulated by tributyltin**. *Arch Toxicol* (2018) **92** 2573-2586. DOI: 10.1007/s00204-018-2248-2 33. Linder A, Hornun V. **Inflammasomes in T cells**. *J Mol Biol* (2022) **434** 167275. DOI: 10.1016/j.jmb.2021.167275 34. Yazdi AS, Ghoreschi K. **The interleukin-1 family**. *Adv Exp Med Biol* (2016) **941** 21-29. DOI: 10.1007/978-94-024-0921-5_2 35. Dinarello CA. **Overview of the IL-1 family in innate inflammation and acquired immunity**. *Immunol Rev* (2018) **281** 8-27. DOI: 10.1111/imr.12621 36. Matarazzo L, Hernandez Santana YE, Walsh PT, Fallon PG. **The IL-1 cytokine family as custodians of barrier immunity**. *Cytokine* (2022) **154** 155890. DOI: 10.1016/j.cyto.2022.155890 37. Hanna VS, Hafez EAA. **Synopsis of arachidonic acid metabolism: A review**. *J Adv Res* (2018) **11** 23-32. DOI: 10.1016/j.jare.2018.03.005 38. Tallima H.. **Clarification of arachidonic acid metabolic pathway intricacies**. *ACS Omega* (2021) **6** 15559-15563. DOI: 10.1021/acsomega.1c01952 39. Tallima H, El Ridi R. **Arachidonic acid: Physiological roles and potential health benefits—A review**. *J Adv Res* (2018) **11** 33-41. DOI: 10.1016/j.jare.2017.11.004 40. Tallima H, Hanna VS, El Ridi R. **Arachidonic acid is a safe and efficacious schistosomicide, and an endoschistosomicide in natural and experimental infections, and cysteine peptidase vaccinated hosts**. *Front Immunol* (2020) **11** 609994. DOI: 10.3389/fimmu.2020.609994 41. Levy M, Castillo SS, Goldkorn T. **nSMase2 activation and trafficking are modulated by oxidative stress to induce apoptosis**. *Biochem Biophys Res Commun* (2006) **344** 900-905. DOI: 10.1016/j.bbrc.2006.04.013 42. Nikolova-Karakashian M, Karakashian A, Rutkute K. **Role of neutral sphingomyelinases in aging and inflammation**. *Subcell Biochem* (2008) **49** 469-486. DOI: 10.1007/978-1-4020-8831-5_18 43. Dotson PP, Karakashian AA, Nikolova-Karakashian MN. **Neutral sphingomyelinase-2 is a redox sensitive enzyme: role of catalytic cysteine residues in regulation of enzymatic activity through changes in oligomeric state**. *Biochem J* (2015) **465** 371-382. DOI: 10.1042/BJ20140665 44. Kang DH, Ha SK. **Uric acid puzzle: dual role as anti-oxidant and pro-oxidant**. *Electrolyte Blood Press* (2014) **12** 1-6. DOI: 10.5049/EBP.2014.12.1.1 45. Nare B, Smith JM, Prichard PK. *Exp Parasitol* (1990) **70** 389-397. DOI: 10.1016/0014-4894(90)90122-s 46. Loverde PT. **Do antioxidants play a role in schistosome host-parasite interactions?**. *Parasitol Today* (1998) **14** 284-289. DOI: 10.1016/s0169-4758(98)01261-7 47. Huang HH, Rigouin C, Williams DL. **The redox biology of schistosome parasites and applications for drug development**. *Curr Pharm Des* (2012) **18** 3595-3611. PMID: 22607149 48. Feingold KR, Soued M, Serio MK, Moser AH, Dinarello CA, Grunfeld C. **Multiple cytokines stimulate hepatic lipid synthesis in vivo**. *Endocrinology* (1989) **25** 267-274. DOI: 10.1210/endo-125-1-267 49. Negrin KA, Flach RJR, DiStefano MT, Matevossian A, Friedline RH, Jung D. **IL-1 signaling in obesity-induced hepatic lipogenesis and steatosis**. *PLoS One* (2014) **9** e107265. DOI: 10.1371/journal.pone.0107265 50. Mirea AM, Tack CJ, Chavakis T, Joosten LAB, Toonen EJM. **IL-1 family cytokine pathways underlying NAFLD: towards new treatment strategies**. *Trends Mol Med* (2018) **24** 458-471. DOI: 10.1016/j.molmed.2018.03.005 51. Bexkens ML, Mebius MM, Houweling M, Brouwers JF, Tielens AGM, van Hellemond JJ. *Int J Parasitol* (2019) **49** 647-656. DOI: 10.1016/j.ijpara.2019.03.005 52. Huang SC, Freitas TC, Amiel E, Everts B, Pearce EL, Lok JB. **Fatty acid oxidation is essential for egg production by the parasitic flatworm**. *PLoS Pathog* (2012) **8** e1002996. DOI: 10.1371/journal.ppat.1002996
--- title: Application of Indian Diabetic Risk Score (IDRS) and Community Based Assessment Checklist (CBAC) as Metabolic Syndrome prediction tools authors: - Manoj Kumar Gupta - Gitashree Dutta - Sridevi G. - Pankaja Raghav - Akhil Dhanesh Goel - Pankaj Bhardwaj - Suman Saurabh - Srikanth S. - Naveen K. H. - Prasanna T. - Neeti Rustagi - Prem Prakash Sharma journal: PLOS ONE year: 2023 pmcid: PMC10042346 doi: 10.1371/journal.pone.0283263 license: CC BY 4.0 --- # Application of Indian Diabetic Risk Score (IDRS) and Community Based Assessment Checklist (CBAC) as Metabolic Syndrome prediction tools ## Abstract ### Background Indian Diabetic Risk Score (IDRS) and Community Based Assessment Checklist (CBAC) are easy, inexpensive, and non-invasive tools that can be used to screen people for Metabolic Syndrome (Met S). The study aimed to explore the prediction abilities of IDRS and CBAC tools for Met S. ### Methods All the people of age ≥30 years attending the selected rural health centers were screened for Met S. We used the International Diabetes Federation (IDF) criteria to diagnose the Met S. ROC curves were plotted by taking Met S as dependent variables, and IDRS and CBAC scores as independent/prediction variables. Sensitivity (SN), specificity (SP), Positive and Negative Predictive Value (PPV and NPV), Likelihood Ratio for positive and negative tests (LR+ and LR-), Accuracy, and Youden’s index were calculated for different IDRS and CBAC scores cut-offs. Data were analyzed using SPSS v.23 and MedCalc v.20.111. ### Results A total of 942 participants underwent the screening process. Out of them, 59 ($6.4\%$, $95\%$ CI: 4.90–8.12) were found to have Met S. Area Under the Curve (AUC) for IDRS in predicting Met S was 0.73 ($95\%$CI: 0.67–0.79), with $76.3\%$ ($64.0\%$-$85.3\%$) sensitivity and $54.6\%$ ($51.2\%$-$57.8\%$) specificity at the cut-off of ≥60. For the CBAC score, AUC was 0.73 ($95\%$CI: 0.66–0.79), with $84.7\%$ ($73.5\%$-$91.7\%$) sensitivity and $48.8\%$ ($45.5\%$-$52.1\%$) specificity at the cut-off of ≥4 (Youden’s Index, 2.1). The AUCs of both parameters (IDRS and CBAC scores) were statistically significant. There was no significant difference ($$p \leq 0.833$$) in the AUCs of IDRS and CBAC [Difference between AUC = 0.00571]. ### Conclusion The current study provides scientific evidence that both IDRS and CBAC have almost $73\%$ prediction ability for Met S. Though CBAC holds relatively greater sensitivity ($84.7\%$) than IDRS ($76.3\%$), the difference in prediction abilities is not statistically significant. The prediction abilities of IDRS and CBAC found in this study are inadequate to qualify as Met S screening tools. ## Introduction An accumulation of clinical, physiological, biochemical, and metabolic factors which increases the risk of cardiovascular disease, type 2 diabetes mellitus (DM), and all-cause mortality is known as Metabolic Syndrome (Met S) [1]. Rapid urbanization, globalization, and the adoption of unhealthy lifestyles, such as unhealthy dietary habits and lack of physical activity, are vital factors in the development of major Non-Communicable Diseases (NCDs), which can manifest as hypertension, diabetes, hyperlipidemia, and obesity. These metabolic risk factors can be modified and controlled if detected early [2, 3]. Due to the unique "atherogenic dyslipidemic profile" and "South Asian phenotype," Indians have a significant chance of developing Met S [4, 5]. The incidence of Met S often parallels the incidences of obesity and type 2 diabetes [11]. Met S also increases the risk of type 2 diabetes and cardiovascular illnesses, including stroke and myocardial infarction, by five and twofold in 5 to 10 years [12, 13]. The Global Burden of Disease (GBD) study revealed the presence of epidemiological transition in India, where $62.7\%$ of the total mortality was attributed to NCDs [6]. Though there are differences in diagnostic criteria of Met S, and the published evidence also varies with respect to the age of study participants and methodology, a significant epidemic of Met S is emerging in the Asia-Pacific region [7]. According to a systematic review of Indian studies, the overall pooled prevalence of Met S among the adult population was $30\%$ ($95\%$CI: $28\%$-$33\%$) [8]. The secondary data analysis of the fourth round of the National Family Health Survey (NFHS-4) presented that the prevalence of Met S was higher among women than men in India. The data also revealed that the chance of developing Met S increases with age [9]. In a country like India, the western world’s influence on dietary patterns, which replace fiber-rich foods with refined carbohydrates and saturated fats, has a significant impact on the expanding obesity epidemic which in turn increases the burden of Met S. A significant association between anthropometric risk factors and Met S among Indian adults has been documented by a systematic review of observational studies which highlighted that people with overweight (pooled OR, 5.47; $95\%$ CI, 3.70–8.09) or obesity (pooled OR, 5.00; $95\%$ CI, 3.61–6.93) had higher odds of having Met S than those of normal or low body weight [10]. Due to the rising burden of obesity, Met S has also emerged as a public health issue among children and adolescents, especially in low and middle-income countries [13]. The Met S is driving the twin global epidemics of type 2 diabetes and CVD. There is an enormous moral, medical, and economic responsibility to identify persons with Met S early so that lifestyle modifications and treatment may avoid the development of diabetes and cardiovascular disease. To prevent and control major NCDs, the Government of India started the National Programme for Prevention & Control of Cancer, Diabetes, Cardiovascular Diseases & Stroke (NPCDCS) in 2010. Grass-root workers use a Community-Based Assessment Checklist (CBAC) under this program. This checklist is based on two non-modifiable (age and family history of NCDs) and four modifiable (waist circumference, smoking, alcohol, and physical inactivity) known risk factors of NCDs. This checklist helps in screening high-risk individuals for NCDs [11]. The Madras Diabetes Research Foundation (MDRF) in Chennai has developed the Indian Diabetes Risk Score (IDRS), a simple and cost-effective diabetes screening tool based on two modifiable (waist circumference and physical inactivity) and two non-modifiable (age and family history of diabetes) risk factors [12, 13]. There are no screening standards or procedures in place in India for Met S. These (IDRS and CBAC) easy and inexpensive tools, which have already been scientifically validated and are widely used to screen patients for NCDs, can also be used to screen people for Met S in the country. Before this, it is necessary to assess those tools’ prediction abilities (sensitivity and specificity). So, the current study was conducted to screen persons for Met S who visited primary health centers and to explore the prediction abilities of IDRS and CBAC tools for Met S. ## Material and methods This facility-based cross-sectional study was carried out from January to December 2019 in the three Rural Health Training Centres (RHTCs) of the Department of Community Medicine and Family Medicine at a tertiary care hospital in Jodhpur, India. Out of these three centers, one is Community Health Centre (CHC), and the other two are Primary Health Centres (PHCs). The three health centers’ catchment areas serve approximately one lakh sixty thousand population. All the patients of age ≥ 30 years attending these health centers for various ailments were included in the study. Pregnant females and patients already diagnosed with cholesterol gallstones, asthma, Polycystic ovary syndrome (PCOS), sleep apnoea, autoimmune disorders, and cancer of colorectal (colon and rectum), gastric, oesophageal, hepatobiliary (liver and gallbladder), pancreas, lung, bladder, thyroid, renal, leukaemia, malignant melanoma, multiple myeloma, and non-*Hodgkin lymphoma* were excluded from the study. The detailed methodology of the study is mentioned in another published article [14]. For operational purposes, the subjective assessments about addictions and physical activity were made based on IDRS and CBAC forms. As per the CBAC form, a drinker is a person who consumes alcohol daily, irrespective of quantity. A smoker was categorized as one who used to smoke in the past/sometimes now and another who smokes on a daily basis. According to CBAC, physical activity was defined as any dedicated physical activity for a minimum of 150 minutes in a week. While in IDRS, regular exercise and strenuous work were considered separately. The presence of a family history (any one of the parents or siblings) of high blood pressure, diabetes, and heart disease was considered a positive family history for NCDs, as per CBAC. While in IDRS, different scores were assigned to having a positive family history of only diabetes in either parent or both parents. We used the International Diabetes Federation (IDF) criteria to diagnose Met S. Central obesity (defined by the waist circumference with ≥90 cm for males and ≥80 cm for females) was the mandatory criterion. If BMI was >30kg/m2, central obesity was assumed irrespective of the waist circumference. The other two criteria used were Blood Pressure (systolic BP ≥ 130 or diastolic BP ≥ 85 mm Hg) and Raised fasting plasma glucose (FPG ≥ 100 mg/dL) [15]. ROC curves were plotted using Met S as dependent variables and IDRS and CBAC scores as independent/prediction variables. Sensitivity (SN), specificity (SP), Positive Predictive Value (PPV), Negative Predictive Value (NPV), Likelihood Ratio for positive test (LR+), Likelihood Ratio for negative test (LR-)Accuracy, and Youden’s index were calculated for different cut-offs of IDRS and CBAC scores. The study was approved by the Institutional Ethics Committee of All India Institute of Medical Sciences, Jodhpur (Letter No. AIIMS/IEC/$\frac{2018}{429}$ dated 19-03-2018). Informed written consent was obtained from all the study participants. Individuals diagnosed with diabetes, hypertension, and Met S were started on treatment and referred to tertiary care hospitals to screen for complications. These patients were given comprehensive counseling about their disease, its complication, and the necessary lifestyle and dietary modification required. Pre-diabetics and pre-hypertensive were explained their risk of developing diabetes or hypertension and were recommended frequent follow-ups. ## Results Out of the total 984 participants who were eligible for the screening, 942 consented to undergo the screening process (non-consent rate; $4.2\%$). The mean age of the participants was 52.3 (SD: 13.5) years and ranged from 30 to 94 years. There were 469 ($49.8\%$) males and 473 ($50.2\%$) females. 629 ($66.8\%$) participants performed regular physical activity. History of smoking and opium consumption was reported by $34.0\%$ and $11.3\%$ of participants, respectively (Table 1). **Table 1** | Variable | n (%) | | --- | --- | | Age (years) (Mean: 52.3±13.5) | Age (years) (Mean: 52.3±13.5) | | ≤40 | 266 (28.2) | | 41–50 | 205 (21.8) | | 51–60 | 196 (20.8) | | >60 | 275 (29.2) | | Gender | Gender | | Male | 469 (49.8) | | Female | 473 (50.2) | | Physical activity | Physical activity | | Regular | 629 (66.8) | | No/Minimal physical activity | 313 (33.2) | | Addiction | Addiction | | Smoking | 320 (34.0) | | Alcohol | 27 (2.9) | | Opium | 106 (11.3) | Of these 942 participants, 514 ($54.6\%$) were found with central obesity as per the criteria mentioned above. A total of 223 ($23.7\%$) were identified as screen positives (RPG ≥140) for diabetes and were invited to undergo FPG. Out of them, 200 (response rate $89.6\%$) reported undergoing FPG. Despite taking measures to counsel the participants to come for the follow-up to screen for the presence of diabetes, 23 participants did not respond. After excluding non-responders, the proportion of participants with FPG ≥100 mg/dl was $18.5\%$ [170]. All the participants [942] were screened for Hypertension, and $42.3\%$ were found with either systolic BP ≥ 130 or diastolic BP ≥ 85 mm Hg. A total of 59 participants ($6.4\%$, $95\%$ CI: 4.90–8.12) were found to have Met S (Fig 1). **Fig 1:** *Flow chart depicting screening of the study participants for Metabolic Syndrome.* Cut-off points of IDRS and CBAC scores were estimated using ROC curves for Met S (Fig 2). **Fig 2:** *ROC curves for depicting prediction ability of Met S with IDRS and CBAC score for Metabolic Syndrome (2a: ROC curves for Males, 2b: ROC curves for Females, 2c: ROC curves for total participants).* On applying univariate analysis, it was observed that age ($$p \leq 0.120$$) and sex ($$p \leq 0.866$$) were not significantly associated with Met S. The Area Under the Curve (AUC) and level of significance for ROC curves for male, female and total are depicted in Table 2. AUC of both the parameters (IDRS and CBAC scores) was found to be statistically significant. **Table 2** | Variables | Metabolic Syndrome | Metabolic Syndrome.1 | Metabolic Syndrome.2 | Metabolic Syndrome.3 | Metabolic Syndrome.4 | Metabolic Syndrome.5 | | --- | --- | --- | --- | --- | --- | --- | | Variables | Male | Male | Female | Female | Total | Total | | Variables | AUC (95% CI) | Sig. | AUC (95% CI) | Sig. | AUC (95% CI) | Sig. | | IDRS | 0.72 (0.64–0.81) | <0.001 | 0.74 (0.66–0.83) | <0.001 | 0.73 (0.67–0.79) | <0.001 | | CBAC | 0.76 (0.68–0.84) | <0.001 | 0.69 (0.60–0.78) | 0.001 | 0.73 (0.66–0.79) | <0.001 | On further analysis, it was noted that there was no significant difference ($$p \leq 0.833$$) in the AUCs of IDRS and CBAC [Difference between AUC = 0.00571] (Table 3). **Table 3** | Variables | Difference b/w AUCs | SE* of difference | p value | | --- | --- | --- | --- | | IDRS (Total) v/s CBAC (Total) | 0.005 | 0.0541 | 0.833 | | IDRS (Male) v/s CBAC (Male) | 0.039 | 0.0749 | 0.602 | | IDRS (Female) v/s CBAC (Female) | 0.053 | 0.0776 | 0.494 | | IDRS (Male) v/s IDRS (Female) | 0.022 | 0.0762 | 0.772 | | CBAC (Male) v/s CBAC (Female) | 0.071 | 0.0764 | 0.359 | Table 4 provides the sensitivity, specificity, PPV, NPV, LR+, LR-, and accuracy of different cut-offs for IDRS and CBAC scores for predicting Met S. An IDRS value ≥60 and CBAC value of ≥4 had the optimum sensitivity ($76.3\%$ and $53.3\%$, respectively) and specificity ($84.7\%$ and $48.8\%$, respectively) for determining Met S. **Table 4** | Score | SN, %(95%CI) | SP, %(95%CI) | PPV (%) | NPV (%) | LR+ (95%CI) | LR- (95%CI) | Accuracy (%) | Youden’s index | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | IDRS Score | | | | | | | | | | ≥ 10 | 100 (93.9–100) | 1.4 (0.8–2.4) | 6.3 | 100.0 | 1.01 (1.01–1.02) | 0.00 | 7.5 | 1.4 | | ≥ 20 | 100 (93.9–100) | 2.2 (1.4–3.3) | 6.4 | 100.0 | 1.02 (1.01–1.03) | 0.00 | 8.3 | 2.2 | | ≥ 30 | 100 (93.9–100) | 6.1 (4.7–7.9) | 6.6 | 100.0 | 1.07 (1.05–1.08) | 0.00 | 12.0 | 6.1 | | ≥ 40 | 98.3 (91.0–99.7) | 17.4 (15.1–20.1) | 7.4 | 99.4 | 1.19 (1.14–1.25) | 0.10 (0.01–0.68) | 22.5 | 15.7 | | ≥ 50 | 94.9 (86.1–98.3) | 31.7 (28.7–34.8) | 8.5 | 98.9 | 1.39 (1.29–1.50) | 0.16 (0.05–0.48) | 35.7 | 26.6 | | ≥ 60 | 76.3 (64.0–85.3) | 54.6 (51.2–57.8) | 10.1 | 97.2 | 1.68 (1.43–1.97) | 0.43 (0.27–0.69) | 55.9 | 30.9 | | ≥ 70 | 59.3 (46.6–70.9) | 77.9 (75.9–81.4) | 15.2 | 96.6 | 2.69 (2.10–3.43) | 0.52 (0.38–0.71) | 76.8 | 37.2 | | ≥ 80 | 22.0 (13.3–34.1) | 92.2 (90.2–93.8) | 15.9 | 94.7 | 2.82 (1.66–4.79) | 0.85 (0.74–0.97) | 87.8 | 14.2 | | ≥ 90 | 0.0 (0.0–6.1) | 99.5 (98.8–99.8) | 0.0 | 93.7 | 0.00 | 1.00 (1.00–1.00) | 93.3 | 0.0 | | CBAC Score | | | | | | | | | | ≥ 1 | 100 (93.9–100) | 2.6 (1.7–3.9) | 6.4 | 100.0 | 1.03 (1.02–1.04) | 0.00 | 8.7 | 2.6 | | ≥ 2 | 100 (93.9–100) | 9.5 (7.7–11.6) | 6.9 | 100.0 | 1.11 (1.08–1.13) | 0.00 | 15.2 | 9.5 | | ≥ 3 | 93.2 (83.8–97.3) | 25.9 (23.1–28.9) | 7.8 | 98.3 | 1.26 (1.16–1.36) | 0.26 (0.10–0.68) | 30.1 | 1.2 | | ≥ 4 | 84.7 (73.5–91.7) | 48.8 (45.5–52.1) | 10.0 | 98.0 | 1.66 (1.46–1.88) | 0.31 (0.17–0.57) | 51.1 | 2.1 | | ≥ 5 | 59.3 (46.6–70.9) | 72.7 (69.7–75.5) | 12.7 | 96.4 | 2.17 (1.71–2.76) | 0.56 (0.41–0.76) | 71.9 | 2.0 | | ≥ 6 | 32.2 (21.7–44.9) | 89.5 (87.3–91.3) | 17.0 | 95.2 | 3.06 (2.01–4.64) | 0.76 (0.63–0.90) | 85.9 | 1.4 | | ≥ 7 | 15.3 (8.2–26.5) | 96.7 (95.3–97.7) | 23.7 | 94.5 | 4.64 (2.31–9.35) | 0.88 (0.79–0.98) | 91.6 | 12.0 | | ≥ 8 | 3.4 (0.9–11.5) | 99.1 (98.2–99.5) | 20.0 | 93.9 | 3.74 (0.81–17.23) | 0.97 (0.93–1.02) | 93.1 | 2.5 | | ≥ 9 | 0.0 (0.0–6.1) | 99.9 (99.3–99.9) | 0.0 | 93.7 | 0.00 (0.00–0.00) | 1.00 (1.00–1.00) | 93.6 | 0.0 | ## Discussion In this study, facility-based screening of rural people of Western Rajasthan was done to estimate the proportion of newly diagnosed cases of Met S. The prediction capacity of the existing non-invasive IDRS and CBAC scores was calibrated to screen for Met S, and the appropriate cut-offs were determined. Met S was diagnosed among $6.4\%$ of participants in the present study. This is relatively lower than the prevalence ($15.6\%$) reported by a large survey in India among the rural adult population [16]. A study from Central India has reported a $5\%$ prevalence of Met S in the rural adult population [17]. Another study from South India has reported $39.7\%$ ($95\%$ CI: 35.3–44.1) prevalence among the rural adult population [18]. According to the secondary data analysis of NFHS-4, the prevalence of Met S was $1.5\%$ among women and $1.1\%$ among men in India. While in Rajasthan, it was $0.8\%$ (0.536–0.966) among women and 1.0 (0.859–1.052) among men [9]. This varied prevalence across the different geographical regions of the country may be because of using different criteria for diagnosing Met S or having different age cut-offs of the study populations. Though IDRS is used to screen for diabetes, we tried to explore its prediction ability to diagnose Met S in this study. In our study, AUC for IDRS score in predicting Met S was 0.73 ($95\%$CI: 0.67–0.79), with $76.3\%$ sensitivity and $54.6\%$ specificity at the cut-off of ≥60, and with $59.3\%$ sensitivity and $77.9\%$ specificity at the cut-off of ≥70. When the major part of the disease iceberg is hidden in the community, and untreated disease could end up with serious complications, a score that maximizes true positives is preferable. So, the screening cut-off is often set at a lower level, which increases the sensitivity value. Thus, IDRS with a cut-off of ≥60 is more appropriate to predict Met S. The present study’s findings are well supported by a large study conducted in the Southern part of India [19]. The study by Mohan V. et al. [ 2013] also concluded that IDRS could help to identify Met S by observing the increased prevalence of Met S among those with high IDRS Scores [20]. But, the prediction ability of IDRS found in this study is inadequate to qualify as a Met S screening tool. The results of this study also show the prediction ability of the CBAC form, which is extensively used in India by front-line healthcare workers to screen for NCDs. Traditionally, in the screening using the CBAC checklist, a score above 4 indicates that the person may be at risk for NCDs. In our study, AUC for CBAC score in predicting Met S was 0.73 ($95\%$CI: 0.66–0.79), with $84.7\%$ sensitivity and $48.8\%$ specificity at the cut-off of ≥4 (Youden’s Index, 2.1). There is a dearth of scientific evidence exploring the prediction capacity of CBAC form for Met S. Based on the findings of this study; the CBAC tool also does not seem promising in identifying people at high risk of developing Met S in primary care settings. The most accurate metric for gauging the effectiveness of a risk score is the total area under the ROC curve. The risk score performs better overall in terms of adequately predicting persons who will get a disease when the area under the curve is higher [21]. The present study depicts that the AUC of IDRS and CBAC scores for predicting Met S were almost similar (0.731 and 0.726, respectively). Based on the detection of true positives (sensitivity) also, IDRS at the cut-off score of ≥60 has slightly lower values ($76.3\%$) than the CBAC checklist at the cut-off of ≥4 ($84.7\%$). There was no significant difference between the AUCs of IDRS and CBAC scores. This slight difference in the AUCs between IDRS and CBAC can be attributed to the appropriate giving consideration to tobacco and alcohol use in the CBAC checklist compared to the IDRS, which gives a substantially higher priority to factors including age, family history, physical activity, and central obesity. This study was non-funded, so the HDL and Triglyceride levels of the study participants could not be assessed. This may underestimate the prevalence of Met S in the study. This is one of the limitations of this study. Besides that, the authors did not collect data related to the diet of study participants, so the effect of diet could not be adjusted for the prediction of Met S using IDRS and CBAC. ## Conclusion and recommendations The current study provides scientific evidence that both IDRS and CBAC have almost $73\%$ prediction ability for Met S. Though CBAC holds relatively greater sensitivity ($84.7\%$) than IDRS ($76.3\%$), the difference in prediction abilities is not statistically significant. The prediction abilities of IDRS and CBAC found in this study are inadequate to qualify as Met S screening tools. Thus, both tools do not seem promising in identifying people at high risk of developing Met S in primary care settings. As these tools are the best available non-invasive tools for screening NCDs and are widely used in India, their prediction abilities need to be explored further with a larger sample size and multicentric basis by considering the limitations mentioned in this study. ## References 1. Alberti KGMM, Eckel RH, Grundy SM, Zimmet PZ, Cleeman JI, Donato KA. **Harmonizing the metabolic syndrome: a joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart**. *Circulation* (2009.0) **120** 1640-1645. DOI: 10.1161/CIRCULATIONAHA.109.192644 2. Mehata S, Shrestha N, Mehta RK, Bista B, Pandey AR, Mishra SR. **Prevalence of the Metabolic Syndrome and its determinants among Nepalese adults: Findings from a nationally representative cross-sectional study**. *Sci Rep.* (2018.0) **8** 14995. DOI: 10.1038/s41598-018-33177-5 3. Saklayen MG. **The Global Epidemic of the Metabolic Syndrome.**. *Curr Hypertens Rep* (2018.0) **20** 12. DOI: 10.1007/s11906-018-0812-z 4. Misra A, Shrivastava U. **Obesity and dyslipidemia in South Asians.**. *Nutrients* (2013.0) **5** 2708-2733. DOI: 10.3390/nu5072708 5. Patel SA, Shivashankar R, Ali MK, Anjana RM, Deepa M, Kapoor D. **Is the “South Asian Phenotype” Unique to South Asians?: Comparing Cardiometabolic Risk Factors in the CARRS and NHANES Studies.**. *Glob Heart* (2016.0) **11** 89-96.e3. DOI: 10.1016/j.gheart.2015.12.010 6. 6Global Burden of Disease (GBD 2019). In: Institute for Health Metrics and Evaluation [Internet]. 17 Mar 2014 [cited 12 May 2022]. Available: https://www.healthdata.org/gbd/2019 7. Ranasinghe P, Mathangasinghe Y, Jayawardena R, Hills AP, Misra A. **Prevalence and trends of metabolic syndrome among adults in the asia-pacific region: a systematic review.**. *BMC Public Health* (2017.0) **17** 101. DOI: 10.1186/s12889-017-4041-1 8. Krishnamoorthy Y, Rajaa S, Murali S, Rehman T, Sahoo J, Kar SS. **Prevalence of metabolic syndrome among adult population in India: A systematic review and meta-analysis.**. *PloS One.* (2020.0) **15** e0240971. DOI: 10.1371/journal.pone.0240971 9. Meher T, Sahoo H. **The epidemiological profile of metabolic syndrome in Indian population: A comparative study between men and women.**. *Clin Epidemiol Glob Health* (2020.0) **8** 1047-1052. DOI: 10.1016/j.cegh.2020.03.018 10. Krishnamoorthy Y.. **Association Between Anthropometric Risk Factors and Metabolic Syndrome Among Adults in India: A Systematic Review and Meta-Analysis of Observational Studies.**. *Prev Chronic Dis.* (2022.0) **19**. DOI: 10.5888/pcd19.210231 11. 11Government of India. Community Based Assessment Checklist (CBAC) Form for Early Detection of NCDs and Tuberculosis (TB). 2018. Available: http://nhsrcindia.org/sites/default/files/Revised%20Community%20Based%20Assessment%20Checklist%2027%20June%202018.pdf 12. Nagarathna R, Tyagi R, Battu P, Singh A, Anand A, Nagendra HR. **Assessment of risk of diabetes by using Indian Diabetic risk score (IDRS) in Indian population.**. *Diabetes Res Clin Pract* (2020.0) **162** 108088. DOI: 10.1016/j.diabres.2020.108088 13. Adhikari P, Pathak R, Kotian S. **Validation of the MDRF-Indian Diabetes Risk Score (IDRS) in another south Indian population through the Boloor Diabetes Study (BDS).**. *J Assoc Physicians India.* (2010.0) **58** 434-436. PMID: 21121209 14. Gupta MK, Raghav P, Tanvir T, Gautam V, Mehto A, Choudhary Y. **Recalibrating the Non-Communicable Diseases risk prediction tools for the rural population of Western India.**. *BMC Public Health.* (2022.0) **22** 376. DOI: 10.1186/s12889-022-12783-z 15. 15IDF Consensus Worldwide Definition of the Metabolic Syndrome. 2006 [cited 8 Apr 2021]. Available: https://www.idf.org/e-library/consensus-statements/60-idfconsensus-worldwide-definitionof-the-metabolic-syndrome.html 16. Barik A, Das K, Chowdhury A, Rai RK. **Metabolic syndrome among rural Indian adults.**. *Clin Nutr ESPEN* (2018.0) **23** 129-135. DOI: 10.1016/j.clnesp.2017.11.002 17. Kamble P, Deshmukh PR, Garg N. **Metabolic syndrome in adult population of rural Wardha, central India**. *Indian J Med Res* (2010.0) **132** 701-705. PMID: 21245618 18. Venugopal V, Dongre AR, Saravanan S. **Prevalence and Determinants of Metabolic Syndrome among the Rural Adult Population of Puducherry.**. *Indian J Community Med Off Publ Indian Assoc Prev Soc Med* (2019.0) **44** 21-25. DOI: 10.4103/ijcm.IJCM_132_18 19. Mohan V, Sandeep S, Deepa M, Gokulakrishnan K, Datta M, Deepa R. **A diabetes risk score helps identify metabolic syndrome and cardiovascular risk in Indians—the Chennai Urban Rural Epidemiology Study (CURES-38).**. *Diabetes Obes Metab.* (2007.0) **9** 337-343. DOI: 10.1111/j.1463-1326.2006.00612.x 20. Mohan V, Anbalagan VP. **Expanding role of the Madras Diabetes Research Foundation—Indian Diabetes Risk Score in clinical practice**. *Indian J Endocrinol Metab* (2013.0) **17** 31-36. DOI: 10.4103/2230-8210.107825 21. Gupta A, Sathiyanarayanan S, Aravindakshan R, Kakkar R. **A Pilot Retrospective Study Validating Noncommunicable Disease Risk Assessment Score (AMNRAS).**. *Int J Prev Med.* (2021.0) **12** 134. DOI: 10.4103/ijpvm.IJPVM_206_19
--- title: 'Association of severe mental illness and septic shock case fatality rate in patients admitted to the intensive care unit: A national population-based cohort study' authors: - Ines Lakbar - Marc Leone - Vanessa Pauly - Veronica Orleans - Kossi Josue Srougbo - Sambou Diao - Pierre-Michel Llorca - Marco Solmi - Christoph U. Correll - Sara Fernandes - Jean-Louis Vincent - Laurent Boyer - Guillaume Fond journal: PLOS Medicine year: 2023 pmcid: PMC10042353 doi: 10.1371/journal.pmed.1004202 license: CC BY 4.0 --- # Association of severe mental illness and septic shock case fatality rate in patients admitted to the intensive care unit: A national population-based cohort study ## Abstract ### Background Patients with severe mental illness (SMI) (i.e., schizophrenia, bipolar disorder, or major depressive disorder) have been reported to have excess mortality rates from infection compared to patients without SMI, but whether SMI is associated with higher or lower case fatality rates (CFRs) among infected patients remains unclear. The primary objective was to compare the 90-day CFR in septic shock patients with and without SMI admitted to the intensive care unit (ICU), after adjusting for social disadvantage and physical health comorbidity. ### Methods and findings We conducted a nationwide, population-based cohort study of all adult patients with septic shock admitted to the ICU in France between January 1, 2014, and December 31, 2018, using the French national hospital database. We matched (within hospitals) in a ratio of 1:up to 4 patients with and without SMI (matched-controls) for age (5 years range), sex, degree of social deprivation, and year of hospitalization. Cox regression models were conducted with adjustment for smoking, alcohol and other substance addiction, overweight or obesity, Charlson comorbidity index, presence of trauma, surgical intervention, Simplified Acute Physiology Score II score, organ failures, source of hospital admission (home, transfer from other hospital ward), and the length of time between hospital admission and ICU admission. The primary outcome was 90-day CFR. Secondary outcomes were 30- and 365-day CFRs, and clinical profiles of patients. A total of 187,587 adult patients with septic shock admitted to the ICU were identified, including 3,812 with schizophrenia, 2,258 with bipolar disorder, and 5,246 with major depressive disorder. Compared to matched controls, the 90-day CFR was significantly lower in patients with schizophrenia (1,$\frac{052}{3}$,269 = $32.2\%$ versus 5,$\frac{000}{10}$,894 = $45.5\%$; adjusted hazard ratio (aHR) = 0.70, $95\%$ confidence interval (CI) 0.65,0.75, $p \leq 0.001$), bipolar disorder ($\frac{632}{1}$,923 = $32.9\%$ versus 2,$\frac{854}{6}$,303 = $45.3\%$; aHR = 0.70, $95\%$ CI = 0.63,0.76, $p \leq 0.001$), and major depressive disorder (1,$\frac{834}{4}$,432 = $41.4\%$ versus 6,$\frac{798}{14}$,452 = $47.1\%$; aHR = 0.85, $95\%$ CI = 0.81,0.90, $p \leq 0.001$). Study limitations include inability to capture deaths occurring outside hospital, lack of data on processes of care, and problems associated with missing data and miscoding in medico-administrative databases. ### Conclusions Our findings suggest that, after adjusting for social disadvantage and physical health comorbidity, there are improved septic shock outcome in patients with SMI compared to patients without. This finding may be the result of different immunological profiles and exposures to psychotropic medications, which should be further explored. ## Author summary ## Introduction Data have consistently indicated that individuals with severe mental illness (SMI) (i.e., schizophrenia, bipolar disorder, or major depressive disorder) are at higher risk of premature mortality than the general population [1,2]. This is mainly attributed to higher rates of physical disease, social disadvantage, unhealthy lifestyle behaviors, and inadequate healthcare in patients with SMI [3–6]. Among somatic diseases, infections are disproportionately more frequent in patients with SMI than in the general population, representing a potentially avoidable contributor to early death [2,7,8]. In a meta-analysis, patients with SMI were reported to have higher mortality rates from infection than the general population [2]. Whether SMI is associated with higher or lower infection-associated case fatality (i.e., the proportion of persons with infection who die from that infection [9]) compared with the general population is unclear. Sepsis (i.e., infection-associated organ dysfunction) is one of the leading causes of death around the world [10], with in-hospital case fatality rates (CFRs) as high as $40\%$ in septic shock, the most severe form of sepsis [11]. Few studies have reported data on sepsis-associated CFR in patients with SMI, showing conflicting results: 2 studies reported higher CFR [12,13] and 4 studies reported lower CFR [14–17]. These latter 4 studies performed additional adjustments but omitted important confounding factors, such as overweight or obesity status, severity of sepsis, and type of hospital. Presence of overweight/obesity may represent a protective factor [18] and is more prevalent in patients with SMI than in the general population [19]. Because of the bias associated with variability and subjectivity of sepsis diagnosis [20–22], there is a need to adjust for severity of illness using an appropriate scoring system [23]. Finally, patients with SMI are more often hospitalized at university hospitals [24–26], which are characterized by higher sepsis case volumes known to be associated with better survival [27], than in smaller hospitals [24,25]. Patient matching within a hospital has been advocated to control best for facility confounders [28]. To the best of our knowledge, to date, no study has determined whether SMI is associated with excess CFR in patients with septic shock after accounting for the most relevant confounding variables. To address this issue, we conducted a nationwide, population-based cohort study using the French national hospital database. The primary objective was to compare 90-day CFRs in septic shock patients with and without SMI admitted to the intensive care unit (ICU), after adjusting for social disadvantage and physical health comorbidity. Secondary objectives were to compare 30- and 365-day CFRs and clinical profiles in septic shock patients with and without SMI. We hypothesized that patients with SMI would have a higher septic shock CFR than patients without SMI. ## Study design, sources, and population In this nationwide, population-based cohort study, we used data from the Programme de Médicalisation des Systèmes d’Information (PMSI database), the French national hospital database in which administrative and medical data are systematically collected for acute (PMSI-MCO) and psychiatric (PMSI-PSY) hospitalizations. The PMSI database is based on diagnosis-related groups (DRGs), with all diagnoses coded according to the 10th revision of the International Classification of Diseases (ICD-10) and using procedural codes from the Classification Commune des Actes Médicaux (CCAM). The PMSI database is used to determine financial resource use and is frequently and carefully verified by its producer as well as the paying party, with possible financial and legal consequences. Data from the PMSI database are anonymized and can be reused for research purposes. A unique anonymous identifier enables different inpatient stays of individual patients to be linked. The study was submitted to the French National Data Protection Commission (N° 2203797) for ethical approval. This manuscript follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [29] (S1 STROBE Checklist). We included all hospital admissions between January 1, 2014, and December 31, 2018, using the following criteria: aged 18 years or older, admitted to the ICU, had a diagnosis of septic shock (ICD-10 code = R572 or a combination of codes corresponding to a severe infection associated with the use of vasopressors). We limited inclusion to patients with an ICU length of stay of at least 48 hours, unless the patient died within 48 hours, in order to avoid overestimating diagnoses of septic shock. Although the coding of septic shock has been strictly regulated since the DRG system was introduced in France, we cannot exclude overcoding due to the high tariff associated with the codes, especially for short stays in the ICU. Indeed, the length of stay for patients with septic shock is about 7 days (IQR 3 to 14 days) [30]. We therefore considered the first quartile (< = 2 days) to be a credible threshold below which the probability of having septic shock was low (excluding patients who died within these 48 hours). ## Outcomes The primary outcome was 90-day CFR (i.e., deaths per 100 cases of septic shock, percentage). Secondary outcomes were 30- and 365-day CFRs and the clinical profiles of patients. ## Collected data We collected the following sociodemographic data: age, sex, and degree of social deprivation (least deprived, less deprived, more deprived, most deprived according to quartiles) based on 4 socioeconomic ecological variables—the proportion who had graduated from high school, median household income, the percentage of blue-collar workers, and the unemployment rate [31]. We also collected data on comorbidities (overweight or obesity, addiction [smoking, alcohol, and other substances], Charlson Comorbidity Index (0, 1 to 2, ≥3 [32]); presence of trauma; surgical intervention; Simplified Acute Physiology Score II (SAPS II) at ICU admission; source of infection and identified pathogens; the type of organ failure (respiratory, renal, neurologic, cardiovascular, hematologic, metabolic); and use of supportive therapies (cardiopulmonary resuscitation, invasive mechanical ventilation, renal replacement therapy, transfusion). Characteristics of the stay were noted, including the source of hospital admission (i.e., where the patient came from [home, transfer from other hospital ward]), the length of time between hospital admission and ICU admission, and durations of ICU and hospital stay; characteristics of the hospital were also recorded (academic, general public, and private). ## Exposures For the purpose of this study, we defined 6 groups: 3 groups with SMI, which included patients with a diagnosis of schizophrenia (ICD-10 codes F20*, F22*, or F25*), bipolar disorder (ICD-10 codes F30*, F31*), or major depressive disorder (ICD-10 codes F33*), and 3 matched groups without SMI (controls). The control groups were created by matching for age (5-year range), sex, degree of social deprivation, and year of hospitalization in a ratio of 1:up to 4 patients with and without SMI within a hospital (to control for confounders at a hospital level). In patients with dual diagnoses, those with codes for schizophrenia and bipolar disorder or major depressive disorder were classified in the schizophrenia group, and those with codes for bipolar disorder and major depressive disorder were classified as bipolar disorder. There was therefore no overlap across the groups. ## Statistical analysis The patients’ characteristics are presented as counts (percentages) and medians (interquartile ranges) for categorical and continuous variables, respectively. CFR was calculated at 30, 90, and 365 days using the total number of patients admitted to the ICU with septic shock as the denominator. Standardized differences were used to compare patients with and without SMI using weights to normalize the distribution of patients. An absolute standardized difference (SD) of ⩽0.20 was chosen to indicate a negligible difference in the mean or prevalence of a variable between groups [33]. The SD helps to understand the magnitude of the differences found, in addition to statistical significance, which examines whether the findings are likely to be due to chance [34]. To study the association between each SMI and outcome, the Kaplan–Meier method and the log-rank statistic were used to estimate and compare the cumulative death rates. Hazard ratios (HRs) and $95\%$ confidence intervals ($95\%$ CIs) were estimated using Cox survival models with a robust variance estimator to account for clustering within matched pairs. Two models were developed for each outcome. Model 1 included SMI only (no adjustment). Model 2 included SMI with additional covariates of smoking, alcohol, and other substance addiction (yes versus no), overweight or obesity (yes versus no), the Charlson comorbidity index (0, 1 to 2, ≥3), presence of trauma (yes versus no), surgical intervention (yes versus no), SAPS II score (modified, without age), organ failures (yes versus no for each of respiratory, renal, neurologic, cardiovascular, hematologic, metabolic, hepatic), the source of hospital admission (home, transfer from other hospital ward), and time between hospital admission and ICU admission (≤1 versus > 1 day). The covariates were selected a priori on the basis of clinical relevance or the results of bivariate outcomes analyses (SD > 0.2). Interactions with SMI were investigated, but associations were negligible. Several sensitivity analyses were performed: model S1 (model 2 with the 17 Charlson comorbidities instead of the Charlson comorbidity index), model S2 (model 2 with infected organs instead of organ failures), model S3 (model 2 with ICU supportive therapies instead of organ failures), model S4 (model 2 with the nature of isolated pathogens), and model S5 on the whole cohort (without matching process) using the same variables as in model 2 and matching variables to consider residual bias from incomplete matching of controls to the respective SMI group. The proportional-hazards assumption for the Cox models was investigated and confirmed graphically through survival functions over time. A $p \leq 0.05$ was considered significant. Data management and analyses were performed using the SAS software. Cox regression analyses were performed using the PROC PHREG in SAS. ## Results The database included a total of 187,587 patients with septic shock (flow chart, Fig 1). The main sociodemographic data of the patients are shown in Table 1. The mean age was 67.1 (±14.3) years and $63.8\%$ were men. A majority of patients (106,941 patients [$57.0\%$]) were socially deprived and most patients (167,738 patients [$89.4\%$]) were hospitalized in public hospitals. Among the 187,587 patients, 3,812 had schizophrenia ($2.0\%$), 2,258 had bipolar disorder ($1.2\%$), and 5,246 had major depressive disorder ($2.8\%$). A total of 3,269 patients with schizophrenia, 1,923 patients with bipolar disorder, and 4,432 patients with major depressive disorder were matched with 10,894, 6,303, and 14,452 controls, respectively. **Fig 1:** *Flow chart of the patients admitted to the intensive care unit (ICU) with septic shock during the study period.* TABLE_PLACEHOLDER:Table 1 ## Comparison of CFRs in septic shock patients with and without SMI Compared to matched controls, the 90-day CFR was significantly lower in patients with schizophrenia (1,$\frac{052}{3}$,269 = $32.2\%$ versus 5,$\frac{000}{10}$,894 = $45.5\%$; adjusted HR (aHR) = 0.70, $95\%$ CI 0.65,0.75, $p \leq 0.001$), bipolar disorder ($\frac{632}{1}$,923 = $32.9\%$ versus 2,$\frac{854}{6}$,303 = $45.3\%$; aHR = 0.70, $95\%$ CI = 0.63,0.76, $p \leq 0.001$), and major depressive disorder (1,$\frac{834}{4}$,432 = $41.4\%$ versus 6,$\frac{798}{14}$,452 = $47.1\%$; aHR = 0.85, $95\%$ CI = 0.81,0.90, $p \leq 0.001$) (Tables 2 and 3). The 30-day and 365-day CFRs were also significantly lower in patients with schizophrenia, bipolar disorder, and major depressive disorder than in matched controls. The sensitivity analyses reported similar findings for 30-, 90-, and 365-day CFRs (S1, S2, and S3 Figs). S4 Fig shows the survival curves in the different groups at 1 year. ## Comparison of clinical profiles in septic shock patients with and without SMI Patients with a major depressive disorder were more likely to have a tobacco (SD = 0.23) and alcohol (SD = 0.32) addiction, and patients with bipolar disorders were more likely to have an addiction to other substance than were their matched controls (SD = 0.22) (Table 4). Patients with schizophrenia and those with bipolar disorder had lower Charlson comorbidity index scores (SD = −0.27 and SD = −0.23, respectively), especially fewer malignancies (SD = −0.32 and SD = −0.26, respectively). Patients with bipolar disorder were more likely to have neurological failure than were their matched controls (SD = 0.25) (S1 Table). Differences in the site of infection or type of pathogen were negligible between SMI patients and their matched controls (S2 Table). **Table 4** | Unnamed: 0 | Patients with schizophrenia | Matched controls | SD† | p-value† | Patients with bipolar disorder | Matched controls.1 | SD‡ | p-value‡ | Patients with major depressive disorder | Matched controls.2 | SD⨎ | p-value⨎ | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | N | 3269 | 10894 | | | 1923 | 6303 | | | 4432 | 14452 | | | | Age–year | | | | | | | | | | | | | | Mean ± SD[95% CI] | 59.6 ± 13.5[59.2–60.1] | 59.9 ± 7.4[59.6–60.1] | 0.02 | 0.495 | 63.5 ± 12.3[62.9–64.1] | 63.7 ± 6.8[63.4–64.0] | −0.02 | 0.629 | 64.7 ± 7.3[64.1–64.9] | 64.5 ± 13.4[64.4–64.9] | −0.02 | 0.552 | | Distribution–n (weighted %)[95% CI] | | | | 0.918 | | | | 0.872 | | | | 0.895 | | 18–44 | 425(13.0%)[11.8–14.2]) | 1,141(12.6%)[11.5–13.8] | 0.01 | | 137(7.1%)[6.0–8.2] | 364(6.8%)[5.6–7.9] | 0.01 | | 307(6.9%)[6.2–7.7] | 778(6.6%)[5.9–7.3] | 0.01 | | | 45–64 | 1,644(50.3%)[48.8–52.0] | 5,504(50.1%)[48.4–51.8] | 0.00 | | 818(42.5%)[40.3–44.7] | 2,610(42.3%)[40.0–44.5] | 0.01 | | 1,867(42.1%)[40.7–43.6] | 5,967(41.9%)[40.4–43.3]] | 0.01 | | | 65–75 | 795(24.3%)[22.8–25.7] | 2,794(24.4)[22.8–25.8] | −0.00 | | 646(33.6%)[31.4–35.7] | 2,189(33.3%)[31.1–35.4] | 0.01 | | 1,233(27.8%)[26.5–29.1] | 4,255(28.2%)[26.5–29.1] | −0.01 | | | >75 | 405(12.4%)[11.3–13.5] | 1,455(12.9)[11.7–14.0] | −0.02 | | 322(16.7%)[15.1–18.4] | 1,140(17.7%)[16.0–19.4] | −0.02 | | 1,025(23.1%)[21.9–24.4] | 3,452(23.3%)[21.9–24.4] | −0.00 | | | Age at death–year | | | | | | | | | | | | | | Mean ± SD[95% CI] | 63.1 ± 13.1[62.3–63.9] | 62.6 ± 12.4[61.7–62.4] | 0.04 | 0.044 | 67.0 ± 12.7[66.1–67.9] | 66.1 ± 11.3[65.3–66.2] | 0.08 | 0.045 | 67.1 ± 12.7[66.5–67.7] | 67.4 ± 12.4[66.6–67.2] | −0.02 | 0.710 | | Sex (women)–n (weighted %)[95% CI] | 1,186(36.3%)[34.6–37.9] | 3,740(36.3%)[34.6–37.9] | 0.00 | 1.000 | 1,003(52.2%)[50.0–54.4] | 3,115(52.2%)[50.0–54.4] | 0.00 | 1.000 | 2,292(51.7%)[50.2–53.2] | 7,144(51.7%)[50.2–53.2] | 0.00 | 1.000 | | Social deprivation, − n (weighted %)[95% CI] | | | | 1.000 | | | | 1.000 | | | | 1.000 | | Least deprived | 1,067(32.6%) [31.0–34.2] | 3,789(32.6%) [31.0–34.2] | 0.00 | | 623(32.4%)[30.3–34.5] | 2,231(32.4%)[30.3–34.5] | 0.00 | | 1,278(28.8%) [27.5–30.2] | 4,547(28.8%) [27.5–30.2] | 0.00 | | | Less deprived | 501(15.3%)[14.1–16.6] | 1,670(15.3%)[14.1–16.6] | 0.00 | | 290(15.1%)[13.5–16.7] | 891(15.1%)[13.5–16.7] | 0.00 | | 661(14.9%)[13.9–16.0] | 2,062(14.9%)[13.9–16.0] | 0.00 | | | More deprived | 990(30.3%)[28.7–31.9] | 3,178(30.3%)[28.7–31.9] | 0.00 | | 629(32.7%)[30.6–34.8] | 1,990(32.7%)[30.6–34.8] | 0.00 | | 1,443(32.6%)[31.2–33.9] | 4,581(32.6%)[31.2–33.9] | 0.00 | | | Most deprived | 711(21.8%)[20.3–23.2] | 2,281(21.8%)[20.3–23.2] | 0.00 | | 381(19.8%) [18.0–21.6] | 1,191 (19.8%) [18.0–21.6] | 0.00 | | 1,050 (23.7%)[22.4–24.9] | 3,262(23.7%)[22.4–24.9] | 0.00 | | | Year–n (weighted %)[95% CI] | | | | 1.000 | | | | 1.000 | | | | 1.000 | | 2014 | 571(17.5%)[16.2–18.8] | 1,844(17.5%) [16.2–18.8] | 0.00 | | 358(18.6%)[16.9–20.4] | 1,157(18.6%)[16.9–20.4] | 0.00 | | 822(18.6%)[17.4–19.7] | 2,602(18.6%)[17.4–19.7] | 0.00 | | | 2015 | 659(20.2%)[18.8–21.5] | 2,184(20.2%) [18.8–21.5] | 0.00 | | 387(20.1%)[18.3–21.9] | 1,243 (20.1%)[18.3–21.9] | 0.00 | | 908(20.5%)[19.3–21.7] | 2,964(20.5%)[19.3–21.7] | 0.00 | | | 2016 | 669(20.5%)[19.1–21.8] | 2,264(20.5%)[19.1–21.8] | 0.00 | | 412(21.4%) [19.6–23.3] | 1,357 (21.4%) [19.6–23.3] | 0.00 | | 915(20.7%)[19.5–21.8] | 3,006(20.7%)[19.5–21.8] | 0.00 | | | 2017 | 678(20.7%) [19.1–22.1] | 2,321(20.7%)[19.1–22.1] | 0.00 | | 371(19.3%)[17.5–21.1] | 1,256(19.3%)[17.5–21.1] | 0.00 | | 889(20.1%)[18.9–21.2] | 2,925(20.1%)[18.9–21.2] | 0.00 | | | 2018 | 692(21.2%)[19.7–22.6] | 2,281(21.2%)[19.7–22.6] | 0.00 | | 395(20.5%)[18.7–22.3] | 1,290 (20.5%)[18.7–22.3] | 0.00 | | 898(20.3%)[19.1–21.5] | 2,955(20.3%)[19.1–21.5] | 0.00 | | | Smoking addiction–n (weighted %)[95% CI] | 766(23.4%)[22.0–24.9] | 2,133(19.6)[18.3–21.0] | 0.09 | <0.001 | 422(21.9%)[20.1–23.8] | 1,098(17.3%)[15.6–19.0]) | 0.12 | <0.001 | 1,199(27.1%) [25.7–28.4] | 2,552(17.7)[16.6–18.9] | 0.23 | <0.001 | | Alcohol addiction–n (weighted %)[95% CI] | 600(18.4%)[17.0–19.7] | 2,136(19.8%)[18.5–21.2] | −0.04 | 0.129 | 445(23.1%)[21.2–25.0] | 1,011(16.2%)[14.6–17.9] | 0.17 | <0.001 | 1,261(28.5%)[27.1–29.8] | 2,225(15.5%)[14.4–16.5] | 0.32 | <0.001 | | Other substance addiction–n (weighted %)[95% CI] | 227(6.9%)[6.1–7.8] | 311(2.9%)[2.3–3.4] | 0.19 | <0.001 | 115(6.0%)[4.9–7.0] | 110(1.7%)[1.2–2.3] | 0.22 | <0.001 | 220(5.0%)[4.2–5.6] | 234(1.6%)[1.3–2.0] | 0.19 | <0.001 | | Opioid-related Disorder | 103(3.2%)[2.6–3.7] | 155(1.6%)[1.2–2.0] | 0.10 | <0.001 | 41(2.1%)[1.5–2.8] | 54(0.9%)[0.5–1.4] | 0.10 | 0.004 | 110(2.5%)[2.0–3.0] | 133(0.6%)[0.3–0.8] | 0.12 | <0.001 | | Cannabis-related Disorder | 79(2.4%)[1.9–2.9] | 69(0.7%)[0.4–1.0] | 0.14 | <0.001 | 34(1.8%)[1.2–2.4] | 31(0.5%)[0.2–0.8] | 0.12 | <0.001 | 44(1.0%)[0.7–1.2] | 66(0.5%)[0.3–0.7] | 0.06 | 0.004 | | Cocaine-related disorder | 32(1.0%)[0.6–1.3] | 33(0.4%)[0.2–0.6] | 0.08 | 0.003 | 15(0.8%)[0.4–1.2] | 8(0.1%)[0.0–0.3] | 0.10 | 0.008 | 22(0.5%)[0.3–0.7] | 30(0.2%)[0.06–0.3] | 0.05 | 0.019 | | Other substances | 115(3.5%)[2.9–4.1] | 100(1.0%)[0.06–1.3%] | 0.17 | <0.001 | 57(3.0%)[2.2–3.7] | 42(0.6%)[0.3–1.0] | 0.18 | <0.001 | 111(2.5%)[2.0–3.0] | 81(0.6%)[0.3–0.8] | 0.16 | <0.001 | | Overweight or obese–n (weighted %)[95% CI] | 533(16.3%)[15.0–17.6] | 1,945(17.7%) [16.3–19.0] | −0.04 | 0.148 | 411(21.4%) [19.5–23.2] | 1,194(18.9%)[17.1–20.6) | 0.06 | 0.053 | 1,019(23.0%)[21.7–24.2] | 2,911(20.5%)[19.4–21.7]) | 0.06 | 0.005 | | Charlson index–n (weighted %)[95% CI] | | | | <0.001 | | | | <0.001 | | | | <0.001 | | 0 | 1,104(33.8%)[32.2–35.4] | 2,223 (21.4 [20.0–22.8]) | 0.28 | | 567(29.5%)[27.4–31.5] | 1,186(20.1%)[18.3–21.9] | 0.22 | | 812(18.3%)[17.2–19.5] | 2,762(19.8%)[17.1–19.5] | −0.04 | | | 1–2 | 1,036(31.7%)[30.1–33.3] | 3,312 (30.7 |29.1–32.3]) | 0.02 | | 621(32.3%)[30.2–34.4] | 1,947(30.5%)[28.4–32.5] | 0.04 | | 1,255(28.3%)[26.9–29.6] | 4,471(31.3%)[26.9–29.6] | −0.07 | | | ≥3 | 1,129(34.5%)[32.9–36.2] | 5,359(47.9%)[46.2–49.6] | −0.27 | | 735(38.2%) [36.0–40.4] | 3,170(49.4%)[47.2–51.7] | −0.23 | | 2,365(53.4%)[51.9–54.8] | 7,219(48.9%)[47.5–50.4] | 0.09 | | | Trauma–n (weighted %)[95% CI] | 98(3.0%)[2.4–3.6] | 252(2.3)[1.8–2.8] | 0.04 | 0.090 | 36(1.9%)[1.2–2.5] | 122(1.8%)[1.2–2.3] | 0.01 | 0.817 | 55(1.2%)[0.9–1.5] | 224(1.5%)[1.2–1.9] | −0.02 | 0.259 | | Surgery–n (weighted %)[95% CI] | 594(18.2%)[16.8–19.5] | 2,459(22.1%)[20.6–23.5] | −0.10 | <0.001 | 351(18.3%)[16.5–20.0] | 1,511(23.8%[21.9–25.7] | −0.14 | <0.001 | 935(21.1%)[19.9–22.3] | 3,397(23.4%)[22.1–24.6] | −0.06 | <0.001 | | SAPS II score at ICU admission, Mean ± SD[95% CI] | 42.8 ± 21.7[42.0–43.5] | 44.8 ± 12.5[44.4–45.2] | −0.11 | <0.001 | 43.3 ± 22.7[42.3–44.3] | 44.0 ± 12.7[43.4–44.5] | −0.04 | 0.347 | 43.2 ± 22.4[42.5–43.9] | 43.8 ± 12.8[43.4–44.2] | −0.03 | 0.210 | | Site of infection–n (weighted %)[95% CI] | | | | | | | | | | | | | | Respiratory | 1,568(48.0%)[46.3–49.7] | 4,537(41.5%)[39.8–43.2] | 0.13 | <0.001 | 826(43.0%)[40.7–45.2] | 2,486(38.8%)[36.6–41.0] | 0.09 | 0.009 | 1,888(42.6%)[41.1–44.0] | 5,719(38.9%)(37.4–40.3] | 0.08 | <0.001 | | Gastrointestinal | 521(15.9%)[14.7–17.2] | 1,714(16.0%)[14.7–17.3] | −0.00 | 0.944 | 290(15.1%)[13.5–16.7] | 1,166(18.7%)[17.0–20.4] | −0.10 | 0.003 | 704(15.9%)[14.8–16.9] | 2,533(17.6%)[16.4–18.7] | −0.05 | 0.029 | | Renal | 311(9.5%)[8.5–10.5] | 894(7.9%)[6.9–8.8] | 0.06 | 0.019 | 216(11.2%)[9.8–12.6] | 533(8.3%)[7.0–9.5] | 0.10 | 0.002 | 472(10.7%)[9.7–11.6] | 1,318(9.2%)[8.3–10.1] | 0.05 | 0.026 | | Cardiac | 306(9.4%)[8.4–10.4] | 1,138(10.4%)[9.3–11.4] | −0.03 | 0.165 | 161(8.4%)[7.1–9.6] | 701(10.9%)[9.4–12.2] | −0.09 | 0.008 | 456(10.3%)[9.4–11.2] | 1,535(10.4%)[9.5–11.3] | −0.00 | 0.862 | | Dermatologic | 180(5.5%)[4.7–6.3] | 783(6.9%)[6.1–7.8] | −0.06 | 0.017 | 96(5.0%)[4.0–6.0] | 427(6.6%)[5.5–7.6] | −0.07 | 0.038 | 271(6.1%)[5.4–6.8] | 1,008(7.0%)[6.2–7.7] | −0.03 | 0.110 | | Organ failures–n (weighted %)[95% CI] | | | | | | | | | | | | | | Respiratory | 2,069(63.3%)[61.6–64.9] | 6,440(58.9%)[57.2–60.6] | 0.09 | <0.001 | 1,197(62.3%)[60.0–64.4] | 3,659(58.7%)[56.5–60.9] | 0.07 | 0.024 | 2,708(61.1%)[59.7–62.5] | 8,368(57.8%)[56.3–59.3] | 0.07 | 0.002 | | Renal | 1,286(39.3%)[37.7–41.0] | 5,373(48.2%)[46.5–49.9] | −0.18 | <0.001 | 815(42.4%)[40.2–44.6] | 3,145(49.3%)[47.0–51.5] | −0.14 | <0.001 | 1,993(45.0%)[43.5–46.4] | 7,277(49.9%)[48.4–51.4] | −0.10 | <0.001 | | Neurologic | 1,051(32.2%)[30.5–33.8] | 2,575(23.7%)[22.2–25.1] | 0.19 | <0.001 | 659(34.3%)[32.1–36.4] | 1,473(23.1%)[21.2–25.0] | 0.25 | <0.001 | 1,288(29.1%)[27.7–30.4] | 3,393(23.3%)[22.0–24.5] | 0.13 | <0.001 | | Cardiovascular | 428(13.1%)[11.9–14.2] | 1,577(13.9%)[12.7–15.1] | −0.02 | 0.359 | 239(12.4%)[10.9–13.9] | 1,036(15.9%)[14.3–17.6] | −0.10 | 0.002 | 660(14.9%) [13.8–15.9] | 2,291(15.8%)[14.7–16.8] | −0.02 | 0.252 | | Hematologic | 395(12.1%)[11.0–13.2] | 1,705(15.9%)[14.7–17.2] | −0.11 | <0.001 | 212(11.0%)[9.6–12.4] | 960(15.2%)[13.6–16.8] | −0.12 | <0.001 | 602(13.6%) [12.6–14.6] | 2,086(14.5%)[13.5–15.5] | −0.03 | 0.212 | | Metabolic | 673(20.6%)[19.2–22.0]) | 2,554(23.1%)[21.7–24.6] | −0.06 | 0.013 | 404(21.0%)[19.2–22.8] | 1,468(23.5%)[21.6–25.3] | −0.06 | 0.068 | 1,018(23.0%)[21.7–24.2] | 3,448(23.5%)[22.2–24.7] | −0.01 | 0.553 | | Hepatic | 237(7.3%)[6.3–8.2] | 1,439(12.9%)[11.7–14.0] | −0.19 | <0.001 | 145(7.5%)[6.4–8.7] | 725(11.5%) [10.0–12.9] | −0.14 | <0.001 | 476(10.7%)[9.8–11.7] | 1,676(11.6%)[10.6–12.5] | −0.03 | 0.213 | | ICU supportive therapies–n (weighted %)[95% CI] | | | | | | | | | | | | | | Cardiopulmonary resuscitation | 161(4.9%)[4.2–5.7] | 669(6.2%)[5.3–7.0] | −0.05 | 0.031 | 85(4.4%)[3.5–5.4] | 349(5.6%)[4.6–6.6] | −0.05 | 0.096 | 177(4.0%)[3.42–4.6] | 829(5.6%)[4.9–6.2] | −0.07 | <0.001 | | Invasive mechanical ventilation | 2,787(85.3%)[84.0–86.5] | 8,960(82.0%)[80.7–83.3] | 0.09 | <0.001 | 1,602(83.3%)[81.6–85.0] | 5,076(80.5%)[78.7–82.2] | 0.07 | 0.023 | 3,564(80.4%)[79.8–81.5] | 11,556(79.8%)[78.6–80.9) | 0.02 | 0.459 | | Renal replacement therapy | 672(20.6%)[19.2–21.9] | 3,278(30.0%)[28.0–31.1] | −0.21 | <0.001 | 452(23.5%)[21.6–25.4] | 1,914(30.0%)[27.9–32.0] | −0.15 | <0.001 | 1,060(23.9%)[22.7–25.2] | 4,149(28.8%)[27.4–30.1]) | −0.11 | <0.001 | | Transfusion | 969(29.6%)[28.1–31.2] | 3,782(34.6%) [33.0–36.3] | −0.11 | <0.001 | 540(28.1%)[26.1–30.1] | 2,228(34.9%)[32.8–37.1] | −0.15 | <0.001 | 1,475(33.3%)[31.9–34.7] | 4,940(33.8%)[32.4–35.2] | −0.01 | 0.578 | | Source of hospital admission–n (weighted %)[95% CI] | | | | | | | | | | | | | | Home | 3,040(93.0%)[92.1–93.9] | 10,577 (97.1%)[96.5–97.6] | −0.19 | <0.001 | 1,814(94.3%)[93.3–95.4] | 6,075(96.2%)[95.3–97.1] | −0.09 | 0.006 | 4,212(95.0%)[94.4–95.7] | 13,952(96.4%)[95.8–96.9] | −0.07 | 0.002 | | Transfer from other hospital | 229(7.0%)[6.1–7.9] | 317(2.9%)[2.4–3.5] | 0.19 | | 109(5.7%)[4.6–6.7] | 228(3.8%)[2.9–4.6] | 0.09 | | 220(5.0%)[4.3–5.6] | 500(3.6%)[3.0–4.1] | 0.07 | | | Time to ICU admission ≤1 day–n (weighted %)[95% CI] | 2,180(67.0%)[65.1–68.3] | 6,734(62.2%)[60.5–63.8] | 0.09 | <0.001 | 1,298(67.5%)[65.4–69.6] | 3,848(61.2%)[59.0–63.4] | 0.13 | <0.001 | 2,721(61.4%)[60.0–62.8] | 8,903(61.2%)[59.7–62.6] | 0.00 | 0.823 | | Hospital characteristics–n (weighted %) [95% CI] | | | | 1.000 | | | | 1.000 | | | | 1.000 | | Academic | 1,532(46.9%)[45.2–48.6] | 5,777 (46.86 [45.2–48.6]) | 0.00 | | 840(43.7%)[41.5–45.9] | 3,187(43.7%)[41.5–45.9] | 0.00 | | 1,871(42.2%)[40.8–43.7] | 7,051(42.2%)[40.8–43.7] | 0.00 | | | Other public hospital | 1,637(50.1%)[48.4–51.8] | 4,898(50.1%)[48.4–51.8] | 0.00 | | 1,006(52.3%)[50.1–54.5] | 2,947(52.3%)[50.1–54.5] | 0.00 | | 2,391(54.0%)[52.5–55.4] | 7,011(54.0%)[52.5–55.4] | 0.00 | | | Private | 100(3.1%)[2.5–3.7] | 219(3.1%)[2.5–3.7] | 0.00 | | 77(4.0%)[3.1–4.8] | 169(4.0%)[3.1–4.8] | 0.00 | | 170(3.8%)[3.3–4.4] | 390(3.8%)[3.3–4.4] | 0.00 | | ## Discussion In this nationwide, population-based cohort study, the 30-, 90-, and 365-day CFRs in patients with septic shock admitted to the ICU were lower in patients with SMI than in other patients, after controlling for multiple potential confounding factors (using intrahospital matching and adjustments for multiple comorbidities and illness severity) and addressing potential biases not considered in previous studies [14–17]. The reasons for the differences in survival between patients with SMI and controls could not be determined in our study but may include differences in immunological profiles [35–39] and exposures to the immunomodulatory effects of psychotropic medications [40]. Immunological characteristics of patients with SMI have been reported for many years, related to effects of the psychiatric disease and the psychotropic treatments. All 3 SMI conditions are associated with dysregulated cytokine responses that may be protective in septic shock [41], as already suggested in autoimmune diseases such as multiple sclerosis [42], rheumatoid arthritis, and Crohn’s disease [40]. Overexpression of specific pro-inflammatory cytokines such as interleukin (IL)-12 and interferon-gamma (IFN-γ) has been reported in SMI, as in autoimmune diseases, and may offset the immunosuppressive state induced by sepsis [40,41]. This finding may in part be related to the treatments received by patients with SMI, with psychotropic drugs including antidepressants [43–45], lithium [46], and antipsychotics [47,48] able to modulate the inflammatory response [35]. This hypothesis has been reinforced during the Coronavirus Disease 2019 (COVID-19) pandemic, during which fluoxetine [49] (an antidepressant) and chlorpromazine [50] (an antipsychotic) were suggested to have beneficial effects. Specifically, a Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) animal model showed independent antiviral and anti-inflammatory effects of fluoxetine [51], in line with several observational studies [44,52,53]. A candidate mechanism, shared by several psychotropic medications and supported by several preclinical [54] and observational studies [44,52,55,56], is the functional inhibition of acid sphingomyelinase (FIASMA) leading to a regulation of apoptosis, cellular differentiation, proliferation, and cell migration. Finally, several RCTs and observational studies have reported evidence of efficacy of fluvoxamine at a daily dose of 200 mg or more against COVID-19 among outpatients with COVID-19 [57–59] and COVID-19 ICU patients [60]. A potential implication of these results is that the frequently observed discontinuation of psychotropic medications on admission to the ICU should be carefully considered given the risks of relapse of the psychiatric disorder as well as the potential benefits of these drugs on mortality in the context of septic shock. Further studies are needed to explore these immune and pharmacological mechanisms. The long-term goal of identifying patient groups with higher case fatality in sepsis than that in the general population is to identify the mechanisms underlying the outcome differences and, critically, modifiable mechanisms that can serve as targets for interventional approaches geared to reduce the outcome disparity of the affected group in reference to the general population. A key finding of our study (and most of the prior ones [14–17]) is that some factors unique to patients with SMI (e.g., possibly baseline immune dysfunction leading to a different, more protective, response to infection) not only negated the adverse prognostic effects of SMI in septic shock patients (which could have resulted in similar case fatality between the groups), but were associated with markedly lower case fatality among these patients. The magnitude of this effect estimate is remarkable, especially in this vulnerable population marked by low socioeconomic status. A major implication is that future work to characterize potential differences in response to infection among patients with and without SMI across key domains of the immune system may identify potential targets for therapeutic interventions to reduce short-term mortality in the general population. However, there were some important differences between patients with and without SMI after matching (e.g., fewer malignancies and fewer comorbid conditions), which may have influenced outcomes. Although these differences were adjusted for, it is possible that residual confounding remained. In addition, the social deprivation indicator is based on the area level and may thus also lead to residual confounding. The lower CFR may have health policy implications on future focus of resource allocation to improve life expectancy in patients with SMI. This finding suggests that the higher mortality rate due to infection/sepsis among patients with SMI reported in previous studies [2] appears to be due to the increased risk of infection/sepsis among patients with SMI and potentially poorer access to timely and adequate care, but not due to greater case fatality once they have been hospitalized for septic shock. As a consequence, our findings suggest that effective primary prevention interventions (i.e., before the onset of infection, to reduce the incidence of infection in patients with SMI) should be prioritized. However, evidence-based strategies for the prevention of infection in patients with SMI are scarce, as highlighted by a recent review on the prevalence rates and immunogenicity of vaccinations in patients with SMI [61]. Future studies should confirm this hypothesis on the full sample of individuals with SMI and sepsis in the population. Our study has several limitations. First, we described only patients who died in hospital, which means that the CFR might be underestimated. Deaths occurring outside the hospital are extremely rare in France but could be differentially experienced by people with SMI [28]. Nonetheless, our findings at 30 and 90 days were similar to those reported in other studies [62]. In addition, the evolution of the CFR between 30 and 90 days and between 90 and 365 days was similar in the patients with and without SMI, supporting a lack of bias to account for the different extrahospital mortality. Second, a weakness of administrative databases is the potential miscoding of diagnoses during hospital stays, which can underestimate important patient features (especially for overweight and obesity, which are insufficiently coded in administrative databases but which allow the most serious cases to be targeted for epidemiological research [63,64]) and disease severity at ICU admission. Missing data are thus assumed to indicate no disease present. In addition, the key exposure in the present study (i.e., SMI) can be misclassified due to use of ICD-10 codes, which could have affected reported effect estimates. Misclassification of mental disorders would be expected to blur the differences between groups and thus diminish outcome differences between septic shock patients with and without SMI. This would suggest that the study’s findings may represent possible underestimation of the magnitude of the better outcomes observed among patients with SMI. However, the coding has been strictly regulated since the DRG system was introduced in France. To control for these weaknesses, we used a matching procedure and adjustment based on a large number of patient characteristics and controlling for confounders at the hospital level. The matching process failed for $15\%$ of patients due to the age imbalance between patients with and without SMI. However, the sensitivity analysis on the whole cohort reported similar findings. There are also limitations associated with the lack of some variables, including specific description of psychotropic medications, body mass index, fitness, and blood lactate levels, which could be useful to categorize our patients. Furthermore, the time between the onset of infection and the need for vasopressor support could not be determined. Some patients may require vasopressor support for a problem other than septic shock. Finally, processes of care for sepsis were not analyzed in detail in our study and may have differed across compared groups, which could have led to residual confounding in modelled effects. Patients with SMI are well documented to receive poorer quality of healthcare, in addition to stigma, stereotyping, and negative attitudes towards these patients by clinicians. Such care differences would be not be expected, however, to result in better outcomes of septic patients with SMI. Such potential differences in care processes would suggest that the study’s findings may represent possible underestimation of the magnitude of the better outcomes observed among patients with SMI. In conclusion, our findings suggest that SMI patients have a better outcome from septic shock in the ICU than those without SMI. This better prognosis may be explained by different immunological mechanisms and exposures to psychotropic medications. Further studies on these mechanisms that may potentially modulate outcomes may have important implications for all septic shock patients. ## References 1. Walker ER, McGee RE, Druss BG. **Mortality in Mental Disorders and Global Disease Burden Implications: A Systematic Review and Meta-analysis**. *JAMA Psychiatry* (2015) **72** 334. DOI: 10.1001/jamapsychiatry.2014.2502 2. Correll CU, Solmi M, Croatto G, Kolton Schneider L, Rohani-Montez SC, Fairley L. **Mortality in people with schizophrenia: a systematic review and meta-analysis of relative risk and aggravating or attenuating factors.**. *World Psychiatry.* (2022) **21** 248-271. DOI: 10.1002/wps.20994 3. Thornicroft G.. **Physical health disparities and mental illness: the scandal of premature mortality.**. *Br J Psychiatry J Ment Sci* (2011) **199** 441-442. DOI: 10.1192/bjp.bp.111.092718 4. Firth J, Siddiqi N, Koyanagi A, Siskind D, Rosenbaum S, Galletly C. **The Lancet Psychiatry Commission: a blueprint for protecting physical health in people with mental illness**. *Lancet Psychiatry* (2019) **6** 675-712. DOI: 10.1016/S2215-0366(19)30132-4 5. Solmi M, Firth J, Miola A, Fornaro M, Frison E, Fusar-Poli P. **Disparities in cancer screening in people with mental illness across the world versus the general population: prevalence and comparative meta-analysis including 4 717 839 people.**. *Lancet Psychiatry.* (2020) **7** 52-63. DOI: 10.1016/S2215-0366(19)30414-6 6. Solmi M, Fiedorowicz J, Poddighe L, Delogu M, Miola A, Høye A. **Disparities in Screening and Treatment of Cardiovascular Diseases in Patients With Mental Disorders Across the World: Systematic Review and Meta-Analysis of 47 Observational Studies.**. *Am J Psychiatry.* (2021) **178** 793-803. DOI: 10.1176/appi.ajp.2021.21010031 7. De Hert M, Correll CU, Bobes J, Cetkovich-Bakmas M, Cohen D, Asai I. **Physical illness in patients with severe mental disorders. I. Prevalence, impact of medications and disparities in health care.**. *World Psychiatry.* (2011) **10** 52-77. DOI: 10.1002/j.2051-5545.2011.tb00014.x 8. De Hert M, Cohen D, Bobes J, Cetkovich-Bakmas M, Leucht S, Ndetei DM. **Physical illness in patients with severe mental disorders. II. Barriers to care, monitoring and treatment guidelines, plus recommendations at the system and individual level.**. *World Psychiatry.* (2011) **10** 138-151. DOI: 10.1002/j.2051-5545.2011.tb00036.x 9. Laupland KB, Tabah A, Holley AD, Bellapart J, Pilcher DV. **Decreasing Case-Fatality But Not Death Following Admission to ICUs in Australia, 2005–2018.**. *Chest* (2021) **159** 1503-1506. DOI: 10.1016/j.chest.2020.11.059 10. Vincent J-L, Marshall JC, Ñamendys-Silva SA, François B, Martin-Loeches I, Lipman J. **Assessment of the worldwide burden of critical illness: the Intensive Care Over Nations (ICON) audit.**. *Lancet Respir Med* (2014) **2** 380-386. DOI: 10.1016/S2213-2600(14)70061-X 11. Vincent J-L, Jones G, David S, Olariu E, Cadwell KK. **Frequency and mortality of septic shock in Europe and North America: a systematic review and meta-analysis**. *Crit Care* (2019) **23** 196. DOI: 10.1186/s13054-019-2478-6 12. Ribe AR, Vestergaard M, Katon W, Charles M, Benros ME, Vanderlip E. **Thirty-Day Mortality After Infection Among Persons With Severe Mental Illness: A Population-Based Cohort Study in Denmark.**. *Am J Psychiatry.* (2015) **172** 776-783. DOI: 10.1176/appi.ajp.2015.14091100 13. Davydow DS, Ribe AR, Pedersen HS, Vestergaard M, Fenger-Grøn M. **The association of unipolar depression with thirty-day mortality after hospitalization for infection: A population-based cohort study in Denmark.**. *J Psychosom Res* (2016) **89** 32-38. DOI: 10.1016/j.jpsychores.2016.08.006 14. Schwarzkopf D, Fleischmann-Struzek C, Rüddel H, Reinhart K, Thomas-Rüddel DO. **A risk-model for hospital mortality among patients with severe sepsis or septic shock based on German national administrative claims data**. *PLoS ONE* (2018) **13** e0194371. DOI: 10.1371/journal.pone.0194371 15. Oud L, Garza J. **Impact of history of mental disorders on short-term mortality among hospitalized patients with sepsis: A population-based cohort study.**. *PLoS ONE.* (2022) **17** e0265240. DOI: 10.1371/journal.pone.0265240 16. Chou EH, Mann S, Hsu T-C, Hsu W-T, Liu CC-Y, Bhakta T. **Incidence, trends, and outcomes of infection sites among hospitalizations of sepsis: A nationwide study.**. *PLoS ONE* (2020) **15** e0227752. DOI: 10.1371/journal.pone.0227752 17. Ford DW, Goodwin AJ, Simpson AN, Johnson E, Nadig N, Simpson KN. **A Severe Sepsis Mortality Prediction Model and Score for Use With Administrative Data.**. *Crit Care Med.* (2016) **44** 319-327. DOI: 10.1097/CCM.0000000000001392 18. Hogue CW, Stearns JD, Colantuoni E, Robinson KA, Stierer T, Mitter N. **The impact of obesity on outcomes after critical illness: a meta-analysis.**. *Intensive Care Med* (2009) **35** 1152-1170. DOI: 10.1007/s00134-009-1424-5 19. Afzal M, Siddiqi N, Ahmad B, Afsheen N, Aslam F, Ali A. **Prevalence of Overweight and Obesity in People With Severe Mental Illness: Systematic Review and Meta-Analysis.**. *Front Endocrinol.* (2021) **12** 769309. DOI: 10.3389/fendo.2021.769309 20. Rhee C, Kadri SS, Danner RL, Suffredini AF, Massaro AF, Kitch BT. **Diagnosing sepsis is subjective and highly variable: a survey of intensivists using case vignettes.**. *Crit Care* (2016) **20** 89. DOI: 10.1186/s13054-016-1266-9 21. Rhee C, Murphy MV, Li L, Platt R, Klompas M. **Comparison of trends in sepsis incidence and coding using administrative claims versus objective clinical data**. *Clin Infect Dis* (2015) **60** 88-95. DOI: 10.1093/cid/ciu750 22. Epstein L, Dantes R, Magill S, Fiore A. **Varying Estimates of Sepsis Mortality Using Death Certificates and Administrative Codes—United States, 1999–2014.**. *MMWR Morb Mortal Wkly Rep.* (2016) **65** 342-345. DOI: 10.15585/mmwr.mm6513a2 23. Strand K, Strand LI, Flaatten H. **The interrater reliability of SAPS II and SAPS 3.**. *Intensive Care Med* (2010) **36** 850-853. DOI: 10.1007/s00134-010-1772-1 24. Fond G, Salas S, Pauly V, Baumstarck K, Bernard C, Orleans V. **End-of-life care among patients with schizophrenia and cancer: a population-based cohort study from the French national hospital database**. *Lancet Public Health* (2019) **4** e583-e591. DOI: 10.1016/S2468-2667(19)30187-2 25. Fond G, Pauly V, Leone M, Llorca P-M, Orleans V, Loundou A. **Disparities in Intensive Care Unit Admission and Mortality Among Patients With Schizophrenia and COVID-19: A National Cohort Study.**. *Schizophr Bull.* (2021) **47** 624-634. DOI: 10.1093/schbul/sbaa158 26. Fabre C, Pauly V, Baumstarck K, Etchecopar-Etchart D, Orleans V, Llorca P-M. **Pregnancy, delivery and neonatal complications in women with schizophrenia: a national population-based cohort study**. *Lancet Reg Health Eur* (2021) **10** 100209. DOI: 10.1016/j.lanepe.2021.100209 27. Maharaj R, McGuire A, Street A. **Association of Annual Intensive Care Unit Sepsis Caseload With Hospital Mortality From Sepsis in the United Kingdom, 2010–2016**. *JAMA Netw Open* (2021) **4** e2115305. DOI: 10.1001/jamanetworkopen.2021.15305 28. Armoiry X, Obadia J-F, Pascal L, Polazzi S, Duclos A. **Comparison of transcatheter versus surgical aortic valve implantation in high-risk patients: A nationwide study in France.**. *J Thorac Cardiovasc Surg* (2018) **156** 1017-1025. DOI: 10.1016/j.jtcvs.2018.02.092 29. Benchimol EI, Smeeth L, Guttmann A, Harron K, Moher D, Petersen I. **The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) Statement.**. *PLoS Med.* (2015) **12** e1001885. DOI: 10.1371/journal.pmed.1001885 30. Sakr Y, Jaschinski U, Wittebole X, Szakmany T, Lipman J, Ñamendys-Silva SA. **Sepsis in Intensive Care Unit Patients: Worldwide Data From the Intensive Care over Nations Audit.**. *Open Forum Infect Dis.* (2018) **5** ofy313. DOI: 10.1093/ofid/ofy313 31. Rey G, Jougla E, Fouillet A, Hémon D. **Ecological association between a deprivation index and mortality in France over the period 1997–2001: Variations with spatial scale, degree of urbanicity, age, gender and cause of death.**. *BMC Public Health.* (2009) **9** 1-12. DOI: 10.1186/1471-2458-9-33 32. Quan H, Li B, Couris CM, Fushimi K, Graham P, Hider P. **Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries**. *Am J Epidemiol* (2011) **173** 676-682. DOI: 10.1093/aje/kwq433 33. Austin PC. **Using the Standardized Difference to Compare the Prevalence of a Binary Variable Between Two Groups in Observational Research.**. *Commun Stat Simul Comput.* (2009) **38** 1228-1234. DOI: 10.1080/03610910902859574 34. Sullivan GM, Feinn R. **Using Effect Size-or Why the P Value Is Not Enough.**. *J Grad Med Educ.* (2012) **4** 279-282. DOI: 10.4300/JGME-D-12-00156.1 35. Riché F, Chousterman BG, Valleur P, Mebazaa A, Launay J-M, Gayat E. **Protracted immune disorders at one year after ICU discharge in patients with septic shock**. *Crit Care* (2018) **22** 42. DOI: 10.1186/s13054-017-1934-4 36. Tolsma V, Schwebel C, Azoulay E, Darmon M, Souweine B, Vesin A. **Sepsis severe or septic shock: outcome according to immune status and immunodeficiency profile**. *Chest* (2014) **146** 1205-1213. DOI: 10.1378/chest.13-2618 37. Niu J, Qin B, Wang C, Chen C, Yang J, Shao H. **Identification of Key Immune-Related Genes in the Progression of Septic Shock.**. *Front Genet* (2021) **12** 668527. DOI: 10.3389/fgene.2021.668527 38. Nakamori Y, Park EJ, Shimaoka M. **Immune Deregulation in Sepsis and Septic Shock: Reversing Immune Paralysis by Targeting PD-1/PD-L1 Pathway.**. *Front Immunol.* (2020) **11** 624279. DOI: 10.3389/fimmu.2020.624279 39. Carnio EC, Moreto V, Giusti-Paiva A, Antunes-Rodrigues J. **Neuro-immune-endocrine mechanisms during septic shock: role for nitric oxide in vasopressin and oxytocin release**. *Endocr Metab Immune Disord Drug Targets* (2006) **6** 137-142. DOI: 10.2174/187153006777442396 40. Sheth M, Benedum CM, Celi LA, Mark RG, Markuzon N. **The association between autoimmune disease and 30-day mortality among sepsis ICU patients: a cohort study.**. *Crit Care* (2019) **23** 93. DOI: 10.1186/s13054-019-2357-1 41. Goldsmith DR, Rapaport MH, Miller BJ. **A meta-analysis of blood cytokine network alterations in psychiatric patients: comparisons between schizophrenia, bipolar disorder and depression**. *Mol Psychiatry* (2016) **21** 1696-1709. DOI: 10.1038/mp.2016.3 42. Oud L, Garza J. **Association of multiple sclerosis with mortality in sepsis: a population-level analysis.**. *J Intensive Care* (2022) **10** 36. DOI: 10.1186/s40560-022-00628-1 43. Tynan RJ, Weidenhofer J, Hinwood M, Cairns MJ, Day TA, Walker FR. **A comparative examination of the anti-inflammatory effects of SSRI and SNRI antidepressants on LPS stimulated microglia**. *Brain Behav Immun* (2012) **26** 469-479. DOI: 10.1016/j.bbi.2011.12.011 44. Hoertel N, Sánchez-Rico M, Vernet R, Beeker N, Jannot A-S, Neuraz A. **Association between antidepressant use and reduced risk of intubation or death in hospitalized patients with COVID-19: results from an observational study**. *Mol Psychiatry* (2021) **26** 5199-5212. DOI: 10.1038/s41380-021-01021-4 45. Rosen DA, Seki SM, Fernández-Castañeda A, Beiter RM, Eccles JD, Woodfolk JA. **Modulation of the sigma-1 receptor–IRE1 pathway is beneficial in preclinical models of inflammation and sepsis**. *Sci Transl Med* (2019) **11** eaau5266. DOI: 10.1126/scitranslmed.aau5266 46. Landén M, Larsson H, Lichtenstein P, Westin J, Song J. **Respiratory infections during lithium and valproate medication: a within-individual prospective study of 50,000 patients with bipolar disorder.**. *Int J Bipolar Disord* (2021) **9** 4. DOI: 10.1186/s40345-020-00208-y 47. Steiner J, Jacobs R, Panteli B, Brauner M, Schiltz K, Bahn S. **Acute schizophrenia is accompanied by reduced T cell and increased B cell immunity**. *Eur Arch Psychiatry Clin Neurosci* (2010) **260** 509-518. DOI: 10.1007/s00406-010-0098-x 48. Maino K, Gruber R, Riedel M, Seitz N, Schwarz M, Müller N. **T- and B-lymphocytes in patients with schizophrenia in acute psychotic episode and the course of the treatment.**. *Psychiatry Res* (2007) **152** 173-180. DOI: 10.1016/j.psychres.2006.06.004 49. Hoertel N.. **Do the Selective Serotonin Reuptake Inhibitor Antidepressants Fluoxetine and Fluvoxamine Reduce Mortality Among Patients With COVID-19?**. *JAMA Netw Open* (2021) **4** e2136510. DOI: 10.1001/jamanetworkopen.2021.36510 50. Plaze M, Attali D, Petit A-C, Blatzer M, Simon-Loriere E, Vinckier F. **Repurposing chlorpromazine to treat COVID-19: The reCoVery study.**. *L’Encephale* (2020) **46** 169-172. DOI: 10.1016/j.encep.2020.05.006 51. Péricat D, Leon-Icaza SA, Sanchez Rico M, Mühle C, Zoicas I, Schumacher F. **Antiviral and Anti-Inflammatory Activities of Fluoxetine in a SARS-CoV-2 Infection Mouse Model.**. *Int J Mol Sci.* (2022) **23** 13623. DOI: 10.3390/ijms232113623 52. Hoertel N, Sánchez-Rico M, Gulbins E, Kornhuber J, Carpinteiro A, Abellán M. **Association between FIASMA psychotropic medications and reduced risk of intubation or death in individuals with psychiatric disorders hospitalized for severe COVID-19: an observational multicenter study.**. *Transl Psychiatry* (2022) **12** 90. DOI: 10.1038/s41398-022-01804-5 53. Sidky H, Sahner DK, Girvin AT, Hotaling N, Michael SG, Kurilla MG. **Assessing the Effect of Selective Serotonin Reuptake Inhibitors in the Prevention of Post-Acute Sequelae of COVID-19**. *MedRxiv Prepr Serv Health Sci* (2022). DOI: 10.1101/2022.11.09.22282142 54. Kornhuber J, Hoertel N, Gulbins E. **The acid sphingomyelinase/ceramide system in COVID-19**. *Mol Psychiatry* (2022) **27** 307-314. DOI: 10.1038/s41380-021-01309-5 55. Fritz BA, Hoertel N, Lenze EJ, Jalali F, Reiersen AM. **Association between antidepressant use and ED or hospital visits in outpatients with SARS-CoV-2.**. *Transl Psychiatry* (2022) **12** 341. DOI: 10.1038/s41398-022-02109-3 56. Hoertel N, Sánchez-Rico M, Kornhuber J, Gulbins E, Reiersen AM, Lenze EJ. **Antidepressant Use and Its Association with 28-Day Mortality in Inpatients with SARS-CoV-2: Support for the FIASMA Model against COVID-19.**. *J Clin Med.* (2022) **11** 5882. DOI: 10.3390/jcm11195882 57. Lenze EJ, Mattar C, Zorumski CF, Stevens A, Schweiger J, Nicol GE. **Fluvoxamine vs Placebo and Clinical Deterioration in Outpatients With Symptomatic COVID-19: A Randomized Clinical Trial**. *JAMA* (2020) **324** 2292-2300. DOI: 10.1001/jama.2020.22760 58. Reis G, Dos Santos Moreira-Silva EA, Silva DCM, Thabane L, Milagres AC, Ferreira TS. **Effect of early treatment with fluvoxamine on risk of emergency care and hospitalisation among patients with COVID-19: the TOGETHER randomised, platform clinical trial**. *Lancet Glob Health* (2022) **10** e42-e51. DOI: 10.1016/S2214-109X(21)00448-4 59. Lee TC, Vigod S, Bortolussi-Courval É, Hanula R, Boulware DR, Lenze EJ. **Fluvoxamine for Outpatient Management of COVID-19 to Prevent Hospitalization: A Systematic Review and Meta-analysis**. *JAMA Netw Open* (2022) **5** e226269. DOI: 10.1001/jamanetworkopen.2022.6269 60. Calusic M, Marcec R, Luksa L, Jurkovic I, Kovac N, Mihaljevic S. **Safety and efficacy of fluvoxamine in COVID-19 ICU patients: An open label, prospective cohort trial with matched controls**. *Br J Clin Pharmacol* (2022) **88** 2065-2073. DOI: 10.1111/bcp.15126 61. Bonkat N, Fellendorf FT, Dalkner N, Reininghaus EZ. **Severe mental disorders and vaccinations—a systematic review**. *World J Biol Psychiatry* (2022) **23** 501-516. DOI: 10.1080/15622975.2021.2013095 62. Bauer M, Gerlach H, Vogelmann T, Preissing F, Stiefel J, Adam D. **Mortality in sepsis and septic shock in Europe, North America and Australia between 2009 and 2019- results from a systematic review and meta-analysis.**. *Crit Care.* (2020) **24** 239. DOI: 10.1186/s13054-020-02950-2 63. Gribsholt SB, Pedersen L, Richelsen B, Thomsen RW. **Validity of ICD-10 diagnoses of overweight and obesity in Danish hospitals.**. *Clin Epidemiol.* (2019) **11** 845-854. DOI: 10.2147/CLEP.S214909 64. Martin B-J, Chen G, Graham M, Quan H. **Coding of obesity in administrative hospital discharge abstract data: accuracy and impact for future research studies.**. *BMC Health Serv Res* (2014) **14** 70. DOI: 10.1186/1472-6963-14-70
--- title: 'Determinants of QuantiFERON Plus-diagnosed tuberculosis infection in adult Ugandan TB contacts: A cross-sectional study' authors: - Jonathan Mayito - Adrian R. Martineau - Divya Tiwari - Lydia Nakiyingi - David P. Kateete - Stephen T. Reece - Irene Andia Biraro journal: PLOS ONE year: 2023 pmcid: PMC10042355 doi: 10.1371/journal.pone.0281559 license: CC BY 4.0 --- # Determinants of QuantiFERON Plus-diagnosed tuberculosis infection in adult Ugandan TB contacts: A cross-sectional study ## Abstract ### Background The tuberculin skin test is commonly used to diagnose latent tuberculosis infection (LTBI) in resource-limited settings, but its specificity is limited by factors including cross-reactivity with BCG vaccine and environmental mycobacteria. Interferon-gamma release assays (IGRA) overcome this problem by detecting M. tuberculosis complex-specific responses, but studies to determine risk factors for IGRA-positivity in high TB burden settings are lacking. ### Methods We conducted a cross-sectional study to determine factors associated with a positive IGRA by employing the QuantiFERON-TB® Gold-plus (QFT Plus) assay in a cohort of asymptomatic adult TB contacts in Kampala, Uganda. Multivariate logistic regression analysis with forward stepwise logit function was employed to identify independent correlates of QFT Plus-positivity. ### Results Of the 202 participants enrolled, $\frac{129}{202}$ ($64\%$) were female, $\frac{173}{202}$ ($86\%$) had a BCG scar, and $\frac{67}{202}$ ($33\%$) were HIV-infected. Overall, $\frac{105}{192}$ ($54\%$, $95\%$ CI 0.48–0.62) participants had a positive QFT Plus result. Increased risk of QFT-Plus positivity was independently associated with casual employment/unemployment vs. non-casual employment (adjusted odds ratio (aOR) 2.18, $95\%$ CI 1.01–4.72), a family vs. non-family relation to the index patient (aOR 2.87, $95\%$ CI 1.33–6.18), living in the same vs. a different house as the index (aOR 3.05, $95\%$ CI 1.28–7.29), a higher body mass index (BMI) (aOR per additional kg/m2 1.09, $95\%$ CI 1.00–1.18) and tobacco smoking vs. not (aOR 2.94, $95\%$ CI 1.00–8.60). HIV infection was not associated with QFT-Plus positivity (aOR 0.91, $95\%$ CI 0.42–1.96). ### Conclusion Interferon Gamma Release Assay positivity in this study population was lower than previously estimated. Tobacco smoking and BMI were determinants of IGRA positivity that were previously unappreciated. ## Background Tuberculosis (TB) ranks second after Covid-19 among the leading infectious causes of death. In 2020, it was responsible for 1.3 million and 214, 000 deaths among HIV-uninfected and HIV-infected people respectively [1]. Estimates indicate that $25\%$ of the global population has latent TB and holds a $10\%$ life time risk of reactivation to active disease, usually within 2–5 years following exposure [2]. Latent TB infection (LTBI) is diagnosed using immunological tests, due to lack of a gold standard microbiologic test, with the interferon gamma release assay (IGRA) being preferred to Tuberculin Skin Test (TST) because of its higher specificity [3]. This arises as a consequence of the IGRA’s reduced potential for cross-reactivity with antigens from *Mycobacterium bovis* and environmental mycobacteria [4]. However, IGRAs have a low positive predictive value (PPV), $2.7\%$, for risk of reactivation to active TB [5]. Therefore the individuals to whom the test is applied should be those at a higher risk of reactivation, who are more likely to benefit from preventive therapy [5]. The TB2 antigen in QuantiFERON-TB Gold PLUS (QFT-PLUS) stimulates both CD4+ and CD8+ T cells. Detection of CD8+ T cells responses improves the PPV to $5.7\%$, which reflects increased responses early after exposure when risk for TB progression is highest [6, 7]. Targeting individuals at high risk of reactivation is critical as a stronger protective effect from isoniazid preventive therapy was observed in individuals with a positive TST than in those with a negative TST [8]. Therefore, provision of TB preventative therapy based on symptom-based screening alone, a practice used in most high burden settings, will expose many individuals at low risk of TB progression to potentially toxic drugs. Specificity of symptom-based screening may be improved by integrating determinants of IGRA positivity. These determinants have largely been evaluated in low-burden settings and are likely distinct from those present in high-burden settings. Male gender, intravenous drug use, CD4 count of 200–349 per μl, and a past history or exposure to M. tuberculosis were associated with a positive IGRA among refugees in Australia [9]. Likewise, older age, Indian ethnicity, being foreign-born, primary or lesser education, and alcohol consumption more than once a week, were associated with a positive IGRA in Singaporean adults aged 18–79 years [10]. Diabetes and pregnancy, conditions associated with reduced immunity, are associated with lower IGRA positivity while the rate of indeterminate IGRA results is increased in pregnancy but not in diabetes [11–13]. On the other hand, alcohol use, advanced age and history of pulmonary TB disease were risk factors for indeterminate IGRA results in a Portuguese national TB survey [14]. Also, young age, malarial infection, iron deficiency anemia and helminth infestation were associated with indeterminate IGRA results [15] while advanced age (> 60 years) and low lymphocyte counts, including CD4 count < 200, were associated with false negative IGRA results [16]. The determinants of TST and IGRA positivity differ with advancing age being the common factor independently associated with the two [16, 17]. The difference might be due to IGRA being based on the region of difference-1 (RD-1) M. tuberculosis specific antigens that make it non-cross reactive with BCG and environmental mycobacteria antigens [18]. Lou et al found that male sex, dental surgery degree studies and close contact with index patient were independently associated with TST positivity among Ugandan medical students [19]. In Morocco, a study by Sabri et al. 2019 showed that male sex, age groups 34–45 years—more so 45–60 years, family history of TB as well as working in a pulmonary unit were strong predictors of TST positivity [17]. Birth in a high-risk country and male gender were also independently associated with a positive TST result [20] while contact with an HIV negative index patient as opposed to an HIV positive one was associated with higher odds of a positive TST [21]. A higher baseline TST induration, older age (>15 years), and a higher epidemiological risk score were independently associated with TST conversion and time to TST conversion in a Ugandan cohort of household contacts of index TB patients. The majority of TST conversions happened within six months, suggesting that TST can be used as a biomarker to detect people who are at risk of developing active TB [22]. On the other hand, HIV infection alters the reactivity of TST because it impairs the cell-mediated immunity that forms the basis of the test [23]. Consequently, the absolute CD4 and total lymphocyte counts have a positive correlation with the TST reactivity [24] and a lower cut of TST reactivity of 5 mm rather than 10 mm or higher for HIV infection is adopted. The TST reactivity is also reduced in other immune suppressive conditions like pregnancy [25], diabetes, renal disease, or organ transplant [26]. We therefore sought to evaluate the determinants of IGRA positivity in asymptomatic close contacts of index pulmonary TB patients in a high burden setting. Close contacts are recently exposed and therefore at a higher risk of infection and progression to active TB. In addition, risk factors associated with recent infection may differ from those of remote infection. ## Materials and methods Ethical approval was obtained from Makerere University School of Biomedical Sciences Institutional Review Board, reference number SBS 595. Informed consent was obtained from all participants before any study related procedures were conducted. This was a cross sectional study carried out between June 2019 and June 2020, among close contacts of index pulmonary TB patients aged 18 years or more with a negative TB symptom screen, a normal CXR and no prior treatment with TB drugs. Participants were recruited from the Infectious Diseases Institute, Mulago National Referral Hospital and Kampala City Council Authority TB clinics. A close contact was taken as an individual who spent a night or frequent or extended daytime periods with an index pulmonary TB patient during the 3 months before they initiated treatment [27]. Identification of bacteriologically confirmed index pulmonary TB patients was carried out in the above TB clinics and they were requested for permission for their contacts to participate in the study. The close contacts were invited to give written consent to participate in the study after being briefed about the study. Consenting participants gave social demographic information and a detailed history of contact with the index pulmonary TB patient including: number of index pulmonary TB patients the close contact had contact with, relationship with the index patient, proximity of contact, intensity of contact and duration the index patient had coughed prior to initiating treatment. Also, information was collected on bacillary load in the index patient’s sputum results, smoking status, BCG immunization status and for HIV positive participants, last viral load results. Formal employment, private business, casual laborer, peasant, and unemployed were the subcategories assessed for occupation. In order to have higher-powered sample numbers, these were re-categorized according to the risk they pose for M. tuberculosis infection, i.e., congestion and other risky working and living conditions. Because they pose comparable hazards, formal employment and private business were classed as non-casual employment, while the rest was categorized as casual employment or unemployment. Recruited participants underwent a physical examination including: measurement of weight, height, and temperature, as well as examination for peripheral lymphadenopathy, chest auscultation, abdominal palpation for masses, and skeletal system examination for joint swelling and vertebral column gibus. Participants then donated 5 ml of blood, of which 4 ml were used for IGRA testing and 1 ml for HIV testing. The IGRA samples were analyzed using QuantiFERON-TB® Gold-plus (QFTPlus) according to the manufacture’s standard operating procedures while HIV testing was done according to the national testing algorithm [28]. Sample size Sample size was predicated on the requirement for 10–20 participants per independent variable included in regression analyses [29]: a total of 202 participants were recruited to investigate associations for 16 independent variables. ## Statistical analysis In descriptive analysis, continuous variables were reported as means with the standard deviation (SD) while categorical variables were reported as proportions in terms of frequencies and percentages. In inferential analysis, a random effect logistic regression analysis was used to determine factors associated with a positive IGRA test. Univariate logistic regression models were used to obtain unadjusted odds ratios for all characteristics potentially associated with a positive IGRA. Multivariate logistic regression analysis with forward stepwise logit function was done with variables included as a priori based on biological plausibility with IGRA positivity. Adjusted odds ratio were reported with $95\%$ confidence intervals ($95\%$ CI). Data was analyzed using Stata/IC 15.0, StataCorp LLC Texas USA. For all comparisons, a two-tailed P-value < 0.05 was considered significant. ## Results Participant socio-demographic characteristics are shown in Table 1. Of the 202 participants enrolled, $\frac{129}{202}$ ($64\%$) were female, $\frac{173}{202}$ ($86\%$) had a BCG scar, and $\frac{67}{202}$ ($33\%$) were from an HIV positive cohort. The average age was 31 (SD: 10) years, while $\frac{130}{202}$ ($64\%$) had a BMI <25 kg/m2. The majority, were related to [$\frac{142}{202}$ ($70\%$)], lived in the same house as [$\frac{150}{202}$ ($74\%$)], or spent more than half of the day [$\frac{156}{202}$ ($77\%$)] with the index patient. Up to $\frac{109}{202}$ ($54\%$) participants had attained post-primary education, $\frac{151}{202}$ ($75\%$) were in casual employment or unemployed while only $\frac{32}{202}$ ($16\%$) were tobacco smokers. Casual employment here referred to an informal employment, which was physical in nature and occurred in a congested work environment. **Table 1** | Independent variable | Category | Number (%) | | --- | --- | --- | | Age, yrs. | 18–29.9 | 103 (51) | | Age, yrs. | 30–49.9 | 85 (42) | | Age, yrs. | >50 | 14 (7) | | Sex | F | 129 (64) | | Sex | M | 73 (36) | | Education | Post-primary | 109 (54) | | Education | Primary or less | 93 (46) | | 1Occupation | Non-casual employment | 51 (25) | | 1Occupation | casual employment or unemployed | 151 (75) | | BMI, kg/m 2 | Underweight (< 18.5) | 16 (8) | | BMI, kg/m 2 | Normal weight (18.5–24.9) | 130 (64) | | BMI, kg/m 2 | Overweight (> 25) | 56 (28) | | 2Index cases exposures (no. index cases) | 1 | 164 (81) | | 2Index cases exposures (no. index cases) | 2 or more | 38 (19) | | Place of contact with index case | Non-household contact | 23 (11) | | Place of contact with index case | Household contact | 179 (87) | | Relation with index | Non-family relation | 60 (30) | | Relation with index | Family relation - | 142 (70) | | Exposure intensity to index case | <50% of day | 45 (23) | | Exposure intensity to index case | >50% of day | 156 (77) | | Proximity to index case | Different houses | 52 (26) | | Proximity to index case | Same house | 150 (74) | | Tobacco smoking | Not a smoker | 170 (84) | | Tobacco smoking | Current smoker | 32 (16) | | HIV status | Uninfected | 135 (67) | | HIV status | Infected | 67 (33) | | QFT status | Negative | 87 (43) | | QFT status | Positive | 105 (52) | | QFT status | Intermediates | 10 (5) | | BCG status | No scar | 28 (14) | | BCG status | Scar present | 173 (86) | | Viral load suppression (cut off 1000 copies/mL) | Suppressed viral load | 61 (91) | | Viral load suppression (cut off 1000 copies/mL) | Non-suppressed viral load | 6 (9) | ## Prevalence of IGRA positivity and associated factors After excluding $\frac{10}{202}$ ($5\%$) participants with indeterminate results, $\frac{105}{192}$ ($54\%$, $95\%$ CI 0.48–0.62) participants had a positive QFT-Plus result (Fig 1). **Fig 1:** *Subject recruitment and IGRA outcomes.* Fifty five percent ($\frac{36}{65}$) of HIV positive participants had a positive QFT-Plus result compared to $\frac{69}{127}$ ($54\%$) of HIV negative participants. More participants aged 50 years and above, $\frac{9}{14}$ ($64\%$), were QFT-Plus positive compared to the younger age groups as were tobacco smokers, $\frac{20}{31}$ ($65\%$), compared to non-smokers, $\frac{85}{161}$ ($53\%$). Using bivariate logistic regression, five factors were independently associated with an increased risk of a positive QFT-Plus result: unemployment or in casual employment vs. non-casual employment (Odds Ratio (OR): 2.02, $95\%$ CI 1.04–3.91), household vs. non-household as place of contact (OR: 2.92, $95\%$ CI 1.13–7.52), family vs. non-family relation to the index patient (OR: 2.96, $95\%$ CI 1.56–5.61), living in the same vs. different houses as the index patient (OR: 2.24, $95\%$ CI 1.16–4.32), and a higher BMI (OR per kg/m2: 1.08, $95\%$ CI 1.00–1.16). Using multivariate logistic regression, in casual employment or unemployment vs. non-casual employment (adjusted OR [aOR]: 2.18, $95\%$ CI 1.01–4.72), family vs. non-family relation to the index patient (aOR: 2.87 CI 1.336.18), living in the same house vs. different houses as the index patient (aOR: 3.05, $95\%$ CI 1.28–7.29), and a high BMI (aOR per kg/m2: 1.09, $95\%$ CI 1.00–1.18) remained independently associated with an increased risk of a positive QFT-Plus result whereas household vs. non-household location of index case contact was no longer significant. In addition, smoking vs. not smoking tobacco (aOR: 2.94 CI 1.00–8.60) was also independently associated with an increased risk of a positive QFT-Plus result (Table 2). **Table 2** | Independent variable | Category | Proportion QFT-positive n/N (%) | Univariate analysis | Univariate analysis.1 | Multivariable analysis | Multivariable analysis.1 | | --- | --- | --- | --- | --- | --- | --- | | Independent variable | Category | Proportion QFT-positive n/N (%) | Unadjusted odds ratio | P-value | Adjusted odds ratio1 | P-value | | Age, yrs. | 18–29.9 | 56/99 (57) | Ref | | | | | Age, yrs. | 30–49.9 | 40/79 (51) | 0.79 [0.43–1.43] | 0.430 | 1.00 [0.97–1.04] | 0.983 | | Age, yrs. | >50 | 9/14 (64) | 1.38 [0.43–4.42] | 0.586 | | | | Sex | Female | 68/124 (55) | Ref | | Ref | | | Sex | Male | 37/68 (54) | 0.98 [0.54–1.78] | 0.955 | 0.93 [0.45–1.94] | 0.854 | | Education | Post-primary | 54/104 (52) | Ref | | Ref | | | Education | Primary or less | 51/88 (58) | 1.28 [0.72–2.26] | 0.403 | 0.96 [0.47–1.95] | 0.906 | | Occupation | Non-casual employment | 20/48 (42) | Ref | | Ref | | | Occupation | Casual employment or unemployed | 85/144 (59) | 2.02 [1.04–3.91] | 0.038 | 2.18 [1.01–4.72] | 0.047 | | BMI, kg/m 2 | | | 1.08 [1.00–1.16] | 0.040 | 1.09 [1.00–1.18] | 0.049 | | Index cases participant was exposed | 1 | 84/155 (54) | Ref | | | | | Index cases participant was exposed | 2 or more | 21/37 (57) | 1.11 [0.54–2.29] | 0.778 | 1.09 [0.47–2.49] | 0.845 | | Place of contact with index case | Non-household contact | 7/22 (32) | Ref | | Ref | | | Place of contact with index case | Household contact | 98/170 (58) | 2.92 [1.13–7.52] | 0.027 | 2.50 [0.75–8.38] | 0.138 | | Relation with index | Non-family relation | 21/58 (36) | Ref | | Ref | | | Relation with index | Family relation | 84/134 (63) | 2.96 [1.56–5.61] | 0.001 | 2.87 [1.336.18] | 0.007 | | Exposure intensity to index case | <50% of day | 24/45 (53) | Ref | | Ref | | | Exposure intensity to index case | >50% of day | 81/147 (55) | 1.07 [0.55–2.10] | 0.835 | 0.72 [0.31–1.69] | 0.451 | | Proximity to index case | Different houses | 20/50 (40) | Ref | | Ref | | | Proximity to index case | Same house | 85/142 (60) | 2.24 [1.16–4.32] | 0.016 | 3.05 [1.28–7.29] | 0.012 | | Tobacco smoking | Not a smoker | 85/161 (53) | Ref | | Ref | | | Tobacco smoking | Current smoker | 20/31 (65) | 1.63 [0.73–3.61] | 0.233 | 2.94 [1.00–8.60] | 0.050 | | HIV status | Uninfected | 69/127 (54) | Ref | | Ref | | | HIV status | Infected | 36/65 (55) | 1.04 [0.57–1.90] | 0.890 | 0.91 [0.42–1.96] | 0.810 | | BCG status | No scar | 13/27 (48) | Ref | | Ref | | | BCG status | Scar present | 92/165 (56) | 1.36 [0.60–3.066] | 0.463 | 1.07 [0.39–2.90] | 0.898 | | 1 Sputum bacillary load, index case | Low | 22/41 (54) | Ref | | | | | 1 Sputum bacillary load, index case | High | 74/133 (56) | 1.08 [0.54–2.19] | 0.824 | | | | 1 Sputum bacillary load, index case | Not known | 9/18 (50) | 0.86 [0.54–2.19] | 0.796 | | | | 2 Cough duration, index case | < 1 month | 16/29 (55) | Ref | | | | | 2 Cough duration, index case | ≥ 1 month | 70/129 (54) | 0.96 [0.43–2.17] | 0.929 | | | | 2 Cough duration, index case | Not known | 19/34 (56) | 1.03 [0.38–2.79] | 0.955 | | | The distribution of BMI for participants with a positive QFTPlus result was slightly positively skewed (Fig 2) compared to that of participants with a negative QFTPlus results which was normally distributed. **Fig 2:** *Distribution boxplots of BMI for participants with positive QFTPlus (A) and negative QFTPlus (B) results.* ## Discussion We found a $54\%$ prevalence of QFT-Plus positivity, which did not differ between HIV positive and HIV negative participants. Tobacco smoking and BMI are hitherto unrecognized factors associated with risk of a positive QFT-Plus result. Family relation, living in the same house as the index patient, and a casual employment or unemployment were the other factors associated with a positive QFT-Plus result. The prevalence of IGRA positivity in our study was less than that reported by Biraro et al ($65\%$) in the same setting [30]. In both studies, the study population was asymptomatic close contacts of index pulmonary TB patients; however, Biraro et al included children less than 18 years. Children, particularly those less than 5 years carry a higher risk for TB infection and disease because of immature immune system [31] but the prevalence is usually higher in adults because of longer periods of exposure. Further, Biraro et al employed QFT-Gold In-Tube compared to QFT-Plus employed in this study. The QFT-*Plus is* considered more sensitive than QFT-Gold In-Tube because of the inclusion of the TB2 antigen which also stimulates CD8+ T cells in addition to CD4+ T cells [32]. The prevalence did not differ between HIV positive and HIV negative participants, despite the heightened risk of TB infection for HIV positive individuals following exposure and for reactivation to active disease [33]. This was also contrary to the observation that IGRAs have sub-optimal performance in immunosuppressive conditions like HIV, particularly in those with reduced CD4 counts [34]. The good performance reflected here among the HIV positive could have resulted from majority of the HIV positive participants having a suppressed viral load, likening their immune response to that of the HIV negatives participants. We also acknowledge a higher HIV positivity among this cohort compared to other TB contacts cohort or to the general population. The HIV prevalence in household contacts was estimated at $13\%$ [35] and that in the general population at $5.8\%$ [36]. Close contact with an index TB patient is a strong risk factor for TB infection [37], and first degree relatives are particularly at a heightened risk [38]. In this study, a family relation and living in the same house as the index case were associated with a significant risk for a positive QFT-Plus result. This was similar to findings by Jung et al study among household contacts in South Korea, where living in the same room as the index patient was associated with a positive IGRA [39], and findings by Hermann et al that family relations including a diseased sibling or sexual partner were associated with an increased risk for positive IGRA [40]. Using IGRA as either a categorical or quantitative trait, sleeping in the same room as a symptomatic index patient or simply contact within household were associated with IGRA-positivity in Colombia [41]. The household risk of TB transmission was dependent on the level of crowding within the households [42]. On the contrary, Shanaube et al didn’t find any association between sleeping in close proximity to the index patient and risk of TB infection [43]. Further, other studies have suggested that in high burden settings, the risk of transmission in the household and the community may not differ [43, 44]. No clear explanation has been put forward for this observation but it may be due to multiple exposures from unidentified index patients in the community. Nutritional status, for which BMI may be a marker, can influence cell-mediated immunity and low BMI is associated with higher TB incidence [45, 46]. Further, a higher BMI has been shown to be protective against incident TB but the effect was not apparent with BMI >30 kg/m2 [47]. Furthermore, cachexia that can occur during active TB results from hypothalamic hormone regulation by peptide YY, grehlin and resistin [48]. In this study, for every 1 Kg/M2 increase in BMI, there was an associated $9\%$ increase in risk of a positive QFT-Plus result. Similar to our finding, Zhang et al showed that a BMI of > 28 Kg/M2 was a risk factor for a positive IGRA and the latent TB prevalence increased from $18.5\%$ in underweight to $23.7\%$ in obese participants [49]. Adipose tissue has previously been shown to harbor M.tb DNA during latent TB infection, with M.tb gaining entry into adipocytes through scavenger receptors and accumulating in cytoplasmic lipid droplets to become predominantly dormant in mature adipocytes [50]. Adiposity is associated with metabolic disturbances which result in defective innate and adaptive immune responses that might aid susceptibility to infections and reactivation of latent disease [51], including TB. The immune defects include reductions in leucocyte numbers, phagocytosis, oxidative burst and proliferative capacity following antigen stimulation [52]. These immune defects are mediated by immunomodulatory adipokines including leptin, adiponectin and pro-inflammatory cytokines: TNFα, IL-6 and IL-1β [52]. Tobacco smoking is associated with an increased risk for M. tuberculosis infection and progression to active disease. Exposure to tobacco smoke damages airway ciliary function and suppresses cytotoxic activity of NK cells and T cell function [53, 54]. For instance, in India, up to $40\%$ of TB infections were attributed to smoking [55] while in a study in Taiwan quitting smoking was associated with a more than $65\%$ decrease in mortality [56]. The World Health Organization estimates that $7\%$ of the TB cases in 2020 were attributable to smoking and lists it among the top five risk factors for TB that should be targeted for TB control through the 5A’s and 5R’s models, to ready and motivate patients for quitting smoking [57]. Risk of M. tuberculosis infection in occupations other than healthcare is understudied [58]. Manual jobs occur in environments likely to promote transmission of TB. For instance, dusty work environments such as mines, stony quarries, building sites, charcoal stores or other exposure to silica are associated with M. tuberculosis infection [59]. Manual workers are also migratory and more likely to work in congested work places, increasing their risk for exposure to M. tuberculosis. Smoking, alcohol and HIV, which independently are risk factors for M. tuberculosis infection and disease, are all more prevalent among manual laborers [60]. On the other hand, having no job is linked to a low socio-economic status, which is characterized by poor living conditions including crowding and poor ventilations. These conditions are associated with a heightened risk for TB transmission [38]. The strengths of our study included the use of IGRA for which immuno-reactivity, unlike TST, is constrained to M. tuberculosis complex-specific antigens, and in particular QFT-Plus, which contains the TB2 antigen that stimulates CD8+ T cells in addition to the CD4+ T cells. The CD8+ T cells responses are more associated with subclinical and active TB than in latent TB [61]. Secondly, we evaluated a large range of risk factors, which strengthened the study’s ability to test for confounders for IGRA positivity. The study also had some limitations. The cross sectional design employed does not indicate a causal relationship between IGRA positivity and the studied factors. Because the exposure and outcome are measured at the same time, the temporal relationship between the two could not be established. Another major limitation was the assumption that contacts in the community could not have been responsible for the M. tuberculosis infection in the study participants. This could have led to a higher estimate of risk of M. tuberculosis infection related to close contact. However, indoor exposure is a stronger risk factor for M. tuberculosis infection than outdoor exposure hence infections were more likely to be from the indoor exposure or close contacts. Secondly, we used CXR as the additional measure to exclude active TB, which is inferior to other digital imaging techniques like the computed tomography scan. Thirdly, we employed a diagnostic test which reflects immune sensitization that may lack specificity as compared to molecular biomarkers like the detection of M.tb DNA in CD34+ peripheral blood mononuclear cells [62] or transcriptional signatures for incipient TB [63], which may have higher PPV. Lastly, our study did not quantify the intensity of smoking in terms of pack years. As a result, the study did not comprehensively evaluate the risk of M. tuberculosis infection posed by smoking. ## Conclusion Interferon Gamma Release Assay positivity in this study population was lower than previously estimated. Tobacco smoking and BMI were determinants of IGRA positivity that were previously unappreciated. ## References 1. 1WHO, Global tuberculosis report 2021. [Internet], 2021. 2. Kiazyk S, Ball TB. **Latent tuberculosis infection: An overview**. *Can Commun Dis Rep* (2017.0) **43** 62-66. DOI: 10.14745/ccdr.v43i34a01 3. Campbell JR. **A systematic review on TST and IGRA tests used for diagnosis of LTBI in immigrants**. *Mol Diagn Ther* (2015.0) **19** 9-24. DOI: 10.1007/s40291-014-0125-0 4. Carranza C. **Diagnosis for Latent Tuberculosis Infection: New Alternatives**. *Front. Immunol* (2020.0). DOI: 10.3389/fimmu.2020.02006 5. Pai M, Denkinger CM, Kik SA. **Gamma Interferon Release Assays for Detection of Mycobacterium tuberculosis Infection**. *Clin Microbiol Rev* (2014.0) **27** 3-20. DOI: 10.1128/CMR.00034-13 6. 6QIAGEN, QuantiFERON-TB Gold Plus (QFT-PLUS), ELISA Package Insert [Internet], 2015 7. Gupta RK. **Evaluation of QuantiFERON-TB Gold Plus for Predicting Incident Tuberculosis among Recent Contacts: A Prospective Cohort Study**. *Ann Am Thorac Soc* (2020.0) **17** 646-650. DOI: 10.1513/AnnalsATS.201905-407RL 8. Boon SD. *AIDS* (2016.0) **30** 797-801. PMID: 26730567 9. Lee S. **Factors Associated with a Positive Interferon-Gamma Release Assay (IGRA) Result in 741 Participants**. *PLOS ONE* (2015.0) 10. Yapa P. **Prevalence of and risk factors associated with latent tuberculosis in Singapore: A cross-sectional survey**. *International Journal of Infectious Diseases* (2018.0) **72** 55-62. PMID: 29758278 11. LaCourse SM. **Effect of pregnancy on interferon gamma release-assay and tuberculin skin test detection of latent TB infection among HIV-infected women in a high burden setting**. *J Acquir Immune Defic Syndr* (2017.0) **75** 128-136. DOI: 10.1097/QAI.0000000000001298 12. Faurholt-Jepsen D. **Diabetes is associated with lower tuberculosis antigen-specific interferon gamma release in Tanzanian tuberculosis patients and non-tuberculosis controls**. *Scand J Infect Dis* (2014.0) **46** 384-91. DOI: 10.3109/00365548.2014.885657 13. Bongomin F. **Latent Tuberculosis Infection Status of Pregnant Women in Uganda Determined Using QuantiFERON TB Gold-Plus**. *Open Forum Infect Dis* (2021.0) **8**. DOI: 10.1093/ofid/ofab241 14. Santosa JA, Duartec R, Nunesa C. **Host factors associated to false negative and indeterminate results in an interferon-γ release assay in patients with active tuberculosis**. *Pulmonary journal* (2020.0) **26** 353-362 15. Banfield S. **Factors Associated with the Performance of a Blood-Based Interferon-γ Release Assay in Diagnosing Tuberculosis**. *PLoS One* (2012.0) **7** e38556. PMID: 22701664 16. Yamasue M. **Factors associated with false negative interferon-γ release assay results in patients with tuberculosis: A systematic review with meta-analysis**. *Sci Rep* (2020.0) **10** 1607. PMID: 32005930 17. Sabri A. **Prevalence and risk factors for latent tuberculosis infection among healthcare workers in Morocco**. *PLoS One* (2019.0) **14** e0221081. DOI: 10.1371/journal.pone.0221081 18. Sharma SK. **Comparison of TST and IGRA in Diagnosis of Latent Tuberculosis Infection in a High TB-Burden Setting**. *PLoS One* (2017.0) **12** e0169539. DOI: 10.1371/journal.pone.0169539 19. Lou JK, Okot-Nwang M, Katamba A. **Prevalence of positive tuberculin skin test and associated factors among Makerere medical students, Kampala, Uganda**. *Afr Health Sci* (2015.0) **15** 1247-1255. DOI: 10.4314/ahs.v15i4.25 20. Koppaka VR. **Risk Factors Associated with Tuberculin Skin Test Positivity among University Students and the Use of Such Factors in the Development of a Targeted Screening Program**. *Clinical Infectious Diseases* (2003.0) **36** 599-607. DOI: 10.1086/367664 21. MacPherson p. **Prevalence and risk factors for latent tuberculosis infection among household contacts of index cases in two South African provinces: Analysis of baseline data from a cluster-randomised trial**. *PLOS ONE* (2020.0) **15** e0230376. DOI: 10.1371/journal.pone.0230376 22. Stein CM, Zalwango S, M LL. **Resistance and Susceptibility to Mycobacterium tuberculosis Infection and Disease in Tuberculosis Households in Kampala, Uganda**. *Am J Epidemiol* (2018.0) **187** 1477-1489. DOI: 10.1093/aje/kwx380 23. Chapman ALN. **Rapid detection of active and latent tuberculosis infection in HIV-positive individuals by enumeration of Mycobacterium tuberculosis-specific T cells**. *AIDS* (2002.0) **16** 2285-93. DOI: 10.1097/00002030-200211220-00008 24. Markowitz N. **Tuberculin and anergy testing in HIV-seropositive and HIV-seronegative persons. Pulmonary Complications of HIV Infection Study Group**. *Ann Intern Med* (1993.0) **119** 185-93. DOI: 10.7326/0003-4819-119-3-199308010-00002 25. Birku M. **Effect of pregnancy and HIV infection on detection of latent TB infection by Tuberculin Skin Test and QuantiFERON-TB Gold In-Tube assay among women living in a high TB and HIV burden setting**. *International Journal of Infectious Diseases* (2020.0) **101** 235-242. DOI: 10.1016/j.ijid.2020.09.1452 26. Kim EY. **Performance of the tuberculin skin test and interferon-γ release assay for detection of tuberculosis infection in immunocompromised patients in a BCG-vaccinated population**. *BMC Infect Dis* (2009.0) **9** 27. Graham SM. **Guidance for National Tuberculosis Programmes on the management of tuberculosis in children—an update**. *Malawi Med J* (2007.0) **19** 82-86. PMID: 23878641 28. 28Health, M.o., CONSOLIDATED GUIDELINES FOR THE PREVENTION AND TREATMENT OF HIV AND AIDS IN UGANDA. [Internet], 2020 Feb: p. 30. 29. Van Voorhis CRW. **Understanding Power and Rules of Thumb for Determining Sample Size**. *Tutorials in Quantitative Methods for Psychology* (2007.0) **3** 30. Biraro IA. **The Use of Interferon Gamma Inducible Protein 10 as a Potential Biomarker in the Diagnosis of Latent Tuberculosis Infection in Uganda**. *PloS One* (2016.0) **11** e0146098. DOI: 10.1371/journal.pone.0146098 31. Nakaoka H. **Risk for Tuberculosis among Children**. *Emerg Infect Dis* (2006.0) **12** 1383-1388. DOI: 10.3201/eid1209.051606 32. Barcellini L. **First independent evaluation of QuantiFERON-TB Plus performance**. *European Respiratory Journal* (2016.0) **47** 1587-1590. DOI: 10.1183/13993003.02033-2015 33. Ji Y. **Risk factors affecting the mortality of HIV-infected patients with pulmonary tuberculosis in the cART era: a retrospective cohort study in China**. *Infectious Diseases of Poverty* (2018.0) **7** 34. Redelman-Sidi G, Sepkowitz KA. **IFN-γ Release Assays in the Diagnosis of Latent Tuberculosis Infection among Immunocompromised Adults**. *American Journal of Respiratory and Critical Care Medicine* (2012.0) **188** 35. Guwatudde D. **Tuberculosis in Household Contacts of Infectious Cases in Kampala, Uganda**. *Am J Epidemiol* (2003.0) **158** 887-898. DOI: 10.1093/aje/kwg227 36. 36ICAP, UGANDA POPULATION-BASED HIV IMPACT ASSESSMENT. [Internet], Aug 2022. 37. Ghanaiee RM. **Household contact investigation for the detection of active tuberculosis and latent tuberculosis: A comprehensive evaluation in two high-burden provinces in Iran**. *New Microbes and New Infections* (2022.0) **45** 100958. DOI: 10.1016/j.nmni.2022.100958 38. Muchuro S. **Integrating interferon-gamma release assay testing into provision of tuberculosis preventive therapy is feasible in a tuberculosis high burden resource-limited setting: A mixed methods study**. *Plos Global Public Health* (2022.0) 39. Jung Seop E. **Household tuberculosis contact investigation in a tuberculosis-prevalent country**. *Medicine* (2018.0) **97** e9681. PMID: 29505017 40. Herzmann C. **Risk for latent and active tuberculosis in Germany**. *Infection* (2017.0) **45** 283-290. DOI: 10.1007/s15010-016-0963-2 41. del Corral H. **IFNγ Response to Mycobacterium tuberculosis, Risk of Infection and Disease in Household Contacts of Tuberculosis Patients in Colombia**. *PLoS One* (2009.0) **4** e8257. PMID: 20011589 42. Boccia D. **Tuberculosis Infection in Zambia: The Association with Relative Wealth**. *Am J Trop Med Hyg* (2009.0) **80** 1004-1011. PMID: 19478266 43. Shanaube K. **Risk Factors Associated with Positive QuantiFERON-TB Gold In-Tube and Tuberculin Skin Tests Results in Zambia and South Africa**. *PloS One* (2011.0) **6** e18206. DOI: 10.1371/journal.pone.0018206 44. Boccia D. **Tuberculosis infection in Zambia: the association with relative wealth**. *Am J Trop Med Hyg* (2009.0) **80** 1004-1011. PMID: 19478266 45. Cegielski P, McMurray DN. **The relationship between malnutrition and tuberculosis: evidence from studies in humans and experimental animals**. *Int J Tuberc Lung Dis* (2004.0) **8** 286-98. PMID: 15139466 46. Lönnroth K. **A consistent log-linear relationship between tuberculosis incidence and body mass index**. *International Journal of Epidemiology* (2010.0) **39** 149-155. DOI: 10.1093/ije/dyp308 47. Kim SJ. **Association of body mass index with incident tuberculosis in Korea**. *PLOS ONE* (2018.0) **13** e0199065. PMID: 29883477 48. Chang SW. **Gut Hormones, Appetite Suppression and Cachexia in Patients with Pulmonary TB**. *PLoS One* (2013.0) **8** e54564. DOI: 10.1371/journal.pone.0054564 49. Zhang H. **Association of Body Mass Index with the Tuberculosis Infection: a Population-based Study among 17796 Adults in Rural China**. *Sci Rep* (2017.0) **7** 41933. DOI: 10.1038/srep41933 50. Neyrolles O. **Is adipose tissue a place for Mycobacterium tuberculosis persistence?**. *PLoS One* (2006.0) **1** e43. DOI: 10.1371/journal.pone.0000043 51. Mancuso P. **Obesity and respiratory infections: Does excess adiposity weigh down host defense?**. *Pulm Pharmacol Ther* (2013.0) **26** 412-419. DOI: 10.1016/j.pupt.2012.04.006 52. Milner JJ, Beck MA. **Micronutrients, immunology and inflammation The impact of obesity on the immune response to infection**. *Proc Nutr Soc* (2012.0) **71** 298-306. PMID: 22414338 53. Fayez EA. **The effect of smoking on latent tuberculosis infection susceptibility in high risk individuals in Iran**. *J Immunoassay Immunochem* (2020.0) **41** 885-895. DOI: 10.1080/15321819.2020.1806075 54. Buskin SE. **Tuberculosis risk factors in adults in King County, Washington, 1988 through 1990**. *Am J Public Health* (1994.0) **84** 1750-1756. DOI: 10.2105/ajph.84.11.1750 55. Lönnroth K, Raviglione M. **Global epidemiology of tuberculosis: prospects for control**. *Semin Respir Crit Care Med* (2008.0) **29** 481-91. DOI: 10.1055/s-0028-1085700 56. Wen C. **The reduction of tuberculosis risks by smoking cessation**. *BMC Infect Dis* (2010.0) **10** 156. DOI: 10.1186/1471-2334-10-156 57. 57WHO, Global tuberculosis report 2022. [Internet], 2022: p. 1–24. 58. Adams S. **Evaluating Latent Tuberculosis Infection Test Performance Using Latent Class Analysis in a TB and HIV Endemic Setting**. *Int J Environ Res Public Health* (2019.0) **16** 2912. DOI: 10.3390/ijerph16162912 59. Gottesfeld P. **Preventing tuberculosis among high-risk workers**. *Lancet Global Health* (2018.0) **6** E1274-E1275. DOI: 10.1016/S2214-109X(18)30313-9 60. Chen D. **Efficacy of HIV interventions among factory workers in low- and middle-income countries: a systematic review**. *BMC Public Health* (2020.0) **20** 1310. DOI: 10.1186/s12889-020-09333-w 61. Rozot V. **Mycobacterium tuberculosis-specific CD8+ T cells are functionally and phenotypically different between latent infection and active disease**. *European Journal of Immunology* (2013.0) **43** 1568-1577. DOI: 10.1002/eji.201243262 62. Belay M. **Detection of Mycobacterium tuberculosis complex DNA in CD34-positive peripheral blood mononuclear cells of asymptomatic tuberculosis contacts: an observational study**. *The Lancet* (2021.0) **2** E267-E275. DOI: 10.1016/S2666-5247(21)00043-4 63. Suliman S. **Four-Gene Pan-African Blood Signature Predicts Progression to Tuberculosis**. *Am J Respir Crit Care Med* (2018.0) **197** 1198-1208. DOI: 10.1164/rccm.201711-2340OC
--- title: Ethnobotanical survey of medicinal plants used by indigenous knowledge holders to manage healthcare needs in children authors: - Peter Tshepiso Ndhlovu - John Awungnjia Asong - Abiodun Olusola Omotayo - Wilfred Otang-Mbeng - Adeyemi Oladapo Aremu journal: PLOS ONE year: 2023 pmcid: PMC10042359 doi: 10.1371/journal.pone.0282113 license: CC BY 4.0 --- # Ethnobotanical survey of medicinal plants used by indigenous knowledge holders to manage healthcare needs in children ## Abstract Childhood diseases remain an increasing health problem in many developing countries and also associated with an enormous financial burden. In South Africa, many people still rely on traditional medicine for their primary healthcare. However, inadequate documentation of medicinal plants used to manage childhood diseases remain a prominent gap. Hence, the current study explored the importance of medicinal plants to treat and manage childhood diseases in the North West Province, South Africa. An ethnobotanical survey was conducted with 101 participants using semi-structured interviews (face-to-face). Ethnobotanical indices such as Frequency of citation (FC), Use-value (UV) and Informed Consensus Factor (ICF) were used for data analysis. A total of 61 plants from 34 families were recorded as medicine used for managing seven [7] categories of diseases resulting from 29 sub-categories. Skin-related and gastro-intestinal diseases were the most prevalent childhood health conditions encountered by the study participants. Based on their FC values that ranged from approximately 0.9–$75\%$, the most popular medicinal plants used by the participants were *Aptosinum elongatum* ($75.2\%$), *Commelina diffusa* ($45.5\%$), *Euphorbia prostrata* ($31.6\%$) and Bulbine frutescens ($31.7\%$). In terms of the UV, A. elongatum (0.75), C. diffusa (0.45), E. prostrata (0.31), H. hemerocallidea (0.19) and E. elephantina (0.19) were the dominant plants used for treating and managing childhood diseases. Based on ICF, skin-related diseases dominated with the highest ICF value of 0.99. This category had 381 use-reports, comprising 34 plants ($55.7\%$ of total plants) used for childhood-related diseases. Particularly, B. frutescens and E. elephantina were the most-cited plants for the aforementioned category. Leaves ($23\%$) and roots ($23\%$) were the most frequently used plant parts. Decoctions and maceration were the main preparation methods, and the plant remedies were mainly administered orally ($60\%$) and topically ($39\%$). The current study revealed the continuous dependence on the plant for primary health care relating to childhood diseases in the study area. *We* generated a valuable inventory of medicinal plants and associated indigenous knowledge for child healthcare needs. However, investigating the biological efficacies, phytochemical profiles and the safety of these identified plants in relevant test systems remain essential in future research. ## 1. Introduction As common with many middle-income countries, South *Africa is* still facing a high child/infant mortality rate [1]. Even though the current new-born mortality rates are within the United Nations Sustainable Development Goal (SDG-3) target of 12 deaths per 1000 live births, the absolute number of deaths is unacceptably high for South Africa [2]. The current infant mortality rate for South Africa in the first quarter of 2022 was 24.3 deaths per 1000 live births, which is a $2.93\%$ decline from 2021 [3]. The overwhelming tide of emerging epidemics and pandemics has increased the strain on the healthcare system. It has also affected resources often allocated for the prevention of diseases and the active promotion of children health, especially in rural areas [4–6]. Even though South Africa has several databases that collect information on neonatal deaths and prevalent diseases, most of these focus on deaths occurring within healthcare facilities [2]. Outside such facilities, the number of deaths is largely unknown, child mortality, compared to the 12 per 1000 reported from the District Health Information *System data* [7]. Globally, the existence of traditional medicine depends fundamentally on the rich diversity of plants and the related knowledge of their use as herbal therapy [8]. These medicinal plants remain indispensable for maintaining health and well-being among several ethnic groups [9–11]. The on-going dire economic situations in most rural areas and limited access to conventional medicine have increased the popularity of traditional medicine [12]. In addition, the use of medicinal plants for childhood diseases in rural areas have been receiving an increased attention among researchers [13–15]. South *Africa is* endowed with a rich wealth of flora and is often acclaimed as a biodiversity hotspot [16]. An estimated 30000 plants are used for traditional medicine for the management of diverse health conditions [13, 17, 18]. In South Africa, the significance of medicinal plants as remedies against many diseases among local communities are well-recognised [19–22]. However, the need to develop a comprehensive database for medicinal plants used for the management and treatment of childhood diseases cannot be overemphasized. In addition, the increasing rate of urbanization and habitat destruction that has a detrimental effect on plant resources especially medicinal plants, remain a global challenge [23]. A recent review by Ndhlovu et al. [ 13] revealed the dearth of scientific investigation on medicinal plants used to treat and manage childhood disease in North West Province of South Africa. Thus, the current study explored the indigenous knowledge and medicinal plants used to manage and treat childhood diseases among local communities in North West Province, South Africa. ## 2.1. Description of the study area The study was conducted in 17 selected communities in the Ngaka Modiri Molema and Bojanala districts in the North West Province, South Africa (Fig 1 and Table 1). We prepared the map of the study area using the free software QGIS version 3.22.14 Bilowieza (available at: www.qgis.org). **Fig 1:** *Selected district municipalities (Ngaka Modiri Molema and Bojanala) in the North West Province, South Africa.The map was prepared using the free software QGIS version 3.22.14 Bilowieza (available at: www.qgis.org). This figure is licensed under Creative Commons Attribution License (CCAL), CC BY 4.0.* TABLE_PLACEHOLDER:Table 1 The selected district lies between latitude 22° South and longitude 28° East of the North West Province, covering 116 320 km2 which is about $9.5\%$ of the total surface area of South Africa. North West Province shares boundaries with four other provinces, namely Northern Cape, Free State, Gauteng and Limpopo provinces [24, 25]. The average annual rainfall is about 360 mm, mostly experienced in the summer months between October and April, while summer temperatures range from 17 to 31°C and winter temperature ranges from 3 to 21°C [26]. Ngaka Modiri Molema and Bojanala Platinum districts were selected due to their high biodiversity and economic activities. Furthermore, the population consist of $94\%$ Black Africans, and *Setswana is* the most spoken language. North West *Province is* one of nine provinces in South Africa and is an important contributor to the South African economy mainly through agricultural and mining activities [27, 28]. According to Stats SA [28], the public health system in North West Province has been in a state of crisis for many years. Particularly, the persistent and widespread medicine stock-outs and shortages remain common in the province. The limited supply of medicine and other pharmaceuticals has resulted in patients visiting the facilities for health care. Furthermore, most of health facilities were built on racial lines during the apartheid era. As a result, there are still challenges of inequitable distribution of health facilities and human capital resources, with a skewed spread of medical officers, professional nurses and allied health professionals across the districts [25]. ## 2.2 Data collection An ethnobotanical survey was conducted between autumn and summer from April to August 2021. In addition, from April 2021-April 2022, follow-up trips were conducted with the key participants. This was done to keep updating participants and to explore whether they could add new data to enrich the study. Prior to the collection of ethnobotanical data, an overview of the focus and the significance of the study was shared with the participants before obtaining their consent to participate. After the engagement, their consent to participate was requested for the study. A total of 101 participants were interviewed in two districts of the North West Province. The data was collected through semi-structured interviews (face-to-face) using Setswana, a widely spoken local language in the study area. These semi-structured interviews were designed to record information about the prevalence of diseases, plants used to manage and treat childhood diseases, plant parts used, and methods of preparation and administration. As a visual aid to assist with identifying diseases among the participants, photographs of known childhood diseases were extracted and compiled from reliable sources [29, 30]. This assisted the participants with the identification of childhood diseases during the survey. ## 2.3 Sampling technique The study involved 101 participants such as traditional health practitioners (specifically those with expertise in managing and treating diseases among children) and herbal vendors in the selected study areas. Participants over 18 years old were selected because the research relies only on adults and knowledge experts. Purposive expert sampling was an advantageous technique because of the ability to generate greater knowledge depth in the field of interest [31, 32]. In addition, participants were selected based on their ability to speak and understand Setswana or English. According to Stats SA [28], both languages are the two main languages on the study sites. The participants of both genders registered with the North West Dingaka Association (regulatory body in the province), whose names also appear in the provincial Traditional Health Practitioners (THPs) database, were targeted. The participants who met the eligibility criteria of the various categories of THPs as defined in the Traditional Health Practitioners Act 22 of 2007 included Traditional Doctors, Diviners (Sangomas), Traditional Surgeons, Traditional Birth Attendants (TBAs) and Herbalists [33]. The study excluded individuals under 18 years as they are considered minors in South Africa and cannot give consent on their own [34]. Due to limited resources, participants from Dr Kenneth Kaunda and Dr Ruth Segomotsi Mopati of the North West Province were excluded. Furthermore, the participants who were not registered with the North West Dingaka Association were excluded. Creswell and Creswell [35] indicated that the exclusion criteria include factors or characteristics that make the recruited population ineligible for a study. Exclusion requirements are a series of predefined meanings used to classify participants who will not be included or will have to withdraw from the research study after they have been included [36]. Upon the agreement by a participants, a digital voice recorder was used to capture the interviews and create an audio pool of information. In addition, photographs of each of the plant species were captured. For botanical identification, voucher specimens for all the plants were collected during the field study and deposited in the herbarium of the South African National Biodiversity Institute (SANBI), Pretoria, South Africa. Furthermore, photos were taken for identification as a conservation mechanism to prevent or reduce the risk of extinction for plants considered as threatened or facing extinction. Botanical names of the collected plant species were identified by an expert SANBI using a detailed regional dichotomous key [37]. ## 2.4 Data analysis Descriptive statistics (mean, frequency and percentage) were used to identify and describe the socio-demographic parameters of the participants [38]. Data were computed and entered into an Excel office sheet, 2016, then exported into Statistical Package for Social Sciences (SPSS version 27). Based on the previous ethnobotanical indices [39, 40], three quantitative parameters: frequency of citation (FC), use-value (UV), and informant consensus factor (ICF), was used to analyse the data. ## 2.4.1 Frequency of citation (FC) Based on the study of Trotter and Logan [41], the frequency of citation (FC) of the plant species was calculated as follows: FC=NpNx100 [1] Where Np = number of times a particular species was mentioned; N = total number of times that all species were mentioned x 100 ## 2.4.2 Use-value (UV) The use-value of a plant species is a measure of the relative importance of how the species is known locally. It ranks the plants according to the number of uses mentioned for a particular species and the number of participants who mentioned the use of the species [39, 40, 42]. This was calculated using the formula below UV=ΣUiN [2] where: UV = use-value of a species; U = number of citations per species; n = number of informants. When the UV is high (closer to 1), it means the species is highly used as medicine by the participants, and when it is very low (closer to 0), it means the plant has a few medicinal uses reported by the participants. ## 2.4.3 Informed consensus factor (ICF) This value was calculated for the different categories of diseases to ascertain the degree of homogeneity of the knowledge among the participants in the study area with respect to the treatment of childhood diseases and the use of plants per disease category [43]. The ICF was calculated using the formula below: ICF=Nur−NtNur−1 [3] Where Nur = number of use citations in each category; nt = number of plants used. All citations were placed into one of the seven categories: Pain and inflammation-related diseases, gastro-intestinal related diseases, oral-related diseases, skin-related diseases, respiratory-related diseases, urinary-related diseases and social/nutritional related diseases. ## 2.5 Ethical approval Prior to conducting the study, ethical clearance (NWU-00485-20-A1) was obtained from the North-West University Health Research Ethics Committee (NWU-HREC), while the permit to collect plant species was obtained from the North West Department of Economic Development, Environment, Conservation and Tourism (ID NW $\frac{27370}{10}$/2020). The local authorities, including traditional leaders (Barolong bo Rra Tshidi), granted us the permission to access the study sites, while the participants signed the informed consent form. This study focused on the knowledge holders who were over the age of 18 and were able to give consent, as the South African laws recognized a person over the age of 18 as an adult. The informed consent form described the terms and conditions that both the researcher and participants needed to follow. For instance, the names of the participants remaining private or anonymous throughout the study, the right to withdraw whenever they feel uncomfortable proceeding with the interview. The ethnobotanical survey were conducted in accordance with the Declaration of Helsinki. In compliance with the Nagoya Protocol on Access to Genetic Resources and the Fair and Equitable Sharing of Benefits Arising from their Utilization to the Convention on Biological Diversity, the study participants from the 17 selected communities in the Ngaka Modiri Molema and Bojanala districts retain the authorship of traditional knowledge documented in this publication. Therefore, any use of the documented information, other than for scientific publications, requires prior consent of the traditional knowledge holders and their agreement on access to benefits resulting from any commercial use. ## 3. Results and discussion Traditional healing using medicinal plants as the core ingredient is well-enriched among many African communities, including the Batswana people [44, 45]. In South Africa, remote areas previously disadvantaged during the former political systems have limited access to modern medicine [28]. Medications, especially for children health problems, are sparingly available to worsen the situation. The dependance and reliability on plants in traditional medicine are unquestionable and taken into cognizance in the current study. ## 3.1 Socio-economic characteristics Table 2 summaries the demographic characteristics of the participants, including THPs and herbal vendors. Gender distribution among the participants was $78\%$ female and $21\%$ male. This signifies the importance of females as active custodians of indigenous knowledge related to childhood health needs. This is also an indication that females play an essential role in managing households in the North West Province. In a similar manner, finding by Omotayo et al. [ 46] reported that the dominance ($56\%$) of female-headed households in the North West Province. In terms of marital status, the majority ($53.5\%$) were single, while $2\%$ of the participants were widowed. **Table 2** | Parameter | Frequency | Percentage (%) | | --- | --- | --- | | Age (years) | | | | 17–37 | 21 | 20.8 | | 38–58 | 53 | 52.5 | | 59–79 | 27 | 26.7 | | Gender | | | | Male | 22 | 21.8 | | Female | 79 | 78.2 | | Marital status | Marital status | Marital status | | Married | 35 | 34.7 | | Single | 54 | 53.5 | | Divorced | 2 | 1.9 | | Widowed | 8 | 7.9 | | Other(s) | 2 | 1.9 | | Educational status | | | | Informal education | 22 | 21.8 | | Primary education | 5 | 5.0 | | Standard education | 64 | 63.3 | | Post-graduate level | 10 | 9.9 | | Household size | | | | Less than or equal to four | 57 | 56.4 | | More than four | 44 | 43.6 | | Type of location | | | | Villages | 80 | 79.2 | | Township | 16 | 15.8 | | Urban area | 5 | 4.9 | | Type of practice | | | | Diviner | 13 | 12.8 | | Herbalist | 24 | 23.9 | | Herbalist and diviner | 38 | 37.6 | | Herbal vendor | 20 | 19.8 | | Traditional birth attendant | 6 | 5.9 | In the current study, the formal education status was high among the knowledgeable participants about herbal-based remedies used to treat and manage childhood diseases. For example, $63\%$ of the participants completed a secondary level of education, $21.8\%$ did not receive formal education and $5\%$ attended primary school. Education plays a crucial role in the socio-economic status of participants and their consciousness of environmental and health-related issues [47–49]. In addition, the household size of the participants revealed that the majority ($56\%$) consisted of less than 5 individuals (Table 2). These findings are consistent with Stats SA [28] and Nkonki-Mandleni et al. [ 50], which reported an average household size of three people in 2011 and 2016, with a decrease in household size due to a lack of employment and urban migration [27]. Although $79\%$ of the participants resided in villages, $15.8\%$ were based in urban areas. A greater proportion of the participants with knowledge on childhood diseases were THPs such as diviners and herbalists. ## 3.2 Overview of the generated plant inventory used for treating childhood diseases The use of plants continue to play a vital role in treating childhood diseases in South Africa, especially in rural areas [13]. In this study, 61 plant species belonging to 34 families were recorded as herbal remedies to treat and manage 29 childhood diseases (Table 3; Fig 2). The number of plants used for childhood diseases was relatively higher compared to 44 plants documented in Nigeria [14] and 50 plants in Uganda [51]. **Fig 2:** *Frequency of plant families used to manage and treat childhood diseases in Ngaka Modiri Molema and Bojanala districts of North West Province, South Africa.* TABLE_PLACEHOLDER:Table 3 Asteraceae [7] and Fabaceae [7] were the most dominant plant families, followed by Solanaceae [4] and Asparagaceae [4], while each of the remaining 30 families were represented by 1–3 plants per family (Fig 2). The dominant botanical families especially the Fabaceae, are known to have the highest number of species and also have the highest number of medicinal plant families in the North West Province [52]. The dominance of Fabaceae as one of the most preferred plant families has been similarly recorded in many ethnobotanical surveys in South Africa [53–55]. In Southern Africa, the importance of *Asteraceae is* well-documented [56]. Members of Asteraceae are widely recognized as weeds in anthropogenic contexts and are among the first species to sprout in the field after the soil has been prepared for planting [57–60]. Furthermore, *Asteraceae is* the most varied and cosmopolitan family of flowering plants, and they are good sources of inulin, a natural polysaccharide with prebiotic solid properties and potent antioxidant, anti-inflammatory, and antimicrobial activity [59, 61]. This could explain the high number of citations for plants in this family in many ethnobotanical studies in rural communities, where these plants from Asteraceae are frequently used in traditional medicine. Several studies have suggested that using plant species from the aforementioned families may be related to the effectiveness of bioactive ingredients against diseases and active biological compounds [42, 60]. ## 3.3 Frequency of citations and use-value pattern for the documented plants The FC values ranged from approximately 0.9–$75\%$, and the top 10 most frequently cited medicinal plants used by the participants were Aptosinum elongatum, Euphorbia prostrata, Commelina diffusa, Bulbine frutescens, Elephantorrhiza elephantina, Harpagophytum procumbens, Hypoxis hemerocallidea, Solanum lichtensteinii, *Siphonochilus aethiopicus* and *Solanum campylacanthum* (Table 3). In terms of the UV, the documented medicinal plants had values that ranged from 0.01–0.75 (Table 3). The most used medicinal plant was A. elongatum, with a UV of 0.75. The plant was mentioned by 76 knowledge holders as treatment for diverse health conditions such as umbilical cord, muscle fits, measles, bladder inflammation, weight loss and appetite. As another example of a well-utilised plant, C. diffusa had a UV of 0.44 with indication as an herbal remedy to treat and manage umbilical cord, purgative in children, while E. prostrata (UV = 0.31) was used to treat stomachaches, fever, and abdominal cramps. Hypoxis hemerocallidea (UV = 0.19) was used for sunken fontanelle, bladder inflammation, kidney failure, urinary tract infection, bronchitis pneumonia, child cleanse influenza and ulcer, gastro-intestinal and appetite. In addition, E. elephantina (UV = 0.17) was used to treat infective eczema, diarrhoea, ulcers, burns, and measles. These findings demonstrate the extensive use of these aforementioned plants in treating various ailments by local inhabitants/healers and the consciousness of indigenous peoples, which makes such medicinal plants the first choice to treat disease [63]. Most THPs undergo common training practices, hence the popularity of the plants among local traditional health practitioners as a treatment preference. Many of the plants with high UV are generally used to treat different diseases in different communities [64–66]. Asong et al. [ 45] reported the use of various plants from North West Province, which included A. elongatum, E. prostrata, H. hemerocallidea and E. elephantina. For instance, the whole plant of A. elongatum is used to treat chickenpox and yaws [45]. Commelina diffusa is prescribed for treating urinary tract infections, swellings, inflammation, diarrhoea, hemorrhoids, enteritis, eye irritation, conjunctivitis, and ophthalmia [67]. It is also used to treat jaundice in children in the Windward Islands and Cuba [68]. On the other hand, E. prostrata, has been used to treat skin problems, migraines, intestinal parasites, and warts as a medicinal herb [69]. Hypoxis hemerocallidea also known as ’African potato’, is widely used in South African traditional medicine to cure, manage various health conditions including childhood convulsions and epilepsy [70]. In the study by Mhlongo et al. [ 71], H. hemerocallidea was identified among the most cited plants with the highest priority. Elephantorrhiza elephantina is used for gastrointestinal diseases, respiratory ailments, pain and inflammation [72]. Factors such as availability and indigenous success may influence the popular use of these medicinal plants within the cultural folklore and primary health care amongst the local communities [73, 74]. Furthermore, Phillips et al. [ 40] hypothesized that a high UV indicates the cultural diversity attached to the plants. Though unknown to the knowledge holders, there is the possible presence of biologically active compounds in these plants that could be responsible for the perceived healing effects, which may be worth investigating. Notably, plants with lower UV, such as *Prunus persica* (UV = 0.9) and Aloe arborescens (UV = 0.01), should not be ignored because local healers indicated that these plants exert potent effects [75]. According to some knowledge holders, some plants with low UV were in high use in the past. However, the continuous destruction of habitats resulting in the scarcity necessitate the search for alternative plants. ## 3.4 Informant census factors The categories of childhood diseases and informant consensus factors (ICF) are shown in Table 4. A total of 29 different childhood diseases are being treated and managed with 61 medicinal plants documented in the study area. The seven ailment categories identified were pain and inflammation diseases, gastro-intestinal related diseases, oral-related diseases, urinary-related diseases, skin-related diseases, respiratory-related diseases, and social/nutritional-related diseases. The ICF of different categories of childhood diseases ranged from 0.62–0.99. Skin-related diseases had the highest ICF value (0.99). This category had 381 use-reports, comprising 34 plants ($55\%$ of total plants) used for childhood diseases (Table 4). The most cited plants were B. frutescens and E. elephantina. This indicates a high consensus on the use reports of these medicinal plants. Asong et al. [ 45] reported using these two plants for skin-related diseases with a high UV. **Table 4** | Category | No of use reports (Nur) | No of plant taxa (Nt) | ICF | | --- | --- | --- | --- | | Gastro-intestinal diseases | 242.0 | 18.0 | 0.92 | | Diarrhoea | | | | | Constipation | | | | | Gastroenteritis | | | | | Pain and inflammation | 127.0 | 10.0 | 0.9 | | Umbilical cord | | | | | Pain and inflammation | | | | | Earache | | | | | Urinary genital diseases | 190.0 | 17.0 | 0.9 | | Bladder inflammation | | | | | Kidney failure | | | | | Urinary tract infection | | | | | Oral-related diseases | 61.0 | 7.0 | 0.9 | | Oral blisters | | | | | Phlegm | | | | | Stop vomiting | | | | | Teething | | | | | Skin-related diseases | 381.0 | 34.0 | 0.99 | | Body rash | | | | | Skin irritation | | | | | Sores | | | | | Impetigo | | | | | Infective eczema | | | | | Measles | | | | | Chicken pox | | | | | Ringworm and warts | | | | | Respiratory diseases | 42.0 | 15.0 | 0.62 | | Pneumonia | | | | | Bronchitis | | | | | Influenza | | | | | Tuberculosis | | | | | Social/nutritional health conditions | 90.0 | 34.0 | 0.65 | | Sunken fontanelle | | | | | Restless and weaning | | | | | Appetite, enhance child growth and gain weight | | | | | Sugar diabetes and cholesterol | | | | The second highest ICF (0.92) was linked to gastrointestinal-related diseases. This category had 242 uses reports and three gastrointestinal-related diseases; with diarrhoea being the most cited. Diarrhoea is the most common clinical manifestation of gastrointestinal diseases and can be caused by both infectious and non-infectious agents [76]. This disease may be abrupt and self-limiting in immune-competent individuals, especially in people with underlying debilitating clinical conditions such as HIV/AIDS and diabetes mellitus. Stats SA [28] indicated that the North West Province suffers from inadequate and lack of access to clean and safe water, as a result this maybe one of the major vehicles for transmission of gastrointestinal-related diseases. The most cited plant in this category was E. elephantina. There was a wide distribution of the plant taxa in this category. Moreover, the high ICF value indicates an agreement among the participants regarding the plant taxa used for treatment and possibly, a high prevalence of diseases in this category [77]. Respiratory-related diseases category recorded the lowest ICF value (Table 4), indicating that there was less knowledge exchange amongst the participants on medicinal plants used to treat and manage this category of diseases. This category had 47 use reports, 25 plant species and 3 respiratory-related diseases. This could be influenced by other respiratory-related diseases such as influenza have known remedies that are used to treat and manage this type of disease is common knowledge and are also based on individuals’ beliefs. According to Gazzaneo et al. [ 42], ICF values range from 0 to1 with values closer to 0, meaning that there is a lesser degree of agreement among the participants on the plants used to treat and manage diseases in a particular category. The lower ICF in this category indicates a lower degree of agreement among the participants regarding the medicinal plants used in the treatment and management of respiratory-related diseases in the study area. These low ICF values recorded in the present study could be ascribed to the recent trends in emerging diseases, including Covid-19 [78]. Besides, the low ICF values for respiratory-related diseases could be explained by the fact that these diseases were not essential health problems at that time of the data collection. However, these types of diseases, mainly skin-related disease (Infective eczema) and gastrointestinal (constipation, diarrhoea), are commonly referred to as THPs and are generally treated and managed with polyherbal medicines; thus, a range of medicinal plants was reported (Table 4). Furthermore, this could be that socio-economic status of living and poor social facilities are poor states in the North West Province [27, 28]. ## 3.5 Plant parts used for childhood diseases In the current study, the participants harvested different plant parts to prepare traditional remedies (Fig 3). Roots and rhizomes were the most frequently used as botanical drugs ($40\%$), followed by leaves ($23\%$) and whole plants ($20\%$). The dominance of underground parts observed in this study resonates with the findings of several studies in, South Africa [72, 79, 80] and other countries [15, 81]. Moichwanetse et al. [ 52] articulated that Batswana people believe in "ditswammung," which means "out of the soil". This is because it is believed that underground parts contain higher concentrations of the active ingredients and are storage organs of secondary metabolites [82, 83]. However, harvesting parts such as bulbs, rhizomes, and bark increase the conservational strains on the ecosystem, which often exacerbate the sustainability of medicinal plants mainly when extensively harvested in the wild [84, 85]. To ensure sustainable utilization of medicinal plant resources, it is necessary to apply proper harvesting strategies and conservation measures. This can be facilitated through increasing public awareness and promoting the cultivation of medicinal plants [86]. **Fig 3:** *Distribution (%) of medicinal plant parts used to manage and treat childhood diseases in Ngaka Modiri Molema and Bojanala districts of North West Province, South Africa (n = 92).* ## 3.6 Preparation, administration, and dosage methods Participants reported a variety of herbal preparations used to prepare traditional medicine for different types of childhood diseases in the study area. Maceration ($39\%$), decoction ($38\%$) are the most utilised forms of preparation, followed by poultice ($14\%$) and infusion ($9\%$), which is used for the least number of plant species. Application of the medicinal plant preparations is done through three different routes of administration. Oral application ($60\%$) was the most frequently cited route of administration, followed by topical ($39\%$), while applications through enema was generally low ($1\%$). Medicinal plants used for skin-related diseases were prepared using maceration and applied topically as lotion. Plants such as A. arborescens, E. elephantina, H. paronychioides, and H. hemerocallidea were popular species used for skin-related diseases. Furthermore, decoction and oral administration were the second most popular methods of preparation and administration, primarily in plants such as E. prostrata, D. anomala, P. granatum, and Z. oxyphylla, reported as a treatment for gastro-intestinal diseases (diarrhoea and constipation). Similar results were also observed in the previous study, where plants such as A. arborescens, E. elephantina, and H. hemerocallidea were applied directly to skin-related diseases [45]. Maceration and decoction are the most common modes of preparation in the current study and correlates with a few previous findings [87, 88]. The largest proportion of remedies prepared as a powder in the present study is not in concordance with some other research in countries such as Uganda [51] and Kenya [89]. ## 3.7 Life-form of the identified plants In the current study, the life-form for the documented plants included herbs, shrubs and trees with a distribution of 68, 17 and $15\%$, respectively. This does not reflect the floristic composition of the vegetation of the North West Province and the weather conditions experienced in the region [26, 90]. Most regions in the North West Province fall within the savannah biome with its associated Bushveld vegetation, and the western region primarily comprises Kalahari thornveld and shrub bushveld. In contrast, the central region is dominated by dry Cymbopogon-Themeda veld [90–92]. The high usage of herbs in treating and managing childhood diseases is a promising finding from a conservation perspective given that herbs re-generate and grow faster after being harvested. In North West Province, $40\%$ of the ecosystems are under severe stress, with 11 of the 61 vegetation and 14 of the 18 river types classified as threatened in terms of ecosystem status [93]. ## 3.8. Commonly encountered childhood diseases in the study areas The nominal group technique and the interviews enabled the researcher to identify 7 disease categories generated from the prevalent 29 sub-diseases as classified by the participants (Table 4). The most prevalent categories of childhood diseases were skin-related diseases (burns, skin irritation and warts). Skin-related diseases are still a huge concern and cause morbidity among children in developing countries, especially in sub-Saharan Africa [94, 95]. The prevalence of skin-related diseases in children under <5 years is well recognised [96, 97]. According to Hay et al. [ 98], skin infections contribute to approximately $34\%$ of occupational health diseases globally. Notably, the high occurrence of skin-related diseases in children could be associated with their low immune systems or low socio-economic status, favorable tropical weather, neglect and poor hygienic living conditions, including the lack of clean water and sanitation, particularly in the remote areas [95]. The high prevalence of skin-related diseases observed in the current study may be partially associated with opportunistic skin infections and HIV/AIDS, which are usually the first signs of HIV infection [77]. North West Province has high inactive cases of HIV/AIDs [99], which are often exacerbated by overcrowding and unhygienic environments [25]. The current findings differ from those of Boschi-Pinto et al. [ 100], whereby gastro-intestinal disease such as diarrhoea was a major cause of morbidity and mortality among children aged 5 in Sub-Saharan Africa. In 2008, among the estimated mortality of 4.2 million children who are 5-year-old in Africa, diarrhoea caused the largest proportion ($19\%$), followed by pneumonia and malaria. Previous studies indicated that co-morbidity, poor nutritional status, dehydration, lack of breastfeeding and prolonged diarrheal duration were risk factors for death in young children in Africa [72, 101]. In this study, gastro-intestinal diseases were the second most prevalent diseases. Gastrointestinal diseases affect the gastrointestinal tract from the mouth to the anus, including nausea/vomiting, food poisoning, lactose intolerance and diarrhoea [102]. According to the World Health Organization (WHO) [95], hygiene may be the leading cause of children-related diseases. Diarrhoea is the most common paediatric sickness and one of the leading causes of infant and child mortality [103, 104]. Diarrhoea morbidity and mortality are prevalent in developing countries, especially in Africa [76, 105]. According to Stats SA [25], approximately $49.2\%$ of the population in the North West Province lives below the upper-bound poverty line with no access to proper housing, water and sanitation. The majority of children reside in formal households, $8.6\%$ in informal dwellings, $13.3\%$ in traditional structures and only $0.3\%$ resided in other types of dwellings [25]. These poor conditions likely contribute to these diseases and may explain the high number of local communities resorting to traditional medicine. Similar findings were observed from previous studies in African countries such as Uganda [106] and Zimbabwe [107]. Pain and inflammation diseases, including urinary, genital diseases (such as bladder inflammation, kidney failure and urinary tract infection), and the umbilical cord was among the largest group behind skin-infections and related gastrointestinal diseases (Fig 4). Amongst the pain and inflammation categories, the umbilical cord (*Lelana la* mokhubu) was the most prevalent disease in this category. About a quarter of the global neonatal deaths are due to umbilical cord infection; $75\%$ of these occur in the first week of life, with the umbilical cord being the cause of death [108]. Most of the participants indicated that umbilical cord infections (omphalitis) and neonatal sepsis are significant contributors to the proportion of neonatal infections that prove fatal in the rural areas. These results were similar to the studies done in Nepal, Bangladesh, and Pakistan, which indicated that umbilical cord was recorded as one of the most prevalent childhood health condition [109, 110]. **Fig 4:** *The range of childhood diseases treated and managed by the participants in Ngaka Modiri Molema and Bojanala districts of North West Province, South Africa.* Other childhood-related diseases included sunken fontanelle diseases (*Tlhogwana ya* bana). It is known to contribute significantly to the mortality and morbidity of children < 5 years in low-middle-income countries [111]. In most developing countries, dehydration and sunken fontanelle are not often attributed to physical conditions. Sunken fontanelle is often associated with forces outside the body, social and spiritual. In many cultures, people believe that small children need special protection because they are very vulnerable to diarrhoea associated with a sunken fontanelle [112]. In the current study, about $90\%$ of the participants indicated that sunken fontanelle is a complex collection of illnesses with diarrhoea as a symptom, and they did not have a concept of dehydration. Depressed fontanelle was recognized as a sign of serious illness, and diarrhoea was linked to symptoms of dehydration through the concept of “*Nogana ya* bana” which means sunken fontanel. ## 3.9 Conservation status In terms of conservation status, the medicinal plants were categorized as “Least Concerned” ($79\%$), “not evaluated” ($11\%$), “Invasive Alien Species” ($6\%$), and “Critical and Endangered species” ($4\%$). The current finding suggests that most medicinal plants used to treat and manage childhood diseases in the study locations are readily available and not under severe conservation strains. Based on a recent review by Ndhlovu et al. [ 13], an estimated $84\%$ of medicinal plants used for childhood-related diseases are listed as ‘least concern’ in South Africa. Both S. aethiopicus and W. salutaris are medicinal plants whose wild populations are reported to be rapidly declining and endangered. Similar results have been observed in different studies [86, 113, 114]. South *Africa is* globally recognised as a prolific habitat for flora and fauna [115], with at least 3000 higher plant species with therapeutic value. However, medicinal plants used to treat and manage the disease are mostly harvested from the wild [60], with the possibility of many facing extinction from uncontrolled harvesting. Several studies have indicated the significance of the conserving the flora and fauna [84, 116]. ## 3.10 Multiple treatment indications and combination of medicinal plants The participants mentioned 6 plants in the inventory with multiple indications (uses), either as single or poly-plant remedies (Table 5). Single-plant remedies with multiple indications included plants such as A. elongatum, D. mooiensis, C. diffusa, E. serpen and S. hyacinthoides. In the study area, these above-mentioned plants had the highest citation as poly-plant remedies for treating and managing childhood diseases (Table 5). The use of two or more plants reflects the concept of synergy, where the association of plants can result in enhanced therapeutic efficacy. Several studies indicate that the mono-substance therapy model has gradually shifted toward adopting combination therapies, in which multiple active components are employed [117, 118]. Moichwanetse et al. [ 52] identified five medicinal plants in North West Province that are prescribed as poly-plant remedies in ethnoveterinary. According to Li and Weng [119], traditional medicine often comprises several ingredients mixed in a given ratio as herbal formula. Each ingredient in isolation sometimes lacks therapeutic activities seen in the holistic formulation, a phenomenon known as the combinatorial effect. **Table 5** | Scientific name | Local name | #Plant part | Application and preparation | Type of disease | | --- | --- | --- | --- | --- | | Commelina diffusa Burm.f. + Aptosinum elongatum Eng. | Ditantanyane + Kgopokgolo | R+B | Decoction, orally | Umbilical cord | | Sansevieria hyacinthoides (L) Druce. + Aptosinum elongatum Eng + Helichrysum paronychioides DC. Humbert. | Mosekelatsebeng + Ditantanyane + Phate ya ngaka | R+ S+ B | Decoction, orally | Reduce weaning and muscular | | Commelina diffusa Burm.f. + Solanum lichtensteinii Willd. | Kgopokgolo + Tolwane | B+R | Infusion with milk, orally | Umbilical cord | | Euphorbia serpen Kunths + Dianthus mooiensis F.N Williams subsp. kirkii (Burtt Davys) SS Hooper. | Lwetsane + Letlhoka la tsela | B+R | Maceration, topical | Sunken fontanel and weaning | | Sansevieria hyacinthoides (L) Druce. + Commelina diffusa Burm.f. | Makgabenyane + Kgopokgolo | B+R | Infusion, orally | Umbilical cord | ## 4. Conclusions The current findings revealed that childhood diseases are still a common problem in rural areas of the North West Province and local communities continue to rely on plants to meet the healthcare needs of their children. This is the first ethnobotanical study in North West Province relating to the use of medicinal plants for managing childhood diseases and wellbeing. We documented 61 medicinal plants and associated indigenous knowledge used to manage diverse diseases. Approximately $89\%$ of medicinal plants were recorded for the first time as medicinal plants used for childhood-related diseases. In comparison with earlier ethnobotanical studies conducted in South Africa, $11\%$ of the medicinal plants were confirmed to be used for childhood-related diseases. These plants included A. arborescens, B. macrostegia, D. cinerea, E. autumnalis, H. hemerocallidea, S. lichtensteinii and W. somnifera. In addition, the medicinal plants had similar uses with the current findings. However, we observed considerable differences with respect to the preparation and administration techniques. The current results highlighted some degree of novelty with regards to the diverse uses associated with medicinal plants in South Africa. We also demonstrated the importance of collecting new ethnobotanical information on well-known medicinal plants. Quantitative analysis revealed that A. elongatum, followed by C. diffusa, and E. prostrata, had the highest UV and were the most cited plant species. Based on ICF, skin-related diseases had the highest ICF value (0.99). This category had 381 use-reports, comprising 34 plants ($55.7\%$ of total plants) used for childhood-related diseases, with B. frutescens and E. elephantina being the most-cited plants in this category. This valuable plant inventory demonstrates a major step towards the ongoing effort of the documentation, preservation, and promotion of indigenous knowledge of the study area. Overall, these results enrich the national and globally pharmacopeia in managing the healthcare needs of children. However, further investigations into the efficacy, safety, biological activities, and phytochemical profiling of these documented plants remain pertinent. Such stringent investigations may uncover novel compounds with pharmaceutical relevance in treating childhood diseases. ## References 1. Sharrow D, Hug L, You D, Alkema L, Black R, Cousens S. **Global, regional, and national trends in under-5 mortality between 1990 and 2019 with scenario-based projections until 2030: a systematic analysis by the UN Inter-agency Group for Child Mortality Estimation**. *The Lancet Global Health* (2022.0) **10** e195-e206. DOI: 10.1016/S2214-109X(21)00515-5 2. Rhoda N, Velaphi S, Gebhardt G, Kauchali S, Barron P. **Reducing neonatal deaths in South Africa: Progress and challenges**. *South African Medical Journal* (2018.0) **108** 9-16 3. 3MacroTrends. South Africa Infant Mortality Rate 1950–2022. United Nations—World Population Prospects; 2022. 4. Mokwena K, Kachabe J. **Profile of mothers whose children are treated for malnutrition at a rural district hospital in the North West province, South Africa**. *South African Journal of Clinical Nutrition* (2021.0) **35** 17-22. DOI: 10.1080/16070658.2021.1921899 5. Lin L-y, Wang J, Ou-Yang X-Y, Miao Q, Chen R, Liang F-x. **The immediate impact of the 2019 novel coronavirus (COVID-19) outbreak on subjective sleep status**. *Sleep Medicine* (2021.0) **77** 348-54. DOI: 10.1016/j.sleep.2020.05.018 6. Killgore WDS, Taylor EC, Cloonan SA, Dailey NS. **Psychological resilience during the COVID-19 lockdown**. *Psychiatry Research* (2020.0) **291** 113216. DOI: 10.1016/j.psychres.2020.113216 7. Garrib A, Stoops N, McKenzie A, Dlamini L, Govender T, Rohde D. **An evaluation of the district health information system in rural South Africa**. *South African Medical Journal* (2008.0) **98** 549-52. PMID: 18785397 8. 8WHO. WHO traditional medicine strategy 2014–2023. Geneva, Switzerland WHO, 2013.. *WHO traditional medicine strategy 2014–2023* (2013.0) 9. Beltreschi L, de Lima RB, da Cruz DD. **Traditional botanical knowledge of medicinal plants in a “quilombola” community in the Atlantic Forest of northeastern Brazil**. *Environment, Development and Sustainability* (2018.0) **21** 1185-203 10. Krishnasis C, Suraj C. **Estimation the presence of heavy metals and trace elements in Indian herbal cosmetics by atomic absorption spectrophotometer**. *Asian Journal of Pharmaceutical Research and Development* (2019.0) **7** 60-5 11. Noorhosseini SA, Fallahi E, Damalas CA. **Promoting cultivation of medicinal and aromatic plants for natural resource management and livelihood enhancement in Iran**. *Environment, Development and Sustainability* (2020.0) **22** 4007-24 12. Haque M, Chowdhury A, Shahjahan M, Harun M, Dostogir G. **Traditional healing practices in rural Bangladesh: a qualitative investigation**. *BMC Complementary and Alternative Medicine* (2018.0) **18** 1-15. PMID: 29295712 13. Ndhlovu PT, Omotayo AO, Otang Mbeng W, Aremu AO. **Ethnobotanical review of plants used for the management and treatment of childhood diseases and well-being in South Africa**. *South African Journal of Botany* (2021.0) **137** 197-215 14. Adeogun OO, Adekunle AA, Ebabhi AM. **Survey and phytochemical analyses of plants use for the treatment of childhood diseases in Ojo Lga, Lagos State**. *International Journal of Herbal Medicine* (2014.0) **2** 109-14 15. Nalumansi P, Kamatenesi-Mugisha M, Anywar G. **Medicinal plants used in paediatric health care in Namungalwe Sub County, Iganga District, Uganda**. *Nova Journal of Medical and Biological Sciences* (2014.0) **2** 1-14 16. Welcome AK, Van Wyk BE. **An inventory and analysis of the food plants of southern Africa**. *South African Journal of Botany* (2019.0) **122** 136-79 17. De Villiers FPR, Ledwaba MJP. **Traditional healers and paediatric care**. *South African Medical Journal* (2003.0) **2003** 664-5. PMID: 14635552 18. Van Wyk BE, Van Oudtshoorn B, Gericke N. *Medicinal Plants of South Africa* (2009.0) 19. Hamilton AC. **Medicinal plants, conservation and livelihoods**. *Journal of Biodiversity and Conservation* (2004.0) **13** 1477-517 20. Van Wyk B-E, Van Oudtshoorn B, Gericke N. *Medicinal plants of South Africa* (1997.0) 21. Aremu AO, Pendota SC. **Medicinal plants for mitigating pain and inflammatory-related conditions: an appraisal of ethnobotanical uses and patterns in South Africa**. *Frontiers in Pharmacology* (2021.0) 2650. DOI: 10.3389/fphar.2021.758583 22. Semenya SS, Potgieter MJ. **Ethnobotanical survey of medicinal plants used by Bapedi traditional healers to treat erectile dysfunction in the Limpopo Province, South Africa**. *Journal of Medicinal Plants Research* (2013.0) **7** 349-57 23. Chen S-L, Yu H, Luo H-M, Wu Q, Li C-F, Steinmetz A. **Conservation and sustainable use of medicinal plants: problems, progress, and prospects**. *Chinese Medicine* (2016.0) **11** 37. DOI: 10.1186/s13020-016-0108-7 24. 24Stats SA. Census 2011. Pretoria, South Africa: Statistics South Africa, 2012 Contract No.: Report No. 03-01-41. 25. Stats SA. *Living conditions of households in South Africa* (2015.0) 26. Kruger AC, Nxumalo MP. **Historical rainfall trends in South Africa: 1921–2015**. *Water SA* (2017.0) **43** 285-97 27. Stats SA. *Quarterly Labour Force Survey* (2018.0) 28. Stats SA. *Provincial profile: North West Community Survey 2016* (2018.0) 29. 29World Health Organization (WHO). WHO guidelines on good herbal processing practices for herbal medicines Geneva, Switzland: WHO; 2019 [cited 2022 04/02].. *WHO guidelines on good herbal processing practices for herbal medicines Geneva* (2019.0) 30. 30Department of Health. Integrated management of childhood illines Pretoria, South Africa 2019. 31. Kothari CR. *Research methodology (Method and Technqiues)* (2004.0) 1-489 32. Tongco MDC. **Purposive sampling as a tool for informant selection**. *Ethnobotany Research and Applications* (2007.0) **5** 147-58 33. 33Constitution of the Republic of South Africa. Republic of South Africa Traditional Health Practitioners Act of 2007. Cape Town, South Africa 2007. 34. 34Constitution of the Republic of South Africa, Act 200 of 1993 (1996). 35. Creswell JW, Creswell JD. *Research design: Qualitative, quantitative, and mixed methods approaches* (2017.0) 36. Salkind NJ. *Encyclopedia of Research Design* (2010.0). DOI: 10.4135/9781412961288 37. Leistner OA, Leistner OA. *Families and genera* (2000.0) **10** 38. Bless C, Higson-Smith C, Kagee A. *Fundamentals of social research methods: An African perspective* (2006.0) 39. Tardío J, Pardo-de-Santayana M. **Cultural importance indices: a comparative analysis based on the useful wild plants of Southern Cantabria (Northern Spain)**. *Economic Botany* (2008.0) **62** 24-39 40. Phillips O, Gentry AH. **The useful plants of Tambopata, Peru: I. Statistical hypotheses tests with a new quantitative technique**. *Economic Botany* (1993.0) **47** 15-32. DOI: 10.1007/BF02862203 41. Trotter RT, Logan MH. *Plants in Indigenous Medicine & Diet* (1986.0) 42. Gazzaneo LRS, Lucena RFP, Albuquerque UP. **Knowledge and use of medicinal plants by local specialists in an region of Atlantic Forest in the state of Pernambuco (Northeastern Brazil)**. *Journal of Ethnobiology and Ethnomedicine* (2005.0) **1** 9. DOI: 10.1186/1746-4269-1-9 43. Heinrich M, Ankli A, Frei B, Weimann C, Sticher O. **Medicinal plants in Mexico: Healers’ consensus and cultural importance**. *Social Science & Medicine* (1998.0) **47** 1859-71. DOI: 10.1016/s0277-9536(98)00181-6 44. Ateba CN, Kaya HO, Pitso FS, Ferim V, Smit JA, Masoga MA. *African Indigenous Knowledge Systems and Sustainable Development: Challenges and Prospects* (2012.0) 68-86 45. Asong JA, Ndhlovu PT, Khosana NS, Aremu AO, Otang-Mbeng W. **Medicinal plants used for skin-related diseases among the Batswanas in Ngaka Modiri Molema District Municipality, South Africa**. *South African Journal of Botany* (2019.0) **126** 11-20 46. Omotayo OA, Ndhlovu PT, Olagunju KO, Tshwane SC, Aremu AO. **Determinants of household’s income and willingness to pay for indigenous plants in North West Province, South Africa: A Two-stage Heckman Approach**. *Sustainability* (2020.0) **13** 5458 47. Imran M, Akhtar S, Chen Y, Ahmad S. *Environmental Education and Women: Voices From Pakistan* (2021.0) **11** 48. Omotayo AO. **Economics of farming household’s food intake and health-capital in Nigeria: A two-stage probit regression approach**. *The Journal of Developing Areas* (2017.0) **51** 109-25 49. Omotayo AO. **Farming households’ environment, nutrition and health interplay in Southwest, Nigeria**. *International Journal of Scientific Research in Agricultural Sciences* (2016.0) **3** 84-98 50. Nkonki-Mandleni B, Ogunkoya FT, Omotayo AO. **Socioeconomic factors influencing livestock production among smallholder farmers in the free state province of south Africa**. *lnternational Journal of Entrepreneurship* (2019.0) **23** 1-17 51. Tugume P, Nyakoojo C. **Ethno-pharmacological survey of herbal remedies used in the treatment of paediatric diseases in Buhunga parish, Rukungiri District, Uganda**. *Journal of Complementary and Alternative Medicine* (2019.0) **19** 353. DOI: 10.1186/s12906-019-2763-6 52. Moichwanetse BI, Ndhlovu PT, Sedupane G, Aremu AO. **Ethno-veterinary plants used for the treatment of retained placenta and associated diseases in cattle among Dinokana communities, North West Province, South Africa**. *South Africa Journal of Botany* (2020.0) **132** 108-16 53. Tshikalange TE, Mophuting BC, Mahore J, Winterboer ST, Namrita L. **An ethnobotanical study of medicinal plants used in villages under Jongilanga Tribal Council, Mpumalanga, South Africa**. *African Journal of Traditional Complementary and Alternative Medicines* (2016.0) **13** 83-9. DOI: 10.21010/ajtcam.v13i6.13 54. Masevhe NA, McGaw LJ, Eloff JN. **The traditional use of plants to manage candidiasis and related infections in Venda, South Africa**. *Journal of Ethnopharmacology* (2015.0) **168** 364-72. DOI: 10.1016/j.jep.2015.03.046 55. Maroyi A.. **Diversity of use and local knowledge of wild and cultivated plants in the Eastern Cape province, South Africa**. *Journal of Ethnobiology and Ethnomedicine* (2017.0) **13** 43. DOI: 10.1186/s13002-017-0173-8 56. Moteetee A, Van Wyk BE. **The medical ethnobotany of Lesotho: A review**. *Bothalia* (2011.0) **41** 209-28 57. Alamgeer Younis W, Asif H, Sharif A, Riaz H, Bukhari IA. **Traditional medicinal plants used for respiratory disorders in Pakistan: a review of the ethno‑medicinal and pharmacological evidence**. *Chinese Medicine* (2018.0) **13** 48. DOI: 10.1186/s13020-018-0204-y 58. Van Wyk BE. **A family-level floristic inventory and analysis of medicinal plants used in Traditional African Medicine**. *Journal of Ethnopharmacology* (2020.0) **249** 112351. DOI: 10.1016/j.jep.2019.112351 59. Rolnik A, Olas B. **The plants of the Asteraceae Family as agents in the protection of human health**. *International Journal of Molecular Sciences* (2021.0) **22** 3009. DOI: 10.3390/ijms22063009 60. Stepp JR, Moerman DE. **The importance of weeds in Ethnopharmacology**. *Journal of ethnopharmacology* (2001.0) **75** 19-23. DOI: 10.1016/s0378-8741(00)00385-8 61. Koc S, Isgor BS, Isgor YG, Shomali Moghaddam N, Yildirim O. **The potential medicinal value of plants from Asteraceae family with antioxidant defense enzymes as biological targets**. *Pharmaceutical Biology* (2015.0) **53** 746-51. DOI: 10.3109/13880209.2014.942788 62. Heydari M, Heydari H, Saadati A, Gharehbeglou M, Tafaroji J, Akbari A. **Ethnomedicine for neonatal jaundice: A cross-sectional survey in Qom, Iran**. *Journal of Ethnopharmacology* (2016.0) **193** 637-42. DOI: 10.1016/j.jep.2016.10.019 63. Alzweiri M, Al Sarhan A, Mansi K, Hudaib M, Aburjai T. **Ethnopharmacological survey of medicinal herbs in Jordan, the Northern Badia region**. *Journal of Ethnopharmacology* (2011.0) **137** 27-35. DOI: 10.1016/j.jep.2011.02.007 64. Mudau TE, Olowoyo JO, Amoo SO. **Ethnobotanical assessment of medicinal plants used traditionally for treating diabetes in Vhembe district, Limpopo Province, South Africa**. *South African Journal of Botany* (2022.0) **146** 304-24. DOI: 10.1016/j.sajb.2021.10.016 65. Semenya SS, Maroyi A. **Ethnobotanical survey of plants used by Bapedi traditional healers to treat tuberculosis and its opportunistic infections in the Limpopo Province, South Africa**. *South African Journal of Botany* (2019.0) **122** 401-21 66. Mongalo NI, Makhafola TJ. **Ethnobotanical knowledge of the lay people of Blouberg area (Pedi tribe), Limpopo Province, South Africa**. *Journal of Ethnobiology and Ethnomedicine* (2018.0) **14** 1-23. PMID: 29316951 67. Rahman MM, Mannan MA, Nijhu RS, Khatun A. **Traditional uses, phytochemistry and pharmacology of Commelina diffusa Burm: An updated systematic review**. *Journal of Pharmacognosy and Phytochemistry* (2021.0) **10** 53-9 68. Beyra Á, León M, Iglesias E, Ferrándiz D, Herrera R, Volpato G. **Estudios etnobotánicos sobre plantas medicinales en la provincia de Camagüey (Cuba)**. *Anales del jardín botánico de Madrid* (2004.0) **61** 185-204 69. Özbigin S, Çitoğlu G. **Uses of some Euphorbia species in traditional medicine in Turkey and their biological activities**. *Turkish Journal of Pharmaceutical Sciences* (2012.0) **9** 241-56 70. Ojewole JAO. *C. A. Mey* (2008.0) **22** 91-6 71. Mhlongo LS, Van Wyk B-E. **Zulu medicinal ethnobotany: new records from the Amandawe area of KwaZulu-Natal, South Africa**. *South African Journal of Botany* (2019.0) **122** 266-90 72. Mashile SP, Tshisikhawe MP, Masevhe NA. **Medicinal plants used in the treatment of maternal health-related problems by the Mapulana of Ehlanzeni District, Mpumalanga province, South Africa**. *Journal of Applied Pharmaceutical Science* (2019.0) **9** 021-9 73. Caetano RdA, Albuquerque UPd, Medeiros PMd. **What are the drivers of popularity and versatility of medicinal plants in local medical systems?**. *Acta Botanica Brasilica* (2020.0) **34** 256-65 74. de Sousa Araújo TA, Alencar NL, de Amorim ELC, de Albuquerque UP. **A new approach to study medicinal plants with tannins and flavonoids contents from the local knowledge**. *Journal of Ethnopharmacology* (2008.0) **120** 72-80. DOI: 10.1016/j.jep.2008.07.032 75. Ullah S, Rashid Khan M, Ali Shah N, Afzal Shah S, Majid M, Asad Farooq M. **Ethnomedicinal plant use value in the Lakki Marwat District of Pakistan**. *Journal of Ethnopharmacology* (2014.0) **158** 412-22. DOI: 10.1016/j.jep.2014.09.048 76. Njume C, Goduka NI. **Treatment of diarrhoea in rural african communities: An overview of measures to maximise the medicinal potentials of indigenous plants**. *International Journal of Environmental Research and Public Health* (2012.0) **9** 3911-393. DOI: 10.3390/ijerph9113911 77. Afolayan AJ, Grierson DS, Otang-Mbeng W. **Ethnobotanical survey of medicinal plants used in the management of skin disorders among the Xhosa communities of the Amathole District, Eastern Cape, South Africa**. *Journal of Ethnpharmacology* (2014.0) **153** 220-32. DOI: 10.1016/j.jep.2014.02.023 78. Banda Chitsamatanga B, Malinga W. **‘A tale of two paradoxes in response to COVID-19’: Public health system and socio-economic implications of the pandemic in South Africa and Zimbabwe**. *Cogent Social Sciences* (2021.0) **7** 1869368 79. Moshi MJ, Otieno DF, Mbabazi PK, Weisheit A. **Ethnomedicine of the Kagera region, North Western Tanzania. Part 2: The medicinal plants used in Katoro Ward, Bukoba District**. *Journal of Ethnobiology and Ethnomedicine* (2010.0) **6** 19. DOI: 10.1186/1746-4269-6-19 80. Olajuyigbe OO, Afolayan A. **Ethnobotanical survey of medicinal plants used in the treatment of gastrointestinal disorders in the Eastern Cape Province, South Africa**. *Journal of Medicinal Plants Research* (2012.0) **6** 3415-24 81. Adhikari PP, Talukdar S, Borah A. **Ethnomedicobotanical study of indigenous knowledge on medicinal plants used for the treatment of reproductive problems in Nalbari district, Assam, India**. *J Ethnopharmacol* (2018.0) **210** 386-407. DOI: 10.1016/j.jep.2017.07.024 82. Flores HE, Flores T. **Underground plant storage organs**. *Functionality of Food Phytochemicals* (2012.0) **31** 113 83. Al-Rowaily SL, Abd-ElGawad AM, Alghanem SM, Al-Taisan WaA, El-Amier YA. **Nutritional Value, Mineral Composition, Secondary Metabolites, And Antioxidant Activity of Some Wild Geophyte Sedges and Grasses**. *Plants* (2019.0) **8** 569. DOI: 10.3390/plants8120569 84. Moyo M, Van Staden J. **Medicinal properties and conservationof**. *Journal of Ethnopharmacology* (2014.0) **152** 243-55. PMID: 24463034 85. Barata AM, Rocha F, Lopes V, Carvalho AM. **Conservation and sustainable uses of medicinal and aromatic plants genetic resources on the worldwide for human welfare**. *Industrial Crops and Products* (2016.0) **88** 8-11 86. Seile BP, Bareetseng S, Koitsiwe MT, Aremu AO. **Indigenous Knowledge on the Uses, Sustainability and Conservation of African Ginger (**. *Diversity* (2022.0) **14** 192 87. Mwinga JL, Makhaga NS, Aremu AO, Mbeng W. **Botanicals used for cosmetic purposes by Xhosa women in the Eastern Cape, South Africa**. *South African Journal of Botany* (2019.0) **126** 4-10 88. Agbodeka K, Gbekley HE, Karou SD, Anani K, Agbonon A, Tchacondo T. **Ethnobotanical study of medicinal plants used for the treatment of malaria in the Plateau Region, Togo**. *Pharmacognosy Research* (2016.0) **8** S12-S8. DOI: 10.4103/0974-8490.178646 89. Geissler PW, Harris SA, Prince RJ, Olsen A, Odhiambo RA, Oketch-Rabah H. **Medicinal plants used by Luo mothers and children in Bondo district, Kenya**. *Journal of Ethnopharmacology* (2002.0) **83** 39-54. DOI: 10.1016/s0378-8741(02)00191-5 90. Davoren E.. *Plant diversity patterns of a settlement in the North–West Province* (2009.0) 91. Esterhuizen A.. *The effect of fire on savanna vegetation dynamics in the semi-arid Molopo Bushveld region of the North-West Province* (2019.0) 92. Lamprecht A, Cilliers S, Götze A, Du Toit M. **Phytosociological description of norite koppies in the Rustenburg area, North-West Province and refinement of the distribution of the Norite Koppies Bushveld on the national vegetation classification map of South Africa**. *Bothalia* (2011.0) **41** 327-39 93. Desmet P, Schaller R, Skowno A. (2009.0) 94. Kiprono SK, Muchunu JW, Masenga JE. **Skin diseases in pediatric patients attending a tertiary dermatology hospital in Northern Tanzania: a cross-sectional study**. *BMC Dermatol* (2015.0) **15** 16. DOI: 10.1186/s12895-015-0035-9 95. 95World Health Organization (WHO). Epidemiology and management of common skin diseases in children in developing countries. Geneva, Switzland: World Health Organization, 2005.. *Epidemiology and management of common skin diseases in children in developing countries* (2005.0) 96. Mahmood A, Eqan M, Pervez S, Alghamdi HA, Tabinda AB, Yasar A. **COVID-19 and frequent use of hand sanitizers; human health and environmental hazards by exposure pathways**. *Science of the Total Environment* (2020.0) **742** 140561. DOI: 10.1016/j.scitotenv.2020.140561 97. Picardo M, Eichenfield LF, Tan J. **Acne and rosacea**. *Dermatology and therapy* (2017.0) **7** 43-52. DOI: 10.1007/s13555-016-0168-8 98. Hay RJ, Johns NE, Williams HC, Bolliger IW, Dellavalle RP, Margolis DJ. **The global burden of skin disease in 2010: an analysis of the prevalence and impact of skin conditions**. *Journal of Investigative Dermatology* (2014.0) **134** 1527-34. DOI: 10.1038/jid.2013.446 99. Rix J.. *Medicine treatment patterns of HIV/AIDS patients at a rural district hospital in the North West province* (2013.0) 100. Boschi-Pinto C, Young M, Black RE. **The Child Health Epidemiology Reference Group reviews of the effectiveness of interventions to reduce maternal, neonatal and child mortality**. *International Journal of Epidemiology* (2010.0) **39** i3-i6. DOI: 10.1093/ije/dyq018 101. Girma B, Berhane Y. **Children who were vaccinated, breast fed and from low parity mothers live longer: a community based case-control study in Jimma, Ethiopia**. *BMC Public Health* (2011.0) **11** 1-7. PMID: 21199570 102. Lucas López R, Grande Burgos MJ, Gálvez A, Pérez Pulido R. **The human gastrointestinal tract and oral microbiota in inflammatory bowel disease: a state of the science review**. *Apmis* (2017.0) **125** 3-10. DOI: 10.1111/apm.12609 103. Mokomane M, Kasvosve I, Melo Ed, Pernica JM, Goldfarb DM. **The global problem of childhood diarrhoeal diseases: emerging strategies in prevention and management**. *Therapeutic Advances in Infectious Disease* (2018.0) **5** 29-43. DOI: 10.1177/2049936117744429 104. Ugboko HU, Nwinyi OC, Oranusi SU, Oyewale JO. **Childhood diarrhoeal diseases in developing countries**. *Heliyon* (2020.0) **6** e03690. DOI: 10.1016/j.heliyon.2020.e03690 105. Agbagbeola DD, Tosin FI, Adetokunbo TS. **Prevalence and determinants of diarrhoea among infants in selected primary health centres in Kaduna north local government area, Nigeria**. *Pan African Medical Journal* (2017.0) **28** 109. DOI: 10.11604/pamj.2017.28.109.8152 106. Anokbonggo W, Odoi-Adome R, Oluju P. **Traditional methods in management of diarrhoeal diseases in Uganda**. *Bulletin of the World Health Organization* (1990.0) **68** 359. PMID: 2393983 107. Pitts M, Mcmaster J, Hartmann T, Mausezahl D. **Lay beliefs about diarrhoeal diseases: their role in health education in a developing country**. *Social Science & Medicine* (1996.0) **43** 1223-8. DOI: 10.1016/0277-9536(95)00430-0 108. López-Medina MD, López-Araque AB, Linares-Abad M, López-Medina IM. **Umbilical cord separation time, predictors and healing complications in newborns with dry care**. *PloS One* (2020.0) **15** e0227209. DOI: 10.1371/journal.pone.0227209 109. Soofi S, Cousens S, Imdad A, Bhutto N, Ali N, Bhutta ZA. **Topical application of chlorhexidine to neonatal umbilical cords for prevention of omphalitis and neonatal mortality in a rural district of Pakistan: a community-based, cluster-randomised trial**. *The Lancet* (2012.0) **379** 1029-36 110. Mullany LC, Darmstadt GL, Khatry SK, LeClerq SC, Katz J, Tielsch JM. **Impact of umbilical cord cleansing with 4.0% chlorhexidine on time to cord separation among newborns in southern Nepal: a cluster-randomized, community-based trial**. *Pediatrics* (2006.0) **118** 1864-71. DOI: 10.1542/peds.2006-1091 111. 111World Health Organization. UNICEF/WHO/The World Bank Group joint child malnutrition estimates: levels and trends in child malnutrition: key findings of the 2020 edition. Geneva, Switzland: 2020 9240003576. 112. Kaltenthaler EC, Drasar BS. **Understanding of hygiene behaviour and diarrhoea in two villages in Botswana**. *Journal of Diarrhoeal Diseases Research* (1996.0) 75-80. PMID: 8870398 113. Adebayo SA, Amoo SO, Mokgehle SN, Aremu AO. **Ethnomedicinal uses, biological activities, phytochemistry and conservation of African ginger (**. *Journal of Ethnopharmacology* (2021.0) **266** 113459. DOI: 10.1016/j.jep.2020.113459 114. Dludlu MN, Dlamini PS, Sibandze GF, Vilane VS, Dlamini CS. **Distribution and conservation status of the Endangered pepperbark tree**. *Oryx* (2017.0) **51** 451-4. DOI: 10.1017/S0030605316000302 115. Street R, Prinsloo G. **Commercially important medicinal plants of South Africa: A review**. *Journal of Chemistry* (2012.0) **2013**. DOI: 10.1155/2013/205048 116. Mugomeri E, Chatanga P, Raditladi T, Makara M, Tarirai C. **Ethnobotanical study and conservation status of local medicinal plants: Towards a repository and monograph of herbal medicines in Lesotho**. *African Journal of Traditional, Complementary and Alternative Medicines* (2016.0) **13** 143-56 117. Rasoanaivo P, Wright CW, Willcox ML, Gilbert B. **Whole plant extracts versus single compounds for the treatment of malaria: synergy and positive interactions**. *Malaria Journal* (2011.0) **10** 1-12. PMID: 21214892 118. Zhou X, Seto SW, Chang D, Kiat H, Razmovski-Naumovski V, Chan K. **Synergistic effects of chinese herbal medicine: A comprehensive review of methodology and current research**. *Frontiers in Pharmacology* (2016.0) **7**. DOI: 10.3389/fphar.2016.00201 119. Li F-s, Weng J-K. **Demystifying traditional herbal medicine with modern approach**. *Nature Plants* (2017.0) **3** 2017109. DOI: 10.1038/nplants.2017.109
--- title: Capturing children food exposure using wearable cameras and deep learning authors: - Shady Elbassuoni - Hala Ghattas - Jalila El Ati - Yorgo Zoughby - Aline Semaan - Christelle Akl - Tarek Trabelsi - Reem Talhouk - Houda Ben Gharbia - Zoulfikar Shmayssani - Aya Mourad journal: PLOS Digital Health year: 2023 pmcid: PMC10042366 doi: 10.1371/journal.pdig.0000211 license: CC BY 4.0 --- # Capturing children food exposure using wearable cameras and deep learning ## Abstract Children’s dietary habits are influenced by complex factors within their home, school and neighborhood environments. Identifying such influencers and assessing their effects is traditionally based on self-reported data which can be prone to recall bias. We developed a culturally acceptable machine-learning-based data-collection system to objectively capture school-children’s exposure to food (including food items, food advertisements, and food outlets) in two urban Arab centers: Greater Beirut, in Lebanon, and Greater Tunis, in Tunisia. Our machine-learning-based system consists of 1) a wearable camera that captures continuous footage of children’s environment during a typical school day, 2) a machine learning model that automatically identifies images related to food from the collected data and discards any other footage, 3) a second machine learning model that classifies food-related images into images that contain actual food items, images that contain food advertisements, and images that contain food outlets, and 4) a third machine learning model that classifies images that contain food items into two classes, corresponding to whether the food items are being consumed by the child wearing the camera or whether they are consumed by others. This manuscript reports on a user-centered design study to assess the acceptability of using wearable cameras to capture food exposure among school children in Greater Beirut and Greater Tunis. We then describe how we trained our first machine learning model to detect food exposure images using data collected from the Web and utilizing the latest trends in deep learning for computer vision. Next, we describe how we trained our other machine learning models to classify food-related images into their respective categories using a combination of public data and data acquired via crowdsourcing. Finally, we describe how the different components of our system were packed together and deployed in a real-world case study and we report on its performance. ## Author summary Capturing food exposure of school children is a challenging task due to recall bias. In this manuscript, we describe a machine-learning-based data-collection tool that can automatically record school children’s exposure to food items, food advertisements and food outlets in their homes, schools and neighborhoods. Our data-collection tool consists of a wearable camera to capture continuous footage of children’s environments during a typical school day, and a series of machine learning models that can extract food-related images from the recorded footage and classify them into images that contain food items consumed by the child wearing the camera, or consumed by others, images that contain food advertisements, and images that contain food outlets. We report on a user-centered design study that assessed the acceptability of using wearable cameras to capture food exposure among school children in two urban Arab centers, namely Greater Beirut in Lebanon and Greater Tunis in Tunisia. We then describe how we trained our various machine learning models to capture food exposure among school children and categorize such food exposure into a predefined typology. Finally, we also report on the results of deploying our data-collection tool in a real-world case study in Tunisia. ## 1 Introduction Children’s food choice drivers are elicited in complex frameworks of interconnected factors at the levels of their homes, schools and neighborhoods. Evidence shows that children’s exposure to food advertisements and shops in their daily environments, particularly on their trajectories to and from school is a main contributing element to their food choices and dietary behaviors [1–4]. Capturing these exposures’ presence and measuring their frequency and potential associations with children’s dietary habits and health and nutrition outcomes is commonly limited by the use of traditional methods of data collections that are subject to information and recall bias [5–7]. Technology-based tools enable an objective and comprehensive measurement of children’s nutrition-related behaviors and experiences around food [6, 8, 9]. Digital technologies can engage participants as active contributors to the research process, objectively document their lived experiences, foster a “people-based” approach to measuring these exposures [10], with children being at the center of the process, which may also lead to more accurate and representative data on their food experiences [9, 11]. There is a wealth of research conducted on technology-based dietary assessment [12]. For example, Lui et al. [ 13] proposed a wearable-sensor platform that can automatically provide information regarding a subject’s dietary habits. Similarly, Signal et al. [ 14] reported on innovative research in New Zealand in which children wore cameras to examine the frequency and nature of everyday exposure to food marketing across multiple media and settings. A follow-up study by McKerchar et al. [ 15] focused on food-store environments and assessed food-product availability, placement, packaging, branding, price promotion, purchases and consumption. Another follow-up study by Liu et al. [ 16] focused on space-time analysis of unhealthy food advertising among the school children using the data captured through the wearable cameras. Gao et al. [ 17] explored the feasibility of applying Simultaneous Localization and Mapping (SLAM) techniques for continuous food volume measurement with a wearable monocular camera. Shroff et al. [ 18] proposed DiaWear, a novel assistive mobile-phone-based calorie-monitoring system to improve the quality of life of individuals with special nutrition management needs. Davies et al. [ 19] explored using wearable cameras to monitor eating and drinking behavior during transport journeys. Similarly, Gage et al. [ 20] analyzed the frequency and context of snacking among children using wearable cameras. Doulah et al. [ 21] proposed a novel wearable sensor system that uses a combination of acceleration and temporalis muscle sensors for accurate detection of food intake and triggering of a wearable camera to capture the food being consumed. Jia et al. [ 22] developed a machine-learning-based approach, which can automatically detect food items from images acquired by an egocentric wearable camera for dietary assessment. Similarly, Jia et al. [ 23] developed a small, chest-worn electronic device called eButton, which automatically took pictures of consumed food for objective dietary assessment. Finally, Hossain et al. [ 24] proposed a machine-learning-based system which classified captured images by wearable egocentric cameras as food or no-food images in real-time, which is the closest to our work. Although many of the above-surveyed technologies have been positively received by younger participants, and are considered to be feasible and acceptable means of documenting food exposures [7–9, 25], they have not been used and validated in low-middle income countries or outside of a Western cultural context. Thus, the purpose of this manuscript is two-fold: The settings of this work are two urban agglomerations in the Arab region; the Greater Beirut Area in Lebanon and Greater Tunis in Tunisia. These two middle-income countries have witnessed a rapidly-proceeding nutrition transition in the past two decades and the estimates of childhood overweight match those of high-income countries, reaching about $30\%$ in Lebanon and $20\%$ in Tunisia [26]. We developed a data collection system that uses wearable cameras to continuously capture videos of the child’s environment during a typical school day. The captured footage acts as a recorded diary of what the child wearing the camera is being exposed to. Since the amount of footage captured is typically large, and only a small percentage corresponds to images with content related to food, we employ machine learning to identify images (i.e., video frames) related to food and discard any other footage. We consider that an image is related to food or is a food exposure image if it displays anything associated with food and/or beverages such as food items, food outlets, restaurants, food brands, food markets, food advertisements, and food vending machines, etc. We also employed machine learning to classify food exposure images into images that contain food items being consumed by the child wearing the camera or others, images that contain food advertisements, and images that contain food outlets. Fig 1 gives an overview of our food exposure typology and Fig 2 (a) shows example images that contain food consumption, (b) food outlets, (c) food advertisements, and (d) both food outlets and food advertisements. **Fig 1:** *Overview of the food exposure typology.* **Fig 2:** *Example images that contain (a) food items being consumed, (b) food outlets, (c) food advertisements, and (d) both food outlets and food advertisements.* In order to develop such a protocol, we first assessed the acceptability of using wearable cameras among school children through a user-centered design study involving school staff, parents and school children (10-12 years old), in 12 interactive design workshops conducted in Greater Beirut and Greater Tunis. These workshops employed discussions, mind mapping and storyboarding activities to identify challenges associated with the use of wearable cameras among school children participating in research, and inform the design of a tool that meets cultural and ethical requirements. Discussions during the workshops were audio-recorded, transcribed and analyzed thematically along with the mind maps and storyboards. Commonly reported challenges were: invading children’s and third-parties’ privacy, distracting school children during classes, and obtaining biased data. To overcome these challenges, participants suggested wearable cameras capturing exposure automatically for a short period of time. Two rounds of image filtering were proposed to safeguard privacy: automated selection of images related to food exposure, followed by parental manual screening. To protect anonymity, participants suggested automated blurring of faces in all captured footage. Building on the ethical and cultural requirements that were identified in our user-centered design study, we developed a machine-learning-based data-collection system that automatically captures images with content related to food using wearable cameras. The system includes: Finally, we deployed our machine-learning-based data-collection system in a real-world case study in Tunisia. In this manuscript, we provide the results of an interim study sample consisting of 265 children aged 11-12 years from 29 schools in Greater Tunis, and we report on the accuracy and precision of the developed system. ## 2.1 User-centered design study Based on recommendations to ensure ethical conduct in visual research [29, 30], we assessed the feasibility and acceptability and aimed to develop a set of design recommendations for using a data-collection tool involving technology to capture exposure to food outlets, items and advertisements among school children in Lebanon and Tunisia. For this purpose, we adopted a user-centered design approach to involve school students, their parents, and school staff, in the design of this data-collection tool. We conducted several qualitative interactive design workshops whereby we discussed potential methods to capture children’s exposure to food in their immediate environment, their advantages and challenges of these methods, and whether using a technology-based tool can be suitable and acceptable in these contexts [31]. In the following, we give an overview of the study, summarize its findings, and describe the data-collection protocol designed collectively with the participants in the qualitative workshops. ## 2.1.1 Workshops and main findings Our user-centered design study was based on a purposive sample of primary schools from Greater Beirut and Greater Tunis. To ensure maximum variation in our sample, schools were selected from different socio-economic backgrounds. Two parallel workshops were conducted in each school: the first one with children from grades 5-6 (aged 10-12 years); and the second one with their parents and the schools’ directors and staff. Written consent and assent were obtained from all adults and children who participated in the workshops. After conducting 12 workshops in six schools ($$n = 2$$ in Greater Beirut and $$n = 4$$ in Greater Tunis), data saturation was reached. In total, 40 students, 31 parents and 17 school staff participated in the workshops, which included a range of activities such as discussions, storyboarding, mind-mapping, and brainstorming. The discussions were recorded and transcribed, the materials (e.g., storyboards, design boards) were collected, and data were thematically analyzed. Commonly reported challenges that we needed to account for in our system were: invading children’s and third parties’ privacy, distracting the school children during classes, and obtaining biased data. To overcome these challenges, most participants suggested that a passive-image capturing tool for a short period of time—such as a wearable camera with continuous footage—could be a safe and efficient method to capture children’s food exposures. This was the instrument of choice because it minimized the challenges that would have been encountered with active image capture, such as children’s distraction at schools and on the roads. However, for it to be acceptable and safe to use, the tool must follow certain conditions as suggested by the participants. In particular, the camera: Additionally, to protect anonymity and avoid privacy invasion, most participants proposed a system that automatically removes all non-food related images from the captured images, followed by parental manual screening to remove unwanted images that were not automatically filtered out. They also suggested automated blurring of faces in all captured footage. ## 2.1.2 Data-collection protocol Based on the ethical frameworks and the findings of the interactive workshops, we propose a study protocol to collect data on children’s food environment during a typical school day. Our protocol concerns children aged 11 to 12 years as our findings and the literature show that this is the youngest age group capable of appropriately using wearable cameras [32]. The protocol involves a machine-learning-based tool, which includes: When conducting research with the proposed tool, ethical approval must be granted from the appropriate ethics committees/institutions. The protocol also specifies that children are only allowed to wear the camera outside the confines of school and have the option to turn the cameras off when needed. Once data are collected, the research team picks up the cameras from the children the following day when children’s parents are also invited to the school. During this session, the cameras’ footage is transferred to a password-protected computer using a USB cable, in a private room and in the presence of the parents only. The desktop application extracts a frame every 10 seconds from the recorded footage and deletes all the video files once extraction is completed. The rationale behind extracting a frame every 10 seconds is to avoid too many duplicate images capturing the same scene. This application also: Then, each folder is transferred into one tablet and deleted from the desktop application. Parents are given a tablet that contains the folder specific to their child and are asked to screen and permanently delete any unwanted images. The researchers will not view any recordings and/or images before the end of the two steps of the filtering process. Finally, the filtered set of images gets transferred from the tablets into a password-protected computer maintained by the primary investigator and are then subjected to subsequent machine learning models that classify the retained images into predefined relevant classes such as food consumption, food advertisements and food outlets. The original folders are then permanently deleted from the tablets following the transfer. We adopted this protocol in a real-world case study in Tunisia, described in Section 3.4. ## 2.2 Machine-learning-based data-collection system Building on the findings of the user-centered design study described in the previous section, we developed a machine-learning-based data-collection system that captures school children’s exposure to food in their immediate environment using wearable cameras and machine learning. This system guarantees user privacy and minimizes the risk of invading children’s private life. In this section, we first give an overview of the camera model and then describe the machine learning models. Finally, we describe other aspects of the system, such as the face blurring technique and the desktop application used to manage the collected data. ## 2.2.1 Camera model To be able to conduct real-world studies to capture food exposure among school children using wearable cameras, the cameras should: The camera’s price was also considered to ensure that the system can be used in large-scale studies. We surveyed over 40 different wearable-camera models retrieved from an online search that was conducted at the time of the user-centered design study (Spring 2019). We ordered a sample of the most adequate models ($$n = 4$$) to pilot-test them in the field and we opted for the MIUFLY 2K Body Pro model as this camera met all the above specifications. It had an acceptable size and weight, a good resolution, a wide-angle image sensor, a long battery life (9 hours of continuous video recording at 720p), an Infrared sensor, a password-protection option, a silent mode that disables all LED lights and sounds, a vibration motor for operation feedback, a 32 GB built-in flash storage and is resistant to water and shock. ## 2.2.2 Food exposure detection model The goal of the food exposure detection model is to automatically classify a given image as a “food exposure” or a “non-food exposure” image. We consider that an image contains food exposure if it displays anything associated with food such as food items, food markets, food outlets, food ads, restaurants, vending machines, etc. Here, the term food also includes beverages. Dataset. To the best of our knowledge, there are no available datasets that contain images labeled as either containing food exposure or not. For this reason, we opted to build our own dataset. The dataset was automatically constructed and labeled using an open source Web crawler that crawls images from Google and Microsoft Bing search engines [33]. We crawled two classes of images as follows: In total, we crawled more than one million images ($$n = 1$$,037,459 images) for around one month from both Google ($85\%$) and Microsoft Bing ($15\%$). This 85-15 split occurred because crawling using Google was faster and returned more images than Microsoft Bing. In addition, the used crawler API provided more filtering and control options when using Google. Each image was labeled as food exposure or not based on the query used to retrieve it. In this work, we relied on the accuracy of the Web search engines to label our dataset. We acknowledge that relying on the search engines to label the data might result in some incorrect labels. However, since we acquired a very large dataset and we selected the queries to retrieve the images in each class carefully, we hypothesized that the percentage of incorrectly labeled images would be negligible and would not adversely affect the training of the machine learning models. To validate this, we randomly selected four subsets of the dataset, three containing 1000 images each and one 2000 images, and we manually validated the labels of each image in the four samples. The percentage of incorrectly labeled images was $10\%$ on average, as shown in Table 1. **Table 1** | Unnamed: 0 | Sample Size | Correctly Labeled | Incorrectly Labeled | Noise Percentage | | --- | --- | --- | --- | --- | | Sample 1 | 1000 | 907 | 93 | 9% | | Sample 2 | 1000 | 894 | 106 | 11% | | Sample 3 | 1000 | 916 | 84 | 8% | | Sample 4 | 2000 | 1762 | 238 | 12% | Next, we performed some pre-processing and filtering on the over one million crawled images. First, we eliminated duplicate images from the dataset, including those with different resolutions, retaining only 702,096 images. Second, we removed all the images with a resolution lower than 224 x 224, and all the images with transparent backgrounds or watermarks. Third and last, we re-balanced the dataset so that the number of food exposure images was equal to that of the non-food exposure ones by randomly deleting images from the bigger class. Thus, the final training dataset consists of approximately 510,000 images, equally balanced between the two classes. Model. We used the dataset described above to train a deep learning model that can automatically classify whether a given image contains food exposure or not (i.e., a binary-classification task). As a first step and as is custom in machine learning projects, we split our dataset as follows: $80\%$ for training (approx. 195,000 images per class), $10\%$ for validation (approx. 25,000 images per class), and $10\%$ for testing (approx. 25,000 images per class). We then trained five different convolutional neural network (CNN) models, using state-of-the-art architectures in deep learning for computer vision, namely VGG16 [34], VGG19 [34], MobileNet V1 [27], and two custom CNNs that we developed from scratch. We then used the validation set to select the best model, the MobileNet V1 model, which we describe in details next. Our food exposure detection model was a pretrained MobileNet V1 model [27] with weights set based on ImageNet [35]. MobileNet V1 is one of the most heavily-used deep learning models these days. It has achieved superior results compared to many other models for various computer-vision tasks such as the ImageNet competition. Moreover, MobileNet is designed to run efficiently on Mobile phones and thus requires much less computational power compared to other popular models. The whole MobileNet network involves only 4.8 million weights. To train our model, we loaded MobileNet V1 with its pre-trained ImageNet weights, then the last layer with 1000 classes, corresponding to the ImageNet classes, was modified to represent two classes instead. We then trained this model by updating the weights of all the layers (i.e., training from scratch). Training the above-described model took place on a local workstation with the following specifications: a Dual Intel Xeon Gold CPU with 28 physical cores and 56 logical cores (threads), a NVIDIA Quadro P2000 GPU with 5 GB of dedicated GPU memory, 64 GB DDR3 RAM, 6 Mbps Network Bandwidth, and 1 TB NVME SSD storage. The total training time on the local workstation took around 2-3 months. We tuned the model’s hyperparameters using the validation set and random search. In brief, we used the Adam optimizer with a learning rate raging between 0.0001 and 0.05, and a learning weight decay of 0.0001. The dropout rate was 0.5 and the batch size was set to 32 since the GPU memory of the workstation was not big enough to support higher batch sizes. Data Augmentation was also employed using different techniques such as zooming, tilting, flipping, mirroring and shifting. Finally, we kept on training the model until the training accuracy either bypassed $98.5\%$ or flattened and stopped increasing by more than 0.01 for five successive epochs. ## 2.2.3 Food exposure classification model The goal of the food exposure model is to classify food-related images captured by the wearable cameras into a hierarchy of food exposure classes. The classes in the food exposure typology are: food consumption, food advertisement and food outlet. The food consumption class consists of images that contain food items that are being consumed or about to be consumed. The food advertisement class includes any ads that are related to food such as billboards, storefront ads, etc. Finally, the food outlet class includes images that contain a food outlet such as a supermarket, a shop, a restaurant, a kiosk, a cafe, etc. Dataset. We used three different datasets to train our food exposure model. The first dataset, which we refer to as the Children Trajectory dataset consists of food-related images captured using wearable cameras of 265 children from 29 schools in Tunisia. This dataset includes 3,560 food-related images and we used the Labelbox crowdsourcing platform [36] to manually classify the images in the dataset into one or more of our three relevant classes, namely food consumption, food advertisement and food outlet. To this end, we trained a team of five annotators, who worked at Labelbox as full-time labelers at the time of annotation, on our annotation task. The task was that for each image, they should select one or more of the following categories: personal food consumption, others food consumption, food outlet, and food advertisement. The segregation of food consumption into personal and others was done since our next machine learning model aims to classify food-consumption images into these two subclasses. To ensure high-quality annotations, we reviewed the labeled images through a voting system, where the incorrect labels were given down votes and then they were corrected by the annotators. Moreover, there was a direct communication with the annotators through a shared document where they can ask about ambiguous images. Table 2 shows the distribution of the annotated images over the different classes. **Table 2** | Class | Count | | --- | --- | | Personal Food Consumption | 1600 | | Others Food Consumption | 340 | | Personal and Others Food Consumption | 940 | | Food Outlet | 380 | | Food Outlet and Advertisement | 70 | Since the number of images that belong to either the food outlet or the food advertisement classes in the Children Trajectory dataset was significantly low, we used a second dataset to train our food exposure model. This dataset consists of images of food outlets or food advertisements that are located in the neighborhoods of the children’s schools (within a range of 800-meter), and that were captured using trained data collectors. We refer to this dataset as the Neighborhood Mapping dataset. Table 3 shows the number of images that belong to each of the two classes (advertisements or outlets). **Table 3** | Class | Count | | --- | --- | | Food Outlet | 2048 | | Food Advertisement | 25 | | Food Outlet and Advertisement | 2130 | Finally, we also used the EgocentricFood dataset [37], which is a publicly-available dataset consisting of 5,038 food-related images that were taken using wearable cameras. From this dataset, we sampled 3000 food-consumption images, and 200 food-outlet images. Since the number of images that belong to the food advertisement class was still very small in all datasets, we additionally crawled food advertisement images using a major Web search engine (Google). Table 4 shows the number of images that belong to the different classes in each dataset. **Table 4** | Dataset | Food Consumption | Food Outlet | Food Outlet + Ads | Food Ads | | --- | --- | --- | --- | --- | | Children Trajectory | 2880 | 380 | 70 | - | | Neighborhood Mapping | - | 2048 | 2130 | 25 | | EgocentricFood | 3000 | 160 | - | - | | Crawled Ads | - | - | - | 512 | | Total | 5880 | 2588 | 2200 | 537 | Model. We combined all the datasets described above and used them to train a deep learning model to classify food exposure images into one or more of the following classes: food consumption (both personal or other), food outlet, and food advertisement. To ensure consistency among the images in the combined datasets, we resized all the images to 224 × 244. We then split our dataset into $80\%$ for training (8,965 images), $10\%$ for validation (1,120 images) and $10\%$ for testing (1,120 images), in a balanced way among the classes. We then trained three state-of-the-art CNN models, namely MobileNet V1 [27], MobileNet V2 [28], and VGG16 [34] and used the validation set to select the best model, the MobileNet V2 model, which we describe in details next. We loaded the MobileNet V2 model with pretrained ImageNet [35] weights. We then replaced the output layer of the pretrained MobileNet V2 model with a GlobalAveragePooling2D layer, followed by a dense layer consisting of 256 neurons and a dropout regularization layer (twice). Finally, we also added one last layer, which is a dense layer, consisting of three neurons, each with a Sigmoid activation function that outputs independent probabilities for each one of our three classes. Finally, to fine tune the model using our combined dataset, we unfroze the last 56 layers of our modified MobileNet V2 model and only trained those. We then tuned the hyperparameters of the model using the validation set and random search. Our best performing model was trained for 20 epochs using the Adam optimizer, with a batch size of 64, and a learning rate of 0.001. We then continued to train that model for 15 more epochs with a learning rate to 0.0001. The model was trained on the same local workstation that we used to train our first model, the food exposure detection model. ## 2.2.4 Food consumption classification model The goal of the food consumption model is to classify food-consumption images into two subsequent subclasses, namely personal food consumption and others food consumption. The personal food consumption class consists of images in which the child wearing the camera is consuming or about to consume food, whereas the others food consumption class consists of images in which other people are consuming or about to consume food. Note that an image can belong to both the personal food consumption and the others food consumption categories, and hence this is a multi-label classification task. Dataset. We constructed the dataset to train our food consumption model by extracting the food consumption images from of the Children Trajectory dataset that was used to train our food exposure classification model. This distribution of the food consumption images over the two subclasses is shown in Table 5. Fig 3 shows sample images for (a) personal food consumption, (b) others food consumption, and (c) personal and others food consumption. **Fig 3:** *Example food consumption images from the Children Trajectory dataset.(a) corresponds to personal food consumption, (b) corresponds to others food consumption, and (c) corresponds to both personal and others food consumption.* TABLE_PLACEHOLDER:Table 5 Model. To train our food consumption model using the subset of the Children Trajectory dataset corresponding to food consumption, we split the dataset into $70\%$ for training (2,016 images), $10\%$ for validation [297] and $20\%$ for testing [576] in a balanced way among the two classes. Our food consumption model was based on a pretrained MobileNet V2 model that has the same architecture as the food exposure model described above, but instead of using three neurons in the output layer, we used only two since we this model aims to classify images into two classes (personal food consumption and others food consumption). The hyperparameters of the model were then tuned using the validation set and random search. The model was trained for 25 epochs, with a learning rate of 0.001, Adam optimizer, and a batch size of 32, on the same local workstation used to train our first two machine learning models. ## 2.2.5 Face blurring technique A critical feature of our system is to blur all faces in any captured images automatically. To select the most suitable face blurring technique, we tested three different approaches: [1] the OPENCV’s library [38]; [2] an open source CNN named MTCNN [39], which was reported to have an accuracy of up to $90\%$ [39]; and [3] a python library called CVLIB, which performs face detection [40]. The first two approaches detect faces and draw a virtual bounding box around them. We ran the three techniques, separately, on two sampled sets of images retrieved from our first dataset crawled from the Web (2000 images each). The OPENCV’s library did not perform well and tended to blur objects instead of human faces. The MCTNN was able to detect and blur $88.91\%$ and $91.53\%$ of faces in the first and second samples, respectively. Only cropped faces or faces in the far background were usually not detected and hence not blurred. The CVLIB library yielded almost the same results as the MTCNN. Thus, we decided to use the MTCNN followed by the CVLIB library to maximize face-detection accuracy for our machine-learning-based data-collection system. To blur faces, we take the x-y coordinates of the virtual bounding boxes that contain faces and apply Gaussian Blurring to the bounding box area. The accuracy of face blurring, when both methods were used in a pipeline, was around $91\%$ on average. ## 2.2.6 Desktop application To retrieve the data stored on the wearable cameras, we developed a desktop application that uses the food exposure detection model (i.e, our first machine learning model) and the face blurring technique described above. The front-end of this application was written in Java, while the back-end used Python. The application was designed to retrieve data from a maximum of 10 cameras simultaneously. Using this application, the research team is able to [1] connect the cameras (on which the data is collected) to the computer through USB ports, and [2] enter the passwords of the cameras to enable USB access. The application then checks if all the cameras have footage and notifies the team of any empty cameras. Then, the research team members enter the school ID, child ID and camera ID onto the application to initiate data transfer. Since the cameras record continuous videos, the application first extracts a frame every 10 seconds and deletes all the video files once extraction is completed. It then employs the fist machine learning model to identify and retain all food exposure images in the collected data and discard any other footage. Next, the application uses the face blurring method to blur all the faces in the retained images. After the images are exported, filtered, and faces in them are blurred, the application generates a brief report of the total number and percentage of retained food exposure images. In case of any error during the process, a warning message is generated at the end describing the error and how it was handled. Our machine-learning-based data-collection tool was thoroughly designed so that no data loss would occur in case of any unexpected failure caused by a system or a hardware error at any point during the process while guaranteeing data protection throughout the entire process. ## 3.1 Food exposure detection model We tested our MobileNet V1 food exposure detection model described in Section 2.2.2 on the test data (i.e., the subset of its dataset consisting of around 25,000 images not used in training nor validation). We obtained a test accuracy of $92.53\%$ and a test F1-score of 0.9204. Table 6 shows the test performance of the model using various metrics. Our trained model and the dataset used to train it are publicly available [41]. **Table 6** | Metric | Accuracy | Precision | Recall | F1-Score | ROC AUC | | --- | --- | --- | --- | --- | --- | | Value | 0.9253 | 0.9541 | 0.8967 | 0.9204 | 0.9504 | ## 3.2 Food exposure classification model Table 7 show the results of our food exposure model described in Section 2.2.3 on its test data. As can be seen from the table, the model achieved very high results in terms of precision, recall and F1-score for all three classes. The food advertisement class had the lowest precision and recall since some of the test images that belong to that class contain food advertisements that are very small and thus barely visible in the images. **Table 7** | Class | Precision | Recall | F1-score | | --- | --- | --- | --- | | Food Consumption | 0.99 | 0.99 | 0.99 | | Food Outlet | 0.98 | 0.99 | 0.98 | | Food Advertisement | 0.95 | 0.93 | 0.93 | | Average | 0.97 | 0.97 | 0.96 | In addition, we used LIME [42], which stands for Local Interpretable Model-agnostic Explanation, to explain the model’s decisions on sample images by extracting the regions that are responsible for the classifier’s predictions. As can be seen in Fig 4, the model was able to correctly classify image (a) as food consumption by focusing on the region of pixels that contain food items. Similarly, the model was able to correctly classify image (b) as food outlet and advertisement by focusing on the region of pixels that corresponds to an outlet, and the region of pixels that corresponds to an advertisement. **Fig 4:** *LIME results on sample images for the food exposure model.* ## 3.3 Food consumption classification model The results of our food consumption model described in Section 2.2.4 on its test data are shown in Table 8. As can be seen from the table, the model achieved very high results in terms of precision, recall and F1-score for the personal food consumption class. However, the others food consumption model had a slightly lower recall since some of the images contained other people consuming food at the same table as the child wearing the camera, however, the camera was directed towards the table and thus the captured image did not clearly show the other people. **Table 8** | Class | Precision | Recall | F1-score | | --- | --- | --- | --- | | Personal Food Consumption | 0.97 | 0.99 | 0.98 | | Others Food Consumption | 0.97 | 0.89 | 0.93 | | Average | 0.97 | 0.94 | 0.95 | Similar to the previous model, we also used LIME to explain the model’s output on a sample image, which is shown in Fig 5. The sample image was classified as personal and others food consumption because there is a dish that is directly in front the camera wearer on one side, and there is another person who is consuming food at the same table on the other side. **Fig 5:** *LIME results on a sample image for the food consumption model.* ## 3.4 Case study Our machine-learning-based data-collection system described in the previous section was deployed in a case study in Tunisia and is currently being deployed in Lebanon [43]. In this section, we present the results from the cross-sectional study that was conducted only in Tunisia. ## 3.4.1 Setting Our case study involved a representative sample of 8-11 years-old school children (grades 4, 5 and 6) registered in schools in the Greater Tunis. A stratified two-stage sampling was used: first, a random sample of 50 schools was selected; then a random sample of 50 children were recruited from each school. Finally, we took a random subsample of around 10 children aged 11-12 years (grades 5 and 6) from each of the 50 schools. This study was approved by the relevant Ethics Committee at the National Institute of Nutritional and Food Technology (INNTA) in Tunisia. We adopted the protocol described in Section 2.1 to collect data using wearable cameras, including obtaining oral consent from schools’ directors, as well as parents and children. Fieldwork took place from January to March 2020 and then from October to November 2020 as Tunisian schools had to close in between in response to the emergent COVID-19 pandemic. We present here the results of the interim study sample that was recruited before the COVID-19 lockdown. In total, 265 children aged 11 to12 years old were recruited from 29 schools in Greater Tunis. Most schools ($86\%$) had 8 or more participants. ## 3.4.2 Results The throughput of our food exposure detection model was around 1,100 images per minute and the throughput of the face blurring module was close to 600 images every 4 minutes (around 150 images per minute). Overall, we managed to collect 788.26 hours (695.29 GB) of footage time and 567,506 images were extracted from this footage. We then ran our machine learning model and ended up with a total of 61,857 images (6.92 GB) related to food exposure (all the rest were deleted). Despite parents being invited to filter out any retained images, none of them attended the data review and manual filtering sessions. To assess the feasibility of our machine-learning-based data-collection system, we computed the school participation rate among our interim study sample. The school participation rate was calculated as the number of empty cameras returned by participants from each school, over the total number of participants in this school. Participation was considered excellent if the ratio ranges between 0 and 0.25, good if it ranges between 0.25 and 0.5, and shy if it ranges between 0.5 and 1. As can be seen in Fig 6, the majority of schools had an excellent participation ($52\%$). **Fig 6:** *School participation rate.* We further computed the participation rate at the child’s level and we show the results in Fig 7. Over $53\%$ of participants had excellent (i.e., more than 4 hours of footage recorded) or good (between 1 and 4 hours of footage) participation rate and only around $8\%$ did not participate at all in the study (i.e., had empty cameras). We received empty cameras as some children forgot to turn their cameras on and others changed their minds and decided not to wear them anymore. **Fig 7:** *Children participation rate.* Finally, to validate the effectiveness of the machine learning model in detecting food exposure images, we took a random sample of 25 children and manually inspected the corresponding final sets of retained images. The size of each set of images ranged between 100 and 500 depending on the amount of recorded footage of each child. Since parents did not filter any images in our case study, we were able to verify the model’s precision by identifying false-positive instances (i.e., images that were not actually related to food exposure but were retained). Since all other images that our machine learning model did not classify as food exposure were immediately deleted, we could not measure the model’s recall (i.e., false-negative rate). The model precision ranged between $82.23\%$ and $93.57\%$ depending on the sample (i.e., the false-positive rate ranged between $6.43\%$ and $17.77\%$). Our machine learning model was thus able to generalize well on real data despite the fact that the model was trained on data collected from the Web. Upon analyzing misclassified images, we observed that many of them were blurry due to fast motion or because they were taken in low-light settings. Fig 8 shows a sample of the food exposure images that were retained by our data-collection tool and in which all the faces were automatically blurred. **Fig 8:** *Some example food exposure images successfully captured by our system.* Next, we applied the food exposure and the food consumption classification models on the food exposure images detected by the first machine learning model as described above. The food exposure model classified the vast majority of the food exposure images as food consumption ($95\%$) and only $4\%$ as food outlets. Among the food consumption images, $69\%$ were classified by the food consumption model as personal food consumption, only $8\%$ were classified as others food consumption, and the remaining $23\%$ were classified as both personal and others food consumption. ## 3.5 Summary and outlook In this manuscript, we described an end-to-end machine-learning-based system to capture food exposure among school children using wearable cameras. Our system has several technical features to ensure that data collection is performed ethically, securely, and in a scalable manner. The development of the system’s features was guided by a user-centered design study on the acceptability and feasibility of using wearable cameras among children. Our system employs three different deep learning models, which are able to automatically detect images related to food exposure and to classify those into fine-grained classes. These classes are images that contain food items consumed by the child wearing the camera or others, images that contain food advertisements and images that contain food outlets. To train the machine learning models used by our system, we constructed various large datasets of labeled images about food exposure, by crawling images from the Web and labeling images captured through wearable cameras using crowdsourcing. We deployed our system in a real case study and provided some analysis of the data collected in that study. We believe that our work paves the way for many more studies of this nature and provides many useful lessons related to the issue of using wearable technologies to document food experiences, particularly among children. We plan to use our trained models in another user study in Greater Beirut, Lebanon. Since our models detect whether an image contains food exposure or not, and for those that contain food exposure, what type of exposure it is, we believe our models do not require any further training and can be deployed successfully in any food context. In the future, we plan to train further machine learning models to automatically classify food consumption images based on their healthiness into various groups using the NOVA food classification system. The filtered and labeled images and metrics relative to each child will then be analyzed along with other individual-level, school-level and community-level variables collected to inform school and community-level policies that foster a healthier food environment for school children in Greater Beirut and Greater Tunis. ## References 1. 1Hoashi H, Joutou T, Yanai K. Image recognition of 85 food categories by feature fusion. In: 2010 IEEE International Symposium on Multimedia. IEEE; 2010. p. 296–301. 2. 2Yang S, Chen M, Pomerleau D, Sukthankar R. Food recognition using statistics of pairwise local features. In: 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. IEEE; 2010. p. 2249–2256. 3. 3Zong Z, Nguyen DT, Ogunbona P, Li W. On the combination of local texture and global structure for food classification. In: 2010 IEEE International Symposium on Multimedia. IEEE; 2010. p. 204–211. 4. 4Matsuda Y, Hoashi H, Yanai K. Recognition of multiple-food images by detecting candidate regions. In: 2012 IEEE International Conference on Multimedia and Expo. IEEE; 2012. p. 25–30. 5. Sobal J, Bisogni CA. **Constructing food choice decisions**. *Annals of Behavioral Medicine* (2009.0) **38** s37-s46. DOI: 10.1007/s12160-009-9124-5 6. Kearney J, Hulshof K, Gibney M. **Eating patterns–temporal distribution, converging and diverging foods, meals eaten inside and outside of the home–implications for developing FBDG**. *Public health nutrition* (2001.0) **4** 693-698. DOI: 10.1079/PHN2001156 7. Boushey CJ, Kerr DA, Wright J, Lutes KD, Ebert DS, Delp EJ. **Use of technology in children’s dietary assessment**. *European journal of clinical nutrition* (2009.0) **63** S50-S57. DOI: 10.1038/ejcn.2008.65 8. Hillier A, Smith T, Cannuscio CC, Karpyn A, Glanz K. **A discrete choice approach to modeling food store access**. *Environment and Planning B: Planning and Design* (2015.0) **42** 263-278. DOI: 10.1068/b39136 9. Williams J, Scarborough P, Matthews A, Cowburn G, Foster C, Roberts N. **A systematic review of the influence of the retail food environment around schools on obesity-related outcomes**. *Obesity reviews* (2014.0) **15** 359-374. DOI: 10.1111/obr.12142 10. Heard AM, Harris JL, Liu S, Schwartz MB, Li X. **Piloting an online grocery store simulation to assess children’s food choices**. *Appetite* (2016.0) **96** 260-267. DOI: 10.1016/j.appet.2015.09.020 11. Gemming L, Doherty A, Utter J, Shields E, Mhurchu CN. **The use of a wearable camera to capture and categorise the environmental and social context of self-identified eating episodes**. *Appetite* (2015.0) **92** 118-125. DOI: 10.1016/j.appet.2015.05.019 12. Zhao X, Xu X, Li X, He X, Yang Y, Zhu S. **Emerging trends of technology-based dietary assessment: a perspective study**. *European Journal of Clinical Nutrition* (2021.0) **75** 582-587. DOI: 10.1038/s41430-020-00779-0 13. 13Liu J, Johns E, Atallah L, Pettitt C, Lo B, Frost G, et al. An intelligent food-intake monitoring system using wearable sensors. In: 2012 Ninth International Conference on Wearable and Implantable Body Sensor Networks. IEEE; 2012. p. 154–160. 14. Signal LN, Stanley J, Smith M, Barr M, Chambers TJ, Zhou J. **Children’s everyday exposure to food marketing: an objective analysis using wearable cameras**. *International Journal of Behavioral Nutrition and Physical Activity* (2017.0) **14** 1-11. DOI: 10.1186/s12966-017-0570-3 15. McKerchar C, Smith M, Stanley J, Barr M, Chambers T, Abel G. **Food store environment examination–FoodSee: a new method to study the food store environment using wearable cameras**. *Global health promotion* (2020.0) **27** 73-81. DOI: 10.1177/1757975919859575 16. Liu W, Barr M, Pearson AL, Chambers T, Pfeiffer KA, Smith M. **Space-time analysis of unhealthy food advertising: New Zealand children’s exposure and health policy options**. *Health promotion international* (2020.0) **35** 812-820. DOI: 10.1093/heapro/daz083 17. 17Gao A, Lo FPW, Lo B. Food volume estimation for quantifying dietary intake with a wearable camera. In: 2018 IEEE 15th International Conference on Wearable and Implantable Body Sensor Networks (BSN). IEEE; 2018. p. 110–113. 18. 18Shroff G, Smailagic A, Siewiorek DP. Wearable context-aware food recognition for calorie monitoring. In: 2008 12th ieee international symposium on wearable computers. IEEE; 2008. p. 119–120. 19. Davies A, Chan V, Bauman A, Signal L, Hosking C, Gemming L. **Using wearable cameras to monitor eating and drinking behaviours during transport journeys**. *European Journal of Nutrition* (2021.0) **60** 1875-1885. DOI: 10.1007/s00394-020-02380-4 20. Gage R, Girling-Butcher M, Joe E, Smith M, Ni Mhurchu C, McKerchar C. **The Frequency and Context of Snacking among Children: An Objective Analysis Using Wearable Cameras**. *Nutrients* (2021.0) **13** 103. DOI: 10.3390/nu13010103 21. Doulah A, Ghosh T, Hossain D, Imtiaz MH, Sazonov E. **Automatic Ingestion Monitor Version 2–A Novel Wearable Device for Automatic Food Intake Detection and Passive Capture of Food Images**. *IEEE Journal of Biomedical and Health Informatics* (2020.0) **25** 568-576. DOI: 10.1109/JBHI.2020.2995473 22. Jia W, Li Y, Qu R, Baranowski T, Burke LE, Zhang H. **Automatic food detection in egocentric images using artificial intelligence technology**. *Public health nutrition* (2019.0) **22** 1168-1179. DOI: 10.1017/S1368980018000538 23. Jia W, Chen HC, Yue Y, Li Z, Fernstrom J, Bai Y. **Accuracy of food portion size estimation from digital pictures acquired by a chest-worn camera**. *Public health nutrition* (2014.0) **17** 1671-1681. DOI: 10.1017/S1368980013003236 24. 24Hossain D, Imtiaz MH, Ghosh T, Bhaskar V, Sazonov E. Real-Time Food Intake Monitoring Using Wearable Egocnetric Camera. In: 2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC). IEEE; 2020. p. 4191–4195. 25. Briggs L, Lake AA. **Exploring school and home food environments: perceptions of 8–10-year-olds and their parents in Newcastle upon Tyne, UK**. *Public health nutrition* (2011.0) **14** 2227-2235. DOI: 10.1017/S1368980011001984 26. 26Joutou T, Yanai K. A food image recognition system with multiple kernel learning. In: 2009 16th IEEE International Conference on Image Processing (ICIP). IEEE; 2009. p. 285–288. 27. 27Howard AG, Zhu M, Chen B, Kalenichenko D, Wang W, Weyand T, et al. Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:170404861. 2017. 28. 28Sandler M, Howard A, Zhu M, Zhmoginov A, Chen LC. Mobilenetv2: Inverted residuals and linear bottlenecks. In: Proceedings of the IEEE conference on computer vision and pattern recognition; 2018. p. 4510–4520. 29. Nebeker C, Lagare T, Takemoto M, Lewars B, Crist K, Bloss CS. **Engaging research participants to inform the ethical conduct of mobile imaging, pervasive sensing, and location tracking research**. *Translational behavioral medicine* (2016.0) **6** 577-586. DOI: 10.1007/s13142-016-0426-4 30. Mok TM, Cornish F, Tarr J. **Too much information: visual research ethics in the age of wearable cameras**. *Integrative Psychological and Behavioral Science* (2015.0) **49** 309-322. DOI: 10.1007/s12124-014-9289-8 31. Semaan A, Zoughby Y, Akl C, Talhouk R, Gharbia HB, Doggui R. **Developing an innovative Tool to Quantify Schoolchildren’s Exposure to Obesogenic Environments**. *Current Developments in Nutrition* (2020.0) **4** 1178-1178 32. Kelly P, Marshall SJ, Badland H, Kerr J, Oliver M, Doherty AR. **An ethical framework for automated, wearable cameras in health behavior research**. *American journal of preventive medicine* (2013.0) **44** 314-319. DOI: 10.1016/j.amepre.2012.11.006 33. 33Li Y, Wang Y, Tian E. A new architecture of an intelligent agent-based crawler for domain-specific deep web databases. In: 2012 IEEE/WIC/ACM International Conferences on Web Intelligence and Intelligent Agent Technology. vol. 1. IEEE; 2012. p. 656–663. 34. 34Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:14091556. 2014. 35. 35Deng J, Dong W, Socher R, Li LJ, Li K, Fei-Fei L. Imagenet: A large-scale hierarchical image database. In: 2009 IEEE conference on computer vision and pattern recognition. Ieee; 2009. p. 248–255. 36. 36Labelbox;. https://labelbox.com/. 37. 37Bolanos M, Radeva P. Simultaneous food localization and recognition. In: 2016 23rd International conference on pattern recognition (ICPR). IEEE; 2016. p. 3140–3145. 38. 38Culjak I, Abram D, Pribanic T, Dzapo H, Cifrek M. A brief introduction to OpenCV. In: 2012 proceedings of the 35th international convention MIPRO. IEEE; 2012. p. 1725–1730. 39. 39Xiang J, Zhu G. Joint face detection and facial expression recognition with MTCNN. In: 2017 4th international conference on information science and control engineering (ICISCE). IEEE; 2017. p. 424–427. 40. 40Ponnusamy A. cvlib-high level Computer Vision library for Python; 2018. 41. 41Food Exposure Detection Model;. https://github.com/ZoulfikarShmayssani/Food-Exposure-Models. 42. Linardatos P, Papastefanopoulos V, Kotsiantis S. **Explainable ai: A review of machine learning interpretability methods**. *Entropy* (2020.0) **23** 18. DOI: 10.3390/e23010018 43. 43Ghattas H, Jamaluddine Z, Semaan A, El-Helou N, Safadi G, Elghossain T, et al. School and Community Drivers of Child Diets in Two Arab Cities: the SCALE protocol and innovative tools to assess children’s food environments;.
--- title: 'Psychosocial health of school-going adolescents during the COVID-19 pandemic: Findings from a nationwide survey in Bangladesh' authors: - Kamrun Nahar Koly - Md. Saiful Islam - Marc N. Potenza - Rashidul Alam Mahumud - Md. Shefatul Islam - Md. Salim Uddin - Md. Afzal Hossain Sarwar - Farzana Begum - Daniel D. Reidpath journal: PLOS ONE year: 2023 pmcid: PMC10042372 doi: 10.1371/journal.pone.0283374 license: CC BY 4.0 --- # Psychosocial health of school-going adolescents during the COVID-19 pandemic: Findings from a nationwide survey in Bangladesh ## Abstract ### Background Common psychosocial health problems (PHPs) have become more prevalent among adolescents globally during the COVID-19 pandemic. However, the psychosocial health of school-going adolescents has remained unexplored in Bangladesh due to limited research during the pandemic. The present study aimed to estimate the prevalence of PHPs (i.e., depression and anxiety) and assess associated lifestyle and behavioral factors among school-going adolescents in Bangladesh during the COVID-19 pandemic. ### Methods A nationwide cross-sectional survey was conducted among 3,571 school-going adolescents (male: $57.4\%$, mean age: 14.9±1.8 years; age range: 10–19 years) covering all divisions, including 63 districts in Bangladesh. A semi-structured e-questionnaire, including informed consent and questions related to socio-demographics, lifestyle, academics, pandemic and PHPs, was used to collect data between May and July 2021. ### Results The prevalence of moderate to severe depression and anxiety were $37.3\%$ and $21.7\%$, respectively, ranging from $24.7\%$ in the Sylhet Division to $47.5\%$ in the Rajshahi Division for depression, and from $13.4\%$ in the Sylhet Division to $30.3\%$ in the Rajshahi Division for anxiety. Depression and anxiety were associated with older age, reports of poor teacher cooperation in online classes, worries due to academic delays, parental comparison of academic performance with other classmates, difficulties coping with quarantine situations, changes in eating habits, weight gain, physical inactivity and having experienced cyberbullying. Moreover, being female was associated with higher odds of depression. ### Conclusions Adolescent psychosocial problems represent a public health problem. The findings suggest a need for generating improved empirically supported school-based psychosocial support programs involving parents and teachers to ensure the well-being of adolescents in Bangladesh. School-based prevention of psychosocial problems that promote environmental and policy changes related to lifestyle practices and active living should be developed, tested, and implemented. ## Background Adolescence is an important developmental transition period between childhood and adulthood that includes multiple physical, cognitive and psychosocial changes [1]. In some cases, these changes have lasting direct and indirect negative effects on mental and physical health, academic performance, and subsequent life opportunities [2–4]. Depression and anxiety are the two most common mental health problems during childhood and adolescence [5, 6]. Morever, the COVID-19 pandemic, and the disease control measures that were implemented created additional, unforeseen stressors for adolescents. Stressors included academic delays, the uncertainty of the future, financial concerns, and worry about becoming infected. These were all reported as additional contributors to mental health issues in adolescents. Based on pre-COVID data, the World Health Organization (WHO) estimated that mental health problems accounted for $16\%$ of the global burden of disease and injury among people aged 10–19 years [1]. A systematic review and meta-analysis reported that the global pooled prevalence of mental health problems was $13.4\%$ among children and adolescents [7]. While mental health problems in adolescence frequently resolve [8, 9], they can lead to complex and severe mental illness in later life if unsolved [10]. In the transitional period of adolescence, commonly reported mental health problems include depression, anxiety, and attention deficit hyperactivity disorder (ADHD), as well as conduct, eating, psychotic, and substance use disorders [1]. The National Mental Health Survey of Bangladesh 2019 reported a $13.6\%$ overall prevalence of any mental health disorders in individuals aged 7–17 years and a $16.8\%$ prevalence in those aged 18 years or older [11]. Moreover, several other pre-COVID (Coronavirus disease 2019) surveys conducted among school-going adolescents in Bangladesh reported frequent anxiety ($18.1\%$), depression ($25\%$-$49\%$), and conduct disorders ($8.9\%$), as well as suicidal behavior ($11.7\%$) [4, 12–16]. COVID-19 was recognized as a global pandemic on March 11, 2020 [17]. To reduce the transmission of the virus, countries around the world implemented various public health measures. In Bangladesh, the measures included social distancing, physical lockdown of areas, and the closure of all educational institutions from March 2020 to September 2021 [18–20]. As a result of the measures, about 38 million students and one million teachers from primary, secondary, and tertiary education institutes were in lockdown [21]. The lockdown period resulted in decreased physical activity, more screen time, irregular sleep patterns, increased loneliness and poorer dietary habits [22, 23]. Globally, during the pandemic period, increases were reported in multiple mental health conditions among adolescents, including anxiety, depression, trauma, grief, suicidal behaviors, and substance [24, 25]. In Bangladesh, several studies investigating COVID-19 impacts on mental health were conducted [26–37]. However, very few studies looked at mental health in adolescents and instead focused on specific groups (university students, healthcare workers) or populations (slum dwellers, general population). The lack of information about pandemic related mental health problems in adolescents is a critical gap. Understanding the gap can inform the development of interventions aimed at promoting wellbeing during and beyond pandemics. This was, in fact, the motivation for the current study, which was to be used to inform a teacher led intervention supporting adolescent mental health [38, 39]. Given the paucity of data on adolescent mental health in Bangladesh during the pandemic, we sought to estimate the prevalence of the two most common issues (depression and anxiety) as well as their sociodemographic and behavioral correlates. We hypothesized that the prevalence estimates of depression and anxiety would be elevated among school-going adolescents during the COVID-19 pandemic in Bangladesh. We also hypothesized that depression and anxiety would be associated with select socio-demographic, educational, and pandemic related factors. ## Study design A cross-sectional survey was conducted by “Aspire to Innovate” (a2i; https://a2i.gov.bd/), an initiative of the Bangladeshi government. The survey investigated the psychosocial health of the Bangladeshi school-going adolescents during the COVID pandemic [40]. The survey was conducted using a2i’s online “edutainment” platform designed to nurture the educational, psychosocial and life skills of school-going adolescents in Bangladesh [41]. ## Study setting and participants a2i is a governmental program in the Information and Communication Technology (ICT) Division of Bangladesh that is supported by the Cabinet Division and United Nations Development Programme (UNDP)[41]. a2i is a program of the government’s Digital *Bangladesh agenda* which has an edutainment online platform named "Kishore Batayan–Konnect" that has been developed for school-going adolescents [41]. a2i aims to nurture the educational, psychosocial, and life skills of school-going adolescents in Bangladesh[42]. Adolescents can share and learn essential life lessons from different creative multimedia content that can help advance their social and personal skills. Working with researchers from icddr,b (International Centre for Diarrhoeal Disease Research, Bangladesh), a2i developed a survey to identify common psychosocial problems among students and their correlates. The data were used to inform the psychosocial skills training for teachers supporting adolescent mental health. The inclusion criteria for gparticipants included being: (ⅰ) adolescents enrolled in secondary or higher secondary school and enrolled in Konnect, (ⅱ) students of the selected teachers involved in the a2i needs assessment, (iii) willing to participate in the survey with parental consent. The cross-sectional survey was conducted between January and August of 2021. Initially, 3,729 surveys were submitted using the Google Forms. After removing incomplete and missing data, overall 3,571 school-going adolescents from the 63 districts (out of 64, one district did not have internet connection) of all eight divisions of Bangladesh were included in the final analysis. ## Study variables A semi-structured questionnaire was developed that included informed consent as well as questions related to socio-demographics, lifestyle, and academic and other factors. Two scales were included to assess depression and anxiety—the Patient Health Questionnaire (PHQ-9) and Generalized Anxiety Disorder (GAD-7). A shareable link using Google Forms was generated for the online survey. The survey was made available to the selected teachers to obtain the data from the selected students who had parents’ consent. The survey took approximately 15–20 minutes for each participant. ## Outcome variables The two outcome variables were depression measured using the PHQ-9 and anxiety measured using the GAD-7. Both of these scales had been used in previous studies in Bangladesh, including school-going adolescents, and had validated and translated Bangla versions [4, 12, 42, 43]. ## Patient Health Questionnaire (PHQ-9) Participants’ levels of depression were assessed using the validated Bangla version of the PHQ-9 scale [44, 45]. The PHQ-9 consists of nine questions assessed using a four-point Likert-type scale (from "0 = not at all" to "3 = almost every day") based on self-reported experiences during the prior two weeks. Total scores are obtained by summing the raw scores of the nine questions, generating a range from 0 to 27 [46]. Higher scores reflect more severe depression. Five predefined cutoff points were used to determine severity levels of depression: i) minimal: 0–4; ii) mild: 5–9; iii) moderate: 10–14; iv) severe: 15–19; and, v) extremely severe: 20 or higher [4, 12, 42, 43, 46]. In analyses, a score of 10 or higher was treated as a positive indicator of moderate to severe depression [4, 12, 42, 43]. The *Cronbach alpha* of the PHQ-9 was 0.89 in the present study. ## Generalized Anxiety Disorder (GAD-7) Participants’ anxiety levels were assessed using the Bangla version of the GAD-7 scale [47]. The GAD-7 consists of seven questions concerning problems related to symptoms of anxiety over the last two weeks. Responses were recorded on a four-point Likert-type scale (from "0 = not at all" to "3 = almost every day"). Total scores were obtained by summing the raw scores of the seven questions giving an individual score ranging from 0 to 21 [48]. Higher scores reflect more severe anxiety. Four predefined cut-off points were used to determine severity levels of anxiety: i) minimal: 0–4; ii) mild: 5–9; and iii) moderate: 10–14; and iv) severe: 15–21 [12, 42, 43, 47]. During analyses, a score of 10 or higher was treated as a positive indicator of moderate to severe anxiety [12, 42, 43, 47]. The *Cronbach alpha* of the GAD-7 was 0.84 in the present study. ## Socio-demographic and academic information Socio-demographics including age, sex (male/ female), and geographic division (Dhaka/ Barisal/ Chittagong/ Khulna/ Mymensingh/ Rajshahi/ Rangpur/ Sylhet) were obtained from participants. In addition, academic information including school type (secondary school/ higher secondary school) and attending online class (yes/ no) were also recorded. The students were also asked about the cooperation by the teachers in online classes (i.e., never/ sometimes/ always) and also if their parents compare them with their classmates (i.e., none/ some/ much). ## Lifestyle measures Participants were asked about their eating behaviors (“Have your eating behaviors changed during the lockdown period?”), body weight changes (“Was there any change in your weight during the lockdown period?”), and physical exercise (“Did you engage in any light physical activity during the lockdown period?”) during the COVID-19-related lockdown period. This section also included questions about the availability of internet connection at home, ways of communicating with peers during the lockdown, and possible problems during the pandemic (loneliness/ missing friends/ disruption of studies/ poor academic performance). ## Cyberbullying To assess cyberbullying, the participants were asked a single-item question (i.e., Do your classmates bully you online?) with three possible responses (i.e., yes/ sometimes/ no). ## Statistical analysis The dataset was cleaned, coded, and sorted using the Microsoft Excel (version 2019) before being imported into the Statistical Package for the Social Sciences (SPSS, version 25.0). Descriptive statistics, including frequencies and percentages, were computed for categorical variables; means and standard deviations were used for continuous variables. Bivariate analyses (i.e., Chi-square tests or t-tests where appropriate) were performed to explore associations between explanatory and outcome variables. The potential multi-collinearity was checked by using tolerance (> 0.1) and variance inflation factor (VIF < 10) before performing regression analysis. Variables were significant at $p \leq 0.05$ in the binary logistic regression analyses were subsequently included in multivariable logistic regression models to determine the factors associated with the outcome variables (depression and anxiety). Crude and adjusted odds ratios (CORs and AORs) were reported from regression models with $95\%$ confidence intervals. A p-value less than 0.05 was considered statistically significant in this exploratory study. ## Ethics This survey was conducted by a2i following the study protocol having been reviewed and approved by the Intuitional Review Board (IRB) of icddr,b. The survey was conducted in accordance with guidelines outlined in the Helsinki declaration. Respondents participated in the survey willingly in an informed fashion without compensation. Students’ parents were informed about the survey. The study’s objectives, risks and benefits of participation, and options to not participate in the survey were presented in the informed consent section. E-written consent were obtained from participants and their parents. ## General characteristics of the sample A total of 3,571 school-going adolescents were included in the final analysis. Students from the 6th through 12th grades and aged 10–19 years (mean age = 14.90 [SD = 1.80]) participated. Respondents were from 63 of the 64 districts that comprised the eight divisions of Bangladesh. Most respondents were male ($57.52\%$), many came from the Sylhet division ($24.67\%$), and most were from secondary schools (grades 6–10; $85.33\%$) (Table 1). **Table 1** | Variables | Number of participants n (%) | Percentages with depression (n = 1332; 37.30%) | Percentages with depression (n = 1332; 37.30%).1 | Percentages with anxiety (n = 774; 21.67%) | Percentages with anxiety (n = 774; 21.67%).1 | Percentages with depression and anxiety (n = 696; 19.49%) | Percentages with depression and anxiety (n = 696; 19.49%).1 | | --- | --- | --- | --- | --- | --- | --- | --- | | Variables | Number of participants n (%) | % (95% CI) | p-value | % (95% CI) | p-value | % (95% CI) | p-value | | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | | Age (mean ± SD) | 14.90±1.80 | 15.44±1.77 | <0.001‡ | 15.56±1.75 | <0.001‡ | 15.60±1.72 | <0.001‡ | | Sex | Sex | Sex | Sex | Sex | Sex | Sex | Sex | | Male | 2054 (57.52) | 35.74 (33.68–37.83) | 0.024† | 20.59 (18.89–22.38) | 0.068† | 18.26 (16.63–19.97) | 0.030† | | Female | 1517 (42.48) | 39.42 (36.98–41.90) | | 23.14 (21.07–25.31) | | 21.16 (19.16–23.27) | | | Division | Division | Division | Division | Division | Division | Division | Division | | Dhaka | 865 (24.22) | 41.73 (38.48–45.04) | <0.001† | 23.24 (20.51–26.14) | <0.001† | 21.39 (18.75–24.21) | <0.001† | | Barisal | 472 (13.22) | 36.44 (32.19–40.85) | | 20.76 (17.29–24.59) | | 18.01 (14.74–21.66) | | | Chittagong | 441 (12.35) | 38.32 (33.87–42.92) | | 24.26 (20.44–28.42) | | 21.32 (17.69–25.32) | | | Khulna | 258 (7.22) | 46.12 (40.11–52.22) | | 28.29 (23.06–34.01) | | 26.74 (21.62–32.38) | | | Mymensingh | 50 (1.40) | 34.00 (22.06–47.74) | | 18.00 (9.30–30.28) | | 18.00 (9.30–30.28) | | | Rajshahi | 333 (9.33) | 47.45 (42.13–52.81) | | 30.33 (25.58–35.42) | | 26.43 (21.91–31.35) | | | Rangpur | 271 (7.59) | 43.54 (37.73–49.49) | | 24.72 (19.87–30.11) | | 23.99 (19.2–29.33) | | | Sylhet | 881 (24.67) | 24.74 (21.98–27.67) | | 13.39 (11.27–15.76) | | 11.46 (9.49–13.69) | | | Academic information | Academic information | Academic information | Academic information | Academic information | Academic information | Academic information | Academic information | | Academic class | Academic class | Academic class | Academic class | Academic class | Academic class | Academic class | Academic class | | Secondary school | 3047 (85.33) | 33.51 (31.85–35.20) | <0.001† | 18.84 (17.48–20.26) | <0.001† | 16.77 (15.48–18.13) | <0.001† | | Higher secondary school | 524 (14.67) | 59.35 (55.10–63.50) | | 38.17 (34.08–42.38) | | 35.31 (31.3–39.47) | | | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | | Bad | 933 (26.13) | 50.16 (46.96–53.36) | <0.001† | 32.37 (29.42–35.42) | <0.001† | 29.15 (26.31–32.13) | <0.001† | | Moderate | 1585 (44.39) | 36.28 (33.94–38.67) | | 20.88 (18.94–22.94) | | 18.93 (17.06–20.91) | | | Good | 1053 (29.49) | 27.45 (24.81–30.2) | | 13.39 (11.43–15.55) | | 11.78 (9.93–13.83) | | | Attending classes online | Attending classes online | Attending classes online | Attending classes online | Attending classes online | Attending classes online | Attending classes online | Attending classes online | | Yes | 2761 (77.32) | 33.14 (31.40–34.91) | <0.001† | 18.18 (16.78–19.65) | <0.001† | 16.04 (14.71–17.45) | <0.001† | | No | 810 (22.68) | 51.48 (48.04–54.91) | | 33.58 (30.39–36.89) | | 31.23 (28.11–34.49) | | | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | | Never | 356 (9.97) | 60.39 (55.25–65.37) | <0.001† | 42.70 (37.63–47.88) | <0.001† | 40.17 (35.17–45.32) | <0.001† | | Sometimes | 654 (18.31) | 46.64 (42.83–50.47) | | 29.05 (25.67–32.62) | | 26.61 (23.33–30.09) | | | Always | 2561 (71.72) | 31.71 (29.93–33.53) | | 16.87 (15.46–18.36) | | 14.80 (13.46–16.21) | | | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | | | 1192 (33.38) | 23.66 (21.31–26.13) | <0.001† | 13.26 (11.42–15.27) | <0.001† | 10.99 (9.31–12.86) | <0.001† | | Some | 800 (22.40) | 34 (30.78–37.34) | | 19.25 (16.63–22.09) | | 17.00 (14.52–19.72) | | | Much | 1579 (44.22) | 49.27 (46.81–51.74) | | 29.26 (27.05–31.54) | | 27.17 (25.02–29.4) | | | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | | | 448 (12.55) | 16.74 (13.5–20.41) | <0.001† | 10.49 (7.91–13.58) | <0.001† | 7.14 (5.03–9.81) | <0.001† | | Some | 975 (27.30) | 18.97 (16.61–21.53) | | 9.44 (7.72–11.39) | | 7.59 (6.05–9.38) | | | Much | 2148 (60.15) | 49.91 (47.79–52.02) | | 29.56 (27.66–31.52) | | 27.47 (25.61–29.38) | | | Lifestyle during the pandemic | Lifestyle during the pandemic | Lifestyle during the pandemic | Lifestyle during the pandemic | Lifestyle during the pandemic | Lifestyle during the pandemic | Lifestyle during the pandemic | Lifestyle during the pandemic | | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | | Poorly | 1225 (34.30) | 54.86 (52.06–57.63) | <0.001† | 36.00 (33.35–38.72) | <0.001† | 33.55 (30.95–36.23) | <0.001† | | Moderately | 1859 (52.06) | 30.23 (28.18–32.35) | | 14.63 (13.08–16.29) | | 12.69 (11.24–14.27) | | | Well | 487 (13.64) | 20.12 (16.75–23.85) | | 12.53 (9.81–15.69) | | 10.06 (7.63–12.97) | | | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | | No | 2002 (56.06) | 37.76 (35.66–39.9) | 0.519† | 22.38 (20.59–24.24) | 0.249† | 20.13 (18.42–21.93) | 0.276† | | Yes | 1569 (43.94) | 36.71 (34.35–39.12) | | 20.78 (18.83–22.84) | | 18.67 (16.8–20.66) | | | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | | Yes | 1378 (38.59) | 46.81 (44.18–49.45) | <0.001† | 30.70 (28.3–33.17) | <0.001† | 27.72 (25.41–30.13) | <0.001† | | Somewhat | 1134 (31.76) | 36.60 (33.83–39.43) | | 19.40 (17.18–21.78) | | 17.81 (15.67–20.12) | | | No | 1059 (29.66) | 25.68 (23.12–28.38) | | 12.37 (10.49–14.46) | | 10.58 (8.83–12.54) | | | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | | Loss | 896 (25.09) | 42.41 (39.2–45.67) | <0.001† | 23.33 (20.65–26.18) | <0.001† | 20.76 (18.2–23.51) | <0.001† | | Gain | 623 (17.45) | 55.86 (51.94–59.72) | | 37.88 (34.14–41.74) | | 34.83 (31.17–38.64) | | | Unaware | 1397 (39.12) | 31.07 (28.68–33.53) | | 17.75 (15.82–19.82) | | 16.11 (14.25–18.1) | | | No change | 655 (18.34) | 25.95 (22.71–29.41) | | 12.37 (10.01–15.05) | | 10.38 (8.22–12.89) | | | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | | Yes | 2100 (58.81) | 30.57 (28.63–32.57) | <0.001† | 16.24 (14.71–17.86) | <0.001† | 14.52 (13.07–16.08) | <0.001† | | No | 1471 (41.19) | 46.91 (44.36–49.46) | | 29.44 (27.15–31.80) | | 26.58 (24.37–28.88) | | | Victimization | Victimization | Victimization | Victimization | Victimization | Victimization | Victimization | Victimization | | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | | Yes | 567 (15.88) | 63.32 (59.29–67.21) | <0.001† | 41.98 (37.96–46.07) | <0.001† | 39.86 (35.89–43.93) | <0.001† | | No | 3004 (84.12) | 32.39 (30.73–34.08) | | 17.84 (16.51–19.24) | | 15.65 (14.38–16.98) | | ## Academic factors during the pandemic More than a quarter of participants reported poor internet connectivity at their homes ($26.1\%$) (Table 1). About $77.3\%$ of adolescents attended online classes. Most participants reported that their teachers were cooperative when they asked questions in online classes ($71.7\%$). Most reported worries due to academic delays ($60.2\%$), and many reported that their parents compared their academic performances with those of other classmates ($44.2\%$). ## Lifestyle during the pandemic Many adolescents ($34.3\%$) reported not coping well with lockdown during the pandemic (Table 1). Many reported feeling lonely ($43.9\%$), not participating in physical exercise ($41.2\%$), and changing their eating behaviors ($38.6\%$) during the lockdown period. Many participants reported weight loss ($25.1\%$) or weight gain ($17.5\%$), and others reported no change ($18.3\%$) or being unaware of changes in body weight ($39.1\%$). A number of participants experienced cyberbullying from their classmates ($15.8\%$). ## Prevalence of depression and anxiety Of the participants, $37.3\%$ reported moderate to extremely severe symptoms of depression; $62.7\%$ reported estimates of no or mild symptoms of depression (Fig 1). Of the participants, $21.7\%$ reported moderate to severe symptoms of anxiety; $78.3\%$ reported estimates of no or mild symptoms of anxiety (Fig 1). **Fig 1:** *Participants’ anxiety and depression levels.* Nearly one-fifth ($19.5\%$) of participants experienced both depression and anxiety, whereas $17.8\%$ had only depression and $2.2\%$ had only anxiety (Fig 2). **Fig 2:** *Presence of psychosocial health problems.* The conditional distribution of symptoms of depression and anxiety by age is shown in Fig 3. **Fig 3:** *Distribution of depression and anxiety with participants’ age (from 10 to 15 years).* ## Factors associated with depression and anxiety In bivariate analyses, depression and anxiety were each associated at $p \leq 0.05$ with most variables (Table 1). Multivariate logistic regression analyses were performed including all significant variables ($p \leq 0.05$) from binary logistic regression analyses (Table 2). The values of Cox & Snell R Square, and Nagelkerke R Square for multivariable logistic regression of depression were 0.27, and 0.36, respectively. Depression was associated with age (AOR: 1.12; $95\%$ CI: 1.05–1.19, $p \leq 0.001$), female gender (AOR: 1.26, $95\%$ CI: 1.05–1.50, $$p \leq 0.011$$), higher secondary school status (AOR: 1.50; $95\%$ CI: 1.14–1.98, $$p \leq 0.004$$), and reports of teachers not being cooperative when students asked questions in online classes (AOR: 1.74; $95\%$ CI: 1.32–2.29, $p \leq 0.001$), poor internet connection at home (AOR = 1.46; $95\%$ CI: 1.12–1.89, $$p \leq 0.005$$), worries about academic delay (AOR: 3.55; $95\%$ CI: 2.62–4.82, $p \leq 0.001$), parental comparisons of their academic performances with those of other classmates (AOR: 2.15; $95\%$ CI: 1.77–2.61, $p \leq 0.001$), difficulties coping with the quarantine situation (AOR: 2.96; $95\%$ CI: 2.21–3.95, $p \leq 0.001$), changes in eating behaviors during the pandemic (AOR: 1.55; $95\%$ CI: 1.26–1.91, $p \leq 0.001$), weight gain (AOR: 2.09; $95\%$ CI: 1.59–2.75, $p \leq 0.001$), physical inactivity (AOR: 1.51; $95\%$ CI: 1.28–1.78, $p \leq 0.001$) and experiencing cyberbullying (AOR: 2.67; $95\%$ CI: 2.16–3.35, $p \leq 0.001$). **Table 2** | Variables | Depression | Depression.1 | Depression.2 | Depression.3 | Anxiety | Anxiety.1 | Anxiety.2 | Anxiety.3 | Depression and anxiety | Depression and anxiety.1 | Depression and anxiety.2 | Depression and anxiety.3 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | Variables | Unadjusted | Unadjusted | Adjusted | Adjusted | Unadjusted | Unadjusted | Adjusted | Adjusted | Unadjusted | Unadjusted | Adjusted | Adjusted | | Variables | OR (95% CI) | p-value | OR (95% CI) | p-value | OR (95% CI) | p-value | OR (95% CI) | p-value | OR (95% CI) | p-value | OR (95% CI) | p-value | | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | Socio-demographic information | | Age | 1.32 (1.27–1.37) | <0.001 | 1.12 (1.06–1.19) | <0.001 | 1.30 (1.25–1.37) | <0.001 | 1.10 (1.03–1.18) | 0.007 | 1.32 (1.25–1.38) | <0.001 | 1.10 (1.02–1.18) | 0.011 | | Sex | Sex | Sex | Sex | Sex | Sex | Sex | Sex | Sex | Sex | Sex | Sex | Sex | | Male | Ref. | | Ref. | | Ref. | | — | — | Ref. | | Ref. | | | Female | 1.17 (1.02–1.34) | 0.024 | 1.26 (1.05–1.50) | 0.011 | 1.16 (0.99–1.36) | 0.068 | — | — | 1.20 (1.02–1.42) | 0.031 | 1.26 (1.02–1.55) | 0.029 | | Division | Division | Division | Division | Division | Division | Division | Division | Division | Division | Division | Division | Division | | Dhaka | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | | Barisal | 0.80 (0.64–1.01) | 0.059 | 1.02 (0.77–1.34) | 0.918 | 0.87 (0.66–1.14) | 0.300 | 1.04 (0.76–1.43) | 0.794 | 0.81 (0.61–1.07) | 0.142 | 1.00 (0.71–1.39) | 0.984 | | Chittagong | 0.87 (0.67–1.10) | 0.235 | 0.73 (0.55–0.97) | 0.031 | 1.06 (0.81–1.39) | 0.680 | 1.01 (0.73–1.37) | 0.976 | 1.00 (0.75–1.32) | 0.976 | 0.90 (0.65–1.25) | 0.514 | | Khulna | 1.20 (0.90–1.58) | 0.211 | 0.84 (0.60–1.18) | 0.326 | 1.30 (0.95–1.78) | 0.097 | 0.93 (0.64–1.34) | 0.679 | 1.34 (0.97–1.85) | 0.072 | 0.97 (0.66–1.41) | 0.852 | | Mymensingh | 0.72 (0.40–1.31) | 0.282 | 0.56 (0.28–1.15) | 0.115 | 0.73 (0.35–1.52) | 0.394 | 0.53 (0.23–1.21) | 0.132 | 0.81 (0.39–1.69) | 0.570 | 0.62 (0.27–1.44) | 0.265 | | Rajshahi | 1.26 (0.98–1.63) | 0.074 | 1.28 (0.95–1.73) | 0.102 | 1.44 (1.09–1.91) | 0.012 | 1.51 (1.09–2.09) | 0.013 | 1.32 (0.99–1.77) | 0.063 | 1.33 (0.94–1.87) | 0.104 | | Rangpur | 1.08 (0.82–1.42) | 0.599 | 0.62 (0.44–0.86) | 0.005 | 1.09 (0.79–1.49) | 0.615 | 0.63 (0.44–0.93) | 0.018 | 1.16 (0.84–1.60) | 0.368 | 0.70 (0.47–1.02) | 0.062 | | Sylhet | 0.46 (0.37–0.56) | <0.001 | 0.66 (0.51–0.84) | 0.001 | 0.51 (0.40–0.66) | <0.001 | 0.77 (0.58–1.03) | 0.082 | 0.48 (0.37–0.62) | <0.001 | 0.68 (0.50–0.93) | 0.014 | | Academic information | Academic information | Academic information | Academic information | Academic information | Academic information | Academic information | Academic information | Academic information | Academic information | Academic information | Academic information | Academic information | | Academic class | Academic class | Academic class | Academic class | Academic class | Academic class | Academic class | Academic class | Academic class | Academic class | Academic class | Academic class | Academic class | | Secondary school | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | | Higher secondary school | 2.98 (2.40–3.50) | <0.001 | 1.50 (1.14–1.98) | 0.004 | 2.66 (2.18–3.24) | <0.001 | 1.39 (1.03–1.86) | 0.029 | 2.77 (2.21–3.32) | <0.001 | 1.32 (0.97–1.79) | 0.076 | | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | Internet connection at home | | Bad | 2.66 (2.21–3.21) | <0.001 | 1.24 (0.99–1.56) | 0.061 | 3.10 (2.47–3.87) | <0.001 | 1.46 (1.12–1.89) | 0.005 | 3.08 (2.44–3.90) | <0.001 | 1.38 (1.04–1.82) | 0.024 | | Moderate | 1.51 (1.27–1.78) | <0.001 | 1.03 (0.84–1.26) | 0.768 | 1.71 (1.38–2.12) | <0.001 | 1.18 (0.93–1.51) | 0.175 | 1.75 (1.40–2.19) | <0.001 | 1.19 (0.92–1.53) | 0.197 | | Good | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | | Attending classes online | Attending classes online | Attending classes online | Attending classes online | Attending classes online | Attending classes online | Attending classes online | Attending classes online | Attending classes online | Attending classes online | Attending classes online | Attending classes online | Attending classes online | | Yes | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | | No | 2.14 (1.83–2.51) | <0.001 | 1.17 (0.96–1.43) | 0.130 | 2.28 (1.91–2.71) | <0.001 | 1.21 (0.98–1.51) | 0.082 | 2.38 (1.99–2.85) | <0.001 | 1.24 (0.99–1.55) | 0.067 | | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | Teachers’ cooperation while asking questions in class | | Never | 3.28 (2.61–4.13) | <0.001 | 1.74 (1.32–2.29) | <0.001 | 3.67 (2.91–4.64) | <0.001 | 1.91 (1.45–2.52) | <0.001 | 3.87 (3.05–4.91) | <0.001 | 2.01 (1.51–2.66) | <0.001 | | Sometimes | 1.88 (1.58–2.24) | <0.001 | 1.32 (1.08–1.63) | 0.008 | 2.02 (1.66–2.46) | <0.001 | 1.46 (1.16–1.84) | 0.001 | 2.09 (1.70–2.56) | <0.001 | 1.46 (1.15–1.85) | 0.002 | | Always | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | Parental comparison of academic performance with other classmates | | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | | Some | 1.66 (1.36–2.03) | <0.001 | 1.35 (1.07–1.70) | 0.010 | 1.56 (1.22–1.99) | <0.001 | 1.25 (0.95–1.64) | 0.110 | 1.66 (1.28–2.15) | <0.001 | 1.32 (0.99–1.77) | 0.063 | | Much | 3.13 (2.66–3.70) | <0.001 | 2.15 (1.77–2.61) | <0.001 | 2.71 (2.22–3.30) | <0.001 | 1.69 (1.34–2.12) | <0.001 | 3.02 (2.44–3.74) | <0.001 | 1.91 (1.50–2.44) | <0.001 | | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | Worries due to academic delay | | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | | Some | 1.17 (0.87–1.57) | 0.312 | 1.23 (0.88–1.72) | 0.225 | 0.89 (0.61–1.29) | 0.534 | 1.02 (0.68–1.54) | 0.917 | 1.07 (0.69–1.64) | 0.766 | 1.26 (0.79–2.02) | 0.339 | | Much | 4.96 (3.81–6.44) | <0.001 | 3.55 (2.62–4.82) | <0.001 | 3.58 (2.61–4.91) | <0.001 | 2.50 (1.75–3.58) | <0.001 | 4.92 (3.39–7.14) | <0.001 | 3.46 (2.28–5.25) | <0.001 | | Lifestyle factors | | | | | | | | | | | | | | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | Coping with the quarantine situation | | Poorly | 4.82 (3.76–6.18) | <0.001 | 2.96 (2.21–3.95) | <0.001 | 3.93 (2.93–5.26) | <0.001 | 2.16 (1.55–3.01) | <0.001 | 4.51 (3.28–6.20) | <0.001 | 2.36 (1.64–3.38) | <0.001 | | Moderately | 1.72 (1.35–2.19) | <0.001 | 1.38 (1.04–1.82) | 0.024 | 1.20 (0.89–1.61) | 0.236 | 0.91 (0.65–1.26) | 0.568 | 1.30 (0.94–1.80) | 0.114 | 0.95 (0.66–1.36) | 0.786 | | Well | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | Feeling loneliness | | No | Ref. | | — | — | Ref. | | — | — | Ref. | | — | — | | Yes | 0.96 (0.83–1.10) | 0.519 | — | — | 0.91 (0.77–1.07) | 0.250 | — | — | 0.91 (0.77–1.08) | 0.276 | — | — | | Worries due to academics | Worries due to academics | Worries due to academics | Worries due to academics | Worries due to academics | Worries due to academics | Worries due to academics | Worries due to academics | Worries due to academics | Worries due to academics | Worries due to academics | Worries due to academics | Worries due to academics | | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | | Some | 1.17 (0.87–1.57) | 0.312 | 1.23 (0.88–1.72) | 0.225 | 0.89 (0.61–1.29) | 0.534 | 1.02 (0.68–1.54) | 0.917 | 1.07 (0.69–1.64) | 0.766 | 1.26 (0.79–2.02) | 0.339 | | Much | 4.96 (3.81–6.44) | <0.001 | 3.55 (2.62–4.82) | <0.001 | 3.58 (2.61–4.91) | <0.001 | 2.50 (1.75–3.58) | <0.001 | 4.92 (3.39–7.14) | <0.001 | 3.46 (2.28–5.25) | <0.001 | | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | Eating habits changed during the locked down period | | Yes | 2.55 (2.14–3.03) | <0.001 | 1.55 (1.26–1.91) | <0.001 | 3.14 (2.53–3.89) | <0.001 | 2.01 (1.57–2.56) | <0.001 | 3.24 (2.58–4.08) | <0.001 | 1.97 (1.52–2.56) | <0.001 | | Somewhat | 1.67 (1.39–2.01) | <0.001 | 1.42 (1.15–1.76) | 0.001 | 1.71 (1.35–2.16) | <0.001 | 1.41 (1.09–1.83) | 0.010 | 1.83 (1.43–2.35) | <0.001 | 1.48 (1.12–1.96) | 0.006 | | No | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | Weight changes during pandemic (self-reported) | | Loss | 2.10 (1.69–2.62) | <0.001 | 1.49 (1.15–1.92) | 0.003 | 2.16 (1.63–2.85) | <0.001 | 1.36 (0.99–1.86) | 0.056 | 2.26 (1.68–3.05) | <0.001 | 1.41 (1.00–1.97) | 0.048 | | Gain | 3.61 (2.85–4.57) | <0.001 | 2.09 (1.59–2.75) | <0.001 | 4.32 (3.26–5.74) | <0.001 | 2.32 (1.68–3.19) | <0.001 | 4.61 (3.42–6.23) | <0.001 | 2.45 (1.74–3.44) | <0.001 | | Unaware | 1.29 (1.04–1.58) | 0.018 | 0.97 (0.76–1.24) | 0.814 | 1.53 (1.17–2.00) | 0.002 | 1.13 (0.84–1.53) | 0.410 | 1.66 (1.24–2.21) | 0.001 | 1.21 (0.88–1.67) | 0.237 | | No change | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | | Yes | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | | No | 2.01 (1.75–2.30) | <0.001 | 1.51 (1.28–1.78) | <0.001 | 2.15 (1.83–2.53) | <0.001 | 1.60 (1.33–1.92) | <0.001 | 2.13 (1.80–2.52) | <0.001 | 1.50 (1.24–1.83) | <0.001 | | Victimization | | | | | | | | | | | | | | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | Cyberbullying | | Yes | 3.60 (2.99–4.34) | <0.001 | 2.67 (2.16–3.35) | <0.001 | 3.33 (2.75–4.03) | <0.001 | 2.24 (1.80–2.80) | <0.001 | 3.57 (2.94–4.34) | <0.001 | 2.52 (2.00–3.17) | <0.001 | | No | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | Ref. | | The values of Cox & Snell R Square, and Nagelkerke R Square for multivariable logistic regression of anxiety were 0.20, and 0.31, respectively. Anxiety was associated with age (AOR: 1.10; $95\%$ CI: 1.03–1.18, $$p \leq 0.007$$) and inversely linked to living in the Sylhet division as compared to the Dhaka division (AOR: 0.66; $95\%$ CI: 0.51–0.84, $$p \leq 0.001$$). Anxiety was associated with higher secondary school status (AOR: 1.39; $95\%$ CI: 1.03–1.86, $$p \leq 0.029$$) and reports of teachers not being cooperative when students asked questions in online classes (AOR: 1.91; $95\%$ CI: 1.45–2.52, $p \leq 0.001$), worries about academic delays (AOR: 2.50; $95\%$ CI: 1.75–3.58, $p \leq 0.001$), parental comparisons of their academic performances with those of other classmates (AOR: 1.69; $95\%$ CI: 1.34–2.12, $p \leq 0.001$), difficulties coping with the quarantine situation (AOR: 2.16; $95\%$ CI: 1.55–3.01, $p \leq 0.001$), changes in eating behaviors during the pandemic (AOR: 2.01; $95\%$ CI: 1.57–2.56, $p \leq 0.001$), weight gain (AOR: 2.32; $95\%$ CI: 1.68–3.19, $p \leq 0.001$), physical inactivity (AOR: 1.60; $95\%$ CI: 1.33–1.92, $p \leq 0.001$), and experiencing cyberbullying (AOR: 2.24; $95\%$ CI: 1.80–2.80, $p \leq 0.001$). The values of Cox & Snell R Square, and Nagelkerke R Square for multivariable logistic regression models of depression and anxiety were 0.21, and 0.33, respectively. Adolescents with both depression and anxiety were more likely to be older, female, and living in the Dhaka division. They were also more likely to report having a poor internet connection at home, lack of teacher cooperation when asking questions in online classes, worries about academic delays, difficulties coping with the quarantine situation, parental comparisons of their academic performances with those of other classmates, changes in eating behaviors, physical inactivity, and cyberbullying. ## Discussion Depression and anxiety are often experienced by adolescents and may compromise their over-all wellbeing and academic functioning [49, 50]. Moreover, the COVID-19 pandemic has further increased the mental health concerns. The present study represents the largest nationwide survey by the Bangladesh government examining psychosocial health correlates of school-going adolescents during the pandemic. Many adolescents experienced moderate to severe depression ($37.3\%$) and anxiety ($21.7\%$). Participants who reported being older, being from higher secondary classes, residing in the Dhaka division or area, perceiving a lack of teacher cooperation in class, having worries about academic delay, having perceived parental comparison of students’ academic performances with those of other classmates, having difficulties coping with quarantine situations, changes in eating behaviors, gaining weight, physical inactivity, or having experienced cyberbullying were more likely to experience depression and anxiety. Being female and loosing weight during the quarantine were also associated with depression, whereas having poor internet connections at home was associated with anxiety. Implications are discussed below. In the present study, and in comparison to pre-COVID findings among Bangladeshi school-going adolescents using a comparable methodology [12], the percentages of students experiencing moderate to severe depression ($37.3\%$ vs. $26.5\%$) and anxiety ($21.7\%$ vs. $18.1\%$) were higher during the pandemic. Other pre-pandemic findings similarly reported slightly lower estimates of depression ($25\%$-$36.6\%$) [4, 13, 51] and other mental health concerns (13.4–$22.9\%$) among children (< 18 years) [52]. A meta-analysis conducted during the pandemic involving twenty-nine studies sampling 80,879 children and adolescents reported a pooled prevalence estimate of depression of $25.2\%$ ($95\%$ CI: $21.2\%$-$29.7\%$) and anxiety of $20.5\%$ ($95\%$ CI: $17.2\%$-$24.4\%$). These estimates were slightly lower than were found in the present study [6]. Other studies conducted among school-going adolescents using the PHQ-9 and GAD-7 also have observed frequent reports of moderate to severe depression and anxiety during the COVID-19 pandemic including studies from China (depression = $17.3\%$, anxiety = $10.4\%$) [53], and the United States (depression = $32\%$, anxiety = $31\%$) [54], consistent with meta-analytic findings from 29 studies (depression = $25.2\%$, anxiety = $20.5\%$) [6]. The high prevalence estimates of depression and anxiety in the current study may reflect experiences during the COVID-19 pandemic, including social isolation, limited peer interactions, online education (e.g., uncertainty, difficulties utilizing online platforms, challenges in complying with online learning standards), reduced contact with buffering supports (e.g., teachers, coaches), and other factors [6, 51, 55]. However, a pre-COVID study conducted in 2012 among 165 adolescents aged 15–19 years selected from two urban schools in Bangladesh reported a higher prevalence of depression ($49\%$ vs. $21.7\%$) compared to the present study [14]. These differences may reflect differences in assessment instruments and sample characteristics [14]. In the current study, nearly one-fifth ($19.5\%$) of participants experienced both depression and anxiety, consistent with the frequent co-occurrence of the two [12, 42, 43]. In the current study, participants’ age was positively associated with depression and anxiety. This is consistent with both the pre-COVID and pandemic research from Bangladesh and elsewhere [3, 12, 13, 51]. Common mental health problems (e.g., depression, anxiety, behavioral problems) are more prevalent in older children associated with puberty- and hormone-related physical and psychological changes [1, 56]. Moreover, social isolation and physical distancing might have impacted older children, who may rely more heavily on peer socialization [6, 57, 58]. However, this present finding contrasts with other reports from Bangladesh that did not find associations between age and depression among school-going adolescents [4, 12, 59]. In the present study, girls were more likely than boys to have higher odds of depression, which is consistent with earlier reports from Bangladesh [3, 4]. A previous study in China among school-going adolescents also reported finding that females had higher prevalence of depression and anxiety compared to males [60]. The finding is also in line with other reports [61–63]. Moreover, cultural practices and gender norms (such as adolescents not being permitted to discuss their pubertal changes with their parents, needing to contribute in household activities, and possibilities of early marriage, violence, and sexual harassment) in South-Asian societies may make adolescent girls more vulnerable to mental health conditions [64]. Participants from the Sylhet division (northeastern Bangladesh) were less likely to have depression and anxiety than those from the Dhaka division in the present study. This finding resonates with previous Bangladeshi reports of participants residing outside the Dhaka division having had lower likelihoods of depression [65]. One possible reason would be a higher percentage of total COVID-19 cases and deaths in the Dhaka division compared to other divisions, with cases and deaths having been frequently broadcast in electronic and print media [66]. Such information may have increased anxieties about oneself or family members becoming infected and about restrictions on usual movement, travel, and other social activities [20]. Following this rationalle, public health awareness messages should be disseminated carefully to reduce unnecessary anxiety. In addition, school-based psychosocial support programs should consider vulnerabilities relating to older age, being female, and geographical location. Higher secondary school students were more likely to have depression than those secondary school students in the present study, which corroborates previous findings [4]. One possible explanation could be that higher secondary school-going adolescents may have more academic pressure related to getting admission offers into universities based on their performance on the national board exam (Higher Secondary School Certificate [HSC]) in the 12th class in Bangladesh [67]. This additional pressure may predispose them to an increased risk of depression and anxiety [51]. Moreover, Higher secondary school students are typically older compared to other age groups which were taken under consideration. They usually experience more physical, mental and social changes and sometimes burdened with familial responsibilities. These additional stressors might increase the risk of developing depression and anxiety among this group [6, 57, 58]. This study also found a higher odds of depression and anxiety among students who reported worries due to academic delays. A previous study conducted among adolescents and young adults during the COVID-19 pandemic reported that increased anxiety and depression symptoms were linked to increased fear of getting COVID-19 and school-related problems [68]. Furthermore, individuals who reported difficulties coping with lockdown during the pandemic were more likely to experience depression and anxiety. The findings corroborate with the previous reports suggesting that pandemic-related issues such as quarantines, physical distancing, isolation, and educational and economic factors may predispose individuals to common mental health problems including depression and anxiety [20, 28–31, 69]. Moreover, all students experienced academic disruptions (e.g., campus closure, exam postponement) related to the sudden shutdown of all Bangladeshi educational institutions due to the COVID-19 pandemic that began on March 17, 2020 [70]. Further research is needed to investigate how best to address and mitigate academic delays and concerns related to the pandemic, mental health concerns and improve mechanisms for coping with quarantines during pandemics. In the present study, participants who reported perceived parental comparison of academic performance with other classmates and not having teachers’ cooperation while asking questions in class were significantly more likely to experience depression and anxiety. Some previous studies also showed that students reporting more parental pressure to study experienced more depression, anxiety, and stress [71, 72]. Teacher support has been linked to better student mental health and higher resilience [73]. These findings suggest that teacher-mediated school-based support programs may help improve adolescent mental health. In the present study, participants who did not engage in physical exercise were more likely to have both depression and anxiety. This finding resonates with pre-COVID *Bangladeshi data* collected from adolescents [4], and another study conducted in China during the COVID-19 pandemic [74]. Physical activity may help adolescents with self-regulation and coping, facilitating mental optimism [74, 75]. Changed eating habits and increased body weight during the pandemic were associated with depression and anxiety, consistent with a prior study conducted among Egyptian youths during the COVID-19, which linked dietary changes to depression and anxiety [76]. Further, a previous study conducted among school-going adolescents in the Indian Kashmir valley during the pandemic reported that being overweight was associated with depression and anxiety [76]. Similarly, a previous Bangladeshi study including a longitudinal study also found that increased body weight or weight gain was linked to depression and anxiety [77, 78]. Thus, programs promoting physical exercise and maintenance of healthy eating habits and limiting weight gain during pandemics should be developed and tested. Of the participants, $15.8\%$ reported experiencing cyberbullying from their classmates. Participants who experienced cyberbullying had approximately twice the odds of depression and anxiety than those who did not. This finding is consistent with several recent reports indicating that adolescents who experienced cyberbullying were more likely to experience depression, anxiety, loneliness, suicidal behavior, and psychosomatic symptoms [79, 80]. A review article found a considerable increase of cyberbullying among children and adolescents during the COVID-19 pandemic, with symptoms of anxiety, depression, and suicide ideation [81]. Thus, cyberbullying should be prevented through various mechanisms including parental and school-based education, and appropriate legislation [82]. ## Strengths and limitations The present study was the first nationwide, Bangladeshi-government-led epidemiological survey of psychosocial health among school-going adolescents during the COVID-19 pandemic. More than 3,500 adolescents from all divisions of Bangladesh, including 63 districts, participated. However, the study has some limitations. First, as the study was cross-sectional in nature, causal relations cannot be established. Future longitudinal studies are warranted. Second, the study used self-reported measures that are vulnerable to multiple biases, including ones related to recall and social desirability. Although over 3,500 school-going adolescents participated, the study may not be regarded as a representative because it utilized an online survey. ## Conclusions and recommendations The present study revealed high prevalence estimates of depression and anxiety among school-going adolescents during the COVID-19 pandemic in Bangladesh. It also highlighted multiple significant relationships between depression and anxiety and socio-demographic, lifestyle, academic and pandemic measures. The findings raise the possibility that initiating psychosocial support programs among school-going adolescents during pandemics may help protect them from common mental health conditions. Adolescents’ perceptions regarding parental and teacher support appear important when considering adolescent depression and anxiety; thus, positive parent-child and teacher-student interactions from adolescent perspectives are important to understand better. Virtual awareness programs could be considered in order to promote physical exercise, healthy eating habits and avoid weight gain. Addressing cyberbullying also is important, and this may best be achieved through collaboration among multiple stakeholders (e.g., parents, teachers, students, governmental groups, healthcare providers). Last but not least, the findings may contribute as baseline information for future research, including longitudinal or interventional studies. ## References 1. 1Organization WH. Adolescent mental health. 2020. 2. Angold A, Costello EJ, Worthman CM. **Puberty and depression: the roles of age, pubertal status and pubertal timing**. *Psychol Med* (1998.0) **28** 51-61. DOI: 10.1017/s003329179700593x 3. Mridha MK, Hossain MM, Khan MSA, Hanif AAM, Hasan M, Mitra D. **Prevalence and associated factors of depression among adolescent boys and girls in Bangladesh: findings from a nationwide survey**. *BMJ Open* (2021.0) **11** e038954-e. DOI: 10.1136/bmjopen-2020-038954 4. Anjum A, Hossain S, Sikder T, Uddin ME, Rahim DA. **Investigating the prevalence of and factors associated with depressive symptoms among urban and semi-urban school adolescents in Bangladesh: a pilot study**. *International Health* (2019.0). DOI: 10.1093/inthealth/ihz092 5. Merikangas KR, He JP, Burstein M, Swanson SA, Avenevoli S, Cui L. **Lifetime prevalence of mental disorders in U.S. adolescents: results from the National Comorbidity Survey Replication—Adolescent Supplement (NCS-A)**. *J Am Acad Child Adolesc Psychiatry* (2010.0) **49** 980-9. DOI: 10.1016/j.jaac.2010.05.017 6. Racine N, McArthur BA, Cooke JE, Eirich R, Zhu J, Madigan S. **Global Prevalence of Depressive and Anxiety Symptoms in Children and Adolescents During COVID-19**. *A Meta-analysis. JAMA Pediatr* (2021.0) 7. Polanczyk GV, Salum GA, Sugaya LS, Caye A, Rohde LA. **Annual research review: A meta-analysis of the worldwide prevalence of mental disorders in children and adolescents**. *Journal of Child Psychology and Psychiatry and Allied Disciplines* (2015.0) **56** 345-65. DOI: 10.1111/jcpp.12381 8. AlAzzam M, Abuhammad S, Tawalbeh L, Dalky H. **Prevalence and Correlates of Depression, Anxiety, and Suicidality Among High School Students: A National Study**. *J Psychosoc Nurs Ment Health Serv* (2021.0) **59** 43-51. DOI: 10.3928/02793695-20210426-02 9. 9Association AP. Diagnostic and Statistical Manual of Mental Disorders (DSM–5)2013. 10. 10Jones PB. Adult mental health disorders and their age at onset. (0960–5371 (Print)). 11. 11World Health O. National Mental Health Survey, Bangladesh 2018–19. 2019. 12. Islam MS, Rahman ME, Moonajilin MS, van Os J. **Prevalence of depression, anxiety and associated factors among school going adolescents in Bangladesh: Findings from a cross-sectional study**. *PLOS ONE* (2021.0) **16** e0247898-e. DOI: 10.1371/journal.pone.0247898 13. Khan A, Ahmed R, Burton NW. **Prevalence and correlates of depressive symptoms in secondary school children in Dhaka city, Bangladesh**. *Ethnicity and Health* (2017.0) **25** 34-46. DOI: 10.1080/13557858.2017.1398313 14. Billah SM, Khan F. **Depression among Urban Adolescent Students of Some Selected Schools**. *Faridpur Medical College Journal* (2015.0) **9** 73-5 15. Rabbani MG, Hossain MM. **Behaviour disorders in urban primary school children in Dhaka, Bangladesh**. *Public Health* (1999.0) **113** 233-6. DOI: 10.1016/s0033-3506(99)00165-1 16. Khan MMA, Rahman MM, Islam MR, Karim M, Hasan M, Jesmin SS. **Suicidal behavior among school-going adolescents in Bangladesh: Findings of the global school-based student health survey**. *Social Psychiatry and Psychiatric Epidemiology* (2020.0) 17. Cucinotta D, Vanelli M. **WHO declares COVID-19 a pandemic**. *Acta bio-medica: Atenei Parmensis* (2020.0) **91** 157-60. DOI: 10.23750/abm.v91i1.9397 18. 18Dhaka T. Coronavirus: Bangladesh declares public holiday from March 26 to April 4. 2020. 19. Banerjee D.. **The COVID-19 outbreak: Crucial role the psychiatrists can play**. *Asian Journal of Psychiatry* (2020.0) **50** 102014. DOI: 10.1016/j.ajp.2020.102014 20. Brooks SK, Webster RK, Smith LE, Woodland L, Wessely S, Greenberg N. **The psychological impact of quarantine and how to reduce it: rapid review of the evidence**. *The Lancet* (2020.0) **395** 912-20. DOI: 10.1016/S0140-6736(20)30460-8 21. 21The World B. Keeping Bangladesh’s Students Learning During the COVID-19 Pandemic. 2021. 22. Wang G, Zhang Y, Zhao J, Zhang J, Jiang F. *Mitigate the effects of home confinement on children during the COVID-19 outbreak* (2020.0) 945-7 23. Rizvi F, Qureshi A, Rajput AM, Afzal M. **Prevalence of Depression, Anxiety and Stress (by DASS Scoring System) among Medical Students in Islamabad, Pakistan**. *British Journal of Medicine & Medical Research* (2015.0) **8** 69-75 24. Guessoum SB, Lachal J, Radjack R, Carretier E, Minassian S, Benoit L. **Adolescent psychiatric disorders during the COVID-19 pandemic and lockdown**. *Psychiatry research* (2020.0) **291** 113264. DOI: 10.1016/j.psychres.2020.113264 25. Meherali S, Punjani N, Louie-Poon S, Abdul Rahim K, Das JK, Salam RA. **Mental Health of Children and Adolescents Amidst COVID-19 and**. *Past Pandemics: A Rapid Systematic Review. International journal of environmental research and public health* (2021.0) **18** 3432. PMID: 33810225 26. Hossain MM, Hsan K, Islam MS, Nath SK. **Psychological states of Bangladeshi people and associated factors during the outbreak of COVID-19: A cross-sectional survey**. *Emerging Trends in Drugs, Addictions, and Health* (2021.0) **1** 100012. DOI: 10.1016/j.etdah.2021.100012 27. Islam MS, Ferdous MZ, Islam US, Mosaddek ASM, Potenza MN, Pardhan S. **Treatment, persistent symptoms, and depression in people infected with COVID-19 in Bangladesh**. *International Journal of Environmental Research and Public Health* (2021.0) **18** 1453. DOI: 10.3390/ijerph18041453 28. Islam MS, Rahman ME, Banik R, Emran MGI, Saiara N, Hossain S. **Financial and Mental Health Concerns of Impoverished Urban-Dwelling Bangladeshi People During COVID-19**. *Frontiers in Psychology* (2021.0) **12** 3326. DOI: 10.3389/fpsyg.2021.663687 29. Islam MS, Sujan MSH, Tasnim R, Mohona RA, Ferdous MZ, Kamruzzaman S. **Problematic smartphone and social media use among Bangladeshi college and university students amid COVID-19: The role of psychological wellbeing and pandemic related factors**. *Frontiers in psychiatry* (2021.0) **12** 647386. DOI: 10.3389/fpsyt.2021.647386 30. Islam MS, Sujan MSH, Tasnim R, Sikder MT, Potenza MN, van Os J. **Psychological responses during the COVID-19 outbreak among university students in Bangladesh**. *PLOS ONE* (2020.0) **15** e0245083-e. DOI: 10.1371/journal.pone.0245083 31. Islam MS, Tasnim R, Sujan MSH, Ferdous MZ, Sikder MT, Masud JHB. **Depressive symptoms associated with COVID-19 preventive practice measures, daily activities in home quarantine and suicidal behaviors: Findings from a large-scale online survey in Bangladesh**. *BMC psychiatry* (2021.0) **21** 273. DOI: 10.1186/s12888-021-03246-7 32. Tasnim R, Sujan MSH, Islam MS, Ritu AH, Siddique MAB, Toma TY. **Prevalence and correlates of anxiety and depression in frontline healthcare workers treating people with COVID-19 in Bangladesh**. *BMC psychiatry* (2021.0) **21** 271. DOI: 10.1186/s12888-021-03243-w 33. Islam MS, Ferdous MZ, Potenza MN. **Panic and generalized anxiety during the COVID-19 pandemic among Bangladeshi people: an online pilot survey early in the outbreak**. *Journal of Affective Disorders* (2020.0) **276** 30-7. DOI: 10.1016/j.jad.2020.06.049 34. Banik R, Islam MS, Ahmed M, Koly KN, Mubarak M, Rahman M. **General psychiatric symptoms among Bangladeshi people approximately one year after the onset of the COVID-19 pandemic**. *BMC Psychiatry* (2022.0) **22** 615. DOI: 10.1186/s12888-022-04232-3 35. Koly KN, Khanam MI, Islam MS, Mahmood SS, Hanifi SMA, Reidpath DD. **Anxiety and Insomnia Among Urban Slum Dwellers in Bangladesh: The Role of COVID-19 and Its Associated Factors**. *Frontiers in Psychiatry* (2021.0) **12**. DOI: 10.3389/fpsyt.2021.769048 36. Hasan MT, Hossain S, Safa F, Anjum A, Khan AH, Koly KN. **Anxiety and depressive symptoms among physicians during the COVID-19 pandemic in Bangladesh: a cross-sectional study**. *Global Mental Health* (2022.0) **9** 285-97. DOI: 10.1017/gmh.2022.30 37. Sultana MS, Islam MS, Sayeed A, Koly KN, Baker K, Hossain R. **Food addiction, orthorexia nervosa and dietary diversity among Bangladeshi university students: a large online survey during the COVID-19 pandemic**. *J Eat Disord* (2022.0) **10** 163. DOI: 10.1186/s40337-022-00680-0 38. Franklin CGS, Kim JS, Ryan TN, Kelly MS, Montgomery KL. **Teacher involvement in school mental health interventions: A systematic review**. *Children and Youth Services Review* (2012.0) **34** 973-82 39. Franklin C, Kim JS, Beretvas TS, Zhang A, Guz S, Park S. **The effectiveness of psychosocial interventions delivered by teachers in schools: A systematic review and meta-analysis**. *Clinical child and family psychology review* (2017.0) **20** 333-50. DOI: 10.1007/s10567-017-0235-4 40. 40a2i. Aspire to Innovate. 2021. 41. 41Konnect. Konnect. 2021. 42. Islam MS, Rahman ME, Moonajilin MS, Griffiths MD. **Validation and evaluation of the psychometric properties of Bangla nine-item Internet Disorder Scale–Short Form**. *Journal of Addictive Diseases* (2020.0) **38** 540-9. DOI: 10.1080/10550887.2020.1799134 43. Moonajilin MS, Rahman ME, Islam MS. **Relationship between overweight/obesity and mental health disorders among Bangladeshi adolescents: a cross-sectional survey**. *Obesity Medicine* (2020.0) **18** 100216 44. Chowdhury A, Ghosh S, Sanyal D. **Bengali adaptation of Brief Patient Health Questionnaire for screening depression at primary care**. *Journal of the Indian Medical Association* (2004.0) **102** 544-7. PMID: 15887819 45. Kroenke K, Spitzer RL, Williams JBW, Löwe B. **The patient health questionnaire somatic, anxiety, and depressive symptom scales: a systematic review**. *General hospital psychiatry* (2010.0) **32** 345-59. DOI: 10.1016/j.genhosppsych.2010.03.006 46. Spitzer RL, Kroenke K, Williams JB. **Validation and utility of a self-report version of PRIME-MD: the PHQ primary care study. Primary Care Evaluation of Mental Disorders. Patient Health Questionnaire**. *JAMA* (1999.0) **282** 1737-44. DOI: 10.1001/jama.282.18.1737 47. Hossain S, Anjum A, Uddin ME, Rahman MA, Hossain MF. **Impacts of socio-cultural environment and lifestyle factors on the psychological health of university students in Bangladesh: A longitudinal study**. *Journal of Affective Disorders* (2019.0) **256** 393-403. DOI: 10.1016/j.jad.2019.06.001 48. Spitzer RL, Kroenke K, Williams JBW, Löwe B. **A brief measure for assessing generalized anxiety disorder: The GAD-7**. *Archives of Internal Medicine* (2006.0) **166** 1092-7. DOI: 10.1001/archinte.166.10.1092 49. Alfoukha MM, Hamdan-Mansour AM, Banihani MA. **Social and Psychological Factors Related to Risk of Eating Disorders Among High School Girls**. *J Sch Nurs* (2019.0) **35** 169-77. DOI: 10.1177/1059840517737140 50. AlAzzam M, Abuhammad S, Abdalrahim A, Hamdan-Mansour AM. **Predictors of Depression and Anxiety Among Senior High School Students During COVID-19 Pandemic: The Context of Home Quarantine and Online Education**. *The Journal of School Nursing* (2021.0) **37** 241-8. DOI: 10.1177/1059840520988548 51. Anjum A, Hossain S, Hasan MT, Alin SI, Uddin ME, Sikder MT. **Depressive Symptom and Associated Factors Among School Adolescents of Urban, Semi-Urban and Rural Areas in Bangladesh: A Scenario Prior to COVID-19**. *Frontiers in Psychiatry* (2021.0) 12. DOI: 10.3389/fpsyt.2021.708909 52. Hossain MD, Ahmed HU, Chowdhury WA, Niessen LW, Alam DS. **Mental disorders in Bangladesh: a systematic review**. *BMC psychiatry* (2014.0) **14** 216. DOI: 10.1186/s12888-014-0216-9 53. Zhou SJ, Zhang LG, Wang LL, Guo ZC, Wang JQ, Chen JC. **Prevalence and socio-demographic correlates of psychological health problems in Chinese adolescents during the outbreak of COVID-19**. *Eur Child Adolesc Psychiatry* (2020.0) **29** 749-58. DOI: 10.1007/s00787-020-01541-4 54. Murata S, Rezeppa T, Thoma B, Marengo L, Krancevich K, Chiyka E. **The psychiatric sequelae of the COVID-19 pandemic in adolescents, adults, and health care workers**. *Depress Anxiety* (2021.0) **38** 233-46. DOI: 10.1002/da.23120 55. Cao W, Fang Z, Hou G, Han M, Xu X, Dong J. **The psychological impact of the COVID-19 epidemic on college students in China**. *Psychiatry Research* (2020.0) **287** 112934. DOI: 10.1016/j.psychres.2020.112934 56. Oldehinkel AJ, Verhulst FC, Ormel J. **Mental health problems during puberty: Tanner stage-related differences in specific symptoms. The TRAILS study**. *J Adolesc* (2011.0) **34** 73-85. DOI: 10.1016/j.adolescence.2010.01.010 57. Loades ME, Chatburn E, Higson-Sweeney N, Reynolds S, Shafran R, Brigden A. **Rapid Systematic Review: The Impact of Social Isolation and Loneliness on the Mental Health of Children and Adolescents in the Context of COVID-19**. *J Am Acad Child Adolesc Psychiatry* (2020.0) **59** 1218-39. DOI: 10.1016/j.jaac.2020.05.009 58. Hall-Lande JA, Eisenberg ME, Christenson SL, Neumark-Sztainer D. **Social isolation, psychological health, and protective factors in adolescence**. *Adolescence* (2007.0) **42** 265-86. PMID: 17849936 59. Anjum A, Hossain S, Hasan MT, Alin S, Uddin ME, Sikder DM. **Depressive Symptoms and Associated Factors Among School Adolescents of Urban, Semi-Urban and Rural Areas in Bangladesh**. *SSRN Electronic Journal* (2021.0) 60. Zhang X, Yang H, Zhang J, Yang M, Yuan N, Liu J. **Prevalence of and risk factors for depressive and anxiety symptoms in a large sample of Chinese adolescents in the post-COVID-19 era**. *Child Adolesc Psychiatry Ment Health* (2021.0) **15** 80. DOI: 10.1186/s13034-021-00429-8 61. Jeelani A, Dkhar SA, Quansar R, Khan SMS. **Prevalence of depression and anxiety among school-going adolescents in Indian Kashmir valley during COVID-19 pandemic**. *Middle East Current Psychiatry* (2022.0) **29** 18 62. Mohammadi MR, Ahmadi N, Yazdi FR, Khaleghi A, Mostafavi SA, Hooshyari Z. **Prevalence, comorbidity and predictors of anxiety disorders among children and adolescents**. *Asian J Psychiatr* (2020.0) **53** 102059. DOI: 10.1016/j.ajp.2020.102059 63. Nolen-Hoeksema S, Girgus JS. **The emergence of gender differences in depression during adolescence**. *Psychol Bull* (1994.0) **115** 424-43. DOI: 10.1037/0033-2909.115.3.424 64. Afzal S, Rana TF, Mehmood S. **Determinants of depression in female adolescents and youth**. *The Professional Medical Journal* (2008.0) **15** 137-42 65. Sultana MS, Khan AH, Hossain S, Ansar A, Sikder MT, Hasan MT. **Prevalence and predictors of post-traumatic stress symptoms and depressive symptoms among Bangladeshi students during COVID-19 mandatory home confinement: A nationwide cross-sectional survey**. *Children and Youth Services Review* (2021.0) **122** 105880. DOI: 10.1016/j.childyouth.2020.105880 66. 66World Health O. WHO Bangladesh COVID-19 Morbidity and Mortality Weekly Update (MMWU). 2021. 67. Bossy S.. **Academic pressure and impact on Japanese students**. *McGill Journal of Education* (2000.0) **35** 71 68. Hawes MT, Szenczy AK, Klein DN, Hajcak G, Nelson BD. **Increases in depression and anxiety symptoms in adolescents and young adults during the COVID-19 pandemic**. *Psychological Medicine* (2022.0) **52** 3222-30. DOI: 10.1017/S0033291720005358 69. Islam MS, Rahman ME, Banik R, Emran MGI, Noshin S, Hossain S. **Correlates of financial concerns, depression symptoms, and posttraumatic stress disorder symptoms among impoverished urban dwelling individuals in Bangladesh during the COVID-19 pandemic: A face-to-face interview approach**. *PsyArXiv* (2020.0) 70. 70Tribune D. Bangladesh decides to keep all educational institutions closed till Dec 19 2020 [Available from: https://www.dhakatribune.com/bangladesh/2020/11/12/all-educational-institutions-to-remain-closed-till-december-19. 71. Quach AS, Epstein NB, Riley PJ, Falconier MK, Fang X. **Effects of Parental Warmth and Academic Pressure on Anxiety and Depression Symptoms in Chinese Adolescents**. *Journal of Child and Family Studies* (2015.0) **24** 106-16 72. Bhasin SK, Sharma R, Saini NK. **Depression, anxiety and stress among adolescent students belonging to affluent families: a school-based study**. *Indian J Pediatr* (2010.0) **77** 161-5. DOI: 10.1007/s12098-009-0260-5 73. Guo J, Liu L, Zhao B, Wang D. **Teacher Support and Mental Well-Being in Chinese Adolescents: The Mediating Role of Negative Emotions and Resilience**. *Frontiers in Psychology* (2020.0) **10**. DOI: 10.3389/fpsyg.2019.03081 74. Chi X, Liang K, Chen S-T, Huang Q, Huang L, Yu Q. **Mental health problems among Chinese adolescents during the COVID-19: The importance of nutrition and physical activity**. *Int J Clin Health Psychol* (2021.0) **21** 100218. DOI: 10.1016/j.ijchp.2020.100218 75. Lubans D, Richards J, Hillman C, Faulkner G, Beauchamp M, Nilsson M. **Physical Activity for Cognitive and Mental Health in Youth: A Systematic Review of Mechanisms**. *Pediatrics* (2016.0) **138**. DOI: 10.1542/peds.2016-1642 76. Alamrawy RG, Fadl N, Khaled A. **Psychiatric morbidity and dietary habits during COVID-19 pandemic: a cross-sectional study among Egyptian Youth (14–24 years)**. *Middle East Current Psychiatry, Ain Shams University* (2021.0) 28 77. Rofey DL, Kolko RP, Iosif A-M, Silk JS, Bost JE, Feng W. **A longitudinal study of childhood depression and anxiety in relation to weight gain**. *Child psychiatry and human development* (2009.0) **40** 517-26. DOI: 10.1007/s10578-009-0141-1 78. Hossain S, Anjum A, Hasan MT, Uddin ME, Hossain MS, Sikder MT. **Self-perception of physical health conditions and its association with depression and anxiety among Bangladeshi university students**. *Journal of Affective Disorders* (2020.0) **263** 282-8. DOI: 10.1016/j.jad.2019.11.153 79. Kumar VL, Goldstein MA. **Cyberbullying and Adolescents**. *Curr Pediatr Rep* (2020.0) **8** 86-92. DOI: 10.1007/s40124-020-00217-6 80. Nixon CL. **Current perspectives: the impact of cyberbullying on adolescent health**. *Adolesc Health Med Ther* (2014.0) **5** 143-58. DOI: 10.2147/AHMT.S36456 81. Pereira IFdM Tissiani AA, Mendonça GJGd Deininger LdSC, Ramos ALBM Oliveira MCCd. **The impact of cyberbullying on the mental health of children and adolescents during the COVID-19 pandemic**. *Research, Society and Development* (2022.0) **11** e226111032446 82. Fineberg N, Dell’Osso B, Demetrovics Z, Chamberlain SR, Corazza O, Zohar J. *Learning to deal with Problematic Usage of the Internet* (2020.0)
--- title: 'Digital behaviour change interventions to increase vegetable intake in adults: a systematic review' authors: - Katherine M. Livingstone - Jonathan C. Rawstorn - Stephanie R. Partridge - Stephanie L. Godrich - Sarah A. McNaughton - Gilly A. Hendrie - Lauren C. Blekkenhorst - Ralph Maddison - Yuxin Zhang - Scott Barnett - John C. Mathers - Maria Packard - Laura Alston journal: The International Journal of Behavioral Nutrition and Physical Activity year: 2023 pmcid: PMC10042405 doi: 10.1186/s12966-023-01439-9 license: CC BY 4.0 --- # Digital behaviour change interventions to increase vegetable intake in adults: a systematic review ## Abstract ### Background Digital interventions may help address low vegetable intake in adults, however there is limited understanding of the features that make them effective. We systematically reviewed digital interventions to increase vegetable intake to 1) describe the effectiveness of the interventions; 2) examine links between effectiveness and use of co-design, personalisation, behavioural theories, and/or a policy framework; and 3) identify other features that contribute to effectiveness. ### Methods A systematic search strategy was used to identify eligible studies from MEDLINE, Embase, PsycINFO, Scopus, CINAHL, Cochrane Library, INFORMIT, IEEE Xplore and Clinical Trial Registries, published between January 2000 and August 2022. Digital interventions to increase vegetable intake were included, with effective interventions identified based on statistically significant improvement in vegetable intake. To identify policy-action gaps, studies were mapped across the three domains of the NOURISHING framework (i.e., behaviour change communication, food environment, and food system). Risk of bias was assessed using Cochrane tools for randomized, cluster randomized and non-randomized trials. ### Results Of the 1,347 records identified, 30 studies were included. Risk of bias was high or serious in most studies ($$n = 25$$/30; $83\%$). Approximately one quarter of the included interventions ($$n = 8$$) were effective at improving vegetable intake. While the features of effective and ineffective interventions were similar, embedding of behaviour change theories ($89\%$ vs $61\%$) and inclusion of stakeholders in the design of the intervention ($50\%$ vs $38\%$) were more common among effective interventions. Only one (ineffective) intervention used true co-design. Although fewer effective interventions included personalisation ($67\%$ vs $81\%$), the degree of personalisation varied considerably between studies. All interventions mapped across the NOURISHING framework behaviour change communication domain, with one ineffective intervention also mapping across the food environment domain. ### Conclusion Few digital interventions identified in this review were effective for increasing vegetable intake. Embedding behaviour change theories and involving stakeholders in intervention design may increase the likelihood of success. The under-utilisation of comprehensive co-design methods presents an opportunity to ensure that personalisation approaches better meet the needs of target populations. Moreover, future digital interventions should address both behaviour change and food environment influences on vegetable intake. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12966-023-01439-9. ## Background Low vegetable and legume consumption is a leading modifiable risk factor for non-communicable diseases globally [1, 2], accounting for over $2\%$ of global deaths in 2017 [1]. International guidelines for vegetable intake recommend at least 3 serves/day (≥ 240 g/day) [3]. However, nationally representative survey data from 162 countries found that, in 2020, an average of $88\%$ of the populations of these countries had an inadequate vegetable intake [4]. Interventions designed to address low vegetable intake often target low fruit intake simultaneously [5]; however, this is more likely to increase fruit intake than vegetable intake [6]. This is largely attributable to interventions not addressing barriers to vegetable intake, which are distinct from those of fruit intake, including lower palatability, lack of cooking confidence, and perceived higher cost and time to purchase, prepare and cook vegetable-rich meals [6–11]. Interventions that specifically focus on vegetables show promise, but are often setting-specific and delivered face-to-face, such as a workplace interventions [12]. While setting-specificity may be an important component of some personalisation approaches, more scalable approaches are needed to ensure interventions can serve large populations across a wide range of settings [13–15]. As an estimated $66\%$ of people globally have access to the internet [16], digital interventions provide an accessible delivery model for increasing vegetable intake in adults [10, 11]. Furthermore, digital interventions are well aligned with the global drive to utilise digital technologies to improve health [17]. For example, $55\%$ of European citizens aged 16–74 reported that they had sought online health information [18], and $88\%$ of Australians reported wanting to access their health information digitally [19]. However, while there is some evidence that digital interventions increase fruit and vegetable intake [20], the effectiveness of digital interventions to increase vegetable intake alone is unclear. Digital interventions offer the ability to personalise content and delivery to the needs and preferences of the user. Although evidence from randomised controlled trials (RCTs) suggest that personalised dietary advice motivates greater improvement in dietary intake than generalised dietary advice [21], personalisation of digital interventions alone may not be sufficient to increase vegetable intake. To help ensure dietary interventions meet the needs of the user, interventions are increasingly being designed with stakeholders, i.e., using co-design practices [22]. Co-design practices involve the lived experiences of the users, and individuals with technical expertise or service providers in the design process [23]. Research suggests that the use of co-design may help improve consumer engagement and satisfaction with a digital intervention by ensuring it meets their needs [23–25]. However, there is limited understanding of whether existing digital interventions to increase vegetable intake have used co-design methods or whether the use of co-design contributes to effectiveness. Mediators of behaviour change, including knowledge of, attitudes towards, and skills in using vegetables, can be targeted in digital interventions to meet the needs of the user [26, 27]. However, achieving higher vegetable intake is also dependent on complex interactions between individual- and environmental-level influences, such as self-efficacy or access to affordable and healthy foods, which require specific policy actions [7, 8]. The NOURISHING framework [28], which maps interventions according to their alignment with policy actions related to behaviour change communications, the food environment or the food system, is a useful framework for considering such approaches. By mapping across each of these domains, gaps, and opportunities for policy actions for achieving behaviour change can be identified and targeted by digital interventions. Therefore, we aimed to systematically review digital interventions to increase vegetable intake in adults to: 1) describe the effectiveness of the interventions in terms of increased consumption; 2) examine links between effectiveness and use of co-design, personalisation, behavioural theories, and/or a policy framework; and 3) identify other features that contribute to effectiveness. ## Methods The protocol for this systematic review is registered with the international prospective register of systematic reviews (PROSPERO; CRD42022290926). The design and reporting of this review were guided by the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement (Additional file 1) and the synthesis without meta-analysis (SWiM) in systematic reviews reporting guidelines [29]. ## Eligibility criteria The population, intervention, comparison, outcome (PICO) framework was used to develop the inclusion and exclusion criteria for study selection. Study designs included RCT, pseudo-RCTs, and pre-post interventions. The population included community-dwelling adults (18 years and older). Studies were excluded if they included pregnant and/or lactating women and/or institutionalised adults. Studies on populations for primary and secondary prevention were included. Interventions were included if they were a digital intervention targeting knowledge of, attitudes towards, and skills in using vegetables. In this review, “digital interventions” were interventions that included any of the following digital components: applications (apps; native, web, progressive and hybrid), websites, computer programs, mobile games, Short Message Services (SMS), Social Networking Services (SNS) and wearable devices [10]. Multi-modal interventions with non-digital components (e.g., face-to-face consultations) were included if digital features represented the primary focus of the intervention. The focus of this review was on vegetable intake, so the primary outcome was change in vegetable intake (i.e., measured as serves, portions, or grams/day). Secondary outcomes considered included changes in attitudes, knowledge, skills, self-efficacy, access and/or intentions related to vegetable intake. Studies were excluded if vegetable intake could not be examined separately. Only peer-reviewed original research articles published in English were included. ## Search strategy The search was developed in consultation with a librarian and undertaken in November 2021 and updated in August 2022. Published literature from January 2000 to August 2022 was searched. The year 2000 was selected as this coincided with an increase in the use of digital technologies in nutrition research and is in alignment with similar reviews of digital interventions [30]. The following databases were searched: MEDLINE (Complete), Embase, PsycINFO, Scopus (only extra searching), CINAHL (EbscoHost), Cochrane Library (Wiley), Rural and Remote Health database (INFORMIT), Health and society database (INFORMIT), IEEE Xplore, ClinicalTrials.gov and the Australian New Zealand Clinical Trial Registry. The full search strategy can be found in Additional file 2. Briefly, search terms were combined using the AND/OR operators for digital (‘digital, ‘smartphone’, ‘website’, ‘app’), intervention (‘intervention’, ‘randomized controlled trial’) and outcomes (‘vegetables’). Reference lists from systematic reviews identified in the search and included records were hand-searched to identify any additional studies. Where relevant protocol papers were identified during the search, an attempt was made to find the accompanying trial papers. ## Data extraction Studies were screened using Covidence software by two members of the team (KML, LA), first by title and abstract and then by full text. Discrepancies were resolved by discussion. Duplicates were removed in Covidence. Data were extracted by one reviewer (KML) and checked by a second reviewer (LA). A data extraction template was developed and piloted in Excel specifically for this review. The following information was extracted from each study: study design (setting, intervention and control conditions, duration), intervention features (digital tools used, co-design methods, behaviour change framework and taxonomies used, personalisation, NOURISHING framework policy domains and areas), population (country, age, sex, rurality, primary or secondary prevention); outcome measures (primary or secondary outcome, change in intake, behaviour, attitude, knowledge, skills, self-efficacy, intention and/or access); results for vegetable intake and effectiveness (yes/no determined based on statistically significant results for vegetable intake). ## Data synthesis A descriptive synthesis of the findings from the included studies was conducted. No meta-analysis was undertaken due to the heterogeneous nature of the digital tools used, characteristics of the populations in the included studies and the indicator of vegetable intake reported. The effectiveness and features of all interventions were summarised to better understand the characteristics that may increase likeliness of effectiveness. Features investigated included the population and study design, such as age, sex, rurality, use of co-design practices, behaviour change theory and personalisation methods. Studies were also mapped against the World Cancer Research Fund International’s NOURISHING framework [28]. This framework comprises three broad domains of policy actions (food environment, food system and behaviour change communication), 10 key policy areas within these domains, and the specific policy actions, which should be identified and implemented by policymakers to fit their national contexts and populations [28]. Examples of policy areas for these three domains included using economic tools to address food affordability (food environment domain), supply chain actions (food systems domain) and nutrition education and skills (behaviour change communication domain). We mapped whether the three broad domains and underlying 10 key policy areas were employed in the design of the intervention. ## Risk of bias assessment Two authors (KML, SP) performed an independent assessment of the risk of bias on the included studies, with any discrepancies resolved by consensus. Three Cochrane Risk of Bias tools were used: for randomized trials (RoB 2), for cluster RCTs (CRCT; RoB 2 CRCT) and for non-randomized studies of interventions (ROBINS-I) [31, 32]. The RoB 2 and RoB 2 CRCT domains for risk of bias assessment included randomization process, deviations from the intended interventions, missing outcome data, measurement of the outcome and selection of the reported result. The judgement within each domain was assessed to carry forward to an overall risk of bias judgement as low risk, some concerns or high risk of bias. The ROBINS-I domains for risk of bias assessment include confounding, selection of participants, classifications of interventions, deviations from intended interventions, missing data, measurements of outcomes and selection of reported results. The judgement within each domain was used to inform an overall risk of bias judgement as either low-risk, moderate-risk, serious risk, critical risk or no information reported. ## Results The search strategy retrieved 1,347 records (Fig. 1). After the removal of duplicates, 1,049 articles were screened for inclusion based on their title and abstract. Of these, the full texts of 97 articles were screened. This review included 30 studies [33–62] (Table 1).Fig. 1PRISMA flow diagram of study selectionTable 1Characteristics of included studies ($$n = 30$$)Author and datePopulationStudyOutcomeIntervention resultsCountrynMean age, sex, ruralityPrimary vs secondary preventionDesignDurationFollow-upPrimarySecondaryVegetable intakeEffectiveAbu-Saad 2019 [33]Israel5053 y$58\%$ femaleNot ruralSecondary – participants with T2DMPilot two-arm un-blinded RCT6 moBL, 3 mo, 6 mo, 12 moDiabetes-related dietary knowledgeVegetable, fruit, wholegrain, added sugars, dietary fibre intakePA, adiposity, HbA1cNS increase vs CG (4.2 vs 3.4 portions/d)NoAlonso-Dominguez 2019 [34]Spain20461 y$46\%$ femaleNot ruralSecondary – participants with T2DMTwo-arm RCT12 moBL, 3 mo, 12 moMediterranean Diet Adherence Screener (including ≥ 2 serves/d vegetables)Diet Quality Index, clinical measuresNS increase adherence at 12 mo ($11\%$) vs BLNoBhurosy 2020 [35]US16519 y$86\%$ femaleNot ruralPrimary – dietRCT3 dDay 1, day 2, day 3Red/orange vegetable intakeS increase from0.9 ± 0.9 times/d on day 1 to 1.6 ± 1.3 times/d on day 2 and to 1.3 ± 1.3 times/d on day 3. NS increase in the CGYesBozorgi 2021 [36]Iran12052 y$40\%$ femaleNot ruralSecondary – participants with hypertensionTwo-arm RCT6 moBL, 2 mo, 6 moAdherence to medicationDASH (including vegetable intake), blood pressure, PAIncrease in IG ($$n = 19$$ by > 2 serves/d; no statistical comparison)NoBrown 2014 [37]US15022 ySex NANot ruralPrimary – dietPilot two-arm RCT7 wkBL, 7 wkMyPlate food group recognitionVegetable, fruit intakeTrend in increase in intervention group (data not shown)NoCantisano 2022 [56]Spain1621 y$100\%$ femaleNot ruralPrimary – diet and lifestylePre-post trial3 moBL, 3 moDietary intake (Global Diet Quality Index, 8 FG including vegetables)PA, lifestyle and wellbeingS increase of 3.75 score vs BL ($$P \leq 0.005$$)YesCelis-Morales 2016 [57]Ireland, NL, Spain, Greece, UK, Poland, Germany126940 y$59\%$ femaleNot ruralPrimary – diet and PAFour-arm RCT6 moLow intensity: BL, 3, 6 moHigh intensity: BL, 1, 2, 3, 6 moDietary intake (9 FG including vegetables),Healthy Eating IndexAnthropometric measures (Weight, BMI, waist), biomarkersNS increase of 2.0 g/day ($$P \leq 0.81$$) between CG and IGNoChan 2020 [38]US16070 y$100\%$ menNot ruralSecondary – participants with prostate cancerPilot four-arm RCT3 moBL, 3, 6 moFeasibilityDiet score, dietary intake (7 FG including cruciferous vegetables), PAS increase in vs CG (0.29 serves/d)YesDebon 2020 [39]Brazil3959 y$82\%$ femaleNot ruralSecondary – participants with hypertensionPilot non-blinded non- randomized, controlled trial3 moBL, 3 moDietary intake (10 FG including vegetables), self-care, biomarkers, blood pressureNS increase in vegetable intake in the IG vs BL (0.95 serves/wk)NoElbert 2016 [40]NL14641 y$73.3\%$ femaleNot ruralPrimary—dietRCT6 moBL, 6 moFruit and vegetable intake overall and by health literacySelf-efficacy in eating fruit and vegetablesNS increase IG vs BL. S increase in participants with high health literacy vs lowNoFjeldsoe 2019 [41]Australia11454 y$67\%$ femaleNot ruralPrimary—lifestyleRCT12 moBL, 6, 12 moFruit, vegetable, SSB intake, takeaway meals, fat, fibre index, weight, PANS increase in serves/day vs BL (0.10; $95\%$ CI: − 0.32 to 0.53)NoGilson 2017 [42]Australia1948 yNot ruralPrimary—diet and physical activityPilot non-randomised uncontrolled trial5 moBL, 5 mo, 2 mo follow upFruit, vegetable, saturated fat, SSB, PASedentary periodsS increase by 1 serve/d vs BL ($$P \leq 0.024$$)YesGoni 2020 [43]Spain72060 y$24\%$ femaleNot ruralSecondary – participants with atrial fibrillationSingle-blind RCT2 yBL, 1, 2 yMediterranean diet (including vegetables)NS increase vs CG (-20 g/day 2-y change)NoHansel 2017 [44]France12057 y$67\%$ femalesNot ruralSecondary – participants with T2DM and abdominal obesityTwo-arm open-label RCT4 moBL, 3, 6 moInternationalDiet Quality Index (including vegetables)Weight, HbA1c, measured maximum oxygen consumptionS increase 0.3 points vs CG (-0.3; $$P \leq 0.01$$)YesHebden 2014 [45]Australia5123 y$81\%$ femaleNot ruralSecondary – participants with overweight or obesityPilot two-arm RCT3 moBL, 3 moWeight, BMIVegetable, SSB intake, takeaway meals, PANS increasevs CGNoHendrie 2020 [58]Australia122448 y$84\%$ femaleNot ruralPrimary—dietPre-post trial90 dBL, 21, 90 dVegetable intake and varietyPsychological variables (attitudes, intentions, self-efficacy, and action planning) and app usageS increase of 0.48 serves/d and 0.35 types /d vs BLYesJahan 2020 [46]Bangla-desh41247 y$86\%$ femaleRuralSecondary – participants with hypertensionTwo-arm open-label RCT5 moBL, 5, 12 moSalt, fruit, vegetable intake, blood pressure, weight, PADietary salt excretion, glucose, quality of lifeNS increase ($1\%$ more increased vs CG)NoKerr 2016 [52]Australia24724 y$65\%$ femaleNot ruralPrimary—dietThree-arm RCT6 moBL, 6 moFruit, vegetables, SSB, energy-dense nutrient-poor foods and beveragesWeight, BMINS decline vs CG (-0.1 serves/d)NoLara 2016 [47]UK7061 y$75\%$ femaleNot ruralPrimary—lifestylePilot two-arm single-blinded RCT2 moBL, 2 moMediterranean diet (including vegetables), PA, healthy ageingDecline (2.6 portions/d) vs BL (2.4 portions/d)(no statistical comparison)NoLombard 2016 [48]Australia64940 y$100\%$ femaleRuralPrimary—weight managementCluster RCT (by town)12 moBL, 12 moWeight lossDiet quality, greater self-managementbehaviours (including vegetables)NS increase in IG by 3 g/dNoPerez-Junkura 2022 [59]Spain2737 y$81\%$ femaleNot ruralPrimary—dietNon-randomised, uncontrolled trial12 moBL, 12 moDietary intake (including vegetables)Gastrointestinal symptomsNS increase vs BL by 0.7 portions/dNoPlaete 2015 [60]Belgium42632 y$60\%$ femaleNot ruralPrimary—dietThree arm- non-randomised controlled trial1 moBL, 1 wk, 1 moFruit, vegetable intakeS increase vs BL (IG1: χ2 1 = 5.3, $$p \leq 0.02$$; IG2: χ2 1 = 12.8, $p \leq 0.001$). NS increase in CGYesPope 2019 [49]US3822 y$74\%$ femaleNot ruralPrimary—lifestyleTwo-arm, RCT3 moBL, 1.5 mo, 3 moFeasibilityFruit, vegetable, wholegrains, SSB, calories, PA, physiology, weightDecline vs BL (no statistical comparison)NoRecio-Redruguez 2016 [50]Spain83352 y$62\%$ femalesNot ruralPrimary—dietTwo-arm RCT3 moBL, 3 moMediterranean diet (including vegetables), PABlood pressure, BMI, biomarkersNS decline vs CG (-$4\%$ ≥ 2 serves/d)NoSchulz 2014 [51]NL505544 y$47\%$ femaleNot ruralPrimary -lifestyleThree-arm RCT2 yBL, 1, 2 yOverall risk scoreFruit, vegetable intake, alcohol, smoking, PANS increase vs CG (β 0.07, $$P \leq 0.62$$)NoTurner-McGrievy 2013 [53]US9643 y$75\%$ femaleNot ruralSecondary – participants who are overweightPost hoc analysis of RCT6 moBL, 3, 6 moWeightFruit, vegetables intake, PANS increase between app, paper journal or website $$P \leq 0.67$$)NoWang 2021 [54]China11018 y$59\%$ femaleNot ruralPrimary—lifestyleNon-randomized controlled trial21 dBL, 21 dDietary intake (including vegetables)PA, fitness, body compositionS increase vs BL ($0\%$ vs $7\%$ ≥ 500 g/d). NS increase in CGYesWang 2020 [55]Mongolia17151 y$57\%$ malesNot ruralSecondary – participants with T2DMTwo-arm RCT12 moBL, 12 moPlasma glucoseFruit, vegetable intake, PA, smoking, weight controlS increase in % who increased intake vs CG ($87\%$ vs $29\%$; $p \leq 0.001$)YesWilliams 2022 [61]Australia47752 y$78\%$ femaleNot ruralPrimary – diet and lifestyleTwo-arm RCT3 moBL, 1, 3 moHealthcare professional visitationsPA, BMI, fruit, vegetable intakeNS increase in meeting guidelines vs CG (0.90 [0.39, 2.10])NoZenun Frano 2022 [62]UK18743 y$84\%$ femaleNot ruralPrimary—dietTwo-arm, single-blinded RCT3 moBL, 3 mom-AHEI (including vegetable scores)Weight, BMI, PANS decline vs CG (-0.32 m-AHEI points)NoAbbreviations:BL baseline, BMI body mass index, CG control group, d. day, DASH Dietary Approaches to Stop Hypertension, FG food groups, IG intervention group, mo month, m-AHEI modified-alternative healthy eating index, NA not available, NS non-significant, PA physical activity, RCT randomized controlled trial, S significant, SSB sugar-sweetened beverages, T2DM type 2 diabetes mellitus, wk week, y year ## Study characteristics The 30 included studies comprised of RCTs ($$n = 22$$) [33–38, 40, 41, 43–47, 49–53, 55, 57, 61, 62], a CRCT ($$n = 1$$) [48] and non-randomized trials ($$n = 7$$) [39, 42, 54, 56, 58–60]. Intervention duration ranged from 3 days [35] to 2 years [43, 51]; more than half ($$n = 17$$; $57\%$) of studies had a follow-up period less than 6 months. Most studies were conducted in Australia [41, 42, 45, 48, 52, 58, 61], followed by the United States [35, 37, 38, 49, 53], Spain [34, 43, 50, 56, 59], the Netherlands [40, 51], the United Kingdom [47, 62], Belgium [60], France [44], pan-European [57], Israel [33], Iran [36], Brazil [39], Bangladesh [46], China [54] and Mongolia [55]. The studies included sample sizes ranging from 16 [56] to 5,055 [51], with 16 studies ($53\%$) including a sample of 150 or more participants. The mean age of participants ranged from 18 years [54] to 70 years [38], with many ($$n = 19$$) conducted in mid-aged and older adult populations (> 40 years). Two studies delivered the digital interventions exclusively in rural areas [46, 48]. Eleven ($37\%$) interventions recruited populations with health conditions, including hypertension [36, 39, 46], type 2 diabetes mellitus [33, 34, 44, 55], heart disease [43] prostate cancer [38] and overweight or obesity [45, 53]. The remaining studies were conducted in generally healthy populations and were designed to improve diet and/or lifestyle ($$n = 18$$) or weight management ($$n = 1$$). Over half of the studies ($$n = 17$$) were published since 2019. ## Risk of bias Risk of bias within 25 ($83\%$) studies was high or serious because of missing outcome data for RCTs or bias due to confounding in non-RCTs (Additional file 3). Most RCTs ($$n = 17$$) and the CRCT adequately generated and concealed allocation resulting in no imbalances apparent between groups. Participant blinding was not possible because of the nature of digital health interventions and was not considered to increase risk of bias. The measure of assessment of vegetable intake was considered appropriate in most RCTs and the CRCT except for three studies where insufficient information was provided. Assessors were blinded to the intervention received by participants in 11 studies. Assessment of the outcome could have been influenced by knowledge of intervention received. However, this was deemed unlikely due to the dietary assessment methods and protocols used to assess vegetable intake, where it is unlikely that dietary coders were aware of the intervention allocation. Finally, seven studies did not reference a protocol or trial registration with a pre-specified analysis plan that was finalized before unblinded outcome data were available for analysis, which may be due to publication preceding the development of reporting guidelines. ## Characteristics of digital tools The most common digital tools used in the included studies were apps ($$n = 19$$; $63\%$), followed by SMS messaging ($$n = 10$$; $33\%$) and websites ($$n = 9$$; $30\%$). Some studies also used phone coaching and emails, and some interventions included a ‘dashboard’ feature to summarise resources and goals [39, 47]. Just under half ($$n = 13$$; $43\%$) used a combination of digital tools (Table 2).Table 2Summary of the features of digital interventions grouped according to effectivenessAuthor and yearControlIntervention featuresInterventionNOURISHING policy domain and policy areaDigital toolCo-designaBehaviour change theoryPersonalisationEffectiveBhurosy 2020 [35]Self-monitoring of dietSelf-monitoring of diet, including red/orange vegetable intake, set a goal to eat 1 more, take pictures of servesDomain: BCCPolicy: nutrition education and skillsAppAssigned goal settingCantisano 2022 [56]ePSICONUT programme – eHealth tools (Headspace, Insight Timer, Fabulous, YouTube channel, WhatsApp, e-mail, and Excel sheets to perform tasks/activities)Domain: BCCPolicy: nutrition education and skillsApp, SMS messages, videos, email, ExcelNot true co-design: researchers and health professionalsGoal settingChan 2020 [38]Generalised dietary and PA adviceTrueNTH Community of Wellness—educational material, links to resources, self-monitoring diet and PADomain: BCCPolicy: nutrition education and skills; nutrition advice and counsellingWebsite, Fitbit, SMS messages, phone callsSocial cognitive theory, goal settingDiet and PA advice, videos and reports, exercise trainer and dietitian, dashboardGilson 2017 [42]Jawbone UP™—financial incentives program. Education materials and self-monitoring PA and healthy dietary choices, and financial incentives for changing behavioursDomains: BCC; food environmentPolicies: nutrition education and skills; economic tools to address affordability and purchase incentivesApp, activity trackerGoal settingSupport and feedback from researchers on goalsHansel 2017 [44]Generalised dietary adviceANODE—dietetic tool providing menus, shopping list, recipes, PA prescribedDomain: BCCPolicy: nutrition education and skills; nutrition advice and counsellingWebsiteMenus, shopping list based on preferences, tastes, calories, needsHendrie 2020 [58]VegEze – motivation and education to increase intake and variety, self-monitoring, gamification, > 50 recipes and meal suggestionsDomain: BCCPolicy: nutrition education and skillsAppNot true co-design: dietitians, researchers, product developers, software engineers, adults (25–45 y)Behaviour change wheel; motivation, goal setting, self-monitoring, social comparison, gamificationFeedback and motivational messages for meeting goalsPlaete 2015 [60]Generalised informationMyPlan 1.0—motivation and education to improve behaviour (group 1 and 2 were recruited by GPs and researchers respectively)Domain: BCCPolicy: nutrition education and skillsWebsiteNot true co-design: researchers, general practitionersSelf-regulation, health action process, goal settingFeedback on health behaviours, action planWang 2021 [54]Health educationWeChat—health education, self-monitoring, reminders, diet, sport advice and supervisionDomain: BCCPolicy: nutrition education and skillsAppTrans-theoreticalModelDietitians, sports coach advice, health reportsWang 2020 [55]Text messages on general health informationText messages covering health awareness, diet control, PA, living habits, weight controlDomain: BCCPolicy: nutrition education and skills; nutrition advice and counsellingSMS messagesNot true co-design: endocrinology, chronic disease, health education, disease prevention expertsTrans-theoreticalModelIneffectiveAbu-Saad 2019 [33]Standard lifestyle counsellingInteractive Lifestyle Assessment, Counselling, and Education (I-ACE)—self-monitoring of dietary intake and PA, dietitian-delivered lifestyle education and adviceDomain: Behaviour Change Communication (BCC)Policy: nutrition education and skills; nutrition advice and counsellingAppNot true co-design: adults, dietitiansMotivational interviewing, goal settingClinical counselling to improve diet based on diet, ethnicity, culture, age, health statusAlonso-Dominguez 2019 [34]Generalised dietary and PA adviceEVIDENT II—self-monitoring diet and PA, in-person food/cooking workshops, walksDomain: BCCPolicy: nutrition education and skills; nutrition advice and counsellingAppNot true co-design: software engineers, dietitians, PA expertsDiet and PA advice based on diet, PA, age, sex, weight, height, strideBozorgi 2021 [36]Usual careEducation and support information on disease management, healthy diet (DASH and low-salt diet), weight loss and motivational messagesDomain: BCCPolicy: nutrition education and skills; nutrition advice and counsellingAppMessages based on patient characteristicsBrown 2014 [37]Brochure containing same informationMobile MyPlate—behavior-directed motivational textmessages on the US DietaryGuidelines and messages of the My-Plate iconDomain: BCCPolicy: nutrition education and skillsApp, SMS messagingNot true co-design: nutrition and health societies, industry, department of agriculture and health and human servicesGoal settingCelis-Morales 2016 [57]Generalised dietary adviceFood4Me—self-monitoring of diet and PA, three levels of feedback report: Level 1: based on diet data; Level 2: based on diet and phenotype data; Level 2: based on diet, phenotype and genotype dataDomain: BCCPolicy: nutrition education and skills; nutrition advice and counsellingWebsite, internet forum, accelerometerBehaviour change wheel; motivation, self-monitoring, assigned goal settingDiet advice based on diet, phenotype (anthropometric; blood biomarkers) and/or genotype (5 nutrient-responsive genes)Debon 2020 [39]Health education workshopsHealth education workshops. Self-monitoring of physical measurements, (e.g. blood pressure, anthropometrics, sleep, mood, PA). Recommendations based on reference values. Alerts and remindersDomain: BCCPolicy: nutrition education and skills; nutrition advice and counsellingAppGoal settingAdvice based on reference values, dashboard summary of health conditionsElbert 2016 [40]No health informationText- or audio-based tailored health information, recipes, testimonialsDomain: BCCPolicy: nutrition education and skillsAppSocial cognitive theory, goal settingAction plan, testimonial matching, advice based on current diet, barriers to fruit and vegetable intake and healthFjeldsoe 2019 [41]Brief written feedbackGet Healthy, Stay Healthy (GHSH)—extended contact intervention with text messages and phone calls with coachDomain: BCCPolicy: nutrition education and skillsSMS messagesGoal settingDiet and PA goals, frequency of goals and texts, phone coachingGoni 2020 [43]Usual clinical carePREDIMAR—nutrition education on the Mediterranean diet, self-monitoring of diet, recipesDomain: BCCPolicy: nutrition education and skills; nutrition advice and counsellingWebsite, app, printed resources, phone calls, cooking videos, testimonialsNot true co-design: dietitians, nutritionists,epidemiologists, doctors, chefs, programmersDietary advice by a dietitianHebden 2014 [45]Booklet from dietitianBooklet, text messages, emails, app and forums, recipes, self-monitoring diet and PADomain: BCCPolicy: nutrition education and skillsSMS messages, e-mails, app, Internet forumsTranstheoretical modelMotivational advice and instantaneous diet and PA feedback based on guidelines, dietitian accessJahan 2020 [46]Brochure on health educationHealth education on DASH, PA, generalised text messages on recommendationsDomain: BCCPolicy: nutrition education and skills; nutrition advice and counsellingSMS messagesKerr 2016 [52]Self-monitoring without feedbackConnecting Health and Technology (CHAT)—text messages, self-monitoring and feedback on diet using images, web links and recipesDomain: BCCPolicy: nutrition education and skillsSMS messaging, appNot true co-design: young adults (18–30 y)Self-determinationFeedback based on diet, nameLara 2016 [47]Usual careLiving, Eating, Activity and Planning through retirement (LEAP)—information on healthy eating (Mediterranean diet), recipes, PA, social roles, self-monitoringDomain: BCCPolicy: nutrition education and skillsWebsiteTrue co-design: researchers, adults (> 55 y), health social care professionalsHealth action process, goal settingContent based on demographics, diet and goals, dashboard summaryLombard 2016 [48]Generalised health sessionHeLP-her—self-management education manual, group session, phone coaching, text messagesDomain: BCCPolicy: nutrition education and skillsSMS messages, phone coachingSelf-determination, cognitive behavioural, motivational interviewing, goal settingDiet and PA goals and action plan, by name, coachingPerez-Junkura 2022 [59]GlutenFreeDiet platform—dietary evaluation, which allows dietitians to measure energy content and nutrient distributionDomain: BCCPolicy: nutrition education and skills; nutrition advice and counsellingWebsite, app counsellingFeedback based on dietPope 2019 [49]Identical, FacebookFacebook group, self-monitoring diet and PA with smartwatchDomain: BCCPolicy: nutrition education and skillsApp, smartwatchSocial cognitive, self-determinationRecio-Redruguez 2016 [50]Generalised informationDiet and PA counselling, leaflet, self-monitoring diet and PADomain: BCCPolicy: nutrition education and skillsApp, counsellingNot true co-design: dietitians, PA experts, software engineersFeedback and plan based on diet and PASchulz 2014 [51]Health risk appraisalMyHealthyBehavior—health risk appraisal, feedback to improve diet, alcohol, PA, smoking (sequentially or simultaneously)Domain: BCCPolicy: nutrition education and skillsWebsiteFeedback based on diet, PA, smoking, by nameTurner-McGrievy 2013 [53]PodcastPodcast, self-monitoring diet and PA, social supportDomain: BCCPolicy: nutrition education and skillsPodcast, appSocial cognitive theory, goal settingWilliams 2022 [61]Static text-based messages and letter from GPDiabetes Online Risk Assessment (DORA) study—video-based story (80–144 s in duration), links to reputable healthy lifestyle resources (e.g. Nutrition Australia)Domain: BCCPolicy: nutrition education and skills; nutrition advice and counsellingWebsite, video, SMS messagesHealth belief modelVideo based on individual T2DM risk factors, gender, and ageZenun Franco 2022 [62]Generalised dietary advice via the eNutri web appEatWellUK study—personalised dietary advice via the eNutri web appDomain: BCCPolicy: nutrition advice and counsellingAppNot true co-design: adults (18 y and over), nutrition professionalsDietary advice based on current dietAbbreviations:BL baseline, BMI body mass index, CG control group, d. day, DASH Dietary Approaches to Stop Hypertension, FG food groups, IG intervention group, mo month, m-AHEI modified-alternative healthy eating index, NA not available, NS non-significant, PA physical activity, RCT randomized controlled trial, S significant, SSB sugar-sweetened beverages, T2DM type 2 diabetes mellitus, wk week, y yeara Design and development input from stakeholders identified as true co-design only if the study used this the term “co-design” ## Vegetable intake As shown in Table 1, vegetable intake was a primary outcome in $63\%$ of studies ($$n = 19$$). Of these, some studies reported vegetable intake as a component of a Mediterranean diet score ($$n = 4$$), International Diet Quality Index ($$n = 1$$), m-Alternate Healthy Eating Index [62] or an overall diet quality index for Dominican adults [56]. Vegetable intake was assessed in most studies using brief diet questions [35, 36, 41, 42, 45, 46, 54–56, 58, 60], followed by a food frequency questionnaire [33, 37–40, 48, 51, 57, 61, 62], 24 h recall [57, 59], Mediterranean diet adherence screener [34, 43, 50], and an image-based dietary assessment tool [52]. ## Co-design practices As shown in Table 2 and Fig. 2, $40\%$ of studies ($$n = 12$$) reported some level of stakeholder input into the intervention design. Only one study, by Lara et al., referred to co-design specifically; a seven-stage, sequential, iterative series of workshops were used for designing, prototyping, testing and optimising the intervention, which was undertaken with researchers, older adults (the target population) and health and social care professionals [47]. This study was designated as using true co-design. Of the studies that reported stakeholder input, health care professionals, such as dietitians and general practitioners, were the most commonly reported stakeholders involved in the design, followed by software engineers. Only five studies reported involving consumers with lived experiences, including young adults (aged 18–30 years) in the Connecting Health and Technology (CHAT) study [52], adults aged over 55 years in the Living, Eating, Activity and Planning through retirement (LEAP) study [47] and Arab adults in a trial of ethnic minority adults with type 2 diabetes mellitis [33].Fig. 2Summary of features of digital interventions to increase vegetable intake ## Personalisation methods Twenty-three studies ($77\%$) included some level of personalised intervention feedback (Table 2 and Fig. 2). The degrees of personalisation ranged from low (e.g., feedback based on assessment of current diet [52]), to moderate (e.g., personalisation of menus and shopping lists [44]), to high (e.g., individual coaching from a dietitian [38]); only one study reported offering participants the opportunity to customise their personalisation, based on preferred frequency and timing of text messaging [41]. Seven studies provided access to diet or physical activity coaching by a health professional via an app [33, 38, 41, 43, 45, 48, 54], phone calls [38, 41, 48], video calls [43], and SMS messages, emails and online forums [45]. One study personalised content to specifically address barriers to vegetable intake based on participant responses [40], another study used a digital program to design a personalised daily or weekly menu based on user preferences such as taste in foods, season and price range [44], while another study created a personalised video to promote healthy lifestyle behaviours based on age, gender and individual type 2 diabetes risk factors [61]. SMS-based interventions often used the participants’ name within the content [34, 36, 39]. Four studies provided personalised feedback and/or action plans based on demographic characteristics (such as age, sex, ethnicity and culture) and/or participant preferences [34, 36, 39] although limited information was provided on how this personalisation was designed or delivered, or whether personalisation was applied to the dietary component of the intervention. Other studies included some aspects of individualised support, although access to advice and support from dietitians was not provided [42, 62]. ## Theoretical underpinning and framework Twenty-one studies ($70\%$) reported embedding behaviour change theories into intervention design and delivery. Social cognitive theory and the trans-theoretical model were the two theories/models used most to underpin the interventions, with behaviour change techniques such as goal setting, motivational interviewing or action planning most frequently used (Table 2 and Fig. 2). When mapping against the NOURISHING framework, all studies aligned with the behaviour change communication domain, with the two policy areas of “nutrition education and skills”, and “nutrition advice and counselling in health care settings” identified. One study also mapped to the food environment domain, with the policy area of “economic tools to address affordability and purchase incentives” identified [42]. In this study, participants accumulated points and received a monetary reward at the end of the intervention relative to the number of healthy dietary choices logged. No studies aligned with the food system domain. ## Effectiveness of digital interventions Only nine studies ($30\%$) reported statistically significant improvements in vegetable intake (i.e., designated as effective interventions) compared with a control group [38, 44, 55] or compared with baseline. In the latter case, this included pre-post interventions [56, 58], uncontrolled randomised trials [42] and RCTs with no statistically significant increase in the control group (and no statistical comparison for between-group changes reported) [35, 54, 60]. There was heterogeneity in the method of reporting improvements in vegetable intake among effective studies, including serves/day and adherence to guidelines. Three studies reported change in serves/day, with the magnitude of this improvement ranging from 0.29 serves/day [38] to 1 serve/day [42]. One study reported that $87\%$ of participants improved vegetable intake compared to $29\%$ of the control group [55], while another study reported a $7\%$ increase in adherence to ≥ 500 g/day of vegetables compared to baseline (and a non-significant increase in the control group) [54]. One pre-post study reporting a 3.75 points increase in vegetable score (as a component of the Global Diet Quality Index; maximum score 100) compared with baseline [56]. Two studies also reported improvements in vegetable intake, but limited data on the magnitude were provided and no statistical comparisons were reported [36, 37]. Three studies reported a decline in vegetable intake compared with baseline, including a 0.2 portion per day decline [47], a $4\%$ decline in participants consuming ≥ 2 serves/day [50] and a further study did not report any data on the magnitude of change [49]. No studies included in this review reported on attitudes towards, knowledge of, skills in respect of, self-efficacy, access to and/or intentions with respect to vegetables. ## Features of effective digital interventions Of the nine effective interventions, sample sizes ranged from 120 to 171 participants (Table 1). A slightly greater percentage of effective interventions were in healthy populations ($$n = 6$$/9; $67\%$) compared with the ineffective interventions ($$n = 13$$/21; $62\%$). Almost half of effective interventions were in younger adults (< 40y; $$n = 4$$, $44\%$), compared with $19\%$ ($$n = 4$$) of ineffective interventions. Neither of the two interventions delivered exclusively in rural communities were effective. Vegetable intake was the primary outcome in $78\%$ ($$n = 7$$) of the effective interventions, compared with $57\%$ ($$n = 12$$) of the ineffective interventions. Of the effective interventions, $33\%$ ($$n = 3$$) utilised an app [35, 54, 58], $22\%$ ($$n = 2$$) used a website [44, 60] and $11\%$ ($$n = 1$$) used SMS messages [55] in isolation, while one study used an app and activity tracker [42] and two studies utilised a combination of four or more delivery modalities (including apps, emails, SMS messages, phone calls, videos and websites) [38, 56]. As shown in Table 2, this contrasted with the ineffective interventions, where $29\%$ ($$n = 6$$) utilised an app [33, 34, 36, 39, 40, 62], $10\%$ ($$n = 2$$) used a website [47, 51], and $10\%$ ($$n = 2$$) used SMS messages [41, 46] in isolation, while $52\%$ ($$n = 11$$) used a combination of delivery modalities [37, 43, 45, 48–50, 52, 53, 57, 59, 61]. The features of effective and ineffective interventions are compared in Fig. 3. Eighty nine percent ($$n = 8$$) of the effective studies referenced behavioural theories in their design (Table 2), including the trans-theoretical model theory [55], the social cognitive theory [38] and the health action process [60]. In contrast, $61\%$ ($$n = 12$$) of the ineffective interventions referenced theories. Sixty-seven percent ($$n = 6$$) effective interventions delivered personalised information, which included personalised dietary advice from a dietitian [34, 54] and personalised menus and food shopping lists based on taste preferences and calorie needs [44]. Of the ineffective interventions, $81\%$ ($$n = 17$$) included personalisation methods. Forty-four percent ($$n = 4$$) of the effective interventions included some level of input from stakeholders into the design of the intervention, compared with $38\%$ ($$n = 8$$) of the ineffective interventions. This included design input from health care professionals, such as dietitians and general practitioners, and software engineers, but rarely involved meaningful consumer involvement. Only one (ineffective) intervention included true co-design, with iterative workshops with researchers, older adults (the target population) and health and social care professionals (Fig. 3).Fig. 3Heat map summary of features of effective and ineffective interventions to increase vegetable intake ## Discussion In this systematic review we identified a paucity of digital interventions that were effective at increasing vegetable intake in adults. Embedding of behaviour change theories and inclusion of stakeholders in the design of the intervention were more common among effective interventions. We also observed that personalisation did not appear to be a feature of effective interventions. However, personalisation methods varied considerably, thus it is possible that the nature or degree of personalisation did not meet the needs of the user. Use of more comprehensive co-design methods may help to ensure that personalisation approaches are informed by the needs of the target population. This review found that researchers used multiple, heterogenous indictors of vegetable intake when reporting outcomes from interventions, which prohibited quantitative synthesis of the magnitude of change in vegetable intake. Nevertheless, in the studies that reported serves/day, vegetable intake increased by between 0.29 to 1 serve/day, which is comparable to evidence from mass media campaigns (0.6 serves/day) [63] and workplace interventions (0.32 serves/day) [64]. Reviews of the effectiveness of interventions to increase vegetable intake specifically are lacking. Our exclusion of studies that did not report intakes of fruit and vegetables separately was critical for discerning how interventions impacted on vegetable intake alone. Given the considerable health and economic benefit at the population level of even a small increase in vegetable intake [65], future research should report these outcomes consistently, and separately from fruit intake. Further, some studies in this review reported vegetable intake as a secondary outcome, or as part of an overall diet quality scores, such as the Mediterranean diet [47, 50]. As a result, interventions targeting more than just vegetable intake may have dedicated less resources to increasing vegetable intake per se and may not have been suitably powered to detect effects on vegetable intake. Although the use of different indicators did not help explain any differences in intervention effectiveness, future interventions should report the magnitude of between-group changes in vegetable intake to ensure that results can be included in a quantitative synthesis. Degrees of personalisation varied considerably between studies, with no clear difference in the type or level of personalisation between effective and ineffective interventions. Moreover, understanding of personalisation methods used in the included studies was limited because the reporting of the design and delivery of personalisation was often minimal. Nonetheless, while many studies used personalised feedback and/or action plans based on demographic characteristics and/or participant preferences, only one study offered participants the ability to customise the timing and delivery of their personalised content [41]. A recent study of the personalisation of digital health information identified that the preferred approach differed by age group, where young adults were more satisfied with user-driven personalisation as distinct from system-driven personalisation [66]. While system-driven personalisation offers the advantage of lower cognitive load for the user, a user-driven approach offers a greater sense of autonomy. As a result, certain population groups, such as those with higher digital health literacy, may wish to exert more control over their personalisation [67]. This degree of autonomy should be considered when designing more sophisticated approaches to personalisation, such as artificial intelligence algorithms and machine learning [68]. Digital technologies are well suited to delivering large-scale personalised dietary support, because the content, frequency and timing of the intervention can be modified to meet the needs and preferences of the user [15]. Thus, future digital interventions for increasing vegetable intake may be improved by better reporting of the use of personalisation methods, ensuring that the tool has sufficient flexibility for the content and modality to be personalised and by considering the use of more sophisticated digital techniques to achieve personalisation. Embedded behaviour change theories were common in both the effective and ineffective interventions. There was no clear difference in the application of these theories between effective or ineffective interventions. However, it is worth noting that all interventions, bar one [42], mapped to the behaviour change communication domain of policy actions outlined in the NOURISHING framework and did not map to the food environment or food system domains. This contrasts with a recent review of settings-based and digital interventions, where studies often mapped to the food environment domain, by including strategies such as free provision of fruit and vegetables in workplaces [5]. In addition, in the review by Wolfenden et al., all interventions that mapped to the food environment domain were effective at increasing fruit and vegetable intake. The lack of behaviour change strategies at the food environment level identified in our review requires further attention in future research. For example, food prescription programs that aim to improve the accessibility and affordability of healthy foods have shown promise for improving vegetable intake and reducing food insecurity [69], and could be integrated into digital healthcare interventions via partnerships with relevant stakeholders, such as health care providers, food markets or foodbanks. This is particularly important in the era of the COVID-19 pandemic, which has increased consumer acceptance and use of digital health initiatives [70], as well as stimulated a concerted global investment in building more food secure communities [71, 72]. A paucity of studies in this review included diverse populations. Similar to other reviews of digital interventions [73], most study populations were female-skewed, and of mid or older age (> 40 years). Disadvantaged populations, such as those with lower socio-economic position and who are culturally and linguistically diverse, were under-represented. Thus, there is potential for selection bias and response bias to have limited the generalisability of the findings from these studies. In addition, the “digital divide” persists, where lower income countries, racial/ethnic minorities, older adults, and individuals who live in lower income households and rural areas have less access to the internet and lower digital literacy [74]. However, global internet use has doubled from 33 to $65\%$ in the last decade [16], and there is some evidence that digital inclusion is increasing [10, 11, 75]. Therefore, there is an opportunity to test the effectiveness of digital interventions in diverse populations to help reduce dietary (and health) inequities and improve digital literacy. Moreover, findings from this review confirm recent research highlighting a lack of nutrition research in rural settings, where there is inequitable access to healthcare and fresh produce, such as fruit and vegetables [13]. As a result, future interventions should consider external validity in other less well-represented population groups such as individuals with lower socioeconomic position and those living in rural settings. Digital interventions are well suited to achieve this because of their potential for linguistic and cultural localisation, national scalability at relatively low cost, and the global drive to improve digital health equity in rural and disadvantaged communities. Fewer than half of included studies reported on interventions that had been developed with some level of design input from stakeholders. In addition, intervention end users were very rarely involved and only one intervention specifically mentioned the use of co-design approaches. Recent reviews on the use of co-design have shown mixed findings, with one review of co-design in health settings showing widespread use [24], and another review of co-design in nutrition and health interventions in community-dwelling adults identifying no interventions implementing a complete co-design process [25]. A more recent review of the use of co-design specifically in nutrition interventions delivered within a healthcare, community or academic setting identified only two studies reporting a partnership with consumers across all stages of research [76]. Taken together, these findings reinforce the need for consistent use of co-design terminology, better reporting of design and development processes and more widespread utilisation of a translational framework for the evaluation of health interventions, such as the NASSS (non-adoption, abandonment, scale-up, spread, sustainability) framework [77]. Future research should include co-design methods at multiple levels (i.e., stakeholders with lived experience as well as technical expertise) and include stakeholders throughout, from project conception to dissemination. Outcomes from this research have implications for the use of digital tools to improve public health nutrition and provide insights into future research needs. Despite the potential for digital tools to improve access to dietary interventions, the persistent threat that digital technologies can exacerbate social inequities of health remains [78]. As such, the inclusion of diverse populations groups in the design and implementation of digital interventions remains a priority. Without this, there is a risk that some population groups may experience barriers to the use of digital technologies, including individuals experiencing socio-economic disadvantage, individuals with disabilities, individuals who require cultural adaptations, and those with low food and digital literacy and self-efficacy [79]. Countries with diverse geographic settings and the potential for disparities in internet access, such as Australia, should ensure that digital interventions are tested in rural settings, which would otherwise be a missed opportunity for addressing widening health disparities [80]. Further, with a paucity of co-design research and consideration of environmental influences, this research suggests that the design of digital interventions to increase vegetable intake is not yet optimal in maximising effectiveness. This review has several strengths and limitations. The main strength was the systematic approach used to search, screen, and synthesise the literature, including the PROSPERO registration of the review protocol and the use of Cochrane risk of bias tools. By limiting the search to articles published in English and including experimental study designs only, it is possible that studies that would be informative for the design of future interventions were missed. As most studies included in this review were rated as high risk for bias, findings should be interpreted with caution. Due to the heterogenous study populations and intervention designs, including small sample sizes, no quantitative synthesis could be performed. Further, intervention outcomes for vegetable intake will be subject to misreporting biases due to the self-report nature of dietary assessment tools available, which includes the potential for participants to introduce bias as their food literacy and understanding of dietary reporting improves. Lastly, grey literature and commercial products for dietary behaviour change were excluded, which may have limited our ability to capture evidence of co-design research and the full range of digital tools designed to increase vegetable intake. ## Conclusions Few digital interventions have been effective in increasing vegetable intake among adults. Embedding behaviour change theories and involving stakeholders in intervention design may increase the likelihood of effectiveness. Personalisation was not a distinctive feature of effective digital interventions, however, this feature remains poorly understood due to considerable variation in its design and reporting. There is an unmet opportunity for the use of more comprehensive co-design methods to ensure personalisation approaches meet the needs of target populations. Furthermore, future digital interventions should consider strategies that address both behaviour change and food environment influences. ## Supplementary Information Additional file 1. PRIMSA checklist. Additional file 2. Search strategy. Additional file 3. Cochrane Risk of Bias for RCTs, cluster RCTs and non-RCTs. ## References 1. Afshin A, Sur P, Fay K, Cornaby L, Ferrara G, Salama J. **Health effects of dietary risks in 195 countries, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017**. *Lancet* (2019.0) **393** 1958-1972. DOI: 10.1016/S0140-6736(19)30041-8 2. 2.Lim SS, Vos T, Flaxman AD, Danaei G, Shibuya K, Adair-Rohani H. A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet. 2012;380:2224–60. 3. 3.Food-Based Dietary Guidelines. Available online: http://www.fao.org/nutrition/education/food-dietary-guidelines/home/en/ (Accessed 9 May 2022). 4. Kalmpourtzidou A, Eilander A, Talsma EF. **Global Vegetable Intake and Supply Compared to Recommendations: A Systematic Review**. *Nutrients* (2020.0) **12** 1558. DOI: 10.3390/nu12061558 5. Wolfenden L, Barnes C, Lane C, McCrabb S, Brown HM, Gerritsen S. **Consolidating evidence on the effectiveness of interventions promoting fruit and vegetable consumption: an umbrella review**. *Int J Behav Nutr Phys Act* (2021.0) **18** 11. DOI: 10.1186/s12966-020-01046-y 6. Glasson C, Chapman K, James E. **Fruit and vegetables should be targeted separately in health promotion programmes: differences in consumption levels, barriers, knowledge and stages of readiness for change**. *Public Health Nutr* (2011.0) **14** 694-701. DOI: 10.1017/S1368980010001643 7. Godrich S, Kent K, Murray S, Auckland S, Lo J, Blekkenhorst L. **Australian Consumer Perceptions of Regionally Grown Fruits and Vegetables: Importance, Enablers, and Barriers**. *Int J Environ Res Public Health* (2019.0) **17** 63. DOI: 10.3390/ijerph17010063 8. Pollard J, Kirk SFL, Cade JE. **Factors affecting food choice in relation to fruit and vegetable intake: a review**. *Nutr Res Rev* (2002.0) **15** 373-387. DOI: 10.1079/NRR200244 9. Livingstone KM, Burton M, Brown AK, McNaughton SA. **Exploring barriers to meeting recommendations for fruit and vegetable intake among adults in regional areas: A mixed-methods analysis of variations across socio-demographics**. *Appetite* (2020.0) **153** 104750. DOI: 10.1016/j.appet.2020.104750 10. Taj F, Klein MCA, van Halteren A. **Digital Health Behavior Change Technology: Bibliometric and Scoping Review of Two Decades of Research**. *JMIR Mhealth Uhealth* (2019.0) **7** e13311. DOI: 10.2196/13311 11. Nour M, Chen J, Allman-Farinelli M. **Efficacy and External Validity of Electronic and Mobile Phone-Based Interventions Promoting Vegetable Intake in Young Adults: Systematic Review and Meta-Analysis**. *J Med Internet Res* (2016.0) **18** e58. DOI: 10.2196/jmir.5082 12. 12.Noy S, Patrick R, Ryan I, Nuttman S, Henderson-Wilson C. Scoping study of community interventions that promote vegetable consumption 2017. Available from: https://www.deakin.edu.au/__data/assets/pdf_file/0007/1251295/2018-01-31-ScopingStudyVegConsumption-WEB.pdf. Accessed 8 Jan 2023. 13. Alston L, Partridge SR. **Limited dietary interventions in rural Australian communities: A systematic review**. *Nutr Diet* (2020.0) **78** 57-68. DOI: 10.1111/1747-0080.12612 14. 14.Rekhy R, McConchie R. Promoting consumption of fruit and vegetables for better health. Have campaigns delivered on the goals? Appetite. 2014;79:113–23. 15. Celis-Morales C, Lara J, Mathers JC. **Personalising nutritional guidance for more effective behaviour change**. *Proc Nutr Soc* (2015.0) **74** 130-138. DOI: 10.1017/S0029665114001633 16. 16.Statistica Research Department. Internet and social media users in the world 2022. 2023. Available from: https://www.statistica.com/statistics/617136/digital-population-worldwide/ (Accessed 9 Jan 2023) 17. 17.Petersen A. Digital Health and Technological Promise: A Sociological Inquiry (1st ed.). Routledge: 2019. 10.4324/9781315200880. Accessed 8 Jan 2023. 18. 18.Eurostat. Internet activities. 2022. Available from: https://ec.europa.eu/eurostat/ (Accessed 9 Jan 2023) 19. 19.Australian Digital Health Agency. Digital Health Survey. 2021. Available from: https://nationalstrategy.digitalhealth.gov.au/ (Accessed 9 Jan 2023) 20. Rodriguez Rocha NP, Kim H. **eHealth Interventions for Fruit and Vegetable Intake: A Meta-Analysis of Effectiveness**. *Health Educ Behav* (2019.0) **46** 947-959. DOI: 10.1177/1090198119859396 21. Jinnette R, Narita A, Manning B, McNaughton SA, Mathers JC, Livingstone KM. **Does Personalized Nutrition Advice Improve Dietary Intake in Healthy Adults? A Systematic Review of Randomized Controlled Trials**. *Adv Nutr* (2021.0) **12** 657-669. DOI: 10.1093/advances/nmaa144 22. 22.O’Brien N, Heaven B, Teal G, Evans EH, Cleland C, Moffatt S, et al. Integrating Evidence From Systematic Reviews, Qualitative Research, and Expert Knowledge Using Co-Design Techniques to Develop a Web-Based Intervention for People in the Retirement Transition. J Med Internet Res. 2016;18(8): e210. 23. 23.Singleton A, Raeside R, Partridge SR, Hayes M, Maka K, Hyun KK, et al. Co-designing a Lifestyle-Focused Text Message Intervention for Women After Breast Cancer Treatment: Mixed Methods Study. J Med Internet Res. 2021;23(6):e27076-e. 24. Slattery P, Saeri AK, Bragge P. **Research co-design in health: a rapid overview of reviews**. *Health Res Policy Syst* (2020.0) **18** 17. DOI: 10.1186/s12961-020-0528-9 25. Tay BSJ, Cox DN, Brinkworth GD, Davis A, Edney SM, Gwilt I. **Co-Design Practices in Diet and Nutrition Research: An Integrative Review**. *Nutrients* (2021.0) **13** 3593. DOI: 10.3390/nu13103593 26. Campbell MK, McLerran D, Turner-McGrievy G, Feng Z, Havas S, Sorensen G. **Mediation of adult fruit and vegetable consumption in the National 5 A Day for Better Health community studies**. *Ann Behav Med* (2008.0) **35** 49-60. DOI: 10.1007/s12160-007-9002-y 27. Brimblecombe J, Ferguson M, Barzi F, Brown C, Ball K. **Mediators and moderators of nutrition intervention effects in remote Indigenous Australia**. *Br J Nutr* (2018.0) **119** 1424-1433. DOI: 10.1017/S0007114518000880 28. Hawkes C, Jewell J, Allen K. **A food policy package for healthy diets and the prevention of obesity and diet-related non-communicable diseases: the NOURISHING framework**. *Obes Rev* (2013.0) **14** 159-168. DOI: 10.1111/obr.12098 29. Campbell M, McKenzie JE, Sowden A, Katikireddi SV, Brennan SE, Ellis S. **Synthesis without meta-analysis (SWiM) in systematic reviews: reporting guideline**. *BMJ* (2020.0) **368** l6890. DOI: 10.1136/bmj.l6890 30. Western MJ, Armstrong MEG, Islam I, Morgan K, Jones UF, Kelson MJ. **The effectiveness of digital interventions for increasing physical activity in individuals of low socioeconomic status: a systematic review and meta-analysis**. *Int J Behav Nutr Phys Act* (2021.0) **18** 148. DOI: 10.1186/s12966-021-01218-4 31. Sterne JAC, Savović J, Page MJ, Elbers RG, Blencowe NS, Boutron I. **RoB 2: a revised tool for assessing risk of bias in randomised trials**. *BMJ* (2019.0) **366** l4898. DOI: 10.1136/bmj.l4898 32. Sterne JA, Hernán MA, Reeves BC, Savović J, Berkman ND, Viswanathan M. **ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions**. *BMJ* (2016.0) **355** i4919. DOI: 10.1136/bmj.i4919 33. Abu-Saad K, Murad H, Barid R, Olmer L, Ziv A, Younis-Zeidan N. **Development and Efficacy of an Electronic, Culturally Adapted Lifestyle Counseling Tool for Improving Diabetes-Related Dietary Knowledge: Randomized Controlled Trial Among Ethnic Minority Adults With Type 2 Diabetes Mellitus**. *J Med Internet Res* (2019.0) **21** e13674. DOI: 10.2196/13674 34. Alonso-Domínguez R, García-Ortiz L, Patino-Alonso MC, Sánchez-Aguadero N, Gómez-Marcos MA, Recio-Rodríguez JI. **Effectiveness of A Multifactorial Intervention in Increasing Adherence to the Mediterranean Diet among Patients with Diabetes Mellitus Type 2: A Controlled and Randomized Study (EMID Study)**. *Nutrients* (2019.0) **11** 162. DOI: 10.3390/nu11010162 35. Bhurosy T, Middlestadt SE, Lin HC, Thiagarajah K, Fly AD. **A Randomized mHealth Trial to Promote Vegetable Intake Through Counting and Goal Setting**. *J Nutr Educ Behav* (2020.0) **52** 1111-1119. DOI: 10.1016/j.jneb.2020.08.009 36. 36.Bozorgi A, Hosseini H, Eftekhar H, Majdzadeh R, Yoonessi A, Ramezankhani A, et al. The effect of the mobile “blood pressure management application” on hypertension self-management enhancement: a randomized controlled trial. Trials. 2021;22(1):413. 37. 37.Brown ON, O’Connor LE, Savaiano D. Mobile MyPlate: a pilot study using text messaging to provide nutrition education and promote better dietary choices in college students. J Am Coll Health. 2014;62(5):320–7. 38. Chan JM, Van Blarigan EL, Langlais CS, Zhao S, Ramsdill JW, Daniel K. **Feasibility and Acceptability of a Remotely Delivered, Web-Based Behavioral Intervention for Men With Prostate Cancer: Four-Arm Randomized Controlled Pilot Trial**. *J Med Internet Res* (2020.0) **22** e19238. DOI: 10.2196/19238 39. 39.Debon R, Bellei EA, Biduski D, Volpi SS, Alves ALSA, Portella MR, et al. Effects of using a mobile health application on the health conditions of patients with arterial hypertension: A pilot trial in the context of Brazil’s Family Health Strategy. Sci Rep. 2020;10(1):6009. 40. Elbert SP, Dijkstra A, Oenema A. **A Mobile Phone App Intervention Targeting Fruit and Vegetable Consumption: The Efficacy of Textual and Auditory Tailored Health Information Tested in a Randomized Controlled Trial**. *J Med Internet Res* (2016.0) **18** e147. DOI: 10.2196/jmir.5056 41. 41.Fjeldsoe BS, Goode AD, Phongsavan P, Bauman A, Maher G, Winkler E, et al. Get Healthy, Stay Healthy: Evaluation of the Maintenance of Lifestyle Changes Six Months After an Extended Contact Intervention. JMIR mHealth uHealth. 2019;7(3):e11070-e. 42. Gilson ND, Pavey TG, Wright OR, Vandelanotte C, Duncan MJ, Gomersall S. **The impact of an m-Health financial incentives program on the physical activity and diet of Australian truck drivers**. *BMC Public Health* (2017.0) **17** 467. DOI: 10.1186/s12889-017-4380-y 43. 43.Goni L, de la O V, Barrio-López MT, Ramos P, Tercedor L, Ibañez-Criado JL, et al. A Remote Nutritional Intervention to Change the Dietary Habits of Patients Undergoing Ablation of Atrial Fibrillation: Randomized Controlled Trial. J Med Internet Res. 2020;22(12):e21436. 44. Hansel B, Giral P, Gambotti L, Lafourcade A, Peres G, Filipecki C. **A Fully Automated Web-Based Program Improves Lifestyle Habits and HbA1c in Patients With Type 2 Diabetes and Abdominal Obesity: Randomized Trial of Patient E-Coaching Nutritional Support (The ANODE Study)**. *J Med Internet Res* (2017.0) **19** e360. DOI: 10.2196/jmir.7947 45. Hebden L, Cook A, Ploeg HP, King L, Bauman A, Allman-Farinelli M. **A mobile health intervention for weight management among young adults: a pilot randomised controlled trial**. *J Hum Nutr Diet* (2014.0) **27** 322-332. DOI: 10.1111/jhn.12155 46. 46.Jahan Y, Rahman MM, Faruque ASG, Chisti MJ, Kazawa K, Matsuyama R, et al. Awareness Development and Usage of Mobile Health Technology Among Individuals With Hypertension in a Rural Community of Bangladesh: Randomized Controlled Trial. J Med Internet Res. 2020;22(12):N.PAG-N.PAG. 47. 47.Lara J, O’Brien N, Godfrey A, Heaven B, Evans EH, Lloyd S, et al. Pilot Randomised Controlled Trial of a Web-Based Intervention to Promote Healthy Eating, Physical Activity and Meaningful Social Connections Compared with Usual Care Control in People of Retirement Age Recruited from Workplaces. PLoS ONE. 2016;11(7): e0159703. 48. Lombard C, Harrison C, Kozica S, Zoungas S, Ranasinha S, Teede H. **Preventing Weight Gain in Women in Rural Communities: A Cluster Randomised Controlled Trial**. *PLoS Med* (2016.0) **13** e1001941. DOI: 10.1371/journal.pmed.1001941 49. 49.Pope ZC, Barr-Anderson DJ, Lewis BA, Pereira MA, Gao Z. Use of Wearable Technology and Social Media to Improve Physical Activity and Dietary Behaviors among College Students: A 12-Week Randomized Pilot Study. Int J Environ Res Public Health. 2019;16(19)3579. 50. Recio-Rodriguez JI, Agudo-Conde C, Martin-Cantera C, González-Viejo MN, Fernandez-Alonso MDC, Arietaleanizbeaskoa MS. **Short-Term Effectiveness of a Mobile Phone App for Increasing Physical Activity and Adherence to the Mediterranean Diet in Primary Care: A Randomized Controlled Trial (EVIDENT II Study)**. *J Med Internet Res* (2016.0) **18** e331. DOI: 10.2196/jmir.6814 51. Schulz DN, Kremers SPJ, Vandelanotte C, van Adrichem MJG, Schneider F, Candel MJJM. **Effects of a web-based tailored multiple-lifestyle intervention for adults: a two-year randomized controlled trial comparing sequential and simultaneous delivery modes**. *J Med Internet Res* (2014.0) **16** e26. DOI: 10.2196/jmir.3094 52. Kerr DA, Harray AJ, Pollard CM, Dhaliwal SS, Delp EJ, Howat PA. **The connecting health and technology study: a 6-month randomized controlled trial to improve nutrition behaviours using a mobile food record and text messaging support in young adults**. *Int J Behav Nutr Phys Act* (2016.0) **13** 52. DOI: 10.1186/s12966-016-0376-8 53. Turner-McGrievy GM, Beets MW, Moore JB, Kaczynski AT, Barr-Anderson DJ, Tate DF. **Comparison of traditional versus mobile app self-monitoring of physical activity and dietary intake among overweight adults participating in an mHealth weight loss program**. *J Am Med Informatics Assoc* (2013.0) **20** 513-518. DOI: 10.1136/amiajnl-2012-001510 54. Wang M, Guo Y, Zhang Y, Xie S, Yu Z, Luo J. **Promoting healthy lifestyle in Chinese college students: evaluation of a social media-based intervention applying the RE-AIM framework**. *Eur J Clin Nutr* (2021.0) **75** 335-344. DOI: 10.1038/s41430-020-0643-2 55. Wang X, Liu D, Du M, Hao R, Zheng H, Yan C. **The role of text messaging intervention in Inner Mongolia among patients with type 2 diabetes mellitus: a randomized controlled trial**. *BMC Med Inform Decis Mak* (2020.0) **20** 90. DOI: 10.1186/s12911-020-01129-7 56. Cantisano LM, Gonzalez-Soltero R, Blanco-Fernández A, Belando-Pedreño N. **ePSICONUT: An e-Health Programme to Improve Emotional Health and Lifestyle in University Students**. *Int J Environ Res Public Health* (2022.0) **19** 9253. DOI: 10.3390/ijerph19159253 57. 57.Celis-Morales C, Livingstone KM, Marsaux CFM, Macready AL, Fallaize R, O’Donovan CB, et al. Effect of personalized nutrition on health-related behaviour change: evidence from the Food4me European randomized controlled trial. Int J Epidemiol. 2016;46(2):578–88. 58. Hendrie GA, Hussain MS, Brindal E, James-Martin G, Williams G, Crook A. **Impact of a Mobile Phone App to Increase Vegetable Consumption and Variety in Adults: Large-Scale Community Cohort Study**. *JMIR Mhealth Uhealth* (2020.0) **8** e14726. DOI: 10.2196/14726 59. Perez-Junkera G, Vázquez-Polo M, Eizagirre FJ, Benjumea L, Tutau C, Esteban B. **Application of a Platform for Gluten-Free Diet Evaluation and Dietary Advice: From Theory to Practice**. *Sensors* (2022.0) **22** 732. DOI: 10.3390/s22030732 60. 60.Plaete J, De Bourdeaudhuij I, Verloigne M, Crombez G. Acceptability, feasibility and effectiveness of an eHealth behaviour intervention using self-regulation: “MyPlan.” Patient Educ Couns. 2015;98(12):1617–24. 61. Williams SL, To Q, Vandelanotte C. **What is the effectiveness of a personalised video story after an online diabetes risk assessment? A Randomised Controlled Trial**. *PLoS ONE* (2022.0) **17** e0264749. DOI: 10.1371/journal.pone.0264749 62. Zenun Franco R, Fallaize R, Weech M, Hwang F, Lovegrove JA. **Effectiveness of Web-Based Personalized Nutrition Advice for Adults Using the eNutri Web App: Evidence From the EatWellUK Randomized Controlled Trial**. *J Med Internet Res* (2022.0) **24** e29088. DOI: 10.2196/29088 63. Pollard CM, Miller MR, Daly AM, Crouchley KE, O’Donoghue KJ, Lang AJ. **Increasing fruit and vegetable consumption: success of the Western Australian Go for 2&5®campaign**. *Public Health Nutr* (2008.0) **11** 314-320. DOI: 10.1017/S1368980007000523 64. Kushida O, Murayama N. **Effects of environmental intervention in workplace cafeterias on vegetable consumption by male workers**. *J Nutr Educ Behav* (2014.0) **46** 350-358. DOI: 10.1016/j.jneb.2014.05.001 65. 65.Deloitte Access Economics. The impact of increasing vegetable consumption on health expenditure 2016 [Available from: https://www2.deloitte.com/au/en/pages/economics/articles/increasing-vegetable-consumption-health-expenditure-impact.html]. 66. Nguyen MH, Bol N, King AJ. **Customisation versus Personalisation of Digital Health Information: Effects of Mode Tailoring on Information Processing Outcomes**. *Eur J Health Commun* (2020.0) **1** 30-54. DOI: 10.47368/ejhc.2020.003 67. Smit ES, Bol N. **From self-reliers to expert-dependents: identifying classes based on health-related need for autonomy and need for external control among mobile users**. *Media Psychol* (2020.0) **23** 391-414. DOI: 10.1080/15213269.2019.1604235 68. Livingstone KM, Ramos-Lopez O, Pérusse L, Kato H, Ordovas JM, Martínez JA. **Precision nutrition: A review of current approaches and future endeavors**. *Trends Food Sci Technol* (2022.0) **128** 253-264. DOI: 10.1016/j.tifs.2022.08.017 69. Little M, Rosa E, Heasley C, Asif A, Dodd W, Richter A. **Promoting Healthy Food Access and Nutrition in Primary Care: A Systematic Scoping Review of Food Prescription Programs**. *Am J Health Promot* (2022.0) **36** 518-536. DOI: 10.1177/08901171211056584 70. Sturgiss E, Desborough J, Hall Dykgraaf S, Matenge S, Dut G, Davis S. **Digital health to support primary care provision during a global pandemic**. *Aust Health Rev* (2022.0) **46** 269-272. DOI: 10.1071/AH21263 71. 71.England PH. Community-centred public health: taking a whole system approach 2020 [Available from: https://www.gov.uk/government/publications/community-centred-public-health-taking-a-whole-system-approach. 72. Gilmore B, Ndejjo R, Tchetchia A, de Claro V, Mago E, Diallo AA. **Community engagement for COVID-19 prevention and control: a rapid evidence synthesis**. *BMJ Glob Health* (2020.0) **5** e003188. DOI: 10.1136/bmjgh-2020-003188 73. Hutchesson MJ, Gough C, Müller AM, Short CE, Whatnall MC, Ahmed M. **eHealth interventions targeting nutrition, physical activity, sedentary behavior, or obesity in adults: A scoping review of systematic reviews**. *Obes Rev* (2021.0) **22** e13295. DOI: 10.1111/obr.13295 74. Nguyen A, Mosadeghi S, Almario CV. **Persistent digital divide in access to and use of the Internet as a resource for health information: Results from a California population-based study**. *Int J Med Inform* (2017.0) **103** 49-54. DOI: 10.1016/j.ijmedinf.2017.04.008 75. Thomas J, Barraket J, Parkinson S, Wilson C, Holcombe-James I, Kennedy J, Mannell K, Brydon A. *Australian Digital Inclusion Index: 2021* (2021.0) 76. 76.Meloncelli N, Young A, Christoffersen A, Rushton A, Zhelnov P, Wilkinson SA, et al. Co-designing nutrition interventions with consumers: A scoping review. J Hum Nutr Diet. 2022. 10.1111/jhn.13082. 77. Abimbola S, Patel B, Peiris D, Patel A, Harris M, Usherwood T. **The NASSS framework for ex post theorisation of technology-supported change in healthcare: worked example of the TORPEDO programme**. *BMC Med* (2019.0) **17** 233. DOI: 10.1186/s12916-019-1463-x 78. Weiss D, Rydland HT, Øversveen E, Jensen MR, Solhaug S, Krokstad S. **Innovative technologies and social inequalities in health: A scoping review of the literature**. *PLoS ONE* (2018.0) **13** e0195447. DOI: 10.1371/journal.pone.0195447 79. Resta P, Laferrière T, McLaughlin R, Kouraogo A, Voogt J, Knezek G, Christensen R, Lai K-W. **Issues and Challenges Related to Digital Equity: An Overview**. *Second Handbook of Information Technology in Primary and Secondary Education* (2018.0) 987-1004 80. Alston L, Jacobs J, Allender S, Nichols M. **A comparison of the modelled impacts on CVD mortality if attainment of public health recommendations was achieved in metropolitan and rural Australia**. *Public Health Nutr* (2020.0) **23** 339-347. DOI: 10.1017/S136898001900199X
--- title: 'Perceived readiness for diabetes and cardiovascular care delivery in Mangochi, Malawi: multicentre study from healthcare providers’ perspectives' authors: - Prosper Lutala - Peter Nyasulu - Adamson S. Muula journal: BMC Primary Care year: 2023 pmcid: PMC10042413 doi: 10.1186/s12875-023-02033-5 license: CC BY 4.0 --- # Perceived readiness for diabetes and cardiovascular care delivery in Mangochi, Malawi: multicentre study from healthcare providers’ perspectives ## Abstract ### Background Despite the expected prevalence rise of $98.1\%$ for diabetes between 2010 and 2030 in sub-Saharan Africa (SSA) and the anticipated rise of both diabetes and cardiovascular diseases (CVDs) in Malawi from their current figures (5.6 and $8.9\%$; respectively), data on the readiness of health facilities to provide diabetes and cardiovascular diseases in Mangochi district is not available. Therefore, this study aimed to assess the readiness of health facilities to provide services for diabetes and cardiovascular diseases. ### Methods An exploratory study was conducted from July to early September 2021 in 34 health facilities in Mangochi, Malawi. Forty-two participants were purposefully selected. They included medical officers, clinical officers, medical assistants, and registered nurses. The study used semi-structured interviews (for qualitative data) with a checklist (for quantitative data) to provide information about the readiness of services (such as guidelines and trained staff, drugs, diagnosis capacity and equipment, essential medicines, community services, and education/counseling).The thematic content analysis and basic descriptive statistics were carried out. ### Results The following main theme emerged from the qualitative part: low use of diabetes-cardiovascular disease (CVD) services. This was due to: health facility factors (shortage of drugs and supplies, poor knowledge, few numbers and lack of training of providers, and absent copies of guidelines), patients factors (poor health-seeking behaviour, lack of education and counseling for many), and community factors (very limited community services for diabetes and CVDs, lack of transport policy and high transportation costs). Data from the checklists revealed low readiness scores across domains (below the $75\%$ target) in diabetes and cardiovascular diseases: trained staff and guidelines ($26.5\%$ vs. $32.4\%$); diagnosis capacity and equipment ($63.7\%$ vs. $66.2\%$); essential medicines ($33.5\%$ vs. $41.9\%$), and community services, and education and counseling ($37.5\%$ vs. $42.5\%$). ### Conclusion There were several noticeable shortfalls identified in the readiness of health facilities to provide diabetes and cardiovascular disease services in Mangochi health facilities. Any future intervention in diabetes-cardiovascular disease care in these areas must include these elements in its basic package. ## Background The burden of noncommunicable diseases (NCD) (generally defined as diseases which are not transmitted, chronic, and related to lifestyle), in general, is becoming a matter of concern globally. For this study, NCD include diabetes and cardiovascular diseases (CVDs), two major causes of death and disability. In defining CVDs, in addition to traditional the traditional definition of stroke and ischemic heart disease, we have included hypertension. Numerous reasons support this expanded definition: firstly, all three diseases are treated under the same roof; secondly, the major cause of stroke is hypertension in sub-Saharan Africa; thirdly, $90.3\%$ of patients with hypertension have a risk of developing stroke; and fourth, stroke represents the main cause of cardiovascular diseases in Malawi [1]. Readiness of service in a facility is defined as availability of a given services declared by staff and verified by the research team the day of the visit [2]. The changes overtime of the burden of the diabetes and cardiovascular diseases are also a reason of great concern. The prevalence of diabetes (which refers mainly to type 2 diabetes as the primary aetiology is lifestyle rather than hereditary origin) and CVDs are rising globally [3–5] and accounting for $70\%$ of mortality [6]. The incidence of diabetes is estimated to increase by at least $98.1\%$ between 2010 and 2030 in sub-Saharan Africa (SSA), which will become the fastest rise worldwide [7–11]. This includes Malawi [12], where the anticipated rise of diabetes and cardiovascular diseases from the current $5.6\%$ and $8.9\%$, respectively [11, 12]. The same apply for their corresponding risk factors such as alcohol abuse (up by $1\%$), smoking ($9\%$), physical inactivity ($21\%$), and insufficient fruit intake [12, 13]. Locally in Mangochi district, although published data on epidemiology doesn’t exist, evidence on review of national prevalence portrays the same high figures in the district [8]. ## Health services provision in Malawi Health services in Malawi are provided at different levels, with specific cadres and activities at each level as demonstrated in Table 1 below. In Mangochi, during data collection, 42 facilities which were reporting to the District Health Office (DHO) were grouped into five zones, each including public, private for profit, and non-profit private facilities. These zones were: Mangochi Boma, Monkey-Bay, Makanjira, Namwera, and Chilipa. Each zone comprises public, mission, and privates facilities. Table 1Public Health Provision in Malawi. This table resents the health provision in Malawi regarding the location in the health system and target population of each, and type of cadre working with activities of each at different levelsVariableCommunity carePrimary careSecondary careTertiary careLocation & target populationCommunity health services (health posts, dispensaries, maternity clinics)Community levelHealth centresPeople in radius of eight kilometers or 10,000 inhabitants andReferrals from the the communityDistrict hospitals, community hospitals and hospitals of the faith- based Christian Health Association of Malawi hospitals (CHAM)Queens (Southern), Zomba(Eastern), Kamuzu (Central) and Mzuzu (Northern)central hospitalsActivitiesdoor-to-door, village outreach clinics and mobile clinicsoutpatient and maternity servicesSecondary level services (outpatient primary and inpatient care, patients referred from health centres in their respective catchment areas (health centres, community hospitals, and hospitals of the faith- based Christian Health Association of Malawi hospitalsSpecialized care and dealing with patients referred from their respective secondary levels’ hospitals. Cadres providing serviceshealth surveillance assistants (HSAs), community midwives and community health volunteers medical assistants or clinical officers, nurses, health surveillance assistants, and community volunteersnon-specialist physicians, clinical officers, medical assistants, nurses/nurse midwives and allied health professionalsConsultants and specialists in diverse domains The public sector offers $60\%$ of health services available. The private services are divided into those for profit and those for non-profit which together provide the remaining $40\%$. The private for-profit services in Malawi are composed of traditional birth attendants, traditional healers and commercial actors, which are still under development in Malawi. Private services are mostly in rural areas. The public sector offers free services at the point of care. However, in private ones, access to care is subject to fees-for-service, albeit at a low rate. In order to increase access to essential health services in Malawi, the government created a list of diseases with high burden to be covered free of charge in both public and private sectors in Malawi called Essential Health Package (EHP) [14, 15]. However, some diseases (or conditions), given their severity and impact in some vulnerable groups (such as maternal child health) have been further included in an agreement document between the government and the Christian Health Association of Malawi (CHAM) called the Service Level Agreement (SLA). Under this agreement, parties are aiming to reduce financial access barriers in faith based facilities. These facilities do charge user-fees (in a population with very low income in general) and are in catchment areas where public health facilities do not exist. The above table includes private care providers in remote areas where there is no close government facility/hospital to provide free key services to their targeted population (pregnant women, children 0–2 years old) and to the surrounding community members. Such private providers are reimbursed by the government through the district assemblies. Despite adopting noncommunicable diseases as one of the conditions in the EHP, discussions are still underway between the government and the CHAM to include noncommunicable disease care in the SLA. A particular diabetes and cardiovascular programme in Malawi were piloted in a district hospital, Kasungu District Hospital. Lessons drawn from this experience were implemented in similar districts and tertiary hospitals. Implementation in health centres followed. However, roll-out in remote areas is going at a slow pace, resulting in several health centres lacking clinics for provision of diabetes and cardiovascular disease care. The government of *Malawi is* committed to implementing diabetes and cardiovascular disease services through different initiatives, including preventive, curative and health promotion as well as policy development [16]. In the context of this study we adopted NCD clinics as representing diabetes and cardiovascular disease care. The integration of other NCD into these clinics is still on-going. For instance, in many districts in Malawi, epilepsy is treated in a mental health clinic, or cancers are cared for in palliative care clinics. Most of these initiatives are taking place in primary and/or secondary level facilities. The primary care level is critical for the successful management of non-communicable diseases [17]. Diabetes management in primary health care is cost-effective [18–20]. However, so far in Malawi, designated diabetes-cardiovascular clinics in health centres are still rare; NCD care is mostly provided by clinics at secondary and tertiary hospitals. Where such services are available in primary or secondary care, the quality of this care, in general, has been either questioned [21–23] or not ascertained. As a response to this scarcity of locations providing NCD care, many health centres refer patients from their catchment areas to the nearest health facility providing diabetes-cardiovascular services. In most cases these services are available but very distant or they just do not exist due to stock-outs. Evidence on the availability and readiness of expected services, in different facilities, to effectively manage patients is still scarce. Readiness of a facility refers to an immediate and long-term adjustment to any introduced innovation focusing on policies, infrastructure, and processes [24]. Several studies assessing the availability and readiness of health systems to provide NCD care have been conducted in Low-and- Middle-Income Countries (LMICs) [22, 25–27]. Their overall result showed a suboptimal quality of NCD care. Recently in Malawi, a national cross-sectional study conducted in 55 health facilities showed, in almost all of them, a lack of educational materials, patient records and adequate resources for treatment and diagnosis of NCD [22]. Lack of knowledge and resources were found in Mangochi in a small study aimed to assess the quality of care patients with diabetes received [28]. Ever since, little development has taken place in the Malawi health system. Furthermore, readiness of a health system, being a dynamic concept which changes over time, is impacted by drugs, supplies, personnel, donations) and even by actual work happening at a given moment. The domains explored by different studies can also differ from study to study depending on design of each. Thus, we conducted this study primarily to assess providers’ perceptions regarding the readiness of the facilities where they work to provide diabetes and cardiovascular care in Mangochi. We secondarily conducted the study to assess the actual readiness of these facilities to provide this care. The findings of this study will generate new insights to be used to direct clinical work, to improve working conditions, or to inform policy-makers and researchers on areas for intervention or on gaps for future research. ## Study design This was an exploratory facility survey conducted in 32 health facilities of the Mangochi district between July 26 and August 25, 2021. The study has a qualitative component using an interview guide and a quantitative component which used a checklist. ## Study setting Mangochi district has a population of 1, 224,716 inhabitants as of 2022 [28]. It has 42 health facilities reporting to the district health office of which, three have stand-alone NCD clinics accredited by the National Ministry of Health. Out of the three, one was a faith-based facility and the other two were public. The remaining 39 facilities had no specific diabetes-CVD clinics and managed patients with these conditions through their general service provision in their outpatient departments. This assessment included both, accredited and non-accredited diabetes-cardiovascular facilities providing these services. The decision to include both accredited and non-accredited facilities in list of facilities providing diabetes-cardiovascular diseases care aligns well with on-going discussion on integration. Current evidence argues that integration of health service delivery within the primary health care context increases implementation efficiency and user-satisfaction. The 32 facilities were conveniently selected based on their locations (in the five zones composing the district), their affiliations (public, private non-profit, or private for profit), and their position in the health system (district hospital, community hospital, faith-based non-profit hospital, health centres, and stand-alone clinics). The numbers of facilities retained in each category were proportionally figured based on their numbers in each zone. ## Study sample and sampling strategy The population in this study was composed of all technical staff categories working in the selected health facilities. We conveniently selected the number and the cadre of participants in each facility. To be included the person must have been in the facility for at least 12 months, and attested to having good knowledge of the programme, the facility and the surrounding community. The person must also have been involved in diabetes-cardiovascular service provision for at least six months in the same facility, and be willing to participate in the study. Study sample size was made of 34 ($81\%$) health facilities out of the 42 across all five zones in the Mangochi district (Fig. 1)., in which 42 healthcare providers (medical assistants, clinical officers, nurse /nurse-midwives technicians, and medical doctors) were interviewed. Fig. 1Map of the Mangochi District’s health facilities [28] We used Hotjar’s free sample size calculator based on the following assumption: total number of health facilities in Mangochi district equals 42. We assumed a confidence level of $95\%$ and a margin of error of $9\%$ to reach a total sample of 32 health facilities [29]. To maximize the inclusion of most senior staff by cadre and get information from all levels in the health system, we purposely included the five hospitals: Mangochi District Hospital and Monkey Bay Community Hospital (both public), and Mulibwanji, St. Martin’s, and Koche Hospitals (all faith-based). Names of other facilities were selected after stratification by level of care in the health system, managing authorities, and geographic locations in the five zones. To account for remoteness and the long history of collaboration with the Kamuzu University of Health Sciences (KUHeS), *Mangochi campus* we purposely added Makanjira and Lungwena Health Centres. At each facility, the in-charge (or his representatives) was systematically the first targeted, unless they declined to participate. By the size of hospitals we included three participants from the Mangochi District Hospital, and two participants in each community or faith-based hospital giving a total of 11 staff. We further decided, during data collection, to complement insufficient information collected, during in depth semi-structured interviews, by including a second participant in four health centres (Namwera, Nankumba, Katuli, and Chilipa.), in all a total of 42 participants. ## Data collection: tool and procedure We collected both qualitative and quantitative data using a checklist and interview guide, both drawn from constructs of the Service Availability and Readiness Assessment (SARA) framework [30]. The SARA framework is a composite framework assessing the five following domains: basic amenities, basic equipment, standard precautions, laboratory capacity, and essential medicines. Each domain further contains specific variable numbers of tracer items; some described in the Table 1 below: basic amenities (7 items), basic equipment (7 items), basic standard precautions (13 items), laboratory capacity (12 items), and essential medicine availability (14 essential medicines) [30]. This research used the locally validated SARA that led to a framework with four domains: training past two years and available copy of guidelines; equipment and diagnosis capacity; availability of essential medicines; and community activities, education and counseling. Prior to its use during the proposal write-up, researchers made some adaptations to the SARA framework. The adapted SARA was finally reviewed through a consultative meeting, attended by a team of 13 senior clinical and nursing members of staff at Mangochi District Hospital, to account for content validity, following researchers’ adaptation. A copy of the tool was distributed to each participant 24 h before the meeting. Individually, each person noted in the margin a few items which were not fitting, firstly, the system in Malawi and secondly, the level of health system where the tool would be administered. Thirteen persons were selected through convenience and intentional sampling. The participants were masters ($$n = 3$$) and bachelor’s ($$n = 11$$) holders in medicine ($$n = 1$$), nursing ($$n = 6$$), clinical medicine [4], and dentistry ($$n = 1$$), anthropology [1]. Their mean work experience was 8 years (SD: 3.1). The research adopted Escobar-Pérez’s Criteria to ascertain the content validation [31]. The validation included following components: sufficiency (The items within the same domain suffice to measure this domain); clarity (the domain can be understood easily; syntax and semantic are appropriate); coherence (items logically related to the domain or indicate what it is measuring); and relevance (items are essential, important and must be included) [32]. Discussions between all experts; including the research team led to some revisions, resulting in a tool fitting the Malawian context with few variations in items between the different levels of the district health system (district hospital, community hospitals (or faith-based hospitals), health centres and private clinics). Changes were the results of agreement on a specific point of this framework adopted by at least $75\%$ of participants. Data collection consisted of administration of the questionnaire (checklist) for quantitative part, a semi-structured interview, and direct observations of items in the facility, for qualitative component. Both qualitative and quantitative data collection approaches three targeted medical assistants, clinical officers, nurses, and medical officers as participants. The quantitative data collection part covered the whole sample size (34 facilities) and was used to assess the availability and readiness of health facilities to provide NCD care [30]. The qualitative data collection was done in 24 health facilities to obtain the perceptions of participants in keys matters concerning readiness for diabetes and/or cardiovascular disease care in Mangochi. The assistant researcher first collected quantitative data from the in-charge (or his representative). The checklist collected information on the demographic data and training of staff as well as presence or absence of guidelines, basic equipment relevant to diabetes and cardiovascular care, diagnostic services, essential medicines, and community services; education and counseling. The qualitative component used a semi-structured interview with a staff (or two to three in hospitals), depending on their eligibility and availability. This interview took place in a corner chosen by the staff, in between consultations or during the lunch time, using an interview guide. The guide drawn from SARA explored perceptions about access to care (transport, distance, affordability in paying facilities), sensed quality of care received, noticed burden of diabetes and cardiovascular diseases in the catchment area (and as a result staff’s perceived workload thereof), health-seeking behaviour, and perception on the functionality of the referral system specific to diabetic or cardiovascular complications. The interview was administered by the principal investigator, digitally recorded, and lasted an average of 32 min. Despite having 42 participants, the saturation point whereby new ideas stopped emerging from the interviews was reached after 24 interviews. The assistant researcher and the senior clinical officer went through a three hour orientation on qualitative and quantitative data collection, interview facilitation, keeping diaries during interviews and coding. The orientation was conducted by the principal investigator and two pilot cases were conducted in the nearest health centre and private clinic. The two sites included were not part of subsequent sites, but results assisted to reveal a few areas for improvement. Observations were conducted by both the assistant researcher and the senior Clinical Officer. under a facility’s staff direction just after the survey and the semi-structured interview. To align the data collection with the SARA spirit guiding the study, each instrument or piece of equipment mentioned, was reviewed to cross-check its current physical presence, numbers, functionality state, and its closeness to the department where it had to be used. Triangulation: concurrently, data from qualitative components (semi-structured interview) were validated by the data collected through the checklist. At the end, a physical verification was conducted to complete the data collection phases. ## Data management and analysis Completed questionnaires from the survey were double entered in an Excel spreadsheet by a clerk and then checked by the principal investigator to catch possible errors. The calculation of readiness was adopted from the approach previously described in Zambia [33] [16]. Facilities’ readiness was defined along the four domains of SARA mentioned above. For each domain, an index score, equivalent to the mean score of items expressed as the percentage of facilities containing all items assessed in a domain [30], was defined. For example, there were 13 equipment items on the survey, and if a facility had 5 functioning equipment items, the basic equipment index for that facility was calculated as 5*$\frac{100}{13}$ = $38.5\%$. The facility readiness index was then calculated as the average of the domain’s indices [30]. We adopted an agreed cut-off of $70\%$ from the same Zambia study [33] on account of proximity and some similar cultural, historical, and health backgrounds. Using this cut-off, a facility with an index below $70\%$ was considered not ready to manage diabetes/cardiovascular diseases. Descriptive analysis for the quantitative data used SPSS for Windows (version 19.0). The transcription of interviews was done verbatim, after listening several times to the recordings, by a professional data clerk. Analysis of transcripts was done manually according to steps of thematic content analysis. The following steps were conducted: familiarization with the first three manuscripts (listening to the recordings, reading several times transcripts, and extraction of repetitive ideas); construction of thematic framework (codes were grouped according to ideas referenced and to the SARA framework’s constructs); coding of all the 24 manuscripts using the framework; charting (elements of one code put together pasted on a blank page); and finally, mapping and interpretation (use of the chart to interpret different themes, reflect on the possible association, and compare and contrast the different themes). Findings from the checklist were compared to the emerging qualitative findings for triangulation’s sake. The data analysis was conducted separately by the principal investigator; the other two were involved following coding for validation of codes. ## Ethical considerations Ethical approval was granted by the College of Medicine Research and Ethics Committee (COMREC reference # P$\frac{.04}{21}$/3312 on June 16th, 2021). Authorization to conduct the study was obtained from the research committee of the Mangochi District Assembly through the Mangochi district’s office of Director of Health and Social Services. Furthermore, data were collected anonymously after a written informed consent by each participant. Privacy was ensured through removal of personal identifiers from data forms just after their collection. To reduce the risk of a participant’s identification, the quotes reported in the findings did not mention the zone in which the participant was working nor his facility. Each participant received 10 US dollars to compensate his time spent for this research. ## Readiness scores are specific to services for diabetes and CVDs education and counseling Compared to our cut-off points only private facilities scored enough to be considered ready to provide care with diagnosis capacity and equipment above our cut-off point of $70\%$. Overall, the four domains’ scores of training/copies of guidelines, diagnosis capacities and equipment, essential medicines, and community activities-education and counseling were low: 5. Total scores were, for diabetes and cardiovascular diseases, respectively in the following domains: 9($26.5\%$) vs. 11($32.4\%$) for training staff and availability of copies of guidelines; 22($63.7\%$) vs. 23($66.2\%$) for diagnostic capacity and equipment; 17($33.5\%$) vs. 14($41.9\%$) for essential medicines; and 13($37.5\%$) vs. 14($42.5\%$) for community activities, education and counseling. Table 2Assessment of facilities’ readiness for diabetes and cardiovascular diseasesDiabetesCardiovascular diseasesAffiliation of facilitiesPublicsn = 22Privatesn = 12Totaln = 34Publicsn = 22Privatesn = 12TotalN = 34Facilities with:n (%)n (%)n (%)n (%)n (%)n (%)Trained staff in diabetes-cardiovascular diseases the past two years and availability of copies of guidelinesTrained staff in the past two years3(13.6)1(8.3)4(11.8)2(9.1)3[25]5(14.7)Diabetes/CVDs copy of guidelines7($31.8\%$)7($58.3\%$)14($41.2\%$)9($40.9\%$)8($66.7\%$)17($50\%$)Domain score training/copy of guidelines5($22.7\%$)4($33.3\%$)9($26.5\%$)6($25\%$)6($45.9\%$)11($32.4\%$)Diagnosis capacity and equipmentBlood glucose9($40.9\%$)8($66.7\%$)17($50\%$)9($40.9\%$)9($75\%$)18($52.9\%$)Urine dipsticks protein4($18.2\%$)6($50\%$)10($29.4\%$)5($22.7\%$)7($58.3\%$)12($35.3\%$)Urine dipsticks ketones3($13.6\%$)6($50\%$)9($26.5\%$)4($18.2\%$)6($50\%$)10($29.4\%$)BP digital machine/sphygmomanometer18(81.8)12($100\%$)30($88.2\%$)18($81.8\%$)12($100\%$)30($88.2\%$)Stethoscope---21($95.5\%$)12($100\%$)33($97.1\%$)Adult scale21($95.5\%$)12($100\%$)33($97.1\%$)21($95.5\%$)11($91.7\%$)32($94.1\%$)Glucometer19($86.4\%$)12($100\%$)31($91.2\%$)---Diagnosis and equipment domain score12($56.1\%$)9($77.8\%$)22($63.7\%$)13($59.1\%$)10($79.2\%$)23($66.2\%$)Essential medicinesCalcium channel blockers‡---2($9.1\%$)10($83.3\%$)12($35.3\%$)Beta-blockers‡‡---14($63.6\%$)9($75\%$)23($67.7\%$)Angiotensin-converting enzymes (ACE)†---3($13.6\%$)6($50\%$)9($26.5\%$)Adrenergic alpha-2 receptor agonists††---4($18.2\%$)4($33.3\%$)8($23.5\%$)Diuretics ‡‡‡---11($50\%$)12($100\%$)23($67.7\%$)Vasodilators¥---4($18.2\%$)5($41.7\%$)9($26.5\%$)Antiplatelets¥¥---15($68.2\%$)12($100\%$27($79.4\%$Lipid-lowering agents¥¥¥---1($4.6\%$)2($16.7\%$)3($8.8\%$)Biguanides †††7($58.3\%$)4($18.2\%$)11($32.4\%$)---Sulfonylurea††††6($50\%$)6($27.3\%$)12($35.3\%$)---Soluble insulin3($25\%$)1($4.6\%$)4($11.8\%$)---IV Glucose solution $50\%$11($91.7\%$)12($54.6\%$)23($67.7\%$)---IV Glucose solution $5\%$12[100]22[100]34($100\%$)---Mean medicines’ domain score8($46.7\%$)9($40.9\%$)17($33.5\%$)6($30.7\%$)8($62.5\%$)14($41.9\%$)Facilities with community services and education-counseling for diabetes and cardiovascular diseasesSchedule/roster of counseling2($9.1\%$)3.0($25.0\%$)5($14.7\%$)3($13.6\%$)6($50\%$)9($26.5\%$)At least one Trained staffx2 years2($9.1\%$)2.0($16.7\%$)4($11.8\%$)3($13.6\%$)3($25\%$)6($17.7\%$)Education materials on modifiable risk factors¥1($4.6\%$)1.0($8.3\%$)2($5.9\%$)1($4.6\%$)0($0.0\%$)1($2.9\%$)Education/counseling sessions on risk behaviours ‡18($81.8\%$)9.0($75.0\%$)27($79.4\%$)21($95.5\%$)10($83.3\%$)31($91.2\%$)Education for self-administration of insulin7($31.8\%$)6.0($50.0\%$)13($38.2\%$)---Education sessions on drugs16($72.7\%$)10($83.3\%$)26($76.5\%$)16($72.7\%$)10($83.3\%$)26($76.5\%$)Education on self-management diabetes or CVDs17($77.3\%$)8($66.7\%$)25($73.5\%$)18($81.8\%$)10($83.3\%$)28($82.4\%$)With community activities service0($0.0\%$)0($0.0\%$)0($0.0\%$)0($0.0\%$)0($0.0\%$)0($0.0\%$)Readiness score index of education-counseling & CA8($35.8\%$)5($40.6\%$)13($37.5\%$)9($35.2\%$)6($46.4\%$)14($42.5\%$)Readiness score diabetes and CVDs services8($48.4\%$)7($40.1\%$)15($40.3\%$)9($37.5\%$)8(58.5)16($45.8\%$)Notes: Mangochi DH, Mangochi District Hospital; CHAM, Christian Health Association in Malawi; SD, standard deviation; n: number of facilities with all items within a specific domain in place; IQR, interquartile range; ‡, amlodipine and nifedipine; ‡‡, atenolol and propranolol; †, enalapril, captopril; ††, alpha-methyl alpha methyldopa dopamine; ‡‡‡, hydrochlorothiazide, furosemide, and spironolactone; ¥, dopamine; ¥¥, aspirin; and ¥¥¥: statin,*, CVDs; IQR, interquartile range; †††: metformin; †, and †††: glibenclamide, CA: community activities 2. Healthcare providers’ perceptions concerning the readiness of the facilities to provide diabetes and cardiovascular care. Qualitative data is represented in diverse themes, further grouped into three: health facility, patient, and community factors. [1] Health facility factors comprise the following themes: lack and/or shortage of drugs and supplies, lack of knowledge by health workers, and deficient education and counseling services for patients. [ 2] Patient factors are low use of NCD services and poor health-seeking behaviour. Finally [3] community factors include absence of community activities and transport policy and costs. Low use of NCDs services can be summarised “deficiencies”. ## Deficiencies /shortage of drugs and supplies and staff A healthcare provider reported his experience of shortage of drugs at the Mangochi district hospital despite its role in supplying the whole district. This shortage is compounded by the high cost of same in private facilities:“Our mother facility is Mangochi District Hospital. (…) Most [patients with NCDs] (…). The concern (…): Mangochi doesn’t have the capacity to stock the NCDs’ medications throughout (…) but it’s a government facility where they [medications] can be given for free. While, ours (CHAM) is a paying institution. ” [ Clinical Officer, CHAM, facility NCDs coordinator] He also emphasized the need for a Service Level Agreement (SLA) to address the drug shortages in CHAM facilities for poor patients in remote catchment areas:[…] some may even come here, but if they don’t have money, it’s a challenge...” [Clinical officer, CHAM, facility NCDs coordinator]. “ *That is* why, all along I have been lobbying. Let’s put the NCDs on SLA (Service Level Agreement) so that the patients can benefit (…)” [Clinical officer, CHAM, facility NCDs coordinator] Provision of diabetes/CVD care in Mangochi was reported as being compromised, in terms of quality of care, by the drug shortages and stock-outs. He went on to express his general observations. “We have many patients with hypertension but access to management, very good management, is very poor because of inadequate resources [drugs and suppliers, mainly].I can say that (…)”. [ Medical assistant, government facility,] Elaborating on the same, he added:“Uhm… we don’t have enough medications. Usually, we don’t have the glucostix, sometimes we don’t have a glucometer, functional weighing scale [with batteries], BP machines…. So, we can have the supplies, but not consistently…” [Registered nurse, coordinator care, government facility] Commenting on human resources shortage:“Very few staff is attached to this clinic! Once the few are out for either supervision or training/orientation; these patients are suffering. Very challenging for us to get people who can take over (…).” [ CHAM, Medical officer, Clinical Officer, Namwera Zone] ## Lack of knowledge by health workers Lack of a guideline’s copy in the facilities and of training the past two years in diabetes and cardiovascular diseases compromised level of knowledge of participants. Several participants complained about lack of in-service training opportunities in diabetes. Many reported that their practice was based on the knowledge acquired during their pre-service education and training. They further stated that the insufficient knowledge was compounded by the lack of a formal copy of guidelines for diabetes and/or CVDs diagnosis and management in their respective facilities. One participant expressed himself this way:“We are very unfortunate (…), with out-dated knowledge. The little we are applying when practising is what we got during our days at the college. Worse again, the copy of guidelines are not even available. [ Medical Assistant, Public Facility, Namwera Zone] They started, instead, to use some pocket books to compensate for this lack of a copy of the guidelines, but these were still inappropriate to fill the gap:We must refer to some of the handbooks we were using while interns like “The Blue Book”. Unfortunately, the pages devoted to diabetes, even hypertension, are very limited in these books. Also, the format also doesn’t allow quick reference; unlike the designated copy of guidelines…” [Medical Assistant, Public Facility, Namwera Zone] Another study participant reported on the lack of trained care providers:(…) of course, we have challenges… mostly it is knowledge. At our facility, there is not even a single person who has undergone specific training focusing on hypertension and diabetes. (…). Things have changed! [ New evidence emerging] Participants felt that their lack of knowledge was even impacting patients’ awareness since non-knowledgeable health workers have little to offer to patients in terms of education.“… they can acquire knowledge from us health workers. If I don’t have the knowledge (...) what can I transmit to patients? …” [ Medical Assistant Namwera zone, CHAM Health Centre] ## Deficient education and counseling services for patients Several participants said that they do counseling from time to time to sensitize their patients living with diabetes and/or CVDs. Others just hand out posters/leaflets (when available) to patients so that they can read, if able on their own. Participants mentioned lack of time, shortage of staff, lack of supporting materials, silent progression of CVDs, and long distances from their homes as key causes of low or absent education/counseling in Mangochi on lifestyle risk factors. One participant noticed that education campaigns and counseling sessions are selectively targeting some specific conditions such as coronavirus or HIV but leaving patients living with diabetes and CVDs without such benefits due to their slow and silent disease progression. He also talked about the lack of interest from donors to support the programme financially.“… They are focusing on coronavirus. (…) But these diseases are long time diseases, hypertension, and diabetes; but (…), nothing on the ground for people [in terms of education-counseling] [nurse, public health centre]“We don’t do[education]…. No poster, no pamphlets, not even a printout on education. Also, most of them [patients] come from far and we are very few staff at the clinic…, we cannot keep them with us for a full morning or day.” [ Nurse, Monkey Bay Zone, Public]“(…) education is not always a routine; we need support materials like posters, and pamphlets which can only be provided through NCD directorate within the Ministry of Health. However, NCDs so far don’t attract donor’s attention (…) [Government, Clinical Officer, Monkey Bay zone] ## Low use of services Almost all healthcare providers acknowledged poor access to care in all health facilities across Mangochi. This low use was due to patients’ loss of trust in a health system characterized by recurrent and frequent stock-outs of drugs and supplies; financial burden due to non-inclusion of NCD care in the service level agreement in private non-profit facilities (CHAM); low awareness of the diseases in the general public (diabetes and CVDs); fear of getting infected in facilities by the on-going coronavirus pandemic and finally, long distances and the high transport fare to reach both public and private facilities. One participant from a community hospital said the following about the level of awareness and distances:“Number one challenge for use as a facility is one, awareness (meaning, low awareness level); two, distances; they are living in remote areas, very far from this hospital” [clinical officer, CHAM, facility NCD’s coordinator] ## Poor health-seeking behaviour Participants expressed mixed and sometime contrasting opinions regarding health-seeking behaviour in NCD clinics. Though late presentation emerged as a predominant feature from most participants, three of them recognized that, in general, health-seeking was early in their catchments. Patients come late to facilities for diverse reasons. The main factors given for late presentation, by patients living with diabetes and cardiovascular diseases, to health facilities were patients’ over-reliance on herbal medicines prior any health facility visit, the cultural norm of waiting for a family decision on whether to seek care or not, loss of trust in the health system, low disease awareness level among patients (and significant others), non-respect for follow-up appointment dates (for those already in care), long distance between facility and patient’s home, and high cost of either transport fare or drugs for those living close to CHAM facilities. A delayed date of appointment interfering with adherence to medications was noted by an NCD care provider in a community hospital:“(…) most of the clients if you give them the appointment they don’t come (…). If you tell them, for example, come on the tenth of August, they may come maybe, next month as they don’t even understand the importance of being kept on drugs throughout. [ meaning in October] …” [Community nurse, provider, Monkey Bay zone].“why to rush here if they will be given a prescription to buy drugs in town […]; and those drugs are almost out of stock for the past six to eight months now…. I don’t trust any more government hospitals” [Medical assistant, provider, Mangochi Central Zone] ## Absence of community activities Regarding services in the community (despite recognising their relevance in diabetes and cardiovascular management), almost all participants noted the limited community activities focusing on diabetes and cardiovascular issues in their respective catchment areas. A participant observed that:“Besides low level of updated knowledge in health providers, community outreaches are history with the current crisis. The past two years and a half, three (…), we have not been able to go into the community to talk about health issues. Therefore, don’t expect patients to change lifestyle, knowledge on drugs and behaviour with few minutes talk during clinic’s days” (Government, Medical Assistant, Makanjira Zone, Medical Assistant) ## Transport policy and costs Mixed views characterised the transport policy for patients living with diabetes or cardiovascular diseases: ## Unclear referral policy Participants observed that, in general, referrals from peripheral health facilities and when needed, ambulances, were free to patients. However, patients’ access to ambulances was not always straight forward, and in some cases considered ambiguous, if not impossible. Ambulance access has a user-fee in CHAM facilities. Other participants observed that these challenges vary from one healthcare provider calling the ambulance to another. One participant noted that it is just a matter of communicating [with the district]:“No challenges in referrals. We used to call the transport officer if we have a patient, where they come and pick the patient to the DHO [meaning district hospital].” [ Medical assistant, Government, Mangochi Boma Zone] Others raised issues related to variations in challenges, with ambulance for referrals, experienced by healthcare providers in remote health facilities depend on: (a) The type of diseases:“You know…, this transport policy is not fully known by some of us. I wonder if the transportation of patients with diabetes or CVDs is really stated there [in policy] (…). You can call an ambulance in the morning for these diseases; they will not show up, even after 24 hours. But call for an ambulance in a case of even a simple incomplete abortion, in one to two hours the patient will be picked-up, (…)” [Medical assistant, Government, Mangochi Boma Zone] or, (b) In terms of type of facility calling for ambulance: CHAM facilities take ownership of referrals of these diseases since transport policy is not included in the on-going Government-CHAM agreement when dealing with patients with emergencies in diabetes or cardiovascular diseases. “NCDs patients [from CHAM facilities] are not eligible for transportation using government ambulances as the agreement [between CHAM facilities and government] is only applied in mothers and children health...” [Medical Officer, CHAM, faith based Hospital, Namwera Zone] Similarly, another medical assistant from a government facility was concerned about the lack of clarity in the application of the policy regarding the ambulance transportation of patients with diabetes/CVDs in districts. He felt the timely-response problem was more based on the type of condition for which the ambulance was called rather than on the ownership (private or public) of the facility:“Others felt that NCDs are not in the mainstream of referral policy in the government as there is still resistance to pick a patient with diabetes/hypertension irrespective of the condition he is in, even for us in the government system. Unless a pregnancy is associated with the emergency condition (…)” [Medical assistant, government, Namwera Zone] Unlike in public hospitals, a user fee is attached to ambulances for referral in CHAM or private for-profit facilities, a major limitation to access for the majority of poor patients. “We have an ambulance here [private for profit]. hum (…); but, due to these financial problems (financial hardship the country is going through), now they are required to pay two thousand Malawi Kwacha (2.5 USD), yes...” [Private facility, nurse-midwife, Mangochi Boma Zone] ## Discussion This study aimed to assess providers’ perceptions regarding the readiness of their facilities to provide diabetes and CVD care in Mangochi. Overall the findings revealed a low level of readiness, below the set threshold in the provision of care for patients with diabetes and CVDs, in different areas studied (human resources, copies of guidelines, diagnosis capacity, essential medicines, and equipment, education and counseling, and community services). The discussion will be around these points: [1] Health facility factors (absence of trained staff and guidelines, low diagnosis capacity and equipment, low supply of essential medicines, and low community education and counseling); [2] Patient factors (low use of NCD services, and poor health-seeking behaviour); and [3] Community factors (absence of community activities, and transport policy and costs). Overall facility factors showed deficiencies in trained staff, copies of guidelines, diagnosis capacity and equipment, l supply of essential medicines, and community activities, education and counseling for diabetes and cardiovascular diseases. Similar results were reported in previous studies [25, 33–36]. Knowledge of providers was compromised by lack of on-job trainings) and absence of copies of guidelines in diabetes and cardiovascular disease clinics in diverse facilities. The same situation was similar to other LMICs: at least one trained staff and copies of guidelines were found in: $1.3\%$ and $1.4\%$ facilities in Nepal [33]; $9\%$ and $42\%$ for guidelines of hypertension in Tanzania outpatients primary care in 2018 [32] and later in 2020; $10.4\%$ and $33.2\%$, respectively [36] against (11.8 and $41.2\%$, respectively), in the present study. More recently, the prevalence of guideline copies in SSA was found below the global average [37]. Diabetes guidelines in general were available in a few sub-Saharan African countries, namely: South Africa [38], Mozambique [39] and Cameroon [40]. More investment will be needed to respond to this rise of diabetes and cardiovascular diseases cases in Mangochi, Malawi in general to increase numbers of trained staff and supply enough copies of guidelines in all facilities. The capacity-building of healthcare providers must be a priority intervention in health systems strengthening. Approaches used for capacitation of healthcare providers, in general, in Low and Middle-Income Countries (LMICs), emphasize training and task-shifting among health workers [16]. Both task-shifting and training have yielded, in diabetes for example to increased diagnosis capacity and adherence to management, early screening, and reduction of uncontrolled diabetes, reduction of inpatient cases with acute metabolic complications, sustained decreases of glycosylated haemoglobin [41], and detection and referral of poorly controlled cases [42]. In several sites, staffs were using pockets books to read to document themselves on the two diseases; somehow helpful. Guidelines are a must in primary health care and their absence creates a handicap to functionality of most peripheral health facilities. However, to fully partially fulfil this role they must not only focus on their availability, but also their usability, applicability, utility [43]; but also, adhesion from users, and wide dissemination and implementation at all levels of a health care system, including primary health care [44–46]. A dissemination of copies of guidelines in these investigated remote health facilities is very critical as it can affect positively the management of patients given the basic low education level of practitioners working in peripheral facilities) [22]. In fact, guidelines are a tool that uphold quality of care, align practice to current evidence and minimises frustrations of providers when they are dealing with borderline or complex cases [22]. Thus, in the future, subsequent studies must go beyond a presence/absence assessment of these guidelines, to investigate their actual use, cost-effectiveness, context-specific roles, implementation, relevance, dissemination, and appropriateness. Unfortunately, low domain readiness for diagnostic capacity was found low in several facilities. Low diagnostic capacity may come from diverse causes depending on the context. The simple urine dipsticks was only in 9 and 10 health facilities, representing less than thirds of facilities visited. While the pure lab test can be out of order, we could expect the cheapest used to detect early complications of diabetes for example to be in stock. Even though being the highest of the domains studied in this study, albeit being combined with equipment, the joint domain equipment-diagnosis capacity is still stand below the set cut-off point of $70\%$ (67.8, $66.2\%$ for diabetes and cardiovascular diseases, respectively). This low diagnostic capacity was reported in research conducted in Malawi [12, 20, 21] and elsewhere in Africa [18, 22, 23, 25]. Low diagnosis capacity domains have been also reported for diabetes and cardiovascular diseases, respectively: in Nepal [mean domain index: 9.0 (± SD 24.3)], 16.6 (± SD 30.0) [25], in Zambia ($2\%$) [33], and in a multicounty study conducted in Bangladesh, Haiti, Kenya, Malawi, Namibia, Nepal, Rwanda, Senegal, Uganda and the United Republic of Tanzania [34]. High reliance on diagnosis as a source of money able to sustain the business in private can be explained by a tight competition imposed by a free fees-for-service mode of payment in public sector and tight regulations in accreditations of services which can generate additional revenues through procedures in private practice such as operating theatre in Malawi. Furthermore, in agreement with the index domains of the diagnostic capacity in the present study in public versus private facilities (diabetes: 56.1 vs. $97.8\%$, cardiovascular diseases: 59.7 vs. $79.2\%$; respectively), Tanzania’s public facilities showed lower figures of diagnosis capacity index domains compared to privates [36]. This can translates low availability of supplies in public facilities, higher socioeconomic level of patients using private’s facilities and therefore increasing demand for test, and non-consistent supplies of reagents and machines in public facilities. This low diagnosis capacity index domain can explain partially the global highest rates of undiagnosed diabetes being reported in Africa ($62\%$), including Malawi [7]., Scarcity of medicines followed the same trends, including essential medicines such as insulin in primary care as reported previously [8, 12, 26]. For example, despite the severity of diabetes and its possible life-threatening complications in case resulting from inadequate treatment, only $32.4\%$, $35.3\%$, and $11.8\%$ have in stock biguanides, sulfonylureas, and insulin respectively, out of the 34 visited. The low means’ domain score index for medicines was as well been found in several other places in LMICs: 5.4 (± SD 15.5) in Nepal [25], 33.3 (± SD 15.5) in Zambia [33], $2\%$ in Bangladesh, Haiti, Kenya, Malawi, Namibia, Nepal, Rwanda, Senegal, Uganda and the United Republic of Tanzania [34]. Again, this availability of essential medicines was more acute in public rather than privates for cardiovascular diseases in the current study as previously reported ($30.7\%$ public and $62.5\%$ in privates). Essential antidiabetic drugs showed reverse trends with high availability in publics (50.3 versus $18.2\%$ for biguanides and 50 vs. 27.5 for sulfonylurea. This trends can just explain the lay severity perception of diabetics which push patients to seek care from public facilities rather than privates, the geographic distribution of diabetes (more in urban than rural) where faith-based facilities are mostly located, the high numbers of small facilities in remote areas that cannot manage diabetes, and finally, the complexity of the diabetes management. More likely also, this can translate an unbalance in trainings’ opportunities which are targeting more providers from the public sector. If we consider the fact that diabetes cases are mostly treated in district, community or faith-based hospitals; unlike hypertension in which initiation of patients to treatment and monitoring is much easier and cheaper; someone can easily understand these differences. The low domain readiness score in essential medicines was due to erratic supplies, but also to affordability (mostly in private non-profit (CHAM) facilities where care is accessed at a cost. Patients are unable to pay due to financial hardship. However, despite having slightly higher readiness scores of medication and of diagnosis capacity domains, private faith-based facilities (compared to public ones (Table 2), and being closer geographically to needy patients, there was not improvement in access in these facilities. Financial barrier to access faith-based private facilities among needy patients in remote areas is likely the cause. Out-of-pocket payment for care as applied in these facilities has been well-documented as one of the deterrent factors of readiness of diabetes services [27]. On the other hand, although existence of a free services policy at the point of care in the public sector in Malawi, the access to facilities remains lower more likely due to frequent stock-outs, general poor quality of care due to deficiencies in diverse areas of care and slow roll-outs of the programme in remote areas due to insufficient funding. An urgent support in procurement, supply chains, and funding seem critical in minimising the issue of availability of drugs, diagnostic capacity and supplies in Malawi public facilities. Furthermore, non-affordability of services in faith-based facilities; main service providers in remote areas where $80\%$ of the population are living (generally poor) must be part of the discussions between stakeholders in the field to address the inequity. To this end, the government of Malawi has to design mechanisms (or leverage existing ones) to increase access to care of these rural people to diabetes-cardiovascular diseases’ services in such areas. For example, Service Level Agreement has shown potential in maternal and child health for more than a decade in Malawi. The SLA is defined as: “A formal agreement between the Government of Malawi (GOM, represented by a District or City Council and a CHAM health facility where the latter provides an agreed package of health services, free of charge, to the population in its catchment area, and is compensated by the former on the basis of a reimbursement mechanism jointly agreed upon with the GOM upon entering the partnership agreement” [47] Page 7. This arrangement could alleviate the above shortfalls, and increase access to diabetes and CVDs care for poor individuals. Provision of this care in rural areas, with focus on the most vulnerable, will increase access and cultural appropriateness of care, and reduce transport costs. In maternal and child health in Malawi, the SLA has increased collaboration between the public and private facilities, ensured equitable access and good quality of care, and built capacity of health workers [47]. The project of implementing the SLA to noncommunicable diseases has been delayed, pending a consensus on some points of the agreement between the government and CHAM. However, there is an urgent need for a speedy approval and effective implementation of the SLA in diabetes and CVD cares. Providers recognised that many patients with diabetes/CVDs are not even given the currently recommended management. The group-counseling at the diabetes-cardiovascular clinics for education in behaviour changes regarding the drugs, lifestyle, and complications was rarely available. They evoked several causes such as time constraints, shortage of staff, lack of supporting materials, silent progression of the diseases, and long distances as possible reasons. A list of factors impeding the conduct of behaviour change interventions in the context of primary/secondary care have been previously reported in the literature [48–51]. Planning implementation of behaviour change approach as an alternative (or complement) to group counseling has to consider these factors at an early stage in order to increase the likelihood of success; and ipso-factor increase the readiness to care in these facilities. Community services were almost absent in this study in diabetes versus cardiovascular diseases apart from self-management of diabetes ($73.5\%$ public vs. $82.1\%$ private); for education on drugs $76.5\%$ vs. $76.5\%$ in diabetes and CVD; and 79.4 vs. $91.2\%$ for diabetes and cardiovascular’s risk factors education, respectively. This is a good observations were good in both public and privates, despite being better in privates. However, the quality again remains unascertained and could may be tell us more about the really services provided. For example, education is vast and can take any form. Assessment without analysing the content, process in this case presents some limits. Future explorations have to look further, for example on the content, the providers, the types of education, and even the areas of focus. Other parameters assessed (staff’s rosters for education ($14\%$ vs. $26.5\%$), educational materials for patients ($5.9\%$ vs. $2.9\%$) although higher than results found in the past national study in Malawi ($0\%$) [22], and community activities in education/counseling on diabetes and cardiovascular diseases ($0\%$). Learning from private-public partnership can provide lessons based on the same community within the region and even within Malawi regading community role in diabetes-cardiovascular fight. This is a big concern as community approach is a critical component in the management of these chronic diseases in general. Zero activity was going on in the community regarding diabetes and cardiovascular preventive measures. May be community staffs were not up to standards to provide such a service. Exploring community health volunteers’ perceptions of their functions, tasks, and fulfilment, a collaborative study was conducted in Lilongwe (Malawi) and Zambia. The study found that community health worker in NCD, in general, can play a critical role in screening, monitoring, and linking patients to the health system [52] in NCD in general. More specifically, the study cited the role of these workers in health care and prevention (lifestyle counseling), monitoring of NCD, management, documentation, and screening [52]. The same experience can be replicated countrywide as health surveillance assistants are well established in each catchment areas, with some experience in provision of community work in similar programmes. However, this approach may require additional funding which the programme doesn’t have for now. While waiting for funding to launch community education and counseling in Mangochi; facility-based, individual, preventive, educative measures in the form of brief behaviour-change advice embedded in routine care and supplementing the on-going group counseling can be explored. If well implemented, evidence has shown that brief behaviour change is a cost effective [53], locally accepted [54], and convenient intervention for primary care [44, 45, 55, 56]. Furthermore, brief behaviour change channelled through approached such as motivational interviewing and the 5As approaches, yields better outcomes [57]. Absence of transport policy and its high cost are impeding the smooth referral of patients. Despite being free, as per ministry of health policy, unpaid transport for referrals is limited in CHAM facilities to cases included in the SLA, such as maternal and child care. Others are subjected to local arrangements or patients’ out-of-pocket costs. However, in some instances, even in the public sectors, participants question the lack of clarity in referring emergencies related to diabetes/cardiovascular diseases, in the speed to pick up the patients, and in the selective nature of which ones based on the type of disease (with low priority transportation for diabetes or CVDs). This as well calls for harmonization in policies, clear communication from the district health office to facilities, and mutual effort to support transportation for those coming from far or even to provide real transportation which could also alleviate the problems. Here again the SLA can play a critical role in increasing access to care. From the patient perspective, the segregation between patients based on the nature of disease, with appropriate drugs out of stock in peripheral facilities doesn’t add-up concerning proper management of diabetes/cardiovascular diseases in the peripheral health facilities. This study must be interpreted in light of several strengths and weaknesses: *This is* an observational study based on a small sample size. Therefore, the findings from this study cannot be generalised to other districts. However, results can generate hypotheses which could guide future large studies on a big scale. We relied on self-reported data which is prone to subjective reporting with risk of desirability bias. Nonetheless, the confirmation of findings with data collected through direct observation validated the responses from the checklist. SARA was validated before being used in Mangochi to fit the local setup. Furthermore, SARA has been previously used elsewhere [58, 59], adapted by our research team, This study has validated these findings. Nonetheless, despite being small the present study expanded into additional areas which were not or partially explored in a prior, more recent, national study [22]. More specifically, beyond diabetes, the current study also added the cardiovascular diseases. Furthermore, it explored additional domains of diabetes and cardiovascular care such as patient education and counseling in non-modifiable risk factors, in adherence to treatment, and in self-management of diabetes, along with community activities in diabetes and cardiovascular disease management. The study assessed availability of different items and services; however, presence cannot automatically mean quality of care provided. Unfortunately, the quality of care was not assessed in the present research. This readiness presented here reflects the situation of a the visit’s day. The findings could give different findings (good or bad) for some facilities in case the team modified the visit’s date. However, the consistency in findings across different visited sites is obvious. Moreover, there are some similarities of the current findings with those of a previous similar study in Malawi and even elsewhere in Africa. These two reasons increase the likelihood of findings reflecting the real state on the ground. ## Conclusion This study found low readiness levels of facilities in terms of staff, copies of guidelines, diagnostic capacity, equipment, medicines, and counseling materials, and other community activities. It demonstrated the need for capacitation of staff, dissemination of copy of guidelines, linkages between the community and facilities as well as implementation of a clear, facility- and evidence-based model of education and counseling. Key results have been produced, which can guide both the district council and the ministry of health in addressing critical issues raised, thus improving the quality of care provided to patients living with diabetes and/or cardiovascular disease. ## References 1. O’Donnell MJ, Xavier D, Liu L, Zhang H, Chin SL, Rao-Melacini P. **Risk factors for ischaemic and intracerebral haemorrhagic stroke in 22 countries (the INTERSTROKE study): a case-control study**. *Lancet* (2010.0) **376** 112-23. DOI: 10.1016/S0140-6736(10)60834-3 2. Moucheraud C. **Service Readiness for Noncommunicable Diseases was low in five countries in 2013–15**. *Health Aff (Millwood)* (2018.0) **37** 1321-30. DOI: 10.1377/hlthaff.2018.0151 3. Engelgau MM, Rosenthal JP, Newsome BJ, Price L, Belis DMG. **Noncommunicable Diseases in low- and Middle-Income Countries**. *Glob Heart* (2018.0) **13** 131-7. DOI: 10.1016/j.gheart.2018.05.001 4. Abbafati C, Abbas KM, Abbasi-Kangevari M, Abd-Allah F, Abdelalim A, Abdollahi M. **Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the global burden of Disease Study 2019**. *Lancet* (2020.0) **396** 1204-22. DOI: 10.1016/S0140-6736(20)30925-9 5. Afshin A, Sur PJ, Fay KA, Cornaby L, Ferrara G, Salama JS. **Health effects of dietary risks in 195 countries, 1990–2017: a systematic analysis for the global burden of Disease Study 2017**. *Lancet* (2019.0) **393** 1958-72. DOI: 10.1016/S0140-6736(19)30041-8 6. 6.World Health Organization. Universal Health coverage factsheet [Internet]. [cited 2023 Jan 25]. Available from: https://www.who.int/news-room/fact-sheets/detail/universal-health-coverage-(uhc) 7. 7.Matthew Gowshall SD, Taylor-Robinson. The increasing prevalence of non-communicable diseases in low-middle income countries: the view from Malawi.Int J Gen Med. 2018 Jun28; 11:255–264. 8. Hu F, Malik V. **Sugar-sweetened beverages and risk of obesity and type 2 diabetes: epidemiologic evidence**. *Physiol Behav* (2010.0) **100** 47-54. DOI: 10.1016/j.physbeh.2010.01.036 9. Malik VS, Popkin BM, Bray GA, Després JP, Willett WC, Hu FB. **Sugar-Sweetened Beverages and risk of metabolic syndrome and type 2 diabetes: a meta-analysis**. *Diabetes Care* (2010.0) **33** 2477-83. DOI: 10.2337/dc10-1079 10. 10.WHO. Global Action Plan for the Prevention and Control of Noncommunicable Diseases. 2013–2020 [Internet]. Geneva, Switzerland: World Health Organization; 2013 Nov [cited 2022 Jan 23] p. 103. (Noncommunicable diseases and Mental Health). Report No.: WT 500. Available from: http://www.who.int/nmh/publications/ncd_action_plan/en/. 11. Msyamboza KP, Ngwira B, Dzowela T, Mvula C, Kathyola D, Harries AD, Bowie C. **The burden of selected chronic non-communicable diseases and their risk factors in Malawi: nationwide STEPS survey**. *PLoS ONE* (2011.0) **6** e20316. DOI: 10.1371/journal.pone.0020316 12. Alemayehu Amberbir, Sabrina H, Lin J, Berman A, Muula D, Jacoby E. **Systematic review of hypertension and diabetes burden, risk factors, and interventions for Prevention and Control in Malawi: the NCD BRITE Consortium**. *Glob Heart* (2019.0) **14** 109-18. DOI: 10.1016/j.gheart.2019.05.001 13. 13.Ng’ambi W, Mwase T, Chinkhumba J, Udedi M, Chigaru F, Banda JC, Nkhoma D, Mfutso-Bengo J. Prevalence of non-communicable diseases risk factors and their determinants in Malawi: Evidence from 2017 WHO STEPwise Survey. medRxiv. 2022:2022-08. 14. 14.Government of the Republic of Malawi, Ministry of Health. Health Sector Strategic Plan II, 2017–2022: Towards Universal Health Coverage [Internet]. 201AD [cited 2022 Aug 13]. Available from: http://www.nationalplanningcycles.org/sites/default/files/planning_cycle_repository/malawi/health_sector_strategic_plan_ii_030417_smt_dps.pdf 15. 15.Government of the. Republic of Malawi, Ministry of Health. Malawi Essential Health Package; 2011. 16. Cundale K, Wroe E, Matanje-Mwagomba BL, Muula AS, Gupta N, Berman J. **Reframing noncommunicable diseases and injuries for the poorest Malawians: the Malawi national NCDI poverty commission**. *Malawi Med J* (2017.0) **29** 194-7. DOI: 10.4314/mmj.v29i2.22 17. Hanson K, Brikci N, Erlangga D, Alebachew A, De Allegri M, Balabanova D. **The Lancet Global Health Commission on financing primary health care: putting people at the centre**. *Lancet Glob Health* (2022.0) **10** e715-72. DOI: 10.1016/S2214-109X(22)00005-5 18. Starfield B, Shi LMJ. **Contribution of primary care to health systems and health**. *Milbank Q* (2005.0) **83** 457-502. DOI: 10.1111/j.1468-0009.2005.00409.x 19. Lawn JE, Rohde J, Rifkin S, Were M, Paul VKCM. **Alma-Ata 30 years on: revolutionary, relevant, and time to revitalise**. *Lancet* (2008.0) **372** 917-27. DOI: 10.1016/S0140-6736(08)61402-6 20. 20.World Health OrganizationThe World Health Report 2008: primary Health Care (now more Than ever)2008GenevaWHO. *The World Health Report 2008: primary Health Care (now more Than ever)* (2008.0) 21. Assayed AA, Muula AS, Nyirenda MJ. **The quality of care of diabetic patients in rural Malawi: a case of Mangochi district**. *Malawi Med J* (2014.0) **26** 109-14. PMID: 26167259 22. Chikowe I, Mwapasa V, Kengne AP. **Analysis of rural health centres preparedness for the management of diabetic patients in Malawi**. *BMC Res Notes* (2018.0) **11** 267. DOI: 10.1186/s13104-018-3369-7 23. Pfaff C, Malamula G, Kamowatimwa G, Theu J, Allain TJ, Amberbir A. **Decentralising diabetes care from hospitals to primary health care centres in Malawi**. *Malawi Med J* (2021.0) **33** 159-68. DOI: 10.4314/mmj.v33i3.3 24. 24.Health System Readiness [Internet]. [cited 2023 Jan 23]. Available from: https://www.healthsystemreadiness.com/ 25. Ghimire U, Shrestha N, Adhikari B, Mehata S, Pokharel Y, Mishra SR. **Health system’s readiness to provide cardiovascular, diabetes and chronic respiratory disease-related services in Nepal: analysis using 2015 health facility survey**. *BMC Public Health* (2020.0) **20** 1163. DOI: 10.1186/s12889-020-09279-z 26. Byambaa AJT, Altangerel E, Batbaatar S, Saw YM, Kariya T, Yamamoto E. **Free and universal access to primary healthcare in Mongolia: the service availability and readiness assessment**. *BMC Health Serv Res* (2019.0) **19** 129. DOI: 10.1186/s12913-019-3932-5 27. Biswas T, Haider MM, Gupta R, Das, Uddin J. **Assessing the readiness of health facilities for diabetes and cardiovascular services in Bangladesh: a cross-sectional survey**. *BMJ Open* (2018.0) **8** e022817. DOI: 10.1136/bmjopen-2018-022817 28. 28.Citypopulation.de. Mangochi district in Malawi [Internet]. 2020 [cited 2021 Jun 1]. Available from: https://www.citypopulation.de/en/malawi/admin/southern/MW301__mangochi/ 29. 29.Hotjar. Sample size calculator [Internet]. Hotjar’s free sample size calculator. Available from: https://www.hotjar.com/poll-survey-sample-size-calculator/ 30. 30.World Health Organization. (2015). Service availability and readiness assessment (SARA): an annual monitoring system for service delivery: implementation guide, version 2.2. World Health Organization. https://apps.who.int/iris/handle/10665/183119 (accessed March 1, 2022] 31. Escobar-Pérez J, Cuervo-Martínez A. **Validez de contenido y juicio de expertos: Una aproximación a su utilización**. *Av Med* (2008.0) **6** 27-36 32. Fernández-Gómez E, Martín-Salvador A, Luque-Vara T, Sánchez-Ojeda MA, Navarro-Prado S, Enrique-Mirón C. **Content validation through Expert Judgement of an instrument on the nutritional knowledge, beliefs, and Habits of pregnant women**. *Nutrients* (2020.0) **12** 1136. DOI: 10.3390/nu12041136 33. Mutale W, Bosomprah S, Shankalala P, Mweemba O, Chilengi R, Kapambwe S. **Assessing capacity and readiness to manage NCDs in primary care setting: gaps and opportunities based on adapted WHO PEN tool in Zambia**. *PLoS ONE* (2018.0) **13** e0200994. DOI: 10.1371/journal.pone.0200994 34. Leslie HH, Spiegelman D, Zhou X, Kruk ME. **Service readiness of health facilities in Bangladesh, Haiti, Kenya, Malawi, Namibia, Nepal, Rwanda, Senegal, Uganda and the United Republic of Tanzania**. *Bull World Health Organ* (2017.0) **95** 738-48. DOI: 10.2471/BLT.17.191916 35. Bintabara D, Mpondo BCT. **Preparedness of lower-level health facilities and the associated factors for the outpatient primary care of hypertension: evidence from Tanzanian national survey**. *PLoS ONE* (2018.0) **13** e0192942. DOI: 10.1371/journal.pone.0192942 36. Bintabara D, Ngajilo D. **Readiness of health facilities for the outpatient management of non-communicable diseases in a low-resource setting: an example from a facility-based cross-sectional survey in Tanzania**. *BMJ Open* (2020.0) **10** e040908. DOI: 10.1136/bmjopen-2020-040908 37. 37.Nuche-Berenguer B, Kupfer L. Readiness of Sub-Saharan Africa Healthcare Systems for the New Pandemic, Diabetes: A Systematic Review. J Diabetes Res. 2018; 2018: 3419290. 38. Gill GV, Price C, Shandu D, Dedicoat M, Wilkinson D. **An effective system of nurse-led diabetes care in rural Africa**. *Diabet Med Diabet Med* (2008.0) **25** 606-11. DOI: 10.1111/j.1464-5491.2008.02421.x 39. Beran D, Silva Matos CYSJ. **The diabetes UK Mozambique twinning programme. Results of improvements in diabetes care in Mozambique: a reassessment 6 years later using the rapid assessment protocol for insulin access**. *Diabet Med* (2010.0) **27** 855-61. DOI: 10.1111/j.1464-5491.2010.03053.x 40. Labhardt ND, Balo JR, Ndam M, Manga ESB. **Improved retention rates with low-cost interventions in hypertension and diabetes management in a rural african environment of nurse-led care: a cluster-randomised trial**. *Trop Med Int Health* (2011.0) **16** 1276-84. DOI: 10.1111/j.1365-3156.2011.02827.x 41. Distiller LA, Brown MA, Joffe MA, Kramer BI. **Striving for the impossible dream: a community-based multi-practice collaborative model of diabetes management**. *Diabet Med* (2010.0) **27** 197-202. DOI: 10.1111/j.1464-5491.2009.02907.x 42. Katz I, Schneider H, Shezi H. **Managing type 2 diabetes in Soweto-the south african chronic disease outreach program experience**. *Prim Care Diabetes* (2009.0) **3** 157-64. DOI: 10.1016/j.pcd.2009.06.007 43. Gagliardi AR, Brouwers MC, Palda VA, Lemieux-Charles LGJ. **How can we improve guideline use? A conceptual framework of implementability**. *Implement Sci* (2011.0) **6** 26. DOI: 10.1186/1748-5908-6-26 44. Katende D, Mutungi G, Baisley K, Biraro S, Ikoona E, Peck R. **Readiness of ugandan health services for the management of outpatients with chronic diseases**. *Trop Med Int Health* (2015.0) **20** 1385-95. DOI: 10.1111/tmi.12560 45. Kratzer J. **Structural barriers to coping with type 1 diabetes mellitus in Ghana: experiences of diabetic youth and their families**. *Ghana Med J* (2012.0) **46** 39-45. PMID: 23661816 46. Mayega RW, Guwatudde D, Makumbi FE, Nelson FENFrederick, Peterson Stefan TG. **Comparison of fasting plasma glucose and haemoglobin A1c point-of_care tests in screening for diabetes and abnormal glucose regulation in a rural low income setting**. *Diabetes Res Clin Pract* (2014.0) **104** 112-20. DOI: 10.1016/j.diabres.2013.12.030 47. 47.Malawi Ministry of Health Christian Health Association of Malawi. Service Level Agreement Guidelines 2016. 2016. (Pages 7, 10) 48. Keyworth C, Epton T, Goldthorpe J, Calam R, Armitage CJ. **It’s difficult, I think it’s complicated’: Health care professionals’ barriers and enablers to providing opportunistic behaviour change interventions during routine medical consultations**. *Br J Health Psychol* (2019.0) **24** 571-92. PMID: 30977291 49. Luxton NA, MacKenzie R, Shih P. **Smoking Cessation Care in cardiothoracic surgery: a qualitative study exploring the views of australian clinicians**. *Heart Lung and Circulation* (2019.0) **28** 1246-52. DOI: 10.1016/j.hlc.2018.04.293 50. Ndejjo R, Musinguzi G, Nuwaha F, Bastiaens H, Wanyenze RK. **Understanding factors influencing uptake of healthy lifestyle practices among adults following a community cardiovascular disease prevention programme in Mukono and Buikwe districts in Uganda: a qualitative study**. *PLoS ONE* (2022.0) **17** e0263867. DOI: 10.1371/journal.pone.0263867 51. Wills J, Kelly M. **What works to encourage student nurses to adopt healthier lifestyles? Findings from an intervention study**. *Nurse Educ Today* (2017.0) **48** 180-4. DOI: 10.1016/j.nedt.2016.10.011 52. Safary E, Mwandeti M, Matanje B, Beiersmann C, Mtaita C, Shiroya V. **Role of community health volunteers in identifying people with elevated blood pressure for diagnosis and monitoring of hypertension in Malawi: a qualitative study**. *BMC Cardiovasc Disord* (2021.0) **21** 361. DOI: 10.1186/s12872-021-02171-7 53. 53.Everett-Murphy K, Mash B, Malan Z. Helping people change. The busy healthcare practitioners guide to providing brief behaviour change counselling on non-communicable diseases (NCD) lifestyle risk factors. Cape Town (SA): Chronic Diseases Initiative for Africa (CDIA); 2014. p.60. 54. Lutala P, Muula A. **Brief behaviour change counselling in non-communicable diseases in Mangochi, Southern Malawi: a hypothetical acceptability study**. *Pilot Feasibility Stud* (2022.0) **8** 69. DOI: 10.1186/s40814-022-01032-0 55. Parker WA, Steyn NP, Levitt NS, Lombard CJ. **They think they know but do they? Misalignment of perceptions of lifestyle modification knowledge among health professionals**. *Public Health Nutr* (2011.0) **14** 1429-38. DOI: 10.1017/S1368980009993272 56. Michie S. **Designing and implementing behaviour change interventions to improve population health**. *J Health Serv Res Policy* (2008.0) **13** 64-9. DOI: 10.1258/jhsrp.2008.008014 57. 57.Abo E, Hamza, Ahmed A, Moustafa. Motivational interviewing for the treatment of addiction. In Moustafa A.A. (edit). Cognitive, Clinical, and Neural Aspects of Drugs Addiction. U.K.: Academic Press. 1st edition; 2020, p. 290–313. 58. 58.Manne-Goehler J, Geldsetzer P, Agoudavi K, Andall-Brereton G, Aryal KK, Bicaba BW et al. Health system performance for people with diabetes in 28 low- and middle-income countries: A cross-sectional study of nationally representative surveys. Wareham NJ, editor. PLOS Med. 2019; 16(3):e1002751. 59. 59.World Health Organization. Service availability and readiness assessment (SARA): An annual monitoring system for service delivery. WHO, Geneva. Switzerland; 2015. p. 55. (Version 2.2). Switzerland: WHO; 2015. P.55. (version 2.2).
--- title: Associations of meeting 24-h movement behavior guidelines with cognitive difficulty and social relationships in children and adolescents with attention deficit/hyperactive disorder authors: - Alyx Taylor - Chuidan Kong - Zhihao Zhang - Fabian Herold - Sebastian Ludyga - Sean Healy - Markus Gerber - Boris Cheval - Matthew Pontifex - Arthur F. Kramer - Sitong Chen - Yanjie Zhang - Notger G. Müller - Mark S. Tremblay - Liye Zou journal: Child and Adolescent Psychiatry and Mental Health year: 2023 pmcid: PMC10042421 doi: 10.1186/s13034-023-00588-w license: CC BY 4.0 --- # Associations of meeting 24-h movement behavior guidelines with cognitive difficulty and social relationships in children and adolescents with attention deficit/hyperactive disorder ## Abstract ### Background Evidence-based 24-h movement behavior (24-HMB) guidelines have been developed to integrate recommendations for the time spent on physical activity, sedentary behavior, and sleep. For children and adolescents, these 24-HMB guidelines recommend a maximum of two hours of recreational screen time (as part of sedentary behavior), a minimum of 60 min per day of moderate to vigorous physical activity (MVPA), and an age-appropriate sleep duration (9–11 h for 5 to 13-year-olds; 8–10 h for 14 to 17-year-olds). Although adherence to the guidelines has been associated with positive health outcomes, the effects of adhering to the 24-HMB recommendations have not been fully examined in children and adolescents with attention eficit/hyperactive disorder (ADHD). Therefore, this study examined potential associations between meeting the 24-HMB guidelines and indicators of cognitive and social difficulties in children and adolescents with ADHD. ### Methods Cross-sectional data on 3470 children and adolescents with ADHD aged between 6 and 17 years was extracted from the National Survey for Children’s Health (NSCH 2020). Adherence to 24-HMB guidelines comprised screen time, physical activity, and sleep. ADHD-related outcomes included four indicators; one relating to cognitive difficulties (i.e., serious difficulties in concentrating, remembering, or making decisions) and three indicators of social difficulties (i.e., difficulties in making or keeping friends, bullying others, being bullied). Logistic regression was performed to determine the associations between adherence to 24-HMB guidelines and the cognitive and social outcomes described above, while adjusting for confounders. ### Results In total, $44.8\%$ of participants met at least one movement behavior guideline, while only $5.7\%$ met all three. Adjusted logistic regressions further showed that meeting all three guidelines was associated with lower odds of cognitive difficulties in relation to none of the guidelines, but the strongest model included only screen time and physical activity as predictors (OR = 0.26, $95\%$ CI 0.12–0.53, $p \leq .001$). For social relationships, meeting all three guidelines was associated with lower odds of difficulty keeping friends (OR = 0.46, $95\%$ CI 0.21–0.97, $$p \leq .04$$) in relation to none of the guidelines. Meeting the guideline for screen time was associated with lower odds of being bullied (OR = 0.61, $95\%$ CI 0.39–0.97, $$p \leq .04$$) in relation to none of the guidelines. While screen time only, sleep only and the combination of both were associated with lower odds of bullying others, sleep alone was the strongest predictor (OR = 0.44, $95\%$ CI 0.26–0.76, $$p \leq .003$$) in relation to none of the guidelines. ### Conclusion Meeting 24-HMB guidelines was associated with reduced likelihood of cognitive and social difficulties in children and adolescents with ADHD. These findings highlight the importance of adhering to healthy lifestyle behaviors as outlined in the 24-HMB recommendations with regard to cognitive and social difficulties in children and adolescents with ADHD. These results need to be confirmed by longitudinal and interventional studies with a large sample size. ## Highlights Meeting the combination of all three 24-HMB guidelines, or a combination of screen time and physical activity, was associated with reduced odds of serious difficulties in concentrating, remembering, or making decisions. Meeting the combination of all three 24-HMB guidelines was associated with reduced odds of difficulties making and keeping friends. Meeting the 24-HMB guideline for screen time was associated with reduced odds of being bullied. Meeting the individual guidelines for screen time or sleep duration or a combination of both was associated with reduced odds of bullying others. ## Introduction Attention deficit/hyperactive disorder [1] is a common neurodevelopmental disorder that affects both children and adults and is characterized by deficits in the domains of attention and hyperactivity- impulsivity [1–4]. Approximately 5 to $7\%$ of children and adolescents are diagnosed with ADHD worldwide [5, 6] and an additional $5\%$ who exhibit symptoms that do not reach diagnostic level [7]. The symptoms of ADHD can have a range of negative consequences. In children and adolescents, the symptoms can reduce the quality of their social interactions, academic and learning activities [8], hindering cognitive development [9], and educational achievement [10, 11]. In addition, there is an increased risk of other mental and physical health conditions including anxiety [12], depression [13], and obesity [14]. When symptoms of ADHD persist into adulthood, difficulties with social life, employment [15], or law enforcement [16] can emerge. Pharmacological treatments, primarily stimulant medication [17], behavioral therapy, and parent training are standard interventions for ADHD [18]. However, approximately $25\%$ of ADHD patients do not respond to stimulant medication and some are unable to tolerate the side effects [19]. Likewise, for some individuals with ADHD, the behavioral interventions are sufficient to manage their symptoms, while others do not seem to benefit in the same way [1, 2, 20]. Therefore, other non-pharmacological interventions have been investigated, including cognitive training (e.g., working memory training) [21, 22] and regular physical activity [23]. There is mounting evidence that physical activity is beneficial for behavior and cognitive performance of children and adolescents with ADHD [24, 25, 48]. In addition to physical activity, other lifestyle factors such as sleep and sedentary behavior may also be related to the symptomatology associated with ADHD [26–29]. For example, in the general population, lower levels of sedentary behaviors, mainly recreational screen time [30], and optimal sleep duration [31] have been positively and independently related to academic achievement in children and adolescents [32]. There is also evidence that these three movement behaviors (i.e., physical activity, sedentary behavior, and sleep) are codependent and thus should be examined simultaneously [33–35]. Together the above-mentioned evidence indicates that integrating non-pharmacological or lifestyle-related interventions aiming to positively influence sedentary behavior, sleep duration and the level of regular physical activity may benefit children and adolescents with ADHD. The 24-hour movement behavior (24-HMB) guidelines for children and adolescents may serve as the basis for considering multi-behavioral interventions that integrate physical activity, sedentary behavior, and sleep [36–42]. Specifically, the guidelines recommend a limit of two hours of non-educational screen time, and a minimum of sixty minutes of moderate-to-vigorous physical activity (MVPA) daily, as well as age-appropriate sleep duration each night. Previous research has shown that children and adolescents with ADHD are typically less active than their peers in the general population [43–45]. For example, Friel et al. [ 2020] reported, based on national survey on children and adolescents aged 6–17 years in the US, that the majority ($91.2\%$) met at least one of the 24-HMB guidelines, but only $8.8\%$ met all three 24-HMB recommendations [46]. In comparison, Wang et al. [ 2022] using the data of the 2018-19 NSCH (National Survey of Children's Health dataset) survey, observed that less than half ($46.8\%$) of the children and adolescents with ADHD (aged between 6 and 17 years) met at least one guideline and that $6.5\%$ met all three 24-HMB recommendations [47]. In summary, children and adolescents with ADHD are less likely to adhere to the 24-HMB recommendations than their peers without ADHD. While positive associations of meeting the 24-HMB guidelines in relation to social interaction and cognitive function have been found in the general population [49–51], these have not been examined specifically in children and adolescents with ADHD. Furthermore, as academic performance is closely related to social interaction and cognitive function, it could be particularly helpful for the young people with ADHD to adhere to the 24-HMB guidelines. Based on the current the literature, we hypothesized that adhering to all three components of the 24-HMB guidelines would be associated with a reduced likelihood of cognitive and social difficulties in children and adolescents with ADHD, when controlling for demographic, socioeconomic and other medical factors. Therefore, this study examined the associations between meeting the 24-HMB guidelines for physical activity, sleep duration and screen time, and measures of cognitive and social difficulties in children and adolescents with ADHD. Results of this study will help health professionals and school administrators deepen an understanding about symptomatic management of children and adolescents with ADHD. ## Study design and data source In this cross-sectional study, we used data from the United States’ 2020 NSCH survey that was collected from June 2020 to January 2021. The survey provides data on health-related measures of the children, their families, and communities, the prevalence of disease type, and associated healthcare needs. In the 2020 NSCH survey, approximately 240,000 households from 50 states and the District of Columbia were invited to complete the screening questionnaire, which selects households with children. Respondents, who are primary caregivers of the child in the household, were invited to participate. Of those who responded, 51,107 were eligible to continue to the data collection and 42,777 provided complete responses. The ethical approval and process for obtaining consent to participate are summarized below under the subheading Declaration, and given in full detail in the 2020 NSCH methodology report [52]. ## Participants and procedure For the current study, data were retrieved on children with a diagnosis of ADHD between the ages of 6 and 17 years and their families from the NSCH 2020 database. The inclusion criteria for the current study were positive responses by the parent or guardian of the child to the following survey items: (i) “*Has a* doctor or other health care provider ever told you that this child has attention deficit/hyperactivity disorder, that is, ADD or ADHD?”; ( ii) “If yes, does the child currently have the condition?” *These criteria* were used to select 3740 children with ADHD and their families from the database. The survey requires the respondents to report on only one child, even when more than one child has the same diagnosis. The demographic data and variables for the exposure of interest and outcomes were then selected for further analysis. ## Demographic and medical data The demographic data selected included the child’s age, sex, ethnicity, preterm birth status, overweight status, ADHD-related medication and behavioral treatment, household poverty level, and the highest level of education of primary caregivers (Table 1). According to the US federal poverty level, family income was coded to one of two levels. The variable called overweight was collected by asking the caregivers to report whether or not the child had been identified by their doctor or another healthcare worker as being overweight. Table 1Participant characteristicsADHD ($$n = 3470$$)CharacteristicsValue1 Age (M, SD)11.97 (3.48)Sex (n, %) Male2376 (69.55) Female1094 (30.45)Ethnicity (n, %) White2839 (69.18) Black or African American250 (17.29) American Indian/Alaska native26 (0.59) Asian55 (1.07) Native Hawaiian & other pacific islander13 (0.86) Two or more ethnic groups287 (11.01)Born 3 weeks or more before due dates (n, %) Yes511 (14.47) No2959 (85.53)*Overweight status* (n, %) Yes469 (14.95) No3001 (85.05)ADHD severity (n, %) Mild1527 (43.20) Moderate1566 (42.29) Severe377 (14.52)ADHD related medication or treatment (n, %) Behavioural treatment only493 (14.39) Medication only1008 (26.18) Behavioural treatment and medication1047 (33.15) Neither905 (26.27)Household poverty level (n, %) ≤ 0–$99\%$ federal poverty level457 (18.67) ≥ $100\%$ federal poverty level3013 (81.33)Highest education level among reported adults (n, %) Less than high school86 (7.02) High school540 (23.74) Some college or associated degree879 (24.06) College degree or higher1965 (45.18)Adherence to the 24-h movement guidelines (n, %) None841 (27.56) Screen time314 (10.87) Sleep1118 (26.87) Physical activity124 (4.93) Screen time + Sleep574 (14.69) Screen time + Physical activity116 (4.00) Sleep + Physical activity182 (5.40) All201 (5.68)Serious difficulty concentrating, remembering, or making decisions (n, %) Yes1858 (55.57) No1612 (44.43)Difficulty making or keeping friends (n, %) No difficulty1545 (45.96) A little difficulty1322 (36.54) A lot of difficulty603 (17.49)Bully others (n, %) Never2374 (69.21) 1 to 2 times in the past 1 year762 (19.73) 1 to 2 times per month198 (5.16) 1 to 2 times per week78 (2.73) Almost every day58 (3.17)Being bullied (n, %) Never1363 (40.85) 1 to 2 times in the past 1 year1190 (33.67) 1 to 2 times per month452 (11.15) 1 to 2 times per week270 (7.40) Almost every day195 (6.93)Values are mean (SD) or n (%). N represents unweighted sample counts and % is weighted to the US population ## Exposure of interest: meeting the 24-HMB guidelines The exposure of interest was meeting the 24-HMB guidelines [41]. These consist of three recommendations: Sufficient sleep for the age group (9–11 h for 5- to 13-year-olds; 8-10 hours for 14- to 17-year-olds); a minimum of 60 min per day of MVPA; and no more than 2 h per day of recreational screen-time. For MVPA, the survey question was: “During the past week, on how many days did this child exercise, play a sport, or participate in physical activity for at least 60 min?” In the current study, the answer of 7 days was considered as meeting the 24-HMB guidelines while lower levels (i.e., 6 or less days) was considered as not meeting the 24-HMB recommendations concerning physical activity. The question related to sleep was: “During the past week, how many hours of sleep did this child get on most weeknights?” With regard to each age range, the minimum number of hours was used to gauge whether the children/adolescents met the sleep recommendation or not. Screen-time was evaluated using the question: “On most weekdays, about how much time did this child spend in front of a TV, computer, cell phone, or other electronic device watching programs, playing games, accessing the internet, or using social media? ( Do not include time spent doing schoolwork.)”. Responses of 2 h or less were considered as meeting the screen time recommendation, while any other responses were classified as not meeting the 24-HMB guidelines. In this study, the number of guidelines that each child met (0 to 3), was used as a continuous variable for further statistical analyses. In addition, four combinations of whether the 24-HMB guidelines were met or not were used as separate variables in follow-up analyses: each pair (physical activity and screen-time; physical activity and sleep; screen-time and sleep) and meeting all guidelines. ## Outcomes: cognitive and social difficulties Cognitive difficulties were evaluated by the question “Does the child have serious difficulty concentrating, remembering, or making decisions?” The binary response options were yes or no. In addition, social difficulties were evaluated using three questions: (i) “Does the child have difficulty in making friends or keeping friends?” The response had three levels, with the options: no, a little, or a lot. ( ii) “During the past 12 months, how often did this child bully others, pick on them, or exclude them?” ( iii) “During the past 12 months, how often was this child bullied, picked on, or excluded by other children?” For the second and third questions, the response options had five levels: never, 1 or 2 times per year, 1 or 2 times per month, 1 or 2 times per week, and almost every day. ## Confounders The potential confounders included in the statistical analyses were age, sex, ethnicity whether or not the child had been born 3 or more weeks earlier than their due date, severity level of ADHD symptoms (mild, moderate, and severe), and ADHD-related medication and/or treatment (neither, behavioral treatment only, medication only, behavioral treatment and medication). ## Statistical analysis Descriptive statistics were calculated for all variables. Continuous variables were described with means and standard deviations, and categorical variables were described using unweighted sample counts and weighted percentages. Multiple logistic regression was used to estimate the odds ratios (with $95\%$ confidence intervals) between meeting 24-HMB guidelines and its component recommendations and four outcomes, including one indicator of cognitive difficulty (i.e., serious difficulties in concentrating, remembering, or making decisions) and three indicators of social difficulties (i.e., difficulties in making or keeping friends, bullying others, being bullied). Separate analyses were carried out, first for the number of 24-HMB guidelines met (continuous variable) and then for specific combinations (physical activity, sleep duration, screening time, screen time + sleep, screen time + physical activity, sleep + physical activity, and physical activity +screen time + sleep) of guideline recommendations (categorical variables) as independent variables in the models. Socio-demographic and medical data (age, sex, ethnicity, preterm birth status, ADHD medication, ADHD behavioral treatment, household poverty level (federal poverty level, FPL), and the highest level of education of the parents/legal guardian of the child) were included as potential confounders. For all statistical analyses, the significance level was set at $p \leq 0.05.$ The statistical analyses were conducted using Stata, version Stata/SE 15.1 (StataCorp LLC., College Station, TX, USA). ## Sample characteristics This study included 3470 children and adolescents with ADHD aged 6–17 years, from the 42,777 households in the USA that provided full data. The children and adolescents had a mean age of 11.97 ± 3.48 years. The distribution of ethnicity in the sample was $69.18\%$ white, $17.29\%$ black or African American, and $11.01\%$ with two or more ethnic groups, while other ethnic groups were smaller (see Table 1). The symptoms of ADHD experienced by the participants were reported as being mild ($43.20\%$), moderate ($42.29\%$), and severe ($14.52\%$). Approximately a quarter of the sample was not receiving treatment at the time of the study ($26.27\%$), a similar proportion received medication only ($26.18\%$), approximately a third of the sample received a combination of both medication and behavioral therapy ($33.15\%$), and a smaller group received behavioral treatment only ($14.39\%$). The household poverty level and the education status of the adults are also presented in Table 1. ## Meeting the 24-HMB guidelines Within our sample, just over a quarter ($27.6\%$) met none of the 24-HMB guidelines, $44.8\%$ met only one guideline, of which sleep was the most common ($26.9\%$) and physical activity the least common ($4.9\%$). Only 201 ($5.7\%$) of the sample met all three 24-HMB guidelines (see Table 1 and Fig 1). The sub-groups meeting one or more of the 24-HMB guidelines are represented in the Venn diagram (Fig 1).Fig. 1Venn diagram showing proportions of participants meeting 24-h movement guidelines. Values are n (%). N represents unweighted sample counts and % is weighted sample sizes. PA physical activity, ST screen time, SL sleep ## Socio-demographic data and medical information Associations between the socio-demographic data and the outcome measures were tested. When the sample was divided into two groups by age, the older group (i.e., aged 14–17 years) were significantly less bullied than the younger group (i.e., aged 6–13 years) (OR = 0.52, $95\%$ CI 0.40–0.68, $p \leq 0.001$). When the data was separated by sex, the odds of being bullied were significantly higher for females than for males (OR = 1.39, $95\%$ CI 1.08–1.78, $$p \leq 0.01$$) and the odds of difficulties in concentrating, remembering, or making decisions were significantly higher for females (OR = 1.41, $95\%$ CI 1.06–1.87, $$p \leq 0.02$$). Concerning ethnicity, the white sub-group was used as the reference group and only the native Hawaiian and other Pacific Islander group were observed to be at higher odds for difficulties in concentrating, remembering, or making decisions (OR = 9.68, $95\%$ CI 1.37–68.41, $$p \leq 0.02$$) and higher odds of difficulties in making or keeping friends (OR = 2.88, $95\%$ CI 1.03–8.08, $$p \leq 0.04$$) (Fig. 2).Fig. 2Associations of meeting 24-h movement behavior guidelines with cognitive function and social relationships among children and adolescents with ADHD Being overweight was associated with higher odds for difficulties in making or keeping friends (OR = 1.48, $95\%$ CI 1.09–2.01, $$p \leq 0.01$$), but no other outcome. With regard to the severity of symptoms of ADHD, the children and adolescents with mild symptoms were used as reference group. Those with moderate symptoms had significantly higher odds of all measured outcomes related cognitive difficulties and social relationships, and those with severe symptoms were associated with the highest odds for each outcome (see Tables 2, 3, 4, 5). Concerning treatment, there was a statistically significant association found between combined behavioral treatment and medication, and higher odds of difficulties in concentrating, remembering, or making decisions (OR = 1.58, $95\%$ CI 1.06–2.37, $$p \leq 0.03$$). Statistically significant associations also occurred between behavioral treatment only and higher odds of all three social relationship outcomes, difficulty making or keeping friends, being bullied, and bullying others (OR = 1.95, $95\%$ CI 1.28–2.98, $$p \leq 0.002$$; OR = 1.78, $95\%$ CI 1.20–2.65, $$p \leq 0.01$$; OR = 1.70, $95\%$ CI 1.02–2.85, $$p \leq 0.04$$) respectively. By contrast, there were no statistically significant associations between being born 3 or more weeks early and the cognitive or social relationship outcomes. There were also no statistically significant associations between household poverty level or highest level of education of the adults and the outcomes. Table 2Associations between meeting 24-h movement behavior guidelines and outcomes of interestSerious difficulties concentrating, remembering, or making decisionsOdds ratio ($95\%$ CI)pIntercept0.55 (0.16–1.85)0.34Age 6–13 years (reference)1 (reference) 14–17 years0.89 (0.65–1.22)0.48Sex Male (reference)1 (reference) Female1.41 (1.06–1.87)0.02Overweight No (reference)1 (reference) Yes1.43 (0.92–2.20)0.11 White (reference)1 (reference) Black/African American0.79 (0.47–1.30)0.35 American Indian/ Alaska native1.49 (0.48–4.67)0.49 Asian1.00 (0.47–2.14)0.99 Native Hawaiian and other pacific islander9.68 (1.37–68.41)0.02 Two or more ethnicities0.80 (0.48–1.31)0.37Born 3 weeks or more weeks before due date No (reference)1 (reference) Yes1.09 (0.76–1.55)0.63ADHD severity level Mild (reference)1 (reference) Moderate3.61 (2.69–4.85) < 0.001 Severe12.90 (7.39–22.50) < 0.001ADHD medication & behavioral treatment Neither behavioral treatment nor medication (reference)1 (reference) Behavioral treatment and medication1.58 (1.06–2.37)0.03 Behavioral treatment only1.50 (0.96–2.36)0.08 Medication only0.91 (0.63–1.32)0.64Household poverty level (0–$99\%$ FPL (reference)1 (reference) 100–$400\%$ FPL0.79(0.50–1.25)0.32Highest level of education among reported adults Less than high school (reference)1 (reference) High school (vocational/trade/business school)1.02 (0.36–2.88)0.97 Some college or associate degree1.19 (0.44–3.21)0.73 College degree or higher1.10 (0.42–2.88)0.8524-HMB guidelines met (categorical) None (reference)1 (reference) Screen time only0.84 (0.50–1.42)0.51 Sleep only0.99 (0.66–1.47)0.94 Physical activity only1.27 (0.48–3.28)0.64 Screen time + Sleep0.67 (0.43–1.03)0.07 Screen time + Physical activity0.26 (0.12–0.53) < 0.001 Sleep + Physical activity0.82 (0.42–1.60)0.56 All three0.43 (0.24–0.78)0.01 Prob > F < 0.001FPL Federal Poverty LevelTable 3Associations between all covariates, meeting 24-h movement guidelines and difficulties in making or keeping friendsDifficulties in making or keeping friendsOdds ratio ($95\%$ CI)pAge 6–13 years (reference)1 (reference) 14–17 years1.21 (0.92–1.59)0.17Sex Male (reference)1 (reference) Female1.12 (0.86–1.45)0.42Overweight No (reference)1 (reference) Yes1.48 (1.09–2.01)0.01 White (reference)1 (reference) Black/African American0.73 (0.48–1.09)0.13 American Indian/ Alaska native0.84 (0.16–4.22)0.83 Asian1.19 (0.46–3.11)0.72 Native Hawaiian and other pacific islander2.88 (1.03–8.08)0.04 Two or more ethnicities0.90 (0.56–1.44)0.65Born 3 weeks or more weeks before due date No (reference)1 (reference) Yes1.40 (0.97–2.02)0.07ADHD severity level Mild (reference)1 (reference) Moderate2.62 (2.01–3.42) < 0.001 Severe8.13 (4.96–13.31) < 0.001ADHD medication & behavioral treatment Neither behavioral treatment nor medication (reference)1 (reference) Behavioral treatment and medication1.38 (0.97–1.95)0.08 Behavioral treatment only1.95 (1.28–2.98)0.002 Medication only0.80 (0.58–1.11)0.18Household poverty level 0–$99\%$ FPL (reference)1 (reference) 100–$400\%$ FPL1.08 (0.74–1.58)0.70Highest level of education among reported adults Less than high school (reference)1 (reference) High school (vocational/trade/business school)1.09 (0.52–2.29)0.82 Some college or associate degree1.48 (0.73–3.00)0.28 College degree or higher1.13 (0.56–2.29)0.7324-HMB guidelines met (categorical) None (reference)1 (reference) Screen time only0.96 (0.59–1.55)0.86 Sleep only1.25 (0.90–1.76)0.19 Physical activity only1.24 (0.55–2.82)0.60 Screen time + Sleep1.23 (0.80–1.88)0.35 Screen time + Physical activity0.59 (0.31–1.14)0.11 Sleep + Physical activity0.78 (0.44–1.40)0.41 All three0.46 (0.21–0.97)0.04 Prob > F < 0.001FPL Federal Poverty LevelTable 4Associations between all covariates, meeting 24-h movement guidelines and bullying othersBullying othersOdds ratio ($95\%$ CI)pAge 6–13 years (reference)1 (reference) 14–17 years0.74 (0.51–1.09)0.13Sex Male (reference)1 (reference) Female1.01 (0.76–1.36)0.93Overweight No (reference)1 (reference) Yes1.19 (0.79–1.36)0.40Ethnicity White (reference)1 (reference) Black/African American1.13 (0.66–1.96)0.65 American Indian/Alaska native2.01 (0.81–5.01)0.14 Asian1.04 (0.49–2.21)0.92 Native Hawaiian and other pacific islander3.94 (0.28–55.57)0.31 Two or more ethnicities0.72 (0.46–1.13)0.16Born 3 weeks or more weeks before due date No (reference)1 (reference) Yes1.14 (0.74–1.75)0.56ADHD severity level Mild (reference)1 (reference) Moderate1.69 (1.02–2.37)0.002 Severe3.00 (1.78–5.04) < 0.001ADHD medication & behavioral treatment Neither behavioral treatment nor medication (reference)1 (reference) Behavioral treatment and medication1.26 (0.81–1.95)0.30 Behavioral treatment only1.70 (1.02–2.85)0.04 Medication only0.66 (0.44–0.99)0.05Household poverty level 0–$99\%$ FPL (reference)1 (reference) 100–$400\%$ FPL0.85(0.51–1.43)0.54Highest level of education among reported adults Less than high school (reference)1 (reference) High school (vocational/trade/business school)0.76 (0.28–2.01)0.58 Some college or associate degree1.19 (0.48–2.96)0.71 College degree or higher0.80 (0.31–2.04)0.6424-HMB guidelines met (categorical) None (reference)1 (reference) Screen time only0.44 (0.26–0.76)0.003 Sleep only0.65 (0.44–0.95)0.03 Physical activity only1.05 (0.52–2.10)0.89 Screen time + Sleep0.60 (0.39–0.93)0.02 Screen time + Physical activity0.91 (0.44–1.89)0.80 Sleep + Physical activity0.94 (0.55–1.59)0.82 All three1.16 (0.45–3.01)0.75 Prob > F < 0.001FPL Federal Poverty LevelTable 5Associations between all covariates, meeting 24-h movement guidelines and being bulliedBeing bulliedOdds ratio ($95\%$ CI)pAge 6–13 years (reference)1 (reference) 14–17 years0.52 (0.40–0.68) < 0.001Sex Male (reference)1 (reference) Female1.39 (1.08–1.78)0.01Overweight No (reference)1 (reference) Yes1.45 (0.99–2.10)0.05Ethnicity White (reference)1 (reference) Black/African American0.74 (0.49–1.13)0.17 American Indian/Alaska native1.31 (0.35–4.95)0.69 Asian0.67 (0.31–1.44)0.30 Native Hawaiian and other pacific islander1.38 (0.12–16.41)0.80 Two or more ethnicities0.92 (0.58–1.46)0.73Born 3 weeks or more weeks before due date No (reference)1 (reference) Yes1.29 (0.92–1.81)0.14ADHD severity level Mild (reference)1 (reference) Moderate1.92 (1.49–2.49) < 0.001 Severe5.13 (3.12–8.41) < 0.001ADHD medication & behavioral treatment Neither behavioral treatment nor medication (reference)1 (reference) Behavioral treatment and medication1.12 (0.79–1.58)0.53 Behavioral treatment only1.78 (1.20–2.65)0.01 Medication only0.80 (0.58–1.10)0.17Household poverty level 0–$99\%$ FPL (reference)1 (reference) 100–$400\%$ FPL0.81 (0.54–1.21)0.31Highest level of education among reported adults Less than high school (reference)1 (reference) High school (vocational/trade/business school)0.93 (0.45–1.93)0.84 Some college or associate degree1.67 (0.82–3.42)0.16 College degree or higher1.14 (0.56–2.32)0.7224-HMB guidelines met (categorical) None (reference)1 (reference) Screen time only0.61 (0.39–0.97)0.04 Sleep only1.33 (0.94–1.88)0.11 Physical activity only2.47 (1.12–5.51)0.03 Screen time + Sleep0.89 (0.60–1.33)0.57 Screen time + Physical activity0.65 (0.35–1.19)0.16 Sleep + Physical activity0.88 (0.51–1.51)0.64 All three0.78 (0.40–1.52)0.47 Prob > F < 0.001FPL Federal Poverty Level ## Social difficulties Concerning difficulties in making or keeping friends (Table 3), when meeting specific combinations of the 24-HMB guidelines were compared with meeting none of the 24-HMB guidelines, only those children and adolescents who met all three guidelines had significantly lower odds of difficulties in making or keeping friends (OR = 0.46, $95\%$ CI 0.21–0.97, $$p \leq 0.04$$). In other words, the social relationships were stronger in those children and adolescents who meet all three 24-HMB guidelines than in peers who met less or none of the 24-HMB guidelines. Children and adolescents who only met the 24-HMB guideline for screen time were found to be at lower odds of being bullied (OR = 0.61, $95\%$ CI 0.39–0.97, $$p \leq 0.04$$). In contrast, there was a significant association between meeting the guideline for physical activity only and increased odds of being bullied (OR = 2.47, $95\%$ CI 1.12–5.51, $$p \leq 0.03$$), as shown in Table 4. Regarding bullying others (Table 5), when meeting specific combinations of 24-HMB guidelines were compared with meeting none of the guidelines, screen time only, sleep only, and the combination of screen time and sleep were all associated with significantly lower odds of bullying others ([OR = 0.44, $95\%$ CI 0.26–0.76, $$p \leq 0.003$$], [OR = 0.65, $95\%$ CI 0.44–0.95, $$p \leq 0.03$$], [OR = 0.60, $95\%$ CI 0.39–0.93, $$p \leq 0.02$$], respectively). ## Main findings This cross-sectional study examined, for the first time, the associations between meeting 24-HMB guidelines by children and adolescents with ADHD aged 6-17 years and four outcome measures relating to cognitive and social difficulties in a large sample of data from the U.S. 2020 NSCH. We showed positive associations between meeting all or specific 24-HMB guidelines and a reduced risk for cognitive and social difficulties. Together, these findings suggest that meeting 24-HMB recommendations may reduce the development of cognitive and social difficulties in children and adolescents with ADHD. While almost half of the children and adolescents with ADHD met at least one of the 24-HMB guidelines ($44.8\%$), only a small proportion of them met all three guidelines ($5.7\%$). Despite these data being collected during the COVID-19 pandemic, our results show a comparable pattern to those being observed during the previous NSCH 2018 cycle of the survey, in which $46.8\%$ of children and adolescents with ADHD met at least one 24-HMB guideline and only $6.5\%$ met all three [47]. In this context, it should also be noted that children and adolescents with ADHD are less likely to meet the 24-HMB guidelines than neurotypical children and adolescents in the same age range in which $91.2\%$ met at least one 24-HMB guideline and $8.8\%$ met all three [46]. Given the the evidence for the health benefits of meeting these 24-HMB guidelines, the latter findings stress the need to support children with ADHD and their caregivers to foster their ability to effectively adopt a healthy lifestyle. ## Cognitive difficulties In our study, the analysis of meeting 24-HMB guidelines as a continuous variable showed that as the number of guidelines met increased, there were significantly lower odds for difficulties in concentrating, remembering, or making decisions. On examination of meeting specific combinations of 24-HMB guidelines, the combination of screen time and physical activity was the strongest predictor of reduced difficulties in concentrating, remembering, or making decisions. Meeting all three guidelines was also significantly associated with these measures of cognitive difficulty, but the reduction in the odds was not as pronounced as that for meeting the guidelines for screen time and physical activity combined. Our results for physical activity and cognitive difficulties broadly agree with prior research that provided evidence for positive effects of physical activity on cognitive function in children and adolescents with ADHD [53–55]. However, previous research included many different types of physical activities and duration of physical activity interventions. For example, Benzing, Chang, and Schmidt, [2018] investigated the effects of acute sessions of 15 min of exergaming on 8–12-year- old children with ADHD and reported a post-exercise improvement of inhibition and switching performance [56]. Suarez-Manzano et al. [ 2018] reviewed studies that investigated the effects of physical activity on cognitive performance in children and adolescents with ADHD and concluded that physical activity for a minimum of 30 min, at a minimum intensity of $40\%$ heart rate reserve, undertaken on a minimum of 3 days per week and a minimum of 5 weeks duration improved attention, inhibition, behavior, emotional and motor control [23]. Another systematic literature review examining the effect of physical activity on executive functions including attention, inhibition, task shifting, and working memory in children and adolescents up to age 18 years with ADHD reported the positive effects of habitual physical activity on all executive functions, but only shifting and working memory were statistically significant [57]. In addition, the authors noted that the positive effects on executive function were greater for physical activities that have a lower cognitive load compared to more cognitively demanding physical activities [57]. In line with the findings of the above-mentioned studies and systematic reviews reporting a positive influence of physical activity on cognitive performance, our study provides support for the practical application of the 24-HMB guideline for physical activity for reducing cognitive difficulties in children and adolescents with ADHD. Concerning screen time, a review of 91 studies showed significant associations between longer screen time and higher scores for symptomatology associated with ADHD in children and adolescents [58]. For instance, Suchert et al. [ 2017] examined the specific activities in sedentary behavior of adolescents aged 13–17 years and found that screen time was associated with symptomatology associated with ADHD, while this was not observed for non-screen sedentary activities [58]. The lack of association with non-sedentary behavior might suggest screen time has effects beyond simple sedentary behavior, possibly due to the lack of short physical activity breaks typically observed in non-recreational sedentary behavior [59], or alternatively that recall of time spent on screen time activity is better than recall of general sedentary behavior [60]. Another possible explanation for the effect of screen time arises from evidence that blue light exposure can delay or disturb sleep [61]. In relation to this, Lissak [2018] reported that an intervention-related reduction of screen time improved ADHD-related behavior and sleep duration in children and adolescents [62]. Studies examining cognitive function in children and adults in the general population have also reported changes in the structure of the brain areas responsible for cognitive control and emotional regulation in association with addictive screen time behavior [63, 64]. Moreover, extended screen time is associated with differences in executive control performance, which, in turn, can increase distractibility [65]. Lissak [2018] reported a case- study in which the intervention included reducing screen time and the results showed reduced symptoms of ADHD behavior and improved sleep duration in the youth who also engaged successfully with school work [62]. Taken together, these findings might explain the current results showing that meeting all three 24-HMB guidelines, including sleep, was associated with reduced cognitive difficulties. With regard to meeting the 24-HMB guideline for sleep duration alone, the association between sleep duration and measures of cognitive difficulties did not reach statistical significance in our study. This is perhaps related to bias arising from the parental self-reports. This assumption is supported by the finding that objective measures of sleep duration using accelerometers showed the mean parental estimate is up to 50.5 minutes less than the objective results [66]. Another study using actigraphy showed improved rate of cognitive processing when the sleep period for adolescents with ADHD was extended to 9.5 h compared to 6.5 h [67]. Thus, in future longitudinal studies examining recommendations for sleep duration should utilize objective measures of sleep (e.g., derived by accelerometers) rather than solely relying on subjective measures (e.g. parental reports). ## Serious difficulties in concentrating, remembering, or making decisions The associations between meeting the 24-HMB guidelines and the measure of cognitive difficulties (serious difficulties in concentrating, remembering or making decisions), are presented in Table 2. Our multivariable regression analysis revealed that the number of guidelines met was associated with significantly lower odds for difficulties concentrating, remembering, or making decisions (OR = 0.76, $95\%$ CI 0.64–0.91, $$p \leq 0.002$$). When specific combinations of the 24-HMB guidelines were compared with meeting none of the guidelines, meeting a combination of both the sleep and the physical activity guidelines, or all three guidelines were associated with significantly lower odds of suffering from difficulties in these cognitive abilities (OR = 0.26, $95\%$ CI 0.12–0.53, $p \leq 0.001$ and OR = 0.43, $95\%$ CI 0.24–0.78, $$p \leq 0.01$$, respectively). ## Making and keeping friends In the current study, children and adolescents who met all three 24-HMB guidelines had significantly lower odds of difficulties in making or keeping friends, reflecting better social relationships with peers. Well-developed social skills are important for success in academic [68] and work environments as well as social relationships for all children and adolescents including those with developmental challenges [69]. Children and adolescents with ADHD whose symptoms may include intrusive, impulsive, or aggressive behavior, can experience barriers to successful social interactions [60]. Such social relationship difficulties can lead to reduced self-esteem and poor mental health, including depression [70]. The latter is supported by a large study that examined longitudinal data from 2950 people who had been diagnosed with ADHD by the age of 7.5 years and observed that symptoms in childhood were associated with an increased risk of depression at age 17.5 years [71]. Furthermore, this increased risk of depression was mediated by both social relationships with peers and academic achievement at 16 years of age [71]. Considering our findings in the context of the previous literature, it seems reasonable to suggest that those who meet the 24-HMB guidelines are more likely to have better social relationships and might also have a lower risk of depression and thus a better chance of academic achievement. However, future longitudinal research is needed to empirically test this hypothesis. ## Being bullied Children and adolescents in our study who met the 24-HMB guideline for screen time only were found to be at lower odds of being bullied. This finding may indicate that those who are less dependent on screen-based activities are also less vulnerable to being bullied. Previous research on adolescents with ADHD indicated that a high dependence on screen-based recreational activities is strongly associated with low self-esteem [72]. Speculatively, a lower self-esteem might make them more vulnerable to being bullied. In contrast, our results revealed an association between meeting the 24-HMB guidelines for physical activity and increased odds of being bullied. A possible explanation might be that the experience of being bullied increased the motivation to engage in physical activity, possibly to increase self-esteem [73, 74]. Bejerot et al. [ 2022] who examined possible associations between ADHD and bullying behaviors in a cross-sectional study, found that for participants who had been diagnosed with ADHD and that also suffer from poor motor skills (i.e, ball dexterity, coordination, or agility performance), have a higher risk to being bullied [75]. Therefore, another possible explanation for the increased odds of meeting the physical activity guidelines might be that these young people sought to improve physical activity skills to prevent the bullying. Longitudinal studies are needed to examine these theoretical assumptions. ## Bullying others Our results showed that meeting the 24-HMB guideline for sleep only, screen time only, and the combination of screen time and sleep were all associated with significantly lower odds of bullying others. Improved sleep has been associated with reduced antisocial behavior in school [62]. Li et al. [ 2021] examined NSCH data from 2011 to 12 for adolescents and found that meeting the age-appropriate sleep target mediated the association between increased MVPA and less bullying behavior [76]. Further, Moreau et al. [ 2013] found that executive functioning was positively associated with sleep duration in children with ADHD [77]. Previously, Unnever and Cornell [2003] had found that those with ADHD taking medication were more likely to bully others, which is perhaps related to a poorer self-control [78]. Taken together, the evidence presented above suggests that a longer sleep duration contributes to reduced bullying behavior, which in turn might be related to a sleep-related increase of inhibition performance. With regard to our results for meeting the 24-HMB guideline for screen time associated with lower odds of bullying others, previous research might provide an explanation for the current findings. Yen et al. [ 2014] found that addictive screen time behavior was associated with decreased social coping in adolescents aged 11 to 18 years old with a diagnosis of ADHD [72]. In addition, there is some evidence to suggest increased use of electronic devices, particularly for rapid response gaming may stimulate increased hyper vigilance and stress response, and increase ADHD symptoms [62]. There is also evidence from a study that investigated the frequency of digital media use in adolescents over 2 years and revealed higher frequency of digital media use was associated with higher level of ADHD symptoms [79]. While the results from our cross-sectional study do not indicate a direction to the association between meeting the 24-HMB guideline for screen time and reduced risk of bullying others, the literature suggests limiting the screen time may support social coping, and/or reduce exposure to stimulation that may cause hyper vigilance, stress or increased ADHD symptoms [28]. ## Implications and practical applications In conjunction with findings of previous research [24, 25, 48], the results of our study suggest that meeting all three of the 24-HMB guidelines is associated with reduced cognitive and social difficulties in children and adolescents with ADHD. Accordingly, our findings support the promotion of the 24-HMB guidelines for children and adolescents with ADHD and their caregivers. A key finding of our study is that meeting the 24-HMB guideline for non-educational (recreational) screen time made a substantial contribution to reduced odds for negative results for all four outcomes relating to cognitive and social difficulties, indicating the children and adolescents are very attracted to using electronic devices for recreation including games [62, 80, 81]. Therefore, it seems reasonable to speculate that some elements that attract them to use the virtual environment might be useful to stimulate learning and specific movement behaviors (e.g. engagement in MVPA). For example, promotion of physical activity through exergaming could be a valuable intervention strategy to reach this cohort [82], while meeting 24-HMB guidelines for non-educational screen time [28, 29]. ## Strengths and limitations A strength of this study is the sample of 3740 sets of data on children and adolescents derived from the 42,777 households who provided full responses to the nationwide collection of the NSCH 2020 survey. However, a disadvantage of the current study is the cross-sectional design which does not provide information on possible causal relationships between variables and thus necessitates further research using longitudinal studies to examine the causal mechanisms supporting our observations. Furthermore, as the current findings are based on information provided by the parent or guardian of the child/ adolescent, our results may be prone to reporting biases. The latter point is particularly applicable to sleep duration which is typically over estimated by the parents, especially for children with poor sleep efficiency [66]. While the measures for cognitive and social difficulties included in the NSCH survey provide some relevant data for the outcomes of interest, other validated measures for cognitive difficulties [83] and social difficulties [84] could be used in future research including controlled studies designed to examine the effects of meeting 24-HMB guidelines on these outcomes in children and adolescents with ADHD. Future studies should be designed to use objective and reliable measures of movement behaviors in order to gain a more nuanced understanding of their influence on health-related outcomes. For example, using objective measures for sleep timing, sleep quality and sleep duration [85, 86], could increase the robustness of the observations and lead to a more fine-graded understanding of the effects of specific movement behaviors. Likewise, prospective controlled research is needed to examine whether the time of day, days of the week, or specific type of physical activities undertaken by the children effect the cognitive or social difficulties outcomes in children with ADHD. Poitras et al. [ 2016], who undertook a systematic review, observed that children and adolescents in the general population could benefit from the recommended amount of MVPA (i.e. 60 minutes per day), even if it was accumulated in small bouts over the day [87]. However, Schmidt et al. [ 2015], showed that while both team games and aerobic exercise in children aged 10–12 years improved measures of aerobic fitness, only the team games improved executive function [88].Thus, the dose-response relationship considering qualitative (i.e. type of physical activity) and quantitative characteristics of movement behaviors (i.e. duration of physical bouts) should be examined in more detail in future studies. In addition, it would be an interesting topic for future research to compare the associations between the 24-HMB guidelines and the same outcome measures for cognitive and social difficulties between the current cohort of children with ADHD and a matched sample of the same survey population without a diagnosis of ADHD. ## Conclusion This cross-sectional study examined whether meeting 24-HMB guidelines—including recommendations concerning physical activity, sedentary behavior, and sleep—is, in a large sample of US children and adolescents with ADHD, associated with specific social and cognitive outcomes. The results of the current study revealed that meeting all three 24-HMB guidelines was associated with reduced odds of the occurrence of one or more negative outcomes for cognitive and social difficulties. Screen time, as a measure of sedentary behavior, was associated with all cognitive and social outcomes of interest, including serious difficulty concentrating, remembering, or making decisions; difficulty in making friends or keeping friends; being bullied; or bullying others. Furthermore, meeting the 24-HMB guideline for physical activity is linked to less cognitive difficulty and less social difficulty (i.e. making and keeping friends). Meeting the sleep recommendation of the 24-HMB guidelines is associated with less social difficulties—namely bullying others. The results of this study together with the previous literature on the benefits of adhering to the 24-HMB guidelines suggest the need to support children with ADHD and their caregivers to foster their ability to effectively adopt a healthy lifestyle. While further research is required to determine causal pathways, our results highlight the relevance of the 24-HMB guidelines as part of healthy lifestyle behaviors to enhance cognitive and social function for children and adolescents with ADHD. ## References 1. 1.Marshall, L. ADHD. Nat Rev Dis Primers. 2015; 1:15027. 10.1038/nrdp.2015.27 2. Faraone SV, Asherson P, Banaschewski T, Biederman J, Buitelaar JK, Ramos-Quiroga JA. **Attention-deficit/hyperactivity disorder**. *Nat Rev Dis Primers* (2015.0) **1** 1-23. DOI: 10.1038/nrdp.2015.20 3. Kazda L, McGeechan K, Bell K, Thomas R, Barratt A. **Association of attention-deficit/hyperactivity disorder diagnosis with adolescent quality of life**. *JAMA Netw Open* (2022.0) **5** e2236364-e2236364. DOI: 10.1001/jamanetworkopen.2022.36364 4. Zou L, Sasaki JE, Wei G-X, Huang T, Yeung AS, Neto OB. **Effects of mind–body exercises (Tai Chi/Yoga) on heart rate variability parameters and perceived stress: a systematic review with meta-analysis of randomized controlled trials**. *J Clin Med* (2018.0) **7** 404. DOI: 10.3390/jcm7110404 5. Polanczyk GV, Willcutt EG, Salum GA, Kieling C, Rohde LA. **ADHD prevalence estimates across three decades: an updated systematic review and meta-regression analysis**. *Int J Epidemiol* (2014.0) **43** 434-42. DOI: 10.1093/ije/dyt261 6. Thomas R, Sanders S, Doust J, Beller E, Glasziou P. **Prevalence of attention-deficit/hyperactivity disorder: a systematic review and meta-analysis**. *Pediatrics* (2015.0) **135** e994-1001. DOI: 10.1542/peds.2014-3482 7. Sayal K, Prasad V, Daley D, Ford T, Coghill D. **ADHD in children and young people: prevalence, care pathways, and service provision**. *Lancet Psychiatry* (2018.0) **5** 175-186. DOI: 10.1016/S2215-0366(17)30167-0 8. Hinshaw SP. **Externalizing behavior problems and academic underachievement in childhood and adolescence: causal relationships and underlying mechanisms**. *Psychol Bull* (1992.0) **111** 127. DOI: 10.1037/0033-2909.111.1.127 9. Arnold LE, Hodgkins P, Kahle J, Madhoo M, Kewley G. **Long-term outcomes of ADHD: academic achievement and performance**. *J Atten Disord* (2020.0) **24** 73-85. DOI: 10.1177/1087054714566076 10. Langberg JM, Becker SP. **Does long-term medication use improve the academic outcomes of youth with attention-deficit/hyperactivity disorder?**. *Clin Child Fam Psychol Rev* (2012.0) **15** 215-233. DOI: 10.1007/s10567-012-0117-8 11. Weyandt LL, Oster DR, Gudmundsdottir BG, DuPaul GJ, Anastopoulos AD. **Neuropsychological functioning in college students with and without ADHD**. *Neuropsychology* (2017.0) **31** 160. DOI: 10.1037/neu0000326 12. Becker SP, Fogleman ND. *Psychiatric co-occurrence (comorbidity) in adolescents with ADHD* (2020.0) 13. Borden LA, Hankinson J, Perry-Parrish C, Reynolds EK, Specht MW, Ostrander R. **Family and maternal characteristics of children with co-occurring ADHD and depression**. *J Atten Disord* (2020.0) **24** 963-972. DOI: 10.1177/1087054716666321 14. Faraone SV, Banaschewski T, Coghill D, Zheng Y, Biederman J, Bellgrove MA. **The world federation of ADHD international consensus statement: 208 evidence-based conclusions about the disorder**. *Neurosci Biobehav Rev* (2021.0) **128** 789-818. DOI: 10.1016/j.neubiorev.2021.01.022 15. Adamou M, Arif M, Asherson P, Aw T-C, Bolea B, Coghill D. **Occupational issues of adults with ADHD**. *BMC Psychiatry* (2013.0) **13** 1-7. DOI: 10.1186/1471-244X-13-59 16. Freckelton I. **Attention deficit hyperactivity disorder (ADHD) and the criminal law**. *Psychiatry Psychol Law* (2019.0) **26** 817-840. DOI: 10.1080/13218719.2019.1695266 17. Storebø OJ, Ramstad E, Krogh HB, Nilausen TD, Skoog M, Holmskov M. **Methylphenidate for children and adolescents with attention deficit hyperactivity disorder (ADHD)**. *Cochrane Database Syst Rev* (2015.0). DOI: 10.1002/14651858.CD009885.pub2 18. Evans SW, Owens JS, Wymbs BT, Ray AR. **Evidence-based psychosocial treatments for children and adolescents with attention deficit/hyperactivity disorder**. *J Clin Child Adolesc Psychol* (2018.0) **47** 157-198. DOI: 10.1080/15374416.2017.1390757 19. Gajria K, Lu M, Sikirica V, Greven P, Zhong Y, Qin P. **Adherence, persistence, and medication discontinuation in patients with attention-deficit/hyperactivity disorder–a systematic literature review**. *Neuropsychiatr Dis Treat* (2014.0) **10** 1543. PMID: 25187718 20. Pelham WE, Fabiano GA, Waxmonsky JG, Greiner AR, Gnagy EM, Pelham WE. **Treatment sequencing for childhood ADHD: A multiple-randomization study of adaptive medication and behavioral interventions**. *J Clin Child Adolesc Psychol* (2016.0) **45** 396-415. DOI: 10.1080/15374416.2015.1105138 21. Chacko A, Feirsen N, Bedard A-C, Marks D, Uderman JZ, Chimiklis A. **Cogmed working memory training for youth with ADHD: a closer examination of efficacy utilizing evidence-based criteria**. *J Clin Child Adolesc Psychol* (2013.0) **42** 769-783. DOI: 10.1080/15374416.2013.787622 22. Steeger CM, Gondoli DM, Gibson BS, Morrissey RA. **Combined cognitive and parent training interventions for adolescents with ADHD and their mothers: A randomized controlled trial**. *Child Neuropsychol* (2016.0) **22** 394-419. DOI: 10.1080/09297049.2014.994485 23. Suarez-Manzano S, Ruiz-Ariza A, de La Torre-Cruz M, Martínez-López EJ. **Acute and chronic effect of physical activity on cognition and behaviour in young people with ADHD: a systematic review of intervention studies**. *Res Dev Disabil* (2018.0) **77** 12-23. DOI: 10.1016/j.ridd.2018.03.015 24. Hoza B, Martin CP, Pirog A, Shoulberg EK. **Using physical activity to manage ADHD symptoms: the state of the evidence**. *Curr Psychiatry Rep* (2016.0) **18** 1-7. DOI: 10.1007/s11920-016-0749-3 25. Ng QX, Ho CYX, Chan HW, Yong BZJ, Yeo W-S. **Managing childhood and adolescent attention-deficit/hyperactivity disorder (ADHD) with exercise: a systematic review**. *Complement Ther Med* (2017.0) **34** 123-128. DOI: 10.1016/j.ctim.2017.08.018 26. Dimakos J, Gauthier-Gagné G, Lin L, Scholes S, Gruber R. **The associations between sleep and externalizing and internalizing problems in children and adolescents with attention-deficit/hyperactivity disorder: empirical findings, clinical implications, and future research directions**. *Child Adolesc Psychiatr Clin* (2021.0) **30** 175-193. DOI: 10.1016/j.chc.2020.08.001 27. Hvolby A. **Associations of sleep disturbance with ADHD: implications for treatment**. *ADHD Atten Defic Hyperact Disord* (2015.0) **7** 1-18. DOI: 10.1007/s12402-014-0151-0 28. Soares PSM, de Oliveira PD, Wehrmeister FC, Menezes AMB, Gonçalves H. **Is screen time throughout adolescence related to ADHD? findings from 1993 pelotas (Brazil) birth cohort study**. *J Atten Disord* (2022.0) **26** 331-339. DOI: 10.1177/1087054721997555 29. Suchert V, Pedersen A, Hanewinkel R, Isensee B. **Relationship between attention-deficit/hyperactivity disorder and sedentary behavior in adolescence: a cross-sectional study**. *ADHD Atten Defic Hyperact Disord* (2017.0) **9** 213-218. DOI: 10.1007/s12402-017-0229-6 30. Adelantado-Renau M, Moliner-Urdiales D, Cavero-Redondo I, Beltran-Valls MR, Martínez-Vizcaíno V, Álvarez-Bueno C. **Association between screen media use and academic performance among children and adolescents: a systematic review and meta-analysis**. *JAMA Pediatr Am Med Assoc* (2019.0) **173** 1058-1067. DOI: 10.1001/jamapediatrics.2019.3176 31. Chaput J-P, Gray CE, Poitras VJ, Carson V, Gruber R, Olds T. **Systematic review of the relationships between sleep duration and health indicators in school-aged children and youth**. *Appl Physiol Nutr Metab* (2016.0) **41** S266-S282. DOI: 10.1139/apnm-2015-0627 32. Tapia-Serrano MA, García-Hermoso A, Sevil-Serrano J, Sánchez-Oliva D, Sánchez-Miguel PA. **Is adherence to 24-hour movement guidelines associated with a higher academic achievement among adolescent males and females?**. *J Sci Med Sport* (2022.0) **25** 155-161. DOI: 10.1016/j.jsams.2021.09.005 33. Chastin SFM, Palarea-Albaladejo J, Dontje ML, Skelton DA. **Combined effects of time spent in physical activity, sedentary behaviors and sleep on obesity and cardio-metabolic health markers: a novel compositional data analysis approach**. *PLoS ONE* (2015.0) **10** e0139984. DOI: 10.1371/journal.pone.0139984 34. Cheval B, Maltagliati S, Sieber S, Cullati S, Sander D, Boisgontier MP. **Physical inactivity amplifies the negative association between sleep quality and depressive symptoms**. *Prev Med* (2022.0) **164** 107233. DOI: 10.1016/j.ypmed.2022.107233 35. Pedišić Ž. **Measurement issues and poor adjustments for physical activity and sleep undermine sedentary behaviour research—the focus should shift to the balance between sleep, sedentary behaviour, standing and activity**. *Kinesiology Kineziološki fakultet* (2014.0) **46** 135-146 36. Guerrero MD, Vanderloo LM, Rhodes RE, Faulkner G, Moore SA, Tremblay MS. **Canadian children’s and youth’s adherence to the 24-h movement guidelines during the COVID-19 pandemic: a decision tree analysis**. *J Sport Health Sci* (2020.0) **9** 313-321. DOI: 10.1016/j.jshs.2020.06.005 37. Liu S, Yu Q, Hossain M, Doig S, Bao R, Zhao Y. **Meeting 24-h movement guidelines is related to better academic achievement: findings from the YRBS 2019 cycle**. *Int J Ment Health Promot* (2022.0) **24** 13-24. DOI: 10.32604/IJMHP.2021.017660 38. Lu S, Cheval B, Yu Q, Hossain MM, Chen S-T, Taylor A. **Associations of 24-hour movement behavior with depressive symptoms and anxiety in children: cross-sectional findings from a Chinese sample**. *Healthcare* (2021.0). DOI: 10.3390/healthcare9111532 39. Paterson DC, Ramage K, Moore SA, Riazi N, Tremblay MS, Faulkner G. **Exploring the impact of COVID-19 on the movement behaviors of children and youth: a scoping review of evidence after the first year**. *J Sport Health Sci* (2021.0). DOI: 10.1016/j.jshs.2021.07.001 40. Rollo S, Antsygina O, Tremblay MS. **The whole day matters: understanding 24-hour movement guideline adherence and relationships with health indicators across the lifespan**. *J Sport Health Sci* (2020.0) **9** 493-510. DOI: 10.1016/j.jshs.2020.07.004 41. Tremblay MS, Carson V, Chaput J-P, Connor Gorber S, Dinh T, Duggan M. **Canadian 24-hour movement guidelines for children and youth: an integration of physical activity, sedentary behaviour, and sleep**. *Appl Physiol Nutr Metab* (2016.0) **41** S311-S327. DOI: 10.1139/apnm-2016-0151 42. Kong C, Chen A, Ludyga S, Herold F, Healy S, Zhao M. **Associations between meeting 24-hour movement guidelines and quality of life among children and adolescents with autism spectrum disorder**. *J Sport Health Sci* (2022.0). DOI: 10.1016/j.jshs.2022.08.003 43. Holton KF, Nigg JT. **The association of lifestyle factors and ADHD in children**. *J Atten Disord* (2020.0) **24** 1511-1520. DOI: 10.1177/1087054716646452 44. Quesada D, Ahmed NU, Fennie KP, Gollub EL, Ibrahimou B. **A review: associations between attention-deficit/hyperactivity disorder, physical activity, medication use, eating behaviors and obesity in children and adolescents**. *Arch Psychiatr Nurs* (2018.0) **32** 495-504. DOI: 10.1016/j.apnu.2018.01.006 45. Tandon PS, Sasser T, Gonzalez ES, Whitlock KB, Christakis DA, Stein MA. **Physical activity, screen time, and sleep in children with ADHD**. *J Phys Act Health* (2019.0) **16** 416-422. DOI: 10.1123/jpah.2018-0215 46. Friel CP, Duran AT, Shechter A, Diaz KM. **US children meeting physical activity, screen time, and sleep guidelines**. *Am J Prev Med* (2020.0) **59** 513-521. DOI: 10.1016/j.amepre.2020.05.007 47. Wang W, Haegele JA, Wu Y, Li C. **Meeting the 24-hour movement guidelines and outcomes in adolescents with ADHD: A cross-sectional observational study**. *Int J Environ Res Public Health* (2022.0) **19** 2132. DOI: 10.3390/ijerph19042132 48. Xie Y, Gao X, Song Y, Zhu X, Chen M, Yang L. **Effectiveness of physical activity intervention on ADHD symptoms: a systematic review and meta-analysis**. *Front Psychiatry Front Media SA* (2021.0) **12** 706625. DOI: 10.3389/fpsyt.2021.706625 49. Walsh JJ, Barnes JD, Cameron JD, Goldfield GS, Chaput J-P, Gunnell KE. **Associations between 24 hour movement behaviours and global cognition in US children: a cross-sectional observational study**. *Lancet Child Adolesc Health* (2018.0) **2** 783-791. DOI: 10.1016/S2352-4642(18)30278-5 50. Carson V, Chaput J-P, Janssen I, Tremblay MS. **Health associations with meeting new 24-hour movement guidelines for canadian children and youth**. *Prev Med* (2017.0) **95** 7-13. DOI: 10.1016/j.ypmed.2016.12.005 51. Bezerra TA, Clark CCT, Souza Filho AN, Fortes LDS, Mota JAPS, Duncan MJ. **24-hour movement behaviour and executive function in preschoolers: a compositional and isotemporal reallocation analysis**. *Eur J Sport Sci* (2021.0) **21** 1064-72. DOI: 10.1080/17461391.2020.1795274 52. 52.United States Census Bureau. 2020 National Survey of Children’s Health: Methodology Report [Internet]. 2020. Available from: https://www2.census.gov/programs-surveys/nsch/technical-documentation/methodology/2020-NSCH-Methodology-Report.pdf 53. Christiansen L, Beck MM, Bilenberg N, Wienecke J, Astrup A, Lundbye-Jensen J. **Effects of exercise on cognitive performance in children and adolescents with ADHD: potential mechanisms and evidence-based recommendations**. *J Clin Med* (2019.0) **8** 841. DOI: 10.3390/jcm8060841 54. Gawrilow C, Stadler G, Langguth N, Naumann A, Boeck A. **Physical activity, affect, and cognition in children with symptoms of ADHD**. *J Atten Disord* (2016.0) **20** 151-162. DOI: 10.1177/1087054713493318 55. Ziereis S, Jansen P. **Effects of physical activity on executive function and motor performance in children with ADHD**. *Res Dev Disabil* (2015.0) **38** 181-191. DOI: 10.1016/j.ridd.2014.12.005 56. Benzing V, Chang Y-K, Schmidt M. **Acute physical activity enhances executive functions in children with ADHD**. *Sci Rep* (2018.0) **8** 1-10. DOI: 10.1038/s41598-018-30067-8 57. Welsch L, Alliott O, Kelly P, Fawkner S, Booth J, Niven A. **The effect of physical activity interventions on executive functions in children with ADHD: a systematic review and meta-analysis**. *Ment Health Phys Act* (2021.0) **20** 100379. DOI: 10.1016/j.mhpa.2020.100379 58. Suchert V, Hanewinkel R, Isensee B. **Sedentary behavior and indicators of mental health in school-aged children and adolescents: a systematic review**. *Prev Med* (2015.0) **76** 48-57. DOI: 10.1016/j.ypmed.2015.03.026 59. Ekelund U, Steene-Johannessen J, Brown WJ, Fagerland MW, Owen N, Powell KE. **Does physical activity attenuate, or even eliminate, the detrimental association of sitting time with mortality? a harmonised meta-analysis of data from more than 1 million men and women**. *Lancet* (2016.0) **388** 1302-1310. DOI: 10.1016/S0140-6736(16)30370-1 60. Gardner DM, Gerdes AC. **A review of peer relationships and friendships in youth with ADHD**. *J Atten Disord* (2015.0) **19** 844-855. DOI: 10.1177/1087054713501552 61. Shechter A, Quispe KA, Mizhquiri Barbecho JS, Slater C, Falzon L. **Interventions to reduce short-wavelength (“blue”) light exposure at night and their effects on sleep: a systematic review and meta-analysis**. *Sleep Adv* (2020.0). DOI: 10.1093/sleepadvances/zpaa002 62. Lissak G. **Adverse physiological and psychological effects of screen time on children and adolescents: literature review and case study**. *Environ Res* (2018.0) **164** 149-157. DOI: 10.1016/j.envres.2018.01.015 63. Loh KK, Kanai R. **Higher media multi-tasking activity is associated with smaller gray-matter density in the anterior cingulate cortex**. *PLoS ONE* (2014.0) **9** e106698. DOI: 10.1371/journal.pone.0106698 64. Pea R, Nass C, Meheula L, Rance M, Kumar A, Bamford H. **Media use, face-to-face communication, media multitasking, and social well-being among 8-to 12-year-old girls**. *Dev Psychol* (2012.0) **48** 327. DOI: 10.1037/a0027030 65. Yang A, Rolls ET, Dong G, Du J, Li Y, Feng J. **Longer screen time utilization is associated with the polygenic risk for attention-deficit/hyperactivity disorder with mediation by brain white matter microstructure**. *EBioMedicine* (2022.0) **80** 104039. DOI: 10.1016/j.ebiom.2022.104039 66. Mazza S, Bastuji H, Rey AE. **Objective and subjective assessments of sleep in children: comparison of actigraphy, sleep diary completed by children and parents’ estimation**. *Front Psychiatry* (2020.0) **11** 495. DOI: 10.3389/fpsyt.2020.00495 67. Becker SP, Epstein JN, Tamm L, Tilford AA, Tischner CM, Isaacson PA. **Shortened sleep duration causes sleepiness, inattention, and oppositionality in adolescents with attention-deficit/hyperactivity disorder: Findings from a crossover sleep restriction/extension study**. *J Am Acad Child Adolesc Psychiatry* (2019.0) **58** 433-442. DOI: 10.1016/j.jaac.2018.09.439 68. DuPaul GJ, Weyandt LL, Janusis GM. **ADHD in the classroom: effective intervention strategies**. *Theory Pract* (2011.0) **50** 35-42. DOI: 10.1080/00405841.2011.534935 69. Spence SH. **Social skills training with children and young people: theory, evidence and practice**. *Child Adolesc Ment Health* (2003.0) **8** 84-96. DOI: 10.1111/1475-3588.00051 70. Humphreys KL, Katz SJ, Lee SS, Hammen C, Brennan PA, Najman JM. **The association of ADHD and depression: mediation by peer problems and parent–child difficulties in two complementary samples**. *J Abnorm Psychol* (2013.0) **122** 854. DOI: 10.1037/a0033895 71. Powell V, Riglin L, Hammerton G, Eyre O, Martin J, Anney R. **What explains the link between childhood ADHD and adolescent depression? investigating the role of peer relationships and academic attainment**. *Eur Child Adolesc Psychiatry* (2020.0) **29** 1581-1591. DOI: 10.1007/s00787-019-01463-w 72. Yen C-F, Chou W-J, Liu T-L, Yang P, Hu H-F. **The association of Internet addiction symptoms with anxiety, depression and self-esteem among adolescents with attention-deficit/hyperactivity disorder**. *Compr Psychiatry* (2014.0) **55** 1601-1608. DOI: 10.1016/j.comppsych.2014.05.025 73. Ash T, Bowling A, Davison K, Garcia J. **Physical activity interventions for children with social, emotional, and behavioral disabilities—a systematic review**. *J Dev Behav Pediatr* (2017.0) **38** 431-445. DOI: 10.1097/DBP.0000000000000452 74. Fard FS, Baniasadi T, Ahmadi TS, Biyabani P, Mofrad SK. **Effects of physical activity on wellbeing among children with ADHD: a mediation by self-esteem**. *J Hum Ins* (2022.0) **6** 1-6 75. Bejerot S, Ståtenhag L, Glans MR. **Below average motor skills predict victimization from childhood bullies: a study of adults with ADHD**. *J Psychiatr Res* (2022.0) **153** 269-275. DOI: 10.1016/j.jpsychires.2022.07.019 76. Li C, Haegele JA, Wang L. **Moderate-to-vigorous physical activity and behavioral outcomes in adolescents with attention deficit and hyperactivity disorder: the role of sleep**. *Disabil Health J* (2021.0) **14** 100970. DOI: 10.1016/j.dhjo.2020.100970 77. Moreau V, Rouleau N, Morin CM. *Sleep, attention, and executive functioning in children with attention-deficit/hyperactivity disorder archives of clinical neuropsychology* (2013.0) 78. Unnever JD, Cornell DG. **Bullying, self-control, and ADHD**. *J Interpers Violence* (2003.0) **18** 129-147. DOI: 10.1177/0886260502238731 79. Ra CK, Cho J, Stone MD, de La Cerda J, Goldenson NI, Moroney E. **Association of digital media use with subsequent symptoms of attention-deficit/hyperactivity disorder among adolescents**. *JAMA Am Med Assoc* (2018.0) **320** 255-263. DOI: 10.1001/jama.2018.8931 80. Tong L, Xiong X, Tan H. **Attention-deficit/hyperactivity disorder and lifestyle-related behaviors in children**. *PLoS ONE* (2016.0) **11** e0163434. DOI: 10.1371/journal.pone.0163434 81. van Egmond-Fröhlich AWA, Weghuber D, de Zwaan M. **Association of symptoms of attention-deficit/hyperactivity disorder with physical activity, media time, and food intake in children and adolescents**. *PLoS ONE* (2012.0) **7** e49781. DOI: 10.1371/journal.pone.0049781 82. Benzing V, Schmidt M. **The effect of exergaming on executive functions in children with ADHD: a randomized clinical trial**. *Scand J Med Sci Sports* (2019.0) **29** 1243-1253. DOI: 10.1111/sms.13446 83. Johnson M, Johnels JÅ, Östlund S, Cedergren K, Omanovic Z, Hjalmarsson K. **Long-term medication for ADHD and development of cognitive functions in children and adolescents**. *J Psychiatr Res* (2021.0) **142** 204-209. DOI: 10.1016/j.jpsychires.2021.07.055 84. Storebø OJ, Andersen ME, Skoog M, Hansen SJ, Simonsen E, Pedersen N. **Social skills training for attention deficit hyperactivity disorder (ADHD) in children aged 5 to 18 years**. *Cochrane Database Syst Rev* (2019.0). DOI: 10.1002/14651858.CD008223.pub3 85. Koinis-Mitchell D, Kopel SJ, Seifer R, LeBourgeois M, McQuaid EL, Esteban CA. **Asthma-related lung function, sleep quality, and sleep duration in urban children**. *Sleep Health* (2017.0) **3** 148-156. DOI: 10.1016/j.sleh.2017.03.008 86. Nováková M, Paclt I, Ptáček R, Kuželová H, Hájek I, Sumová A. **Salivary melatonin rhythm as a marker of the circadian system in healthy children and those with attention-deficit/hyperactivity disorder**. *Chronobiol Int* (2011.0) **28** 630-637. DOI: 10.3109/07420528.2011.596983 87. Poitras VJ, Gray CE, Borghese MM, Carson V, Chaput J-P, Janssen I. **Systematic review of the relationships between objectively measured physical activity and health indicators in school-aged children and youth**. *Appl Physiol Nutr Metab* (2016.0) **41** S197-239. DOI: 10.1139/apnm-2015-0663 88. Schmidt M, Jäger K, Egger F, Roebers CM, Conzelmann A. **Cognitively engaging chronic physical activity, but not aerobic exercise, affects executive functions in primary school children: a group-randomized controlled trial**. *J Sport Exerc Psychol* (2015.0) **37** 575-591. DOI: 10.1123/jsep.2015-0069
--- title: Association of metformin, aspirin, and cancer incidence with mortality risk in adults with diabetes authors: - Suzanne G Orchard - Jessica E Lockery - Jonathan C Broder - Michael E Ernst - Sara Espinoza - Peter Gibbs - Rory Wolfe - Galina Polekhina - Sophia Zoungas - Holli A Loomans-Kropp - Robyn L Woods - John McNeil - John McNeil - Robyn Woods - Rory Wolfe - Anne Murray - Andrew Chan - Suzanne Orchard - Jessica Lockery - Mark Nelson - Christorpher Reid - Raj Shah - Anne Newmann - Elsdon Storey - Nigel Stocks - Andrew Tonkin - Sara Espinoza journal: JNCI Cancer Spectrum year: 2023 pmcid: PMC10042437 doi: 10.1093/jncics/pkad017 license: CC BY 4.0 --- # Association of metformin, aspirin, and cancer incidence with mortality risk in adults with diabetes ## Body Cancer is a leading cause of death worldwide [1,2]. Cancer incidence is expected to increase in the next decade, with older people (eg, those aged over 70 years) at higher risk of incident cancer and cancer mortality [3]. In the context of an aging population [4], prevention and treatment of cancer are a public health imperative. Type 2 diabetes is a complex disease characterized by β-cell failure in the setting of insulin resistance [5]; it is a known risk factor for several types of cancer, including liver, pancreatic, colorectal, breast, endometrial, and kidney cancer [6,7]. In type 2 diabetes, systemic insulin resistance results in adaptive increases in β-cell mass and function, which initially conserve glucose homeostasis at the expense of elevated insulin levels. When this compensatory mechanism fails, hyperglycemia occurs [5]. Thus, in most people with type 2 diabetes, hyperglycemia is associated with endogenous hyperinsulinemia. Although the underlying mechanism behind type 2 diabetes and cancer risk remains unclear, both hyperglycemia and hyperinsulinemia are associated with increases in the prevalence and mortality of malignancies [6,8-11], and both contribute to carcinogenic processes, including enhanced cellular proliferation, invasion, and apoptosis inhibition [12-14]. Metformin, an oral antihyperglycemic agent, is the recommended first-line treatment for type 2 diabetes in the absence of contraindications [12]. Metformin acts by suppressing hepatic glucose production and increasing peripheral glucose uptake [13], thereby lowering blood glucose levels without increasing circulating insulin [14]. This specific trait differentiates metformin from other antihyperglycemic medications, such as sulfonylureas and insulin therapy, which lower blood glucose levels by increasing plasma insulin concentrations [15]. Recent analyses have suggested that their use may be associated with increased risk of cancer [16-18]. In contrast, several studies have shown that metformin may protect against the development and progression of a variety of malignancies [19-26]. Other observational studies, however, have reported no association between metformin use and cancer incidence or outcome, with authors citing methodological biases as tending to exaggerate the benefit of metformin [27-32]. The conflicting evidence suggests that metformin may exercise different effects on cancer at different anatomical sites or, alternatively, that analyses of the effect of metformin in clinical practice may be complicated by other factors, such as co-prescribed medications or residual confounding caused by comorbid conditions. Aspirin is commonly co-prescribed with metformin for prevention of cardiovascular disease in people with diabetes [33]. Recent meta-analyses have found that low-dose aspirin, taken regularly for 4 to 5 years, could reduce cancer incidence, risk of metastatic spread, and cancer mortality over the subsequent 10 or more years [34-36]. That said, a recent clinical trial of aspirin in older adults, the ASPirin in Reducing Events in the Elderly (ASPREE) study, showed no effect of aspirin on cancer incidence but an increased risk of cancer-related death [37,38]. Furthermore, the A Study of Cardiovascular Events in Diabetes (ASCEND) clinical trial found no evidence of reduction in gastrointestinal or other cancer incidences in people with diabetes who were randomized to aspirin vs placebo after 7 years of treatment and follow-up [39]. Bearing this is mind, it is possible that these medications have opposing effects on cancer prevention, with aspirin increasing and metformin decreasing risk. Disentangling the effects of metformin and aspirin may assist in explaining the conflicting evidence about metformin and cancer. In this analysis, we aimed to use the randomization of participants to aspirin or placebo in the ASPREE trial to examine in older adults with diabetes 1) the association between metformin and cancer incidence and mortality, 2) the effect of aspirin (vs placebo) in metformin users on cancer incidence and mortality, and 3) whether the effect of aspirin (vs placebo) differs between those who do and do not use metformin. ## Abstract ### Background Metformin and aspirin are commonly co-prescribed to people with diabetes. Metformin may prevent cancer, but in older people (over 70 years), aspirin has been found to increase cancer mortality. This study examined whether metformin reduces cancer mortality and incidence in older people with diabetes; it used randomization to 100 mg aspirin or placebo in the ASPirin in Reducing Events in the Elderly (ASPREE) trial to quantify aspirin’s impact on metformin users. ### Methods Analysis included community-dwelling ASPREE participants (aged ≥70 years, or ≥65 years for members of US minority populations) with diabetes. Diabetes was defined as a fasting blood glucose level greater than 125 mg/dL, self-report of diabetes, or antidiabetic medication use. Cox proportional hazards regression models were used to analyze the association of metformin and a metformin-aspirin interaction with cancer incidence and mortality, with adjustment for confounders. ### Results Of 2045 participants with diabetes at enrollment, 965 were concurrently using metformin. Metformin was associated with a reduced cancer incidence risk (adjusted hazard ratio [HR] = 0.68, $95\%$ confidence interval [CI] = 0.51 to 0.90), but no conclusive benefit for cancer mortality (adjusted HR = 0.72, $95\%$ CI = 0.43 to 1.19). Metformin users randomized to aspirin had greater risk of cancer mortality compared with placebo (HR = 2.53, $95\%$ CI = 1.18 to 5.43), but no effect was seen for cancer incidence (HR = 1.11, $95\%$ CI = 0.75 to 1.64). The possible effect modification of aspirin on cancer mortality, however, was not statistically significant (interaction $$P \leq .11$$). ### Conclusions In community-dwelling older adults with diabetes, metformin use was associated with reduced cancer incidence. Increased cancer mortality risk in metformin users randomized to aspirin warrants further investigation. ### ASPREE Trial Registration ClinicalTrials.gov ID NCT01038583 ## The ASPREE clinical trial This ASPREE trial was a secondary, intention-to-treat analysis of ASPREE clinical trial data (ClinicalTrials.gov ID NCT01038583). The ASPREE study enrolled community-dwelling individuals 70 years of age or older (≥65 years of age for members of US minority groups) with no major cardiovascular disease in Australia and the United States. Preexisting cancer was not an exclusion if life expectancy was beyond 5 years [$19\%$ of participants had preexisting cancer [40]]. Details regarding trial methods, recruitment, and outcomes have been described previously [37,41-44]. Briefly, 19 114 participants were randomly assigned to aspirin (100 mg) or matching placebo and followed for a median of 4.7 years. Demographic data, including sex, race, ethnicity, smoking status, alcohol use and previous aspirin use were collected by participant self-report. Race/ethnicity categories are Caucasian/White and other, where other includes Aboriginal/Torres Strait Islanders, American Indian, Asian, Black/African American, Hispanic/Latino or Native Hawaiian/Other Pacific Islander/Maori. Ethics committees at each participating center approved the trial, and all participants provided written informed consent before enrollment. ## Event data collection and adjudication Cancer was defined as diagnosis of any new primary cancer, excluding nonmelanoma skin cancer, that had been histopathologically confirmed or clinically evident on imaging. Cancer mortality was defined as death where the primary cause was attributable to cancer. Participants completed a questionnaire designed to record new cancer events at 6-month intervals, and clinical records were searched annually for new cancer diagnoses. All in-trial event reports (cancer and death) triggered the collection of clinical evidentiary documentation (eg, histopathology, specialist letters, imaging, and death certificates) from hospitals, pathology services, and responsible physicians. These clinical documents were compiled into an event summary and presented to a committee of international clinical experts specializing in oncology, for adjudication. Where histopathological confirmation was not undertaken clinically (eg, in the setting of diffuse metastatic disease or patient refusal of surgical intervention of any kind), cancer cases were considered to reach the cancer endpoint only if strong clinical evidence of disease was present on imaging (computed tomography, positron emission tomography, magnetic resonance imaging, or bone scans showing clear primary or diffuse metastatic disease) or blood biomarkers. Alternatively, clinically documented treatment for metastatic disease was considered sufficient to confirm the cancer endpoint. If the results of imaging investigations were unclear, suspicious, or inconclusive imaging, the cancer case was not considered a cancer endpoint. Further details of the cancer and cause-of-death adjudication processes have been published elsewhere [37,38]. ## Collection and coding of medications The ASPREE study defined baseline medications as any medications prescribed by a physician (or any nonsteroidal anti-inflammatory drug) and taken regularly at the time of randomization. Baseline medication data were collected directly from ASPREE participants, who brought their medications to the enrollment visit that immediately preceded randomization. Medication data were cross-checked with the participant’s medical record (when available), then transcribed into the ASPREE data system [45] and coded according to the World Health Organization Anatomical Therapeutic Chemical coding system [46]. Detailed methods for the coding process have been published elsewhere [47]. ## Definitions Metformin use refers to the prescription of a medication with an Anatomical Therapeutic Chemical code of A10BA02. Diabetes was defined as the presence at study entry of a high fasting blood glucose level (FBGL) (>125 mg/dL) [48], a self-report of diabetes, or prescription of an antihyperglycemic medication (see Supplementary Table 1, available online, for the full list). See Figure 1 for a flow diagram of the baseline cohort. **Figure 1.:** *Cohort at baseline included in the current analysis. Elevated blood glucose refers to a fasting blood glucose level >125 mg/dL. Counts of participants with self-reported diabetes, elevated blood glucose, and anti-diabetic medication use are not mutually exclusive.* ## Statistical analysis The purpose of this secondary data analysis was to explore the long-term associations between metformin use and cancer based on the principles of intention to treat. Therefore, metformin exposure was defined as baseline metformin use only. A review of metformin use or nonuse over the follow-up period revealed that $83\%$ of participants maintained consistency of either use or non-use of metformin. Cox proportional hazards regression models were used to analyze the relationship between metformin exposure at study baseline and cancer outcomes. For cancer incidence, the analysis was performed on the first cancer event (date of diagnosis) of any in-trial cancer, and censoring was defined at death if non–cancer-related death occurred (as a non–cancer-related death strongly indicated that cancer was not present) or the last date on which clinical event data were collected. For cancer mortality, the date of death was used as the event date, and censoring was defined at the end of the study, when the National Death Indices search was performed. Adjusted hazard ratios (HRs) were determined for incident cancer and mortality, which controlled for baseline factors identified as potential confounders. Because of limited sample size, cancer location site and stage were not analyzed. Competing-risks Nelson-Aalen cumulative incidence curves of cancer incidence and mortality are presented for participants with diabetes who do and do not use metformin. To assess whether the association of metformin on outcome varied by therapeutic efficacy, additional Cox proportional hazard regression models included an interaction term between baseline blood glucose and metformin. Using these models, the log adjusted hazard ratio of metformin, across varying levels of blood glucose, were visualized using line plots. The random allocation of ASPREE participants to aspirin or placebo was used to compare the aspirin effect between those who do and do not use metformin. Thus, these Cox proportional hazard regression models included an interaction between metformin and aspirin and were not adjusted for baseline factors. Supplementary analysis was conducted using competing risks regression through Fine-Gray subdistribution hazard models. Deaths that occurred when participants were still at risk of cancer incidence were considered a competing risk of cancer incidence, but non–cancer-related deaths in participants with a cancer diagnosis were considered a competing risk of cancer mortality. The proportional hazards assumption was assessed using tests of the Schoenfeld residuals against time [49], which showed that the assumption was satisfied in all models. Analyses, performed in R, version 4.0.2 (R Foundation for Statistical Computing), were 2-sided, with $P \leq .05$ considered statistically significant. ## Results Of the 2045 participants with diabetes, 965 used metformin at baseline (median [Interquartile range, IQR] follow-up = 4.6 [3.5-5.5] years) and 1080 did not (median [IQR] follow-up = 4.5 [3.3-5.5] years) (Figure 1). Most participants with diabetes stayed within their baseline groups over follow-up (1698 of 2045 [$83\%$]), although 107 ($11\%$) participants using metformin at baseline stopped use during follow-up for at least 1 year, and 240 of 1080 ($22\%$) participants not using metformin at baseline subsequently commenced metformin during follow-up. Table 1 shows baseline characteristics of ASPREE participants with diabetes, stratified by metformin use. Metformin users were more likely to be younger and not White, report previous regular aspirin use, have polypharmacy, have a body mass index of 25 kg/m2 or higher, and never have used alcohol compared with those who did not use metformin. Metformin users were also more likely to use other diabetes medications and have lower FBGLs than those not using metformin. Supplementary Table 2 (available online) shows baseline characteristics for participants who did not have diabetes. **Table 1.** | Unnamed: 0 | Diabetes | Diabetes.1 | Diabetes.2 | Diabetes.3 | | --- | --- | --- | --- | --- | | Characteristic | Metformin (n = 965) | No metformin (n = 1080) | Total (N = 2045) | P | | Age at randomization, No. (%) | Age at randomization, No. (%) | Age at randomization, No. (%) | Age at randomization, No. (%) | Age at randomization, No. (%) | | 65-69 y | 83 (9) | 62 (6) | 145 (7) | .009 | | 70-74 y | 496 (51) | 541 (50) | 1037 (51) | .009 | | 75-79 y | 255 (26) | 288 (27) | 543 (27) | .009 | | 80-84 y | 103 (11) | 134 (12) | 237 (12) | .009 | | ≥85 y | 28 (3) | 55 (5) | 83 (4) | .009 | | Sex, No. (%) | Sex, No. (%) | Sex, No. (%) | Sex, No. (%) | Sex, No. (%) | | Female | | | | | | Male | 497 (52) | 549 (51) | 1046 (51) | .762 | | Ethnicity and race,b No. (%) | Ethnicity and race,b No. (%) | Ethnicity and race,b No. (%) | Ethnicity and race,b No. (%) | Ethnicity and race,b No. (%) | | White/Caucasian | 755 (78) | 909 (84) | 1664 (81) | <.001 | | Other | | | | | | BMI category, No. (%) | BMI category, No. (%) | BMI category, No. (%) | BMI category, No. (%) | BMI category, No. (%) | | ≥25 | 854 (89) | 917 (85) | 1771 (87) | .032 | | Smoking status, No. (%) | Smoking status, No. (%) | Smoking status, No. (%) | Smoking status, No. (%) | Smoking status, No. (%) | | Current | 47 (5) | 51 (5) | 98 (5) | .947 | | Former | 424 (44) | 482 (45) | 906 (44) | .947 | | Never | 494 (51) | 547 (51) | 1041 (51) | .947 | | Alcohol use, No. (%) | Alcohol use, No. (%) | Alcohol use, No. (%) | Alcohol use, No. (%) | Alcohol use, No. (%) | | Current | 613 (64) | 767 (71) | 1380 (67) | <.001 | | Former | 105 (11) | 83 (8) | 188 (9) | <.001 | | Never | 247 (26) | 230 (21) | 477 (23) | <.001 | | Clinical features | Clinical features | Clinical features | Clinical features | Clinical features | | Previous regular aspirin use,c No. (%) | 190 (20) | 169 (16) | 359 (18) | .017 | | CKD,d No. (%) | 345 (38) | 368 (36) | 713 (37) | .385 | | Polypharmacy (≥5), No. (%) | 641 (66) | 424 (39) | 1065 (52) | <.001 | | Personal cancer history, No. (%) | 182 (19) | 204 (19) | 386 (19) | .971 | | Family cancer history,e No. (%) | 538 (56) | 636 (59) | 1174 (57) | .152 | | Physical component summary score,f median (IQR)g | 47.4 (39.4-53.5) | 47.5 (39.8-54.1) | 47.4 (39.7-53.7) | .671 | | Randomized treatment group, No. (%) | Randomized treatment group, No. (%) | Randomized treatment group, No. (%) | Randomized treatment group, No. (%) | Randomized treatment group, No. (%) | | Aspirin | 516 (53) | 508 (47) | 1024 (50) | – | | Placebo | 449 (47) | 572 (53) | 1021 (50) | – | | FBGL | FBGL | FBGL | FBGL | FBGL | | FBGL, mean (SD), mg/dL | 132.8 (37.4) | 129.5 (34.9) | 131.0 (36.1) | .042 | | FBGL, mean (SD), mmol/L | 7.4 (2.1) | 7.2 (1.9) | 7.3 (2.0) | – | | Diabetes treatment, No. (%) | Diabetes treatment, No. (%) | Diabetes treatment, No. (%) | Diabetes treatment, No. (%) | Diabetes treatment, No. (%) | | Insulin | 83 (9) | 74 (7) | 157 (8) | .138 | | Other antihyperglycemic medication use | 364 (38) | 137 (13) | 501 (24) | <.001 | | Diabetes self-report, No. (%) | Diabetes self-report, No. (%) | Diabetes self-report, No. (%) | Diabetes self-report, No. (%) | Diabetes self-report, No. (%) | | Self-report diabetes only | – | 413 (38) | 413 (20) | – | Table 2 describes the relationship between metformin use and cancer incidence and mortality in participants with and without diabetes. After adjustment for baseline characteristics, including FBGL, there was a lower rate of cancer incidence in the metformin group than in the no metformin group (adjusted HR = 0.68, $95\%$ confidence interval [CI] = 0.51 to 0.90), but no significant differences were observed for cancer mortality (adjusted HR = 0.72, $95\%$ CI = 0.43 to 1.19). Metformin users had similar event rates to people without diabetes for cancer incidence (adjusted HR = 1.09, $95\%$ CI = 0.88 to 1.35) and cancer mortality (adjusted HR = 1.39, $95\%$ CI = 0.96 to 2.02), while people with diabetes who did not use metformin had higher rates of cancer incidence (adjusted HR = 1.35, $95\%$ CI = 1.13 to 1.62) and cancer mortality (adjusted HR = 1.55, $95\%$ CI = 1.12 to 2.15). Supplementary analysis with competing-risks regression were consistent with these results (Supplementary Table 3, available online). The cumulative-incidence curves show that metformin users have lower cumulative cancer incidence over time but not lower rates of cancer mortality (Figure 2). **Figure 2.:** *Nelson-Aalen cumulative-incidence curves ($95\%$ confidence interval) for cancer incidence and mortality in people with diabetes by metformin use. P values (top-left corner) were calculated using Gray tests.* TABLE_PLACEHOLDER:Table 2. The association of metformin with cancer incidence and mortality as well as continuous FBGL measures are shown in Figure 3. Visually, there was a suggestion that FBGL modified the association between metformin and cancer incidence (interaction effect $$P \leq .06$$), suggesting that the benefit of metformin may be more pronounced in those with lower FBGLs. **Figure 3.:** *Log-adjusted hazard ratios (HRs) (95% confidence interval [CI]) of metformin (vs no metformin) across varying levels of baseline fasting blood glucose. The log-adjusted hazard ratios were determined by using Cox regression models, with an interaction between metformin and blood glucose, adjusting for baseline confounders: age at randomization, sex, ethnicity (Caucasian/White vs other), body mass index (as continuous), smoking status (current and former vs never), alcohol status (current and former vs never), previous aspirin use, chronic kidney disease, treatment group, polypharmacy, family cancer history, physical component summary score, personal cancer history, insulin use, and other oral antihyperglycemic medication use.* The combined effect of metformin and aspirin is shown in Table 3. Among those using metformin, those randomized to aspirin had a similar rate of cancer incidence (HR = 1.11, $95\%$ CI = 0.75 to 1.64) but a significantly greater rate of cancer mortality (HR = 2.53, $95\%$ CI = 1.18 to 5.43) compared with placebo. For those without metformin exposure, event rates were similar between participants randomized to aspirin and placebo for cancer incidence (HR = 1.10, $95\%$ CI = 0.79 to 1.52) and mortality (HR = 1.16, $95\%$ CI = 0.64 to 2.09). The possible effect modification on cancer mortality, however, was not statistically significant (interaction $$P \leq .11$$). Competing-risks regression supplementary analysis produced similar results (Supplementary Table 4, available online). **Table 3.** | Unnamed: 0 | Metformin | Metformin.1 | Metformin.2 | No metformin | No metformin.1 | No metformin.2 | P interaction of metformin and aspirina | | --- | --- | --- | --- | --- | --- | --- | --- | | | Aspirin | Placebo | HRa | Aspirin | Placebo | HRa | P interaction of metformin and aspirina | | | No. (rate)b | No. (rate)b | (95% CI) | No. (rate)b | No. (rate)b | (95% CI) | P interaction of metformin and aspirina | | Incident cancerc | 56 (26.94) | 45 (24.37) | 1.11 (0.75 to 1.64) | 70 (34.55) | 74 (31.76) | 1.10 (0.79 to 1.52) | .97 | | Cancer mortalityd | 25 (10.99) | 9 (4.47) | 2.53 (1.18 to 5.43) | 22 (9.98) | 22 (8.67) | 1.16 (0.64 to 2.09) | .11 | ## Discussion In older people with diabetes, we found that a relationship exists between metformin and cancer prevention that may be modified by lower FBGLs. Overall, we found that people with diabetes whose physician had prescribed metformin had a lower cancer incidence risk than those whose physicians had not prescribed metformin over 4.5 years of follow up. We found no conclusive associations between metformin and cancer mortality, however, likely because of small event numbers. Furthermore, the rate of incident cancer in those on metformin was similar in ASPREE participants who did not have diabetes, indicating that the benefits associated with metformin may potentially attenuate diabetes as a risk factor for cancer. Several potential explanations exist for the risk reduction associated with metformin use that we observed. A chemoprevention effect of metformin has been attributed to several biological mechanisms, including 1) activation of the liver kinase B-1–adenyl-monophosphate protein kinase pathway and subsequent suppression of hepatic glucose production leading to a reduction in insulin requirements [50,51] and 2) direct effect on cancer cells through reduction in insulin and/or insulin-like growth factor-I (IGFI) receptor signaling [52,53] and inhibition the mammalian target of rapamycin pathway by Adenosine Monophosphate-activated protein kinase–dependent mechanisms reducing adenosine triphosphate synthesis [52,54,55]. Therefore, metformin’s mechanism of chemoprevention is not thought to be solely attributable to adequate control of blood glucose but also to its ability to reduce hyperinsulinemia and subsequent insulin signaling pathway activity. Our data indicated that the beneficial associations of metformin on cancer incidence may not be observed in those with high FBGLs (>150 mg/dL), suggesting that if hyperglycemia and hyperinsulinemia persist, then metformin may have limited clinical effect on cancer risk. Although our analysis of the interaction between FBGL and metformin on cancer incidence was not conclusive, previous studies have shown that blood glucose control is essential for minimizing the risk of microvascular complications, a condition that emerging evidence shows is associated with future risk of cancer [56,57]. Our results are also broadly consistent with the American Diabetes Association recommendations for target glucose levels to minimize diabetes-related morbidity (FBGL <150 mg/dL or hemoglobin A1c [HbA1c] below equivalent cutoff). Although not conclusive, our results suggest that metformin may make little difference to outcomes if FBGLs are above the American Diabetes Association recommended level. The ASPREE clinical trial found no difference between aspirin and placebo for cancer incidence but an increased risk of cancer mortality with aspirin [37]. In particular, ASPREE demonstrated an increased risk of cancer-related mortality with aspirin regardless of diabetes status, especially for stage III and above cancers [37,38]. Given that our analysis used the same data but focused on the subgroup with diabetes, we expected to observe an increased cancer mortality risk with aspirin. Our goal was specifically to explore whether metformin use modified this risk. We found that for metformin users, aspirin use compared with placebo was associated with a significantly increased risk of cancer mortality. Theoretically, aspirin could increase cancer risk through hyperinsulinemia. Several clinical trials conducted in the 1980s demonstrated a detrimental effect of aspirin therapy on insulin sensitivity in people with [58] and without diabetes [59,60]. A more recent clinical study in healthy obese people showed that high-dose aspirin reduced hepatic glucose production and peripheral plasma glucose levels, but these effects were at the expense of a $47\%$ increase in plasma insulin concentrations [61]. Therefore, it is plausible that the insulin-attenuating action of metformin may present only in the absence of an aspirin-induced increase in plasma insulin concentration and that aspirin use could result in net harm for cancer outcomes. Given that the magnitude of the elevated cancer mortality risk observed within the metformin group was pronounced and greater than the hazard ratio observed for the overall cohort [37], we explored whether the risks of aspirin on cancer mortality were modified or indeed magnified with metformin use. Our results do not, however, provide sufficient evidence to draw this conclusion. Although we observed markedly different hazard ratios for the estimated effect of aspirin on cancer mortality within the metformin (HR = 2.53, $95\%$ CI = 1.18 to 5.43) and no metformin groups (HR = 1.16, $95\%$ CI = 0.64 to 2.09), our sample size was limited, and the interaction effect comparing the hazard ratios was low ($$P \leq .113$$) but not statistically significant. A relatively small proportion of ASPREE participants had diabetes ($10.6\%$); of these, fewer than half were prescribed metformin, and a smaller proportion still experienced cancer mortality. Thus, although our data showed significantly increased risk of cancer mortality with aspirin among metformin users, we cannot be sure whether the differences in aspirin effects we observed between the metformin and no metformin groups were the result of a true effect modification by metformin or of other factors. Previous meta-analyses of aspirin clinical trials conducted in middle-aged individuals (ranging in median age at randomization from 57.5 to 66.9 years) found that aspirin treatment prevented cancer, particularly colorectal cancer, over the next 20 years [34,35]. The majority of the studies in these meta-analyses, however, were conducted before the introduction of metformin into mainstream use in the United States, which occurred in 1995 [62]; as such, they will not have metformin as a confounder. Within the United States today, however, approximately $61.7\%$ of people with diabetes who are older than 60 years of age and likely now taking metformin use aspirin for primary prevention, and this number is increasing with time [63]. Taken together, then, much of the evidence supporting aspirin for cancer prevention in middle-aged people was gathered from metformin-naive populations, and much of the recent observational data being used to examine metformin chemoprevention were likely gathered from aspirin-enriched populations. Our results are not conclusive, but we believe that they provide incentive to better understand the relationship among metformin, aspirin, and cancer outcomes, particularly in older individuals with diabetes, through research using larger cohorts and trials. A key strength of our study was its prospective design, with regular clinical screening and robust clinical event adjudication that minimized ascertainment bias. Our cohort had detailed baseline data collection with limited missing data, including concise ascertainment of medication use ($83\%$ of the study population maintained their baseline status of metformin use or nonuse throughout the follow-up period), and we were able to adjust for a wide range of demographic, lifestyle, and known risk factors. Randomization of participants to aspirin or placebo enabled us to analyze the effect of aspirin among metformin users while minimizing confounding bias. We were limited by the data available to define diabetes, however. Only a single measure of FBGL was collected at enrollment, and HbA1c was not collected. Therefore, diabetes was defined using a single FBGL measure rather than serial FBGLs or HbA1c. Consequently, the proportion of people with baseline diabetes may be overestimated. Regardless, the total number of participants with diabetes was limited; hence, event numbers in those with metformin exposure was low. This limitation prevented further statistical testing of the effect of metformin and aspirin on cancer by anatomical location. Additionally, we did not capture pre-enrollment diabetes duration (date of diagnosis) nor commencement date of metformin; thus, we could not address the concept of metformin treatment latency effects. In community-dwelling older people with diabetes, metformin use was associated with reduced cancer incidence. Aspirin use was associated with increased cancer mortality risk in metformin users, but the modification effect of metformin and aspirin did not reach statistical significance. Further research is required to understand the relationship among metformin, aspirin, and cancer risk. ## Data availability The data underlying this article cannot be shared because the detail, complexity, and size make them reidentifiable, and privacy of the individuals who participated in the study must be maintained. However, the underlying data can be accessed and analyzed in a secure environment on reasonable request via application through ASPREE.AMS@monash.edu. Applications will be reviewed for scientific merit and successful applicants provided access to participant-level data within a secured data sharing platform. The ASPREE protocol can be publicly accessed via https://aspree.org/usa/wp-content/uploads/sites/$\frac{3}{2021}$/07/ASPREE-Protocol-Version-9_-Nov2014_FINAL.pdf. ## Funding This work was supported by grants U01AG029824 and U19AG062682 from the National Institute on Aging and the National Cancer Institute at the National Institutes of Health, by grants 334047 and 1127060 from the National Health and Medical Research Council of Australia, and by Monash University and the Victorian Cancer Agency. J.E.L. was funded by a Fulbright Postdoctoral Fellowship sponsored by the Australian American Fulbright Commission and funded by RMIT University. ## Conflicts of interest The authors declared no potential conflicts of interest concerning this article’s research, authorship, and/or publication. ## Author contributions Suzanne Gaye Orchard, PhD (Conceptualization; Formal analysis; Project administration; Supervision; Writing—original draft), Jessica E. Lockery, PhD (Conceptualization; Methodology; Software; Supervision; Writing—review & editing), Jonathan C. Broder, MStat&OpRes (Data curation; Formal analysis; Methodology; Visualization; Writing—review & editing), Michael E. Ernst, PharmD (Conceptualization; Methodology; Writing—review & editing), Sara Espinoza, MD (Methodology; Writing—review & editing), Peter Gibbs, MD (Investigation; Writing—review & editing), Rory Wolfe, PhD (Conceptualization; Data curation; Formal analysis; Methodology; Supervision; Writing—review & editing), Galina Polekhina, PhD (Formal analysis; Visualization; Writing—review & editing), Sophia Zoungas, PhD (Methodology; Writing—review & editing), Holli A. Loomans-Kropp, PhD (Methodology; Writing—review & editing), Robyn L. Woods, PhD (Conceptualization; Funding acquisition; Investigation; Methodology; Project administration; Supervision; Writing—review & editing). ## References 1. Fitzmaurice C, Allen C, Barber RM. **Global, regional, and national cancer incidence, mortality, years of life lost, years lived with disability, and disability-adjusted life-years for 32 cancer groups, 1990 to 2015: a systematic analysis for the global burden of disease study**. *JAMA Oncol* (2017.0) **3** 524-548. PMID: 27918777 2. Sung H, Ferlay J, Siegel RL. **Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries**. *CA Cancer J Clin* (2021.0) **71** 209-249. PMID: 33538338 3. Fitzmaurice C, Dicker D, Pain A. **The global burden of cancer 2013**. *JAMA Oncol* (2015.0) **1** 505-527. PMID: 26181261 4. Christensen K, Doblhammer G, Rau R, Vaupel JW.. **Ageing populations: the challenges ahead**. *Lancet* (2009.0) **374** 1196-1208. PMID: 19801098 5. Alejandro EU, Gregg B, Blandino-Rosano M, Cras-Méneur C, Bernal-Mizrachi E.. **Natural history of β-cell adaptation and failure in type 2 diabetes**. *Mol Aspects Med* (2015.0) **42** 19-41. PMID: 25542976 6. Vigneri P, Frasca F, Sciacca L, Pandini G, Vigneri R.. **Diabetes and cancer**. *Endocr Relat Cancer* (2009.0) **16** 1103-1123. PMID: 19620249 7. Giovannucci E, Harlan DM, Archer MC. **Diabetes and cancer: a consensus report**. *Diabetes Care* (2010.0) **33** 1674-1685. PMID: 20587728 8. Li W, Zhang X, Sang H. **Effects of hyperglycemia on the progression of tumor diseases**. *J Exp Clin Cancer Res* (2019.0) **38** 327. PMID: 31337431 9. Huang Y, Cai X, Qiu M. **Prediabetes and the risk of cancer: a meta-analysis**. *Diabetologia* (2014.0) **57** 2261-2269. PMID: 25208757 10. Ma J, Giovannucci E, Pollak M. **A prospective study of plasma C-peptide and colorectal cancer risk in men**. *J Natl Cancer Inst* (2004.0) **96** 546-553. PMID: 15069117 11. Colangelo LA, Gapstur SM, Gann PH, Dyer AR, Liu K.. **Colorectal cancer mortality and factors related to the insulin resistance syndrome**. *Cancer Epidemiol Biomarkers Prev* (2002.0) **11** 385-391. PMID: 11927499 12. **9. Pharmacologic approaches to glycemic treatment:**. *Diabetes Care* (2021.0) **44(suppl 1)** S111-S124. PMID: 33298420 13. Pernicova I, Korbonits M.. **Metformin—mode of action and clinical implications for diabetes and cancer**. *Nat Rev Endocrinol* (2014.0) **10** 143-156. PMID: 24393785 14. Freemark M, Bursey D.. **The effects of metformin on body mass index and glucose tolerance in obese adolescents with fasting hyperinsulinemia and a family history of type 2 diabetes**. *Pediatrics* (2001.0) **107** E55. PMID: 11335776 15. Sola D, Rossi L, Carnevale Schianca GP. **Sulfonylureas and their use in clinical practice**. *Arch Med Sci* (2015.0) **11** 840-848. PMID: 26322096 16. Chen Y, Du L, Li L. **Cancer risk of sulfonylureas in patients with type 2 diabetes mellitus: a systematic review**. *J Diabetes* (2017.0) **9** 482-494. PMID: 27273872 17. Vigneri R, Sciacca L, Vigneri P.. **Rethinking the relationship between insulin and cancer**. *Trends Endocrinol Metab* (2020.0) **31** 551-560. PMID: 32600959 18. Bosetti C, Rosato V, Li D. **Diabetes, antidiabetic medications, and pancreatic cancer risk: an analysis from the International Pancreatic Cancer Case-Control Consortium**. *Ann Oncol* (2014.0) **25** 2065-2072. PMID: 25057164 19. Bao B, Azmi AS, Ali S, Zaiem F, Sarkar FH.. **Metformin may function as anti-cancer agent via targeting cancer stem cells: the potential biological significance of tumor-associated miRNAs in breast and pancreatic cancers**. *Ann Transl Med* (2014.0) **2** 59. PMID: 25333034 20. Tseng CH.. **Metformin and lung cancer risk in patients with type 2 diabetes mellitus**. *Oncotarget* (2017.0) **8** 41132-41142. PMID: 28456789 21. Tseng CH.. **Metformin reduces ovarian cancer risk in Taiwanese women with type 2 diabetes mellitus**. *Diabetes Metab Res Rev* (2015.0) **31** 619-626. PMID: 25820555 22. Tsai MJ, Yang CJ, Kung YT. **Metformin decreases lung cancer risk in diabetic patients in a dose-dependent manner**. *Lung Cancer* (2014.0) **86** 137-143. PMID: 25267165 23. Kim YI, Kim SY, Cho SJ. **Long-term metformin use reduces gastric cancer risk in type 2 diabetics without insulin treatment: a nationwide cohort study**. *Aliment Pharmacol Ther* (2014.0) **39** 854-863. PMID: 24612291 24. Wan G, Sun X, Li F. **Survival benefit of metformin adjuvant treatment for pancreatic cancer patients: a systematic review and meta-analysis**. *Cell Physiol Biochem* (2018.0) **49** 837-847. PMID: 30184547 25. Shi YQ, Zhou XC, Du P. **Relationships are between metformin use and survival in pancreatic cancer patients concurrent with diabetes**. *Medicine (Baltimore)* (2020.0) **99** e21687. PMID: 32925714 26. Roshan MH, Shing YK, Pace NP.. **Metformin as an adjuvant in breast cancer treatment**. *SAGE Open Med* (2019.0) **7** 2050312119865114. PMID: 31360518 27. Mamtani R, Pfanzelter N, Haynes K. **Incidence of bladder cancer in patients with type 2 diabetes treated with metformin or sulfonylureas**. *Diabetes Care* (2014.0) **37** 1910-1917. PMID: 24496803 28. Azoulay L, Dell’Aniello S, Gagnon B, Pollak M, Suissa S.. **Metformin and the incidence of prostate cancer in patients with type 2 diabetes**. *Cancer Epidemiol Biomarkers Prev* (2011.0) **20** 337-344. PMID: 21148757 29. Smiechowski B, Azoulay L, Yin H, Pollak MN, Suissa S.. **The use of metformin and colorectal cancer incidence in patients with type II diabetes mellitus**. *Cancer Epidemiol Biomarkers Prev* (2013.0) **22** 1877-1883. PMID: 23966577 30. Smiechowski BB, Azoulay L, Yin H, Pollak MN, Suissa S.. **The use of metformin and the incidence of lung cancer in patients with type 2 diabetes**. *Diabetes Care* (2013.0) **36** 124-129. PMID: 22923670 31. Lega IC, Austin PC, Gruneir A, Goodwin PJ, Rochon PA, Lipscombe LL.. **Association between metformin therapy and mortality after breast cancer: a population-based study**. *Diabetes Care* (2013.0) **36** 3018-3026. PMID: 23633525 32. Margel D, Urbach DR, Lipscombe LL. **Metformin use and all-cause and prostate cancer-specific mortality among men with diabetes**. *J Clin Oncol* (2013.0) **31** 3069-3075. PMID: 23918942 33. **10. Cardiovascular disease and risk management:**. *Diabetes Care* (2021.0) **44(suppl 1)** S125-S150. PMID: 33298421 34. Rothwell PM, Wilson M, Elwin CE. **Long-term effect of aspirin on colorectal cancer incidence and mortality: 20-year follow-up of five randomised trials**. *Lancet* (2010.0) **376** 1741-1750. PMID: 20970847 35. Rothwell PM, Fowkes FGR, Belch JFF, Ogawa H, Warlow CP, Meade TW.. **Effect of daily aspirin on long-term risk of death due to cancer: analysis of individual patient data from randomised trials**. *Lancet* (2011.0) **377** 31-41. PMID: 21144578 36. Rothwell PM, Wilson M, Price JF, Belch JFF, Meade TW, Mehta Z.. **Effect of daily aspirin on risk of cancer metastasis: a study of incident cancers during randomised controlled trials**. *Lancet* (2012.0) **379** 1591-1601. PMID: 22440947 37. McNeil JJ, Nelson MR, Woods RL. **Effect of aspirin on all-cause mortality in the healthy elderly**. *N Engl J Med* (2018.0) **379** 1519-1528. PMID: 30221595 38. McNeil JJ, Gibbs P, Orchard SG. **Effect of aspirin on cancer incidence and mortality in older adults**. *J Natl Cancer Inst* (2021.0) **113** 258-265. PMID: 32778876 39. Bowman L, Mafham M, Wallendszus K. **Effects of aspirin for primary prevention in persons with diabetes mellitus**. *N Engl J Med* (2018.0) **379** 1529-1539. PMID: 30146931 40. Orchard SG, Lockery JE, Gibbs P. **Cancer history and risk factors in healthy older people enrolling in the ASPREE clinical trial**. *Contemp Clin Trials* (2020.0) **96** 106095. PMID: 32739494 41. **Study design of ASPirin in Reducing Events in the Elderly (ASPREE): a randomized, controlled trial**. *Contemp Clin Trials* (2013.0) **36** 555-564. PMID: 24113028 42. Lockery JE, Collyer TA, Abhayaratna WP. **Recruiting general practice patients for large clinical trials: lessons from the Aspirin in Reducing Events in the Elderly (ASPREE) study**. *Med J Aust* (2019.0) **210** 168-173. PMID: 30835844 43. McNeil JJ, Woods RL, Nelson MR. **Effect of aspirin on disability-free survival in the healthy elderly**. *N Engl J Med* (2018.0) **379** 1499-1508. PMID: 30221596 44. McNeil JJ, Wolfe R, Woods RL. **Effect of aspirin on cardiovascular events and bleeding in the healthy elderly**. *N Engl J Med* (2018.0) **379** 1509-1518. PMID: 30221597 45. Lockery JE, Collyer TA, Reid CM. **Overcoming challenges to data quality in the ASPREE clinical trial**. *Trials* (2019.0) **20** 686. PMID: 31815652 46. 46ATC/DDD index 2023. WHO Collaborating Centre for Drug Statistics Methodology. Updated January 23, 2023. AccessedOctober 3, 2019. https://www.whocc.no/atc_ddd_index/ 47. Lockery JE, Rigby J, Collyer TA. **Optimising medication data collection in a large-scale clinical trial**. *PLoS One* (2019.0) **14** e0226868. PMID: 31881040 48. **2. Classification and diagnosis of diabetes:**. *Diabetes Care* (2020.0) **44(suppl 1)** S15-S33 49. Grambsch PM, Therneau TM.. **Proportional hazards tests and diagnostics based on weighted residuals**. *Biometrika* (1994.0) **81** 515-526 50. Shaw RJ, Lamia KA, Vasquez D. **The kinase LKB1 mediates glucose homeostasis in liver and therapeutic effects of metformin**. *Science* (2005.0) **310** 1642-1646. PMID: 16308421 51. Salvatore T, Marfella R, Rizzo MR, Sasso FC.. **Pancreatic cancer and diabetes: a two-way relationship in the perspective of diabetologist**. *Int J Surg* (2015.0) **21(suppl 1)** S72-S77. PMID: 26123386 52. Kisfalvi K, Eibl G, Sinnett-Smith J, Rozengurt E.. **Metformin disrupts crosstalk between G protein–coupled receptor and insulin receptor signaling systems and inhibits pancreatic cancer growth**. *Cancer Res* (2009.0) **69** 6539-6545. PMID: 19679549 53. Karnevi E, Said K, Andersson R, Rosendahl AH.. **Metformin-mediated growth inhibition involves suppression of the IGF-I receptor signalling pathway in human pancreatic cancer cells**. *BMC Cancer* (2013.0) **13** 235. PMID: 23663483 54. Rozengurt E, Sinnett-Smith J, Kisfalvi K.. **Crosstalk between insulin/insulin-like growth factor-1 receptors and g protein-coupled receptor signaling systems: a novel target for the antidiabetic drug metformin in pancreatic cancer**. *Clin Cancer Res* (2010.0) **16** 2505-2511. PMID: 20388847 55. Broadhurst PJ, Hart AR.. **Metformin as an adjunctive therapy for pancreatic cancer: a review of the literature on its potential therapeutic use**. *Dig Dis Sci* (2018.0) **63** 2840-2852. PMID: 30159732 56. Jin K.. **Does coronary microvascular dysfunction have a role in cardiovascular oncology?**. *Eur J Prev Cardiol* (2023.0) **30** 206-208. PMID: 36200324 57. Toya T, Sara JD, Corban MT. **Assessment of peripheral endothelial function predicts future risk of solid-tumor cancer**. *Eur J Prev Cardiol* (2020.0) **27** 608-618. PMID: 31668110 58. Bratusch-Marrain PR, Vierhapper H, Komjati M, Waldhäusl WK.. **Acetyl-salicylic acid impairs insulin-mediated glucose utilization and reduces insulin clearance in healthy and non-insulin-dependent diabetic man**. *Diabetologia* (1985.0) **28** 671-676. PMID: 3905478 59. Newman WP, Brodows RG.. **Aspirin causes tissue insensitivity to insulin in normal man**. *J Clin Endocrinol Metab* (1983.0) **57** 1102-1106. PMID: 6355137 60. Giugliano D, Sacca L, Scognamiglio G, Ungaro B, Torella R.. **Influence of acetylsalicylic acid on glucose turnover in normal man**. *Diabete Metab* (1982.0) **8** 279-282. PMID: 6819169 61. Hundal RS, Petersen KF, Mayerson AB. **Mechanism by which high-dose aspirin improves glucose metabolism in type 2 diabetes**. *J Clin Invest* (2002.0) **109** 1321-1326. PMID: 12021247 62. Bailey CJ.. **Metformin: historical overview**. *Diabetologia* (2017.0) **60** 1566-1576. PMID: 28776081 63. Liu EY, Al-Sofiani ME, Yeh HC, Echouffo-Tcheugui JB, Joseph JJ, Kalyani RR.. **Use of preventive aspirin among older US adults with and without diabetes**. *JAMA Netw Open* (2021.0) **4** e2112210. PMID: 34152419
--- title: 'Prevalence of characteristics associated with sarcopenia in elders: a cross-sectional study' authors: - Caroline Ribeiro de Sousa - Janaína Fonseca Victor Coutinho - Marília Braga Marques - Rachel Gabriel Bastos Barbosa - Jarbas de Sá Roriz - Edson Silva Soares - Charlys Barbosa Nogueira - Rodrigo Lopes de Paula Souza journal: Revista Brasileira de Enfermagem year: 2023 pmcid: PMC10042476 doi: 10.1590/0034-7167-2022-0209 license: CC BY 4.0 --- # Prevalence of characteristics associated with sarcopenia in elders: a cross-sectional study ## ABSTRACT ### Objectives: to identify the prevalence and characteristics associated with sarcopenia in elders in Primary Health Care Units. ### Methods: cross-sectional study with 384 elders. To evaluate sarcopenia, we measured: strength and muscle mass, and physical performance. The elderly were classified as having: probable sarcopenia; sarcopenia; or severe sarcopenia. The chi-squared test and the multinomial logistic regression method were used. ### Results: the prevalence of probable sarcopenia was $25.52\%$; of sarcopenia, $11.98\%$; and of severe sarcopenia, $9.90\%$. Probable sarcopenia is 1.75 times more prevalent in men; osteoporosis is 2.16 times more prevalent in people with severe sarcopenia; polypharmacy is 1.57 times more likely in individuals with probable sarcopenia; and calf circumference below 31 cm is 2.24 times more likely in patients with sarcopenia and 2.19 times more likely in patients with severe sarcopenia. ### Conclusions: the highest prevalence was of probable sarcopenia, and the characteristics associated with sarcopenia were sex, osteoporosis, polypharmacy, overweight, obesity, and calf circumference. ## Objetivos: identificar a prevalência e as características associadas à sarcopenia em pessoas idosas de Unidades de Atenção Primária à Saúde. identificar prevalencia y características relacionadas a la sarcopenia en personas ancianas de Unidades de Atención Primaria de Salud. ## Métodos: estudo transversal, com 384 pessoas idosas. Para avaliação de sarcopenia, mediu-se: força e massa muscular, desempenho físico. Classificaram se pessoas idosas com: sarcopenia provável; sarcopenia; e sarcopenia grave. Analisou-se com teste de qui quadrado e método de regressão logística multinomial. estudio transversal, con 384 personas ancianas. Para evaluación de sarcopenia, medidos: fuerza y masa muscular, desempeño físico. Clasificadas personas ancianas con: sarcopenia probable; sarcopenia; y sarcopenia grave. Analizado con prueba chi cuadrado y método de regresión logística multinomial. ## Resultados: a prevalência de provável sarcopenia foi de 25,$52\%$; sarcopenia, 11,$98\%$; e sarcopenia grave, 9,$90\%$. Homens são 1,75 vez mais prevalentes em indivíduos com provável sarcopenia; osteoporose é 2,16 vezes mais prevalente na sarcopenia grave; polifarmácia, 1,57 vez mais prevalente na provável sarcopenia; circunferência da panturrilha menor que 31 cm é 2,24 vezes mais prevalente na sarcopenia e 2,19 vezes na sarcopenia grave. la prevalencia de probable sarcopenia fue de 25,$52\%$; sarcopenia, 11,$98\%$; y sarcopenia grave, 9,$90\%$. Hombres son 1,75 vez más prevalentes en individuos con probable sarcopenia; osteoporosis es 2,16 veces más prevalente en la sarcopenia grave; polifarmacia, 1,57 vez más prevalente en la probable sarcopenia; circunferencia de los gemelos menor que 31 cm es 2,24 veces más prevalente en la sarcopenia y 2,19 veces en la sarcopenia grave. ## Conclusões: houve maior prevalência de provável sarcopenia, e as características associadas à sarcopenia foram: sexo, osteoporose, polifarmácia, sobrepeso, obesidade e circunferência da panturrilha. ## Conclusiones: Hubo mayor prevalencia de probable sarcopenia, y las características relacionadas a la sarcopenia fueron: sexo, osteoporosis, polifarmacia, sobrepeso, obesidad y circunferencia de los gemelos. ## INTRODUCTION The loss of muscle mass has been studied for nearly 30 years, receiving the name of sarcopenia[1]. However, in 2016 it was recognized as a muscle disease (ICD-10-MC-M62.84), characterized by the loss of strength and muscle amount. Its causes are multifactorial and involve: aging, genetics, hormone and muscle tissue alterations, neurological decline, increased levels of pro-inflammatory cytokines, and mitochondrial dysfunctions[2-4]. Worldwide, the prevalence of sarcopenia may vary from $3\%$ to $86.5\%$[5]. In Brazil, its prevalence is of $15.4\%$, albeit with differences between cities. In Florianópolis, its prevalence is $33.3\%$; in São Paulo, $4.8\%$; in Salvador, $17.8\%$; and in Natal $10.7\%$[6-9]. This variation is due to ethnicity, place of residence (urban or rural), area researched (community, hospital, outpatient clinic, or long-permanence institution), instruments, methods, and cutoff points for diagnosis[10-11]. Studies indicate that some characteristics, such as age, sex, level of physical activity, and the presence of chronic diseases are associated with the presence of sarcopenia[3,10]. Nonetheless, it is still necessary to understand its many possible causes[12]. Furthermore, when it is not treated, this condition has severe personal, social, and economic tolls, due to the fact it impairs daily-life activities and leads to lower functional capabilities, falls, fractures, institutionalization, hospitalization, and death[13]. In this context, recognizing the issue and intervening as soon as possible leads to better outcomes in sarcopenia patients. Consequently, the importance of evaluations in Primary Health *Care is* evident, as this level of care is the one responsible for stratifying, screening, embracing, developing actions, and ensuring integral and continuous care to the elderly[14]. Considering the above, increasing our understanding about this disease and raising awareness about its characteristics in different contexts is essential to develop diagnostic possibilities and interventions that can prevent it and promote health, which, in turn, will lead to better care and more quality of life for the elderly. ## OBJECTIVES To identify the prevalence and characteristics associated with sarcopenia in the elderly of Primary Health Care Units. ## Ethical aspects This study followed all ethical precepts. Its protocol was approved by the Research Ethics Committee in 2018. ## Design, period, and place of study Cross-sectional epidemiological study. The guidelines of the EQUATOR network were followed through the use of the tool Strengthening the Reporting of Observational Studies in Epidemiology (STROBE)[15]. The study was carried out from April 2018 to June 2019, with elders being attended in six Primary Health Care Units (PHCUs) in Fortaleza, a city in the state of Ceará (CE). ### criteria for inclusion and exclusion The study population was formed by 105,833 elders registered in the PHCUs in Fortaleza. The sample was calculated using the formula of cross-sectional studies with infinite populations, a population proportion of $50\%$, $5\%$ error, and confidence interval of $95\%$, to a total of 384 elders. Fortaleza is divided in six Regional Secretariats (RS), and we chose the metho d of stratified sampling due to the heterogeneous subdivision of elders among them. The PHCUs that attended to the highest number of elders in each secretariat were chosen, and the value from the sampling calculation was divided according with the percentage of elders registered in each, according to data made available by the Fortaleza Health Secretariat (SR I, $25\%$; SR II, $13\%$; SR III, $11\%$; SR IV, $9\%$; SR V, $27\%$; e SR VI, $15\%$). Elders who went to the unit randomly were invited to participate in the research. Individuals aged 60 years or older who received attention in the PHCUs were included. Elders diagnosed with dementia according to their companions or to medical reports were excluded. 414 elders were recruited. 5 of them were excluded due to dementia diagnosis, and 25 instruments were not filled in properly by the researchers. As a result, the study counted on the participation of 384 people. ## Study protocol Sociodemographic and clinical data were collected using a structured instrument in the form of a self-report. We investigated: age, sex, educational level, income in minimum wages (R$ 954.00 in 2018 and R$ 998.00 in 2019), retirement, marital status, housing, physical activity, number of falls in the last 12 months, drinking, smoking, comorbidities (hypertension, diabetes, cancer, osteoarthritis, cardiopathy, chronic kidney disease, osteoporosis, dyslipidemia, depression, anxiety, Parkinson’s disease, glaucoma, hypothyroidism, and schizophrenia Anthropometric evaluations were carried out by measuring the weight, height, and the Body Mass Index (BMI), using the classification criteria determined by the Pan-American Health Organization[16]. The calf circumference was evaluated, with values below 31 cm being indicative of decrease in muscle mass. Sarcopenia was evaluated using criteria from the European Working Group on Sarcopenia in Older People 2 (EWGSOP2), which determine the following classification: probable sarcopenia, when the only symptom is low muscle strength; sarcopenia, when low muscle quantity/quality is confirmed; and severe sarcopenia, when it is possible to detect low muscle strength, low muscle quantity/quality, and low physical performance[3]. The most commonly used method to measure physical performance is the measurement of gait speed in the 10-Meter Test[17], whose cutoff point indicative of lower physical performance is 0.8 m/s[3]. Grip strength was measured using a Jamar hydraulic dynamometer adjusted to level 2, a level in which the grip strength performance is the highest[18]. The cutoff points vary according with gender, with 27kgf for men and 16 kgf for women. Lower values indicate low muscle strength[3]. Among methods available to evaluate muscle mass, we chose using the anthropometric equation to calculate Total Muscle Mass (TMM). This equation was developed, validated, and compare with body composition evaluation results, as calculated using Dual Energy X-Ray Absorptiometry (DEXA). The DEXA is considered to be the most recommended method. However, it is very costly and requires specialized professionals and equipment, meaning it is not accessible to all levels of health[19]. The MMT (kg) is established using the formula: TMM = (0.244 × body mass) + (7.8 × stature) - (0.098 × age) + (6.6 × sex) + (ethnicity - 3.3). For the variable sex: 0 = women, 1 = men; for self-referred ethnicity, which as categorized later, the following values were adopted: 0=white (white, mixed, and native), -1,2 = Asian; and -1.4 = African ascent (black and brown)[19]. According with TMM, the Muscle Mass Index was calculated [MMI = TMM/stature 2]. Later, it was classified according with the cutoff points proposed by European Consensus: men < 7.0 kg/m2, and women < 5.5 kg/m2[3,19]. ## Analysis of results and statistics For data analysis, at first, we chose to describe predictor variables and outcomes, using absolute and relative frequencies. The normality of data was analyzed using the Kolmogorov-Smirnov test, using medians and interquartile amplitudes for age and educational level. After the variables were described, we analyzed the association between sociodemographic/clinical characteristics and the outcome “sarcopenia” using the chi-squared test, considering as significant associations where $p \leq 0.005.$ For the multivariate analysis, all variables where $p \leq 0.20$ in the bivariate analysis were considered[20]. Furthermore, we applied the multinomial logistic regression model, since the distribution of the outcome has four categories[21]. This type of regression allows us to verify the association of each outcome category. In this research, the “no sarcopenia” category was adopted as a reference. Furthermore, it is important to highlight that logistic regression results are presented as odds ratio (OR). Nonetheless, since this is a prevalence study, the delta method was used, which converts the OR to prevalence ratio (PR) of adjusted variables in the final model. Therefore, the effect size was found using PR, and the association strength was found using the confidence interval of $95\%$ (CI$95\%$). We considered associations where $p \leq 0.05$ were significant. All analyses were carried out using the software Stata 13. ## RESULTS Sociodemographic variables showed a median age of 69 years (interquartile interval of 10), with the age group from 60 t0 79 years as predominant ($87.5\%$; $$n = 336$$). Most participants were female ($67.5\%$; $$n = 255$$), income of up to one minimum wage ($55.5\%$; $$n = 213$$), retired ($71.6\%$; $$n = 275$$), with no partner ($65.4\%$; $$n = 251$$) and lived with family ($77.3\%$; $$n = 297$$). The educational level median was 5 years (interquartile interval of 8), with 66.4 ($$n = 255$$) having up to 8 years of study and $10.2\%$ ($$n = 39$$) illiterate. Clinical data showed that $53.4\%$ ($$n = 205$$) did not practice physical activity, $11.2\%$ ($$n = 43$$) drank, $9.1\%$ ($$n = 35$$) smoked, $47.7\%$ ($$n = 183$$) had fallen in the last 12 months. From these, $28.6\%$ ($$n = 110$$) had fallen twice or more in this period. Polypharmacy was present in $24.5\%$ ($$n = 94$$) of participants. According to the Body Mass Index (BMI) $25.2\%$ ($$n = 97$$) had low weight, $39.3\%$ ($$n = 44$$) normal weight, and $24\%$ ($$n = 92$$) were obese. Regarding their calf circumference, $12.8\%$ ($$n = 49$$) of participants had it below 31cm. Regarding comorbidities, $64.8\%$ ($$n = 249$$) had hypertension; $39.8\%$, ($$n = 153$$) diabetics; $23.4\%$ ($$n = 80$$), osteoarthritis; $24.7\%$ ($$n = 95$$), osteoporosis; $12.5\%$ ($$n = 48$$), dyslipidemia; $10.5\%$ ($$n = 39$$), cardiovascular disease; $4.9\%$ ($$n = 19$$), cancer; $4.9\%$ ($$n = 19$$), hypothyroidism; $3.6\%$ ($$n = 14$$), anxiety; e $3.1\%$ ($$n = 12$$), depression. Table 1 shows the descriptive data related with the criteria established by the EWGSOP2 to evaluate sarcopenia. **Table 1** | Variable | Unnamed: 1 | Category | n(%) | Mean (DP ±) | Min. - Max. | | --- | --- | --- | --- | --- | --- | | Grip Strength (dynamometer) | Female | < 16 kgf ≥ a 16 kgf | 122 (31.8)142 (37) | 19.33 (±7.14) | 2 - 46.5 | | Grip Strength (dynamometer) | Male | < 27 kgf ≥ 27kgf | 62 (16.1)58 (15.1) | 19.33 (±7.14) | 2 - 46.5 | | Body Mass Index | Female | < 5.5 kg/m2 ≥ 5.5 kg/m2 | 126 (32.3)138 (35.9) | 6.46 (±1.39) | 3.54 - 13.51 | | | Male | < 7 kg/m2 ≥ 7 kg/m2 | 19 (4.9)101 (26.8) | 6.46 (±1.39) | 3.54 - 13.51 | | Gait Speed Test | | ≤ 0.8 m/s | 140 (36.5) | 7.31s (±2.16) | 3.98 - 28s | | Gait Speed Test | | > 0.8 m/s | 244 (63.5) | 7.31s (±2.16) | 3.98 - 28s | Figure 1 shows the evaluation criteria and the prevalence of sarcopenia in elders registered in the PHCUs. Figure 1Criteria for the evaluation of the prevalence of sarcopenia in elders attended in the Primary Health Care Units ($$n = 384$$), Fortaleza, Ceará, Brazil, 2019 A bivariate analysis using sociodemographic and clinical data was carried out, with the possible outcomes no sarcopenia, probable sarcopenia, sociodemographic, and severe sarcopenia. Its main findings are presented in Table 2. **Table 2** | Variables | SARCOPENIA | SARCOPENIA.1 | SARCOPENIA.2 | SARCOPENIA.3 | SARCOPENIA.4 | | --- | --- | --- | --- | --- | --- | | Variables | No sarcopenia (%) | Probable sarcopenia (%) | Sarcopenia (%) | Severe sarcopenia (%) | p * | | Age | | | | | 0.377 | | 80 years or older | 24 (50.0) | 10 (20.8) | 6 (12.5) | 8 (16.7) | | | 60 to 79 years | 178 (53.0) | 88 (26.2) | 40 (11.9) | 30 (8.9) | | | Sex | | | | | 0.005 | | Male | 64 (51.2) | 44 (35.2) | 8 (6.4) | 9 (7.2) | | | Female | 138 (53.3) | 54 (20.8) | 38 (14.7) | 29 (11.2) | | | Years of formal education | | | | | 0.437 | | 9 years or more | 71 (55.0) | 36 (27.9) | 12 (9.3) | 10 (7.7) | | | Up to 8 years | 131 (51.4) | 62 (24.3) | 34 (13.3) | 28 (11.0) | | | Marital Status | | | | | 0.020 | | Has a partner | 70 (52.6) | 40 (30.1) | 18 (13.5) | 5 (3.8) | | | Does not have a partner | 132 (52.6) | 58 (23.1) | 28 (11.2) | 33 (11.1) | | | Physical exercise | | | | | 0.128 | | Yes | 100 (55.9) | 38 (21.2) | 26 (14.5) | 15 (8.4) | | | No | 102 (49.8) | 60 (29.2) | 20 (9.8) | 23 (11.2) | | | Hypertension | | | | | 0.187 | | Yes | 121 (48.6) | 70 (28.1) | 31 (12.5) | 27 (10.8) | | | No | 81 (60.0) | 20 (20.7) | 15 (11.1) | 11 (8.2) | | | Diabetes | | | | | 0.049 | | Yes | 68 (44.4) | 48 (31.4) | 22 (14.4) | 15 (9.8) | | | No | 134 (58.0) | 50 (21.6) | 24 (10.4) | 23 (10.0) | | | Osteoarthritis | | | | | 0.006 | | Yes | 39 (43.3) | 21 (23.3) | 13 (14.5) | 17 (18.9) | | | No | 163 (55.5) | 77 (26.2) | 33 (11.2) | 21 (7.1) | | | Osteoporosis | | | | | 0.166 | | Yes | 43 (45.3) | 24 (25.3) | 14 (14.7) | 14 (174.7) | | | No | 159 (55.0) | 74 (25.6) | 32 (11.1) | 24 (8.3) | | | Dyslipidemia | | | | | 0.012 | | Yes | 23 (47.9) | 7 (14.6) | 12 (25.0) | 6 (12.5) | | | No | 179 (53.3) | 91 (27.1) | 34 (10.1) | 32 (9.5) | | | Polypharmacy | | | | | < 0.001 | | Five or more medications | 31 (33.0) | 31 (33.0) | 16 (17.0) | 16 (17.0) | | | Up to four medications | 171 (59.0) | 67 (23.1) | 30 (10.3) | 22 (7.6) | | | Body Mass Index | | | | | < 0.001 | | Low weight | 48 (49.5) | 9 (9.3) | 19 (19.6) | 21 (21.6) | | | Regular | 80 (53.0) | 35 (23.2) | 22 (14.6) | 14 (9.2) | | | Overweight | 22 (50.0) | 17 (38.6) | 9 (9.1) | 1 (2.3) | | | Obese | 52 (56.5) | 37 (40.2) | 1 (1.1) | (2.2) | | | Calf | | | | | < 0.001 | | < 31 | 14 (28.6) | 8 (16.3) | 13 (26.5) | 14 (28.6) | | | ≥ 31 | 188 (56.1) | 90 (26.9) | 33 (9.8) | 24 (7.2) | | The proportion of non-sarcopenic elders was similar between the sexes. Nonetheless, males showed $35.2\%$ ($$n = 44$$) of probable sarcopenia cases, $6.4\%$ ($$n = 8$$) of sarcopenia cases, and $7.2\%$ ($$n = 9$$) of severe sarcopenia cases ($$p \leq 0.0005$$). There was an equal distribution of elders with and without partners among non-sarcopenic elders, but the difference was larger among those with severe sarcopenia. The prevalence of severe sarcopenia was $11.1\%$ ($$n = 33$$) in those with no partners and $3.8\%$ ($$n = 5$$) in those with partners ($$p \leq 0.020$$). Most elders with some degree of sarcopenia received up to one minimum wage ($$p \leq 0.437$$), being: $27.7\%$ ($$n = 59$$) of those with probable sarcopenia, $13.6\%$ ($$n = 29$$) of those with sarcopenia, and $10.3\%$ ($$n = 22$$), of those with severe sarcopenia. From those who lived alone, $16.5\%$ ($$n = 14$$) had probable sarcopenia; $11.8\%$ [10] had sarcopenia; and $10.6\%$ ($$n = 9$$) had severe sarcopenia ($$p \leq 0.435$$). Regarding the number of falls, there was no association ($$p \leq 0.202$$). The prevalence of elders who had fallen twice or more in the last year was $45.5\%$ ($$n = 50$$), among the non-sarcopenic, $31.8\%$ ($$n = 35$$) in the probable sarcopenic, $10.9\%$ ($$n = 12$$) in the sarcopenic, and $11.8\%$ ($$n = 13$$) in the severe sarcopenic. Regarding drinking, its prevalence was higher in non-sarcopenic elders ($67.4\%$; $$n = 29$$), while $20.9\%$ ($$n = 9$$) had probable sarcopenia ($$p \leq 0.171$$). The prevalence of smokers was also higher in the non-sarcopenic ($45.7\%$, $$n = 16$$), while $28.6\%$ ($$n = 10$$) had probable sarcopenia, $17.1\%$ ($$n = 6$$) sarcopenia, and $8.6\%$ ($$n = 3$$) had severe sarcopenia ($$p \leq 0.701$$). Regarding comorbidities, the prevalence of elders with osteoarthritis was higher among the sarcopenic ($14.5\%$; $$n = 13$$) and severe sarcopenic ($18.9\%$; $$n = 17$$) ($$p \leq 0.006$$). Similarly, the proportion of dyslipidemia was lower in the non-sarcopenic ($47.9\%$; $$n = 23$$) and probable sarcopenic ($14.6\%$; $$n = 7$$), while in the sarcopenic, the proportion was $25\%$ ($$n = 12$$), and, in the severe sarcopenic, it was $12.5\%$ ($$n = 6$$) ($$p \leq 0.012$$). Polypharmacy was more frequent among the sarcopenic ($33\%$; $$n = 31$$), the probable sarcopenic ($17\%$; $$n = 16$$), and the severe sarcopenic ($17\%$; $$n = 16$$) ($p \leq 0.001$). Regarding BMI, all categories had a higher number of non-sarcopenic elders. Among low weight elders, most had severe sarcopenia (21,$6\%$; $$n = 21$$) and sarcopenia ($19.6\%$; $$n = 19$$); in elders with regular weight, most were probable sarcopenic ($23.2\%$; $$n = 35$$) and sarcopenic ($14.6\%$; $$n = 22$$). Among overweight and obese elders, most were probable sarcopenic: $38.6\%$ ($$n = 17$$) and $40.2\%$ ($$n = 37$$), respectively. Finally, elders whose calf circumference was below 31 were more often the severe sarcopenic ($28.6\%$; $$n = 14$$) and sarcopenic ($26.5\%$; $$n = 13$$) ($p \leq 0.001$). Table 3 shows the prevalence ratio of the sarcopenia outcome, after a multivariate analysis. **Table 3** | Variables | Sarcopenia * PR (IC95%) | Sarcopenia * PR (IC95%).1 | Sarcopenia * PR (IC95%).2 | Sarcopenia * PR (IC95%).3 | Sarcopenia * PR (IC95%).4 | Sarcopenia * PR (IC95%).5 | | --- | --- | --- | --- | --- | --- | --- | | Variables | Probable sarcopenia | p | Sarcopenia | p | Severe sarcopenia | p | | Sex (male) | 1.75 (1.25 - 2.46) | 0.001 | 0.50 (0.24 - 1.08) | 0.077 | 1.14 (0.60 - 2.17) | 0.679 | | Marital status (has a partner) | 0.78 (0.56 - 1.09) | 0.146 | 1.13 (0.63 - 2.01) | 0.678 | 1.97 (0.88 - 4.38) | 0.097 | | Physical exercise (yes) | 1.05 (0.75 - 1.47) | 0.796 | 1.53 (0.91 - 2.59) | 0.110 | 0.39 (0.17 - 0.92) | 0.031 | | Drinks (yes) | 0.77 (0.56 - 1.08) | 0.126 | 1.48 (0.87 - 2.51) | 0.145 | 0.62 (0.35 - 1.09) | 0.100 | | Hypertension (yes) | 0.65 (0.36 - 1.18) | 0.162 | 0.51 (0.13 - 1.93) | 0.320 | 1.00 (0.39 - 2.56) | 0.996 | | Diabetes (yes) | 0.91 0.74 - 1.120 | 0.384 | 1.05 (0.72 - 1.55) | 0.710 | 1.10 (0.60 - 2.01) | 0.705 | | Osteoarthritis (yes) | 1.39 (0.99 - 1.96) | 0.060 | 0.96 (0.55 - 1.67) | 0.873 | 0.70 (0.38 - 1.27) | 0.241 | | Osteoporosis (yes) | 0.96 (0.63 - 1.48) | 0.869 | 1.02 (0.54 - 1.93) | 0.948 | 2.16 (1.17 - 3.97) | 0.013 | | Dyslipidemia (yes) | 0.48 (0.24 - 0.98) | 0.043 | 2.05 (1.13 - 3.73) | 0.018 | 0.83 (0.37 - 1.86) | 0.650 | | Polypharmacy (≥ 5 medications) | 1.57 (1.10 - 2.56) | 0.014 | 1.17 (0.62 - 2.22) | 0.619 | 1.61 (0.86 - 3.05) | 0.136 | | Body Mass Index | | | | | | | | Low weight | 0.46 (0.24 - 0.90) | 0.023 | 1.34 (0.76 - 2.38) | 0.312 | 1.66 (0.89 - 3.10) | 0.109 | | Regular | 1 | - | 1 | - | 1 | - | | Overweight | 1.78 (1.20 - 2.61) | 0.003 | 0.91 (0.37 - 2.32) | 0.832 | 0.35 (0.06 - 2.02) | 0.242 | | Obesity | 1.72 (1.19 - 2.50) | 0.003 | 0.12 (0.02 - 0.88) | 0.037 | 0.29 (0.08 - 1.09) | 0.067 | | Calf (< 31) | 0.98 (0.53 - 1.80) | 0.936 | 2.24 (1.23 - 4.07) | 0.009 | 2.19 (1.16 - 4.17) | 0.016 | Males are 1.75 times (CI$95\%$: 1.25 - 2.46) more likely to be probable sarcopenic as opposed to non-sarcopenic. Regarding the other outcomes, there was no difference regarding gender. When it comes to physical activities, elders who practice some form of exercise had a lower prevalence of severe sarcopenia (PR: 0.39; CI$95\%$:0.17 - 0.92). Sarcopenic elders were 2.05 times (CI$95\%$: 1.13-3.73) more likely to show dyslipidemia. Osteoporosis was 2.16 times (CI$95\%$:1.17-3.97) more prevalent in those with severe sarcopenia. Polypharmacy, in turn, increased the prevalence of probable sarcopenia in 1.57 (1.10-2.56). Regarding BMI, the prevalence of overweight elders was 1.78 times (CI$95\%$: 1.20-2.61) higher in those with probable sarcopenia. Finally, obesity was 1.72 (CI$95\%$:1.19-2.50) times more prevalent in elders with probable sarcopenia. Finally, calf circumferences lower than 31 cm increased the prevalence of sarcopenia in 2.24 times (CI$95\%$:1.23 - 4.07) and the prevalence of severe sarcopenia in 2.19 times (CI$95\%$:1.16 - 4.17). ## DISCUSSION The prevalence of probable sarcopenia, sarcopenia, and severe sarcopenia are similar to that found in other Brazilian cities and in international studies[22-24], albeit the lack of a single classification system makes it harder to find more robust information[5]. Finding a consensual diagnosis would facilitate not only research, but also the discovery of treatment options and the translation of the investigation results into practice[10]. To do so, professionals must be able to incorporate evaluation actions to be able to determine adequate interventions[25]. The methods used in this study can be used by primary health care workers - especially by the nurse, due to their relevant role in identifying the attention needs of individuals in primary care, as well as in health promotion and protection[26]. In this study, males were more likely to develop probable sarcopenia. Further studies also found higher prevalence and risk for sarcopenia in men[27-28]. Still, in general, literature states that being female is the risk factor for sarcopenia[29], considering that, starting with 50 years, the loss of strength in women is hastened due to hormonal changes in their non-reproductive stages[30]. Nonetheless, primary care is directed towards children and women. This feminine environment causes feelings of invulnerability in men, which in turn leads them to seek mostly emergency services and specialized consultations when they lose functionality[31-32]. That said, the findings of this study open space for new research, specifically with male elders. It was found that dyslipidemia is associated with probable sarcopenia and sarcopenia. Studies show associations between dyslipidemia and the development of sarcopenia[33-34]. Nonetheless, pathologic mechanisms are still unknown. It has been suggested that higher fat levels cause inflammatory cytokines, such as tumor necrosis factor alpha and interleukins, to be secreted, reducing muscle tissue[33-34]. The association of physical exercise and severe sarcopenia should be remarked upon. Studies show that, the less physical activity, the less the muscle mass, and the greater the prevalence of physical disabilities[35-36]. The regular practice of exercise delays muscle loss and increases muscle strength, preventing sarcopenia. Literature shows that the best results are achieved with Progressive Resisted Exercise (PRE)[35-36]; furthermore, when the exercise is supervised by professionals, it is more beneficial to treat sarcopenia, improving muscle mass, strength, and physical performance when compared to exercise carried out at home unsupervised[37]. Regarding comorbidities, research shows the association between osteoporosis and the development of sarcopenia[38]. It stands out that sarcopenia and osteoporosis have the same etiology (inflammation, hormonal and nutritional deficiency, and lack of physical exercise) and the same risk factors for muscle incapacity[39]. Also, osteoarthritis limits mobility due to pain and rigidity, reducing muscle strength. Regarding the presence of diabetes, there was no association, in disagreement with the results from other studies[28,40]. Nonetheless, it is necessary to study this element in more depth in future investigations, since endocrine changes and the liberation of inflammatory cytokines in diabetes lead to muscle degradation. Additionally, insulin resistance is a multifactorial condition, and aging and obesity are associated with a chronic inflammatory state that causes skeletal muscle loss, being this muscle the main target-tissue responsive to insulin, contributing for sarcopenia[41]. In this study, polypharmacy was the most prevalent characteristic for elders with any degree of sarcopenia, corroborating international studies[42-43]. The use of several medications is common in aging and increases the risk of adverse reactions. These can interfere in the metabolism and homeostasis, causing mitochondrial dysfunction and hydroelectric and endocrine unbalances, as well as gastrointestinal absorption dysfunctions, all of which are factors that lead to the development of sarcopenia[42-43]. In regard to findings involving body composition, Lee’s equation method, which considers BMI and calf circumference, is considered to be an effective method to evaluate weight reduction and sarcopenia[44]. Body composition may hide sarcopenia obesity, characterized by high levels of body fat, which catalyzes the reduction of lean body mass and muscle force[45-46]. It was worth saying that identifying sarcopenic obesity is difficult for health workers[46]. Still, measuring muscle mass is a challenge to be implemented in practice due to the limitations intrinsic to evaluation instruments (such as cost, availability, and ease of use), which are often more useful for research than useful in clinical practice, especially for Primary Care[47]. Furthermore, although European *Consensus is* advertised as a valid way to evaluate sarcopenia, it has limitations in the evaluation of elders with physical and cognitive restrictions. As a result, the perception about the actual prevalence of sarcopenia is mistaken[48]. Bedridden and wheelchair bound elders are more likely to develop comorbidities and, thus, to lose functionality[48-49]. This corroborates the need for research to develop diagnostic methods and strategies capable of including elders with physical and cognitive restrictions. Considering the above, we must note that studies of this nature generate important information that can serve as guide for managers and professionals to plan and develop interventions for people of specific ages. Sarcopenia is a topic that has become increasingly relevant and requires studies that address new treatment options and preventive interventions. ## Study limitations Limitations include the fact that some independent variables (educational level and comorbidities) were self-reported, making it impossible to establish cause relations due to the study design. Furthermore, it was not possible to evaluate sarcopenia in elderly with cognitive restrictions, and in those who were bedridden or could not go to the units. Also, we suggest longitudinal studies to be carried out in order to evaluate in depth the development of sarcopenia in the elderly population, both in urban and rural areas. ## Contributions to the Field of Nursing The proposal of evaluating the prevalence and characteristics of sarcopenia in primary care will make it possible to replicate the study in other settings. Furthermore, we expect encouraging nurses to carry out further studies on the topic, which is still seldom discussed by nursing workers. In addition, this study stands out due to its originality, and the information generated will give support for interventions for the prevention and promotion of the health of the elder. ## CONCLUSIONS In the elders attended in the Primary Health Care at Fortaleza, the prevalence of probable sarcopenia was $25.52\%$; sarcopenia, $11.8\%$; and severe sarcopenia, $9.90\%$. We suggest using the same methods used in this study for evaluations in primary care. Sex, osteoporosis, polypharmacy, overweight, obesity, and calf circumference below 31 cm are the most present characteristics of some degree of sarcopenia, while physical activity is less prevalent among those with severe sarcopenia. Some of these characteristics are modifiable conditions; therefore, systematic evaluations, in addition to lifestyle changes, could prevent the installation and the repercussions of sarcopenia on the quality of life of elders and their families. ## References 1. Rosenberg IH.. **Summary comments**. *The American J Clin Nutrit* (1989) **50** 1231-1233. DOI: 10.1093/ajcn/50.5.1231 2. Anker SD, Morley JE, von Haehling S.. **Bem-vindo ao código CID-10 para sarcopenia**. *J Cachexia Sarcopenia Muscle* (2016) **7** 512-514. DOI: 10.1002/jcsm.12147 3. Cruz-Jentoft AJ, Bahat G, Bauer J, Boirie Y, Bruyère O, Cederholm T. **Sarcopenia: revised European consensus on definition and diagnosis**. *Age Ageing* (2019) **48** 16-31. DOI: 10.1093/ageing/afy169 4. Veronese N, Demurtas J, Soysal P, Smith L, Torbahn G, Schoene D. **Sarcopenia and health-related outcomes: an umbrella review of observational studies**. *Eur Geriatr Med* (2019) **10** 853-862. DOI: 10.1007/s41999-019-00233-w 5. Petermann RF, Balntzi V, Gray SR, Lara J, Ho FK, Pell JP. **Global prevalence of sarcopenia and severe sarcopenia: a systematic review and meta-analysis**. *J Cachexia Sarcopenia Muscle* (2022) **13** 86-99. DOI: 10.1002/jcsm.12783 6. Pelegrini A, Mazo GZ, Pinto AA, Benedetti TRB, Silva DAS, Petroski EL.. **Sarcopenia: prevalence and associated factors among elderly from a Brazilian capital**. *Fisioter Mov* (2018) **31** 1-8. DOI: 10.1590/1980-5918.031.ao02 7. Alexandre TS, Duarte YAO, Santos JLF, Lebrão ML.. **Prevalence and associated factors of sarcopenia, dynapenia, and sarcodynapenia in community dwelling elderly in São Paulo - SABE Study**. *Rev Bras Epidemiol* (2018) **21** 1-13. DOI: 10.1590/1980-549720180009.supl.2 8. Confortin SC, Ono LM, Barbosa AR, d'Orsi E. **Sarcopenia and its association with changes in socioeconomic, behavioral, and health factors: the EpiFloripa Elderly Study**. *Cad Saude Publica* (2018) **34** 1-13. DOI: 10.1590/0102-311X00164917 9. Dutra T, Pinheiro PA, Carneiro JAO, Coqueiro RS, Fernandes MH. **Prevalence and factors associated with sarcopenia in elderly women living in the community**. *Rev Bras Cineantropometria Desemp Hum* (2015) **17** 460-471. DOI: 10.5007/1980-0037.2015v17n4p460 10. Papadopoulou SK.. **Sarcopenia: a contemporary health problem among older adult populations**. *Nutrients* (2020) **12** 1-20. DOI: 10.3390/nu12051293 11. Beaudart C, McCloskey E, Bruyère O, Cesari M, Rolland Y, Rizzoli R.. **Sarcopenia in daily practice: assessment and management**. *BMC Geriatr* (2016) **16** 1-10. DOI: 10.1186/s12877-016-0349-4 12. Sayer AA, Syddall HE, Gilbody HJ, Dennison EM, Cooper C.. **A sarcopenia se origina no início da vida? Resultados do estudo de coorte de Hertfordshire**. *J Gerontol A Biol Sci Med Sci* (2004) **59** 930-934. DOI: 10.1093/gerona/59.9.M930 13. Pillatt AP, Patias RS, Berlezi EM, Schneider RH.. **Which factors are associated with sarcopenia and frailty in elderly persons residing in the community?**. *Rev Bras Geriatr Gerontol* (2018) **21** 755-766. DOI: 10.1590/1981-22562018021.180165 14. Costa NR.. **A Estratégia de Saúde da Família, a atenção primária e o desafio das metrópoles brasileiras**. *Ciênc Saúde Coletiva* (2016) **21** 1389-1398. DOI: 10.1590/1413-81232015215.24842015 15. Cuschieri S.. **The STROBE guidelines**. *Saudi J Anaesth [Internet]* (2019) **13** 31-34 16. **División de Promoción y Protección de la Salud (HPP). Encuesta Multicentrica salud beinestar y envejecimiento (SABE) em América Latina el Caribe: Informe Preliminar**. *In: XXXVI Reunión del Comité asesor de investigaciones em Salud [Internet]* (2001) 17. Rogers ME, Rogers NL, Takeshima N, Islam MM.. **Methods to assess and improve the physical parameters associated with fall risk in older adults**. *Prev Med* (2003) **36** 255-264. DOI: 10.1016/S0091-7435(02)00028-2 18. Crosby CA, Wehbé MA, Mawr B.. **Hand strength: normative values**. *J Hand Surg Am* (1994) **19** 665-670. DOI: 10.1016/0363-5023(94)90280-1 19. Lee RC, Wang Z, Heo M, Ross R, Janssen I, Heymsfield SB.. **Total-body skeletal muscle mass: development and cross-validation of anthropometric prediction models**. *Am J Clin Nutr* (2000) **72** 796-803. DOI: 10.1093/ajcn/72.3.796 20. D’orsi E, Xavier AJ, Ramos LR.. **Work, social support and leisure protect the elderly from functional loss: epidoso study**. *Rev Saúde Pública* (2011) **45** 685-692. DOI: 10.1590/S0034-89102011000400007 21. Biesheuvel CJ, Vergouwe Y, Steyerberg EW, Grobbee DE, Moons KG.. **Polytomous logistic regression analysis could be applied more often in diagnostic research**. *J Clin Epidemiol* (2008) **61** 125-134. DOI: 10.1016/j.jclinepi.2007.03.002 22. Sui SX, Holloway-Kew KL, Hyde NK, Williams LJ, Tembo MC, Leach S, Pasco JA.. **Prevalence of Sarcopenia Employing Population-Specific Cut-Points: Cross-Sectional Data from the Geelong Osteoporosis Study, Australia**. *J Clin Med* (2021) **10** 343. DOI: 10.3390/jcm10020343 23. Sousa-Santos AR, Afonso C, Borges N, Santos A, Padrão P, Moreira P, Amaral TF.. **Factors associated with sarcopenia and undernutrition in older adults**. *Nutr Diet* (2019) **76** 604-612. DOI: 10.1111/1747-0080.12542 24. Borges VS, Lima-Costa MFF, Andrade FB.. **A nationwide study on prevalence and factors associated with dynapenia in older adults: ELSI-Brazil**. *Cad Saude Publica* (2020) **36** 1-10. DOI: 10.1590/0102-311X00107319 25. Paula JA, Wamser EL, Gomes ARS, Valderramas SR, Cardoso NJ, Schieferdecker MEM.. **Analysis of methods for detecting sarcopenia in independent community-dwelling elderly women. Rev Bras Geriatr**. *Gerontol* (2016) **19** 235-246. DOI: 10.1590/1809-98232016019.140233 26. Sousa PHSF, Souza RF, Souza Costa MRSS, Azevedo MVC, Torres RC, Nascimento GC.. **Nursing protagonism in basic health care**. *Braz J Develop* (2020) **6** 76157-70. DOI: 10.34117/bjdv6n10-151 27. Bijlsma AY, Meskers CG, Van N, Westendorp RG, Sipila S, Stenroth L.. **Diagnostic criteria for sarcopenia and physical performance**. *Age* (2014) **36** 275-285. DOI: 10.1007/s11357-013-9556-5 28. Pacifico J, Geerlings MAJ, Reijnierse EM, Phassouliotis C, Lim WK, Maier AB.. **Prevalence of sarcopenia as a comorbid disease: a systematic review and meta-analysis**. *Exp Gerontol* (2020) **131** 1-19. DOI: 10.1016/j.exger.2019.110801 29. Berlezi EM, Gross CB, Pimentel JJ, Pagno AR, Fortes CK, Pillatt AP.. **Estudo do fenótipo de fragilidade em idosos residentes na comunidade**. *Ciênc Saúde Coletiva* (2019) **24** 4201-4210. DOI: 10.1590/1413-812320182411.31072017 30. Buckinx F, Aubertin-Leheudre M.. **Sarcopenia in menopausal women: current perspectives**. *Int J Womens Health* (2022) **14** 805-819. DOI: 10.2147/IJWH.S340537 31. Lima CS, Aguiar RS.. **Men’s access to primary health care services: an integrative review**. *Res, Soc Develop* (2020) **9** 1-20. DOI: 10.33448/rsd-v9i4.3027 32. Gomes RN, Araújo ED, Carvalho F.. **Por que os homens buscam menos os serviços de saúde do que as mulheres? as explicações de homens com baixa escolaridade e homens com ensino superior**. *Cad Saúde Pública* (2007) **23** 565-574. DOI: 10.1590/S0102-311X2007000300015 33. Lee DY, Shin S.. **Sarcopenia is associated with metabolic syndrome in Korean adults aged over 50 years: a cross-sectional study**. *Int J Environ Res Public Health* (2022) **19** 1330. DOI: 10.3390/ijerph19031330 34. Mo D, Hsieh P, Yu H, Zhou L, Gong J, Xu L. **Osteosarcopenic obesity and its relationship with dyslipidemia in women from different ethnic groups of China**. *Arch Osteoporos* (2018) **65** 1-9. DOI: 10.1007/s11657-018-0481-1 35. Law TD, Clark LA, Clark BC.. **Resistance exercise to prevent and control sarcopenia and dinapenia**. *Annu Rev Gerontol Geriatr* (2016) **36** 205-228. DOI: 10.1891/0198-8794.36.205 36. Rubio del Peral JA, Gracia JMA.. **Ejercicios de resistencia en el tratamiento y prevención de la sarcopenia en ancianos**. *Gerokomos* (2018) **29** 133-137. DOI: 10.15517/pensarmov.v16i1.3000 37. Tsekoura M, Billis E, Tsepis E, Dimitriadis Z, Matzaroglou C, Tyllianakis M.. **The effects of group and home exercise programs on elderly people with sarcopenia: a randomized controlled test**. *J Clin Med* (2018) **7** 1-18. DOI: 10.3390/jcm7120480 38. Pérez-Sousa MA, Pozo-Cruz JD, Cano-Gutiérrez CA, Izquierdo M, Ramírez-Vélez R.. **High prevalence of probable sarcopenia in a representative sample from Colombia: implications for geriatrics in Latin America**. *J Am Med Dir Assoc* (2021) **22** 859-864. DOI: 10.1016/j.jamda.2022.02.012 39. Khongsri N, Tongsuntud S, Limampai P, Kuptniratsaikul V. **The prevalence of sarcopenia and related factors in a community-dwelling elders Thai population**. *Osteoporosis Sarcopenia* (2016) **2** 110-115. DOI: 10.1016/j.afos.2016.05.001 40. Freitas MM, Oliveira VLP, Grassi T, Valduga K, Miller MEP, Schuchmann RA.. **Difference in sarcopenia prevalence and associated factors according to 2010 and 2018 European consensus (EWGSOP) in elderly patients with type 2 diabetes mellitus**. *Exp Gerontol* (2020) **132** 1-7. DOI: 10.1016/j.exger.2020.110835 41. Frasca D, Blomberg BB, Paganelli R.. **Aging, obesity, and inflammatory age-related diseases**. *Front Immunol* (2017) **8** 1-10. DOI: 10.3389/fimmu.2017.01745 42. König M, Spira D, Demuth I, Steinhagen-Thiessen E, Norman K.. **Polypharmacy as a risk factor for clinically relevant sarcopenia: results from the Berlin Aging Study II**. *J Gerontol A Biol Sci Med Sci* (2017) **73** 117-122. DOI: 10.1093/gerona/glx074 43. Pana A, Sourtzi P, Kalokairinou A, Velonaki VS.. **Sarcopenia and polypharmacy among older adults: a scoping review of the literature**. *Arch Gerontol Geriatr* (2021) **98** 1-10. DOI: 10.1016/j.archger.2021.104520 44. Pagotto V, Santos KFD, Malaquias SG, Bachion MM, Silveira EA.. **Calf circumference: clinical validation for evaluation of muscle mass in the elderly**. *Rev Bras Enferm* (2018) **71** 322-328. DOI: 10.1590/0034-7167-2017-0121 45. Campos GC, Lopes CS, Lourenço RA.. **Obesidade sarcopênica e funcionalidade: uma revisão da literatura**. *Rev Hosp Univ Pedro Ernesto* (2018) **16** 102-109. DOI: 10.12957/rhupe.2017.37650 46. Batsis JA, Villareal DT.. **Sarcopenic obesity in the elderly: etiology, epidemiology and treatment strategies**. *Nat Rev Endocrinol* (2018) **14** 513-537. DOI: 10.1038/s41574-018-0062-9 47. Buckinx F, Landi F, Cesari M. **Pitfalls in the measurement of muscle mass: a need for a reference standard**. *J Cachexia Sarcopenia Muscle* (2018) **9** 269-278. DOI: 10.1002/jcsm.12268 48. Danielewicz AL, Barbosa AR, Duca GFD.. **Nutritional status, physical performance and functional capacity in an elderly population in southern Brazil**. *Rev Assoc Med Bras* (2014) **60** 242-248. DOI: 10.1590/1806-9282.60.03.0013 49. Ochi M, Tabara Y, Kido T, Uetani E, Ochi N, Igase M. **Quadriceps sarcopenia and visceral obesity are risk factors for postural instability in the middle-aged to elderly population**. *Geriatr Gerontol Int* (2010) **10** 233-243. DOI: 10.1111/j.1447-0594.2010.00610.x
--- title: 'RESVECH 2.0: cross-cultural adaptation for Brazil, reliability and validity for the evaluation of venous ulcers' authors: - Marina Rosa Menegon - Suelen Gomes Malaquias - Joana Aragão da Silva - Beatriz Guitton Renaud Baptista de Oliveira - Juan Carlos Restrepo Medrano - José Verdú-Soriano - Maria Márcia Bachion journal: Revista Brasileira de Enfermagem year: 2023 pmcid: PMC10042478 doi: 10.1590/0034-7167-2022-0185 license: CC BY 4.0 --- # RESVECH 2.0: cross-cultural adaptation for Brazil, reliability and validity for the evaluation of venous ulcers ## ABSTRACT ### Objectives: to cross-culturally adapt the scale Resultados en la valoración y evolución de la cicatrización de las heridas - RESVECH 2.0 for Brazilian Portuguese; to estimate the internal consistency and construct and criterion validity of the scale in the evaluation of venous ulcers. ### Methods: methodological study, based on international guidelines for studies of this type. Wounds were evaluated using the RESVECH 2.0 and Pressure Ulcer Scale of Healing 3.0 (PUSH). Descriptive analysis, confirmatory factor analysis, Cronbach’s alpha and Spearman’s correlation ($p \leq 0.05$) were used. ### Results: 12 nurses and 77 people with 153 venous ulcers participated in the study. The translation was successful, the proposed factor model was validated, and Cronbach ‘s alpha = 0.832 ($95\%$CI, 0.780-0.880) and correlation coefficient (RESVECH 2.0 and PUSH 3.0) = 0.74 were obtained. ### Conclusions: the adaptation of RESVECH 2.0 to Brazilian *Portuguese is* robust. Reliability and validity show compatibility for use in the country in the evaluation of venous ulcers. ## Objetivos: adaptar transculturalmente a escala Resultados en la valoración y evolución de la cicatrización de las heridas - RESVECH 2.0 para o português do Brasil; estimar sua consistência interna, validade de construto e de critério para utilização em úlceras venosas. adaptar transculturalmente la escala “Resultados en la valoración y evolución de la cicatrización de heridas”, RESVECH 2.0 al portugués de Brasil; estimar su consistencia interna, validez de constructo y de criterio para su utilización en úlceras varicosas. ## Métodos: estudo metodológico, baseado em diretrizes internacionais para estudos dessa natureza. Realizou-se avaliação das feridas por meio da RESVECH 2.0 e da Pressure Ulcer Scale of Healing 3.0 (PUSH). Empregou-se análise descritiva, análise fatorial confirmatória, alfa de Cronbach e correlação de Spearman ($p \leq 0$,05). es un estudio metodológico, basado en directivas internacionales sobre investigaciones de esta naturaleza. Se evaluaron las heridas por medio de la RESVECH 2.0 y de la Escala de Cicatrización de Úlceras por Presión 3.0 (PUSH). Se llevó a cabo con análisis descriptivo, análisis factorial confirmatorio, alfa de Cronbach y correlación de Spearman ($p \leq 0$,05). ## Resultados: participaram 12 enfermeiros e 77 pessoas com 153 úlceras venosas. A tradução foi bem-sucedida, o modelo fatorial proposto foi validado, obteve-se alfa de Cronbach = 0,832 (IC$95\%$=0,780-0,880) e coeficiente de correlação (RESVECH 2.0 e PUSH 3.0) = 0,74. participaron 12 enfermeros y 77 personas que tenían 153 úlceras venosas. La traducción fue exitosa, el modelo factorial propuesto fue validado, el alfa de Cronbach = 0,832 ($95\%$CI=0,780-0,880) y el coeficiente de correlación (RESVECH 2.0 y PUSH 3.0) = 0,74. ## Conclusões: a adaptação da RESVECH 2.0 para o português do Brasil é robusta. A confiabilidade e validade evidenciam compatibilidade para utilização no país e avaliação de úlceras venosas. ## Conclusiones: la adaptación de la RESVECH 2.0 al portugués brasileño es sólida. La fiabilidad y la validez demuestran la compatibilidad para su utilización en el país en la evaluación de las úlceras varicosas. ## INTRODUCTION Venous ulcers (VUs) are the most common type of leg ulcer and are usually associated with chronic venous insufficiency[1-2]. This type of injury often appears in the leg, between the knee and the ankle, but sometimes it also appears in areas below the ankle[3]. It usually heals slowly[1-3] and has a high chance of relapse[4]. Seeing that these injuries are chronic and are monitored by a multidisciplinary team in the primary and secondary care of the Unified National Health System, it is essential to use scales to assess the evolution of healing, in order to allow standardized clinical evaluation records, ensure effective communication among professionals, provide an accurate assessment of the outcomes of the care provided and help the decision-making process regarding the care techniques to be used. There are dozens of instruments to assess wound healing[5-7]. Most of them are focused on pressure injuries or chronic wounds in general, but very few are aimed at the evaluation of leg ulcers, and none address venous ulcers specifically. Despite the variety of scales available, no instrument is considered the gold standard to assess every type of wound, including venous ulcers. The Pressure Ulcer Scale for Healing (PUSH)[8-9] is one of the most widely used and recognized scales in the world[10]. Originally developed to assess the healing of pressure injuries, it has since been used to assess other types of chronic wounds, such as leg ulcers in general[11-12] and venous ulcers[13-15], in several countries [12,15-16], including Brazil[11,15]. The scale has been cross-culturally adapted to Brazilian Portuguese[17] and has a very good inter-rater reliability in the evaluation of leg ulcers[11] and venous ulcers[15] and good responsiveness in chronic wounds (pressure injuries, neuropathic ulcers and venous ulcers)[18]. However, the characteristics of venous ulcers are not fully considered in the PUSH assessment parameters, as these injuries can affect extensive areas, result in infection[14,19], and be painful[19-20]. In an attempt to develop a more comprehensive instrument to evaluate chronic wounds, researchers from the Grupo *Nacional para* el Estudio y Asesoramiento en Úlceras por Presión y Heridas Crónicas (GNEAUPP) in Spain developed the Resultados Esperados de la Cicatrización de las Heridas Crônicas (RESVECH) scale[5,21]. The version 1.0 of the scale contained nine items; however, after the first clinical validation tests, three of them were excluded (periwound maceration, tunneling and pain), resulting in version 2.0[21]. Items currently evaluated include: dimensions of lesion; depth and tissues involved; edge features; tissues in the wound bed; exudate; and signs of infection/inflammation[21-22], which are relevant for the evaluation of venous ulcers[19-20]. The final assessment is based on the score obtained, which can range from 0 (healed wound) to 35 points (worst possible condition)[22]. The scale has been cross-culturally adapted for use in other countries such as Colombia[23], Portugal[24] and Brazil[25]. The first translation and cross-cultural adaptation into Brazilian Portuguese was carried out in a city in the state Minas Gerais[25] and resulted in changes in the scale, without showing whether this occurred with the author’s consent. In that research, there was no testing of psychometric properties when applied to the population. Valid and reliable instruments can contribute to clinical practice, health assessment and research, support decision-making[25-26], favor the assessment of the healing process and contribute to a standardized and effective communication of the results observed during and at the end of the treatment of people with VU. Thus, the cross-cultural adaptation of the RESVECH 2.0 in a broader context and the verification of the reliability of the Brazilian version are essential to provide better evidence to support its use in the evaluation of people with venous ulcers in Brazil. The results of this study are intended to contribute to the clinical performance of nursing and health professionals, through a more accurate assessment of the wound healing process, allowing the analysis of the effectiveness of treatments and supporting professional nursing practice and evidence-based health care. Additionally, this study may contribute to teaching in the area of evaluation and treatment of wounds in undergraduate and lato sensu graduate programs, and enable the use of the scale in a protocol for the evaluation of healing of venous ulcers in future research. ## OBJECTIVES To cross-culturally adapt the scale Resultados Esperados de la Valoración y Evaluación de la Cicatrización de las Heridas Crônicas (RESVECH 2.0) into a Brazilian Portuguese version. To estimate the internal consistency and construct and criterion validity of the Brazilian version of RESVECH 2.0 in the evaluation of venous ulcers. ## Ethical aspects This is an excerpt from a matrix project called “Tradução, adaptação transcultural, confiabilidade e responsividade de escalas de avaliação de capacidade funcional, cicatrização e qualidade de vida de pessoas com úlceras venosas”, funded by CNPq [Process $\frac{312093}{2013}$-6]. The study follows the recommendations of Resolution No. 466 of 2012 of the National Health Council[27] and was approved by the Research Ethics Committee of the Hospital das Clínicas of the Federal University of Goiás. ## Type of study Methodological multicenter study carried out in Goiânia-GO and Niterói-RJ, from 2016 to 2018. ## Cross-Cultural Adaptation The recommendations from the literature on translation and cross-cultural adaptation were followed in the process[28-30]. Thus, the steps shown in Figure 1 were followed: Figure 1Flowchart of the translation process ## Data collection Data was collected from 2016 to 2018, in Goiânia-GO and Niterói-RJ, in two stages. The cognitive debriefing of the pre-final version (translated) was conducted with nurses with experience in the treatment of people with venous ulcers, according to the recommendations in the literature[29-31]. Professionals were invited to fill out a characterization form and carefully analyze the translated instrument, recording any doubts. The analysis of the psychometric properties was conducted with people with venous ulcers undergoing outpatient treatment in the study settings. The following inclusion criteria were applied: age ≥ 18 years; satisfactory score in the mini-mental state examination[32], according to the level of education; medical diagnosis of chronic venous insufficiency; presence of clinical signs of venous insufficiency; active venous ulcer. Those with signs of moderate or severe arterial impairment were excluded. After receiving training, the researchers applied the translated version of RESVECH 2.0 and PUSH 3.0 in people with venous ulcers. This scale was chosen as reference because it is used worldwide and has acceptable psychometric properties for use in this population, as described above. ## Data analysis The collected data were organized in an electronic database using the software Statistical Package for the Social Sciences (SPSS). For construct (or concept) validation, confirmatory factor analysis (CFA) was used to evaluate the factor structure of the RESVECH 2.0 with 19 items (Figure 2). Dimensions 1 to 5 consist of isolated items and dimension 6 (inflammation/infection) includes 14 items. CFA was chosen because it is considered the most appropriate method for evaluating dichotomous variables and minimizing the risks of high standard error in the correlation coefficients, allowing the grouping of items that are associated with each other and determining the relationship between a group of variables[33]. Figure 2Factorial structure to be tested for the validation of the scale Resultados en la valoración y evolución de la cicatrización de las heridas - RESVECH 2.0 DL - Dimensions of the lesion; DTI - Depth/ tissues involved; TWB -Tissues in the wound bed; Exudate; R - item of the RESVECH scale. Initially, the data was tested to verify the adequacy of the CFA model, using a matrix of correlation coefficients with the 19 items to assess the degree of correlation between them. A tetrachoric correlation matrix was used to verify the correlation between the nominal items and the the rank-biserial correlation coefficient[34] with p-value<0.05 was used to verify the correlation between ordinal and nominal, and ordinal and ordinal items. Then, the Bartlett test of sphericity[35] and Kaiser-Meyer-Olkin (KMO) test were performed to verify the adequacy of the CFA model. For the Bartlett test, a p-value<0.05 was adopted. In turn, in the KMO test, a result above 0.600 was considered adequate[36]. The model’s Cronbach’s alpha was also calculated, considering values>0.78 as acceptable[31,37]. Spearman’s correlation coefficient (rs) was used to analyze the validity related to the criterion between the data obtained from RESVECH 2.0 and from PUSH 3.0. A value of $p \leq 0.05$ was adopted. Values above 0.70 were considered acceptable[37]. ## Cross-cultural adaptation of RESVECH 2.0 to Brazilian Portuguese There were few difficulties in the translation of the RESVECH 2.0 scale, and, as expected, they occurred on the part of the professional that was not from the health area, who used, for example, the term escara to refer to ulcer. In the stages of expert committee analysis and testing of the pre-final version, some terms were discussed as, even though they exist in Brazilian Portuguese, they are not common in the research scenario or do not represent phenomena for which there is more current terminology in clinical practice. A total of 12 nurses participated in the cognitive debriefing stage. The participants considered that the use of the term “danificadas” to refer to the conditions of the edges was odd. After discussion of semantics (danificadas = that which has been flawed or harmed; lesada or deteriorada = changed for the worse; damaged, according to Oxford Languages) and considerations about the most common term in clinical practice, the term “deteriorada” was chosen to refer to the wound edges. Another term discussed in the context of evaluation of wound edges, was “bordas engrossadas (envelhecidas ou evertidas)”. After discussion with the authors of the original scale, these terms were maintained and more details were provided in the description of these items on the guiding instrument for application of the scale (Appendix 1 - Supplementary material). Still regarding the evaluation of wound edges, the cognitive debriefing participants found it difficult to understand the difference between “bordas não distinguíveis (não há bordas)” and “bordas não delimitadas”, or “bordas não distinguíveis x fechada/cicatrizada”. Upon clarification based on stage I pressure ulcers, in which there is no rupture of the epidermis, and therefore no wound edges, the participants understood the use of the expression in the context, but highlighted that, in the case of venous ulcers, the category “bordas não distinguíveis” would not be applicable. Exudate was evaluated in the RESVESCH 2.0 based on the terms “seco”, “úmido”, “molhado”, ”saturado” and “com fuga de exsudato”. Likewise, although the terms are understandable, they are not common in clinical practice in the setting of the present investigation, where the professionals usually assess the exudate as “ausente”, “pequeno”, “moderado” or “grande”. The authors of the original scale did not authorize changes in the options for the evaluation of exudate. They consider that, as in other scales, adequate training must be provided so that the instrument is perceived in its scope as originally envisioned. There were no issues with the item “tecido compatível com biofilme”, but later, in the researchers’ training, further clarification was required for the standardized evaluation of this item. Thus, the following description was added to guide the assessment of the presence of tissue compatible with biofilm: tejido compatível con biofilm es una capa de “sustância” sobre la ferida de color blanco-amarillento, ou transparente pero brillante (habitualmente clasificado como esfacelo o como fibrina, pero que en este caso se retira fácilmente con una torunda o gasa) (José Verdu Soriano). All other items of the scale were successfully translated, with no issues in semantics or cultural understanding, resulting in a culturally adapted version of the RESVECH 2.0 (Appendix 2 - Supplementary material) for further analysis of internal consistency. To assess internal consistency, 153 VUs presented by 77 participants were evaluated. Of these, 36 were linked to the research center of UFF and recruited in the referral outpatient clinic for wound care in Niterói-RJ, and 41 were linked to the UFG research center and recruited in the outpatient wound care network in Goiânia-GO. Among the participants, 42 ($54.5\%$) were female and 35 ($45.5\%$) were male. Approximately half ($54.9\%$) of the lesions had an area equal to or greater than 24 cm2 and $19\%$ had an area equal to or greater than 100 cm2. ## Confirmatory Factor Analysis The KMO test value was 0.615, demonstrating the suitability of data for the CFA. Also, the probability of Bartlett’s test of sphericity suggested the factorability of the correlation matrix (chi-square: 648.006; p-value<0.001)[35]. Confirmatory factor analysis showed that most items had a significant correlation coefficient >0.3 (Table 1), indicating the factorability of the matrix, according to the proposed model (Figure 3). **Table 1** | Variable | β | 95%CI | Standard Error | p value | FL | | --- | --- | --- | --- | --- | --- | | Item | | | | | | | 1 Dimensions of the lesion | 0.687 | 0.513 - 0.862 | 0.089 | <0.001 | 0.487 | | 2 Depth/tissues involved | 0.307 | 0.119 - 0.406 | 0.096 | 0.001 | 0.401 | | 3 Edges | 0.303 | 0.112 - 0.494 | 0.097 | 0.002 | 0.494 | | 4 Slough | 0.249 | 0.050 - 0.446 | 0.1 | 0.014 | 0.589 | | 5 Exudate | 0.578 | 0.402 - 0.752 | 0.089 | <0.001 | 0.566 | | 6. Inflammation/Infection | | | | | | | 6.1 Increased pain | 0.157 | 0.032 - 0.346 | 0.096 | 0.045 | 0.485 | | 6.2 Perilesional erythema | 0.515 | 0.370 - 0.659 | 0.073 | <0.001 | 0.517 | | 6.3 Perilesional edema | 0.209 | 0.025 - 0.396 | 0.094 | 0.026 | 0.433 | | 6.4 Increased temperature | 0.652 | 0.512 -0.703 | 0.072 | <0.001 | 0.562 | | 6.5 Increased exudate | 0.128 | 0.057 - 0.313 | 0.094 | 0.048 | 0.327 | | 6.6 Purulent exudate | 0.372 | 0.188 - 0.556 | 0.094 | <0.001 | 0.58 | | 6.7 Friable tissue | 0.204 | 0.187 - 0.206 | 0.1 | 0.924 | 0.442 | | 6.8 Stagnant wound | 0.609 | 0.450 - 0.766 | 0.079 | <0.001 | 0.567 | | 6.9 Biofilm compatible tissue | 0.408 | 0.306 -0.655 | 0.088 | <0.001 | 0.619 | | 6.10 Odor | 0.567 | 0.407 - 0.728 | 0.082 | <0.001 | 0.642 | | 6.11 Hypergranulation | 0.203 | 0.017 - 0.422 | 0.112 | 0.041 | 0.41 | | 6.12 Size increase | 0.489 | 0.294 - 0.683 | 0.099 | <0.001 | 0.411 | | 6.13 Satellite injuries | 0.208 | 0.029 - 0.386 | 0.091 | 0.023 | 0.386 | | 6.14 Paleness of the tissue | 0.147 | 0.067 - 0.273 | 0.107 | 0.042 | 0.321 | Figure 3Path diagram of the factorial structure of the scale Resultados en la valoración y evolución de la cicatrización de las heridas - RESVECH 2.0 validated with 19 items ɛ/B - Standard error; * - Factor loading. All factor loadings had values > 0.4 (Table 1), except for item 6.5, items 6.13 and 6.14. However, these items still presented acceptable factor loading values (> 0.3). ## Internal consistency of the translated version of RESVECH 2.0 Internal consistency was estimated using Cronbach’s alpha with a $95\%$ confidence interval. The proposed factorial model showed that the estimated internal consistency of REVESH 2.0 was 0.832 ($95\%$ CI = 0.780-0.880), indicating good internal reliability. ## Criterion-related validity of the translated version of RESVECH 2.0 A correlation coefficient of 0.74 supported the criterion validity of the scale. It was found that the items “dimensão”, “tecido”, “exsudato”, total RESVESCH 2.0 score and respective PUSH items presented a strong (rs >0.70) correlation with each other (Table 2). As for the other items of the RESVECH 2.0, it is worth noting that the size of the lesion showed a moderate correlation (rs 0.40 to 0.69) with other items of the PUSH 3.0. **Table 2** | REVESCH 2.0 | PUSH 3.0 | PUSH 3.0.1 | PUSH 3.0.2 | PUSH 3.0.3 | | --- | --- | --- | --- | --- | | REVESCH 2.0 | Area | Exudate | Tissue | Total | | Dimensions of the lesion(rs) | 0.892 | 0.609 | 0.121 | 0.878 | | p value | <0.001 | <0.001 | 0.135 | <0.001 | | Depth/ tissues involved (rs) | 0.243 | 0.236 | 0.120 | 0.257 | | p value | 0.003 | 0.003 | 0.139 | 0.001 | | Edges (rs) | 0.174 | 0.270 | 0.106 | 0.254 | | p value | <0.001 | 0.001 | 0.193 | 0.002 | | Slough (rs) | 0.087 | 0.138 | 0.789 | 0.204 | | p value | 0.284 | 0.090 | <0.001 | 0.011 | | Exudate (rs) | 0.402 | 0.624 | 0.186 | 0.536 | | p value | <0.001 | <0.001 | 0.021 | <0.001 | | Item 6 (rs) | 0.325 | 0.327 | 0.137 | 0.331 | | p value | <0.001 | <0.001 | 0.069 | <0.001 | | Total (rs) | 0.683 | 0.641 | 0.247 | 0.740 | | p value | <0.001 | <0.001 | 0.002 | <0.001 | ## DISCUSSION The translation of RESVECH 2.0 was successful and no major difficulties were found in the process. Ease of translation for other contexts has also been reported[23-25]. The fact that some terms were considered odd for use in clinical practice in Brazil may be due to the specificity of the present investigation, which addressed only the context of VU assessment. This version of the instrument is robust, as the principles of good practices were followed[28-30], and represents a relevant resource for clinical practice in Brazil, as there is evidence it can be applied in VU treatment, enabling the integration of signs and symptoms of infection, which are of great relevance for monitoring the healing process of these wounds. It should be noted that the RESVECH 2.0 allows recording the size(area) of wounds up to 100 cm 2[5]. This is important data, considering that VUs usually have areas larger than 24 cm2[20], as considered in PUSH and corroborated in the present study. Two items of the second category “Depth/tissues affected” would not be applicable in the context of VU: “involvement of the muscle”; “involvement of bones and/or surrounding tissue”, as most VUs are superficial[20,38]. However, this is not a limiting factor for the use of the scale in this population. In other settings, the applicability of the term “borda” in clinical practice was not questioned, but in another region of Brazil, where the first RESVECH 2.0 translation was carried out, the term was changed to “margem”, which, according to the authors[25], was more common in Brazilian culture, which was not confirmed in the present study. It is more common to describe the edges as: “epitelizadas”, “maceradas”, “com hiperceratose”, “hiperemiadas” and/or “com crostas” [39], and these options generally apply to venous ulcers. Perhaps that is why nurses were surprised when seeing the option “bordas engrossadas (envelhecidas ou evertidas)”. As previously highlighted, it is expected that adequate training will be sufficient to overcome any difficulties in the understanding of the clinical phenomena to which the items refer. There was no difficulty in translating or understanding the items in the category “type of tissue in the wound bed”. Most of the options available meet the characteristics of venous ulcers, in which slough is usually present[20], as venous hypertension reduces blood flow in the capillary network, triggering a decrease in oxygen circulation, causing adhered neutrophils to activate and release free radicals and chemotactic substances, which damage the tissue, leading to tissue death[40]. In other studies that translated the REVECH 2.0, there was also discussion regarding the description of exudate[23-24], as occurred among the nurses participating in the present study. In the first translation to Brazilian Portuguese, the description of this item was changed to “pequena”, “média” and “grande quantidade”, according to the opinion of the experts[25]. In the description of the item exudate in RESVECH 2.0, the present translation resulted in the terms: “seco”, “úmido”, “molhado”, “saturado” and “com fuga de exsudato”, as the authors did not authorize the alteration of the response options. This set provides more options, allowing a more refined assessment of the evolution of the amount of exudate. This means that a small change in the amount of exudate could be observed through RESVECH 2.0 when one evaluation indicates “com fuga de exsudato” and the subsequent one indicates “saturado”. Both options would be reported as “grande quantidade” if using the PUSH. The evaluation and description of the saturation of dressings can be an advantage, since, when compared to the PUSH, it admits broader answers for situations of high exuding wounds. In VUs, this may occur due to prolonged periods with legs down, low adherence to compression therapy and congestive heart failure, phenomena related to increased capillary permeability and osmotic hydrostatic pressure[41]. The category “infection/inflammations” has 14 items, including pain assessment, allowing a refined evaluation[3,42]. It includes aspects commonly observed in the care of people with venous ulcers, therefore, no difficulties in understanding were mentioned in the cognitive debriefing. The analysis of the factor model indicates the validity of the construct, with aspects 1 to 5 consisting of isolated items and aspect 6 “infection/inflammation” encompassing 14 items. Although item 2 (“depth/tissues involved”) had a loading factor within the acceptable limit, little correlation was observed with the other items on the scale. This can be explained by the fact that, as verified in the present study and mentioned in other studies[20,41], most VUs are superficial, with involvement of the subcutaneous tissue. It is a predominant, relatively stable condition, whose evolution would be epithelialization itself. In this sense, although clinically relevant, the assessment of this item remains stable, while other items, such as the wound area, evolve substantially, contributing to the absent correlation in the case of venous ulcers. The REVECH 2.0 depth evaluation includes an option that is not very applicable to venous ulcers, the destruction of tissues reaching the muscles, tendons, or bones. It is a condition that can occur in pressure injuries, or in arterial or mixed ulcers, but is not common in VUs[20,41]. In turn, the presence of necrotic/devitalized tissue, especially slough, is a predominant characteristic of VUs, as shown in this and other studies[22,41,43], which may explain the fact that, despite presenting a high factor loading (0.589), item 4 “type of tissue in the wound bed” had little correlation with the other items on the scale. In the domain 6 of the scale, items 6.5, 6.13 and 6.14 (“increasing exudate”, “satellite lesions” and “paleness of the tissue do tecido” respectively), presented lower factor loadings, indicating that they may not be significantly associated with the other items of the scale[33] considering people with venous ulcers, who are the specific population studied. The International Wound Infection Institute[42] indicates the following signs/symptoms infection, based on consensus: erythema, local warmth, edema, purulent discharged, delayed wound healing, increasing pain, increasing malodor. This data corroborates the findings of the present study, which showed that these items have significant factor loadings, demonstrating their relevance for wound assessment and clinical decision-making. In the case of venous ulcers, wound infections must be managed[42], as they can reduce the chance of healing by up to $42\%$[44] and lead to new or increasing pain[45]. The Cronbach’s alpha coefficient in the factor analysis model indicated that the internal consistency of the Brazilian Portuguese version of RESVECH 2.0 was 0.832. This result cannot be compared to Cronbach’s alpha coefficients obtained in studies that did not use a factorial model. Having made this observation, the value obtained in the present study meets the recommended values, that is, above 0.70[37], and is within the $95\%$ confidence interval ($95\%$CI = 0.780-0.880), indicating good internal consistency and demonstrating the cohesion and coherence of the items to assess the construct validity of the RESVECH 2.0, that is, the conditions of the healing process in VU cases. The criterion validity was supported by a coefficient of 0.74 between RESVECH 2.0 and PUSH 3.0, indicating a strong correlation and reiterating the relevance of RESVECH 2.0 - adapted version for clinical practice in the care of people with chronic wounds, which include venous ulcers. Considering that care for people with venous ulcers is predominantly provided in the outpatient care network, in primary care, where staff turnover is high, the use of a reliable and standardized instrument to assess the healing of these lesions can contribute to effective communication between professionals, supporting the evaluation of the outcomes of the care provided and the decision-making process regarding the care technologies to be used. ## Study limitations A limitation of this study was that it was carried out in only two setting, in a country of continental size such as Brazil. However, the fact that it was carried out in more than one region of the country is already an advance. ## Contributions to the Area Considering that care for people with venous ulcers is predominantly provided in the outpatient care network, in primary care, where staff turnover is high, the use of a reliable and standardized instrument to assess the healing of these lesions can contribute to effective communication between professionals, supporting the evaluation of the outcomes of the care provided and the decision-making process regarding the care technologies to be used. ## CONCLUSIONS A robust adaptation of the instrument to Brazilian Portuguese, with good reliability (internal consistency) and criterion-related validity was produced. The scale had appropriate psychometric properties for use in the country in the evaluation of venous ulcers. It will allow the evaluation of lesions with areas larger than 24 cm2 and the integration of signs and symptoms of infection and pain assessment, which is of great clinical relevance for monitoring the healing of this type of wounds. ## Unknown https://doi.org/10.48331/scielodata.RCMAU5 ## References 1. Eberhardt RT, Raffetto JD.. **Chronic venous insufficiency**. *Circulation* (2014) **130** 333-346. DOI: 10.1161/CIRCULATIONAHA.113.006898 2. Atkin L, Bucko Z, Conde Montero E, Cutting K, Moffatt C, Probst A. **Implementing TIMERS : the race against hard-to-heal wounds Inflammation / infection Social factors Edge Regeneration Moisture Tissue**. *J Wound Care* (2019) **28** S1-50. DOI: 10.12968/jowc.2019.28.Sup3a.S1 3. Wounds UK.. *Best Practice Statement: addressing complexities in the management of venous leg ulcers [Internet]* (2019) 4. Borges EL, Ferraz AF, Carvalho DV, Matos SS, Lima VLAN.. **Prevenção de Recidiva de úlcera varicosa: um estudo de coorte**. *Acta Paul Enferm* (2016) **1** 9-16. DOI: 10.1590/1982-0194201600003 5. Restrepo-Medrano JC, Soriano JV.. **Desarrollo de un índice de medida de la evolución hacia la cicatrización de las heridas crónicas**. *Gerokomos* (2011) **22** 176-183. DOI: 10.4321/S1134-928X2011000400005 6. Garbuio DC, Zamarioli CM, Silva NCM, Oliveira-Kumakura ARS, Carvalho EC.. **Instrumentos para avaliação da cicatrização de lesões de pele: revisão integrativa**. *Rev Eletrôn Enferm* (2018) **20** 16. DOI: 10.5216/ree.v20.49425 7. Cardinelli CC, Lopes LPN, Di Piero KC, Freitas ZMF.. **Instrumentos para avaliação de feridas: scoping review**. *Res Soc Dev* (2021) **10** 1-13. DOI: 10.33448/rsd-v10i11.19246 8. Maklebust J.. **PUSH Tool reality check: audience response. Pressure Ulcer Scale for Healing**. *Adv Wound Care* (1997) **10** 102-106. PMID: 9362592 9. Thomas D, Rodeheaver G, Bartolucci A, Franz R, Sussman C, Ferrell B. **Pressure ulcer scale for healing: derivation and validation of the PUSH tool. The PUSH Task Force**. *Adv Wound Care* (1997) **10** 96-101. PMID: 9362591 10. Arndt JV, Kelechi TJ.. **An overview of instruments for wound and skin assessment and healing**. *J Wound, Ostomy Cont Nurs* (2014) **41** 17-23. DOI: 10.1097/01.WON.0000438020.28853.c1 11. Santos V, Sellmer D, Massulo M.. **Inter rater reliability of Pressure Ulcer Scale for Healing (PUSH) in patients with chronic leg ulcers**. *Rev Latino-Am Enfermagem* (2007) **15** 391-396. DOI: 10.1590/S0104-11692007000300005 12. Almeida S, Moreira C, Salome G.. **Pressure Ulcer Scale for Healing in monitoring of wound healing in elderly patients with leg ulcer**. *Rev Bras Cir Plást* (2014) **29** 120-127. DOI: 10.5935/2177-1235.2014RBCP0020 13. Ratliff CR, Rodeheaver GT.. **Use of the PUSH tool to measure venous ulcer healing**. *Ostomy Wound Manag[Internet]* (2005) **51** 62-63 14. Malaquias SG, Bachion MM, Sant’Ana SMSC, Dallarmi CCB, Lino RS, Ferreira PS.. **Pessoas com úlceras vasculogênicas em atendimento ambulatorial de enfermagem: estudo das variáveis clínicas e sociodemográficas**. *Rev Esc Enferm USP* (2012) **46** 302-310. DOI: 10.1590/S0080-62342012000200006 15. Alves GR, Malaquias SG, Bachion MM.. **Pressure Ulcer Scale for Healing (PUSH): interobserver reliability in the evaluation of venous ulcers**. *Rev Eletrônica Enferm* (2018) **20**. DOI: 10.5216/ree.v20.49411 16. Janice H, Karen L, Ann-Marie M, Deirdre O, Lyndsay O, Pamela EH. **A prospective, multicenter study to validate use of the PUSH in patients with diabetic, venous, and pressure ulcers**. *Ostomy Wound Manag[Internet]* (2010) **56** 26-36 17. Santos V, Azevedo M, Silva T, Carvalho V, Carvalho V.. **Adaptação transcultural do pressure ulcer scale for healing (PUSH) para a língua portuguesa**. *Rev Latino-Am Enfermagem* (2005) **13** 305-313. DOI: 10.1590/S0104-11692005000300004 18. Choi E, Chin W, Wan E, Lam C.. **Evaluation of the internal and external responsiveness of the Pressure Ulcer Scale for Healing (PUSH) tool for assessing acute and chronic wounds**. *J Adv Nurs* (2016) **72** 1134-1143. DOI: 10.1111/jan.12898 19. Santos MC, Tibola J, Marques CMG.. **Tradução, revalidação e confiabilidade da Escala de Cicatrização de Vancouver para língua portuguesa - Brasil**. *Rev Bras Queimaduras [Internet]* (2014) **13** 26-30 20. Sant’Ana S, Nunes C, Santos Q, Bachion M, Malaquias S, Oliveira B.. **Úlceras venosas: caracterização clínica e tratamento em usuários atendidos em rede ambulatorial**. *Rev Bras Enferm* (2012) **65** 637-644. DOI: 10.1590/S0034-71672012000400013 21. Restrepo-Medrano J.. **Instrumentos de monitorización clínica y medida de la cicatrización en úlceras por presión (UPP) y úlceras de la extremidad inferior (UEI): desarrollo y validación de um índice de medida**. *Universidade de Alicante;* (2010) 22. Restrepo-Medrano J, Soriano J.. **Development of a wound healing index for chronic wounds**. *EWMA J[Internet]* (2012) **12** 39-44 23. Restrepo-Medrano JC.. **Validación para Colombia del índice RESVECH 2.0 para la valoración de cicatrización en heridas crónicas**. *Rev Av Salud* (2019) **3** 7-14. DOI: 10.21897/25394622.1748 24. Marques J.. *Adaptação cultural e validação para a população portuguesa de um instrumento de monitorização de feridas crónicas - Escala RESVECH 2.0[Internet]* (2015) 25. Domingues EAR, Carvalho MRF, Kaizer UAO.. **Adaptação transcultural de um instrumento de avaliação de feridas**. *Cogitare Enferm* (2018) **23** 1-9. DOI: 10.5380/ce.v23i3.54927 26. Coluci MZO, Alexandre NMC, Milani D.. **Construção de instrumentos de medida na área da saúde**. *Cienc Saude Coletiva* (2015) **20** 925-936. DOI: 10.1590/1413-81232015203.04332013 27. **Conselho Nacional de Saúde**. *Resolução No 466: Dispõe sobre diretrizes e normas regulamentadoras de pesquisas envolvendo seres humanos* (2012) 28. Beaton DE, Guillemin F.. **Guidelines for the Process of Cross-Cultural Adaptation of Self-Report Measures**. *Spine (Phila Pa 1976)* (2000) **25** 3186-3191. PMID: 11124735 29. Beaton D, Bombardier C, Guillemin F, Ferraz M.. **Recommendations for the Cross-Cultural Adaptation of the DASH & QuickDASH Outcome Measures**. *Institute for Work & Health* (2007) 1-45 30. Wild D, Grove A, Martin M, Eremenco S, McElroy S, Verjee-Lorenz A. **Principles of Good Practice for the Translation and Cultural Adaptation Process for Patient-Reported Outcomes (PRO) Measures: report of the ISPOR Task Force for Translation and Cultural Adaptation**. *VALUE Health* (2005) **8** 94-104. DOI: 10.1111/j.1524-4733.2005.04054.x 31. Echevarría-Guanilo ME, Gonçalves N, Romanoski PJ.. **Propriedades psicométricas de instrumentos de medidas: bases conceituais e métodos de avaliação - Parte II**. *Texto Contexto Enferm* (2019) **8** 1-14. DOI: 10.1590/1980-265X-tce-2017-0311 32. Bertolucci P, Brucki SM, Campacci SR, Juliano Y. O. **Mini-Exame do Estado Mental em uma população geral: impacto da escolaridade**. *Arq Neuropsiquiatr* (1994) **52** 1-7. DOI: 10.1590/S0004-282X1994000100001 33. Hongyu K.. **Análise Fatorial Exploratória: resumo teórico, aplicação e interpretação**. *Eng Sci* (2018) **7** 88-103. DOI: 10.18607/ES201877599 34. Lira SA, Chaves A.. **Coeficientes de correlação para variáveis ordinais e dicotômicas derivados do coeficiente linear de pearson**. *RECIE [Internet]* (2006) **15** 45-53 35. Bartlett M.. **Tests of significance in factor analysis**. *Br J Stat Psychol* (1950) **3** 77-85 36. Kaiser HF.. **An index of factorial simplicity**. *Psychometrika* (1974) **39** 31-36 37. Souza AC, Alexandre NMC, Guirardello EB.. **Propriedades psicométricas na avaliação de instrumentos: avaliação da confiabilidade e da validade**. *Epidemiol Serv Saude* (2017) **26** 649-659. DOI: 10.5123/S1679-49742017000300022 38. Aldunate JLCB, Isaac C, Ladeira PRS de C, Fernandes V, Ferreira Castro M. **Úlceras venosas em membros inferiores**. *Rev Med* (2010) **89** 158-163. DOI: 10.11606/issn.1679-9836.v89i3/4p158-163 39. Oliveira B, Nogueira G, Carvalho M, Abreu A.. **Caracterização dos pacientes com úlcera venosa acompanhados no Ambulatório de Reparo de Feridas**. *Rev Eletrôn Enferm* (2012) **14** 156-163. DOI: 10.5216/ree.v14i1.10322 40. Swanson T, Angel D, Sussman G, Cooper R, Haesler E, Ousey K. **International Wound Infection Institute (IWII)**. *J Stomal Ther Aust* (2016) **32** 41. Wounds UK.. *Best Practice Statement: holistic management of venous leg ulceration* (2015) 42. **Wound Infection in clinical practice: principles of best practice**. *Wounds International;* (2022) 52 43. Sergio FR, Silveira IA, Oliveira BGRB.. **Avaliação clínica de pacientes com úlceras de perna acompanhados em ambulatório**. *Esc Anna Nery* (2021) **25** 1-6. DOI: 10.1590/2177-9465-ean-2020-0139 44. Scotton MF, Miot HA, Abbade LPF.. **Factors that influence healing of chronic venous leg ulcers: a retrospective cohort**. *An Bras Derm* (2014) **89** 414-422. DOI: 10.1590/abd1806-4841.20142687 45. Ligi D, Mosti G, Croce L, Raffetto J, Mannello F.. **Chronic venous disease - Part I: inflammatory biomarkers in wound healing**. *Biochim Biophys Acta* (2016) **1862** 1964-1974. DOI: 10.1016/j.bbadis.2016.07.018
--- title: Outcomes of Patients Hospitalized for Acute Diverticulitis With Comorbid Generalized Anxiety Disorder journal: Cureus year: 2023 pmcid: PMC10042514 doi: 10.7759/cureus.35461 license: CC BY 3.0 --- # Outcomes of Patients Hospitalized for Acute Diverticulitis With Comorbid Generalized Anxiety Disorder ## Abstract Introduction Diverticular disease and anxiety disorders are common in the general population. Prior research on diverticular disease showed that these patients have an increased frequency of anxiety and depression. The objective of this study was to explore the impact of generalized anxiety disorder (GAD) on the outcomes of adult patients admitted with acute diverticulitis. Methods Using the National Inpatient Sample database from the year 2014 and International Classification of Diseases, Ninth Edition Revision, Clinical Modification (ICD-9 CM) codes, acute diverticulitis patients were selected. The outcomes of diverticulitis patients with and without GAD were explored. The outcomes of interest included inpatient mortality, hypotension/shock, acute respiratory failure, acute hepatic failure, sepsis, intestinal abscess, intestinal obstruction, myocardial infarction, acute renal failure, and colectomy. A multivariate logistic regression analysis was performed to determine if GAD is an independent predictor for the outcomes. Results Among 77,520 diverticulitis patients in the study, 8,484 had comorbid GAD. GAD was identified as a risk factor for intestinal obstruction (adjusted odds ratio (aOR) 1.22, $95\%$ CI: 1.05-1.43, $p \leq 0.05$), and intestinal abscess (aOR 1.19, $95\%$ CI: 1.10-1.29, $p \leq 0.05$). GAD was found to be a protective factor for hypotension/shock (aOR 0.83, $95\%$ CI: 0.76-0.91, $p \leq 0.05$) and acute respiratory failure (aOR 0.76, $95\%$ CI: 0.62-0.93, $p \leq 0.05$). The aORs of sepsis, inpatient mortality, myocardial infarction, acute renal failure, and colectomy were not statistically significant. Conclusions Patients with acute diverticulitis who are also diagnosed with GAD are at increased risk for intestinal obstruction and intestinal abscess, which may be due to the influence GAD has on the gut microbiota as well as the impact of GAD pharmacotherapy on gut motility. There was also a decreased risk for acute respiratory failure and hypotension/shock appreciated in the GAD cohort which may be attributable to the elevated healthcare resource utilization seen generally in GAD patients, which may allow for presentation to the emergency department, hospitalization, and treatment earlier in the diverticulitis disease course. ## Introduction During colonoscopy, diverticulosis is a frequent finding in adult patients. The prevalence of diverticulosis is as high as ~$50\%$ among patients above the age of 60 years [1]. Approximately $10\%$ to $25\%$ of patients with diverticulosis develop a symptomatic disease during their lifetime. There are multiple non-modifiable risk factors for diverticulosis, which include advanced age and some genetic abnormalities such as Ehlers-Danlos and renal polycystic disease [2-4]. A particularly common non-modifiable risk factor for diverticulosis is the male sex [2]. Possible modifiable risk factors include elevated body mass index, smoking, and medication use including steroids, non-steroidal anti-inflammatory drugs, and opiates [2,5-8]. Less well-studied possible risk factors for diverticulosis include colonic dysmotility due to neuronal degeneration and altered colonic neuromuscular activity due to changes in serotonin signaling [2,9,10]. Complications of diverticular disease can include acute diverticulitis, abscess, fistula, bowel obstruction, and perforation. Prior studies demonstrated that a diagnosis of diverticular disease is associated with an elevated frequency of anxiety and depressive disorders [11]. Notably, anxiety and major depressive disorders are relatively well studied regarding their relationship with inflammatory bowel disease (IBD) and irritable bowel syndrome (IBS) with these patients having more severe disease and acute flares [12,13]. Generalized anxiety disorder (GAD) is considered a common psychiatric diagnosis with a lifetime prevalence of $4.6\%$ and $7.7\%$ in male and female patients of ages 18-64 years, respectively [14]. The specific pathophysiology of GAD is unclear; however, there is thought to be an association with the serotonin and noradrenergic systems [15]. Notably, while anxiety disorders are associated with worse outcomes in IBD and IBS patients, standard pharmacologic treatment for anxiety disorders can also impact the gastrointestinal tract. Pharmacologic treatment of GAD includes serotonin-norepinephrine reuptake inhibitors (SNRIs) and selective serotonin reuptake inhibitors (SSRIs) as first-line therapeutic interventions, with second-line agents including buspirone, benzodiazepines, and pregabalin and second-generation antipsychotics. While these therapeutics are intended to alleviate anxiety, several of these medications have known gastrointestinal adverse effects. In particular, SSRIs and SNRIs are theorized to affect gut motility due to their impact on serotonin receptors and serotonin levels [16]. The gastrointestinal side effect profile for SSRIs and SNRIs includes nausea, vomiting, diarrhea, and weight changes. Patients using SSRIs have also been found to be at increased risk of irritable bowel syndrome [17]. Despite the link between diverticular disease and anxiety disorders, there has been little research about the impact anxiety has on diverticulitis. Therefore, we aimed to identify the clinical outcomes of patients admitted for diverticulitis who also have comorbid GAD. This research was previously presented as a poster at the Annual American College of Gastroenterology conference on October 24, 2022. ## Materials and methods A retrospective cohort study was performed for all adult patients (defined as patients 18 years old and older) who were hospitalized due to diverticulitis in the year 2014. Institutional review board approval was not required for this research project in light of no patient-level data being utilized. The data were extracted from the National Inpatient Sample (NIS), a database developed for the Healthcare Cost and Utilization Project, which is sponsored by the Agency for Healthcare Research and Quality [18]. The NIS database is widely recognized as the biggest all-payer inpatient database in the United States of America. The International Classification of Diseases, Ninth Edition Revision, Clinical Modification (ICD-9 CM) codes were utilized to identify all of the diagnoses from the NIS database. The patients included in this study were stratified into two groups: those with a history of GAD and those who lack a history of GAD. Between these two groups, demographic information and data about their hospitalization including age, sex, race, length of stay, and hospitalization cost were extracted and subsequently compared. The Charlson comorbidity index, which is an established tool that is used to adjust for confounding variables, was also compared between these groups [19,20]. All of the statistical analyses were performed using Windows, Version 28 (Released 2021; IBM Corp., Armonk, New York, United States). The outcomes of interest collected for these two groups were hypotension/shock, myocardial infarction, acute renal failure, acute respiratory failure, acute hepatic failure, sepsis, intestinal abscess, intestinal obstruction, colectomy, and inpatient mortality. These outcomes were then compared between these groups. Means and proportions were compared using independent T-tests and chi-squared tests, respectively. The statistical analyses performed were two-tailed, with a p-value threshold of under 0.05 being considered statistically significant. Categorical variables were described as numbers (N) and percentages (%), while continuous variables were reported as means ± standard deviation (SD). A multivariate logistic regression analysis was also conducted to establish whether GAD is an independent predictor of the aforementioned outcomes, after age, sex race, and Charlson comorbidity index had been adjusted for. ## Results During the 2014 year, 77,520 adults were hospitalized due to diverticulitis. Among these diverticulitis patients, 8,484 of them had a history of GAD. As seen in Table 1, this subgroup of diverticulitis patients with GAD were younger (62.68 years old vs. 63.24 years old, $p \leq 0.05$), more likely to be female ($72.6\%$ vs. $43.5\%$, $p \leq 0.05$), more likely to be Caucasian ($83.9\%$ vs. $75.5\%$, $p \leq 0.05$), and had a longer length of stay (4.86 days vs. 4.53 days, $p \leq 0.05$). No statistically significant differences were identified in the Charlson comorbidity index (2.89 with GAD vs. 2.85 without, $$p \leq 0.15$$) and total hospital charge ($40,003.19 with GAD vs. $39,659.51 without, $$p \leq 0.54$$). **Table 1** | Unnamed: 0 | With generalized anxiety disorder | Without generalized anxiety disorder | p-value | | --- | --- | --- | --- | | N = 77,520 | N = 8,484 | N = 69,036 | | | Patient age, mean (SD) | 62.68 (14.62) | 63.24 (15.63) | <0.05 | | Sex, N (%) | | | <0.05 | | Female | 6,160 (72.6%) | 30,033 (43.5%) | | | Male | 2,321 (27.4%) | 38,973 (56.5%) | | | Race, N (%) | | | <0.05 | | White | 6,836 (83.9%) | 49,907 (75.5%) | | | Black | 429 (5.3%) | 6,292 (9.5%) | | | Hispanic | 678 (8.3%) | 7,217 (10.9%) | | | Asian or Pacific Islander | 32 (0.4%) | 907 (1.4%) | | | Native American | 21 (0.3%) | 271 (0.4%) | | | Other | 148 (1.8%) | 1,542 (2.3%) | | | Length of stay, in days (SD) | 4.86 (4.49) | 4.53 (4.45) | <0.05 | | Total hospital charges, in $ (SD) | 40,003.19 (46,843.99) | 39,659.51 (53,898.20) | 0.54 | | Charlson comorbidity index (SD) | 2.89 (0.02) | 2.85 (2.23) | 0.15 | In Table 2, the outcomes of diverticulitis patients with and without comorbid GAD were compared. The diverticulitis patients with a history of GAD had an increased likelihood of acute respiratory failure ($1.4\%$ vs. $1.1\%$, $p \leq 0.05$), and hypotension/shock ($7.1\%$ vs. $5.9\%$, $p \leq 0.05$). Diverticulitis patients without a history of GAD were more likely to have an intestinal abscess ($10.7\%$ vs. $9.0\%$, $p \leq 0.05$), colectomy ($2.8\%$ vs. $2.2\%$, $p \leq 0.05$), sepsis ($5.6\%$ vs. $5.1\%$, $p \leq 0.05$), and had a higher inpatient mortality ($0.6\%$ vs. $0.4\%$, $p \leq 0.05$). No statistically significant difference in intestinal obstruction ($$p \leq 0.12$$), acute renal failure ($$p \leq 0.08$$), and myocardial infarction ($$p \leq 0.39$$) was found between diverticulitis patients with and without comorbid GAD. Because of the small sample size for acute hepatic failure, further analysis of this outcome could not be performed. **Table 2** | Outcomes | With generalized anxiety disorder | Without generalized anxiety disorder | p-value | | --- | --- | --- | --- | | Intestinal obstruction | 51 (0.6%) | 329 (0.5%) | 0.12 | | Intestinal abscess | 761 (9.0%) | 7,354 (10.7%) | <0.05 | | Colectomy | 190 (2.2%) | 1,932 (2.8%) | <0.05 | | Sepsis | 430 (5.1%) | 3,871 (5.6%) | <0.05 | | Acute hepatic failure | * | 66 (0.1%) | 0.97 | | Acute respiratory failure | 119 (1.4%) | 765 (1.1%) | <0.05 | | Acute renal failure | 593 (7.0%) | 5,186 (7.5%) | 0.08 | | Myocardial infarction | 55 (0.6%) | 506 (0.7%) | 0.39 | | Hypotension/shock | 600 (7.1%) | 4,057 (5.9%) | <0.05 | | Inpatient mortality | 33 (0.4%) | 398 (0.6%) | <0.05 | The adjusted odds ratios (aORs) of the different outcomes, after controlling for the Charlson comorbidity index, race, sex, and age, are displayed in Table 3. GAD was subsequently identified as an independent risk factor for intestinal abscess (aOR 1.19, $95\%$ confidence interval (CI): 1.10-1.29, $p \leq 0.05$) and intestinal obstruction (aOR 1.22, $95\%$ CI: 1.05-1.43, $p \leq 0.05$). In addition, GAD was found to be a protective factor for acute respiratory failure (aOR 0.76, $95\%$ CI: 0.62-0.93, $p \leq 0.05$), and hypotension/shock (aOR 0.83, $95\%$ CI: 0.76-0.91, $p \leq 0.05$). The p-values for the aORs of sepsis (aOR 1.07, $95\%$ CI: 0.97-1.19, $$p \leq 0.19$$), inpatient mortality (aOR 1.34, $95\%$ CI: 0.93-1.92, $$p \leq 0.11$$), myocardial infarction (aOR 1.05, $95\%$ CI: 0.78-1.40, $$p \leq 0.77$$), acute renal failure (aOR 1.02, $95\%$ CI: 0.93-1.11, $$p \leq 0.76$$), and colectomy (aOR 0.75, $95\%$ CI: 0.55-1.02, $$p \leq 0.07$$) did not meet the cutoff for statistical significance. **Table 3** | Outcomes | Adjusted odds ratio* | 95% Confidence interval | p-value | | --- | --- | --- | --- | | Intestinal obstruction | 1.22 | 1.05-1.43 | <0.05 | | Intestinal abscess | 1.19 | 1.10-1.29 | <0.05 | | Colectomy | 0.75 | 0.55-1.02 | 0.07 | | Sepsis | 1.07 | 0.97-1.19 | 0.19 | | Acute respiratory failure | 0.76 | 0.62-0.93 | <0.05 | | Acute renal failure | 1.02 | 0.93-1.11 | 0.76 | | Myocardial infarction | 1.05 | 0.78-1.40 | 0.77 | | Hypotension/shock | 0.83 | 0.76-0.91 | <0.05 | | Inpatient mortality | 1.34 | 0.93-1.92 | 0.11 | While Table 2 and Table 3 outline the same outcomes, the data initially appear to be in conflict. For example, in Table 2, intestinal abscess is seen to occur less commonly in the GAD group with a statistically significant p-value. In comparison, Table 3 demonstrates this same outcome occurring more commonly in the GAD group with a statistically significant p-value. This difference can be explained by Table 3 displaying data that have been adjusted for many potential confounding factors. ## Discussion Previous studies showed that diverticular disease is associated with an elevated prevalence of anxiety and depressive disorders [12,13]. Despite this association, the impact anxiety has on medical outcomes of diverticulitis was unclear. This study is the first to investigate the outcomes of diverticulitis among inpatients with comorbid GAD. The findings in this study demonstrated that patients hospitalized due to diverticulitis who have comorbid GAD have an increased risk of intestinal abscess and intestinal obstruction. This represents a critical finding due to the increased mortality risk associated with intestinal obstruction and intestinal abscess [21,22]. These findings may be attributable in part to the underlying pathophysiology of GAD and the impact of the pharmacologic interventions used for GAD. Serotonin signaling in the enteric tract can modulate colonic peristalsis [23]. Close to $95\%$ of the body’s production of serotonin is attributed to the gastrointestinal epithelium [24]. Patients with GAD are noted to have an altered serotonergic signaling in the central nervous system and the use of SSRIs can increase the synaptic level of monoamines and lead to more activation of postsynaptic receptors [24]. Patients with diverticulitis have been observed to have a significant decrease in SERT (a serotonin transporter) transcript levels and have changes in attenuation of SERT expression and function [10]. Diverticular disease, possibly as a consequence, has also been associated with an increased number of serotonin-producing cells in the colonic mucosa [25]. Chronic SSRI and SNRI use may upregulate colonic serotonin signaling leading to increased colonic phasic contractility, which affects colonic motility [26]. Changes in colonic motility can lead to dysmotility, which may contribute to the increased risk of intestinal obstruction and intestinal abscess in diverticulitis patients with comorbid GAD. The other possible pathophysiology leading to increased complications of intestinal obstruction and abscess in patients with diverticulitis and GAD is dysregulation of the gut-brain axis. Animal models for depression have been seen to have altered enteric microbiota [24]. A study on patients with GAD noted that these patients also have altered enteric microbiota as compared to those without GAD [27]. While the relationship between enteric microbiota and acute diverticulitis requires further investigation, early studies show that elevated levels of certain intestinal bacteria, such as Subdoligranulum species and Marvinbryantia species, can lead to inflammation [28,29]. This increased colonic inflammation may possibly contribute to the elevated risk of intestinal obstruction and abscess in the GAD cohort [28]. Our study found that GAD was also associated with a decreased likelihood of acute respiratory failure and hypotension/shock. These outcomes may be less likely as a result of an earlier diagnosis and treatment of diverticulitis. In prior studies, patients who have comorbid GAD as well as other anxiety disorders were noted to have higher healthcare utilization, including primary care, emergency department visits, and hospitalizations [30]. Therefore, GAD patients may seek out medical care earlier and more frequently if they experience symptoms of diverticulitis, allowing intervention earlier in the disease process and decreasing the likelihood of progression to acute respiratory failure and hypotension/shock. There are several limitations to this study. One limitation of this study relates to the process of conducting research using the NIS database, which depends on billing codes input by medical providers, which may lack precision. Inaccurate billing code use may result in over or underrepresentation of the subgroup of diverticulitis patients with GAD. In addition, due to lack of ICD-9 coding for SSRI and SNRI use, there was no way to explore which patients were taking anti-anxiety medications at the time of their hospitalization. Another limiting factor is that the NIS database only includes hospitalized patients. Therefore, any acute diverticulitis patients cared for exclusively as outpatients were not captured. While this study has these limitations, a strength of this study is the ability to assess patient demographics and outcomes on a national scale. This research was also strengthened by the utilization of the multivariate logistic regression analysis, which was able to adjust for possible confounding variables. ## Conclusions In conclusion, acute diverticulitis patients with comorbid GAD had a higher risk of intestinal abscess and intestinal obstruction and a decreased risk for acute respiratory failure and hypotension/shock. Given that patients with GAD who present to the hospital with acute diverticulitis are more likely to have intestinal obstruction and abscess, providers should consider having a lower threshold to work up or start treatment for these pathologies as both are associated with elevated mortality. Further investigation into whether well-controlled GAD has improved outcomes in acute diverticulitis as compared to poorly controlled GAD would provide important information on inpatient healthcare providers in terms of risk stratification decisions in the hospital as well as guidance to outpatient providers about the urgency to initiate treatment for GAD patients who have known diverticulosis. ## References 1. Weizman AV, Nguyen GC. **Diverticular disease: epidemiology and management**. *Can J Gastroenterol* (2011) **25** 385-389. PMID: 21876861 2. Violi A, Cambiè G, Miraglia C. **Epidemiology and risk factors for diverticular disease**. *Acta Biomed* (2018) **89** 107-112. PMID: 30561403 3. Lederman ED, McCoy G, Conti DJ, Lee EC. **Diverticulitis and polycystic kidney disease**. *Am Surg* (2000) **66** 200-203. PMID: 10695753 4. Strate LL, Erichsen R, Baron JA. **Heritability and familial aggregation of diverticular disease: a population-based study of twins and siblings**. *Gastroenterology* (2013) **144** 736-742. PMID: 23313967 5. Strate LL, Liu YL, Aldoori WH, Syngal S, Giovannucci EL. **Obesity increases the risks of diverticulitis and diverticular bleeding**. *Gastroenterology* (2009) **136** 115-122. PMID: 18996378 6. Aune D, Sen A, Leitzmann MF, Tonstad S, Norat T, Vatten LJ. **Tobacco smoking and the risk of diverticular disease - a systematic review and meta-analysis of prospective studies**. *Colorectal Dis* (2017) **19** 621-633. PMID: 28556447 7. Yuhara H, Corley DA, Nakahara F. **Aspirin and non-aspirin NSAIDs increase risk of colonic diverticular bleeding: a systematic review and meta-analysis**. *J Gastroenterol* (2014) **49** 992-1000. PMID: 24221694 8. Kvasnovsky CL, Papagrigoriadis S, Bjarnason I. **Increased diverticular complications with nonsteriodal anti-inflammatory drugs and other medications: a systematic review and meta-analysis**. *Colorectal Dis* (2014) **16** 0-96 9. Simpson J, Sundler F, Humes DJ, Jenkins D, Scholefield JH, Spiller RC. **Post inflammatory damage to the enteric nervous system in diverticular disease and its relationship to symptoms**. *Neurogastroenterol Motil* (2009) **21** 847-840. PMID: 19453515 10. Costedio MM, Coates MD, Danielson AB, Buttolph TR 3rd, Blaszyk HJ, Mawe GM, Hyman NH. **Serotonin signaling in diverticular disease**. *J Gastrointest Surg* (2008) **12** 1439-1445. PMID: 18491196 11. Loosen SH, Paffenholz P, Luedde T, Kostev K, Roderburg C. **Diverticular disease is associated with an increased incidence rate of depression and anxiety disorders**. *Int J Colorectal Dis* (2021) **36** 2437-2443. PMID: 33937941 12. Häuser W, Janke KH, Klump B, Hinz A. **Anxiety and depression in patients with inflammatory bowel disease: comparisons with chronic liver disease patients and the general population**. *Inflamm Bowel Dis* (2011) **17** 621-632. PMID: 20848528 13. Nahon S, Lahmek P, Durance C, Olympie A, Lesgourgues B, Colombel JF, Gendre JP. **Risk factors of anxiety and depression in inflammatory bowel disease**. *Inflamm Bowel Dis* (2012) **18** 2086-2091. PMID: 22294486 14. Kessler RC, Petukhova M, Sampson NA, Zaslavsky AM, Wittchen H -U. **Twelve-month and lifetime prevalence and lifetime morbid risk of anxiety and mood disorders in the United States**. *Int J Methods Psychiatr Res* (2012) **21** 169-184. PMID: 22865617 15. Munir S, Takov V. **Generalized Anxiety Disorder**. *StatPearls [Internet]* (2022) 16. Wang SM, Han C, Bahk WM, Lee SJ, Patkar AA, Masand PS, Pae CU. **Addressing the side effects of contemporary antidepressant drugs: a comprehensive review**. *Chonnam Med J* (2018) **54** 101-112. PMID: 29854675 17. Lin WT, Liao YJ, Peng YC, Chang CH, Lin CH, Yeh HZ, Chang CS. **Relationship between use of selective serotonin reuptake inhibitors and irritable bowel syndrome: a population-based cohort study**. *World J Gastroenterol* (2017) **23** 3513-3521. PMID: 28596687 18. **Overview of the National (Nationwide) Inpatient Sample (NIS)**. (2021) 19. Austin SR, Wong YN, Uzzo RG, Beck JR, Egleston BL. **Why summary comorbidity measures such as the Charlson comorbidity index and Elixhauser score work**. *Med Care* (2015) **53** 0-72 20. Charlson ME, Pompei P, Ales KL, MacKenzie CR. **A new method of classifying prognostic comorbidity in longitudinal studies: development and validation**. *J Chronic Dis* (1987) **40** 373-383. PMID: 3558716 21. Chen XZ, Wei T, Jiang K. **Etiological factors and mortality of acute intestinal obstruction: a review of 705 cases**. *Chin J Integr Med* (2008) **6** 1010-1016 22. Gregersen R, Andresen K, Burcharth J, Pommergaard HC, Rosenberg J. **Long-term mortality and recurrence in patients treated for colonic diverticulitis with abscess formation: a nationwide register-based cohort study**. *Int J Colorectal Dis* (2018) **33** 431-440. PMID: 29511842 23. Spencer NJ, Hu H. **Enteric nervous system: sensory transduction, neural circuits and gastrointestinal motility**. *Nat Rev Gastroenterol Hepatol* (2020) **17** 338-351. PMID: 32152479 24. Jenkins TA, Nguyen JC, Polglaze KE, Bertrand PP. **Influence of tryptophan and serotonin on mood and cognition with a possible role of the gut-brain axis**. *Nutrients* (2016) **8** 56. PMID: 26805875 25. Banerjee S, Akbar N, Moorhead J, Rennie JA, Leather AJ, Cooper D, Papagrigoriadis S. **Increased presence of serotonin-producing cells in colons with diverticular disease may indicate involvement in the pathophysiology of the condition**. *Int J Colorectal Dis* (2007) **22** 643-649. PMID: 17086395 26. Tack J, Broekaert D, Corsetti M, Fischler B, Janssens J. **Influence of acute serotonin reuptake inhibition on colonic sensorimotor function in man**. *Aliment Pharmacol Ther* (2006) **23** 265-274. PMID: 16393306 27. Jiang HY, Zhang X, Yu ZH, Zhang Z, Deng M, Zhao JH, Ruan B. **Altered gut microbiota profile in patients with generalized anxiety disorder**. *J Psychiatr Res* (2018) **104** 130-136. PMID: 30029052 28. Tursi A, Scarpignato C, Strate LL, Lanas A, Kruis W, Lahat A, Danese S. **Colonic diverticular disease**. *Nat Rev Dis Primers* (2020) **6** 20. PMID: 32218442 29. Daniels L, Budding AE, de Korte N. **Fecal microbiome analysis as a diagnostic test for diverticulitis**. *Eur J Clin Microbiol Infect Dis* (2014) **33** 1927-1936. PMID: 24894339 30. Horenstein A, Heimberg RG. **Anxiety disorders and healthcare utilization: a systematic review**. *Clin Psychol Rev* (2020) **81** 101894. PMID: 32818687
--- title: Fasting Versus Non-Fasting Total Testosterone Levels in Women During the Childbearing Period journal: Cureus year: 2023 pmcid: PMC10042517 doi: 10.7759/cureus.35462 license: CC BY 3.0 --- # Fasting Versus Non-Fasting Total Testosterone Levels in Women During the Childbearing Period ## Abstract Background Total testosterone in men should be measured in the fasting state early in the morning with at least two samples according to guidelines. For women, no such a recommendation is available despite the importance of testosterone in this demographic. The aim of this study is to evaluate the effect of fasting versus non-fasting state on the total testosterone levels in women during the reproductive period. Methods This study was conducted at Faiha Specialized Diabetes, Endocrine and Metabolism Center in Basrah, (Southern Iraq) between January 2022 to November 2022. The total enrolled women were 109; their age was 18-45 years. The presentation was for different complaints; 56 presented for medical consultation with 45 apparently healthy women accompanying the patients as well as eight volunteering female doctors. Testosterone levels were measured by electrochemiluminescence immunoassays using the Roche Cobas e411 platform (Roche Holding, Basel, Switzerland). Two samples were collected from each woman; one was fasting and another was non-fasting the following day, and all samples were taken before 10 am. Results For all of the participants, the mean ± SD fasting was significantly higher as compared to the non-fasting testosterone (27.39±18.8 ng/dL and 24.47±18.6 ng/dL respectively, p-value 0.01). The mean fasting testosterone level was also significantly higher in the apparently healthy group, (p-value 0.01). In women who presented with hirsutism, menstrual irregularities and or hair fall, no difference was seen in the testosterone levels between fasting and non-fasting states (p-value 0.4). Conclusion In the apparently healthy women of childbearing age, serum testosterone levels were higher in the fasting versus the non-fasting states. In women who presented with complaints of hirsutism, menstrual irregularities, and or hair fall, the serum testosterone levels were not affected by the fasting states. ## Introduction The total testosterone levels in men should be measured in the fasting state in early in the morning with at least two samples according to guidelines [1,2]. In women no such solid recommendation has been made despite the importance of testosterone; it acts directly as an androgen and is an obligatory precursor for the synthesis of estradiol [3]. Testosterone exerts a physiological impact on reproductive and non-reproductive tissues in women. Hyperandrogenism is a cardinal feature for women with polycystic ovary syndrome (PCOS) and it is agreed to by all to be an important component for diagnosing this condition, in association with oligo-anovulation and/or polycystic ovarian morphology [4]. Rising evidence indicates PCOS with hyperandrogenism may have a more unfavourable metabolic profile than patients with the normal androgenic phenotype, suggesting a potential increment in the risk of cardiovascular disease [5]. The effect of endogenous androgens as a risk for type 2 diabetes is well established in women with documented hyperandrogenism as polycystic ovary syndrome (PCOS) is a condition linked to increased insulin resistance [6]. Sexual function in women is positively associated with the levels of testosterone [7]. During the reproductive period, testosterone is produced by the ovaries and by peripheral conversion of dehydroepiandrosterone (DHEA) and androstenedione; both are pre-androgens synthesized by the adrenal glands and ovaries [8]. About $50\%$ of testosterone is produced by ovaries and the adrenal glands ($25\%$ for each), while the peripheral conversion of androgens contributes equally to the rest of the testosterone that circulates in the blood [9]. Testosterone concentration begins to rise in girls, approximately at the age of six to eight years, when the adrenal zona reticularis maturation leads to increased production of DHEA and dehydroepiandrosterone sulfate (DHEA-S) [8], indicating the onset of adrenarche. With the onset of ovulation, cyclical production of testosterone by the ovaries begins when the concentrations peak in mid-cycle and stay high during the luteal phase [10]. Endogenous androgen concentration decreases steadily after the age of 30 years [11]. Women will lose about $60\%$ of their total androgen pool when reaching menopause [8]. Most circulating testosterone is conjugated with proteins; about $66\%$ is bound to sex hormone-binding globulin (SHBG), $30\%$ to albumin and only 2-$4\%$ of testosterone remains unbound (free testosterone), the form which is considered to be active. Plasma levels of androgens (testosterone and dihydrotestosterone (DHT) are very low in women, however, they have high androgen receptor affinity and therefore have solid androgenic properties, while the adrenal androgen DHEA-S, despite its high plasma concentrations has a low affinity for androgen receptors, and is considered mainly a precursor androgen [12]. To evaluate excess androgen as well as deficiency, reference ranges based on accurately measured levels of total testosterone are indispensable [7]. The highly sensitive and specific liquid chromatography-tandem mass spectrometry (LC-MS/MS) was evaluated to be superior to the conventional immunoassays for the estimation of the sex hormone at low concentrations, especially in women [13, 14]. However, measurement of testosterone using conventional immunoassays in clinical practice is appropriate if liquid chromatography and tandem mass spectrometry assay are not available, as recommended by Global Consensus Position Statement on the Use of Testosterone Therapy for Women [15]. The normal range of testosterone in women measured by conventional immunoassays during reproductive age is 15-46 ng/dL [16]. Evaluation of testosterone in men may be affected by circadian variation, whereas food intake mildly suppresses testosterone concentrations [17,18]. Various stressors may cause testosterone concentrations to fluctuate; up to $30\%$ of reductions can occur during the acute phase of illness [19]. Furthermore, several medications, like glucocorticoids, may also temporarily affect testosterone secretion. Testosterone concentrations in women vary according to the phase of the menstrual cycle and the body mass index, and this makes establishing normal ranges more cumbersome [20]. For increasing specificity, most societies recommend evaluating androgen excess in women during the follicular phase [21]. The aim of this study is to evaluate the effect of fasting versus non-fasting on the total testosterone levels in women during the reproductive period. ## Materials and methods Study design, place and time A cross-sectional study was conducted at the Faiha Specialized Diabetes, Endocrine and Metabolism Center (FDEMC) in Basrah, Southern Iraq, between January 2022 and November 2022. Participants *In this* study, 109 women were recruited, their ages ranging from 18-45 years. Of the total women, 56 presented for different medical consultations (18 of them for hirsutism, 25 with irregular menstrual cycles, 13 with hair fall), 45 apparently healthy women accompanying the patients as well as eight female doctors. Verbal consent was taken from each participant after explaining the aim of the study in accordance with the ethical standards of the FDEMC Research Committee, from which the ethical approval was obtained (ref #$\frac{60}{33}$/21), and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. The inclusion criteria included female gender and being between 18-45 years of age and the exclusion criteria were being below 18 years and above 45 years of age, pregnant women, women on oral contraceptive pills or other hormonal therapy, on antiandrogenic drugs, prolactinoma, pituitary surgery, or women with hypo- or hypergonadotrophic hypogonadism, premature ovarian insufficiency or history of oophorectomy, as well as those who did not attend the second appointment on the following day. Every woman was evaluated for measurement of height (meter), and weight (kilogram) from which the body mass index (BMI) was calculated. Biochemical analysis A random sample of venous blood is taken during fasting or not fasting to measure the level of total testosterone and other related hormones, as required to diagnose the diseased condition of the woman; then a second appointment to measure testosterone only is set in a reverse manner regarding fasting state. A sample of 5 ml of venous blood was taken at each appointment for the time 9:00-10:00 am and was labelled as fasting or non-fasting, accordingly. The above samples were not specifically collected for the study and were taken as part of a routine evaluation and management of the included patients. Total testosterone was measured from separated serum by electrochemiluminescence immunoassays (ECLIA), Roche® Cobas e411 platform (Roche Holding, Basel, Switzerland). The normal reference range of testosterone in women in this study was 15-45 ng/dL (coefficient variance <$6\%$) [16]. Statistical analysis *All data* were analyzed using the statistical package of the social science program (SPSS v. 26, IBM Corp., Armonk, NY). Mean was used as an expression of numerical data and the percentage was used to express nominal data. A two-sided paired T-test used for comparing means with (p-value <0.05) was considered to be significant. ## Results The means of age, body mass index (BMI), luteinizing hormone (LH), follicular stimulating hormone (FSH) and estradiol, with frequency (percentage) of marital status, complaints and diabetic status of the total number of 109 participants are presented in Table 1. Of these, 69 ($63.4\%$) were married and 40 ($36.6\%$) were unmarried. The mean age was 28.46±8.02 years, BMI 29.2±6.9 kg/m2, LH 6.88±5.69 mIU/mL, FSH 5.84±3.87 mIU/mL and estradiol 88.78±61.86 pg/dl. Twenty women had diabetes ($18.3\%$). Participants who were apparently healthy were 53 ($48.6\%$), and the other 56 ($51.4\%$) complained of hirsutism (18; $16.5\%$), irregular menstrual cycle (25; $22.9\%$) and hair fall (13; $11.9\%$). **Table 1** | General characteristics of participants(N=109) | General characteristics of participants(N=109).1 | General characteristics of participants(N=109).2 | | --- | --- | --- | | Age (years) | Mean ± SD | 28.46±8.02 | | Marital status, N (%) | Married | 69(63.4) | | Marital status, N (%) | Unmarried | 40(36.6) | | Complaints, N (%) | Hirsutism | 18(16.5) | | Complaints, N (%) | Irregular cycle | 25(22.9) | | Complaints, N (%) | Hair fall | 13(11.9) | | Complaints, N (%) | Apparently healthy | 53(48.6) | | DM, N (%) | Yes | 20(18.3) | | DM, N (%) | No | 89(81.7) | | BMI, (kg/m2) | Normal weight (%) | 27(24.8) | | BMI, (kg/m2) | Overweight (%) | 38(34.9) | | BMI, (kg/m2) | Obese (%) | 44(40.4) | | LH (mIU/mL) | Mean±SD | 6.88±5.69 | | FSH (mIU/mL) | Mean±SD | 5.84±3.87 | | Estradiol (pg/dL) | Mean±SD | 88.78±61.86 | For all of the participants, the mean±SD fasting was significantly higher as compared to the non-fasting testosterone (27.39±18.8 ng/dL and 24.47±18.6 ng/dL respectively, p-value 0.01) as shown in Table 2. The mean fasting testosterone level was also significantly higher in the apparently healthy group, (p-value 0.01). In women presenting with hirsutism, menstrual irregularities and or hair fall, no difference was seen in the testosterone levels between fasting and non-fasting states (p-value 0.4). **Table 2** | Group | Fasting testosterone (mean ± SD) | Non-fasting testosterone (mean ± SD) | Mean difference (mean ±SD) | P value | | --- | --- | --- | --- | --- | | All participants (N 109) | 27.39 ± 18.84 | 24.46 ± 18.62 | 2.92 ± 12.36 | 0.01 | | Apparently healthy (N 53) | 25.0 ± 15.82 | 20.38 ± 15.17 | 4.61 ± 12.59 | 0.01 | | Patients with hirsutism, irregular menstrual cycle, and or hair fall (56) | 29.65 ± 21.20 | 28.33 ± 20.78 | 1.32 ± 12.04 | 0.4 | To study the effect of BMI on fasting and non-fasting testosterone levels, the BMI was subdivided into three groups (18-24.9kg/m2, 25-30kg/m2 and >30kg/m2). The mean ± SD of fasting and non-fasting testosterone was 27.39±18.8 and 24.47±18.6 respectively. Women with obesity tend to have higher mean testosterone levels as compared to other BMI groups both in fasting and non-fasting state. However, these differences were not significant. The fasting testosterone level was higher independent of BMI categories (Table 3). **Table 3** | BMI | Fasting testosterone | Fasting testosterone.1 | Non-fasting testosterone | Non-fasting testosterone.1 | | --- | --- | --- | --- | --- | | BMI | Number | Mean±SD | Number | Mean±SD | | Normal weight | 27 | 26.63± 18.87 | 27 | 22.24± 13.98 | | Overweight | 38 | 25.20± 18.06 | 38 | 22.62± 16.18 | | Obese | 44 | 29.74± 19.63 | 44 | 27.42± 22.64 | | Total | 109 | 27.392±18.8 | 109 | 24.469±18.6 | | p-value | 0.243 | 0.243 | 0.445 | 0.445 | ## Discussion The decision to measure total testosterone in women in the state of fasting or not is important from the clinical point of view as food may affect testosterone levels. This was clearly stated by guidelines in men [1,2] but not in women. Despite the evidence that testosterone and other androgens are essential for reproductive function and health [22], there are a lot of obstacles to explaining the physiology of androgen in females. There were difficulties in accurately measuring low levels of testosterone in women; finding a study comparing and investigating the effect of fasting states on testosterone measurement between men and women was also difficult. This study showed that in the apparently healthy women, fasting testosterone was higher than non-fasting testosterone but not in women presented with hirsutism, menstrual irregularities, and or hair fall. We found fasting would increase total testosterone in women by around $9\%$. Compared to male studies, Tremellen et al. found that eating fast food, mixed meal or high fat produced a $25\%$ reduction in serum total testosterone during the first hour of eating, with levels continuing to be suppressed below the fasting level for about four hours, without significant effect on the levels of gonadotropin; however, consuming intravenous fat had no effect on testosterone levels, suggesting that the mechanism may be operated through the gastrointestinal tract that provokes an indirectly mediated response [23]. A decrease in circulating testosterone and free testosterone after a fat-containing meal was also described by Meikle et al [24]. In another study, Gagliano-Jucá et al. showed that testosterone reduced after oral ingestion of 75 g glucose, the mean drop had a nadir of 100.8 ng/dl, but levels appeared to be reactive and returning towards baseline within 120 min [25]. Similarly, Caronia et al. [ 26] also showed a reduction of $25\%$ in the level of testosterone following an oral glucose load; however, there was no change in cortisol, LH or SHBG levels. Food intake induces the release of gastro-pancreatic hormones that modulate splanchnic blood flow with the modulation of an affinity of testosterone to binding proteins. This lead to increased testosterone translocation to the peripheral tissues with higher plasma clearance. Another possible mechanism to explain is the presence of specific factors in food that may enhance the intracellular testosterone uptake directly or by testosterone complex formation, which is translocated into the cells [27]. These effects of meal intake in male studies appeared comparable to our study in the apparently healthy female group and not in the hyperandrogenism group. This finding may be explained by the differences between men and women in the testosterone level, mechanisms and major organs of secretion, compositions of the androgens, and the pathophysiology of female hyperandrogenism. A study comparing the change in testosterone levels as an effect of different food constituents in women with PCOS which was conducted by Katcher et al., showed a reduction in testosterone levels of $27\%$ within two hours after ingestion of a Western meal, or a low-fat, high-fat and high-fibre meal in women with PCOS. However, testosterone levels continue to drop for two hours more after the Western meal and are high-fat compared with the high-fibre and low-fat meal, indicating that the postprandial testosterone is affected by the meal composition independent of caloric load [28]. But in the presented study, no difference was seen in the testosterone level as a result of the fasting or non-fasting state in women complaining of hirsutism, menstrual irregularities, and or hair fall. In this study, random non-fasting samples were taken after the usual regular breakfast and not after a high-fat meal like in the previous study. The effect of BMI with fasting and non-fasting testosterone level showed no significant difference in this study, consistent with Panidis et al. [ 29] that reported that serum testosterone levels changes occur similarly after oral administration of 75 g dextrose to the normal weight, overweight and obese women, as a result of Parra et al [30]. This study has some limitations. First, the sample size was small. Second, the unavailability of gold standard testosterone measurement technique by LC/MS. Third, data from other ethnicities should form part of future studies. ## Conclusions In the apparently healthy women of childbearing age, serum testosterone levels were higher in the fasting versus the non-fasting states. In women who presented complaining of hirsutism, menstrual irregularities, and or hair fall, the serum testosterone levels were not affected by the fasting states. ## References 1. Bhasin S, Brito JP, Cunningham GR. **Testosterone therapy in men with hypogonadism: an endocrine society clinical practice guideline**. *J Clin Endocrinol Metab* (2018) **103** 1715-1744. PMID: 29562364 2. Hackett G, Kirby M, Edwards D. **British Society for Sexual Medicine guidelines on adult testosterone deficiency, with statements for UK practice**. *J Sex Med* (2017) **14** 1504-1523. PMID: 29198507 3. Simpson ER, Davis SR. **Minireview: aromatase and the regulation of estrogen biosynthesis--some new perspectives**. *Endocrinology* (2001) **142** 4589-4594. PMID: 11606422 4. Azziz R, Carmina E, Dewailly D. **The Androgen Excess and PCOS Society criteria for the polycystic ovary syndrome: the complete task force report**. *Fertil Steril* (2009) **91** 456-488. PMID: 18950759 5. Yang R, Yang S, Li R, Liu P, Qiao J, Zhang Y. **Effects of hyperandrogenism on metabolic abnormalities in patients with polycystic ovary syndrome: a meta-analysis**. *Reprod Biol Endocrinol* (2016) **14** 67. PMID: 27756332 6. Rubin KH, Glintborg D, Nybo M, Abrahamsen B, Andersen M. **Development and risk factors of type 2 diabetes in a nationwide population of women with polycystic ovary syndrome**. *J Clin Endocrinol Metab* (2017) **102** 3848-3857. PMID: 28938447 7. Wåhlin-Jacobsen S, Pedersen AT, Kristensen E. **Is there a correlation between androgens and sexual desire in women?**. *J Sex Med* (2015) **12** 358-373. PMID: 25475395 8. Davis SR, Panjari M, Stanczyk FZ. **Clinical review: DHEA replacement for postmenopausal women**. *J Clin Endocrinol Metab* (2011) **96** 1642-1653. PMID: 21411558 9. Labrie F, Luu-The V, Labrie C, Bélanger A, Simard J, Lin SX, Pelletier G. **Endocrine and intracrine sources of androgens in women: inhibition of breast cancer and other roles of androgens and their precursor dehydroepiandrosterone**. *Endocr Rev* (2003) **24** 152-182. PMID: 12700178 10. Scarabin-Carré V, Canonico M, Brailly-Tabard S. **High level of plasma estradiol as a new predictor of ischemic arterial disease in older postmenopausal women: the three-city cohort study**. *J Am Heart Assoc* (2012) **1** 0 11. Labrie F, Bélanger A, Cusan L, Gomez JL, Candas B. **Marked decline in serum concentrations of adrenal C19 sex steroid precursors and conjugated androgen metabolites during aging**. *J Clin Endocrinol Metab* (1997) **82** 2396-2402. PMID: 9253307 12. Labrie F, Martel C, Bélanger A, Pelletier G. **Androgens in women are essentially made from DHEA in each peripheral tissue according to intracrinology**. *J Steroid Biochem Mol Biol* (2017) **168** 9-18. PMID: 28153489 13. Taieb J, Mathian B, Millot F. **Testosterone measured by 10 immunoassays and by isotope-dilution gas chromatography-mass spectrometry in sera from 116 men, women, and children**. *Clin Chem* (2003) **49** 1381-1395. PMID: 12881456 14. Moal V, Mathieu E, Reynier P, Malthièry Y, Gallois Y. **Low serum testosterone assayed by liquid chromatography-tandem mass spectrometry. Comparison with five immunoassay techniques**. *Clin Chim Acta* (2007) **386** 12-19. PMID: 17706625 15. Davis SR, Baber R, Panay N. **Global consensus Position Statement on the use of testosterone therapy for women**. *J Clin Endocrinol Metab* (2019) **104** 4660-4666. PMID: 31498871 16. Braunstein GD, Reitz RE, Buch A, Schnell D, Caulfield MP. **Testosterone reference ranges in normally cycling healthy premenopausal women**. *J Sex Med* (2011) **8** 2924-2934. PMID: 21771278 17. Plymate SR, Tenover JS, Bremner WJ. **Circadian variation in testosterone, sex hormone-binding globulin, and calculated non-sex hormone-binding globulin bound testosterone in healthy young and elderly men**. *J Androl* (1989) **10** 366-371. PMID: 2592266 18. Mostafa T, Rashed L, Kotb K, Taymour M. **Effect of testosterone and frequent low-dose sildenafil/tadalafil on cavernous tissue oxidative stress of aged diabetic rats**. *Andrologia* (2012) **44** 411-415. PMID: 22489795 19. Muehlenbein MP, Hirschtick JL, Bonner JZ, Swartz AM. **Toward quantifying the usage costs of human immunity: Altered metabolic rates and hormone levels during acute immune activation in men**. *Am J Hum Biol* (2010) **22** 546-556. PMID: 20309883 20. Chang WY, Knochenhauer ES, Bartolucci AA, Azziz R. **Phenotypic spectrum of polycystic ovary syndrome: clinical and biochemical characterization of the three major clinical subgroups**. *Fertil Steril* (2005) **83** 1717-1723. PMID: 15950641 21. Rosner W, Auchus RJ, Azziz R, Sluss PM, Raff H. **Utility, limitations, and pitfalls in measuring testosterone: an Endocrine Society position statement**. *J Clin Endocrinol Metab* (2007) **92** 405-413. PMID: 17090633 22. Davis SR, Wahlin-Jacobsen S. **Testosterone in women--the clinical significance**. *Lancet Diabetes Endocrinol* (2015) **3** 980-992. PMID: 26358173 23. Tremellen K, Hill A, Pearce K. **Mechanistic insights into the aetiology of post-prandial decline in testosterone in reproductive-aged men**. *Andrologia* (2019) **51** 0 24. Meikle AW, Stringham JD, Woodward MG, McMurry MP. **Effects of a fat-containing meal on sex hormones in men**. *Metabolism* (1990) **39** 943-946. PMID: 2392062 25. Gagliano-Jucá T, Li Z, Pencina KM, Beleva YM, Carlson OD, Egan JM, Basaria S. **Oral glucose load and mixed meal feeding lowers testosterone levels in healthy eugonadal men**. *Endocrine* (2019) **63** 149-156. PMID: 30191441 26. Caronia LM, Dwyer AA, Hayden D, Amati F, Pitteloud N, Hayes FJ. **Abrupt decrease in serum testosterone levels after an oral glucose load in men: implications for screening for hypogonadism**. *Clin Endocrinol (Oxf)* (2013) **78** 291-296. PMID: 22804876 27. Lehtihet M, Arver S, Bartuseviciene I, Pousette A. **S-testosterone decrease after a mixed meal in healthy men independent of SHBG and gonadotrophin levels**. *Andrologia* (2012) **44** 405-410. PMID: 22524522 28. Katcher HI, Kunselman AR, Dmitrovic R, Demers LM, Gnatuk CL, Kris-Etherton PM, Legro RS. **Comparison of hormonal and metabolic markers after a high-fat, Western meal versus a low-fat, high-fiber meal in women with polycystic ovary syndrome**. *Fertil Steril* (2009) **91** 1175-1182. PMID: 18331737 29. Panidis D, Rousso D, Skiadopoulos S, Vavilis D, Karayannis B, Petropoulos P. **Does postprandial hypersinsulinemia contribute to hyperandrogenism in patients with polycystic ovary syndrome?**. *Clin Exp Obstet Gynecol* (1997) **24** 88-91. PMID: 9342471 30. Parra A, Barrón J, Mota-González M, Ibarra V, Espinosa de los Monteros A. **Serum androgen changes during meal-induced hyperinsulinemia and after acute sequential blockade and hyperstimulation of insulin release in women with polycystic ovary syndrome**. *Arch Med Res* (1995) **26 Spec No** 0-17
--- title: Chronic Kidney Disease of Unknown Etiology in a Tertiary Care Teaching Hospital journal: Cureus year: 2023 pmcid: PMC10042529 doi: 10.7759/cureus.35446 license: CC BY 3.0 --- # Chronic Kidney Disease of Unknown Etiology in a Tertiary Care Teaching Hospital ## Abstract Background Several primary studies have looked at the burden of chronic kidney disease among diabetic patients, but their results have shown significant variance in India. In order to determine the combined prevalence of chronic kidney disease and associated risk factors among patients with diabetes, this study used a combination of methods. Methods Over the course of two years, a cross-sectional observational study was undertaken in the Tertiary Care Teaching Hospital's Department of General Medicine including all chronic kidney disease patients of 18 years of age and above of either gender. People not suffering from the disease were chosen as controls. Kidney Injury Molecule-1 (KIM-1) and neutrophil gelatinase-associated lipocalin-ELISA (NGAL-ELISA) sample analysis by the kit method was done. The study was carried out in accordance with Schedule Y, ICH GCP principles, and the Helsinki Declaration after receiving approval from the institutional ethics committee. Results In our study, the urinary mean KIM-1 was 49.75±4.35 μg/g Cr in the Chronic Kidney Disease of Unknown etiology (CKDu) group and 1.43±0.15 μg/g Cr in the controls group. The mean NGAL levels of the CKDu Group and the controls group were 8.94±1.31 μg/g and 0.41±0.05 μg/g, respectively. In CKDu and the controls group, the mean eGFR (ml/min/1.73m2) was 69.83±7.91 and 108±3.7, respectively. The mean serum creatinine (mg/dL) was reported 3.79 in the CKDu group and 1.0 in the controls group. Conclusion Despite the urban centers previously being thought of as a non-endemic location, for the first time in the city, 60 CKDu patients are reported in this study. This is the first study to use the urinary biomarkers KIM-1 and NGAL to find suspected cases of CKDu and early kidney damage in local communities in the urban centers. ## Introduction A decreased glomerular filtration rate (GFR) of 60 ml/min/1.73 m2 for three months is stated to be chronic kidney disease (CKD), which is characterized as functional or structural anomalies of the kidney. It is a major health issue in global public health [1]. Globally, there were 697.5 million instances of all-stage CKD in 2017, and 1.2 million people per year expired as a result of expensive medical care [2]. Furthermore, it is predicted that being unable to obtain renal replacement therapy will cause between 2.3 and 7.1 million adult deaths before 2030 [3]. Particularly in Latin America, sub-Saharan Africa, and India, the prevalence of CKD has been rising [4]. Studies indicate that a number of factors, including obesity, advanced age, hypertension, diabetes mellitus, male gender, dyslipidemia, HIV infection, usage of nephrotoxic drugs, heavy alcohol use, smoking, family history of kidney disease, electrolyte, and acid-base abnormalities, low-paying employment use of conventional nonallopathic medications, and low hemoglobin, are to blame for CKD, despite the fact that the exact cause of the condition is still unknown [5]. Some of the possible risk factors can easily be identified and treated early, generally at a low cost. Patients with CKD frequently experience reduced quality of life, significantly higher healthcare expenses, and cardiovascular mortality, stroke, ischemic heart disease, gout, depression peripheral vascular disease, and anxiety are all at a higher risk [6]. The proximal tubule apical membrane expresses kidney injury molecule-1 (KIM-1), a type 1 transmembrane protein, when there is kidney injury [7]. KIM-1 was demonstrated to be an exceptional predictor of histological alterations in serum creatinine and blood urea nitrogen in the proximal tubule in response to various pathophysiological circumstances or toxicants [8]. The 25 kDa protein known as neutrophil gelatinase-associated lipocalin (NGAL) is a member of the lipocalin superfamily. Although many other cells, such as renal tubular cells, can create NGAL in response to numerous stresses, it was initially discovered in active neutrophils. Additionally, it has been found to be involved in kidney development and damaged tubular regeneration. Later clinical studies found urine NGAL to be an early marker of acute kidney injury (AKI). In light of recent data, it may also serve as an investigation for a number of various renal and non-renal disorders [9]. ## Materials and methods It is a prospective observational study done in the Department of General Medicine at Tertiary Care Teaching Hospital over a period of two years. Inclusion and exclusion criteria The inclusion and exclusion criteria are summarized in Table 1. **Table 1** | Inclusion Criteria | Exclusion Criteria | | --- | --- | | All chronic kidney disease patients above 18 years of age of either gender | Known hypertensive patients | | | Nephrotic or nephritic syndrome patients | | | Kidney damage caused by a snake bite or other nephrogenic toxins. | | | Urological disease of known etiology | Methodology The constituents of the study that are most probable to produce reliable results are organized in a systematic manner by the research methodology. This study provides a concise overview of the materials and procedures used to assess the prevalence of CKD with an unknown etiology among hospitalized patients. Sixty different CKDu patients from the city of Southern Province were used in the case group. The groups were segregated using the random sampling technique and the two groups were based on relevant factors such as age, gender, medical history, and other relevant variables to ensure that the two groups were as similar as possible, except for the intervention being tested. KIM-1 and NGAL-ELISA sample analysis (kit method) was used. Statistical analysis The gathered information was organized in a Microsoft Excel 2010 (Microsoft® Corp., Redmond, WA) spreadsheet, and SPSS version 27.0 (IBM Corp., Armonk, NY) was used to analyze it. Frequency and percentage were used to represent the qualitative data. Additionally, it was displayed using typical images like bar diagrams and pie charts. ## Results In our study, the subjects' demographics are presented in Table 2, the mean age of the subjects, and the percentage of males and females were statistically not significant differences between the CKD and control groups (all $P \leq 0.05$). In CKDu out of 60, 36 ($60\%$) males and 24 ($40\%$) females were in the CKDu group, and in the control group, 42 ($70\%$) were males and 18 were females ($30\%$) (Table 2). Most of the patients were 30-40 years old, i.e., 20 out of 60 ($33.3\%$), followed by 41-50 years old, i.e., 15 out of 60 ($25\%$) in the CKDu Group (Table 3). The characteristics of the patients in the CKDu group and the control group are shown in Table 4. The distribution of various CKD stages and the distribution of the urinary albumin/creatinine ratio (ACR) are mentioned in Table 5 and Table 6, respectively. In Table 7, we have summarized data of KIM-1 normalized to serum creatinine, ACR, KIM-1, NGAL, and eGFR. The urinary KIM-1 was 49.75±4.35 μg/g Cr in the CKDu group and 1.43±0.15 μg/g Cr in the controls group. NGAL levels of the CKDu group and controls group were 8.94±1.31 μg/g and 0.41±0.05 μg/g, respectively. eGFR (ml/min/1.73m2) in CKDu group and controls group were 69.83±7.91 and 108±3.7, respectively. The mean serum creatinine (mg/dL) in the CKDu group was 3.79 and in the controls group was 1.0. ## Discussion The present study helps to shed light on the function of innovative urine biomarkers (NGAL and KIM-1) in the earliest discovery of CKDu in Andhra Pradesh. Additionally, this is the first comparative study in India to examine chronic kidney disease with unknown etiology (CKDu) using a case definition based on WHO criteria. Males are more susceptible to CKDu in our study than females. In a similar study by Harambat et al. done in NCP, men are more likely to have CKDu ($6\%$) than women ($2.9\%$) [10]. The prevalence of CKDu was higher in males ($25.7\%$) than in females ($11.8\%$) in a study by Mills et al. too [11]. According to a meta-analysis-based inquiry that looked at 68 papers, men with non-diabetic renal illness had much faster kidney function decline over time than women [12]. Males' faster progression from the initial stages of kidney damage to the chronic phases of kidney injury was possibly caused by continual experience due to their rigorous environmental or occupational stressors [13]. As a result, the above-mentioned meta-analysis study excluded women and children and specifically concentrated on male farmers [12]. Our region, which is in the dry zone, uses farming methods that are very similar to those of the CKDu endemic region. As a result, the risk of CKDu appearing in our city is growing. Based on the WHO's classification of CKDu as well as elevated levels of KIM-1 and NGAL, we present here a study report of 60 different CKDu patients from the city of Southern Province. Diabetes, hypertension, pyelonephritis, renal calculi, etc. are a few examples of co-morbid conditions that may affect urine KIM-1 and NGAL levels [14]. In order to identify and remove instances with co-morbid conditions, we used questionnaires and assessments based on personal medical histories. A recognized initial non-invasive investigation to identify CKD is the measurement of albumin levels [15]. Albuminuria testing as a broad population indicator of renal disease is also supported by an epidemiological investigation [16]. The integrity of the kidneys' proximal tubules and glomerulus is determined by the amount of albumin in the urine [17], which is a gradient that clearly distinguishes the NGAL and KIM-1 values in the cases from the controls. *In* general, higher urinary KIM-1 levels may signify proximal tubular injury, whereas higher urinary NGAL levels may be caused by observable damage to the distal convoluted tubule and Henle loop. Drastically increased levels of NGAL and KIM-1 in the urine were also used to confirm instances with CKDu. Both markers made it simple to identify the suspected instances. Farmers in emerging regions of urban centers with normal serum creatinine and eGFR as well as in apparent good health displayed higher levels of NGAL and KIM-1 in comparison to the two control groups, suggesting probable initial kidney injury. It might imply that excessive albumin filtration precedes tubular damage expressed by KIM-1. Similar cases with high urine NGAL, IL-18, and NAG levels have been reported among Nicaraguan sugarcane cutters [18]. Recently, the detection of urinary KIM-1 utilising the micro-urine nanoparticle revealing approach has been reported, however, comparisons cannot be made because of the study's smaller sample size. Due to ischemia injury, KIM-1 is noticeably upregulated in the kidneys [19, 20]. However, only in CKDu with an eight-fold increase over the control group, did NGAL elevation become apparent. Similar findings were found by Bernardi et al., where NGAL levels in CKD sufferers were $26\%$ higher [21]. In another study, sugarcane farmers had 1.49 times greater NGAL, according to Alobaidi et al. [ 22]. However, no research utilising NGAL in local populations in India has been published. Elevated NGAL levels suggest that damaged tubules will be re-epithelialized, and iron that was lost owing to injury to proximal epithelial tubule cells will be absorbed, and iron-dependent nephrogenesis will be triggered subsequently [23]. Limitations The absence of established urine biomarker levels that suggest subclinical impairment in nephropathy in the local regions in which our study was carried out was a major study constraint. With the exception of a few recent studies, no prior research has been conducted in the region using comparable occupational cohorts. A follow-up study is necessary because it was also uncertain how each subject's short-term individual variance changed over time. Because only native male farmers in specific farming areas of our city were included in the current study, generalizing the results to the general population and other geographic places may be difficult. ## Conclusions This is the first study to use the urine biomarkers KIM-1 and NGAL to identify suspected instances of CKDu and to detect early kidney impairment in our local populations using the mentioned markers. According to the findings of our cross-sectional study, high urinary ACR levels were substantially linked with the tubular damage indicated by urinary NGAL and KIM-1 markers. Urinary tubular indicators, found in farming communities in the city, confirm tubulointerstitial illness with tubular injury. ## References 1. Kopple JD. **National kidney foundation K/DOQI clinical practice guidelines for nutrition in chronic renal failure**. *Am J Kidney Dis* (2001) **37** 0 2. Ebrahim Z, Glorieux G, Rafique Moosa M, Blaauw R. **Effect of simplified dietary advice on nutritional status and uremic toxins in chronic kidney disease participants**. *South Afr J Clin Nutr* (2022) **35** 168-176 3. Liyanage T, Ninomiya T, Jha V. **Worldwide access to treatment for end-stage kidney disease: a systematic review**. *Lancet* (2015) **385** 1975-1982. PMID: 25777665 4. Areses Trapote R, Sanahuja Ibáñez MJ, Navarro M. **Epidemiology of chronic kidney disease in Spanish pediatric population. REPIR II Project [Article in Spanish]**. *Nefrologia* (2010) **30** 508-517. PMID: 20613854 5. Okwuonu CG, Chukwuonye II, Adejumo OA, Agaba EI, Ojogwu LI. **Prevalence of chronic kidney disease and its risk factors among adults in a semi-urban community of South-East Nigeria**. *Niger Postgrad Med J* (2017) **24** 81-87. PMID: 28762361 6. De Broe ME. **Chinese herbs nephropathy and Balkan endemic nephropathy: toward a single entity, aristolochic acid nephropathy**. *Kidney Int* (2012) **81** 513-515. PMID: 22373701 7. Fangel MV, Nielsen PB, Kristensen JK, Larsen TB, Overvad TF, Lip GY, Jensen MB. **Albuminuria and risk of cardiovascular events and mortality in a general population of patients with type 2 diabetes without Ccardiovascular disease: a Danish cohort study**. *Am J Med* (2020) **133** 269-279 8. Halle MP, Essomba N, Djantio H. **Clinical characteristics and outcome of HIV infected patients with chronic kidney disease in Sub Saharan Africa: an example from Cameroon**. *BMC Nephrol* (2019) **20** 253. PMID: 31288761 9. Landau D, Schreiber R, Kleinman A, Vodonos A, Shalev H. **Pediatric chronic kidney disease rates in Southern Israel are higher than reported**. *F1000Res* (2013) **2** 186. PMID: 24555088 10. Harambat J, van Stralen KJ, Kim JJ, Tizard EJ. **Epidemiology of chronic kidney disease in children**. *Pediatr Nephrol* (2012) **27** 363-373. PMID: 21713524 11. Mills KT, Xu Y, Zhang W. **A systematic analysis of worldwide population-based data on the global burden of chronic kidney disease in 2010**. *Kidney Int* (2015) **88** 950-957. PMID: 26221752 12. Jha V, Garcia-Garcia G, Iseki K. **Chronic kidney disease: global dimension and perspectives**. *Lancet* (2013) **382** 260-272. PMID: 23727169 13. Gadde P, Sanikommu S, Manumanthu R, Akkaloori A. **Uddanam nephropathy in India: a challenge for epidemiologists**. *Bull World Health Organ* (2017) **95** 848-849. PMID: 29200526 14. Pourhoseingholi MA, Vahedi M, Rahimzadeh M. **Sample size calculation in medical studies**. *Gastroenterol Hepatol Bed Bench* (2013) **6** 14-17. PMID: 24834239 15. Belcher JM, Garcia-Tsao G, Sanyal AJ. **Association of AKI with mortality and complications in hospitalized patients with cirrhosis**. *Hepatology* (2013) **57** 753-762. PMID: 22454364 16. Wong F, O'Leary JG, Reddy KR. **New consensus definition of acute kidney injury accurately predicts 30-day mortality in patients with cirrhosis and infection**. *Gastroenterology* (2013) **145** 1280-1288. PMID: 23999172 17. Angeli P, Gines P, Wong F. **Diagnosis and management of acute kidney injury in patients with cirrhosis: revised consensus recommendations of the International Club of Ascites**. *Gut* (2015) **64** 531-537. PMID: 25631669 18. Elia C, Graupera I, Barreto R. **Severe acute kidney injury associated with non-steroidal anti-inflammatory drugs in cirrhosis: a case-control study**. *J Hepatol* (2015) **63** 593-600. PMID: 25872166 19. Emlet DR, Shaw AD, Kellum JA. **Sepsis-associated AKI: epithelial cell dysfunction**. *Semin Nephrol* (2015) **35** 85-95. PMID: 25795502 20. Wiest R, Lawson M, Geuking M. **Pathological bacterial translocation in liver cirrhosis**. *J Hepatol* (2014) **60** 197-209. PMID: 23993913 21. Bernardi M, Moreau R, Angeli P, Schnabl B, Arroyo V. **Mechanisms of decompensation and organ failure in cirrhosis: from peripheral arterial vasodilation to systemic inflammation hypothesis**. *J Hepatol* (2015) **63** 1272-1284. PMID: 26192220 22. Alobaidi R, Basu RK, Goldstein SL, Bagshaw SM. **Sepsis-associated acute kidney injury**. *Semin Nephrol* (2015) **35** 2-11. PMID: 25795495 23. Prowle JR, Bellomo R. **Sepsis-associated acute kidney injury: macrohemodynamic and microhemodynamic alterations in the renal circulation**. *Semin Nephrol* (2015) **35** 64-74. PMID: 25795500
--- title: 'Associations of ABO and Rhesus D blood groups with phenome-wide disease incidence: A 41-year retrospective cohort study of 482,914 patients' authors: - Peter Bruun-Rasmussen - Morten Hanefeld Dziegiel - Karina Banasik - Pär Ingemar Johansson - Søren Brunak journal: eLife year: 2023 pmcid: PMC10042530 doi: 10.7554/eLife.83116 license: CC BY 4.0 --- # Associations of ABO and Rhesus D blood groups with phenome-wide disease incidence: A 41-year retrospective cohort study of 482,914 patients ## Abstract ### Background: Whether natural selection may have attributed to the observed blood group frequency differences between populations remains debatable. The ABO system has been associated with several diseases and recently also with susceptibility to COVID-19 infection. Associative studies of the RhD system and diseases are sparser. A large disease-wide risk analysis may further elucidate the relationship between the ABO/RhD blood groups and disease incidence. ### Methods: We performed a systematic log-linear quasi-Poisson regression analysis of the ABO/RhD blood groups across 1,312 phecode diagnoses. Unlike prior studies, we determined the incidence rate ratio for each individual ABO blood group relative to all other ABO blood groups as opposed to using blood group O as the reference. Moreover, we used up to 41 years of nationwide Danish follow-up data, and a disease categorization scheme specifically developed for diagnosis-wide analysis. Further, we determined associations between the ABO/RhD blood groups and the age at the first diagnosis. Estimates were adjusted for multiple testing. ### Results: The retrospective cohort included 482,914 Danish patients ($60.4\%$ females). The incidence rate ratios (IRRs) of 101 phecodes were found statistically significant between the ABO blood groups, while the IRRs of 28 phecodes were found statistically significant for the RhD blood group. The associations included cancers and musculoskeletal-, genitourinary-, endocrinal-, infectious-, cardiovascular-, and gastrointestinal diseases. ### Conclusions: We found associations of disease-wide susceptibility differences between the blood groups of the ABO and RhD systems, including cancer of the tongue, monocytic leukemia, cervical cancer, osteoarthrosis, asthma, and HIV- and hepatitis B infection. We found marginal evidence of associations between the blood groups and the age at first diagnosis. ### Funding: Novo Nordisk Foundation and the Innovation Fund Denmark ## Introduction Still 100 years after the discovery of the ABO and Rhesus D (RhD) blood group systems, the selective forces that may have attributed to the observed blood group population differences remain elusive (Anstee, 2010). The pathophysiological mechanisms behind the observed relationship between blood groups and diseases are not well understood either. The ABO system has been associated with susceptibility to multiple diseases, including gastrointestinal- and cardiovascular diseases and pancreatic-, gastric-, and ovarian cancers (Vasan et al., 2016; Liumbruno and Franchini, 2014; Wolpin et al., 2010; Groot et al., 2020; Edgren et al., 2010; Dahlén et al., 2021; Li and Schooling, 2020). The ABO system has also been associated with the susceptibility, progression, and severity of COVID-19 (Ellinghaus et al., 2020). In contrast, apart from hemolytic disease of the newborn, reported associations between the RhD blood group and disease development are sparser (Anstee, 2010). Specifically, higher levels of factor VIII (FVIII) and von Willebrand factor (vWF) observed in individuals with a non-O blood group have been suggested to affect the development of cardiovascular disease (Jenkins and O’Donnell, 2006; Franchini and Lippi, 2015). Additionally, blood group-related antigens have been suggested to be involved in the adhesion of trophoblast, inflammatory cells, and metastatic tumor cells to the endothelial cells of the vasculature (Ravn and Dabelsteen, 2000). The endothelial cells of the vasculature have also been suggested to contribute to the initiation and propagation of severe clinical manifestations of COVID-19 (Teuwen et al., 2020). Recently, an associative disease-wide risk analysis of the ABO and RhD blood groups was conducted in a large Swedish cohort (Dahlén et al., 2021). The study generated further support for previous findings and suggested new associations. Here, we further uncover the relationship between the ABO and RhD blood groups and disease susceptibility using a Danish cohort of 482,914 patients. In contrast to previous studies, we use up to 41 years of follow-up data, and a disease categorization scheme specifically developed for disease-wide analysis called phecodes (Wu et al., 2019). Further, we determine the uniqueness of each individual ABO blood group as opposed to using blood group O as the reference. We estimate incidence rate ratios of 1312 phecodes (diagnoses) for the ABO and RhD blood groups. Further, we determine associations between the ABO/RhD blood groups and the age at the first diagnosis to better disclose the temporal life course element of disease development. ## Study design This retrospective cohort study was based on the integration of the Danish National Patient Registry (DNPR) and data on ABO/RhD blood groups of hospitalized patients. We included Danish patients who had their ABO/RhD blood group determined in the Capital Region or Region Zealand (covering ~$45\%$ of the Danish population Nordjylland, 2022), between January 1, 2006, and April 10, 2018. A blood type determination is commonly done for patients who may require a blood transfusion during hospitalization for example, anemic patients and women in labor. In the inclusion period, approximately $90\%$ of the population in the Capital Region and $97\%$ of the Region Zealand population were of European ancestry (Supplementary file 1). The DNPR provided the International Classification of Diseases 8th and 10th revision (ICD-8 and ICD-10) diagnosis codes, dates of diagnosis, date of birth, date of potential emigration, and sex of patients, with records dating back to 1977. Similar to a case-control study, the patients were included retrospectively. Here, selection into the study was based on an in-hospital ABO/RhD blood group determination. That is, the person-time and the entire disease history back to 1977 of patients hospitalized between 2006 and 2018 with known ABO/RhD blood groups were included retrospectively. We defined diseased and non-diseased individuals using the phecode mapping from ICD-10 diagnosis codes (Wu et al., 2019). Before categorizing the assigned ICD diagnosis codes into phecodes, the ICD-8 codes were converted to ICD-10 codes (Pedersen et al., 2023). Further, referral diagnoses were excluded. Pregnancy- and perinatal diagnosis (ICD-10 chapters 15–16) assigned before or after age 10 were excluded, or, when possible, rightly assigned to the mother or newborn, respectively. The disease categories of injuries, poisonings, and symptoms were deemed unlikely to be associated with the blood groups and excluded from the analyses (phecode categories: ‘injuries and poisonings’, ‘symptoms’ and phecodes above 999). Only phecodes with at least 100 cases in the study sample were included. The patients were followed from the entry in the DNPR to the date of death, emigration, the first event of the studied phecode, or end study period (April 10, 2018), whichever came first. Thus, follow-up was up to 41 years. The patients were allowed to contribute events and time at risk to multiple phecode analysis. ## Diagnosis-wide incidence rate ratios We used a log-linear quasi-Poisson regression model to estimate incidence rate ratios (IRRs) of each phecode among individuals with blood groups A, B, AB, and O relative to the other blood groups, respectively (e.g. A vs. B, AB, and O) (Dewey et al., 1995; Ver Hoef and Boveng, 2007). Further, we compared individuals with positive RhD type relative to negative RhD type. The analyses of diseases developed by both males and females were adjusted for sex, while analyses of sex-restricted diseases (e.g. cervical cancer) only included a subgroup of individuals of the restricted sex. Sex-restricted diseases were pre-defined by the phecode terminology. Sex was adjusted for as prior studies have found sex differences in the incidence rates of multiple diseases (Westergaard et al., 2019). Further, we adjusted for the year of birth and attained age, both modeled using restricted cubic splines with five knots. Attained age was split into 1 year intervals and treated as a time-dependent covariate, thus allowing individuals to move between categories with time. Herewith, age was used as the underlying time scale. Further, an interaction between attained age and sex was modeled for non-sex-restricted analyses. Patients were excluded from the analysis if they were assigned the phecode under study at the start of the DNPR. For analysis of congenital phecodes (e.g. sickle cell disease), prevalence ratios were estimated instead of IRRs by using the cohort size as the offset (see Supplementary file 2 for a list of the congenital phecodes). The analyses of ABO blood groups were adjusted for RhD type, and the RhD-analyses were adjusted for the ABO blood group. Adjustment for the birth year was done to control for societal changes and was used instead of the calendar period of diagnosis. The robust quasi-Poisson variance formula was used to control for over-dispersion (Ver Hoef and Boveng, 2007). We conducted a supplemental analysis using the same methodology but where blood group O was instead used as the reference to enable direct comparison and meta-analysis with previous studies. ## Age of first hospital diagnosis We estimated differences in age of first phecode of individuals with blood group A, B, AB, and O relative to any other blood group, respectively. Similar analyses were done for RhD-positive individuals relative to RhD negative individuals. We used a linear regression model adjusted for sex and birth year (as a restricted cubic spline with five knots). Analysis of sex-restricted phecodes was not adjusted for sex. Individuals who were assigned the studied phecode at the start date of the DNPR were excluded as the age of diagnosis was uncertain. Further, congenital- and pregnancy-related phecodes were not included. Statistical analyses were performed in R (version 3.6.2) using the survival and rms package. p-values were two-sided. p-values and confidence intervals were adjusted for multiple testing by the false discovery rate (FDR) approach, accounting for the number of performed tests (5 blood groups times 1312 phecodes; Benjamini and Hochberg, 1995; Altman and Bland, 2011). FDR adjusted p-values <0.05 were deemed statistically significant. The analysis pipeline was made in python (anaconda$\frac{3}{5.3.0}$) using snakemake for reproducibility (Köster and Rahmann, 2012). The analyses code is available through https://www.github.com/peterbruun/blood_type_study (copy archived at Bruun-Rasmussen, 2023). The manuscript complies with the STROBE reporting guidelines. ## Results In total, 482,914patients ($60.4\%$females) were included and 1312 phecodes (diagnosis codes) were examined (Figure 1, and Table 1). The median follow-up time for all phecode analyses was 17,555,322 person-years (Q1-Q3: 17,324,597–17,615,142). The cohort held a wide age distribution of patients born from 1901 to 2015 (Table 1, and Supplementary file 3). The ABO/RhD blood group distribution of the patients was similar to that of a previously summarized reference population of 2.2 million Danes (Table 1; Barnkob et al., 2020; Banks, 2022). **Figure 1.:** *Selection of patients for the 41-year retrospective cohort study on ABO/RhD blood groups and associations with disease incidence in 482,914 Danish patients.* TABLE_PLACEHOLDER:Table 1. ## Incidence rate ratios After adjustment for multiple testing, we found the incidence rate ratios (IRRs) of 101 and 28 phecodes (116 unique) to be statistically significant for the ABO and RhD blood groups, respectively. The statistically significant IRRs are given with $95\%$ confidence intervals in Table 2. The estimates of all examined phecodes are given in Supplementary file 4. Further, Manhattan plots of the p-values and disease categories are presented in Figures 2—6. **Figure 2.:** *Manhattan plot for blood group A with phecodes included by category.The vertical axis shows the -log10 transformed FDR adjusted p-values on a log10-scale. The horizontal axis shows the phecodes by category. The red line indicates the statistically significant level of <0.05 for FDR adjusted p-values. Associations with p-value >0.8 are not displayed. Coloured and annotated associations were deemed statistically significant. The direction of the triangles indicates positive or inverse associations (upward: IRR >1, downward: IRR <1). The size of the triangles indicates the size of the incidence rate ratio.* **Figure 3.:** *Manhattan plot for blood group B with phecodes included by category.The vertical axis shows the -log10 transformed FDR adjusted p-values on a log10-scale. The horizontal axis shows the phecodes by category. The red line indicates the statistically significant level of <0.05 for FDR adjusted p-values. Associations with p-value >0.8 are not displayed. Coloured and annotated associations were deemed statistically significant. The direction of the triangles indicates positive or inverse associations (upward: IRR >1, downward: IRR <1). The size of the triangles indicates the size of the incidence rate ratio.* **Figure 4.:** *Manhattan plot for blood group AB with phecodes included by category.The vertical axis shows the -log10 transformed FDR adjusted p-values on a log10-scale. The horizontal axis shows the phecodes by category. The red line indicates the statistically significant level of <0.05 for FDR adjusted p-values. Associations with p-value >0.8 are not displayed. Coloured and annotated associations were deemed statistically significant. The direction of the triangles indicates positive or inverse associations (upward: IRR >1, downward: IRR <1). The size of the triangles indicates the size of the incidence rate ratio.* **Figure 5.:** *Manhattan plot for blood group O with phecodes included by category.The vertical axis shows the -log10 transformed FDR adjusted p-values on a log10-scale. The horizontal axis shows the phecodes by category. The red line indicates the statistically significant level of <0.05 for FDR adjusted p-values. Associations with p-value >0.8 are not displayed. Coloured and annotated associations were deemed statistically significant. The direction of the triangles indicates positive or inverse associations (upward: IRR >1, downward: IRR <1). The size of the triangles indicates the size of the incidence rate ratio.* **Figure 6.:** *Manhattan plot for the Rhesus D blood group with phecodes included by category.The vertical axis shows the -log10 transformed FDR adjusted p-values on a log10-scale. The horizontal axis shows the phecodes by category. The red line indicates the statistically significant level of <0.05 for FDR adjusted p-values. Associations with p-value >0.8 are not displayed. Coloured and annotated associations were deemed statistically significant. The direction of the triangles indicates positive or inverse associations (upward: IRR >1, downward: IRR <1). The size of the triangles indicates the size of the incidence rate ratio.* TABLE_PLACEHOLDER:Table 2. The number of statistically significant IRRs for A, B, AB, O, and RhD were 50, 38, 11, 53, and 28, respectively. However, a between blood group comparison on the number of statistically significant IRRs is problematic because the analyses of blood group A and O had the highest power given that these blood groups were most frequent in the study sample (Table 1). For 13 phecodes, an association was found for both the ABO blood group and the RhD blood group. The ABO blood groups were found positively associated with 75 phecodes and inversely associated with 67 phecodes. The RhD-positive blood group was found to have 16 positive- and 12 inverse associations. Blood groups A and O were associated with diseases of the circulatory and digestive system. Blood group B was associated with several infectious, metabolic, and musculoskeletal diseases. The associations of the RhD blood group included cancers, infectious diseases, and pregnancy complications. The results of the supplementary analyses where blood group O was used as the reference is shown in Supplementary files 6 and 7. ## Age at first diagnosis We found the B blood group to be associated with a later diagnosis of viral infection. Further, blood group O was associated with a later diagnosis of phlebitis and thrombophlebitis (Table 3 and Supplementary file 5). The RhD-positive group was associated with a later diagnosis of acute and chronic tonsilitis diagnosis. **Table 3.** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | Blood group A | Blood group A.1 | Blood group B | Blood group B.1 | Blood group AB | Blood group AB.1 | Blood group 0 | Blood group 0.1 | Blood group RhD | Blood group RhD.1 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | Phecode | Phenotype | N | Estimate (95% CI) | p-value | Estimate (95% CI) | p-value | Estimate (95% CI) | p-value | Estimate (95% CI) | p-value | Estimate (95% CI) | p-value | | 079 | Viral infection | 25075 | –0.26 (–3.06, 2.54) | 0.864 | 0.92 (0.13, 1.71) | 0.022 | 0.53 (–6.18, 7.23) | 0.887 | –0.23 (–3.22, 2.75) | 0.887 | 0.27 (–5.22, 5.75) | 0.93 | | 451 | Phlebitis and thrombophlebitis | 16748 | –0.58 (–1.02,–0.13) | 0.011 | –0.24 (–5.71, 5.22) | 0.936 | –0.6 (–5.87, 4.66) | 0.833 | 0.91 (0.57, 1.25) | <0.001 | 0.01 (–0.33, 0.34) | 0.97 | | 451.2 | Phlebitis and thrombophlebitis of lower extremities | 15650 | –0.53 (–1.08, 0.01) | 0.055 | –0.27 (–5.85, 5.31) | 0.93 | –0.7 (–6.35, 4.95) | 0.82 | 0.9 (0.55, 1.25) | <0.001 | 0.08 (–3.89, 4.06) | 0.97 | | 474 | Acute and chronic tonsillitis | 41428 | –0.29 (–1.45, 0.87) | 0.634 | 0.42 (–1.97, 2.8) | 0.744 | 0.38 (–5.71, 6.48) | 0.909 | 0.05 (–2.15, 2.24) | 0.97 | 0.67 (0.15, 1.19) | 0.011 | | 474.1 | Acute tonsillitis | 18162 | 0.1 (–4.42, 4.61) | 0.97 | 0.41 (–7.17, 7.98) | 0.923 | 0.75 (–7.23, 8.74) | 0.864 | –0.42 (–3.76, 2.93) | 0.82 | 1.34 (0.64, 2.04) | <0.001 | ## Discussion We found the ABO/RhD blood groups to be associated with a wide spectrum of diseases including cancers and musculoskeletal-, genitourinary-, endocrinal-, infectious-, cardiovascular-, and gastrointestinal diseases. Associations of the ABO blood groups included monocytic leukemia, tonsilitis, renal dialysis, diseases of the female reproductive system, and osteoarthrosis. Associations of the RhD blood group included cancer of the tongue, malignant neoplasm (other), tuberculosis-, HIV-, hepatitis B infection, type 2 diabetes, hereditary hemolytic anemias, major puerperal infection, anxiety disorders, and contracture of tendon. The blood groups may reflect their corresponding genetic markers; thus, our findings may indicate an association between disease and the ABO locus on chromosome 9 and the RH locus on chromosome 1, respectively. Alternatively, the associations may indicate that the blood groups are involved in disease mechanisms at the molecular level mediated either through the blood group antigens or by the blood group reactive antibodies. However, our findings have a compromised causal interpretation given the retrospective inclusion of individuals (and person-time) after an in-hospital blood group test. Our results support several previously observed associations including positive associations between the non-O blood groups and prothrombotic diseases of the circulatory system (phecodes: 411.1–459.9), associations with gastroduodenal ulcers, associations of blood group O and lower risk of type 2 diabetes, and positive association between blood group B and tuberculosis (Vasan et al., 2016; Edgren et al., 2010; Dahlén et al., 2021; Fagherazzi et al., 2015; Rao, 2012). Further, our results support findings associating non-O blood groups with increased risk of pancreatic cancer (Liumbruno and Franchini, 2014). The role of the ABO blood group in HIV susceptibility remains controversial; we only observed a positive association for the RhD-positive blood group (Davison et al., 2020). We found blood group B to be positively associated with ‘ectopic pregnancy’, ‘excessive vomiting in pregnancy, and ‘abnormality of organs and soft tissues of pelvis complicating pregnancy’ indicating that blood group B mothers may be more likely to experience pregnancy complications. Further, we found positive associations of blood group A with both ‘mucous polyp of cervix’, and blood group AB with ‘cervicitis and endocervicitis’. Taken together these findings may indicate that the ABO blood groups are associated with diseases of the female reproductive system. However, the study design does not allow for any causal interpretation. Only a few statistically significant associations were found for the analyses of the age of the first diagnosis; thus, indicating that the blood group’s involvement in disease onset may be marginal. However, we assumed a linear relationship with age because assessing potential non-linear relationships for each disease would be unfeasible given the large number of tests performed. The linearity assumption may not hold for all analyses which limits the interpretation of the estimates. A strength of our approach is that we utilized the phecode disease classification scheme that is specifically developed for disease-wide risk analyses (Wu et al., 2019) The phecode mapping scheme combines ICD-10 codes that clinical domain experts have deemed to cover the same disease. For example, respiratory tuberculosis (A16), tuberculosis of nervous system (A17), and miliary tuberculosis (A19), are combined into the phecode tuberculosis (phecode 10). Phecodes may therefore provide increased power and precision compared with using ICD-10 categories (Denny et al., 2010). Further, contrary to previous studies, we compared each blood group to all other blood groups, instead of determining effect estimates relative to blood group O. Thus, here we better capture the uniqueness of each individual ABO blood group. ## Limitations Our study has some important limitations, firstly, the retrospective inclusion of patients and person-time may have introduced an immortal time bias from deaths before enrollment (in-hospital ABO/RhD blood group test) (Yadav and Lewis, 2021). The findings are therefore conditioned on patients surviving until the enrollment period. This implies, for example, that if a specific blood group causes a higher incidence of a deadly disease, then patients with such blood group are more likely to have died before enrollment, and therefore fewer individuals having both that blood group and the disease will be present in our cohort. If so, the direction of the estimates for deadly diseases strongly related to any blood group will have been lowered or even flipped, relative to any causal relationship. The study design, however, enabled 41 year of follow-up and was deemed reasonable because the blood groups have not been associated with mortality differences. Moreover, the blood group distribution in our cohort was found to be almost identical to a reference population of 2.2 million Danish blood donors. Further, we replicated several findings of associations between the blood groups and severe diseases, including pancreatic cancer (Vasan et al., 2016; Liumbruno and Franchini, 2014). This may indicate that the potential bias was less prevalent. Further, by controlling for year of birth, the potential effects of immortal time bias were likely reduced, however, this could not be tested. Immortal time biases are potentially applicable in many biobanks studies, e.g. when using the UK Biobank for retrospective studies (Yadav and Lewis, 2021). *The* generalizability of our findings is limited further because our cohort solely included hospitalized patients with known ABO and RhD blood groups. These are patients whom the treating doctor has deemed likely to potentially require a blood transfusion during hospitalization. The patients under study might therefore suffer from other diseases than patients without a determined blood group, and than never hospitalized individuals. Further, diseases that do not require hospitalization could not be examined. If the effect sizes are modified by factors which are more common in our cohort than in the general population then the estimates may not be generalizable. However, it is unclear if such effect modifier exists. Lastly, it was not possible to adjust for possible confounding from the geographical distribution or ethnicity of the patients (Anstee, 2010). This may have biased some estimates because the distribution of blood groups varies between ethnicities while ethnicity is also associated with differences in disease susceptibility. Particularly, ethnicity has been associated with differences in prevalence of infectious-, cardiovascular-, sickle cell disease, and thasalamia (Kurian and Cardarelli, 2007; McQuillan et al., 2004). Thus, the estimate of these disease groups should be interpreted with caution. The Danish population is however quite homogenous and approximately $94\%$ of Danes have European ancestry (Supplementary file 1). Therefore, a potential bias from ethnicity may be less prevalent in our cohort as compared with studies in populations of more admixed origin. In conclusion, we found the ABO/RhD blood groups to be associated with a wide spectrum of diseases, including cardiovascular-, infectious-, gastrointestinal- and musculoskeletal diseases. This may indicate that some of the potential selective pressure on the blood groups can be attributed to disease susceptibility differences. We found few associations between the blood groups and age of first diagnosis. ## Funding Information This paper was supported by the following grants: ## Data availability Anonymized patient data was used in this study. Due to national and EU regulations, the data cannot be shared with the wider research community. However, data can be accessed upon relevant application to the Danish authorities. The Danish Patient Safety Authority and the Danish Health Data Authority have permitted the use of the data in this study; whilst currently, the appropriate authority for journal data use in research is the regional committee ("Regionsråd"). The statistical summary data used to create the tables and graphs are available as Table 2—source data 1 and Table 3—source data 1. The analysis code is publicly available through https://www.github.com/peterbruun/blood_type_study (copy archived at Bruun-Rasmussen, 2023). ## References 1. Altman DG, Bland JM. **How to obtain the confidence interval from a P value**. *BMJ* (2011) **343**. DOI: 10.1136/bmj.d2090 2. Anstee DJ. **The relationship between blood groups and disease**. *Blood* (2010) **115** 4635-4643. DOI: 10.1182/blood-2010-01-261859 3. Banks TDB. **Blodtypes**. (2022) 4. Barnkob MB, Pottegård A, Støvring H, Haunstrup TM, Homburg K, Larsen R, Hansen MB, Titlestad K, Aagaard B, Møller BK, Barington T. **Reduced prevalence of SARS-cov-2 infection in ABO blood group O**. *Blood Advances* (2020) **4** 4990-4993. DOI: 10.1182/bloodadvances.2020002657 5. Benjamini Y, Hochberg Y. **Controlling the false discovery rate: A practical and powerful approach to multiple testing**. *Journal of the Royal Statistical Society* (1995) **57** 289-300. DOI: 10.1111/j.2517-6161.1995.tb02031.x 6. Bruun-Rasmussen P. *Software Heritage* (2023) 7. Dahlén T, Clements M, Zhao J, Olsson ML, Edgren G. **An agnostic study of associations between ABO and rhd blood group and phenome-wide disease risk**. *eLife* (2021) **10**. DOI: 10.7554/eLife.65658 8. Davison GM, Hendrickse HL, Matsha TE. **Do blood group antigens and the red cell membrane influence human immunodeficiency virus infection?**. *Cells* (2020) **9**. DOI: 10.3390/cells9040845 9. Denny JC, Ritchie MD, Basford MA, Pulley JM, Bastarache L, Brown-Gentry K, Wang D, Masys DR, Roden DM, Crawford DC. **PheWAS: Demonstrating the feasibility of a phenome-wide scan to discover gene-disease associations**. *Bioinformatics* (2010) **26** 1205-1210. DOI: 10.1093/bioinformatics/btq126 10. Dewey M, Clayton D, Hills M. **Statistical models in epidemiology**. *Journal of the Royal Statistical Society. Series A Statistics in Society* (1995) **158**. DOI: 10.2307/2983301 11. Edgren G, Hjalgrim H, Rostgaard K, Norda R, Wikman A, Melbye M, Nyrén O. **Risk of gastric cancer and peptic ulcers in relation to ABO blood type: a cohort study**. *American Journal of Epidemiology* (2010) **172** 1280-1285. DOI: 10.1093/aje/kwq299 12. Ellinghaus D, Degenhardt F, Bujanda L, Buti M, Albillos A, Invernizzi P, Fernández J, Prati D, Baselli G, Asselta R, Grimsrud MM, Milani C, Aziz F, Kässens J, May S, Wendorff M, Wienbrandt L, Uellendahl-Werth F, Zheng T, Yi X, de Pablo R, Chercoles AG, Palom A, Garcia-Fernandez A-E, Rodriguez-Frias F, Zanella A, Bandera A, Protti A, Aghemo A, Lleo A, Biondi A, Caballero-Garralda A, Gori A, Tanck A, Carreras Nolla A, Latiano A, Fracanzani AL, Peschuck A, Julià A, Pesenti A, Voza A, Jiménez D, Mateos B, Nafria Jimenez B, Quereda C, Paccapelo C, Gassner C, Angelini C, Cea C, Solier A, Pestaña D, Muñiz-Diaz E, Sandoval E, Paraboschi EM, Navas E, García Sánchez F, Ceriotti F, Martinelli-Boneschi F, Peyvandi F, Blasi F, Téllez L, Blanco-Grau A, Hemmrich-Stanisak G, Grasselli G, Costantino G, Cardamone G, Foti G, Aneli S, Kurihara H, ElAbd H, My I, Galván-Femenia I, Martín J, Erdmann J, Ferrusquía-Acosta J, Garcia-Etxebarria K, Izquierdo-Sanchez L, Bettini LR, Sumoy L, Terranova L, Moreira L, Santoro L, Scudeller L, Mesonero F, Roade L, Rühlemann MC, Schaefer M, Carrabba M, Riveiro-Barciela M, Figuera Basso ME, Valsecchi MG, Hernandez-Tejero M, Acosta-Herrera M, D’Angiò M, Baldini M, Cazzaniga M, Schulzky M, Cecconi M, Wittig M, Ciccarelli M, Rodríguez-Gandía M, Bocciolone M, Miozzo M, Montano N, Braun N, Sacchi N, Martínez N, Özer O, Palmieri O, Faverio P, Preatoni P, Bonfanti P, Omodei P, Tentorio P, Castro P, Rodrigues PM, Blandino Ortiz A, de Cid R, Ferrer R, Gualtierotti R, Nieto R, Goerg S, Badalamenti S, Marsal S, Matullo G, Pelusi S, Juzenas S, Aliberti S, Monzani V, Moreno V, Wesse T, Lenz TL, Pumarola T, Rimoldi V, Bosari S, Albrecht W, Peter W, Romero-Gómez M, D’Amato M, Duga S, Banales JM, Hov JR, Folseraas T, Valenti L, Franke A, Karlsen TH. **Genomewide association study of severe covid-19 with respiratory failure**. *The New England Journal of Medicine* (2020) **383** 1522-1534. DOI: 10.1056/NEJMoa2020283 13. Fagherazzi G, Gusto G, Clavel-Chapelon F, Balkau B, Bonnet F. **Abo and rhesus blood groups and risk of type 2 diabetes: Evidence from the large E3N cohort study**. *Diabetologia* (2015) **58** 519-522. DOI: 10.1007/s00125-014-3472-9 14. Franchini M, Lippi G. **The intriguing relationship between the ABO blood group, cardiovascular disease, and cancer**. *BMC Medicine* (2015) **13**. DOI: 10.1186/s12916-014-0250-y 15. Groot HE, Villegas Sierra LE, Said MA, Lipsic E, Karper JC, van der Harst P. **Genetically determined ABO blood group and its associations with health and disease**. *Arteriosclerosis, Thrombosis, and Vascular Biology* (2020) **40** 830-838. DOI: 10.1161/ATVBAHA.119.313658 16. Jenkins PV, O’Donnell JS. **Abo blood group determines plasma von Willebrand factor levels: A biologic function after all?**. *Transfusion* (2006) **46** 1836-1844. DOI: 10.1111/j.1537-2995.2006.00975.x 17. Köster J, Rahmann S. **Snakemake -- a scalable bioinformatics workflow engine**. *Bioinformatics* (2012) **28** 2520-2522. DOI: 10.1093/bioinformatics/bts480 18. Kurian AK, Cardarelli KM. **Racial and ethnic differences in cardiovascular disease risk factors: A systematic review**. *Ethnicity & Disease* (2007) **17** 143-152. PMID: 17274224 19. Li S, Schooling CM. **A phenome-wide association study of ABO blood groups**. *BMC Medicine* (2020) **18**. DOI: 10.1186/s12916-020-01795-4 20. Liumbruno GM, Franchini M. **Hemostasis, cancer, and ABO blood group: The most recent evidence of association**. *Journal of Thrombosis and Thrombolysis* (2014) **38** 160-166. DOI: 10.1007/s11239-013-1027-4 21. McQuillan GM, Kruszon-Moran D, Kottiri BJ, Curtin LR, Lucas JW, Kington RS. **Racial and ethnic differences in the seroprevalence of 6 infectious diseases in the United States: Data from NHANES III, 1988-1994**. *American Journal of Public Health* (2004) **94** 1952-1958. DOI: 10.2105/ajph.94.11.1952 22. Nordjylland R. **Regioner i Danmark**. (2022) 23. Pedersen MK, Eriksson R, Pedersen HK, Collin C, Reguant R, Simon C, Sørup FKH, Damgaard KA, Birch AM, Larsen M, Nielsen AP, Kirstine Belling SB. *Under Rev* (2023) 24. Rao D. **The ABO blood group distribution and pulmonary tuberculosis**. *Journal of Clinical and Diagnostic Research* (2012) **6** 943-946. DOI: 10.7860/JCDR.2012.4370.2298 25. Ravn V, Dabelsteen E. **Tissue distribution of histo-blood group antigens**. *APMIS* (2000) **108** 1-28. DOI: 10.1034/j.1600-0463.2000.d01-1.x 26. Teuwen LA, Geldhof V, Pasut A, Carmeliet P. **COVID-19: The vasculature unleashed**. *Nature Reviews. Immunology* (2020) **20** 389-391. DOI: 10.1038/s41577-020-0343-0 27. Vasan SK, Rostgaard K, Majeed A, Ullum H, Titlestad KE, Pedersen OBV, Erikstrup C, Nielsen KR, Melbye M, Nyrén O, Hjalgrim H, Edgren G. **ABO blood group and risk of thromboembolic and arterial disease: A study of 1.5 million blood donors**. *Circulation* (2016) **133** 1449-1457. DOI: 10.1161/CIRCULATIONAHA.115.017563 28. Ver Hoef JM, Boveng PL. **Quasi-poisson vs. negative binomial regression: How should we model overdispersed count data?**. *Ecology* (2007) **88** 2766-2772. DOI: 10.1890/07-0043.1 29. Westergaard D, Moseley P, Sørup FKH, Baldi P, Brunak S. **Population-wide analysis of differences in disease progression patterns in men and women**. *Nature Communications* (2019) **10**. DOI: 10.1038/s41467-019-08475-9 30. Wolpin BM, Kraft P, Gross M, Helzlsouer K, Bueno-de-Mesquita HB, Steplowski E, Stolzenberg-Solomon RZ, Arslan AA, Jacobs EJ, Lacroix A, Petersen G, Zheng W, Albanes D, Allen NE, Amundadottir L, Anderson G, Boutron-Ruault MC, Buring JE, Canzian F, Chanock SJ, Clipp S, Gaziano JM, Giovannucci EL, Hallmans G, Hankinson SE, Hoover RN, Hunter DJ, Hutchinson A, Jacobs K, Kooperberg C, Lynch SM, Mendelsohn JB, Michaud DS, Overvad K, Patel AV, Rajkovic A, Sanchéz MJ, Shu XO, Slimani N, Thomas G, Tobias GS, Trichopoulos D, Vineis P, Virtamo J, Wactawski-Wende J, Yu K, Zeleniuch-Jacquotte A, Hartge P, Fuchs CS. **Pancreatic cancer risk and ABO blood group alleles: Results from the pancreatic cancer cohort Consortium**. *Cancer Research* (2010) **70** 1015-1023. DOI: 10.1158/0008-5472.CAN-09-2993 31. Wu P, Gifford A, Meng X, Li X, Campbell H, Varley T, Zhao J, Carroll R, Bastarache L, Denny JC, Theodoratou E, Wei WQ. **Mapping icd-10 and icd-10-cm codes to phecodes: Workflow development and initial evaluation**. *JMIR Medical Informatics* (2019) **7**. DOI: 10.2196/14325 32. Yadav K, Lewis RJ. **Immortal time bias in observational studies**. *JAMA* (2021) **325** 686-687. DOI: 10.1001/jama.2020.9151
--- title: Transferred mitochondria accumulate reactive oxygen species, promoting proliferation authors: - Chelsea U Kidwell - Joseph R Casalini - Soorya Pradeep - Sandra D Scherer - Daniel Greiner - Defne Bayik - Dionysios C Watson - Gregory S Olson - Justin D Lathia - Jarrod S Johnson - Jared Rutter - Alana L Welm - Thomas A Zangle - Minna Roh-Johnson journal: eLife year: 2023 pmcid: PMC10042539 doi: 10.7554/eLife.85494 license: CC BY 4.0 --- # Transferred mitochondria accumulate reactive oxygen species, promoting proliferation ## Abstract Recent studies reveal that lateral mitochondrial transfer, the movement of mitochondria from one cell to another, can affect cellular and tissue homeostasis. Most of what we know about mitochondrial transfer stems from bulk cell studies and have led to the paradigm that functional transferred mitochondria restore bioenergetics and revitalize cellular functions to recipient cells with damaged or non-functional mitochondrial networks. However, we show that mitochondrial transfer also occurs between cells with functioning endogenous mitochondrial networks, but the mechanisms underlying how transferred mitochondria can promote such sustained behavioral reprogramming remain unclear. We report that unexpectedly, transferred macrophage mitochondria are dysfunctional and accumulate reactive oxygen species in recipient cancer cells. We further discovered that reactive oxygen species accumulation activates ERK signaling, promoting cancer cell proliferation. Pro-tumorigenic macrophages exhibit fragmented mitochondrial networks, leading to higher rates of mitochondrial transfer to cancer cells. Finally, we observe that macrophage mitochondrial transfer promotes tumor cell proliferation in vivo. Collectively these results indicate that transferred macrophage mitochondria activate downstream signaling pathways in a ROS-dependent manner in cancer cells, and provide a model of how sustained behavioral reprogramming can be mediated by a relatively small amount of transferred mitochondria in vitro and in vivo. ## Introduction It has been previously described that mitochondria can undergo lateral transfer between cells (Torralba et al., 2016; Antanavičiūtė et al., 2014; Lou et al., 2012; Rebbeck et al., 2011; Tan et al., 2015; Wang and Gerdes, 2012; Wang and Gerdes, 2015; Lampinen et al., 2022). Mitochondria are dynamic organelles, known to provide energy for the cell, but more recently shown to have a variety of additional essential cellular functions (Zong et al., 2016). In animal models, a series of seminal studies revealed that cancer cells void of mitochondrial DNA still form tumors by obtaining mitochondria from stromal cells, thereby restoring cancer cell mitochondrial function, cellular respiration, and tumor formation (Tan et al., 2015; Dong et al., 2017). Other experiments suggest that mitochondrial transfer not only restores bioenergetics, but can alter the metabolic state of recipient cells (Brestoff et al., 2021; Nicolás-Ávila et al., 2020; Phinney et al., 2015; Saha et al., 2022; Crewe et al., 2021; Korpershoek et al., 2022; Liu et al., 2022; van der Vlist et al., 2022; Yang et al., 2022; Liu et al., 2021), allowing recipient cells to adapt to stressors or changes in the environment, prompting the development of methods targeting mitochondrial dysfunction in disease (Patel et al., 2023; Caicedo et al., 2015). Although these studies elegantly demonstrate that mitochondrial transfer alters recipient cellular behavior, many aspects of this process remain unclear. For instance, the rescue of cellular function is commonly attributed to enhanced mitochondrial energetic or metabolic profiles; however, the fate and function of transferred mitochondria in recipient cells are under-explored. Furthermore, it is unclear how cells respond to laterally transferred mitochondria if the recipient cells already have a fully functioning mitochondrial network, and in particular, if the transferred mitochondria only comprise a small subset of the overall mitochondrial network in the recipient cell. Given that metastasis is a low-frequency event and is the consequence of changes in cellular behavior on the single-cell level, we aimed to examine the function and behavior of transferred mitochondria within individual recipient cells that have functioning endogenous mitochondrial networks. Using a combination of in vitro high-resolution microscopy, optogenetics, imaging flow cytometry, and in vivo tumor models, we demonstrate a previously undescribed mechanism of mitochondrial transfer-associated cellular reprogramming. Collectively, our data explain how a relatively small amount of transferred mitochondria can impact cellular behavior in the recipient cell with fully functioning endogenous mitochondria – Transferred macrophage mitochondria in cancer cells are dysfunctional, ROS accumulates at the site of transferred mitochondria, promoting ERK-mediated cancer cell proliferation. ## Cancer cells with macrophage mitochondria exhibit increased proliferation We previously reported that macrophages transfer cytoplasmic contents to cancer cells in vitro and in vivo (Roh-Johnson et al., 2017), and hypothesized that a macrophage/cancer cell system would be ideal for probing mitochondrial transfer in cells with functioning mitochondrial networks. Our studies employed blood-derived human macrophages and a human breast cancer cell line, MDA-MB-231 (231 cells), stably expressing a mitochondrially localized mEmerald or red fluorescent protein (mito-mEm or mito-RFP, respectively; Figure 1a). We observed mitochondrial transfer from macrophages to 231 cells using live cell confocal microscopy (Figure 1b, arrowheads) and flow cytometry (Figure 1c–d; flow cytometry scheme in Figure 1—figure supplement 1a). Control gates were set to $0.2\%$, based on confirmation of mitochondrial transfer by FACS-isolation of distinct mEmerald+ populations (see methods for more information). With these methods, a range of transfer efficiencies were observed, which we attribute to donor-to-donor variability (Figure 1d), yet mitochondrial transfer was consistently observed in 231 cells, as well as to another breast cancer line, MDA-MB-468, and a melanoma cell line, A375 (Figure 1—figure supplement 1b). To determine whether macrophage mitochondrial transfer was unique to cancer cells, we tested a non-malignant breast epithelial cell line, MCF10A. We observed reduced mitochondrial transfer efficiencies to MCF10A cells, with no significant differences compared to control (Figure 1—figure supplement 1c), suggesting that macrophages exhibit higher mitochondrial transfer efficiencies to malignant cells. Transferred mitochondria contain a key outer mitochondrial membrane protein, TOMM20 (Figure 1—figure supplement 1d, arrowhead) and mitochondrial DNA (Figure 1—figure supplement 1e, arrowhead), suggesting that intact organelles are transferred to 231 cells. To better define the requirements for transfer, we performed trans-well experiments in which we cultured 231 cells either physically separated from macrophages by a 0.4 μm trans-well insert or in contact with macrophages (scheme in Figure 1—figure supplement 1f), or with conditioned media (Figure 1—figure supplement 1g, h). These data showed that mitochondrial transfer increased dramatically under conditions where 231 cells could contact macrophages directly (Figure 1—figure supplement 1g and h). Taken together, these results suggest that macrophage mitochondrial transfer to cancer cells likely requires cell-to-cell contact. Furthermore, while mitochondrial transfer may not be unique to cancer cells, macrophages transfer mitochondria to cancer cells at higher frequencies. Thus, due to the low rates of mitochondrial transfer across macrophage donors ($0.84\%$, Figure 1d), we subsequently took advantage of single-cell, high-resolution approaches – rather than bulk approaches – to follow the fate and functional status of transferred mitochondria. **Figure 1.:** *Cell-contact-mediated transfer of macrophage mitochondria leads to increased cancer cell proliferation.(a) CD14+ monocytes harvested from human blood are transduced and differentiated for 6 days. Mito-mEm +macrophages (green) are co-cultured with MDA-MB-231 cells (231 cells) expressing mito-RFP (magenta; right image). (b) Confocal image showing transferred mitochondria (green, arrowhead) in a 231 cell (magenta, cell outline in white). (c) Representative flow cytometry plots depicting mitochondrial transfer (black box) within a population of co-cultured mito-RFP+ 231 cells (right) compared to monoculture control (left) with background level of mEmerald (mEm) fluorescence set at 0.2%. (d) Aggregate data of mitochondrial transfer rates across macrophage donors. Each data point represents one replicate (N=14 donors). (e) Analysis of proliferative capacity by quantifying Ki-67 levels and DNA content in co-cultured 231 cells after 24 hr. Percentage of cancer cells within a specific cell cycle phase with or without transfer is shown. A significantly different percent of recipient cells occupies G2/M (black) phases of the cell cycle compared to non-recipient cells (N=4 donors; statistics for G2/M only). (f) Co-cultured recipient 231 cells have a significantly higher specific growth rate compared to non-recipients (N=60 cells (control), 115 (recipient) over 4 donors indicated as shades of gray). (g) Schematic of mitochondrial isolation and bath application on MDA-MB-231 cells. Mitochondria are isolated from mito-mEmerald expressing THP-1 monocytes and bath applied at 20–30 µg/mL for 24 hr. Cancer cells which had taken up mEm+ mitochondria are then FACS-isolated and plated for 48 hr for further analyses. (h) Representative confocal image showing mito-RFP-expressing 231 cell (magenta) that had taken up macrophage mitochondria (green, grey arrow). (i) 48 hr after FACS-isolating 231 cells with macrophage mitochondria, flow cytometry was used to determine percent of daughter cells which still contain mEm+ mitochondria. N=3 biological replicates. (j), Cell cycle analysis of daughter cells 48 hr after FACS-isolation of 231 cells that had taken up macrophage mitochondria. N=3 biological replicates. For all panels, standard error of the mean (SEM) is displayed and scale bars are 10 µm. Mann-Whitney (d), two-way ANOVA (e, j), Welch’s t-test (f, i), *p<0.05; **p<0.01; ****p<0.0001.* To determine the effects of macrophage mitochondrial transfer on cancer cells, we performed single cell RNA-sequencing on cancer cells that received macrophage mitochondria. These data revealed that mitochondrial transfer induced significant changes in canonical cell proliferation-related pathways (Figure 1—figure supplement 2a). To follow up on the RNA-sequencing results, we used flow cytometry to evaluate proliferation changes, and found significant increases in the percent of cells within the G2 and Mitotic (M) phases of the cell cycle in recipient cells, as compared to their co-cultured counterparts that did not receive mitochondria (Figure 1e; Figure 1—figure supplement 2b-d). These cells were not undergoing cell cycle arrest, as we found that recipient cells completed cytokinesis at rates equivalent to their co-cultured non-recipient counterparts (Figure 1—figure supplement 2e). For further confirmation of this proliferative phenotype, we used quantitative phase imaging (QPI) to detect changes in dry mass of co-cultured 231 cells over time (Zangle and Teitell, 2014). With this approach, we could obtain growth rate information of a large number of cancer cells over time ($$n = 60$$ control cells; $$n = 115$$ recipient cancer cells). Consistent with the flow cytometry-based cell cycle analysis, the specific growth rates increased significantly in 231 cells with macrophage mitochondria compared to 231 cells that did not receive mitochondria (Figure 1f). To examine whether the effects of mitochondrial transfer was sustained in recipient cells, we also measured the growth rates of daughter cells born from recipient 231 cells containing macrophage mitochondria (Zangle et al., 2014). We identified five ‘parent’ cancer cells with macrophage mitochondria, for which we were able to reliably follow both daughter cells upon division. Daughter cells that inherited the ‘parent’s’ macrophage mitochondria exhibited an increase in their rate of change of dry mass over time versus sister cells that did not inherit macrophage mitochondria (Figure 1—figure supplement 3a-c). These experiments indicate that the proliferation phenotype in recipient cancer cells is sustained. Our results so far suggest that either macrophage mitochondrial transfer increases cancer cell proliferation, or that more proliferative cells are simply more capable of receiving macrophage mitochondria. Thus, to test between these hypotheses, we first blocked cells in the G1-phase of the cell cycle by treating co-cultures with a CDK$\frac{4}{6}$ inhibitor, Palbociclib (Figure 1—figure supplement 3d), and we observed no changes in mitochondrial transfer rates (Figure 1—figure supplement 3e). These data indicate that the enhanced proliferation observed in recipient cells is not due to proliferative cells more readily receiving transfer. We then performed experiments to rigorously test whether transferred macrophage mitochondria causes cancer cell proliferation, rather than mitochondrial receipt and proliferation being correlative events in cancer cells. We also wanted to determine whether the observed proliferative phenotype is due to macrophage mitochondria, and not other molecules that are passed along with the macrophage mitochondria. Thus, we biochemically purified mitochondria from a macrophage cell line, THP-1, and directly applied these macrophage mitochondria to cancer cells for 24 hr (Figure 1g). We then FACS-isolated cancer cell populations that contained purified macrophage mitochondria, and allowed this population to undergo additional rounds of cell division, and then reanalyzed the proliferative capacity of cancer cells that had retained the macrophage mitochondria versus cancer cells that had lost the macrophage mitochondria over this time. We first confirmed that cancer cells retained the macrophage mitochondria by imaging (Figure 1h). We also found that cancer cells that had retained the macrophage mitochondria exhibited an increased percentage of cells in the G2/M phase of the cell cycle compared to cancer cells that had lost the macrophage mitochondria (Figure 1i-j). Together with the QPI results, these results support the model that macrophage mitochondrial transfer promotes a sustained pro-growth and proliferative effect in both recipient and subsequent daughter cells. ## Transferred mitochondria are dysfunctional and accumulate ROS We next sought to understand how donated mitochondria can stimulate a proliferative response in recipient cells. We performed time-lapse confocal microscopy on co-cultures and found that in cancer cells with macrophage mitochondria, macrophage-derived mito-mEm+ mitochondria remained distinct from the recipient host mitochondrial network. Cancer cells were cocultured with macrophages for 12 hr and subjected to an additional 15 hr of timelapse microscopy, and we observed no detectable loss of the fluorescent signal at transferred mitochondria throughout the course of imaging (Figure 2a, arrowhead; Video 1). Thus, transferred macrophage mitochondria did not appear to fuse with the existing endogenous mitochondrial network in recipient cells. To probe the functional state of the donated mitochondria, we performed live imaging with MitoTracker Deep Red (MTDR), a cell-permeable dye that is actively taken up by mitochondria with a membrane potential (Poot et al., 1996). To our surprise, all of the transferred mitochondria were MTDR-negative (Figure 2b, top left). This was also confirmed using a different mitochondrial membrane potential-sensitive dye, Tetramethylrhodamine Methyl Ester (TMRM; Figure 1—figure supplement 1e). These results suggested that the transferred mitochondria lacked membrane potential. To determine whether these membrane potential-deficient transferred mitochondria were subjected to lysosomal degradation, we labeled lysosomes and acidic vesicles with a dye, LysoTracker, and found that the majority of transferred mitochondria ($57\%$) did not co-localize with the LysoTracker signal (Figure 2b, top right). The status of transferred mitochondria was unexpected because mitochondria typically maintain strong membrane potentials, and dysfunctional mitochondria that lack membrane potential are normally degraded or repaired by fusion with healthy mitochondrial networks (Phinney et al., 2015). Next, we utilized another dye which stains cellular membranes, MemBrite, and observed that $91\%$ of transferred mitochondria were not encapsulated by a membranous structure, thus also excluding sequestration as a mechanism for explaining the lack of degradation or interaction with the endogenous mitochondrial network (Figure 2—figure supplement 1a). These data, taken together with the long-lived observation of the transferred mitochondria in Figure 2a, suggest that transferred macrophage mitochondria lack membrane potential, yet remain as a distinct population in recipient cancer cells, not fusing with the endogenous host mitochondrial network nor subjected to degradation. **Figure 2.:** *Transferred macrophage mitochondria are long-lived, depolarized, and accumulate reactive oxygen species, promoting cancer cell proliferation.(a) Stills from time-lapse imaging depicting the longevity of the transferred mitochondria (green, arrowhead) within a 231 cell (magenta, cell outline in white). Time elapsed listed in left corner. (b) Confocal image of a mito-RFP+ 231 cell (magenta) containing macrophage mitochondria (green, arrowhead) stained with MTDR (yellow) and LysoTracker (teal). MTDR does not accumulate in 100% of donated mitochondria (N=25 cells, 5 donors). Majority (57%) of donated mitochondria do not colocalize with LysoTracker signal (N=24 cells, 4 donors). (c) Ratiometric quantification of mito-Grx1-roGFP2 biosensor mapped onto the recipient 231 cell with fire LUT (top panel). Confocal image of mito-Grx1-roGFP2-expressing 231 cell (bottom right, green and yellow) containing a macrophage mitochondria (bottom left, red, arrowhead). (d) Ratiometric measurements of the mito-Grx1-roGFP2 sensor per 231 cell (paired dots) at a region of interest containing the host mitochondrial network (host) or a transferred mitochondria (transfer). Cells were co-cultured for 24 hr (N=27 cells, 3 donors indicated in shades of gray). (e) Exogenous purified macrophage mitochondria (green) is void of mitochondrial membrane potential (MitoTracker Deep Red-negative, yellow, arrowhead) in cancer cells. (f) Cell cycle analysis of cancer cells with exogenous purified macrophage mitochondria versus sister cells that did not take up exogenous purified mitochondria, either treated with vehicle or 100 μM mitoTEMPO (mitochondrially-targeted superoxide scavenger. N=3 donors; statistics for G2/M only). (g) Schematic of optogenetic experiments to generate data in (h). Cells expressing mito-KillerRed are photobleached in a specific ROI containing either cytoplasm only (left) or mito-KillerRed+ mitochondria (right). Following photobleaching, cells are imaged over time to quantify the amount of cell division. (h) Quantification of cell division after photobleaching. Each data point is the average within a field of view (N=13 experiments), with control (cyto) and experimental (mito) data shown as paired dots per experiment. Scale bars are 10 µm. Wilcoxon matched-pairs signed rank test (d, h), two-way ANOVA (f), *p<0.05; ****p<0.0001.* **Video 1.:** *Macrophage mitochondria are long-lived and remain distinct in recipient cancer cells.Video depicting a recipient mito-RFP expressing 231 cell (magenta) that contains mito-mEm macrophage mitochondria (green in magenta cell, center of frame). 231 cells were co-cultured with macrophages for 7 hr prior to the start of imaging for a duration ~15 hr with a time interval of 5 min. Maximum intensity projections of images are displayed at 12 frames per second, timestamp in upper left corner in hours (h), and scale bar is 10 μm.* Given the surprising observation that transferred mitochondria lack membrane potential, we hypothesized that instead of providing a metabolic or energetic advantage, the donated mitochondria may act as a signal source to promote sustained changes in cancer cell behavior. This hypothesis could offer insight into how this rare event, in which a relatively small amount of mitochondria is transferred, could mediate sustained changes in the proliferative capacity of recipient cancer cells. One signaling molecule associated with mitochondria is reactive oxygen species (ROS), which occur normally as byproducts of mitochondrial respiration, and can be produced at high levels during organellar dysfunction (Schieber and Chandel, 2014). Using a genetically encoded biosensor, mito-Grx1-roGFP2, as a live readout of the mitochondrial glutathione redox state (Gutscher et al., 2008), we found that after 24 and 48 hr, significantly higher ratios of oxidized:reduced protein were associated with the transferred mitochondria versus the host network (Figure 2c–d; Figure 2—figure supplement 1b). These data indicate that transferred macrophage mitochondria in recipient cells are associated with higher levels of oxidized glutathione, suggesting that they are accumulating higher amounts of ROS. Consistent with these results, a second biosensor that is specific for the reactive oxygen species H2O2, mito-roGFP2-Orp1 (Gutscher et al., 2009), also reported more oxidation at the transferred mitochondria compared to the host network (Figure 2—figure supplement 1c–d) after 48 hr of co-culture. At 24 hr, we observed a similar trend, but no statistically significant difference (Figure 2—figure supplement 1d). These results indicate that ROS accumulate at the site of transferred mitochondria in recipient cancer cells. It is unclear whether the observed ROS accumulation is generated by the transferred mitochondria themselves, or generated elsewhere in the recipient cancer cell and accumulating locally at transferred mitochondria. Regardless of the source, we observed robust ROS accumulation specifically at the site of transferred mitochondria and with this unexpected finding, we next tested whether this ROS accumulation could serve as a molecular signal, regulating cell proliferation. To rigorously test the model that transferred macrophage mitochondria accumulate ROS, promoting cancer cell proliferation, we turned toward purified macrophage mitochondria approaches as in Figure 1g and sought approaches to reduce ROS levels. First, to better model the macrophage mitochondrial transfer to cancer cells that occurs in coculture conditions, we determined conditions for cancer cells to internalize exogenous macrophage mitochondria at rates similar to in vitro mitochondrial transfer conditions at 24 hr – $0.68\%$ ± $0.36\%$ internalization rate, $$n = 3$$ biological replicates (compare to Figure 1d). We next determined that purified mitochondria taken up by cancer cells remain distinct, are not encapsulated by membranes after 24 hr (Figure 2—figure supplement 1e), and do not exhibit membrane potential (Figure 2e). Similar to our previous proliferation results with purified macrophage mitochondrial uptake at longer time points (Figure 1j), we found that cancer cells with internalized purified macrophage mitochondria (which, under these conditions, comprise ~$1\%$ of the total population) exhibited a significant increase in proliferative cells in the G2/M phase of the cell cycle, compared to sister cells that did not internalize mitochondria (Figure 2f, comparing black bars in lanes 1&2), and that this increase was ameliorated when ROS is quenched with a mitochondrially localized superoxide scavenger, mitoTEMPO (Figure 2f; comparing black bars in lanes 2&4). Importantly, cancer cells that did not internalize mitochondria were not affected by ROS quenching (Figure 2f; comparing black bars in lanes 1&3). These results indicate that transferred mitochondria promote proliferation in a ROS-dependent manner. To test whether ROS accumulation can induce cancer cell proliferation directly, we stably expressed a mitochondrially localized photosensitizer, mito-KillerRed, which generates ROS when photobleached with 547 nm light (Bulina et al., 2006). As expected, photobleaching mito-KillerRed+ regions of interest induced ROS (Bass et al., 1983; Figure 2—figure supplement 2a). We then drew mito-KillerRed+ regions of interest that mimicked the size of macrophage mitochondrial transfer to induce local ROS in cancer cells, and analyzed the rate of cell division by imaging these cells over 18 hr (Figure 2g). We found that cells with induced ROS (by photobleaching mito-KillerRed+ regions) exhibited an increased percentage of dividing cells compared to negative control photobleached cells (mito vs. cyto bleach; Figure 2h; Figure 2—figure supplement 2b–c). These results indicate that induction of mitochondrially localized ROS can directly promote cancer cell proliferation. ## ROS accumulation leads to ERK-dependent proliferation We next aimed to determine how ROS induction may regulate cell proliferation. ROS is known to induce several downstream signaling pathways (Schieber and Chandel, 2014; Brillo et al., 2021), including ERK/MAPK signaling, a pathway known to regulate proliferation and tumorigenesis (Dhillon et al., 2007). Thus, we sought to determine if cancer cells that had received macrophage mitochondria exhibited increased ERK signaling. We stably expressed the ERK-Kinase Translocation Reporter (ERK-KTR) (Regot et al., 2014), which translocates from the nucleus to the cytoplasm when ERK is activated, in 231 cells (231-ERK-KTR). After co-culturing 231-ERK-KTR cells with macrophages, we used the imaging flow cytometer, Amnis ImageStream, to compare relative ERK-KTR translocation values in hundreds of cells that had or had not received macrophage mitochondria (ERK-KTR quantification and ERK signaling validation described in Figure 3—figure supplements 1–2). These data show that cancer cells with macrophage mitochondria have significantly higher cytoplasmic to nuclear (C/N) ERK-KTR ratios compared to cells that did not receive mitochondrial transfer, indicating increased ERK activity (Figure 3a–b; Figure 3—figure supplement 2a–b). **Figure 3.:** *Recipient cancer cells exhibit ERK-dependent proliferation.(a) ImageStream was used to measure the MFI of an ERK-Kinase Translocation Reporter (ERK-KTR, orange) in the nucleus (DAPI, blue) or cytoplasm of co-cultured 231 cells that did (right) or did not (left) receive mitochondria (green, arrowhead). Below: representative line scans (white dotted lines) of ERK-KTR (orange) and DAPI (blue). (b) Average ERK activity from data displayed in (d) (cytoplasm/nucleus (C/N) mean fluorescence intensity (MFI); N=3 donors indicated as shades of gray). (c) Confocal images of 231 cells expressing ERK-KTR (green) and Mito-KillerRed (magenta) with Hoechst 33342 (blue), after control cytoplasmic bleach (cyto, left) or mito-KillerRed+ bleach (mito, right). Below: representative line scans (white dotted lines) of ERK-KTR (green) and Hoechst (blue). (d) Quantification of ERK-KTR translocation 40 min post-bleach (cyto vs. mito), normalized to time 0. Each dot represents a measurement from a single cell. (e) Analysis of proliferative capacity by quantifying Ki-67 and DNA levels of co-cultured 231 cells treated with vehicle or ERK inhibitor (ERKi) with or without transfer or (f), mitochondrial internalization after mitochondrial bath application (N=3 donors; statistics for G2/M only). Error bars represent SEM and scale bars are 10 µm., Welch’s t-test (b), Mann-Whitney (d), two-way ANOVA (e–f), *p<0.05; **p<0.01; ****p<0.0001.* Due to our observations that cells that receive macrophage mitochondria exhibit increased ERK activation and that local ROS induction is sufficient to induce cell proliferation, we then asked whether cancer cell mitochondrial ROS would directly enhance ERK activation. By expressing both mito-KillerRed and ERK-KTR in 231 cells, we induced ROS by photobleaching mito-KillerRed+ regions and found that ROS induction increased ERK-KTR translocation, indicating that ROS induction is sufficient to increase ERK activity in cancer cells (Figure 3c–d; Figure 3—figure supplement 2c). We next tested whether ERK signaling is required for the mitochondrial transfer-induced cancer cell proliferation. We first determined an effective concentration of SCH772984, an ERK inhibitor (ERKi), that still inhibits ERK activity, but does not dramatically affect 231 proliferation, as we sought to determine whether inhibiting ERK affects mitochondrial transfer-induced proliferation, not proliferation more generally. We first confirmed that treatment with this effective concentration of ERKi led to decreased ERK activity, as determined by the ERK-KTR translocation reporter (Figure 3—figure supplement 4a–b). We then found that treatment with ERKi significantly decreased proliferation of recipient 231 cells when compared to vehicle control-treated recipient cells (Figure 3e; Figure 3—figure supplement 4c). We further noted that the decrease in proliferation with this concentration of ERK inhibitor was observed only in cancer cells that received macrophage mitochondria, and not in cancer cells that did not receive macrophage mitochondria (bars 2&4 in Figure 3e, compared to bars 1&3), suggesting that the ERK-dependent cell proliferation specifically occurs in cancer cells that received mitochondrial transfer. As a control, we also confirmed that ERKi treatment did not alter mitochondrial transfer efficiencies, showing that ERK signaling does not influence mitochondrial transfer (Figure 3—figure supplement 4d). Finally, similarly to Figure 2f, we bath applied purified macrophage mitochondria to cancer cells in the presence of vehicle or ERKi and compared the proliferative capacity of cells that had internalized macrophage mitochondria versus cells that did not (Figure 3f). We found that, as before, uptake of purified macrophage mitochondria increased the percentage of cancer cells in the G2/M phase of the cell cycles (Figure 3f, bars 1&2), but that this process is ameliorated by the inhibition of ERK signaling (Figure 3f, bars 2&4). We also found that ERK inhibition did not affect the cell cycle state of cancer cells that had not taken up purified macrophage mitochondria (Figure 3f, bars 1&3). Thus, these results indicate that mitochondrial transfer promotes cancer cells proliferation through a ROS/ERK-dependent mechanism. ## M2-like macrophages exhibits enhanced mitochondrial transfer rates In many solid tumors, it has long been appreciated that macrophage density is associated with disease progression and poor patient prognosis (Pollard, 2004). Macrophages are highly plastic, altering their phenotypes, expression profiles and function, depending on environmental stimuli and conditional requirements (Pan et al., 2020). Accordingly, macrophages exist in a spectrum of activation states but are canonically simplified by the two ends of spectrum: pro-inflammatory and anti-tumorigenic M1-like macrophages; or anti-inflammatory and pro-tumorigenic M2-like macrophages (Huang et al., 2018). Since the ways in which M2-like macrophages promote tumor progression continue to be elucidated and given that there remains a dearth of understanding of how donor cell biology affects mitochondrial transfer, we aimed to determine how macrophage activation status affects intracellular mitochondrial dynamics and transfer efficiencies to cancer cells. Activated macrophages were co-cultured with 231 cells, and we first quantified mitochondrial networks using Mitochondrial Network Analyses (MiNA) (Valente et al., 2017). We found that M2-like macrophages contain significantly more fragmented mitochondria when compared to M1-like or M0 (non-activated) macrophages (Figure 4a–b; Figure 4—figure supplement 1a–d). We then co-cultured 231 cells with either M0, M1-like, or M2-like macrophages and using flow cytometry, we found that mitochondrial transfer efficiencies were significantly increased from M2-like macrophages when compared to M1-like or non-activated M0 macrophages (Figure 4c). Given that M2-like macrophages exhibited fragmented mitochondrial networks and enhanced mitochondrial transfer rates, we hypothesized that smaller mitochondrial fragments might be transferred more readily than larger networks. To test this hypothesis, we directly manipulated mitochondrial morphology by modulating a key regulator of mitochondrial fission, DRP1 (Fonseca et al., 2019). Macrophages transduced with DRP1-shRNA containing lentivirus exhibited hyper-fused mitochondrial networks (Figure 4d and e) and exhibited decreased mitochondrial transfer (Figure 4f). Together these findings reveal that macrophage activation alters mitochondrial dynamics, and that altering mitochondrial dynamics directly affects mitochondrial transfer rates. Finally, to determine whether the functionality of transferred mitochondria differ between macrophage subtypes, we evaluated the membrane potential of transferred mitochondria, and found that transferred mitochondria from M1-like and M2-like macrophages were similarly depolarized (Figure 4—figure supplement 1e), as to what we observed with M0 macrophages (Figure 2b). Taken together, these results suggest that pro-tumorigenic M2-like macrophages exhibit increased mitochondrial fragmentation, promoting mitochondrial transfer to cancer cells. **Figure 4.:** *M2-like macrophages exhibit increased mitochondrial fragmentation and increased mitochondrial transfer to cancer cells.(a) Representative images of mito-mEm+ macrophages that were non-stimulated (M0, left) or activated to become M1-like (middle) or M2-like (right). (b) Mitochondrial network analyses (MiNA) were used to determine number of mitochondrial fragments per cell (N=2 donors). (c) Macrophages were co-cultured with mito-RFP 231 cells for 24 hr and mitochondrial transfer was quantified with flow cytometry (N=4 donors). (d) Representative images of mito-mEm (green) macrophages in macrophages with control nt-shRNA KD and DRP1 KD. (e) q-RT-PCR of DRP1 knockdown (DRP1-KD) macrophages (N=3 donors). (f) Rates of mitochondrial transfer with control and DRP1-knockdown macrophages (N=3 donors). For all panels, individual donors are indicated as shades of gray with each cell as a data point, error bars represent SEM and scale bars are 10 µm. Two-way ANOVA (b, c), unpaired t-test (e, f), ***p<0.001; ****p<0.0001.* To assess whether mitochondrial transfer also occurs in a clinically relevant cancer model, we used three-dimensional stable organoid cultures generated from patient-derived xenografts (PDxOs) (Guillen et al., 2021). We examined organoids from a recurrent primary breast tumor (HCI-037) and a bone metastasis (HCI-038) derived from the same breast cancer patient. PDxOs grown in 3D (Figure 4—figure supplement 2, top) were dissociated, combined with mito-mEm+ macrophages (Figure 4—figure supplement 2a, bottom), and then embedded in Matrigel (experimental scheme in Figure 4—figure supplement 2b). After 72 hr, mitochondrial transfer was assayed by live imaging (Figure 4—figure supplement 2c) and quantified with flow cytometry (Figure 4—figure supplement 2d, e,). Mitochondrial transfer was observed from macrophages to both HCI-037 and HCI-038 PDxO cells (Figure 4—figure supplement 2e), although intriguingly, M2-like macrophages preferentially transferred mitochondria to the bone metastasis PDxO cells (HCI-038), whereas M0 and M2-like macrophages transferred mitochondria to primary breast tumor PDxO cells (HCI-037) at the same rate. In all cases, transferred macrophage mitochondria lacked membrane potential (Figure 4—figure supplement 2c), consistent with our results in 231 recipient cells. ## Cancer cells with macrophage mitochondria exhibit increased proliferation in vivo Next, to better model a tumor environment, we examined macrophage mitochondrial transfer to cancer cells in two separate in vivo models of metastatic breast cancer. We first injected E0771 murine adenocarcinoma cells expressing mito-mEm into wildtype C57BL/6J mice that had received lethal irradiation with subsequent bone marrow reconstitution from mito::mKate2 mice (mito:mKate2→WT), restricting mKate2 expression to immune cells (experimental schematic in Figure 5—figure supplement 1a). We found that in vivo mitochondrial transfer occurred at a rate of $4.8\%$, compared to control transplantation studies at $0.46\%$ (Figure 5a). We also performed experiments in mice that restrict GFP-labeled macrophage mitochondria to the myeloid lineage by using the LysM-Cre transgenic mouse crossed to the lox-stop-lox-MitoTag mouse (experimental schematic Figure 5b, see methods for more details). We injected E0771 cells expressing mito-RFP into these mice with GFP-labeled macrophage mitochondria and observed E0771 cells containing macrophage mitochondria using immunohistochemistry approaches of tumor sections (Figure 5c; Figure 5—figure supplement 1b). Using similar cell proliferation analyses as previously described (Figure 1e), we also observed that recipient tumor cells exhibited enhanced proliferative capacity compared to the tumor cells that did not receive transfer (Figure 5d; Figure 5—figure supplement 1c). These results show that mammary adenocarcinoma cells with macrophage mitochondria exhibit increased proliferation in vivo. **Figure 5.:** *Macrophage mitochondrial transfer promotes tumor cell proliferation in vivo.(a) Quantification of E0771 mammary adenocarcinoma cells from in vivo tumors with mKate2+ mitochondria in bone marrow reconstitution experiments versus control mice. N=10 mice per condition. (b) Schematic representation of a second mouse model to quantify proliferation in cancer cells with macrophage mitochondria in vivo. Myeloid lineages were specifically labeled with mito-GFP by crossing a Loxp-Stop-Loxp-MitoTag-GFP mouse to a LysM-Cre mouse. E0771 cells expressing mito-RFP were injected into the mammary fat pad of mice with MitoTag-GFP expression in myeloid cells, and tumors were isolated and analyzed for direct observation of transfer through fluorescent microscopy (c) and Ki67/DNA to quantify proliferative index (d). (c) Representative immunofluorescence image of E0771 tumor cell expressing mito-RFP (magenta) containing GFP+ macrophage mitochondria (arrowheads) from mice in which GFP+ mitochondria are restricted to the myeloid lineage (‘LysM-Cre’). (d) Cell cycle analysis of E0771 in vivo tumor cells with and without GFP+ macrophage mitochondria in ‘LysM-Cre’ model in which GFP+ mitochondria are restricted to the myeloid lineage. N=3 mice. (e) Working model for macrophage mitochondrial transfer to breast cancer cells. For all panels, individual donors are indicated as shades of gray with each cell as a data point, error bars represent SEM and scale bars are 10 µm. Welch’s t- test (a), two-way ANOVA (d), **p<0.01; ****p<0.0001.* Finally, to determine the impact of macrophage mitochondrial transfer on population growth over time, we derived a relationship between overall growth of the cell population and the fraction of cells with macrophage mitochondria that experience an increase in growth rate (Figure 5—figure supplement 2). We used this analysis to predict the increase in population size due to transferred mitochondria as a function of the number of population doublings. The fraction of the population receiving mitochondria was assumed to be $5\%$ based on our in vivo studies (Figure 5a), with the tumor cell population exhibiting a $15\%$ increase in baseline growth rate due to transferred mitochondria based on QPI growth rate measurements (Figure 1f). We also assumed that half of the population loses transferred mitochondria, and the associated growth increase, with every division given that our QPI measurements indicated that typically only one of the daughter cells inherit the parent’s exogenous macrophage mitochondria (Figure 1—figure supplement 3a–c). By 20 divisions, with a $15\%$ increase in growth rate (brown line), even with only $5\%$ of the cancer cell population with macrophage mitochondria at any given time, the model already predicts a $15\%$ increase in population size compared to baseline population rates (comparing the brown line to the blue dotted line). These results highlight the significance of macrophage mitochondrial transfer on the growth of a cell population over time. Taken together, our work supports a model (Figure 5f) whereby M2-like macrophages exhibit fragmented mitochondria leading to increased mitochondrial transfer. In the recipient cancer cell, transferred mitochondria are long-lived, depolarized, and accumulate ROS, leading to increased ERK activity and subsequent cancer cell proliferation. ## Discussion Lateral mitochondrial transfer is a relatively young and rapidly evolving field. Previously literature had shown that healthy mitochondria are transferred, enhancing recipient cell viability by increasing ATP production and stimulating metabolic processes. Our observations, however, suggest that transferred mitochondria promote tumor cell proliferation as a byproduct of their potential dysfunctionality. This model raises several fascinating questions, including when and how transferred mitochondria become depolarized and accumulate ROS, where the ROS is generated in the recipient cell, and why depolarized mitochondria are not repaired or degraded in the recipient cell, given that 231 cells are capable of performing mitophagy (Biel and Rao, 2018). Impaired mitophagy and enhanced mitochondrial dysfunction are hallmarks of age (Chen et al., 2020), yet little is known about how age-related mitochondrial dysfunction influences mitochondrial transfer. Interestingly, instead of degrading dysfunctional mitochondria through mitophagy, neurons in an Alzheimer’s disease mouse model have been shown to transfer dysfunctional mitochondria to neighboring astrocytes (Lampinen et al., 2022), which contributes to neuronal mitochondrial homeostasis. Given that age is the greatest known risk factor of Alzheimer’s disease, and most cancers are also age-related, these data collectively warrant broader investigations into how age-associated mitochondrial dysfunction contributes to mitochondrial transfer and how this form of communication may have specific influences on distinct diseased states. Cellular stress occurs throughout biological systems, and cells have evolved a myriad of mechanisms to cope with disadvantageous cellular conditions, including mitochondrial stress (Ma et al., 2020). Our results suggest that transferred mitochondria are a source for downstream signal activation through a ROS-ERK-mediated mechanism. The origin of ROS generation in recipient cells is still unclear, and how ROS locally accumulate at macrophage mitochondria is an open question. It is possible that ROS are generated by the transferred mitochondria themselves, as previous reports have shown that mitochondria with reduced membrane potential can generate ROS (Feng et al., 2022; Franco-Iborra et al., 2018; Nakai et al., 2003). However, there are multiple mechanisms for ROS production (Zhao et al., 2019), and it is possible that the ROS are generated elsewhere in the cell and accumulating at transferred mitochondria. A previous report showed that the endoplasmic reticulum can produce ROS in the presence of dysfunctional mitochondria (Leadsham et al., 2013), suggesting another possible explanation. The questions of how ROS is generated, and how ROS can be spatially restricted to a specific subcompartment of the cell are exciting avenues of investigation and much further study. Although high levels of ROS are cytotoxic to cells (Stadtman and Levine, 2000; Fruhwirth and Hermetter, 2008; Auten et al., 2002), physiological levels of ROS are known second messenger molecules stimulating various pro-survival signaling cascades (Schieber and Chandel, 2014; Brillo et al., 2021). Additionally, a modest increase of mitochondrial-derived ROS has been shown to exhibit protective mechanisms through mitohormesis (Crewe et al., 2021; Ristow and Schmeisser, 2014), a process in which cellular defense mechanisms are stimulated by sub-lethal stress levels, protecting cells to withstand a secondary exposure. Mitochondrial transfer has been shown to promote resistance to subsequent chemotherapeutic treatments in healthy neurons (English et al., 2020) and tumor cells (Wang et al., 2018; Boukelmoune et al., 2018), however the mechanism of this resistance is unclear. It is possible that mitochondrial transfer mediates this protective response through mitohormesis, promoting longevity and proliferation of the recipient cancer cells. More studies are required to connect mitochondrial transfer, sub-lethal cellular stress, and resistance to chemotherapeutic treatments in disease progression and tissue homeostasis. The role of mitochondrial transfer has been largely studied in recipient cells. There remains a dearth of information describing how donor cells regulate mitochondrial transfer. Although intercellular mitochondrial transport has been implicated in the process of mitochondrial transfer (Boukelmoune et al., 2018; Ahmad et al., 2014), we show that macrophage differentiation directly affects mitochondrial transfer through changes in their mitochondrial morphology. The relationship between macrophage differentiation and metabolism has been partially defined, with anti-tumor-like (M1-like) macrophages exhibiting more glycolytic metabolism, and pro-tumor-like (M2-like) macrophages upregulating oxidative phosphorylation (Van den Bossche et al., 2017; Mortezaee and Majidpoor, 2022). But how mitochondrial morphology regulates metabolism is less understood. Studies have indicated that mitochondrial fusion supports increases oxidative phosphorylation in fibroblasts (Yao et al., 2019), however other studies have shown that increasing mitochondrial fission upregulates oxidative phosphorylation in hepatocytes (Zhou et al., 2022). Thus, the correlation between mitochondrial morphology and cellular metabolic status is unclear, and these differences are likely due to different cell types and environmental conditions. While how the metabolic status of donor cells influences mitochondria is still unknown, our results support the hypothesis that pro-tumorigenic M2-like macrophage activation promotes mitochondrial fragmentation, and that mitochondrial fragmentation directly promotes mitochondrial transfer. Our findings are consistent with several studies describing a metastatic advantage in cancer cells that receive exogenous mitochondria (Zampieri et al., 2021; van der Merwe et al., 2021; Dong et al., 2017; Tan et al., 2015). However, the mechanism underlying this behavior is unexpected. Studies examining mitochondrial transfer have typically used recipient cells with damaged or non-functional mitochondria, and the fate and function of donated mitochondria are rarely followed in recipient cells. Furthermore, it was largely unclear how transferred mitochondria can affect the behavior of recipient cells with functioning endogenous mitochondrial networks, particularly if the donated mitochondria only account for a small fraction of the total mitochondrial network in the recipient cell. Our work detailing how transferred mitochondria can activate downstream signaling pathways in response to ROS provides an explanation for how a relatively small amount of transferred mitochondria can generate a sustained behavioral response in recipient cells. ## Cell culture of cell lines and peripheral blood mononuclear cells (PBMCs) Human cell lines MDA-MB-231 (HTB-26), MDA-MB-468 (HTB-132), A375 (CRL-1619), THP-1 (TIB-202), MCF10A (CRL-10317), and the murine cell lines E0771 (CRL-3461) were directly purchased from American Type Culture Collection and cultured according to their recommendations. Cell lines are authenticated through STR profiling, and all cultured cell lines are subjected to mycoplasma testing every 6 months using the Universal Mycoplasma Detection Kit (30–1012 K, ATCC). Base medias used were DMEM, high glucose (11965118, ThermoFisher), RPMI (11875119, ThermoFisher) and $10\%$ heat-inactivated fetal bovine serum (FBS; F4135, ThermoFisher). All cell lines were kept in culture for no more than 25 passages total. ## Genetic modification of PBMCs and differentiation into macrophages PBMCs were isolated from leukocyte filters obtained from de-identified human blood donors (ARUP Blood Services). CD14 +monocytes were isolated from buffy coats and genetically modified with lentiviral vectors in the presence of virus-like particles packaging Vpx (to overcome restriction in myeloid cells) as previously described (Johnson et al., 2020; Greiner et al., 2022). Briefly, freshly harvested CD14 +monocytes were plated at a density of 4–5 M cells per 10 cm plate in ‘macrophage culture media’ containing: RPMI (11875119, ThermoFisher), $10\%$ FBS (26140079, Thermo Fisher), $0.5\%$ penicillin/streptomycin (P/S; P4333, Thermo Fisher), 10 mM HEPES (15630080, ThermoFisher), $0.1\%$ 2-Mercaptoethanol (21985023, Thermo Fisher), recombinant human GM-CSF at 20 ng/ml (300–03, Peprotech) with the addition of polybrene (1 μg/ml), and supernatant containing Vpx particles (0.5 mL per 4 M cells) to facilitate viral transduction. Thirty min after plating, 100–200 µL of concentrated lentiviral stock was added to the plated monocytes. $50\%$ of the media was replaced on day 2 and a full media replacement occurred on day 4. Macrophages were used for experiments starting on day 6 or 7 after harvest of PBMCs unless otherwise noted. ## Distinction of biological and technical replicates Each human blood donor (referred to as ‘donors’ or ‘experiments’) is a biological replicate. Multiple samples from each donors run in parallel are defined as technical replicates (typically in triplicate for each biological replicate). ## Generation of mito-FP and FP-TOMM20 stable cell lines *We* generated a modified pLKO.1 plasmid backbone with an accessible multiple cloning site (pLKO.1_MCS) for generation of fluorescent reporters. For mito-FP expression, we cloned the cytochrome oxidase subunit VIII mitochondrial targeting sequence and tagged it to mEmerald (referred to as mito-mEm) or tagRFPt (referred to as mito-RFP) and introduced these into the pLKO.1_MCS backbone in order to generate lentiviruses. pLKO.1 mito-mEmerald and pLKO.1 mito-TagRFP-T are available on Addgene (#174542 and 174543, respectively). For FP-TOMM20 expression, inserts containing the sequence of either mEmerald (mEmerald-TOMM20) or mcherry (mCherry-TOMM20) fused to TOMM20 and cloned into the pLKO.1 backbone and used to generate lentiviruses. Stable lines were generated through lentiviral transduction. For transduction, approximately 50,000 cells were plated into one well of a 6-well plate directly into the appropriate lentivirus supernatant diluted 1:5 in DMEM complete media with a final concentration of 10 µg/mL polybrene (TR-1003-G, Sigma). After 48–72 hr, cells were expanded, and multiclonal populations were flow sorted for appropriate levels of fluorescent expression. All other transgenic cell lines were generated as outlined in subsequent sections. mEmerald-TOMM20-N-10 (Addgene plasmid # 54282) and mCherry-TOMM20-N-10 (Addgene plasmid # 55146) were a gift from Michael Davidson. ## Lentivirus production pLKO.1_MCS plasmids containing the appropriate transgene were used to generate lentivirus as outlined in Johnson et al., 2020. Briefly, 293 FT cells in 15 cm plates were transfected with PEI-max (24765, Polysciences) and plasmids for pCMV-VSV-G, psPax2, and transgene cassettes. The following day, cells were washed and cells were grown for an additional 36 hr in fresh media. Supernatants were harvested, passed through 0.45 μm filters, and either used fresh or concentrated by ultracentrifugation as previously described (Johnson et al., 2020). Lentiviral supernatants were used to transduce cell lines as outlined in ‘generation of mito-FP’ section unless otherwise noted. ## Flow cytometry The following flow cytometry machines were used: a BD FACS Aria (equipped with 4 Lasers: 405, 488, 561, 640) referred to as Aria, or a BD LSR Fortessa (5 Lasers: UV, 405, 488, 561, 640) referred to as the Fortessa. Technical details per experiment type are listed below. ## Stable line generation Cells were enzymatically dissociated using trypsin-EDTA (25200056, ThermoFisher) and resuspended in buffer consisting of $0.5\%$ Bovine Serum Albumin (BSA; Sigma, A9418) in DPBS (14190250, ThermoFisher). Cells were sorted according to fluorescent intensity on the Aria and collected in the appropriate media containing $0.5\%$ P/S. ## Mitochondrial transfer quantification For MDA-MB-231 and MCF10A cell lines: cells were enzymatically dissociated using trypsin-EDTA and stained as follows: cells were resuspended in ‘staining buffer’ (DPBS + $2\%$ FBS) containing a human antibody against CD11b conjugated to the fluorophore Brilliant Violet 711 (BV711-CD11b; macrophage marker; Biolegend, 301344) at a 1:20–40 dilution. After a 30 min incubation on ice, cells were washed and resuspended in cold DPBS for analysis on the Fortessa. The background level of mEmerald fluorescence was set at $0.2\%$ based on a fully stained co-culture control where macrophages were not transduced with mito-mEmerald. This gate was defined by FACS-isolating co-cultures of mito-RFP MDA-MB-231/mito-mEm macrophages and determining a gate that accurately isolated MDA-MB-231 cells containing macrophage mitochondria. We found that the cancer cell population with the highest mEm signal were cancer/macrophage fusions, and we therefore removed this population from downstream analysis. Setting the gate to $0.2\%$ predominantly led to isolation of cancer cells with fragments of macrophage mitochondria, as visualized by microscopy. ## Mitochondrial transfer quantification of PDxO containing co-cultures Hanging drop co-cultures suspended in Growth Factor Reduced Matrigel (354230, Corning) were pooled and dissociated using a solution of Dispase II (50 U/mL; 17105041, Fisher Scientific) followed by TrypLE Express (12605010, Thermo Fisher). Cells were then incubated in TrueStain FcX (422301, ThermoFisher) at 1:33 dilution with staining buffer for 10 min at room temperature. Primary human antibodies against CD326 conjugated to PE (PE-EpCam; PDxO marker; 369806, Biolegend) and BV711-CD11b were added at 1:20 and 1:40, respectively. After 30 min on ice, cells were washed and resuspended in cold DPBS for analysis on the Fortessa. The background level of mEmerald fluorescence in the ‘transfer gate’ was set at $0.2\%$ based on a fully stained monoculture control. ## Quantification of Ki67 and DNA content Co-cultures were enzymatically dissociated with trypsin-EDTA and incubated in staining buffer containing anti-human BV711-CD11b at 1:40 for 30 min on ice. Cells were then fixed and stained using the eBioscience Foxp3/Transcription Factor Staining Buffer Set (00-5523-00, ThermoFisher) according to manufacturer’s instructions. Cell were stained with an APC conjugated Ki67 antibody (APC-Ki67; 17-5699-42, ThermoFisher) at 1:20 for 30 min followed by a 3 µM DAPI (D9542, Sigma) solution for 10 min. Cells were resuspended in cold DPBS for analysis on the Fortessa. The background level of mEmerald fluorescence in the ‘transfer gate’ was set at 0–$0.2\%$ based on a fully stained co-culture control where macrophages were not transduced with mito-mEmerald. ## Single-cell RNA-seq Mito-mEm macrophages were cocultured with mito-RFP 231 cells for 24 hr. Two populations were FACS-isolated: [1] Mito-RFP 231 cells containing mito-mEm macrophage mitochondria; [2] mito-RFP 231 cells not containing mito-mEm macrophage mitochondria. From FACS-isolated populations, a cDNA library was generated using the 10 X genomics Single Cell 3’ Gene Expression Library V3 and amplified according to the manufacturer’s protocol. The resulting libraries were sequenced on a NovaSeq 6000 resulting in approximately 100 K mean reads per cell. The raw sequencing data were processed using CellRanger 3.02 (https://support.10xgenomics.com/) to generate FASTQ files, aligned to GRCh38 (Ensemble 93), and a gene expression matrix for individual cells based on the unique molecular indices was generated. The resultant filtered gene-cell barcode matrix was imported into SEURAT version 4 (Hao et al., 2020) with R studio version 1.3.1093 and R version 4.03. We first performed quality control by determining the mean and standard deviation of genes per cell and filtered out all cells that were more than 1.5 standard deviations away from the mean. The reads were then scaled and normalized using SEURAT ‘sctransform’ function (Hafemeister and Satija, 2019). Using the normalized data, we determined differential gene expression in the MDA-MB-231 population that received macrophage mitochondria compared to those that did not, using a non-parametric Wilcoxon rank sum test with the SEURAT ‘FindMarkers’ function. Lastly, the differential expression data were exported from R and pathway enrichment analysis was performed using Qiagen’s Ingenuity Pathway Analysis software (Krämer et al., 2014). Single-cell RNA-sequencing data are available with GEO accession number GSE181410. The analysis code for single-cell RNA-sequencing analysis is available on GitHub (https://github.com/rohjohnson-lab/kidwell_casalini_2021; RRID:SCR_002630 (version number 1)). ## Trans-well experiments Approximately 40,000 mito-RFP MDA-MB-231 and 80,000 mito-mEm macrophages were plated in trans-wells (3401, Corning) under the conditions listed in Figure 1—figure supplement 1e–h. Cells were analyzed after 24 hr with flow cytometry as indicated in ‘mitochondrial transfer quantification’ section. ## Live cell imaging of co-cultures with cell-permeable dyes Imaging was performed using either a Zeiss LSM 880 with AiryScan technology (Carl Zeiss, Germany) and a 63 x/1.4 NA oil objective or a Leica Yokogawa CSU-W1 spinning disc confocal microscope with a Leica Plan-Apochromat 63 x/1.4 NA oil objective and iXon Life 888 EMCCD camera. Images taken on the LSM 880 were acquired using the AiryScan Fast mode. For all live imaging, cells were maintained at 37 °C, $5\%$ CO2 with an on-stage incubator. MDA-MB-231 cells and primary macrophages stably expressing the appropriate transgenes were mixed in a 1:2 ratio and plated at an approximate density of 300,000 cells directly onto 35 mm glass bottom dishes (FD35-100, World Precision Instruments) for all live imaging experiments unless otherwise noted. Duration of co-culture is indicated in main text or figure legend. For detection of nuclear and mitochondrial DNA, Hoechst 33342 (B2261, Sigma) was diluted into the culture media to a final concentration of 5 μg/mL. After 10 min at 37 °C, cells were washed, and complete media was replaced before imaging. For detection of mitochondrial membrane potential with MitoTracker Deep Red (MTDR; M22426, ThermoFisher), MTDR was diluted into serum-free DMEM media (11965118, ThermoFisher) at a final concentration of 25 nM and incubated at 37 °C for 30 min. Following incubation, cells were washed with warm PBS, and warmed complete media was replaced before imaging. For detection of mitochondrial membrane potential with Tetramethylrhodamine, Methyl Ester, Perchlorate (TMRM; T668, ThermoFisher), TMRM was diluted into serum-free DMEM media at a final concentration of 100 nM and incubated at 37 °C for 30 min. Following incubation, cells were washed with warm PBS, and warmed complete media was replaced before imaging. For detection of lysosomes and acidic vesicles, LysoTracker Blue (L7525, ThermoFisher) was diluted to a final concentration of 75 nM in serum-free DMEM media and incubated at 37 °C for 30 min. Following incubation, cells were washed with warm PBS and warmed complete media was replaced before imaging. For detection of plasma/vesicular membranes, MemBrite $\frac{640}{660}$ (Biotium, 30097) was used at a final concentration of 1:1000 and stained according to manufacturer’s instructions. To preferentially label intracellular membrane compartments, cells were allowed to rest for 45 min after Membrite $\frac{640}{660}$ staining before imaging, as indicated in the manufacturer’s instructions. For detection of ROS, Carboxy-H2DCFDA (C400, ThermoFisher) was diluted to 5 μM into warmed HBSS (14025092, ThermoFisher) and incubated at 37 °C for 15–30 min. After incubation, cells were washed with HBSS and warmed complete media was replaced before imaging. ## Live imaging of sorted recipient cells MDA-MB-231 cells were harvested and stained as indicated in ‘mitochondrial transfer’ section of flow cytometry methods. Cells were sorted on the Aria directly into media containing $0.5\%$ P/S. Sorted cells were plated directly onto imaging dishes coated with CellTak (354240, Corning) and allowed to attach at 37 °C for up to 4 hr before staining and live imaging. ## Quantitative phase imaging (QPI) Mito-RFP MDA-MB-231 cells and mito-mEm macrophages were seeded in a 1:2 ratio at a density between 90,000 and 120,000 cells directly onto imaging dishes 24 hr prior to the start of imaging. QPI images were acquired on Olympus IX83 inverted microscope (Olympus Corporation, Japan) with Phasics SID4 camera (Phasics, France) and Thorlabs 623 nm wavelength DC2200 LED (Thorlabs, USA). The microscope was operated in brightfield with Olympus UPLFLN 40 X objective and a 1.2 X magnifier in front of camera, giving ×48 magnification. Fluorescence images were acquired using X-Cite 120LED illumination (Excelitas technologies, USA) and an R1 *Retiga camera* (Cairn research Ltd, UK) with GFP (Olympus Corporation U-FBNA) and RFP (IDEX health & science, USA mCherry-B-000) filter cubes. Cells were maintained at 37 °C temperature, $5\%$ CO2 and $90\%$ humidity with an Okolab (Okolab, Italy) on-stage incubator on a Prior III Proscan microscope stage (Prior Scientific Instruments Ltd., UK). Automation was performed with MicroManager open-source microscopy software via MATLAB 2012b. QPI images of 40 positions per imaging set, four replicate (biological replicate) imaging sets total, were acquired every 15 min with fluorescence images acquired in an alternate subset of locations every 15 min for 48 hr to reduce phototoxicity. ## QPI data analysis QPI and fluorescent images were analyzed with MATLAB 2019a. Cell phase shift images were background corrected using sixth order polynomial surface fitting, and converted to dry mass (m) map, using, m=∫1αϕλdA, where λ, is the wavelength of source light = 0.623 µm, α, specific refractive increment = 0.185 µm3/pg, A, image pixel area = 0.36 µm2/pixel, and ϕ is the phase shift in fraction of a wavelength at each pixel. Cell dry mass maps were then segmented using a Sobel filter for edge detection and tracked over time (Crocker and Grier, 1996). Specific growth rate of each tracked cell was computed as the slope of a linear, least-squares best fit line to mass over time data normalized by cell average mass. Fluorescent mitochondria images were resized to match QPI images and overlaid with corresponding QPI image segmentation mask to measure the integrated fluorescence intensity of every cell, normalized by cell area. Macrophage mitochondria high frequency punctae signal in MDA-MB-2321 cells were separated from the high intensity, low spatial frequency of the macrophage mitochondria network fluorescence signal using the rolling ball filter in MATLAB. The size of the rolling ball was 4.8–9 μm, chosen to be just above the average size of mitochondrial punctae based on the quantity of mitochondria transferred and retained in the MDA-MB-231 cells. Cells with RFP signal 1.5 times more than the background were identified as MDA-MB-231 cells, and with mEmerald fluorescent signal double that of background as macrophages. MDA-MB-231 cell tracks were then binned based on the presence or absence of mEmerald +mitochondrial punctae, indicating transfer from macrophages. The specific growth rate of each cell was calculated as the slope of a least-squares linear fit to QPI mass vs time data divided by the average mass of the cell. The code for automated tracking of cell mass from QPI and fluorescence data and computing growth rates of the different groups of cells is available on GitHub (https://github.com/Zangle-Lab/Macrophage_tumor_mito_transfer, copy archived at ZangleLab, 2023). ## Cytokinesis analysis Cytokinesis rate was calculated by tracking cells manually to confirm division of cells in less than the maximum doubling time expected (40 hours). Cells leaving the imaging frame in less than 30 hr were omitted from the cytokinesis calculation. ## Lineage analysis The average specific growth rate of MDA-MB-231 parent and daughter cell was calculated by manually annotating mass versus time tracks from mass tracking based on the presence of mEmerald +punctae. The difference in growth of daughter cells that did or did not inherit mitochondria from mitochondria containing parents was observed by normalizing the mass of each daughter by its initial mass at birth. ## ROS biosensor line generation, imaging, and quantification MDA-MB-231 cells were transfected with the following plasmids: pLPCX mito-Grx1-roGFP2 (Gutscher et al., 2008) and pLPCX mito-roGFP2-Orp1 (Gutscher et al., 2009) (Addgene, plasmid #64977 and #64992, respectively) using the Polyplus-transfection jetPRIME DNA/siRNA transfection kit (55–131, Genesee Scientific) according to the manufacturer’s instructions. Cells were allowed to recover for 3–7 days and then sorted for expression. Cells were passaged every 3–5 days and sorted as needed to maintain a high percentage of expressing cells. Biosensor-expressing MDA-MB-231 lines were co-cultured with mito-RFP expressing macrophages for 24 or 48 hr and imaged on the Zeiss LSM 880. Cells were sequentially imaged (per z-plane) for the presence of transferred macrophage mitochondria (Ex. 561 nm, Em. BP 570–620nm +LP 645 nm) and the biosensor in its a reduced (Ex. 488 nm, Em. BP 420–480nm +BP 495–550 nm) and oxidized (Ex. 405 nm, Em. BP 420–480nm +BP 495–550 nm) form. Images were initially processed using Zen software (see image analysis section) and further analysis was performed using FIJI as indicated in Morgan et al., 2011. pLPCX mito Grx1-roGFP2 (Addgene plasmid # 64977) and pLPCX mito roGFP2-Orp1 (Addgene plasmid #64992) were a gift from Tobias Dick. ## Mito-KillerRed line generation and imaging *To* generate 3xHA-killerred-OMP25, a plasmid containing 3xHA-EGFP-OMP25 (Chen et al., 2016) was used as a template and the sequence of KillerRed replaced EGFP. The entire transgene was then cloned into the pLKO.1_MCS backbone. pLKO.1 3xHA-KillerRed-OMP25 is available on Addgene (#174544). MDA-MB-231 cells were transduced with lentiviral supernant that packaged the 3xHA-killerred-OMP25 transgene, allowed to recover and were cell sorted to select for the appropriate level of fluorescent expression. Cells expressing both mito-mEm and mito-KillerRed were generated in parallel to confirm the correct localization of the mito-KillerRed (data not shown). pMXs-3XHA-EGFP-OMP25 was a gift from David Sabatini (Addgene plasmid #83356). ## For generation of mt-ROS with the mito-KillerRed cell line MDA-MB-231 mito-KillerRed-expressing cells were labeled with Carboxy-H2DCFDA as described above. Using a Leica Yokogawa CSU-W1 spinning disc confocal microscope equipped with a 2D-VisiFRAP Galvo System Multi-Point FRAP/Photoactivation module, MDA-MB-231 mito-KillerRed-expressing cells were imaged at 488 nm (for DCFDA detection) and 561 nm (for mito-KillerRed detection) at a time interval of 2 seconds. After 2 frames, a~2µm x 2µm region of interest (ROI) of mito-KillerRed was photobleached using a 561 laser ($100\%$ laser power, 5ms, 1 cycle), and continuous imaging at 488 nm and 561 nm allowed for DCFDA quantification and mito-KillerRed photobleaching, respectively. ## To quantify cell division upon ROS production Cells were stained with 5 μg/mL Hoescht 33342 as described above to visualize nuclei. Multiple stage positions were established such that control experiments, in which a cytoplasmic ROI without mito-KillerRed expression that was photobleached using identical parameters, as well as a no-photobleaching control, could be imaged simultaneously with experimental photobleached cells. Approximately 8–10 cells of each category – photobleached in mito-KillerRed-expressing regions, photobleached in control cytoplasmic non-expressing regions, or not photobleached – were imaged by acquiring Z-stacks (1 µm step size) every 15 min for 18 hr. Cell division was quantified by visualizing nuclear division with FIJI software. ## ERK-KTR generation MDA-MB-231 cells were transduced with lentiviral supernant that was packaged using either pLentiPGK Blast DEST ERKKTRmRuby2 or pLentiPGK Puro DEST ERKKTRClover plasmids (Kudo et al., 2018) as outlined in ‘generation of mito-FP’ section. Cells were allowed to recover post-infection, sorted for fluorescent expression, and maintained as stable cell lines. pLentiPGK Blast DEST ERKKTRmRuby2 (Addgene plasmid # 90231) and pLentiPGK Puro DEST ERKKTRClover (Addgene plasmid # 90227) were a gift from Markus Covert. ## ERK-KTR-mClover and mito-KillerRed generation and imaging A stable MDA-MB-231 line expressing mito-KillerRed was transduced with lentivirus that was packaged using a pLentiPGK Puro DEST ERKKTRClover plasmid. Cells were sorted for expression of both mito-KillerRed and ERK-KTR-mClover and maintained as a stable line. ## For mt-ROS generation and ERK-KTR imaging MDA-MB-231 cells expressing mito-KillerRed and ERK-KTR-mClover were stained with 5 μg/mL Hoescht 33342 as described above to visualize nuclei. Using a Leica Yokogawa CSU-W1 spinning disc confocal microscope equipped with a 2D-VisiFRAP Galvo System Multi-Point FRAP/Photoactivation module, MDA-MB-231 cells expressing mito-KillerRed and ERK-KTR-mClover were imaged every 1 minute with 561 nm (for mito-KillerRed) and 488 nm (for ERK-KTR-mClover) and 405 nm (for nuclei) lasers. A~2 µm x 2 µm ROI of KillerRed + mitochondria was photobleached using a 561 laser ($100\%$ laser power, 5ms, 1 cycle), and continuous imaging at 488 nm and 561 nm allowed for visualization of ERK-KTR-mClover translocation and mito-KillerRed photobleaching, respectively. Multiple stage positions were set such that control experiments, in which a cytoplasmic region without mito-KillerRed expression that was photobleached using identical parameters, could be imaged simultaneously with experimental photobleached cells. ## ERK-KTR quantification with FIJI ERK-KTR-mClover translocation was quantified every 10 minutes by taking maximum projections of Z-planes only encompassing the cell nucleus. Using FIJI software, a ROI was drawn in the nucleus guided by the Hoescht staining, and the MFI of ERK-KTR-mClover was quantified in this region. The same ROI was moved outside of the nucleus to a cytoplasmic region devoid of mitochondria, and the MFI of ERK-KTR-mClover was quantified. This analysis was performed for each timepoint after photobleaching. The values were then used to calculate a cytoplasmic:nuclear ratio at each time point, and normalized to 1 at time point zero. ## Quantification of ERK-KTR using the Amnis ImageStream To quantify translocation of the ERK-KTR-mRuby we used the Amnis Imagestream mk II with ISX software (version 201.1.0.725). Mito-mEm macrophages were co-cultured with ERK-KTR-mRuby+MDA MB-231 cells for 24 hr. Samples were prepared as indicated in ‘quantification of Ki67 and DNA content’ section with the exception that we did not stain for intracellular markers. Images were captured with the 40 x objective and sample collect flow was set to low, as this allows for higher image resolution. Using Image Data Exploration and Analyses Software (IDEAS; version 6), we quantified translocation using two metrics: [1] the IDEAS translocation Wizard and [2] custom-generated program to detect cytoplasmic (cyto) and nuclear (nuc) ERK-KTR mean fluorescent intensities (MFI) to calculate a cyto:nuc ratio as indicated in Figure 3—figure supplements 1 and 2. The translocation wizard is a pre-built program made to detect the nuclear translocation of a probe. It does this by making a pixel-by-pixel correlation between the probe of interest (ERK-KTR) and the nuclear image (DAPI). The program gives each cell a score indicating how similar the two fluorescent images are. A high score suggests the images are similar (more nuclear translocation) and a low score suggests that the images are less similar (less nuclear translocation). We also quantified ERK-KTR translocation by generating a custom masking strategy to quantify the mRuby MFI in the cytoplasm and nucleus using IDEAS software. To identify nuclear mRuby, we manually set a threshold of DAPI signal and reported the mRuby MFI of pixels within that threshold range. To quantify the cytoplasmic mRuby fraction we reported the mRuby MFI from outside the threshold. These values are then used to calculate a cyto:nuc ratio. ## Drug treatments: ERKi, PMA and MitoTEMPO SCH772984 (ERKi; 7101, SelleckChem) and Phorbol 12-myristate 13-acetate (PMA; S7791, Selleckchem) was dissolved in $100\%$ DMSO to make 10 mM stock solutions and stored at –80 °C. No individual aliquot went through more than 2 freeze-thaw cycles. The stock solution was thawed and then diluted directly into complete media for a final concentration of 1 μM for ERKi and 100 nM (cancer cell treatment) or 162 nM (THP-1 differentiation) for PMA. For all ERKi experiments, co-cultures were treated at the time of plating and for a duration of 24 hr. For PMA treatment of cancer cells the cells were plated the day prior and were treated for 1 hr prior to harvest and analysis. THP-1 cells were differentiated for 24 hr in PMA prior to harvest. To quench mitochondrial ROS, MitoTEMPO (Cayman Chemical, 16621) was formulated at 200 mM in $100\%$ DMSO and diluted directly into warm complete media for a final concentration of 100 μM. MDA-MB-468 cells were treated for the duration of 24 hours as described in 'Mitochondrial isolation and bath application', and cells were harvested for proliferative capacity analyses as previously in ‘Quantification of Ki67 and DNA content’. MitoTEMPO aliquots were stored at –20 °C, remained protected from light and never underwent a freeze-thaw cycle. ## Macrophage activation and verification For macrophage activation, macrophages were harvested and differentiated as indicated in ‘cell culture of PBMCs’ section. Between days 6–7 of differentiation, IFN-γ (3000–02, Peprotech, 20 ng/mL) for M1 activation or IL-4 +IL-13 (200–04, 200–13, Peprotech, 20 ng/mL) for M2 activation were added to culture media for 48 hr before experiments were conducted. To confirm M1 and M2 activation, macrophages were collected and stained for known surface markers for M1 (CD86; 62-0869-42, Thermofisher) and M2 (CD206; 321110, Biolegend) activation. Flow cytometry was performed on the Fortessa to observe changes in fluorescent intensities across M0, M1, and M2 macrophages populations (Figure 4—figure supplement 1a). ## Immunofluorescence and analysis of mitochondrial morphology Mito-mEm expressing macrophages were co-cultured with mito-RFP +MDA MB-231 cells for 24 hr and fixed with warm $4\%$ PFA with $5\%$ sucrose in 1 x DPBS for 20 min and permeabilized with $0.2\%$ Triton X-100 in 1 x DPBS (9002-93-1, Sigma). Cells were stained with chicken α-GFP (AB13970, Abcam) and Rabbit α-RFP (AB62341, Abcam) antibodies at 1:500 and 1:1000, respectively. The following secondary antibodies were used: Alexa Fluor 488 AffiniPure Goat anti-Chicken (103-545-155, Jackson ImmunoResearch) and IgG (H+L) Cross-Adsorbed Goat anti-Rabbit Alexa Fluor 555 (A21428, Invitrogen) both at 1:500. Cells were subsequently stained with 1 μg/mL DAPI in DPBS for 10 min. Cells were then mounted with ProLong Diamond Antifade Mountant (P36965, ThermoFisher) and stored at 4 °C before imaging. Imaging was performed using the Zeiss LSM 880 using the AiryScan fast mode. AiryScan processed images (see image analysis section) were used to quantify mitochondrial morphologies with the FIJI plug-in, Mitochondrial Network Analyses (MiNA; Figure 4—figure supplement 1b–d). Pre-processing parameters: Manually select top and bottom of the cell of interest, exclude any space above and below the cell as this can introduce background noise. 3D project cell. Unmask sharp Radius [5], Mask Weight (0.6), Median 3D (0.5, 0.5, 0.5), Make binary (Otsu), Skeletonize, Analyze skeleton 2D/3D. A ‘mitochondrial fragment’ was defined as a mitochondrion with 0–1 branches, 0 junctions, and a length greater than 0 µm and a maximum length of 2 µm. ## DRP1 knockdown Monocytes were isolated as indicated in ‘cell culture of PBMCs’ section and transduced with lentiviruses to express mito-mEm and either non-target (nt) short hairpin (sh) RNA (SHC002, Sigma), or DRP1-shRNA (TRCN0000001097, Sigma; gene target HGNC ID 2973). All constructs were either produced or cloned into the pLKO.1 backbone. ## rt-qPCR verification of genetic knockdown RNA from nt-shRNA and DRP1-shRNA expressing macrophages were isolated from 3 independent macrophage donors. To isolate RNA, we used standard TRIzol/chloroform RNA isolation techniques. cDNA libraries were made using SuperScript III Reverse Transcriptase (18080093, ThermoFisher), according to manufacturer’s instructions. DRP1-knockdown was verified via qRT-PCR with Power SYBR Green Mast Mix (4368511, ThermoFisher). Primers were designed with NCBI primer design, commercially produced by Integrated DNA Technologies and tested for specificity with standard PCR. Primers were as follows; DRP1-F: AGAAAATGGGGTGGAAGCAGA, DRP1-R: AAGTGCCTCTGATGTTGCCA, GAPDH-F: AGCCACATCGCTCAGACA, GAPDH-R: ACATGTAAACCATGTAGTTGAGGT. Cycle Thresholds (CT) values were determined by averaging 3 technical replicates from 3 biological samples. Control ΔCT: expression was normalized to GAPDH by subtracting the DRP1 CT value of the nt-shRNA expressing macrophages from the GAPDH CT value of the same sample. *Target* gene, DRP1ΔCT: DRP1 CT values of the DRP1-shRNA expressing macrophages were subtracted from the GAPDH CT values of the same sample. The ΔΔCT values was calculated by subtracting DRP1ΔCT – control ΔCT. Normalized target gene expression was calculated (2-ΔΔCT) and used to determine % knockdown ((1–2-ΔΔCT)*100). ## PDxO culture and co-culture with macrophages PDxO cell lines HCI-037 and HCI-038 were generated and maintained as described in Guillen et al., 2021. Like MDA-MB-231 cells, these models are estrogen and progesterone receptor negative and HER2 negative (triple negative breast cancer). For co-culture with macrophages, mature PDxOs were dissociated from Growth Factor Reduced Matrigel with a Dispase II solution followed by treatment with TrypLE Express to generate a suspension of single cells. PDxO cells were then mixed with mito-mEm macrophages (differentiated for 7–9 days) in a 1:2 ratio at a density of 90,000 cells total per hanging drop culture. Macrophage media was used for hanging drops (for media components, see isolation of PBMCs section) and they were suspended from the lid of a tissue culture plate to allow for cell aggregation for 24 hr and then pooled and embedded into Growth Factor Reduced Matrigel. Embedded hanging drop cultures were then allowed to incubate for 72 hr and were then analyzed for mitochondrial transfer with flow cytometry (see flow cytometry section). ## Mitochondrial isolation and bath application For data represented in Figure 1g–j: 150–200x106 mito-mEm expressing THP-1 cells were pelleted by centrifugation for 5 min at 300 g. Pellets were resuspended in 2 mL of mitochondrial isolation buffer (70 mM sucrose, 220 mM D-mannitol, 2 mM HEPES, 1 x protease inhibitor, pH 7.4) and incubated on ice for 15–30 min. Suspended cells were dounce homogenized 100–150 times in a Potter-Elvehjem PTFE pestle and glass tube (Sigma, P7734). Cell homogenates were centrifuged at least twice (700 g for 10 min, 4 °C) to pellet and remove unwanted cellular material, until no pellet was observable – as many as 7 centrifugation cycles. Final supernatants were centrifuged at 20,000 g for 15 min at 4 °C to pellet isolated mitochondria. Mitochondrial pellets were suspended in ~250 μL of ice cold mitochondrial isolation buffer +protease inhibitor, and relative mitochondrial concentrations were determined via standard BCA protein concentration assay (ThermoFisher, 23225). 20–30 µg/mL of mitochondria were applied to pre-plated mito-RFP expressing MDA-MB-231 cells for 18–24 hours. After mitochondrial incubation cells were thoroughly washed to remove any un-internalized mitochondria and mitochondrial percent internalization was determined via flow cytometry and cells were FACS isolated with BD FACS Aria as described above under ‘Flow cytometry – Stable line generation’. FACS isolated cells were either imaged on the LSM880 Airy Scan Confocal as described in ‘Live cell imaging of co-cultures with cell-permeable dyes’, or plated for an additional 48 hr. After roughly two cell cycles, the cells were harvested and cell cycle analyses were conducted as described in ‘Quantification of Ki67 and DNA content’. For data represented in Figures 2e–f,–3f: Mito-mEm expressing THP-1 monocytes were differentiated with 162 nM Phorbol 12-myristate 13-acetate (PMA - SelleckChem, #S7791) for 24 hr. Cells were trypsinized and washed with ice cold PBS and centrifuged at 300 g for 5 min. Cell pellets were suspended in 500–1000 μL mitochondrial isolation buffer +protease inhibitor and dounce homogenized as reported above. Final supernatants were centrifuged at 20,000 g for 15 min at 4 °C to pellet isolated mitochondria. Mitochondrial pellets were suspended in 110 μL of ice cold mitochondrial isolation buffer +protease inhibitor, and mitochondrial concentrations were determined via standard BCA (as above). Pre-plated MDA-MB-468 cells were bath applied with concentrations 3–5 µg/mL of exogenous mitochondria for 5–6 hr which was then removed to eliminate any un-internalized mitochondria. Twenty-four hr after initial mitochondrial addition, cells were either imaged on the LSM880 Airy Scan Confocal as described in ‘Live cell imaging of co-cultures with cell-permeable dyes’, treated with ERK inhibitor as described in ‘Drug treatments: ERKi, PMA and MitoTEMPO’ and harvested for cell cycle analyses were conducted as described in ‘Quantification of Ki67 and DNA content’. All drug treatments (ERK inhibitor and MitoTEMPO) were applied at the time of mitochondrial application and were maintained until harvest. ## In vivo models All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Utah (protocol # 19–12001) and at the Cleveland Clinic (protocol #2179). In accordance to approved protocols, all animals were anesthetized appropriately to assure maximum comfort throughout the duration of procedures. When tumors were grown to approved volumes, mice were humanely euthanized with slow C02 gas exchange for 5 min. We calculated how many animals would be required for each experiment using G*Power3.1 – Based on our in vitro studies, we considered a $5\%$ increase in mitochondrial transfer or a $5\%$ increase in the percentage of cells in the G2/M phase of the cell cycle as statistically significant, with a $1\%$ standard deviation, thus, we required a minimum of three animals per treatment group. With the variability in tumor growth, we injected at least five animals per treatment group such that we could ensure to complete studies with at least three animals. Regarding Figure 5a: Six-week-old C57BL/6J (The Jackson Laboratory, Stock #000664) and Tg(CAG-mKate2)1Poche/J (The Jackson Laboratory, mito::mKate2, stock #032188) female mice were purchased from the Jackson Laboratory as required and housed in the Cleveland Clinic Biological Research Unit Facility. Wild-type mice were treated with 11 Gy radiation split into two fractions. 2x106 bone marrow cells from mito:mKate2 or wild-type mice were retro-orbitally injected for reconstitution. Drinking water was supplemented with Sulfatrim (Pharmaceutical Associates, Inc) during the first 10 days, and mice were monitored for an additional 6 weeks. A total of 250,000 mito-mEm E0771 cells were mixed with 1:50 diluted Geltrex (ThermoFisher) and implanted to 4th mammary pad in 100 μl RPMI. Mice were treated with Buprenorphine and Ibuprofen for 3 days, and monitored for endpoint symptoms. Animals were euthanized when the tumors reached 1 cm3 or $10\%$ of the body weight was lost. Resected tumors were minced and incubated with Collagenase IV (StemCell Technologies) containing DNAseI (Roche) for 30 min at 37 °C. Single cells were strained through 70 μm filter (FisherBrand) and stained with 1:1000 diluted LIVE/DEAD Fixable Stains (ThermoFisher) for 10 min on ice. Samples were acquired with BD Fortessa. For in vivo cell cycle analysis upon macrophage mitochondrial transfer in Figure 5d: 8–12 week old B6N.Cg-Gt(ROSA)26Sortm1(CAG-EGFP*)Thm/J (also known as MitoTag mice, The Jackson Laboratory stock 032675 Fecher et al., 2019 and B6.129P2-Lyz2tm1(cre)Ifo/J) (also known as LysMcre mice, The Jackson Laboratory, stock 004781) were ordered and crossed accordingly, producing offspring which were heterozygous for both transgenes. These heterozygous siblings were crossed to produce both experimental (MitoTag/cre) and control (WT/cre) animals in the same litter. 250,000 Mito-RFP E0071 cells were injected into the mammary fat pad at a 1:1 ratio of matrigel (Corning) and sterile 1 x PBS into 6–8 week old mice of the appropriate genotypes. When the largest tumor reached 1cm3 the mice were euthanized and the tumors were homogenized as above and processed for Ki67 flow cytometry as listed in ‘Quantification of Ki67 and DNA content’. ## Agent-based model for impact of mitochondrial transfer on cell division over time The agent based model performs a Monte-Carlo simulation of individual cell ‘agents’ over time. At every timepoint, cells increase in mass, m, over the simulated time interval Δt according to an exponential growth law:[1]mt+1=mt+k⋅mt⋅Δt here k is the exponential growth constant. Over every time interval, a fraction of cells, fΔt, gain transferred mitochondria:[2]fΔt=fΔtTd where f is the overall fraction of the population gaining mitochondria (set to $5\%$ based on our observation that this fraction of the population has transferred mitochondria) over the cell doubling time, Td. The exponential growth constant, k, is then equal to k0, the baseline growth rate, for cell agents without transferred mitochondria, or k0r, where r is the factor of growth rate increase for cells with transferred mitochondria. If the mass of a cell is greater than double its baseline mass, then it divides into two new daughter cells, each at half the mass of the parent, that are then tracked in the simulation. We assume that half the population loses transferred mitochondria (and the associated growth increase) with every division. The mass of the population, mP, is then found by summing the mass of all individual cell agents at a given time. This result can be compared to the overall final tumor mass in the baseline case, mB, after a given number, d, of doublings based on pure exponential growth:[3]mB=m0 edln⁡2 This result is plotted in Figure 5—figure supplement 2 for the case of $5\%$ of the tumor cell population receiving macrophage mitochondria. ## Data and materials availability The code for QPI analysis is available on GitHub (https://github.com/Zangle-Lab/Macrophage_tumor_mito_transfer). Single-cell RNA-sequencing data are available in GEO accession number GSE181410. The code for single-cell RNA-sequencing analysis is available on GitHub (https://github.com/rohjohnson-lab/kidwell_casalini_2021; RRID:SCR_002630 (Version 1)). All other data are available in the main text or in the Supplementary Data. ## Image analysis All images taken with the Airyscan detector on the Zeiss LSM 880 were subjected to deconvolution using the Zen software (Carl Zeiss) with 'auto' settings (referred to as AiryScan processed). Maximum intensity projections of selected z-planes were generated using Zen or FIJI software (Schindelin et al., 2012). Linear adjustments to the brightness and contrast were made using FIJI. Images were cropped and panels were assembled using Adobe Photoshop and Illustrator, respectively (Adobe, Inc). ## Graphical representations and statistical analysis All graphs were generated using Prism software (v9, GraphPad). All graphs show mean with standard error of the mean. Statistical analyses were performed using both Excel (v16.51, Microsoft) and Prism. Statistical tests used and p-value ranges are indicated in each figure legend. Nested statistical tests were used to take into account the technical replicates within each biological replicate in the analysis of variance tests. Flow cytometry data and representations were analyzed and generated using FlowJo software (v10.7, BD). Welches t-test was used when the goal was to compare mean values of data with normal distribution, and Mann-Whitney analyses was applied when the data was not normally distributed. Two-way ANOVA was utilized when comparing how two independent variables influence a dependent variable. All statistical methodologies were performed under the guidance of biostatistician, Dr. Kenneth M. Boucher. ## Funding Information This paper was supported by the following grants: ## Data availability The code for QPI analysis is available on GitHub (https://github.com/Zangle-Lab/Macrophage_tumor_mito_transfer; copy archived at ZangleLab, 2023) for Figure 1.Single-cell RNA-sequencing data are available in GEO accession number GSE181410. The code for single-cell RNA-sequencing analysis is available on GitHub (https://github.com/rohjohnson-lab/kidwell_casalini_2021; copy archived at rohjohnson-lab, 2023) for Figure 1. The following previously published dataset was used: Roh-JohnsonM GreinerD 2021Macrophage and MDA-MB-231 coculture and mitochondrial transferNCBI Gene Expression OmnibusGSE181410 ## References 1. Ahmad T, Mukherjee S, Pattnaik B, Kumar M, Singh S, Kumar M, Rehman R, Tiwari BK, Jha KA, Barhanpurkar AP, Wani MR, Roy SS, Mabalirajan U, Ghosh B, Agrawal A. **Miro1 regulates intercellular mitochondrial transport & enhances mesenchymal stem cell rescue efficacy**. *The EMBO Journal* (2014) **33** 994-1010. DOI: 10.1002/embj.201386030 2. Antanavičiūtė I, Rysevaitė K, Liutkevičius V, Marandykina A, Rimkutė L, Sveikatienė R, Uloza V, Skeberdis VA. **Long-distance communication between laryngeal carcinoma cells**. *PLOS ONE* (2014) **9**. DOI: 10.1371/journal.pone.0099196 3. Auten RL, Whorton MH, Nicholas Mason S. **Blocking neutrophil influx reduces DNA damage in hyperoxia-exposed newborn rat lung**. *American Journal of Respiratory Cell and Molecular Biology* (2002) **26** 391-397. DOI: 10.1165/ajrcmb.26.4.4708 4. Bass DA, Parce JW, Dechatelet LR, Szejda P, Seeds MC, Thomas M. **Flow cytometric studies of oxidative product formation by neutrophils: A graded response to membrane stimulation**. *Journal of Immunology* (1983) **130** 1910-1917. PMID: 6833755 5. Biel TG, Rao VA. **Mitochondrial dysfunction activates lysosomal-dependent mitophagy selectively in cancer cells**. *Oncotarget* (2018) **9** 995-1011. DOI: 10.18632/oncotarget.23171 6. Boukelmoune N, Chiu GS, Kavelaars A, Heijnen CJ. **Mitochondrial transfer from mesenchymal stem cells to neural stem cells protects against the neurotoxic effects of cisplatin**. *Acta Neuropathologica Communications* (2018) **6**. DOI: 10.1186/s40478-018-0644-8 7. Brestoff JR, Wilen CB, Moley JR, Li Y, Zou W, Malvin NP, Rowen MN, Saunders BT, Ma H, Mack MR, Hykes BL, Balce DR, Orvedahl A, Williams JW, Rohatgi N, Wang X, McAllaster MR, Handley SA, Kim BS, Doench JG, Zinselmeyer BH, Diamond MS, Virgin HW, Gelman AE, Teitelbaum SL. **Intercellular mitochondria transfer to macrophages regulates white adipose tissue homeostasis and is impaired in obesity**. *Cell Metabolism* (2021) **33** 270-282. DOI: 10.1016/j.cmet.2020.11.008 8. Brillo V, Chieregato L, Leanza L, Muccioli S, Costa R. **Mitochondrial dynamics, ROS, and cell signaling: A blended overview**. *MDPI* (2021) **11**. DOI: 10.3390/life11040332 9. Bulina ME, Chudakov DM, Britanova OV, Yanushevich YG, Staroverov DB, Chepurnykh TV, Merzlyak EM, Shkrob MA, Lukyanov S, Lukyanov KA. **A genetically encoded photosensitizer**. *Nature Biotechnology* (2006) **24** 95-99. DOI: 10.1038/nbt1175 10. Caicedo A, Fritz V, Brondello JM, Ayala M, Dennemont I, Abdellaoui N, de Fraipont F, Moisan A, Prouteau CA, Boukhaddaoui H, Jorgensen C, Vignais ML. **MitoCeption as a new tool to assess the effects of mesenchymal stem/stromal cell mitochondria on cancer cell metabolism and function**. *Scientific Reports* (2015) **5**. DOI: 10.1038/srep09073 11. Chen WW, Freinkman E, Wang T, Birsoy K, Sabatini DM. **Absolute quantification of matrix metabolites reveals the dynamics of mitochondrial metabolism**. *Cell* (2016) **166** 1324-1337. DOI: 10.1016/j.cell.2016.07.040 12. Chen G, Kroemer G, Kepp OM. **Mitophagy: An emerging role in aging and age-associated diseases**. *Frontiers in Cell and Developmental Biology* (2020) **8**. DOI: 10.3389/fcell.2020.00200 13. Crewe C, Funcke JB, Li S, Joffin N, Gliniak CM, Ghaben AL, An YA, Sadek HA, Gordillo R, Akgul Y, Chen S, Samovski D, Fischer-Posovszky P, Kusminski CM, Klein S, Scherer PE. **Extracellular vesicle-based interorgan transport of mitochondria from energetically stressed adipocytes**. *Cell Metabolism* (2021) **33** 1853-1868. DOI: 10.1016/j.cmet.2021.08.002 14. Crocker JC, Grier DG. **Methods of digital video microscopy for colloidal studies**. *J Colloid Interface Sci* (1996) **179** 298-310. DOI: 10.1006/jcis.1996.0217 15. Dhillon AS, Hagan S, Rath O, Kolch W. **Map kinase signalling pathways in cancer**. *Oncogene* (2007) **26** 3279-3290. DOI: 10.1038/sj.onc.1210421 16. Dong LF, Kovarova J, Bajzikova M, Bezawork-Geleta A, Svec D, Endaya B, Sachaphibulkij K, Coelho AR, Sebkova N, Ruzickova A, Tan AS, Kluckova K, Judasova K, Zamecnikova K, Rychtarcikova Z, Gopalan V, Andera L, Sobol M, Yan B, Pattnaik B, Bhatraju N, Truksa J, Stopka P, Hozak P, Lam AK, Sedlacek R, Oliveira PJ, Kubista M, Agrawal A, Dvorakova-Hortova K, Rohlena J, Berridge MV, Neuzil J. **Horizontal transfer of whole mitochondria restores tumorigenic potential in mitochondrial DNA-deficient cancer cells**. *eLife* (2017) **6**. DOI: 10.7554/eLife.22187 17. English K, Shepherd A, Uzor NE, Trinh R, Kavelaars A, Heijnen CJ. **Astrocytes rescue neuronal health after cisplatin treatment through mitochondrial transfer**. *Acta Neuropathologica Communications* (2020) **8**. DOI: 10.1186/s40478-020-00897-7 18. Fecher C, Trovò L, Müller SA, Snaidero N, Wettmarshausen J, Heink S, Ortiz O, Wagner I, Kühn R, Hartmann J, Karl RM, Konnerth A, Korn T, Wurst W, Merkler D, Lichtenthaler SF, Perocchi F, Misgeld T. **Cell-type-specific profiling of brain mitochondria reveals functional and molecular diversity**. *Nature Neuroscience* (2019) **22** 1731-1742. DOI: 10.1038/s41593-019-0479-z 19. Feng Y, Liu Y, Ma X, Xu L, Ding D, Chen L, Wang Z, Qin R, Sun W, Chen H. **Intracellular marriage of bicarbonate and mn ions as “ immune ion reactors” to regulate redox homeostasis and enhanced antitumor immune responses**. *Journal of Nanobiotechnology* (2022) **20**. DOI: 10.1186/s12951-022-01404-x 20. Fonseca TB, Sánchez-Guerrero Á, Milosevic I, Raimundo N. **Mitochondrial fission requires drp1 but not dynamins**. *Nature* (2019) **570** E34-E42. DOI: 10.1038/s41586-019-1296-y 21. Franco-Iborra S, Cuadros T, Parent A, Romero-Gimenez J, Vila M, Perier C. **Defective mitochondrial protein import contributes to complex I-induced mitochondrial dysfunction and neurodegeneration in parkinson’s disease**. *Cell Death & Disease* (2018) **9**. DOI: 10.1038/s41419-018-1154-0 22. Fruhwirth GO, Hermetter A. **Mediation of apoptosis by oxidized phospholipids**. *Sub-Cellular Biochemistry* (2008) **49** 351-367. DOI: 10.1007/978-1-4020-8831-5_13 23. Greiner D, Scott TM, Olson GS, Aderem A, Roh-Johnson M, Johnson JS. **Genetic modification of primary human myeloid cells to study cell migration, activation, and organelle dynamics**. *Current Protocols* (2022) **2**. DOI: 10.1002/cpz1.514 24. Guillen KP, Fujita M, Butterfield AJ, Scherer SD, Bailey MH, Chu Z, DeRose YS, Zhao L, Cortes-Sanchez E, Yang CH, Toner J, Wang G, Qiao Y, Huang X, Greenland JA, Vahrenkamp JM, Lum DH, Factor RE, Nelson EW, Matsen CB, Poretta JM, Rosenthal R, Beck AC, Buys SS, Vaklavas C, Ward JH, Jensen RL, Jones KB, Li Z, Oesterreich S, Dobrolecki LE, Pathi SS, Woo XY, Berrett KC, Wadsworth ME, Chuang JH, Lewis MT, Marth GT, Gertz J, Varley KE, Welm BE, Welm AL. **A breast cancer patient-derived xenograft and organoid platform for drug discovery and precision oncology**. *bioRxiv* (2021). DOI: 10.1101/2021.02.28.433268 25. Gutscher M, Pauleau AL, Marty L, Brach T, Wabnitz GH, Samstag Y, Meyer AJ, Dick TP. **Real-time imaging of the intracellular glutathione redox potential**. *Nature Methods* (2008) **5** 553-559. DOI: 10.1038/nmeth.1212 26. Gutscher M, Sobotta MC, Wabnitz GH, Ballikaya S, Meyer AJ, Samstag Y, Dick TP. **Proximity-based protein thiol oxidation by H2O2-scavenging peroxidases**. *The Journal of Biological Chemistry* (2009) **284** 31532-31540. DOI: 10.1074/jbc.M109.059246 27. Hafemeister C, Satija R. **Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression**. *Genome Biology* (2019) **20**. DOI: 10.1186/s13059-019-1874-1 28. Hao Y, Hao S, Andersen-Nissen E, Mauck WM, Zheng S, Butler A, Lee MJ, Wilk AJ, Darby C, Zagar M, Hoffman P, Stoeckius M, Papalexi E, Mimitou EP, Jain J, Srivastava A, Stuart T, Fleming LB, Yeung B, Rogers AJ, McElrath JM, Blish CA, Gottardo R, Smibert P, Satija R. **Integrated analysis of multimodal single-cell data**. *bioRxiv* (2020). DOI: 10.1101/2020.10.12.335331 29. Huang X, Li Y, Fu M, Xin HB. **Polarizing macrophages in vitro**. *Methods in Molecular Biology* (2018) **1784** 119-126. DOI: 10.1007/978-1-4939-7837-3_12 30. Johnson JS, De Veaux N, Rives AW, Lahaye X, Lucas SY, Perot BP, Luka M, Garcia-Paredes V, Amon LM, Watters A, Abdessalem G, Aderem A, Manel N, Littman DR, Bonneau R, Ménager MM. **A comprehensive map of the monocyte-derived dendritic cell transcriptional network engaged upon innate sensing of HIV**. *Cell Reports* (2020) **30** 914-931. DOI: 10.1016/j.celrep.2019.12.054 31. Korpershoek JV, Rikkers M, Wallis FSA, Dijkstra K, Te Raa M, de Knijff P, Saris DBF, Vonk LA. **Mitochondrial transport from mesenchymal stromal cells to chondrocytes increases DNA content and proteoglycan deposition in vitro in 3D cultures**. *Cartilage* (2022) **13** 133-147. DOI: 10.1177/19476035221126346 32. Krämer A, Green J, Pollard J, Tugendreich S. **Causal analysis approaches in ingenuity pathway analysis**. *Bioinformatics* (2014) **30** 523-530. DOI: 10.1093/bioinformatics/btt703 33. Kudo T, Jeknić S, Macklin DN, Akhter S, Hughey JJ, Regot S, Covert MW. **Live-cell measurements of kinase activity in single cells using translocation reporters**. *Nature Protocols* (2018) **13** 155-169. DOI: 10.1038/nprot.2017.128 34. Lampinen R, Belaya I, Saveleva L, Liddell JR, Rait D, Huuskonen MT, Giniatullina R, Sorvari A, Soppela L, Mikhailov N, Boccuni I, Giniatullin R, Cruz-Haces M, Konovalova J, Koskuvi M, Domanskyi A, Hämäläinen RH, Goldsteins G, Koistinaho J, Malm T, Chew S, Rilla K, White AR, Marsh-Armstrong N, Kanninen KM. **Neuron-astrocyte transmitophagy is altered in alzheimer’s disease**. *Neurobiology of Disease* (2022) **170**. DOI: 10.1016/j.nbd.2022.105753 35. Leadsham JE, Sanders G, Giannaki S, Bastow EL, Hutton R, Naeimi WR, Breitenbach M, Gourlay CW. **Loss of cytochrome c oxidase promotes ras-dependent ROS production from the ER resident NADPH oxidase, yno1p, in yeast**. *Cell Metabolism* (2013) **18** 279-286. DOI: 10.1016/j.cmet.2013.07.005 36. Liu D, Gao Y, Liu J, Huang Y, Yin J, Feng Y, Shi L, Meloni BP, Zhang C, Zheng M, Gao J. **Intercellular mitochondrial transfer as a means of tissue revitalization**. *Signal Transduction and Targeted Therapy* (2021) **6**. DOI: 10.1038/s41392-020-00440-z 37. Liu Y, Wu M, Zhong C, Xu B, Kang L. **M2-like macrophages transplantation protects against the doxorubicin-induced heart failure via mitochondrial transfer**. *Biomaterials Research* (2022) **26**. DOI: 10.1186/s40824-022-00260-y 38. Lou E, Fujisawa S, Morozov A, Barlas A, Romin Y, Dogan Y, Gholami S, Moreira AL, Manova-Todorova K, Moore MAS. **Tunneling nanotubes provide a unique conduit for intercellular transfer of cellular contents in human malignant pleural mesothelioma**. *PLOS ONE* (2012) **7**. DOI: 10.1371/journal.pone.0033093 39. Ma K, Chen G, Li W, Kepp O, Zhu Y, Chen Q. **Mitophagy, mitochondrial homeostasis, and cell fate**. *Frontiers in Cell and Developmental Biology* (2020) **8**. DOI: 10.3389/fcell.2020.00467 40. Morgan B, Sobotta MC, Dick TP. **Measuring E (GSH) and H2O2 with rogfp2-based redox probes**. *Free Radical Biology & Medicine* (2011) **51** 1943-1951. DOI: 10.1016/j.freeradbiomed.2011.08.035 41. Mortezaee K, Majidpoor J. **Dysregulated metabolism: A friend-to-foe skewer of macrophages**. *International Reviews of Immunology* (2022) **1** 1-17. DOI: 10.1080/08830185.2022.2095374 42. Nakai M, Mori A, Watanabe A, Mitsumoto Y. **1-methyl-4-phenylpyridinium (MPP+) decreases mitochondrial oxidation-reduction (redox) activity and membrane potential (deltapsi (M)) in rat striatum**. *Experimental Neurology* (2003) **179** 103-110. DOI: 10.1006/exnr.2002.8056 43. Nicolás-Ávila JA, Lechuga-Vieco AV, Esteban-Martínez L, Sánchez-Díaz M, Díaz-García E, Santiago DJ, Rubio-Ponce A, Li JL, Balachander A, Quintana JA, Martínez-de-Mena R, Castejón-Vega B, Pun-García A, Través PG, Bonzón-Kulichenko E, García-Marqués F, Cussó L, A-González N, González-Guerra A, Roche-Molina M, Martin-Salamanca S, Crainiciuc G, Guzmán G, Larrazabal J, Herrero-Galán E, Alegre-Cebollada J, Lemke G, Rothlin CV, Jimenez-Borreguero LJ, Reyes G, Castrillo A, Desco M, Muñoz-Cánoves P, Ibáñez B, Torres M, Ng LG, Priori SG, Bueno H, Vázquez J, Cordero MD, Bernal JA, Enríquez JA, Hidalgo A. **A network of macrophages supports mitochondrial homeostasis in the heart**. *Cell* (2020) **183** 94-109. DOI: 10.1016/j.cell.2020.08.031 44. Pan Y, Yu Y, Wang X, Zhang T. **Tumor-associated macrophages in tumor immunity**. *Frontiers in Immunology* (2020) **11**. DOI: 10.3389/fimmu.2020.583084 45. Patel SP, Michael FM, Gollihue JL, Brad Hubbard W, Sullivan PG, Rabchevsky AG. **Delivery of mitoceuticals or respiratory competent mitochondria to sites of neurotrauma**. *Mitochondrion* (2023) **68** 10-14. DOI: 10.1016/j.mito.2022.11.001 46. Phinney DG, Di Giuseppe M, Njah J, Sala E, Shiva S, St Croix CM, Stolz DB, Watkins SC, Di YP, Leikauf GD, Kolls J, Riches DWH, Deiuliis G, Kaminski N, Boregowda SV, McKenna DH, Ortiz LA. **Mesenchymal stem cells use extracellular vesicles to outsource mitophagy and shuttle micrornas**. *Nature Communications* (2015) **6**. DOI: 10.1038/ncomms9472 47. Pollard JW. **Tumour-educated macrophages promote tumour progression and metastasis**. *Nature Reviews. Cancer* (2004) **4** 71-78. DOI: 10.1038/nrc1256 48. Poot M, Zhang YZ, Krämer JA, Wells KS, Jones LJ, Hanzel DK, Lugade AG, Singer VL, Haugland RP. **Analysis of mitochondrial morphology and function with novel fixable fluorescent stains**. *The Journal of Histochemistry and Cytochemistry* (1996) **44** 1363-1372. DOI: 10.1177/44.12.8985128 49. Rebbeck CA, Leroi AM, Burt A. **Mitochondrial capture by a transmissible cancer**. *Science* (2011) **331**. DOI: 10.1126/science.1197696 50. Regot S, Hughey JJ, Bajar BT, Carrasco S, Covert MW. **High-sensitivity measurements of multiple kinase activities in live single cells**. *Cell* (2014) **157** 1724-1734. DOI: 10.1016/j.cell.2014.04.039 51. Ristow M, Schmeisser K. **Mitohormesis: Promoting health and lifespan by increased levels of reactive oxygen species (ROS)**. *Dose-Response* (2014) **12** 288-341. DOI: 10.2203/dose-response.13-035.Ristow 52. Roh-Johnson M, Shah AN, Stonick JA, Poudel KR, Kargl J, Yang GH, di Martino J, Hernandez RE, Gast CE, Zarour LR, Antoku S, Houghton AM, Bravo-Cordero JJ, Wong MH, Condeelis J, Moens CB. **Macrophage-dependent cytoplasmic transfer during melanoma invasion in vivo**. *Developmental Cell* (2017) **43** 549-562. DOI: 10.1016/j.devcel.2017.11.003 53. rohjohnson-lab 2023Kidwell_casalini_2021swh:1:rev:58c92cd368dc3a561631b46e21c7e9dd67f9e9ebSoftware Heritagehttps://archive.softwareheritage.org/swh:1:dir:c5ee0aca070eddd321916062b5c709fcb0584eac;origin=https://github.com/rohjohnson-lab/kidwell_casalini_2021;visit=swh:1:snp:a0e85d4d7afddff7561a6b236c07cf52a530c45b;anchor=swh:1:rev:58c92cd368dc3a561631b46e21c7e9dd67f9e9eb. *Software Heritage* (2023) 54. Saha T, Dash C, Jayabalan R, Khiste S, Kulkarni A, Kurmi K, Mondal J, Majumder PK, Bardia A, Jang HL, Sengupta S. **Intercellular nanotubes mediate mitochondrial trafficking between cancer and immune cells**. *Nature Nanotechnology* (2022) **17** 98-106. DOI: 10.1038/s41565-021-01000-4 55. Schieber M, Chandel NS. **Ros function in redox signaling and oxidative stress**. *Current Biology* (2014) **24** R453-R462. DOI: 10.1016/j.cub.2014.03.034 56. Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, Preibisch S, Rueden C, Saalfeld S, Schmid B, Tinevez JY, White DJ, Hartenstein V, Eliceiri K, Tomancak P, Cardona A. **Fiji: an open-source platform for biological-image analysis**. *Nature Methods* (2012) **9** 676-682. DOI: 10.1038/nmeth.2019 57. Stadtman ER, Levine RL. **Protein oxidation**. *Annals of the New York Academy of Sciences* (2000) **899** 191-208. DOI: 10.1111/j.1749-6632.2000.tb06187.x 58. Tan AS, Baty JW, Dong LF, Bezawork-Geleta A, Endaya B, Goodwin J, Bajzikova M, Kovarova J, Peterka M, Yan B, Pesdar EA, Sobol M, Filimonenko A, Stuart S, Vondrusova M, Kluckova K, Sachaphibulkij K, Rohlena J, Hozak P, Truksa J, Eccles D, Haupt LM, Griffiths LR, Neuzil J, Berridge MV. **Mitochondrial genome acquisition restores respiratory function and tumorigenic potential of cancer cells without mitochondrial DNA**. *Cell Metabolism* (2015) **21** 81-94. DOI: 10.1016/j.cmet.2014.12.003 59. Torralba D, Baixauli F, Sánchez-Madrid F. **Mitochondria know no boundaries: Mechanisms and functions of intercellular mitochondrial transfer**. *Frontiers in Cell and Developmental Biology* (2016) **4**. DOI: 10.3389/fcell.2016.00107 60. Valente AJ, Maddalena LA, Robb EL, Moradi F, Stuart JA. **A simple imagej macro tool for analyzing mitochondrial network morphology in mammalian cell culture**. *Acta Histochemica* (2017) **119** 315-326. DOI: 10.1016/j.acthis.2017.03.001 61. Van den Bossche J, O’Neill LA, Menon D. **Macrophage immunometabolism: Where are we (going)?**. *Trends in Immunology* (2017) **38** 395-406. DOI: 10.1016/j.it.2017.03.001 62. van der Merwe M, van Niekerk G, Fourie C, du Plessis M, Engelbrecht A-M. **The impact of mitochondria on cancer treatment resistance**. *Cellular Oncology* (2021) **44** 983-995. DOI: 10.1007/s13402-021-00623-y 63. van der Vlist M, Raoof R, Willemen H, Prado J, Versteeg S, Martin Gil C, Vos M, Lokhorst RE, Pasterkamp RJ, Kojima T, Karasuyama H, Khoury-Hanold W, Meyaard L, Eijkelkamp N. **Macrophages transfer mitochondria to sensory neurons to resolve inflammatory pain**. *Neuron* (2022) **110** 613-626. DOI: 10.1016/j.neuron.2021.11.020 64. Wang X, Gerdes HH. **Long-distance electrical coupling via tunneling nanotubes**. *Biochimica et Biophysica Acta* (2012) **1818** 2082-2086. DOI: 10.1016/j.bbamem.2011.09.002 65. Wang X, Gerdes HH. **Transfer of mitochondria via tunneling nanotubes rescues apoptotic PC12 cells**. *Cell Death and Differentiation* (2015) **22** 1181-1191. DOI: 10.1038/cdd.2014.211 66. Wang J, Liu X, Qiu Y, Shi Y, Cai J, Wang B, Wei X, Ke Q, Sui X, Wang Y, Huang Y, Li H, Wang T, Lin R, Liu Q, Xiang AP. **Cell adhesion-mediated mitochondria transfer contributes to mesenchymal stem cell-induced chemoresistance on T cell acute lymphoblastic leukemia cells**. *Journal of Hematology & Oncology* (2018) **11**. DOI: 10.1186/s13045-018-0554-z 67. Yang C, Yokomori R, Chua LH, Tan SH, Tan DQ, Miharada K, Sanda T, Suda T. **Mitochondria transfer mediates stress erythropoiesis by altering the bioenergetic profiles of early erythroblasts through CD47**. *The Journal of Experimental Medicine* (2022) **219**. DOI: 10.1084/jem.20220685 68. Yao CH, Wang R, Wang Y, Kung CP, Weber JD, Patti GJ. **Mitochondrial fusion supports increased oxidative phosphorylation during cell proliferation**. *eLife* (2019) **8**. DOI: 10.7554/eLife.41351 69. Zampieri LX, Silva-Almeida C, Rondeau JD, Sonveaux P. **Mitochondrial transfer in cancer: a comprehensive review**. *International Journal of Molecular Sciences* (2021) **22**. DOI: 10.3390/ijms22063245 70. Zangle TA, Teitell MA. **Live-cell mass profiling: an emerging approach in quantitative biophysics**. *Nature Methods* (2014) **11** 1221-1228. DOI: 10.1038/nmeth.3175 71. Zangle TA, Teitell MA, Reed J. **Live cell interferometry quantifies dynamics of biomass partitioning during cytokinesis**. *PLOS ONE* (2014) **9**. DOI: 10.1371/journal.pone.0115726 72. ZangleLab 2023Macrophage_tumor_mito_transferswh:1:rev:f9af8d52768b3f28c2ff0c9edc63997883541905Software Heritagehttps://archive.softwareheritage.org/swh:1:dir:845ecdf0d508aa287e93df65cef90be9b7a000fd;origin=https://github.com/Zangle-Lab/Macrophage_tumor_mito_transfer;visit=swh:1:snp:d76b7a36cf8b831c9cab2162a8462664c804d4ad;anchor=swh:1:rev:f9af8d52768b3f28c2ff0c9edc63997883541905. *Software Heritage* (2023) 73. Zhao RZ, Jiang S, Zhang L, Yu ZB. **Mitochondrial electron transport chain, ROS generation and uncoupling (review)**. *International Journal of Molecular Medicine* (2019) **44** 3-15. DOI: 10.3892/ijmm.2019.4188 74. Zhou Y, Long D, Zhao Y, Li S, Liang Y, Wan L, Zhang J, Xue F, Feng L. **Oxidative stress-mediated mitochondrial fission promotes hepatic stellate cell activation via stimulating oxidative phosphorylation**. *Cell Death & Disease* (2022) **13**. DOI: 10.1038/s41419-022-05088-x 75. Zong WX, Rabinowitz JD, White E. **Mitochondria and cancer**. *Molecular Cell* (2016) **61** 667-676. DOI: 10.1016/j.molcel.2016.02.011
--- title: 'Hypertension Control Cascade and Regional Performance in India: A Repeated Cross-Sectional Analysis (2015-2021)' journal: Cureus year: 2023 pmcid: PMC10042544 doi: 10.7759/cureus.35449 license: CC BY 3.0 --- # Hypertension Control Cascade and Regional Performance in India: A Repeated Cross-Sectional Analysis (2015-2021) ## Abstract Background The weak control cascade of hypertension from the time of screening till the attainment of optimal blood pressure (BP) control is a public health challenge, particularly in resource-limited settings. The study objectives were to [1] estimate the change in the rate of prevalence of hypertension, the yield of newly diagnosed cases, initiation of treatment, and attainment of BP control in the age group 15 to 49 years; [2] ascertain the magnitude and predictors of undiagnosed hypertension, lack of initiation of treatment, and poor control of those on antihypertensive therapy; and [3] estimate the regional variation and state-level performance of the hypertension control cascade in India. Methodology We analyzed demographic and health surveillance (DHS) data from India’s National Family Health Survey Fifth Series (NFHS-5), 2019-2021, and NFHS-4 [2015-2016]. The NFHS-5 sample comprised 695,707 women and 93,267 men in the age group of 15 to 49 years. Multiple logistic regressions were performed to find the associated predictors, and respective adjusted odds ratios (aORs) were reported. Results The prevalence of hypertension (cumulative previously diagnosed and new cases) among individuals aged 15 to 49 years was $22.8\%$ ($22.6\%$, $23.1\%$; $$n = 172$$,532), out of which $52.06\%$ were newly diagnosed cases. In contrast, in NFHS-4, the prevalence of hypertension among individuals aged 15 to 49 years was $20.4\%$ ($20.2\%$, $20.6\%$; $$n = 153$$,384), of which $41.65\%$ were newly diagnosed cases. In NFHS-5, $40.7\%$ ($39.8\%$ and $41.6\%$) of the previously diagnosed cases were on BP-lowering medications compared to $32.6\%$ ($31.8\%$, $33.6\%$) in NFHS-4. Furthermore, in NFHS-5, controlled BP was observed in $73.7\%$ ($72.7\%$ and $74.7\%$) of the patients on BP-lowering medication compared to $80.8\%$ ($80.0\%$, $81.6\%$) in NFHS-4. Females compared to males (aOR = 0·72 and 0·007), residents of rural areas (aOR = 0·82 and 0·004), and those belonging to the socially disadvantaged groups were not initiated on treatment despite awareness of their hypertension status indicative of poor treatment-seeking behavior. Furthermore, increasing age (aOR = 0·49, $P \leq 0$·001), higher body mass index (aOR = 0·51, $P \leq 0$·001), and greater waist-to-hip ratio (aOR = 0·78, $$P \leq 0$$·047) were associated with uncontrolled hypertension in patients on antihypertensive drug therapy. Conclusions Hypertension control cascade in *India is* largely ineffectual although screening yield and initiation of antihypertensive treatment have improved in NFHS-5 compared to NFHS-4. Identification of high-risk groups for opportunistic screening, implementing community-based screening, strengthening primary care, and sensitizing associated practitioners are urgently warranted. ## Introduction Hypertension is a major cause of cardiovascular disease and deaths worldwide, especially in low- and middle-income countries (LMICs). In 2019, around 1.28 billion adults aged between 30 and 79 years are estimated to be affected by hypertension worldwide, with the prevalence of hypertension being $32\%$ among women and $34\%$ among men [1,2]. As per the global burden of diseases 2019 estimates, hypertensive heart disease considering all ages and sexes accounts for $0.85\%$ of total Disability Adjusted Life Years (DALYs) globally and has shown an upward trend [3]. The aging of the population and increased exposure to lifestyle risk factors, such as unhealthy diets (high salt and low potassium intake) and lack of physical activity, contribute to the increase [4]. In India, too, the estimated number of DALYs associated with hypertension has increased from 21 million in 1990 to 39 million in 2016 (+$89\%$) [5]. The prevalence of hypertension as per the National Family Health Survey Series Four (NFHS-4) was found to be $18.1\%$ in 2015-2016 [6], while $21\%$ of females aged over 15 years had hypertension compared to $24\%$ of males of the same age range, as estimated in NFHS-5 [2019-2021] [7]. According to the rule of halves for hypertension, half the people with high blood pressure (BP) are not known (rule 1), half of those who are known are not treated (rule 2), and half of those who are treated are not controlled (rule 3) [8]. The weak care cascade of hypertension from the time of screening, diagnosis, treatment initiation, and the attainment of optimal BP control is thereby a public health challenge, particularly in resource-limited settings. Consequently, despite the availability of safe, well-tolerated, and cost-effective BP (BP)-lowering therapies, <$14\%$ of adults with hypertension have BP controlled to a systolic BP (SBP)/diastolic BP (DBP) <$\frac{140}{90}$ mmHg [1]. The secondary data analysis of NFHS-4 revealed that among patients with hypertension in India, only $63.2\%$ had their BP measured earlier, while only $21.5\%$ were aware of their diagnosis [6]. In another repeated cross-sectional survey in the National Capital Region of India, among 3,048 individuals, the prevalence of hypertension was reported to have increased over 20 years with no improvement in its management [9]. Untreated hypertension or resistant hypertension can substantially increase the chances of heart attack, stroke, and kidney failure [10]. Also, long-standing and uncontrolled hypertension is a strong risk factor for microvascular and macrovascular complications such as ischemic heart disease, stroke, chronic kidney disease, retinopathy, etc. [ 11]. Only a few studies have explored the care cascades and treatment-seeking behavior of patients with hypertension in India [9]. Furthermore, a comparison of state health performance with treatment-seeking and hypertension control has not been assessed previously in Indian health settings. These data are pertinent to inform policy and programs for hypertension control in India. Consequently, analysis of nationally representative empirical data for understanding the existing barriers and challenges in the hypertension control cascade in India and ways of strengthening the same through a focus on effective public health interventions is urgently warranted. The study objectives were to estimate in the age group of 15-49 years in India [1] the change in the rate of prevalence of hypertension, the yield of newly diagnosed cases, initiation of treatment, and attainment of BP control; [2] ascertain the magnitude and predictors of undiagnosed hypertension, lack of initiation of treatment, and poor control of those on antihypertensive therapy. Additionally, we evaluate regional estimates of the hypertension control cascade and compared them state-wise after stratifying them with a comprehensive health system performance index. ## Materials and methods Data source and study population The study was carried out on demographic and health surveillance (DHS) data from India’s NFHS-5 [2019-2021] and NFHS-4 [2015-2016] for comparative analysis. NFHS surveys provide data on India’s population and health for 707 districts, 28 states, and eight union territories. NFHS-5 is a two-stage stratified sample. Primary sampling units (PSUs) were villages in rural areas, and Census Enumeration Blocks (CEBs) in urban areas, and these PSUs were selected based on the probability proportional to size (PPS) sampling method. NFHS-5 included a sample of 788,974 participants, while NFHS-4 consisted of a sample of 770,783 participants. Men and women questionnaires collected information from candidates aged 15-54 and 15-49 years, respectively. Two sets of questionnaires (district and state module) were used for women while men had just one questionnaire (state module only) [12]. In this analysis, information was collected from a sample of men and women aged 15-49 years whose BP information was available in the biomarker dataset. We excluded men aged >49 years and pregnant women in this analysis. Measurement of BP All participants aged 15 years or more had their BP measured three times, with a five-minute gap between readings, using an Omron BP monitor (OMRON, Kyoto, Japan) [12]. Outcome variables and operational definitions Hypertensive We excluded the initial BP reading and calculated the average of the last two BP readings in the dataset. Individuals detected with average SBP >= 140 mmHg or DBP >= 90 mmHg on screening, or were previously told they had hypertension on two or more occasions (by a healthcare professional), or were taking antihypertensive medication were classified as hypertensive. New Cases New cases of hypertension were defined as individuals detected with hypertension on screening and responded no to the following two statements: [1] told they had high BP on two or more occasions by the doctor or other health professionals, and [2] currently taking prescribed medicine to lower BP. Awareness of Hypertension Individuals who responded yes to the following statement were considered as being aware of their hypertensive status: Told had high BP on two or more occasions by the doctor or other health professionals. On Hypertension Treatment Individuals who responded yes to the following statement were considered to be on hypertension treatment: Currently taking a prescribed medicine to lower BP. Controlled Hypertension Individuals who were currently taking antihypertensive medication and were detected with SBP < 140 mmHg and DBP < 90 mmHG on screening were classified as having controlled hypertension. Uncontrolled Hypertension Individuals who were currently taking antihypertensive medication and were detected with SBP >= 140 mmHg or DBP >= 90 mmHg on screening were classified as having uncontrolled hypertension. Independent variables The predictor variables were selected based on literature reviews such as age, gender, education level (no education, primary, and secondary or higher education), place of residence (urban or rural), religion (Hindu, Muslim, Christian, or Others), lifestyle factors (smoking and alcohol), and marital status. The frequency of having fried food and aerated drinks was categorized as less frequent (never or occasionally) and more frequent (weekly or daily), presence of comorbidities such as diabetes and heart disease, wealth index, etc., were also considered. The healthcare facility was categorized into three groups: all public facilities as Public, all private facilities as Private, and nongovernmental organization (NGO) along with other facilities as Other. Subgroup analysis The regional variation in the prevalence and control of hypertension in India was estimated in all the states and union territories of India, wherein the outcomes were stratified as those individuals who were aware of their hypertensive condition and were on antihypertensive medication. Individuals were considered as having good control of their BP if SBP < 140 mmHg and DBP < 90 mmHg. The states of India were categorized as per their National Health Index score for the year 2019-2020, which classified them as High (HI score > 55), medium (HI score 45 to 55), and low (HI score < 45) [13]. Since 2017, the National Institution for Transforming India (NITI) Aayog, the apex public policy thinktank of the government of India has been leading the health index program to assess the annual performance of states and union territories on several metrics such as governance, procedures, and health results, although it includes predominant maternal and child health-related indicators as a proxy for the overall health status [14]. Data and statistical analysis All the values of the variables were checked for their plausibility. Individual Men (IAMR7AFL) and Individual Women (IAIR7ADT) files were used for this analysis because the wide range of predictors incorporated in the study was not available in household files (e.g., type of healthcare facility utilized). Hence, the total sample size was 770,783 (women and men) in NFHS-4, while 788,974 (695,707 women and 93,267 men) in NFHS-5. There were several improbable values in the body mass index (BMI) of individuals, for which we replaced those with missing values, i.e., BMI values >80 and <7 were set as missing values. Appropriate weights were applied throughout the analysis for calculating the adjusted proportions and their $95\%$ confidence interval (CI). Variables with a statistically significant association in bivariate analysis were included in the regression model. Multiple logistic regression was performed to find the predictors for hypertension awareness, treatment seeking, and its control. All the assumptions and prerequisites were checked for the logistic regression analysis. Predictor variables were assessed for multicollinearity. Toward the end, the model was assessed for its fitness. The analysis was performed in STATA version 15.1 (Stata Corp., College Station, TX, USA). Ethical considerations This study is a secondary data analysis of publicly available NFHS-5 data. All the respondents who took the survey provided their voluntary written and informed consent. NFHS-5 received ethical clearance from the ethical review board of the International Institute of Population Sciences (IIPS), Mumbai, India. Permission was also obtained from IIPS to conduct this analysis. ## Results We analyzed a sample of 695,707 women and 93,267 men from the NFHS-5 data set to evaluate the prevalence of hypertensive individuals, awareness of their hypertensive status, and use of medication among aware individuals. Furthermore, we estimated the prevalence of uncontrolled hypertension among those taking the medication and the determinants of uncontrolled hypertension. The baseline characteristics of the study population segregated by previously diagnosed and newly diagnosed cases are reported in Table 1. In NFHS-5, a total of 82,718 previously diagnosed hypertension cases and 89,814 new hypertension cases were detected on screening. The prevalence of hypertension (cumulative previously diagnosed and new cases) among individuals aged 15 to 49 years was $22.8\%$ ($22.6\%$, $23.1\%$; $$n = 172$$,532), out of which $52.06\%$ were newly diagnosed cases. In contrast, in NFHS-4, a total of 83,997 previously diagnosed hypertension cases and 69,387 new hypertension cases were detected on screening. The prevalence of hypertension (cumulative previously diagnosed and new cases) among individuals aged 15 to 49 years was $20.4\%$ ($20.2\%$, $20.6\%$; $$n = 153$$,384), out of which $41.65\%$ were newly diagnosed cases. **Table 1** | Variables | NFHS-4 (2015-2016) | NFHS-4 (2015-2016).1 | NFHS-5 (2019-2021) | NFHS-5 (2019-2021).1 | | --- | --- | --- | --- | --- | | | Previously diagnosed cases (n = 83,997) | New cases (n = 69,387) | Previously diagnosed cases (n = 82,718) | New cases (n = 89,814) | | Sex | Sex | Sex | Sex | Sex | | Male | 47.09 (45.6, 48.59) | 52.91 (51.41, 54.40) | 43.59 (41.64, 45.57) | 56.41 (54.43, 58.36) | | Female | 60.36 (59.61, 61.11) | 39.64 (38.89, 40.39) | 52.80 (52.02, 53.57) | 47.20 (46.43, 47.98) | | Age group (in years) | Age group (in years) | Age group (in years) | Age group (in years) | Age group (in years) | | Adolescent (15-19) | 64.82 (63.24, 66.37) | 35.18 (33.63, 36.76) | 59.47 (57.87, 61.06) | 40.53 (38.94, 42.13) | | Young (20-39) | 60.47 (59.61, 61.32) | 39.53 (38.68, 40.39) | 52.86 (51.89, 53.83) | 47.14 (46.17, 48.11) | | Middle-aged (≥40 to 49) | 53.72 (52.88, 54.57) | 46.28 (45.43, 47.12) | 47.16 (46.38, 47.95) | 52.84 (52.05, 53.62) | | Anemia (NHFS-4, n = 138,121; NFHS-5, n = 138,006) | Anemia (NHFS-4, n = 138,121; NFHS-5, n = 138,006) | Anemia (NHFS-4, n = 138,121; NFHS-5, n = 138,006) | Anemia (NHFS-4, n = 138,121; NFHS-5, n = 138,006) | Anemia (NHFS-4, n = 138,121; NFHS-5, n = 138,006) | | Severe | 65.82 (61.53, 69.88) | 34.18 (30.12, 38.47) | 49.08 (46.30, 51.85) | 50.92 (48.15, 53.71) | | Moderate | 59.52 (57.99, 61.03) | 40.48 (38.97, 42.01) | 41.71 (40.58, 42.82) | 58.30 (57.18, 59.42) | | Mild | 56.65 (55.61, 57.68) | 43.35 (42.32, 44.39) | 39.81 (38.71, 40.92) | 60.19 (59.08, 61.29) | | Not anemic | 48.44 (47.54. 49.34) | 51.56 (50.66, 52.46) | 35.87 (35.03, 36.71) | 64.13 (63.29, 64.97) | | BMI (NFHS-4, n = 139,521; NFHS-5, n = 144,726) | BMI (NFHS-4, n = 139,521; NFHS-5, n = 144,726) | BMI (NFHS-4, n = 139,521; NFHS-5, n = 144,726) | BMI (NFHS-4, n = 139,521; NFHS-5, n = 144,726) | BMI (NFHS-4, n = 139,521; NFHS-5, n = 144,726) | | Underweight | 53.85 (52.55, 55.14) | 46.15 (44.86, 47.45) | 39.63 (38.29, 40.99) | 60.37 (59.01, 61.71) | | Normal weight | 52.02 (51.03, 53.01) | 47.98 (46.99, 48.97) | 37.39 (36.47, 38.32) | 62.61 (61.68, 63.53) | | Overweight/Obese | 53.14 (52.16, 54.12) | 46.86 (45.88, 47.84) | 41.36 (40.51, 42.21) | 58.64 (57.79, 59.49) | | Waist-to-hip ratio (NFHS-4, n = 144,392) | Waist-to-hip ratio (NFHS-4, n = 144,392) | Waist-to-hip ratio (NFHS-4, n = 144,392) | Waist-to-hip ratio (NFHS-4, n = 144,392) | Waist-to-hip ratio (NFHS-4, n = 144,392) | | <=0.9 for males and <=0.8 for females | - | - | 36.51 (35.28, 37.76) | 63.49 (62.24, 64.72) | | Cutoffs > 0.9 for males and >0.8 for females | - | - | 39.58 (38.81, 40.38) | 60.42 (59.62, 61.20) | | Marital status | Marital status | Marital status | Marital status | Marital status | | Never in union | 61.93 (60.53, 63.31) | 38.07 (36.69, 39.47) | 58.31 (56.89, 59.68) | 41.70 (40.32, 43.11) | | Currently married | 57.94 (57.18, 58.71) | 42.06 (41.30, 42.82) | 50.20 (49.42, 50.98) | 49.81 (49.02, 50.58) | | Widowed/Divorced/Separated | 54.54 (52.73, 56.33) | 45.46 (43.67, 47.27) | 45.71 (44.18, 47.24) | 54.29 (52.76, 55.82) | | Education | Education | Education | Education | Education | | No education/Preprimary education | 52.39 (51.47, 53.31) | 47.61 (46.70, 48.53) | 47.38 (46.48, 48.27) | 52.62 (51.73, 53.52) | | Primary | 53.98 (52.78, 55.18) | 46.02 (44.82, 47.22) | 46.75 (45.51, 47.99) | 53.25 (52.01, 54.49) | | Secondary education | 60.69 (59.77, 61.61) | 39.31 (38.39, 40.23) | 51.65 (50.71, 52.61) | 48.35 (47.39, 49.31) | | Higher education | 66.65 (65.11, 68.16) | 33.35 (31.84, 34.89) | 61.19 (59.58, 62.78) | 38.81 (37.22, 40.42) | | Residence | Residence | Residence | Residence | Residence | | Urban | 63.36 (62.01, 64.68) | 36.64 (35.32, 37.99) | 56.75 (55.29, 58.21) | 43.25 (41.81, 44.71) | | Rural | 54.72 (53.90, 55.53) | 45.28 (44.47, 46.10) | 48.32 (47.43, 49.21) | 51.68 (50.79, 52.57) | | Cast/Tribe (NFHS-4, n = 144,521; NFHS-5, n = 161,782) | Cast/Tribe (NFHS-4, n = 144,521; NFHS-5, n = 161,782) | Cast/Tribe (NFHS-4, n = 144,521; NFHS-5, n = 161,782) | Cast/Tribe (NFHS-4, n = 144,521; NFHS-5, n = 161,782) | Cast/Tribe (NFHS-4, n = 144,521; NFHS-5, n = 161,782) | | Scheduled caste | 59.06 (57.61, 60.49) | 40.94 (39.51, 42.39) | 51.05 (49.77, 52.34) | 48.95 (47.66, 50.23) | | Scheduled tribe | 44.37 (42.68, 46.08) | 55.63 (53.92, 57.32) | 40.02 (38.52, 41.54) | 59.98 (58.46, 61.48) | | Other backward caste | 60.24 (59.21, 61.27) | 39.76 (38.73, 40.79) | 53.52 (52.60, 54.43) | 46.48 (45.57, 47.41) | | Others | 59.93 (58.83, 61.03) | 40.07 (38.97, 41.17) | 54.58 (53.16, 56.00) | 45.42 (44.00, 46.84) | | Religion | Religion | Religion | Religion | Religion | | Hindu | 58.38 (57.54, 59.21) | 41.62 (40.79, 42.46) | 50.78 (49.96, 51.60) | 49.22 (48.41, 50.04) | | Muslim | 59.16 (57.74, 60.55) | 40.84 (39.45, 42.26) | 52.33 (50.22, 54.44) | 47.67 (45.56, 49.78) | | Others | 56.26 (54.41, 58.1) | 43.74 (41.90, 45.59) | 57.88 (55.70, 60.02) | 42.12 (39.98, 44.31) | | Wealth index | Wealth index | Wealth index | Wealth index | Wealth index | | Poorest | 46.82 (45.64, 48.00) | 53.18 (52.00, 54.36) | 46.91 (45.40, 48.41) | 53.11 (51.59, 54.61) | | Poorer | 53.83 (52.69, 54.97) | 46.17 (45.03, 47.31) | 46.53 (45.33, 47.73) | 53.47 (52.27, 54.67) | | Middle | 58.37 (57.22, 59.51) | 41.63 (40.49, 42.78) | 48.08 (46.95, 49.22) | 51.92 (50.78, 53.05) | | Richer | 60.57 (59.35, 61.78) | 39.43 (38.22, 40.65) | 51.52 (50.33, 52.71) | 48.48 (47.31, 49.67) | | Richest | 64.64 (63.28, 65.97) | 35.36 (34.03, 36.72) | 61.18 (59.76, 62.57) | 38.82 (37.43, 40.24) | | Type of healthcare facility accessed (NFHS-4, n = 44,611; NFHS-5, n = 58,407) | Type of healthcare facility accessed (NFHS-4, n = 44,611; NFHS-5, n = 58,407) | Type of healthcare facility accessed (NFHS-4, n = 44,611; NFHS-5, n = 58,407) | Type of healthcare facility accessed (NFHS-4, n = 44,611; NFHS-5, n = 58,407) | Type of healthcare facility accessed (NFHS-4, n = 44,611; NFHS-5, n = 58,407) | | Public facility | 60.33 (59.06, 61.58) | 39.67 (38.42, 40.94) | 48.69 (47.55, 49.83) | 51.31 (50.17, 52.45) | | Private facility | 61.46 (60.15, 62.75) | 38.54 (37.25, 39.85) | 54.84 (52.41, 56.25) | 45.16 (43.75, 46.59) | | NGO/Other | 54.82 (48.97, 60.54) | 45.18 (39.46, 51.03) | 48.17 (42.07, 54.32) | 51.83 (45.68, 57.93) | | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | | No | 57.67 (56.89, 58.44) | 42.33 (41.56, 43.11) | 52.87 (51.95, 53.79) | 47.13 (46.21, 48.05) | | Yes | 60.52 (59.18, 61.85) | 39.48 (38.15, 40.82) | 48.09 (47.11, 49.08) | 51.91 (50.92, 52.89) | | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | | Diabetes (NFHS-4, n = 150,618; NFHS-5, n = 170,399) | 64.54 (62.02, 66.97) | 35.46 (33.03, 37.98) | 60.18 (58.16, 62.17) | 39.82 (37.83, 41.84) | | Heart disease (NFHS-4,n = 152,395; NFHS-5, n = 171,603) | 69.95 (67.71, 72.09) | 30.05 (27.91, 32.29) | 63.33 (59.90, 66.63) | 36.67 (33.37, 40.10) | | Any tobacco use | Any tobacco use | Any tobacco use | Any tobacco use | Any tobacco use | | No | 60.41 (59.65, 61.18) | 39.59 (38.82, 40.35) | 52.62 (51.81, 53.42) | 47.38 (46.58, 48.19) | | Yes | 45.23 (44, 46.47) | 54.77 (53.53, 56) | 41.00 (39.43, 42.59) | 59.01 (57.41, 60.57) | | Alcohol usage current | Alcohol usage current | Alcohol usage current | Alcohol usage current | Alcohol usage current | | No | 59.46 (58.72, 60.19) | 40.54 (39.81, 41.28) | 52.17 (51.36, 52.97) | 47.83 (47.03, 48.64) | | Yes | 43.23 (41.09, 45.4) | 56.77 (54.6, 58.91) | 37.31 (34.86, 39.81) | 62.70 (60.20, 65.14) | | Fried food | Fried food | Fried food | Fried food | Fried food | | Less frequent | 59.12 (58.25, 59.98) | 40.88 (40.02, 41,75) | 52.01 (51.16, 52.86) | 47.99 (47.14, 48.84) | | More frequent | 57.47 (56.57, 58.35) | 42.53 (41.65, 43.43) | 50.76 (49.70, 51.83) | 49.24 (48.17, 50.30) | | Aerated drinks | Aerated drinks | Aerated drinks | Aerated drinks | Aerated drinks | | Less frequent | 56.74 (56, 57.48) | 43.26 (42.52, 44) | 50.78 (49.97, 51.58) | 49.22 (48.42, 50.03) | | More frequent | 62.71 (61.48, 63.92) | 37.31 (36.08, 38.52) | 54.72 (53.27, 56.16) | 45.28 (43.84, 46.73) | | Region | Region | Region | Region | Region | | North | 65.31 (63.82, 66.75) | 34.71 (33.25, 36.18) | 59.28 (58.21, 60.35) | 40.72 (39.65, 41.79) | | Central | 52.51 (51.41, 53.6) | 47.49 (46.4, 48.59) | 57.19 (56.03, 58.33) | 42.81 (41.67, 43.97) | | East | 55.11 (53.68, 56.53) | 44.90 (43.47, 46.32) | 50.22 (47.99, 52.46) | 49.78 (47.54, 52.01) | | Northeast | 46.28 (44.8, 47.77) | 53.72 (52.23, 55.2) | 42.34 (40.65, 44.06) | 57.66 (55.94, 59.35) | | West | 50.67 (48.74, 52.59) | 49.33 (47.41, 51.26) | 45.71 (43.04, 48.41) | 54.29 (51.59, 56.96) | | South | 65.83 (64.09, 67.53) | 34.17 (32.47, 35.91) | 47.55 (45.95, 49.14) | 52.45 (50.86, 54.05) | In NFHS-5, among the hypertension cases, a majority of males ($56.41\%$), middle-aged ($52.84\%$), rural residents ($51.68\%$), tobacco users ($59\%$), and alcohol users ($62.7\%$) were new cases that were previously undiagnosed. A majority of patients in the lower wealth quintiles ($53.1\%$) and having lower educational status (53.25 % with just primary education) were newly diagnosed cases on screening compared to those from higher wealth quintiles and higher educational status, respectively. The region-wide distribution of the old and new cases was fairly similar across all categories except the northeastern region where new cases were the highest ($57.66\%$). The sociodemographic, anthropometric, and lifestyle characteristics associated with hypertension awareness status are reported in Table 2. Middle-aged individuals (adjusted odds ratios [aOR] = 2.51 and $P \leq 0.001$) and overweight or obese participants (aOR = 1.98 and $P \leq 0.001$) were found to be more aware of their hypertension status. Although the residents of rural areas, individuals possessing health insurance, and those not drinking alcohol have higher odds of being aware of their hypertension status; however, these associations were not statistically significant. **Table 2** | Variables | Previously not told they had high BP (n = 34,972) | Previously told they had high BP on two or more occasions (n = 47,746) | Unadjusted odds | P-value | Adjusted odds | P-value.1 | | --- | --- | --- | --- | --- | --- | --- | | Sex | Sex | Sex | Sex | Sex | Sex | Sex | | Male | 69.11 (66.66, 71.45) | 30.89 (28.55, 33.34) | Ref | | Ref | | | Female | 44.73 (43.69, 54.77) | 55.27 (54.23, 56.31) | 2.76 (2.47, 3.08) | <0.001 | 1.31 (0.95, 1.81) | 0.09 | | Age group (in years) | Age group (in years) | Age group (in years) | Age group (in years) | Age group (in years) | Age group (in years) | Age group (in years) | | Adolescent (15-19) | 81.40 (80.17, 82.56) | 18.61 (17.44, 19.83) | Ref | | Ref | | | Young (20-39) | 50.56 (49.29, 51.84) | 49.44 (48.16, 50.71) | 4.27 (3.96, 4.61) | | 1.30 (0.98, 1.74) | | | Middle-aged (≥40 to 49) | 31.01 (29.78, 32.27) | 68.99 (67.73, 70.22) | 9.73 (8.94, 10.58) | <0.001 | 2.51 (1.80, 3.51) | <0.001 | | Anemia (n = 51,680) | Anemia (n = 51,680) | Anemia (n = 51,680) | Anemia (n = 51,680) | Anemia (n = 51,680) | Anemia (n = 51,680) | Anemia (n = 51,680) | | Severe | 11.18 (9.04, 13.76) | 88.82 (86.24, 90.96) | Ref | | - | | | Moderate | 12.30 (11.48, 13.16) | 87.71 (86.84, 88.52) | 0.89 (0.70, 1.14) | | | | | Mild | 12.28 (11.44, 13.19) | 87.72 (86.81, 88.56) | 0.89 (0.70, 1.15) | | | | | Not anemic | 11.53 (10.88, 12.21) | 88.47 (87.79, 98.12) | 0.96 (0.75, 1.23) | 0.29 | | | | BMI (n = 55,169) | BMI (n = 55,169) | BMI (n = 55,169) | BMI (n = 55,169) | BMI (n = 55,169) | BMI (n = 55,169) | BMI (n = 55,169) | | Underweight | 23.82 (22.31, 25.4) | 76.18 (74.61, 77.69) | Ref | | Ref | | | Normal weight | 16.06 (15.31, 16.84) | 83.94 (83.16, 84.70) | 1.63 (1.48, 1.79) | | 1.24 (1.03, 1.50) | | | Overweight/Obese | 8.15 (7.60, 8.73) | 91.85 (91.27, 92.40) | 3.52 (3.12, 3.93) | <0.001 | 1.98 (1.57, 2.50) | <0.001 | | Waist-to-hip ratio (n = 54,874) | Waist-to-hip ratio (n = 54,874) | Waist-to-hip ratio (n = 54,874) | Waist-to-hip ratio (n = 54,874) | Waist-to-hip ratio (n = 54,874) | Waist-to-hip ratio (n = 54,874) | Waist-to-hip ratio (n = 54,874) | | <=0.9 for males and <=0.8 for females | 18.02 (16.83, 19.28) | 81.98 (80.72, 83.17) | Ref | | Ref | | | Cutoffs > 0.9 for males and >0.8 for females | 12.18 (11.67, 12.71) | 87.82 (87.29, 88.33) | 1.58 (1.45, 1.73) | <0.001 | 1.03 (0.87, 1.22) | 0.65 | | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | | Never in union | 80.83 (79.67, 81.94) | 19.17 (18.06, 20.33) | Ref | | Ref | | | Currently married | 39.31 (38.26, 40.35) | 60.71 (59.65, 61.74) | 6.50 (6.10, 6.94) | | 2.06 (1.59, 2.66) | | | Widowed/Divorced/Separated | 32.74 (30.47, 35.09) | 67.26 (64.91, 69.53) | 8.66 (7.69, 9.74) | <0.001 | 2.35 (1.57, 3.53) | <0.001 | | Education | Education | Education | Education | Education | Education | Education | | No education/Preprimary education | 36.76 (35.48, 38.05) | 63.24 (61.95, 64.52) | Ref | | Ref | | | Primary | 38.24 (36.59, 39.92) | 61.76 (60.08, 63.41) | 0.93 (0.86, 1.01) | | 0.93 (0.74, 1.17) | | | Secondary education | 49.63 (48.45, 50.81) | 50.37 (49.19, 51.55) | 0.58 (0.55, 0.62) | | 0.98 (0.82, 1.17) | | | Higher education | 62.06 (59.88, 64.21) | 37.94 (54.80, 40.12) | 0.35 (0.32, 0.39) | <0.001 | 1.18 (0.91, 1.53) | 0.34 | | Residence | Residence | Residence | Residence | Residence | Residence | Residence | | Urban | 56.62 (54.66, 58.55) | 43.38 (41.45, 45.34) | Ref | | Ref | | | Rural | 41.48 (40.51, 42.46) | 58.52 (57.54, 59.49) | 1.84 (1.68, 2.01) | <0.001 | 0.91 (0.75, 1.11) | 0.40 | | Cast/Tribe (n = 78,206) | Cast/Tribe (n = 78,206) | Cast/Tribe (n = 78,206) | Cast/Tribe (n = 78,206) | Cast/Tribe (n = 78,206) | Cast/Tribe (n = 78,206) | Cast/Tribe (n = 78,206) | | Scheduled caste | 43.93 (42.32, 45.55) | 56.07 (54.45, 57.68) | 1.28 (1.16, 1.41) | | 1.11 (0.88, 1.39) | | | Scheduled tribe | 51.28 (48.82, 53.76) | 48.72 (46.24, 51.22) | 0.95 (0.84, 1.08) | | 1.06 (0.84, 1.41) | | | OBC | 46.62 (45.25, 47.99) | 52.38 (52.01, 54.75) | 1.15 (1.05, 1.25) | | 1.10 (0.91, 1.33) | | | Others | 50.15 (48.16, 52.15) | 49.85 (57.85, 51.84) | Ref | <0.001 | Ref | 0.75 | | Religion | Religion | Religion | Religion | Religion | Religion | Religion | | Hindu | 46.29 (45.14, 47.44) | 53.71 (52.56, 54.86) | Ref | | Ref | | | Muslim | 53.76 (51.50, 56.01) | 46.24 (43.99, 48.51) | 0.74 (0.67, 0.81) | | 0.95 (0.77, 1.17) | | | Others | 49.41 (45.79, 53.01) | 50.61 (46.99, 54.21) | 0.88 (0.76, 1.01) | <0.001 | 0.94 (0.69, 1.27) | 0.83 | | Wealth index | Wealth index | Wealth index | Wealth index | Wealth index | Wealth index | Wealth index | | Poorest | 46.76 (45.14, 48.39) | 53.24 (51.61, 54.86) | Ref | | Ref | | | Poorer | 42.85 (41.33, 44.38) | 57.15 (55.62, 58.67) | 1.17 (1.08, 1.26) | | 1.29 (1.04, 1.60) | | | Middle | 41.24 (39.68, 42.81) | 58.76 (57.19, 60.32) | 1.25 (1.14, 1.36) | | 1.32 (1.05, 1.66) | | | Richer | 45.83 (44.11, 47.56) | 54.17 (52.44, 55.90) | 1.03 (0.94, 1.13) | | 1.38 (1.04, 1.82) | | | Richest | 56.78 (54.74, 58.80) | 43.22 (41.21, 45.26) | 0.67 (0.60, 0.74) | <0.001 | 1.32 (0.96, 1.83) | 0.09 | | Type of healthcare facility accessed (n = 28,715) | Type of healthcare facility accessed (n = 28,715) | Type of healthcare facility accessed (n = 28,715) | Type of healthcare facility accessed (n = 28,715) | Type of healthcare facility accessed (n = 28,715) | Type of healthcare facility accessed (n = 28,715) | Type of healthcare facility accessed (n = 28,715) | | Public facility | 35.94 (34.61, 37.32) | 64.06 (62.71, 65.49) | Ref | | Ref | | | Private facility | 41.63 (39.56, 43.73) | 58.37 (56.27, 60.44) | 0.78 (0.71, 0.86) | | 1.05 (0.91, 1.22) | | | NGO/Other | 33.95 (26.27, 42.55) | 66.06 (57.45, 73.73) | 1.09 (0.75, 1.57) | <0.001 | 2.26 (1.11, 4.58) | 0.06 | | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | | No | 49.43 (48.24, 50.62) | 50.57 (49.38, 51.76) | Ref | | Ref | | | Yes | 43.17 (41.64, 44.71) | 56.83 (55.29, 58.36) | 1.28 (1.20, 1.37) | <0.001 | 0.88 (0.76, 1.02) | 0.11 | | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | | Diabetes (n = 81,845) | Diabetes (n = 81,845) | Diabetes (n = 81,845) | Diabetes (n = 81,845) | Diabetes (n = 81,845) | Diabetes (n = 81,845) | Diabetes (n = 81,845) | | No | 49.17 (48.08, 50.27) | 50.83 (49.73, 51.92) | Ref | | Ref | | | Yes | 17.04 (14.89, 19.44) | 82.96 (80.56, 85.11) | 4.7 (4.01, 5.51) | <0.001 | 1.34 (0.93, 1.93) | 0.11 | | Heart disease (n = 82,372) | Heart disease (n = 82,372) | Heart disease (n = 82,372) | Heart disease (n = 82,372) | Heart disease (n = 82,372) | Heart disease (n = 82,372) | Heart disease (n = 82,372) | | No | 48.12 (47.04, 49.21) | 51.88 (50.80, 52.96) | Ref | | Ref | | | Yes | 20.59 (16.88, 24.87) | 79.41 (75.13, 83.12) | 3.57 (2.81, 4.55) | <0.001 | 1.25 (0.69, 2.24) | 0.44 | | Any tobacco use | Any tobacco use | Any tobacco use | Any tobacco use | Any tobacco use | Any tobacco use | Any tobacco use | | No | 47.55 (46.46, 48.65) | 52.45 (51.35, 53.54) | 1.08 (0.98, 1.19) | 0.10 | - | | | Yes | 49.53 (47.07, 51.99) | 50.47 (48.01, 52.93) | Ref | | | | | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | | No | 47.36 (46.28, 48.44) | 52.64 (51.56, 53.72) | Ref | | Ref | | | Yes | 57.60 (53.38, 61.72) | 42.40 (38.28, 46.62) | 0.66 (0.55, 0.78) | <0.001 | 1.02 (0.65, 1.60) | 0.90 | | Fried food | Fried food | Fried food | Fried food | Fried food | Fried food | Fried food | | Less frequent | 47.40 (46.20, 48.61) | 52.60 (51.39, 53.80) | Ref | | - | | | More frequent | 48.12 (46.74, 49.50) | 51.88 (50.50, 53.26) | 0.97 (0.91, 1.02) | 0.317 | | | | Aerated drinks | Aerated drinks | Aerated drinks | Aerated drinks | Aerated drinks | Aerated drinks | Aerated drinks | | Less frequent | 45.68 (44.66, 46.71) | 54.32 (53.32, 55.34) | Ref | | Ref | | | More frequent | 56.64 (54.49, 58.76) | 43.36 (41.24, 45.51) | 0.64 (0.59, 0.69) | <0.001 | 0.79 (0.66, 0.94) | 0.009 | | Region | Region | Region | Region | Region | Region | Region | | North | 44.10 (42.57, 45.63) | 55.91 (54.37, 57.43) | Ref | | Ref | | | Central | 49.69 (47.86, 51.53) | 50.31 (48.47, 52.14) | 0.79 (0.72, 0.87) | | 0.97 (0.77, 1.21) | | | East | 38.17 (36.37, 40.00) | 61.83 (60.00, 63.63) | 1.27 (1.15, 1.41) | | 1.01 (0.79, 1.28) | | | Northeast | 37.01 (34.67, 39.41) | 62.99 (60.59, 65.33) | 1.34 (1.19, 1.51) | | 0.53 (0.40, 0.70) | | | West | 65.71 (62.11, 69.13) | 34.31 (30.87, 37.91) | 0.41 (0.34, 0.48) | | 0.40 (0.30, 0.53) | | | South | 48.59 (46.14, 51.04) | 51.41 (48.96, 53.86) | 0.83 (0.74, 0.93) | <0.001 | 0.77 (0.64, 0.98) | <0.001 | In NFHS-5, $40.7\%$ ($39.8\%$, $41.6\%$) of the previously diagnosed cases were initiated on BP-lowering medications. However, in NFHS-4, about $32.6\%$ ($31.8\%$, $33.6\%$) of the previously diagnosed cases were on BP-lowering medications. The determinants of positive treatment-seeking behavior considered in patients on BP-lowering medications are reported in Table 3. A significantly higher proportion of middle-aged ($53.47\%$), those who were separated from their partners ($53.49\%$), and those who belonged to the richest wealth quintiles ($45.08\%$) were on BP-lowering medications. Contrary to this, females compared to males (aOR = 0.72 and $$P \leq 0.007$$), residents of rural areas (aOR = 0.82 and $$P \leq 0.004$$), and those belonging to the socially disadvantaged groups - scheduled caste [SC]/scheduled caste [ST]/another backward caste [OBC] - were less likely to be on BP-lowering medications. **Table 3** | Variables | Not taking medicine (n = 32,223) | Taking medicine (n = 20,697) | Unadjusted odds | P-value | Adjusted odds | P-value.1 | | --- | --- | --- | --- | --- | --- | --- | | Sex | Sex | Sex | Sex | Sex | Sex | Sex | | Male | 50.98 (47.72, 54.24) | 49.02 (45.76, 52.28) | Ref | | Ref | | | Female | 59.95 (59.06, 60.84) | 40.05 (39.16, 40.94) | 0.69 (0.61, 0.79) | <0.001 | 0.72 (0.57, 0.91) | 0.007 | | Age group (in years) | Age group (in years) | Age group (in years) | Age group (in years) | Age group (in years) | Age group (in years) | Age group (in years) | | Adolescent (15-19) | 58.01 (55.36, 60.62) | 41.99 (39.38, 44.64) | Ref | | Ref | | | Young (20-39) | 69.80 (68.89, 70.79) | 30.20 (29.21, 31.20) | 0.59 (0.53, 0.66) | | 0.91 (0.70, 1.18) | | | Middle-aged (≥40 to 49) | 46.53 (45.42, 47.64) | 53.47 (52.36, 54.58) | 1.58 (1.42, 1.77) | <0.001 | 2.15 (1.64, 2.83) | <0.001 | | Anemia (n = 50,424) | Anemia (n = 50,424) | Anemia (n = 50,424) | Anemia (n = 50,424) | Anemia (n = 50,424) | Anemia (n = 50,424) | Anemia (n = 50,424) | | Severe | 63.35 (59.51, 67.03) | 36.65 (32.97, 40.49) | Ref | | - | | | Moderate | 59.20 (57.83, 60.55) | 40.80 (39.45, 42.17) | 1.19 (1.01, 1.41) | | | | | Mild | 60.61 (59.18, 62.02) | 39.39 (37.98, 40.82) | 1.12 (0.94, 1.32) | | | | | Not anemic | 60.30 (59.12, 61.46) | 39.70 (38.54, 40.88) | 1.13 (0.96, 1.34) | 0.12 | | | | BMI (n = 52,778) | BMI (n = 52,778) | BMI (n = 52,778) | BMI (n = 52,778) | BMI (n = 52,778) | BMI (n = 52,778) | BMI (n = 52,778) | | Underweight | 65.44 (63.61, 67.23) | 34.56 (32.77, 36.4) | Ref | | Ref | | | Normal weight | 64.15 (63.05, 65.25) | 35.85 (34.75, 36.95) | 1.05 (0.97, 1.15) | | 0.95 (0.82, 1.11) | | | Overweight/obese | 51.91 (50.73, 53.09) | 48.09 (46.91, 49.27) | 1.75 (1.60, 1.91) | <0.001 | 1.39 (1.18, 1.63) | <0.001 | | Waist-to-hip ratio (n = 52,755) | Waist-to-hip ratio (n = 52,755) | Waist-to-hip ratio (n = 52,755) | Waist-to-hip ratio (n = 52,755) | Waist-to-hip ratio (n = 52,755) | Waist-to-hip ratio (n = 52,755) | Waist-to-hip ratio (n = 52,755) | | <=0.9 for males and <=0.8 for females | 63.01 (61.31, 64.67) | 36.99 (35.33, 38.69) | Ref | | Ref | | | Cutoffs > 0.9 for males and >0.8 for females | 58.51 (57.56, 59.45) | 41.51 (40.55, 42.44) | 1.21 (1.12, 1.29) | <0.001 | 1.01 (0.89, 1.15) | 0.77 | | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | | Never in union | 59.95 (57.79, 62.08) | 40.05 (37.92, 42.21) | Ref | | Ref | | | Currently married | 60.09 (59.16, 61) | 39.91 (39.00, 40.84) | 0.99 (0.91, 1.08) | | 0.78 (0.63, 0.98) | | | Widowed/Divorced/Separated | 46.51 (43.98, 49.06) | 53.49 (50.94, 56.02) | 1.72 (1.51, 1.95) | <0.001 | 1.12 (0.85, 1.49) | <0.001 | | Education | Education | Education | Education | Education | Education | Education | | No education/Preprimary education | 59.15 (57.80, 60.49) | 40.85 (39.51, 42.20) | Ref | | Ref | | | Primary | 56.29 (54.54, 58.03) | 43.71 (41.97, 45.46) | 1.12 (1.03, 1.22) | | 1.09 (0.94, 1.26) | | | Secondary education | 59.06 (57.95, 60.17) | 40.94 (39.83, 42.05) | 1.00 (0.94, 1.06) | | 1.02 (0.91, 1.15) | | | Higher education | 63.33 (61.27, 65.35) | 36.67 (34.65, 38.73) | 0.83 (0.75, 0.92) | <0.001 | 0.80 (0.66, 0.98) | 0.02 | | Residence | Residence | Residence | Residence | Residence | Residence | Residence | | Urban | 53.49 (51.68, 55.28) | 46.51 (44.72, 48.32) | Ref | | Ref | | | Rural | 62.21 (61.21, 63.22) | 37.79 (36.81, 38.79) | 0.69 (0.64, 0.75) | <0.001 | 0.82 (0.72, 0.94) | 0.004 | | Cast/Tribe (n = 50,035) | | | | | | | | SC | 63.61 (62.01, 65.17) | 36.39 (34.83, 37.99) | 0.74 (0.68, 0.81) | | 0.79 (0.68, 0.92) | | | ST | 57.27 (54.78, 59.72) | 42.73 (40.28, 45.22) | 0.97 (0.86, 1.09) | | 1.02 (0.84, 1.24) | | | OBC | 60.94 (59.72, 62.17) | 39.06 (37.83, 40.30) | 0.83 (0.77, 0.91) | <0.001 | 0.90 (0.80, 1.02) | 0.01 | | Non-SC/non-ST/non-OBC | 56.71 (55.14, 58.26) | 43.29 (41.74, 44.86) | Ref | | Ref | | | Religion | Religion | Religion | Religion | Religion | Religion | Religion | | Hindu | 60.18 (59.21, 61.15) | 39.82 (38.85, 40.8) | Ref | | Ref | | | Muslim | 53.69 (51.35, 56.01) | 46.31 (43.99, 48.65) | 1.30 (1.18, 1.43) | | 1.12 (0.96, 1.30) | | | Others | 60.66 (57.94, 63.31) | 39.34 (36.69, 42.06) | 0.98 (0.87, 1.10) | <0.001 | 1.01 (0.84, 1.21) | 0.31 | | Wealth index | Wealth index | Wealth index | Wealth index | Wealth index | Wealth index | Wealth index | | Poorest | 64.50 (62.74, 66.23) | 35.51 (33.77, 37.26) | Ref | | Ref | | | Poorer | 64.01 (62.52, 65.47) | 35.99 (34.53, 37.48) | 1.02 (0.93, 1.11) | | 1.03 (0.88, 1.20) | | | Middle | 59.13 (57.56, 60.67) | 40.87 (39.33, 42.44) | 1.25 (1.14, 1.37) | | 1.23 (1.04, 1.46) | | | Richer | 56.18 (54.62, 57.75) | 43.82 (42.25, 45.41) | 1.41 (1.28, 1.56) | | 1.21 (1.02, 1.44) | | | Richest | 54.92 (53.25, 56.57) | 45.08 (43.43, 46.75) | 1.49 (1.34, 1.64) | <0.001 | 1.14 (0.93, 1.39) | 0.03 | | Type of healthcare facility accessed (n = 20,663) | Type of healthcare facility accessed (n = 20,663) | Type of healthcare facility accessed (n = 20,663) | Type of healthcare facility accessed (n = 20,663) | Type of healthcare facility accessed (n = 20,663) | Type of healthcare facility accessed (n = 20,663) | Type of healthcare facility accessed (n = 20,663) | | Public facility | 60.16 (58.64, 61.67) | 39.84 (38.33, 41.36) | Ref | | Ref | | | Private facility | 58.63 (56.89, 60.34) | 41.37 (39.66, 43.11) | 1.06 (0.97, 1.16) | | 0.98 (0.89, 1.07) | | | NGO/Other | 47.00 (38.11, 56.09) | 53.00 (43.91, 61.91) | 1.70 (1.18, 2.45) | 0.008 | 1.63 (1.10, 2.42) | 0.04 | | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | | Yes | 52.43 (51.07, 53.79) | 47.57 (46.21, 48.93) | 1.49 (1.40, 1.59) | <0.001 | 1.55 (1.40, 1.71) | <0.001 | | No | 62.21 (61.18, 63.23) | 37.79 (36.77, 38.82) | Ref | | Ref | | | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | | Diabetes (n = 52,214) | Diabetes (n = 52,214) | Diabetes (n = 52,214) | Diabetes (n = 52,214) | Diabetes (n = 52,214) | Diabetes (n = 52,214) | Diabetes (n = 52,214) | | No | 61.15 (60.23, 62.05) | 38.85 (37.95, 39.77) | Ref | | Ref | | | Yes | 32.71 (30.29, 35.23) | 67.29 (64.77, 69.71) | 3.23 (2.88, 3.63) | <0.001 | 2.22 (1.85, 2.65) | <0.001 | | Heart disease (n = 52,648) | Heart disease (n = 52,648) | Heart disease (n = 52,648) | Heart disease (n = 52,648) | Heart disease (n = 52,648) | Heart disease (n = 52,648) | Heart disease (n = 52,648) | | No | 59.69 (58.79, 60.59) | 40.31 (39.41, 41.21) | Ref | | Ref | | | Yes | 42.01 (37.54, 46.61) | 57.99 (53.39, 62.46) | 2.04 (1.69, 2.46) | <0.001 | 1.44 (1.06, 1.97) | 0.01 | | Usage of any tobacco | Usage of any tobacco | Usage of any tobacco | Usage of any tobacco | Usage of any tobacco | Usage of any tobacco | Usage of any tobacco | | No | 59.71 (58.79, 60.62) | 40.29 (39.38, 41.21) | 0.79 (0.71, 0.88) | <0.001 | 0.85 (0.70, 1.03) | 0.10 | | Yes | 54.11 (51.53, 56.70) | 45.89 (43.30, 48.50) | Ref | | Ref | | | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | | No | 59.57 (58.67, 60.47) | 40.43 (39.53, 41.33) | Ref | | Ref | | | Yes | 49.31 (44.67, 53.93) | 50.70 (46.07, 55.33) | 1.51 (1.25, 1.82) | <0.001 | 1.08 (0.79, 1.47) | 0.61 | | Fried food | Fried food | Fried food | Fried food | Fried food | Fried food | Fried food | | Less frequently | 59.44 (58.38, 60.51) | 40.56 (39.51, 41.62) | Ref | | - | | | More frequently | 59.07 (57.84, 60.28) | 40.93 (39.72, 42.16) | 1.01 (0.95, 1.07) | 0.59 | | | | Aerated drinks | Aerated drinks | Aerated drinks | Aerated drinks | Aerated drinks | Aerated drinks | Aerated drinks | | Less frequently | 59.71 (58.77, 60.64) | 40.29 (39.36, 41.23) | Ref | | Ref | | | More frequently | 57.00 (55.11, 58.87) | 43.00 (41.13, 44.89) | 1.11 (1.03, 1.20) | 0.006 | 1.18 (1.03, 1.34) | 0.01 | | Region | Region | Region | Region | Region | Region | Region | | North | 66.34 (64.94, 67.72) | 33.66 (32.28, 35.06) | Ref | | Ref | | | Central | 73.26 (71.86, 74.62) | 26.74 (25.38, 28.14) | 0.71 (0.65, 0.79) | | 0.71 (0.60, 0.84) | | | East | 60.47 (58.55, 62.36) | 39.53 (37.64, 41.45) | 1.28 (1.16, 1.42) | | 1.35 (1.12, 1.62) | | | Northeast | 47.16 (44.87, 49.47) | 52.84 (50.53, 55.13) | 2.21 (1.97, 2.46) | | 2.23 (1.79, 2.77) | | | West | 40.28 (37.56, 43.06) | 59.72 (56.94, 62.44) | 2.92 (2.56, 3.33) | | 2.62 (2.02, 3.39) | | | South | 46.09 (43.61, 48.60) | 53.91 (51.42, 56.42) | 2.30 (2.04, 2.59) | <0.001 | 2.39 (2.01, 2.85) | <0.001 | In NFHS-5, controlled BP was observed in $73.7\%$ ($72.7\%$, $74.7\%$) of patients on BP-lowering medication, while $80.8\%$ ($80.0\%$, $81.6\%$) of patients taking hypertension medication had controlled BP in NFHS-4. Among patients on antihypertensive medication, those reporting consuming alcohol, tobacco smoking, frequent consumption of fried food, presence of diabetes comorbidity, and lacking higher education had significantly lower odds of BP control compared to their counterparts. Factors such as increasing age (aOR = 0.49 and $P \leq 0.001$), higher BMI (aOR = 0.51, $P \leq 0.001$ for obese/overweight), and greater waist-to-hip ratio (aOR = 0.78 and $$P \leq 0.047$$) were also associated with poor control of hypertension despite medication therapy. Only females (aOR = 1.7 and $$P \leq 0.003$$) and individuals with higher education levels (aOR = 1.5 and $P \leq 0.004$) when on drug treatment were associated with higher odds of achieving control over their BP levels (Table 4). **Table 4** | Variables | Uncontrolled hypertension (n = 5,580) | Controlled hypertension (n = 15,117) | Unadjusted odds | P-value | Adjusted odds | P-value.1 | | --- | --- | --- | --- | --- | --- | --- | | Sex | Sex | Sex | Sex | Sex | Sex | Sex | | Male | 34.28 (29.89, 38.96) | 65.72 (61.04, 70.11) | Ref | | Ref | | | Female | 25.47 (24.53, 26.44) | 74.53 (73.56, 75.47) | 1.52 (1.24, 1.87) | <0.001 | 1.70 (1.21, 2.41) | 0.003 | | Age group (in years) | Age group (in years) | Age group (in years) | Age group (in years) | Age group (in years) | Age group (in years) | Age group (in years) | | Adolescent (15-19) | 8.21 (6.12, 10.94) | 91.79 (89.06, 93.88) | Ref | | Ref | | | Young (20-39) | 19.07 (17.67, 20.56) | 80.93 (79.44, 82.33) | 0.37 (0.27, 0.52) | | 0.88 (0.49, 1.56) | | | Middle-aged (≥40 to 49) | 33.04 (31.75, 34.36) | 66.96 (65.64, 68.25) | 0.18 (0.13, 0.24) | <0.001 | 0.49 (0.27, 0.89) | <0.001 | | Anemia (n = 19,321) | Anemia (n = 19,321) | Anemia (n = 19,321) | Anemia (n = 19,321) | Anemia (n = 19,321) | Anemia (n = 19,321) | Anemia (n = 19,321) | | Severe | 20.87 (16.22, 26.42) | 79.13 (73.58, 83.78) | Ref | | - | | | Moderate | 20.49 (18.91, 22.18) | 79.51 (77.82, 81.11) | 1.02 (0.74, 1.41) | | | | | Mild | 21.71 (20.01, 23.48) | 78.30 (76.52, 79.99) | 0.95 (0.68, 1.31) | | | | | Not anemic | 23.39 (21.95, 24.89) | 76.61 (75.11, 78.05) | 0.86 (0.63, 1.18) | 0.06 | | | | BMI (n = 20,635) | BMI (n = 20,635) | BMI (n = 20,635) | BMI (n = 20,635) | BMI (n = 20,635) | BMI (n = 20,635) | BMI (n = 20,635) | | Underweight | 11.36 (9.60, 13.39) | 88.64 (86.61, 90.39) | Ref | | Ref | | | Normal weight | 23.36 (22.04, 24.73) | 76.64 (75.27, 77.96) | 0.42 (0.34, 0.51) | | 0.63 (0.44, 0.89) | | | Overweight/obese | 31.58 (30.14, 33.06) | 68.42 (66.94, 69.86) | 0.22 (0.23, 0.34) | <0.001 | 0.51 (0.35, 0.72) | <0.001 | | Waist-to-hip ratio (n = 20,630) | Waist-to-hip ratio (n = 20,630) | Waist-to-hip ratio (n = 20,630) | Waist-to-hip ratio (n = 20,630) | Waist-to-hip ratio (n = 20,630) | Waist-to-hip ratio (n = 20,630) | Waist-to-hip ratio (n = 20,630) | | <=0.9 for males and <=0.8 for females | 19.23 (16.98, 21.70) | 80.77 (78.3, 83.02) | Ref | | Ref | | | Cutoffs > 0.9 for males and >0.8 for females | 27.55 (26.5, 28.62) | 72.45 (71.38, 73.5) | 0.62 (0.53, 0.73) | <0.001 | 0.78 (0.61, 0.99) | 0.04 | | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | | Never in union | 13.01 (10.95, 15.42) | 86.99 (84.6, 89.05) | Ref | | Ref | | | Currently married | 27.03 (25.97, 28.12) | 72.97 (71.88, 74.03) | 0.40 (0.33, 0.49) | | 0.78 (0.53, 1.14) | | | Widowed/Divorced/Separated | 33.67 (30.41, 37.12) | 66.33 (62.9, 69.59) | 0.29 (0.23, 0.37) | <0.001 | 0.62 (0.39, 1.00) | 0.11 | | Education | Education | Education | Education | Education | Education | Education | | No education/Preprimary education | 29.97 (28.26, 31.73) | 70.03 (68.27, 71.74) | Ref | | Ref | | | Primary | 28.53 (26.32, 30.87) | 71.74 (69.13, 73.70) | 1.07 (0.93, 1.22) | | 1.16 (0.92, 1.46) | | | Secondary education | 24.38 (23.02, 25.79) | 75.62 (74.21, 76.98) | 1.33 (1.19, 1.47) | | 1.36 (1.13, 1.65) | | | Higher education | 22.44 (19.62, 25.53) | 77.56 (74.47, 80.38) | 1.48 (1.22, 1.78) | <0.001 | 1.58 (1.15, 2.15) | 0.004 | | Residence | Residence | Residence | Residence | Residence | Residence | Residence | | Urban | 28.12 (26.31, 30.01) | 71.88 (69.99, 73.69) | Ref | | Ref | | | Rural | 25.11 (24.03, 26.23) | 74.89 (73.77, 75.97) | 1.16 (1.04, 1.30) | 0.005 | 0.96 (0.80, 1.14) | 0.66 | | Cast/Tribe (n = 19,162) | Cast/Tribe (n = 19,162) | Cast/Tribe (n = 19,162) | Cast/Tribe (n = 19,162) | Cast/Tribe (n = 19,162) | Cast/Tribe (n = 19,162) | Cast/Tribe (n = 19,162) | | SC | 25.43 (23.52, 27.43) | 74.57 (72.57, 76.48) | 1.02 (0.89, 1.17) | | - | | | ST | 23.41 (20.74, 26.31) | 76.59 (73.69, 79.26) | 1.14 (0.95, 1.36) | | | | | OBC | 25.36 (23.93, 26.85) | 74.64 (73.15, 76.07) | 1.02 (0.91, 1.16) | | | | | Non-SC/non-ST/non-OBC | 25.89 (24.08, 27.79) | 74.11 (72.21, 75.92) | Ref | 0.53 | | | | Religion | Religion | Religion | Religion | Religion | Religion | Religion | | Hindu | 25.24 (24.19, 26.32) | 74.76 (73.68, 75.81) | Ref | | Ref | | | Muslim | 29.95 (27.36, 32.68) | 70.05 (67.32, 72.64) | 0.78 (0.68, 0.90) | | 0.81 (0.65, 1.00) | | | Others | 29.30 (26.03, 32.80) | 70.71 (67.21, 73.97) | 0.81 (0.68, 0.96) | <0.001 | 0.87 (0.65, 1.15) | 0.66 | | Wealth index | Wealth index | Wealth index | Wealth index | Wealth index | Wealth index | Wealth index | | Poorest | 22.25 (20.05, 24.61) | 77.75 (75.39, 79.95) | Ref | | Ref | | | Poorer | 23.93 (21.93, 26.05) | 76.07 (73.95, 78.07) | 0.91 (0.77, 1.07) | | 0.94 (0.70, 1.26) | | | Middle | 26.77 (24.87, 28.76) | 73.23 (71.24, 75.13) | 0.78 (0.66, 0.92) | | 0.81 (0.60, 1.09) | | | Richer | 27.40 (25.58, 29.29) | 72.69 (70.71, 74.42) | 0.75 (0.64, 0.88) | | 0.83 (0.60, 1.15) | | | Richest | 28.51 (26.49, 30.61) | 71.49 (69.39, 73.51) | 0.71 (0.61, 0.84) | <0.001 | 0.66 (0.45, 0.96) | 0.15 | | Type of healthcare facility accessed (n = 8,103) | Type of healthcare facility accessed (n = 8,103) | Type of healthcare facility accessed (n = 8,103) | Type of healthcare facility accessed (n = 8,103) | Type of healthcare facility accessed (n = 8,103) | Type of healthcare facility accessed (n = 8,103) | Type of healthcare facility accessed (n = 8,103) | | Public facility | 26.39 (24.58, 28.28) | 73.61 (71.72, 75.42) | Ref | | - | | | Private facility | 26.52 (24.31, 28.86) | 73.48 (71.14, 75.7) | 0.99 (0.85, 1.14) | | | | | NGO/Other | 41.91 (28.39, 56.74) | 58.11 (43.26, 71.61) | 0.49 (0.27, 0.91) | 0.07 | | | | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | Health insurance coverage | | Yes | 27.71 (26.19, 29.29) | 72.29 (70.71, 73.81) | 0.89 (0.81, 0.98) | 0.02 | 1.10 (0.94, 1.28) | 0.23 | | No | 25.49 (24.31, 26.71) | 74.51 (73.29, 75.69) | Ref | | Ref | | | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | | Diabetes (n = 20,443) | Diabetes (n = 20,443) | Diabetes (n = 20,443) | Diabetes (n = 20,443) | Diabetes (n = 20,443) | Diabetes (n = 20,443) | Diabetes (n = 20,443) | | No | 25.28 (24.27, 26.32) | 74.72 (73.68, 75.73) | Ref | | Ref | | | Yes | 34.22 (31.25, 37.33) | 65.78 (62.67, 68.75) | 0.65 (0.56, 0.75) | <0.001 | 0.79 (0.63, 0.99) | 0.04 | | Heart disease (n = 20,583) | Heart disease (n = 20,583) | Heart disease (n = 20,583) | Heart disease (n = 20,583) | Heart disease (n = 20,583) | Heart disease (n = 20,583) | Heart disease (n = 20,583) | | No | 26.21 (25.23, 27.22) | 73.79 (72.78, 74.77) | Ref | | - | | | Yes | 26.43 (21.49, 32.04) | 73.57 (67.96, 78.51) | 0.98 (0.75, 1.30) | 0.93 | | | | Usage of any tobacco | Usage of any tobacco | Usage of any tobacco | Usage of any tobacco | Usage of any tobacco | Usage of any tobacco | Usage of any tobacco | | No | 25.81 (24.85, 26.81) | 74.19 (73.20, 75.15) | 1.29 (1.07, 1.55) | | 0.85 (0.63, 1.15) | 0.31 | | Yes | 31.05 (27.33, 35.02) | 68.95 (64.98, 72.67) | Ref | 0.006 | Ref | | | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | Alcohol usage currently | | No | 25.83 (24.88, 26.81) | 74.17 (73.19, 75.12) | Ref | | Ref | | | Yes | 38.21 (30.51, 46.53) | 61.81 (53.47, 69.49) | 0.56 (0.39, 0.79) | 0.001 | 0.80 (0.48, 1.34) | 0.41 | | Fried food | Fried food | Fried food | Fried food | Fried food | Fried food | Fried food | | Less frequently | 25.56 (24.34, 26.81) | 74.44 (73.19, 75.66) | Ref | | - | | | More frequently | 27.20 (25.71, 28.75) | 72.80 (71.25, 74.30) | 0.92 (0.83, 1.01) | 0.09 | | | | Aerated drinks | Aerated drinks | Aerated drinks | Aerated drinks | Aerated drinks | Aerated drinks | Aerated drinks | | Less frequently | 26.42 (25.37, 27.49) | 73.58 (72.51, 74.63) | Ref | | - | | | More frequently | 25.53 (23.41, 27.76) | 74.47 (72.24, 76.59) | 1.04 (0.92, 1.18) | 0.45 | | | | Region | Region | Region | Region | Region | Region | Region | | North | 26.71 (24.98, 28.51) | 73.29 (71.49, 75.02) | Ref | | Ref | | | Central | 23.29 (21.41, 25.27) | 76.71 (74.73, 78.59) | 1.20 (1.04, 1.38) | | 0.90 (0.69, 1.17) | | | East | 25.99 (23.81, 28.31) | 74.01 (71.70, 76.19) | 1.03 (0.89, 1.20) | | 0.69 (0.53, 0.89) | | | Northeast | 33.71 (31.05, 36.47) | 66.29 (63.53, 68.95) | 0.71 (0.61, 0.83) | | 0.50 (0.37, 0.66) | | | West | 23.73 (20.38, 27.44) | 76.27 (72.56, 79.62) | 1.17 (0.94, 1.45) | | 0.96 (0.68, 1.35) | | | South | 27.85 (26.00, 29.77) | 72.15 (20.23, 74.00) | 0.94 (0.82, 1.07) | <0.001 | 1.08 (0.86, 1.34) | <0.001 | Table 5 compares the proportion of hypertensive patients on treatment and achieving optimal BP control between various states and union territories of India stratified by their health index (performance) scores. Individuals in the majority of the Indian states have poor (<$50\%$) treatment-seeking behavior due to the noninitiation of regular antihypertensive treatment despite awareness of their hypertension status. In empowered action group states such as Jharkhand and Uttar Pradesh, less than $25\%$ of previously diagnosed hypertensives were on treatment at the time of the survey. However, a majority of states achieved BP control in $65\%$ or more of hypertensive patients taking BP-lowering medications. **Table 5** | States | Health index reference year (2019-2020) | Awareness of the hypertension status and on treatment, row% (95% CI) NFHS-5 | On treatment with optimal control, row% (95% CI) NFHS-5 | Awareness of the hypertension status and on treatment, row% (95% CI) NFHS-4 | On treatment with optimal control, row% (95% CI) NFHS-4 | | --- | --- | --- | --- | --- | --- | | High health index | High health index | High health index | High health index | High health index | High health index | | Kerala | 82.21 | 59.96 (55.62, 64.14) | 72.71 (68.02, 76.93) | 49.59 (45.78, 53.41) | 84.72 (80.34, 88.27) | | Mizoram | 75.77 | 44.57 (37.83, 51.52) | 86.29 (78.69, 91.47) | 24.13 (20.38, 28.33) | 78.25 (70.61, 84.35) | | Tamil Nadu | 72.42 | 38.97 (34.64, 43.49) | 71.06 (66.95, 74.86) | 18.94 (17.18, 20.84) | 80.24 (77.25, 82.93) | | Tripura | 70.16 | 51.64 (46.03, 57.20) | 73.35 (66.68, 79.11) | 46.12 (40.31, 52.05) | 81.14 (75.63, 85.63) | | Telangana | 69.96 | 39.78 (39.03, 40.54) | 72.96 (69.31, 73.81) | 30.27 (26.22, 34.66) | 79.09 (74.29, 83.21) | | Andhra Pradesh | 69.95 | 64.50 (59.82, 68.93) | 76.46 (72.28, 80.19) | 33.61 (29.17, 38.35) | 79.58 (74.89, 83.59) | | Maharashtra | 69.14 | 61.94 (58.06, 65.67) | 77.17 (72.55, 81.28) | 43.53 (38.84, 48.34) | 83.31 (79.56, 86.47) | | Dadra and Nagar Haveli and Daman and Diu | 66.21 | 53.98 (42.77, 64.81) | 69.44 (51.31, 83.05) | 62.54 (51.54, 72.39) | 84.94 (73.18, 92.11) | | Gujarat | 63.59 | 53.63 (50.06, 57.17) | 79.31 (75.09, 82.96) | 45.57 (41.94, 49.24) | 84.14 (79.94, 87.61) | | Himachal Pradesh | 63.17 | 39.46 (35.39, 43.68) | 81.13 (76.06, 85.33) | 35.25 (31.65, 39.02) | 76.15 (70.92, 80.71) | | Chandigarh | 62.53 | 28.11 (17.91, 41.21) | 93.73 (68.81, 99.02) | 28.51 (20.53, 38.12) | 72.73 (56.30, 84.67) | | Punjab | 58.08 | 28.11 (25.84, 39.48) | 67.81 (63.76, 71.61) | 32.56 (30.06, 35.16) | 65.02 (60.77, 69.05) | | Karnataka | 57.93 | 55.58 (50.39, 60.66) | 67.92 (63.34, 72.18) | 43.61 (38.70, 48.66) | 74.59 (70.29, 78.46) | | Sikkim | 55.53 | 44.22 (35.00, 53.86) | 62.71 (46.98, 76.12) | 32.72 (28.69, 37.02) | 67.23 (60.41, 73.41) | | Medium health index | Medium health index | Medium health index | Medium health index | Medium health index | Medium health index | | Goa | 53.68 | 73.57 (62.36, 82.39) | 44.76 (31.33, 58.99) | 62.54 (51.54, 72.39) | 79.50 (68.02, 87.61) | | Lakshadweep | 51.87 | 67.13 (51.68, 79.59) | 72.42 (57.14, 83.79) | 49.00 (38.43, 59.66) | 72.99 (57.61, 84.31) | | Puducherry | 50.83 | 64.02 (50.36, 75.73) | 77.41 (64.51, 86.61) | 15.13 (9.21, 23.86) | 75.14 (65.47, 82.81) | | Chhattisgarh | 50.71 | 38.26 (33.89, 42.81) | 74.75 (69.62, 79.27) | 40.57 (36.46, 44.82) | 79.55 (75.00, 83.45) | | Delhi | 49.84 | 46.96 (42.78, 51.21) | 68.04 (63.31, 72.43) | 33.29 (26.22, 41.21) | 82.44 (74.05, 88.54) | | Haryana | 49.26 | 29.44 (26.94, 32.08) | 75.32 (71.59, 78.71) | 23.45 (20.49, 26.71) | 82.87 (79.00, 86.15) | | Assam | 47.74 | 54.81 (51.66, 57.91) | 64.08 (60.41, 67.59) | 41.83 (39.04, 44.67) | 63.26 (59.71, 66.68) | | Jharkhand | 47.55 | 21.04 (18.35, 24.00) | 81.02 (76.35, 84.95) | 16.87 (14.99, 18.93) | 84.74 (80.87, 87.95) | | Jammu and Kashmir | 46.99 | 48.66 (44.12, 53.23) | 59.81 (55.27, 64.21) | 42.77 (39.86, 45.73) | 80.77 (77.98, 83.27) | | Low health index | Low health index | Low health index | Low health index | Low health index | Low health index | | A&N islands | 44.74 | 71.11 (62.07, 70.71) | 69.36 (58.81, 78.21) | 57.87 (48.41, 66.81) | 71.62 (63.06, 78.86) | | Odisha | 44.31 | 44.81 (41.24, 48.43) | 76.33 (72.79, 79.54) | 37.07 (33.9, 40.36) | 82.92 (80.17, 85.36) | | Uttarakhand | 44.21 | 39.68 (34.76, 44.81) | 67.27 (58.76, 74.79) | 39.71 (36.16, 43.35) | 78.97 (74.86 (82.56) | | Meghalaya | 43.05 | 72.39 (67.07, 77.15) | 77.28 (71.81, 81.95) | 52.21 (45.5, 58.85) | 86.04 (82.16, 98.18) | | Rajasthan | 41.33 | 32.08 (29.16, 35.15) | 81.70 (77.82, 85.02) | 33.86 (31.48, 36.32) | 78.25 (70.61, 84.35) | | Madhya Pradesh | 36.72 | 38.76 (35.57, 42.06) | 76.74 (72.99, 80.11) | 48.53 (38.84, 48.34) | 84.78 (82.75 (86.61) | | Manipur | 34.26 | 36.09 (31.53, 40.91) | 66.67 (57.82, 74.48) | 20.29 (17.67, 23.18) | 67.19 (61.18, 72.69) | | Arunachal Pradesh | 33.92 | 30.09 (26.98, 33.40) | 68.33 (63.63, 72.69) | 32.89 (29.30, 36.70) | 64.08 (59.22, 68.66) | | Bihar | 31.00 | 29.34 (27.13, 31.65) | 85.99 (83.46, 88.19) | 42.19 (39.05, 45.41) | 89.11 (86.98, 90.93) | | Uttar Pradesh | 30.57 | 23.34 (21.86, 24.89) | 76.57 (73.97, 78.99) | 26.47 (24.92, 28.08) | 85.79 (83.97, 87.42) | | Nagaland | 27.01 | 31.59 (26.16, 37.59) | 52.00 (43.79, 60.11) | 26.47 (23.26, 29.96) | 51.91 (45.28, 58.46) | | West Bengal | Not available | 62.43 (58.69, 66.04) | 62.05 (57.79, 66.13) | 42.87 (39.01, 46.81) | 73.90 (69.92, 77.52) | Figures 1 and 2 display the change in state-wise prevalence of individuals with awareness of their hypertension status, on treatment, and with controlled BP. Furthermore, the correlation coefficient of the NITI Aayog health index score with a state-wise prevalence of aware individuals on treatment was 0.36 ($$P \leq 0.037$$) while correlation with the prevalence of individuals with controlled BP was 0.19 ($$P \leq 0.26$$). **Figure 1:** *Prevalence of hypertension awareness, treatment, and control across states in India (NFHS-4, 2015-2016).Figure credits: All the authors of this study.NFHS-4, National Family Health Survey Fourth Series* **Figure 2:** *Prevalence of hypertension awareness, treatment, and control across states in India (NFHS-5, 2019-2021).Figure credits: All the authors of this study.NFHS-5, National Family Health Survey Fifth Series* ## Discussion This nationally representative survey in the age group of 15 to 49 years indicates a weak hypertension control cascade in India. The burden of hypertension increased slightly from $20.4\%$ (NFHS-4) to $22.8\%$ (NFHS-5), while the proportion of new cases detected on screening increased from $41.6\%$ to $52.1\%$. The proportion of previously diagnosed cases on antihypertensive therapy increased from $32.6\%$ to $40.7\%$. However, the proportion of cases of BP-lowering medication attaining controlled BP modestly reduced from $80.8\%$ to $73.7\%$. The overall prevalence of hypertension in India in the age group of 15-49 years as per NFHS-5 was $22.80\%$, an estimate similar to that reported in the District Level Health Survey and the Annual Health Survey [2012-2014], similar nationally representative surveys from India [15]. On comparing DHS surveys across countries, the prevalence of hypertension in *India is* higher than in Peru ($19.77\%$) [16] and Nepal ($19.99\%$) [17] but lower than in Bangladesh ($27.5\%$) [18]. A majority (~$58\%$) of existing hypertension cases in India are undiagnosed as per the current round of the NFHS, a finding almost identical to that in Bangladesh [2017-2018] [18]. The burden of undiagnosed cases was significantly higher in males, middle-aged, lower education level, poorer wealth quintiles, STs, and rural inhabitants compared to females, younger, higher education level, richer wealth quintiles, non-ST, and urban inhabitants, respectively. In contrast, evidence from a study conducted in China [19] and an intervention trial conducted in Nepal [20] reported an increasing trend in hypertension status awareness with the advancing age of individuals. However, previous studies from multiple LMICs also indicate that populations having low education and socioeconomic status (SES) have reduced awareness of their hypertension status, although, in Bangladesh, education was protective against a lack of awareness of the actual hypertension status [21]. Availability of health insurance influences an individual’s decision to seek treatment for their health condition, a finding consistent with our study that corroborates prior evidence suggesting those without health insurance had lower odds of availing treatment for hypertension [19]. Furthermore, a higher proportion of men compared to women were not on BP-lowering medication, a finding consistent with NFHS-4 [2015-2016] [22]. Similar to previous studies, this study's findings also suggest that older adults [23], males [20], and obese/overweight individuals [23] were less likely to attain optimal BP due to biological risk. The waist-to-hip ratio is also now emerging as a better correlate for developing both hypertension and suboptimal BP control when on medication [24]. Consequently, patients with diabetes experience greater challenges in achieving BP control due to the high prevalence of obesity and/or high waist-to-hip ratio in these comorbid patients [25]. In this study, low education was a predictor of poorly controlled hypertension. There is growing recognition that an educational gradient predisposes individuals with a lower educational level to a higher risk of onset and progression of cardiovascular disease due to improper health-seeking behavior and poor medication adherence [26]. Northeastern states of India have the highest prevalence of hypertension [27]. We also found that most states in the northeastern region of India had poor treatment-seeking behavior and poor BP control, which also correlated with their low health index. Strengthening primary health systems in low-resource settings may translate into an effective treatment cascade for hypertension care in India. Our study has certain important public health policy implications. First, a large subset of the population in India remains undiagnosed with hypertension indicative of a lack of effective screening and missed opportunities in primary care outpatient settings despite policy directives in this regard. Additionally, screening of adolescents and young adults must be initiated as part of the medical routine as a greater proportion of these subgroups tend to remain unaware of their hypertensive status and have poor treatment-seeking behavior [22]. Patients with risk factors such as those with a higher waist-to-hip ratio should be prioritized for screening as they have an increased risk of remaining undiagnosed. Physicians should provide an enhanced focus on individuals with comorbidities such as diabetes who are less likely to have control over their BP levels, which further accelerates their risk of disease progression. Greater advocacy is needed in the National Program for Noncommunicable Diseases (NCDs) prevention in India to meet the modified strategies related to prevention and behavior change [28]. Second, six in 10 patients despite having awareness of a hypertension diagnosis are not initiated on treatment suggestive of poor treatment-seeking behavior, signifying the requirement for appropriate educational and behavioral interventions from the time of initial diagnosis. Third, there has been a significant improvement in the proportion of patients on antihypertensive treatment ($40.7\%$) compared to the previous NFHS-4 (2015-2016; $32.6\%$) round suggestive of improved drug accessibility that could be secondary to schemes such as the Pradhan Mantri Jan Aushadi Yojana (PMJAY) that promote people’s access to high-quality generic medicines at affordable prices [29]. Finally, India’s health performance index does not correlate with core elements of the hypertension treatment cascade, signifying optimal maternal and child health indicators are not an appropriate proxy for the effectiveness of NCDs management that requires the incorporation of specific and relevant parameters. There are certain limitations of this study. First, NFHS does not include the geriatric population. However, analysis from a large population data set also reflects a suboptimal treatment cascade among the elderly in India with similar loss of hypertension care at multiple stages, including measurement of hypertension ($72.5\%$), diagnosis/awareness ($57.3\%$), on treatment ($50.5\%$), and control ($27.5\%$) albeit comparatively better than younger age groups as observed in our analysis [30]. Second, the information on adherence to antihypertensive treatment, which is a key determinant of BP control was unavailable and could not be estimated in this analysis. Third, the survey did not assess the physical activity of the individuals, and therefore, we could not assess its association with BP control. Fourth, clinical diagnosis of hypertension was not established in the NFHS surveys and only reflects a statistical estimate of the surveyed population. ## Conclusions The hypertension control cascade in younger and middle-aged groups has major lacunae at every stage, from screening and diagnosis to initiation of effective antihypertensive treatment and attainment of safe BP levels although significant improvements were observed in the screening yield and initiation of antihypertensive treatment. Identification of high-risk groups for opportunistic screening, implementation of community-based screening, strengthening primary care, and sensitizing associated practitioners are urgently warranted. ## References 1. **Hypertension - Fact Sheet**. (2023) 2. Nguyen TN, Chow CK. **Global and national high blood pressure burden and control**. *Lancet* (2021) **398** 932-933. PMID: 34450082 3. **GBD Compare**. (2023) 4. Mills KT, Bundy JD, Kelly TN. **Global disparities of hypertension prevalence and control: a systematic analysis of population-based studies from 90 countries**. *Circulation* (2016) **134** 441-450. PMID: 27502908 5. Gupta R, Gaur K, S Ram CV. **Emerging trends in hypertension epidemiology in India**. *J Hum Hypertens* (2019) **33** 575-587. PMID: 30254382 6. Prenissl J, Manne-Goehler J, Jaacks LM. **Hypertension screening, awareness, treatment, and control in India: a nationally representative cross-sectional study among individuals aged 15 to 49 years**. *PLoS Med* (2019) **16** 0 7. **India Fact Sheet: National Family Health Survey-5 (2019-21)**. (2023) 8. Hooker RC, Cowap N, Newson R, Freeman GK. **Better by half: hypertension in the elderly and the "rule of halves": a primary care audit of the clinical computer record as a springboard to improving care**. *Fam Pract* (1999) **16** 123-128. PMID: 10381016 9. Roy A, Praveen PA, Amarchand R. **Changes in hypertension prevalence, awareness, treatment and control rates over 20 years in National Capital Region of India: results from a repeat cross-sectional study**. *BMJ Open* (2017) **7** 0 10. Mahapatra R, Kaliyappan A, Chinnakali P, Hanumanthappa N, Govindarajalou R, Bammigatti C. **Prevalence and risk factors for resistant hypertension: cross-sectional study from a tertiary care referral hospital in South India**. *Cureus* (2021) **13** 0 11. **Hypertension: Standard Treatment Guidelines.**. (2023) 12. **India Report. International Institute for Population Sciences (IIPS) and ICF: National Family Health Survey (NFHS-5), 2019-21. IIPS, Mumbai, India**. (2023) 13. **Health Performance: NITI Aayog**. (2023) 14. **Health Index. NITI Aayog**. (2023) 15. Geldsetzer P, Manne-Goehler J, Theilmann M. **Diabetes and hypertension in India: a nationally representative study of 1.3 million adults**. *JAMA Intern Med* (2018) **178** 363-372. PMID: 29379964 16. Chambergo-Michilot D, Rebatta-Acuña A, Delgado-Flores CJ, Toro-Huamanchumo CJ. **Socioeconomic determinants of hypertension and prehypertension in Peru: evidence from the Peruvian Demographic and Health Survey**. *PLoS One* (2021) **16** 0 17. Hasan M, Sutradhar I, Akter T, Das Gupta R, Joshi H, Haider MR, Sarker M. **Prevalence and determinants of hypertension among adult population in Nepal: data from Nepal Demographic and Health Survey 2016**. *PLoS One* (2018) **13** 0 18. Iqbal A, Ahsan KZ, Jamil K, Haider MM, Khan SH, Chakraborty N, Streatfield PK. **Demographic, socioeconomic, and biological correlates of hypertension in an adult population: evidence from the Bangladesh demographic and health survey 2017-18**. *BMC Public Health* (2021) **21** 1229. PMID: 34174868 19. Lu J, Lu Y, Wang X. **Prevalence, awareness, treatment, and control of hypertension in China: data from 1.7 million adults in a population-based screening study (China PEACE Million Persons Project)**. *Lancet* (2017) **390** 2549-2558. PMID: 29102084 20. Khanal MK, Dhungana RR, Bhandari P, Gurung Y, Paudel KN. **Prevalence, associated factors, awareness, treatment, and control of hypertension: findings from a cross sectional study conducted as a part of a community based intervention trial in Surkhet, Mid-western region of Nepal**. *PLoS One* (2017) **12** 0 21. Lyu YY, Zhang B, Wang HJ. **Changing trends regarding the rates of prevalence, awareness and treatment of hypertension and the differences by demographic and economic factors, among adult farmers in nine provinces in China, from 1991 to 2015. (Article in Chinese)**. *Zhonghua Liu Xing Bing Xue Za Zhi* (2020) **41** 498-503. PMID: 32344471 22. Krishnamoorthy Y, Rajaa S, Sulgante S, Chinnakali P, Jaswal N, Goel S. **Prevalence of hypertension and determinants of treatment-seeking behaviour among adolescents and young adults in India: an analysis of NFHS-4**. *J Public Health (Oxf)* (2022) 0 23. Li W, He Y, Xia L. **Association of age-related trends in blood pressure and body composition indices in healthy adults**. *Front Physiol* (2018) **9** 1574. PMID: 30534075 24. Wang Y, Howard AG, Adair LS, Wang H, Avery CL, Gordon-Larsen P. **Waist circumference change is associated with blood pressure change independent of BMI change**. *Obesity (Silver Spring)* (2020) **28** 146-153. PMID: 31755247 25. Schmieder RE, Ruilope LM. **Blood pressure control in patients with comorbidities**. *J Clin Hypertens (Greenwich)* (2008) **10** 624-631. PMID: 18772645 26. Di Chiara T, Scaglione A, Corrao S, Argano C, Pinto A, Scaglione R. **Education and hypertension: impact on global cardiovascular risk**. *Acta Cardiol* (2017) **72** 507-513. PMID: 28657499 27. Gupta K, Jain V, Qamar A, Singal AK, Ramakrishnan S, Gupta R, Bajaj NS. **Regional impact of updated guidelines on prevalence and distribution of blood pressure categories for hypertension in India: results from the National Family Health Survey 4**. *Indian Heart J* (2021) **73** 481-486. PMID: 34474762 28. **National Programme for prevention & Control of Cancer, Diabetes, Cardiovascular Diseases & stroke (NPCDCS): Operational Guidelines: Revised (2013-17)**. (2023) 29. Chaudhary RK, Philip MJ, Santhosh A, Karoli SS, Bhandari R, Ganachari MS. **Health economics and effectiveness analysis of generic anti-diabetic medication from jan aushadhi: an ambispective study in community pharmacy**. *Diabetes Metab Syndr* (2023) **15** 102303. PMID: 34626923 30. Kothavale A, Puri P, Sangani PG. **Quantifying population level hypertension care cascades in India: a cross-sectional analysis of risk factors and disease linkages**. *BMC Geriatr* (2022) **22** 98. PMID: 35114935
--- title: 'Oncological healthcare providers’ perspectives on appropriate melanoma survivorship care: a qualitative focus group study' authors: - Nadia C. W. Kamminga - Marlies Wakkee - Rianne J. De Bruin - Astrid. A. M. van der Veldt - Arjen Joosse - Suzan W. I. Reeder - Peter W. Plaisier - Tamar Nijsten - Marjolein Lugtenberg journal: BMC Cancer year: 2023 pmcid: PMC10042579 doi: 10.1186/s12885-023-10759-9 license: CC BY 4.0 --- # Oncological healthcare providers’ perspectives on appropriate melanoma survivorship care: a qualitative focus group study ## Abstract ### Background The increasing group of melanoma survivors reports multiple unmet needs regarding survivorship care (SSC). To optimise melanoma SSC, it is crucial to take into account the perspectives of oncological healthcare providers (HCPs) in addition to those of patients. The aim of this study is to gain an in-depth understanding of HCPs’ perspectives on appropriate melanoma SSC. ### Methods Four online focus groups were conducted with mixed samples of oncological HCPs (dermatologists, surgeons, oncologists, oncological nurse practitioners, support counsellors and general practitioners) (total $$n = 23$$). A topic guide was used to structure the discussions, focusing on perspectives on both SSC and survivorship care plans (SCPs). All focus groups were recorded, transcribed verbatim, and subjected to an elaborate thematic content analysis. ### Results Regarding SSC, HCPs considered the current offer minimal and stressed the need for broader personalised SSC from diagnosis onwards. Although hardly anyone was familiar with SCPs, they perceived various potential benefits of SCPs, such as an increase in the patients’ self-management and providing HCPs with an up-to-date overview of the patient’s situation. Perceived preconditions for successful implementation included adequate personalisation, integration in the electronic health record and ensuring adequate funding to activate and provide timely updates. ### Conclusions According to HCPs there is considerable room for improvement in terms of melanoma SSC. SCPs can assist in offering personalised and broader i.e., including psychosocial SSC. Aside from personalisation, efforts should be focused on SCPs' integration in clinical practice, and their long-term maintenance. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12885-023-10759-9. ## Background The incidence of cutaneous melanoma has been steadily increasing and reached over 300.000 new cases worldwide in 2020 [1]. Novel therapies, including immune checkpoint inhibitors (ICIs) and targeted therapies (BRAF + MEK inhibitors), have led to an increased overall survival of patients with advanced disease [2, 3], converting metastasised melanoma into one of the first potentially curable cancers [4]. Consequently, more patients with melanoma are able to resume their lives, albeit with very variable prognoses, both within and between stages [5]. In-depth qualitative research has shown that metastatic melanoma patients have unmet needs in terms of survivorship care (SSC) such as the need for tailored information and broader supportive care [6]. Furthermore, even in patients with a thin (lower stage) melanoma – for whom the prognosis in most cases is excellent – the impact of the diagnosis is often significant and they need more SSC than currently provided [7]. SSC is defined as care provided to cancer survivors, focusing on prevention and identification of treatable cancer recurrences, second cancers, late effects and improving quality of life [8]. A central component of and a tool to provide SSC is a survivorship care plan (SCP), which aims at informing patients about the disease, treatment and its possible effects and improving coordination of care [8]. The recommended categories of SSC [8, 9] and components of SCPs [8, 10] are listed in Table 1. Providing adequate SSC can fulfil the unmet needs of cancer survivors including those with melanoma [6].Table 1Recommended categories of SSC and components of SCPs by the IOM [8]Recommended categories SSCa8, 9Recommended components of SCPsb8, 101. Information and education about the disease, its treatment and the possible early and late effects;2. Identification and treatment of the disease and therapy effects on all possible domains (i.e. physical and psychosocial, including work- and insurance-related);3. Oncological follow-up with surveillance for cancer progression, recurrences or second cancers;4. Coordination between all the healthcare providers involved in the care process, to make sure all of the survivor’s health needs are met• Cancer type, treatments received, and their potential consequences;• Specific information about the timing and content of recommended follow-up;• Recommendations regarding preventive practices and how to maintain health and well-being;• Information regarding employment and health insurance; and• Information on the availability of psychosocial, nutritional, and other supportive servicesaSSC = Survivorship care, i.e. the care provided by either specialists or primary care providers to all cancer survivors, focusing on prevention, ensuring access to effective interventions and helping patients to improve their quality of lifebSCP = Survivorship care plan, i.e. comprehensive care summary and follow-up plan that is clearly and effectively explained and consists of critical information needed for the survivor’s long-term care Despite the recommendation of the American Institute of Medicine (IOM) to provide all cancer survivors with an SCP [8], adopted by the *Dutch melanoma* guideline [11] as well as international (melanoma) guidelines [9, 11], implementation in clinical practice is limited [12, 13]. Only 5–$52\%$ of the healthcare providers (HCPs) provide their patients with an SCP [12]. When comparing cancer types, patients with melanoma appear least likely to receive an SCP [14]. These data, together with the reported unmet SSC needs, signal suboptimal provision and implementation of SSC and SCPs in melanoma care. Furthermore, little is known about how HCPs should provide SSC to patients with metastatic diseases with such varying prognoses as melanoma. To optimise melanoma SSC, it is crucial to take into account the perspective of oncological HCPs, in addition to those of patients. However, how oncological HCPs view melanoma SSC and what they consider important regarding this topic has not yet been described in current literature. Therefore, the aim of the present study is to gain an in-depth understanding of the perspectives of oncological HCPs on appropriate melanoma SSC. This will enable tailored melanoma SSC to both patients’ and HCPs’ needs, which may lead to better implementation and effectiveness in clinical practice [8, 12, 15]. ## Study design and methodological considerations A qualitative online focus group design was chosen as qualitative research is particularly suitable for in-depth exploration of experiences and perspectives on a particular subject [16]. Moreover, focus groups are expected to provide rich and diverse data because of the interaction between participants [17]. Due to the COVID-19 pandemic, we decided to organise online focus groups instead of face-to-face meetings to prevent unnecessary group gatherings. The reporting of this study followed the *Consolidated criteria* for Reporting Qualitative research (COREQ) [18]. ## Setting and participant selection This study was conducted as part of a project in which a personalised SCP for melanoma survivors will be developed. In this project, melanoma survivors are defined as individuals diagnosed with melanoma (stage I—IV) [10, 19]. However, ‘survivor’ and ‘patient’ are used interchangeably throughout this article. This project was performed in the region of Groot-Rijnmond with one participating academic hospital (Erasmus Medical Center) and four non-academic hospitals (Albert Schweitzer Hospital, Franciscus Gasthuis & Vlietland and Maasstad Hospital). Within this region, melanoma care is often multidisciplinary in nature and shared between specialists and primary-care providers [20]. For stage I follow-up is mainly GP-led, for stage II and III both dermatologist- and surgical oncologist-led and mainly oncologist-led for stage IV melanoma. This long-term care is covered by mandatory state insurance and if ongoing income is not covered by an employer, a sickness benefit provides a temporary income for these patients for a maximum period of 2 years [21]. To select participants, each participating hospital was asked to provide a list of all their HCPs involved in melanoma care i.e., dermatologists, surgeons, oncologists, oncological nurse practitioners, support counsellors and general practitioners (GPs). Eligible participants had to work within one of the participating hospitals or larger region and had to speak the Dutch language well. Potential participants received information about the study by email and could apply by filling in an online form (Additional file 1). They were offered a small thank-you gift for participation. Thirty-two HCPs were willing to participate. Based on their availability and using a purposive sampling method [22], 24 HCPs were invited to participate. The aim was having a variable sample in terms of age, sex, hospital type, profession and field of medicine in each of the focus groups. All participants signed a consent form. After four focus groups, varying from five to six HCPs per session, data saturation was reached i.e., no new themes were identified from the data [23]. ## Data collection Prior to the focus groups participants completed a short-self-administered questionnaire to collect demographics and received instructions on how to participate in the online sessions. The focus groups were held through Microsoft Teams® and moderated by at least two of four researchers, including a female medical doctor (N.K.), female medical student (R.B.), female psychologist (M.L.) and female dermatologist (M.W.). All participants knew the researchers’ background and reasons for doing the research. One researcher (N.K. or R.B.) took notes during the focus groups. The focus groups lasted 90 to 120 min. A topic guide, which was based on relevant literature [6, 8, 12, 15, 24], was used to structure the discussion (Additional file 2). Four main topics were addressed: perceived impact of melanoma on patients, current SSC/SCP practices, opportunities for improvements and perceived facilitators and barriers using SCPs. All focus groups were both audio- and video recorded. ## Data analysis All recordings were transcribed verbatim in anonymised form. Video recordings were used to link statements to the correct participant. All transcripts were analysed using Nvivo version 12®. An elaborate thematic content analysis was performed consisting of several phases [25]: first, the researchers familiarised themselves with the data by rereading and summarizing each transcript [25, 26]. Subsequently, the transcripts were coded by one researcher (N.K. or R.B.) and then checked by a second researcher (N.K. or R.B.). During this initial coding process, the researchers identified all potential relevant themes [25]. The resulting unstructured list of initial codes was discussed with a third researcher (M.L.). In the second phase of analysis, the codes were sorted into a more structured coding list: relationships between all initial codes were identified and organised into main- and subthemes by two researchers (R.B. and N.K.). The resulting, structured list of candidate themes was discussed within the multidisciplinary research team (N.K., R.B., M.L.) until consensus was reached. In the final phase the list of candidate themes was reviewed, further refined and named (N.K., R.B., M.L. and M.W.), followed by checking them in accordance with the complete data set (N.K. and M.L.) [25]. ## Participant characteristics Characteristics of participating HCPs and group compositions are displayed in Table 2.Table 2Characteristics of focus group participants ($$n = 23$$)ParticipantSexAgeHospitalSpecialtyFrequency of involvement in melanoma careFocus group 1 HCP1F51EMCSupport counsellorMonthly HCP2M53EMCSurgeonDaily HCP3M37EMCOncologistDaily HCP4F33EMCOncology nurse practitionerWeekly HCP5M50-General practitionerYearly HCP6M58MSZDermatologistWeeklyFocus group 2 HCP7F37EMCSurgeonDaily HCP8F41EMCOncologistDaily HCP9F55EMC, MSZSupport counsellorMonthly HCP10F43MSZSurgeonWeekly HCP11M31EMCOncology nurse practitionerDaily HCP12F41ASZDermatologistDailyFocus group 3 HCP13M46MSZDermatologistWeekly HCP14F50FGVSurgeonWeekly HCP15F43-General practitionerYearly HCP16F39ASZDermatologistDaily HCP17M55ASZSurgeonWeekly HCP18M43EMCSurgeonWeeklyFocus group 4 HCP19M38EMCDermatologistWeekly HCP20M44ASZSurgeonWeekly HCP21F56EMCSupport counsellorMonthly HCP22M57FGVSurgeonWeekly HCP23F52-General practitionerYearlyF Female, M Male, EMC Erasmus medical centre, MSZ Maasstad hospital, ASZ Albert Schweitzer Hospital, FGV Franciscus Gasthuis & Vlietland hospital ## Perspectives on appropriate melanoma SSC including SCPs The analysis resulted in 4 main themes and 13 sub-themes (Fig. 1), which are discussed below. Fig. 1Overview of themes and subthemes ## Minimal offer of SSC HCPs regarded the current offer of melanoma SSC minimal and highlighted the difference with SSC offered to survivors of other types of cancer (e.g., breast- and colorectal cancer).I think it’s limited, especially compared to the SSC for other malignancies […] So no, I’m not that impressed by it [current melanoma SSC]. To be honest, my follow-up consults with melanoma patients never take very long. – Surgeon, male, 57 (HCP22) Furthermore, they considered current SSC mainly medically oriented, while optimal SSC should also include non-medical care and address psychosocial issues, such as work-related problems (see also 2.3). ## Varying problems and unmet SSC needs of patients HCPs emphasised that melanoma can have a significant impact, but is variable among patients. For example, in stage I-II melanoma, some feel reassured after the excision, whereas others continue to feel afraid despite the melanoma being removed. I have always noticed a big difference when talking about the outcome of the excision. There are people who say: ‘Oh well, you cut it out, is there anything else we need to do or are we done?’ And there are people who think they’ll be dead by next week. – Dermatologist, male, 38 (HCP19) Consequently, unmet SSC needs vary largely among patients and therefore SSC should differ per individual, firstly by their disease stage, but also on setting of the disease, their prior knowledge about melanoma and their way of coping. ## Need for personalised SSC from diagnosis onwards Because of these varying needs, that may change throughout the disease trajectory, HCPs indicated different options of additional care and support should be offered, tailored to the patient’s needs. Whereas some patients are doing well and consider short check-ups sufficient, others need broader SSC. Although the timing of the need for SSC varies among patients, HCPs indicated that challenges already can arise during the diagnostic phase, and therefore SSC should be offered from diagnosis onwards. I think SSC should start at the moment of diagnosis. And then, depending on how the patient reacts... some people will immediately need tools, answers and guidance. For others, that comes later. – Dermatologist, female, 41 (HCP12) Furthermore, HCPs mentioned that by adequately informing patients about their options and signalling their problems from the beginning, worsening of any occurring problems may be prevented. This in turn may result in healthcare cost reductions. ## Need for consistent information, personalised and in combination with guidance According to HCPs, information provision related to SSC should be improved. They mentioned that because current (local) guidelines of all medical specialties involved in melanoma care contain different information regarding melanoma and especially follow-up, there is a lot of practice variation in information provision. HCPs emphasised this can cause stress among patients, and suggested multiple options for improvement: updating, adjusting and streamlining relevant guidelines, adequate continuing medical education (CME) for HCPs involved in melanoma care and centralisation of information. This could ensure that all specialties involved provide equal information, within one but also between different hospitals. The surgery guideline contains different information than the dermatology guideline. That is not acceptable. That creates confusion […] Practice variation causes a lot of stress for patients. – Surgeon, male, 53 (HCP2) Furthermore, HCPs stressed the importance of tailoring information to the individual patient: content of information should differ not only between, but also within all melanoma stages since even within one disease stage, treatments and prognoses can be different. Right after diagnosis, a standard leaflet with information about melanoma in general would be sufficient, while further along the patient journey patients should be informed accordingly. Therefore, they stressed information should be tailored to the patients’ disease stage, their specific situation and their individual needs. Furthermore, since they noticed patients are often left with a lot of questions and concerns, they suggested to accompany information by actual guidance: since patients tend to forget a lot of the information received during the first and usually overwhelming consultation, they considered an extra appointment around one week after diagnosis useful for all patients. HCPs mentioned that, particularly for patients with lower stages (for whom follow-up usually consists of 1 follow-up check), putting more effort in adequate information provision around diagnosis, could prevent extra, unnecessary appointments at the dermatologist’s. Furthermore, sufficiently informing patients could prevent them from looking for (incorrect) information online. However, they indicated that it should be borne in mind that by offering patients a standard extra consultation, they can come up with many extra (unrelated) questions, which could take a lot of HCPs’ already limited time. ## Need for improved identification and treatment of disease or therapy effects in non-medical domains HCPs agreed current SSC is mainly medically focused and pays too little attention to non-medical issues (see also 1.1). They stressed more efforts should be focused on psychosocial aspects including work-related problems. Such issues should at least be identified, for example by using short questionnaires or patient reported outcome measures (PROMs). Advantages of PROMs include signalling of patients’ problems, making them easier to discuss and encourage patients to prepare discussion points before the consultation. HCPs however, indicated they experience focusing on psychosocial topics as difficult, since they lack time to discuss topics as e.g., sexuality during consultations and do not know where to refer the patient if needed. The outcomes of these questionnaires would help them to refer patients to appropriate care if necessary. What I think, and notice among colleagues: they sometimes find it difficult to bring things up because they are afraid they’ll open some sort of cesspool […] of which they think: what do I do with this? […] it starts with identification, but you have to be able to do something about it. And I think that as long as practical tools to do something are lacking, we will never start identifying properly. – Dermatologist, male, 38 (HCP19) HCPs also indicated that patients rarely actively ask for extra support themselves, but also often do not know what their options and possibilities are. They indicated a referral guide could inform patients about these options and provide clarity about where to go with (both medical and non-medical) questions and complaints, strengthening patient’s self-management. Furthermore, they highlighted the potential role of support-counsellors in guiding patients to the right non-medical care and support. In addition, HCPs saw a potential role for the patient’s GP for more counselling, in view of their (closer) relationship with the patient. ## Need for uniform and patient driven oncological follow-up Currently every HCP adheres to the (local) guidelines used in their hospital. Because these guidelines slightly differ per center, potentially causing unnecessary stress among patients, they emphasised the need for more uniform follow-up. HCPs stressed patients should be informed about their (expected) follow-up scheme to manage expectations. However, they also indicated the need for patient-driven follow-up where possible, adapted to the individual patient’s needs: they stated patients should be able to visit for skin checks at low threshold, as this contributes to regaining confidence in their skin. Furthermore, they perceived moving the oncological follow-up of low-risk patients from specialists to GPs as an option. In addition, they argued that follow-up could be offered on indication instead of standard, since currently almost no recurrences or new tumors are found during scheduled follow-up checks of these patients. Additionally, according to HCPs, more and more patients prefer to perform these short follow-up checks either at the GP or from home, because hospital visits are often associated with disadvantages (e.g. time, travel distance and costs).Distance… transport… parking costs…. time… quite a few elements that people bring up to justify to prefer coming to us [GPs] for short consultations. – General practitioner, female, 43 (HCP15) However, not all HCPs agreed on moving oncological follow-up from secondary to primary care, because in their opinion not all GPs recognise skin abnormalities correctly, not all patients trust their GP and GPs might be too busy for this. The participating GPs agreed there is considerable variability of skills between GPs and moving this responsibility to them would have to be accompanied by proper training. ## Need for improved coordination: intensified cooperation and fixed contact person Regarding coordination, HCPs stressed the importance of improving contact and cooperation between primary and secondary care, in which the GP should also play a bigger role. Although the cooperation between secondary and tertiary care is generally sufficient in melanoma care, they mentioned the feedback of information from tertiary to secondary care could be improved. Furthermore, they mentioned that patients now sometimes accidentally visit both the surgeon and dermatologist in one week, while that should be alternating. According to HCPs, one shared patient file would be the ideal solution. Moreover, they mentioned that during follow-up the role division of all involved HCPs is clear, but not afterwards, in the period when the late effects occur. According to HCPs, region-wide agreements should be made with regional multidisciplinary rounds (MDRs) and a uniform and homogeneous care pathway (see also 2.4). They indicated melanoma care should be organised as one melanoma team. By working as a team and providing patients with a clear overview of the steps that will be taken, they can take away the patient’s uncertainty. I think that if […] you present yourself as a melanoma team, you can take away some of the patient’s uncertainty. So, you could say: “I'm the dermatologist, I've removed it, I've got bad news, but someone else from our team is ready and waiting to discuss the next steps with you... they know about you, they know what I've discussed with you and then you can ask your questions following this conversation”. – Surgeon, male, 43 (HCP18) Furthermore, they suggested a fixed contact person or case manager could give patients clarity where to go with questions. They considered a doctor’s assistant or a clinical nurse specialist (CNS) as options to fulfil this role, with the prerequisite that they must be properly trained. ## Varying perceived benefits of melanoma SCPs Almost all HCPs were unfamiliar with the term SCP and were unaware that an SCP is recommended in (Dutch) melanoma guidelines, yet they perceived several benefits of melanoma SCPs. First, it could ensure that all relevant information can be found in one place, which is convenient for patients and could also prevent them from seeking (incorrect) information online. Moreover, it could contribute to more responsibility and empowerment of patients and thus increase their self-management. In addition, HCPs also saw benefits for themselves: patients receiving the information they need may have fewer questions for the HCP during and in between consultations. Moreover, it could provide HCPs with a summary of the patient’s disease and (received) treatments. This ensures always having an up-to-date overview of the patient’s situation, even if for example, feedback of information failed. One advantage of course, could be that if a lot of HCPs can access it, you’d always be up-to-date with the latest developments […] I think if someone comes to my consult and I don’t have an up-to-date letter and they say, look at my app [SCP], that says this… that could be an advantage. – General practitioner, female, 52 (HCP23). ## Should target all melanoma patients (I-IV) and medical specialists involved in melanoma care HCPs believed an SCP should be provided to all melanoma survivors given the varying problems they may experience and the associated unmet SSC needs. However, they currently considered an SCP easiest to develop and implement for stage I to III melanoma given more uniform (follow-up) trajectories. Regarding stage IV melanoma, insufficient knowledge exists on standardised, uniform follow-up: HCPs were uncertain how follow-up e.g., after treatment with immunotherapy, should be organised, which could also vary from patient to patient. Since SSC needs start from diagnosis (see also 2.1), HCPs indicated SCPs should also be provided from that point on. In terms of HCPs, they believed that the target group should include all medical specialists involved in melanoma care i.e., the dermatologist, medical oncologist, surgeon and general practitioner. ## Content should focus on all categories of SSC From the recommended categories of SSC (Table 1), HCPs considered several elements useful for inclusion in the melanoma SCP. Regarding category 1 they suggested including (links to) reliable information: about the disease and its treatment, but also on non-medical topics such as the potential psychosocial impact. The SCP should answer frequently asked questions about practical things and immediately refer to relevant information. Specifically for patients with stage III melanoma, HCPs suggested a decision aid for choosing or declining adjuvant therapy. In order to adequately detect problems of both disease and treatment (category 2), HCPs stressed that PROMS should be included in the SCP. In addition, the benefit of including a referral guide with contact information for support regarding both medical and non-medical problems was discussed. Concerning category 3, they suggested an overview of the patient’s (personal) follow-up schedule. Additionally, they suggested including a tool to take photos of skin abnormalities. According to HCPs, this could reduce fear among patients, facilitate early diagnoses and might be used for digital follow-up. Furthermore, enabling the patient to take and save photos themselves could also contribute to better cooperation between HCPs (category 4), as this would help them to distinguish between new and old (diagnosed by other HCPs) skin abnormalities. It would be nice if you can store photos in it […] ‘Where was it?’, ‘ What did it look like?’, ‘ *Is this* an in-transit metastasis or a new, second primary?’ That sort of things. It’s nice if the patient – some already do that – has the photos with them. – Surgeon, female, 37 (HCP7) In addition, HCPs indicated a tool to make (audio) recordings of consultations would be useful so patients can share them with close relatives to inform them about their disease, treatment and its (potential) impact. ## Personalised SCP tailored to stage, phase of disease and needs of individual patients According to HCPs, the SCP must meet a number of preconditions in order to be successfully implemented in practice. First, melanoma SCPs should be personalised, whereby HCPs stressed they should not only be adapted to the disease stage, but also to the treatment trajectory the patient is in and tailored to individual needs. In terms of information provision, that’s different for each stage and even within a stage […] I think that each stage, each situation requires certain information. – Oncologist, female, 41 (HCP8) In order to really meet these individual needs, the patients should be involved in the process around the SCP – both in the design and the actual activation (see also 4.4) – and be able to decide for themselves if, and how they want to use it. ## Digital SCP, integrated in the electronic health record Second, although currently a paper version of the SCP would still be convenient for people with limited digital literacy, HCPs agreed that a digital SCP would eventually work best in practice. Especially linking it to the electronic health record (EHR) would increase its use and effectiveness. Then, the patient would have all important information together at one location i.e., information about the patients’ disease and treatment, but also all other reliable (non-medical) information relevant to the patient (see also subtheme 3.2). Furthermore, linking the SCP to the EHR would make it easier (i.e., cost less time and effort) for HCPs to implement it in practice. Ideally, for example, if the HCP enters the stage and treatment in the patient’s EHR, this information should automatically appear in their SCP.If you could easily link it with the EHR […] and there you can assign something at the push of a button […] Then, during your consultations you can immediately click on something and say, okay, that needs to be added […] that would make it very easy. – Oncology nurse practitioner, male, 31 (HCP11) ## Easy to use SCP with reliable and understandable information As a third precondition, HCPs emphasised the SCP must be easy to use for both patient and HCP. It should be self-explanatory and activation of the SCP should cost as little time as possible. Furthermore, they indicated that one single tool for all diseases combined, would be the easiest to use for patients, as this would provide them with an overview of information and care for all their comorbidities. In addition, the importance of incorporating reliable information that is of high quality, but at the same time easy to understand for the patient was emphasised. I think reliable and understandable information, easy to access and easy to work with. – Oncologist, female, 41 (HCP8) ## Adequate funding to activate and provide timely updates of SCPs Thinking carefully about the financial part of the SCP and making sure there is enough funding, was mentioned as a fourth precondition for successfully implementing the SCP. In addition, consideration must be given to who should activate the SCP; HCPs believed this should be very simple, preferably fillable by the patient him/herself. However, while they indicated the patient’s characteristics could be entered by the patient, HCPs stressed that entering the patient’s medical information cannot be done by the patient alone, as this could be dangerous (e.g., a patient might think he has stage IV when it is stage II). They stressed that either the patient’s specialist (i.e., dermatologist, oncologist or oncological surgeon) or CNS should provide help in this. Moreover, they stressed the importance of keeping the SCP up-to-date, because the content could quickly become outdated. Look at where we were with melanoma five years ago. It would of course be ridiculous if the app from five years ago was still in the app store. That would even be pretty dangerous. – Support counsellor, female, 55 (HCP9) In order to achieve timely updates, for which they stressed adequate funding is necessary, they indicated the information should be linked to existing information websites that are already regularly being updated. Moreover, also for keeping the content up-to-date they saw a role for the CNS. ## Discussion This study reports HCPs’ perspectives on appropriate melanoma SSC including SCPs. The importance of personalisation was a central theme, both for SSC and SCPs. According to HCPs, melanoma SSC needs to be tailored to both characteristics (e.g., disease stage and type of treatment) and needs of individual patients. This fits the current trend towards personalisation of healthcare [27, 28] and aligns with previous research highlighting the importance of taking the individual patient' needs into account [6]. Research showed not everyone benefits from the same amount of information, where too detailed information could even have a negative effect on some patients [29]. Furthermore, as also indicated by melanoma survivors [6], HCPs stressed the need of focusing on broader SSC i.e., including psychosocial care. This is consistent with other studies both within [30–32] and outside the field of (skin)cancer [33], in which the need for non-medical care was emphasised. Our study is the first to show that HCPs share this view. An important study finding is that HCPs perceive a potentially important role of melanoma SCPs in optimising SSC and tailoring it to individual patient’s needs. That is, when moving away from the current static, non-personalised (i.e., not needs-based) models [34]. They see its added value for patients as well as for HCPs themselves, especially in offering an (up-to-date) overview of relevant information on the disease and patients’ (medical) situation including e.g., diagnosis received care and schedule of follow-up. Giving patients more control over their health data has the potential to improve their self-management [36, 37]. Moreover, it allows them to share an up-to-date overview of their situation with other HCPs. The latter is particularly important in the multidisciplinary melanoma care and could improve coordination between HCPs [8]. Furthermore, an SCP can facilitate referring survivors to appropriate psychosocial care. HCPs in our study explained that identification of psychosocial problems will never be optimal as long as it is unclear where patients can be referred to. A practical locally adapted referral guide in an SCP could provide a solution for this. Despite their promises, there are also several preconditions to be met to achieve successful implementation of SCPs. Aside from personalisation, HCPs stressed the importance of integrating the SCP in the EHR to facilitate its use in daily practice. By doing so, relevant patient data (like diagnosis and treatment) can be easily imported into the SCP and updated if the patients’ situation changes [38, 39]. Another important identified precondition is ensuring adequate long-term funding to activate, integrate and facilitate maintenance of the SCP. This aligns with previous studies emphasising the importance of sufficient organisational resources [40], as current SCPs are often not sufficiently integrated in care processes [38] thereby failing to ensure its continuity. To facilitate updating its content, HCPs in our study suggested [1] linking the SCP’s information to existing, reliable websites and [2] assigning someone, such as a CNS, to check on a timely basis whether its content is still in line with the current and constantly evolving medical knowledge and guidelines. This is consistent with previous research concluding that nurses are well-placed to provide education, care planning and support to cancer survivors [32, 41]. Our results showing that HCPs prefer a digital SCP, linked to the EHR, fits the current digital transformation of healthcare [42]. Although digital technologies such as a melanoma SCP, facilitate the delivery of personalised care [42], carefully addressing the variability in digital health literacy levels is warranted [43]. To develop an inclusive melanoma SCP, it is pivotal to tailor its content as well as adapt it to the needs of patients with lower levels of (digital) health literacy, and involve them in its development and implementation [43]. More research is needed on the most suitable ways of personalising digital technologies in (cancer) care to reach inclusive care. Although they saw the potential value of SCPs, most HCPs were not familiar with the term SCP, let alone used it in practice [14]. More awareness on the existence of SCPs among HCPs as well as on the importance of providing adequate SSC is needed. This can be done by providing CME for all HCPs involved in melanoma care. Within this CME efforts should also be focused on working agreements in melanoma care to address the perceived practice variation and lack of uniformity. However, using an SCP as an HCP would probably reduce practice variation in itself, if it for example contains a uniform follow-up plan for that patient’s stage. In line with our finding that HCPs considered current follow-up for low-risk patients too frequent, previous research suggested a less-frequent follow-up schedule than currently is recommended in the Dutch guidelines as appropriate and safe [44] and a more patient-driven follow-up model should be considered when providing personalised SSC. As emphasised by the HCPs, further knowledge is required regarding the optimal organisation of follow-up care for stage IV patients. To our knowledge, this is the first study providing an in-depth understanding of HCPs’ perspectives on appropriate SSC for patients with melanoma. We not only explored areas of needed improvement, but also provided suggestions for solutions. Previous studies investigating SSC were mostly aimed at patients with breast- or colorectal cancer, or cancer in general [12, 45]. Investigating this topic for melanoma is important as this group seems to lack proper SSC [6], despite increasing survival rates. By investigating the perspectives of a variety of HCPs, in addition to those of patients [6], a more broad, complete understanding on (needed improvement of) melanoma SSC was gained. As melanoma is one of the first metastatic diseases of which patients are starting to be considered cured after systemic treatment we believe our results can be used as blue-print for other metastatic diseases with a similar disease course (i.e., having substantially improved yet varying prognoses) and organisation of care to melanoma. In so doing, common challenges in implementing SCPs, such as adequate funding and updating its content, but also unique themes such as addressing the need for a personalised SCP, bearing in mind the prognosis switch [6], must be taken into account. Due to COVID19 pandemic restrictions, online instead of face-to-face focus groups were performed. Whereas online focus groups could have been hindered by technical problems and by participants having insufficient digital skills [46], this did not occur, presumably as participating HCPs were already experienced in meeting online and were informed in advance. Although non-verbal communication in the online setting may not as easily be picked up as compared to face-to-face, the moderators felt they had a good overview, with all participants visible in one screen, and could therefore easily pick-up non-verbal cues and respond to them. Another advantage of the online setting was that it removed barriers such as travel distance and timing, making it easier to bring together a diverse group of professionals from different centers [47]. Since all participating HCPs were not familiar with an SCP, we were not able to describe barriers to its implementation like we intended. However, our results from these discussions provided valuable preconditions which can facilitate successful implementation. Finally, although this study is set within the context of the Dutch healthcare system, we believe the identified themes are transferable to other countries, especially to those in which melanoma care is organized in similar networks as is recommended in several guidelines. In conclusion, according to HCPs, current melanoma SSC needs improvement and they emphasised the importance of offering personalised, broader (i.e., including psychosocial) care, which can be facilitated by (digital) SCPs. In addition to personalisation, inclusivity, integration in daily clinical practice, and long-term maintenance of the SCP need consideration. This will foster its implementation, long-term existence and use in clinical practice. ## Supplementary Information Additional file 1.Additional file 2. ## References 1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A. **Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries**. *CA Cancer J Clin* (2021.0) **71** 209-249. DOI: 10.3322/caac.21660 2. Leeneman B, Schreuder K, Uyl-de Groot CA, van Akkooi ACJ, Haanen J, Wakkee M. **Stage-specific trends in incidence and survival of cutaneous melanoma in the Netherlands (2003–2018): A nationwide population-based study**. *Eur J Cancer* (2021.0) **154** 111-119. DOI: 10.1016/j.ejca.2021.06.007 3. Wolchok JD, Chiarion-Sileni V, Gonzalez R, Grob JJ, Rutkowski P, Lao CD. **Long-Term Outcomes With Nivolumab Plus Ipilimumab or Nivolumab Alone Versus Ipilimumab in Patients With Advanced Melanoma**. *J Clin Oncol* (2022.0) **40** 127-137. DOI: 10.1200/JCO.21.02229 4. Weiss SA, Kluger H. **CheckMate-067: Raising the Bar for the Next Decade in Oncology**. *J Clin Oncol* (2022.0) **40** 111-113. DOI: 10.1200/JCO.21.02549 5. Gershenwald JE, Scolyer RA, Hess KR, Sondak VK, Long GV, Ross MI. **Melanoma staging: Evidence-based changes in the American Joint Committee on Cancer eighth edition cancer staging manual**. *CA Cancer J Clin.* (2017.0) **67** 472-92. DOI: 10.3322/caac.21409 6. 6.Kamminga NCW, van der Veldt AAM, Joosen MCW, de Joode K, Joosse A, Grunhagen DJ, et al. Experiences of resuming life after immunotherapy and associated survivorship care needs: a qualitative study among patients with metastatic melanoma. Br J Dermatol. 2022;187(3):381–91. 7. Husson O, Holterhues C, Mols F, Nijsten T, van de Poll-Franse LV. **Melanoma survivors are dissatisfied with perceived information about their diagnosis, treatment and follow-up care**. *Br J Dermatol* (2010.0) **163** 879-881. DOI: 10.1111/j.1365-2133.2010.09895.x 8. 8.Hewitt M, Greenfield S, Stovall E. Lost in transition. From cancer patient to cancer survivor. National Academies Press, 2005:1–506. 9. 9.Netherlands Health Council of the Netherlands. Follow-up in Oncology. Identify objectives, substantiate actions. The Hague: Health Council of the Netherlands, 2007; publication no. 2007/10. 10. 10.President’s Cancer Panel. Living Beyond Cancer: Finding a New Balance: Bethesda, MD: National Cancer Institute; 2004. 11. 11.Nederlandse Melanoomwerkgroep. Landelijke Richtlijn Melanoom. IKNL, 2019. 12. Hill RE, Wakefield CE, Cohn RJ, Fardell JE, Brierley ME, Kothe E. **Survivorship Care Plans in Cancer: A Meta-Analysis and Systematic Review of Care Plan Outcomes**. *Oncologist* (2020.0) **25** e351-e372. DOI: 10.1634/theoncologist.2019-0184 13. Mayer DK, Birken SA, Check DK, Chen RC. **Summing it up: an integrative review of studies of cancer survivorship care plans (2006–2013)**. *Cancer* (2015.0) **121** 978-996. DOI: 10.1002/cncr.28884 14. Jabson JM, Bowen DJ. **Cancer treatment summaries and follow-up care instructions: which cancer survivors receive them?**. *Cancer Causes Control* (2013.0) **24** 861-871. DOI: 10.1007/s10552-013-0163-7 15. de Rooij BH, Ezendam NP, Nicolaije KA, Vos MC, Pijnenborg JM, Boll D. **Factors influencing implementation of a survivorship care plan-a quantitative process evaluation of the ROGY Care trial**. *J Cancer Surviv* (2017.0) **11** 64-73. DOI: 10.1007/s11764-016-0562-3 16. 16.Green J, Thorogood N. Qualitative methodology in health research. Qualitative methods for health research. Fourth edition. London: SAGE Publications Ltd.; 2018. 17. Rabiee F. **Focus-group interview and data analysis**. *Proc Nutr Soc* (2004.0) **63** 655-660. DOI: 10.1079/PNS2004399 18. Tong A, Sainsbury P, Craig J. **Consolidated criteria for reporting qualitative research (COREQ): a 32-item checklist for interviews and focus groups**. *Int J Qual Health Care* (2007.0) **19** 349-357. DOI: 10.1093/intqhc/mzm042 19. 19.NCI (National Cancer Institute). About Survivorship Research: Survivorship Definitions. 2004 [Available from: http://dccps.nci.nih.gov/ocs/definitions.html. 20. Jefford M, Howell D, Li Q, Lisy K, Maher J, Alfano CM. **Improved models of care for cancer survivors**. *Lancet* (2022.0) **399** 1551-1560. DOI: 10.1016/S0140-6736(22)00306-3 21. 21.UWV. Sickness benefit [Available from: https://www.uwv.nl/en/individuals/sickness-benefit. 22. 22.Bloor M, Wood F. Keywords in Qualitative Methods: Purposive sampling. Qualitative evaluation and research methods (2nd ed). Newbury Park, CA: SAGE Publications; 2016. 23. Hennink MM, Kaiser BN, Marconi VC. **Code Saturation Versus Meaning Saturation: How Many Interviews Are Enough?**. *Qual Health Res* (2017.0) **27** 591-608. DOI: 10.1177/1049732316665344 24. Dunn J, Watson M, Aitken JF, Hyde MK. **Systematic review of psychosocial outcomes for patients with advanced melanoma**. *Psychooncology* (2017.0) **26** 1722-1731. DOI: 10.1002/pon.4290 25. Braun V, Clarke V. **Using thematic analysis in psychology**. *Qual Res Psychol* (2006.0) **3** 77-101. DOI: 10.1191/1478088706qp063oa 26. Dierckx de Casterle B, Gastmans C, Bryon E, Denier Y. **QUAGOL: a guide for qualitative data analysis**. *Int J Nurs Stud.* (2012.0) **49** 360-71. DOI: 10.1016/j.ijnurstu.2011.09.012 27. Berry L, Seiders K, Wilder S. **Innovations in access to care: a patient-centered approach**. *Ann Intern Med* (2003.0) **139** 569-574. DOI: 10.7326/0003-4819-139-7-200310070-00009 28. Moran J, Bekker H, Latchford G. **Everyday use of patient-centred, motivational techniques in routine consultations between doctors and patients with diabetes**. *Patient Educ Couns* (2008.0) **73** 224-231. DOI: 10.1016/j.pec.2008.07.006 29. de Rooij BH, Ezendam NPM, Vos MC, Pijnenborg JMA, Boll D, Kruitwagen R. **Patients' information coping styles influence the benefit of a survivorship care plan in the ROGY Care Trial: New insights for tailored delivery**. *Cancer* (2019.0) **125** 788-797. DOI: 10.1002/cncr.31844 30. 30.Bennett AE, O'Neill L, Connolly D, Guinan E, Boland L, Doyle S, et al. Perspectives of Esophageal Cancer Survivors on Diagnosis, Treatment, and Recovery. Cancers (Basel). 2020;13(1). 31. 31.de Wind A, Tamminga SJ, Bony CAG, Diether M, Ludwig M, Velthuis MJ, et al. Loss of Paid Employment up to 4 Years after Colorectal Cancer Diagnosis-A Nationwide Register-Based Study with a Population-Based Reference Group. Cancers (Basel). 2021;13(12). 32. 32.Gunn KM, Olver I, Skrabal Ross X, Harrison N, Livingston PM, Wilson C. Improving Survivors' Quality of Life Post-Treatment: The Perspectives of Rural Australian Cancer Survivors and Their Carers. Cancers (Basel). 2021;13(7). 33. Roodbeen RTJ, Lugtenberg M, Postges H, Lansink KWW, Theeuwes HP, de Jongh MAC. **Experiences of recovery and posthospital care needs of working-age adults after physical trauma: a qualitative focus group study**. *BMJ Open* (2022.0) **12** e053330. DOI: 10.1136/bmjopen-2021-053330 34. Keesing S, McNamara B, Rosenwax L. **Cancer survivors' experiences of using survivorship care plans: a systematic review of qualitative studies**. *J Cancer Surviv* (2015.0) **9** 260-268. DOI: 10.1007/s11764-014-0407-x 35. Klemanski DL, Browning KK, Kue J. **Survivorship care plan preferences of cancer survivors and health care providers: a systematic review and quality appraisal of the evidence**. *J Cancer Surviv* (2016.0) **10** 71-86. DOI: 10.1007/s11764-015-0452-0 36. Rea S, Pathak J, Savova G, Oniki TA, Westberg L, Beebe CE. **Building a robust, scalable and standards-driven infrastructure for secondary use of EHR data: the SHARPn project**. *J Biomed Inform* (2012.0) **45** 763-771. DOI: 10.1016/j.jbi.2012.01.009 37. 37.Peng C, Goswami P. Meaningful Integration of Data from Heterogeneous Health Services and Home Environment Based on Ontology. Sensors (Basel). 2019;19(8). 38. Boer H, Nuver J, Lefrandt JD, Duijzer AW, Bunskoek S, Berendsen AJ. **Easy navigating through the forest of survivorship care**. *J Clin Oncol* (2014.0) **32** 61-62. DOI: 10.1200/JCO.2013.52.9933 39. Parry C, Kent EE, Forsythe LP, Alfano CM, Rowland JH. **Can't see the forest for the care plan: a call to revisit the context of care planning**. *J Clin Oncol* (2013.0) **31** 2651-2653. DOI: 10.1200/JCO.2012.48.4618 40. Birken SA, Mayer DK, Weiner BJ. **Survivorship care plans: prevalence and barriers to use**. *J Cancer Educ* (2013.0) **28** 290-296. DOI: 10.1007/s13187-013-0469-x 41. Anbari AB, Wanchai A, Graves R. **Breast cancer survivorship in rural settings: a systematic review**. *Support Care Cancer* (2020.0) **28** 3517-3531. DOI: 10.1007/s00520-020-05308-0 42. Wyatt JC, Sullivan F. **eHealth and the future: promise or peril?**. *BMJ* (2005.0) **331** 1391-1393. DOI: 10.1136/bmj.331.7529.1391 43. Kemp E, Trigg J, Beatty L, Christensen C, Dhillon HM, Maeder A. **Health literacy, digital health literacy and the implementation of digital health technologies in cancer care: the need for a strategic approach**. *Health Promot J Austr* (2021.0) **32** 104-114. DOI: 10.1002/hpja.387 44. Deckers EA, Hoekstra-Weebers J, Damude S, Francken AB, Ter Meulen S, Bastiaannet E. **The MELFO Study: A Multicenter, Prospective, Randomized Clinical Trial on the Effects of a Reduced Stage-Adjusted Follow-Up Schedule on Cutaneous Melanoma IB-IIC Patients-Results After 3 Years**. *Ann Surg Oncol* (2020.0) **27** 1407-1417. DOI: 10.1245/s10434-019-07825-7 45. Ke Y, Ng T, Chan A. **Survivorship care models for breast cancer, colorectal cancer, and adolescent and young adult (AYA) cancer survivors: a systematic review**. *Support Care Cancer* (2018.0) **26** 2125-2141. DOI: 10.1007/s00520-018-4197-y 46. Konttila J, Siira H, Kyngas H, Lahtinen M, Elo S, Kaariainen M. **Healthcare professionals' competence in digitalisation: A systematic review**. *J Clin Nurs* (2019.0) **28** 745-761. DOI: 10.1111/jocn.14710 47. Green J, Thorogood N. *Qualitative research in a digital age* (2018.0)
--- title: Effects of biological sex and oral contraceptive pill use on cutaneous microvascular endothelial function and nitric oxide-dependent vasodilation in humans authors: - Casey G. Turner - Anna E. Stanhewicz - Karen E. Nielsen - Jeffrey S. Otis - Rafaela G. Feresin - Brett J. Wong journal: Journal of Applied Physiology year: 2023 pmcid: PMC10042598 doi: 10.1152/japplphysiol.00586.2022 license: CC BY 4.0 --- # Effects of biological sex and oral contraceptive pill use on cutaneous microvascular endothelial function and nitric oxide-dependent vasodilation in humans ## Abstract The purpose of this study was to evaluate in vivo endothelial function and nitric oxide (NO)-dependent vasodilation between women in either menstrual or placebo pill phases of their respective hormonal exposure [either naturally cycling (NC) or using oral contraceptive pills (OCPs)] and men. A planned subgroup analysis was then completed to assess endothelial function and NO-dependent vasodilation between NC women, women using OCP, and men. Endothelium-dependent and NO-dependent vasodilation were assessed in the cutaneous microvasculature using laser-Doppler flowmetry, a rapid local heating protocol (39°C, 0.1 °C/s), and pharmacological perfusion through intradermal microdialysis fibers. Data are represented as means ± standard deviation. Men displayed greater endothelium-dependent vasodilation (plateau, men: 71 ± 16 vs. women: 52 ± $20\%$CVCmax, $P \leq 0.01$), but lower NO-dependent vasodilation (men: 52 ± 11 vs. women: 63 ± $17\%$NO, $$P \leq 0.05$$) compared with all women. Subgroup analysis revealed NC women had lower endothelium-dependent vasodilation (plateau, NC women: 48 ± $21\%$CVCmax, $$P \leq 0.01$$) but similar NO-dependent vasodilation (NC women: 52 ± $14\%$NO, $P \leq 0.99$), compared with men. Endothelium-dependent vasodilation did not differ between women using OCP and men ($$P \leq 0.12$$) or NC women ($$P \leq 0.64$$), but NO-dependent vasodilation was significantly greater in women using OCP (74 ± $11\%$NO) than both NC women and men ($P \leq 0.01$ for both). This study highlights the importance of directly quantifying NO-dependent vasodilation in cutaneous microvascular studies. This study also provides important implications for experimental design and data interpretation. NEW & NOTEWORTHY This study supports differences in microvascular endothelial function and nitric oxide (NO)-dependent vasodilation between women in low hormone phases of two hormonal exposures and men. However, when separated into subgroups of hormonal exposure, women during placebo pills of oral contraceptive pill (OCP) use have greater NO-dependent vasodilation than naturally cycling women in their menstrual phase and men. These data improve knowledge of sex differences and the effect of OCP use on microvascular endothelial function. ## INTRODUCTION Microvascular function plays an important role in blood pressure and blood glucose regulation, by influencing total peripheral resistance and glucose uptake, respectively [1]. Cutaneous microvascular function is representative of systemic microvascular function [2], is easily accessible, and yields reproducible results [2]; however, basic physiology of sex differences in cutaneous microvascular responses remains incompletely understood. To date, a limited number of studies have directly compared cutaneous microvascular function between young men and women (3–7). A smaller number of studies have investigated this question in response to rapid local heating (3–5), which is a common stimulus to induce microvascular vasodilation, in part due to a large reliance on nitric oxide (NO) [8, 9]. These studies suggest similar microvascular endothelium-dependent vasodilation between men and women (3–5) but possible differences in contributing mechanisms, specifically NO [4, 5]. Additional investigation of sex differences in cutaneous microvascular function in response to local heating is warranted for experimental design, data interpretation, and translation of findings. In vascular research, premenopausal women are often restricted to experimental testing during low hormone phases, because female sex hormones may impact vascular function [10, 11]. During the natural menstrual cycle, testing is often completed during the menstrual/early follicular (M/EF) phase (days 1–7), as this corresponds to the lowest circulating levels of 17β-estradiol (E2) and progesterone. Premenopausal women in the low hormone phase of oral contraceptive pill (OCP) use (i.e., placebo pills) are often grouped together with naturally cycling (NC) women as an assumed homogenous group, as circulating concentrations of both endogenous and exogenous hormones should be low in these women during this time. However, it remains unclear if vascular function is similar between women in these hormonal exposures, especially in the microvasculature. Previous work has investigated this question in large arteries via brachial artery flow-mediated dilation [12], but large vessel function may not accurately reflect microvascular function [13, 14]. Furthermore, in the cutaneous microvasculature, when a study cohort included only NC women, NO-dependent vasodilation was lower in women compared with men [4]; however, when a study cohort included NC women and women using OCP, NO-dependent vasodilation in women did not differ from men [5]. This suggests that NO-dependent vasodilation may differ between women using OCP and NC women, regardless of testing during respective low hormone phases. The primary purpose of this study was to investigate cutaneous microvascular endothelium-dependent and NO-dependent vasodilation between women in low hormone phases of their respective hormonal exposure (either NC or using OCP) and men. A secondary purpose of this study was to compare cutaneous microvascular endothelium-dependent and NO-dependent vasodilation between subgroups of NC women, women using OCP, and men, in a planned subgroup analysis. We hypothesized that 1) the magnitude of endothelium-dependent and NO-dependent vasodilation would be similar between all women and men, 2) endothelium-dependent vasodilation would be similar between the two subgroups of women (i.e., NC women and women using OCP), and 3) NO-dependent vasodilation would be greater in women using OCP compared with NC women in their respective low hormone phases. An exploratory aim was also included to assess sex differences in cutaneous sensory nerve-mediated vasodilation, which can be assessed using the same methodology [3, 5, 9, 15, 16], for which we hypothesized that sensory nerve-mediated vasodilation would be 1) similar between all women and men and 2) similar between the two subgroups of women. ## Ethical Approval All protocols and procedures were approved by Advarra Institutional Review Board (Columbia, MD; No. Pro00024265, No. Pro00056105) and the United States Food and Drug Administration (IND 138231, IND 157532). All experimental procedures conformed with the Declaration of Helsinki. Each participant provided written and verbal consent before participating in any experimental procedure. ## Participants Participants included men ($$n = 18$$) and women ($$n = 18$$ total) who were either NC ($$n = 9$$) or using OCP ($$n = 9$$). All women using OCPs were taking monophasic, combination formulations for at least 3 mo. The average duration of OCP use was 30 ± 32 mo (range: 3–108 mo). Further details about the OCPs participants were using, including components, dosages, and generation, are shown in Table 1. Women were tested during commonly used windows to represent low hormone phases [NC, days 2–5 of natural menstrual cycle; OCP, days 1–2 of placebo pills (3–7, 17)], where circulating endogenous (e.g., NC) or exogenous (e.g., OCP) hormones should be low. NC women were tested on day 2 ($$n = 2$$), day 3 ($$n = 3$$), day 4 ($$n = 3$$), or day 5 ($$n = 1$$) of the natural menstrual cycle. Women using OCP were tested on day 1 ($$n = 3$$) or day 2 ($$n = 6$$) of placebo pills. Phase was determined by self-report and confirmed via self-report cycle tracking or presentation of OCP pack. Plasma E2 was also assessed. All women were required to submit a urine pregnancy test (McKesson hCG Combo Test Cassette, Consult Diagnostics; Richmond, VA) to confirm negative pregnancy status. Participants were recruited and included to achieve a similar proportion of White and non-White young adults within each group due to previous reports of an effect of racial identity on cutaneous microvascular function (5, 18–21). **Table 1.** | Number of Participants | EE Dose | Progestin Dose | Generation | | --- | --- | --- | --- | | 2 | 0.01 mg EE | 1 mg Norethindrone acetate | 1st | | 2 | 0.02 mg EE | 1 mg Norethindrone acetate | 1st | | 1 | 0.035 mg EE | 0.25 mg Norgestimate | 2nd | | 1 | 0.03 mg EE | 0.15 mg Desogestrel | 3rd | | 2 | 0.02 mg EE | 3 mg Drospirenone | 4th | | 1 | 0.03 mg EE | 3 mg Drospirenone | 4th | Participants were young adults (< 40 yr old, range: 18–36 yr for the entire cohort) to mitigate potential age-related declines in NO-dependent vasodilation [22] and circulating female sex hormones [23] and normotensive (systolic blood pressure < 120 mmHg and diastolic blood pressure < 80 mmHg). Self-report health history was obtained, and all participants were free of cardiovascular, pulmonary, and metabolic diseases and had no history of nerve pain/damage, cancer (chemotherapy or radiation therapy), or skin disorders (e.g., psoriasis). No participants used tobacco products, nicotine products, supplements, or medications (except women using OCPs). Further exclusion criteria included body mass index (BMI) > 30 kg/m2 that was clearly due to excessive adiposity, active COVID-19 infection, <1-mo postknown COVID-19 infection, or long-lasting symptoms after known COVID-19 infection [24]. All participants reported engaging in moderate physical activity at least 3 days/wk. Participants were asked to refrain from alcohol, vigorous exercise, caffeine, and high-fat meals for at least 8 h before the experimental protocol. ## Instrumentation Participants were seated in the semirecumbent position. The experimental arm (left) was positioned and secured at heart level to minimize the effect of hydrostatic pressure on blood flow. Skin blood flow data were analyzed at one intradermal microdialysis fiber site (CMA 31; Harvard Apparatus, Hollister, MA) on the dorsal forearm. An ice pack was used to numb the skin [25], and a 23-gauge needle was used to make an entry and exit point in the dermal layer of the skin. The microdialysis fiber was threaded through the lumen of the needle, the microdialysis membrane was left in the dermal layer, and the needle was removed. To control local skin temperature, a local heater unit (VHP1 heater units and VMS-HEAT controller; Moor Instruments, Axminster, UK) was placed directly over the microdialysis membrane. An integrated laser-Doppler probe (VP7b probes and VMS-LDF2 monitor; Moor Instruments) was placed in the center of the local heating unit to obtain red blood cell (RBC) flux, an index of skin blood flow, at the microdialysis site. Blood pressure was measured every 10 min from the contralateral arm (right) using an automated brachial oscillometric device, and heart rate was derived from pulse detection (Welch Allyn Vital Signs Series 6000; Skaneateles Falls, NY). Mean arterial pressure (MAP) was calculated as one-third pulse pressure plus diastolic pressure. ## Experimental Protocol Women participants were asked to supply a venous blood sample to assess plasma levels of E2. Blood samples were obtained by a certified phlebotomist or registered nurse via basic venipuncture and using aseptic protocols. Approximately 20 mL of blood was collected in EDTA vacutainers. Samples were centrifuged at 1,000 RPM, plasma was aliquoted into microtubules, and plasma was stored at −80°C until analyzed (estradiol ELISA kit, Product No. 501890, Cayman Chemical, Ann Arbor, MI). The microdialysis site was perfused with lactated Ringer solution (Baxter Healthcare, Deerfield, IL) through a trauma resolution period after microdialysis fiber placement (∼45–60 min), baseline data collection, and a rapid local heating protocol. The local heater unit was first set to thermoneutral temperature (33°C), and baseline skin blood flow was assessed for ∼15 min. After the baseline measurement, a rapid local heating protocol was conducted to elicit endothelium-dependent vasodilation, where local heater temperature was increased to 39°C at a rate of 0.1°C/s [8]. No participants reported pain sensations during the local heating protocol. Once a plateau in skin blood flow was achieved (∼30–40 min into local heating), 20 mM Nω-nitro-l-arginine methyl ester (l-NAME), a nonspecific NO synthase (NOS) inhibitor, was perfused through the microdialysis fiber to assess the contribution of NO to vasodilation (9, 18, 19, 26–28). Once a new plateau after l-NAME perfusion (i.e., post-l-NAME plateau) was achieved (∼30 min into l-NAME infusion), maximal vasodilation was induced by heating the skin to 43°C (0.1°C/s) and infusing 28 mM sodium nitroprusside [29]. All solutions were perfused at a rate of 2 μL/min (Beehive Controller and Baby Bee syringe pumps; Bioanalytical Systems, West Lafayette, IN). Pharmacological agents were diluted with sterile lactated Ringer solution [30] and drawn through filter needles (BD Filter Needle; Becton Dickinson, Franklin Lakes, NJ). ## Data Analysis Skin blood flow data (RBC flux) were continuously recorded at 40 Hz using commercially available hardware and software (PowerLab $\frac{16}{35}$ data acquisition and LabChart 8 software; ADInstruments, Colorado Springs, CO). Cutaneous vascular conductance (CVC) was calculated as RBC flux divided by MAP and standardized to site-specific maximal vasodilation (%CVCmax). Five main periods of skin blood flow data were analyzed: 1) baseline, 2) initial peak [i.e., sensory nerve-mediated vasodilation [9, 15, 16]], 3) plateau (i.e., endothelium-dependent vasodilation [8, 9, 31]), 4) post-l-NAME plateau (used to calculate NO-dependent vasodilation), and 5) maximal vasodilation. The initial peak is a rapid, but transient, phase, and thus ∼60 s of data were analyzed for this period. The following 3-min windows of data were analyzed for the remaining four phases: immediately preceding the onset of the local heating protocol for baseline, immediately preceding the infusion of l-NAME for the plateau, immediately preceding initiation of maximal vasodilation for the post-l-NAME plateau, and before the cessation of the experimental protocol for maximal vasodilation. NO-dependent vasodilation (%NO) was calculated as the percent change from plateau to the post-l-NAME plateau [27]. ## Statistical Analysis We completed an initial analysis comparing women and men, followed by a planned subgroup analysis. Primary outcome variables included plateau and calculated NO-dependent vasodilation and were, therefore, used to power these analyses. Sample size for both analyses was determined with a priori power analysis. Effect size was specified based on preliminary data collected in our laboratory. Our initial preliminary data suggested a large effect size ($d = 1.4$) of the mean difference in plateau between women (either NC or using OCP) in their respective low hormone phases and men (+$18\%$CVCmax in men vs. women). Assuming an α level of 0.05 and $95\%$ power, this resulted in a sample size of $$n = 14$$ per group. To protect against potential underestimation of the standard deviation (SD) in the preliminary data, we increased the sample size by $25\%$, to result in a final sample size of $$n = 18$$ per group. Sample size estimation for the subgroup analysis was based on preliminary data for NO-dependent vasodilation between subgroups of women (NC women vs. women using OCP). Our preliminary data suggested a large effect size ($d = 2.3$) of the mean difference in NO-dependent vasodilation between NC women during the menstrual phase and women using OCP during the placebo pill phase (−$31\%$NO in NC women vs. women using OCP). Assuming an α level of 0.05 and $95\%$ power, this resulted in a sample size of $$n = 7$$ per group. Sample size was, again, increased by $25\%$ in case of underestimation of the SD, to result in a final sample size of $$n = 9$$ per group. Secondary outcome variables included baseline, initial peak, post-l-NAME plateau, and maximal skin blood flow. All outcome measures were tested for equal variance between groups (Levene’s test) and normality (Shapiro–Wilk test) before analysis. Variance was not statistically different in any comparison. Only baseline, plateau, and post-l-NAME plateau data were determined to be normally distributed overall and within subgroups. Therefore, parametric statistical tests were used to analyze baseline, plateau, and post-l-NAME data (independent samples t test, one-way ANOVA), and nonparametric statistical tests were used to analyze maximal, initial peak, and %NO data (Mann–Whitney U test, Kruskal–Wallis test). Furthermore, in the subgroup analysis, Tukey’s correction factors were used to account for multiple pairwise comparisons for one-way ANOVA and Dunn’s correction factors for Kruskal–Wallis tests. In addition, an independent samples t test was used to compare plasma E2 between NC and OCP women, and a preliminary Pearson’s correlation was used to assess the potential correlation between duration of OCP use and plateau or %NO-dependent vasodilation in women using OCP. The level of significance was set at α ≤ 0.05 for all statistical tests. All data are presented as means ± SD with $95\%$ confidence intervals (CI), and all data were analyzed and graphed using commercially available software (SAS, Cary, NC and GraphPad Prism 8, San Diego, CA). ## Participant Hemodynamics Participant demographic and hemodynamic information is shown in Table 2. There was no statistical difference in measured E2 between subgroups of women (NC women: 116 ± 106 pg/mL, OCP women: 82 ± 61 pg/mL, $$P \leq 0.50$$). **Table 2.** | Unnamed: 0 | Women (n = 18) | Women (n = 18).1 | Men (n = 18) | | --- | --- | --- | --- | | | NC Women | OCP Women | Men (n = 18) | | | (n = 9) | (n = 9) | Men (n = 18) | | Age, yr | 22 ± 3 | 22 ± 3 | 25 ± 6 | | Self-identified race/ethnicity | 5 NHB, 4 NHW | 2 H, 3 NHB, 4 NHW | 10 NHB, 8 NHW | | Height, m | 1.64 ± 0.08 | 1.68 ± 0.06 | 1.78 ± 0.09 | | Mass, kg | 71.5 ± 22.2 | 70.0 ± 10.7 | 77.8 ± 12.7 | | Body mass index (BMI) kg/m2 | 25 ± 6 | 25 ± 3 | 25 ± 3 | | Resting heart rate, beats/min | 69 ± 6 | 71 ± 6 | 60 ± 7 | | Systolic blood pressure, mmHg | 111 ± 7 | 115 ± 3 | 116 ± 5 | | Diastolic blood pressure, mmHg | 72 ± 5 | 74 ± 3 | 70 ± 5 | | Mean arterial pressure, mmHg | 85 ± 5 | 87 ± 3 | 85 ± 4 | ## All Women versus Men Maximal CVC and baseline (%CVCmax) data for the complete cohort is shown in Table 3. There were no statistically significant differences between groups in maximal CVC; however, baseline was greater in men than women (Table 3). Figure 1, A–C shows data comparing all women and men. The magnitude of the plateau (Fig. 1A) was greater in men (71 ± $16\%$CVCmax; $95\%$ CI: $63\%$–$80\%$CVCmax) than in women (52 ± $20\%$CVCmax; $95\%$ CI: $43\%$–$62\%$CVCmax; $P \leq 0.01$). The post-l-NAME plateau (Fig. 1B) was greater in men (34 ± $10\%$CVCmax; $95\%$ CI: $29\%$–$39\%$CVCmax) than in women (18 ± $10\%$CVCmax; $95\%$ CI: $13\%$–$23\%$CVCmax; $P \leq 0.01$), suggesting men have greater absolute NO-independent dilation in response to local heating than women. NO-dependent vasodilation (Fig. 1C) was lower in men (52 ± $11\%$NO; $95\%$ CI: $46\%$–$57\%$NO) compared with women (63 ± $17\%$NO; $95\%$ CI: $55\%$–$72\%$NO; $$P \leq 0.05$$). **Figure 1.:** *Responses to local heating between women and men. A: plateau. B: post-l-NAME plateau. C: NO-dependent vasodilation. Purple bars, all women ($$n = 18$$). Gray bars, men ($$n = 18$$). Data are represented as means ± SD. NO-dependent vasodilation data was analyzed with Mann–Whitney test. Plateau and post-l-NAME plateau data were analyzed with independent samples t tests. Level of significance was set at 0.05. Statistical significance is indicated as a horizontal line over respective group bars and labeled with the P value of the comparison. %CVCmax, percent of maximal cutaneous vascular conductance; l-NAME, Nω-nitro-l-arginine methyl ester; NO, nitric oxide.* TABLE_PLACEHOLDER:Table 3. ## Subgroup Analysis Maximal CVC ($$P \leq 0.10$$) and baseline ($$P \leq 0.05$$, but no significant post hoc comparisons) between subgroups are shown in Table 4. There were no statistically significant differences in maximal CVC or baseline blood flow between subgroups (Table 4). Figure 2, A–C shows data between NC women, OCP women, and men. The plateau was 48 ± $21\%$CVCmax in NC women ($95\%$ CI: $32\%$–$65\%$CVCmax), 56 ± $19\%$CVCmax in women using OCP ($95\%$ CI: $42\%$–$71\%$CVCmax), and 71 ± $16\%$CVCmax in men (Fig. 2A). There was a statistically significant difference between groups for plateau ($$P \leq 0.01$$), where plateau was greater in men compared with NC women ($$P \leq 0.01$$). There was no statistically significant difference in plateau between women using OCP and men ($$P \leq 0.12$$) or NC women ($$P \leq 0.64$$). The post-l-NAME plateau was 22 ± $11\%$CVCmax in NC women ($95\%$ CI: $14\%$–$31\%$CVCmax), 14 ± $7\%$CVCmax in women using OCP ($95\%$ CI: $9\%$–$19\%$CVCmax), and 34 ± $10\%$CVCmax in men (Fig. 2B). There was a statistically significant difference between groups for post-l-NAME plateau ($P \leq 0.01$), where men displayed a greater magnitude of post-l-NAME plateau compared with NC women ($$P \leq 0.01$$) and women using OCP ($P \leq 0.01$), again suggesting men have greater absolute NO-independent dilation in response to local heating compared with both subgroups of women. There was no statistically significant difference for post-l-NAME plateau between subgroups of women ($$P \leq 0.15$$). NO-dependent vasodilation was 52 ± $14\%$NO in NC women ($95\%$ CI: $42\%$–$63\%$NO), 74 ± $11\%$NO in women using OCP ($95\%$ CI: $66\%$–$83\%$NO), and 52 ± $11\%$NO in men (Fig. 2C). There was a statistically significant difference between groups for NO-dependent vasodilation ($P \leq 0.01$), where women using OCP displayed greater %NO compared with NC women ($P \leq 0.01$) and men ($P \leq 0.01$). There was no statistically significant difference in NO-dependent vasodilation between men and NC women ($P \leq 0.99$). Furthermore, for women using OCP, there was no significant correlation between duration of OCP use and endothelium-dependent vasodilation (plateau; r = −0.30, $$P \leq 0.47$$) or NO-dependent vasodilation ($r = 0.00$, $P \leq 0.99$; data not shown). **Figure 2.:** *Responses to local heating between NC women, women using OCP, and men. A: plateau. B: post-l-NAME plateau. C: NO-dependent vasodilation. Pink bars, NC women ($$n = 9$$). Blue bars, women using OCP ($$n = 9$$). Gray bars, men ($$n = 18$$). Data are represented as means ± SD. NO-dependent vasodilation data was analyzed with Kruskal–Wallis test. Plateau and post-l-NAME plateau data were analyzed with one-way ANOVAs. Tukey’s correction factors were used to account for multiple pairwise comparisons for one-way ANOVAs and Dunn’s correction factors for Kruskal–Wallis test. Level of significance was set at 0.05. Statistical significance is indicated as a horizontal line over respective group bars and labeled with the P value of the comparison. l-NAME, Nω-nitro-l-arginine methyl ester; NC, naturally cycling; NO, nitric oxide; OCP, oral contraceptive pill.* TABLE_PLACEHOLDER:Table 4. ## Sensory Nerve-Mediated Vasodilation Figure 3 shows results addressing the exploratory aim to assess sensory nerve-mediated vasodilation in this cohort. The magnitude of the initial peak (Fig. 3A) was greater in men (58 ± $18\%$CVCmax; $95\%$ CI: $49\%$–$67\%$CVCmax) than all women (41 ± $13\%$CVCmax; $95\%$ CI: $34\%$–$47\%$CVCmax; $P \leq 0.01$). When divided into subgroups, the initial peak was 38 ± $16\%$CVCmax in NC women ($95\%$ CI: $26\%$–$50\%$CVCmax), 43 ± $10\%$CVCmax in women using OCP ($95\%$ CI: $36\%$–$51\%$CVCmax), and 58 ± $18\%$CVCmax in men (Fig. 3B). There was a statistically significant difference between groups for initial peak ($$P \leq 0.01$$), where initial peak was greater in men than NC women ($$P \leq 0.02$$). There was no statistically significant difference in initial peak between women using OCP and men ($$P \leq 0.08$$) or NC women ($P \leq 0.99$). **Figure 3.:** *Sensory nerve-mediated vasodilation in response to local heating. A: initial peak in men vs. women. Purple bars, all women (n = 18). Gray bars, men (n = 18). Data was analyzed with Mann–Whitney test. B: initial peak between subgroups. Pink bars, NC women (n = 9). Blue bars, women using OCP (n = 9). Gray bars, men (n = 18). Data was analyzed with Kruskal–Wallis test. Dunn’s correction factors were used to account for multiple pairwise comparisons for Kruskal–Wallis tests. All data are represented as means ± SD. Level of significance was set at 0.05. Statistical significance is indicated as a horizontal line over respective group bars and labeled with the P value of the comparison. %CVCmax, percent of maximal cutaneous vascular conductance; l-NAME, Nω-nitro-l-arginine methyl ester; NC, naturally cycling; NO, nitric oxide; OCP, oral contraceptive pills.* A greater magnitude of initial peak was observed in men compared with all women (Fig. 3A), suggesting greater sensory nerve-mediated [9, 15, 16] vasodilation in men. In contrast, previous studies suggest no sex differences in initial peak [3, 5]. Again, the near-maximal stimulus of 42°C may result in similar sensory nerve-mediated responses in men and women, but local heating to lower thermal stimuli (such as 39°C) may reveal physiological sex differences. ## DISCUSSION There are two key findings from this study. First, in contrast with our hypothesis, there are sex differences in microvascular responses to rapid local heating of the skin to 39°C when women are tested during the menstrual phase of the natural menstrual cycle or during placebo phase of OCP. Second, in agreement with our hypothesis, NO-dependent vasodilation is greater in women during the placebo pill phase of OCP use compared with both NC women in the menstrual phase of the natural menstrual cycle and men. These findings address current gaps in the literature that are pertinent to experimental design, data interpretation, and translation of findings. ## Overall Sex Differences In this sample of women and men, there are sex differences in microvascular responses to rapid local heating of the skin to 39°C. We observed a greater magnitude of plateau in men compared with women in low hormone phases, suggesting greater endothelium-dependent vasodilation [8, 9, 31] in men. In contrast with this study, previous studies suggest no sex differences in plateau when heating to 39°C [5] or 42°C [3, 4]. Although end temperatures of both 39°C and 42°C represent submaximal thermal stimuli, local heating to 39°C elicits ∼$50\%$CVCmax and 42°C elicits near-maximal vasodilation (∼$85\%$–$95\%$CVCmax) in the skin [32]. Therefore, when heating to 42°C, near-maximal vasodilation responses may be similar between sexes, but local heating to lower thermal stimuli may reveal physiological sex differences. The present data indirectly supports this with no observed difference in maximal CVC between groups (Tables 3 and 4). Other reports suggest there is no sex difference in the endothelium-dependent or endothelium-independent reactivity to vasodilator pharmacological stimuli in the cutaneous microvasculature, further suggesting that methodology and specific stimuli may be important considerations for interpretation of findings [7]. However, we did observe greater baseline blood flow in men compared with women in our sample (Table 3). This agrees with a previous report of greater sympathetic regulation of basal blood flow in women compared with men [33]. Therefore, women may have to overcome an initial withdrawal of sympathetic regulation that men do not and may, thus, yield a lower response to a submaximal stimulus, such as heating the skin to 39°C. Previous data suggest greater cutaneous microvascular vasodilation in White young adults compared with young adults who identify as Black or African American (5, 18–21). Therefore, it is possible the effect of biological sex observed in this study may have been influenced by racial identity/ethnicity of the participants in each group, but we were not powered to assess the effect of race/ethnicity. However, given that we recruited a similar proportion of White and non-White young adults within each subgroup, any effect due to race/ethnicity would be balanced across groups and would be unlikely to be a major explanation for our current findings. Although men exhibited a greater magnitude of the plateau to local heating than women, the magnitude of the post-l-NAME plateau was also greater in men than in women, resulting in lower NO-dependent vasodilation in men versus women. Previous reports suggest greater endothelial NOS [eNOS, the protein responsible for NO production in the skin in response to local heating [31]] expression and mRNA in female compared with male endothelial cells [34], but it is currently unclear how eNOS expression may differ between men and women within the cutaneous microvasculature. In the present study, the pattern of the local heating response observed in men relative to women suggests a greater dependence on NO-independent mechanisms, such as endothelium-derived hyperpolarizing factors (EDHF) [35], in men compared with women in low hormone phases. The present study was not designed to examine NO-independent mechanisms, and this may therefore warrant further investigation. ## Hormonal Exposures in Women The subgroup analysis revealed that differences in plateau between all women and men were largely driven by NC women (Fig. 2A), whereas differences in NO-dependent vasodilation between all women and men were largely driven by women using OCP (Fig. 2C). These findings suggest that inclusion of both NC women and women using OCP in the same sample may minimize the ability to accurately interpret cutaneous microvascular function data for sex differences, regardless of testing women during respective low hormone phases. For instance, when considering all women compared with men, the data indicated significantly greater NO-dependent vasodilation in women (63 ± $17\%$NO) compared with men (52 ± $11\%$NO, Fig. 1C). However, during the subgroup analysis, NO-dependent vasodilation between NC women during the menstrual phase and men was nearly identical within our cohort (NC women, 52 ± $14\%$NO, Fig. 2C). These findings are important for experimental design, as the inclusion of women in both hormonal exposures (in total, as well as in proportion to each other) may confound results or yield studies that are not reproducible. Present findings also indicate that NO-dependent vasodilation differs between NC women and women using OCP during their respective low hormone phases (NC women: $52\%$NO, OCP women: $74\%$NO; Fig. 2C). Although this study does not provide insight into the mechanism(s) underlying this difference in NO-dependent vasodilation, women using OCP have previously been shown to have increased eNOS mRNA expression within the skin compared with NC women [36]. There is similar evidence suggesting an increased reliance on NO in women using OCP compared with NC women, such that women using OCP showed increased vasoconstrictor responses to NOS inhibition [37]. However, this increase in eNOS expression may not coincide with increases in vasodilation responses [36]. This may be related to other reports of increased renin-angiotensin-aldosterone system activity, oxidative stress, and inflammation in women using OCP (36, 38–41) that may counteract or blunt NO effects on vascular function. Future studies assessing underlying mechanisms and these responses during high hormone exposure may provide additional insight into this question. The NO contribution was the only portion of the response for which we found a statistically significant difference between subgroups of women in this analysis. The magnitude of endothelium-dependent vasodilation (i.e., plateau) was not statistically discernible between groups (NC women, $48\%$CVCmax; women using OCP, $56\%$CVCmax). This is analogous with previous data regarding brachial artery endothelium-dependent vasodilation between NC women and women using OCP [12]. Though not statistically significant, there was a large effect size for the post-l-NAME plateau between groups of women (NC vs. OCP, mean difference, +$8\%$CVCmax, $d = 0.9$) within the present study. Collectively, the marginally greater plateau and lower post-l-NAME plateau contributed to a greater calculated NO-dependent vasodilation in women using OCP. These data underscore why it is important to quantify NO-dependent vasodilation, as solely measuring the magnitude of the plateau may not address the underlying contribution of NO. In addition, the finding of similar plateau values, yet different contributions from NO, between subgroups of women may indicate an increase in EDHF-mediated vasodilation in NC women during the menstrual phase. Although NO is implicated as the main contributor to thermal hyperemia in the skin, several pathways contribute to the response [5, 9, 18, 29, 35, 42, 43], and the contribution of mediating mechanisms has yet to be defined clearly in subgroups of women. Therefore, although grouping women in various hormonal exposures together increases the inclusion of women in research, it may hinder the ability to delineate effects of endogenous and exogenous hormones on vascular and endothelial function in women, which may have practical and/or clinical relevance. Exogenous hormone exposure is a regular and common aspect of life for premenopausal women. Indeed, ∼$28\%$ of premenopausal women currently use some form of hormonal contraceptive method in the United States, with ∼$14\%$ attributable to OCP [44]. Specific to the age range included in the present study, roughly $20\%$ of women aged 15–19 yr and 20–29 yr use OCP whereas ∼$11\%$ aged 30–39 yr use OCP [44]. Therefore, excluding all women using hormonal contraceptive methods from vascular research is also not an appropriate solution. We suggest that participant inclusion be designed with targeted research aims in mind. When research is aimed to assess population-based outcomes, it may be appropriate to include women in various hormonal states or phases, as this increases external validity. However, when research is aimed to assess between or within sex differences, especially regarding mechanistic pathways, hormonal exposures should be taken into consideration and inclusion should be planned accordingly. ## Limitations Although the sample sizes in the present study sizes determined by a priori power analysis, evaluation within a larger sample may be warranted to confirm results and investigate potential mechanisms. Furthermore, criteria to determine experimental time frames for phasic testing in premenopausal women are inconsistent within the relevant literature. During OCP use, peak circulating concentrations of exogenous hormones are typically reached within 1–2 h of oral consumption, with molecules clearing the system within 10–24 h on average [45]. Elevations in endogenous E2 and follicular maturation have been recorded toward the end of the placebo pill week of OCP use [46, 47], indicating later days within the placebo pill week may allow reawakening of the hypothalamus-pituitary-ovarian axis, accounting for rises in endogenous hormone production within 7 days. Therefore, days 1 and 2 of placebo pills in the OCP monthly cycle, as used in the present study, should correspond to low circulating concentrations of exogenous, as well as endogenous, hormones. However, it may also be warranted to assess the phenomenon measured in the present study during other days of the placebo pill week of OCP use for confirmation over the entire placebo pill period. Further limitations of this study are as follows. First, we did not perform blood analyses for fasting blood glucose or lipid profile. Therefore, we cannot objectively conclude that no participants had altered glucose or lipid status. However, all participants were normotensive and did not use any medications (except for women using OCP). Furthermore, no participants reported being diagnosed with prediabetes, type 1 or 2 diabetes, hypercholesterolemia, or dyslipidemia. Second, women did not undergo cycle tracking before experimental testing. However, the menstrual phase of the natural menstrual cycle is an easily self-tracked phase. All NC women verbally confirmed current day of cycle at experimental visits based on menstrual bleeding, which was, in all cases, confirmed by self-reported cycle tracking via mobile-based apps. Third, day/phase of the natural menstrual cycle or OCP use was self-reported. Therefore, these studies operated under the assumption that included women were honest about day/phase of their natural menstrual cycle or OCP use. In addition, experimental design relied on the assumption that women using OCP were compliant with their pill administration (i.e., not missing pills) and administered their pills/doses at consistent times across days. Fourth, women using any monophasic, combination OCP were allowed to participate in this study. It is common within the field of cutaneous microvascular research to include women using any variety of monophasic, combination OCP. Therefore, this choice in participant inclusion reflects this current practice within the field; however, investigation of the effect of other types of OCP or hormonal contraception methods on mechanisms of vascular function is warranted. Finally, this study did not employ an eNOS-specific inhibitor to quantify NO-dependent vasodilation. Therefore, it is not certain whether these results reflect eNOS function; however, previous studies suggest the eNOS isoform is largely responsible for NO production during local, nonpainful heating of the skin [31, 48, 49]. ## Implications These data highlight the importance of quantifying NO-dependent vasodilation, instead of inferring based on the overall magnitude of the plateau. Although NO may be a major contributor to the local heating response within the skin under many circumstances, it is not the sole known contributor, and an extensive breakdown of mediating mechanisms across subgroups within the populations has not been completed. Furthermore, when research is aimed to investigate underlying mechanisms of endothelial function or to deduce between or within sex differences, hormonal exposure of women participants should be considered, and inclusion should be planned carefully. ## Conclusions The submaximal thermal stimulus of 39°C elicits a greater magnitude of endothelium-dependent vasodilation in men compared with women in the menstrual phase of the natural menstrual cycle or placebo pill phase of OCP use when grouped together. Cutaneous microvascular NO-dependent vasodilation is greater in women during the placebo pill phase of monophasic, combined OCP use compared with NC women in the menstrual phase and men. However, NO-dependent vasodilation is similar between men and NC women in the menstrual phase. Data from this study further the understanding of between and within sex differences in microvascular endothelial function, as well as the potential impact of OCP on mechanisms contributing to microvascular function, highlighting areas in need of further investigation. ## DATA AVAILABILITY Data will be made available upon reasonable request. ## GRANTS This study was funded by the National Heart, Lung, and Blood Institute Grant R01 HL141205 (to B.J.W.). ## DISCLOSURES No conflicts of interest, financial or otherwise, are declared by the authors. ## AUTHOR CONTRIBUTIONS C.G.T. and B.J.W. conceived and designed research; C.G.T. performed experiments; C.G.T. and K.E.N. analyzed data; C.G.T., A.E.S., K.E.N., J.S.O., R.G.F., and B.J.W. interpreted results of experiments; C.G.T. prepared figures; C.G.T. drafted manuscript; C.G.T., A.E.S., K.E.N., J.S.O., R.G.F., and B.J.W. edited and revised manuscript; C.G.T., A.E.S., K.E.N., J.S.O., R.G.F., and B.J.W. approved final version of manuscript. ## References 1. Jonk AM, Houben AJ, de Jongh RT, Serné EH, Schaper NC, Stehouwer CD. **Microvascular dysfunction in obesity: a potential mechanism in the pathogenesis of obesity-associated insulin resistance and hypertension**. *Physiology (Bethesda)* (2007) **22** 252-260. DOI: 10.1152/physiol.00012.2007 2. Holowatz L, Thompson-Torgerson C, Kenney W. **The human cutaneous circulation as a model of generalized microvascular function**. *J Appl Physiol (1985)* (2008) **105** 370-372. DOI: 10.1152/japplphysiol.00858.2007 3. Martin Z, Shannon C, Kistler B, Nagelkir P, Del Pozzi A. **Effect of sex and menstrual cycle on skin sensory nerve contribution to local heating**. *Int J Exerc Sci* (2019) **12** 1265-1279. PMID: 31839845 4. Stanhewicz A, Greaney J, Kenney W, Alexander L. **Sex- and limb-specific differences in the nitric oxide-dependent cutaneous vasodilation in response to local heating**. *Am J Physiol Regul Integr Comp Physiol* (2014) **307** R914-R919. DOI: 10.1152/ajpregu.00269.2014 5. Patik J, Curtis B, Nasirian A, Vranish J, Fadel P, Brothers R. **Sex differences in the mechanisms mediating blunted cutaneous microvascular function in young black men and women**. *Am J Physiol Heart Circ Physiol* (2018) **315** H1063-H1071. DOI: 10.1152/ajpheart.00142.2018 6. Greaney J, Stanhewicz A, Kenney W, Alexander L. **Lack of limb or sex differences in the cutaneous vascular responses to exogenous norepinephrine**. *J Appl Physiol (1985)* (2014) **117** 1417-1423. DOI: 10.1152/japplphysiol.00575.2014 7. Gagnon D, Crandall C, Kenny G. **Sex differences in postsynaptic sweating and cutaneous vasodilation**. *J Appl Physiol (1985)* (2013) **114** 394-401. DOI: 10.1152/japplphysiol.00877.2012 8. Choi P, Brunt V, Fujii N, Minson C. **New approach to measure cutaneous microvascular function: an improved test of NO-mediated vasodilation by thermal hyperemia**. *J Appl Physiol (1985)* (2014) **117** 277-283. DOI: 10.1152/japplphysiol.01397.2013 9. MAinson C, Berry L, Joyner M. **Nitric oxide and neurally mediated regulation of skin blood flow during local heating**. *J Appl Physiol (1985)* (2001) **91** 1619-1626. DOI: 10.1152/jappl.2001.91.4.1619 10. Williams MR, Westerman RA, Kingwell BA, Paige J, Blombery PA, Sudhir K, Komesaroff PA. **Variations in endothelial function and arterial compliance during the menstrual cycle**. *J Clin Endocrinol Physiol* (2001) **86** 5389-5395. DOI: 10.1210/jcem.86.11.8013 11. Charkoudian N, Stephens D, Pirkle K, Kosiba W, Johnson J. **Influence of female reproductive hormones on local thermal control of skin blood flow**. *J Appl Physiol (1985)* (1999) **87** 1719-1723. DOI: 10.1152/jappl.1999.87.5.1719 12. Shenouda N, Priest S, Rizzuto V, MacDonald M. **Brachial artery endothelial function is stable across a menstrual and oral contraceptive pill cycle but lower in premenopausal women than in age-matched men**. *Am J Physiol Heart Circ Physiol* (2018) **315** H366-H374. DOI: 10.1152/ajpheart.00102.2018 13. Sandoo A, Carroll D, Metsios GS, Kitas GD, Veldhuijzen van Zanten JJ. **The association between microvascular and macrovascular endothelial function in patients with rheumatoid arthritis: a cross-sectional study**. *Arthritis Res Ther* (2011) **13**. DOI: 10.1186/ar3374 14. Ibrahimi K, De Graaf Y, Draijer R, Jan Danser A, VanDenBrink A, van den Meiracker A. **Reproducibility and agreement of different non-invasive methods of endothelial function assessment**. *Microvasc Res* (2018) **117** 50-56. DOI: 10.1016/j.mvr.2018.01.004 15. Magerl W, Treede R. **Heat-evoked vasodilatation in human hair skin: axon reflexes due to low-level activity of nociceptive afferents**. *J Physiol* (1996) **497** 837-848. DOI: 10.1113/jphysiol.1996.sp021814 16. Hodges G, Pozzi AD, McGarr G, Mallette M, Cheung S. **The contribution of sensory nerves to cutaneous vasodilatation of the forearm and leg to local skin heating**. *Eur J Appl Physiol* (2015) **115** 2091-2098. DOI: 10.1007/s00421-015-3188-7 17. Sims S, Heather A. **Myths and Methodologies: Reducing scientific design ambiguity in studies comparing sexes and/or menstrual cycle phases**. *Exp Physiol* (2018) **103** 1309-1317. DOI: 10.1113/EP086797 18. Miller J, Turner C, Otis J, Sebeh Y, Hayat M, Quyyumi A, Wong BJ. **Inhibition of iNOS augments cutaneous endothelial NO-dependent vasodilation in prehypertensive non-Hispanic Whites and in non-Hispanic Blacks**. *Am J Physiol Heart Circ Physiol* (2021) **320** H190-H199. DOI: 10.1152/ajpheart.00644.2020 19. Wong BJ, Turner CG, Miller JT, Walker DC, Sebeh Y, Hayat MJ, Otis JS, Quyyumi AA. **Sensory nerve-mediated and nitric oxide-dependent cutaneous vasodilation in normotensive and prehypertensive non-Hispanic blacks and whites**. *Am J Physiol Heart Circ Physiol* (2020) **319** H271-H281. DOI: 10.1152/ajpheart.00177.2020 20. Turner C, Miller J, Otis J, Hayat M, Quyyumi A, Wong B. **Cutaneous sensory nerve-mediated microvascular vasodilation in normotensive and prehypertensive non-Hispanic Blacks and Whites**. *Physiol Rep* (2020) **8**. DOI: 10.14814/phy2.14437 21. Brothers R, Fadel P, Keller D. **Racial disparities in cardiovascular disease risk: mechanisms of vascular dysfunction**. *Am J Physiol Heart Circ Physiol* (2019) **317** H777-H789. DOI: 10.1152/ajpheart.00126.2019 22. Holowatz LA, Thompson-Torgerson CS, Kenney WL. **Altered mechanisms of vasodilation in aged human skin**. *Exerc Sport Sci Rev* (2007) **35** 119-125. DOI: 10.1097/jes.0b013e3180a02f85 23. Randolph J, Sowers M, Bondarenko I, Harlow S, Luborsky J, Little R. **Change in estradiol and follicle-stimulating hormone across the early menopausal transition: effects of ethnicity and age**. *J Clin Endocrinol Metab* (2004) **89** 1555-1561. DOI: 10.1210/jc.2003-031183 24. Dillon GA, Wolf ST, Alexander LM. **Nitric oxide-mediated cutaneous microvascular function is not altered in young adults following mild-to-moderate SARS CoV-2 infection**. *Am J Physiol Heart Circ Physiol* (2022) **322** H319-H327. DOI: 10.1152/ajpheart.00602.2021 25. Hodges G, Chiu C, Kosiba W, Zhao K, Johnson J. **The effect of microdialysis needle trauma on cutaneous vascular responses in humans**. *J Appl Physiol (1985)* (2009) **106** 1112-1118. DOI: 10.1152/japplphysiol.91508.2008 26. Keen J, Levitt E, Hodges G, Wong B. **Short-term dietary nitrate supplementation augments cutaneous vasodilatation and reduces mean arterial pressure in healthy humans**. *Microvasc Res* (2015) **98** 48-53. DOI: 10.1016/j.mvr.2014.12.002 27. Wong B, Fieger S. **Transient receptor potential vanilloid type-1 (TRPV-1) channels contribute to cutaneous thermal hyperaemia in humans**. *J Physiol* (2010) **588** 4317-4326. DOI: 10.1113/jphysiol.2010.195511 28. Wolf S, Dillon G, Alexander L, Jablonski N, Kenney W. **Skin pigmentation is negatively associated with circulating vitamin D concentration and cutaneous microvascular endothelial function**. *Am J Physiol Heart Circ Physiol* (2022) **323** 490-498. DOI: 10.1152/ajpheart.00309.2022 29. Stanhewicz A, Jandu S, Santhanam L, Alexander L. **Increased angiotensin II sensitivity contributes to microvascular dysfunction in women who have had preeclampsia**. *Hypertension* (2017) **70** 382-389. DOI: 10.1161/HYPERTENSIONAHA.117.09386 30. Smith C, Craighead D, Alexander L. **Effects of vehicle microdialysis solutions on cutaneous vascular responses to local heating**. *J Appl Physiol (1985)* (2017) **123** 1461-1467. DOI: 10.1152/japplphysiol.00498.2017 31. Kellogg D, Zhao J, Wu Y. **Endothelial nitric oxide synthase control mechanisms in the cutaneous vasculature of humans in vivo**. *Am J Physiol Heart Circ Physiol* (2008) **295** H123-H129. DOI: 10.1152/ajpheart.00082.2008 32. Cracowski J, Minson C, Salvat-Melis M, Halliwill J. **Methodological issues in the assessment of skin microvascular endothelial function in humans**. *Trends Pharmacol Sci* (2006) **27** 503-508. DOI: 10.1016/j.tips.2006.07.008 33. Cooke J, Creager M, Osmundson P, Shepherd J. **Sex differences in control of cutaneous blood flow**. *Circulation* (1990) **82** 1607-1615. DOI: 10.1161/01.cir.82.5.1607 34. Cattaneo MG, Vanetti C, Decimo I, Di Chio M, Martano G, Garrone G, Bifari F, Vicentini LM. **Sex-specific eNOS activity and function in human endothelial cells**. *Sci Rep* (2017) **7**. DOI: 10.1038/s41598-017-10139-x 35. Brunt V, Minson C. **KCa channels and epoxyeicosatrienoic acids: major contributors to thermal hyperaemia in human skin**. *J Physiol* (2012) **590** 3523-3534. DOI: 10.1113/jphysiol.2012.236398 36. Cherney DZI, Scholey JW, Cattran DC, Kang AK, Zimpelmann J, Kennedy C, Lai V, Burns KD, Miller JA. **The effect of oral contraceptives on the nitric oxide system and renal function**. *Am J Physiol Renal Physiol* (2007) **293** F1539-F1544. DOI: 10.1152/ajprenal.00351.2007 37. John S, Jacobi J, Schlaich M, Delles C, Schmieder R. **Effects of oral contraceptives on vascular endothelium in premenopausal women**. *Am J Obstet Gynecol* (2000) **183** 28-33. DOI: 10.1067/mob.2000.105739 38. Cauci S, Buligan C, Marangone M, Francescato M. **Oxidative stress in female athletes using combined oral contraceptives**. *Sports Med Open* (2016) **2**. DOI: 10.1186/s40798-016-0064-x 39. Cauci S, Francescato M, Curcio F. **Combined oral contraceptives increase high-sensitivity C-reactive protein but not haptoglobin in female athletes**. *Sports Med* (2017) **47** 175-185. DOI: 10.1007/s40279-016-0534-9 40. Cauci S, Xodo S, Buligan C, Colaninno C, Barbina M, Barbina G, Francescato MP. **Oxidative stress is increased in combined oral contraceptives users and is positively associated with high-sensitivity C-reactive protein**. *Molecules* (2021) **26**. DOI: 10.3390/molecules26041070 41. Lobysheva II, van Eeckhoudt S, Dei Zotti F, Rifahi A, Pothen L, Beauloye C, Balligand J-L. **Heme-nitrosylated hemoglobin and oxidative stress in women consuming combined contraceptives. Clinical application of the EPR spectroscopy**. *Free Radic Biol Med* (2017) **108** 524-532. DOI: 10.1016/j.freeradbiomed.2017.03.039 42. Wenner M, Taylor H, Stachenfeld N. **Androgens influence microvascular dilation in PCOS through ET-A and ET-B receptors**. *Am J Physiol Endocrinol Physiol* (2013) **305** E818-E825. DOI: 10.1152/ajpendo.00343.2013 43. Hodges G, Kosiba W, Zhao K, Johnson J. **The involvement of norepinephrine, neuropeptide Y, and nitric oxide in the cutaneous vasodilator response to local heating in humans**. *J Appl Physiol (1985)* (2008) **105** 233-240. DOI: 10.1152/japplphysiol.90412.2008 44. Daniels K, Abma J. **Current contraceptive status among women aged 15-49: United States, 2017-2019**. *NCHS Data Brief* (2020) **388** 1-8. PMID: 33151146 45. Terlinden R, Uragg H, Göhler K, Kneip C. **Pharmacokinetics of chlormadinone acetate following single and multiple oral dosing of chlormadinone acetate (2 mg) and ethinylestradiol (0.03 mg) and elimination and clearance of a single dose of radiolabeled chlormadinone acetate**. *Contraception* (2006) **74** 239-244. DOI: 10.1016/j.contraception.2006.03.011 46. Baerwald A, Pierson R. **Ovarian follicular development during the use of oral contraception: a review**. *J Obstet Gynaecol Can* (2004) **26** 16-24. DOI: 10.1016/s1701-2163(16)30692-2 47. Legro RS, Pauli JG, Kunselman AR, Meadows JW, Kesner JS, Zaino RJ, Demers LM, Gnatuk CL, Dodson WC. **Effects of continuous versus cyclical oral contraception: a randomized controlled trial**. *J Clin Endocrinol Metab* (2008) **93** 420-429. DOI: 10.1210/jc.2007-2287 48. Kellogg D, Zhao J, Wu Y. **Roles of nitric oxide synthase isoforms in cutanenous vasodilation induced by local warming of the skin and whole body heat stress in humans**. *J Appl Physiol (1985)* (2009) **107** 1438-1444. DOI: 10.1152/japplphysiol.00690.2009 49. Bruning RS, Santhanam L, Stanhewicz AE, Smith CJ, Berkowitz DE, Kenney WL, Holowatz LA. **Endothelial nitric oxide synthase mediates cutaneous vasodilation during local heating and is attenuated in middle-aged human skin**. *J Appl Physiol (1985)* (2012) **112** 2019-2026. DOI: 10.1152/japplphysiol.01354.2011
--- title: Endothelin A receptor inhibition increases nitric oxide-dependent vasodilation independent of superoxide in non-Hispanic Black young adults authors: - Casey G. Turner - Matthew J. Hayat - Caroline Grosch - Arshed A. Quyyumi - Jeffrey S. Otis - Brett J. Wong journal: Journal of Applied Physiology year: 2023 pmcid: PMC10042601 doi: 10.1152/japplphysiol.00739.2022 license: CC BY 4.0 --- # Endothelin A receptor inhibition increases nitric oxide-dependent vasodilation independent of superoxide in non-Hispanic Black young adults ## Abstract Young non-Hispanic Black adults have reduced microvascular endothelial function compared with non-Hispanic White counterparts, but the mechanisms are not fully elucidated. The purpose of this study was to investigate the effect of endothelin-1 A receptor (ETAR) and superoxide on cutaneous microvascular function in young non-Hispanic Black ($$n = 10$$) and White ($$n = 10$$) adults. Participants were instrumented with four intradermal microdialysis fibers: 1) lactated Ringer’s (control), 2) 500 nM BQ-123 (ETAR antagonist), 3) 10 μM tempol (superoxide dismutase mimetic), and 4) BQ-123 + tempol. Skin blood flow was assessed via laser-Doppler flowmetry (LDF), and each site underwent rapid local heating from 33°C to 39°C. At the plateau of local heating, 20 mM l-NAME [nitric oxide (NO) synthase inhibitor] was infused to quantify NO-dependent vasodilation. Data are means ± standard deviation. NO-dependent vasodilation was decreased in non-Hispanic Black compared with non-Hispanic White young adults ($P \leq 0.01$). NO-dependent vasodilation was increased at BQ-123 sites (73 ± $10\%$ NO) and at BQ-123 + tempol sites (71 ± $10\%$NO) in non-Hispanic Black young adults compared with control (53 ± $13\%$NO, $$P \leq 0.01$$). Tempol alone had no effect on NO-dependent vasodilation in non-Hispanic Black young adults (63 ± $14\%$NO, $$P \leq 0.18$$). NO-dependent vasodilation at BQ-123 sites was not statistically different between non-Hispanic Black and White (80 ± $7\%$NO) young adults ($$P \leq 0.15$$). ETAR contributes to reduced NO-dependent vasodilation in non-Hispanic Black young adults independent of superoxide, suggesting a greater effect on NO synthesis rather than NO scavenging via superoxide. NEW & NOTEWORTHY Endothelin-1 A receptors (ETARs) have been shown to reduce endothelial function independently and through increased production of superoxide. We show that independent ETAR inhibition increases microvascular endothelial function in non-Hispanic Black young adults. However, administration of a superoxide dismutase mimetic alone and in combination with ETAR inhibition had no effect on microvascular endothelial function suggesting that, in the cutaneous microvasculature, the negative effects of ETAR in non-Hispanic Black young adults are independent of superoxide production. ## INTRODUCTION The balance between vasodilator and vasoconstrictor mechanisms is important for cardiovascular health [1]. Across the lifespan, non-Hispanic Black adults often display an imbalance in these mechanisms, characterized by blunted vasodilator (2–14) and enhanced vasoconstrictor (15–20) responses, relative to non-Hispanic White adults [21]. Likewise, the prevalence of cardiovascular disease (CVD) and several risk factors for CVD (i.e., hypertension, diabetes) is higher in non-Hispanic Black adults relative to non-Hispanic White adults [22]. Understanding mechanisms underlying this disparity in vascular function is necessary to improve cardiovascular outcomes and clinical care in non-Hispanic Black adults across the lifespan. There is a close interdependence between endothelial-derived nitric oxide (NO), a cardioprotective vasodilator [23], and endothelin-1 (ET-1), a powerful vasoconstrictor [24], often representing opposing mechanisms within the vasculature. However, ET-1 action depends on receptor subtype binding and location, where ET-1 binding to ET A (ETAR) or ET B (ETBR) receptor subtypes on vascular smooth muscle cells leads to vasoconstriction, but ET-1 binding to ETBR on endothelial cells yields vasodilation [25]. Data from our laboratory and others have shown reduced microvascular endothelium- and NO-dependent vasodilation in healthy, non-Hispanic Black young adults relative to non-Hispanic White counterparts (2–6, 9). Increased ET-1 signaling is associated with reduced NO production and increased oxidative stress (26–30), in part through increased superoxide generation [24, 27]. This phenotype is reflected in non-Hispanic Black adults across several ages and health statuses (2–6, 9, 10, 12, 31–34), and recent data [35] suggests inhibition of the ETAR can increase microvascular function in young, non-Hispanic Black and White women. However, it is currently unclear if ETAR inhibition influences microvascular function by affecting superoxide generation or if this mechanism is modified by self-identified racial background. Microvascular endothelial function can be measured via local heating of the skin. Nonpainful, rapid heating of the skin elicits robust, biphasic, and reliable vasodilation (36–38). The initial vasodilation is a rapid, transient response mediated largely by sensory nerves and TRPV-1 channels with a modest contribution from NO [37, 39, 40]. The second phase is a sustained plateau mediated largely (∼$60\%$–$80\%$) by endothelial NO-dependent mechanisms with contributions from other pathways, including endothelial-derived hyperpolarizing factors (EDHFs), adenosine receptors, and histamine receptors (37, 40–44). The ETAR subtype is functional in human skin and has been effectively inhibited during local heating (35, 45–49). Considering that cutaneous microvascular function can be used as a surrogate of systemic microvascular function [50], the present in vivo assessment of cutaneous microvascular vasodilation yields important context of the influence of ET-1 binding to the ETAR subtype and possible subsequent superoxide generation on vascular function in young, healthy non-Hispanic Black and non-Hispanic White adults. The purpose of this study was to investigate the effect of ETAR antagonism, alone and in combination with a superoxide dismutase mimetic, on endothelium-dependent vasodilation and NO-dependent vasodilation in the cutaneous microvasculature of young non-Hispanic Black and White adults. We hypothesized that ETAR antagonism would increase endothelium-dependent vasodilation and NO-dependent vasodilation in non-Hispanic Black adults. We further hypothesized that combined inhibition of ETAR and superoxide would increase NO-dependent vasodilation to a greater extent than either treatment alone. ## Ethical Approval This study was approved by Advarra Institutional Review Board (Columbia, MD; No. Pro00024265), the Georgia State University Institutional Review Board, and the United States Food and Drug Administration (IND 138231). All experimental procedures conformed with the Declaration of Helsinki. Each participant provided written and verbal informed consent before participating in any procedure. ## Participants Participant characteristics are shown in Table 1. Participants who self-identified as either non-Hispanic Black ($$n = 10$$) or non-Hispanic White ($$n = 10$$) were recruited and tested. Women ($$n = 9$$ total) were tested during the menstrual phase of the natural menstrual cycle ($$n = 4$$) or during the placebo pill phase of oral contraceptives (OCP; $$n = 5$$). Menstrual cycle and oral contraceptive pill phase were determined by self-report and confirmed by cycle tracking via phone-based apps or presentation of pill pack. Women were required to submit a urine pregnancy test (McKesson hCG Combo Test Cassette, Consult Diagnostics; Richmond, VA) to confirm negative pregnancy status. Self-report health history and health habit information was obtained (Table 1). All participants were normotensive and were free of cardiovascular, pulmonary, and metabolic diseases and had no history of nerve pain/damage, cancer (or cancer treatment), or skin disorders (e.g., psoriasis). No participants used tobacco products, nicotine products, supplements, or medications (except for women using OCP). Further exclusion criteria included active COVID-19 infection, <1-mo post-COVID-19 infection via self-report date of positive test result, and long-lasting symptoms following known COVID-19 infection. Three participants reported a positive COVID-19 test (one non-Hispanic Black man, two non-Hispanic White women). All three indicated minor symptoms (cold-like symptoms, headache), and all three tested positive >10 mo before participation in this study. Recent data suggests there is little to no reduction in cutaneous NO-dependent vasodilation in young adults with mild-to-moderate COVID-19 within ∼1–8 mo after COVID-19 infection [51]. Data from the participants who tested positive for COVID-19 were not the lowest in their respective cohort, suggesting mild COVID-19 did not affect their overall microvascular responses. **Table 1.** | Unnamed: 0 | Non-Hispanic Black Young adults (n = 10) | Non-Hispanic Black Young adults (n = 10).1 | Non-Hispanic White Young adults (n = 10) | Non-Hispanic White Young adults (n = 10).1 | | --- | --- | --- | --- | --- | | | Women (n = 4) | Men (n = 6) | Women (n = 5) | Men (n = 5) | | Age, yr | 22 ± 4 | 21 ± 2 | 21 ± 4 | 22 ± 3 | | Cycle information of women | 2 NM | | 2 NM | | | Cycle information of women | 2 OCP | | 3 OCP | | | Height, m | 1.64 ± 0.09 | 1.75 ± 0.05 | 1.61 ± 0.06 | 1.80 ± 0.05 | | Mass, kg | 56.95 ± 10.86 | 71.30 ± 14.08 | 54.05 ± 5.43 | 81.88 ± 5.88 | | Body mass index, kg/m2 | 21.45 ± 4.58 | 23.28 ± 3.72 | 20.78 ± 2.07 | 25.42 ± 2.04 | | Systolic blood pressure, mmHg | 112 ± 4 | 112 ± 7 | 112 ± 4 | 116 ± 4 | | Diastolic blood pressure, mmHg | 69 ± 3 | 68 ± 3 | 69 ± 3 | 67 ± 6 | | Mean arterial pressure, mmHg | 84 ± 3 | 83 ± 4 | 83 ± 3 | 83 ± 2 | | Heart rate, beats/min | 67 ± 6 | 61 ± 5 | 64 ± 4 | 58 ± 9 | | Positive COVID-19 test (months before participating in the study) | | 1 (11 mo) | 2 (10 and 11 mo) | | | Physical activity, min/wk | 133 ± 41 | 138 ± 25 | 130 ± 27 | 115 ± 49 | | Sleep, h/night | 7.0 ± 1.4 | 7.2 ± 0.4 | 6.8 ± 0.5 | 7.3 ± 1.5 | | Alcohol, drinks/wk | 0.5 ± 0.5 | 0.3 ± 0.5 | 1.2 ± 2.2 | 3.0 ± 2.6 | | Caffeine, drinks/wk | 7 ± 8 | 2 ± 3 | 10 ± 8 | 13 ± 3 | ## Instrumentation Participants were asked to refrain from alcohol, vigorous exercise, and caffeine for at least 8 h before the experimental protocol. For the duration of the experiment, participants were seated in the semirecumbent position, and the experimental arm was positioned and secured at heart level to minimize the effect of hydrostatic pressure on perfusion pressure and, thus, blood flow. Participants were instrumented with four microdialysis fibers (CMA 31; Harvard Apparatus, Hollister, MA) on the dorsal forearm. The microdialysis fibers had a membrane 10 mm in length with a 55 kDa molecular weight cutoff. Microdialysis fibers were gas sterilized in ethylene oxide for 24-h using Georgia State University Core Facilities (Anprolene AN74 sterilizer and AN7916 gas kit; Andersen Sterilizers; Haw River, NC). An ice pack was used to numb the skin [52], and a 23-gauge needle was then placed into the dermal layer of the skin. The microdialysis fiber was threaded through the lumen of the needle, the microdialysis membrane was left in the dermal layer, and the needle was removed. Microdialysis sites were randomly assigned to receive 1) lactated Ringer’s solution (Baxter Healthcare, Deerfield, IL) to serve as a control [53], 2) 500 nM BQ-123 [54], an ETAR antagonist (AdipoGen Life Sciences, San Diego, CA), 3) 10 μM tempol (5, 55–58), a superoxide dismutase mimetic (Sigma Aldrich, St. Louis, MO), or 4) BQ-123 + tempol. All drugs were diluted in sterile lactated Ringer’s solution and drawn through sterile filter needles (BD Filter Needle; Becton Dickinson, Franklin Lakes, NJ) or sterile syringe filters (Acrodisc, 13 mm disk, 0.2 μm pore, hydrophilic PES membrane, USP Class VI; Pall Corporation, Port Washington, NY). Stock solutions of BQ-123 and tempol were prepared, separated, and stored in sterile vials at −20°C. Stock solutions were allowed to thaw to room temperature and protected from light before use. Stock solutions were kept for no longer than 30 days and each separate vial was only used once. Trauma from microdialysis fiber placement was allowed to resolve (∼45–60 min), and drugs were infused for at least 30 min before the experimental protocol at a rate of 2 μL/min (Beehive Controller and Baby Bee syringe pumps; Bioanalytical Systems, West Lafayette, IN). To control local skin temperature, local heater units (VHP1 heater units and VMS-HEAT controller; Moor Instruments, Axminster, UK) were placed directly over each microdialysis membrane. Integrated laser-Doppler probes (VP7b probes and VMS-LDF2 monitor; Moor Instruments) were placed in the center of the local heating unit to obtain red blood cell flux, an index of skin blood flow, at each microdialysis site. Blood pressure was measured from the contralateral (right) arm using an automated brachial oscillometric device, and heart rate was derived from pulse detection (Welch Allyn Vital Signs Series 6000; Skaneatelles Falls, NY). Blood pressure and heart rate measurements were made every 10 min and mean arterial pressure (MAP) was calculated as one-third pulse pressure plus diastolic pressure. ## Experimental Protocol Local heater units were first set at a thermoneutral temperature (33°C), and baseline skin blood flow was assessed for 10–15 min. Following baseline measurements, local heater temperature was increased from 33°C to 39°C at a rate of 0.1°C/s [59]. No participants reported pain sensation during the local heating protocol. Once a plateau in skin blood flow was achieved (∼30–40 min into local heating), 20 mM Nω-nitro-l-arginine methyl ester (l-NAME; NO synthase inhibitor) was perfused through all microdialysis sites to quantify the percent contribution of NO to vasodilation [2, 4, 39, 60]. When a plateau to l-NAME (i.e., post-l-NAME plateau) was achieved at all sites (∼30 min into l-NAME infusion), maximal vasodilation was induced by heating the skin from 39°C to 43°C (0.1°C/s) and infusing 54 mM sodium nitroprusside (SNP; NO donor and endothelium-independent vasodilator) (2–4, 60). ## Data Analysis Skin blood flow data were continuously recorded at 40 Hz using commercially available hardware and software (PowerLab $\frac{16}{35}$ data acquisition and Lab Chart 8 software; ADInstruments, Colorado Springs, CO). Cutaneous vascular conductance (CVC) was calculated as red blood cell flux divided by MAP and standardized to site-specific maximal vasodilation (%CVCmax). Skin blood flow data used for analysis were averaged over a 3-min window as follows: 1) baseline immediately preceding the onset of local heating, 2) local heating plateau immediately preceding the infusion of l-NAME, 3) post-l-NAME plateau immediately preceding initiation of maximal vasodilation, and 4) maximum immediately preceding the end of the protocol. We also quantified NO-dependent vasodilation at all sites using the following equation: [(plateau magnitude – post-l-NAME plateau magnitude)/(plateau magnitude)] × 100 [39]. ## Statistical Analysis Sample size was determined with an a priori power analysis. Effect sizes were specified based on preliminary data from pilot studies completed in our laboratory. Assuming an α level of 0.05, $90\%$ power, and mean %NO-dependent vasodilation of $52\%$ (SD ± $12\%$) and $75\%$ (SD ± $14\%$) for non-Hispanic Black and White young adults, respectively, the required sample size to detect this difference in means is 8 per group. Because the standard deviation (SD) from pilot data may not accurately reflect the SD from the study population, we increased our sample size by $25\%$. Thus, our final sample size was 10 participants per group. This sample size is similar to a recent study investigating the effect of BQ-123 on cutaneous NO-dependent vasodilation in non-Hispanic Black and White young women [35]. Skin blood flow (%CVCmax) and %NO-dependent vasodilation data were analyzed using a general linear model (i.e., two-way analysis of variance) with factors for racial identity (non-Hispanic Black and non-Hispanic White) and microdialysis treatment (control, BQ-123, tempol, and BQ-123 + tempol). Tukey’s post hoc test was used to estimate pairwise comparisons. All data were analyzed and graphed using commercially available software (SAS, Cary, NC and GraphPad Prism 8, San Diego, CA). The level of significance was set at P ≤ 0.05. All data are presented as means ± SD. ## RESULTS Baseline, post-l-NAME plateau, and maximal CVC data are shown in Table 2. There was a main effect of race for maximal CVC ($$P \leq 0.01$$), but there were no other significant main or interaction effects ($P \leq 0.2$ for all). **Table 2.** | Unnamed: 0 | Non-Hispanic Black Young Adults (n = 10) | Non-Hispanic White Young Adults (n = 10) | | --- | --- | --- | | | Non-Hispanic Black Young Adults (n = 10) | Non-Hispanic White Young Adults (n = 10) | | | Non-Hispanic Black Young Adults (n = 10) | Non-Hispanic White Young Adults (n = 10) | | Baseline, %CVCmax | | | | Control | 13 ± 5 | 13 ± 6 | | BQ-123 | 21 ± 9 | 11 ± 5 | | Tempol | 18 ± 10 | 15 ± 8 | | BQ-123 + Tempol | 22 ± 8 | 11 ± 5 | | Post-l-NAME, %CVCmax | | | | Control | 22 ± 9 | 16 ± 9 | | BQ-123 | 17 ± 8 | 17 ± 12 | | Tempol | 19 ± 11 | 14 ± 5 | | BQ-123 + Tempol | 18 ± 7 | 15 ± 5 | | Maximal, CVC* | | | | Control | 2.20 ± 0.96 | 2.03 ± 0.70 | | BQ-123 | 2.32 ± 0.54 | 1.99 ± 0.64 | | Tempol | 2.63 ± 0.47 | 1.94 ± 0.97 | | BQ-123 + Tempol | 2.72 ± 0.67 | 2.05 ± 0.83 | ## Plateau Data Plateau data (i.e., endothelium-dependent vasodilation) is shown in Fig. 1. In non-Hispanic Black young adults, plateau was 50 ± $9\%$CVCmax at control sites, 70 ± $14\%$CVCmax at BQ-123 sites, 62 ± $12\%$CVCmax at tempol sites, and 64 ± $15\%$CVCmax at BQ-123 + tempol sites. In non-Hispanic White young adults, plateau was 78 ± $14\%$CVCmax at control sites, 82 ± $14\%$CVCmax at BQ-123 sites, 78 ± $15\%$CVCmax at tempol sites, and 84 ± $10\%$CVCmax at BQ-123 + tempol sites. There was not a significant interaction effect of race × treatment for plateau data ($$P \leq 0.28$$). However, there were significant main effects of self-identified race ($P \leq 0.01$) indicating the plateau was lower in non-Hispanic Black young adults relative to non-Hispanic White young adults ($P \leq 0.01$). There was also a significant main effect of treatment ($$P \leq 0.04$$) indicating ETAR inhibition with BQ-123 increased the plateau in the entire cohort compared with plateau at control sites in the entire cohort ($$P \leq 0.05$$). **Figure 1.:** *Endothelium-dependent vasodilation (i.e., plateau) between non-Hispanic Black (n = 10 participants) and non-Hispanic White (n = 10 participants) young adults. Data are presented as means ± SD. Circles represent non-Hispanic Black young adults. Triangles represent non-Hispanic White young adults. Open symbols, women. Closed symbols, men. Data were analyzed using a two-way ANOVA with factors of self-identified racial identity and treatment site. Symbols of significance represent main effects. aP ≤ 0.05 compared with all sites in non-Hispanic White young adults. bP ≤ 0.05 compared with control sites in the overall cohort. Exact P values are provided in the main text. BQ-123, endothelin A receptor antagonist; Tempol, superoxide dismutase mimetic.* ## NO-Dependent Vasodilation NO-dependent vasodilation is shown in Fig. 2. In non-Hispanic Black young adults, NO-dependent vasodilation was 53 ± $13\%$ NO at control sites, 73 ± $10\%$ NO at BQ-123 sites, 63 ± $14\%$ NO at tempol sites, and 71 ± $10\%$ NO at BQ-123 + tempol sites. In non-Hispanic White young adults, NO-dependent vasodilation was 78 ± $13\%$ NO at control sites, 80 ± $11\%$ NO at BQ-123 sites, 82 ± $7\%$ NO at tempol sites, and 82 ± $7\%$ NO at BQ-123 + tempol sites. There was a main effect of racial identity ($P \leq 0.01$), where NO-dependent vasodilation overall was lower in non-Hispanic Black relative to non-Hispanic White young adults. There was also a main effect of treatment ($P \leq 0.01$), where BQ-123 perfusion, alone or in combination with tempol, increased NO-dependent vasodilation in the entire cohort compared with control sites. Furthermore, there was a significant interaction effect of race × treatment for NO-dependent vasodilation ($$P \leq 0.04$$). NO-dependent vasodilation was lower in non-Hispanic Black relative to non-Hispanic White young adults at control ($P \leq 0.01$), tempol ($P \leq 0.01$), and BQ-123 + tempol sites ($$P \leq 0.03$$), but not at BQ-123 sites alone ($$P \leq 0.15$$). In non-Hispanic Black young adults, BQ-123 ($P \leq 0.01$) and BQ-123 + tempol ($P \leq 0.01$) increased NO-dependent vasodilation compared with respective control site, whereas tempol alone did not have a significant effect ($$P \leq 0.18$$). In non-Hispanic White young adults, no treatment significantly affected NO-dependent vasodilation. **Figure 2.:** *Nitric oxide (NO)-dependent vasodilation between non-Hispanic Black (n = 10 participants) and non-Hispanic White (n = 10 participants) young adults. Data are presented as means ± SD. Circles represent non-Hispanic Black young adults. Triangles represent non-Hispanic White young adults. Open symbols, women. Closed symbols, men. Data were analyzed using a two-way ANOVA with factors of self-identified racial identity and treatment site. Tukey’s post hoc test was used to assess pairwise comparisons. Symbols of significance represent post hoc analyses of the interaction effect. cP ≤ 0.05 compared with respective treatment site in non-Hispanic White young adults. dP ≤ 0.05 compared with control site within group. Exact P values are provided in the main text. BQ-123, endothelin A receptor antagonist; Tempol, superoxide dismutase mimetic.* ## DISCUSSION The main findings of this study are as follows: 1) ETAR antagonism increased endothelium-dependent vasodilation (plateau; Fig. 1) overall in this cohort of young, normotensive adults independent of self-identified racial group, 2) ETAR inhibition increased the percent contribution of NO in non-Hispanic Black young adults to a level not different than that observed in non-Hispanic White young adults (Fig. 2), and 3) ET-1 signaling through ETAR appears to affect endothelium-dependent vasodilation and NO-dependent vasodilation independent of superoxide in young, normotensive non-Hispanic Black adults. ## Endothelium-Dependent Vasodilation and NO-Dependent Vasodilation Results from the present study are consistent with previous data showing decreased endothelium-dependent microvascular vasodilation in young, healthy non-Hispanic Black adults relative to non-Hispanic White counterparts (2–6, 9). Neither administration of an ETAR antagonist (BQ-123) nor a superoxide dismutase mimetic (tempol) influenced endothelium-dependent vasodilation uniquely in either group, but ETAR antagonism did increase endothelium-dependent vasodilation in the overall cohort (Fig. 1). Despite a ∼$40\%$ increase in the magnitude of endothelium-dependent vasodilation in non-Hispanic Black young adults with ETAR antagonism, there was also greater variability within the response than at control sites. However, there was a greater observed maximal CVC in non-Hispanic Black compared with non-Hispanic White young adults (Table 2), and this may suggest augmented vascular smooth muscle function in non-Hispanic Black young adults, perhaps to compensate for lower endothelial function. However, this response to nitroprusside may depend on the size of the blood vessel and/or the vascular bed (e.g., muscle vs. skin) as some data have shown a reduced forearm blood flow response to nitroprusside in non-Hispanic Black adults [10]. Findings from the present study are also consistent with previous data indicating decreased contribution of NO to endothelium-dependent microvascular vasodilation in young, healthy non-Hispanic Black relative to non-Hispanic White adults (2–6, 9) (Fig. 2). ETAR antagonism increased the percentage of NO-dependent vasodilation in non-Hispanic Black young adults to a level similar to that observed in non-Hispanic White young adults at control sites (Fig. 2). This could indicate that ET-1 signaling through ETAR does not uniquely impact the overall capacity of the endothelium to elicit dilation in young, normotensive non-Hispanic Black adults but instead regulates changes in the underlying mechanisms. The present data also suggest that when using local heating of the skin to characterize endothelium-dependent vasodilator mechanisms and to interpret conclusions about NO-dependent vasodilation, it is important to inhibit NO synthase and directly quantify the contribution of NO to vasodilation, as a change (or lack thereof) in the plateau may not coincide with a change in NO-dependent vasodilation in magnitude or direction. Several pathways are known to contribute to the local heating plateau, and although NO is a major contributor, EDHFs have also been shown to make a significant contribution (∼$15\%$–$30\%$) to the plateau [59, 61]. Previous work suggests EDHF mechanisms are preserved, but NO mechanisms are reduced, in healthy, middle-aged, Black adults compared with White counterparts [10]. Therefore, it is possible EDHF mechanisms compensate for reductions in NO within healthy non-Hispanic Black young adults as well, and that this contribution from EDHF is amended when NO bioavailability increases during ETAR antagonism. ET-1 signaling through the ETAR subtype is implicated in the pathogenesis of vascular dysfunction [62, 63], and Black adults have additional alterations in the endothelin system that may contribute to vascular dysfunction and disease. Hypertensive non-Hispanic Black adults exhibit greater ETAR subtype-dependent vasoconstriction compared with hypertensive non-Hispanic White counterparts [62]. Black adults with diagnosed CVD show greater ET-converting enzyme protein levels than White counterparts [64]. Furthermore, young non-Hispanic Black men have greater resting plasma ET-1 levels and greater ET-1 generation in response to acute psychological and physiological stressors compared with young non-Hispanic White men [65]. As such, the present data support a mechanism for ETAR in blunted vasodilator responses in young, normotensive non-Hispanic Black adults. However, as this study was designed to antagonize the ETAR receptor, and not to decrease ET-1 itself, it is possible that ETAR antagonism allowed greater ET-1 signaling through endothelial ETBR, which mediates vasodilation and serves as an ET-1 clearance mechanism [45, 46, 49]. In this subset of participants, ETAR appeared to have a greater effect on NO mechanisms. As NO is cardioprotective, these findings further suggest that ET-1 signaling through ETAR may be an optimal early intervention strategy to preserve, or restore, NO mechanisms in non-Hispanic Black young adults with early onset cardiovascular risk or disease. From the present data, it is unclear how ET-1 receptor subtype expression may contribute to microvascular function in normotensive non-Hispanic Black young adults. ET-1 receptor subtype expression is altered in various conditions, such as hypertension [63, 64], coronary artery disease [66], and diabetes [67]; however, changes in receptor expression appear to differ across these conditions. A greater vasoconstrictor phenotype appears to rely more on relative expression of ETAR to ETBR and cell type distribution than a generalized increase or decrease in ETAR expression alone [63, 64, 66, 67]. Receptor subtype expression does appear to be independently related with ethnicity/identified race in disease states [64, 67], but this remains unknown in young, healthy individuals. Previous data also suggest an influence of sex hormones on ET-1 receptor-mediated responses in women [45, 46, 48, 49, 54], and ETAR antagonism appears to improve cutaneous microvascular endothelium-dependent and NO-dependent vasodilation in both young Black and White women [35]. Furthermore, there appear to be mechanistic differences in the regulation of vascular function between non-Hispanic Black men and women [9, 68]. However, ET-1 receptor-mediated vascular responses have not been directly investigated for sex differences within the cutaneous microvasculature, and it is currently unclear how sex and racial identity may interact to affect these responses. Although there did not appear to be stark sex differences within our sample (symbols in Figs. 1 and 2 delineated to represent men and women), this study was not powered to analyze sex differences. Nevertheless, a fully powered study exploring sex differences in the effect of ETAR in microvascular function in non-Hispanic Black and non-Hispanic White young adults is warranted. ## The Role of Superoxide Some of the detrimental properties of ET-1 have been attributed to its propensity to increase superoxide production by activating nicotinamide adenine dinucleotide phosphate (NADPH) oxidase (NOX) and/or inducing eNOS uncoupling [24, 27]. In the cutaneous microvasculature, superoxide contributes to reductions in NO-dependent vasodilation in non-Hispanic Black young adults relative to non-Hispanic White young adults [5]. The present study included administration of a superoxide dismutase mimetic (tempol), both alone and in combination with ETAR antagonism, to assess the related and independent effects of superoxide within this cohort. Tempol alone did not change either endothelium-dependent or NO-dependent vasodilation compared with respective control sites in either group (Figs. 1 and 2). Furthermore, tempol + BQ-123 did not differ from BQ-123 alone sites for either endothelium-dependent or NO-dependent vasodilation in either group (Figs. 1 and 2). It is unclear why tempol had no effect in the present study. Our data are in contrast with data previously reported by Hurr et al. [ 5], which showed an increase in local heating plateau and NO contribution [reported as the difference (Δ) between the local heating plateau and the post-l-NAME plateau] with tempol in Black compared with White young adults. Of note, the present study reports calculated %NO-dependent vasodilation (equation in methods). However, the difference between the local heating plateau and the post-l-NAME plateau (i.e., Δ) at tempol sites in non-Hispanic Black young adults in our study (43 ± $13\%$CVCmax) is similar to that reported for Black young adults in the Hurr et al. study (48 ± $25\%$CVCmax) [5]. Furthermore, the magnitude of the local heating plateau in non-Hispanic Black young adults in the present study at control (50 ± $9\%$CVCmax) and tempol (62 ± $12\%$CVCmax) sites are also similar to those of Black young adults in the study from Hurr et al. [ 5] (control: 47 ± $15\%$ CVCmax, tempol: 63 ± $27\%$ CVCmax). Therefore, minor differences in observed values between these studies may account for differing statistical outcomes, but a similar mechanistic foundation is suggested by both studies. Previous findings suggest NOX and xanthine oxidase (XO) as major sources of superoxide generation in young non-Hispanic Black adults, though there may be differences between young Black men and women [9]. Patik et al. [ 9] reported increases in local heating plateau and NO contribution (Δ) to cutaneous microvascular vasodilation in Black young adults with inhibition of NOX and XO. Patik et al. [ 9] also found that young Black men responded to NOX and XO inhibition whereas young Black women did not. Although the present study was not powered to determine sex differences, both young Black women (plateau: $30\%$ increase from control, %NO: $14\%$ increase from control) and young Black men (plateau: $11\%$ increase from control, %NO: $19\%$ increase from control) appeared to respond positively to tempol administration. Furthermore, Akins et al. [ 35] recently found ETAR antagonism with BQ-123 to increase plateau and NO contribution (Δ) in both Black and White young women. These data collectively suggest ETAR and superoxide affect microvascular endothelial function and NO-dependent vasodilation independently in non-Hispanic Black young adults. However, it is unclear from the present study whether ET-1 signaling through ETAR affects NO bioavailability more greatly via changes in NO synthesis and/or eNOS expression, phosphorylation, or coupling, or through NO scavenging via reactive oxygen species in normotensive, non-Hispanic Black young adults. Regardless, trends within our data further supports the necessity of direct investigation of sex differences in the effect of ETAR in Black young adults. ## Limitations There are a few limitations to this study that warrant consideration. First, social determinants of health are known to affect physiological mechanisms and responses and could contribute to the observations in the present study. Recent data from Wolf et al. [ 69] showed that social economic status (SES) was lower in a cohort of young African American participants relative to European American participants. However, there was no significant association between SES and cutaneous vasodilation in response to local heating [69]. We did not directly assess SES or other social determinants of health but recognize that differences in these variables could account for the observed physiology. We did assess some health habits (Table 1), but these were self-report measures and were only assessed on the day of the study. Nevertheless, these self-report measures indicate groups were well-matched for these select self-report health habits. To fully assess a potential association between social determinants of health and microvascular responses would require a substantially larger participant cohort. Second, we recruited participants who self-identified their racial identity as either non-Hispanic Black or non-Hispanic White, as genotyping participants was beyond the scope of this study. We excluded participants who reported multiracial/ethnic and Hispanic backgrounds, but, because we relied on self-identification, it is possible there were participants of multiracial/ethnic or Hispanic lineage. Third, we did not perform blood analyses for fasting blood glucose or lipid profile. All participants were normotensive, and no participants reported having type 1 or 2 diabetes, hypercholesterolemia, or dyslipidemia. Furthermore, the magnitude of responses in both groups was similar to results of previous studies where glucose and lipids were analyzed [5, 9]. Therefore, it is unlikely that our findings are the result of altered glucose or lipid status; however, we cannot objectively conclude that no participants had altered glucose or lipid status. ## Conclusions This study presents evidence of a role for ETAR in reduced NO-dependent vasodilation in young, normotensive non-Hispanic Black adults compared with non-Hispanic White adults. Furthermore, this study suggests that ET-1 signaling through ETAR affects NO-dependent vasodilation independent of superoxide in this sample. Therefore, as NO is cardioprotective, upregulated ET-1 signaling through ETAR may be an early characteristic of cardiovascular risk in non-Hispanic Black young adults. Likewise, ETAR may be a possible target intervention for non-Hispanic Black young adults with early onset cardiovascular disease or vascular dysfunction. ## DATA AVAILABILITY Data will be made available on reasonable request. ## GRANTS This study was funded by NIH grant R01 HL141205 (to B.J.W.). ## DISCLOSURES No conflicts of interest, financial or otherwise, are declared by the authors. ## AUTHOR CONTRIBUTIONS C.G.T., A.A.Q., J.S.O., and B.J.W. conceived and designed research; C.G.T. performed experiments; C.G.T., M.J.H., C.G., and B.J.W. analyzed data; C.G.T., M.J.H., C.G., and B.J.W. interpreted results of experiments; C.G.T. and M.J.H. prepared figures; C.G.T. drafted manuscript; C.G.T., M.J.H., C.G., A.A.Q., J.S.O., and B.J.W. edited and revised manuscript; C.G.T., M.J.H., C.G., A.A.Q., J.S.O., and B.J.W. approved final version of manuscript. ## References 1. Spieker L, Flammer A, Lüscher T. **The vascular endothelium in hypertension**. *Handb Exp Pharmacol* (2006) **176** 249-283. DOI: 10.1007/3-540-36028-x_8 2. Miller J, Turner C, Otis J, Sebeh Y, Hayat M, Quyyumi A, Wong BJ. **Inhibition of iNOS augments cutaneous endothelial NO-dependent vasodilation in prehypertensive non-Hispanic Whites and in non-Hispanic Blacks**. *Am J Physiol Heart Circ Physiol* (2021) **320** H190-H199. DOI: 10.1152/ajpheart.00644.2020 3. Turner C, Miller J, Otis J, Hayat M, Quyyumi A, Wong B. **Cutaneous sensory nerve-mediated microvascular vasodilation in normotensive and prehypertensive non-Hispanic Blacks and Whites**. *Physiol Rep* (2020) **8**. DOI: 10.14814/phy2.14437 4. Wong B, Turner C, Miller J, Walker D, Sebeh Y, Hayat M, Otis JS, Quyyumi AA. **Sensory nerve-mediated and nitric oxide-dependent cutaneous vasodilation in normotensive and prehypertensive non-Hispanic blacks and whites**. *Am J Physiol Heart Circ Physiol* (2020) **319** H271-H281. DOI: 10.1152/ajpheart.00177.2020 5. Hurr C, Patik J, Kim K, Christmas K, Brothers R. **Tempol augments the blunted cutaneous microvascular thermal reactivity in healthy young African Americans**. *Exp Physiol* (2018) **103** 343-349. DOI: 10.1113/EP086776 6. Kim K, Hurr C, Patik J, Brothers R. **Attenuated cutaneous microvascular function in healthy young African Americans: role of intradermal**. *Microvasc Res* (2018) **118** 1-6. DOI: 10.1016/j.mvr.2018.02.001 7. Cardillo C, Kilcoyne CM, Cannon RO, Panza JA. **Racial differences in nitric oxide-mediated vasodilator response to mental stress in the forearm circulation**. *Hypertension* (1998) **31** 1235-1239. DOI: 10.1161/01.hyp.31.6.1235 8. Pienaar P, Micklesfield L, Gill J, Shore A, Gooding K, Levitt N, Lambert EV. **Ethnic differences in microvascular function in apparently healthy South African men and women**. *Exp Physiol* (2014) **99** 985-994. DOI: 10.1113/expphysiol.2014.078519 9. Patik J, Curtis B, Nasirian A, Vranish J, Fadel P, Brothers R. **Sex differences in the mechanisms mediating blunted cutaneous microvascular function in young black men and women**. *Am J Physiol Heart Circ Physiol* (2018) **315** H1063-H1071. DOI: 10.1152/ajpheart.00142.2018 10. Ozkor M, Rahman A, Murrow J, Kavtaradze N, Lin J, Manatunga A, Hayek S, Quyyumi AA. **Differences in vascular nitric oxide and endothelium-derived hyperpolarizing factor bioavailability in blacks and whites**. *Arterioscler Thromb Vasc Biol* (2014) **34** 1320-1327. DOI: 10.1161/ATVBAHA.113.303136 11. Morris A, Patel R, Binongo J, Poole J, Al Mheid I, Ahmed Y, Stoyanova N, Vaccarino V, Din-Dzietham R, Gibbons GH, Quyyumi A. **Racial differences in arterial stiffness and microcirculatory function between Black and White Americans**. *J Am Heart Assoc* (2013) **2**. DOI: 10.1161/JAHA.112.002154 12. Melikian N, Wheatcroft S, Ogah O, Murphy C, Chowienczyk P, Wierzbicki A, Sanders TA, Jiang B, Duncan ER, Shah AM, Kearney MT. **Asymmetric dimethylarginine and reduced nitric oxide bioavailability in young black African men**. *Hypertension* (2007) **49** 873-877. DOI: 10.1161/01.HYP.0000258405.25330.80 13. Campia U, Choucair W, Bryant M, Waclawiw M, Cardillo C, Panza J. **Reduced endothelium-dependent and -independent dilation of conductance arteries in African Americans**. *J Am Coll Cardiol* (2002) **40** 754-760. DOI: 10.1016/s0735-1097(02)02015-6 14. Strain W, Chaturvedi N, Nihoyannopoulos P, Bulpitt C, Rajkumar C, Shore A. **Differences in the association between type 2 diabetes and impaired microvascular function among Europeans and African Caribbeans**. *Diabetologia* (2005) **48** 2269-2277. DOI: 10.1007/s00125-005-1950-9 15. Ray C, Monahan K. **Sympathetic vascular transduction is augmented in young normotensive blacks**. *J Appl Physiol (1985)* (2002) **92** 651-656. DOI: 10.1152/japplphysiol.00788.2001 16. Stein C, Lang C, Singh I, He H, Wood A. **Increased vascular adrenergic vasoconstriction and decreased vasodilation in blacks. Additive mechanisms leading to enhanced vascular reactivity**. *Hypertension* (2000) **36** 945-951. DOI: 10.1161/01.hyp.36.6.945 17. Vranish J, Holwerda S, Young B, Credeur D, Patik J, Barbosa T, Keller DM, Fadel PJ. **Exaggerated vasoconstriction to spontaneous bursts of muscle sympathetic activity in healthy young black men**. *Hypertension* (2018) **71** 192-198. DOI: 10.1161/HYPERTENSIONAHA.117.10229 18. Adefurin A, Ghimire L, Kohli U, Muszkat M, Sofowora G, Paranjape S, Stein CM, Kurnik D. **Blacks have a greater sensitivity to α1-adrenoceptor-mediated venoconstriction compared with whites**. *Hypertension* (2013) **61** 915-920. DOI: 10.1161/HYPERTENSIONAHA.111.00854 19. Sherwood A, Hill L, Blumenthal J, Johnson K, Hinderliter A. **Race and sex differences in cardiovascular alpha-adrenergic and beta-adrenergic receptor responsiveness in men and women with high blood pressure**. *J Hypertens* (2017) **35** 975-981. DOI: 10.1097/HJH.0000000000001266 20. Thomas K, Nelesen R, Malcarne V, Ziegler M, Dimsdale J. **Ethnicity, perceived discrimination, and vascular reactivity to phenylephrine**. *Psychosom Med* (2006) **68** 692-697. DOI: 10.1097/01.psy.0000238214.80871.e6 21. Brothers RM, Fadel PJ, Keller DM. **Racial disparities in cardiovascular disease risk: mechanisms of vascular dysfunction**. *Am J Physiol Heart Circ Physiol* (2019) **317** H777-H789. DOI: 10.1152/ajpheart.00126.2019 22. Virani S, Alonso A, Aparicio H, Benjamin E, Bittencourt M, Callaway C. **Heart Disease and Stroke Statistics—2021 Update: a report from the American Heart Association**. *Circulation* (2021) **143** e254-e743. DOI: 10.1161/CIR.0000000000000950 23. Vallance P, Chan N. **Endothelial function and nitric oxide: clinical relevance**. *Heart* (2001) **85** 342-350. DOI: 10.1136/heart.85.3.342 24. Pollock D, Pollock J. **Endothelin and oxidative stress in the vascular system**. *Curr Vasc Pharmacol* (2005) **3** 365-367. DOI: 10.2174/157016105774329408 25. Khalil R. **Modulators of the vascular endothelin receptor in blood pressure regulation and hypertension**. *Curr Mol Pharmacol* (2012) **4** 176-186. DOI: 10.2174/1874467211104030176 26. Wedgwood S, Dettman R, Black S. **ET-1 stimulates pulmonary arterial smooth muscle cell proliferation via induction of reactive oxygen species**. *Am J Physiol Lung Cell Mol Physiol* (2001) **281** L1058-L1067. DOI: 10.1152/ajplung.2001.281.5.L1058 27. Loomis ED, Sullivan J, Osmond D, Pollock D, Pollock J. **Endothelin mediates superoxide production and vasoconstriction through activation of NADPH oxidase and uncoupled nitric-oxide synthase in the rat aorta**. *J Pharmacol Exp Ther* (2005) **315** 1058-1064. DOI: 10.1124/jpet.105.091728 28. Wedgwood S, Black S. **Endothelin-1 decreases endothelial NOS expression and activity through ETA receptor-mediated generation of hydrogen peroxide**. *Am J Physiol Lung Cell Mol Physiol* (2005) **288** L480-L487. DOI: 10.1152/ajplung.00283.2004 29. Sud N, Black S. **Endothelin-1 impairs nitric oxide signaling in endothelial cells through a protein kinase Cδ-dependent activation of STAT3 and decreased endothelial nitric oxide synthase expression**. *DNA Cell Biol* (2009) **28** 543-553. DOI: 10.1089/dna.2009.0865 30. Ramzy D, Rao V, Tumiati L, Xu N, Sheshgiri R, Miriuka S, Delgado DH, Ross HJ. **Elevated endothelin-1 levels impair nitric oxide homeostasis through a PKC-dependent pathway**. *Circulation* (2006) **114** I319-I326. DOI: 10.1161/CIRCULATIONAHA.105.001503 31. Kalinowski L, Dobrucki I, Malinski T. **Race-specific differences in endothelial function: predisposition of African Americans to vascular diseases**. *Circulation* (2004) **109** 2511-2517. DOI: 10.1161/01.CIR.0000129087.81352.7A 32. Morris A, Zhao L, Patel R, Jones D, Ahmed Y, Stoyanova N, Gibbons GH, Vaccarino V, Din-Dzietham R, Quyyumi A. **Differences in systemic oxidative stress based on race and metabolic syndrome: the Morehouse and Emory Team up to Eliminate Health Disparities (META-Health) Study**. *Metab Syndr Relat Disord* (2012) **10** 252-259. DOI: 10.1089/met.2011.0117 33. Deo S, Holwerda S, Keller D, Fadel P. **Elevated peripheral blood mononuclear cell-derived superoxide production in healthy young black men**. *Am J Physiol Heart Circ Physiol* (2015) **308** H548-H552. DOI: 10.1152/ajpheart.00784.2014 34. Feairheller D, Park J-Y, Sturgeon K, Williamson S, Diaz K, Veerabhadrappa P, Brown MD. **Racial difference in oxidative stress and inflammation: in vitro and in vivo**. *Clin Transl Sci* (2011) **4** 32-37. DOI: 10.1111/j.1752-8062.2011.00264.x 35. Akins J, Richey R, Campbell J, Martin Z, Olvera G, Brothers R. **Contributions of endothelin-1 and L-arginine to blunted cutaneous microvascular function in young, black women**. *Am J Physiol Heart Circ Physiol* (2022) **322** H260-H268. DOI: 10.1152/ajpheart.00457.2021 36. Kenney WL. **Edward F. Adolph Distinguished Lecture: skin-deep insights into vascular aging**. *J Appl Physiol (1985)* (2017) **123** 1024-1038. DOI: 10.1152/japplphysiol.00589.2017 37. Minson CT, Berry LT, Joyner MJ. **Nitric oxide and neurally mediated regulation of skin blood flow during local heating**. *J Appl Physiol (1985)* (2001) **91** 1619-1626. DOI: 10.1152/jappl.2001.91.4.1619 38. Kellogg DL, Liu Y, Kosiba IF, O'Donnell D. **Role of nitric oxide in the vascular effects of local warming of the skin in humans**. *J Appl Physiol (1985)* (1999) **86** 1185-1190. DOI: 10.1152/jappl.1999.86.4.1185 39. Wong BJ, Fieger SM. **Transient receptor potential vanilloid type-1 (TRPV-1) channels contribute to cutaneous thermal hyperaemia in humans**. *J Physiol* (2010) **588** 4317-4326. DOI: 10.1113/jphysiol.2010.195511 40. Wong BJ, Minson CT. **Altered thermal hyperaemia in human skin by prior desensitization of neurokinin-1 receptors**. *Exp Physiol* (2011) **96** 599-609. DOI: 10.1113/expphysiol.2011.057356 41. Fieger SM, Wong BJ. **Adenosine receptor inhibition with theophylline attenuates the skin blood flow response to local heating in humans**. *Exp Physiol* (2010) **95** 946-954. DOI: 10.1113/expphysiol.2010.053538 42. Kellogg DL, Zhao JL, Wu Y. **Endothelial nitric oxide synthase control mechanisms in the cutaneous vasculature of humans in vivo**. *Am J Physiol Heart Circ Physiol* (2008) **295** H123-H129. DOI: 10.1152/ajpheart.00082.2008 43. Kellogg DL, Zhao JL, Wu Y. **Roles of nitric oxide synthase isoforms in cutaneous vasodilation induced by local warming of the skin and whole body heat stress in humans**. *J Appl Physiol (1985)* (2009) **107** 1438-1444. DOI: 10.1152/japplphysiol.00690.2009 44. Wong BJ, Williams SJ, Minson CT. **Minimal role for H1 and H2 histamine receptors in cutaneous thermal hyperemia to local heating in humans**. *J Appl Physiol (1985)* (2006) **100** 535-540. PMID: 16195389 45. Wenner M, Taylor H, Stachenfeld N. **Androgens influence microvascular dilation in PCOS through ET-A and ET-B receptors**. *Am J Physiol Endocrinol Physiol* (2013) **305** E818-E825. DOI: 10.1152/japplphysiol.00902.2005 46. Wenner M, Sebzda K, Kuczmarski A, Pohlig R, Edwards D. **ETB receptor contribution to vascular dysfunction in postmenopausal women**. *Am J Physiol Regul Integr Comp Physiol* (2017) **313** R51-R57. DOI: 10.1152/ajpregu.00410.2016 47. Stanhewicz A, Jandu S, Santhanam L, Alexander L. **Alterations in endothelin type B receptor contribute to microvascular dysfunction in women who have had preeclampsia**. *Clin Sci (Lond)* (2017) **131** 2777-2789. DOI: 10.1042/CS20171292 48. Sebzda K, Kuczmarski A, Pohlig R, Lennon S, Edwards D, Wenner M. **Ovarian hormones modulate endothelin-1 receptor responses in young women**. *Microcirculation* (2018) **25**. DOI: 10.1111/micc.12490 49. Shoemaker L, Haigh K, Kuczmarski A, McGinty S, Welti L, Hobson J, Edwards DG, Feinberg RF, Wenner MM. **ET**. *Am J Physiol Heart Circ Physiol* (2021) **321** H592-H598. DOI: 10.1152/ajpheart.00087.2021 50. Holowatz L, Thompson-Torgerson C, Kenney W. **The human cutaneous circulation as a model of generalized microvascular function**. *J Appl Physiol (1985)* (2008) **105** 370-372. DOI: 10.1152/japplphysiol.00858.2007 51. Dillon GA, Wolf ST, Alexander LM. **Nitric oxide-mediated cutaneous microvascular function is not altered in young adults following mild-to-moderate SARS CoV-2 infection**. *Am J Physiol Heart Circ Physiol* (2022) **322** H319-H327. DOI: 10.1152/ajpheart.00602.2021 52. Hodges G, Chiu C, Kosiba W, Zhao K, Johnson J. **The effect of microdialysis needle trauma on cutaneous vascular responses in humans**. *J Appl Physiol (1985)* (2009) **106** 1112-1118. DOI: 10.1152/japplphysiol.91508.2008 53. Smith C, Craighead D, Alexander L. **Effects of vehicle microdialysis solutions on cutaneous vascular responses to local heating**. *J Appl Physiol (1985)* (2017) **123** 1461-1467. DOI: 10.1152/japplphysiol.00498.2017 54. Wenner M, Taylor H, Stachenfeld N. **Endothelin B receptor contribution to peripheral microvascular function in women with polycystic ovary syndrome**. *J Physiol* (2011) **589** 4671-4679. DOI: 10.1113/jphysiol.2011.216218 55. Fujii N, Brunt V, Minson C. **Tempol improves cutaneous thermal hyperemia through increasing nitric oxide bioavailability in young smokers**. *Am J Physiol Heart Circ Physiol* (2014) **306** H1507-H1511. DOI: 10.1152/ajpheart.00886.2013 56. DuPont J, Ramick M, Farquhar W, Townsend R, Edwards D. **NADPH oxidase-derived reactive oxygen species contribute to impaired cutaneous microvascular function in chronic kidney disease**. *Am J Physiol Renal Physiol* (2014) **306** F1499-F1506. DOI: 10.1152/ajprenal.00058.2014 57. Medow M, Aggarwal A, Baugham I, Messer Z, Stewart J. **Modulation of the axon-reflex response to local heat by reactive oxygen species in subjects with chronic fatigue syndrome**. *J Appl Physiol (1985)* (2013) **114** 45-51. DOI: 10.1152/japplphysiol.00821.2012 58. Medow M, Bamji N, Clarke D, Ocon A, Stewart J. **Reactive oxygen species (ROS) from NADPH and xanthine oxidase modulate the cutaneous local heating response in healthy humans**. *J Appl Physiol (1985)* (2011) **111** 20-26. DOI: 10.1152/japplphysiol.01448.2010 59. Choi P, Brunt V, Fujii N, Minson C. **New approach to measure cutaneous microvascular function: an improved test of NO-mediated vasodilation by thermal hyperemia**. *J Appl Physiol (1985)* (2014) **117** 277-283. DOI: 10.1152/japplphysiol.01397.2013 60. Keen J, Levitt E, Hodges G, Wong B. **Short-term dietary nitrate supplementation augments cutaneous vasodilatation and reduces mean arterial pressure in healthy humans**. *Microvasc Res* (2015) **98** 48-53. DOI: 10.1016/j.mvr.2014.12.002 61. Brunt V, Minson C. **KCa channels and epoxyeicosatrienoic acids: major contributors to thermal hyperaemia in human skin**. *J Physiol* (2012) **590** 3523-3534. DOI: 10.1113/jphysiol.2012.236398 62. Campia U, Cardillo C, Panza J. **Ethnic differences in the vasoconstrictor activity of endogenous endothelin-1 in hypertensive patients**. *Circulation* (2004) **109** 3191-3195. DOI: 10.1161/01.CIR.0000130590.24107.D3 63. Cao L, Cao Y, Xu C, Edvinsson L. **Altered endothelin receptor expression and affinity in spontaneously hypertensive rat cerebral and coronary arteries**. *PLoS One* (2013) **8**. DOI: 10.1371/journal.pone.0073761 64. Grubbs A, Anstadt M, Ergul A. **Saphenous vein endothelin system expression and activity in African American patients**. *Arterioscler Thromb Vasc Biol* (2002) **22** 1122-1127. DOI: 10.1161/01.atv.0000023160.67766.f0 65. Treiber F, Jackson R, Davis H, Pollock J, Kapuku G, Mensah G, Pollock DM. **Racial differences in endothelin-1 at rest and in response to acute stress in adolescent males**. *Hypertension* (2000) **35** 722-725. DOI: 10.1161/01.hyp.35.3.722 66. Sutherland A, Nataatmadja M, Walker P, Cuttle L, Garlick R, West M. **Vascular remodeling in the internal mammary artery graft and association with in situ endothelin-1 and receptor expression**. *Circulation* (2006) **113** 1180-1188. DOI: 10.1161/CIRCULATIONAHA.105.582890 67. Brothers T, Grubbs A, Zhang Y, Ergul A. **Selective downregulation of endothelin-B receptors in diabetic African-American patients**. *Ethn Dis* (2002) **12** 46-50. PMID: 11913620 68. Akins J, Martin Z, Patik J, Curtis B, Campbell J, Olvera G, Brothers RM. **Young, non-Hispanic Black men and women exhibit divergent peripheral and cerebral vascular reactivity**. *Exp Physiol* (2022) **107** 450-461. DOI: 10.1113/EP090168 69. Wolf ST, Jablonski NG, Ferguson SB, Alexander LM, Kenney WL. **Four weeks of vitamin D supplementation improves nitric oxide-mediated microvascular function in college-aged African Americans**. *Am J Physiol Heart Circ Physiol* (2020) **319** H906-H914. DOI: 10.1152/ajpheart.00631.2020
--- title: Analysis of a genetic region affecting mouse body weight authors: - Connie L. K. Leung - Subashini Karunakaran - Michael G. Atser - Leyla Innala - Xiaoke Hu - Victor Viau - James D. Johnson - Susanne M. Clee journal: Physiological Genomics year: 2023 pmcid: PMC10042608 doi: 10.1152/physiolgenomics.00137.2022 license: CC BY 4.0 --- # Analysis of a genetic region affecting mouse body weight ## Abstract Genetic factors affect an individual’s risk of developing obesity, but in most cases each genetic variant has a small effect. Discovery of genes that regulate obesity may provide clues about its underlying biological processes and point to new ways the disease can be treated. Preclinical animal models facilitate genetic discovery in obesity because environmental factors can be better controlled compared with the human population. We studied inbred mouse strains to identify novel genes affecting obesity and glucose metabolism. BTBR T+ Itpr3tf/J (BTBR) mice are fatter and more glucose intolerant than C57BL/6J (B6) mice. *Prior* genetic studies of these strains identified an obesity locus on chromosome 2. Using congenic mice, we found that obesity was affected by a ∼316 kb region, with only two known genes, pyruvate dehydrogenase kinase 1 (Pdk1) and integrin α 6 (Itga6). *Both* genes had mutations affecting their amino acid sequence and reducing mRNA levels. *Both* genes have known functions that could modulate obesity, lipid metabolism, insulin secretion, and/or glucose homeostasis. We hypothesized that genetic variation in or near Pdk1 or Itga6 causing reduced Pdk1 and Itga6 expression would promote obesity and impaired glucose tolerance. We used knockout mice lacking Pdk1 or Itga6 fed an obesigenic diet to test this hypothesis. Under the conditions we studied, we were unable to detect an individual contribution of either Pdk1 or Itga6 to body weight. During our studies, with conditions outside our control, we were unable to reproduce some of our previous body weight data. However, we identified a previously unknown role for Pdk1 in cardiac cholesterol metabolism providing the basis for future investigations. The studies described in this paper highlight the importance and the challenge using physiological outcomes to study obesity genes in mice. ## INTRODUCTION Obesity and type 2 diabetes are complex diseases that share common risk factors. These metabolic diseases share inter-related traits such as increased BMI, adiposity, hyperinsulinemia, insulin resistance, and mitochondrial dysfunction. The discovery of genetic factors affecting these traits in humans has been productive but challenging, due to the relatively small effects of many genetic variants, and the effects of interindividual variation in environmental and lifestyle factors that can impact the phenotypic expression of these genetic changes. Genome-wide association studies (GWAS), and advances in sequencing technology have allowed the identification of loci that affect obesity susceptibility in the general population. However, the known common variants affecting the general population, added together, have modest effects on body weight. For example, GWAS data from a cohort of 694,649 samples found 346 loci associated with BMI had relatively small individual effect sizes, and the combined loci only accounted for ∼$3.9\%$ of the variation in BMI [1]. In addition to genetic factors, lifestyle factors such as stress [2, 3], availability to and access to healthy food, or socioeconomic status [4] have been associated with obesity. This complexity confounds the discovery of obesity genes in humans. As an alternative approach to overcome some of these challenges, we have capitalized on the natural genetic differences that occur between inbred mouse strains, and the strong evolutionary conservation of obesity mechanisms across species. The BTBR T+ Itpr3tf/J (BTBR) mouse strain has a high propensity for obesity. Prior studies have shown that this strain harbors alleles that promote metabolic disease compared with the C57BL/6J (B6) strain (5–8). Genetic studies identified several quantitative trait loci (QTLs) that affect metabolic disease traits in these strains [7, 8], and genes affecting these traits have been identified [9, 10]. Additional insights about factors affecting metabolic disease in these strains have come from multiple “omic” studies and their integration with genetics (11–19). One quantitative trait locus (QTL) identified using these strains is Modifier of obese 1 (Moo1) [7]. At this locus, BTBR alleles were associated with increased body weight. We have previously localized this QTL to a ∼6 Mb region of chromosome 2 and have shown that it is associated with multiple metabolic traits [7, 20]. Here, we report that Moo1 comprises at least two QTLs, localized the major effect to a 316 kb region of chromosome 2, described the environmental modification of this locus by stress, as well as insights into lipid metabolism when PDK1 (pyruvate dehydrogenase kinase 1) levels are reduced. ## Animals Animals were housed in ventilated microisolator cages in environmentally controlled facilities with 12-h light:dark cycles, unless otherwise noted. All mice were generated from crosses between heterozygotes (HETs) such that littermates were used as controls. Congenic mice with B6 regions of chromosome 2 replacing those of the BTBR background strain were generated from an in-house breeding colony. Subcongenic strains were created from recombinants of the Moo1C strain previously described [20]. A heterozygous breeder pair of Pdk1tm1.1(KOMP)Vlcg knockout (KO), global knockout mice, were obtained from the Knockout Mouse Project (KOMP), then bred in-house. In these mice, the Pdk1 coding region is replaced by LacZ. These mice were originally on the C57BL/6N background but were backcrossed to C57BL/6J as part of colony expansion and maintenance. Because prior studies showed no effect of this locus in females, experiments used exclusively male animals (personal communication from Dr. Susanne Clee). Floxed Itga6 mice were obtained from the laboratory of Dr. Elisabeth Georges-Labouesse, where the *Itga6* gene was flanked by two loxP sites. When Cre recombinase is expressed, the 3′ end of the transmembrane domain and the cytoplasmic A and B exons of Itga6 are deleted [21, 22]. The mice were on a mixed C57BL/6J x 129 Sv background and were re-derived into CDM (with 2 additional backcrosses to C57BL/6J), but the remaining 129 *Sv* genetic contribution is unknown. Heterozygous Itga6 floxed mice were bred with mice heterozygous for cre recombinase driven by the ubiquitously expressing β-actin promoter (B6N.FVB-Tmem163Tg(ACTB-cre)2Mrt/CjDswJ mice; Stock Number 019099, The Jackson Laboratory; breeding animals were received from an in-house colony as a kind gift from the laboratory of Dr. Timothy Kieffer) to knockout Itga6 in all cells from early in development. The goal was to obtain heterozygous Itga6 KO (HET KO) animals along with littermate heterozygous floxed (FLOX), cre-expressing (CRE) and wild-type (WT) animals. Homozygous Itga6 KO animals die soon after birth from severe epithelial blistering [21], thus we used heterozygous Itga6 KO mice in our studies. Over the course of these studies, mice were housed in multiple facilities and rooms within these facilities, Center for Disease Modelling (CDM) room 1, room 2, and room 3 and Wesbrook Mouse Facility. The majority of the studies were completed within the CDM at the University of British Columbia (UBC). Analysis of the subcongenic strains was performed concurrently with mice housed in room 1 in CDM. However, CDM underwent reorganization, and the mice were moved to CDM room 2. The stress study was conducted in CDM room 2. CDM then underwent construction and thus when CDM reopened, repeat analysis of Moo1V strain concurrent with Itga6 KO and repeat analysis Pdk1 KO in CDM room 3 was conducted. At weaning, all experimental animals were placed on a diet high in fat with $20\%$ kcal sucrose, and $60\%$ kcal fat (D12492i, Research Diets). Mice were weighed weekly, at a standardized time of day. At the specified times, mice were euthanized by an overdose of isoflurane anesthesia in accordance with the Animal Care guidelines. At this time a cardiac blood sample was withdrawn, and tissues were rapidly harvested, weighed, and flash frozen. Blood samples were kept on ice, then plasma was separated by centrifugation at 4°C at 10,000 rpm for 8 min. Tissue and plasma samples were stored at −80°C until analysis. All procedures were approved by the University of British Columbia Animal Care and Use Committee and were performed in accordance with Canadian Council on Animal Care Guidelines (Protocol No. A19-0267). ## Genotyping DNA was extracted from ear notches using a commercially available kit (Puregene, Qiagen). Congenic mice were genotyped at the first and last markers shown to be B6 within the congenic insert. Recombinants were confirmed, then additional markers were genotyped to fine-map the recombination. Genotyping of single-nucleotide polymorphism (SNP) markers was performed using high-resolution melt curve analysis (Qiagen kit) using the Rotor-gene Q thermocycler (Qiagen). Microsatellite markers were genotyped by PCR amplification and resolution on polyacrylamide gels, as described [23]. Pdk1 KO mice were genotyped by PCR amplification of a 195 bp band for the WT allele and a 639 bp band for the KO allele as recommended by KOMP [23]. Genotyping of Itga6 mutant mice was conducted with primers flanking the region of the loxP sites to amplify a fragment of 120 bp in WT animals and, 150 bp when the loxP site is present. Genotyping for the presence of cre recombinase was performed using a set of primers that amplify across the deleted region were used to amplify a fragment of 600 bp, as described [24]. Primer sequences are provided in Supplemental Table S1. All genotyping reactions contained DNA samples known to be of each genotype as controls, along with a no template control. ## Gene Expression Analysis Tissue samples were processed, and RNA was extracted using E.Z.N.A total RNA extraction kit (Omega Bio-tek, Norcross, GA), after which cDNA was generated using the RevertAid First Strand cDNA Kit (Thermo Fisher Scientific, Cat. No. K1631, Waltham, MA). Gene expression was measured using real-time qPCR with SYBR green. β-actin (Actb) was used as a reference gene as it more consistently amplified in the same manner across different experimental groups and tissues compared with glyceraldehyde 3-phosphate dehydrogenase (Gapdh) or cyclophilin (Ppib). A single product was verified using melt curve analysis. Cycle threshold value (Ct) was subtracted from the control gene Ct value, to give the ΔCt value (dCt). Then ΔΔCt values (ddCt) for each mouse were calculated by subtracting each dCt value from the average dCt of the control group. No reverse transcriptase (no RT) and no template control (water) were used as negative controls and were determined for each gene and tissue assessed. Primer sequences are provided in Supplemental Table S2. ## Body Weight and Blood Sampling Mice were weighed weekly at a standardized time of day from weaning at 3 wk until 10 wk of age. To determine changes in body composition, dual-energy X-ray absorptiometry (DEXA) analysis was performed using a PIXImus Mouse Densitometer (Inside Outside Sales, Madison, WI). Body weight, body length, and cardiac blood samples were also obtained at the time of euthanasia. ## Stress Study To directly assess the interaction between stress and genotype on obesity, we used chronic restraint stress, a well-established model of stress in rodents [25]. At 10 wk of age, the mice were singly housed for the duration of the study. At 11 wk of age, restraint stress was induced in half the mice by placing the mice in a plexiglass restrainer for 90 min from 9:00 AM to 10:30 AM daily, for 7 days. The plexiglass restrainer confined the mice into a small space, with no room to move, but it did not specifically immobilize them. The control mice were singly housed during the week of stress but were not placed in the restrainers; however, all the other hormone and metabolic measurements were performed concurrently with the stress group. Age-matched littermates were used in both the control and stressed groups. To assess food intake during the experiment, mice were given a pre-weighed amount of food at the start of the study. The weight of the food remaining at the end of the study was determined again to calculate the total food intake during the 1 wk of restraint stress. There were no visible crumbs of food in the cages when the food weight was measured. Corticosterone levels were measured in plasma from the cardiac bleeds of the mice using a Corticosterone Double Antibody radioimmunoassay kit (MP Biomedicals, Solon, OH) as previously described [26]. Thirty-six hours after the last stress induction, the mice were euthanized and a standard set of tissue samples for all the mice were collected, including brain, heart, stomach, intestines, spleen, fat deposits [epididymal, renal, mesenteric, brown adipose tissue (BAT)], soleus muscle, and testes. These tissues were weighed and then flash frozen on dry ice as soon as they were collected. Immediately after euthanasia, before tissue collection, cardiac blood samples were collected and placed into microcentrifuge tubes containing 6 µL of 25 mM EDTA and kept on ice until plasma separation. All plasma samples were separated by centrifugation at 4°C at 10,000 rpm for 8 min, and all tissue and plasma samples were stored at −80°C until analysis. ## Glucose and Triglycerides Metabolite levels were measured in plasma from the cardiac blood samples collected from mice at the time of euthanasia. The mice were fasted for 4 h before euthanization. Glucose levels were measured with a colorimetric assay as described earlier (Autokit Glucose Cat. No. 997-03001, Wako Diagnostics, Richmond, VA). Insulin was measured by ELISA (Mouse Ultrasensitive Insulin ELISA, Cat. No. 80-INSMSU-E01, Alpco, Salem, NH). Plasma β-hydroxybutyrate levels (Beta Hydroxybutyrate Assay kit, Cat. No. ab83390, Abcam, Cambridge, MA), cholesterol (Cholesterol-SL Assay, Cat. No. 234-60 Sekisui Diagnostics, PEI, Canada), triglyceride and glycerol (Triglyceride-SL Assay, Cat. No. 236-60, Sekisui Diagnostics, PEI, Canada), and pyruvate (Pyruvate Assay Kit, Cat. No. MAK071, Sigma-Aldrich, Oakville, ON, Canada) were measured using colorimetric assays according to the manufacturers’ directions. The products of the colorimetric assay were measured spectrophotometrically at 505 nm for glucose and cholesterol, at 520 nm for triglycerides, at 450 nm for insulin and β-hydroxybutyrate, and at 570 nm for pyruvate using an Infinite M1000 microplate reader (Tecan, Durham, NC). Tissue cholesterol and triglyceride levels were measured in liver and heart samples [27]. Tissues (∼100 mg) were homogenized using a tissue homogenizer in 3 mL of chloroform and methanol (2:1), and extracted using 1.5 mL of ice-cold water and a second time with 750 µL of ice-cold water. A third of the measured organic layer from the liver extraction was dried with 30 µL of Thesit (hydroxypolyethoxydodecane) neat (Sigma-Aldrich, Oakville, ON, Canada) under nitrogen gas. The volume used for dehydration was measured precisely after the aliquot, as to not disturb the organic layer. For the heart samples, the entire organic layer was dried. Standards with cholesterol (Wako Diagnostics, Richmond, VA) and triolein (1:250 uL) (Sigma-Aldrich, Oakville, ON, Canada) were dried under nitrogen gas. Once dried, the samples were mixed with 300 µL of water and incubated at 37°C with vigorous vortexing twice partway through the 30-min incubation. Samples were stored at 4°C until analysis. Before the analysis, samples were brought up to room temperature and diluted with $10\%$ Thesit as needed, as determined by first testing the undiluted samples and then repeating the assay on samples that were not on the standard curve [28]. Tissue cholesterol was measured using a colorimetric assay (Cholesterol E, Cat. No. 439-17501, Wako Diagnostics, Richmond, VA). The cholesterol reagents were reconstituted according to the manufacturer’s instructions. In a flat bottom 96-well plate (Microtest plate 96 well, Cat. No. 82.158, Sarstedt, Nümbrecht, Germany), 20 µL of extracted lipids and standards were added, then 200 µL of cholesterol E reagent (Cholesterol E, Cat. No. 439-17501, Wako Diagnostics, Richmond, VA) was added and incubated at room temperature for 5 min. The products of the colorimetric assay were measured spectrophotometrically at 590 nm using an Infinite M1000 microplate reader (Tecan, Durham, NC). The values were expressed per amount of starting protein and accounted for the dried volume. Tissue triglyceride levels were measured using an enzymatic triglyceride assay (Glycerol Reagent, Cat. No. F6428, Triglyceride Reagent, Cat. No. T2449, Sigma-Aldrich, Oakville, ON, Canada). In a flat bottom 96-well plate, (as described earlier), 5 µL of extracted lipids and standards and 200 µL Glycerol Reagent (Glycerol Reagent, Cat. No. F6428, Sigma-Aldrich, Oakville, ON, Canada) were added and first measured to obtain a reading for glycerol at 540 nm. Then 50 µL of Triglyceride Reagent (Triglyceride Reagent, Cat. No. T2449, Sigma-Aldrich, Oakville, ON, Canada), which contains lipase to release glycerol from the fatty acids, was added to each of the samples. The absorbance values of the plate were measured at 540 nm using an Infinite M1000 microplate reader (Tecan, Durham, NC). The samples were incubated for 5 min at 37°C before each absorbance reading. The last absorbance read was then subtracted from the initial plate absorbance to obtain the final absorbance due to glycerol from triglycerides. To examine triglyceride synthesis and secretion in Pdk1 KO mice, a standard lipogenesis test (triglyceride synthesis and secretion) was conducted using poloxamer 407 [29] at 12 wk of age. Poloxamer 407 is a chemical that blocks triglyceride breakdown and clearance from the blood by inhibiting lipoprotein lipase [29]. Thus, during fasting, increasing plasma triglyceride concentrations result from increased triglyceride synthesis and secretion. The mice were fasted for 4 h before testing. Blood was collected at time 0, and 30, 60, 120, 240, and 360 min posttreatment of injection with poloxamer 407 (1 g/kg ip). Saphenous blood was collected into tubes containing 6 µL of 25 mM EDTA and kept on ice until plasma separation. All plasma samples were separated by centrifugation at 4°C at 10,000 rpm for 8 min. Plasma was stored at −80°C until further analysis. To assess lipid clearance from circulation in Pdk1 KO mice, an oral lipid tolerance test was performed at 14 wk of age. Mice were fasted overnight (5:00 PM to 9:00 AM). At the time food was removed, the mice were placed in clean cages with water before the lipid tolerance tests and kept in these cages throughout the testing to ensure that the triglycerides in the plasma were from the oral gavage instead of from food crumbs. For the lipid tolerance, saphenous blood was collected at time 0, 30, 60, 120, 240, and 360 min post oral gavage of olive oil (300 µL) [30]. Saphenous blood was collected into tubes containing 6 µL of 25 mM EDTA and kept on ice until plasma separation. All plasma samples were separated by centrifugation at 4°C at 10,000 rpm for 8 min and stored at −80°C until further analysis. ## Western Blotting Western blotting was performed to analyze the protein levels of PDK1. Hearts were isolated from mice as previously described, flash frozen in liquid nitrogen, and ground with a liquid nitrogen-chilled mortar and pestle. A portion of the ground tissue was put in Radioimmunoprecipitation assay lysis buffer: 150 mM NaCl, $1\%$ Nonidet P-40, $0.5\%$ sodium deoxycholate, $0.1\%$ sodium dodecyl sulfate (SDS), 50 mM Tris, pH 7.4, 2 mM EGTA, 2 mM Na3VO4, and 2 mM NaF supplemented with complete mini protease inhibitor cocktail (Roche, Laval, QC), then frozen at −80°C and thawed. Protein concentration in the lysates were measured using a Pierce bicinchoninic acid (BCA) protein assay kit (Cat. No. 23225, Thermo Fisher Scientific). Protein lysates were incubated in Laemmli loading buffer (Thermo, J61337AC) containing dithiothreitol at 95°C for 5 min. Fifteen micrograms of each sample were resolved on a $12\%$ SDS-PAGE. Proteins were then transferred to PVDF membranes (Bio-Rad, CA) and probed with antibodies against PDK1 (1:2,500, Cat. No. ab202468, Abcam), β-tubulin (1:2,000, Cat. No. T8328, Sigma). The signals were detected by secondary horseradish peroxidase (HRP)-conjugated antibodies (1:10,000, Anti-mouse, Cat. No. 7076; Anti-rabbit, Cat. No. 7074; Cell Signaling Technologies) and Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific). ## Statistical Analyses Statistical analyses, including two-way ANOVA and Student’s t test, were performed using GraphPad Prism Software (v.6.0, La Jolla, CA). A P value of less than 0.05 was considered significant. ## Localization of Moo1 Previous studies in Dr. Alan Attie’s laboratory at the University of Wisconsin had mapped a region that causes an increase in body weight in BTBR mice compared with B6 mice on the ob/ob background, despite the extreme hyperplasia of the mice without functional leptin, to chromosome 2. Fine mapping was done using congenic mice to localize a region on chromosome 2 that causes an increase in body weight when BTBR alleles are inherited [7]. The chromosome 2 region found in the ob/ob mouse background was also found to affect high-fat diet (HFD)-induced obesity. A recent independent genetic cross also showed a significant linkage peak in this region (Supplemental Fig. S1) [16, 17]. We have previously described the localization of Moo1 to a ∼6 Mb region of chromosome 2 contained within a congenic strain, Moo1C [20]. To further genetically localize Moo1, we created a panel of subcongenic strains from recombinants of the Moo1C strain (Fig. 1). **Figure 1.:** *Modifier of obese 1 (Moo1) subcongenic strains. Subcongenic strains with B6 congenic inserts in a BTBR genomic background, drawn to scale. Positions on chromosome 2 are from the mm9 genome assembly. The markers defining the boundaries of the Moo1V strain are shown on the zoomed version. White boxes represent the B6 congenic insert, gray boxes are undetermined genotype between the last markers tested as B6 and the first marker tested as BTBR, black = BTBR region as throughout the rest of the genome. On the zoomed version, the location of the known genes (Itga6, Pdk1; purple) and predicted genes (from GENCODE VM20; gray). The upper gray line includes five predicted genes: Platr26, Gm13647, Gm13663, Gm17250, Gm13662; the gray arrow at the bottom of the region is Gm13746. None of these predicted genes are shown to have orthologues in other species, including rat which is very closely related to mouse.* Before these mice were moved to the Clee laboratory at UBC, the congenic mice were backcrossed to remove the ob/ob allele background. We assessed body weights weekly in mice from each strain, as in our prior studies in the genetic screens (Fig. 2) [20]. These studies found that at least two independent loci within the Moo1C region contributed to the QTL, even without the ob/ob allele background. A major locus was localized to a 316 kb region in the Moo1V strain, and has been termed Moo1a. A second locus having a smaller effect was localized telomeric to this region in the nonoverlapping Moo1G strain and has been termed Moo1b. For the studies described in this paper, we focused on Moo1V strain (Moo1a). **Figure 2.:** *Body weight analysis of Modifier of obese 1 (Moo1) subcongenic strains. Male mice were fed the same high-fat diet (HFD) housed in Center for Disease Modelling (CDM) room 1 with the Moo1C and Moo1V mice, and weighed weekly. Neither Moo1G strain mice (A), Moo1I strain mice (B), Moo1K strain mice (C), nor Moo1L strain mice (D) had significant differences in body weight between genotypes. Data are means ± SE analyzed by two-way ANOVA. ns, not significant.* To identify genetic differences between the strains, we designed primers flanking each exon and spanning ∼2 kb of promoter for Itga6 and Pdk1, and performed Sanger sequencing. Numerous differences were identified (Table 1). Notably, within the coding region of Itga6 was a nonsynonymous change that results in a leucine in B6 mice and a valine in BTBR (now rs13464795). Notably, valine is present in many other species, suggesting it may be the ancestral allele (Supplemental Fig. S2). Within the coding region of Pdk1 was a 6 bp insertion/deletion (indel) that results in the presence of an alanine and serine at amino acids 27 and 28 in B6 mice, but their absence in the BTBR strain (now rs222188523). Interestingly, the N-terminal region of PDK1 varies in length across species (Supplemental Fig. S3), so the functional importance of this change is unclear, however in the mouse strains with sequence data available, these two amino acids appear only in B6 (Supplemental Figs. S3 and S4). In both genes, we also identified many variants in noncoding regions, synonymous variants, and variants in the untranslated regions (Table 1). Since this work was performed, many mouse strains in addition to B6 have been sequenced, including BTBR. These data, retrieved from the Mouse Phenome Database, list a total of 3,416 variants (SNPs and indels) differing between the B6 and BTBR strains in the 316 kb region of the Moo1V strain. Because we identified many noncoding variants around Itga6 and Pdk1, we examined whether these may affect their expression in multiple tissues. These studies showed that Moo1BTBR/BTBR mice have a ∼$50\%$ reduction in gene expression of Pdk1 and Itga6 in most tissues compared with Moo1B6/B6 mice (Fig. 3). These data are consistent with a recent genetic analysis of these strains that identified cis-eQTLs [16, 17] for both Pdk1 and Itga6 (Supplemental Fig. S1). Thus, we considered both Pdk1 and Itga6 as positional candidate genes. ## Environmental and Stress Modulation of Moo1 During our studies of the Moo1C strain, the mice were relocated to a new university (UBC) and multiple housing locations (see, methods). The effect of Moo1 on body weight in these strains varied between locations, even becoming undetectable in one location (Fig. 4, A–C). Subsequently, similar environmental effects were observed for the Moo1V strain housed in different rooms in the same facility (Supplemental Fig. S5). Collectively, these observations indicate that gene-environment interactions play a strong role in the penetrance of this phenotype. **Figure 4.:** *Environmental and stress modulation of the body weight phenotype. Mice were moved from the University of Wisconsin-Madison (UW) to the University of British Columbia (UBC-Wesb) and into a second facility at UBC (UBC-CDM). Modifier of obese 1 (Moo1)C growth curves in UW (A), Wesbrook (B), and Center for Disease Modelling (CDM) facilities (C). Orexin (Hcrt) (D), Orexin receptor 1 (Hcrtr1) (E), and Orexin receptor 2 (Hcrtr2) (F) expression levels in the brain were assessed in Moo1C congenic mice from the Wesbrook (Wesb) facility compared with the CDM facility. Data are ΔΔCt (ddCt) values relative to Moo1CB6/B6 mice in CDM. n = 6–9. G: corticosterone levels in cardiac blood samples of Moo1V strain mice at 12 wk of age did not differ. Mice corticosterone levels were not significantly different in all four groups. Mice were housed in CDM room 2 and were fed a high-fat diet (HFD). H: body weight was assessed in 12 wk mice 24 h before and 24 h after the last stress treatments. Moo1VB6/B6(B6/B6) control n = 8, stress n = 12; Moo1VBTBR/BTBR (BTBR/BTBR) stress and control group n = 17/group. Body composition was assessed by dual-energy X-ray absorptiometry (DEXA) before and after the one week experiment. Percent change in percent body fat (I), fat mass (J), total mass (K), and lean mass (L). Moo1VB6/B6 control n = 8, Moo1VB6/B6 stress n = 12, Moo1VBTBR/BTBR stress n = 17, and Moo1VBTBR/BTBR control group n = 17. M: food intake was assessed in each group during the week of stress, day 7 food weight was subtracted from that given on day 1 of the experiment. Moo1VB6/B6 control n = 8, Moo1VB6/B6 stress n = 12, Moo1VBTBR/BTBR stress n = 17 and Moo1VBTBR/BTBR control group n = 17. Glucose tolerance tests (GTTs) were conducted both 24 h before and after the 1 wk of stress treatment and at the same time in nonstressed controls. Glucose tolerance in mice before (N) and after (O) the stress treatment were measured and area under the curve (AUC) (P) was calculated as area under the curve to t = 0. Moo1VB6/B6 (B6/B6) control n = 8, Moo1VB6/B6 stress n = 12, Moo1VBTBR/BTBR (BTBR/BTBR) stress n = 17, and Moo1VBTBR/BTBR control group n = 17. Data are shown as mean percent change in each group ± SE and were analyzed by two-way ANOVA with stress treatment and genotype as the factors (shown below each graph, along with the P value for their interaction). Bonferroni pairwise comparisons between groups that were significantly different are shown (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001) or ns for not significant.* One notable difference between the housing locations was the noise and potential disturbances to the mice due to adjacent construction. We assessed the expression of genes involved in pathways by which stress is known to affect food intake in the brains from Moo1C mice housed at the two UBC facilities, CDM and Wesbrook. Orexin mRNA levels were four- to eightfold higher in mice housed in the Wesbrook facility compared with the mice housed in the CDM, suggesting that differences between the facilities affected this pathway (Fig. 4, D–F). To directly test whether stress modifies the effect of Moo1 on body weight, we conducted controlled stress experiments to identify stress effects on body weight and composition, food intake, and glucose tolerance. To determine the degree to which mice were stressed due to their placement in the restrainers, circulating cortisol levels were measured at the end of the study. Corticosterone levels were the same in both the control and the stress group, suggesting that the mice did not respond differently to stress (Fig. 4G). The mean levels in all groups were elevated compared with what is found in normal, nonstressed mice, typically <46 ng/mL and stressed levels are typically >100 ng/mL (31–34). To measure food intake, the mice were singly housed throughout the study, which could increase stress in all groups. In addition, the corticosterone levels were measured in blood collected 36 h after the last stress induction. This collection would be after the expected peak corticosterone levels [35], which may explain the smaller differences in the control compared with the stress groups, resulting in no differences between the groups. Body weight increased in the nonstressed control group during this 1-wk period. In contrast, their littermates which were subjected to stress lost weight. These effects were not significantly different between genotypes (Fig. 4H). We also assessed the effects of stress on body composition. The effect on fat mass was proportionally greater, resulting in significant decreases in percent body fat (Fig. 4I). Notably, both total mass and fat mass were reduced significantly more in response to stress in the Moo1VBTBR/BTBR genotype compared with Moo1VB6/B6 mice (Fig. 4, J and K). Stress significantly reduced fat mass (Fig. 4J) and lean mass (Fig. 4L) in both genotypes. In the nonstressed control mice, the change in these parameters did not differ between genotypes. This suggests that stress affects each genotype differently. Food intake was significantly reduced in Moo1VBTBR/BTBR mice compared with Moo1VB6/B6 mice (Fig. 4M). The difference between genotypes was similar in both control and stress groups. As we expected with a decrease in adiposity in the stressed mice, we found that under stress, the Moo1V mice had improved glucose tolerance (Fig. 4, N–P). Our observations that the Moo1 phenotype is modulated by housing environment, and body weight analysis was performed in CDM room 2 before we obtained the Pdk1 and Itga6 mutant mice, we repeated the Moo1-V strain body weight analysis in the newly renovated facility, CDM room 3. This analysis was performed in the same location and simultaneously with Pdk1 and Itga6 mutant mice. In this housing location, although the effects of Moo1a on body weight were blunted, body fat was still reduced in mice with the Moo1VB6/B6 mice (Supplemental Fig. S5). ## Effects of Reduced Pdk1 on Body Weight Pdk1 encodes pyruvate dehydrogenase kinase 1. There are four kinases in this family that phosphorylate and inhibit pyruvate dehydrogenase (PDH), blocking the conversion of pyruvate to acetyl-CoA for entry into the Krebs cycle for glucose metabolism [36, 37]. Reduced PDK1 is expected to increase acetyl-CoA formation. Acetyl-CoA is a substrate for lipogenesis and its activation could thereby promote fat storage. Thus, Pdk1 is both a positional and functional candidate gene, with a reduction in PDK1 activity expected to promote obesity. We tested this hypothesis using global Pdk1 KO mice and validated these KO mice using qPCR and western blots (Supplemental Fig. S6). We first analyzed the role of Pdk1 KO in body weight and measured weekly body weights to 10 wk of age. Growth curves from Pdk1 KO mice were not significantly different from those of their HET or wildtype (WT) littermate controls (Fig. 5A). A lack of difference in total body weight does not necessarily indicate that there are no changes in adiposity. We measured adiposity as well. We did not detect a significant difference in body composition measured by DEXA in the KO mice compared with littermate HET and WT controls (Fig. 5A). We also measured body length to identify if there were differences but found no significant differences (Fig. 5B). We also found no significant differences in brain, inguinal, and epididymal fat weight (Fig. 5, C–E). However, we found decreases in mesenteric fat in Pdk1 KO mice when compared with HET mice (Fig. 5F). This contrasts with our expectation from the congenic mice, where the genotype with reduced Pdk1 expression had increased adiposity. We then assessed the weights of perirenal and brown adipose tissue and observed no differences (Fig. 5, G and H). The liver is a highly metabolic tissue, which may accumulate fat causing hepatic steatosis. However, when we weighed livers, we found no differences in the Pdk1 KO mice (Fig. 5I). Together, these studies, under the described experimental conditions, were unable to confirm that loss of Pdk1 mediates the effect of the Moo1 locus on obesity (Fig. 5, J–M). **Figure 5.:** *Effect of Pdk1 on body weight, composition, and tissue weight. A: Pdk1 knockout (KO), heterozygous (HET), and wildtype (WT) littermate mice were assessed weekly for body weight from 4 to 10 wk of age. n = 6–14. At 16 wk, Pdk1 mice were euthanized and were measured for their body length (B), brain weight (C), inguinal fat (D), epididymal fat (E), mesenteric fat (F), perirenal fat (G), brown adipose tissue (BAT) (H), and liver weight (I). KO n = 9, HET n = 18, WT n = 16. Body composition analysis in 16-wk-old mice measuring fat mass (J), lean mass (K), % body fat (L), and total mass (M) did not differ between KO, HET, and WT mice. KO n = 10, HET n = 18, WT n = 16. Data are expressed as means ± SE and were analyzed using one-way ANOVA (P value shown). Tukey’s pairwise comparisons between groups that were significantly different are shown (*P < 0.05).* ## Effects of Reduced Pdk1 on Lipid Metabolism Studies have shown that reducing other isoforms of PDKs may contribute to changes in lipid metabolism and may differ in fed and fasted states (38–40). However, the in vivo effects of Pdk1 loss on lipid metabolism were not known. We performed the analysis of lipid clearance and triglyceride synthesis in Pdk1 KO mice to assess lipid metabolism. We analyzed the effects of reduced Pdk1 (Supplemental Fig. S6) on metabolites in both fasted and nonfasted states. Plasma and liver triglyceride and cholesterol levels did not significantly differ in mice lacking Pdk1 (Fig. 6, A–J). Although heart cholesterol levels were significantly increased (Fig. 6K), triglyceride levels were unaffected (Fig. 6L). To further our understanding of the change in lipids in Pdk1 KO mice, we performed an oral lipid tolerance test. Although plasma triglyceride levels in these KO were not significantly different (Fig. 6M), KO mice were found to have decreased lipid clearance, increased triglyceride levels compared with the HET mice at 4- and 6-h post oral lipid challenge (Fig. 6N). The triglyceride appearance can be an indicator of lipid digestion and absorption from the intestine, and was similar in all genotypes. Changes in plasma lipid levels could indicate that there may be changes in lipid synthesis and secretion by the liver. To test this, we measured triglyceride secretion in Pdk1 KO mice using a standard assay [29]. This assay assesses lipid production during fasting by blocking triglyceride clearance and uptake using poloxamer 407. The WT mice were similar to the KO mice, which is expected because this cohort of Pdk1 KO mice did not have differences in plasma triglyceride levels (Fig. 6N). Interestingly, the HET mice had reduced triglyceride secretion compared with the KO and WT mice throughout; however, the data were only statistically significant at 6-h post-injection of poloxamer 407 (Fig. 6N). Collectively, these observations reveal a potential novel role for PDK1 protein in cardiac cholesterol accumulation. **Figure 6.:** *Lipid levels in fasted and nonfasted Pdk1 mice. Plasma cholesterol levels were measured in nonfasted mice at 6 wk (A), fasted mice at 8 wk (B), nonfasted mice at 16 wk (C), and fasted mice at 16 wk (D). Plasma triglyceride levels were measured in nonfasted mice at 6 weeks (E), fasted mice at 8 wk (F), nonfasted mice at 16 wk (G), and fasted mice at 16 weeks (H). n = 14–30. Liver cholesterol (I) and triglyceride (J) and heart cholesterol (K) and triglyceride (L) were measured in fasted Pdk1 knockout (KO), heterozygous (HET), and wildtype (WT) mice at 16 wk. n = 10 in each group. M: oral lipid tolerance tests were performed on 14-wk-old Pdk1 KO, HET, and WT mice. n = 12–20. N: triglyceride appearance in the plasma during fasting, following lipase inhibition with poloxamer 407 was measured in 12 wk old mice. n = 12–18. Data are expressed as means ± SE and were analyzed using repeated-measures ANOVA with genotype and time as factors (P value shown). Tukey’s pairwise comparisons between groups at each time point that were significantly different are shown (*P < 0.05, ***P < 0.001, for HET vs. KO).* ## Effects of Reduced Itga6 on Body Weight Integrins are a family of cell adhesion molecules comprised of a heterodimer of an α and β subunit. Integrin α 6 (ITGA6) binds with either the β1 or β4 subunits to form heterodimers, which are involved in cell-to-cell adhesion and cell adhesion to the extracellular matrix [41]. ITGA6 is known to affect neuronal development and patterning [42, 43]. Global Itga6 knockout mice show disrupted neurulation and axial development [44]. Altered neuronal development in areas controlling food intake and/or affecting energy expenditure could theoretically lead to obesity. Thus, Itga6 is also both a functional and positional candidate gene to mediate the obesity effects at the Moo1 locus. We hypothesized that reduced ITGA6 activity would promote obesity and tested this hypothesis using mice with partial deletion of Itga6. Moo1VBTBR/BTBR mice were heavier than the Moo1VB6/B6 mice and the Moo1VBTBR/BTBR had a $50\%$ reduced *Itga6* gene expression. Thus, we hypothesized the Itga6 HET mice would have increased body weight. We used heterozygous knockout mice in our studies because a homozygous KO Itga6 mouse is not viable, dying shortly after birth261. To determine if reduced Itga6 affects body weight, we measured growth curves as in our previous studies. HET KO mice growth curves were not significantly different compared with their WT, FLOX, or CRE littermates. To ensure differences in adiposity were not missed, e.g., due to differences in body size (length) or changes in only specific fat pads, we also directly assessed body fat. In contrast to our expectations, we did not detect differences in body fat measured by DEXA, nor differences in the individual fat pad weights (Fig. 7, A–K). These data potentially identify that a global $50\%$ reduction of Itga6 alone is insufficient to modify body weight, at least using this Cre model under the conditions we studied. An important caveat of these studies is that at least in the liver in our studies, we were unable to detect a significant reduction of Itga6 in our HET KO mice compared with their controls (Supplemental Fig. S7). **Figure 7.:** *Effect of Itga6 on body weight, composition, and tissue weight. A: growth curves of heterozygous knockout Itga6fl/wt;Cretg/wt (HET KO), Itga6wt/wt;Cretg/wt (CRE), flox positive Itga6fl/fl;Crewt/wt (FLOX), and wildtype Itga6wt/wt;Crewt/wt (WT) mice, graphed as mean weekly body weight until 10 wk of age, as assessed in our congenic studies; n = 21–35. Lean mass (B), fat mass (C), total fat mass (D), and % body fat (E), Inguinal (F), epididymal (G), mesenteric (H), and perirenal fat pad weights (I), body weight (J), and body length (K) were measured in 12-wk-old Itga6 KO and control animals. Data are expressed as means ± SE and were analyzed using one-way ANOVA (P value shown).* ## DISCUSSION The goal of the present study was to elucidate the causal genes that contribute to obesity at the Moo1 QTL. Unfortunately, despite conducting in vivo genetic loss-of-function studies on the two leading candidate genes, Pdk1 and Itga6, we were unable to unambiguously point to a single causal gene that mediates the obesity phenotype. Nevertheless, a QTL affecting body weight was localized to the region contained in the Moo1V strain (Moo1a). Our inability to identify the mechanism by which this locus affects obesity could have many explanations. The effects of this locus on obesity appear to be susceptible to modulation by the environment (i.e., stress and diet), which likely complicate our efforts to identify the causal gene. It is also possible that both genes must be reduced to see the effects on obesity. Another possibility is that a SNP in the Moo1 region may have long-range interaction or gene × gene interactions within the locus that affect a gene or genes outside the locus. There is precedent for variants affecting genes even 1 Mb away [45, 46]. One of the expressed sequence tags (ESTs), perhaps a directly adjacent gene, Rapgef4, or an unknown element causes phenotype; however, there are no known miRNAs in the region. More studies will be needed to elucidate this mechanism. There were coding sequence changes in Pdk1 and Itga6 (Table 1) that were not predicted using biochemical models to have any function. However, the effects of these sequence changes on protein function were not tested. Thus, there is a possibility that these sequence changes leading to a gain or loss of function or other ESTs in the Moo1V strain region may alter obesity, instead of reduced Pdk1 or reduced Itga6 expression. Moo1 QTL has multiple loci: one in the region of Moo1V (Moo1a), at least one in the region of the Moo1U strain, that is most likely located within the Moo1G strain but not within the Moo1L strain (Moo1b). This region extends from rs13476570 (top of gray zone for Moo1G) to rs27982530 (top of known B6 region in the Moo1L strain). This ∼1.8 Mb region where Moo1b is likely located is largely identical by descent between B6 and BTBR mice, with only 111 variants (of 55349 known SNPs and indels) listed as polymorphic between these two strains (retrieved from https://phenome.jax.org/snp/retrievals), none of which are annotated to affect protein-coding regions or splice sites. This is markedly in contrast to the 3416 (of 9593 known) variants retrieved from the 316 kb region containing Moo1a that differ between B6 and BTBR. Interestingly, Moo1b contains part of Rapgef4, also known as Epac2, which regulates cAMP signaling and is another potential candidate gene. Given the larger effect size, we focused on Moo1a. The Moo1T strain, which contains Rapgef4, does not have a clear obesity phenotype. It is possible that SNPs in Moo1T and not in Moo1V affect this gene causing effect on body weight in opposite direction. In the current studies, we found that the housing environment modified the obesity phenotype of Moo1. We found potential suggestive evidence that stress may differentially affect the alleles of Moo1a. An increase in stress would counteract the increased adiposity associated with the inheritance of BTBR alleles in this region. Preliminary findings also suggested that the orexin pathway may be involved in the interaction between stress and chromosome 2 genotype. Interestingly, Pdk1 and Itga6 have been implicated in the orexin pathway and response to restraint stress, respectively (47–51), providing additional support for the findings of these studies. The significant positive finding from this work was that reduced Pdk1 affects cholesterol changes in the heart. PDK1 inhibitors have been proposed as potential cancer drugs [52, 53] and consistent with our findings, the PDK inhibitor AZ10219759 was reported to cause lipid accumulation in the heart with subsequent necrosis and atrial tissue degeneration [54]. However, the mechanism by which PDK contributes to lipid accumulation is unknown [55]. Pdk1 may have metabolic effects in the heart that should be considered when using Pdk1 inhibitors as a potential therapeutic. Further studies are needed to understand the role of PDK in cardiac lipid accumulation, ideally using cardiomyocyte-specific knockout mice. Some of the limitations of these studies were that we were unable to replicate body weight data from one location to another. There have been known challenges where scientists are unable to replicate results when studies are repeated in different laboratories [56, 57]. In our studies, we used only high-fat diets to be consistent throughout the genetic studies, but a chow diet was not used. Differences in methodology, environment, diet, genes or genetic background, microbiome changes, seasonal changes, sample size, and statistical power, can all affect the results of an experiment [58]. In our studies, there have been differences in environment that were beyond our control, including changing mouse housing locations due to construction and stress from the surrounding environment that was beyond our control (e.g., construction) that likely affected our ability to repeat the same measurements and results. We were unable to detect a role for a HET knockout of Itga6 on increasing body weight, but we are unable to make firm conclusions about Itga6 because Itga6 expression in livers collected at the end of the study was not significantly reduced in HET knockout mice. However, Itga6 is an essential gene, whereby global knockout of Itga6 is lethal. Homozygous Itga6 knockout mice die shortly after birth due to extensive skin blistering [24]. Thus, Itga6 is an important gene and should not be disregarded. Although we were unable to demonstrate evidence of the contribution of Pdk1 and *Itag6* genes on body weight changes, there has been evidence of associations of this genetic region that are potentially associated with metabolic phenotypes. There is an association between Pdk1 and various fat mass traits [59]. SNPs near Pdk1 and Itga6 are potentially associated with many metabolic phenotypes in the human population, such as BMI, fasting glucose, type 2 diabetes, total cholesterol, and triglycerides [60]. These SNPs are located between genes, so it is not clear which gene in the region they affect. However, this provides support that this region of the genome could affect metabolic traits in humans. Additional support for Itga6 comes from genetic studies of other inbred strains (A/J and SM/J) to identify genes involved in impaired glucose metabolism, which identified a congenic strain affecting these traits encompassing Itga6 but not Pdk1 [61]. The studies conducted in this paper highlight the importance, and the challenge of physiological studies in mice. It is important to note when there are noise windows, construction or facility changes during experiments. There are changes within the facility and environmental factors are hard to control in mouse studies, but may have significant and profound effects on phenotypes and the ability to reproduce phenotypes. Despite these challenges, these studies showed that PDK1 may be important for cholesterol changes in the heart. ## DATA AVAILABILITY Data will be made available upon reasonable request. ## GRANTS This work was supported by a Tier 2 Canada Research Chair Award and a Canadian Institutes of Health Research Grant (to S. M. Clee). ## DISCLOSURES No conflicts of interest, financial or otherwise, are declared by the authors. ## AUTHOR CONTRIBUTIONS C.L.K.L. and S.M.C. conceived and designed research; C.L.K.L., S.K., L.I., and X.H. performed experiments; C.L.K.L., S.K., and L.I. analyzed data; C.L.K.L., V.V., and S.M.C. interpreted results of experiments; C.L.K.L. and M.G.A. prepared figures; C.L.K.L. drafted manuscript; C.L.K.L., S.K., M.G.A., J.D.J., and S.M.C. edited and revised manuscript; C.L.K.L., S.K., M.G.A., L.I., X.H., V.V., and J.D.J. approved final version of manuscript. ## References 1. Pulit SL, Stoneman C, Morris AP, Wood AR, Glastonbury CA, Tyrrell J, Yengo L, Ferreira T, Marouli E, Ji Y, Yang J, Jones S, Beaumont R, Croteau-Chonka DC, Winkler TW, Hattersley AT, Loos RJF, Hirschhorn JN, Visscher PM, Frayling TM, Yaghootkar H, Lindgren CM. **Meta-analysis of genome-wide association studies for body fat distribution in 694 649 individuals of European ancestry**. *Hum Mol Genet* (2019.0) **28** 166-174. DOI: 10.1093/hmg/ddy327 2. Koch F-S, Sepa A, Ludvigsson J. **Psychological stress and obesity**. *J Pediatr* (2008.0) **153** 839-844. DOI: 10.1016/j.jpeds.2008.06.016 3. Tomiyama AJ. **Stress and obesity**. *Annu Rev Psychol* (2019.0) **70** 703-718. DOI: 10.1146/annurev-psych-010418-102936 4. Volaco A, Cavalcanti AM, Filho RP, Précoma DB. **Socioeconomic status: the missing link between obesity and diabetes mellitus?**. *Curr Diabetes Rev* (2018.0) **14** 321-326. DOI: 10.2174/1573399813666170621123227 5. Clee SM, Nadler ST, Attie AD. **Genetic and genomic studies of the BTBR ob/ob mouse model of type 2 diabetes**. *Am J Ther* (2005.0) **12** 491-498. DOI: 10.1097/01.mjt.0000178781.89789.25 6. Ranheim T, Dumke C, Schueler KL, Cartee GD, Attie AD. **Interaction between BTBR and C57BL/6J genomes produces an insulin resistance syndrome in (BTBR x C57BL/6J) F1 mice**. *Arterioscler Thromb Vasc Biol* (1997.0) **17** 3286-3293. DOI: 10.1161/01.atv.17.11.3286 7. Stoehr JP, Byers JE, Clee SM, Lan H, Boronenkov IV, Schueler KL, Yandell BS, Attie AD. **Identification of major quantitative trait loci controlling body weight variation in ob/ob mice**. *Diabetes* (2004.0) **53** 245-249. DOI: 10.2337/diabetes.53.1.245 8. Stoehr JP, Nadler ST, Schueler KL, Rabaglia ME, Yandell BS, Metz SA, Attie AD. **Genetic obesity unmasks nonlinear interactions between murine type 2 diabetes susceptibility loci**. *Diabetes* (2000.0) **49** 1946-1954. DOI: 10.2337/diabetes.49.11.1946 9. Bhatnagar S, Oler AT, Rabaglia ME, Stapleton DS, Schueler KL, Truchan NA, Worzella SL, Stoehr JP, Clee SM, Yandell BS, Keller MP, Thurmond DC, Attie AD. **Positional cloning of a type 2 diabetes quantitative trait locus; tomosyn-2, a negative regulator of insulin secretion**. *PLoS Genet* (2011.0) **7**. DOI: 10.1371/journal.pgen.1002323 10. Clee SM, Yandell BS, Schueler KM, Rabaglia ME, Richards OC, Raines SM, Kabara EA, Klass DM, Mui ET, Stapleton DS, Gray-Keller MP, Young MB, Stoehr JP, Lan H, Boronenkov I, Raess PW, Flowers MT, Attie AD. **Positional cloning of Sorcs1, a type 2 diabetes quantitative trait locus**. *Nat Genet* (2006.0) **38** 688-693. DOI: 10.1038/ng1796 11. Davis DB, Lavine JA, Suhonen JI, Krautkramer KA, Rabaglia ME, Sperger JM, Fernandez LA, Yandell BS, Keller MP, Wang IM, Schadt EE, Attie AD. **FoxM1 is up-regulated by obesity and stimulates beta-cell proliferation**. *Mol Endocrinol* (2010.0) **24** 1822-1834. DOI: 10.1210/me.2010-0082 12. Ferrara CT, Wang P, Neto EC, Stevens RD, Bain JR, Wenner BR, Ilkayeva OR, Keller MP, Blasiole DA, Kendziorski C, Yandell BS, Newgard CB, Attie AD. **Genetic networks of liver metabolism revealed by integration of metabolic and transcriptional profiling**. *PLoS Genet* (2008.0) **4**. DOI: 10.1371/journal.pgen.1000034 13. Keller MP, Choi Y, Wang P, Davis DB, Rabaglia ME, Oler AT, Stapleton DS, Argmann C, Schueler KL, Edwards S, Steinberg HA, Chaibub Neto E, Kleinhanz R, Turner S, Hellerstein MK, Schadt EE, Yandell BS, Kendziorski C, Attie AD. **A gene expression network model of type 2 diabetes links cell cycle regulation in islets with diabetes susceptibility**. *Genome Res* (2008.0) **18** 706-716. DOI: 10.1101/gr.074914.107 14. Lan H, Rabaglia ME, Stoehr JP, Nadler ST, Schueler KL, Zou F, Yandell BS, Attie AD. **Gene expression profiles of nondiabetic and diabetic obese mice suggest a role of hepatic lipogenic capacity in diabetes susceptibility**. *Diabetes* (2003.0) **52** 688-700. DOI: 10.2337/diabetes.52.3.688 15. Nadler ST, Stoehr JP, Schueler KL, Tanimoto G, Yandell BS, Attie AD. **The expression of adipogenic genes is decreased in obesity and diabetes mellitus**. *Proc Natl Acad Sci USA* (2000.0) **97** 11371-11376. DOI: 10.1073/pnas.97.21.11371 16. Tian J, Keller MP, Oler AT, Rabaglia ME, Schueler KL, Stapleton DS, Broman AT, Zhao W, Kendziorski C, Yandell BS, Hagenbuch B, Broman KW, Attie AD. **Identification of the bile acid transporter Slco1a6 as a candidate gene that broadly affects gene expression in mouse pancreatic islets**. *Genetics* (2015.0) **201** 1253-1262. DOI: 10.1534/genetics.115.179432 17. Tu Z, Keller MP, Zhang C, Rabaglia ME, Greenawalt DM, Yang X, Wang IM, Dai H, Bruss MD, Lum PY, Zhou YP, Kemp DM, Kendziorski C, Yandell BS, Attie AD, Schadt EE, Zhu J. **Integrative analysis of a cross-loci regulation network identifies App as a gene regulating insulin secretion from pancreatic islets**. *PLoS Genet* (2012.0) **8**. DOI: 10.1371/journal.pgen.1003107 18. Zhao E, Keller MP, Rabaglia ME, Oler AT, Stapleton DS, Schueler KL, Neto EC, Moon JY, Wang P, Wang IM, Lum PY, Ivanovska I, Cleary M, Greenawalt D, Tsang J, Choi YJ, Kleinhanz R, Shang J, Zhou YP, Howard AD, Zhang BB, Kendziorski C, Thornberry NA, Yandell BS, Schadt EE, Attie AD. **Obesity and genetics regulate microRNAs in islets, liver, and adipose of diabetic mice**. *Mamm Genome* (2009.0) **20** 476-485. DOI: 10.1007/s00335-009-9217-2 19. Dong C, Simonett SP, Shin S, Stapleton DS, Schueler KL, Churchill GA, Lu L, Liu X, Jin F, Li Y. **INFIMA leverages multi-omics model organism data to identify effector genes of human GWAS variants**. *Genome Biol* (2021.0) **22**. DOI: 10.1186/s13059-021-02450-8 20. Karunakaran S, Manji A, Yan CS, Wu ZJ, Clee SM. **The Moo1 obesity quantitative trait locus in BTBR T+ Itpr3tf/J mice increases food intake**. *Physiol Genomics* (2013.0) **45** 191-199. DOI: 10.1152/physiolgenomics.00159.2012 21. Georges-Labouesse E, Messaddeq N, Yehia G, Cadalbert L, Dierich A, Le Meur M. **Absence of integrin α6 leads to epidermolysis bullosa and neonatal death in mice**. *Nat Genet* (1996.0) **13** 370-373. DOI: 10.1038/ng0796-370 22. De Arcangelis A, Hamade H, Alpy F, Normand S, Bruyère E, Lefebvre O, Méchine-Neuville A, Siebert S, Pfister V, Lepage P, Laquerriere P, Dembele D, Delanoye-Crespin A, Rodius S, Robine S, Kedinger M, Van Seuningen I, Simon-Assmann P, Chamaillard M, Labouesse M, Georges-Labouesse E. **Hemidesmosome integrity protects the colon against colitis and colorectal cancer**. *Gut* (2017.0) **66** 1748-1760. DOI: 10.1136/gutjnl-2015-310847 23. Skarnes WC, Rosen B, West AP, Koutsourakis M, Bushell W, Iyer V, Mujica AO, Thomas M, Harrow J, Cox T, Jackson D, Severin J, Biggs P, Fu J, Nefedov M, de Jong PJ, Stewart AF, Bradley A. **A conditional knockout resource for the genome-wide study of mouse gene function**. *Nature* (2011.0) **474** 337-342. DOI: 10.1038/nature10163 24. Georges-Labouesse E, Mark M, Messaddeq N, Gansmüller A. **Essential role of α6 integrins in cortical and retinal lamination**. *Curr Biol* (1998.0) **8** 983-986. DOI: 10.1016/s0960-9822(98)70402-6 25. Ball KT, Best O, Luo J, Miller LR. **Chronic restraint stress causes a delayed increase in responding for palatable food cues during forced abstinence via a dopamine D1-like receptor-mediated mechanism**. *Behav Brain Res* (2017.0) **319** 1-8. DOI: 10.1016/j.bbr.2016.11.020 26. Goel N, Innala L, Viau V. **Sex differences in serotonin (5-HT) 1A receptor regulation of HPA axis and dorsal raphe responses to acute restraint**. *Psychoneuroendocrinology* (2014.0) **40** 232-241. DOI: 10.1016/j.psyneuen.2013.11.020 27. Briaud I, Harmon JS, Kelpe CL, Segu VB, Poitout V. **Lipotoxicity of the pancreatic beta-cell is associated with glucose-dependent esterification of fatty acids into neutral lipids**. *Diabetes* (2001.0) **50** 315-321. DOI: 10.2337/diabetes.50.2.315 28. Huynh FK, Levi J, Denroche HC, Gray SL, Voshol PJ, Neumann UH, Speck M, Chua SC, Covey SD, Kieffer TJ. **Disruption of hepatic leptin signaling protects mice from age-and diet-related glucose intolerance**. *Diabetes* (2010.0) **59** 3032-3040. DOI: 10.2337/db10-0074 29. Millar JS, Cromley DA, McCoy MG, Rader DJ, Billheimer JT. **Determining hepatic triglyceride production in mice: comparison of poloxamer 407 with Triton WR-1339**. *J Lipid Res* (2005.0) **46** 2023-2028. DOI: 10.1194/jlr.D500019-JLR200 30. Excoffon KJ, Liu G, Miao L, Wilson JE, McManus BM, Semenkovich CF, Coleman T, Benoit P, Duverger N, Branellec D, Denefle P, Hayden MR, Lewis ME. **Correction of hypertriglyceridemia and impaired fat tolerance in lipoprotein lipase–deficient mice by adenovirus-mediated expression of human lipoprotein lipase**. *Arterioscler Thromb Vasc Biol* (1997.0) **17** 2532-2539. DOI: 10.1161/01.atv.17.11.2532 31. Park SE, Park D, Song K-I, Seong J-K, Chung S, Youn I. **Differential heart rate variability and physiological responses associated with accumulated short-and long-term stress in rodents**. *Physiol Behav* (2017.0) **171** 21-31. DOI: 10.1016/j.physbeh.2016.12.036 32. Valentine H, Williams WO, Maurer KJ. **Sedation or inhalant anesthesia before euthanasia with CO2 does not reduce behavioral or physiologic signs of pain and stress in mice**. *J Am Assoc Lab Anim Sci* (2012.0) **51** 50-57. PMID: 22330868 33. Flint MS, Tinkle SS. **C57BL/6 mice are resistant to acute restraint modulation of cutaneous hypersensitivity**. *Toxicol Sci* (2001.0) **62** 250-256. DOI: 10.1093/toxsci/62.2.250 34. Shanks N, Griffiths J, Zalcman S, Zacharko RM, Anisman H. **Mouse strain differences in plasma corticosterone following uncontrollable footshock**. *Pharmacol Biochem Behav* (1990.0) **36** 515-519. DOI: 10.1016/0091-3057(90)90249-h 35. Goel N, Philippe TJ, Chang J, Koblanski ME, Viau V. **Cellular and serotonergic correlates of habituated neuroendocrine responses in male and female rats**. *Psychoneuroendocrinology* (2022.0) **136**. DOI: 10.1016/j.psyneuen.2021.105599 36. Korotchkina LG, Patel MS. **Probing the mechanism of inactivation of human pyruvate dehydrogenase by phosphorylation of three sites**. *J Biol Chem* (2001.0) **276** 5731-5738. DOI: 10.1074/jbc.M007558200 37. Korotchkina LG, Patel MS. **Site specificity of four pyruvate dehydrogenase kinase isoenzymes toward the three phosphorylation sites of human pyruvate dehydrogenase**. *J Biol Chem* (2001.0) **276** 37223-37229. DOI: 10.1074/jbc.M103069200 38. Jeoung NH, Rahimi Y, Wu P, Lee WP, Harris RA. **Fasting induces ketoacidosis and hypothermia in PDHK2/PDHK4-double-knockout mice**. *Biochem J* (2012.0) **443** 829-839. DOI: 10.1042/BJ20112197 39. Wu C-Y, Tso S-C, Chuang JL, Gui W-J, Lou M, Sharma G, Khemtong C, Qi X, Wynn RM, Chuang DT. **Targeting hepatic pyruvate dehydrogenase kinases restores insulin signaling and mitigates ChREBP-mediated lipogenesis in diet-induced obese mice**. *Mol Metab* (2018.0) **12** 12-24. DOI: 10.1016/j.molmet.2018.03.014 40. Jeoung NH, Harris RA. **Pyruvate dehydrogenase kinase-4 deficiency lowers blood glucose and improves glucose tolerance in diet-induced obese mice**. *Am J Physiol Endocrinol Metab* (2008.0) **295** E46-E54. DOI: 10.1152/ajpendo.00536.2007 41. Giancotti FG, Ruoslahti E. **Integrin signaling**. *Science* (1999.0) **285** 1028-1032. DOI: 10.1126/science.285.5430.1028 42. Marchetti G, De Arcangelis A, Pfister V, Georges-Labouesse E. **α6 integrin subunit regulates cerebellar development**. *Cell Adh Migr* (2013.0) **7** 325-332. DOI: 10.4161/cam.25140 43. Pellegatta M, De Arcangelis A, D'Urso A, Nodari A, Zambroni D, Ghidinelli M, Matafora V, Williamson C, Georges-Labouesse E, Kreidberg J, Mayer U, McKee KK, Yurchenco PD, Quattrini A, Wrabetz L, Feltri ML. **α6β1 and α7β1 integrins are required in Schwann cells to sort axons**. *J Neurosci* (2013.0) **33** 17995-18007. DOI: 10.1523/JNEUROSCI.3179-13.2013 44. Lallier TE, Whittaker CA, DeSimone DW. **Integrin alpha 6 expression is required for early nervous system development in Xenopus laevis**. *Development* (1996.0) **122** 2539-2554. DOI: 10.1242/dev.122.8.2539 45. Rask-Andersen M, Almén MS, Schiöth HB. **Scrutinizing the FTO locus: compelling evidence for a complex, long-range regulatory context**. *Hum Genet* (2015.0) **134** 1183-1193. DOI: 10.1007/s00439-015-1599-5 46. Smemo S, Tena JJ, Kim K-H, Gamazon ER, Sakabe NJ, Gómez-Marín C, Aneas I, Credidio FL, Sobreira DR, Wasserman NF, Lee JH, Puviindran V, Tam D, Shen M, Son JE, Vakili NA, Sung H-K, Naranjo S, Acemel RD, Manzanares M, Nagy A, Cox NJ, Hui C-C, Gomez-Skarmeta JL, Nóbrega MA. **Obesity-associated variants within FTO form long-range functional connections with IRX3**. *Nature* (2014.0) **507** 371-375. DOI: 10.1038/nature13138 47. Gao Y-Z, Guo S-Y, Yin Q-Z, Cui X-Q, Hisamitsu T, Jiang X-H. **Possible involvement of integrin signaling pathway in the process of recovery from restraint stress in rats**. *Neurosci Bull* (2007.0) **23** 229-235. DOI: 10.1007/s12264-007-0034-x 48. Liu Y, Zhao Y, Guo L. **Effects of orexin A on glucose metabolism in human hepatocellular carcinoma in vitro via PI3K/Akt/mTOR-dependent and-independent mechanism**. *Mol Cell Endocrinol* (2016.0) **420** 208-216. DOI: 10.1016/j.mce.2015.11.002 49. Orsetti M, Di Brisco F, Canonico P, Genazzani A, Ghi P. **Gene regulation in the frontal cortex of rats exposed to the chronic mild stress paradigm, an animal model of human depression**. *Eur J Neurosci* (2008.0) **27** 2156-2164. DOI: 10.1111/j.1460-9568.2008.06155.x 50. Sikder D, Kodadek T. **The neurohormone orexin stimulates hypoxia-inducible factor-1 activity**. *Genes Dev* (2007.0) **21** 2995-3005. DOI: 10.1101/gad.1584307 51. Wan X, Liu Y, Zhao Y, Sun X, Fan D, Guo L. **Orexin A affects HepG2 human hepatocellular carcinoma cells glucose metabolism via HIF-1α-dependent and-independent mechanism**. *Plos one* (2017.0) **12**. DOI: 10.1371/journal.pone.0184213 52. Sutendra G, Michelakis ED. **Pyruvate dehydrogenase kinase as a novel therapeutic target in oncology**. *Front Oncol* (2013.0) **3**. DOI: 10.3389/fonc.2013.00038 53. Zhang M, Zhao Y, Li Z, Wang C. **Pyruvate dehydrogenase kinase 4 mediates lipogenesis and contributes to the pathogenesis of nonalcoholic steatohepatitis**. *Biochem Biophys Res Commun* (2018.0) **495** 582-586. DOI: 10.1016/j.bbrc.2017.11.054 54. Jones HB, Reens J, Johnson E, Brocklehurst S, Slater I. **Myocardial steatosis and necrosis in atria and ventricles of rats given pyruvate dehydrogenase kinase inhibitors**. *Toxicol Pathol* (2014.0) **42** 1250-1266. DOI: 10.1177/0192623314530195 55. Khan AUH, Allende-Vega N, Gitenay D, Gerbal-Chaloin S, Gondeau C, Vo D-N, Belkahla S, Orecchioni S, Talarico G, Bertolini F, Bozic M, Valdivielso JM, Bejjani F, Jariel I, Lopez-Mejia IC, Fajas L, Lecellier C-H, Hernandez J, Daujat M, Villalba M. **The PDK1 inhibitor dichloroacetate controls cholesterol homeostasis through the ERK5/MEF2 pathway**. *Sci Rep* (2017.0) **7**. DOI: 10.1038/s41598-017-10339-5 56. Cox AR, Barrandon O, Cai EP, Rios JS, Chavez J, Bonnyman CW, Lam CJ, Yi P, Melton DA, Kushner JA. **Resolving discrepant findings on ANGPTL8 in β-cell proliferation: a collaborative approach to resolving the betatrophin controversy**. *PLoS One* (2016.0) **11**. DOI: 10.1371/journal.pone.0159276 57. von Herrath M, Pagni PP, Grove K, Christoffersson G, Tang-Christensen M, Karlsen AE, Petersen JS. **Case reports of pre-clinical replication studies in metabolism and diabetes**. *Cell Metab* (2019.0) **29** 795-802. DOI: 10.1016/j.cmet.2019.02.004 58. Li A, Meyre D. **Challenges in reproducibility of genetic association studies: lessons learned from the obesity field**. *Int J Obes (Lond)* (2013.0) **37** 559-567. DOI: 10.1038/ijo.2012.82 59. Li H, Rukina D, David FPA, Li TY, Oh C-M, Gao AW, Katsyuba E, Bou Sleiman M, Komljenovic A, Huang Q, Williams RW, Robinson-Rechavi M, Schoonjans K, Morgenthaler S, Auwerx J. **Identifying gene function and module connections by the integration of multispecies expression compendia**. *Genome Res* (2019.0) **29** 2034-2045. DOI: 10.1101/gr.251983.119 60. MacArthur D, Neale B, Bloom J, Seed C, Solomonson M 61. Kobayashi M, Ohno T, Ihara K, Murai A, Kumazawa M, Hoshino H, Iwanaga K, Iwai H, Hamana Y, Ito M, Ohno K, Horio F. **Searching for genomic region of high-fat diet-induced type 2 diabetes in mouse chromosome 2 by analysis of congenic strains**. *PLoS One* (2014.0) **9**. DOI: 10.1371/journal.pone.0096271
--- title: Suppression of lncRNA OIP5-AS1 Attenuates Apoptosis and Inflammation, and Promotes Proliferation by Mediating miR-25-3p Expression in Lipopolysaccharide-Induced Myocardial Injury authors: - Jiaju Ma - Hebu Qian - Han Zou journal: Analytical Cellular Pathology (Amsterdam) year: 2023 pmcid: PMC10042636 doi: 10.1155/2023/3154223 license: CC BY 4.0 --- # Suppression of lncRNA OIP5-AS1 Attenuates Apoptosis and Inflammation, and Promotes Proliferation by Mediating miR-25-3p Expression in Lipopolysaccharide-Induced Myocardial Injury ## Abstract ### Purpose Long non-coding RNAs (LncRNAs) OIP5-AS1 and miR-25-3p play important roles in myocardial injury, whereas their roles in lipopolysaccharide (LPS)-induced myocardial injury remain unknown. The purpose of our study was to investigate the functional mechanisms of OIP5-AS1 and miR-25-3p in LPS-induced myocardial injury. ### Methods Rats and H9C2 cells were treated with LPS to establish the model of myocardial injury in vivo and in vitro, respectively. The expression levels of OIP5-AS1 and miR-25-3p were determined by quantitative reverse transcriptase-polymerase chain reaction. Enzyme-linked immunosorbent assay was performed to measure the serum levels of IL-6 and TNF-α. The relationship between OIP5-AS1 and miR-25-3p/NOX4 was determined by luciferase reporter assay and/or RNA immunoprecipitation assay. The apoptosis rate was detected by flow cytometry, and cell viability was detected by 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2-H-tetrazolium bromide assay. Western blot was performed to detect the protein levels of Bax, Bcl-2, caspase3, c-caspase3, NOX4, and p-NF-κB p65/NF-κB p65. ### Results OIP5-AS1 was up-regulated, and miR-25-3p was down-regulated in myocardial tissues of LPS-induced rats and LPS-treated H9C2 cells. Knockdown of OIP5-AS1 relieved the myocardial injury in LPS-induced rats. Knockdown of OIP5-AS1 also inhibited the inflammation and apoptosis of myocardial cells in vivo, which was subsequently confirmed by in vitro experiments. In addition, OIP5-AS1 targeted miR-25-3p. MiR-25-3p mimics reversed the effects of OIP5-AS1 overexpression on promoting cell apoptosis and inflammation and on inhibiting cell viability. Besides, miR-25-3p mimics blocked the NOX4/NF-κB signalling pathway in LPS-induced H9C2 cells. ### Conclusion Silencing of lncRNA OIP5-AS1 alleviated LPS-induced myocardial injury by regulating miR-25-3p. ## 1. Introduction Myocardial injury is an important cause of adverse cardiovascular outcomes following myocardial ischemia and circulatory arrest [1]. The underlying pathogenesis of myocardial injury includes exaggerated inflammation and oxidative stress, and changes in the mitochondrial permeability transition pores [2]. Although surgery and various drugs have been widely used in the treatment of myocardial injury, the therapeutic efficacy remains unsatisfactory [3–5]. Thus, it is imperative to develop more effective therapeutic strategies for myocardial injury. Long non-coding RNAs (lncRNAs), comprising at least 200 nucleotides, are a subset of non-coding RNA transcripts with multiple known regulatory functions [2]. Recent studies have proved that lncRNAs are important regulators in myocardial injury. For example, lncRNA KCNQ1OT1 attenuates lipopolysaccharide (LPS)-induced myocardial injury via increasing the viability and decreasing the apoptosis of injured cardiomyocytes [6]. lncRNA NEAT1 promotes the progression of septic myocardial cell injury by targeting miR-144-3p [3] or miR-590-3p [4]. lncRNA CYTOR attenuates LPS-induced myocardial injury by regulating miR-24/XIAP axis [5]. Previous studies have also reported that lncRNA OIP5 antisense RNA 1 (OIP5-AS1) promotes endothelial vascular injury by sponging miR-195-5p [7] and mitigates reactive oxygen species (ROS)-driven mitochondrial injury and apoptosis following myocardial injury [8]. In addition, Niu et al. have revealed that OIP5-AS1 attenuates myocardial ischemia/reperfusion (I/R) injury via regulating miR-29a-SIRT1/AMPK/PGC1α pathway [6]. However, the potential role of OIP5-AS1 in LPS-induced myocardial injury remains unclear. Because the action mechanisms of OIP5-AS1 in myocardial injury are complex, more downstream axes still need to be revealed. MicroRNAs (miRNAs) are also important regulators in the pathogenesis and progression of myocardial injury and related diseases [9, 10]. For example, miR-214 alleviates myocardial injury in a septic mouse model [11]. Down-regulation of miR-23b prevents the cardiac dysfunction associated with polymicrobial sepsis [12]. MiR-146a mitigates LPS-induced myocardial injury by inhibiting NF-κB activation and the production of inflammatory factors [8]. Additionally, miR-25 protects cardiomyocytes against hypoxia/reoxygenation (H/R)-induced fibrosis and apoptosis [13], and inhibits LPS-induced apoptosis of cardiomyocytes by targeting PTEN [14]. However, the interaction between OIP5-AS1 and miR-25-3p in the pathogenesis of myocardial injury is rarely known. In this study, we determined the expression levels of OIP5-AS1 and miR-25-3p in LPS-induced myocardial cells, evaluated their possible interaction, and further analyzed the potential downstream pathways of miR-25-3p in vitro. We also explored the function of OIP5-AS1 in a rat model of LPS-induced myocardial injury. Our results may provide a novel therapeutic target for myocardial injury. ## 2.1. Animals Sprague–Dawley rats ($$n = 32$$) with an weight of 200 ± 20 g were purchased from the Beijing Vital River Laboratory Animal Technologies Co. Ltd. All rats were housed in an independent environment at room temperature with a 12 hour light/dark cycle and free access to food and water. This study was approved by the ethics committee of our hospital, and the experimental procedures complied with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. ## 2.2. Establishment of a Rat Model of LPS-Induced Myocardial Injury The short hairpin RNA against OIP5-AS1 (sh-OIP5-AS1) and the negative control (sh-NC) were synthesized by the GeneChem Company (Shanghai, China) and packaged with lentiviral vector pGLV3-GFP (LV-sh-OIP5-AS1 and LV-sh-NC). To knockdown OIP5-AS1 in vivo, rats were infected with 1 × 108 plaque forming units (PFUs) LV-sh-OIP5-AS1 (in 50 μL phosphate-buffered saline [PBS]; a retention time of more than one month in vivo) via intratracheal injection for 7 days before LPS (L2630, Sigma–Aldrich, Shanghai, China) treatment, and sepsis was then triggered by peritoneal injection of LPS, as previously described [12]. Briefly, rats were randomly divided into four groups ($$n = 8$$ in each group): sham, LPS Blank, LPS + sh-NC, and LPS + sh-OIP5-AS1 group. Rats in the experimental groups were intravenously injected with LPS (5 mg/kg, in 100 μL PBS), whereas rats in the sham group were injected with 100 μL PBS. At the end of the experiment, all rats were euthanized via injecting phenobarbital sodium at a dosage of 100 mg/kg. Blood samples (5 mL) were collected from the abdominal aorta of rats, placed for 30 minutes, and centrifuged at 3000 rpm for 10 minutes. The serum samples (supernatant) were used for enzyme-linked immunosorbent assay (ELISA). In addition, myocardial tissues were obtained from the left ventricle of the heart and subjected to hematoxylin and eosin (H&E) staining and quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR). ## 2.3. Cell Culture An embryonic rat ventricular myocardial cell line (H9C2) was purchased from the American Type Culture Collection (Manassas, VA, USA), and cultured in Dulbecco's Modified Eagle's Medium supplemented with $10\%$ fetal bovine serum, 100 IU/mL penicillin, and 10 μg/mL streptomycin. H9C2 cells were maintained in a humidified incubator supplied with $5\%$ CO2 at 37°C (Supplementary Figure S1). ## 2.4. Transfection OIP5-AS1 overexpression plasmid was obtained from GeneChem Company, and the miR-25-3p mimics and its counterpart negative control were synthesized by the Life Technologies Corporation (Carlsbad, CA, USA). The OIP5-AS1 shRNAs (a retention time of about one week in vitro) and overexpression plasmids were transfected into cells using Lipofectamine 3000 (Thermo Fisher Scientific, Carlsbad, CA, USA) for 48 hours following the manufacturer's instructions. ## 2.5. Luciferase Reporter Assay The 3′ untranslated region (3′-UTR) of OIP5-AS1/NOX4 containing the predicted binding sequences of miR-25-3p was synthesized and cloned into psiCHECK-2 vector (Promega, Madison, WI, USA), referring as OIP5-AS1/NOX4 wild-type (WT). The 3′-UTR of OIP5-AS1/NOX4 containing the mutant binding sequences of miR-25-3p was constructed as OIP5-AS1/NOX4 mutation (MUT). Afterward, miRNA mimics/NC was transfected into the cells with OIP5-AS1/NOX4 WT/MUT using Lipofectamine 3000 (Thermo Fisher Scientific). Relative luciferase activity was finally measured by a Multimode Detector (Beckman Coulter, Fullerton, CA, USA). ## 2.6. RNA Immunoprecipitation Assay H9C2 cells were transfected with miR-25-3p mimics. After 48 hours of transfection, RNA immunoprecipitation (RIP) assay was performed using the Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore, Bedford, MA, USA). RIP lysis buffer supplemented with protease and RNase inhibitor was used to lyse the cells. Afterward, the cell extracts were incubated with RIP buffer containing magnetic beads conjugated with anti-Ago2 antibody (Cell Signaling, Danvers, MA, USA) or negative control Immunoglobulin G (IgG) (Millipore). The co-precipitated RNAs were then synthesized into complementary DNA (cDNA) and evaluated by qRT-PCR. ## 2.7. Enzyme-Linked Immunosorbent Assay The levels of TNF-α and IL-6 were measured using a rat TNF-α ELISA kit (catalog No: K1052-100; Biovision, San Francisco, CA, USA) and a rat IL-6 ELISA kit (catalog No: K4145-100; Biovision), respectively. The contents of creatine kinase (CK)-MB and cardiac troponin I (cTnI) in the serum of rats were measured using a CK-MB isoenzyme Assay Kit (catalog No: H197-1-1; Nanjing Jiancheng Bioengineering, Inc., Nanjing, China) and a rat Cardiac Troponin I ELISA ELISA kit (catalog No: ab246529; Abcam, Cambridge, MA, USA), respectively. ## 2.8. Quantitative Reverse Transcriptase-Polymerase Chain Reaction Total RNAs were extracted from tissues and cells using TRIzol reagent (Life Technologies Corporation). A Bio-Rad SYBR Green PCR Master Mix (Bio-Rad, Hercules, CA, USA) was then used for qRT-PCR. The primers were listed as follows: OIP5-AS1, forward: 5′-AAAGCAAGGTCTCCCCACAAG-3′, reverse: 5′-GGTCTGTGCTAGATCAAAAGGCA-3′; miR-25-3p, forward: 5′-CATTGCACTTGTCTCGGTCTGA-3′; reverse: 5′-GCTGTCAACGATACGCTACGTAACG-3′; glyceraldehyde-3-phosphate dehydrogenase (GAPDH), forward: 5′-TCCGCCCCTTCCGCTGATG-3′, reverse: 5′-CACGGAAGGCCATGCCAGTGA-3′; and U6, forward: 5′- CTCGCTTCGGCAGCACA-3′, reverse: 5′-AACGCTTCAGAATTTGCGT-3′. The expression of miR-25-3p was normalized to U6, and the expression of OIP5-AS1 was normalized to GAPDH. The fold changes were calculated following the 2−ΔΔCt method. ## 2.9. H&E Staining The myocardial tissues from rats were fixed in $10\%$ neutral formaldehyde solution for 24 hours at 4°C and then dehydrated and vitrified. Following this, they were embedded in paraffin and cut into 6 μm sections. The sections were then de-waxed and stained with H&E. Finally, the stained sections were imaged under a BX50 bioluminescent microscope (Olympus, Tokyo, Japan) to reveal the morphological changes. ## 2.10. TUNEL Staining The apoptosis of myocardial cells in rats was detected using an In Situ Cell Death Detection Kit (Roche, Basel, Switzerland) according to the manufacturer's instructions. Simply, the paraffin-embedded sections were dewaxed with xylene, dehydrated with gradient alcohol, incubated with 20 μg/mL protease K to enhance membrane permeability, and further incubated with $3\%$ H2O2 to block endogenous peroxidase. Subsequently, the sections were incubated with TUNEL reaction mixture for 1 hour at 37°C under darkness. After counterstained with 4,6-diamino-2-phenyl indole (DAPI), the fluorescence was observed under a microscope (Olympus). ## 2.11. Western Blot Cells were lysed in RIPA buffer (Cell Signaling Technology), and equivalent amounts of protein extracts were separated using $10\%$ sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE). Protein samples were then transferred to polyvinylidene fluoride membranes and blocked with $5\%$ non-fat milk. These membranes were then incubated with primary antibodies, including rabbit anti-Bcl-2 (1: 2000, ab182858, Abcam), rabbit anti-Bax (1: 1000, ab32503, Abcam), rabbit anti-c-caspase3 (1: 500, ab13847, Abcam), rabbit anti-caspase 3 (1: 5000, ab32351, Abcam), rabbit anti-NOX4 (1: 2000, ab133303, Abcam), rabbit NF-κB p65 (1: 2000, ab32536, Abcam), rabbit p-NF-κB p65 (1: 1000, ab194726, Abcam), and rabbit anti-β-actin (1: 200; ab115777, Abcam) overnight at 4°C. Afterward, the membranes were gently washed and then incubated with horseradish peroxidase-labelled secondary antibody (goat anti-rabbit, 1: 2000, ab205718, Abcam) at 25°C for 1 hour. Finally, relative expression level was quantified using a Gel-Pro Analyzer (Media Cybernetics) and normalized to β-actin. ## 2.12. Cellular Apoptosis Assay Cell apoptosis was examined by annexin V-phycoerythrin staining (BD Science, Franklin Lakes, NJ, USA). Briefly, cells were rinsed in PBS and stained with annexin V-PE solution for 25 minutes at 37°C in the dark. Subsequently, the stained cells were detected on a flow cytometry (Beckman Coulter), and the data were analyzed using the FlowJo software (Tree Star, Ashland, OR, USA). ## 2.13. Cell Viability Assay Cell viability was detected by 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2-H-tetrazolium bromide (MTT) assay. Cells were plated into 96-well plates at 3000 cells/100 μL/per well and treated with LPS and/or transfected with mimics or vectors. These samples then incubated with 10 μL MTT (5 mg/mL) for 4 hours at 37°C. Following this, the samples were further incubated with 150 μL dimethyl sulfoxide for 10 minutes. The optical density of each well was measured at 490 nm using a microplate reader (MG Labtech, Durham, NC, USA). ## 2.14. Statistical Analysis Each experiment was performed in at least triplicates, and the data were presented as the mean ± standard deviation. One-way analysis of variance (one-way ANOVA) was used for the comparisons among multiple groups (a test for Gaussian distribution was applied before any analysis), and Tukey's multiple comparisons test was used for pairwise comparisons. The t-test was used for comparisons between two groups. The results were considered statistically significant if the P-value was <0.05. ## 3.1. Construction of a Rat Model of Myocardial Injury The LPS-induced myocardial injury was first assessed in rats. H&E staining showed that LPS treatment resulted in irregular arrangement, myofibrillar fragmentation, and cardiocyte degeneration in myocardial tissues (Figure 1(a)). TUNEL staining revealed more apoptotic myocardial cells in the LPS group than those in the sham group (Figure 1(b)). CK-MB and cTnl are biomarkers of myocardial injury. LPS-induced pro-inflammatory cytokines, IL-6 and TNF-α, are also correlated with myocardial injury [15]. We detected the serum levels of CK-MB, cTnl, IL-6, and TNF-α. The results showed that these indexes in LPS-treated rats were remarkably increased compared with those in control rats ($P \leq 0.01$; Figures 1(c), 1(d), 1(e), and 1(f)). These results indicated that the rat model of LPS-induced myocardial injury was successfully constructed. ## 3.2. OIP5-AS1 Is Up-Regulated by LPS in Myocardial Tissues, and Knockdown of OIP5-AS1 Suppresses Inflammation in LPS-Induced Rats Subsequently, we investigated the expression of sh-OIP5-AS1. qRT-PCR results demonstrated that OIP5-AS1 in myocardial tissues was up-regulated in LPS-induced rats ($P \leq 0.01$; Figure 2(a)). The efficiency of OIP5-AS1 knockdown was evaluated, and sh-OIP5-AS1 significantly decreased the expression of OIP5-AS1 in myocardial tissues ($P \leq 0.01$; Figure 2(b)). In addition, the serum levels of IL-6 and TNF-α (two inflammatory factors) were increased by the treatment of LPS in a time-dependent manner ($P \leq 0.01$). The increased serum levels of IL-6 and TNF-α at different time points were all significantly weakened by the intervention of sh-OIP5-AS1 in LPS-induced rats ($P \leq 0.01$; Figures 2(c) and 2(d)). Similarly, LPS-induced elevation of CK-MB and cTnl levels was significantly offset by sh-OIP5-AS1 ($P \leq 0.01$; Figures 2(e) and 2(f)). Besides, as illustrated in Figure 2(g), H&E staining demonstrated that knockdown of OIP5-AS1 alleviated LPS-induced inflammatory cell infiltration, irregular arrangement, and cardiomyocyte degeneration. TUNEL staining also determined that the LPS-induced myocardial apoptosis was weakened by the down-regulation of OIP5-AS1 (Figure 2(h)). ## 3.3. Sh-OIP5-AS1 Inhibits LPS-Induced Production of Inflammatory Cytokines in H9C2 Cells Next, we investigate the role of OIP5-AS1 in H9C2 cells and found that the relative expression of OIP5-AS1 was increased in LPS-treated H9C2 cells ($P \leq 0.01$; Figure 3(a)). We then silenced OIP5-AS1 and discovered that cells transfected with sh-OIP5-AS1-1 and sh-OIP5-AS1-2 both exhibited decreased OIP5-AS1 expression compared with the control ($P \leq 0.01$; Figure 3(b)). We also measured the levels of two inflammatory cytokines, including IL-6 and TNF-α. The levels of IL-6 and TNF-α in LPS-treated cells were approximately three times higher than those in the control group. Meanwhile, the promoting effects of LPS on the levels of IL-6 and TNF-α were reversed by sh-OIP5-AS1-1 ($P \leq 0.01$; Figures 3(c) and 3(d)). In summary, knockdown of OIP5-AS1 could partially attenuate LPS-induced production of inflammatory cytokines, including IL-6 and TNF-α in H9C2 cells. ## 3.4. Knockdown of OIP5-AS1 Inhibits the Apoptosis and Promotes the Proliferation of LPS-Treated H9C2 Cells To explore the function of OIP5-AS1 on the apoptosis of H9C2 cells, OIP5-AS1 was knockdown, and the apoptosis rate of H9C2 cells exposed to LPS was detected. Flow cytometry assay demonstrated that the apoptosis rate of H9C2 cells was increased after LPS treatment, whereas it was decreased due to the transfection of sh-OIP5-AS1-1 ($P \leq 0.01$; Figure 4(a)). Bcl-2 is an apoptotic suppressor, whereas c-caspase 3 is a promoter of apoptosis [16, 17]. We then measured the expression of these two proteins in H9C2 cells using western blot. The results showed that the expression of Bcl-2 was decreased, and the expression of Bax and c-caspase3/caspase3 ratio was increased following LPS treatment. Moreover, the effects of LPS on these proteins were reversed by the transfection of sh-OIP5-AS1-1 in H9C2 cells ($P \leq 0.01$; Figure 4(b)). In addition, knockdown of OIP5-AS1 also reversed LPS-induced inhibition on the viability of H9C2 cells ($P \leq 0.01$; Figure 4(c)). Taken together, these results indicated that the knockdown of OIP5-AS1 could prevent H9C2 cells from LPS-induced apoptosis and inhibition of cell proliferation. ## 3.5. OIP5-AS1 Directly Targets miR-25-3p Based on the predicting results of starBase4.0, WT OIP5-AS1 could target miR-25-3p ($P \leq 0.01$; Figure 5(a)). The expression of miR-25-3p was increased by up to threefold after OIP5-AS1 knockdown ($P \leq 0.01$; Figure 5(b)). The transfection of miR-25-3p mimics effectively up-regulated miR-25-3p in cells ($P \leq 0.01$; Figure 5(c)). The potential interaction between OIP5-AS1 and miR-25-3p was then verified by luciferase reporter assay. Cells co-transfected with OIP5-AS1-wt and miR-25-3p mimics displayed decreased relative luciferase activity compared with other groups, which indicated that OIP5-AS1 interacted with miR-25-3p ($P \leq 0.01$; Figure 5(d)). Furthermore, the expression of miR-25-3p was negatively correlated with the expression of OIP5-AS1 ($$P \leq 0.0004$$; Figure 5(f)). This finding was supported by the results of RIP assay that more OIP5-AS1 was co-precipitated when miR-25-3p was overexpressed. These results indicated that OIP5-AS1 directly interacts with miR-25-3p. ## 3.6. The Regulatory Role of OIP5-AS1 in Inflammation Is Associated with miR-25-3p MiR-25-3p was down-regulated both in LPS-induced rats and H9C2 cells compared with controls ($P \leq 0.01$; Figures 6(a) and 6(b)). Relative expression of OIP5-AS1 in H9C2 cells was significantly increased after transfection with pcDNA-OIP5-AS1 ($P \leq 0.01$; Figure 6(c)). Overexpression of miR-25-3p reduced IL-6 and TNF-α levels in LPS-treated H9C2 cells. The same effects of miR-25-3p were also discovered even though OIP5-AS1 was overexpressed in LPS-treated H9C2 cells, which indicated that OIP5-AS1 contributed to LPS-induced inflammation via regulating miR-25-3p ($P \leq 0.01$; Figures 6(d) and 6(e)). Taken together, these results indicated that the regulatory effect of OIP1-AS1 on LPS-induced inflammation was associated with miR-25-3p. ## 3.7. OIP5-AS1 Promotes Cellular Apoptosis and Represses Cellular Proliferation via Regulating miR-25-3p in LPS-Induced H9C2 Cells The apoptosis of LPS-induced H9C2 cells was significantly repressed by miR-25-3p mimics in cells transfected or not transfected with pcDNA-OIP5-AS1 ($P \leq 0.01$; Figure 7(a)). Western blot assays showed that overexpression of miR-25-3p increased Bcl-2 expression, and reduced Bax expression and the c-caspase3/caspase3 ratio in LPS-induced H9C2 cells regardless of OIP5-AS1 overexpression or not ($P \leq 0.01$; Figure 7(b)). The viability of LPS-induced H9C2 cells was increased when cells were transfected with miR-25-3p mimics ($P \leq 0.01$; Figure 7(c)). Collectively, these results demonstrated that OIP5-AS1 promoted the apoptosis and repressed the proliferation of LPS-induced H9C2 cells via regulating miR-25-3p. ## 3.8. MiR-25-3p Blocks the NOX4/NF-κB Signalling Pathway in LPS-Induced H9C2 Cells To further reveal the downstream mechanisms of miR-25-3p in myocardial injury, the target mRNAs of miR-25-3p were predicated by miRDB. NOX4 was determined as a potential target of miR-25-3p (Figure 8(a)). Luciferase reporter assay showed that the luciferase activity was lower in cells co-transfected with NOX4 WT and miR-25-3p mimics than those co-transfected with NOX4 WT and miR-NC ($P \leq 0.01$; Figure 8(b)), illustrating the target relationship between NOX4 and miR-25-3p. In addition, the protein expression of NOX4 and p-NF-κB p65/NF-κB p65 was elevated in LPS-induced H9C2 cells compared with the controls ($P \leq 0.01$). The transfection of miR-25-3p mimics significantly decreased the high expression of NOX4 and p-NF-κB p65/NF-κB p65 in LPS-induced H9C2 cells ($P \leq 0.01$; Figure 8(c)). ## 4. Discussion Myocardial injury is a serious pathological change associated with a high risk of mortality. Until now, many strategies have been developed in establishing the model of myocardial injury, such as isoproterenol [18], Adriamycin [13], I/R [14], and sepsis [19]. Since myocardial injury is one of the dominant symptoms of sepsis, the model of sepsis-induced myocardial injury is widely used in animal and cell experiments [20]. In inducing the model of sepsis-induced myocardial injury, LPS is the most commonly used agent with the advantages of simplicity, repeatability, and rapidity, which can activate acute inflammatory response, oxidative stress, and apoptosis [16]. In this study, the model of LPS-induced myocardial injury was established in rats and H9C2 cells. We demonstrated that OIP5-AS1 was up-regulated, but miR-25-3p was down-regulated by the treatment of LPS in rats and H9C2 cells. Loss-of-function analyses indicated that knockdown of OIP5-AS1 relieved LPS-induced apoptosis and inflammation, and LPS-induced inhibition of proliferation by interacting with miR-25-3p. Besides, the NOX4/NF-κB signalling pathway was blocked by miR-25-3p in LPS-induced H9C2 cells. Recent studies have shown that myocardial injury is associated with the abnormal expression of diverse lncRNAs [17, 21]. OIP5-AS1 is a lncRNA that overexpressed in a human umbilical vein endothelial cell (HUVEC) model of atherosclerosis. In HUVECs treated with oxidative low-density lipoprotein, OIP5-AS1 promotes apoptosis and inhibits proliferation [22]. In our study, we demonstrated that OIP5-AS1 was up-regulated in a rat model of LPS-induced myocardial injury and also in LPS-induced H9C2 cells. In addition, knockdown of OIP5-AS1 increased the viability and decreased the apoptosis of LPS-induced H9C2 cells, indicating that OIP5-AS1 may be involved in the progression of myocardial injuries. It has been demonstrated that down-regulation of OIP5-AS1 promotes cell proliferation and suppresses apoptosis in oxidative low-density lipoprotein-mediated endothelial cell injury [23]. OIP5-AS1 can also regulate the inflammatory response. In LPS-activated nucleus pulposus cells, inflammation response was repressed when OIP5-AS1 was silenced [24], which is in agreement with our results that knockdown of OIP5-AS1 led to a decrease in the expression of IL-6 and TNF-α. Similar expression alteration of OIP5-AS1 and relevant effects on cell proliferation, apoptosis, and inflammation are also observed in LPS-induced acute lung injury [6]. Thus, silencing of OIP5-AS1 may be a suppressor in LPS-induced myocardial injury and a meaningful target for treatments. MiRNAs have drawn increasing attention as possible therapeutic targets for myocardial injury [9, 25, 26]. MiR-25-3p is poorly expressed in ox-LDL-induced coronary vascular endothelial cells (CVECs) and vascular tissues. Exosomal miRNA-25-3p derived from platelets represses the inflammatory response in CVECs [27]. MiR-25 protects cardiomyocytes from LPS-induced injury via regulating the expression of inflammatory cytokines [28]. In addition, it has been reported that miR-25 is decreased in LPS-induced cardiomyocytes and its overexpression can inhibit the apoptosis of cardiomyocyte [29]. These results are consistent with our findings that miR-25-3p was down-regulated in LPS-induced H9C2 cells, and its overexpression suppresses the apoptosis and the production of inflammatory cytokines (IL-6 and TNF-α), and promotes the viability of LPS-induced H9C2 cells. Taken together, our results indicate that miR-25-3p may relieve LPS-induced myocardial injury by inhibiting inflammation and apoptosis, and promoting proliferation. MiR-25-3p has also been reported to interact with various lncRNAs in diverse diseases [30–33]. LncRNA can act as competitive endogenous RNA via directly sponging target miRNAs. Here, we found that lncRNA OIP5-AS1 interacted with miR-25-3p, which reduced the expression of miR-25-3p in LPS-treated rat and H9C2 cells. Overexpression of miR-25-3p repressed apoptosis and inflammatory response, and enhanced proliferation in the presence of overexpressed OIP5-AS1. These results indicate that OIP5-AS1 may regulate cell apoptosis, proliferation, and inflammatory response by sponging miR-25-3p. Similar interactions between miR-25-3p and other lncRNAs in cardiovascular diseases have also been reported in previous studies, for example, MALAT1 plays an essential role in myocardial infarction by sponging miR-25-3p [34]. Here, we identified the regulatory interaction of OIP5-AS1 with miR-25-3p in LPS-induced myocardial injury. LncRNAs and miRNAs are involved in the progression of myocardial injury by targeting diverse mRNAs. It has been reported that Krüppel-like factor 4 (KLF4) is targeted by miR-25-3p, and overexpression of KLF4 promotes hypoxia-induced injury in H9C2 cells [35]. Up-regulation of miR-25-3p facilitates the activation of the PI3K/Akt pathway and represses myocardial I/R injury [8]. In this study, NOX4 was determined as a downstream target of miR-25-3p. The following assays revealed that miR-25-3p mimics blocked the NOX4/NF-κB signalling pathway in LPS-induced H9C2 cells. NOX4 is a key enzyme for the production of ROS, which contributes to myocardial injury through inducing ROS [36]. Fan et al. have found that CAPE-pNO2 relieves heart injury in mice with diabetic cardiomyopathy via inhibiting the NOX4/NF-κB pathway [37]. Therefore, the blocking of the NOX4/NF-κB pathway may be involved in the action mechanisms of OIP5-AS1-miR-25-3p axis in LPS-induced myocardial injury. However, this study still has some limitations. For example, whether there is a correlation between OIP5-AS1 expression and clinical characteristics of myocardial injury in human. The function of the NOX4/NF-κB pathway in LPS-induced myocardial injury is not verified. The NOX4/NF-κB pathway is also not the only downstream target of OIP5-AS1-miR-25-3p. More downstream regulators involving cell inflammation, apoptosis, and proliferation in LPS-induced myocardial injury remain need to be studied. In addition, the upstream mechanisms of OIP5-AS1 in LPS-induced myocardial injury are also unclear. The up-regulation of lncRNAs involves frequent somatic copy number alteration, transcription factors, histone modification, and posttranscriptional destabilization [38–41]. Therefore, the upstream mechanisms of LPS-induced up-regulation of OIP5-AS1 also need further investigation. In summary, our study demonstrates that knockdown of OIP5-AS1 may function as a suppressor in apoptosis and inflammation, and as an inhibitor in proliferation in LPS-induced myocardial injury via down-regulating miR-25-3p. ## Data Availability Data supporting this research article are available from the corresponding author or first author on reasonable request. ## Ethical Approval The animal experiments were approved by the Ethics Committee of Suzhou Ninth People's Hospital (Jiangsu, China), and the experimental procedures complied with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. ## Conflicts of Interest The author(s) declare(s) that they have no conflicts of interest. ## Authors' Contributions JM conceived and designed the present study. HQ and HZ performed the experiments, analyzed the data, and drafted the article. JM revised the article critically for important intellectual content. JM, HQ, and HZ confirm the authenticity of all the raw data. All authors read and approved the final manuscript. ## References 1. Zhao Y. B., Zhao J., Zhang L. J.. **MicroRNA-370 protects against myocardial ischemia/reperfusion injury in mice following sevoflurane anesthetic preconditioning through PLIN5-dependent PPAR signaling pathway**. (2019) **113** p. 108697. DOI: 10.1016/j.biopha.2019.108697 2. Zhang X., Wang W., Zhu W.. **Mechanisms and functions of long non-coding RNAs at multiple regulatory levels**. (2019) **20** p. 5573. DOI: 10.3390/ijms20225573 3. Wei J. L., Wu C. J., Chen J. J.. **LncRNA NEAT1 promotes the progression of sepsis-induced myocardial cell injury by sponging miR-144-3p**. (2020) **24** 851-861. DOI: 10.26355/eurrev_202001_20069 4. Liu L., Liu F., Sun Z., Peng Z., You T., Yu Z.. **LncRNA NEAT1 promotes apoptosis and inflammation in LPS-induced sepsis models by targeting miR-590-3p**. (2020) **20** 3290-3300. DOI: 10.3892/etm.2020.9079 5. Chen T., Zhu C., Ye C.. **LncRNA CYTOR attenuates sepsis-induced myocardial injury via regulating miR-24/XIAP**. (2020) **38** 976-985. DOI: 10.1002/cbf.3524 6. Niu X., Pu S., Ling C.. **lncRNA Oip5-as1 attenuates myocardial ischaemia/reperfusion injury by sponging miR-29a to activate the SIRT1/AMPK/PGC1alpha pathway**. (2020) **53** p. e12818. DOI: 10.1111/cpr.12818 7. Zhang J., Zhao T., Tian L., Li Y.. **LncRNA OIP5-AS1 promotes the proliferation of hemangioma vascular endothelial cells via regulating miR-195-5p/NOB1 axis**. (2019) **10** p. 449. DOI: 10.3389/fphar.2019.00449 8. An R., Feng J., Xi C., Xu J., Sun L.. **MiR-146a attenuates sepsis-induced myocardial dysfunction by suppressing IRAK1 and TRAF6 via targeting ErbB4 expression**. (2018) **2018**. DOI: 10.1155/2018/7163057 9. Henning R. J.. **Cardiovascular exosomes and microRNAs in cardiovascular physiology and pathophysiology**. (2021) **14** 195-212. DOI: 10.1007/s12265-020-10040-5 10. Kalayinia S., Arjmand F., Maleki M., Malakootian M., Singh C. P.. **MicroRNAs: roles in cardiovascular development and disease**. (2021) **50** p. 107296. DOI: 10.1016/j.carpath.2020.107296 11. Ge C., Liu J., Dong S.. **MiRNA-214 protects sepsis-induced myocardial injury**. (2018) **50** 112-118. DOI: 10.1097/SHK.0000000000000978 12. Zhou H., Wang X., Zhang B.. **Depression of lncRNA NEAT1 antagonizes LPS-evoked acute injury and inflammatory response in alveolar epithelial cells via HMGB1-RAGE signaling**. (2020) **2020**. DOI: 10.1155/2020/8019467 13. Wang F., Shu F., Wang X. Q.. **Sevoflurane ameliorates adriamycin-induced myocardial injury in rats through the PI3K/Akt/GSK-3beta pathway**. (2021) **25** 968-975. DOI: 10.26355/eurrev_202101_24666 14. Zhou M., Yu Y., Luo X.. **Myocardial ischemia-reperfusion injury: therapeutics from a mitochondria-centric perspective**. (2021) **146** 781-792. DOI: 10.1159/000518879 15. Ward P. A.. **The sepsis seesaw: seeking a heart salve**. (2009) **15** 497-498. DOI: 10.1038/nm0509-497 16. Xianchu L., Lan P. Z., Qiufang L.. **Naringin protects against lipopolysaccharide-induced cardiac injury in mice**. (2016) **48** 1-6. DOI: 10.1016/j.etap.2016.09.005 17. Wang X., Li X. L., Qin L. J.. **The lncRNA XIST/miR-150-5p/c-Fos axis regulates sepsis-induced myocardial injury via TXNIP-modulated pyroptosis**. (2021) **101** 1118-1129. DOI: 10.1038/s41374-021-00607-4 18. Song L., Srilakshmi M., Wu Y., Saleem T. S. M.. **Sulforaphane attenuates isoproterenol-induced myocardial injury in mice**. (2020) **2020**. DOI: 10.1155/2020/3610285 19. Bi C. F., Liu J., Yang L. S., Zhang J. F.. **Research progress on the mechanism of sepsis induced myocardial injury**. (2022) **15** 4275-4290. DOI: 10.2147/JIR.S374117 20. Kakihana Y., Ito T., Nakahara M., Yamaguchi K., Yasuda T.. **Sepsis-induced myocardial dysfunction: pathophysiology and management**. (2016) **4** p. 22. DOI: 10.1186/s40560-016-0148-1 21. Zhang R., Niu Z., Liu J.. **LncRNA SNHG1 promotes sepsis-induced myocardial injury by inhibiting Bcl-2 expression via DNMT1**. (2022) **26** 3648-3658. DOI: 10.1111/jcmm.17358 22. Wang M., Liu Y., Li C., Zhang Y., Zhou X., Lu C.. **Long noncoding RNA OIP5-AS1 accelerates the ox-LDL mediated vascular endothelial cells apoptosis through targeting GSK-3**. (2019) **11** 1827-1834. PMID: 30972206 23. Zhang C., Yang H., Li Y., Huo P., Ma P.. **LNCRNA OIP5-AS1 regulates oxidative low-density lipoprotein-mediated endothelial cell injury via miR-320a/LOX1 axis**. (2020) **467** 15-25. DOI: 10.1007/s11010-020-03688-9 24. Wang S. M., Liu G. Q., Xian H. B., Si J. L., Qi S. X., Yu Y. P.. **LncRNA NEAT1 alleviates sepsis-induced myocardial injury by regulating the TLR2/NF-kappaB signaling pathway**. (2019) **23** 4898-4907. DOI: 10.26355/eurrev_201906_18078 25. Li Y., Li Z., Liu J., Liu Y., Miao G.. **MiR-190-5p alleviates myocardial ischemia-reperfusion injury by targeting PHLPP1**. (2021) **2021**. DOI: 10.1155/2021/8709298 26. Kura B., Szeiffova Bacova B., Kalocayova B., Sykora M., Slezak J.. **Oxidative stress-responsive microRNAs in heart injury**. (2020) **21** p. 358. DOI: 10.3390/ijms21010358 27. Yao Y., Sun W., Sun Q.. **Platelet-derived exosomal microRNA-25-3p inhibits coronary vascular endothelial cell inflammation through Adam10 via the NF-kappaB signaling pathway in ApoE(-/-) mice**. (2019) **10** p. 2205. DOI: 10.3389/fimmu.2019.02205 28. Zhang H., Caudle Y., Shaikh A., Yao B., Yin D.. **Inhibition of microRNA-23b prevents polymicrobial sepsis-induced cardiac dysfunction by modulating TGIF1 and PTEN**. (2018) **103** 869-878. DOI: 10.1016/j.biopha.2018.04.092 29. Yao Y., Sun F., Lei M.. **MiR-25 inhibits sepsis-induced cardiomyocyte apoptosis by targetting PTEN**. (2018) **38**. DOI: 10.1042/BSR20171511 30. Che Z., Xueqin J., Zhang Z.. **LncRNA OIP5-AS1 accelerates intervertebral disc degeneration by targeting miR-25-3p**. (2021) **12** 11201-11212. DOI: 10.1080/21655979 31. Qian W., Zheng Z. Q., Nie J. G.. **LncRNA SNHG12 alleviates hypertensive vascular endothelial injury through miR-25-3p/SIRT6 pathway**. (2021) **110** 651-661. DOI: 10.1002/JLB.1A0820-501R 32. Chen X., Yang J., Wang Y.. **LncRNA JPX promotes cervical cancer progression by modulating miR-25-3p/SOX4 axis**. (2020) **20** p. 441. DOI: 10.1186/s12935-020-01486-3 33. Ren Z., Tang L., Ding Z., Song J., Zheng H., Li D.. **Knockdown of lncRNA JPX suppresses IL-1beta-stimulated injury in chondrocytes through modulating an miR-25-3p/PPID axis**. (2022) **24** p. 388. DOI: 10.3892/ol.2022.13508 34. Chen B., Luo L., Wei X.. **M1 bone marrow-derived macrophage-derived extracellular vesicles inhibit angiogenesis and myocardial regeneration following myocardial infarction via the MALAT1/MicroRNA-25-3p/CDC42 axis**. (2021) **2021**. DOI: 10.1155/2021/9959746 35. Lu G., Cheng Z., Wang S.. **Knockdown of long noncoding RNA SNHG14 protects H9c2 cells against hypoxia-induced injury by modulating miR-25-3p/KLF4 axis**. (2021) **77** 334-342. DOI: 10.1097/FJC.0000000000000965 36. Wang F., Wang H., Liu X.. **Neuregulin-1 alleviate oxidative stress and mitigate inflammation by suppressing NOX4 and NLRP3/caspase-1 in myocardial ischaemia-reperfusion injury**. (2021) **25** 1783-1795. DOI: 10.1111/jcmm.16287 37. Fan L., Xiao Q., Zhang L.. **CAPE-pNO(2) attenuates diabetic cardiomyopathy through the NOX4/NF-kappaB pathway in STZ-induced diabetic mice**. (2018) **108** 1640-1650. DOI: 10.1016/j.biopha.2018.10.026 38. Hu X., Feng Y., Zhang D.. **A functional genomic approach identifies FAL1 as an oncogenic long noncoding RNA that associates with BMI1 and represses p21 expression in cancer**. (2014) **26** 344-357. DOI: 10.1016/j.ccr.2014.07.009 39. Xie J. J., Jiang Y. Y., Jiang Y.. **Super-enhancer-driven long non-coding RNA LINC01503, regulated by TP63, is over-expressed and oncogenic in squamous cell carcinoma**. (2018) **154** 2137-2151. DOI: 10.1053/j.gastro.2018.02.018 40. Zhang E., Han L., Yin D.. **H3K27 acetylation activated-long non-coding RNA CCAT1 affects cell proliferation and migration by regulating SPRY4 and HOXB13 expression in esophageal squamous cell carcinoma**. (2017) **45** 3086-3101. DOI: 10.1093/nar/gkw1247 41. Hammerle M., Gutschner T., Uckelmann H.. **Posttranscriptional destabilization of the liver-specific long noncoding RNA HULC by the IGF2 mRNA-binding protein 1 (IGF2BP1)**. (2013) **58** 1703-1712. DOI: 10.1002/hep.26537
--- title: 'Patients with Hypocortisolism Treated with Continuous Subcutaneous Hydrocortisone Infusion (CSHI): An Option for Poorly Controlled Patients' authors: - Malene Lyder Mortensen - Marie Juul Ornstrup - Claus H. Gravholt journal: International Journal of Endocrinology year: 2023 pmcid: PMC10042637 doi: 10.1155/2023/5315059 license: CC BY 4.0 --- # Patients with Hypocortisolism Treated with Continuous Subcutaneous Hydrocortisone Infusion (CSHI): An Option for Poorly Controlled Patients ## Abstract ### Objective Despite appropriate oral glucocorticoid replacement therapy, patients with hypocortisolism often suffer from impaired health and frequent hospitalizations. Continuous subcutaneous hydrocortisone infusion (CSHI) has been developed as an attempt to improve the health status of these patients. The objective of this study was to compare the effects of CSHI to conventional oral treatment on hospitalizations, glucocorticoid doses, and subjective health status. Patients. Nine Danish patients (males: 4 and females: 5) with adrenal insufficiency (AI) were included, with a median age of 48 years, due to Addison ($$n = 4$$), congenital adrenal hyperplasia ($$n = 1$$), steroid induced secondary adrenal insufficiency ($$n = 2$$), morphine induced secondary adrenal insufficiency ($$n = 1$$), and Sheehan's syndrome ($$n = 1$$). Only patients with severe symptoms of cortisol deficit on oral treatment were selected for CSHI. Their usual oral hydrocortisone doses varied from 25–80 mg per day. The duration of follow-up depended on when the treatment was changed. The first patient started CSHI in 2009 and the last in 2021. ### Design A retrospective case series comparing hospitalizations and glucocorticoid doses before and after treatment with CSHI. In addition, patients were retrospectively interviewed about their health-related quality of life (HRQoL) after the change of treatment modality. ### Results Patients significantly reduced their daily dose of glucocorticoids by 16.1 mg ($$p \leq 0.02$$) after changing to CSHI. The number of hospital admission due to adrenal crisis decreased by 1.3 per year on CSHI, which was a $50\%$ reduction ($$p \leq 0.04$$). All patients found it easier to handle an adrenal crisis with CSHI, and almost all patients found it easier to overcome everyday activities and had fewer symptoms of cortisol deficit such as abdominal pain and nausea (7-8 out of 9 patients). ### Conclusions The change of treatment from conventional oral hydrocortisone to CSHI resulted in a reduced daily dose of glucocorticoids and a reduced number of hospitalizations. Patients reported regain of energy, achievement of better disease control, and better handling of adrenal crisis. ## 1. Introduction Patients with adrenal insufficiency (AI) have inadequate production of cortisol [1]. The condition requires life-long glucocorticoid (GC) replacement therapy and stress adaptation to prevent adrenal crisis. Despite adequate treatment, patients with AI have an increased mortality risk and report fatigue in addition to reduced quality of life (QoL), questioning the quality of current treatment regimens [2–5]. The conventional replacement therapy shows heterogeneity across Europe, but the most frequently used is oral hydrocortisone (OHC) administered twice or thrice daily, with the highest dose administered in the morning [6]. Under normal physiological conditions, the circulating levels of cortisol follow a distinct circadian rhythm with the lowest concentration in the evening and during the night, replaced by rising levels in the early morning. Around awakening, a peak in the cortisol level appears, followed by declining concentrations during the daytime [7]. The impaired health and QoL in patients with AI may reflect that the conventional OHC replacement therapy fails to mimic this physiological pattern of cortisol secretion. In fact, treatment with OHC often leads to alternately unphysiological high or low levels of circulating GC throughout the day, and the average patient on conventional treatment receives a daily dose of GC that exceeds the expected need [8]. Excessive exposure to GC is associated with many adverse effects including hypertension, obesity, osteoporosis, and neuropsychiatric symptoms [9, 10]. An alternative and novel treatment possibility of AI is continuous subcutaneous hydrocortisone infusion (CSHI), which has been able to imitate the physiological cortisol circadian rhythmicity better than oral treatment [11–13]. The treatment with CSHI has so far been reserved for special cases due to a lack of scientific evidence and the complexity of living with a pump, in addition to the high price of pumps and inaccessibility to getting a pump. At present, there is a paucity of studies investigating the treatment of AI with CSHI. Here, we report the experience of 9 Danish patients with AI treated with CSHI, which contributes to the current understanding of the possibilities CSHI provides in treating AI. We report patient's experiences with CSHI and investigate the effect on daily GC dose, body mass index (BMI), hospital admissions, and HRQoL. ## 2. Materials and Methods Nine patients with different forms of AI were treated with CSHI based on clinical evaluation at Aarhus University Hospital, Denmark. Patients were referred from hospitals in all parts of Denmark, and only patients who suffered from severe symptoms of cortisol deficit on standard OHC treatment were selected for CSHI and are presented in this case series (Table 1). In February 2022, medical records were reviewed to gather information about hospitalizations and GC dose before and after treatment with CSHI. Hospitalizations were counted as admissions per year, starting from the patients' first admission due to AI. All hospital admissions were included except elective procedures, and they were divided into 3 groups: emergency department visits, hospital admissions due to adrenal crisis, and hospital admissions without an adrenal crisis (Supplementary information 1). Hospital admissions with adrenal crisis were defined as admissions due to symptoms of adrenal crisis such as nausea, vomiting, and abdominal pain, and/or where patients were treated with intravenous GCs (see further details in Supplementary information 1). Patients were retrospectively interviewed about their experiences with the change of treatment regimen. Questionnaires were developed with inspiration from AddiQoL, which is an Addison-specific health-related questionnaire [14]. The questionnaire in this study consists of 20 questions with different topics including fatigue, psychological aspects, AI-symptoms, sleep, and everyday life (Supplementary information 2). The questionnaire was used to evaluate how the CSHI treatment affected QoL and the subjective well-being of the patients. All patients were asked to focus their question response on before and after the change of treatment. Prior to the CSHI treatment, all patients were treated with oral GC replacement therapy with different regimens. Four patients received hydrocortisone thrice a day with the largest dose administered in the morning. Five patients were treated with Plenadren in the morning, a once-daily dose with a dual-release system, however, also supplemented with hydrocortisone in the afternoon (Supplementary information 3). For CSHI we used hydrocortisone (Solu-Cortef) of 100 mg/2 ml solution in an insulin pump (model 740G, Medtronic). The pump had a basal rate setting, providing the patients with a basal daily dose of hydrocortisone. Basal infusion rates were calculated with inspiration from the feasibility study of Løvås and Husebye [11]. To determine the initial total daily dose, we considered the body weight of the patient, the previous oral dose of hydrocortisone, and a thorough discussion with the patient concerning the need for GC in their current situation. Infusion rates were divided into time slots, and the main principle was to provide cortisol in a physiological pattern with the highest doses around awakening and decreasing doses throughout the day. The individual doses were continuously adjusted based on clinical symptomatology to reach the lowest possible dose. The pump also allowed bolus infusions, which made it possible for patients to get extra cortisol when needed. Patients were educated to self-manage the pump, including knowledge on how to manage pump failure, how to change the cannulas, how to self-administer bolus doses in case of fever or other stress situations, and sick days. ## 3. Statistics Changes in GC doses, hospital admissions, and BMI before and after the introduction of CSHI were analyzed by paired t- test in Excel and considered significant at a level of $p \leq 0.05.$ We present data with median and interquartile range, or mean and standard deviation, as appropriate, depending on normality or non-normality distribution of data. ## 3.1. Ethics The case series was performed as part of an evaluation of CSHI and was approved by the hospital review board. All patients consented to participate in this study. ## 4. Results The patients in this case series were men and women aged between 36 and 58 years (Table 1). All patients were diagnosed with AI, and most of them had additional comorbidity including type 1 diabetes, ulcerative colitis, and chronic pain conditions. They had been treated with CSHI for varying periods of time. The first patient started CSHI in 2009 and the last patient in September 2021. On average, the number of hospital admissions due to adrenal crisis decreased by $50\%$ after switching from conventional treatment to the CSHI regimen (from 2.6 hospital admissions per year to 1.3 hospital admissions per year, $$p \leq 0.04$$) (Table 2). Four patients had not been hospitalized in the years around the change of treatment. The other 5 patients had all reduced their number of hospital admissions after switching to CSHI. Patients 5 and 8 had been hospitalized with adrenal crisis before CSHI was initiated, but neither have had any hospitalizations with an adrenal crisis after the change of treatment to CSHI. Eight out of nine patients received a lower daily dose of GCs with CSHI compared to OHC treatment (Figure 1). The average daily dose equivalent during conventional treatment was 47.5 mg hydrocortisone, which was significantly reduced by $34\%$ to 31.4 mg hydrocortisone on CSHI ($$p \leq 0.02$$). Five out of eight patients lost weight after changing to CSHI (Table 3), however, the average BMI did not change ($$p \leq 0.4$$). One patient was not included because of missing information. ## 4.1. Subjective Health Status All patients found it easier to handle symptoms of adrenal crisis after changing to CSHI, and they all preferred CSHI over standard oral treatment (Table 4). In addition, most found it easier to overcome daily activities, experienced a positive mood change, and had less abdominal pain and nausea (8 out of 9 patients). Furthermore, most felt more rested in the morning and more energetic during the day (7 out of 9). Most experienced a positive effect on their cognitive function (5 out of 9), while potential symptoms of GC deficit, such as muscle pain and headache, were reduced in three and four patients, respectively, however not all patients experienced those symptoms before initiating CSHI. Some experienced better quality of sleep and surplus of energy for social activities (4 out of 9), and an increased sex drive (4 out of 9), while none reported fewer infections. When considering the potential downsides of having a CSHI system, six of nine patients found CSHI more troublesome than conventional oral treatment, and two of nine reported psychological problems with regard to wearing the pump. Finally, all patients were asked to grade their general well-being; “How are you today, compared to before you got the pump, on a scale from 1 to 10”? One represents “miserable” and 10 represents “great.” All patients experienced considerable and significant improvements regarding their overall well-being (Table 5). They graded themselves with very low well-being while on conventional OHC (range 1–4), with a marked increase after switching to CSHI (range 4–10), with a total average change of +6.8 ($p \leq 0.001$) (Table 5). ## 5. Discussion In this case series, we present nine patients with hypocortisolism selected for treatment with CSHI due to very poor disease-control on standard OHC. The patients generally received high doses of OHC before switching to CSHI, with an average dose equaling 47.5 mg hydrocortisone per day, and some of these were treated with doses as high as 80 mg OHC per day. It is common that patients treated with OHC often receive too high doses of GCs, but conversely have an insufficient cortisol coverage when higher doses are needed during infections and stressful situations [8, 15]. In adults with Addison's disease (AD), the recommended daily dosage of hydrocortisone is approximately 15–25 mg per day [16]; thus, our study cohort was grossly overtreated with supraphysiological doses of OHC at baseline, and the high dose was actually one of the reasons for switching the patients to CSHI. Chronic excessive GC treatment has many adverse side effects such as osteoporosis, gain of weight, infections, increased risk of the metabolic syndrome, and Cushing's syndrome [9, 15]. When compared to the general population, patients with AD have a 2-fold higher mortality [17]. It is reasonable to suggest that the reduced subjective health status and the higher mortality in patients with AI may relate to long-term treatment with supraphysiological doses of GCs, and undoubtedly, a reduction in GC dose could be beneficial for the patients [17]. Patients in the present study significantly reduced their daily dose of hydrocortisone by $34\%$, corresponding to a daily dose reduction of 16.1 mg hydrocortisone after changing to CSHI; however, many still received a dose that is higher than recommended for which we have no ready explanations. Several other case series also found that patients were able to reduce their GC dose with CSHI. In 2007 Løvås and Husebye [11] did a pilot study aiming to investigate the technical feasibility, tolerability, and safety of CSHI. They treated 7 patients with AD with CSHI for up to three months and found that patients were able to reduce their GC dose by 40–$85\%$. Khanna et al. [ 18] treated 3 AI patients with CSHI for an average of 17 months. Patients in this study reduced the daily dose of hydrocortisone by $14\%$ on average compared to the previous OHC dose. Cardini et al. [ 19] used CSHI to treat a 42-year-old woman with secondary AI. After 14 months, the steroid dose was lowered by $15\%$ and in addition to this beneficial effect, an AI-specific quality-of-life questionnaire (AddiQoL) score also improved. Standard OHC replacement treatment fails to mimic the physiological circadian rhythm of circulating cortisol levels, and this may also be part of the explanation for why patients with AI suffer from fatigue, reduced quality of life, and impaired health, especially after many years of AI [20, 21]. Our data show that all 9 patients in the current study felt much better on CSHI than on standard OHC treatment, with a greater ability to overcome everyday activities and the feeling of having more energy during the day. However, there is a risk of recall bias, since the evaluation was performed retrospectively. A problem with OHC replacement therapy is that the early morning rise in cortisol levels seen in healthy individuals is absent [22]. The lack of cortisol peak in the morning may explain why patients wake up feeling unrested and unable to do anything before several hours after awakening. Six out of nine patients in the present audit felt more rested in the morning when receiving CSHI treatment, on which we know the circadian rhythm of cortisol is restored, including the early morning peak. Björnsdottir et al. [ 13] conducted a randomized crossover study and compared insulin sensitivity and glucose levels under CSHI versus OHC in 15 patients. All patients completed 8 weeks in each arm of the study. Insulin sensitivity was evaluated by a euglycaemic-hyperinsulinaemic clamp, a method used to quantify insulin resistance. Nighttime glucose levels were more stable in patients receiving CSHI compared to OHC, without compromising insulin sensitivity. In addition, CSHI provided a more physiological circadian cortisol pattern. Løvås and Husebye [11] reported 24-hour salivary and serum cortisol profiles that proved CSHI capable of re-establishing the circadian rhythm of cortisol levels. Our study is not the first to document how a change of treatment from OHC to CSHI improves subjective well-being in patients with AI. Nella et al. [ 23] compared CSHI to conventional OHC treatment in 8 patients with congenital adrenal hyperplasia. The treatment was evaluated by SF-36 Vitality score, AddiQoL, and a fatigue questionnaire, in addition to 24-hour hormonal samplings. After 6 months of follow-up, CSHI had positively affected HRQoL scores, as well as adrenal steroid control. Six patients chose to continue long-term CSHI, and after 18 months, the subjective health status was still improved. The included patients were all classified as “difficult-to-treat,” and when entering the study, all patients had one or more comorbidities. In the pilot study by Løvås and Husebye [11], CSHI improved HRQoL measurements, although only statistically significant for physical functioning and vitality subscales. Patients in the present study were closely monitored by physicians, especially around the change of treatment to CSHI. It is a risk that the reported subjective improvement in quality of life found in this and other studies could be related to closer follow-up by the physicians rather than being caused by the treatment itself. In addition, the fact that patients are suffering from other conditions than AI makes it harder to differentiate between the causes of the experienced changes in symptoms. As an example, two patients suffering from a chronic pain condition, and three patients were diagnosed with type 1 diabetes. It is reasonable to suggest that these conditions could interfere with the quality of life. Ultimately, randomized controlled trials should determine the efficacy of CSHI as a treatment alternative for patients with AI. However, the fact that the patients in our study felt better after the introduction of CSHI was also reflected in their number of hospitalizations. After a change of treatment to CSHI, the number of hospital admissions per year due to the adrenal crisis was halved. This might be explained by a better handling of cortisol deficit and thus avoidance of hospital admission. The duration of treatment with CSHI varies among the patients in this study. Three out of nine patients started CSHI within the last year. It is reasonable to believe that the reduction in hospitalizations per year due to adrenal crisis would be more pronounced with a longer follow-up period since some hospital admissions with adrenal crisis appear around the change of treatment to CSHI that does require more technical skills. In the study by Khanna et al. [ 18], the CSHI system also reduced the number of hospital admissions in 2 out of 3 patients, as well as days spent in the hospital, and thus lowered the estimated cost of care per patient. The authors argued that CSHI was cost-effective compared to standard OHC treatment if the candidates are selected carefully (e.g., if oral administration is unreliable or ineffective). Before changing to CSHI, the included patients had frequently been hospitalized due to treatment failure and adrenal crisis, which was also the case with several patients in our case series. Symptoms of cortisol deficit include nausea and vomiting, which makes oral delivery of cortisol challenging, especially when higher doses are needed. Eight out of nine patients in this cohort found it easier to handle symptoms of ensuing adrenal crisis after they got the pump. Oral treatment is of cause less invasive, and several patients answered that they found treatment with CSHI more troublesome compared to treatment with OHC (6 out of 9). Despite this, all patients preferred CSHI over OHC due to the positive effect on their general well-being. In the study by Khanna et al. [ 18], the treatment failure on OHC also occurred because of gastrointestinal disease and vomiting. Not only did the change of treatment to CSHI reduce the number of hospital admissions in the three patients, but it also improved the subjective well-being of the patients (retrospectively). This emphasizes the thought that CSHI may reduce the number of hospital admissions and improve the quality of life in selected patients experiencing severe symptoms of cortisol deficit on OHC. We performed a rough calculation comparing the cost of pump, medication, and utensils, comparing this with the cost of acute hospitalizations, which most patients experienced many times, and showed that avoidance of just 1-2 such acute hospitalization per year could more than cover the cost of CSHI. Needless to say, such cost-effect calculations should be performed in more depth by health economics researchers that have more experience with also including quality of life, and other costs. There is a lack of larger trials comparing the effect of CSHI with conventional OHC, and besides the study by Björnsdottir et al. [ 13], only two randomized clinical trials (RCTs) have been conducted. The first was a multicenter, crossover, and open-labeled RCT conducted by Oksnes et al. [ 24] in 2014. The study showed promising results when 3 months of treatment with thrice-daily OHC was compared to CSHI in 32 patients with AD. The primary outcome was ACTH levels, and CSHI yielded a normalization of morning ACTH and serum cortisol, in contrast to low serum cortisol levels found on OHC, thus re-establishing the circadian cortisol rhythm. In addition, AddiQoL, and certain subscales of more generic questionnaires, improved with CSHI, although the effect on the HRQoL scores was not as obvious as in the pilot study conducted by Løvås et al. [ 11]. Patients in the pilot study were included due to poor functioning on standard OHC, while in the RCT study by Oksnes et al. [ 24], patients were included regardless of their health status. The more significant rise in AddiQol scores in the pilot study indicates that CSHI is more beneficial for patients with a poorly controlled disease on OHC. Gagliardi et al. [ 25] also published a randomized double-blind placebo-controlled crossover trial in 2014. Patients received CSHI plus oral placebo, or OHC plus placebo infusion, in 4 weeks in a random order, with a 2-week washout period. Ten patients with AD completed the trial. The primary outcome was subjective health status measured with several different health-related questionnaires. The study casted some doubt on the potential benefit from CSHI, as they found no significant change in subjective health status when OHC was compared to CSHI. However, the included patients had a better baseline subjective health status compared with other AD cohorts. The low statistical power was another limitation, as only 10 patients completed the study. Furthermore, it is possible that 4 weeks of treatment was not enough time to detect changes in the subjective health status. Overall, results from former studies indicate that AI patients with a bad baseline status benefit from a change of treatment from OHC to CSHI, but this might not be the case for well-functioning patients, and results from the RCTs emphasize the importance of selecting the right patients for CSHI. The present study is limited by the nature of it being a case series. To infer a more precise cause-effect relationship, RCTs are needed, but since AI is a rare disease RCTs can be difficult to perform due to a lack of patients. Thus, we do believe this study offers useful knowledge for elucidating the use of CSHI. Patients in this study experienced symptoms of inappropriate GC replacement on OHC and were strictly selected for treatment with CSHI. *In* general, findings in case reports provide little basis for generalization; however, this was not just a single case. Several AI patients with a poorly controlled disease have been found to benefit from CSHI. This strengthens the evidence for the use of CSHI in selected patients and it emphasizes the demand for further research. We conclude that the change of treatment from oral GCs to CSHI resulted in better disease control on a lower daily dose of GCs and improved quality of life. In addition, the dose of GC was reduced, and likewise was the number of hospital admissions. These results provide further evidence for the use of CSHI in treating selected patients with poorly controlled AI. However, the sparse literature calls for larger studies to investigate both short-and long-term health benefits offered by CSHI in poor functioning patients with AI. Although the recent studies are promising, and CSHI seems to offer a more physiological cortisol replacement than OHC, in addition to a reduction in the daily GC exposure, a bigger randomized placebo-controlled trial on selected patients with poorly controlled AI is needed. Focus should center around effects on HRQoL, hospitalization, and signs of under or overreplacement with GC. ## Data Availability The data that supports the findings of this study are available in the supplementary material of this article. ## Conflicts of Interest The authors declare that they have no conflicts of interest. ## References 1. Charmandari E., Nicolaides N. C., Chrousos G. P.. **Adrenal insufficiency**. (2014) **383** 2152-2167. DOI: 10.1016/s0140-6736(13)61684-0 2. Hahner S., Loeffler M., Fassnacht M.. **Impaired subjective health status in 256 patients with adrenal insufficiency on standard therapy based on cross-sectional analysis**. (2007) **92** 3912-3922. DOI: 10.1210/jc.2007-0685 3. Bancos I., Hazeldine J., Chortis V.. **Primary adrenal insufficiency is associated with impaired natural killer cell function: a potential link to increased mortality**. (2017) **176** 471-480. DOI: 10.1530/eje-16-0969 4. Løvås K., Loge J. H., Husebye E. S.. **Subjective health status in Norwegian patients with Addison’s disease**. (2002) **56** 581-588. DOI: 10.1046/j.1365-2265.2002.01466.x 5. Chantzichristos D., Persson A., Eliasson B.. **Mortality in patients with diabetes mellitus and Addison’s disease: a nationwide, matched, observational cohort study**. (2017) **176** 31-39. DOI: 10.1530/eje-16-0657 6. Murray R. D., Ekman B., Uddin S., Marelli C., Quinkler M., Zelissen P. M.. **Management of glucocorticoid replacement in adrenal insufficiency shows notable heterogeneity - data from the EU-AIR**. (2017) **86** 340-346. DOI: 10.1111/cen.13267 7. Oster H., Challet E., Ott V.. **The functional and clinical significance of the 24-hour rhythm of circulating glucocorticoids**. (2017) **38** 3-45. DOI: 10.1210/er.2015-1080 8. Simon N., Castinetti F., Ouliac F., Lesavre N., Brue T., Oliver C.. **Pharmacokinetic evidence for suboptimal treatment of adrenal insufficiency with currently available hydrocortisone tablets**. (2010) **49** 455-463. DOI: 10.2165/11531290-000000000-00000 9. Oray M., Abu Samra K., Ebrahimiadib N., Meese H., Foster C. S.. **Long-term side effects of glucocorticoids**. (2016) **15** 457-465. DOI: 10.1517/14740338.2016.1140743 10. Walker B. R.. **Glucocorticoids and cardiovascular disease**. (2007) **157** 545-559. DOI: 10.1530/eje-07-0455 11. Løvås K., Husebye E. S.. **Continuous subcutaneous hydrocortisone infusion in Addison’s disease**. (2007) **157** 109-112. DOI: 10.1530/eje-07-0052 12. Mallappa A., Nella A. A., Sinaii N.. **Long-term use of continuous subcutaneous hydrocortisone infusion therapy in patients with congenital adrenal hyperplasia**. (2018) **89** 399-407. DOI: 10.1111/cen.13813 13. Björnsdottir S., Øksnes M., Isaksson M.. **Circadian hormone profiles and insulin sensitivity in patients with Addison’s disease: a comparison of continuous subcutaneous hydrocortisone infusion with conventional glucocorticoid replacement therapy**. (2015) **83** 28-35. DOI: 10.1111/cen.12670 14. Øksnes M., Bensing S., Hulting A. L.. **Quality of life in European patients with Addison’s disease: validity of the disease-specific questionnaire AddiQoL**. (2012) **97** 568-576. DOI: 10.1210/jc.2011-1901 15. Han T. S., Walker B. R., Arlt W., Ross R. J.. **Treatment and health outcomes in adults with congenital adrenal hyperplasia**. (2014) **10** 115-124. DOI: 10.1038/nrendo.2013.239 16. Bornstein S. R., Allolio B., Arlt W.. **Diagnosis and treatment of primary adrenal insufficiency: an endocrine society clinical practice guideline**. (2016) **101** 364-389. DOI: 10.1210/jc.2015-1710 17. Bergthorsdottir R., Leonsson-Zachrisson M., Odén A., Johannsson G.. **Premature mortality in patients with Addison’s disease: a population-based study**. (2006) **91** 4849-4853. DOI: 10.1210/jc.2006-0076 18. Khanna A., Khurana R., Kyriacou A., Davies R., Ray D. W.. **Management of adrenocortical insufficiency with continuous subcutaneous hydrocortisone infusion: long-term experience in three patients**. (2015) **2015**. DOI: 10.1530/edm-15-0005 19. Cardini F., Torlone E., Bini V., Falorni A.. **Continuous subcutaneous hydrocortisone infusion in a woman with secondary adrenal insufficiency**. (2019) **63** 398-400. DOI: 10.1007/s12020-018-1780-4 20. Bleicken B., Hahner S., Loeffler M., Ventz M., Allolio B., Quinkler M.. **Impaired subjective health status in chronic adrenal insufficiency: impact of different glucocorticoid replacement regimens**. (2008) **159** 811-817. DOI: 10.1530/eje-08-0578 21. Forss M., Batcheller G., Skrtic S., Johannsson G.. **Current practice of glucocorticoid replacement therapy and patient-perceived health outcomes in adrenal insufficiency - a worldwide patient survey**. (2012) **12** p. 8. DOI: 10.1186/1472-6823-12-8 22. Merza Z., Rostami-Hodjegan A., Memmott A.. **Circadian hydrocortisone infusions in patients with adrenal insufficiency and congenital adrenal hyperplasia**. (2006) **65** 45-50. DOI: 10.1111/j.1365-2265.2006.02544.x 23. Nella A. A., Mallappa A., Perritt A. F.. **A phase 2 study of continuous subcutaneous hydrocortisone infusion in adults with congenital adrenal hyperplasia**. (2016) **101** 4690-4698. DOI: 10.1210/jc.2016-1916 24. Oksnes M., Björnsdottir S., Isaksson M.. **Continuous subcutaneous hydrocortisone infusion versus oral hydrocortisone replacement for treatment of addison’s disease: a randomized clinical trial**. (2014) **99** 1665-1674. DOI: 10.1210/jc.2013-4253 25. Gagliardi L., Nenke M. A., Thynne T. R. J.. **Continuous subcutaneous hydrocortisone infusion therapy in Addison’s disease: a randomized, placebo-controlled clinical trial**. (2014) **99** 4149-4157. DOI: 10.1210/jc.2014-2433
--- title: 'Navigating biosafety concerns within COVID-19 do-it-yourself (DIY) science: an ethnographic and interview study' authors: - Anna Wexler - Rebekah Choi - Alex Pearlman - Lisa M. Rasmussen journal: Biosocieties year: 2023 pmcid: PMC10042665 doi: 10.1057/s41292-023-00301-2 license: CC BY 4.0 --- # Navigating biosafety concerns within COVID-19 do-it-yourself (DIY) science: an ethnographic and interview study ## Anna Wexler is an Assistant Professor in the Department of Medical Ethics and Health Policy at the Perelman School of Medicine at the University of Pennsylvania. Her research centers on the ethical, legal, and social issues surrounding emerging technology, with a focus on do-it-yourself medicine, direct-to-consumer health products, and neuroethics. ## Rebekah Choi received her MPH from the University of Pennsylvania in 2017. She has worked on various social science research projects and large-scale randomized controlled trials. She currently works as a Project Manager in the Department of Medical Ethics and Health Policy at the Perelman School of Medicine. ## Alex Pearlman is a bioethicist and journalist. Her research focuses on emerging ethical and regulatory issues in the health, science, and biotechnology sectors. She is currently the Director of Communications at Concentric by Gingko, as well as a Research Affiliate with the Community Biotechnology Initiative at the MIT Media Lab. ## Lisa M. Rasmussen is a Professor in the Department of Philosophy and Faculty Fellow in the Graduate School at the University of North Carolina at Charlotte. Her research focuses on research ethics, particularly in unregulated research areas like citizen science and DIY biology. ## Abstract Non-establishment or do-it-yourself (DIY) science involves individuals who may not have formal training conducting experiments outside of institutional settings. While prior scholarship has examined the motivations and values of those involved in the subset of DIY science known as “DIY biology,” little research has addressed how these individuals navigate ethical issues in practice. The present study therefore aimed to understand how DIY biologists identify, approach, and resolve one particular ethical issue—biosafety—in their work. We conducted a digital ethnography of Just One Giant Lab (JOGL), the primary hub for DIY biology during the COVID-19 pandemic, and subsequently conducted interviews with individuals involved with JOGL. We found that JOGL was the first global DIY biology initiative to create a Biosafety Advisory Board and develop formal biosafety guidelines that applied to different groups in multiple locations. There was disagreement, however, regarding whether the Board should have an advisory role or provide mandatory oversight. We found that JOGL practiced ethical gatekeeping of projects that fell outside the limits defined by the Board. Our findings show that the DIY biology community recognized biosafety issues and tried to build infrastructure to facilitate the safe conduct of research. ### Supplementary Information The online version contains supplementary material available at 10.1057/s41292-023-00301-2. ## Introduction Biomedical research is most commonly conducted in institutional settings by researchers with formal training. By contrast, what has been termed “non-establishment research” (Rasmussen et al. 2020) involves individuals who may not have scientific training conducting experiments outside of traditional settings (Landrain et al. 2013; Meyer 2013). This research has been alternately referred to as “do-it-yourself science” (Ferretti 2019), “biomedical citizen science” (Wiggins and Wilbanks 2019; Guerrini, Wexler, et al. 2019; Trejo et al. 2020), “biohacking” (Delfanti 2013; Wexler 2017), “participant-led research” (Kempner and Bailey 2019; Vayena et al. 2016; Grant et al. 2019), or “community biology” (Walker et al. 2020). It includes practices as varied as hacking diabetes devices (Lee et al. 2016; Omer 2016), self-administering unapproved medical treatments (Ekekezie et al. 2020; Wexler 2016), developing treatments for disease (Kempner and Bailey 2019), and experimenting with gene-editing techniques (Pauwels and Denton 2018). This paper focuses on the subset of non-establishment research that has been termed “DIY Biology” or “DIYbio” (Meyer 2013; Landrain et al. 2013; Sundaram 2021), which encompasses a loose-knit group of individuals undertaking a wide range of activities related to biology, which they conduct at home or in local biology labs (Erikainen 2022; Delfanti 2013). Many of those involved with DIY biology refer to themselves as a “community” or as a grassroots movement, and have built relationships with one other, through both in-person interactions (e.g., at community labs and at national and international meetings) as well as via extensive online communication (Wexler 2017; Meyer and Vergnaud 2020; Seyfried et al. 2014). As there is no consensus regarding terminology even among participants (Trejo et al. 2020), here we utilize the term DIY Biology or DIYbio when discussing the abovementioned group, and DIY science when referring to citizen science writ large. Because DIY science is not typically federally funded or conducted at federally sponsored institutions, there are no systematic layers of regulatory oversight to ensure that it has met particular ethical standards (Rothstein et al. 2015; Guerrini et al. 2018; Rasmussen 2021). In the absence of regulations, many have argued that the ethical conduct of DIY science poses unique concerns for DIY communities, policymakers, and conventional research ethics gatekeepers (Edwards 2017; Grant et al. 2019; Vayena and Tasioulas 2013b; Rasmussen 2017; Fiske et al. 2018; Wiggins and Wilbanks 2019; Resnik 2019). Some have proposed that a new “social contract” is needed to enable this kind of research and minimize its pitfalls (Vayena et al. 2016) and have developed frameworks illustrating how ethical issues may differ depending on the degree of involvement of lay individuals (Vayena and Tasioulas 2013a). Other work has assessed DIY biologists’ attitudes toward ethics, finding that they prioritize issues such as failure to return results and power imbalance (Guerrini et al. 2021), and that they are most open to a form of ethical oversight that is voluntary and community-driven (Trejo et al. 2021). One of the most frequently expressed ethical concerns regarding DIY biology is that of biosafety. Concerns have been raised that individuals may cause physical harm to themselves (i.e., by conducting unsafe experiments) or to the public, due to the risk of pathogens escaping the laboratory, either accidentally or deliberately (Kuiken 2016, 2020; Kolodziejczyk 2017; Lim 2021; Sundaram 2021). Those involved with DIY biology have responded to such concerns by pointing to the long history of efforts toward biosafety within the movement, which has included locally established biosafety codes, the “*Ask a* Biosafety Expert” feature on DIYbio.org, and a biosafety training camp that featured involvement from a leading international biosafety organization (Sundaram 2021; Grushkin et al. 2013; Kuiken 2016; Grushkin 2018). Still, efforts to quell biosafety concerns have done little to quash them. Recently, the COVID-19 pandemic brought renewed attention to issues of biosafety within DIY biology, with the media and scholars expressing concerns about the dangers of self-experimentation with DIY vaccines (Caplan and Bateman-House 2020; Shah and Jamrozik 2020; Guerrini, Sherkow, et al. 2020; Murphy 2020). To date, despite the longstanding public interest in biosafety within DIY science, there has been little empirical research on how citizen scientists navigate biosafety issues in practice. While prior sociological scholarship has examined the motivations, values, and politics of DIY biology (Meyer 2013; Barba 2014; Delgado and Callén 2017; McGowan et al. 2017; Roosth 2017; Grant et al. 2019; Guerrini, Trejo, et al. 2020) and other work has cataloged the history of biosafety efforts within DIY biology (Grushkin et al. 2013; Kuiken 2016; Sundaram 2021; Lim 2021), this work has not examined how individuals anticipate and respond to biosafety challenges in their own scientific investigations. The present study aimed to fill this gap by empirically assessing how individuals involved in DIY biology navigated challenges related to biosafety. The COVID-19 pandemic offered a unique setting to examine this question, as it spurred efforts from DIY biologists to coordinate—virtually and globally—on efforts to develop COVID-19 diagnostics and preventatives. Because those involved in DIYbio were forced to communicate primarily via online outlets, both out of geographic necessity and due to quarantine restrictions, there was rich ground for digital ethnography. In addition, the international nature of DIY efforts during the pandemic meant that participants did not necessarily share a common ethical framework, thereby offering an opportunity to study issues as they arose in a global setting that was not dependent on shared ethical values. The present study employed a two-phased approach. In the first phase, we conducted a digital ethnography of Just One Giant Lab (JOGL), the online platform that became the de facto hub for DIYbio during the COVID-19 pandemic. In the second phase, we sought to gain a deeper understanding of our data by conducting follow-up interviews with those involved in biosafety-related issues on the JOGL platform. Our aim was to advance knowledge of how those involved in DIY biology identify, approach, and resolve biosafety-related considerations in their work. ## Just One Giant Lab (JOGL) Just One Giant Lab (JOGL) is an online platform that was launched in 2019 by three longstanding members of the DIY biology community to foster open science collaboration across borders (JOGL 2022). In February 2020, the platform introduced the “OpenCovid19 Initiative,” a dedicated area of the site where individuals could network with others to address COVID-19 issues across five broad areas: diagnostics, validation, treatment, prevention, and data (JOGL 2020a). Individuals could freely create personal profiles as well as “project pages” that described specific endeavors, such as efforts to develop a one-hour diagnostic test, face shields made from recycled plastic, and an app that would use machine learning to compare a normal cough to that of someone infected with COVID (Jorgenson 2020; Lowe 2021; Bektas 2020; Morales 2020). Given the lack of a messaging feature on the JOGL platform, project participants began turning to Slack, a popular workplace communication tool, to converse with one another. Though Slack was initially designed for the quick exchange of messages, both in public and private “channels,” as well as through direct messages, JOGL participants began utilizing Slack for their growing project needs. Activity on the JOGL Slack channel peaked in late March 2020, when there were approximately 600 active members, nearly two thousand registered users, and several dozen channels, which were dedicated both to individual projects and general topics (e.g., events, announcements, onboarding, community, “looking for resources,” and “looking for skills”). Hundreds of new members flooded the JOGL Slack every few days, and by April 2020, Slack had become the primary communication method for all JOGL participants. ## Phase 1: digital ethnography Due to the large volume of conversations across JOGL Slack channels, we employed several methods to narrow our focus to biosafety-related content. First, we met with JOGL leadership and described our project at a weekly public meeting; through this and by browsing JOGL Slack channels, we identified entire channels of interest (such as one dedicated entirely to biosafety). Second, keywords related to biosafety were developed iteratively by all four authors in a process that involved generating terms, revising them during group discussion, testing them for relevance on the JOGL Slack channels, and revising again until all authors concurred with the final keyword list (see Appendix 1). Next, all conversations occurring on public Slack channels between March 1 and August 31, 2020 were searched using these keywords, and conversational snippets occurring before and after the use of this keyword were captured. To make JOGL participants aware of our research, we announced our project at a weekly JOGL Zoom call and pinned a post in the public JOGL Slack announcement channel during data collection (August-December 2020), offering users the opportunity to opt out of having their textual data collected. Search results for each keyword were compiled into individual documents, which were then reviewed, annotated, and discussed jointly by all authors. For both methods (whole channel review and conversational review based on search terms), we sought out and reviewed additional data as necessary (i.e., websites related to specific projects, videos of JOGL meetings, and documents created by JOGL members). In total, we analyzed approximately 1,300 pages of data. Results were analyzed using a codebook which was developed inductively by all four authors based on the principles of grounded theory (Belgrave and Seide 2019; Strauss and Glaser 1967), and iteratively revised through group discussion. ## Phase 2: interviews To gain a deeper understanding of the data gathered in Phase 1, and to clarify and extend our primary observations, we conducted hour-long interviews with those involved with JOGL. We recruited from three subgroups of interest: those involved with formulating biosafety guidelines or those who participated in biosafety discussions (“biosafety group”), those conducting specific COVID-19-related projects (“project participants”), and those holding leadership positions at JOGL (“JOGL leadership”). Detailed information about sampling and recruitment is presented in Appendix 2. The interview guides (Appendix 3) were informed by the data gathered in Phase 1, as well as by preliminary consultations with those involved in either studying or conducting DIYbio. Specific interview questions varied by subgroup, and assessed participants’ prior involvement with DIYbio, their experiences with JOGL, their involvement in or awareness of biosafety guidelines, ethical issues they encountered, and attitudes toward ethics in DIY biology. All interviews were conducted by a single interviewer (AP) via webconference and audio-recorded with participant’s consent. Participants were offered a $40 Amazon e-gift card for completing the interview as compensation for their time. A total of 31 interviews were conducted between February and June 2021 across the biosafety group ($$n = 13$$), project participants ($$n = 12$$), and JOGL leadership ($$n = 6$$). Most interviewees were male ($58.1\%$; $$n = 18$$) and the majority had prior experience with DIYbio ($80.6\%$; $$n = 25$$). Interviewees were located in North America ($$n = 20$$), Europe ($$n = 6$$), Africa ($$n = 2$$), Asia ($$n = 1$$), or reported splitting time between multiple locations ($$n = 2$$). All interviews were transcribed via SpeechPad. The interview codebook was developed inductively following an examination of a subset of transcripts by two authors (AP, RC), and iteratively revised by all four authors. During the first round of coding, two coders (AP, RC) categorized all relevant text through inductive coding using the qualitative analysis software Dedoose; these excerpts were exported and cleaned in Excel. Next, two coders (AW, RC) used code mapping, the process of theming or recategorizing groups of codes, to analyze and combine first-level categories that were identified in the initial coding round. All themes that emerged were independently assessed for accuracy and consistency by two coders (AW, RC). Any disagreements were resolved through discussion and by referencing other data samples. Both Phase 1 and Phase 2 of this study were deemed exempt from review by the University of Pennsylvania Institutional Review Board. ## Results Five primary themes related to biosafety emerged from our data. First, biosafety on JOGL was approached in a unique way: namely, JOGL was the first global DIY biology initiative to have a Biosafety Advisory Board, as well as a detailed set of biosafety guidelines that applied across multiple groups and projects. Second, while there was recognition that these achievements were unprecedented in the realm of open science, those involved with the Biosafety Board or with drafting the guidelines did not view their efforts as particularly novel, but rather as the natural culmination of nearly a decade of work to enhance biosafety practices within DIY biology. Third, both on Slack discussions and in interviews, there was disagreement regarding whether the Board should have an advisory role or provide mandatory oversight of projects. Fourth, in at least one instance, JOGL practiced a kind of ethical gatekeeping, excluding a potential project because it was not in keeping with ethical limits defined by the Biosafety Board. Fifth, individuals differed in their perception of the utility of the biosafety guidelines: while there was an awareness of them among project participants, members of the Biosafety Board and JOGL leadership questioned whether they had provided real-world value to individual projects. Each of these themes are discussed below, drawing from both ethnographic and interview data. Direct quotes are attributed to interviewees according to subject number and group (i.e., “B1” is a biosafety group member, “P1” is a project participant, and “J1” is a member of JOGL leadership). Given the international nature of our sample, some interviewees were not native English speakers. Attributions are not provided for Slack quotations to protect participants’ privacy. ## Theme 1. A first for DIY science: a global biosafety board and biosafety guidelines The early days of the OpenCovid19 Initiative were full of promise: thousands of people interested in open science came together virtually to fill gaps in SARS-COV-2 diagnostics, prevention, and treatment. Whereas previous open science efforts had been disjointed and fractured, with individuals working locally on different projects, in March 2020, individuals focused their attention, jointly, on common problems related to COVID-19. For some in JOGL leadership, the sudden surge of interest and feeling of global unity was emotionally moving; one described it as a quasi-religious experience, as a “feeling of togetherness that you can touch somehow; it was amazing” (J2). At the same time, however, those involved in the OpenCovid19 Initiative began to worry about the prudence and safety of some of the projects that were being proposed on JOGL. One interviewee noted that upon seeing the details of some of the diagnostic and vaccine projects, “my tentacles that this could go wrong were kind of tingling.” ( B4). Another recalled:I think really early on people were talking about like, "Let's go out and collect samples," I think of, like, swabs, swabbing other places, in essence, which would have brought, you know, the virus into their facilities. And really quickly, people started raising alarm bells about being like, "Hold on. Let's think that through" (B2). Because JOGL was an open platform, it presented a specific problem: there was no clear vetting mechanism, which meant anyone could join and propose a project. One interviewee noted that while an open platform like JOGL welcomed everyone, there was a concomitant “risk of welcoming, not necessarily evil doers, but incompetent doers” (B9). On March 11, 2020, just a few weeks after the launch of the OpenCovid19 Initiative, a weekly JOGL community Zoom meeting was held during which Thomas Landrain, the co-founder and CEO of JOGL, proposed the creation of a Biosafety Advisory Board that could both review projects and serve as a community resource to answer questions (JOGL 2020c). The idea was met with enthusiasm by others at the meeting, and that same day, a Slack channel devoted to biosafety was created by JOGL’s program coordinator, with its description stating that it is was dedicated to “discussions/questions about biosafety and biosecurity concerns for the program.” On the channel, Landrain noted that he foresaw the Board as having three main goals: publishing general recommendations, reviewing projects within the initiative and providing guidance and warnings, and answering biosafety and biosecurity questions asked by JOGL members. On the weekly call, an individual who had long been involved in biosafety in DIY science was nominated to serve on the Board; she in turn recruited others, most of whom had been involved in previous biosafety endeavors in DIY biology. The Board included individuals such as Todd Kuiken, a researcher with extensive experience studying and advocating for enhanced biosafety practices within DIYbio; Chris Monaco, a microbiologist and lab equipment designer at the Center for Disease Control (CDC); and various founders and directors of community labs, among others. Efforts coalesced quickly, and by early April, the ten-member Board had published a fifteen-page document outlining biosafety guidelines for the OpenCovid19 Initiative (JOGL 2020b). The guidelines provided both general information about biological risk assessment as well as specific recommendations (e.g., for the use of positive controls and sample collection). They also specified that no human or animal testing should be performed without prior approval, and that testing should not be offered or marketed to the general public. The guidelines were the first detailed, public biosafety guidance to have emerged from the DIY science movement. The speed with which JOGL’s efforts toward biosafety impressed even the Board members:I was happy to see how quickly the folks that sort of were in that JOGL community in the beginning, how quickly they recognized that they needed some additional guidance around biosafety, and how really fast they formulated that committee in case issues came up. I was really glad to see that happened as quickly as it did (B2). While there had been previous efforts toward biosafety within DIY science, JOGL’s version was unique for several reasons. First, while prior endeavors toward guiding the practice of biosafety had largely taken place at community levels (Guerrini et al. 2019a, b), the international nature of JOGL meant that the Biosafety Advisory Board and guidelines spanned boundaries in ways that previous efforts had not. Second, local efforts toward biosafety often occurred informally: the head of a community lab might instruct a new member in best practices in their specific lab by training them, rather than referring to an established manual. The JOGL biosafety guidelines, however, represented a kind of formal, written knowledge of biosafety practices, albeit ones geared specifically toward COVID-19 efforts. Third, the Board and its guidelines had a measure of external credibility, as Dee Zimmerman, the former President of Association for Biosafety and Biosecurity (ABSA), the largest international biosafety association, was a Board member and is listed as a coauthor of the biosafety guidelines. These accomplishments were recognized as significant by those involved in DIY biology. One Board member noted:The fact that the OpenCovid19 Initiative formed a biosafety group composed of people from around the world in open science communities and then they actually made a deliverable happen is huge. Because that wasn't able to be done very well in the past (B6). Another Board member, who has long been involved in biosafety efforts in DIY biology, described JOGL’s biosafety group as a fundamental advance in the sophistication with which open science efforts had evolved in the last decade because “the community itself [recognized]” (B2) the need for ethical oversight and subsequently developed an organizational architecture suited for that purpose. ## Theme 2. Natural culmination of a decade of efforts Despite the uniqueness of JOGL’s achievements in the realm of biosafety, members of the Board did not see their efforts as extraordinary. Rather, as one put it, “it was just a culmination of years of people working on these things” (B2). Some drew a direct line between JOGL’s biosafety endeavors in 2020 and early efforts toward biosafety in open science, which included the development of the Codes of Ethics at European and U.S. DIYbio workshops in 2011 (Eggleson 2014; DIYBio.org 2011). Though these codes were relatively limited in scope and the relevant mention of biosafety comprised just four words (“Safety: adopt safe practices”), it was the first time that the DIY science community had publicly and formally recognized that biosafety should be a guiding principle. The ensuing years saw other endeavors related to biosafety in open science, many of which were supported by external funding. In 2012, the Federal Bureau of Investigation (FBI) held a joint meeting with members of the DIY biology community, in which it paid for individuals to attend and discuss issues related to biosafety (Scroggins 2013; Lempinen 2011). The Wilson Center funded a study on DIY biology through its Synthetic Biology Project (Grushkin et al. 2013; Kuiken et al. 2018), as well as the “*Ask a* Biosafety Expert” feature on DIYbio.org in 2013–2014, an open forum where anyone could ask biosafety-related questions. A grant from Open Philanthropy allowed Todd Kuiken and Dan Grushkin, the co-founder of New York City’s community lab Genspace, to hold a three-day biosafety “boot camp” in 2019 for DIY Biology and community labs, which was presented in partnership with ABSA (Baltimore Under Ground Science Space 2019; ABSA 2019). This same funding also supported their endeavors to evaluate community biosafety procedures and develop best practices. In October 2020, approximately six months after JOGL’s guidelines were released, they and several additional co-authors released a 250-page Community Biology Biosafety Handbook at the Global Community Bio Summit, an annual global DIYbio meeting (Armendariz et al. 2020). In reflecting upon JOGL’s efforts toward biosafety, members of the Board described the Board and its biosafety guidelines as coming together “pretty easily” (B1), “naturally” (B3), and “smoothly” (B4). Interviewees pointed to four main factors that contributed to the ease with which the Board and guidelines came to fruition. First, Board members and JOGL leadership had extensive experience with issues related to biosafety, both in institutional settings as well as in DIYbio. Board member Kuiken, for example, had conducted the Code of Conduct DIYbio workshops, and had led the Wilson Center reports and Open Philanthropy efforts; Board member Jorgenson was the co-founder of the first community DIY science lab in New York City; and Elena Perez-Nadales worked as a biomedical researcher at Maimonides Biomedical Research Institute in Cordoba, Spain. JOGL co-founder Landrain had led the DIY Bio Congress in Paris (Synenergene 2014). This gave them not only experiential expertise but also access to written biosafety documents. As one Board member noted:And yeah, of course, from my experience here at the [redacted] with making [redacted], I have whole like SOPs, standard operating procedures of how to, you know, double glove and all this kind of stuff... Yeah, I used some of my old things (B3). The second main factor was that Board members were individuals who, for the most part, knew each other and had established positive working relationships, largely due to their previous experience with biosafety in DIYbio. In interviews, they typically referred to each other by first name, and often mentioned the esteem and trust in which they held their fellow Board members. These prior working relationships were also what allowed them to coalesce quickly as a group, as they were able to quickly recruit others to the JOGL efforts. One Board member recalled reaching out to other co-authors of the Community Handbook after hearing about JOGL’s interest in biosafety guidelines:As soon as I saw the announcements and had realized the type of work that some of the JOGL community were interested in doing, I emailed the other biosafety authors and said, "Hey, have you seen this? Maybe we should at least like, definitely make sure they have access to the chapters, sort of, even if they're not all online right now as public links. And, you know, would it be worth doing something?" And then I think we agreed on an email thread that that would be worth doing (B4). The third factor cited as facilitating the rapid formation of the Board and the development of the guidelines was prior relationships with outside experts. Because Kuiken and others had already interacted with ABSA at the biosafety boot camp, the Board had a direct line to the organization, and to its past president, Dee Zimmerman. This enabled the Board to obtain informal advice and feedback, thereby ensuring the guidelines they had put together were valid and credible. As one Board member noted:Connections that we had through ABSA… I think that was pretty huge having that. I don't know if I'd say it's an endorsement, but having a name like ABSA, be, I guess, somewhat affiliated with it, I think, gave it a lot of credibility… (B3). The fourth factor was that in March 2020, when JOGL expressed interest in creating COVID-19 biosafety guidelines, there had already been a significant amount of work put into the drafting of the Community Biosafety Lab Handbook that would be released just months later:Most of the work [on the handbook] had been completed at that time. I think the copy editing was still ongoing, and like it hadn't been put into one giant document at that moment… So, definitely, having the draft or un-copy-edited version of the biosafety handbook was helpful (B4). Many Board members mentioned that they drew largely on this handbook when crafting the biosafety guidelines. While only four authors out of the dozen on the Community Biosafety handbook were also on JOGL’s Biosafety Advisory Board, drawing on the handbook effectively allowed the Board to benefit from the efforts of the larger Community Biosafety handbook group. However, Board members also sourced additional material from other outlets, as the handbook did not contain specific information related to COVID-19. On the Slack biosafety channel, Board members posted links to guidelines from the Center for Disease Control (CDC), the World Health Organization (WHO), and the National Institutes of Health (NIH); in interviews members confirmed drawing on this material as well as additional documents, such as those published by Public Health England. Thus, while the Biosafety Advisory Board and guidelines were a significant achievement for open science, Board members made it clear that their work was indebted to previous efforts over the last decade, many of which had been externally funded. The accomplishments of JOGL in the realm of biosafety therefore built on prior relationships within the community and with outside experts, as well as both formal and informal bodies of knowledge related to biosafety. ## Theme 3. Role of the Biosafety Advisory Board: advisory vs. mandatory When Landrain, the CEO of JOGL, had initially proposed the Biosafety Advisory Board, he saw it as having three main goals: publishing ethics guidelines, reviewing projects, and advising JOGL members. While those involved in biosafety were largely in agreement with the goals of establishing guidelines and advising JOGL members, the extent to which the Board should provide oversight proved to be a thornier issue. Some felt that the Board should act like a review committee, providing a seal or stamp of approval for JOGL projects. In this view, the Board would have the power to both approve projects and exclude them. As one member of JOGL leadership noted on the biosafety Slack channel:Projects are free to join but if a project is following a dangerous road and its leaders ignore the warning of the board, I think it’s important to be able to say that they can’t be part of this initiative anymore. Others, however, maintained that the Board should act in a purely advisory role. As the Board was forming in early March, one member summed up the debate on the biosafety Slack channel:I think what needs to be decided on sooner rather than later is if this board is acting in a similar capacity to that of a review committee at an institution. Will projects need to come to us for approval before they proceed? Or are we simply providing guidelines, recommendations and access to experts that can answer questions and weigh in. While on its surface, the question of the role of the Board was an administrative one, it also cut to the heart of a deeper issue: what modes of gatekeeping, if any, should be acceptable on an open science platform? Banning projects from participating in the OpenCovid19 Initiative would go against the radical openness that is critical to the ideology of DIY biology, yet including a potentially dangerous project could harm the reputation of JOGL and DIY science as a whole. One Board member described it as a delicate balancing act:You’re trying to, balance that line, between, making sure that people are doing things safe, but also not preventing them from pursuing their ideas and interests (B1). Most Board members ultimately favored an advisory role: one where the Board would provide guidance but not mandatory oversight. They maintained different justifications for holding this position. Some felt that there would not be “buy-in” from the open science community for mandatory oversight, and it would therefore not only be impractical but would also drive people away from JOGL:From my standpoint, it's kind of, why alienate people when you... get the same sort of benefit from an advisory board that's giving that advice without saying, “You must do X, Y, and Z to do this” (B2). Along similar lines, other Board members expressed hesitancy about mandatory oversight, either because they felt the Board lacked a legitimate claim to authority, or because they felt such authority would not have been widely accepted by members of the open science community. Still others felt that there was no clear mechanism or process by which the Board could have exercised its authority, as establishing a review system would have required a system of rules: regarding how projects would obtain initial approval from the Board, processes for projects that lacked approval or those that subsequently violated biosafety standards, and means of appealing Board decisions. Without a comprehensive system that outlined consequences for actions—and an elucidation of how such consequences would be enforced—some Board members felt that a mandatory review process would be unsustainable. Others felt that the fundamental goal of biosafety guidance was not to be punitive, but rather to facilitate the conduct of research in as low-risk manner as possible:It has to be supportive. And I can honestly say, in my career, there have been very, very few times where I have looked across at a researcher and said, "No. You cannot do this." I'll probably say maybe four or five times in my career because there's always a way that you can do it. The whole point of biosafety isn't to stop the research, it's to ensure that it's done safely (B7). A final concern was related to accountability. If the Board approved a project, and that project ultimately caused some kind of physical harm, to what extent would Board members be liable? This was not only an issue for Board members from within the DIY biology community, but also one that could hamper the Board’s ability to recruit outside experts. As one Board member put it:I was keen that the board was advisory I think, partly because we actually wanted to bring in some biosafety professionals who have actually got some experience in the field. And I personally felt like that, having something more formalized would be challenging to persuade them to do because there's quite a bit of exposure given the particular scenario... I mean, as soon as you assume some kind of formal responsibility, there is a kind of liability that you're accepting as well, to some extent (B4). Ultimately, it was agreed that the Board would assume an advisory rather than an oversight role, due to the fact that most Board members held this position. As one member of JOGL leadership reflected:They [the Board] came on knowing that it was mostly for advising and creating biosafety and biosecurity guidelines. And so, they didn't accept that position knowing that they could actually be accountable. So, it was clear that then this would become more of an advice that we at JOGL would actually use and execute. So, we would be the ones accountable for making decisions, as we are accountable when we are kicking out members out of the platform because that person is not behaving well (J5). Thus, as will be discussed below, the responsibility for drawing lines with regard to biosafety—between who could participate and who could not—ultimately fell on JOGL, rather than on the Biosafety Board. ## Theme 4. Ethical gatekeeping In positioning itself as a platform for those interested in open science, JOGL drew its membership largely from those who had previously participated in the DIY biology community. Indeed, on Slack, weekly JOGL calls, and throughout our interviews, individuals continuously referred to themselves and their peers as a “community.” But there were no criteria that conferred membership in the community, nor were there grounds for exclusion. As one interviewee put it:I mean, the community is not a community. The community is a bunch of people who self-identify as being in the community. So, there's no way to keep people out of the community. But if people start to do things that are deceptive, dangerous, and under the name of being groundbreaking, but not really paying attention to science......it just becomes problematic for the entire community because the press loves a story about people acting dangerously (B5). Because JOGL provided a platform for the DIY biology community, it was faced with a dilemma: should the organization itself draw boundaries to exclude potentially dangerous projects, or should it leave such decisions up to the community writ large? Some in JOGL leadership felt that it was not JOGL’s place to police unsafe projects—and that the community would naturally exclude those who were not interested in upholding the ethical norms of DIY biology. As one member of JOGL leadership put it:The community that makes up JOGL is less likely to do or practice unethical experiments. And so even if there was one unethical person who had some idea, or maybe they didn't realize it was unethical… they may be able to connect with other people who do have expertise in what they're doing and realize that JOGL was not the platform for it because of the general culture there. And then they go off and do their own thing as a secret little group (B11). Others in JOGL leadership, however, felt that informal social controls would not be sufficient to maintain order, and that JOGL, as an organization, needed to take on the responsibility of creating its own boundaries:I think it was an argument between us [in JOGL leadership], like, "We need a code of conduct," and people are like, "Oh, no, don't worry, it will be okay." " Well, we need a code of conduct." I think, very practically, we needed one to be allowed as an organization to exclude people. The inclusion criteria of our community, there is none. But I think a code of conduct is a way of saying, "*This is* our principles, if you don't abide by it, then you leave." ( J1) Ultimately JOGL did create a code of conduct—outlining inappropriate behavior (harassment, violence, trolling) that were grounds for exclusion from the platform—and all new members of JOGL were required to check a box, agreeing to abide by it. However, the code did not encompass unsafe experiments, and since the biosafety guidelines were merely advisory, a violation of them was not grounds for exclusion. JOGL’s policies were tested several months into the pandemic, when an open-source vaccine collectively known as RaDVaC began to discuss the possibility of joining the JOGL platform. Because RaDVaC had published the instructions for developing and self-administering a DIY COVID-19 vaccine on their website (RaDVaC 2022), some in JOGL worried that they were not upholding the ethical and safety norms of the community. As one Board member recalled:To me, the most worrisome thing was when the RaDVaC people dunked in, because I felt that they aren’t actually following any of those guidelines that the DIY community follows, and that they were more interested in just getting more people to test their vaccine out on (B2). By the time RaDVaC expressed interest in joining JOGL in late 2020, there had been no activity on the public Slack biosafety channel for months, and the Board was no longer in active communication about JOGL-related biosafety issues. There was therefore no formal assessment of whether RaDVaC’s approach to ethics and safety was in keeping with JOGL’s standards. However, a Slack channel was created dedicated to open-source vaccines, where one member of JOGL leadership expressed optimism about an open-source vaccine, but suggested a further conversation about what JOGL “could do to support your initiative, [and] how we could improve safety measures.” Soon after, an informal Zoom meeting was held among RaDVaC, JOGL leadership, and two members of the Biosafety Board who had become aware of RaDVaC’s desire to join JOGL. Regarding what transpired on the call, one interviewee recalled that JOGL “did actually not go with them except we didn't kick them out” (J5); another, however, recalled that JOGL “shut down” RaDVaC (B13). Following the call, members of RaDVaC were no longer active on JOGL’s Slack, and it was clear that JOGL and RaDVaC had not embraced one another. Less clear, however, is what role the biosafety guidelines themselves played in denying RaDVaC a platform on JOGL. While the guidelines stated that no animal or human testing should be performed unless “you have gone through relevant approval processes,” there had been no official biosafety review. One interviewee who was on the call made no mention of the guidelines themselves as having been a reason for RaDVaC’s exclusion. But two others felt that having set out ethical norms—via the Biosafety Board having developed the biosafety guidelines—was key to being able to deny RaDVaC a prominent platform at JOGL. As they put it:The RaDVaC projects which was, you know, very cool, you know, on paper because, you know, having...developing open-source vaccines is...would be really, really important...but the way they approached the problem was not compatible with the ethical limits that we had defined where there was no self-testing (J5).JOGL kind of said no to them [RaDVaC] when they came up with that. And so, to me, that sort of shows the success of these programs [the Biosafety Board and guidelines] because when they're presented with sort of a project or an opportunity that clearly would go against those norms, they stuck to those norms and said, "No, thanks" (B2). Thus, even after the height of the pandemic, JOGL practiced a kind of ethical gatekeeping. While there was no formal biosafety review, the process of defining ethical guidelines resulted in a decision to exclude a project that was not adhering to defined ethical boundaries. ## Theme 5. Utility of the biosafety guidelines While the guidelines appear to have played a role in allowing JOGL to exclude projects, the extent to which they provided positive value to individual projects is less clear. Some Biosafety Board members and JOGL leadership doubted whether the community even was aware of them, noting they themselves couldn’t “even really find the guidelines” (B13) and wondered how anyone else would find them. In addition, both Board members and JOGL leadership repeatedly highlighted the lack of attention to dissemination of and follow-through regarding the guidelines. Some felt that there should have been additional efforts to increase awareness of them. As one Board member stated:I don't think there's been any sort of accountability if that makes sense…. Or any sort of, like, follow up to make sure that people are following them. I think it's important that they [biosafety guidelines] are there and that people have access to them but I don't know if... Maybe there could be a little bit more, like, push to let people know that they're there (B1). In a similar vein, some in JOGL leadership felt that the Board had ended its efforts too early, and that it was the Board’s responsibility to have been more engaged throughout the lifetime of JOGL:The fact that I know they [the Board] aren't active anymore, haven't been for a long time, to me says that maybe they felt once the guidelines were produced, that their job was done. But the job wasn't done because then they had to get the information out and visible. ( J6). One member of the Board acknowledged the validity of these concerns, but felt that the underlying problem was related to the lack of time and funding, not necessarily the personal dedication of Board members:Just having a link to a manual is not enough. Having an advisory board there helps, but is that enough? Is it really [enough] knowing if people are trained? No. But do we have the resources to do all that? No. ( B2). Although most Board members and JOGL leadership believed that the JOGL community was likely unaware of the biosafety guidelines, this view was not supported by our interviews with project participants: of the 12 individuals interviewed, nine reported being aware of the guidelines. This awareness may have been due to one significant step JOGL had taken to make individuals aware of the biosafety guidelines. The organization had raised money from the AXA Research Fund (AXA 2020), and in April 2020, it began to offer “micro-grants” of several thousand dollars for individual projects. As part of the grant application process, JOGL required applicants to check a box, acknowledging that they had read the biosafety guidelines and agreed to adhere to them (JOGL 2021). This simple endeavor—requiring any individual applying for a JOGL microgrant to pledge to abide by the biosafety guidelines—appears to have had a significant effect on awareness. Eight of the nine project participants who reported being aware of the guidelines had also been funded by a JOGL microgrant, whereas all three who were not aware of the biosafety guidelines had not received any JOGL grant. It is worth noting, however, that awareness itself is not a measure of utility or value. None of the project participants mentioned using the guidelines. For some of the projects—such as those that used machine learning to develop software applications, or those that focused on educational projects—the biosafety guidelines themselves were irrelevant to their project, as their projects did not involve the use of biological samples. The more significant explanation for the questionable utility of the guidelines, however, may have rested on the fact that most projects—running on the good will of volunteers and lacking any resources—never progressed to any kind of advanced stage where biosafety concerns would have been relevant. And of the few projects that had advanced, several interviewees noted that such projects were led by individuals who had significant prior biology experience or an “institutional foot” in the door (J2). Thus, they felt that those whose projects had reached an advanced stage were already practicing good biosafety practices—that “they're doing what it says in those guidelines, whether they know it or not” (B2). As one Board Member put it:Most of the people who would need them [the guidelines] sort of knew it already… I think probably everybody who needed them, knew (B3). Thus, while the project leaders’ awareness of the biosafety guidelines was likely influenced by the requirement to acknowledge them in JOGL’s microgrant application, there was agreement among Board members and JOGL leadership that more could have been done to disseminate them. The guidelines did not appear to play a role in educating newcomers about biosafety practices, but rather, as discussed in the previous section, played a more symbolic role in upholding the ethical norms of JOGL. ## Discussion While prior scholarship has highlighted ethical issues in DIY science and assessed participants’ attitudes toward ethics (Guerrini, Trejo, et al. 2020; Trejo, et al. 2021; Guerrini et al. 2021; Rasmussen 2021, 2017), the present study is, to our knowledge, the first to assess how DIY biologists navigated one ethical issue, biosafety, in real-world practice. Our results align with previous findings showing that rather than proceeding recklessly and with disregard for risks, most in the DIY biology community value safety and endeavor to prioritize it in their practices (Grushkin et al. 2013; Kuiken et al. 2018; Meyer and Vergnaud 2020; Lim 2021; Trejo et al. 2021; Seyfried et al. 2014). Specifically, in this study, we demonstrate how leaders of the JOGL platform, the main online hub for DIY science during the COVID-19 pandemic, quickly recognized the potential biosafety risks and actively tried to build the infrastructure to help facilitate the safe conduct of research. The fact that an international organization like JOGL succeeded in forming a Biosafety Advisory Board and publishing a set of biosafety guidelines—ones that had a measure of external credibility due to the participation of outside experts—marks a significant achievement in the realm of DIY science. These accomplishments, however, did not come in a void; they built upon a decade of prior efforts related to open science and biosafety that were largely supported by external funding. These past endeavors had resulted in both outside partnerships and resources that facilitated the rapid development of JOGL’s biosafety guidelines. That the DIY Biology community not only welcomed relationships with experts, but also actively sought them out, is in line with the findings from Trejo et al. [ 2021] that the community values such partnerships, as long as outsiders respect the independent nature and culture of DIYbio. In that sense, our findings lend empirical support to policy proposals that have called for transparent and collaborative engagement on the part of experts and government agencies to enhance the ethical conduct and safety of DIY science (Guerrini, Sherkow, et al. 2020). Still, compared to the manner in which biosafety is regulated in institutional settings, JOGL’s Biosafety Board lacked teeth: by maintaining a purely advisory role, it lacked the authority to provide any kind of significant oversight. The Board was also not active for long, likely due to both the waning of interest of JOGL members in COVID-related projects by late 2020 (as mainstream science was successfully developing diagnostic tests and vaccines) as well as the lack of progression of projects to stages where biosafety concerns would have become relevant. The reluctance of the Board to adopt a mandatory approach to ethical oversight is not surprising, as a prior study assessing DIY biologists’ preference for ethical oversight found that participants were “especially opposed to mandatory, rule-based oversight mechanisms” (Trejo et al. 2021). Many Biosafety Board members in our interviews appeared in-tune with this sentiment, as their reluctance to adopt an authoritative approach was driven by their perception that it would not have been accepted by the wider DIY biology community. Even though the power of JOGL Biosafety Board’s was limited, its output—the biosafety guidelines—appears to have provided a formal foundation for JOGL leadership to exclude an open-source vaccine project. While there was no assessment by the Biosafety Board of the open-source vaccine project—and it is possible that even without the biosafety guidelines, JOGL would have excluded the project—the guidelines offered JOGL leadership a concrete expression of ethical norms and standards, ones that were developed by representative members of the community and outside experts. The guidelines themselves therefore had a measure of authority, and could be referenced in any exclusion decisions. That JOGL was able to exercise ethical gatekeeping was likely due to two structural features. First, it provided access to an international network of DIY science projects and peers through an online platform that it controlled. Second, it also acted as a funder, providing modest microgrants for COVID-19 projects. At both of these points—access to the platform and the provision of funding—JOGL required individuals to certify that their behavior and projects adhered to certain ethical norms and standards. In this way, JOGL could define borders between who could participate and who could not. In at least one respect, JOGL attempted to have those lines drawn by representative members of the community (i.e., through the development of the biosafety guidelines). In the future, recognition and identification of sites of ethical gatekeeping may be helpful in terms of enhancing the ethical conduct of DIY science. The fact that there was ethical gatekeeping on JOGL does not mean that actors not allowed to be part of the “community” will cease their efforts, but rather such efforts may occur in more isolated settings, without the full support of (or enculturation from) those involved DIY biology. Thus, efforts to support biosafety within DIY biology should focus not only on improving practices within the community, but also identifying and addressing risks from splinter groups or individual actors. This project has a number of limitations. First, because we had access only to public Slack channels, it is possible that additional conversations regarding ethics and biosafety occurred privately (i.e., via direct message or email) and were therefore not captured by our methods. Second, since most projects did not reach the stage of encountering significant biosafety concerns, our data on how individual project members may have resolved biosafety issues are limited. Third, as our interviews were conducted nearly a year after the height of activity on JOGL, participants’ recollections may have been incomplete; however, whenever possible, we confirmed participants’ recollections with data from our digital ethnography. Fourth, our study focuses on a subset of individuals involved with DIY biology and may therefore not be generalizable to the wider DIY biology community. ## Conclusion The present study reinforces the findings of previous scholarship that has pointed to a strong interest among DIY biologists in incorporating safe and ethical practices—albeit in a way that is respectful of their core values of openness and flexibility. For the DIY biology community, our findings point to the importance of community-developed standards and attention to sites of ethical gatekeeping; for those outside it, such as policymakers, our study points to the importance of fostering partnerships between outside experts and DIY biologists. Our results also point to the critical role that external funding has played in providing the community with resources to develop approaches and training for addressing biosafety issues. Finally, given the challenges of regulating non-establishment research, and the public health interest in addressing any concomitant potential biosafety risks, it is imperative that future research continues to examine the practices of DIY science. ## Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 109 kb) ## References 1. ABSA International: The Association of Biosafety and Biosecurity & Baltimore Under Ground Science Space. 2019. Principles and Practices of Biosafety, August 25, https://www.eventbrite.com/e/principles-and-practices-of-biosafety-tickets-64426483346. Accessed 21 Feb 2022. 2. Armendariz, A. et al. 2020. Community Biology Biosafety Handbook, Just One Giant Lab. https://docs.google.com/document/d/1Qkc2uCAcLX45b0GjSGZohweelJ-vDOhX5MDSf6F4MEI/edit#. Accessed 21 Feb 2022. 3. AXA Research Fund. 2020. The AXA Research Fund commits to the COVID-19 effort. April. https://www.axa-research.org/en/news/the-axa-research-fund-commits-to-the-COVID19-effort. Accessed 21 Feb 2022. 4. Baltimore Under Ground Science Space. 2019. Checking Ourselves Before Wrecking Ourselves: Co-Evolving Innovation and Safety in the DIYBio Community. https://bugssonline.org/community/diybio-biosafety/. Accessed 21 Feb 2022. 5. Barba, G.A.S. 2014. We are biohackers: Exploring the collective identity of the DIYbio movement. Delft: Delft University of Technology. Master of Science Thesis. 6. Bektas, A. 2020. ALERT (Accessible LAMP-Enabled Rapid Test). Just One Giant Lab, 5 April, https://app.jogl.io/project/187/ALERT. Accessed 21 Feb 2022. 7. Belgrave LL, Seide K. *Handbook of Research Methods in Health Social Sciences* (2019.0) 299-316 8. Caplan AL, Bateman-House A. **The Danger of DIY Vaccines**. *Science* (2020.0) **369** 1035-1035. DOI: 10.1126/science.abe4440 9. Delfanti A. *Biohackers: The Politics of Open Science* (2013.0) 10. Delgado A, Callén B. **Do-It-Yourself Biology and Electronic Waste Hacking: A Politics of Demonstration in Precarious Times**. *Public Understanding of Science* (2017.0) **26** 179-194. DOI: 10.1177/0963662516647348 11. DIYBio.org. 2011. DIY Bio Codes. https://diybio.org/codes/. Accessed 21 Feb 2022. 12. Edwards, K. 2017. Ethics in citizen science. Presented at PEPH Webinar: Ethics & Citizen Science hosted by the National Institute of Environmental and Health Sciences; 5 October, USA. 13. Eggleson K. **Transatlantic Divergences in Citizen Science Ethics—Comparative Analysis of the DIYbio Code of Ethics Drafts of 2011**. *NanoEthics* (2014.0) **8** 187-192. DOI: 10.1007/s11569-014-0197-7 14. Ekekezie C. **Understanding the Scope of Do-It-Yourself Fecal Microbiota Transplant**. *The American Journal of Gastroenterology* (2020.0) **115** 1. DOI: 10.14309/ajg.0000000000000499 15. Erikainen S. **The Promissory Visions of DIYbio: Reimaging Science from the Fringe**. *Science as Culture* (2022.0) **31** 287-310. DOI: 10.1080/09505431.2022.2028135 16. Ferretti F. **Mapping Do-It-Yourself Science**. *Life Sciences, Society and Policy* (2019.0) **15** 1. DOI: 10.1186/s40504-018-0090-1 17. Fiske A. **Conceptual and Ethical Considerations for Citizen Science in Biomedicine**. *Personal Health Science* (2018.0). DOI: 10.1007/978-3-658-16428-7_10 18. Grant AD, Wolf GI, Nebeker C. **Approaches to Governance of Participant-Led Research: A Qualitative Case Study**. *British Medical Journal Open* (2019.0) **9** e025633 19. Grushkin, D. 2018. Biohackers are about open-access to science, not DIY pandemics. Stop misrepresenting us. Stat News, 4 June: online. 20. Grushkin D, Kuiken T, Millet P. *Seven Myths and Realities of Do-It-Yourself Biology* (2013.0) 21. Guerrini CJ. **A best-worst scaling experiment to prioritize concern about ethical issues in citizen science reveals heterogeneity on people-level v. Data-level issues**. *Scientific Reports* (2021.0) **11** 19119. DOI: 10.1038/s41598-021-96743-4 22. Guerrini CJ, Wexler A. **Biomedical Citizen Science or Something Else? Reflections on Terms and Definitions**. *The American Journal of Bioethics* (2019.0) **19** 17-19. DOI: 10.1080/15265161.2019.1619880 23. Guerrini CJ, Spencer GE, Zettler PJ. **DIY CRISPR**. *North Carolina Law Review* (2019.0) **97** 1399-1460 24. Guerrini CJ. **Citizen Science, Public Policy**. *Science* (2018.0) **361** 134-136. DOI: 10.1126/science.aar8379 25. Guerrini CJ, Trejo M. **Core Values of Genomic Citizen Science: Results from a Qualitative Interview Study**. *BioSocieties* (2020.0) **17** 203-228. DOI: 10.1057/s41292-020-00208-2 26. Guerrini CJ, Sherkow JS. **Self-experimentation, Ethics, and Regulation of Vaccines**. *Science* (2020.0) **369** 1570-1572. DOI: 10.1126/science.abe1963 27. Jorgenson, E. 2020. One Hour Covid Test Using LAMP. Just One Giant Lab, 26 March, https://app.jogl.io/project/163/1HourCovidTest. Accessed 21 Feb 2022. 28. Just One Giant Lab. 2022. Meet Our Team. https://jogl.io/team. Accessed 21 Feb 2022. 29. Just One Giant Lab. 2021. OpenCovid19—Grant Review Round 5. 23 February, https://app.jogl.io/challenge/oc19grants. Accessed 21 Feb 2022. 30. Just One Giant Lab. 2020a. OpenCovid19 Initiative. 29 February, https://app.jogl.io/program/opencovid19. Accessed 21 Feb 2022. 31. Just One Giant Lab. 2020b. OpenCovid19 Initiative Biosafety and Biosecurity Guidelines v1.0, Just One Giant Lab. https://docs.google.com/document/d/1ldQ1aRxHjuj-U7kNqwWlkwwwW54U6ls8xMI9l0LBFrQ/edit. Accessed 21 Feb 2022. 32. Just One Giant Lab. 2020c. OpenCovid19 Initiative [Community Call #2], Just One Giant Lab, March 11, https://www.youtube.com/watch?v=QOtYq9Jn7sU. Accessed 21 Feb 2022. 33. Kempner J, Bailey J. **Collective Self-experimentation in Patient-Led Research: How Online Health Communities Foster Innovation**. *Social Science & Medicine* (2019.0) **238** 112366. DOI: 10.1016/j.socscimed.2019.112366 34. Kolodziejczyk B. *Do-It-Yourself Biology Shows Safety Risks of an Open Innovation Movement* (2017.0) 35. Kuiken T, Trump BD. **Biology Without Borders: Need for Collective Governance?**. *Synthetic Biology 2020: Frontiers in Risk Analysis and Governance. Risk, Systems and Decisions* (2020.0) 269-295 36. Kuiken T. **Learn from DIY Biologists: The Citizen-Science Community has a Responsible, Proactive Attitude that is Well Suited to Gene-Editing**. *Nature* (2016.0) **531** 167-168. DOI: 10.1038/531167a 37. Kuiken, T., E. Pauwels, and S.W. Denton. 2018. The Rise of the New Bio-Citizen: Ethics, Legitimacy, and Responsible Governance in Citizen-Driven Biomedical Research and Innovation, https://www.wilsoncenter.org/article/the-rise-the-new-bio-citizen-workshop. Accessed 21 Feb 2022. 38. Landrain T. **Do-It-Yourself Biology: Challenges and Promises for an Open Science and Technology Movement**. *Systems and Synthetic Biology* (2013.0) **7** 115-126. DOI: 10.1007/s11693-013-9116-4 39. Lee JM, Hirschfeld E, Wedding J. **A Patient-Designed Do-It-Yourself Mobile Technology System for Diabetes: Promise and Challenges for a New Era in Medicine**. *JAMA* (2016.0) **315** 1447-1448. DOI: 10.1001/jama.2016.1903 40. Lempinen, E.W. 2011. FBI, AAAS Collaborate on Ambitious Outreach to Biotech Researchers and DIY Biologists, American Association for the Advancement of Science, 1 April, https://www.aaas.org/news/fbi-aaas-collaborate-ambitious-outreach-biotech-researchers-and-diy-biologists. 41. Lim, Y.-B. 2021. For Liberty and Accessible Science for All: Building a Better Understanding of U.S. Community Laboratories and the Do-It-Yourself (DIY) Biology Movement. 42. Lowe, N. 2021. Face Shields From Recycled Plastic. Just One Giant Lab, 17 March, https://app.jogl.io/project/573/WasteShields. Accessed 21 Feb 2022. 43. McGowan ML. **“Let’s Pull These Technologies Out of the Ivory Tower”: The Politics, Ethos, and Ironies of Participant-Driven Genomic Research**. *BioSocieties* (2017.0) **12** 494-519. DOI: 10.1057/s41292-017-0043-6 44. Meyer M. **Domesticating and Democratizing Science: A Geography of Do-It-Yourself Biology**. *Journal of Material Culture* (2013.0) **18** 117-134. DOI: 10.1177/1359183513483912 45. Meyer M, Vergnaud F. **The Rise of Biohacking: Tracing the Emergence and Evolution of DIY Biology Through Online Discussions**. *Technological Forecasting and Social Change* (2020.0) **160** 120206. DOI: 10.1016/j.techfore.2020.120206 46. Morales, H. 2020. Coughcheck App. Just One Giant Lab, 18 March, https://app.jogl.io/project/132/coughcheck. Accessed 21 Feb 2022. 47. Murphy, H. 2020. These Scientists Are Giving Themselves D.I.Y. Coronavirus Vaccines. The New York Times, 1 September: online. 48. Omer T. **Empowered Citizen ‘Health Hackers’ who are not Waiting**. *BMC Medicine* (2016.0) **14** 118. DOI: 10.1186/s12916-016-0670-y 49. Pauwels, E., and S.W. Denton. 2018. The Rise of the Bio-Citizen, Wilson Center. 50. RaDVaC. 2022. RaDVaC: Rapid Deployment Vaccine Collaborative. 51. Rasmussen, L. 2017. Filling the ‘Ethics Gap’ in Citizen Science Research. Presented at PEPH Webinar: Ethics & Citizen Science hosted by the National Institute of Environmental and Health Sciences; 5 October, USA. 52. Rasmussen LM. **Realizing Present and Future Promise of DIY Biology and Medicine through a Trust Architecture**. *Hastings Center Report* (2020.0) **50** 10-14. DOI: 10.1002/hast.1194 53. Rasmussen LM, Iltis AS, MacKay D. **Research Ethics in Citizen Science**. *The Oxford Handbook of Research Ethics* (2021.0) 54. Resnik DB. **Citizen Scientists as Human Subjects: Ethical Issues**. *Citizen Science: Theory and Practice* (2019.0) **4** 11 55. Roosth S. *Synthetic: How Life Got Made* (2017.0) 56. Rothstein MA, Wilbanks JT, Brothers KB. **Citizen Science on Your Smartphone: An ELSI Research Agenda**. *The Journal of Law, Medicine & Ethics* (2015.0) **43** 897-903. DOI: 10.1111/jlme.12327 57. Scroggins, M. 2013. DIYbio and the “New FBI,” BioCoder. 58. Seyfried G, Pei L, Schmidt M. **European Do-It-Yourself (DIY) Biology: Beyond the Hope, Hype and Horror**. *BioEssays* (2014.0) **36** 548-551. DOI: 10.1002/bies.201300149 59. Shah, S., and E. Jamrozik. 2020. Do-It-Yourself Vaccines for COVID-19. Scientific American, August 30, 2020, https://www.scientificamerican.com/article/do-it-yourself-vaccines-for-covid-19/. Accessed 21 Feb 2022. 60. Strauss AL, Glaser B. *The Discovery of Grounded Theory* (1967.0) 61. Sundaram LS. **Biosafety in DIY-bio Laboratories: From Hype to Policy**. *EMBO Reports* (2021.0) **22** e52506. PMID: 33761157 62. SynenergeneWhat can “do-it-yourself biology” do? 13 December 20142014ParisFrance. *What can “do-it-yourself biology” do? 13 December 2014* (2014.0) 63. Trejo M. **“A Cohort of Pirate Ships”: Biomedical Citizen Scientists’ Attitudes Toward Ethical Oversight**. *Citizen Science: Theory and Practice* (2021.0) **6** 15. PMID: 35474712 64. Trejo M. **How Biomedical Citizen Scientists Define What They Do: It’s All in the Name**. *AJOB Empirical Bioethics* (2020.0) **12** 1-8. PMID: 32981477 65. Vayena E. **Research Led by Participants: A New Social Contract for a New Kind of Research**. *Journal of Medical Ethics* (2016.0) **42** 216-219. DOI: 10.1136/medethics-2015-102663 66. Vayena E, Tasioulas J. **Adapting Standards: Ethical Oversight of Participant-Led Health Research**. *PLoS Medicine* (2013.0) **10** e1001402. DOI: 10.1371/journal.pmed.1001402 67. Vayena E, Tasioulas J. **The Ethics of Participant-Led Biomedical Research**. *Nature Biotechnology* (2013.0) **31** 786-787. DOI: 10.1038/nbt.2692 68. Walker, J.T. et al. 2020. Proceedings of the Global Community Bio Summit 4.0. https://www.biosummit.org/proceedings. Accessed 21 Feb 2022. 69. Wexler A. **The Practices of Do-It-Yourself Brain Stimulation: Implications for Ethical Considerations and Regulatory Proposals**. *Journal of Medical Ethics* (2016.0) **42** 211. DOI: 10.1136/medethics-2015-102704 70. Wexler A. **The Social Context of “Do-It-Yourself” Brain Stimulation: Neurohackers, Biohackers, and Lifehackers**. *Frontiers in Human Neuroscience* (2017.0) **11** 224. DOI: 10.3389/fnhum.2017.00224 71. Wiggins A, Wilbanks J. **The Rise of Citizen Science in Health and Biomedical Research**. *The American Journal of Bioethics* (2019.0) **19** 3-14. DOI: 10.1080/15265161.2019.1619859
--- title: High expression of TARS is associated with poor prognosis of endometrial cancer authors: - Lihui Si - Lianchang Liu - Ruiqi Yang - Wenxin Li - Xiaohong Xu journal: Aging (Albany NY) year: 2023 pmcid: PMC10042687 doi: 10.18632/aging.204558 license: CC BY 3.0 --- # High expression of TARS is associated with poor prognosis of endometrial cancer ## Abstract Introduction: Endometrial cancer is the second largest and most common cancer in the world. It is urgent to explore novel biomarkers. Methods: Data were obtained from The Cancer Genome Atlas (TCGA) database. The receiver operating characteristic (ROC) curves, Kaplan-Meier curves and Cox analysis, nomograms, gene set enrichment analysis (GSEA) were conducted. Cell proliferation experiments were performed in Ishikawa cell. Results: TARS was significantly highly expressed in serous type, G3 grade, and deceased status. Significant association was between high TARS expression with poor overall survival ($$P \leq 0.0012$$) and poor disease specific survival ($$P \leq 0.0034$$). Significant differences were observed in advanced stage, G3 and G4, and old. The stage, diabetes, histologic grade, and TARS expression showed independent prognostic value for overall survival of endometrial cancer. The stage, histologic grade, and TARS expression showed independent prognostic value for disease specific survival of endometrial cancer. Activated CD4+ T cell, effector memory CD4+ T cell, memory B cell and type 2 T helper cell may participate in the high TARS expression related immune response in endometrial cancer. The CCK-8 results showed significantly inhibited cell proliferation in si-TARS ($P \leq 0.05$) and promoted cell proliferation in O-TARS ($P \leq 0.05$), confirmed by the colony formation and live/dead staining. Conclusion: High TARS expression was found in endometrial cancer with prognostic and predictive value. This study will provide new biomarker TARS for diagnosis and prognosis of endometrial cancer. ## INTRODUCTION Endometrial cancer is a common cancer in female reproductive systems, and is the second largest and most common cancer in the world second only to cervical cancer [1]. There can be 319,500 new cases in the world each year, and the mortality rate is higher than $23\%$ [2]. The rate of incidence is rising among younger females [3]. Endometrial cancer begins in the endometrium, located at the innermost layer of the uterus and is the result of abnormal growth of cells with invasion or spread to other parts of the body [4]. Endometrial cancer is divided into type I hormone dependencies and type II non-hormone dependencies [5]. Type I endometrium cancer is an endometrium-like adenocarcinoma in tissue classification, which is the most common subtype with good prognosis [6]. Meanwhile, type II endometrial cancer carries mutant genes such as P53, P16, etc., and has a high risk of metastasis and poor prognosis [7]. The most common clinical manifestation of endometrial cancer is abnormal uterine bleeding, but this symptom can also be caused by many other diseases [8]. In some cases, endometrial cancer may have already developed into advanced stage when signs and symptoms can be noticed. At present, the treatment of endometrial cancer is mainly surgery, followed by comprehensive treatment with auxiliary methods such as radiotherapy, chemotherapy, and hormone therapy [9]. However, the local high recurrence rate, high metastasis rate, and hormone therapy resistance are still the predicament of clinical treatment [10]. Therefore, it is urgent to explore the potential molecular mechanism of endometrial cancer progression, and find novel biomarkers and effective treatment targets. Amino acids are attached to their corresponding tRNAs by enzymes called aminoacyl-tRNA synthetases (ARSs), which play important roles in protein synthesis [11]. The threonyl-tRNA synthetase (TARS) is one of the ARSs and serves as an important therapeutic target [12]. TARS was discovered as an active enzyme in the mid-1950s, which can produce a carboxyl activated complex that combined with enzymes [13]. However, the role of TRAS in endometrial cancer has not been illuminated yet. In this study, we examined the association between clinicopathologic characteristics and TARS expression in endometrial cancer using data from The Cancer Genome Atlas (TCGA) database. The receiver operating characteristic (ROC) curves were plotted to study the diagnostic value of TARS expression. The Kaplan-Meier curves and *Cox analysis* were used to study the overall survival and disease specific survival. The nomograms were used to study the predictive value of TARS expression. *The* gene set enrichment analysis (GSEA) was conducted, and high TARS expression-enriched pathways indicated their influence on the immune response and cell proliferation of endometrial cancer. ## Data mining The TCGA database (https://www.cancergenome.nih.gov) was used to get the whole RNA-Seq expression files as well as any related clinical features [14]. The TARS mRNA expression data was converted into RSEM-normalized values using log2 (x + 1). The non-parametric rank sum test was employed to evaluate the levels of TARS mRNA expression. The Wilcoxon rank sum test was used to compare two groups, and the Kruskal-Wallis test was used to compare multiple groups. The Fisher’s exact test and chi-square test were employed to evaluate the association between clinical traits and TARS expression. ## Diagnostic value of TARS expression Using the pROC application to show ROC curves, we estimated the area under the ROC curves (AUC) values and established the proper cutoff threshold for the evaluation of TARS diagnostic capabilities [15]. The patients were divided into high or low TARS expression according to the cutoff threshold. The AUC of ROC curve between normal and tumor was 0.901 (Supplementary Figure 1A). Besides, the AUC was 0.890 for stage I (Supplementary Figure 1B), 0.864 for stage II (Supplementary Figure 1C), 0.936 for stage III (Supplementary Figure 1D), and 0.957 for stage IV (Supplementary Figure 1E). The results indicated promising diagnostic value of TARS expression. ## Predictive value of TARS expression The patients with endometrial cancer were grouped, followed by comparison of histological type, stage, histologic grade, menopause status, and residual tumor for overall survival and disease specific survival. ## GSEA analysis First, a search of the TCGA database was conducted, and then an online GSEA analysis was performed to look into the relationship between TARS expression and enriched pathways [16]. ## Cell culture and plasmid transfection The human endometrial carcinoma cell line Ishikawa was obtained from Shanghai Biochemical Cell Institute (Shanghai, China). Ishikawa cells were cultured in RPMI 1640 medium, which contains $10\%$ fetal bovine serum and $1\%$ penicillin-streptomycin solution, at 37°C in the $5\%$ CO2 humidified atmosphere [17]. The si-TARS and si-control plasmids were purchased from Miaoling Bio (Wuhan, China) for transfection into the cells. ## Real-time quantitative PCR The total RNA extraction was performed using the Invitrogen kit (Thermo Fisher Scientific, MA, USA), followed by the reverse transcription. The real-time quantitative PCR (qRT-PCR) was conducted for detecting TARS expression. The 2−ΔΔCt approach was used for quantification. The primers were as follows: TARS forward primer, 5′-TGTGTGCCATTGAATAAGGA-3′; TARS reverse primer, 5′-CACCTTCATTATCAAGATAC-3′; β-actin forward primer, ACCCCAAAGCCAACAGA; β-actin reverse primer, CCAGAGTCCATCACAATACC [18]. ## Cell proliferation Plasmids were introduced to the Ishikawa cells and cultured for 24 hours. 10 μL of CCK-8 reagent was added and reacted for 0.5 h. Cell viability was determined using the 490 nm absorbance measurement. The Calcein AM and PI co-staining, as well as colony formation assay, was carried out as previously reported [19]. ## Statistical analysis The analysis was performed using R3.5.1 [20]. The survival rate was examined using the Kaplan-Meier curve [21]. The independent prognostic potential of TARS was examined using univariate and multivariate Cox models. The correlation between TARS expression and immune cells were analyzed. $P \leq 0.05$ was statistically significant. ## High TARS expression in endometrial cancer The mRNA expression of TARS was first examined in endometrial cancer. Significant high TARS expression was found in tumor compared with paired normal tissue ($$P \leq 0.016$$, Figure 1A), and normal endometrial tissue ($P \leq 0.001$, Figure 1B). **Figure 1:** *High TARS expression in endometrial cancer. (A) TARS expression in tumor vs. paired normal tissue. (B) TARS expression in tumor vs. normal tissue.* ## Characteristics of patients with endometrial cancer The characteristics of patients with endometrial cancer were studied using TCGA data. Totally, 370 patients with endometrial cancer were analyzed (Supplementary Table 1). There were 72 patients ($19.46\%$) less than 55 years old. Endometrioid type with 303 patients ($81.89\%$) was most in patients with endometrial cancer. Notably, histologic grade ($$P \leq 0.0015$$) and vital status ($$P \leq 0.0017$$) were significantly different in high TARS group and low TARS group. Meanwhile, age ($$P \leq 0.9828$$), histological type ($$P \leq 0.5773$$), stage ($$P \leq 0.0662$$), diabetes ($$P \leq 0.0916$$), hypertension ($$P \leq 0.6759$$), menopause status ($$P \leq 0.5987$$), and residual tumor ($$P \leq 0.5075$$) showed no statistical differences. ## TARS expression in subgroups The TARS expression grouped by age (Figure 2A), diabetes (Figure 2B), hypertension (Figure 2C), histological type (Figure 2D), histologic grade (Figure 2E), stage (Figure 2F), menopause status (Figure 2G), residual tumor (Figure 2H), and vital status (Figure 2I) were exhibited. TARS was significantly highly expressed in serous type ($$P \leq 0.0011$$), G3 grade ($P \leq 0.001$), and deceased status ($$P \leq 0.0012$$). However, the other subgroups showed no statistical differences. **Figure 2:** *TARS expression in subgroups. TARS expression grouped by (A) age, (B) diabetes, (C) hypertension, (D) histological type, (E) histologic grade, (F) stages, (G) menopause status, (H) residual tumor, and (I) vital status.* ## High TARS expression is associated with poor survival Kaplan–Meier curves were plotted to evaluate the overall survival (Figure 3A) and disease specific survival (Figure 3B). The results showed significant association between high TARS expression with poor overall survival ($$P \leq 0.0012$$) and poor disease specific survival ($$P \leq 0.0034$$). **Figure 3:** *High TARS expression is associated with poor survival. (A) Overall survival group by TARS in all tumors. (B) Disease specific survival group by TARS in all tumors.* ## Overall survival grouped TARS expression The subgroup analysis of overall survival was performed (Figure 4A–4F). Significant differences were observed in advanced stage ($$P \leq 0.0053$$), G3 and G4 ($$P \leq 0.015$$), and old ($$P \leq 0.0032$$). Nevertheless, early stage, G1 and G2, and young subgroups showed no statistical differences. **Figure 4:** *Overall survival grouped TARS expression. Overall survival group by GJB3 in (A) early stage, (B) advanced stage, (C) G1 and G2, (D) G3 and G4, (E) young, and (F) old.* The variables identified by univariate analysis (Figure 5A) were confirmed by multivariate analysis (Figure 5B). The stage [hazard ratio (HR): 1.589, $95\%$ confidence interval (CI): 1.246–2.027, $P \leq 0.001$], diabetes (HR: 1.556, $95\%$ CI: 1.068–2.267, $$P \leq 0.021$$), histologic grade (HR: 2.078, $95\%$ CI: 1.289–3.352, $$P \leq 0.003$$), and TARS expression (HR: 4.912, $95\%$ CI: 1.765–13.674, $$P \leq 0.002$$) showed independent prognostic value for overall survival of endometrial cancer. **Figure 5:** *Cox analysis of overall survival. (A) Univariate analysis of overall survival. (B) Multivariate analysis of overall survival.* ## Disease specific survival grouped TARS expression The subgroup analysis of disease specific survival was performed (Figure 6A–6F). Significant differences were observed in advanced stage ($$P \leq 0.0076$$), G3 and G4 ($$P \leq 0.026$$), and old ($$P \leq 0.0092$$). Nevertheless, early stage, G1 and G2, and young subgroups showed no statistical differences. **Figure 6:** *Disease specific survival grouped TARS expression. Disease specific survival group by GJB3 in (A) early stage, (B) advanced stage, (C) G1 and G2, (D) G3 and G4, (E) young, and (F) old.* The variables identified by univariate analysis (Figure 7A) were confirmed by multivariate analysis (Figure 7B). The stage (HR: 2.109, $95\%$ CI: 1.536–2.897, $P \leq 0.001$), histologic grade (HR: 2.646, $95\%$ CI: 1.335–5.244, $$P \leq 0.005$$), and TARS expression (HR: 6.723, $95\%$ CI: 1.606–28.152, $$P \leq 0.009$$) showed independent prognostic value for disease specific survival of endometrial cancer. **Figure 7:** *Cox analysis of disease specific survival. (A) Univariate analysis of disease specific survival. (B) Multivariate analysis of disease specific survival.* ## Predictive value of TARS expression in overall survival The nomogram was used to study the predictive value of TARS expression. High TARS expression had shorter overall survival (Figure 8A). Higher stage, histologic grade, pre-menopause status, or more residual tumor exhibited shorter overall survival. The ROC curves showed moderate diagnostic capability (Figure 8B). The nomogram-predicted probability of 1-year (Figure 8C), 3-year (Figure 8D), and 5-year (Figure 8E) overall survival was close to the corresponding actual overall survival, respectively. Moreover, the decision curves reflecting the prediction model confirmed that high TARS expression could predict shorter overall survival (Figure 8F–8H). **Figure 8:** *Predictive value of TARS expression in overall survival. (A, B) ROC curves evaluating the TARS expression for predicting overall survival. (C) Nomogram predicted 1-year overall survival vs. actual 1-year overall survival. (D) Nomogram predicted 3-year overall survival vs. actual 3-year overall survival. (E) Nomogram predicted 5-year overall survival vs. actual 5-year overall survival. (F–H) Decision curve analysis reflects the feasibility of TARS expression in predicting 1-year, 3-year, and 5-year overall survival.* ## Predictive value of TARS expression in disease specific survival High TARS expression had shorter disease specific survival (Figure 9A). Higher stage, histologic grade, or more residual tumor had shorter disease specific survival. The ROC curves showed moderate diagnostic capability (Figure 9B). The nomogram-predicted probability of 1-year (Figure 9C), 3-year (Figure 9D), and 5-year (Figure 9E) disease specific survival was close to the corresponding actual disease specific survival, respectively. Moreover, the decision curves reflecting the prediction model confirmed that high TARS expression could predict shorter disease specific survival (Figure 9F–9H). **Figure 9:** *Predictive value of TARS expression in disease specific survival. (A, B) ROC curves evaluating the TARS expression for predicting disease specific survival. (C) Nomogram predicted 1-year disease specific survival vs. actual 1-year disease specific survival. (D) Nomogram predicted 3-year disease specific survival vs. actual 3-year disease specific survival. (E) Nomogram predicted 5-year disease specific survival vs. actual 5-year disease specific survival. (F–H) Decision curve analysis reflects the feasibility of TARS expression in predicting 1-year, 3-year, and 5-year disease specific survival.* ## High TARS expression-enriched pathways High TARS expression-enriched pathways were screened by GSEA analysis (Supplementary Table 2). High TARS expression was significant correlated with unfolded protein response, MTORC1 signaling, protein secretion, G2M checkpoint, mitotic spindle, Myc targets v1, DNA repair, E2F targets, oxidative phosphorylation, Myc targets v2, and androgen response (Figure 10A–10K). These high TARS expression-enriched pathways may influence the immune response and cell proliferation of endometrial cancer. **Figure 10:** *High TARS expression-enriched pathways. (A) Androgen response. (B) DNA repair. (C) E2F targets. (D) G2M checkpoint. (E) Mitotic spindle. (F) MTORC1 signaling. (G) Myc targets v1. (H) Myc targets v2. (I) Protein secretion. (J) Unfolded protein response. (K) Oxidative phosphorylation.* ## Correlation between TARS expression and immune cells Based on the results of GSEA analysis, the correlation between TARS expression and immune cells were evaluated. After screening, only 4 types of immune cells showed significant correlation with TARS expression ($P \leq 0.001$), including activated CD4+ T cell (Figure 11A), effector memory CD4+ T cell (Figure 11B), memory B cell (Figure 11C), and type 2 T helper cell (Figure 11D). The results suggested that these 4 types of immune cells may participate in the high TARS expression related immune response in endometrial cancer. **Figure 11:** *Correlation between TARS expression and immune cells. (A) Activated CD4+ T cell. (B) Effector memory CD4+ T cell. (C) Memory B cell. (D) Type 2 T helper cell.* ## High TARS expression in tissue and cell Compared with adjacent normal tissue, TRAS expression was significantly higher ($P \leq 0.001$) in endometrial cancer (Figure 12A). Also, significant higher TARS expression was observed in endometrial cancer cell lines (Figure 12B). Of note, Ishikawa showed the highest TARS expression, therefore used in the subsequent cell proliferation experiments. The effect of si-TRAS and O-TARS on TARS expression was verified (Figure 12C). TARS expression was significantly lower in si-TARS group ($P \leq 0.01$), and higher in O-TARS group ($P \leq 0.01$). **Figure 12:** *High TARS expression in tissue and cell. (A) TARS expression in 30 endometrial cancer tissues and adjacent normal tissues by qRT-PCR. (B) Relative TARS expression in HaCaT, hEEC, KLE, HEC1-A, HEC1-B, Ishikawa, and RL95-2 by qRT-PCR. (C) Relative TARS expression in Ishikawa cells transfected with control, si-NC, si-TARS, and O-TARS by qRT-PCR. Abbreviation: NS; no significance; **P < 0.01; ***P < 0.001.* ## TARS knockdown inhibits cell proliferation The function of TARS on Ishikawa cell proliferation was studied using strategies of knockdown and over expression. The CCK-8 results (Figure 13A) showed significantly inhibited cell proliferation in si-TARS ($P \leq 0.05$) and promoted cell proliferation in O-TARS ($P \leq 0.05$). Besides, the colony formation results (Figure 13B) showed decreased colonies in si-TARS ($P \leq 0.01$) and increased colonies in O-TARS ($P \leq 0.05$). Finally, the live/dead staining further confirmed the results (Figure 13C). By quantification (Figure 13D), there were fewer live cells in si-TARS ($P \leq 0.01$) and more live cells in O-TARS ($P \leq 0.01$). **Figure 13:** *TARS knockdown inhibits cell proliferation. (A) Relative cell proliferation of Ishikawa cell by CCK-8 assay. (B) Relative number of colonies of Ishikawa cell. (C) Co-staining of calcein AM and PI of Ishikawa cell, and (D) relative live cells. The live cells were stained with green fluorescence, and the dead cells were stained with red fluorescence. Scale bar = 50 μm. Abbreviation: NS; no significance; *P < 0.05; **P < 0.01.* ## DISCUSSION The evolution of endometrial cancer is involved in multiple genes and develops in multiple steps, which is mainly related to the activation of protooncogenes, and loss or mutation of tumor suppressor genes [22]. The cancer heterogeneity and individual differences have brought additional difficulties to the diagnosis and precise treatment. The rapid breakthrough of the whole genome sequencing technology provides new ideas for the clinical problems and the study of related pathological mechanism [23]. The bioinformatics analysis based on a few samples increases the risk of obtaining fake positive results. Therefore, this study searched and downloaded the expression data of endometrial cancer and normal endometrial tissue genes from TCGA database. In this study, high TARS expression was found in endometrial cancer. TARS was also significantly highly expressed in serous type ($$P \leq 0.0011$$), G3 grade ($P \leq 0.001$), and deceased status ($$P \leq 0.0012$$). The results showed significant association between high TARS expression with poor overall survival ($$P \leq 0.0012$$) and poor disease specific survival ($$P \leq 0.0034$$). In recent years, the incidence and disease mortality of endometrial cancer have been increased around the world [24]. The pathogenesis of endometrial cancer is not clear, but it is generally believed that hypertension and diabetes are high-risk factors for endometrial cancer [25]. Here, significant differences were observed in advanced stage ($$P \leq 0.0053$$), G3 and G4 ($$P \leq 0.015$$), and old ($$P \leq 0.0032$$). Nevertheless, early stage, G1 and G2, and young subgroups showed no statistical differences. The stage (HR: 1.589, $95\%$ CI: 1.246–2.027, $P \leq 0.001$), diabetes (HR: 1.556, $95\%$ CI: 1.068–2.267, $$P \leq 0.021$$), histologic grade (HR: 2.078, $95\%$ CI: 1.289–3.352, $$P \leq 0.003$$), and TARS expression (HR: 4.912, $95\%$ CI: 1.765–13.674, $$P \leq 0.002$$) showed independent prognostic value for overall survival of endometrial cancer. Significant differences were observed in advanced stage ($$P \leq 0.0076$$), G3 and G4 ($$P \leq 0.026$$), and old ($$P \leq 0.0092$$). Nevertheless, early stage, G1 and G2, and young subgroups showed no statistical differences. The stage (HR: 2.109, $95\%$ CI: 1.536–2.897, $P \leq 0.001$), histologic grade (HR: 2.646, $95\%$ CI: 1.335–5.244, $$P \leq 0.005$$), and TARS expression (HR: 6.723, $95\%$ CI: 1.606–28.152, $$P \leq 0.009$$) showed independent prognostic value for disease specific survival of endometrial cancer. Although the diagnosis and treatment methods and prognosis of endometrial cancer have made considerable progress in recent years, the incidence and mortality of endometrial cancer have not been reduced [26]. There is an emergency need to effectively predict the prognostic indicators to improve the survival of patients with endometrial cancer [27]. However, endometrial cancer has no specific serum biomarkers, and the prognostic biomarkers are also limited [28]. Our results found high TARS expression could predict shorter overall survival and disease specific survival. High TARS expression was confirmed in tissue and cell. Biomarkers are valuable for screening women with high risk of endometrial cancer, dividing patients into different prognosis risks, and evaluating the prognosis differences to achieve personalized treatment [29, 30]. The tumor biomarker CA125 (Carbohydrate Antigen 125) contributes to the diagnosis of endometrial cancer [31, 32]. Compared with early endometrial cancer, the concentration of CA125 usually increases in type II or advanced endometrial cancer [33]. The 5-year survival rate of endometrial cancer metastasis dropped to $17\%$. So far, there are no biomarkers with high specialty and strong sensitivity as an early diagnosis or prognostic evaluation indicator [34]. In our study of TARS, the AUC of ROC curve between normal and tumor was 0.901. Besides, the AUC was 0.890 for stage I, 0.864 for stage II, 0.936 for stage III, and 0.957 for stage IV. The results indicated promising diagnostic value of TARS expression. High TARS expression was significant correlated with unfolded protein response, MTORC1 signaling, protein secretion, G2M checkpoint, mitotic spindle, Myc targets v1, DNA repair, E2F targets, oxidative phosphorylation, Myc targets v2, and androgen response. The results suggested that activated CD4+ T cell, effector memory CD4+ T cell, memory B cell and type 2 T helper cell may participate in the high TARS expression related immune response in endometrial cancer. Autoantibodies directed against one or more ARSs are present in anti-synthetase syndrome (ASSD), an autoimmune illness that is also defined by clinical symptoms [35]. Zhou et al. reported the tumor mutation burden and immune infiltrates in endometrial cancer [36]. The function of TARS on Ishikawa cell proliferation was studied using strategies of knockdown and over expression. The CCK-8 results showed significantly inhibited cell proliferation in si-TARS ($P \leq 0.05$) and promoted cell proliferation in O-TARS ($P \leq 0.05$). Besides, the colony formation results showed decreased colonies in si-TARS ($P \leq 0.01$) and increased colonies in O-TARS ($P \leq 0.05$). Finally, the live/dead staining further confirmed the results. By quantification, there were fewer live cells in si-TARS ($P \leq 0.01$) and more live cells in O-TARS ($P \leq 0.01$). Studies of Vivacqua et al. have found that active metabolites of selective estrogen receptor modulator TAM, 4-hydroxyl Moqifen, promotes the cell proliferation of Ishikawa and HEC1-A through the GPR30 pathway rather than rely on ERα's rapid response pathway [37]. The molecular mechanisms of TARS in immune response and cell proliferation need to be further explored. The study is in lack of prospective follow-up data for further verification. ## CONCLUSION High TARS expression was found in endometrial cancer with prognostic and predictive value. High TARS expression is significantly associated with poor overall survival and poor disease specific survival. Activated CD4+ T cell, effector memory CD4+ T cell, memory B cell, and type 2 T helper cell may participate in the high TARS expression related immune response in endometrial cancer. TARS knockdown inhibits cell proliferation. This study will provide new biomarker TARS for diagnosis and prognosis of endometrial cancer. ## References 1. Amant F, Moerman P, Neven P, Timmerman D, Van Limbergen E, Vergote I. **Endometrial cancer.**. *Lancet* (2005) **366** 491-505. DOI: 10.1016/S0140-6736(05)67063-8 2. Felix AS, Yang HP, Bell DW, Sherman ME. **Epidemiology of Endometrial Carcinoma: Etiologic Importance of Hormonal and Metabolic Influences.**. *Adv Exp Med Biol* (2017) **943** 3-46. DOI: 10.1007/978-3-319-43139-0_1 3. Crosbie EJ, Kitson SJ, McAlpine JN, Mukhopadhyay A, Powell ME, Singh N. **Endometrial cancer.**. *Lancet* (2022) **399** 1412-28. DOI: 10.1016/S0140-6736(22)00323-3 4. Lu KH, Broaddus RR. **Endometrial Cancer.**. *N Engl J Med* (2020) **383** 2053-64. DOI: 10.1056/NEJMra1514010 5. Urick ME, Bell DW. **Clinical actionability of molecular targets in endometrial cancer.**. *Nat Rev Cancer* (2019) **19** 510-21. DOI: 10.1038/s41568-019-0177-x 6. Makker V, MacKay H, Ray-Coquard I, Levine DA, Westin SN, Aoki D, Oaknin A. **Endometrial cancer.**. *Nat Rev Dis Primers* (2021) **7** 88. DOI: 10.1038/s41572-021-00324-8 7. Vermij L, Smit V, Nout R, Bosse T. **Incorporation of molecular characteristics into endometrial cancer management.**. *Histopathology* (2020) **76** 52-63. DOI: 10.1111/his.14015 8. Matteson KA, Robison K, Jacoby VL. **Opportunities for Early Detection of Endometrial Cancer in Women With Postmenopausal Bleeding.**. *JAMA Intern Med* (2018) **178** 1222-3. DOI: 10.1001/jamainternmed.2018.2819 9. Brooks RA, Fleming GF, Lastra RR, Lee NK, Moroney JW, Son CH, Tatebe K, Veneris JL. **Current recommendations and recent progress in endometrial cancer.**. *CA Cancer J Clin* (2019) **69** 258-79. DOI: 10.3322/caac.21561 10. Terzic M, Aimagambetova G, Kunz J, Bapayeva G, Aitbayeva B, Terzic S, Laganà AS. **Molecular Basis of Endometriosis and Endometrial Cancer: Current Knowledge and Future Perspectives.**. *Int J Mol Sci* (2021) **22** 9274. DOI: 10.3390/ijms22179274 11. Kwon NH, Fox PL, Kim S. **Aminoacyl-tRNA synthetases as therapeutic targets.**. *Nat Rev Drug Discov* (2019) **18** 629-50. DOI: 10.1038/s41573-019-0026-3 12. Jeong SJ, Park S, Nguyen LT, Hwang J, Lee EY, Giong HK, Lee JS, Yoon I, Lee JH, Kim JH, Kim HK, Kim D, Yang WS. **A threonyl-tRNA synthetase-mediated translation initiation machinery.**. *Nat Commun* (2019) **10** 1357. DOI: 10.1038/s41467-019-09086-0 13. Freist W, Gauss DH. **Threonyl-tRNA synthetase.**. *Biol Chem Hoppe Seyler* (1995) **376** 213-24. PMID: 7626230 14. Cai H, Jiao Y, Li Y, Yang Z, He M, Liu Y. **Low CYP24A1 mRNA expression and its role in prognosis of breast cancer.**. *Sci Rep* (2019) **9** 13714. DOI: 10.1038/s41598-019-50214-z 15. Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez JC, Müller M. **pROC: an open-source package for R and S+ to analyze and compare ROC curves.**. *BMC Bioinformatics* (2011) **12** 77. DOI: 10.1186/1471-2105-12-77 16. Liu J, Zhou S, Li S, Jiang Y, Wan Y, Ma X, Cheng W. **Eleven genes associated with progression and prognosis of endometrial cancer (EC) identified by comprehensive bioinformatics analysis.**. *Cancer Cell Int* (2019) **19** 136. DOI: 10.1186/s12935-019-0859-1 17. He J, Zhang H. **The antitumor effect of lobaplatin against Ishikawa endometrial cancer cells in vitro and in vivo.**. *Biomed Pharmacother* (2019) **114** 108762. DOI: 10.1016/j.biopha.2019.108762 18. Cao Z, Wang H, Mao X, Luo L. **Noncanonical function of threonyl-tRNA synthetase regulates vascular development in zebrafish.**. *Biochem Biophys Res Commun* (2016) **473** 67-72. DOI: 10.1016/j.bbrc.2016.03.051 19. Cai H, Dai X, Guo X, Zhang L, Cao K, Yan F, Ji B, Liu Y. **Ataxia telangiectasia mutated inhibitor-loaded copper sulfide nanoparticles for low-temperature photothermal therapy of hepatocellular carcinoma.**. *Acta Biomater* (2021) **127** 276-86. DOI: 10.1016/j.actbio.2021.03.051 20. Heinemann J. **Cluster Analysis of Untargeted Metabolomic Experiments.**. *Methods Mol Biol* (2019) **1859** 275-85. DOI: 10.1007/978-1-4939-8757-3_16 21. Ranstam J, Cook JA. **Kaplan-Meier curve.**. *Br J Surg* (2017) **104** 442. DOI: 10.1002/bjs.10238 22. McAlpine J, Leon-Castillo A, Bosse T. **The rise of a novel classification system for endometrial carcinoma; integration of molecular subclasses.**. *J Pathol* (2018) **244** 538-49. DOI: 10.1002/path.5034 23. Stratton MR, Campbell PJ, Futreal PA. **The cancer genome.**. *Nature* (2009) **458** 719-24. DOI: 10.1038/nature07943 24. Murali R, Soslow RA, Weigelt B. **Classification of endometrial carcinoma: more than two types.**. *Lancet Oncol* (2014) **15** e268-78. DOI: 10.1016/S1470-2045(13)70591-6 25. Onstad MA, Schmandt RE, Lu KH. **Addressing the Role of Obesity in Endometrial Cancer Risk, Prevention, and Treatment.**. *J Clin Oncol* (2016) **34** 4225-30. DOI: 10.1200/JCO.2016.69.4638 26. Arend RC, Jones BA, Martinez A, Goodfellow P. **Endometrial cancer: Molecular markers and management of advanced stage disease.**. *Gynecol Oncol* (2018) **150** 569-80. DOI: 10.1016/j.ygyno.2018.05.015 27. Hutt S, Tailor A, Ellis P, Michael A, Butler-Manuel S, Chatterjee J. **The role of biomarkers in endometrial cancer and hyperplasia: a literature review.**. *Acta Oncol* (2019) **58** 342-52. DOI: 10.1080/0284186X.2018.1540886 28. Sanderson PA, Critchley HO, Williams AR, Arends MJ, Saunders PT. **New concepts for an old problem: the diagnosis of endometrial hyperplasia.**. *Hum Reprod Update* (2017) **23** 232-54. DOI: 10.1093/humupd/dmw042 29. Zhou L, Wang W, Wang F, Yang S, Hu J, Lu B, Pan Z, Ma Y, Zheng M, Zhou L, Lei S, Song P, Liu P. **Plasma-derived exosomal miR-15a-5p as a promising diagnostic biomarker for early detection of endometrial carcinoma.**. *Mol Cancer* (2021) **20** 57. DOI: 10.1186/s12943-021-01352-4 30. Li J, Dowdy S, Tipton T, Podratz K, Lu WG, Xie X, Jiang SW. **HE4 as a biomarker for ovarian and endometrial cancer management.**. *Expert Rev Mol Diagn* (2009) **9** 555-66. DOI: 10.1586/erm.09.39 31. Bian J, Sun X, Li B, Ming L. **Clinical Significance of Serum HE4, CA125, CA724, and CA19-9 in Patients With Endometrial Cancer.**. *Technol Cancer Res Treat* (2017) **16** 435-9. DOI: 10.1177/1533034616666644 32. Li J, Wang X, Qu W, Wang J, Jiang SW. **Comparison of serum human epididymis protein 4 and CA125 on endometrial cancer detection: A meta-analysis.**. *Clin Chim Acta* (2019) **488** 215-20. DOI: 10.1016/j.cca.2018.11.011 33. Jiang T, Huang L, Zhang S. **Preoperative serum CA125: a useful marker for surgical management of endometrial cancer.**. *BMC Cancer* (2015) **15** 396. DOI: 10.1186/s12885-015-1260-7 34. Pabalan N, Kunjantarachot A, Ruangpratheep C, Jarjanazi H, Christofolini DM, Barbosa CP, Bianco B. **Potential of RASSF1A promoter methylation as biomarker for endometrial cancer: A systematic review and meta-analysis.**. *Gynecol Oncol* (2017) **146** 603-8. DOI: 10.1016/j.ygyno.2017.06.017 35. Galindo-Feria AS, Notarnicola A, Lundberg IE, Horuluoglu B. **Aminoacyl-tRNA Synthetases: On Anti-Synthetase Syndrome and Beyond.**. *Front Immunol* (2022) **13** 866087. DOI: 10.3389/fimmu.2022.866087 36. Zhou H, Chen L, Lei Y, Li T, Li H, Cheng X. **Integrated analysis of tumor mutation burden and immune infiltrates in endometrial cancer.**. *Curr Probl Cancer* (2021) **45** 100660. DOI: 10.1016/j.currproblcancer.2020.100660 37. Vivacqua A, Bonofiglio D, Recchia AG, Musti AM, Picard D, Andò S, Maggiolini M. **The G protein-coupled receptor GPR30 mediates the proliferative effects induced by 17beta-estradiol and hydroxytamoxifen in endometrial cancer cells.**. *Mol Endocrinol* (2006) **20** 631-46. DOI: 10.1210/me.2005-0280
--- title: Identification of fatty acid metabolism-related clusters and immune infiltration features in hepatocellular carcinoma authors: - Zhixuan Ren - Duan Gao - Yue Luo - Zhenghui Song - Guojing Wu - Na Qi - Aimin Li - Xinhui Liu journal: Aging (Albany NY) year: 2023 pmcid: PMC10042688 doi: 10.18632/aging.204557 license: CC BY 3.0 --- # Identification of fatty acid metabolism-related clusters and immune infiltration features in hepatocellular carcinoma ## Abstract Hepatocellular Carcinoma (HCC) is a type of liver cancer which is characterized by inflammation-associated tumor. The unique characteristics of tumor immune microenvironment in HCC contribute to hepatocarcinogenesis. It was also clarified that aberrant fatty acid metabolism (FAM) might accelerate tumor growth and metastasis of HCC. In this study, we aimed to identify fatty acid metabolism-related clusters and establish a novel prognostic risk model in HCC. Gene expression and corresponding clinical data were searched from the Cancer Genome Atlas (TCGA) and the International Cancer Genome Consortium (ICGC) portal. From the TCGA database, by unsupervised clustering method, we determined three FAM clusters and two gene clusters with distinct clinicopathological and immune characteristics. Based on 79 prognostic genes identified from 190 differentially expressed genes (DEGs) among three FAM clusters, five prognostic DEGs (CCDC112, TRNP1, CFL1, CYB5D2, and SLC22A1) were determined to construct risk model by least absolute shrinkage and selection operator (LASSO) and multivariate cox regression analysis. Furthermore, the ICGC dataset was used to validate the model. In conclusion, the prognostic risk model constructed in this study exhibited excellent indicator performance of overall survival, clinical feature, and immune cell infiltration, which has the potential to be an effective biomarker for HCC immunotherapy. ## INTRODUCTION As a major health burden in the world, liver cancer is expected to affect more than one million people by 2025 [1]. The most common primary liver cancer, hepatocellular carcinoma (HCC), ranks fourth among all cancer-related deaths [1]. HCC patients in early stage can be cured by resection, transplantation, thermal ablation and TACE [2]. Early detection of HCC can increase the possibility of potentially curative treatment. Nevertheless, since early HCC diagnosis is challenging, the prognosis of HCC patients remains dismal. HCC patients with intermediate-stage can benefit from catheter-based locoregional treatment [3]. The multitargeted Tyrosine kinase inhibitors (TKI) sorafenib and lenvatinib were approved for the treatment of advanced-stage HCC [4]. A subset of patients treated with immune checkpoint inhibitors has demonstrated strong anti-tumor activity [5]. Identifying and validating predictive biomarkers is a major challenge for HCC immunotherapy. Thus, it is imperative to search novel molecular biomarkers to improve the diagnostic accuracy and guide therapies for HCC patients. In HCC, cancer cells undergo considerable metabolic reprogramming when preparing to proliferate [6]. It has been clear that lipid metabolic rewiring is an influential metabolic alteration in cancer cells. Fatty acid is an integral component of lipid metabolism, it participates in membrane synthesis, storage of energy, and production of signaling molecules [7]. Over the past few years, there has been expanding understanding of the role of fatty acid metabolism (FAM) in tumor progression [8]. Cancer cells can obtain fatty acids from both intracellular and extracellular sources, and changes in fatty acid metabolism are characteristics of oncogenesis and metastasis [9]. By enhancing lipid synthesis, storage and degradation, aberrant fatty acid metabolism impacts the biology of cancer cells to drive tumorigenesis and disease progression [8]. A recent study has uncovered that fatty acid level influenced by cancer cell fatty acid metabolism can change CD8+ T cell activity [10]. It also found that tumor and immune cells compete for fatty acids, which promotes tumor growth [10]. According to a study, fatty acid chain lengthening has been determined as a distinguishing feature of lung cancer [11]. Increasing evidences indicated that fatty acids may contribute to the cancer initiation and development such as gastric cancer, colorectal cancer and breast cancer [12–14]. Deregulated fats can also affect the efficacy of chemotherapy and radiation therapy for cancer patients [15, 16], as well as the effectiveness of immunotherapy. Treatments that target deregulated fatty acids and the inhibition of immune checkpoints in cancer may augment each other’s effects [17]. HCC prevention and treatment may benefit greatly from an understanding of fatty acid metabolism heterogeneity, nonetheless, few studies that investigate possible mechanism and prognostic value of fatty acid metabolism-related genes (FAMs) have been conducted in HCC. In this study, we explored the fatty acid metabolism-related clusters and assessed the composition of tumor microenvironment (TME) in HCC. First, based on expression of 49 FAMs, we identified 3 FAM clusters with distinct biological pathways and immune characteristics. Then 2 gene clusters were determined according to 190 DEGs retrieved from 3 FAM clusters. Afterward, based on the prognostic value of 190 DEGs, we established a prognostic model. Finally, the reliability of the model and the immune landscape of HCC samples were determined. ## Data source On TCGA website (https://portal.gdc.cancer.gov/), gene expression information (fragments per kilobase million, FPKM) and clinical characteristics of 371 HCC patients were obtained. From ICGC database (https://icgc.org/), we acquired information of another 231 HCC patients, including RNA-seq data and clinical features [18]. Based on previous descriptions, we have transformed the LIHC (liver hepatocellular carcinoma) FPKM values into TPM (transcripts per kilobase million) values [19]. Through the GeneCards database, using “fatty acid metabolism” as a keyword, the fatty acid metabolism-related genes (FAMs) was searched and screened. Then, with a relevance score ≥ 50, 49 FAMs were retrieved for the next analyses and provided in Supplementary Table 1. In order to assess mutation states of FAM-related genes in HCC samples, mutation data was processed by “maftools” R package [20]. ## Consensus clustering for FAMs As a result of consensus unsupervised clustering analysis, HCC patients were categorized into different clusters by the R package “ConsensusClusterPlus” based on the FAMs expression [21]. Using the R packages “survival” and “survminer”, we tested whether there are any differences in survival time between distinct clusters using Kaplan-Meier curves. A heatmap plot of the clinical and pathological characteristics was created using R’s “pheatmap” package. From the MSigDB (molecular signatures database) (https://www.gsea-msigdb.org/gsea/msigdb) we extracted the hallmark gene sets (c2.cp.kegg.v7.5.1) and performed gene set variation analysis (GSVA) to determine different biological processes between distinct clusters. ## Immune landscape analysis From previous literature, we gathered the gene sets for immune cells [22], and collected cancer-related gene signatures using the MSigDB. The level of immune cell infiltration and cancer-related gene signatures in the HCC tumor microenvironment were evaluated using single-sample gene set enrichment analysis (ssGSEA) [23]. As well, we compared expression levels of several immune checkpoints and HLA genes between different clusters. ## Analysis of DEGs With an adjusted p-value of 0.001, we identified DEGs among different FAM clusters by the “limma” package in R. By using the “clusterprofiler” R package, we performed functional enrichment analyses on the DEGs to determine their potential functions and pathways. Furthermore, by using a method of unsupervised clustering based prognostic DEGs expression, HCC samples were classified into different clusters (FAM gene cluster D and FAM gene cluster E) for deeper analysis. ## Generation of the FAM-related prognostic model Prognostic analysis of DEGs was conducted using univariate cox regression. Afterward, a prognostic model was established by lasso regression analysis and multivariate cox analysis. The risk score was calculated based on the follow formula: risk score = Σ (Expi × coefi). Expi means expression of each gene, and coefi represents the risk coefficient. HCC samples were categorized by the median risk score into high and low risk groups. Kaplan-Meier curves (K-M curves) were generated using the “survival” and “survminer” R packages in order to investigate the differences in survival between distinct groups. Based on clinical characteristics and risk scores, the “rms” package in R is used to plot the nomogram to predict survival outcomes [24]. Calibrating plot was used to determine the accuracy of the nomogram. In order to verify the model, we divided the ICGC set into high- and low-risk groups, and performed Kaplan–Meier curve and receiver operating characteristic (ROC) curve of ICGC set. ## Statistical analyses Our statistical analyses were performed using R version 4.1.2. Differential clinical characteristics among distinct groups were analyzed by the Chi-squared test. Cox regression analysis (univariate and multivariate) was conducted to identify the independent prognostic factors. Comparison between the two groups was performed using Wilcox rank sum test. The significance level was set at p × 0.05, and two-tailed p values were applied. ## Landscape of genetic variation and transcriptional alterations of FAMs in HCC 49 FAMs obtained from the Genecards website were included in this study. Based on analysis of somatic mutation incidence, the TCGA set of 49 FAMs displayed a relatively high rate of somatic mutations. FAMs mutations were detected in 119 ($32.69\%$) of the 364 HCC samples (Figure 1A). Among these, ALB was found with the highest mutation frequency ($13\%$), followed by APOB. **Figure 1:** *Multi-omics landscape of FAM-related genes in HCC based on TCGA cohort. (A) The mutation frequency of 49 FAMs in TCGA-LIHC cohort. Each column of the figure represents an individual patient. (B) The CNV variation frequency of FAMs (Red and green plots separately represent CNV gain and CNV loss). (C) Locations of CNV alterations in FAMs on 23 chromosomes. (D) The mRNA expression levels of 49 FAMs between HCC and normal tissues. Abbreviations: FAM: fatty acid metabolism; HCC: hepatocellular carcinoma; FAMs: fatty acid metabolism-related genes; TCGA: The Cancer Genome Atlas; LIHC: liver hepatocellular carcinoma; CNV: copy number variant. *p < 0.05; **p < 0.01; ***p < 0.001.* Afterward, we examined somatic copy number alterations (CNVs) in these 49 FAMs and found widespread alterations in all 49 FAMs. Among them, FASN, ACOX1 and MTR showed increased CNVs, while FABP3, ACADVL, HADH, FAAH, and ACADS showed decreases in CNVs (Figure 1B). The CNVs in the FAMs on their respective chromosomes were showed in Figure 1C. Moreover, a comparison of mRNA levels of FAMs was made between HCC tumor and normal tissues, and as showed in Figure 1D, most FAMs expression levels were positively correlated with CNV gain or loss and significantly different in tumor tissues. Consequently, while CNVs can be the primary cause of FAM expression changes, they are not the only factor that regulates mRNA expression [25]. Gene expression can also be affected by transcription factors and DNA methylation [26, 27]. We found HCC and normal samples have remarkably different genetic landscapes and mRNA expression levels of FAMs, indicating that FAMs may play an undiscovered role in HCC. Furthermore, Supplementary Figure 1A shows that the association between each FAM was highly correlated. Similarly, the infiltration levels of immune cells were assessed by ssGSEA algorithm and they showed high correlation in HCC (Supplementary Figure 1B). In summary, the above results indicated that FAMs are strongly correlated with HCC. ## Identification of FAM cluster in HCC Through a FAMs network (Figure 2A), the full scope of FAMs interactions and their prognostic value in HCC patients was displayed. Next, consensus clustering analysis was used to investigate interactions between FAMs and HCC. Using a consensus clustering algorithm, HCC patients were categorized into different clusters (Supplementary Figure 2). Using $k = 3$, we were able to sort the entire cohort into cluster A ($$n = 197$$), B ($$n = 72$$) and C ($$n = 102$$) (Figure 2B). A principal component analysis (PCA) of the FAMs transcription profiles highlighted significant differences among the three clusters (Figure 2C). The Kaplan-Meier curves for the three FAM clusters indicated that cluster C had the most prominent survival advantage, while cluster A had the worst ($p \leq 0.05$) (Figure 2D). Moreover, as shown in heatmap of clinicopathological features and expression of FAMs in HCC patients, cluster A displayed the lowest level of FAMs expression (Figure 2E). **Figure 2:** *FAM clusters and relevant clinical features. (A) The interaction of expression on 49 FAMs in HCC. The line connecting the FAMs represents their interactions, with the line thickness indicating the strength of the association between FAMs. Red dots, fatty acid metabolism-related genes; Purple dots, risk factors for HCC; Green dots, favorable factors for HCC; Pink edges, positive correlation with P < 0.0001; Blue edges, negative correlation with P < 0.0001. (B) Consensus matrices of 49 FAMs in HCC for k = 3. (C) PCA analysis showing a remarkable difference in transcriptomes between the three FAM clusters in TCGA cohort. (D) K-M curve for the three FAM clusters. (E) The heatmap of clinical characteristics and expression levels of FAMs in different clusters. Abbreviations: FAM: fatty acid metabolism; FAMs: fatty acid metabolism-related genes; HCC: hepatocellular carcinoma; PCA: principal component analysis.* ## GSVA enrichment analysis and immune infiltration estimation in distinct clusters GSVA enrichment analysis was conducted among different clusters in order to identify potential biological pathways in HCC. The top 20 pathways in each cluster were visualized (Figure 3A–3C). Cluster C was significantly enriched in fatty acid metabolism pathway and immune-related pathways, such as PPAR signaling pathway, Toll-like, B cell receptor signaling pathway, Fc-gamma-R-mediated phagocytosis pathway and Nod-like receptor signaling pathway (Figure 3C). Furthermore, in order to assess whether FAMs contribute to TME of HCC, we used the ssGSEA algorithm to calculate connection between the three clusters and 23 kinds of immune cells of every HCC sample. Among the three clusters, there were significant differences in the infiltration of immune cells (Figure 3D). Besides, HCC patients in cluster A had the highest expression level of most immune checkpoints among three FAM clusters (Figure 3E), that implied an exhausted immune TME in cluster A patients. **Figure 3:** *The results of GSVA and immune infiltration analysis in three clusters. GSVA results of biological pathways between: (A) cluster A vs. cluster B, (B) cluster B vs. cluster C, and (C) cluster A vs. cluster C, red and blue represent activated and inhibited pathways, respectively. (D) The infiltration levels of 23 immune cells in the three FAM clusters. (E) Significant differences in expression of immune checkpoint in the three FAM clusters. Abbreviations: GSVA: gene set variation analysis; FAM: fatty acid metabolism. *p < 0.05; **p < 0.01; ***p < 0.001.* ## Identified of gene clusters based on FAM cluster-related DEGs in HCC In the previous steps, three clusters were determined, then significant DEGs with adjusted p value < 0.001 were identified by differential analyses between any two clusters. The Venn diagram (Figure 4A) illustrated the following intersections which resulted in 190 DEGs. A functional enrichment analysis was employed to research the potential biological behavior of 190 DEGs. According to GO (gene ontology) and KEGG (the *Kyoto encyclopedia* of genes and genomes) analysis, these FAM cluster-related genes were significantly enriched in metabolism pathways (Supplementary Figure 3). **Figure 4:** *Identification of gene clusters based on DEGs in the TCGA-LIHC cohort. (A) Venn diagram showed the DEGs among the three FAM clusters. (B) HCC samples were divided into two clusters based on the consensus clustering (k = 2). (C) The OS analysis of HCC samples between gene cluster D and E. (D) The heatmap of clinical characteristics of HCC patients in different clusters. (E) The mRNA expression levels of 49 FAMs between gene cluster D and E. Abbreviations: DEGs: different expressed genes; TCGA: the Cancer Genome Atlas; LIHC: liver hepatocellular carcinoma; FAM: fatty acid metabolism; HCC: hepatocellular carcinoma; OS: overall survival; FAMs: fatty acid metabolism-related genes. *p < 0.05; **p < 0.01; ***p < 0.001.* Furthermore, 190 genes were screened for prognostic value by univariate cox regression analysis, and among them, 79 genes were found to be associated with overall survival (OS) in HCC (Supplementary Table 2). According to 79 prognostic genes, two genomic clusters named gene clusters D and E were identified by consensus clustering algorithm (Figure 4B and Supplementary Figure 4). According to Kaplan-Meier curves, HCC patients in gene cluster D had poorer overall survival compared to those in gene cluster E (Figure 4C). Afterward, HCC patients in FAM gene cluster D were related with higher FAM gene expression, advanced stage, advanced grade, and higher dead risk (Figure 4D). The result of further expression analysis was consistent with that in heatmap (Figure 4E). In addition, the immune analysis between two gene clusters revealed that gene cluster D tend to have higher infiltration level of most immune cells such as activated B cell, activated CD8+ T cells and activated CD4+ T cells (Figure 5A). Consistent with this, patients in gene cluster D also had higher expression level of immune checkpoints (Figure 5B). Interestedly, we estimated the relative abundance of several important cancer-related signatures by ssGSEA algorithm in different gene clusters (Figure 5C). The results showed that HCC patients in gene cluster D had higher abundance levels of bad prognostic signatures, including EMT (epithelial-mesenchymal transition), poor survival, proliferation, vascular invasion, recurrent, metastasis signatures, and immune microenvironment signatures, such as innate immune response, pan-F-TBRS, co-inhibition antigen presenting cell (APC), co-stimulation APC, co-inhibition T cell, co-stimulation T cell, MHC-I HLA (major histocompatibility complex-I human leukocyte antigen), MHC-II HLA, antigen processing machinery, and immune checkpoint, compared to those in gene cluster E. Figure 5D showed that gene cluster D had higher expression levels of HLA genes. **Figure 5:** *Different immune and cancer-related characteristics in gene cluster D and E. (A) The 23 kinds of immune cells in the two gene clusters. (B) Significant differences in expression of immune checkpoint between the two gene clusters. (C) The enrichment levels of cancer-related signatures in the two gene clusters. (D) Expression levels of HLA genes between gene cluster D and E. *p < 0.05; **p < 0.01; ***p < 0.001.* ## Construction and verification of the prognostic risk model in HCC By lasso regression analysis and multivariate cox analyses in 79 prognostic DEGs, we identified 5 genes including three risk factors (CCDC112, TRNP1, CFL1) and two protective factors (CYB5D2, SLC22A1) and created a prognostic model in HCC according to these five genes (Supplementary Figure 5 and Supplementary Table 3). The risk score of HCC patients was calculated as follows: risk score = 0.382912 × TRNP1 + 0.65021 × CCDC112 + 1.885657 × CFL1 + (−1.23099) × CYB5D2 + (−0.29032) × SLC22A1. In TCGA-LIHC set, the median cut-off value was used to stratify the patients into two groups: high-risk score ($$n = 182$$) and low-risk score ($$n = 183$$). Figure 6A displayed the distribution of HCC patients across three FAM clusters, two gene clusters, and two risk score groups. There was a significant risk score difference between FAM clusters and gene clusters. The risk score of cluster B was the lowest, while that of cluster A was the highest (Figure 6B). Cluster D had a higher risk score than cluster E (Figure 6C). In TCGA-LIHC set, high-risk patients had a worse outcome than low-risk patients, and AUC (Area under curve) values of 0.708, 0.682, and 0.650 respectively represent 1-, 2-, and 3-year survival rates of risk scores (Figure 6D–6F). **Figure 6:** *Construction and validation of prognostic risk model. (A) Alluvial diagram depicting the relationship of FAMcluster, genecluster, risk score (FAMscore) group and survival state. Boxplots of risk score in different FAMclusters (B) and geneclusters (C) Risk score distribution and scatter plots showing the risk score distribution and patient survival status in TCGA (D); Kaplan–Meier analysis of OS between the two groups in TCGA (E); ROC curves to predict the sensitivity and specificity of 1-, 3-, 5-year survival according the risk score in TCGA (F). Risk score distribution and scatter plots (G), Kaplan–Meier curves (H), ROC curves (I) of the risk model in ICGC cohort. The univariate (J) and multivariate (K) independent prognostic analysis of the model in TCGA cohort. Abbreviations: TCGA: the cancer genome atlas database; ICGC: International Cancer Genome Consortium; OS: overall survival; ROC: receiver operating characteristic.* As an external validation cohort, patients in ICGC-JP (ICGC-Japan) cohort were categorized, by the median risk score, into high- and low-risk groups. Consistently, in ICGC-JP cohort, high-risk patients had worse outcomes than low-risk patients, and the corresponding AUC values of 1-, 2-, and 3-year survival rates were 0.777, 0.718, 0.695, respectively, which indicated a good efficiency (Figure 6G–6I). Furthermore, cox regression analysis, both univariate (Figure 6J) and multivariate (Figure 6K), revealed the prognostic risk model is a reliable independent prognostic factor of HCC patients. We have done more exploration of five genes on other databases, such as TCGA (Supplementary Figure 6A–6D), ICGC (Supplementary Figure 6E–6G) and GEO database (GSE25097, GSE112790, GSE102079, GSE45267, GSE39791 datasets) (Supplementary Figure 7). Moreover, we verified IHC on HPA database (Supplementary Figure 8A) and protein expression levels on CPTAC database (Supplementary Figure 8B). Interestingly, all results are consistent with our study, which TRNP1, CCDC112, CFL1 were risk factors (compared to normal tissues, there was a significant upregulation of TRNP1, CCDC112, CFL1 expression in HCC tissues. K-M curves showed that upregulated TRNP1, CCDC112, CFL1 were associated with poor OS) and CYB5D2, SLC22A1 were protective genes (expression of CYB5D2, SLC22A1 were decreased in tumor tissues, and higher expression of CYB5D2, SLC22A1 was associated with good OS). ## Relationship of TME and the prognostic risk model in HCC In the TCGA-LIHC cohort, we assessed the abundance of immune cells and cancer-related signatures by using the ssGSEA algorithm. Through the spearman method, the association among risk score and immune cells, cancer-related signatures levels were evaluated. As shown in the boxplots, the levels of immune cells (Supplementary Figure 9) and immune checkpoints (Figure 7A) in high-risk patients were higher than low-risk patients. Moreover, Figure 7B showed that high-risk patients also had higher abundance levels of bad prognostic signatures, such as EMT, poor survival, proliferation, vascular invasion, recurrent, metastasis signatures, and immune microenvironment signatures, such as innate immune response, pan-F-TBRS, co-inhibition APC, co-inhibition T cell, co-stimulation APC, co-stimulation T cell, MHC-I HLA, MHC-II HL, antigen processing machinery, and immune checkpoint compared to low-risk patients. Also, we conducted gene set enrichment analysis (GSEA) of HCC patients in different risk groups, and the result showed that high-risk group was enriched in *Fc gamma* R mediated phagocytosis, T cell receptor signaling pathway, Nod-like receptor signaling pathway, Fc epsilon Ri signaling pathway, while low risk group was enriched in PPAR signaling pathway and drug metabolism pathway (Figure 7C). **Figure 7:** *Connection among prognostic risk model and immune or cancer-related characteristics of HCC patients. (A) Significant differences in expression of immune checkpoints between the two groups. (B) The enrichment level of cancer-related signatures in the two groups. (C) Immune-related pathways enriched in the high-risk group. The correlation between genes in prognostic risk model and the infiltration level of 23 immune cells (D) and immune-related pathway (E). Red for positive associations and green for negative associations. Abbreviation: HCC: hepatocellular carcinoma. *p < 0.05; **p < 0.01; ***p < 0.001.* Furthermore, the relationship between five genes in the model and immune cells was analyzed (Figure 7D). We observed that three high-risk genes (CCDC112, TRNP1 and CFL1) were significantly positively correlated with most immune cells, whereas significant negative correlation was observed between two low-risk genes (CYB5D2 and SLC22A1) and infiltration of immune cells. Consistently, Figure 7E displayed the result of correlation between five genes, risk score and immune related pathways. ## Construction of nomogram in HCC Comparison of genes mutations between the two risk groups revealed that high-risk patients had significantly higher mutation rates of TP53, MUC4, FLG, CSMD3, ARID1A, FAT3 than low-risk patients (Figure 8A, 8B). Moreover, high-risk HCC patients were remarkably associated with worse outcome, more advanced tumor stage and worse pathological grade (Figure 8C, 8D). To identify the reliability of this risk model in HCC patients, the prognostic nomogram plot containing the risk score and stage was constructed in TCGA-LIHC cohort (Figure 8E). Furthermore, calibration plot indicated excellent agreement between prediction and actual risk (Figure 8F). Overall, the risk model showed good prognostic value in HCC samples. **Figure 8:** *Connections between prognostic risk model and clinical characteristics of HCC patients. The waterfall plot of tumor somatic mutation established in (A) high risk group and (B) low risk group. (C) The heatmap of the model and clinical characteristics in TCGA-LIHC cohort. (D) Stacked bar plot of HCC survival state, pathological grade and tumor stage. (E) Nomogram for predicting the 1-,3- and 5- year OS of HCC patients. (F) Calibration curve of the program for predicting of 1-,3- and 5-year OS of HCC patients.* ## DISCUSSION HCC seriously threatens human health with high mortality rate. While HCC can be managed with multiple treatments, patients with the disease have extremely low 5-year survival rates due to the fact that it is commonly diagnosed in advanced stages [28]. Currently, a number of immune checkpoint inhibitors (ICIs) have been approved by the FDA (Food and Drug Administration) to treat advanced HCC, including nivolumab [29] and pembrolizumab [30]. However, there are numerous disadvantages of ICI treatment, including low response rates and side effects. Therefore, new therapeutic targets and novel prognostic models are essential for HCC patients. Metabolism reprogramming is critical for tumor initiation and progression, especially during HCC development [31]. Synthesis of fatty acids has been involved in energy metabolism and membrane production of tumor cells. Deregulated fatty acid metabolism has been regarded as a vital metabolic regulator in supporting cancer cell proliferation [32]. A remolded microenvironment caused by abnormally fatty acid metabolism could promote HCC progression. In this study, our objective was to assess the association of FAMs and the risk of HCC. First, we explored the mutation and correlation state of 49 FAMs obtained from the Genecard database. The top three frequently mutated genes were ALB, APOB, and FASN. Missense mutation and C>T of FAMs were the most common mutations in HCC. Due to the high expression of ALB ($20\%$) [33] and APOB’s ability to facilitate VLDL secretion [34] (which consumes large amounts of energy), mutation of ALB or APOB may be inactivated to divert energy into cancer-relevant metabolic pathways [35]. According to the expression profiles of 49 FAMs, we determined 3 FAM clusters. Among 3 FAM clusters, cluster C had highest level of immune infiltration. Subsequently, differential analyses among 3 FAM clusters were employed. We screened 190 DEGs and showed them in a Venn plot. Based on the expression of 79 prognostic genes identified from 190 DEGs, HCC patients were grouped into 2 different gene clusters. Gene cluster D had worse survival rate, higher expression level of FAMs, and higher infiltration level of immune cells. Interestingly, gene cluster D also had higher enrichment of poor prognostic signatures, such as poor survival, liver cancer recurrent related signatures, cancer progression related signatures such as EMT, proliferation, vascular invasion, metastasis signatures, and several immune signatures, such as innate immune response, pan-F-TBRS, co-inhibition APC, co-inhibition T cell, co-stimulation APC, co-stimulation T cell, MHC-I HLA, MHC-II HLA, antigen processing machinery and immune checkpoint related signatures. These results indicated that FAMs appear to affect TME of HCC. Moreover, based on 79 prognostic FAMs, a FAM-related model containing 5 genes (TRNP1, CCDC112, CFL1, CYB5D2, SLC22A1) was constructed by LASSO and multivariate Cox regression analysis in TCGA-LICH cohort. And we successfully confirmed the model using ICGC-JP cohort. HCC patients were categorized into two groups, high risk and low risk group. In both the TCGA and ICGC cohorts, the K-M curves showed that patients in the low group had better outcomes than those in the high group. The 1 year AUC of the model was 0.708, 0.777 in TCGA and ICGC cohort, respectively, which demonstrated that the accuracy of the risk model was excellent. The relation of our model and immune infiltration was also assessed. And the infiltration levels of immune cells were evaluated by ssGSEA. The analysis of relationship revealed that risk score was significantly positively correlated with infiltration of immune cell in HCC patients, especially CD56 bright natural killer cell, activated CD4 T cell and activated dendritic cell. The results of correlation analysis between each gene in model and HCC immunity were consistent with the properties of genes. For example, TRNP1, CCDC112, CFL1 are risk factors, then they were positively correlated with the infiltration levels of most of immune cells, whereas the results of CYB5D2, SLC22A1 were contrary to this. Patients in high risk group had higher enrichment level of poor prognostic signatures, such as poor survival, liver cancer recurrent related signatures, cancer progression related signatures such as EMT, proliferation, vascular invasion, metastasis signatures, and several immune signatures, such as innate immune response, pan-F-TBRS, co-inhibition APC, co-inhibition T cell, co-stimulation APC, co-stimulation T cell, MHC-I HLA, MHC-II HLA, antigen processing machinery and immune checkpoint related signatures. TRNA1, CCDC112, and CFL1 expression were substantially higher in HCC tissues than in normal tissues, whereas CYB5D2 and SLC22A1 expression were significantly lower. These results were found in the TCGA, ICGC, GEO, HPA, and CPTAC databases. Liu et al. reported TRNP1 as a risk factor of four-gene model for predicting OS in HCC patients [36]. TRNP1 is essential for neural development and cell self-renewal [37]. As a hypoxia-responsive gene, CFL1 contributes to hypoxia-induced HCC progression by activating PLD1/AKT signals [38]. In a mechanism study, knockdown of CFL1 increased F-actin levels and disrupted the balance between F-actin and G-actin, which resulted in aggressiveness inhibition of HCC cells [39]. Researches have reported that decreased level of CYB5D2 is associated with breast cancer progression [40]. SLC22A1 downregulation correlates with worse patient outcomes and tumor progression [41]. It is thought that the development of HCC is accompanied by aberrant SLC22A1 variants, which may greatly affect the sorafenib levels in the affected intracellular concentrations in HCC [42]. However, there is still a lack of knowledge about how TRNP1 and CCDC112 affect the development and prognosis of HCC. In recognition of the clinical utility of the model in predicting over survival in HCC patients, using the risk score and stage together, a nomogram was created to predict the 1-, 3-, and 5-year survival rates for HCC in TCGA cohort. The calibration plot verified the accuracy of nomograms. Nevertheless, our study has several shortcomings. First, molecular mechanisms of these genes need to be uncovered by additional functional experiments. Second, an additional experiment is needed for further verification of model genes. Finally, since the study was analyzed on data from public database, the risk model needs to be validated by our own clinical cohort. Taken together, we identified 3 FAM clusters, 2 gene clusters and established a novel 5-gene prognostic model for HCC patients. Fatty acid metabolism-related genes exhibited synergy with immune activation. We hope the prognostic model may help improve immunotherapy for HCC in the future. ## CONCLUSION Our findings investigated molecular cluster and prognostic model about fatty acid metabolism in HCC, and highlighted a potential strategy for targeting the immunometabolism of HCC. ## References 1. Llovet JM, Kelley RK, Villanueva A, Singal AG, Pikarsky E, Roayaie S, Lencioni R, Koike K, Zucman-Rossi J, Finn RS. **Hepatocellular carcinoma.**. *Nat Rev Dis Primers* (2021) **7** 6. DOI: 10.1038/s41572-020-00240-3 2. Vogel A, Martinelli E. **Updated treatment recommendations for hepatocellular carcinoma (HCC) from the ESMO Clinical Practice Guidelines.**. *Ann Oncol* (2021) **32** 801-5. DOI: 10.1016/j.annonc.2021.02.014 3. Yang JD, Hainaut P, Gores GJ, Amadou A, Plymoth A, Roberts LR. **A global view of hepatocellular carcinoma: trends, risk, prevention and management.**. *Nat Rev Gastroenterol Hepatol* (2019) **16** 589-604. DOI: 10.1038/s41575-019-0186-y 4. Zhao Y, Zhang YN, Wang KT, Chen L. **Lenvatinib for hepatocellular carcinoma: From preclinical mechanisms to anti-cancer therapy.**. *Biochim Biophys Acta Rev Cancer* (2020) **1874** 188391. DOI: 10.1016/j.bbcan.2020.188391 5. Sangro B, Sarobe P, Hervás-Stubbs S, Melero I. **Advances in immunotherapy for hepatocellular carcinoma.**. *Nat Rev Gastroenterol Hepatol* (2021) **18** 525-43. DOI: 10.1038/s41575-021-00438-0 6. Hall Z, Chiarugi D, Charidemou E, Leslie J, Scott E, Pellegrinet L, Allison M, Mocciaro G, Anstee QM, Evan GI, Hoare M, Vidal-Puig A, Oakley F. **Lipid Remodeling in Hepatocyte Proliferation and Hepatocellular Carcinoma.**. *Hepatology* (2021) **73** 1028-44. DOI: 10.1002/hep.31391 7. Currie E, Schulze A, Zechner R, Walther TC, Farese RV. **Cellular fatty acid metabolism and cancer.**. *Cell Metab* (2013) **18** 153-61. DOI: 10.1016/j.cmet.2013.05.017 8. Hoy AJ, Nagarajan SR, Butler LM. **Tumour fatty acid metabolism in the context of therapy resistance and obesity.**. *Nat Rev Cancer* (2021) **21** 753-66. DOI: 10.1038/s41568-021-00388-4 9. Nagarajan SR, Butler LM, Hoy AJ. **The diversity and breadth of cancer cell fatty acid metabolism.**. *Cancer Metab* (2021) **9** 2. DOI: 10.1186/s40170-020-00237-2 10. Ringel AE, Drijvers JM, Baker GJ, Catozzi A, García-Cañaveras JC, Gassaway BM, Miller BC, Juneja VR, Nguyen TH, Joshi S, Yao CH, Yoon H, Sage PT. **Obesity Shapes Metabolism in the Tumor Microenvironment to Suppress Anti-Tumor Immunity.**. *Cell* (2020) **183** 1848-66.e26. DOI: 10.1016/j.cell.2020.11.009 11. Marien E, Meister M, Muley T, Gomez Del Pulgar T, Derua R, Spraggins JM, Van de Plas R, Vanderhoydonc F, Machiels J, Binda MM, Dehairs J, Willette-Brown J, Hu Y. **Phospholipid profiling identifies acyl chain elongation as a ubiquitous trait and potential target for the treatment of lung squamous cell carcinoma.**. *Oncotarget* (2016) **7** 12582-97. DOI: 10.18632/oncotarget.7179 12. Madak-Erdogan Z, Band S, Zhao YC, Smith BP, Kulkoyluoglu-Cotul E, Zuo Q, Santaliz Casiano A, Wrobel K, Rossi G, Smith RL, Kim SH, Katzenellenbogen JA, Johnson ML. **Free Fatty Acids Rewire Cancer Metabolism in Obesity-Associated Breast Cancer via Estrogen Receptor and mTOR Signaling.**. *Cancer Res* (2019) **79** 2494-510. DOI: 10.1158/0008-5472.CAN-18-2849 13. Wang YN, Zeng ZL, Lu J, Wang Y, Liu ZX, He MM, Zhao Q, Wang ZX, Li T, Lu YX, Wu QN, Yu K, Wang F. **CPT1A-mediated fatty acid oxidation promotes colorectal cancer cell metastasis by inhibiting anoikis.**. *Oncogene* (2018) **37** 6025-40. DOI: 10.1038/s41388-018-0384-z 14. Wang L, Li C, Song Y, Yan Z. **Inhibition of carnitine palmitoyl transferase 1A-induced fatty acid oxidation suppresses cell progression in gastric cancer.**. *Arch Biochem Biophys* (2020) **696** 108664. DOI: 10.1016/j.abb.2020.108664 15. Han S, Wei R, Zhang X, Jiang N, Fan M, Huang JH, Xie B, Zhang L, Miao W, Butler AC, Coleman MA, Vaughan AT, Wang Y. **CPT1A/2-Mediated FAO Enhancement-A Metabolic Target in Radioresistant Breast Cancer.**. *Front Oncol* (2019) **9** 1201. DOI: 10.3389/fonc.2019.01201 16. Wu Y, Fabritius M, Ip C. **Chemotherapeutic sensitization by endoplasmic reticulum stress: increasing the efficacy of taxane against prostate cancer.**. *Cancer Biol Ther* (2009) **8** 146-52. DOI: 10.4161/cbt.8.2.7087 17. Veglia F, Tyurin VA, Blasi M, De Leo A, Kossenkov AV, Donthireddy L, To TKJ, Schug Z, Basu S, Wang F, Ricciotti E, DiRusso C, Murphy ME. **Fatty acid transport protein 2 reprograms neutrophils in cancer.**. *Nature* (2019) **569** 73-8. DOI: 10.1038/s41586-019-1118-2 18. Fujimoto A, Furuta M, Totoki Y, Tsunoda T, Kato M, Shiraishi Y, Tanaka H, Taniguchi H, Kawakami Y, Ueno M, Gotoh K, Ariizumi S, Wardell CP. **Whole-genome mutational landscape and characterization of noncoding and structural mutations in liver cancer.**. *Nat Genet* (2016) **48** 500-9. DOI: 10.1038/ng.3547 19. Conesa A, Madrigal P, Tarazona S, Gomez-Cabrero D, Cervera A, McPherson A, Szcześniak MW, Gaffney DJ, Elo LL, Zhang X, Mortazavi A. **Erratum to: A survey of best practices for RNA-seq data analysis.**. *Genome Biol* (2016) **17** 181. DOI: 10.1186/s13059-016-1047-4 20. Mayakonda A, Lin DC, Assenov Y, Plass C, Koeffler HP. **Maftools: efficient and comprehensive analysis of somatic variants in cancer.**. *Genome Res* (2018) **28** 1747-56. DOI: 10.1101/gr.239244.118 21. Wilkerson MD, Hayes DN. **ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking.**. *Bioinformatics* (2010) **26** 1572-3. DOI: 10.1093/bioinformatics/btq170 22. Long Q, Huang C, Meng Q, Peng J, Yao F, Du D, Wang X, Zhu W, Shi D, Xu X, Qi X, Deng W, Chen M. **TNF Patterns and Tumor Microenvironment Characterization in Head and Neck Squamous Cell Carcinoma.**. *Front Immunol* (2021) **12** 754818. DOI: 10.3389/fimmu.2021.754818 23. Rooney MS, Shukla SA, Wu CJ, Getz G, Hacohen N. **Molecular and genetic properties of tumors associated with local immune cytolytic activity.**. *Cell* (2015) **160** 48-61. DOI: 10.1016/j.cell.2014.12.033 24. Iasonos A, Schrag D, Raj GV, Panageas KS. **How to build and interpret a nomogram for cancer prognosis.**. *J Clin Oncol* (2008) **26** 1364-70. DOI: 10.1200/JCO.2007.12.9791 25. Sebestyén E, Singh B, Miñana B, Pagès A, Mateo F, Pujana MA, Valcárcel J, Eyras E. **Large-scale analysis of genome and transcriptome alterations in multiple tumors unveils novel cancer-relevant splicing networks.**. *Genome Res* (2016) **26** 732-44. DOI: 10.1101/gr.199935.115 26. Koch A, Joosten SC, Feng Z, de Ruijter TC, Draht MX, Melotte V, Smits KM, Veeck J, Herman JG, Van Neste L, Van Criekinge W, De Meyer T, van Engeland M. **Analysis of DNA methylation in cancer: location revisited.**. *Nat Rev Clin Oncol* (2018) **15** 459-66. DOI: 10.1038/s41571-018-0004-4 27. Lambert SA, Jolma A, Campitelli LF, Das PK, Yin Y, Albu M, Chen X, Taipale J, Hughes TR, Weirauch MT. **The Human Transcription Factors.**. *Cell* (2018) **172** 650-65. DOI: 10.1016/j.cell.2018.01.029 28. Zheng J, Kuk D, Gönen M, Balachandran VP, Kingham TP, Allen PJ, D'Angelica MI, Jarnagin WR, DeMatteo RP. **Actual 10-Year Survivors After Resection of Hepatocellular Carcinoma.**. *Ann Surg Oncol* (2017) **24** 1358-66. DOI: 10.1245/s10434-016-5713-2 29. El-Khoueiry AB, Sangro B, Yau T, Crocenzi TS, Kudo M, Hsu C, Kim TY, Choo SP, Trojan J, Welling TH, Meyer T, Kang YK, Yeo W. **Nivolumab in patients with advanced hepatocellular carcinoma (CheckMate 040): an open-label, non-comparative, phase 1/2 dose escalation and expansion trial.**. *Lancet* (2017) **389** 2492-502. DOI: 10.1016/S0140-6736(17)31046-2 30. Verset G, Borbath I, Karwal M, Verslype C, Van Vlierberghe H, Kardosh A, Zagonel V, Stal P, Sarker D, Palmer DH, Vogel A, Edeline J, Cattan S. **Pembrolizumab Monotherapy for Previously Untreated Advanced Hepatocellular Carcinoma: Data from the Open-Label, Phase II KEYNOTE-224 Trial.**. *Clin Cancer Res* (2022) **28** 2547-54. DOI: 10.1158/1078-0432.CCR-21-3807 31. Rui L. **Energy metabolism in the liver.**. *Compr Physiol* (2014) **4** 177-97. DOI: 10.1002/cphy.c130024 32. Li Z, Zhang H. **Reprogramming of glucose, fatty acid and amino acid metabolism for cancer progression.**. *Cell Mol Life Sci* (2016) **73** 377-92. DOI: 10.1007/s00018-015-2070-4 33. Uhlén M, Fagerberg L, Hallström BM, Lindskog C, Oksvold P, Mardinoglu A, Sivertsson Å, Kampf C, Sjöstedt E, Asplund A, Olsson I, Edlund K, Lundberg E. **Proteomics. Tissue-based map of the human proteome.**. *Science* (2015) **347** 1260419. DOI: 10.1126/science.1260419 34. Egusa G, Beltz WF, Grundy SM, Howard BV. **Influence of obesity on the metabolism of apolipoprotein B in humans.**. *J Clin Invest* (1985) **76** 596-603. DOI: 10.1172/JCI112011 35. Fernandez-Banet J, Lee NP, Chan KT, Gao H, Liu X, Sung WK, Tan W, Fan ST, Poon RT, Li S, Ching K, Rejto PA, Mao M, Kan Z. **Decoding complex patterns of genomic rearrangement in hepatocellular carcinoma.**. *Genomics* (2014) **103** 189-203. DOI: 10.1016/j.ygeno.2014.01.003 36. Liu J, Zhang SQ, Chen J, Li ZB, Chen JX, Lu QQ, Han YS, Dai W, Xie C, Li JC. **Identifying Prognostic Significance of RCL1 and Four-Gene Signature as Novel Potential Biomarkers in HCC Patients.**. *J Oncol* (2021) **2021** 5574150. DOI: 10.1155/2021/5574150 37. Esgleas M, Falk S, Forné I, Thiry M, Najas S, Zhang S, Mas-Sanchez A, Geerlof A, Niessing D, Wang Z, Imhof A, Götz M. **Trnp1 organizes diverse nuclear membrane-less compartments in neural stem cells.**. *EMBO J* (2020) **39** e103373. DOI: 10.15252/embj.2019103373 38. Yao B, Li Y, Chen T, Niu Y, Wang Y, Yang Y, Wei X, Liu Q, Tu K. **Hypoxia-induced cofilin 1 promotes hepatocellular carcinoma progression by regulating the PLD1/AKT pathway.**. *Clin Transl Med* (2021) **11** e366. DOI: 10.1002/ctm2.366 39. Zhang L, Chai Z, Kong S, Feng J, Wu M, Tan J, Yuan M, Chen G, Li Z, Zhou H, Cheng S, Xu H. **Nujiangexanthone A Inhibits Hepatocellular Carcinoma Metastasis via Down Regulation of Cofilin 1.**. *Front Cell Dev Biol* (2021) **9** 644716. DOI: 10.3389/fcell.2021.644716 40. Ojo D, Rodriguez D, Wei F, Bane A, Tang D. **Downregulation of CYB5D2 is associated with breast cancer progression.**. *Sci Rep* (2019) **9** 6624. DOI: 10.1038/s41598-019-43006-y 41. Heise M, Lautem A, Knapstein J, Schattenberg JM, Hoppe-Lotichius M, Foltys D, Weiler N, Zimmermann A, Schad A, Gründemann D, Otto G, Galle PR, Schuchmann M, Zimmermann T. **Downregulation of organic cation transporters OCT1 (SLC22A1) and OCT3 (SLC22A3) in human hepatocellular carcinoma and their prognostic significance.**. *BMC Cancer* (2012) **12** 109. DOI: 10.1186/1471-2407-12-109 42. Herraez E, Lozano E, Macias RI, Vaquero J, Bujanda L, Banales JM, Marin JJ, Briz O. **Expression of SLC22A1 variants may affect the response of hepatocellular carcinoma and cholangiocarcinoma to sorafenib.**. *Hepatology* (2013) **58** 1065-73. DOI: 10.1002/hep.26425
--- title: Associations between serum total cholesterol level and bone mineral density in older adults authors: - Sheng Hu - Silin Wang - Wenxiong Zhang - Lang Su - Jiayue Ye - Deyuan Zhang - Yang Zhang - Qiang Guo - Dongliang Yu - Jinhua Peng - Jianjun Xu - Yiping Wei journal: Aging (Albany NY) year: 2023 pmcid: PMC10042689 doi: 10.18632/aging.204514 license: CC BY 3.0 --- # Associations between serum total cholesterol level and bone mineral density in older adults ## Abstract Background: *Osteoporosis is* a major clinical problem in elderly men and women. The correlation between total cholesterol and bone mineral density remains controversial. NHANES is the cornerstone for national nutrition monitoring to inform nutrition and health policy. Methods: Sample sizes and the location of the study and the time when it was conducted: we obtained 4236 non-cancer elderly from NHANES (National Health and Nutrition Examination Survey) database from 1999 to 2006. Data were analyzed with the use of the statistical packages R and EmpowerStats. We analyzed the relationship between total cholesterol and lumbar bone mineral density. We performed research population description, stratified analysis, single factor analysis, multiple equation regression analysis, smooth curve fitting, and threshold effect and saturation effect analysis. Results: A significant negative association between serum cholesterol levels and bone mineral density of the lumbar spine in US non cancer affected older adults aged 60 years or older. Older adults ≥ 70 years of age had an inflection point at 280 mg / dl, and those with moderate physical activity had an inflection point at 199 mg / dl, The smooth curves they fitted were all U-shaped. Conclusions: *There is* a negative association between total cholesterol and lumbar spine bone mineral density in non-cancer elderly greater than or equal to 60 years of age. ## INTRODUCTION Makovey and Chen [1] studied the relationship of cholesterol and bone mineral density in women before and after menopause with a modest inverse correlation. And Yang, Liu [2] found an inverse correlation between bone mineral density and serum cholesterol levels in type 2 diabetes. Ackert-Bicknell [3] found an inverse correlation between serum HDL levels and bone status in vitamin D-deficient postmenopausal women. However, these are small population-specific studies. A study by Tang et al. [ 4] observed a negative correlation between high-density lipoprotein cholesterol (HDL-C) levels and BMD. In contrast, Zolfaroli et al. [ 5] concluded that HDL-C levels were positively correlated with BMD in the lumbar spine and femoral neck in a postmenopausal female population. In addition to this, Cui et al. [ 6] concluded that HDL-C levels in premenopausal and postmenopausal subjects were not associated with BMD at any site. The conclusions of these studies remain controversial. And, what is the relationship between serum total cholesterol and bone mineral density in normal people without diabetes and cancer? Among the normal elderly without diabetes and cancer? In the large sample of normal elderly without cancer? *In a* large sample of normal older adults without cardiovascular disease? Hence, the relationship between cholesterol levels and BMD, and whether cholesterol levels have potential value in predicting osteoporosis is worth exploring. Therefore, our aim in this study was to evaluate the relationship between total cholesterol and bone mineral density using a representative sample of normal elderly without diseases such as cancer in the national health and Nutrition Examination Survey [7] (NHANES). NHANES is the cornerstone for national nutrition monitoring to inform nutrition and health policy. Nutritional assessment in NHANES is described with a focus on dietary data collection, analysis, and uses in nutrition monitoring [7]. ## Study population The National Health and Nutrition Examination Survey (NHANES, National Health and Nutrition Examination Survey) is a population-based cross-sectional survey designed to collect information about the United States. Information on the health and nutrition of the country’s family population. The project surveys a nationally representative sample each year, and these populations are located in counties across the country. NHANES interviews include demographics, socioeconomics, diet and health related issues. The physical examination part includes physiological measurements, laboratory examinations, etc. [ 8]. Our analysis is based on 1999-2006 data, which represents the three cycles of NHANES. The specific exclusion process is shown in the screening flowchart (Figure 1). In the end, a total of 3290 cancer-free participants over 60 years of age were included in our analysis. **Figure 1:** *Flow chart of sample selection from the NHANES.* ## Variables The exposure variable for this study was serum total cholesterol. In the mobile examination center (MEC) laboratory, blood specimens are processed, stored, and shipped to the Johns Hopkins University Lipoprotein Analytical Laboratory for analysis. The outcome variable was lumbar spine BMD, measured by DEXA. The following categorical variables were included as covariates in our analysis: age, sex, race / ethnicity, physical activity. Continuous covariates were included in our analysis: income poverty rate, blood urea nitrogen (BUN), total protein, serum uric acid, blood calcium and body mass index (BMI). Detailed information on total cholesterol, lumbar spine BMD, and covariates is available in http://www.cdc.gov/nchs/nhanes/ Available publicly. ## Statistical data analysis Data were analyzed with the use of the statistical packages R (The R Foundation; http://www.r-project.org; version 3.6.3) and EmpowerStats (https://www.empowerstats.net, X&Y solutions, Inc. Boston, MA). And P values < 0.05 were considered statistically significant. We performed weighted and variance estimation analyses to account for significant variance in our dataset. Weighted multiple logistic regression models were used to assess the association between total cholesterol and BMD of the lumbar spine. We calculated differences between groups using a weighted chi square test for categorical variables or a weighted linear regression model for continuous variables. Subgroup analysis was performed by stratified multiple regression analysis. In addition, smooth curve fitting and generalized additive models were used to address the nonlinear relationship between total cholesterol and lumbar spine BMD. For non-linearity in the model, a recursive algorithm was used to calculate the inflection point in the relationship of total cholesterol and BMD when non-linearity was detected, with a bi segmented linear regression model on either side of the inflection point. ## Data availability statement All data and materials were sourced from public databases. The datasets for this study can be found at https://www.cdc.gov/nchs/nhanes/. ## RESULTS A total of 3290 participants aged 60-85 were included in our analysis. The weighted characteristics of the participants were subdivided according to the tertiles of serum total cholesterol (low: ≥75 mg/dL to 191 mg/dL; medium: ≥ 192 mg/dL to <224 mg/dL; High: ≥225 mg/dL to <704 mg/dL). As shown in Table 1, there were significant differences in baseline characteristics between the tertiles of total cholesterol except for race / ethnicity. Compared with other subgroups, participants in the highest tertile of total cholesterol were more likely to be female, 60-69 years old, and sedentary. Participants in the top tertile of total cholesterol had lower income to poverty ratio, blood urea nitrogen, serum uric acid, and body mass index levels but higher total protein and serum calcium levels. **Table 1** | Total cholesterol (mg/dL) | Total | Low (≥75 to 191) | Middle (≥192 to <224) | High (≥225 to <704) | P-value | | --- | --- | --- | --- | --- | --- | | Sex (%) | | | | | <0.001 | | Male | 50.608 | 64.537 | 50.229 | 37.534 | | | Female | 49.392 | 35.463 | 49.771 | 62.466 | | | Age (%) | | | | | 0.034 | | 60-69 years | 52.219 | 49.259 | 52.521 | 54.781 | | | >= 70 years | 47.781 | 50.741 | 47.479 | 45.219 | | | Race/ethnicity (%) | | | | | 0.167 | | Non-Hispanic White | 76.869 | 74.537 | 79.560 | 76.497 | | | Non-Hispanic Black | 16.687 | 18.241 | 15.032 | 16.801 | | | Other Hispanic | 3.404 | 3.611 | 2.750 | 3.843 | | | Other races - Including multi-racial | 3.040 | 3.611 | 2.658 | 2.860 | | | Physical activity (%) | | | | | 0.074 | | Sedentary | 33.818 | 34.394 | 31.938 | 35.094 | | | Low | 25.962 | 26.938 | 24.440 | 26.509 | | | Moderate | 15.228 | 14.115 | 15.482 | 16.038 | | | High | 24.992 | 24.553 | 28.140 | 22.358 | | | Income to poverty ratio | 2.443 ± 1.497 | 2.489 ± 1.502 | 2.498 ± 1.520 | 2.345 ± 1.466 | 0.038 | | Blood urea nitrogen (mg/dL) | 16.410 ± 7.092 | 16.957 ± 8.111 | 16.091 ± 6.668 | 16.091 ± 6.668 | 0.008 | | Total protein (mg/dL) | 7.316 ± 0.516 | 7.251 ± 0.527 | 7.338 ± 0.511 | 7.359 ± 0.504 | <0.001 | | Serum uric acid (mg/dL) | 5.656 ± 1.460 | 5.787 ± 1.511 | 5.619 ± 1.439 | 5.565 ± 1.421 | 0.001 | | Serum calcium (mg/dL) | 9.469 ± 0.413 | 9.410 ± 0.418 | 9.460 ± 0.397 | 9.533 ± 0.415 | <0.001 | | Body mass index | 28.343 ± 5.459 | 28.711 ± 5.816 | 28.045 ± 5.156 | 28.280 ± 5.376 | 0.017 | | Lumber spine BMD (g/cm2) | 1.014 ± 0.195 | 1.044 ± 0.204 | 1.044 ± 0.204 | 0.982 ± 0.181 | <0.001 | The results of multiple regression equation analysis are shown in Table 2. In the unadjusted model, total cholesterol was negatively correlated with lumbar spine BMD (β=-0.0007, $95\%$ CI: -0.0008, -0.0005, $P \leq 0.0001$). After adjusting for confounding factors, this positive correlation still exists in Model 2 (β=-0.0004, $95\%$ CI: -0.0005, -0.0002, $P \leq 0.0001$) and Model 3(β=-0.0002, $95\%$ CI: -0.0004, -0.0001, $$P \leq 0.0034$$), P for trend is less than 0.001. We also stratified by continuous variables (third quantiles) and categorical variables; The results are shown in Table 2. **Table 2** | Outcome | Crude model | Crude model.1 | Model I | Model I.1 | Model II | Model II.1 | | --- | --- | --- | --- | --- | --- | --- | | Outcome | β (95%CI) | P-value | β (95%CI) | P-value | β (95%CI) | P-value | | Total cholesterol (10 mg/dL) | -0.007 (-0.008, -0.005) | <0.0001 | -0.004 (-0.005, -0.002) | <0.0001 | -0.002 (-0.004, -0.001) | 0.0034 | | Total cholesterol (tertiles) | Total cholesterol (tertiles) | | | | | | | Low | Reference | | Reference | | Reference | | | Middle | -0.237 (-0.392, -0.082) | 0.0027 | -0.047 (-0.195, 0.100) | 0.5293 | 0.012 (-0.143, 0.167) | 0.8818 | | High | -0.678 (-0.831, -0.525) | <0.0001 | -0.340 (-0.490, -0.191) | <0.0001 | -0.184 (-0.340, -0.028) | 0.0210 | | P for trend | <0.001 | | <0.001 | | 0.020 | | | Sex | | | | | | | | Male | -0.004 (-0.006, -0.002) | 0.0002 | -0.004 (-0.006, -0.002) | 0.0005 | -0.003 (-0.005, -0.001) | 0.0058 | | Female | -0.003 (-0.005, -0.001) | 0.0026 | -0.003 (-0.005, -0.001) | 0.0028 | -0.001 (-0.003, -0.001) | 0.3472 | | Age | | | | | | | | 60-69 years | -0.005 (-0.007, -0.003) | <0.0001 | -0.003 (-0.005, -0.001) | 0.0018 | -0.002 (-0.004, 0.000) | 0.1152 | | >= 70 years | -0.009 (-0.011, -0.007) | <0.0001 | -0.004 (-0.006, -0.002) | 0.0006 | -0.003 (-0.005, -0.001) | 0.0115 | | Race/ethnicity | | | | | | | | Non-Hispanic White | -0.007 (-0.009, -0.006) | <0.0001 | -0.004 (-0.006, -0.002) | <0.0001 | -0.003 (-0.005, -0.001) | 0.0007 | | Non-Hispanic Black | -0.003 (-0.007, -0.001) | 0.1439 | -0.000 (-0.005, 0.004) | 0.8174 | -0.000 (-0.004, 0.004) | 0.9697 | | Other Hispanic | -0.001 (-0.009, 0.007) | 0.7496 | -0.000 (-0.008, 0.008) | 0.9834 | 0.005 (-0.003, 0.014) | 0.2047 | | Other races - Including multi-racial | -0.003 (-0.012, 0.005) | 0.4869 | -0.002 (-0.010, 0.007) | 0.7036 | -0.001 (-0.010, 0.008) | 0.8562 | | Serum uric acid (tertiles) | Serum uric acid (tertiles) | | | | | | | Low | -0.007 (-0.010, -0.004) | <0.0001 | -0.004 (-0.007, -0.002) | 0.0004 | -0.003 (-0.006, -0.001) | 0.0150 | | Middle | -0.006 (-0.008, -0.003) | <0.0001 | -0.004 (-0.006, -0.001) | 0.0074 | -0.003 (-0.005, -0.000) | 0.0452 | | High | -0.006 (-0.008, -0.003) | <0.0001 | -0.003 (-0.005, -0.000) | 0.0250 | -0.001 (-0.003, 0.002) | 0.4791 | | Physical activity (tertiles) | | | | | | | | Sedentary | -0.007 (-0.010, -0.004) | <0.0001 | -0.004 (-0.007, -0.001) | 0.0025 | -0.003 (-0.005, -0.000) | 0.0729 | | Low | -0.006 (-0.009, -0.003) | <0.0001 | -0.003 (-0.006, -0.001) | 0.0125 | -0.001 (-0.003, 0.002) | 0.7098 | | Moderate | -0.010 (-0.015, -0.006) | <0.0001 | -0.006 (-0.010, -0.002) | 0.0080 | -0.004 (-0.008, -0.000) | 0.0769 | | High | -0.005 (-0.008, -0.002) | 0.0010 | -0.003 (-0.006, -0.000) | 0.0573 | -0.002 (-0.005, 0.001) | 0.1684 | | Income to poverty ratio (tertiles) | Income to poverty ratio (tertiles) | | | | | | | Low | -0.003 (-0.005, -0.001) | 0.0115 | -0.001 (-0.004, -0.001) | 0.2478 | -0.000 (-0.002, 0.002) | 0.8079 | | Middle | -0.007 (-0.010, -0.005) | <0.0001 | -0.004 (-0.007, -0.001) | 0.0032 | -0.003 (-0.005, -0.000) | 0.0642 | | High | -0.008 (-0.0011, -0.005) | <0.0001 | -0.005 (-0.007, -0.002) | 0.0017 | -0.004 (-0.007, -0.001) | 0.01591 | | Blood urea nitrogen (tertiles) | Blood urea nitrogen (tertiles) | | | | | | | Low | -0.005 (-0.008, -0.002) | 0.0003 | -0.003 (-0.005, 0.000) | 0.0610 | -0.002 (-0.005, 0.001) | 0.2595 | | Middle | -0.007 (-0.010, -0.004) | <0.0001 | -0.003 (-0.006, -0.001) | 0.0148 | -0.002 (-0.005, -0.001) | 0.1428 | | High | -0.008 (-0.010, -0.005) | <0.0001 | -0.005 (-0.007, -0.002) | <0.0001 | -0.003 (-0.005, -0.000) | 0.0181 | | Total protein (tertiles) | | | | | | | | Low | -0.006 (-0.009, -0.003) | <0.0001 | -0.003 (-0.006, -0.001) | 0.0081 | -0.002 (-0.004, 0.001) | 0.1597 | | Middle | -0.006 (-0.008, -0.003) | <0.0001 | -0.003 (-0.006, -0.000) | 0.0280 | -0.002 (-0.005, -0.000) | 0.1077 | | High | -0.008 (-0.010, -0.005) | <0.0001 | -0.004 (-0.006, -0.001) | 0.0035 | -0.002 (-0.005, -0.000) | 0.0539 | | Serum calcium (tertiles) | | | | | | | | Low | -0.008 (-0.011, -0.005) | <0.0001 | -0.005 (-0.008, -0.002) | 0.0011 | -0.005 (-0.008, -0.002) | 0.0033 | | Middle | -0.005 (-0.007, -0.002) | 0.0003 | -0.002 (-0.004, 0.001) | 0.2133 | -0.001 (-0.003, 0.002) | 0.4794 | | High | -0.007 (-0.009, -0.004) | <0.0001 | -0.004 (-0.006, -0.002) | 0.0004 | -0.003 (-0.005, -0.000) | 0.0390 | | BMI (tertiles) | | | | | | | | Low | -0.006 (-0.009, -0.004) | <0.0001 | -0.003 (-0.006, -0.001) | 0.072 | -0.004 (-0.006, -0.001) | 0.0053 | | Middle | -0.005 (-0.008, -0.003) | <0.0001 | -0.002 (-0.004, 0.000) | 0.904 | -0.000 (-0.003, 0.002) | 0.8110 | | High | -0.007 (-0.010, -0.005) | <0.0001 | -0.004 (-0.006, -0.001) | 0.016 | -0.004 (-0.007, -0.002) | 0.0007 | Table 3 is a stratified analysis of total cholesterol versus lumbar spine bone mineral density. Except for Non-Hispanic Black and Other Hispanic in race/ethnicity, with the lowest tertile of total cholesterol as the reference, the other items of β in the highest tertile of total cholesterol were negative, and the P was less than 0.01, and the difference was statistically significant. **Table 3** | Unnamed: 0 | Total cholesterol (tertiles) | Total cholesterol (tertiles).1 | Total cholesterol (tertiles).2 | | --- | --- | --- | --- | | | Low | Middle β (95%CI) P value | High β (95%CI) P value | | Age | | | | | 60-69 years | Reference | -0.01 (-0.03, 0.01) 0.2533 | -0.05 (-0.07, -0.03) <0.0001 | | >= 70 years | Reference | -0.04 (-0.06, -0.02) 0.0009 | -0.10 (-0.13, -0.08) <0.0001 | | Sex | | | | | Male | Reference | -0.00 (-0.02, 0.02) 0.9288 | -0.01 (-0.03, 0.01) 0.3765 | | Female | Reference | -0.01 (-0.03, 0.01) 0.3765 | -0.03 (-0.05, -0.01) 0.0021 | | Race ethnicity | | | | | Non-Hispanic White | Reference | -0.02 (-0.04, -0.01) 0.0059 | -0.08 (-0.10, -0.06) <0.0001 | | Non-Hispanic Black | Reference | -0.03 (-0.07, 0.02) 0.2688 | -0.02 (-0.06, 0.03) 0.4768 | | Other Hispanic | Reference | -0.00 (-0.08, 0.08) 0.9424 | 0.02 (-0.05, 0.10) 0.5491 | | Other races - Including multi-racial | Reference | -0.05 (-0.14, 0.04) 0.2654 | -0.07 (-0.15, 0.01) 0.0969 | | Physical activity | | | | | Sedentary | Reference | -0.04 (-0.07, -0.01) 0.0096 | -0.07 (-0.10, -0.05) <0.0001 | | Low | Reference | -0.07 (-0.10, -0.03) <0.0001 | -0.08 (-0.11, -0.05) <0.0001 | | Moderate | Reference | -0.02 (-0.06, 0.03) 0.4257 | -0.09 (-0.13, -0.04) <0.0001 | | High | Reference | 0.01 (-0.02, 0.04) 0.6042 | -0.04 (-0.07, -0.01) 0.0139 | | Income to poverty ratio (tertiles) | Income to poverty ratio (tertiles) | Income to poverty ratio (tertiles) | | | 0-1.37 | Reference | -0.03 (-0.06, -0.00) 0.0432 | -0.04 (-0.07, -0.02) 0.0024 | | 1.38-2.91 | Reference | -0.03 (-0.06, 0.00) 0.0799 | -0.07 (-0.10, -0.04) <0.0001 | | 2.93-5.00 | Reference | -0.03 (-0.05, -0.00) 0.0475 | -0.07 (-0.10, -0.04) <0.0001 | | Blood urea nitrogen (tertiles, mg/dL) | Blood urea nitrogen (tertiles, mg/dL) | Blood urea nitrogen (tertiles, mg/dL) | | | 2.00 - 12.00 | Reference | -0.04 (-0.06, -0.01) 0.0147 | -0.05 (-0.08, -0.02) 0.0003 | | 13.00 - 16.00 | Reference | -0.01 (-0.03, 0.02) 0.6859 | -0.06 (-0.09, -0.03) <0.0001 | | 17.00 - 98.00 | Reference | -0.03 (-0.06, -0.01) 0.0120 | -0.09 (-0.11, -0.06) <0.0001 | | Total protein (tertiles, mg/dL) | Total protein (tertiles, mg/dL) | | | | 5.40 - 7.00 | Reference | -0.03 (-0.05, 0.00) 0.0764 | -0.07 (-0.10, -0.04) <0.0001 | | 7.10 - 7.40 | Reference | -0.03 (-0.05, 0.00) 0.0509 | -0.05 (-0.07, -0.02) 0.0002 | | 7.50 - 11.00 | Reference | -0.02 (-0.05, 0.01) 0.1544 | -0.08 (-0.11, -0.05) <0.0001 | | Serum uric acid (tertiles, mg/dL) | Serum uric acid (tertiles, mg/dL) | | | | 1.50 - 4.80 | Reference | -0.00 (-0.03, 0.02) 0.7715 | -0.07 (-0.09, -0.04) <0.0001 | | 4.90 - 6.00 | Reference | -0.01 (-0.03, 0.02) 0.6510 | -0.04 (-0.07, -0.01) 0.0051 | | 6.10 - 13.70 | Reference | -0.04 (-0.07, -0.02) 0.0016 | -0.07 (-0.10, -0.05) <0.0001 | | Serum calcium (tertiles, mg/dL) | Serum calcium (tertiles, mg/dL) | | | | 6.70 - 9.20 | Reference | -0.05 (-0.08, -0.02) 0.0014 | -0.07 (-0.10, -0.04) <0.0001 | | 9.30 - 9.50 | Reference | 0.00 (-0.03, 0.03) 0.9714 | -0.07 (-0.09, -0.04) <0.0001 | | 9.60 – 11.30 | Reference | -0.02 (-0.05, 0.00) 0.0913 | -0.06 (-0.08, -0.03) <0.0001 | | Body mass index (tertiles) | Body mass index (tertiles) | | | | 15.18 - 25.72 | Reference | -0.00 (-0.03, 0.02) 0.7332 | -0.06 (-0.09, -0.03) <0.0001 | | 25.73 – 29.86 | Reference | -0.00 (-0.03, 0.02) 0.6066 | -0.06 (-0.09, -0.04) <0.0001 | | 29.87 – 57.31 | Reference | -0.03 (-0.06, -0.01) 0.0111 | -0.06 (-0.09, -0.04) <0.0001 | We performed a univariate analysis of lumbar bone mineral density (Supplementary Table 1). Compared with men, women’s lumbar spine BMD is reduced by 0.119, which is a relatively large difference, and the difference is statistically significant (P value less than 0.0001). Other results are shown in Supplementary Table 1. We also performed weighted generalized additive models and smoothing curve fitting to assess their association (Figure 2). Figure 2 shows that total cholesterol is linearly negatively correlated with lumbar bone mineral density. We also performed smooth curve fitting in subgroups stratified by categorical variables (Figure 3). Figure 3A shows that there is a turning point in the fitting curve for elderly people older than or equal to 70 years old, at a total cholesterol of 280 mg/dL (Supplementary Table 2). Figure 3D shows that the fitting curve of older people with more than moderate physical activity has a turning point, at a total cholesterol of 199 mg/dL (Supplementary Table 2). Other turning points are shown in Figure 4 and Supplementary Table 2. **Figure 2:** *The association between serum total cholesterol and lumbar bone mineral density. (A) Each black dot represents a sample. (B, C) The solid arc line represents the smooth curve fit between the variables. The blue bar represents the 95% confidence interval of the fit. Adjustments were made for age, gender, race/ethnicity, physical activity, income poverty rate, blood urea nitrogen, serum urea, total protein, blood phosphorus, and blood calcium.* **Figure 3:** *The association between serum total cholesterol and lumbar spine BMD stratified by different categorical variables. (A) Stratified by age. Sex, race / ethnicity, physical activity, income poverty rate, blood urea nitrogen, total protein, serum uric acid, blood calcium and body mass index were adjusted. (B) Stratified by sex. Age, race / ethnicity, physical activity, income poverty rate, blood urea nitrogen, total protein, serum uric acid, blood calcium and body mass index were adjusted. (C) Stratified by race / ethnicity. Age, sex, physical activity, income poverty rate, blood urea nitrogen, total protein, serum uric acid, blood calcium and body mass index were adjusted. (D) Stratified by physical activity. Age, sex, race / ethnicity, income poverty rate, blood urea nitrogen, total protein, total protein, serum uric acid, blood calcium and body mass index were adjusted.* **Figure 4:** *The association between serum total cholesterol and lumbar spine BMD stratified by tertiles of different continuous variables. (A) Stratified by income poverty rate tertiles. Age, Sex, race / ethnicity, physical activity, blood urea nitrogen, total protein, serum uric acid, blood calcium and body mass index were adjusted. (B) Stratified by blood urea nitrogen tertiles. Age, Sex, race / ethnicity, physical activity, income poverty rate, total protein, serum uric acid, blood calcium and body mass index were adjusted. (C) Stratified by total protein tertiles. Age, Sex, race / ethnicity, physical activity, income poverty rate, blood urea nitrogen, serum uric acid, blood calcium and body mass index were adjusted. (D) Stratified by serum uric acid tertiles. Age, Sex, race / ethnicity, physical activity, income poverty rate, blood urea nitrogen, total protein, blood calcium and body mass index were adjusted. (E) Stratified by blood calcium tertiles. Age, Sex, race / ethnicity, physical activity, income poverty rate, blood urea nitrogen, total protein, serum uric acid and body mass index were adjusted. (F) Stratified by body mass index tertiles. Age, Sex, race / ethnicity, physical activity, income poverty rate, blood urea nitrogen, total protein, serum uric acid and blood calcium were adjusted.* ## DISCUSSION Samelson, Cupples [9] believes that the cholesterol levels of women and men from young adult to middle-aged do not seem to have long-term clinical significance for later osteoporosis. However, Zolfaroli, Ortiz [5]’s research suggests that there is a positive correlation between cholesterol and bone mineral density in the lumbar spine and femoral neck in postmenopausal women. Some researchers also found a moderate negative correlation between BMD and serum cholesterol levels [1]. Therefore, the relationship between bone mineral density and serum total cholesterol level is complex and unclear. The primary objective of this study was to investigate whether total cholesterol was independently associated with lumbar spine BMD. In this study, we used a nationally representative sample of older Americans without cancer ($$n = 3290$$). Our results suggest a significant inverse association between serum cholesterol levels and bone mineral density of the lumbar spine in US non cancer affected older adults aged 60 years or older. Older adults ≥ 70 years of age had an inflection point at 280 mg / dl, log likelihood ratio tests was 0.002. And those with moderate physical activity had an inflection point at 199 mg / dl, the p value from the log likelihood ratio test was 0.005. There was an inflection point at 277 mg / dl for older adults in the highest tertile of serum uric acid (the p value from the log likelihood ratio test was 0.012) and at 275 mg / dl for older adults in the middle tertile of serum calcium (the p value from the log likelihood ratio test was 0.023). The smooth curves they fitted were all U-shaped. Their fitted smooth curves were all U-shaped, all are statistically significant. The exact mechanism between total cholesterol and bone metabolism is unclear, the correlation between total cholesterol and bone mineral density remains controversial [10–12]. There may be the following reasons. Bones are highly innervated and vascularized. The seemingly closed system of the skeleton, tightly linked to systemic metabolic homeostasis, is dynamically regulated by hormones and nutrients. A large number of epidemiological studies have demonstrated a positive association between the risk of cardiovascular disease and osteoporosis [13–16]. Bone metabolism is a continuous cycle of bone formation and resorption, which is coordinated by osteoblasts, osteocytes, and osteoclasts. Free cholesterol may inhibit BMP2, thereby blocking Runx2, Alpl, and COL1A1 expression in osteoblasts and subsequently inhibiting osteoblast differentiation [12, 17]. Cholesterol and its metabolites influence the bone homeostasis through modulating the differentiation and activation of osteoblasts and osteoclasts [12]. It has been shown that inhibition of cholesterol biosynthesis inhibits mRNA expression in the precursor stromal bone marrow cells of osteoblasts, thereby preventing osteogenic differentiation and achieving improved BMD [10, 18, 19]. Treatment of rodents with the cholesterol lowering drugs simvastatin and lovastatin both increases bone formation [20]. In an animal study, a high cholesterol diet showed a decrease in femoral BMD accompanied by an increase in serum OCN and carboxy terminal collagen cross-linking (CTX), suggesting that high cholesterol may increase bone turnover [17, 20]. High fat diet also induced cathepsin K-Positive osteoclasts and RANKL expression, leading to enhanced osteoclastogenesis, resulting in decreased bone mineral density [21]. Mazidi et al. [ 22] found that cholesterol levels increased indicators of inflammation in humans, and, in a study by Wang et al. found that inflammatory factors decreased BMD by affecting osteoclast activation or function [23], therefore, high cholesterol levels may affect BMD by activating the inflammatory response. In this study, bone mineral density levels in older adults were the outcome variable and serum cholesterol levels were the exposure variable. After adjusting for more than a dozen variables such as age, sex, and race, we found that high serum cholesterol levels were an independent risk factor for reduced BMD levels in older adults. In this study, we analyzed a representative sample of a multi-ethnic population to better generalize the American population. This large sample size allows us to perform further subgroup analysis. This is the main advantage of this research. Our findings suggest that older adults should be told that cholesterol is controlled at an appropriate level and that older patients with hypercholesterolemia have reduced bone mineral density. Our study still has some deficiencies. First, other confounding factors not included in this study may have had an impact on the results. Second, because this study used a cross-sectional design, it is difficult to determine whether there is a causal relationship between total cholesterol and lumbar spine BMD. Therefore, the role of total cholesterol in bone metabolism requires further studies with large samples. ## CONCLUSIONS There is a negative association between total cholesterol and lumbar spine BMD in non-cancer elderly greater than or equal to 60 years of age. ## References 1. Makovey J, Chen JS, Hayward C, Williams FM, Sambrook PN. **Association between serum cholesterol and bone mineral density.**. *Bone* (2009) **44** 208-13. DOI: 10.1016/j.bone.2008.09.020 2. Yang Y, Liu G, Zhang Y, Xu G, Yi X, Liang J, Zhao C, Liang J, Ma C, Ye Y, Yu M, Qu X. **Association Between Bone Mineral Density, Bone Turnover Markers, and Serum Cholesterol Levels in Type 2 Diabetes.**. *Front Endocrinol (Lausanne)* (2018) **9** 646. DOI: 10.3389/fendo.2018.00646 3. Ackert-Bicknell CL. **HDL cholesterol and bone mineral density: is there a genetic link?**. *Bone* (2012) **50** 525-33. DOI: 10.1016/j.bone.2011.07.002 4. Tang Y, Wang S, Yi Q, Xia Y, Geng B. **High-density Lipoprotein Cholesterol Is Negatively Correlated with Bone Mineral Density and Has Potential Predictive Value for Bone Loss.**. *Lipids Health Dis* (2021) **20** 75. DOI: 10.1186/s12944-021-01497-7 5. Zolfaroli I, Ortiz E, García-Pérez MÁ, Hidalgo-Mora JJ, Tarín JJ, Cano A. **Positive association of high-density lipoprotein cholesterol with lumbar and femoral neck bone mineral density in postmenopausal women.**. *Maturitas* (2021) **147** 41-6. DOI: 10.1016/j.maturitas.2021.03.001 6. Cui LH, Shin MH, Chung EK, Lee YH, Kweon SS, Park KS, Choi JS. **Association between bone mineral densities and serum lipid profiles of pre- and post-menopausal rural women in South Korea.**. *Osteoporos Int* (2005) **16** 1975-81. DOI: 10.1007/s00198-005-1977-2 7. Ahluwalia N, Dwyer J, Terry A, Moshfegh A, Johnson C. **Update on NHANES Dietary Data: Focus on Collection, Release, Analytical Considerations, and Uses to Inform Public Policy.**. *Adv Nutr* (2016) **7** 121-34. DOI: 10.3945/an.115.009258 8. Curtin LR, Mohadjer LK, Dohrmann SM, Montaquila JM, Kruszan-Moran D, Mirel LB, Carroll MD, Hirsch R, Schober S, Johnson CL. **The National Health and Nutrition Examination Survey: Sample Design, 1999-2006.**. *Vital Health Stat 2* (2012) 1-39. PMID: 22788053 9. Samelson EJ, Cupples LA, Hannan MT, Wilson PW, Williams SA, Vaccarino V, Zhang Y, Kiel DP. **Long-term effects of serum cholesterol on bone mineral density in women and men: the Framingham Osteoporosis Study.**. *Bone* (2004) **34** 557-61. DOI: 10.1016/j.bone.2003.11.024 10. Zhang Q, Zhou J, Wang Q, Lu C, Xu Y, Cao H, Xie X, Wu X, Li J, Chen D. **Association Between Bone Mineral Density and Lipid Profile in Chinese Women.**. *Clin Interv Aging* (2020) **15** 1649-64. DOI: 10.2147/CIA.S266722 11. Li GH, Cheung CL, Au PC, Tan KC, Wong IC, Sham PC. **Positive effects of low LDL-C and statins on bone mineral density: an integrated epidemiological observation analysis and Mendelian randomization study.**. *Int J Epidemiol* (2020) **49** 1221-35. DOI: 10.1093/ije/dyz145 12. Yin W, Li Z, Zhang W. **Modulation of Bone and Marrow Niche by Cholesterol.**. *Nutrients* (2019) **11** 1394. DOI: 10.3390/nu11061394 13. Szekanecz Z, Raterman HG, Pethő Z, Lems WF. **Common mechanisms and holistic care in atherosclerosis and osteoporosis.**. *Arthritis Res Ther* (2019) **21** 15. DOI: 10.1186/s13075-018-1805-7 14. Gaudio A, Xourafa A, Rapisarda R, Castellino P, Signorelli SS. **Peripheral artery disease and osteoporosis: Not only age-related (Review).**. *Mol Med Rep* (2018) **18** 4787-92. DOI: 10.3892/mmr.2018.9512 15. Sennerby U, Farahmand B, Ahlbom A, Ljunghall S, Michaëlsson K. **Cardiovascular diseases and future risk of hip fracture in women.**. *Osteoporos Int* (2007) **18** 1355-62. DOI: 10.1007/s00198-007-0386-0 16. Marcovitz PA, Tran HH, Franklin BA, O’Neill WW, Yerkey M, Boura J, Kleerekoper M, Dickinson CZ. **Usefulness of bone mineral density to predict significant coronary artery disease.**. *Am J Cardiol* (2005) **96** 1059-63. DOI: 10.1016/j.amjcard.2005.06.034 17. You L, Sheng ZY, Tang CL, Chen L, Pan L, Chen JY. **High cholesterol diet increases osteoporosis risk via inhibiting bone formation in rats.**. *Acta Pharmacol Sin* (2011) **32** 1498-504. DOI: 10.1038/aps.2011.135 18. Parhami F, Tintut Y, Beamer WG, Gharavi N, Goodman W, Demer LL. **Atherogenic high-fat diet reduces bone mineralization in mice.**. *J Bone Miner Res* (2001) **16** 182-8. DOI: 10.1359/jbmr.2001.16.1.182 19. Parhami F, Jackson SM, Tintut Y, Le V, Balucan JP, Territo M, Demer LL. **Atherogenic diet and minimally oxidized low density lipoprotein inhibit osteogenic and promote adipogenic differentiation of marrow stromal cells.**. *J Bone Miner Res* (1999) **14** 2067-78. DOI: 10.1359/jbmr.1999.14.12.2067 20. Mandal CC. **High Cholesterol Deteriorates Bone Health: New Insights into Molecular Mechanisms.**. *Front Endocrinol (Lausanne)* (2015) **6** 165. DOI: 10.3389/fendo.2015.00165 21. Prieto-Potín I, Roman-Blas JA, Martínez-Calatrava MJ, Gómez R, Largo R, Herrero-Beaumont G. **Hypercholesterolemia boosts joint destruction in chronic arthritis. An experimental model aggravated by foam macrophage infiltration.**. *Arthritis Res Ther* (2013) **15** R81. DOI: 10.1186/ar4261 22. Mazidi M, Mikhailidis DP, Banach M. **Associations between risk of overall mortality, cause-specific mortality and level of inflammatory factors with extremely low and high high-density lipoprotein cholesterol levels among American adults.**. *Int J Cardiol* (2019) **276** 242-7. DOI: 10.1016/j.ijcard.2018.11.095 23. Wang T, He C. **TNF-α and IL-6: The Link between Immune and Bone System.**. *Curr Drug Targets* (2020) **21** 213-27. DOI: 10.2174/1389450120666190821161259
--- title: A novel risk score model based on pyroptosis-related genes for predicting survival and immunogenic landscape in hepatocellular carcinoma authors: - Hongyu Wang - Bo Zhang - Yanan Shang - Fei Chen - Yumei Fan - Ke Tan journal: Aging (Albany NY) year: 2023 pmcid: PMC10042690 doi: 10.18632/aging.204544 license: CC BY 3.0 --- # A novel risk score model based on pyroptosis-related genes for predicting survival and immunogenic landscape in hepatocellular carcinoma ## Abstract Background: *Hepatocellular carcinoma* (HCC) is the third leading cause of cancer worldwide, with high incidence and mortality. Pyroptosis, a form of inflammatory-regulated cell death, is closely associated with oncogenesis. Methods: Expression profiles of HCC were downloaded from the TCGA database and validated using the ICGC and GEO databases. Consensus clustering analysis was used to determine distinct clusters. The pyroptosis-related genes (PRGs) included in the pyroptosis-related signature were selected by univariate Cox regression and LASSO regression analysis. Kaplan-Meier and receiver operating characteristic (ROC) analyses were performed to estimate the prognostic potential of the model. The characteristics of infiltration of immune cells between different groups of HCC were explored. Results: Two independent clusters were identified according to PRG expression. Cluster 2 showed upregulated expression, poor prognosis, increased immune cell infiltration and worse immunotherapy response than cluster 1. A prognostic risk signature consisting of five genes (GSDME, NOD1, PLCG1, NLRP6 and NLRC4) was identified. In the high-risk score group, HCC patients showed decreased survival rates. In particular, multiple clinicopathological characteristics and immune cell infiltration were significantly associated with the risk score. Notably, the 5 PRGs in the risk score have been implicated in carcinogenesis, immunological pathways and drug sensitivity. Conclusions: A prognostic signature comprising five PRGs can be used as a potential prognostic factor for HCC. The PRG-related signature provides an in-depth understanding of the association between pyroptosis and chemotherapy or immunotherapy for HCC patients. ## INTRODUCTION Liver cancer ranks as the third leading cause of cancer-associated death according to GLOBOCAN 2020 [1]. Hepatocellular carcinoma (HCC), which has seriously affected human health, is the major primary liver cancer [2, 3]. In the past decade, despite great progress in surgery and various treatments, such as radiotherapy, chemotherapy, transarterial chemoembolization (TACE), molecular targeted therapy and minimally invasive surgery, the overall 5-year survival rate is only $18\%$, and the long-term prognosis of HCC patients still needs to be improved [4–6]. During the process of tumor initiation, development and metastasis, cancer cells gradually form an adaptive tumor immune microenvironment and begin to avoid programmed death and escape immunity. Pyroptosis, a newly identified type of cell death triggered by inflammation, exhibits morphological characteristics of both necrosis and apoptosis [7, 8]. Under physiological conditions, pyroptosis defends against pathogen or bacterial infections. However, excessive pyroptosis tends to lead to sustained amplified inflammatory responses that are involved in various human diseases, such as infectious diseases, cardiovascular diseases, atherosclerosis, diabetic kidney disease, renal ischemia-reperfusion injury and neurodegenerative diseases [9, 10]. Pyroptosis provides new therapeutic strategies for human diseases. More importantly, previous studies have elucidated that pyroptosis is of great significance to tumor progression, and its anti-cancer effects have gradually attracted worldwide attention [11]. Morphologically, the main characteristics of pyroptotic cells include bubble-like protrusions, cellular swelling, and the formation of membrane pores by the gasdermin (GSDM) protein family [12]. The formation of GSDM pores on the plasma membrane eventually leads to cell lysis, releasing many damage-associated molecular patterns (DAMPs), such as ATP, interleukin-1 beta (IL-1β), S100 family proteins, heat shock proteins (HSPs) and high mobility group box protein 1 (HMGB1) [12, 13]. The occurrence of pyroptosis leads to a strong inflammatory response in the body, which then affects the tumor immune microenvironment [14, 15]. Nucleotide-binding domain and leucine-rich repeat-containing receptors (NLRs) and the GSDM family play essential roles in pyroptosis signaling pathways. Noncanonical pathways triggered by caspase 11 in mice and caspase $\frac{4}{5}$ in humans and canonical pathways triggered by caspase-1 are generally two modes of pyroptosis [16]. In the canonical pathway, inflammasomes play a role in recruiting apoptosis-associated speck-like protein containing a caspase recruitment domain (ASC) to activate caspase-1, leading to cytokine secretion and GSDMD cleavage [17]. The N-terminus of GSDMD forms pores on the membrane to cause the release of inflammatory factors and cell lysis [13]. Recent studies suggested that caspase-3 could be activated by some stimuli to promote the cleavage of GSDME, leading to pore formation [18]. NLRs, a family of proteins that play a key role in host defense, not only recognize conserved pathogen-associated molecular patterns (PAMPs) but also identify DAMPs [19]. NLRs can induce inflammasome formation [20]. The inflammasome can process signals to trigger a cascade of inflammatory responses. Thus, there are significant associations between NLRs and multiple human diseases related to infection and immunity [21]. NLRs exhibit diverse molecular functions under both physiological and pathological conditions, such as inflammasome assembly, signal transduction, transcription activation and autophagy [22]. Since these novel links between pyroptosis and human diseases may improve our understanding of the pathogenesis of diseases and promote the development of new ways to prevent and treat these diseases, pyroptosis is also receiving widespread attention from clinicians [9, 10]. Recently, numerous studies have demonstrated that inflammasome-regulated pyroptosis is closely interlinked with the pathogenesis of cancer [23]. For example, NLRP6 expression was decreased in gastric cancer and obviously associated with *Helicobacter pylori* infection, lymph node metastasis, tumor stage and survival rate [24]. Overexpression of NLRP6 reduced cell growth, decreased invasion and migration, and promoted cell apoptosis in gastric cancer cells [24]. Moreover, decreased level of NLRP6 was correlated with unfavorable prognosis in patients with head and neck squamous cell carcinoma (NHSCC), revealing the tumor suppressive role of NLRP6 in gastric cancer and NHSCC [25]. In addition, the protein level of NLRC4 was upregulated and linked with unfavorable prognosis in glioma patients, demonstrating that NLRC4 is a diagnostic biomarker and potential therapeutic target for glioma [26]. Furthermore, loss of NLRC4 impeded colon cancer liver metastasis accompanied by reduced infiltration level of M2 macrophages and IL-1β expression in mice with high-fat diet-triggered nonalcoholic fatty liver disease (NAFLD) [27]. The protein levels of GSDMD were markedly upregulated in non-small cell lung cancer (NSCLC), and upregulated GSDMD was markedly correlated with invasive characteristics and worse prognosis [19, 28]. GSDME protein levels were increased in esophageal squamous cell carcinoma (ESCC) and positively corresponded to a favorable prognosis [29]. Cotreatment with the PLK1 inhibitor BI2536 and cisplatin triggered caspase-3/GSDME axis-dependent pyroptosis in ESCC cells [29]. The high expression of GSDME in tumors can effectively promote the infiltration of different immune cells, and correspondingly, the immune cell infiltration and immune response in GSDME-deficient tumors tend to decrease. This GSDME-dependent pyroptosis, a novel nonapoptotic mechanism of eliminating cancer cells, is downstream of the activated mitochondria-mediated caspase pathway [18, 30]. Nevertheless, the association between pyroptosis-related genes (PRGs) and immunity in HCC remains unclear, and it is vital to construct a new prognostic model of PRGs. In this study, we performed a comprehensive systematic analysis of PRGs in HCC using TCGA, ICGC and GEO databases. Two independent HCC clusters established by consensus clustering analysis were shown to have different immune cell infiltration and prognostic survival. To further assess the effects of the PRGs in HCC, a five-PRG risk model, including GSDME, NOD1, PLCG1, NLRP6 and NLRC4, was identified to be greatly linked with the overall survival (OS) of HCC patients. We also determined the significance of the signature by exploring the associations between the risk score and immune cell infiltration, clinical features, drug sensitivity and immunotherapy response in HCC patients. These results provide an in-depth understanding of the prognostic power of PRGs and provide immunotherapy strategies and treatments for HCC patients. ## Expression of pyroptosis-related genes (PRGs) in HCC in the TCGA database The flow chart depicting the analysis procedure for the present study is shown in Supplementary Figure 1. According to previous studies, we first analyzed the expression of thirty-three PRGs that were of great significance in modulating pyroptotic cell death in HCC. We evaluated the expression of these PRGs to investigate the functions of pyroptosis in HCC using the TCGA database. The expression of most PRGs, including CASP3, CASP4, CASP6, CASP8, CASP9, GPX4, GSDMA, GSDMB, GSDMC, GSDMD, GSDME, NLRP1, NLRP7, NOD1, NOD2, PJVK, PRKACA, PYCARD, PLCG1, SCAF11 and TIRAP, was significantly increased in HCC tissues compared with normal tissues (Figure 1A, 1B). Conversely, markedly decreased expression of AIM2, IL1B, IL6, NLRP3 and NLRP6 was observed in HCC tissues compared with normal tissues (Figure 1A, 1B). **Figure 1:** *Differential expression of 33 PRGs in HCC according to the TCGA database. (A) Heatmap of the differential expression of PRGs in HCC samples and normal samples. (B) Box diagram of the differential expression of PRGs in HCC samples and normal samples. *p < 0.05, **p < 0.01, ***p < 0.001.* To confirm the altered expression of the thirty-three PRGs, we also investigated the transcriptional level of these PRGs according to the ICGC database. The expression of AIM2, IL18, IL1B, IL6, NLRC4, NLRP2, NLRP3, NLRP6, NLRP7 and TNF was obviously downregulated in HCC tissues (Supplementary Figure 2A, 2B). However, the expression of CASP3, CASP6, CASP8, CASP9, GPX4, GSDMA, GSDMC, GSDMD, GSDME, NLRP1, NOD1, PLCG1, PYCARD and TIRAP was markedly elevated in HCC tissues compared with normal tissues (Supplementary Figure 2A, 2B). ## Construction of an interactive network of PRGs and signaling pathway analysis To further investigate the mechanisms of PRGs with differential expression in HCC, signaling pathway analysis was carried out using the Metascape database. Consistent with our speculation, these PRGs were strongly and positively associated with pyroptosis (Figure 2A). Additionally, these PRGs were also remarkably involved in various immunity-related pathways, including the nucleotide-binding oligomerization domain (NOD) pathway, NOD-like receptor signaling pathway, response to lipopolysaccharide, NOD$\frac{1}{2}$ signaling pathway, AIM2 inflammasome, negative regulation of cytokine production, regulation of cytokine-mediated signaling pathway, regulation of inflammatory response to antigenic stimulus, neutrophil-mediated immunity, and positive regulation of leukocyte differentiation (Figure 2A). Moreover, PPI networks were generated through the Metascape and STRING databases (Figure 2B, 2C). Additionally, the core interactions of PRGs are shown in Figure 2D. We then assessed the correlations among these PRGs. There were positive or negative correlations among the thirty-three PRGs according to the TCGA and ICGC databases (Figure 2E, 2F). **Figure 2:** *Functional analysis of 33 PRGs in HCC. (A) The enriched signaling pathways of 33 PRGs were obtained from the Metascape database. (B) A PPI network was constructed using the STRING database. (C) A gene-gene interactive network was constructed using the Metascape database. (D) The hub genes were selected from the PPI network using the Metascape database. (E, F) Heatmaps demonstrating the correlations among 33 PRGs with Spearman analysis in the TCGA and ICGC databases.* ## Consensus clustering analysis of PRGs in HCC Depending on the diverse expression levels of PRGs, we performed consensus clustering analysis. We identified $k = 2$ as the variable clustering stability, suitably dividing HCC patients into two subgroups (Figure 3A). Considering the transcriptome data of these 2 clusters, PCA was conducted (Figure 3B, 3C). The expression of these PRGs in the two clusters was further estimated (Figure 3D and Supplementary Figure 3). Most PRGs were highly expressed in cluster 2 (C2) compared with cluster 1 (C1) (Figure 3D and Supplementary Figure 3). Additionally, obvious differences in survival between the two clusters and worse OS were observed in HCC patients in C2 compared with those in C1 (Figure 3E). In addition, there were obvious differences in multiple clinicopathological parameters between the two clusters, including grade, T stage and TNM stage (Table 1). **Figure 3:** *Identification of distinct clusters of PRGs in HCC. (A, B) Two clusters were defined by consensus clustering analysis. (C) Cumulative distribution curves for $k = 2$-6. (D) Heatmap showing the expression pattern of PRGs in the two clusters. (E) KM analysis showed the OS for the two clusters of HCC patients.* TABLE_PLACEHOLDER:Table 1 ## Immune cell infiltration in two different clusters Because PRGs were closely associated with the immune response (Figure 2A), we then explored the relationship between different clusters and the tumor immune microenvironment. The results of the TIMER algorithm illuminated that the infiltration scores of six major immune cells, including CD4+ T cells, B cells, CD8+ T cells, macrophages, neutrophils and dendritic cells, in C1 were obviously lower than those in C2 (Figure 4A, 4B). The percentage abundance of infiltrated immune cells in each HCC patient is shown with different colors and immune cell types (Figure 4C). We also investigated the influence of different clusters on the expression levels of well-known important immune checkpoint genes and observed that the expressions of CD274, PDCD1, PDCD1LG2, CTLA4, LAG3, HAVCR2 and TIGIT were markedly downregulated in C1 compared with C2 (Figure 4D). More importantly, the TIDE score was lower in C1 than in C2, suggesting a better response to immunotherapy in C1 (Figure 4E). **Figure 4:** *Evaluation of immune cell infiltration abundance in different clusters of HCC samples by the TIMER algorithm. (A, B) Heatmap and box diagram showing the differential infiltration abundance of six types of immune cells in C1 and C2. (C) Bar plot demonstrating the composition of a great variety of immune cells in every HCC patient from C1 and C2 analyzed by the TIMER algorithm. (D) Box plots indicating the altered expression of immune checkpoint genes in C1 and C2. (E) Box plots showing the TIDE scores in the two clusters. ***p < 0.001.* Furthermore, the relationships between patient clusters and immune cell infiltration were also confirmed by the CIBERSORT algorithm. The infiltration abundances of activated CD4+ memory T cells, resting memory CD4+ T cells, regulatory T cells (Tregs), M0 macrophages, resting NK cells, activated mast cells, naïve B cells, memory B cells, neutrophils and resting mast cells in C1 and C2 were obvious different (Supplementary Figure 4A, 4B). The percentage abundance of infiltrated immune cells in each HCC patient is indicated with different colors and immune cell types according to the CIBERSORT algorithm (Supplementary Figure 4C). ## Molecular and functional enrichment analyses of the differences in two clusters of PRGs To further elucidate the molecular mechanism underlying the difference between C1 and C2, we then investigated the alteration of gene expression between these two clusters. As shown in Figure 5A, 5B, 486 genes were significantly upregulated and 6643 genes were downregulated in C1 compared with C2. Next, GO and KEGG analyses were performed to explore the different signaling pathways between C1 and C2 using upregulated or downregulated genes. The top 5 enriched KEGG pathways for upregulated genes were complement and coagulation cascades, metabolism of xenobiotics by cytochrome P450, drug metabolism-cytochrome P450, retinol metabolism and bile secretion (Figure 5C). The top 5 enriched pathways for upregulated genes were small molecule catabolic process, fatty acid metabolic process, carboxylic acid catabolic process, organic acid catabolic process and carboxylic acid biosynthetic process (Figure 5D). The top 5 enriched KEGG pathways for downregulated genes were endocytosis, Salmonella infection, human cytomegalovirus infection, human T-cell leukemia virus 1 infection and chemokine signaling pathway (Figure 5E). Additionally, the top 5 enriched GO terms for downregulated gene pathways were T-cell activation, covalent chromatin modification, regulation of cell–cell adhesion, actin filament organization, and positive regulation of cell adhesion (Figure 5F). These data imply that the difference between C1 and C2 is linked with metabolism- and immunity-associated signaling pathways. **Figure 5:** *(A) Volcano plot displaying the upregulated and downregulated genes in C2 compared with C1. (B) A clustering heatmap showing the changed expression of genes in two clusters after the deep filtration of genes with p < 0.05 and |log2 (fold change)|> 1.5 as thresholds. (C, D) KEGG and GO analyses were applied to explore the different signaling pathways for the upregulated genes. (E, F) KEGG and GO analyses were applied to explore the different signaling pathways for the downregulated genes.* ## Genetic mutation and drug sensitivity prediction of the two clusters in HCC We then generated the mutation profiles of HCC patients in C1 and C2 using the TCGA database. In C1, the top five genes with high mutation rates were CTNNB1, TTN, TP53, MUC16 and PCLO (Figure 6A). In contrast, TP53, TTN, MUC16, CSMD3 and PCLO were the most common mutation cohorts of genes altered in C2 (Figure 6B). In addition, missense was the primary type of mutation, and SNP was the major variant in both C1 and C2 (Figure 6A, 6B). The results of SNV class revealed that the most common type of the two risk groups was C > T (Figure 6A, 6B). **Figure 6:** *Mutational landscape and drug sensitivity of two clusters. (A) The landscape of mutation profiles in C1. (B) The landscape of mutation profiles in C2. Variant classification, variant types and SNV classification are shown. (C) Comparison of drug sensitivity in the two clusters. ***p < 0.001.* We also evaluated the difference in sensitivity to chemotherapeutic drugs in these two clusters. There was a significant difference in the IC50 values of sorafenib, sunitinib, paclitaxel, gefitinib, etoposide, 5-fluorouracil, docetaxel, doxorubicin, vinblastine and gemcitabine between the two clusters, suggesting that C1 was more resistant to these drugs (Figure 6C). ## Identification and construction of a PRG-related prognostic signature in the TCGA database We carried out univariate Cox regression analysis of 33 pyroptosis regulators to select PRGs with prognostic value. Seven hub genes, including CASP3, CASP4, GSDME, NLRC4, NLRP6, NOD1 and PLCG1, were selected with a cutoff of $p \leq 0.05$ (Figure 7A). Consistently, the altered expression of CASP3, CASP4, GSDME, NLRC4, NLRP6, NOD1 and PLCG1 also corresponded with favorable or unfavorable OS in HCC patients according to the KM analysis (Figure 7B). These seven candidate genes were considered as prognostic factors in HCC. **Figure 7:** *Construction of a five-PRG signature model in the TCGA-HCC cohort. (A) Forest plot of univariate Cox regression to select the genes with prognostic potential. (B) KM analysis revealed the prognostic value of CASP3, CASP4, GSDME, NLRC4, NLRP6, NOD1 and PLCG1 with the log-rank test. (C, D) A prognostic model containing 5 PRGs was built using LASSO Cox regression analysis. (E) The risk score and OS status of each case. (F) KM analysis for OS between the low-risk group and high-risk group. (G) The AUC of time-dependent ROC curves was shown.* LASSO Cox regression analysis based on the optimum λ value was then performed to build a prognostic model for the candidate PRGs to address collinearity. The risk score was calculated as follows: risk score = (0.1974 × GSDME) + (0.1926 × NLRC4) + (-0.1947 × NLRP6) + (0.0581 × NOD1) + (0.0272 × PLCG1) (Figure 7C, 7D). The risk score, survival outcome and 5 PRG gene expression of each HCC patient in TCGA database are vividly shown (Figure 7E). Of note, KM curve analysis results revealed that HCC patients in the high-risk group had unfavorable OS compared with those in the low-risk group (Figure 7F). ROC analysis was applied to verify the sensitivity and specificity of the prognostic model. The areas under the ROC curve (AUCs) were 0.699 for 1-year survival, 0.649 for 3-year survival and 0.66 for 5-year survival (Figure 7G). ## Confirmation of the PRG-related prognostic signature in the ICGC database A similar calculation was applied to the data from the ICGC database to verify the availability of the prognostic signature. The risk score was calculated as follows: risk score = (0.2485 × GSDME) + (-0.5484 × NLRC4) + (-0.9524 × NLRP6) + (0.1684 × NOD1) + (-0.3558 × PLCG1) (Supplementary Figure 5). We also divided the HCC patients into high-risk and low-risk subgroups in the ICGC database. HCC patients with high risk had a reduced survival time and a higher risk of mortality (Supplementary Figure 5A). The HCC patients in the high-risk subgroup had poor prognosis compared with those in the low-risk subgroup based on KM analysis, indicating good accuracy of this prognostic signature (Supplementary Figure 5B). The AUC values were 0.635 for 1-year survival, 0.687 for 2-year survival and 0.724 for 3-year survival (Supplementary Figure 5C). Collectively, these data suggested that the pyroptosis-related prognostic signature model could distinguish favorable prognoses in HCC patients. ## Independent prognostic potential of the PRG signature according to various clinicopathological parameters To further certify the prognostic value of the 5-PRG signature, the association between various clinicopathological parameters and risk score was explored. The high-risk score of the 5-PRG signature was obviously associated with worse OS in young (< 60 years), old (> 60 years), female, male, early stage (T1 + T2), advanced stage (T3 + T4), early grade (G1 and G2), advanced grade (G3), M0, N0, TNM stage I+II and TNM stage III HCC patients (Figure 8). Together, these data suggest that the 5-FRG signature can predict OS among each stratum of age, sex, stage and grade and further prove the good stratification ability of the 5-PRG prognostic model. **Figure 8:** *Prognostic potential of the risk score with different clinical parameters. KM analysis of OS between two subgroups stratified by age < 60, age > 60, male, female, T1 + T2, T3 + T4, G1, G2, G3, M0, N0, TNM I+II and TNM III with the log-rank test according to the TCGA database.* ## Univariate and multivariate Cox regression analyses and construction of the nomogram To further assess the prognostic value of the PRG-related prognostic signature in HCC patients, we performed univariate and multivariate Cox regression analyses (Figure 9A, 9B). Following univariate Cox regression analysis, GSDME, NLRC4, NLRP6, NOD1, PLCG1, T stage and M stage were clearly related to OS (Figure 9A). Following the results of multivariate Cox regression analysis, T stage and grade had obvious correlations with OS (Figure 9B). A nomogram model integrating T stage and grade was further constructed to predict the OS of HCC patients based on multivariate regression analysis (Figure 9C). The calibration plots of the nomogram illuminated good agreement between the nomogram-predicted and actual 1-, 3- and 5-year survival rates (Figure 9D). **Figure 9:** *Univariate and multivariate Cox regression analyses for risk score and construction of a nomogram. (A, B) Univariate Cox regression and multivariate Cox regression analyses of five PRGs and clinical features. (C) A nomogram containing the prognostic signature and different clinicopathological parameters was constructed. (D) Calibration curve of the actual 1-, 3-, and 5-year OS. (E) Association between the risk score and the infiltration abundances of six immune cells. (F) Heatmap depicting the correlations between the risk score and five PRGs and the infiltrated abundances of six types of immune cells. *p < 0.05, **p < 0.01.* ## Immune cell infiltration analysis of the risk model We then estimated the relationship between the immune cell infiltration and the risk score in HCC. The infiltrated levels of six major immune cell types were investigated utilizing the TIMER method. The risk score was strongly linked with the infiltrated levels of B cells, neutrophils, macrophages, CD4+ T cells, CD8+ T cells and dendritic cells (Figure 9E). In addition to the risk score, GSDME, NOD1, PLCG1 and NLRC4 were also significantly positively correlated with the infiltration abundances of these immune cells, whereas NLRP6 was negatively linked with the infiltrated abundances of B cells, CD4+ T cells and dendritic cells (Figure 9F). ## Expression of 5 hub PRGs The expression of the five-gene signature was obviously elevated in HCC tissues compared with normal tissues (Figure 10A). The expression of this signature was much higher in metastatic tissues (Figure 10A). Next, the transcriptional levels of these five hub genes were separately examined based on the HCCDB database. Increased expression of GSDME and PLCG1, and decreased expression of NLRC4 were found in HCC tissues in most GEO datasets (Supplementary Figure 6). Moreover, NOD1 expression was increased and NLRP6 expression was decreased in three different datasets (Supplementary Figure 6). **Figure 10:** *Expression of the prognostic signature and PRGs in HCC samples and normal liver samples. (A) The expression of the prognostic signature in HCC was examined using the TMNplot database. (B) The protein levels of GSDME, NLRC4 and PLCG1 were examined using the CPTAC database. (C) IHC analysis of the protein levels of GSDME, PLCG1, NLRC4 and NLRP6 using the HPA database. (D) Differential expression and distribution of GSDME, PLCG1, NLRC4 and NOD1 in HCC based on single-cell RNA-sequence analysis using the Human Liver Browser database. ***p < 0.001.* The protein expression level of these genes was examined according to the CPTAC database. The protein levels of GSDME and PLCG1 were higher, while the protein level of NLRC4 was lower in HCC than in normal tissues (Figure 10B). IHC results were obtained from the HPA database to further estimate the protein expression levels of GSDME, NLRC4, PLCG1 and NLRP6. The protein levels of GSDME and PLCG1 were upregulated in HCC, which was consistent with the CPTAC data. In contrast, the protein levels of NLRC4 and NLRP6 were downregulated in HCC compared with normal liver tissues (Figure 10C). We further examined these gene expressions using single-cell RNA-sequence data. Elevated expression levels of GSDME, NOD1 and PLCG1 in HCC tissues were observed (Figure 10D). Interestingly, these four PRGs were expressed not only in liver cancer cells but also in some immune cells, which may be one reason for the immune cell infiltration of the risk score (Figure 10D). ## Genetic mutation and drug sensitivity of 5 PRGs in HCC We then explored the genetic mutation profiles of these five PRGs using the cBioPortal online tool. The PLCG1 gene had the highest mutation frequency ($2\%$), followed by GSDME ($1.1\%$), NLRP6 ($1.1\%$), NLRC4 ($1.1\%$) and NOD1 ($0.3\%$) (Supplementary Figure 7A, 7B). The mutation was the primary type for these 5 genes (Supplementary Figure 7B). We also investigated the correlations between the five PRGs and several tumorigenesis-associated pathways, including the cell cycle, apoptosis, DNA damage response, EMT, hormone ER, hormone AR, RAS/MAPK, PI3K/AKT, RTK and TSC/mTOR pathways. The 5 hub PRGs were essentially linked with the inhibition or activation of these signaling pathways (Supplementary Figure 7C). We then evaluated whether the 5 PRGs affected the sensitivity of chemotherapy drugs using the GDSC database (Figure 11). According to the median expression of these 5 PRGs, HCC patients were separated into high-expression and low-expression groups. The IC50 values of all these chemotherapeutic drugs were significantly different between the high-expression group and the low-expression group (Figure 11). These data illustrate that HCC patients with increased expression of 5 PRGs are more sensitive to common chemotherapeutic agents. **Figure 11:** *Correlations between the 5 PRGs and drug sensitivity in HCC. (A–F) The relationships between the expression of 5 PRGs and drug sensitivity were explored based on the GDSC database through the pRRophetic package. *p < 0.05, **p < 0.01, ***p < 0.001.* ## DISCUSSION Liver cancer is the third leading cause of cancer-related death in the world, and its high incidence rate and mortality seriously threaten human health [1]. The HCC patients do not have obvious diagnostic symptoms at the early stage, so the opportunity for surgery may have been lost at the time of diagnosis, and the survival rate of HCC patients is still unsatisfactory [3]. Thus, it is urgent to deeply and systematically understand the molecular mechanism and open up new diagnostic strategies and treatment methods for HCC. Pyroptosis is a new form of regulated cell death (RCD) that causes cell membrane rupture and death via continuous cell expansion, resulting in the release of cell contents, which in turn activates a strong inflammatory response [11]. Compared with apoptosis, pyroptosis occurs faster and is accompanied by the release of a large number of proinflammatory factors, leading to the rapid death of cancer cells [31]. A growing number of studies have illuminated that pyroptosis-related molecules play a role in promoting tumor development and provide a new idea for the treatment of liver cancer [32–34]. In this study, two independent clusters were identified using consensus clustering analysis according to the expression levels of 33 PRGs. Importantly, PRGs in C2 had increased expression, and patients in C2 exhibited a worse prognosis than those in C1. Meanwhile, essential differences in terms of grade, T stage and TNM stage between C1 and C2 were observed (Table 1). C2 was enriched in immunity-related biological pathways and strongly correlated with prognosis and immune infiltration. In addition, to obtain a novel prognostic signature to predict OS, we selected pyroptosis regulators related to prognosis in HCC. Based on the prognostic potential of PRGs in HCC patients, we established and validated the risk prediction models of five PRGs (GSDME, NLRC4, NLRP6, NOD1 and PLCG1) and separated the HCC patients into a high-risk group and a low-risk group. The univariate and multivariate Cox regression analysis results illustrated that the established PRG risk model was an independent prognostic model for HCC patients. Combined with the established clinicopathological characteristics, ROC analysis proved that the risk model had more benefits in predicting the OS of patients with HCC. To expand our risk model, we further established a novel nomogram model to predict OS. At the same time, the actual OS is highly consistent with the model predictions as the results of the calibration curve. Chemotherapy and immunotherapy are important adjunct treatments for patients with HCC [4]. The development of new anticancer drugs is a time-consuming, high-investment and high-risk project. Often, the birth of new anticancer drugs requires several years or even decades of research, development and validation. Sorafenib, a multitargeted tumor drug, can selectively target the receptors of certain signaling pathways to facilitate apoptosis, suppress angiogenesis and inhibit cancer cell proliferation [35]. Sorafenib is an effective first-line therapy for late-stage HCC [36, 37]. Although sorafenib is less toxic and well tolerated, it still has some special adverse effects, which should be considered in clinical research and application. Moreover, according to clinical observations, the overall effective rate of treatment for liver cancer is relatively low. Additionally, sorafenib resistance is becoming more common. Fortunately, many other broad-spectrum anticancer drugs, including 5-fluorouracil, docetaxel, doxorubicin, etoposide, gefitinib, gemcitabine, paclitaxel, vinblastine and sunitinib, are also used as treatment strategies for liver cancer patients. In the current study, according to the GDSC database, the relationships between clusters and chemotherapeutic drug sensitivity were investigated. The sensitivity of the two clusters to common chemotherapeutic drugs was obviously different, and cluster 2 HCC patients may benefit from these drugs. In addition, the risk model containing five PRGs was also significantly correlated with sensitivity to these drugs. In the high PRG expression group, the IC50 value of chemotherapeutic agents was obviously decreased, indicating that HCC patients with elevated PRG expression may gain more therapeutic benefits from these drugs through pyroptosis, which may make the treatment of HCC more effective and have fewer side effects. As an inflammatory type of RCD, pyroptosis was identified by cell swelling, membrane rupture and pore formation, leading to the release of intracellular contents, including IL-1β and IL-18, and ultimately causing a cascade-amplified inflammatory action [7]. The essential components of pyroptosis, including inflammasomes, GSDM proteins and cytokines, are all associated with the development, invasion and metastasis of tumors [15]. Cleavage of GSDM family members, such as GSDMD and GSDME, mediated by cysteine proteases is the key process that causes pyroptosis [38]. Previous studies have shown that GSDME is downregulated in some human cancers and might act as a tumor suppressor [39, 40]. The DNA methylase inhibitor decitabine (5-aza-2'-deoxycytosine) could downregulate the expression of GSDME, thereby preventing the proliferation and colony formation ability of melanoma, gastric cancer and CRC cells, and may reduce the invasive ability of breast cancer cells [41]. In addition, GSDME is associated with etoposide resistance [42]. Loss of GSDME facilitates the resistance of melanoma cell lines to etoposide, which can be rescued by overexpression of GSDME [42]. Treatment of lung cancer cells with inhibitors of KRAS, EGFR or ALK results in caspase-3-regulated activation of GSDME, thereby increasing the anticancer efficacy of these drugs [43, 44]. In mouse tumor models, knockdown of GSDME enhanced tumor growth, whereas ectopic expression of GSDME inhibited tumor growth [45]. Importantly, the tumor inhibitory effect of GSDME was dominated by killing cytotoxic lymphocytes, as this effect was markedly abolished in mice with loss of perforin or in mice deficient in CD8+ T and NK cells [45]. CAR-T cells induce pyroptosis by sequentially releasing granzyme B, activating caspase-3 and cleaving GSDME [46]. Pyroptosis-associated factors in turn activate caspase-1 in macrophages, leading to cleavage of GSDMD, which ultimately induces cytokine release syndrome [46]. Consistently, knockout of the GMEDE gene in B16 melanoma greatly decreased the survival rate in tumor-implanted mice. Therefore, GSDM genes not only trigger pyroptosis in tumor cells but also activate antitumor immunity [44, 47]. NOD1, a member of the pattern recognition receptor (PRR) family, is involved in various pathologies, especially cancer. NOD1 is expressed in some types of cells, including endothelial cells, hematopoietic cells and various immune cells (e.g., neutrophils, macrophages, monocytes, NK cells, and lymphocytes) [48]. These findings are consistent with our observations. NOD1 activation elicits antigen-specific T-cell immune responses primarily through Th2 polarization [49]. Additionally, NOD1 stimulates Th1, Th2 and Th17 immune responses along with other innate immune TLRs [49]. Additionally, NOD1 activation also contributes to the B-cell antigen receptor-assisted survival of mature B cells [50]. Activation of NOD1 also promoted chemokine production and specific recruitment of neutrophils in mice [51]. A recent study demonstrated that activation of NOD1 facilitated oncogenesis by promoting autophagy-dependent macrophage reprogramming and triggering myeloid-derived suppressor cell (MDSC) expansion and immunosuppressive ability through arginase-1 activity in colorectal cancer [52]. In contrast, NOD1 expression was markedly decreased in HCC tissues, and overexpression of NOD1 greatly prevented tumorigenesis and increased the response to chemotherapeutic drugs through suppression of the SRC/MAPK pathway in vitro and in vivo [53]. These results imply that NOD1 exerts its tumor-suppressive effect on HCC. In the current study, we observed that NOD1 expression was markedly increased in HCC tissues in some datasets and that NOD1 was also expressed in immune cells through single-cell RNA sequence analysis. PLCG1, a primary subtype of phospholipase C (PLC), is directly activated by diverse membrane receptors. Upon T-cell activation, as a phospholipase, PLCG1 can cleave phosphatidylinositol 4,5-diphosphate in the plasma membrane into two second messengers: inositol 1,4,5 triphosphate and diacylglycerol. Inositol 1,4,5-triphosphate causes calcium release from the endoplasmic reticulum, increases the intracellular calcium concentration and activates NFAT, while diacylglycerol activates specific isoforms of protein kinase C (PKC) [54, 55]. Recent bioinformatics analysis identified that PLCG1 was frequently highly expressed and mutated in various cancers and was involved in tumorigenesis as an oncogene [56]. Elevated expression of PLCG1 was linked with poor survival and tumor progression in lower-grade glioma (LGG) patients [57]. Knockdown of PLCG1 significantly reduced the proliferation, migration and invasiveness of IDH wild-type LGG cells [57]. The PLCG1-mediated signaling pathway also regulated tumor metastasis. The PLCG1/PKCθ axis accelerated STAT3 activation and promoted the proliferation and survival of cutaneous T-cell lymphoma cells [55]. These results have highlighted the important role of these PRGs in immunity and oncogenesis. A growing body of research has revealed that the five core prognostic PRGs are also closely related to various human diseases. GSDME expression was elevated in the renal tubules of patients with systemic lupus erythematosus (SLE) and pristane-induced lupus mice. Knockout of GSDME significantly alleviated SLE pathogenesis by suppressing GSDME-regulated pyroptosis of renal cells [58]. These data suggest that GSDME-mediated pyroptosis is involved in the pathogenesis of SLE and that GSDME may be a potential therapeutic target for SLE. Loss of GSDME effectively ameliorated cisplatin- or ischemia–reperfusion-induced inflammation and acute kidney injury by inhibiting caspase-3/GSDME-induced pyroptosis [59]. In fact, some chemotherapeutic drugs, including cisplatin and doxorubicin, can trigger GSDME cleavage in human renal cells. Knockdown of GSDME attenuated doxorubicin- or cisplatin-triggered cell pyroptosis [60]. Therefore, GSDME-modulated pyroptosis may play a vital role in chemotherapy-induced nephrotoxicity. Moreover, loss of GSDME also aggravated skin damage in UVB-treated mice by promoting the infiltration and activation of neutrophils [61]. Previous studies have identified NOD1 as a key player in host-microbial defense and multiple inflammatory diseases. There is a direct link between NOD1 and atherosclerosis. In vivo experiments indicated that deficiency of NOD1 reduced the risk of atherosclerotic thrombosis by inhibiting leukocyte infiltration and decreasing macrophage apoptosis [62]. NOD1 expression was upregulated in the adipose tissue of patients with metabolic syndrome or gestational diabetes [63, 64]. Interestingly, the polymorphism in NOD1 (Glu266Lys) was associated with dietary saturated fat and insulin sensitivity in humans aged 20-29 years [65]. Whole body or hematopoietic depletion of NOD1 significantly decreased high-fat diet (HFD)-associated glucose and insulin resistance in mice [66, 67]. Another study indicated that, loss of NOD1 accelerated obesity in mice fed a HFD, accompanied by increased levels of free thyroidal T4, reduced expression of uncoupling protein 1 (UCP1) in brown adipose tissues, and enhanced infiltration of inflammatory cells in white adipose tissues and liver tissues, suggesting a protective role of NOD1 against inflammation and obesity [68]. Infection with *Japanese encephalitis* virus (JEV) greatly elevated the transcriptional and protein expression of NOD1 in mice, and knockout of NOD1 enhanced resistance to JEV infection by inhibiting the neuroinflammatory response and multiple downstream signaling pathways [69]. Knockout of NOD1 also significantly decreased the number of isolated lymphoid follicles in the distal ileum and colon of mice and greatly increased the total number of bacteria in the ileum to affect intestinal homeostasis [70]. Compared with wild-type (WT) mice, mice lacking NOD1 are more likely to be infected with early pneumococcal septicemia, which implies that NOD1 plays a key role in initiating innate defense and promoting a rapid response to infection [71]. These results imply that the physiological function of NOD1 in the intestine is crucial to maintain the homeostasis between the microbiota and host immune system. PLCG1 is a vital regulator of cellular signaling. In mice, specific knockout of PLCG1 in neural progenitor cells resulted in axonal guidance defects in the dorsal midbrain during embryogenesis. Moreover, in adult PLCG1-deficient mice, structural changes in the corpus callosum, olfactory tubercle, and substantia innominate were observed. These data indicated that PLCG1 may play key roles in the development of white matter structure by regulating the netrin-1/deleted in colorectal cancer (DCC) complex signaling pathway [72]. Mice with GABAergic neuron-specific deletion of PLCG1 showed handling-induced recurrent seizures with a reduced number of GABAergic synapses, decreased hippocampal inhibitory synaptic transmission, anxiety alleviation and fear memory disorder [73]. Numerous studies have extensively investigated the immune functions of NLRC4 in response to bacterial infection. For example, mice with NLRC4 deficiency had low resistance to Salmonella Typhimurium and *Legionella pneumophila* infections and exhibited elevated bacterial burden [74]. When mice were infected with Shigella, intestinal mucosa thickening, shrinkage of the cecum, macroscopic edema, and acute weight loss were observed in NLRC4-/- mice, suggesting that NLRC4 conferred resistance to Shigella infection [75]. In NLRC4 knockout mice, bacterial flagellin, one of the main innate immune activators in the intestine, failed to induce the expression of IL-18 and IL-1β, indicating that NLRC4 was necessary to rapidly generate inflammasome cytokines [76]. Lack of NLRC4 also aggravated dextran sulfate sodium (DSS)-induced acute colitis and increased flagellate-caused mortality in mice [76]. Recently, increasing evidence has revealed the important functions of NLRP6 in microbial infection-associated inflammation. Mice lacking NLRP6 were highly resistant to infection with a variety of bacterial pathogens, such as Salmonella typhimurium, *Listeria monocytogenes* and Escherichia coli. When NLRP6-deficient mice were infected with these bacterial pathogens, the number of circulating monocytes and neutrophils increased, accompanied by activation of the mitogen-activated protein kinase (MAPK) and nuclear factor-κB (NF-κB) signaling pathways. In contrast, NLRP6-/- mice showed increased parasite shedding and significant susceptibility to Cryptosporidium infection compared with WT control mice [77]. NLRP6 knockout mice exhibited spontaneous intestinal hyperplasia, large recruitment of inflammatory cells, and deterioration of DSS-induced colitis. The lack of NLRP6 in mouse colon epithelial cells led to a decrease in IL-18 levels and a change in fecal microbiota composition. Compared with WT controls, NLRP6-/- mice infected with encephalomyocarditis virus or murine norovirus 1 had increased mortality and viremia [78]. Mechanistically, NLRP6 bound to viral RNA in cooperation with Asp-Glu-Ala-His (DEAH) box helicase 15 (DHX15) to induce the expression of interferons and interferon-stimulated genes [78]. Additionally, the expression of NLRP6 was increased in intestinal tissues when mice were infected with Candida albicans. The colonization of Candida albicans facilitated HCC growth in WT mice, but this effect disappeared in NLRP6-/- mice, suggesting that NLRP6 could promote the occurrence and development of HCC [79]. Although there are some contradictory experimental results, NLRP6 undoubtedly participates in the regulation of innate immunity [80]. Immunotherapy aims to activate the human immune system to kill tumor cells and inhibit tumor growth. The targets of immunotherapy are not tumor cells and tissues but the human body's own immune system [81, 82]. The increased expression of immune checkpoint molecules on cancer cells and/or tumor-infiltrating immune cells can inhibit antitumor immunity. Previous studies have confirmed the clinical efficacy of the application of PD-1 or PD-L1 in inhibiting the progression of advanced HCC [83, 84]. Thus, immunotherapy has become a novel treatment approach representing an effective and promising option against HCC. In our current research, both TIMER and CIBERSORT analyses demonstrated that the two clusters exhibited different infiltrated abundances of various immune cells. Interestingly, C2 exhibited higher immune cell infiltration and immune checkpoint gene expression. Moreover, the risk score of the 5-PRG signature was also markedly and positively linked with the infiltrated abundances of six major immune cells. Moreover, the single-cell RNA sequencing analysis results indicated that the core PRGs in the prognostic signature were expressed in both liver cancer cells and different immune cells. More importantly, we also observed that patients in C2 corresponded to higher TIDE scores according to the TIDE algorithm, indicating a worse response to immunotherapy. In summary, the identified distinct clusters and prognostic signature play a critical role in mediating immune cell infiltration and immunotherapy response. Despite the promising findings obtained, our study still has several limitations. First, a pyroptosis-related prognostic model was constructed by using retrospective data from different databases to predict the survival rate of HCC patients. More large-scale data are needed to assess the application potential of the five PRG-based risk score models. Second, the expression of PRGs in different databases is not consistent. Most HCC patients in the TCGA-HCC database were Caucasian, and it is not clear whether the expression of PRGs and the prognostic signature has a similar tendency in other races and datasets. Third, the molecular functions of the five PRGs identified in this study need to be verified by more in-depth in vitro and in vivo experiments and clinical data to further explore their roles and their impact on immune cell infiltration and immunotherapy in HCC. In summary, our analysis results provided insight into the expression pattern of the PRGs and constructed a risk score model and nomogram for prognosis prediction. The two independent clusters and the 5-PRG risk score, which integrated pyroptosis and immunological features with GSDME, NOD1, PLCG1, NLRP6 and NLRC4, could reliably predict prognosis and immunotherapy response in HCC patients. Additionally, the risk score-based nomogram model has promising clinical applications. ## Data collection and process The mRNA expression data and relevant clinical information for patients with HCC (371 HCC samples and 50 normal samples) were downloaded from the TCGA database. RNA sequencing (RNA-seq) data from the ICGC (International Cancer Gene Consortium) database, containing 202 normal samples and 240 HCC samples, were downloaded and used as the validation cohort. Moreover, expression of 5 PRGs in different GEO datasets were downloaded from the HCCDB database [36, 85]. ## Identification of differentially expressed PRGs PRGs were collected from a previous study [86]. The expression profiles of 33 PRGs were directly downloaded from TCGA and ICGC databases. The “ggplot2” and “pheatmap” packages of R language were used to identify differentially expressed PRGs with a P value <0.05. The online STRING (https://string-db.org/) and Metascape (https://metascape.org/gp/index.html#/main/step1) platforms were used to construct the gene-gene interaction and protein-protein interaction (PPI) networks. ## Consensus clustering analysis of PRGs The PRGs were subjected to unsupervised clustering analysis with the R package “ConsensusClusterPlus”. Principal component analysis (PCA) was carried out to estimate the gene expression patterns among different clusters. Clustering heatmaps were generated using the “pheatmap” package. Kaplan-Meier (KM) analysis was performed to reveal the difference in survival among different clusters by using the “survival” and “survminer” packages. ## Construction of the risk score Univariate regression analysis was first applied to select PRGs that were correlated with prognosis in HCC. Then, LASSO (least absolute shrinkage and selection operator) regression analysis with the R package “glmnet” was applied to construct the risk score model after univariate regression analysis. The equation was established as follows: risk score = sum of coefficients × prognostic PRG expression levels. KM curves and receiver operating characteristic (ROC) curves were further utilized to examine the prognostic ability of the risk model. ## Construction of the nomogram Univariate Cox regression and multivariate Cox regression analyses were applied to verify whether the risk model was linked with prognosis in HCC. In addition, a nomogram was constructed based on age, sex, tumor (T), node (N), metastasis (M) and risk score using the R package “rms”. ## Mutation landscapes in two clusters Tumor mutation burden (TMB) could predict the response to some different forms of immunotherapy and across multiple types of cancer. The mutation landscapes of the two clusters were visualized and compared through the R package “maftools”. ## Analysis of differentially expressed genes The R package “limma” was utilized to acquire the differentially expressed genes between different clusters with |log2 (fold change)|> 1.5 and $p \leq 0.05.$ Additionally, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment and Gene Ontology (GO) analyses were performed to explore the potential function of differentially expressed genes using the “ClusterProfiler” package. Heatmaps and boxplots were generated using the “pheatmap” and “ggplot2” packages, respectively. ## Immune cell infiltration abundance in HCC The infiltration abundances of a variety of immune cells between the two risk groups were investigated using the CIBERSORT and TIMER algorithms. The infiltrated abundance of various immune cells in every HCC sample was explored using the “immunedeconv” package. The heatmap results are shown by the R package “pheatmap”. ## Associations between clusters and immunotherapy response The expression levels of major immune checkpoint genes between cluster 1 and cluster 2 were compared to show the difference under immunotherapy between the two subgroups. Additionally, the responses to immunotherapy were assessed with the TIDE (tumor immune dysfunction and exclusion) algorithm using the R packages “ggplot2” and “ggpubr”. The TIDE score of HCC was obtained from http://tide.dfci.harvard.edu. ## Immunohistochemistry analysis Immunohistochemical (IHC) staining results were directly obtained from the HPA (Human Protein Atlas) database (https://www.proteinatlas.org/) as described previously [36, 87]. The protein levels of PRGs in normal liver tissues and HCC tissues were compared through IHC staining. ## Targeted therapy drug prediction The chemotherapeutic response for each sample was predicted according to the largest publicly available Genomics of Drug Sensitivity in Cancer (GDSC) database (https://www.cancerrxgene.org/) with the R package “pRRophetic”. The IC50 (half-maximal inhibitory concentration) was assessed through ridge regression. ## Statistical analysis All statistics were performed using R software (version 4.0.3). The Wilcoxon test was used for comparisons between two different subgroups. In KM analysis, the log-rank test was applied to estimate the difference in survival rate between subgroups. A p value less than 0.05 was set as statistically significant for all analyses. ## Data availability The datasets analyzed for this study can be downloaded from the TCGA database (https://portal.gdc.cancer.gov), ICGC database (https://dcc.icgc.org/releases/current/Projects) and HCCDB database (http://lifeome.net/database/hccdb/home.html). All data generated or analyzed during this study are included in this article and its Supplementary Material files. Further inquiries can be directed to the corresponding authors. ## Consent for publication All authors approved the manuscript for publication. ## References 1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, Bray F. **Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries.**. *CA Cancer J Clin* (2021) **71** 209-49. DOI: 10.3322/caac.21660 2. Llovet JM, Kelley RK, Villanueva A, Singal AG, Pikarsky E, Roayaie S, Lencioni R, Koike K, Zucman-Rossi J, Finn RS. **Hepatocellular carcinoma.**. *Nat Rev Dis Primers* (2021) **7** 6. DOI: 10.1038/s41572-020-00240-3 3. Forner A, Reig M, Bruix J. **Hepatocellular carcinoma.**. *Lancet* (2018) **391** 1301-14. DOI: 10.1016/S0140-6736(18)30010-2 4. Zhu XD, Li KS, Sun HC. **Adjuvant therapies after curative treatments for hepatocellular carcinoma: Current status and prospects.**. *Genes Dis* (2020) **7** 359-69. DOI: 10.1016/j.gendis.2020.02.002 5. Siegel RL, Miller KD, Jemal A. **Cancer statistics, 2020.**. *CA Cancer J Clin* (2020) **70** 7-30. DOI: 10.3322/caac.21590 6. Hu J, Gong N, Li D, Deng Y, Chen J, Luo D, Zhou W, Xu K. **Identifying hepatocellular carcinoma patients with survival benefits from surgery combined with chemotherapy: based on machine learning model.**. *World J Surg Oncol* (2022) **20** 377. DOI: 10.1186/s12957-022-02837-2 7. Shi J, Gao W, Shao F. **Pyroptosis: Gasdermin-Mediated Programmed Necrotic Cell Death.**. *Trends Biochem Sci* (2017) **42** 245-54. DOI: 10.1016/j.tibs.2016.10.004 8. Zhou Z, He H, Wang K, Shi X, Wang Y, Su Y, Wang Y, Li D, Liu W, Zhang Y, Shen L, Han W, Shen L. **Granzyme A from cytotoxic lymphocytes cleaves GSDMB to trigger pyroptosis in target cells.**. *Science* (2020) **368** eaaz7548. DOI: 10.1126/science.aaz7548 9. Pan Y, Cai W, Huang J, Cheng A, Wang M, Yin Z, Jia R. **Pyroptosis in development, inflammation and disease.**. *Front Immunol* (2022) **13** 991044. DOI: 10.3389/fimmu.2022.991044 10. Ma Y, Jiang J, Gao Y, Shi T, Zhu X, Zhang K, Lu K, Xue B. **Research progress of the relationship between pyroptosis and disease.**. *Am J Transl Res* (2018) **10** 2213-9. PMID: 30093958 11. Lu L, Zhang Y, Tan X, Merkher Y, Leonov S, Zhu L, Deng Y, Zhang H, Zhu D, Tan Y, Fu Y, Liu T, Chen Y. **Emerging mechanisms of pyroptosis and its therapeutic strategy in cancer.**. *Cell Death Discov* (2022) **8** 338. DOI: 10.1038/s41420-022-01101-6 12. Yu P, Zhang X, Liu N, Tang L, Peng C, Chen X. **Pyroptosis: mechanisms and diseases.**. *Signal Transduct Target Ther* (2021) **6** 128. DOI: 10.1038/s41392-021-00507-5 13. Kovacs SB, Miao EA. **Gasdermins: Effectors of Pyroptosis.**. *Trends Cell Biol* (2017) **27** 673-84. DOI: 10.1016/j.tcb.2017.05.005 14. Shi J, Zhao Y, Wang K, Shi X, Wang Y, Huang H, Zhuang Y, Cai T, Wang F, Shao F. **Cleavage of GSDMD by inflammatory caspases determines pyroptotic cell death.**. *Nature* (2015) **526** 660-5. DOI: 10.1038/nature15514 15. Wei X, Xie F, Zhou X, Wu Y, Yan H, Liu T, Huang J, Wang F, Zhou F, Zhang L. **Role of pyroptosis in inflammation and cancer.**. *Cell Mol Immunol* (2022) **19** 971-92. DOI: 10.1038/s41423-022-00905-x 16. Khanova E, Wu R, Wang W, Yan R, Chen Y, French SW, Llorente C, Pan SQ, Yang Q, Li Y, Lazaro R, Ansong C, Smith RD. **Pyroptosis by caspase11/4-gasdermin-D pathway in alcoholic hepatitis in mice and patients.**. *Hepatology* (2018) **67** 1737-53. DOI: 10.1002/hep.29645 17. Liu X, Zhang Z, Ruan J, Pan Y, Magupalli VG, Wu H, Lieberman J. **Inflammasome-activated gasdermin D causes pyroptosis by forming membrane pores.**. *Nature* (2016) **535** 153-8. DOI: 10.1038/nature18629 18. Wang Y, Gao W, Shi X, Ding J, Liu W, He H, Wang K, Shao F. **Chemotherapy drugs induce pyroptosis through caspase-3 cleavage of a gasdermin.**. *Nature* (2017) **547** 99-103. DOI: 10.1038/nature22393 19. Geddes K, Magalhães JG, Girardin SE. **Unleashing the therapeutic potential of NOD-like receptors.**. *Nat Rev Drug Discov* (2009) **8** 465-79. DOI: 10.1038/nrd2783 20. Lamkanfi M. **Emerging inflammasome effector mechanisms.**. *Nat Rev Immunol* (2011) **11** 213-20. DOI: 10.1038/nri2936 21. Guo H, Callaway JB, Ting JP. **Inflammasomes: mechanism of action, role in disease, and therapeutics.**. *Nat Med* (2015) **21** 677-87. DOI: 10.1038/nm.3893 22. Kufer TA, Sansonetti PJ. **NLR functions beyond pathogen recognition.**. *Nat Immunol* (2011) **12** 121-8. DOI: 10.1038/ni.1985 23. Wang L, Qin X, Liang J, Ge P. **Induction of Pyroptosis: A Promising Strategy for Cancer Treatment.**. *Front Oncol* (2021) **11** 635774. DOI: 10.3389/fonc.2021.635774 24. Wang Q, Wang C, Chen J. **NLRP6, decreased in gastric cancer, suppresses tumorigenicity of gastric cancer cells.**. *Cancer Manag Res* (2018) **10** 6431-44. DOI: 10.2147/CMAR.S182980 25. Shen Y, Li X, Wang D, Zhang L, Li X, Xia T, Shang X, Yang X, Su L, Fan X. **Novel prognostic model established for patients with head and neck squamous cell carcinoma based on pyroptosis-related genes.**. *Transl Oncol* (2021) **14** 101233. DOI: 10.1016/j.tranon.2021.101233 26. Lim J, Kim MJ, Park Y, Ahn JW, Hwang SJ, Moon JS, Cho KG, Kwack K. **Upregulation of the NLRC4 inflammasome contributes to poor prognosis in glioma patients.**. *Sci Rep* (2019) **9** 7895. DOI: 10.1038/s41598-019-44261-9 27. Ohashi K, Wang Z, Yang YM, Billet S, Tu W, Pimienta M, Cassel SL, Pandol SJ, Lu SC, Sutterwala FS, Bhowmick N, Seki E. **NOD-like receptor C4 Inflammasome Regulates the Growth of Colon Cancer Liver Metastasis in NAFLD.**. *Hepatology* (2019) **70** 1582-99. DOI: 10.1002/hep.30693 28. Gao J, Qiu X, Xi G, Liu H, Zhang F, Lv T, Song Y. **Downregulation of GSDMD attenuates tumor proliferation via the intrinsic mitochondrial apoptotic pathway and inhibition of EGFR/Akt signaling and predicts a good prognosis in non-small cell lung cancer.**. *Oncol Rep* (2018) **40** 1971-84. DOI: 10.3892/or.2018.6634 29. Wu M, Wang Y, Yang D, Gong Y, Rao F, Liu R, Danna Y, Li J, Fan J, Chen J, Zhang W, Zhan Q. **A PLK1 kinase inhibitor enhances the chemosensitivity of cisplatin by inducing pyroptosis in oesophageal squamous cell carcinoma.**. *EBioMedicine* (2019) **41** 244-55. DOI: 10.1016/j.ebiom.2019.02.012 30. Rogers C, Fernandes-Alnemri T, Mayes L, Alnemri D, Cingolani G, Alnemri ES. **Cleavage of DFNA5 by caspase-3 during apoptosis mediates progression to secondary necrotic/pyroptotic cell death.**. *Nat Commun* (2017) **8** 14128. DOI: 10.1038/ncomms14128 31. Jorgensen I, Miao EA. **Pyroptotic cell death defends against intracellular pathogens.**. *Immunol Rev* (2015) **265** 130-42. DOI: 10.1111/imr.12287 32. Rao Z, Zhu Y, Yang P, Chen Z, Xia Y, Qiao C, Liu W, Deng H, Li J, Ning P, Wang Z. **Pyroptosis in inflammatory diseases and cancer.**. *Theranostics* (2022) **12** 4310-29. DOI: 10.7150/thno.71086 33. Wu D, Wei C, Li Y, Yang X, Zhou S. **Pyroptosis, a New Breakthrough in Cancer Treatment.**. *Front Oncol* (2021) **11** 698811. DOI: 10.3389/fonc.2021.698811 34. Wang YY, Liu XL, Zhao R. **Induction of Pyroptosis and Its Implications in Cancer Management.**. *Front Oncol* (2019) **9** 971. DOI: 10.3389/fonc.2019.00971 35. Llovet JM, Ricci S, Mazzaferro V, Hilgard P, Gane E, Blanc JF, de Oliveira AC, Santoro A, Raoul JL, Forner A, Schwartz M, Porta C, Zeuzem S. **Sorafenib in advanced hepatocellular carcinoma.**. *N Engl J Med* (2008) **359** 378-90. DOI: 10.1056/NEJMoa0708857 36. Zhang B, Zhao J, Liu B, Shang Y, Chen F, Zhang S, He J, Fan Y, Tan K. **Development and Validation of a Novel Ferroptosis-Related Gene Signature for Prognosis and Immunotherapy in Hepatocellular Carcinoma.**. *Front Mol Biosci* (2022) **9** 940575. DOI: 10.3389/fmolb.2022.940575 37. Keating GM. **Sorafenib: A Review in Hepatocellular Carcinoma.**. *Target Oncol* (2017) **12** 243-53. DOI: 10.1007/s11523-017-0484-7 38. Magnani L, Colantuoni M, Mortellaro A. **Gasdermins: New Therapeutic Targets in Host Defense, Inflammatory Diseases, and Cancer.**. *Front Immunol* (2022) **13** 898298. DOI: 10.3389/fimmu.2022.898298 39. Ibrahim J, De Schutter E, Op de Beeck K. **GSDME: A Potential Ally in Cancer Detection and Treatment.**. *Trends Cancer* (2021) **7** 392-4. DOI: 10.1016/j.trecan.2020.12.002 40. Yao L, Li J, Xu Z, Yan Y, Hu K. **GSDMs are potential therapeutic targets and prognostic biomarkers in clear cell renal cell carcinoma.**. *Aging (Albany NY)* (2022) **14** 2758-74. DOI: 10.18632/aging.203973 41. Li F, Xia Q, Ren L, Nie Y, Ren H, Guo X, Yu J, Xing Y, Chen Z. **GSDME Increases Chemotherapeutic Drug Sensitivity by Inducing Pyroptosis in Retinoblastoma Cells.**. *Oxid Med Cell Longev* (2022) **2022** 2371807. DOI: 10.1155/2022/2371807 42. Lage H, Helmbach H, Grottke C, Dietel M, Schadendorf D. **DFNA5 (ICERE-1) contributes to acquired etoposide resistance in melanoma cells.**. *FEBS Lett* (2001) **494** 54-9. DOI: 10.1016/s0014-5793(01)02304-3 43. Lu H, Zhang S, Wu J, Chen M, Cai MC, Fu Y, Li W, Wang J, Zhao X, Yu Z, Ma P, Zhuang G. **Molecular Targeted Therapies Elicit Concurrent Apoptotic and GSDME-Dependent Pyroptotic Tumor Cell Death.**. *Clin Cancer Res* (2018) **24** 6066-77. DOI: 10.1158/1078-0432.CCR-18-1478 44. Tan Y, Chen Q, Li X, Zeng Z, Xiong W, Li G, Li X, Yang J, Xiang B, Yi M. **Pyroptosis: a new paradigm of cell death for fighting against cancer.**. *J Exp Clin Cancer Res* (2021) **40** 153. DOI: 10.1186/s13046-021-01959-x 45. Zhang Z, Zhang Y, Xia S, Kong Q, Li S, Liu X, Junqueira C, Meza-Sosa KF, Mok TM, Ansara J, Sengupta S, Yao Y, Wu H, Lieberman J. **Gasdermin E suppresses tumour growth by activating anti-tumour immunity.**. *Nature* (2020) **579** 415-20. DOI: 10.1038/s41586-020-2071-9 46. Liu Y, Fang Y, Chen X, Wang Z, Liang X, Zhang T, Liu M, Zhou N, Lv J, Tang K, Xie J, Gao Y, Cheng F. **Gasdermin E-mediated target cell pyroptosis by CAR T cells triggers cytokine release syndrome.**. *Sci Immunol* (2020) **5** eaax7969. DOI: 10.1126/sciimmunol.aax7969 47. Li L, Jiang M, Qi L, Wu Y, Song D, Gan J, Li Y, Bai Y. **Pyroptosis, a new bridge to tumor immunity.**. *Cancer Sci* (2021) **112** 3979-94. DOI: 10.1111/cas.15059 48. Fernández-García V, González-Ramos S, Martín-Sanz P, Laparra JM, Boscá L. **NOD1-Targeted Immunonutrition Approaches: On the Way from Disease to Health.**. *Biomedicines* (2021) **9** 519. DOI: 10.3390/biomedicines9050519 49. Fritz JH, Le Bourhis L, Sellge G, Magalhaes JG, Fsihi H, Kufer TA, Collins C, Viala J, Ferrero RL, Girardin SE, Philpott DJ. **Nod1-mediated innate immune recognition of peptidoglycan contributes to the onset of adaptive immunity.**. *Immunity* (2007) **26** 445-59. DOI: 10.1016/j.immuni.2007.03.009 50. Hayakawa K, Formica AM, Zhou Y, Ichikawa D, Asano M, Li YS, Shinton SA, Brill-Dashoff J, Núñez G, Hardy RR. **NLR Nod1 signaling promotes survival of BCR-engaged mature B cells through up-regulated Nod1 as a positive outcome.**. *J Exp Med* (2017) **214** 3067-83. DOI: 10.1084/jem.20170497 51. Masumoto J, Yang K, Varambally S, Hasegawa M, Tomlins SA, Qiu S, Fujimoto Y, Kawasaki A, Foster SJ, Horie Y, Mak TW, Núñez G, Chinnaiyan AM. **Nod1 acts as an intracellular receptor to stimulate chemokine production and neutrophil recruitment**. *J Exp Med* (2006) **203** 203-13. DOI: 10.1084/jem.20051229 52. Maisonneuve C, Tsang DKL, Foerster EG, Robert LM, Mukherjee T, Prescott D, Tattoli I, Lemire P, Winer DA, Winer S, Streutker CJ, Geddes K, Cadwell K. **Nod1 promotes colorectal carcinogenesis by regulating the immunosuppressive functions of tumor-infiltrating myeloid cells.**. *Cell Rep* (2021) **34** 108677. DOI: 10.1016/j.celrep.2020.108677 53. Ma X, Qiu Y, Zhu L, Zhao Y, Lin Y, Ma D, Qin Z, Sun C, Shen X, Li T, Han L. **NOD1 inhibits proliferation and enhances response to chemotherapy via suppressing SRC-MAPK pathway in hepatocellular carcinoma.**. *J Mol Med (Berl)* (2020) **98** 221-32. DOI: 10.1007/s00109-019-01868-9 54. Vaqué JP, Gómez-López G, Monsálvez V, Varela I, Martínez N, Pérez C, Domínguez O, Graña O, Rodríguez-Peralto JL, Rodríguez-Pinilla SM, González-Vela C, Rubio-Camarillo M, Martín-Sánchez E. **PLCG1 mutations in cutaneous T-cell lymphomas.**. *Blood* (2014) **123** 2034-43. DOI: 10.1182/blood-2013-05-504308 55. García-Díaz N, Casar B, Alonso-Alonso R, Quevedo L, Rodríguez M, Ruso-Julve F, Esteve-Codina A, Gut M, Gru AA, González-Vela MC, Gut I, Rodriguez-Peralto JL, Varela I. **PLCγ1/PKCθ Downstream Signaling Controls Cutaneous T-Cell Lymphoma Development and Progression.**. *J Invest Dermatol* (2022) **142** 1391-400.e15. DOI: 10.1016/j.jid.2021.09.024 56. Jang HJ, Suh PG, Lee YJ, Shin KJ, Cocco L, Chae YC. **PLCγ1: Potential arbitrator of cancer progression.**. *Adv Biol Regul* (2018) **67** 179-89. DOI: 10.1016/j.jbior.2017.11.003 57. Li T, Yang Z, Li H, Zhu J, Wang Y, Tang Q, Shi Z. **Phospholipase Cγ1 (PLCG1) overexpression is associated with tumor growth and poor survival in IDH wild-type lower-grade gliomas in adult patients.**. *Lab Invest* (2022) **102** 143-53. DOI: 10.1038/s41374-021-00682-7 58. Luo G, He Y, Yang F, Zhai Z, Han J, Xu W, Zhang J, Zhuang L, Zhang Y, Li Y, Song R, Luo X, Liang J, Sun E. **Blocking GSDME-mediated pyroptosis in renal tubular epithelial cells alleviates disease activity in lupus mice.**. *Cell Death Discov* (2022) **8** 113. DOI: 10.1038/s41420-022-00848-2 59. Xia W, Li Y, Wu M, Jin Q, Wang Q, Li S, Huang S, Zhang A, Zhang Y, Jia Z. **Gasdermin E deficiency attenuates acute kidney injury by inhibiting pyroptosis and inflammation.**. *Cell Death Dis* (2021) **12** 139. DOI: 10.1038/s41419-021-03431-2 60. Shen X, Wang H, Weng C, Jiang H, Chen J. **Caspase 3/GSDME-dependent pyroptosis contributes to chemotherapy drug-induced nephrotoxicity.**. *Cell Death Dis* (2021) **12** 186. DOI: 10.1038/s41419-021-03458-5 61. Chen Y, Lian N, Chen S, Xiao T, Ke Y, Zhang Y, Song C, Yang Y, Xu S, Gu H, Chen X. **GSDME deficiency leads to the aggravation of UVB-induced skin inflammation through enhancing recruitment and activation of neutrophils.**. *Cell Death Dis* (2022) **13** 841. DOI: 10.1038/s41419-022-05276-9 62. González-Ramos S, Fernández-García V, Recalde M, Rodríguez C, Martínez-González J, Andrés V, Martín-Sanz P, Boscá L. **Deletion or Inhibition of NOD1 Favors Plaque Stability and Attenuates Atherothrombosis in Advanced Atherogenesis**. *Cells* (2020) **9** 2067. DOI: 10.3390/cells9092067 63. Lappas M. **NOD1 expression is increased in the adipose tissue of women with gestational diabetes.**. *J Endocrinol* (2014) **222** 99-112. DOI: 10.1530/JOE-14-0179 64. Zhou YJ, Liu C, Li CL, Song YL, Tang YS, Zhou H, Li A, Li Y, Weng Y, Zheng FP. **Increased NOD1, but not NOD2, activity in subcutaneous adipose tissue from patients with metabolic syndrome.**. *Obesity (Silver Spring)* (2015) **23** 1394-400. DOI: 10.1002/oby.21113 65. Cuda C, Badawi A, Karmali M, El-Sohemy A. **Effects of polymorphisms in nucleotide-binding oligomerization domains 1 and 2 on biomarkers of the metabolic syndrome and type II diabetes.**. *Genes Nutr* (2012) **7** 427-35. DOI: 10.1007/s12263-012-0287-5 66. Amar J, Chabo C, Waget A, Klopp P, Vachoux C, Bermúdez-Humarán LG, Smirnova N, Bergé M, Sulpice T, Lahtinen S, Ouwehand A, Langella P, Rautonen N. **Intestinal mucosal adherence and translocation of commensal bacteria at the early onset of type 2 diabetes: molecular mechanisms and probiotic treatment.**. *EMBO Mol Med* (2011) **3** 559-72. DOI: 10.1002/emmm.201100159 67. Chan KL, Tam TH, Boroumand P, Prescott D, Costford SR, Escalante NK, Fine N, Tu Y, Robertson SJ, Prabaharan D, Liu Z, Bilan PJ, Salter MW. **Circulating NOD1 Activators and Hematopoietic NOD1 Contribute to Metabolic Inflammation and Insulin Resistance.**. *Cell Rep* (2017) **18** 2415-26. DOI: 10.1016/j.celrep.2017.02.027 68. González-Ramos S, Paz-García M, Fernández-García V, Portune KJ, Acosta-Medina EF, Sanz Y, Castrillo A, Martín-Sanz P, Obregon MJ, Boscá L. **NOD1 deficiency promotes an imbalance of thyroid hormones and microbiota homeostasis in mice fed high fat diet.**. *Sci Rep* (2020) **10** 12317. DOI: 10.1038/s41598-020-69295-2 69. Chen Z, Zhao Z, Liu Y, Imran M, Rao J, Cai N, Ye J, Cao S. **Nucleotide-Binding Oligomerization Domain 1 (NOD1) Positively Regulates Neuroinflammation during Japanese Encephalitis Virus Infection.**. *Microbiol Spectr* (2022) **10** e0258321. DOI: 10.1128/spectrum.02583-21 70. Bouskra D, Brézillon C, Bérard M, Werts C, Varona R, Boneca IG, Eberl G. **Lymphoid tissue genesis induced by commensals through NOD1 regulates intestinal homeostasis.**. *Nature* (2008) **456** 507-10. DOI: 10.1038/nature07450 71. Clarke TB, Davis KM, Lysenko ES, Zhou AY, Yu Y, Weiser JN. **Recognition of peptidoglycan from the microbiota by Nod1 enhances systemic innate immunity.**. *Nat Med* (2010) **16** 228-31. DOI: 10.1038/nm.2087 72. Kang DS, Yang YR, Lee C, Park B, Park KI, Seo JK, Seo YK, Cho H, Cocco L, Suh PG. **Netrin-1/DCC-mediated PLCγ1 activation is required for axon guidance and brain structure development.**. *EMBO Rep* (2019) **20** e48117. DOI: 10.15252/embr.201948117 73. Kim HY, Yang YR, Hwang H, Lee HE, Jang HJ, Kim J, Yang E, Kim H, Rhim H, Suh PG, Kim JI. **Deletion of PLCγ1 in GABAergic neurons increases seizure susceptibility in aged mice.**. *Sci Rep* (2019) **9** 17761. DOI: 10.1038/s41598-019-54477-4 74. Sundaram B, Kanneganti TD. **Advances in Understanding Activation and Function of the NLRC4 Inflammasome.**. *Int J Mol Sci* (2021) **22** 1048. DOI: 10.3390/ijms22031048 75. Mitchell PS, Roncaioli JL, Turcotte EA, Goers L, Chavez RA, Lee AY, Lesser CF, Rauch I, Vance RE. **NAIP-NLRC4-deficient mice are susceptible to shigellosis.**. *Elife* (2020) **9** e59022. DOI: 10.7554/eLife.59022 76. Carvalho FA, Nalbantoglu I, Aitken JD, Uchiyama R, Su Y, Doho GH, Vijay-Kumar M, Gewirtz AT. **Cytosolic flagellin receptor NLRC4 protects mice against mucosal and systemic challenges.**. *Mucosal Immunol* (2012) **5** 288-98. DOI: 10.1038/mi.2012.8 77. Sateriale A, Gullicksrud JA, Engiles JB, McLeod BI, Kugler EM, Henao-Mejia J, Zhou T, Ring AM, Brodsky IE, Hunter CA, Striepen B. **The intestinal parasite**. *Proc Natl Acad Sci U S A* (2021) **118** e2007807118. DOI: 10.1073/pnas.2007807118 78. Wang P, Zhu S, Yang L, Cui S, Pan W, Jackson R, Zheng Y, Rongvaux A, Sun Q, Yang G, Gao S, Lin R, You F. **Nlrp6 regulates intestinal antiviral innate immunity.**. *Science* (2015) **350** 826-30. DOI: 10.1126/science.aab3145 79. Liu Z, Li Y, Li C, Lei G, Zhou L, Chen X, Jia X, Lu Y. **Intestinal**. *Front Microbiol* (2022) **13** 812771. DOI: 10.3389/fmicb.2022.812771 80. Hara H, Seregin SS, Yang D, Fukase K, Chamaillard M, Alnemri ES, Inohara N, Chen GY, Núñez G. **The NLRP6 Inflammasome Recognizes Lipoteichoic Acid and Regulates Gram-Positive Pathogen Infection.**. *Cell* (2018) **175** 1651-64.e14. DOI: 10.1016/j.cell.2018.09.047 81. Rosenberg SA. **Progress in human tumour immunology and immunotherapy.**. *Nature* (2001) **411** 380-4. DOI: 10.1038/35077246 82. Hamilton PT, Anholt BR, Nelson BH. **Tumour immunotherapy: lessons from predator-prey theory.**. *Nat Rev Immunol* (2022) **22** 765-75. DOI: 10.1038/s41577-022-00719-y 83. Zhu AX, Finn RS, Edeline J, Cattan S, Ogasawara S, Palmer D, Verslype C, Zagonel V, Fartoux L, Vogel A, Sarker D, Verset G, Chan SL. **Pembrolizumab in patients with advanced hepatocellular carcinoma previously treated with sorafenib (KEYNOTE-224): a non-randomised, open-label phase 2 trial.**. *Lancet Oncol* (2018) **19** 940-52. DOI: 10.1016/S1470-2045(18)30351-6 84. Macek Jilkova Z, Aspord C, Decaens T. **Predictive Factors for Response to PD-1/PD-L1 Checkpoint Inhibition in the Field of Hepatocellular Carcinoma: Current Status and Challenges.**. *Cancers (Basel)* (2019) **11** 1554. DOI: 10.3390/cancers11101554 85. Fan Y, Hou J, Liu X, Han B, Meng Y, Liu B, Chen F, Shang Y, Cao P, Tan K. **Integrated Bioinformatics Analysis Identifies Heat Shock Factor 2 as a Prognostic Biomarker Associated With Immune Cell Infiltration in Hepatocellular Carcinoma.**. *Front Genet* (2021) **12** 668516. DOI: 10.3389/fgene.2021.668516 86. Xu D, Ji Z, Qiang L. **Molecular Characteristics, Clinical Implication, and Cancer Immunity Interactions of Pyroptosis-Related Genes in Breast Cancer.**. *Front Med (Lausanne)* (2021) **8** 702638. DOI: 10.3389/fmed.2021.702638 87. Liu X, Liu B, Shang Y, Cao P, Hou J, Chen F, Zhang B, Fan Y, Tan K. **Decreased TMPRSS2 expression by SARS-CoV-2 predicts the poor prognosis of lung cancer patients through metabolic pathways and immune infiltration.**. *Aging (Albany NY)* (2022) **14** 73-108. DOI: 10.18632/aging.203823
--- title: ANGPTL2 promotes VEGF-A synthesis in human lung cancer and facilitates lymphangiogenesis authors: - Po-I Liu - Ya-Jing Jiang - An-Chen Chang - Chang-Lun Huang - Yi-Chin Fong - Jeng-Hung Guo - Chun-Lin Liu - Shih-Wei Wang - Ju-Fang Liu - Sunny Li-Yu Chang - Chih-Hsin Tang journal: Aging (Albany NY) year: 2023 pmcid: PMC10042695 doi: 10.18632/aging.204581 license: CC BY 3.0 --- # ANGPTL2 promotes VEGF-A synthesis in human lung cancer and facilitates lymphangiogenesis ## Abstract Lung cancer is an extremely common cancer and metastatic lung cancer has a greatly low survival rate. Lymphangiogenesis is essential for the development and metastasis of lung cancer. The adipokine angiopoietin-like protein 2 (ANGPTL2) regulates tumor progression and metastasis, although the functions of ANGPTL2 in lung cancer are unknown. Analysis of data from TCGA genomics program, the GEPIA web server and the Oncomine database revealed that higher levels of ANGPTL2 expression were correlated with progressive disease and lymph node metastasis. ANGPTL2 enhanced VEGF-A-dependent lymphatic endothelial cell (LEC) tube formation and migration. Integrin α5β1, p38 and nuclear factor (NF)-κB signaling mediated ANGPTL2-regulated lymphangiogenesis. Importantly, overexpression ANGPTL2 facilitated tumor growth and lymphangiogenesis in vivo. Thus, ANGPTL2 is a promising therapeutic object for treating lung cancer. ## INTRODUCTION Lung cancer is one of the extremely common cancers worldwide and the resulting cause of cancer death [1]. The majority of newly diagnosed patients have locally advanced or metastatic disorder, and many who undergo surgery for early-stage disease experience recurrence within the first 5 years postoperatively [2]. Lung cancer metastasis involves several processes; the establishment of hypoxia, the formation of lymphangiogenesis and angiogenesis, cancer cell migratory and invasive activities, and the appearance of distant metastasis [3–5]. In particular, lymphangiogenesis enables lymphatic endothelial cells (LECs) to form lymphatic tubes around cancers, facilitating the invasion of cancer cells into the lymph node [6, 7]. Vascular endothelial growth factor (VEGF) reportedly enhances LEC-mediated lymphangiogenesis and lung cancer metastasis [8, 9]. Tumor-secreted VEGF family proteins, such as VEGF-A, -C and -D, are critical mediators in the regulation of LEC proliferation and promotion of lymphangiogenesis [6, 7]. Levels of VEGF production are higher in lung cancer patients than in normal healthy controls [10, 11]. Thus, it is critical to investigate the mechanisms underlying VEGF overexpression and VEGF-induced promotion of LEC-mediated lymphangiogenesis and lung cancer metastasis. Adipokines, proteins secreted by adipocyte tissue, facilitate tumorigenesis and distant metastasis in different types of cancers [12–14]. The adipokine angiopoietin-like protein 2 (ANGPTL2) is member of the angiopoietin-like family and acts as a growth factor of vascular endothelium [15]. ANGPTL2 is greatly produced in adipose tissue and obese mice exhibit upregulated levels of ANGPTL2 mRNA and circulating protein [16]. Abnormally high levels of ANGPTL2 have been found in lung cancer cells and serum from patients with colorectal or gastric cancer [17–19], with evidence showing that upregulated ANGPTL2 expression promotes the growth, drug resistance and metastasis of colorectal cancer [18, 20]. In addition, ANGPTL2 has reported to enhance the progression and metastasis of lung cancer [21–23]. However, the regulating effects of ANGPTL2 upon lymphangiogenesis remain unknown in lung cancer. In current report, we found higher levels of ANGPTL2 and LYVE-1 (a LEC marker) expression in lung cancer tissue than in normal control tissue. We also indicate that ANGPTL2 promotes VEGF-A-dependent lymphangiogenesis via integrin α5β1, p38 MAP kinase (MAPK) and NF-κB signaling, indicating that ANGPTL2 may be worth targeting when applying tumor-associated lymphangiogenesis. ## Materials ANGPTL2 (ab199133) antibody was purchased from Abcam (Cambridge, MA, USA), VEGF-A (A17877) was purchased from Abclonal (Cambridge, MA, USA), p-p65 [3033] antibody was purchased from Cell Signaling Technology (Danvers, MA, USA), β-actin (GT5512) and p65 (GTX102090) antibodies were purchased from GeneTex (Hsinchu, Taiwan). Small interfering RNAs (siRNAs) against integrin α5 (sc-29372) and β1(sc-35674), and immunoglobulin (Ig)-like transcript 4 (ILT4) (sc-45200) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). VEGF-A recombinant protein was acquired from PeproTech (Rocky Hill, NJ, USA). The p38 MAPK inhibitor SB203580 (HY-10256) and NF-κB inhibitor PDTC (P8765-1G) was acquired from Sigma-Aldrich (St. Louis, MO, USA). The ANGPTL2 overexpression plasmid (ANGPTL2 cDNA) was commercially synthesized by NCFB (Academia Sinica, Taiwan). ANGPTL2 (clone ID: TRCN0000158500) knockdown (sh-ANGPTL2) was purchased from the National RNAi Core Facility (RNAi Core, Academia Sinica, Taiwan). ## Cell culture The human lung adenocarcinoma cell line A549 was purchased from the American Type Culture Collection (ATCC; Manassas, VA, USA). CL1-0 and CL1-5 cell lines were provided by Dr. Shun-Fa Yang (Chung Shan Medical University, Taiwan). Cells were cultured in DMEM medium containing streptomycin (100 μg/mL), penicillin (100 U/mL) and $10\%$ FBS and maintained at 37° C and $5\%$ CO2. The human LEC cell line was purchased from Lonza (Walkersville, MD, USA). Cells were grown in EGM™-2 MV BulletKit™ Medium consisting of basal medium plus the EBM™-2 MV SingleQuot™ Kit (Lonza; Walkersville, MD, USA). Cells were seeded onto $1\%$ gelatin-coated plastic ware and cultured at 37° C and $5\%$ CO2. ## Quantification of ANGPTL2 and LYVE-1 expression Levels of ANGPTL2 and LYVE-1 expression in normal and tumor human tissue specimens were quantified using The Cancer Genome Atlas (TCGA) genomics program, the Gene Expression Profiling Interactive Analysis (GEPIA) web server and the Oncomine database [24, 25]. A total of 11 lung adenocarcinoma tissue samples from the Gene Expression Omnibus (GEO) database (GDS$\frac{4402}{1431848}$_at) were analyzed for levels of the ANGPTL2 gene in lymph node metastatic tumor cells ($$n = 8$$) and primary lung tumor cells ($$n = 3$$). The online database Kaplan–Meier Plotter (http://www.kmplot.com) was used to examine the association between ANGPTL2 expression and overall survival in human lung cancer. ## Collection of lung cancer conditioned medium Lung cancers were pretreated or transfected with the pharmacological inhibitors or genetic siRNAs, then treated with ANGPTL2. The medium was collected as conditioned medium (CM) and stored at −80° C until use. ## Measurement of LEC migration LEC migratory activity was evaluated using Transwell inserts in 24-well dishes (Costar, NY, USA), as according to our previous research [26]. Migratory cells were imaged under ×200 magnification using an Eclipse Ti2 microscope (Nikon, Tokyo, Japan). ## Measurement of LEC tube formation LECs (3 × 105 cells) were cultured in $50\%$ EGM™-2MV medium and $50\%$ lung cancer CM, then applied to plates precoated with Matrigel. LEC tube formation was photographed after 6 h and the number of tube branches was counted manually [27]. ## Transient transfection and luciferase assays Lung cancer cells were transfected with integrin α5, β1 and ILT4 siRNA or nuclear factor-κB (NF-κB) luciferase plasmid (Stratagene; St. Louis, MO, USA) using Lipofectamine 2000, then treated with ANGPTL2. The Dual-luciferase® Reporter Assay System was performed to analyze luciferase activity. ## Establishment of ANGPTL2 knockdown CL1-5 cells and overexpression CL1-0 cells To establish ANGPTL2 knockdown CL1-5 cells and overexpression CL1-0 cells, a lentivirus was prepared according to a standard protocol [28]. For infection, CL1-5 and CL1-0 cells were seeded in a 6-well dish and the lentivirus was added to the medium (multiplicity of infection = 10). After 24 h, the culture medium was changed and then at 48 h, 2 μg/mL of puromycin was added to select for ANGPTL2 knockdown and overexpression cells. ## Immunohistochemistry (IHC) staining The tissues were stained with VEGF-A or LYVE-1 antibodies and quantified according to the protocol described in our previous work [29, 30]. The sum of the intensity and percentage scores was used as the final staining score, as described previously [25]. ## Statistical analysis All values are expressed as the mean ± standard deviation (S.D.). Statistical differences between the experimental groups were assessed for significance using the Student’s t-test. Statistical comparisons of more than two groups were performed using one-way analysis of variance (ANOVA) with the Bonferroni post hoc test. Between-group differences were considered to be significant if the p-value was less than 0.05. Materials and Methods relating to western blot assay, reverse transcription-quantitative PCR (RT-qPCR) assay and tumor xenograft study are all obtainable within Supplementary Information. ## Data availability statement The data sets used and analyzed during the current study are available from the corresponding author on reasonable request. ## ANGPTL2 is highly expressed in patients with progressive lung cancer disease and lymph node metastasis Adipokines can promote cancer progression and metastasis [12–14]. To examine the effects of adipokine levels upon lung cancer progression, we investigated the clinical significance 13 of adipokines identified in lung adenocarcinoma samples from the TCGA database. We found higher levels of ANGPTL2 mRNA expression in tumor tissue than in adjacent normal tissue (Figure 1). Conversely, levels of apelin, CCL2, progranulin, interleukin-6, chemerin, retinol binding protein 4 (RBP4), resistin, and plasminogen activator inhibitor-1 (PAI-1) expression were lower in tumor tissue than in adjacent normal tissue (Figure 1). Thus, ANGPTL2 is a more important adipokine than others in lung cancer progression. Similarly, IHC data confirmed upregulated expression of ANGPTL2 in lung cancer tissue (Figure 2A, 2B), while TCGA data revealed significant associations between high levels of ANGPTL2 expression and regional lymph node metastasis (Figure 2C). Furthermore, higher levels of ANGPTL2 expression in patients with lung adenocarcinoma were associated with lower survival (Figure 2D). **Figure 1:** *Adipokine levels in lung cancer tissue and normal healthy samples. (A–M) Adipokine mRNA expression in human lung cancer tissue and adjacent normal tissue was analyzed in records from the TCGA database. *p < 0.05 compared with normal tissue.* **Figure 2:** *Levels of ANGPTL2 expression correlate with clinicopathologic features of lung adenocarcinoma tissue infiltrated by lymphatic vessels. (A, B) ANGPTL2 expression in human lung cancer tissue and adjacent normal tissue samples was analyzed by IHC staining. (C) The association between ANGPTL2 expression and regional lymph node metastasis was analyzed in samples from the TCGA database. (D) Associations between ANGPTL2 expression and overall survival rates of lung cancer patients were analyzed using the Kaplan-Meier Plotter database. (E, F) LYVE-1 expression in human lung cancer tissue and adjacent normal tissue samples was analyzed by IHC staining. (G, H) Data obtained from the GEO database (GDS4402/1431848_at) were analyzed for ANGPTL2 expression in primary lung tumor tissue and lung cancer with lymph node metastasis tissue samples. *p < 0.05 compared with normal tissue.* Cancer cells increase lymphatic vessel density in the tumor microenvironment and in lymph nodes by secreting lymphangiogenic factors, thereby promoting lymph node metastasis and poor prognosis [31, 32]. To characterize lymphatic vessels in lung tumors, we used an IHC assay with LYVE-1 antibody. This detected intratumor lymphatic vessels in the tumor peripheral region of lung cancer patients and none in tissue from normal healthy controls (Figure 2E, 2F). Records from the Gene Expression Omnibus (GEO) database (GDS$\frac{4402}{1431848}$_at) showed higher ANGPTL2 levels in lymph node metastatic tumor cells than in primary lung tumor cells from animal lung adenocarcinoma tissue (Figure 2G, 2H). Our results indicate that high levels of ANGPTL2 are positively associated with poor survival and lymph node metastasis in lung cancer. ## ANGPTL2 facilitates VEGF-A-dependent LEC tube formation and migration To examine the effects of ANGPTL2 in lung cancer lymph node metastasis, we measured basal migratory activities of lung cancer cell lines CL1-0, CL1-5, and A549. CL1-5 displayed higher migratory ability than A549 and CL1-0 (Figure 3A). Levels of ANGPTL2 protein and mRNA expression were also higher in CL1-5 cells than in A549 and CL1-0 cells (Figure 3B, 3C), implying that ANGPTL2 level is associated with migratory ability in lung cancer cells. After transfecting CL1-0 cells with ANGPTL2 cDNA, ANGPTL2 expression increased (Figure 3D). In vitro LEC tube formation and migration is a well-established model for mimicking lymphangiogenesis [7]. CM from ANGPTL2 cDNA-transfected lung cancer cells markedly facilitated tube formation and migration in LECs (Figure 3E, 3F), while transfecting CL1-5 and A549 cells with ANGPTL2 shRNA reduced ANGPTL2 expression, LEC tube formation and migration (Figure 3D, 3G–3J), indicating that ANGPTL2 promotes lymphangiogenesis in lung cancer cells. **Figure 3:** *ANGPTL2 levels correlate with lung cancer cell migratory activity and facilitate LEC tube formation and migration. (A) The migratory ability of human lung cancer cell lines (CL1-0, CL1-5 and A549) was measured by the Transwell assay. (B, C) ANGPTL2 mRNA and protein expression in lung cancer cell lines was examined by Western blot (n=3) and qPCR. (D) Cells were transfected with ANGPTL2 cDNA or shRNA, then ANGPTL2 expression was measured by Western blot (n=3). *p < 0.05 compared with CL1-0 cells. (E–J) Cells were transfected with ANGPTL2 cDNA or shRNA. The CM was collected and applied to the LECs. LEC tube formation and migration was examined. *p < 0.05 compared with controls.* We next investigated lymphangiogenic factors in ANGPTL2-mediated effects. Records from the TCGA database revealed that VEGF-A, VEGF-C, hepatocyte growth factor (HGF) and fibroblast growth factor-2 (FGF-2), but not C-fos-induced growth factor (FIGF), were positively correlated with ANGPTL2 expression (Figure 4A–4E). Overexpression and knockdown of ANGPTL2 markedly regulated VEGF-A synthesis but not that of other growth factors (Figure 4F–4H). Antibody to VEGF-A, but not to VEGF-C, antagonized ANGPTL-2-induced LEC lymphangiogenesis (Figure 4I, 4J and Supplementary Figure 1). Notably, treatment with VEGF-A reversed ANGPTL-2 shRNA-induced reductions in LEC tube formation and migration (Figure 4K–4N), implying that ANGPTL2 facilitates VEGF-A-dependent lymphangiogenesis in lung cancer cells. **Figure 4:** *ANGPTL2 promotes VEGF-A-dependent LEC tube formation and migration. (A–E) Associations between ANGPTL2 expression and lymphangiogenic factors (VEGF-A, VEGF-C, FIGF, HGF and FGF2) were analyzed using data from the TCGA database. (F–H) CL1-0 cells were transfected with ANGPTL2 cDNA; CL1-5 and A549 cells were transfected with ANGPTL2 shRNA. Levels of mRNA expression were examined by qPCR. *p < 0.05 compared with controls. (I, J) CL1-0 cells were transfected with ANGPTL2 cDNA. The CM was collected and applied to the LECs with VEGF-A or VEGF-C antibody. LEC tube formation and migration was examined. (K–N) Cells were transfected with ANGPTL2 shRNA. The CM was collected and applied to the LECs with VEGF-A. LEC tube formation and migration was examined. *p < 0.05 compared with CL1-0 CM; #p < 0.05 compared with Control; #p < 0.05 compared with CL1-5 or A549 with ANGPTL2 shRNA CM.* ## Integrin α5β1, p38 and NF-κB signaling pathway is mediated ANGPTL2-facilitated VEGF-A expression Leukocyte immunoglobulin-like receptor B2 (LILRB2) and integrin α5β1 are major functional receptors of ANGPTL2 [15]. Transfection of CL1-0 cells with integrin α5β1, but not LILRB2 siRNA, antagonized ANGPTL2-induced increases in VEGF-A expression (Figure 5A, 5B). The p38/NF-κB pathway regulates ANGPTL2-mediated cellular functions [16]. Treatment with p38 (SB203580) and NF-κB (PDTC) inhibitors both prevented ANGPTL2-induced increases in VEGF-A synthesis (Figure 5C). Integrin α5β1 siRNA and SB203580 reduced ANGPTL2-induced promotion of p65 phosphorylation and NF-κB luciferase activity (Figure 5D, 5E). Thus, ANGPTL2 enhances VEGF-A synthesis in lung cancer cells via integrin α5β1, p38 and NF-κB signaling. **Figure 5:** *ANGPTL2 increases VEGF-A synthesis in lung cancer cells via the integrin α5β1 receptor, p38 and NF-κB signaling. (A–C) Cells were transfected with ANGPTL2 cDNA, then stimulated with LILRB2 and integrin α5β1 siRNA or SB203580 and PDTC; VEGF-A expression was measured by Western blot (n=3). Quantitative data of the protein level were obtained using ImageJ software. Densitometric analysis of protein expression was normalized to β-actin. (D, E) Cells were transfected with ANGPTL2 cDNA, then stimulated with integrin α5β1 siRNA and SB203580; p38 and NF-κB activation was measured by Western blot (n=3) and NF-κB luciferase activity. Quantitative data of the protein level were obtained using ImageJ software. Densitometric analysis of protein expression was normalized to β-actin. *p < 0.05 compared with CL1-0; #p < 0.05 compared with Control.* ## ANGPTL2 promotes lymphangiogenesis in vivo Next, we examined whether ANGPTL2 regulates lung cancer lymphangiogenesis in tumor xenograft mouse models. Overexpression of ANGPTL2 in CL1-0 cells increased tumor growth (Figure 6A–6C). IHC staining demonstrated that ANGPTL2 overexpression significantly increased the expression of ANGPTL2, LYVE-1 and VEGF-A (Figure 6D). In contrast, ANGPTL2 blockade suppressed tumor growth and the levels of ANGPTL2, LYVE-1 and VEGF-A (Figure 6). These results support the targeting of ANGPTL-2 for regulating tumor growth and lymphangiogenesis in lung cancer. **Figure 6:** *ANGPTL2 promotes tumor growth and lymphangiogenesis in vivo. (A–C) CL1-0 and CL1-5 cells were subcutaneously injected into the right flanks of BALB/c-nu mice. Four weeks later, the mice were sacrificed and the tumors were excised and weighed (n=5). (D) Tumor sections were immunostained using ANGPTL2, LYVE-1 and VEGF-A antibodies. *p < 0.05 compared with controls.* ## DISCUSSION It is well established that advanced lung cancer is extremely aggressive and associated with metastasis [33]. Lymphangiogenesis is essential for the development and metastasis of lung cancer [34]. Understanding the underlying mechanisms of lung cancer lymphangiogenesis may assist with the development of novel treatment approaches for lung cancer, which has a low overall five-year survival rate, despite improvements in surgical management and therapeutic combinations of radiation and chemotherapy. The adipokine ANGPTL2 is associated with cancer progression and metastasis [21, 22]. High levels of ANGPTL2 have been found in lung and gastric cancer patients [17–19]. In addition, ANGPTL2 enhanced the progression and drug resistance in colorectal cancer [18, 20]. We report finding high levels of ANGPTL2 expression in human specimens of lymph node metastatic lung cancer. Upregulation of ANGPTL2 indicates a poor prognosis with reduced survival in lung cancer patients. Our findings show that ANGPTL2 facilitates VEGF-A-dependent LEC lymphangiogenesis in lung cancer cells and that the integrin α5β1, p38 and NF-κB signaling cascade is involved in ANGPTL2-enhanced promotion of VEGF-A synthesis. Importantly, ANGPTL2 promoted lung cancer growth and lymphangiogenesis in vivo. Thus, ANGPTL2 is a promising therapeutic target for treating lung cancer progression and metastasis. VEGF-A, VEGF-C and VEGF-D are major lymphangiogenic factors that regulate lymphangiogenic progression during tumor development and metastasis [6, 7, 35]. Amongst these growth factors, VEGF-C is strongly associated with the control of lymphatic vessel invasion and lymphangiogenesis [36]. In this study, overexpression or knockdown of ANGPTL2 significantly regulated VEGF-A production but not that of VEGF-C, HGF, or FGF-2. ANGPTL2-induced facilitation of LEC tube formation and migration was antagonized by treatment with VEGF-A antibody, but not VEGF-C antibody. Recombinant VEGF-A rescued the inhibition of LEC lymphangiogenesis following ANGPTL2 blockade. Our mouse xenograft model revealed that ANGPTL2 overexpression promotes the upregulation of LYVE-1 and VEGF-A expression in lung cancer tissue. Thus, ANGPTL2 promotes VEGF-A-dependent LEC lymphangiogenesis. Similarly, VEGF-A reportedly mediates simvastatin-induced regulation of tumor lymphangiogenesis and lymph node metastasis [37]; recombinant canstatin suppresses VEGF-A-mediated lymphangiogenesis in an animal model of oral squamous cell carcinoma [38]. Thus, VEGF-A is a critical regulator of lymphangiogenesis during tumor growth and metastasis. A previous report has indicated that ANGPTL2-induced regulation of cellular functions occurs via LILRB2 and the integrin α5β1 receptor [15]. When this study used siRNAs against LILRB2 and integrin α5β1, we found that integrin α5β1 siRNA but not LILRB2 siRNA antagonized ANGPTL2-facilitated VEGF-A production, suggesting that the integrin α5β1 receptor controls ANGPTL2-enhanced promotion of VEGF-A synthesis and lymphangiogenesis. It is known that p38 activation is crucial for regulating different cellular events [39], such as the promotion of lymphangiogenesis and metastasis [39, 40, 41]. Our data found that ANGPTL2 enhances p38 phosphorylation, while the p38 inhibitor reversed ANGPTL2-regulated VEGF-A production. NF-κB signaling is an important downstream molecule (or mediator) of p38 in the regulation of ANGPTL2-induced inflammatory responses [16]. Our data showed that the NF-κB inhibitor antagonized ANGPTL2-mediated VEGF-A synthesis in lung cancer cells. Our results also showed that ANGPTL2 enhances p65 phosphorylation and NF-κB luciferase activity, which was reduced by the integrin α5β1 siRNA and p38 inhibitor, indicating that integrin α5β1 receptor-dependent p38/NF-κB activation regulates ANGPTL2-induced mediation of VEGF-A expression and lymphangiogenesis in human lung cancer cells. The limitations should be noted in this study. The study results could have been strengthened statistically by using more lung cancer patient tissues. In addition, the more detail clinicopathologic data (age, sex and pathology diagnoses) should be enrolled. Secondly, although our data strongly suggest that ANGPTL2 promotes VEGF-A-mediated lymphangiogenesis in human lung cancer cells, we cannot exclude the possibility that ANGPTL2 also promotes the activities of other angiogenetic factors. In summary, our study indicates that ANGPTL2 promotes VEGF-A synthesis in human lung cancer cells and subsequently facilitates LEC lymphangiogenesis via the integrin α5β1 receptor, p38 and NF-κB signaling (Figure 7). These results improve our understanding about how ANGPTL2-mediated VEGF-A synthesis contributes to lymphangiogenesis, which may help scientists design more effective therapy for metastatic lung cancer. **Figure 7:** *Schematic diagram summarizes the mechanisms by which ANGPTL2 facilitates lymphangiogenesis in human lung cancer cells. ANGPTL2 increases VEGF-A production and subsequently facilitates LEC lymphangiogenesis in human lung cancer cells via the integrin α5β1 receptor, p38 and NF-κB signaling.* ## References 1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, Bray F. **Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries.**. *CA Cancer J Clin* (2021) **71** 209-49. DOI: 10.3322/caac.21660 2. Uramoto H, Tanaka F. **Recurrence after surgery in patients with NSCLC.**. *Transl Lung Cancer Res* (2014) **3** 242-9. DOI: 10.3978/j.issn.2218-6751.2013.12.05 3. Paduch R. **The role of lymphangiogenesis and angiogenesis in tumor metastasis.**. *Cell Oncol (Dordr)* (2016) **39** 397-410. DOI: 10.1007/s13402-016-0281-9 4. Lin CC, Chen KB, Tsai CH, Tsai FJ, Huang CY, Tang CH, Yang JS, Hsu YM, Peng SF, Chung JG. **Casticin inhibits human prostate cancer DU 145 cell migration and invasion via Ras/Akt/NF-κB signaling pathways.**. *J Food Biochem* (2019) **43** e12902. DOI: 10.1111/jfbc.12902 5. Chen YF, Chang WH, Tsai HY, Lee MM, Wu TS. **Bioactivities and action mechanisms of active compounds from**. *Biomedicine (Taipei)* (2021) **11** 40-6. DOI: 10.37796/2211-8039.1219 6. Kataru RP, Kim H, Jang C, Choi DK, Koh BI, Kim M, Gollamudi S, Kim YK, Lee SH, Koh GY. **T lymphocytes negatively regulate lymph node lymphatic vessel formation.**. *Immunity* (2011) **34** 96-107. DOI: 10.1016/j.immuni.2010.12.016 7. Stacker SA, Williams SP, Karnezis T, Shayan R, Fox SB, Achen MG. **Lymphangiogenesis and lymphatic vessel remodelling in cancer.**. *Nat Rev Cancer* (2014) **14** 159-72. DOI: 10.1038/nrc3677 8. Nishino K, Yoshimatsu Y, Muramatsu T, Sekimoto Y, Mitani K, Kobayashi E, Okamoto S, Ebana H, Okada Y, Kurihara M, Suzuki K, Inazawa J, Takahashi K. **Isolation and characterisation of lymphatic endothelial cells from lung tissues affected by lymphangioleiomyomatosis.**. *Sci Rep* (2021) **11** 8406. DOI: 10.1038/s41598-021-88064-3 9. Berta J, Török S, Tárnoki-Zách J, Drozdovszky O, Tóvári J, Paku S, Kovács I, Czirók A, Masri B, Megyesfalvi Z, Oskolás H, Malm J, Ingvar C. **Apelin promotes blood and lymph vessel formation and the growth of melanoma lung metastasis.**. *Sci Rep* (2021) **11** 5798. DOI: 10.1038/s41598-021-85162-0 10. Lazzari C, Bulotta A, Damiano G, Mirabile A, Viganó M, Veronesi G, Gregorc V. **Angiogenesis inhibition in lung cancer: emerging novel strategies.**. *Curr Opin Oncol* (2022) **34** 107-14. DOI: 10.1097/CCO.0000000000000807 11. Tian W, Cao C, Shu L, Wu F. **Anti-Angiogenic Therapy in the Treatment of Non-Small Cell Lung Cancer.**. *Onco Targets Ther* (2020) **13** 12113-29. DOI: 10.2147/OTT.S276150 12. Barb D, Neuwirth A, Mantzoros CS, Balk SP. **Adiponectin signals in prostate cancer cells through Akt to activate the mammalian target of rapamycin pathway.**. *Endocr Relat Cancer* (2007) **14** 995-1005. DOI: 10.1677/ERC-06-0091 13. Cascio S, Bartella V, Auriemma A, Johannes GJ, Russo A, Giordano A, Surmacz E. **Mechanism of leptin expression in breast cancer cells: role of hypoxia-inducible factor-1alpha.**. *Oncogene* (2008) **27** 540-7. DOI: 10.1038/sj.onc.1210660 14. Yunusova NV, Kondakova IV, Kolomiets LA, S G Af, Chernyshova AL, Shatokhina OV, Frolova AE, Zhou Z, Wang W. **[Serum adipokines and their receptors in endometrial and colon cancer patients: relationship with tumor invasion and metastasis].**. *Vopr Onkol* (2015) **61** 619-23. PMID: 26571833 15. Kadomatsu T, Endo M, Miyata K, Oike Y. **Diverse roles of ANGPTL2 in physiology and pathophysiology.**. *Trends Endocrinol Metab* (2014) **25** 245-54. DOI: 10.1016/j.tem.2014.03.012 16. Tabata M, Kadomatsu T, Fukuhara S, Miyata K, Ito Y, Endo M, Urano T, Zhu HJ, Tsukano H, Tazume H, Kaikita K, Miyashita K, Iwawaki T. **Angiopoietin-like protein 2 promotes chronic adipose tissue inflammation and obesity-related systemic insulin resistance.**. *Cell Metab* (2009) **10** 178-88. DOI: 10.1016/j.cmet.2009.08.003 17. Liu X, Yu X, Xie J, Zhan M, Yu Z, Xie L, Zeng H, Zhang F, Chen G, Yi X, Zheng J. **ANGPTL2/LILRB2 signaling promotes the propagation of lung cancer cells.**. *Oncotarget* (2015) **6** 21004-15. DOI: 10.18632/oncotarget.4217 18. Yoshinaga T, Shigemitsu T, Nishimata H, Kitazono M, Hori E, Tomiyoshi A, Takei T, Yoshida M. **Angiopoietin-like protein 2 as a potential biomarker for colorectal cancer.**. *Mol Clin Oncol* (2015) **3** 1080-4. DOI: 10.3892/mco.2015.577 19. Yoshinaga T, Shigemitsu T, Nishimata H, Takei T, Yoshida M. **Angiopoietin-like protein 2 is a potential biomarker for gastric cancer.**. *Mol Med Rep* (2015) **11** 2653-8. DOI: 10.3892/mmr.2014.3040 20. Toiyama Y, Tanaka K, Kitajima T, Shimura T, Kawamura M, Kawamoto A, Okugawa Y, Saigusa S, Hiro J, Inoue Y, Mohri Y, Goel A, Kusunoki M. **Elevated serum angiopoietin-like protein 2 correlates with the metastatic properties of colorectal cancer: a serum biomarker for early diagnosis and recurrence.**. *Clin Cancer Res* (2014) **20** 6175-86. DOI: 10.1158/1078-0432.CCR-14-0007 21. Odagiri H, Kadomatsu T, Endo M, Masuda T, Morioka MS, Fukuhara S, Miyamoto T, Kobayashi E, Miyata K, Aoi J, Horiguchi H, Nishimura N, Terada K. **The secreted protein ANGPTL2 promotes metastasis of osteosarcoma cells through integrin α5β1, p38 MAPK, and matrix metalloproteinases.**. *Sci Signal* (2014) **7** ra7. DOI: 10.1126/scisignal.2004612 22. Aoi J, Endo M, Kadomatsu T, Miyata K, Nakano M, Horiguchi H, Ogata A, Odagiri H, Yano M, Araki K, Jinnin M, Ito T, Hirakawa S. **Angiopoietin-like protein 2 is an important facilitator of inflammatory carcinogenesis and metastasis.**. *Cancer Res* (2011) **71** 7502-12. DOI: 10.1158/0008-5472.CAN-11-1758 23. Sasaki H, Suzuki A, Shitara M, Hikosaka Y, Okuda K, Moriyama S, Yano M, Fujii Y. **Angiopoietin-like protein ANGPTL2 gene expression is correlated with lymph node metastasis in lung cancer.**. *Oncol Lett* (2012) **4** 1325-8. DOI: 10.3892/ol.2012.924 24. Liu SC, Tsai CH, Wu TY, Tsai CH, Tsai FJ, Chung JG, Huang CY, Yang JS, Hsu YM, Yin MC, Wu YC, Tang CH. **Soya-cerebroside reduces IL-1 beta-induced MMP-1 production in chondrocytes and inhibits cartilage degradation: implications for the treatment of osteoarthritis.**. *Food Agr Immunol* (2019) **30** 620-32. DOI: 10.1080/09540105.2019.1611745 25. Achudhan D, Liu SC, Lin YY, Lee HP, Wang SW, Huang WC, Wu YC, Kuo YH, Tang CH. **Antcin K inhibits VEGF-dependent angiogenesis in human rheumatoid arthritis synovial fibroblasts.**. *J Food Biochem* (2022) **46** e14022. DOI: 10.1111/jfbc.14022 26. Chou KY, Chang AC, Ho CY, Tsai TF, Chen HE, Chen PC, Hwang TI. **Thrombospondin-4 promotes bladder cancer cell migration and invasion via MMP2 production.**. *J Cell Mol Med* (2021) **25** 6046-55. DOI: 10.1111/jcmm.16463 27. Lien MY, Tsai HC, Chang AC, Tsai MH, Hua CH, Wang SW, Tang CH. **Chemokine CCL4 Induces Vascular Endothelial Growth Factor C Expression and Lymphangiogenesis by miR-195-3p in Oral Squamous Cell Carcinoma.**. *Front Immunol* (2018) **9** 412. DOI: 10.3389/fimmu.2018.00412 28. Chen JC, Huang C, Lee IN, Wu YP, Tang CH. **Amphiregulin enhances cell migration and resistance to doxorubicin in chondrosarcoma cells through the MAPK pathway.**. *Mol Carcinog* (2018) **57** 1816-24. DOI: 10.1002/mc.22899 29. Su CH, Lin CY, Tsai CH, Lee HP, Lo LC, Huang WC, Wu YC, Hsieh CL, Tang CH. **Betulin suppresses TNF-α and IL-1β production in osteoarthritis synovial fibroblasts by inhibiting the MEK/ERK/NF-κB pathway.**. *J Funct Foods* (2021) **86** 104729. DOI: 10.1016/j.jff.2021.104729 30. Lee KT, Su CH, Liu SC, Chen BC, Chang JW, Tsai CH, Huang WC, Hsu CJ, Chen WC, Wu YC, Tang CH. **Cordycerebroside A inhibits ICAM-1-dependent M1 monocyte adhesion to osteoarthritis synovial fibroblasts.**. *J Food Biochem* (2022) **46** e14108. DOI: 10.1111/jfbc.14108 31. Skobe M, Hawighorst T, Jackson DG, Prevo R, Janes L, Velasco P, Riccardi L, Alitalo K, Claffey K, Detmar M. **Induction of tumor lymphangiogenesis by VEGF-C promotes breast cancer metastasis.**. *Nat Med* (2001) **7** 192-8. DOI: 10.1038/84643 32. Al-Rawi MA, Jiang WG. **Lymphangiogenesis and cancer metastasis.**. *Front Biosci (Landmark Ed)* (2011) **16** 723-39. DOI: 10.2741/3715 33. Nowroozi A, Khalili N, Razi S, Keshavarz-Fathi M, Rezaei N. **Tumor-infiltrating lymphocyte therapy for lung cancer and its future paradigms.**. *Expert Opin Biol Ther* (2022) **22** 735-45. DOI: 10.1080/14712598.2022.2072206 34. Rouzaut A, Irigoyen M, Montuenga LM. **Lymphangiogenesis and lung cancer.**. *J Thorac Oncol* (2007) **2** 384-6. DOI: 10.1097/01.JTO.0000268670.72249.ad 35. Lahdenranta J, Hagendoorn J, Padera TP, Hoshida T, Nelson G, Kashiwagi S, Jain RK, Fukumura D. **Endothelial nitric oxide synthase mediates lymphangiogenesis and lymphatic metastasis.**. *Cancer Res* (2009) **69** 2801-8. DOI: 10.1158/0008-5472.CAN-08-4051 36. Zhu T, Hu X, Wei P, Shan G. **Molecular background of the regional lymph node metastasis of gastric cancer.**. *Oncol Lett* (2018) **15** 3409-14. DOI: 10.3892/ol.2018.7813 37. Ji RC, Eshita Y, Kobayashi T, Hidano S, Kamiyama N, Onishi Y. **Role of simvastatin in tumor lymphangiogenesis and lymph node metastasis.**. *Clin Exp Metastasis* (2018) **35** 785-96. DOI: 10.1007/s10585-018-9940-8 38. Hwang-Bo J, Park JH, Bae MG, Chung IS. **Recombinant canstatin inhibits VEGF-A-induced lymphangiogenesis and metastasis in an oral squamous cell carcinoma SCC-VII animal model.**. *Cancer Med* (2016) **5** 2977-88. DOI: 10.1002/cam4.866 39. Kudaravalli S, den Hollander P, Mani SA. **Role of p38 MAP kinase in cancer stem cells and metastasis.**. *Oncogene* (2022) **41** 3177-85. DOI: 10.1038/s41388-022-02329-3 40. Lin CY, Wang SW, Guo JH, Tai HC, Sun WC, Lai CT, Yang CY, Liu SC, Fong YC, Tang CH. **WISP-3 Stimulates VEGF-C-Dependent Lymphangiogenesis in Human Chondrosarcoma Cells by Inhibiting miR-196a-3p Synthesis.**. *Biomedicines* (2021) **9** 1330. DOI: 10.3390/biomedicines9101330 41. Wang X, Jim Sheu JC, Lai MT, Cherry Chang YY, Sheng X, Wei L, Gao Y, Wang X, Liu N, Xie W, Chen CM, Wandy Ding Y, Sun L. **RSF-1 overexpression determines cancer progression and drug resistance in cervical cancer.**. *Biomedicine (Taipei)* (2018) **8** 4. DOI: 10.1051/bmdcn/2018080104
--- title: AAV1.NT-3 gene therapy prevents age-related sarcopenia authors: - Burcak Ozes - Lingying Tong - Morgan Myers - Kyle Moss - Alicia Ridgley - Zarife Sahenk journal: Aging (Albany NY) year: 2023 pmcid: PMC10042697 doi: 10.18632/aging.204577 license: CC BY 3.0 --- # AAV1.NT-3 gene therapy prevents age-related sarcopenia ## Abstract Sarcopenia is progressive loss of muscle mass and strength, occurring during normal aging with significant consequences on the quality of life for elderly. Neurotrophin 3 (NT-3) is an important autocrine factor supporting Schwann cell survival and differentiation and stimulating axon regeneration and myelination. NT-3 is involved in the maintenance of neuromuscular junction (NMJ) integrity, restoration of impaired radial growth of muscle fibers through activation of the Akt/mTOR pathway. We tested the efficacy of NT-3 gene transfer therapy in wild type (WT)-aged C57BL/6 mice, a model for natural aging and sarcopenia, via intramuscular injection 1 × 1011 vg AAV1.tMCK.NT-3, at 18 months of age. The treatment efficacy was assessed at 6 months post-injection using run to exhaustion and rotarod tests, in vivo muscle contractility assay, and histopathological studies of the peripheral nervous system, including NMJ connectivity and muscle. AAV1.NT-3 gene therapy in WT-aged C57BL/6 mice resulted in functional and in vivo muscle physiology improvements, supported by quantitative histology from muscle, peripheral nerves and NMJ. Hindlimb and forelimb muscles in the untreated cohort showed the presence of a muscle- and sex-dependent remodeling and fiber size decrease with aging, which was normalized toward values obtained from 10 months old WT mice with treatment. The molecular studies assessing the NT-3 effect on the oxidative state of distal hindlimb muscles, accompanied by western blot analyses for mTORC1 activation were in accordance with the histological findings. Considering the cost and quality of life to the individual, we believe our study has important implications for management of age-related sarcopenia. ## INTRODUCTION Sarcopenia is a common geriatric syndrome, defined as generalized, progressive loss of muscle mass and strength, occurring during normal aging with significant consequences on the quality of life for the elderly [1–4]. Age-related degradation of muscle leads to a slow but continuous loss in lean mass, starting after the age of 40 and thereafter is accelerated by age 70 [5, 6]. The prevalence of sarcopenia between age 60 to 70 is reported to be 5–$13\%$ but increases to 11–$50\%$ in people older than 80 [7, 8]. Sarcopenic patients face decreased strength and mobility, therefore are at higher risk for falls, bone fractures and increased mortality. Since the number and proportion of the global aging population is rapidly growing, preventing or delaying sarcopenia should have a great impact on the quality of life for elderly, as well as its socio-economic burden on individuals and society. Current strategies are mainly focused on exercise and nutrition, and despite some molecular pathways involved in aging have been suggested with potential for drug development, no specific medications have been developed with therapeutic impact in sarcopenia. Underlying pathophysiologic mechanisms contributing to sarcopenia are highly complex and multifactorial. In addition to behavioral or “extrinsic” factors, such as a sedentary lifestyle, the remaining contributing/causal factors appear directly related to the aging process of muscle to include mitochondrial dysfunction with oxidative damage, lysosomal dysfunction, decreased protein synthesis, decreased anabolic hormones, inflammaging/immune-senescence and satellite cells dysfunction (decreased number and regenerative capacity). Mitochondria, as the “power-house” of cells, is considered as a central player in overall aging process, and certainly in muscle going through sarcopenic change. In a recent study using neurotrophin 3 (NT-3) gene therapy, we demonstrated that NT-3 increases muscle fiber diameter through direct activation of mTORC1 pathway [9]. In addition, NT-3 was shown to improve hypomyelination state, neuromuscular junction (NMJ) connectivity, have anti-inflammatory, antioxidant, antiapoptotic properties, and enhances mitochondria biogenesis [10–14]. This suggests a potential application of NT-3 gene therapy for muscle wasting conditions including age-related sarcopenia. In this study, we used a triple muscle-specific creatine kinase (tMCK) promoter to restrict NT-3 expression to the skeletal muscle and self-complimentary adeno-associated virus serotype 1 (scAAV1) as vector to assess the therapeutic efficacy of AAV1.NT-3 in wild type-aged C57BL/6J mice, a model for natural aging and sarcopenia. Quantitative histopathologic parameters served to address age-related changes in muscle, peripheral nerve and NMJ. These changes include muscle fiber size and fiber type switch, myelin thickness and the innervated status of the NMJ. Functional studies and in vivo muscle physiology were used to assess the motor strength of the mice. The results show that AAV1.NT-3 gene therapy in the wild type C57BL/6J mice at 2 years of age, unequivocally improved the function of sarcopenic muscle, increased muscle fiber size, myelin thickness and NMJ connectivity. In addition, we found attenuation of age-related kyphosis and coat changes as well. ## rAAV.NT-3 vector production and potency scAAV1.tMCK.NT-3 design (Supplementary Figure 1A), and production followed previously described methods at Nationwide Children`s Hospital, Columbus [10]. scAAV1.tMCK.NT-3, at 1 × 1011 vg dose, was delivered to the gastrocnemius muscle of 18 months old C57BL/6 mice. Blood samples from terminally anesthetized treated and untreated mice were collected by cardiac puncture at six months post gene injection, and serum was assayed for NT-3 levels using a capture ELISA, as previously reported [10]. No detectable NT-3 levels were found in the untreated C57BL/6 mice at 2 years of age, in contrast to the treated cohort (Supplementary Figure 1B), showing significant improvements in functional and histologic outcome measures as described below. Similar to our previous observations in younger age groups, we found no effect of sex on serum NT-3 levels in the aged C57BL/6 mice. ## NT-3 gene therapy improved function, attenuated age-related musculoskeletal and skin changes in the aged C57BL/6 mouse Run to exhaustion treadmill test was used to assess the efficacy of NT-3 gene therapy on motor function at 2-, 4- and 6-months post gene delivery. AAV1.NT-3 improved treadmill performance, leading to remarkable increases in running distances and time until exhaustion compared to the untreated counterparts at all time points tested. We found a $63\%$ (NT-3, 135.8 ± 17.5, $$n = 12$$ vs. UT, 83.5 ± 11.1, $$n = 11$$; $$p \leq 0.0184$$), $73\%$ (NT-3, 180.6 ± 16.2, $$n = 12$$ vs. UT, 104.4 ± 4.8, $$n = 11$$; $$p \leq 0.0003$$) and $63.5\%$ (NT-3, 166.1 ± 13.8 $$n = 11$$, vs. UT, 101.6 ± 7.2, $$n = 8$$; $$p \leq 0.0077$$) increase in distance run at 2-, 4- and 6-months post gene delivery, respectively (Figure 1A). Treated females performed significantly better at 4 months post-injection ($$p \leq 0.0394$$) compared to treated males (Figure 1B), while there was no sex difference found in the untreated cohort (Figure 1C). We also assessed the efficacy of NT-3 gene therapy on motor coordination of aged C57BL/6 mice with rotarod test. NT-3 treated cohort displayed significantly better motor coordination with $39.2\%$ longer duration compared to untreated at endpoint (NT-3, 56.9 ± 5.9, $$n = 10$$ vs. UT, 40.9 ± 6.9, $$n = 8$$; $$p \leq 0.038$$, Figure 1D). Treated females performed better than males without reaching statistical significance (Supplementary Figure 2A). **Figure 1:** *Functional and in vivo muscle physiology improvements in 2-year-old C57BL/6 mice with AAV1.NT-3 gene transfer therapy. (A–C) Treadmill performance test performed at 2-, 4-, and 6-months post-injection (PI). (A) AAV1.NT-3 treated mice showed significant improvement at the time points tested (2 months PI, NT-3: 135.8 m, n = 12 vs. UT: 83.5 m, n = 11, p = 0.0184; 4 months PI, NT-3: 180.6 m, n = 12 vs. UT: 104.4 m, n = 11, p = 0.0003; 6 months PI, NT-3: 166.1 m, n = 11, vs. UT: 101.6 m, n = 8; p = 0.0077). (B) Treadmill performance of the female mice was better than the males in the treated cohort (2 months PI, F: 166.7 m, n = 6 vs. M: 104.9 m, n = 6; 4 months PI, F: 215.3 m, n = 6 vs. M: 145.9 m, n = 6, p = 0.0394; 6 months PI, F: 192.9 m, n = 6, vs. M: 133.9 m, n = 5) while no sex effect was observed in the (C) untreated cohort (2 months PI, F: 93.4 m, n = 5 vs. M: 75.3 m, n = 6; 4 months PI, F: 105.5 m, n = 5 vs. M: 103.5 m, n = 6; 6 months PI, F: 95.6 m, n = 3, vs. M: 105.2 m, n = 5). (D) AAV1.NT-3 treated mice showed significant improvement in the rotarod at end point (NT-3: 56.9 sec, n = 10 vs. UT: 40.9 sec, n = 8, p = 0.0374). In vivo muscle contractility assay showed a higher force output in (E) maximum twitch response in the treated cohort whereas (F) increase in the maximum tetanic measurement did not reach significance levels (Max twitch, NT-3: 2.22 mN*m, n = 11 vs. UT: 1.57 mN*m, n = 8, p = 0.0149; Max tetanic, NT-3: 11.09 mN*m, n = 10 vs. UT: 10.06 mN*m, n = 9). Data is represented as mean ± SEM; Two-way ANOVA, Sidak’s multiple comparisons test for (A–C) and student t-test for (D–F); *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.* In vivo muscle contractility assay performed in the gastrocnemius muscle of C57BL/6 mice at endpoint showed an increase in the maximum twitch response with NT-3 gene therapy (NT-3, 2.22 ± 0.2, $$n = 11$$ vs. UT, 1.57 ± 0.1, $$n = 8$$; $$p \leq 0.015$$, Figure 1E) while the maximum tetanic response did not improve significantly (NT-3, 11.1 ± 0.4, $$n = 10$$ vs. UT, 10.1 ± 0.6, $$n = 9$$; $$p \leq 0.142$$, Figure 1F). When treatment effect on genders was analyzed, both the maximum twitch and maximum tetanic responses were found to be higher in males, without reaching statistical significance (Supplementary Figure 2B, 2C). The treatment did not alter the endpoint weight of the treated cohort (NT-3, $$n = 11$$, 37.0 ± 1.95 g; UT, $$n = 10$$, 37.6 ± 1.94 g; $$p \leq 0.84$$, Supplementary Figure 2D). Two untreated mice, aged 84 and 100 weeks, and one treated mouse, aged 99 weeks, died before reaching to the end point of the experiment. We also documented improvements in age related musculoskeletal and skin changes in the NT-3 treated C57BL/6 mice including reduction of kyphosis, dermatitis, and alopecia (Supplementary Figure 3A). AAV1.NT-3 injected, and untreated cohorts were photographed at the end point, and the severity of kyphosis, dermatitis, and alopecia were assessed using a semiquantitative scoring system (0: none, 0.5: mild-moderate, 1: severe). Varying degrees of kyphosis were present in $83\%$ of untreated mice, half of which were severe, while only $50\%$ of the AAV1.NT-3 injected mice had kyphosis and one third were severe. Dermatitis, observed in $33\%$ of untreated mice, was not seen in the treated group. Moderate to severe alopecia around the ears was present in $67\%$ of the untreated mice, whereas this ratio decreased to $33\%$ percent in treated mice (Supplementary Figure 3B–3D). ## Reversal of age-related neuromuscular histopathology with NT-3 gene therapy We examined the efficacy of NT-3 gene therapy upon muscle fiber size and fiber type composition in tibialis anterior, gastrocnemius, quadriceps, and triceps muscles at 6 months post gene injection. Succinic dehydrogenase stain was used for quantification [15], which delineates fibers according to mitochondria content: fatigue-resistant slow-twitch oxidative (STO) fibers with darkest staining, the fast-twitch oxidative (FTO) with intermediate staining and the fast-twitch glycolytic (FTG) with lightest staining intensity (Figure 2A). Fiber size measurements were done on the representative images from deep, intermediate, and superficial zones of untreated and NT-3-treated samples as illustrated in the tibialis anterior and gastrocnemius muscles [15] (Figure 2B, 2C). Both muscles from the untreated cohort showed atrophic angular fibers with the majority belonging to fast twitch type 2 fibers, typical of sarcopenic muscle (Figure 2A). Compared to younger 10-month-old C57BL/6 mice, the mean muscle fiber size of all fiber types from the hindlimb muscles was smaller at 2 years of age, corroborating previous reports [16] and an overall fiber size increase was evident with treatment. Fiber type diameters (mean ± SEM μm; derived from the mean of measurements individually made in each mouse) from untreated and treated 2-year-old C57BL/6 mice of both sexes as combined, and separately for males and females, along with inclusion of data from the same strain at 10 months of age in all 4 muscles are detailed in Supplementary Figures 4–7 and Supplementary Tables 1–12. **Figure 2:** *Muscle fiber size increase in aged C57BL/6 mice with AAV1.NT-3 gene therapy. (A) Succinic dehydrogenase (SDH)-stained skeletal muscle showing muscle fiber types based on mitochondria content and angular atrophic type 2 fibers (arrows). Dark (D) fibers are fatigue-resistant slow twitch oxidative (STO) or type1 fibers, intermediate (I)-stained fibers are fast twitch oxidative (FTO/type 2A) fibers, and light (L) fibers are fast twitch glycolytic (FTG/type 2B). Representative images from (B) tibialis anterior and (C) gastrocnemius muscles showing three different zones (deep, intermediate, and superficial; designated as Zone 1, 2, and 3) from the untreated (UT) and NT-3 treated cohorts. Scale bar = 30 μm.* With AAV1.NT-3 gene therapy, we observed an overall reversal of the sarcopenic effect on fiber size and alterations in fiber type composition of all muscles examined. Changes in fiber type distribution of STO, FTO, and FTG (as a percent of total) and percent loss as “sarcopenic” (compared to 10-month-old) and percent gain as “NT-3 effect” in the glycolytic (FTG) and oxidative (STO) fiber size are shown in a summary format in Figure 3. Interestingly, alterations in fiber type composition showed different patterns for anterior and posterior compartment muscles of the distal hindlimb. In the tibialis anterior muscle from males, there was a reversal of sarcopenia-related decline in STO fibers, with a notable percentage increase with treatment, along with a decrease in FTG fiber type; suggesting a switch from FTG to STO fiber type. This glycolytic to oxidative fiber type switch resulted in a significant increase in STO fiber size (from $19.76\%$ loss to $15.03\%$ increase). In female tibialis anterior muscle however, an opposite pattern, from oxidative to glycolytic fiber type switch was noted, which was associated with prominent FTG fiber size increase (from $13.1\%$ loss to $18.54\%$ increase, Figure 3A, 3B, Supplementary Table 2). Contrarily, in the gastrocnemius muscle, FTG to STO fiber type switch was notable in females along with a significant increase in STO fiber diameter (from 12. $47\%$ loss to $20.19\%$ increase), while neither the fiber type composition nor the fiber diameter was altered with treatment in males significantly (Figure 3C, 3D, Supplementary Tables 5, 6). **Figure 3:** *NT-3 effect on fiber diameter and fiber type switch. Line graphs represent the changes in fiber type contribution to total as percent (STO, FTO and FTG) in the muscles analyzed. Data are represented as mean ± SEM. Bar graphs represent percent changes in average fiber diameter of sarcopenic mice compared to 10-month-old mice to show age-related fiber size changes, and NT-3-treated mice compared to untreated 2-year-old mice to show the effect of treatment in average fiber size change. Mean fiber size is calculated for each cohort and percent change is determined based on these mean fiber size values for each muscle. Green arrow heads mark bars for treated mice showing no fiber size increase, yellow arrow heads depicting bars for untreated mice did show fiber size increase compared to 10-month- old mice, likely as compensatory change. (A, B) Tibialis anterior (UT, n = 8; NT-3, n = 9; WT, n = 8; with equal sex distribution), (C, D) gastrocnemius (UT, n = 6; NT-3, n = 6; WT, n = 8; with equal sex distribution) (E, F) quadriceps (UT, n = 6; NT-3, n = 4; WT, n = 7; with equal sex distribution) and (G, H) triceps (UT, n = 8; NT-3, n = 9; WT, n = 8; with equal sex distribution) muscles from 10-month-old mice, UT cohort and treatment cohort, shown as sexes combined (grey) and separated (red: females, blue: males). Two-way ANOVA, Tukey’s multiple comparisons test. Data is represented as mean ± SEM; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 showing only comparison to UT cohort.* Compared to 10 months old C57BL/6 mice, the greatest size reduction in FTG and STO occurred in quadriceps muscle from 2-year-old in both sexes and was more prominent in males. With treatment, the fiber type switch from glycolytic to oxidative was associated with reversal of sarcopenic effect on oxidative fiber size, in both sexes. The size increase in FTG was meager for both sexes (Figure 3E, 3F, Supplementary Tables 8, 9). Interestingly, triceps muscle appeared more resistant to the age-related size change contrasting with other muscles. The decrease in FTG fiber diameter, compared to those from 10 months old C57BL/6 mice was milder; a size increase occurred in STO and FTO fibers in males, while the STO fiber size decrease was minimal in females (Supplementary Tables 10–12). With treatment, the fiber type switch from glycolytic to oxidative reflected as size increase for both STO and FTO fibers in both sexes, was more prominent for FTO than STO ($11.3\%$ in FTO vs. $4\%$ in STO) in males. In addition, males showed 11.38 % size increase in FTG. A trend for STO and FTO fiber type increase was present with aging in males when untreated 2-year-old samples were compared to those from 10-month-old (Figure 3G, 3H). Aging process in Schwann cells, can easily be identified on the high resolution semi-thin cross sections of peripheral nerves which includes an overall thinning of myelin [17, 18] and age-related myelin alterations, such as increased myelin corrugation, infoldings and outfoldings reflecting axonal atrophy. We observed these changes in sciatic and tibial nerves distally in 2-year-old C57BL/6 mice. To assess the efficacy of NT-3 gene therapy on the aging process in peripheral nerves, we chose to evaluate G ratio (axon diameter/fiber diameter), a measure of myelin thickness in the distal tibial nerve, from aged C57BL/6 mice at 6 months post gene delivery, and compared to untreated counterparts (Figure 4A, 4B). Thinning of myelin in rodent nerves with aging has been reported previously, although mean G ratio in sciatic nerves from aged group was not found to be significantly different from younger groups, explained by the presence of concomitant axon size decrease [17–19]. In distal tibial nerves however, we found significantly increased mean G ratio in aged C57BL/6 mice, reflecting an age related thinning of myelin, when compared to data obtained from 1 year old mice (Figure 4C, 4D; 2-year-old, 0.727 ± 0.0013 vs. 1-year old, 0.665 ± 0.0015; $p \leq 0.0001$). The percent of fibers with a G ratio ≥ 0.7, a reflection of thin myelin in aged mice, constituted $67.79\%$, whereas in 1 year old mice, this was only $27.32\%$ of total fibers. NT-3 gene therapy resulted in an attenuation of age-related myelin change, indicated by treated and untreated groups having significantly different slopes of G ratio scatterplots (Figure 4E). The mean G ratio was significantly improved toward optimal peripheral nerve G ratio of 0.6 reflecting thicker myelin at 6 months post AAV1.NT-3 gene delivery (NT-3, 0.672 ± 0015 vs. untreated, 0.727 ± 0.0013; $p \leq 0.0001$). The percent fibers with G ratio ≥ 0.7 representing thin myelin reduced to $32.62\%$, which constituted $67.79\%$ of total fibers in the untreated cohort (Figure 4F). **Figure 4:** *NT-3 gene transfer improves myelin thickness of peripheral nerves and neuromuscular junction assembly in C57BL/6 mice. Representative semithin, toluidine blue–stained cross-sections of tibial nerves from (A) treated and (B) untreated C57BL/6 mice. Scale bar = 10 μm. (C) Myelin thickness significantly decreased with age in 2-year-old mice compared to 1-year-old counterparts (1 yo, r2 = 0.0363; 2 yo, r2 = 0.2245; Linear regression, slopes are significantly different, p < 0.0001). (D) The shift toward thinner myelin can also be observed in percent distribution graph with aging. (E) A notable increase of myelin thickness was observed with treatment compared to samples from untreated (NT-3, r2 = 0.0989; UT, r2 = 0.2245; Linear regression, slopes are significantly different, p = 0.0037) and (F) percent analysis of g ratio of treated cohort displayed a distribution that peaks at 0.6 (n = 6 for both treated and untreated 2-year-old mice, n = 4 for 1-year-old control mice, with even sex distribution). (G) Representative image showing innervated (I), partially innervated (PI) and denervated (D) NMJs from the lumbrical muscles of the aged C57BL/6 mice. Scale bar = 10 μm. (H) Percent of the innervated NMJs in the treated mice was significantly higher than the untreated mice (*p = 0.0123). We evaluated an average of 41.2 NMJs per mouse (n = 4 mice for each cohort with equal sex distribution). Data is represented as mean ± SEM; Two-way ANOVA, Sidak’s multiple comparisons test; *p < 0.05.* We also assessed the status of NMJ in response to NT-3 gene therapy in this model as there is evidence that changes in endplate morphology and NMJ remodeling occur with aging and precede loss of fast motor units [20]. Using immunohistochemistry-based parameters [11], we analyzed a total of 330 NMJs derived from intrinsic foot muscles of NT-3 treated and untreated aged C57BL/6 mice. This analysis showed that AAV1.NT-3 gene therapy, at 6 months of treatment gave rise to a $29.7\%$ increase of innervated NMJs ($$p \leq 0.0123$$). A decrease in the denervated and partially denervated/innervated NMJs was also noted (Figure 4G, 4H). We believe that these quantitative histopathological assessments of muscle, NMJ and nerve collectively provide strong support to the functional improvements observed in aged C57BL/6 mice with NT-3 gene therapy. ## NT-3 induced remodeling of muscle metabolism in age-related sarcopenia Studies have shown that mtDNA and mRNA abundance and mitochondrial ATP production, all decline with advancing age [21–23]. We first started by investigating whether the mitochondria biogenesis marker Pgc1α transcripts and mtDNA copy number per genomic DNA are altered in muscle at 6 months post NT-3 gene therapy in aged C57BL/6 mice. In the untreated tibialis anterior muscle, Pgc1α transcripts showed a significant decline in both sexes at 2 years of age, compared to 10 months old mouse samples (Figure 5A). The response to NT-3 gene therapy, however, showed sex difference; with males having significantly more Pgc1α transcripts than females. The mitochondria copy number in females from the untreated cohort was significantly lower than 10-month-old samples, while the treated cohort showed an increase. Interestingly, mitochondria copy number in the untreated and treated muscle from males did not differ from the 10-month-old males (Figure 5B). We then analyzed the expression levels of Cox1 and Cox3, which are mtDNA encoded subunits of cytochrome c oxidase of respiratory complex IV and Atp5d subunit of complex V, encoded by genomic DNA. We also investigated if intensity-quantification of histochemical staining for cytochrome c oxidase (COX) enzyme activity in these muscles show any correlation with the expression levels of these subunits, and/or directly reflect the functional mitochondria content in the muscle. As expected, Cox1 and Cox3 transcripts significantly declined in the untreated tibialis anterior muscle at 2 years of age compared to the levels from younger C57BL/6 controls (Figure 5C, 5D). AAV1.NT-3 treatment gave rise to a trend of increased Cox1 expression, contrasting with highly significant increases in nuclear encoded Atp5d transcripts in both sexes, which was more prominent in females (Figure 5C–5E). We also found COX enzyme stain-intensity quantification correlating well with increased Atp5d expression levels and mitochondria content, more notable in females compared to the untreated muscle (Figure 5F–5J). **Figure 5:** *NT-3-treatment-induced changes on markers of mitochondrial biogenesis, oxidative phosphorylation and mTORC1 pathway in tibialis anterior muscle. Bar graphs represent (A) relative expression levels of Pgc1α, (B) mtDNA copy number/genomic DNA, relative expression levels of (C) Cox1, (D) Cox3, and (E) Atp5d genes of tibialis anterior muscle in treated and untreated C57BL/6 mice (n = 8, 10-month-old (mo) mice; n = 8, 2-year-old untreated mice (2 yr); n = 9, 2-year-old NT-3 treated (2 yr-NT-3) mice; with equal sex distribution). (F–I) Representative images of COX-stained sections of tibialis anterior muscle in the treated and untreated female and male mice. Scale bar: 25 μm, applies to all images (J) Bar graphs showing the intensity analysis on COX-stained sections (n = 6, untreated mice; n = 6, NT-3 treated mice; with equal sex distribution). (K) Western blots showing the expression level of p-S6, and p-4E-BP1 proteins. Protein levels of (L) p-S6, and (M) p-4E-BP1 normalized to Actin (n = 6 for both cohorts with equal sex distribution, blots were cropped for conciseness). Relative expression levels of (N) Hk-1 and (O) Pfkm enzymes (n = 8, 2-year-old untreated mice; n = 9, 2-year-old NT-3 treated mice). Student t-test for the analysis marked with purple asterisk. Two-way ANOVA, Tukey’s multiple comparisons test. Data is represented as mean ± SEM; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.* Considering the significant size increase in FTG fibers, along with increased glycolytic fiber type in the female tibialis anterior muscle, we next explored the possibility of mTORC1 activation, as previously shown with NT-3 in the neurogenic muscle from trembler J mouse (TrJ) [9]. In western blot analysis, we found significant increases in the downstream targets of mTORC1, phosphorylated S6 kinase and 4E-BP1 proteins, indicating the activation of mTORC1 which correlate with increased radial growth of FTG fibers (Figure 5K–5M). In addition, we found a trend for increased expression of glycolytic enzymes Pfkm and Hk-1 in females, suggesting increased carbohydrate metabolism in muscle. In contrast, expression of phosphorylated S6 protein and HK-1 transcripts were lower in males compared to untreated samples, in keeping with the sex-dependent pattern of changes in this muscle (Figure 5N, 5O). We found similar changes with aging in the untreated gastrocnemius muscle at 2 years of age for both Pgc1α transcripts as well as the mitochondria copy number, which were significantly lower than 10-month-old counterparts (Figure 6A, 6B). At 6 months post NT-3 gene therapy, although not statistically significant, there was further decline in Pgc1α relative expression levels compared to untreated counterparts, without sex difference (Figure 6A). The mitochondria copy number increased in the female muscle without reaching significance, contrasting with a decline in the samples from males (Figure 6B). Cox1 and Cox3 transcripts from untreated muscles at 2 years of age significantly declined, while Atp5d expression increased compared to the levels from 10 months old C57BL/6 muscle (Figure 6C–6E). With NT-3 gene therapy, we found no significant change in the expression levels of these transcripts, although there were sex-dependent differences in staining intensity of COX enzyme in the NT-3 treated muscle, compared to Ringer’s lactate-treated controls (Figure 6F–6I). Females exhibited a stronger staining intensity with treatment, while an opposite pattern was present in males, showing stronger staining intensity in the untreated muscle (Figure 6J). The COX enzyme stain-intensity quantification reflected the changes in Cox1, Cox3, and Atp5d transcripts and mitochondria content between males and females, compared to untreated muscle. Contrasting with tibialis anterior, in western blot analysis, phosphorylated S6 kinase and 4E-BP1 proteins showed significant increases in males, which reflected a small radial growth only in the FTO fibers, while no size increase was noted in FTG fibers in response to treatment (Figure 6K–6M, Figure 3D, Supplementary Table 6). Moreover, the Pfkm and Hk-1 transcripts were lower in males compared to untreated samples suggesting that males did not respond to NT-3 induced anabolic stimulation by increasing carbohydrate metabolism in the gastrocnemius muscle. Females, nonetheless, showed an increase trend for phosphorylated S6 kinase protein levels and the Pfkm and Hk-1 transcripts without reaching significance (Figure 6L–6O). **Figure 6:** *NT-3-treatment-induced changes on markers of mitochondrial biogenesis, oxidative phosphorylation and mTORC1 pathway in gastrocnemius muscle. Bar graphs represent (A) relative expression levels of Pgc1α, (B) mtDNA copy number/genomic DNA, relative expression levels of (C) Cox1, (D) Cox3, and (E) Atp5d genes of gastrocnemius muscle in treated and untreated aged C57BL/6 mice. (n = 8, 10-month-old (mo) mice; n = 8, 2-year-old (2 yr) untreated mice; n = 9, 2-year-old NT-3 treated (2 yr-NT-3) mice; with equal sex distribution). (F–I) Representative images of COX-stained sections of tibialis anterior muscle in the treated and untreated female and male mice. Scale bar: 25 μm, applies to all images (J) Bar graphs showing the intensity analysis on COX-stained sections (n = 7, untreated mice; n = 8, NT-3 treated mice; with equal sex distribution). (K) Western blots showing the expression level of p-S6, and p-4E-BP1 proteins. Protein levels of (L) p-S6, and (M) p-4E-BP1 normalized to Actin (n = 6 for both cohorts with equal sex distribution, blots were cropped for conciseness). Relative expression levels of (N) Hk-1 and (O) Pfkm enzymes (n = 8, 2-year-old untreated mice; n = 9, 2-year-old NT-3 treated mice). Student t-test for part L. Two-way ANOVA, Tukey’s multiple comparisons test. Data is represented as mean ± SEM; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.* ## DISCUSSION In this study, we provide strong evidence for the efficacy of AAV1.NT-3 gene therapy in sarcopenia and age related hypomyelination of peripheral nerves, as well as the maintenance of NMJ connectivity in 2-year-old C57BL/6 mouse, a model for natural aging. At 6 months post gene delivery, the NT-3 effects on the neuromuscular system manifested as significant histopathological improvements, reflecting upon functional performance and in vivo muscle physiology. Attenuation of other age related musculoskeletal and skin changes with treatment including kyphosis, dermatitis, and alopecia was also striking. In the untreated C57BL/6 mice, detailed morphometric data from all four muscles, although confirmed the previous observations of type 2 fiber atrophy in general [24–26], revealed some differences between muscles in response to aging process. The most severe glycolytic fiber atrophy occurred in both male and female quadriceps, and the most severe oxidative fiber atrophy was present in male tibialis anterior muscles; triceps were the least affected muscle by sarcopenia. At 6 months post NT-3 gene therapy, quantitative data from all four muscles revealed an overall normalization of fiber size towards values observed in 10 months of age. In the quadriceps and triceps (proximal-extensor muscles) of the treated aged cohort, there was a switch to fatigue-resistant oxidative fibers occurred in both sexes, when compared to untreated. However, in distal hindlimb muscles (anterior compartment/flexor muscle vs. posterior compartment/extensor muscle), we found what appears to be a sex-dependent muscle remodeling. The switch from oxidative to glycolytic fiber type observed in the tibialis anterior muscle in females was associated with prominent FTG size increase. In the same muscle from males, oxidative fiber size increase was greater than glycolytic, along with an increase in oxidative fiber type. Contrary to the tibialis anterior, in the female gastrocnemius muscle, there was a switch from glycolytic to oxidative fiber type, in addition to a prominent STO size increase. We were intrigued with these observations that the most prominent muscle fiber size increase occurred in the direction of fiber type switch; to oxidative vs. to glycolytic. Interestingly however, male gastrocnemius muscle failed to show any size increase in STO or FTG, while a small FTO percentage gain was detected along in addition to a meager size increase in this fiber subtype. Overall, these observations show the presence of a muscle- and sex-dependent remodeling with aging, as well as response to NT-3 treatment. The results from the molecular studies assessing the NT-3 effect on the oxidative state of distal hind limb muscles, accompanied by western blot analyses for mTORC1 activation are in accordance with the histologic findings. We like to emphasize here that NT-3 effect is not directed to well-differentiated or normal functioning cells, but rather is operative upon remodeled-cell metabolism that may result from a pathological process [9, 27]. In this context, it is not surprising to see NT-3, through mTORC1 activation, exerts an anabolic effect in the female tibialis anterior by increasing FTG fiber size and the percentage of glycolytic fibers, and by doing so, overcoming the age-related remodeling by normalizing the muscle towards to a younger age. Even though Pgc1α transcripts at this time point was found not elevated, we showed evidence that NT-3 prevented the age-related mitochondria loss by increasing mitochondria copy number towards 10-month levels. This was associated with significantly increased Atp5d transcripts and COX stain intensity, suggesting increased mitochondrial function. Decreased Pgc1α levels can be interpreted as correlative of decreased STO percentage, which was also associated with an increase in FTG percentage along with a significant size increase of FTG fibers. We believe that the total mtDNA increase in tissue might be linked to these changes in FTG subtype. In contrast, NT-3 in male tibialis anterior showed no mTORC1 activation, but preserved mitochondria content and function. This was supported by an increase in Pgc1α and Atp5d transcripts as well as COX stain intensity, which was reflected in muscle histology as fiber type switch from glycolytic, back to the oxidative levels, as in the 10-month-old samples. The findings in gastrocnemius, when contrasted with tibialis anterior muscle, appear unique, as males were less responsive to NT-3. Even though with treatment, mTORC1 was activated, the male gastrocnemius muscle failed to respond with size increase in any of the fiber types. NT-3 also did not alter mitochondria copy number, or markers of oxidative phosphorylation expression significantly. In female gastrocnemius, although did not reach to significance level (likely due to small n number), NT-3 lead to some increases in the phosphorylated S6 protein, as well as in mitochondria content, which were associated with STO greater than FTG fiber size increase, along with increased STO percentage. Studies have shown that mtDNA and mRNA abundance and mitochondrial ATP production, all decline with advancing age [21]. PGC1α is an important transcriptional coactivator of mitochondrial biogenesis and respiration [28]. Previous studies have shown that PGC1α drives the formation of fatigue resistant STO muscle fibers [29]. Decreased Pgc1α expression in sarcopenic muscle from old rodents has been reported [30], as well as with aging in sedentary persons compared with physically active individuals [31, 32]. Recent transcriptomic data from skeletal muscle found reduced expression of genes related to all electron transport complexes and pyruvate dehydrogenase complexes with aging in humans [33]. Our molecular studies from untreated aged C57BL/6 mice agree with these previous reports by showing decreased Pgc1α expression and mtDNA content as well as reduced Cox1 and Cox3 transcripts compared to 10-month-old muscle. Lack of Pgc1α increase at 6 months post gene injection, although seemingly not fitting with the finding of increased mitochondria content in female tibialis, might be explained with the possibility of Pgc1α activation in an earlier time point followed by an equilibrium state. In a previous study of NT-3 gene therapy in young male trembler J (TrJ) mice, at 4 months post AAV.NT-3 injection we found increased Pgc1α expression, accompanied by increased phosphorylation levels of 4E-BP1 and S6 proteins as evidence of mTORC1 activation in the gastrocnemius muscle [9]. AAV.NT-3 treatment was capable of reversing the defective expression levels of Pgc1α seen in the TrJ neurogenic muscle along with enhanced levels of activated 4E-BP1. In the skeletal muscle, mTORC1 can regulate mitochondrial biogenesis and metabolism through 4E-BP1/PGC1α [34]. Therefore, it is conceivable that NT-3 might also promote oxidative phosphorylation through activation of 4E-BP1 and PGC1α in the muscle. PGC1α is known to co-regulate several genes, and its expression alone is considered sufficient to increase mitochondrial mass [35]; however, it was also reported that PGC1α is dispensable for exercise induced mitochondrial biogenesis in skeletal muscle [36]. Therefore, further studies are needed to assess the link between NT-3 and Pgc1α expression in the sarcopenic muscle, including earlier time point experiments in this model. Physical exercise, aerobic or anaerobic, has been well accepted to be a countermeasure for sarcopenia, but also recognized as an important practice in the prevention and treatment of not only chronic and degenerative diseases, but also age-associated multisystem diseases, including cardiovascular, neurodegenerative, chronic pulmonary diseases, cancer, diabetes, and morbid obesity. Aerobic/endurance exercise helps to maintain and improve cardiovascular fitness, respiratory function and muscle oxidative capacity, whereas strength/resistance-exercise programs primarily increase type 2 muscle fiber size, muscle strength, and function. Although the underlying sequential events that lead to these muscle adaptations are poorly understood, it has been emphasized that the mechanisms that regulate these processes involve the “quality” of skeletal muscle mitochondria [37]. Cellular antioxidant capacity and oxidative stress are postulated to be critical factors in the aging process. In one study, whole body resistance exercise training in aged men and women was shown to induce significantly higher complex IV activity and decreased oxidative stress markers, compared to before training [38]. Notably, there were no apparent changes in normal mtDNA content or mtDNA deletion products, suggesting that regular resistance exercise decreases oxidative stress, but does not affect mtDNA. It was also postulated that increases in complex IV of the electron transport chain may have an indirect antioxidant effect in older adults [39]. We think these observations are relevant to this current study regarding the similarities between exercise and NT-3-induced changes in sarcopenic muscle. COX stain intensity quantification obtained from tibialis and gastrocnemius muscles in our study appears to be a sensitive indicator of functioning mitochondria, as in the case of resistance exercise, correlating well with treadmill functional tests being more prominent in females. Activation of the mTORC1 pathway leading to radial growth in glycolytic fiber type, that we see in female tibialis muscle, is also a supportive finding. Instead, an overall increase of oxidative fibers in quadriceps and triceps of both sexes, and in female gastrocnemius muscle with NT-3, is reminiscent to the effects of aerobic/endurance exercise. Moreover, exercise induced mechanical loading has been shown to induce production of IGF-1, VEGF, and hepatocyte growth factor in osteocytes cell lines, which may play roles in regulating muscle growth [40]. In addition, muscle NT-3 levels increased by exercise training have been shown to contribute to improvement in various conditions [40–44]. Moreover, although it has not been studied in the context of sarcopenia, we predict that NT-3, via its known anti-inflammatory and immunomodulatory properties, [14, 45–47] may also have an attenuating effect on age-related inflammation; presumably a contributing factor to sarcopenia. Exercise intervention, despite its effectiveness as therapeutic regime for sarcopenia, is only available to patients or elderly who are reasonably mobile and can participate in such intervention without safety concerns. When considering the burden of sarcopenia on the lifestyle of elderly, and on the healthcare system, we believe this preclinical study is providing strong support for AAV.NT-3 gene therapy in the successful management of sarcopenia, as a serious and plausible option in the future. ## Animals and treatment groups Naturally aged C57BL/6 mice were included in the study (JAX stock #000664) and all animal experiments were performed according to the guidelines approved by The Research Institute at Nationwide Children’s Hospital Animal Care and Use Committee that operates in full accordance with the Animal Welfare Act and the Health Research Extension Act (IACUC approval number = AR18-00076). 18 months old C57BL/6 (6 males and 6 females, $$n = 12$$) mice received 1 × 1011 vg dose of AAV1.tMCK.NT-3, via intramuscular (IM) injection into the gastrocnemius muscle. Age- and sex-matched C57BL/6 (6 males and 6 females, $$n = 12$$) mice were injected with Ringer’s lactate as controls. Mice were tested functionally with treadmill, rotarod, and in vivo muscle contractility assay and they were sacrificed six months post-injection by an over-dosage of xylazine/ketamine anesthesia for harvesting blood, sciatic nerves, as well as upper and lower limb muscles including lumbricals, at six months post gene injection. ## AAV1.tMCK.NT-3 vector production and potency Construct of self-complimentary (sc) AAV serotype 1 vector with muscle specific tMCK promoter was described previously [10]. The vector was produced in our Viral Vector Core at Nationwide Children’s Hospital, Columbus (Andelyn Biosciences). Aliquots of virus were stored at −80°C until used. Serum was separated from blood samples that were collected by cardiac puncture at six months post gene injection and serum NT-3 levels were detected by ELISA as previously reported (NT-3, $$n = 9$$; UT, $$n = 8$$) [10]. ## Run to exhaustion test Run to exhaustion treadmill performance test was performed at two-, four- and six-months post injection. Mice were exercised to exhaustion via treadmill (Columbus Instruments, Exer-6M Treadmill), as described previously [15]. Mice were acclimated to the treadmill prior to data collection. Mice were run to exhaustion with increasing treadmill speed by 1 meter/min each minute, starting at an initial 7 meter per minute velocity. Lanes have a shock plate that pulses at a frequency of ~3 Hz. Mice were considered at “exhaustion” level when they were unable to re-engage the treadmill for 3 seconds after resting on the shock-plate. Run duration was recorded and used to calculate the distance ran (2 months post-injection: NT-3, $$n = 12$$; UT, $$n = 11$$; 4 months post-injection, NT-3, $$n = 12$$; UT, $$n = 11$$; 6 months post-injection: NT-3, $$n = 11$$; UT, $$n = 8$$). ## In vivo muscle contractility assay In vivo muscle contractility assay was performed at the endpoint, as described previously [15]. Hind paw of the anesthetized mouse was placed on footplate, which was attached to a dual-mode lever, and the tibia was aligned perpendicular to the lever. Subcutaneous EMG electrodes were used to stimulate the tibial nerve. Gastrocnemius muscle torque around the ankle joint was measured by muscle physiology apparatus (Aurora Scientific, ON, Canada) using isometric contraction (maximum twitch response) and fatigue (maximum tetanic response) protocols (Max twitch: NT-3, $$n = 11$$; UT, $$n = 8$$; Max tetanic: NT-3, $$n = 10$$; UT, $$n = 9$$). ## Rotarod Rotarod test was performed at six months post-injection to assess motor function and balance of the mice. Mice were acclimated to rotarod apparatus (Columbus Instruments, Ohio, USA) prior to data collection. Rotarod protocol included a 5-rpm run with a constant acceleration of 0.2 rpm/s. The averages of the best two out of three runs were calculated (NT-3, $$n = 10$$; UT, $$n = 8$$). ## Musculoskeletal and skin changes Age-related musculoskeletal and skin changes including kyphosis, dermatitis, and alopecia were documented semi-quantitatively. Six mice, from both treated and untreated cohorts, were photographed at the end point and the severity of kyphosis, dermatitis, and alopecia were scored. Severe changes were scored as 1, mild-moderate changes as 0.5, and no changes as 0. For kyphosis, arbitrary lines were drawn tangential to the scoliosis curvature to delineate the scoliosis angle (Supplementary Figure 3A). Angle range 90°–110° is considered as severe, 11°–150° as mild-moderate and >150° as none. ## Muscle histology Tibialis anterior, gastrocnemius, quadriceps, and triceps muscles from treated and untreated mice were collected and 12 μm thick cross cryostat-sections were cut. Succinic dehydrogenase (SDH) enzyme histochemistry was performed as previously described to assess metabolic fiber type distribution and myofiber size changes in the aging muscle [15]. Deep, intermediate, and superficial zones of the muscles were analyzed to represent fibers at various oxidative states equally. One representative area from each zone were photographed at 20× magnification using an Olympus BX41 microscope and SPOT Insight 12 Mp sCMOS camera. Sample selection was based on the suitability of the tissue sections, including staining quality, contrast, and lack of artifacts and not based on outcomes of behavioral or physiological analyses. Shortest distance across the muscle fiber was measured as fiber diameter (Zeiss Axiovision LE4 software V4.9.1.0) and mean fiber diameter (mean ± SEM) was calculated for each fiber type (STO, FTO, FTG) as well as for combination of all fiber types. Fiber type percent distribution of total fibers was determined for each mouse from each treatment group. Data were obtained from a total of 2747 ($$n = 9$$), 2959 ($$n = 9$$), 1722 ($$n = 5$$), and 1357 ($$n = 4$$) fibers of the treated cohort, and 2815 ($$n = 8$$), 2557 ($$n = 8$$), 2040 ($$n = 6$$), and 2164 ($$n = 6$$) fibers of the untreated cohort for tibialis anterior, triceps, gastrocnemius and quadriceps muscles respectively. ## G ratio of the myelinated fibers G ratio was calculated to assess the myelin thickness of fibers in tibialis nerve, as described previously [48]. Semithin, toluidine blue-stained cross sections of tibial nerve were prepared for each mouse ($$n = 6$$ for both treated and untreated 2-year-old mice, $$n = 4$$ for 1-year-old control mice, with even sex distribution) and three randomly selected nonoverlapping areas were photographed randomly at 100× magnification using an Olympus BX41 microscope and SPOT Insight 12Mp sCMOS camera. Myelin interior and exteriors were outlined in Axiovision (AxioVs40 × 64 V 4.9.1.0) to determine the area, which was used to calculate diameters to determine g ratio. A total of 1833 fibers for treated, 1928 for untreated and 1658 for 1-year-old control mice were measured to generate scattergrams and the percent g ratio distribution histograms. Slopes of treated vs. untreated and 1-year-old vs. 2-year-old mice were compared using GraphPad Prism (9.0.0). ## Immunohistochemical analysis of neuromuscular junctions (NMJ) Lumbrical muscles collected from treated and untreated mice ($$n = 4$$ for both cohorts with equal sex distribution) were processed as described previously [11]. Muscles were fixed and stained with primary antibodies (Acetylcholine receptor (AChR) antibody, α Bungarotoxin, T1175, 1:500; Anti Neurofilament 200 antibody, N4142, 1:500; SV2 antibody, AB_2315387, 1:50) [49, 50], followed by incubation with secondary antibodies (Alexa Fluor 488 conjugated anti-rabbit and anti-mouse IgG, 1:500). Samples were imaged at 60× magnification using a Zeiss LSM 800 confocal microscope. NMJs were considered to be innervated when nerve completely overlapped the AChRs, as partially innervated when some parts of the AChRs were not overlapping with nerve, and as denervated when there was not any nerve co localizing with AChRs [11, 51]. An average of 41.3 NMJs per mouse were evaluated from NT-3-treated and UT-mice ($$n = 4$$ mice per group with equal sex distribution). ## Protein extraction and western blot analysis Twenty micrometer thick sections from frozen TA and GAS muscle blocks (20 section per block, $$n = 3$$ per group) were put into 2 ml centrifuge tubes and homogenized in homogenization buffer [125 mM Tris-HCL pH6.8, $4\%$ SDS, 4 M Urea solution with 1X Halt protease inhibitor (ThermoFisher) and 1× phosphatase inhibitor (Sigma)] using a disposable pestle. The lysate was then incubated on a rotary spin cycle at 4°C for 2 hours, followed by centrifugation at 10,000 g for 10 min at 4°C. The supernatant was then transferred to a new tube. Protein samples were run in Novex 10–$20\%$ Tricine mini protein gel (ThermoFisher) and transferred to PDVF membranes (GE Healthcare). Membranes were blocked for 1 h at room temperature with $5\%$ milk in TBS-T (TBS buffer with $0.05\%$ Tween-20). Membranes were then incubated with primary antibodies in TBS-T buffer overnight at 4°C. After 5 min of three times wash with TBS-T, membranes were incubated with secondary antibodies in $5\%$ milk in TBS-T for 1 h. Membranes were washed again with TBS-T for 3 times and TBS for 2 times with 5 min each wash. ECL WesternSure premium chemiluminescent substrate (LI-COR) was used for band detection followed by exposure using Chemidoc Imaging system (Bio-Rad) and band intensities were quantified using ImageJ (NIH). Primary antibodies: anti-phospho S6 protein Ser$\frac{235}{236}$ (#4858), and anti-phospho 4EBP1 thr$\frac{37}{46}$ (#2855) were from Cell Signaling Technology and anti-actin antibody (sc-47778) was from Santa Cruz; secondary antibody: HRP-linked anti-rabbit/mouse IgG (#$\frac{7074}{7076}$) was purchased from Cell Signaling Technology. ## RNA isolation and mRNA expression Total RNA was extracted from frozen muscle blocks (20 micrometer thick sections, 20 section per block; NT-3, $$n = 9$$; UT, $$n = 8$$; WT, $$n = 8$$) using Mini RNeasy Plus Universal Kit (Qiagen). cDNAs were synthesized using ProtoScript II First Strand cDNA Synthesis Kit (BioLabs). Primer sets (synthesized by IDT) for Pgc-1α, Cox1, Cox3 and Atp5d were obtained from previous publications [52–54] and new primers were designed for Pfkm (F-GAAGATACCAACTCGGACCAC, R-ATGACCCATGAAGAGCATCA) and Hk-1 genes (F-CGGAATGGGGAGCCTTTGG, R-GCCTTCCTTATCCGTTTCAATGG). All qPCR were performed using PowerUp SYBR Green Master Mix (ThermoFisher) according to the manufacturer’s instructions. qPCR assays were performed using QuantStudio 6 Flex (Applied Biosystem). Expression data were normalized to mouse Gapdh mRNA level and data were analyzed by ΔΔCt method. ## COX staining density analysis Histochemical enzyme activity of cytochrome c oxidase (COX) was assessed by quantifying the COX stain intensity in tibialis anterior and gastrocnemius muscles from treated and untreated mice. 12 μm thick-fresh frozen sections were stained using COX enzyme histochemistry protocol established in our clinical neuromuscular pathology laboratory. Muscle sections representing deep, intermediate, and superficial zones from gastrocnemius ($$n = 8$$ treated, $$n = 7$$ untreated) and tibialis anterior ($$n = 6$$ treated, $$n = 6$$ untreated) muscles were photographed at 10× magnification, using an Olympus BX41 microscope and SPOT camera, with even distribution of sex. Image data were collated and processed utilizing the Python programming language. [ 55] Images were calibrated using representative selections from each image to establish ranges of intensity for each fiber-type; dark (oxidative/higher mitochondria content/type 1) and light (glycolytic/lower mitochondria content/type 2) fibers, including COX deficient pale fibers. Ratio of dark to light fibers was calculated and presented as mean ± SEM. ## Statistics Adequate sample size was determined according to our previous studies that performed analogous experiments [11, 15]. All statistical analyses were performed in GraphPad Prism 9.0 software. Two tail Student t-test, one-way ANOVA with Tukey’s multiple comparison test, two-way ANOVA with Sidak’s multiple comparison test or linear regression analysis were performed when applicable, and significance level was set at P ≤ 0.05. The tests that meet the best assumptions of the data were chosen. Results were given as mean ± SEM in all experiments and the number of animals was mentioned in figure legends along with the name of the statistical analysis performed. Other than the functional tests and quantitative muscle analysis, no blinding was used. ## References 1. Wiedmer P, Jung T, Castro JP, Pomatto LCD, Sun PY, Davies KJA, Grune T. **Sarcopenia - Molecular mechanisms and open questions.**. *Ageing Res Rev* (2021) **65** 101200. DOI: 10.1016/j.arr.2020.101200 2. Cruz-Jentoft AJ, Landi F, Schneider SM, Zúñiga C, Arai H, Boirie Y, Chen LK, Fielding RA, Martin FC, Michel JP, Sieber C, Stout JR, Studenski SA. **Prevalence of and interventions for sarcopenia in ageing adults: a systematic review. Report of the International Sarcopenia Initiative (EWGSOP and IWGS).**. *Age Ageing* (2014) **43** 748-59. DOI: 10.1093/ageing/afu115 3. Janssen I, Shepard DS, Katzmarzyk PT, Roubenoff R. **The healthcare costs of sarcopenia in the United States.**. *J Am Geriatr Soc* (2004) **52** 80-5. DOI: 10.1111/j.1532-5415.2004.52014.x 4. Clark BC, Manini TM. **Functional consequences of sarcopenia and dynapenia in the elderly.**. *Curr Opin Clin Nutr Metab Care* (2010) **13** 271-6. DOI: 10.1097/MCO.0b013e328337819e 5. Mitchell WK, Williams J, Atherton P, Larvin M, Lund J, Narici M. **Sarcopenia, dynapenia, and the impact of advancing age on human skeletal muscle size and strength; a quantitative review.**. *Front Physiol* (2012) **3** 260. DOI: 10.3389/fphys.2012.00260 6. Murton AJ. **Muscle protein turnover in the elderly and its potential contribution to the development of sarcopenia.**. *Proc Nutr Soc* (2015) **74** 387-96. DOI: 10.1017/S0029665115000130 7. Porter MM. **The effects of strength training on sarcopenia.**. *Can J Appl Physiol* (2001) **26** 123-41. DOI: 10.1139/h01-009 8. Vandervoot AA, Symons TB. **Functional and metabolic consequences of sarcopenia.**. *Can J Appl Physiol* (2001) **26** 90-101. DOI: 10.1139/h01-007 9. Yalvac ME, Amornvit J, Chen L, Shontz KM, Lewis S, Sahenk Z. **AAV1.NT-3 gene therapy increases muscle fiber diameter through activation of mTOR pathway and metabolic remodeling in a CMT mouse model.**. *Gene Ther* (2018) **25** 129-38. DOI: 10.1038/s41434-018-0009-8 10. Sahenk Z, Galloway G, Clark KR, Malik V, Rodino-Klapac LR, Kaspar BK, Chen L, Braganza C, Montgomery C, Mendell JR. **AAV1.NT-3 gene therapy for charcot-marie-tooth neuropathy.**. *Mol Ther* (2014) **22** 511-21. DOI: 10.1038/mt.2013.250 11. Ozes B, Moss K, Myers M, Ridgley A, Chen L, Murrey D, Sahenk Z. **AAV1.NT-3 gene therapy in a CMT2D model: phenotypic improvements in**. *Brain Commun* (2021) **3** fcab252. DOI: 10.1093/braincomms/fcab252 12. Yalvac ME, Arnold WD, Hussain SA, Braganza C, Shontz KM, Clark KR, Walker CM, Ubogu EE, Mendell JR, Sahenk Z. **VIP-expressing dendritic cells protect against spontaneous autoimmune peripheral polyneuropathy.**. *Mol Ther* (2014) **22** 1353-63. DOI: 10.1038/mt.2014.77 13. Sahenk Z, Yalvac ME, Amornvit J, Arnold WD, Chen L, Shontz KM, Lewis S. **Efficacy of exogenous pyruvate in Trembler**. *Brain Behav* (2018) **8** e01118. DOI: 10.1002/brb3.1118 14. Yalvac ME, Arnold WD, Braganza C, Chen L, Mendell JR, Sahenk Z. **AAV1.NT-3 gene therapy attenuates spontaneous autoimmune peripheral polyneuropathy.**. *Gene Ther* (2016) **23** 95-102. DOI: 10.1038/gt.2015.67 15. Sahenk Z, Ozes B, Murrey D, Myers M, Moss K, Yalvac ME, Ridgley A, Chen L, Mendell JR. **Systemic delivery of AAVrh74.tMCK.hCAPN3 rescues the phenotype in a mouse model for LGMD2A/R1.**. *Mol Ther Methods Clin Dev* (2021) **22** 401-14. DOI: 10.1016/j.omtm.2021.06.010 16. Gheller BJ, Riddle ES, Lem MR, Thalacker-Mercer AE. **Understanding Age-Related Changes in Skeletal Muscle Metabolism: Differences Between Females and Males.**. *Annu Rev Nutr* (2016) **36** 129-56. DOI: 10.1146/annurev-nutr-071715-050901 17. Ceballos D, Cuadras J, Verdú E, Navarro X. **Morphometric and ultrastructural changes with ageing in mouse peripheral nerve.**. *J Anat* (1999) **195** 563-76. DOI: 10.1046/j.1469-7580.1999.19540563.x 18. Verdú E, Ceballos D, Vilches JJ, Navarro X. **Influence of aging on peripheral nerve function and regeneration.**. *J Peripher Nerv Syst* (2000) **5** 191-208. DOI: 10.1046/j.1529-8027.2000.00026.x 19. Chung T, Park JS, Kim S, Montes N, Walston J, Höke A. **Evidence for dying-back axonal degeneration in age-associated skeletal muscle decline.**. *Muscle Nerve* (2017) **55** 894-901. DOI: 10.1002/mus.25267 20. Gonzalez-Freire M, de Cabo R, Studenski SA, Ferrucci L. **The Neuromuscular Junction: Aging at the Crossroad between Nerves and Muscle.**. *Front Aging Neurosci* (2014) **6** 208. DOI: 10.3389/fnagi.2014.00208 21. Short KR, Bigelow ML, Kahl J, Singh R, Coenen-Schimke J, Raghavakaimal S, Nair KS. **Decline in skeletal muscle mitochondrial function with aging in humans.**. *Proc Natl Acad Sci U S A* (2005) **102** 5618-23. DOI: 10.1073/pnas.0501559102 22. Barchiesi A, Vascotto C. **Transcription, Processing, and Decay of Mitochondrial RNA in Health and Disease.**. *Int J Mol Sci* (2019) **20** 2221. DOI: 10.3390/ijms20092221 23. Hebert SL, Lanza IR, Nair KS. **Mitochondrial DNA alterations and reduced mitochondrial function in aging.**. *Mech Ageing Dev* (2010) **131** 451-62. DOI: 10.1016/j.mad.2010.03.007 24. Lexell J, Taylor CC, Sjöström M. **What is the cause of the ageing atrophy? Total number, size and proportion of different fiber types studied in whole vastus lateralis muscle from 15- to 83-year-old men.**. *J Neurol Sci* (1988) **84** 275-94. DOI: 10.1016/0022-510x(88)90132-3 25. Joseph AM, Adhihetty PJ, Buford TW, Wohlgemuth SE, Lees HA, Nguyen LM, Aranda JM, Sandesara BD, Pahor M, Manini TM, Marzetti E, Leeuwenburgh C. **The impact of aging on mitochondrial function and biogenesis pathways in skeletal muscle of sedentary high- and low-functioning elderly individuals.**. *Aging Cell* (2012) **11** 801-9. DOI: 10.1111/j.1474-9726.2012.00844.x 26. Lexell J, Taylor CC. **Variability in muscle fibre areas in whole human quadriceps muscle: effects of increasing age.**. *J Anat* (1991) **174** 239-49. PMID: 2032938 27. Young C, Miller E, Nicklous DM, Hoffman JR. **Nerve growth factor and neurotrophin-3 affect functional recovery following peripheral nerve injury differently.**. *Restor Neurol Neurosci* (2001) **18** 167-75. PMID: 11847440 28. Handschin C, Spiegelman BM. **Peroxisome proliferator-activated receptor gamma coactivator 1 coactivators, energy homeostasis, and metabolism.**. *Endocr Rev* (2006) **27** 728-35. DOI: 10.1210/er.2006-0037 29. Lin J, Wu H, Tarr PT, Zhang CY, Wu Z, Boss O, Michael LF, Puigserver P, Isotani E, Olson EN, Lowell BB, Bassel-Duby R, Spiegelman BM. **Transcriptional co-activator PGC-1 alpha drives the formation of slow-twitch muscle fibres.**. *Nature* (2002) **418** 797-801. DOI: 10.1038/nature00904 30. Kang C, Chung E, Diffee G, Ji LL. **Exercise training attenuates aging-associated mitochondrial dysfunction in rat skeletal muscle: role of PGC-1α.**. *Exp Gerontol* (2013) **48** 1343-50. DOI: 10.1016/j.exger.2013.08.004 31. Krämer DK, Ahlsén M, Norrbom J, Jansson E, Hjeltnes N, Gustafsson T, Krook A. **Human skeletal muscle fibre type variations correlate with PPAR alpha, PPAR delta and PGC-1 alpha mRNA.**. *Acta Physiol (Oxf)* (2006) **188** 207-16. DOI: 10.1111/j.1748-1716.2006.01620.x 32. Lanza IR, Short DK, Short KR, Raghavakaimal S, Basu R, Joyner MJ, McConnell JP, Nair KS. **Endurance exercise as a countermeasure for aging.**. *Diabetes* (2008) **57** 2933-42. DOI: 10.2337/db08-0349 33. Su J, Ekman C, Oskolkov N, Lahti L, Ström K, Brazma A, Groop L, Rung J, Hansson O. **A novel atlas of gene expression in human skeletal muscle reveals molecular changes associated with aging.**. *Skelet Muscle* (2015) **5** 35. DOI: 10.1186/s13395-015-0059-1 34. Tsai S, Sitzmann JM, Dastidar SG, Rodriguez AA, Vu SL, McDonald CE, Academia EC, O'Leary MN, Ashe TD, La Spada AR, Kennedy BK. **Muscle-specific 4E-BP1 signaling activation improves metabolic parameters during aging and obesity.**. *J Clin Invest* (2015) **125** 2952-64. DOI: 10.1172/JCI77361 35. Wu Z, Puigserver P, Andersson U, Zhang C, Adelmant G, Mootha V, Troy A, Cinti S, Lowell B, Scarpulla RC, Spiegelman BM. **Mechanisms controlling mitochondrial biogenesis and respiration through the thermogenic coactivator PGC-1.**. *Cell* (1999) **98** 115-24. DOI: 10.1016/S0092-8674(00)80611-X 36. Rowe GC, El-Khoury R, Patten IS, Rustin P, Arany Z. **PGC-1α is dispensable for exercise-induced mitochondrial biogenesis in skeletal muscle.**. *PLoS One* (2012) **7** e41817. DOI: 10.1371/journal.pone.0041817 37. Barbieri E, Agostini D, Polidori E, Potenza L, Guescini M, Lucertini F, Annibalini G, Stocchi L, De Santi M, Stocchi V. **The pleiotropic effect of physical exercise on mitochondrial dynamics in aging skeletal muscle.**. *Oxid Med Cell Longev* (2015) **2015** 917085. DOI: 10.1155/2015/917085 38. Parise G, Brose AN, Tarnopolsky MA. **Resistance exercise training decreases oxidative damage to DNA and increases cytochrome oxidase activity in older adults.**. *Exp Gerontol* (2005) **40** 173-80. DOI: 10.1016/j.exger.2004.09.002 39. Melov S, Tarnopolsky MA, Beckman K, Felkey K, Hubbard A. **Resistance exercise reverses aging in human skeletal muscle.**. *PLoS One* (2007) **2** e465. DOI: 10.1371/journal.pone.0000465 40. Juffer P, Jaspers RT, Lips P, Bakker AD, Klein-Nulend J. **Expression of muscle anabolic and metabolic factors in mechanically loaded MLO-Y4 osteocytes.**. *Am J Physiol Endocrinol Metab* (2012) **302** E389-95. DOI: 10.1152/ajpendo.00320.2011 41. Ying Z, Roy RR, Edgerton VR, Gómez-Pinilla F. **Exercise restores levels of neurotrophins and synaptic plasticity following spinal cord injury.**. *Exp Neurol* (2005) **193** 411-9. DOI: 10.1016/j.expneurol.2005.01.015 42. Wang BL, Jin H, Han XQ, Xia Y, Liu NF. **Involvement of brain-derived neurotrophic factor in exercise-induced cardioprotection of post-myocardial infarction rats.**. *Int J Mol Med* (2018) **42** 2867-80. DOI: 10.3892/ijmm.2018.3841 43. Domínguez-Sanchéz MA, Bustos-Cruz RH, Velasco-Orjuela GP, Quintero AP, Tordecilla-Sanders A, Correa-Bautista JE, Triana-Reina HR, García-Hermoso A, González-Ruíz K, Peña-Guzmán CA, Hernández E, Peña-Ibagon JC, Téllez-T LA. **Acute Effects of High Intensity, Resistance, or Combined Protocol on the Increase of Level of Neurotrophic Factors in Physically Inactive Overweight Adults: The BrainFit Study.**. *Front Physiol* (2018) **9** 741. DOI: 10.3389/fphys.2018.00741 44. Sharma NK, Ryals JM, Gajewski BJ, Wright DE. **Aerobic exercise alters analgesia and neurotrophin-3 synthesis in an animal model of chronic widespread pain.**. *Phys Ther* (2010) **90** 714-25. DOI: 10.2522/ptj.20090168 45. Sekimoto M, Tsuji T, Matsuzaki J, Chamoto K, Koda T, Nemoto K, Degawa M, Nishimura S, Nishimura T. **Functional expression of the TrkC gene, encoding a high affinity receptor for NT-3, in antigen-specific T helper type 2 (Th2) cells.**. *Immunol Lett* (2003) **88** 221-6. DOI: 10.1016/s0165-2478(03)00080-4 46. Beutner C, Lepperhof V, Dann A, Linnartz-Gerlach B, Litwak S, Napoli I, Prinz M, Neumann H. **Engineered stem cell-derived microglia as therapeutic vehicle for experimental autoimmune encephalomyelitis.**. *Gene Ther* (2013) **20** 797-806. DOI: 10.1038/gt.2012.100 47. Yang J, Yan Y, Xia Y, Kang T, Li X, Ciric B, Xu H, Rostami A, Zhang GX. **Neurotrophin 3 transduction augments remyelinating and immunomodulatory capacity of neural stem cells.**. *Mol Ther* (2014) **22** 440-50. DOI: 10.1038/mt.2013.241 48. Ozes B, Myers M, Moss K, Mckinney J, Ridgley A, Chen L, Bai S, Abrams CK, Freidin MM, Mendell JR, Sahenk Z. **AAV1.NT-3 gene therapy for X-linked Charcot-Marie-Tooth neuropathy type 1.**. *Gene Ther* (2022) **29** 127-37. DOI: 10.1038/s41434-021-00231-3 49. Lin W, Burgess RW, Dominguez B, Pfaff SL, Sanes JR, Lee KF. **Distinct roles of nerve and muscle in postsynaptic differentiation of the neuromuscular synapse.**. *Nature* (2001) **410** 1057-64. DOI: 10.1038/35074025 50. Valdez G, Heyer MP, Feng G, Sanes JR. **The role of muscle microRNAs in repairing the neuromuscular junction.**. *PLoS One* (2014) **9** e93140. DOI: 10.1371/journal.pone.0093140 51. Seburn KL, Nangle LA, Cox GA, Schimmel P, Burgess RW. **An active dominant mutation of glycyl-tRNA synthetase causes neuropathy in a Charcot-Marie-Tooth 2D mouse model.**. *Neuron* (2006) **51** 715-26. DOI: 10.1016/j.neuron.2006.08.027 52. Wang L, Xu X, Jiang C, Ma G, Huang Y, Zhang H, Lai Y, Wang M, Ahmed T, Lin R, Guo W, Luo Z, Li W. **mTORC1-PGC1 axis regulates mitochondrial remodeling during reprogramming.**. *FEBS J* (2020) **287** 108-21. DOI: 10.1111/febs.15024 53. Chaignat E, Yahya-Graison EA, Henrichsen CN, Chrast J, Schütz F, Pradervand S, Reymond A. **Copy number variation modifies expression time courses.**. *Genome Res* (2011) **21** 106-13. DOI: 10.1101/gr.112748.110 54. Jousse C, Muranishi Y, Parry L, Montaurier C, Even P, Launay JM, Carraro V, Maurin AC, Averous J, Chaveroux C, Bruhat A, Mallet J, Morio B, Fafournoux P. **Perinatal protein malnutrition affects mitochondrial function in adult and results in a resistance to high fat diet-induced obesity.**. *PLoS One* (2014) **9** e104896. DOI: 10.1371/journal.pone.0104896 55. 55Python Software Foundation. 2022. https://www.python.org/.. (2022)
--- title: ACSL1, CH25H, GPCPD1, and PLA2G12A as the potential lipid-related diagnostic biomarkers of acute myocardial infarction authors: - Zheng-Yu Liu - Fen Liu - Yan Cao - Shao-Liang Peng - Hong-Wei Pan - Xiu-Qin Hong - Peng-Fei Zheng journal: Aging (Albany NY) year: 2023 pmcid: PMC10042701 doi: 10.18632/aging.204542 license: CC BY 3.0 --- # ACSL1, CH25H, GPCPD1, and PLA2G12A as the potential lipid-related diagnostic biomarkers of acute myocardial infarction ## Abstract Lipid metabolism plays an essential role in the genesis and progress of acute myocardial infarction (AMI). Herein, we identified and verified latent lipid-related genes involved in AMI by bioinformatic analysis. Lipid-related differentially expressed genes (DEGs) involved in AMI were identified using the GSE66360 dataset from the Gene Expression Omnibus (GEO) database and R software packages. Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted to analyze lipid-related DEGs. Lipid-related genes were identified by two machine learning techniques: least absolute shrinkage and selection operator (LASSO) regression and support vector machine recursive feature elimination (SVM-RFE). The receiver operating characteristic (ROC) curves were used to descript diagnostic accuracy. Furthermore, blood samples were collected from AMI patients and healthy individuals, and real-time quantitative polymerase chain reaction (RT-qPCR) was used to determine the RNA levels of four lipid-related DEGs. Fifty lipid-related DEGs were identified, 28 upregulated and 22 downregulated. Several enrichment terms related to lipid metabolism were found by GO and KEGG enrichment analyses. After LASSO and SVM-RFE screening, four genes (ACSL1, CH25H, GPCPD1, and PLA2G12A) were identified as potential diagnostic biomarkers for AMI. Moreover, the RT-qPCR analysis indicated that the expression levels of four DEGs in AMI patients and healthy individuals were consistent with bioinformatics analysis results. The validation of clinical samples suggested that 4 lipid-related DEGs are expected to be diagnostic markers for AMI and provide new targets for lipid therapy of AMI. ## INTRODUCTION Acute myocardial infarction (AMI) has become the main cause of hospitalization and death worldwide, seriously threatening human health. Previous studies have suggested that AMI is a complex syndrome with multifactorial disorders. Its risk factors include early family history, smoking, hypertension, dyslipidemia, and diabetes [1–4]. The rupture of vulnerable and lipid-overloaded coronary atherosclerotic plaques can induce the formation of acute thrombus, leading to acute occlusion of blood vessels and progressing to AMI [5]. Dyslipidemia, especially elevated levels of low-density lipoprotein (LDL) cholesterol, is believed to play a key role in the pathogenesis of atherosclerosis [6, 7]. Atherogenesis begins when residues of LDL cholesterol, chylomicron, and very low-density lipoprotein (VLDL) cholesterol molecules enter the artery intima. Then, lipid radicals are oxidized and endocytosed by macrophages, followed by the formation of foam cells [6]. Previous studies have demonstrated that every $1\%$ reduction in LDL cholesterol levels is associated with a $1\%$ reduction in AMI risk [8, 9]. Currently, statin therapy has become the cornerstone for cholesterol-lowering drug therapy for coronary heart disease and AMI patients. Moreover, the combination of PCSK9 inhibitors with statins is recommended to reduce the risk of major adverse cardiovascular events (MACEs) in AMI patients [10]. However, the combination of lipid-lowering therapy still cannot eliminate the risk of MACEs in AMI patients. This might be partly because some lipid-related genes have not been identified. Therefore, identifying new lipid metabolism-related genes associated with AMI will help develop new lipid-lowering drugs to reduce the risk of MACEs in AMI patients. Microarray analysis is an innovative and practical method to discern susceptibility genes to deal with coronary heart disease [11] and AMI [12]. Nevertheless, microarray analysis using differentially expressed genes (DEGs) might have limitations in reproducibility and sensitivity [13, 14]. Machine learning can enhance the prediction and accuracy of these key genes discerned using traditional microarrays or next-generation sequencing data [15]. The most frequently used machine learning techniques include the least absolute shrinkage and selection operator (LASSO) regression and support vector machine recursive feature elimination (SVM-RFE) algorithm [16]. Meanwhile, the combined application of LASSO regression and SVM-RFE algorithm in identifying new lipid-related genes involved in AMI has not been conducted. Therefore, in the present study, we analyzed the GSE66360 dataset from different perspectives: 1) DEGs among lipid-related AMI genes were identified using the “limma” R package and GO and KEGG pathway enrichment analyses. 2) Machine learning methods were applied to large-scale screening and diagnostic identification of AMI-related molecular markers. 3) Lipid-related gene expression levels screened by machine learning were validated using clinical cases. ## Identification of lipid-related DEGs in GSE66360 Due to the limitations of chip detection technology, a total of 673 lipid-related genes in 49 AMI samples and 50 normal samples were used to analyze DEGs. After data normalization and removal of batch differences, 50 lipid-related genes were identified, 28 upregulated and 22 downregulated (Table 1). These 50 lipid-related DEGs can be visualized in the volcano plot and heatmap (Figure 1A, 1B). The expression pattern between the two groups was highlighted based on the box plot (Figure 2). Among them, the top three upregulated genes were ACSSL1 (Acyl-CoA Synthetase Long-Chain Family Member 1), PLBD1 (Phospholipase B Domain Containing 1), and CH25H (Cholesterol 25-Hydroxylase). Meanwhile, the top three downregulated genes were ELOVL4 (ELOVL Fatty Acid Elongase 4), TNFAIP8L2 (TNF Alpha Induced Protein 8 Like 2), and CYP2E1 (Cytochrome P450 Family 2 Subfamily E Member 1). ## Correlation analysis of lipid-related DEGs The 50 lipid-related DEGs in the GSE66360 were significantly correlated by the Pearson correlation analysis (Figure 3). The positive correlation between the PLBD1 and ACSL1 was strongest; and the CYP4F2 had the most obvious negative correlation with PTPN13. Moreover, the ACSL1 was positively related to GPCPD1 and CH25H; the CH25H was positively related to GPCPD1 and ACSL1; the PLA2G12A was positively related to GPCPD1 while negatively related to CH25H. **Figure 3:** *Pearson correlation analysis of the 50 lipid-related differentially expressed genes (DEGs).* ## Functional analyses of lipid-related DEGs The biological functions of lipid-related DEGs were determined by GO and KEGG enrichment analyses using R software. The most enriched GO terms were fatty acid metabolic, glycerolipid metabolic, and phospholipid metabolic processes (biological processes, Figure 4A); intrinsic component of endoplasmic reticulum membrane, integral component of endoplasmic reticulum membrane, and lipid droplet (cellular components, Figure 4B); oxidoreductase activity, iron ion binding, and acyltransferase activity (molecular functions, Figure 4C). The KEGG enrichment analyses showed that lipid-related DEGs were involved in the arachidonic acid metabolism, glycerophospholipid metabolism, chemical carcinogenesis-DNA adducts, and the PPAR signaling pathway (Figure 5). The details of these analyses can also be found in Supplementary Table 1. **Figure 4:** *Gene Ontology (GO) enrichment analysis of 50 differentially expressed genes (DEGs). Abbreviations: (A) BP: biological process; (B) CC: cellular component; (C) MF: molecular function.* **Figure 5:** *Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of 50 lipid-related differentially expressed genes (DEGs).* ## Screening key DEGs by LASSO logistic regression and SVM-RFE The LASSO logistic regression identified 15 lipid-related genes based on the optimum λ value (Figure 6A), whereas the SVM-RFE algorithm identified four genes (Figure 6B). Four overlap genes including ACSL1, CH25H, GPCPD1 (Glycerophosphocholine Phosphodiesterase 1), and PLA2G12A (Phospholipase A2 Group XIIA) were identified by LASSO and SVM-RFE algorithm as key lipid-related DEGs for subsequent analysis (Figure 6C). In addition, other 14 genes identified by LASSO regression are listed in Supplementary Table 2. **Figure 6:** *Identification of key lipid-related DEGs by machine learning methods. (A) Least absolute shrinkage and selection operator (LASSO) logistic regression screening of key lipid-related DEGs. (B) Support vector machine-recursive feature elimination (SVM-RFE) algorithm screening of key lipid-related DEGs. (C) Venn diagram of the intersection of diagnostic markers obtained by the two algorithms.* ## ROC curves between AMI and control groups The expression of the four lipid-related DEGs between the AMI and control samples of the GSE66360 dataset were analyzed using R software, and the ROC curves were constructed. The area under the curve (AUC), unified with specificity and sensitivity, verified the intrinsic validity of diagnostic tests [17]. Four lipid-related DEGs had a superior diagnostic value for AMI. Among them, the gene with the most significant diagnostic value was ACSL1 (AUC = 0.878). The other genes are: CH25H (AUC = 0.853), GPCPD1 (AUC = 0.819), and PLA2G12A (AUC = 0.747) (Figure 7). These four lipid-related DEGs can be considered underlying diagnostic biomarkers for AMI. **Figure 7:** *ROC curve analysis. ROC curve of ACSL1 (A), CH25H (B), GPCPD1 (C), PLA2G12A (D) in GSE66360 dataset.* ## Validation by RT-qPCR To verify the dependability of the GSE66360 dataset, RT-qPCR was used to validate the expression levels of the above four lipid-related genes in clinical samples. The clinical data of AMI and control groups are summarized in Table 2. The RT-qPCR results suggested in Figure 8, that the expression level of PLA2G12A ($$p \leq 2.8$$e-08) increased in controls compared to the AMI group, while ACSL1 ($$p \leq 2.4$$e-09), CH25H ($$p \leq 0.023$$), and GPCPD1 ($$p \leq 1.3$$e-08) were higher in AMI group. Hence, the RT-qPCR results performed were consistent with the main bioinformatic analysis. ## Verification of the potential biomarkers for AMI Furthermore, we analyzed the gene expression levels in the AMI group and healthy individuals using the ROC curve to verify the diagnostic value of the four screened lipid-related genes as shown in Figure 9. The AUC values of ACSL1, CH25H, GPCPD1 and PLA2G12A were 0.846 [$95\%$ confidence interval (CI): 0.764–0.929], 0.632 ($95\%$ CI: 0.517–0.747), 0.830 ($95\%$ CI: 0.749–0.912), and 0.822 ($95\%$ CI: 0.744–0.901), respectively. This result showed that these lipid-related DEGs are diagnostic biomarkers for AMI. **Figure 9:** *ROC curve analysis. ROC curve of ACSL1 (A), CH25H (B), GPCPD1 (C), PLA2G12A (D) in clinical samples.* ## DISCUSSION Acute myocardial infarction (AMI) refers to hypoxia caused by coronary atherosclerosis stenosis and myocardial necrosis caused by acute and persistent ischemia [18]. Dyslipidemia is a known risk factor for AMI [19], and lipid-lowering therapy is the treatment cornerstone. Several convincing studies have shown that the combined effect of lowering triglyceride, LDL cholesterol, and total cholesterol levels yield higher cardiovascular risk than lowering LDL cholesterol levels alone [20–22]. The accumulated molecular genetic data indicate that many genes are related to AMI occurrence, including lipid-related genes [23]. However, the lipid-related genes involved in AMI have not been completely identified. Thus, it is necessary to comprehend the role of lipid-related genes in AMI diagnosis and treatment. Herein, we retrieved data of AMI patients (GSE66360) and subjected it to differential genes analysis, and identified lipid-related DEGs associated with AMI. Lipid-related DEGs were then subjected to GO and KEGG enrichment analyses. LASSO regression is a machine learning method that recognizes variables by looking for a λ value for a minimal classification error [24]. SVM-RFE is another machine learning method that finds optimal variables through subtracting SVM-generated feature vectors [25]. We used these two algorithms to screen characteristic variables and created an optimal classification model. Four lipid-related genes (ASCL1, CH25H, GPCPD1, and PLA2G12A) were identified based on these two methods, which significantly impact AMI diagnosis. Moreover, the findings of CH25H were controversial compared with previous studies and should be interpreted with caution. Nevertheless, the p values of these four lipid-related genes were < 0.05, verified by RT-qPCR and consistent with our bioinformatic analysis results. ACSL1 is a key rate-limiting enzyme in lipid metabolism [26], catalyzing the energy production of fatty acids or the production of phospholipids, cholesterol esters, and triglycerides [27]. Previous studies have shown that heart-specific overexpression of ACSL1 in mice increases triglyceride accumulation in cardiomyocytes [28]. Li et al. demonstrated that inhibiting ACSL1 expression in the heart can reduce lipid metabolism and promote the regeneration of cardiomyocytes [29]. A cohort study has shown that the expression level of ACSL1 in peripheral blood leukocytes of AMI patients was higher than that of healthy controls, and this high expression was a risk factor for AMI [30]. A recent study confirmed that the overexpression of ACSL1 can reduce fatty acid β-oxidation and increase plasma triglyceride levels by regulating the PPARγ pathway, which is one of the mechanisms that can promote the pathogenesis of AMI [31]. These results supported the findings of our bioinformatic analysis and suggested that ACSL1 plays a pathological role in AMI through lipid metabolism and might be a promising AMI biomarker. Moreover, PLA2G12A is a secreted phospholipase A2, but its physiological function is largely unclear. In humans, there is a suggestive association between a PLA2G12A polymorphism and response to anti-vascular endothelial growth factor therapy in patients with exudative age-related macular degeneration [32]. Alexandros et al. showed that PLA2G12A is highly expressed in aortic endothelial cells in vivo and may inhibit atherosclerosis by reducing the adhesion properties of vascular endothelial cells, which confirmed PLA2G12A as a candidate gene for atherosclerosis protection [33]. This was consistent with our findings that PLA2G12A was downregulated in AMI samples and was a protective gene, possibly by reducing vascular adhesion to decrease AMI incidence. CH25H regulates cholesterol and lipid metabolism by converting cholesterol to 25-HC, and plays an important role in regulating cellular inflammatory states and cholesterol biosynthesis in endothelial cells and monocytes [34]. CH25H and 25-HC were traditionally regarded as key regulators to maintain cholesterol homeostasis by inhibiting sterol regulator-binding protein (SREBP) and liver X receptor (LXR) [35]. Elizabeth et al. showed that 25-HC production promotes the formation of macrophage foam cells and increases susceptibility to atherosclerosis, thereby increasing AMI risk [36]. However, the pro-inflammatory role of CH25H in atherosclerosis remains controversial. Other studies have shown that CH25H is involved in macrophages’ functional endothelium and anti-inflammatory phenotype and that CH25H ablation increases susceptibility to atherosclerosis [37]. Our current study suggested that CH25H was upregulated in AMI samples, consistent with the Elizabeth et al. results. This contradiction might be partly due to different experimental conditions requiring further study. Additionally, GPCPD1 is a key enzyme in choline and phospholipid metabolism. GPCPD1 has also been reported to be involved in the complex network of enzymatic reactions regulating choline metabolism [38]. It can cleave glycerophosphocholine to form glycerol-3-phosphate and choline [39]. GPCPD1 has been reported to promote cell migration, metastasis, adhesion, and diffusion in breast, endometrial, and ovarian cancers. However, its biological role in cardiovascular disease remains unclear. Hence, more studies are needed to further verify our current findings. In the current research, several correlations between four key lipid-related DEGs were also noticed. Correlation analysis indicated that the ACSL1, CH25H, GPCPD1, and PLA2G12A genes may influence the occurrence of AMI by synergistically regulating the same lipid metabolic pathway. Meanwhile, the functional analyses were also performed to evaluate the potential biological functions of lipid-related DEGs. The GO enrichment analysis showed that these genes were closely related to fatty acid metabolism. Furthermore, the KEGG enrichment analysis revealed that the lipid-related genes were primarily associated with the PPAR signaling pathway. PPAR is activated by fatty acids and their derivatives, thereby creating a lipid signaling network between the cell surface and the nucleus [40]. As lipid sensors and master regulators, PPAR controls the expression of genes that function in lipid metabolism [41]. The PPAR signaling pathway, a crossing regulator of lipid signaling and inflammation, [40] was enriched, indicating that it plays a crucial role in lipid metabolism response to AMI. A previous study has found that the downregulation of PPARγ contributes to the activation and aggregation, eventually forming micro-thromboses, finally leading to myocardial dysfunction [42]. These results indicated that these lipid-related genes might affect AMI occurrence through the PPAR signaling pathway. However, further research is required to confirm the correlations between these key genes. However, our current study also has some limitations. First, we used the dataset from circulating endothelial cells to perform the bioinformatics analysis, and used the peripheral blood mononuclear cells from myocardial infarction and normal people for verification. Although there was some sample heterogeneity, our research, like other studies using GSE66360 dataset [43–45], obtained a satisfactory result, which fully supported our conclusion. However, more studies are needed to further confirm our findings. Second, the included clinical samples were relatively small. Therefore, our conclusions must be verified by a larger AMI cohort. Third, lipid-related DEGs were only confirmed in clinical samples, and their potential functions were not demonstrated in AMI cells or animal models. Hence, more in vivo and in vitro studies are needed to clarify the underlying mechanisms of these key genes in AMI. In summary, four lipid-related genes involved in AMI were confirmed by bioinformatics analysis and machine learning methods. *These* genes might influence AMI occurrence by regulating lipid metabolism. Our current findings might help understand the mechanisms of lipid metabolism-related genes in AMI and develop future lipid-lowering treatment strategies for AMI. ## Lipid-related gene dataset The workflow diagram of this research was shown in Supplementary Figure 1. A total of 742 lipid-related genes were retrieved from Gene Set Enrichment Analysis (https://www.gsea-msigdb.org/gsea/index.jsp) (Supplementary Table 3). The AMI dataset (GSE66360) was downloaded from the GEO website (https://www.ncbi.nlm.nih.gov/geo/). The GSE66360 dataset including a total of 99 circulating endothelial cell samples that collected from 49 AMI and 50 control subjects, and this dataset was based on the GPL570 platform (Affymetrix Human Genome U133 Plus 2.0 Array). A total of 21629 genes were detected in the GSE66360 dataset. ## Identification of lipid-related DEGs The “limma” R package was used to identify lipid-related DEGs between AMI patients and normal participants. The threshold values were $p \leq 0.05$ and |log fold change (FC)| > 0.585 [46–48]. Heatmaps, volcano plots, and boxplot charts were plotted using “heatmap” and “ggplot2” R packages. ## Functional enrichment analysis of lipid-related genes The GO and KEGG pathway enrichment analyses were conducted using the “enrichplot” R package. Cell composition, biological processes, and molecular functions were included in the GO analysis. ## Screening of lipid-related genes through SVM-RFE and LASSO logistic regression The “glmnet” R package was used to perform LASSO logistic regression which the response type set as binomial and alpha set as 1 to identify lipid-related genes [49]. LASSO regression is a regularized penalty regression method, combining ridge regression and subset selection. It applies ordinary least squares, but the sum of absolute values of the regression coefficients is less than the predetermined constant value [50]. LASSO logistic regression is a generalization of the binomial distribution of the LASSO output variable. Herein, we used LASSO to screen lipid-related genes. Moreover, SVM-RFE is a machine learning method based on support vector machines that identify optimal variables by removing SVM-generated feature vectors [51], and the thresholds were set as follows: halve.above = 100 and $k = 5.$ The “E1071” R package was used to establish the SVM module to sift lipid-related genes. Then, the intersections of lipid-related genes sifted by LASSO and SVM-RFE were applied to AMI diagnostic analysis, and the ROC curves were plotted. ## Clinical validation samples From September 2021 to May 2022, 50 AMI patients (AMI group) and healthy participants (control group) were recruited from the Hunan Provincial People’s Hospital. The blood samples were collected from AMI patients within hours of admission with chest pain and before using antiplatelet or anticoagulant to eliminate the influence of possible changes in blood status after pharmacological intervention. All AMI patients underwent percutaneous coronary intervention (PCI) within 12 h of the chest pain onset. The AMI patients were diagnosed based on the 2018 guidelines for diagnosing AMI patients [52]. A total of 50 healthy individuals were enrolled in the hospital physical examination center in the same period. The exclusion criteria were: (i) active inflammation; (ii) patients receiving thrombolysis and with other underlying heart diseases (e.g., severe valvular abnormalities, cardiomyopathy, or congenital heart disease); and (iii) patients who had hepatic and/or renal dysfunction, tumors, and autoimmune diseases. All participants provided written informed consent before the beginning of the study. This research was approved by the Ethics Committee of the Hunan Provincial People’s Hospital (approval number: [2021]-41). ## Real-time quantitative polymerase chain reaction (RT-qPCR) Peripheral blood was obtained from blood samples of patients using RNeasy™ Mini Kit (QIAGEN, Frankfurt, Germany) to extract total RNA. Total RNA was reverse transcribed into cDNA using the PrimeScript RT reagent kit (Takara Bio, Japan). RT-qPCR was performed with a LightCycler 480 II Real-time PCR instrument (Roche, Switzerland) using the TransStart Top Green qPCR SuperMix (AQ131-03, Transgen, Beijing, China). ## Statistical analysis All bioinformatics and Pearson’s correlation analyses were performed using R software (version 4.6.0, http://www.R-project.org). SPSS software (version 22.0) was used to analyze clinical data. Clinical characteristic data were analyzed using Student’s t-test and χ2 test. R and Grap Pad Prism software were used for ROC curve analysis. A $p \leq 0.05$ was considered statistically significant. ## References 1. Yamada Y, Matsui K, Takeuchi I, Fujimaki T. **Association of genetic variants with coronary artery disease and ischemic stroke in a longitudinal population-based genetic epidemiological study.**. *Biomed Rep* (2015) **3** 413-9. DOI: 10.3892/br.2015.440 2. Abram S, Arruda-Olson AM, Scott CG, Pellikka PA, Nkomo VT, Oh JK, Milan A, Abidian MM, McCully RB. **Frequency, Predictors, and Implications of Abnormal Blood Pressure Responses During Dobutamine Stress Echocardiography.**. *Circ Cardiovasc Imaging* (2017) **10** e005444. DOI: 10.1161/CIRCIMAGING.116.005444 3. Chiu MH, Heydari B, Batulan Z, Maarouf N, Subramanya V, Schenck-Gustafsson K, O'Brien ER. **Coronary artery disease in post-menopausal women: are there appropriate means of assessment?**. *Clin Sci (Lond)* (2018) **132** 1937-52. DOI: 10.1042/CS20180067 4. Madhavan MV, Gersh BJ, Alexander KP, Granger CB, Stone GW. **Coronary Artery Disease in Patients ≥80 Years of Age.**. *J Am Coll Cardiol* (2018) **71** 2015-40. DOI: 10.1016/j.jacc.2017.12.068 5. Anderson JL, Morrow DA. **Acute Myocardial Infarction.**. *N Engl J Med* (2017) **376** 2053-64. DOI: 10.1056/NEJMra1606915 6. Andreadou I, Tsoumani M, Vilahur G, Ikonomidis I, Badimon L, Varga ZV, Ferdinandy P, Schulz R. **PCSK9 in Myocardial Infarction and Cardioprotection: Importance of Lipid Metabolism and Inflammation.**. *Front Physiol* (2020) **11** 602497. DOI: 10.3389/fphys.2020.602497 7. Michos ED, McEvoy JW, Blumenthal RS. **Lipid Management for the Prevention of Atherosclerotic Cardiovascular Disease.**. *N Engl J Med* (2019) **381** 1557-67. DOI: 10.1056/NEJMra1806939 8. Grundy SM, Cleeman JI, Merz CN, Brewer HB, Clark LT, Hunninghake DB, Pasternak RC, Smith SC, Stone NJ. **Implications of recent clinical trials for the National Cholesterol Education Program Adult Treatment Panel III Guidelines.**. *J Am Coll Cardiol* (2004) **44** 720-32. DOI: 10.1016/j.jacc.2004.07.001 9. Gotto AM, Brinton EA. **Assessing low levels of high-density lipoprotein cholesterol as a risk factor in coronary heart disease: a working group report and update.**. *J Am Coll Cardiol* (2004) **43** 717-24. DOI: 10.1016/j.jacc.2003.08.061 10. Gencer B, Koskinas KC, Räber L, Karagiannis A, Nanchen D, Auer R, Carballo D, Carballo S, Klingenberg R, Heg D, Matter CM, Lüscher TF, Rodondi N. **Eligibility for PCSK9 Inhibitors According to American College of Cardiology (ACC) and European Society of Cardiology/European Atherosclerosis Society (ESC/EAS) Guidelines After Acute Coronary Syndromes.**. *J Am Heart Assoc* (2017) **6** e006537. DOI: 10.1161/JAHA.117.006537 11. Sinnaeve PR, Donahue MP, Grass P, Seo D, Vonderscher J, Chibout SD, Kraus WE, Sketch M, Nelson C, Ginsburg GS, Goldschmidt-Clermont PJ, Granger CB. **Gene expression patterns in peripheral blood correlate with the extent of coronary artery disease.**. *PLoS One* (2009) **4** e7037. DOI: 10.1371/journal.pone.0007037 12. Muse ED, Kramer ER, Wang H, Barrett P, Parviz F, Novotny MA, Lasken RS, Jatkoe TA, Oliveira G, Peng H, Lu J, Connelly MC, Schilling K. **A Whole Blood Molecular Signature for Acute Myocardial Infarction.**. *Sci Rep* (2017) **7** 12268. DOI: 10.1038/s41598-017-12166-0 13. Ntzani EE, Ioannidis JP. **Predictive ability of DNA microarrays for cancer outcomes and correlates: an empirical assessment.**. *Lancet* (2003) **362** 1439-44. DOI: 10.1016/S0140-6736(03)14686-7 14. Ein-Dor L, Kela I, Getz G, Givol D, Domany E. **Outcome signature genes in breast cancer: is there a unique set?**. *Bioinformatics* (2005) **21** 171-8. DOI: 10.1093/bioinformatics/bth469 15. Chen R, Liu X, Jin S, Lin J, Liu J. **Machine Learning for Drug-Target Interaction Prediction.**. *Molecules* (2018) **23** 30200333. DOI: 10.3390/molecules23092208 16. Wei S, Lu J, Lou J, Shi C, Mo S, Shao Y, Ni J, Zhang W, Cheng X. **Gastric Cancer Tumor Microenvironment Characterization Reveals Stromal-Related Gene Signatures Associated With Macrophage Infiltration.**. *Front Genet* (2020) **11** 663. DOI: 10.3389/fgene.2020.00663 17. Kumar R, Indrayan A. **Receiver operating characteristic (ROC) curve for medical researchers.**. *Indian Pediatr* (2011) **48** 277-87. DOI: 10.1007/s13312-011-0055-4 18. Moreira DM, da Silva RL, Vieira JL, Fattah T, Lueneberg ME, Gottschall CA. **Role of vascular inflammation in coronary artery disease: potential of anti-inflammatory drugs in the prevention of atherothrombosis. Inflammation and anti-inflammatory drugs in coronary artery disease.**. *Am J Cardiovasc Drugs* (2015) **15** 1-11. DOI: 10.1007/s40256-014-0094-z 19. Boateng S, Sanborn T. **Acute myocardial infarction.**. *Dis Mon* (2013) **59** 83-96. DOI: 10.1016/j.disamonth.2012.12.004 20. Ferrières J, Amber V, Crisan O, Chazelle F, Jünger C, Wood D. **Total lipid management and cardiovascular disease in the dyslipidemia international study.**. *Cardiology* (2013) **125** 154-63. DOI: 10.1159/000348859 21. Chapman MJ, Ginsberg HN, Amarenco P, Andreotti F, Borén J, Catapano AL, Descamps OS, Fisher E, Kovanen PT, Kuivenhoven JA, Lesnik P, Masana L, Nordestgaard BG. **Triglyceride-rich lipoproteins and high-density lipoprotein cholesterol in patients at high risk of cardiovascular disease: evidence and guidance for management.**. *Eur Heart J* (2011) **32** 1345-61. DOI: 10.1093/eurheartj/ehr112 22. Can M, Acikgoz S, Mungan G, Ugurbas E, Ankarali H, Sumbuloglu V, Demirtas S, Karaca L. **Is direct method of low density lipoprotein cholesterol measurement appropriate for targeting lipid lowering therapy?**. *Int J Cardiol* (2010) **142** 105-7. DOI: 10.1016/j.ijcard.2008.11.141 23. Zdravkovic S, Wienke A, Pedersen NL, de Faire U. **Genetic susceptibility of myocardial infarction.**. *Twin Res Hum Genet* (2007) **10** 848-52. DOI: 10.1375/twin.10.6.848 24. Zhao YP, Wang JJ, Li XY, Peng GJ, Yang Z. **Extended least squares support vector machine with applications to fault diagnosis of aircraft engine.**. *ISA Trans* (2020) **97** 189-201. DOI: 10.1016/j.isatra.2019.08.036 25. Wan JJ, Chen BL, Kong YX, Ma XG, Yu YT. **An Early Intestinal Cancer Prediction Algorithm Based on Deep Belief Network.**. *Sci Rep* (2019) **9** 17418. DOI: 10.1038/s41598-019-54031-2 26. Goldenberg JR, Wang X, Lewandowski ED. **Acyl CoA synthetase-1 links facilitated long chain fatty acid uptake to intracellular metabolic trafficking differently in hearts of male versus female mice.**. *J Mol Cell Cardiol* (2016) **94** 1-9. DOI: 10.1016/j.yjmcc.2016.03.006 27. Li LO, Ellis JM, Paich HA, Wang S, Gong N, Altshuller G, Thresher RJ, Koves TR, Watkins SM, Muoio DM, Cline GW, Shulman GI, Coleman RA. **Liver-specific loss of long chain acyl-CoA synthetase-1 decreases triacylglycerol synthesis and beta-oxidation and alters phospholipid fatty acid composition.**. *J Biol Chem* (2009) **284** 27816-26. DOI: 10.1074/jbc.M109.022467 28. Chiu HC, Kovacs A, Ford DA, Hsu FF, Garcia R, Herrero P, Saffitz JE, Schaffer JE. **A novel mouse model of lipotoxic cardiomyopathy.**. *J Clin Invest* (2001) **107** 813-22. DOI: 10.1172/JCI10947 29. Li Y, Yang M, Tan J, Shen C, Deng S, Fu X, Gao S, Li H, Zhang X, Cai W. **Targeting ACSL1 promotes cardiomyocyte proliferation and cardiac regeneration.**. *Life Sci* (2022) **294** 120371. DOI: 10.1016/j.lfs.2022.120371 30. Yang L, Yang Y, Si D, Shi K, Liu D, Meng H, Meng F. **High expression of long chain acyl-coenzyme A synthetase 1 in peripheral blood may be a molecular marker for assessing the risk of acute myocardial infarction.**. *Exp Ther Med* (2017) **14** 4065-72. DOI: 10.3892/etm.2017.5091 31. Li T, Li X, Meng H, Chen L, Meng F. **ACSL1 affects Triglyceride Levels through the PPARγ Pathway.**. *Int J Med Sci* (2020) **17** 720-7. DOI: 10.7150/ijms.42248 32. Wang VM, Rosen RB, Meyerle CB, Kurup SK, Ardeljan D, Agron E, Tai K, Pomykala M, Chew EY, Chan CC, Tuo J. **Suggestive association between PLA2G12A single nucleotide polymorphism rs2285714 and response to anti-vascular endothelial growth factor therapy in patients with exudative age-related macular degeneration.**. *Mol Vis* (2012) **18** 2578-85. PMID: 23112570 33. Nicolaou A, Northoff BH, Sass K, Ernst J, Kohlmaier A, Krohn K, Wolfrum C, Teupser D, Holdt LM. **Quantitative trait locus mapping in mice identifies phospholipase Pla2g12a as novel atherosclerosis modifier.**. *Atherosclerosis* (2017) **265** 197-206. DOI: 10.1016/j.atherosclerosis.2017.08.030 34. Tuong ZK, Lau P, Du X, Condon ND, Goode JM, Oh TG, Yeo JC, Muscat GE, Stow JL. **RORα and 25-Hydroxycholesterol Crosstalk Regulates Lipid Droplet Homeostasis in Macrophages.**. *PLoS One* (2016) **11** e0147179. DOI: 10.1371/journal.pone.0147179 35. Zhao J, Chen J, Li M, Chen M, Sun C. **Multifaceted Functions of CH25H and 25HC to Modulate the Lipid Metabolism, Immune Responses, and Broadly Antiviral Activities.**. *Viruses* (2020) **12** 727. DOI: 10.3390/v12070727 36. Gold ES, Ramsey SA, Sartain MJ, Selinummi J, Podolsky I, Rodriguez DJ, Moritz RL, Aderem A. **ATF3 protects against atherosclerosis by suppressing 25-hydroxycholesterol-induced lipid body formation.**. *J Exp Med* (2012) **209** 807-17. DOI: 10.1084/jem.20111202 37. Li Z, Martin M, Zhang J, Huang HY, Bai L, Zhang J, Kang J, He M, Li J, Maurya MR, Gupta S, Zhou G, Sangwung P. **Krüppel-Like Factor 4 Regulation of Cholesterol-25-Hydroxylase and Liver X Receptor Mitigates Atherosclerosis Susceptibility.**. *Circulation* (2017) **136** 1315-30. DOI: 10.1161/CIRCULATIONAHA.117.027462 38. Marchan R, Büttner B, Lambert J, Edlund K, Glaeser I, Blaszkewicz M, Leonhardt G, Marienhoff L, Kaszta D, Anft M, Watzl C, Madjar K, Grinberg M. **Glycerol-3-phosphate Acyltransferase 1 Promotes Tumor Cell Migration and Poor Survival in Ovarian Carcinoma.**. *Cancer Res* (2017) **77** 4589-601. DOI: 10.1158/0008-5472.CAN-16-2065 39. Glunde K, Penet MF, Jiang L, Jacobs MA, Bhujwalla ZM. **Choline metabolism-based molecular diagnosis of cancer: an update.**. *Expert Rev Mol Diagn* (2015) **15** 735-47. DOI: 10.1586/14737159.2015.1039515 40. Wahli W, Michalik L. **PPARs at the crossroads of lipid signaling and inflammation.**. *Trends Endocrinol Metab* (2012) **23** 351-63. DOI: 10.1016/j.tem.2012.05.001 41. Michalik L, Auwerx J, Berger JP, Chatterjee VK, Glass CK, Gonzalez FJ, Grimaldi PA, Kadowaki T, Lazar MA, O'Rahilly S, Palmer CN, Plutzky J, Reddy JK. **International Union of Pharmacology. LXI. Peroxisome proliferator-activated receptors.**. *Pharmacol Rev* (2006) **58** 726-41. DOI: 10.1124/pr.58.4.5 42. Zhou H, Li D, Zhu P, Hu S, Hu N, Ma S, Zhang Y, Han T, Ren J, Cao F, Chen Y. **Melatonin suppresses platelet activation and function against cardiac ischemia/reperfusion injury via PPARγ/FUNDC1/mitophagy pathways.**. *J Pineal Res* (2017) **63** e12438. DOI: 10.1111/jpi.12438 43. Zheng PF, Liao FJ, Yin RX, Chen LZ, Li H, Nie RJ, Wang Y, Liao PJ. **Genes associated with inflammation may serve as biomarkers for the diagnosis of coronary artery disease and ischaemic stroke.**. *Lipids Health Dis* (2020) **19** 37. DOI: 10.1186/s12944-020-01217-7 44. Cheng M, An S, Li J. **Identifying key genes associated with acute myocardial infarction.**. *Medicine (Baltimore)* (2017) **96** e7741. DOI: 10.1097/MD.0000000000007741 45. Ji Z, Zhang R, Yang M, Zuo W, Yao Y, Qu Y, Su Y, Liu Z, Gu Z, Ma G. **Accuracy of triggering receptor expressed on myeloid cells 1 in diagnosis and prognosis of acute myocardial infarction: a prospective cohort study.**. *PeerJ* (2021) **9** e11655. DOI: 10.7717/peerj.11655 46. Xia C, Chen L, Sun W, Yan R, Xia M, Wang Y, Yang D. **Total saponins from Paris forrestii (Takht) H. Li. show the anticancer and RNA expression regulating effects on prostate cancer cells.**. *Biomed Pharmacother* (2020) **121** 109674. DOI: 10.1016/j.biopha.2019.109674 47. Huang J, Liu L, Qin L, Huang H, Li X. **Weighted Gene Coexpression Network Analysis Uncovers Critical Genes and Pathways for Multiple Brain Regions in Parkinson's Disease.**. *Biomed Res Int* (2021) **2021** 6616434. DOI: 10.1155/2021/6616434 48. Xu M, Li Z, Yang L, Zhai W, Wei N, Zhang Q, Chao B, Huang S, Cui H. **Elucidation of the Mechanisms and Molecular Targets of**. *Biomed Res Int* (2020) **2020** 5848497. DOI: 10.1155/2020/5848497 49. McEligot AJ, Poynor V, Sharma R, Panangadan A. **Logistic LASSO Regression for Dietary Intakes and Breast Cancer.**. *Nutrients* (2020) **12** 2652. DOI: 10.3390/nu12092652 50. Pierre JF, Akbilgic O, Smallwood H, Cao X, Fitzpatrick EA, Pena S, Furmanek SP, Ramirez JA, Jonsson CB. **Discovery and predictive modeling of urine microbiome, metabolite and cytokine biomarkers in hospitalized patients with community acquired pneumonia.**. *Sci Rep* (2020) **10** 13418. DOI: 10.1038/s41598-020-70461-9 51. Sundermann B, Bode J, Lueken U, Westphal D, Gerlach AL, Straube B, Wittchen HU, Ströhle A, Wittmann A, Konrad C, Kircher T, Arolt V, Pfleiderer B. **Support Vector Machine Analysis of Functional Magnetic Resonance Imaging of Interoception Does Not Reliably Predict Individual Outcomes of Cognitive Behavioral Therapy in Panic Disorder with Agoraphobia.**. *Front Psychiatry* (2017) **8** 99. DOI: 10.3389/fpsyt.2017.00099 52. Thygesen K, Alpert JS, Jaffe AS, Chaitman BR, Bax JJ, Morrow DA, White HD. **Fourth Universal Definition of Myocardial Infarction (2018).**. *Circulation* (2018) **138** e618-51. DOI: 10.1161/CIR.0000000000000617
--- title: Cellular senescence with SASP in periodontal ligament cells triggers inflammation in aging periodontal tissue authors: - Kuniko Ikegami - Motozo Yamashita - Mio Suzuki - Tomomi Nakamura - Koki Hashimoto - Jirouta Kitagaki - Manabu Yanagita - Masahiro Kitamura - Shinya Murakami journal: Aging (Albany NY) year: 2023 pmcid: PMC10042704 doi: 10.18632/aging.204569 license: CC BY 3.0 --- # Cellular senescence with SASP in periodontal ligament cells triggers inflammation in aging periodontal tissue ## Abstract The direct cause of periodontitis is periodontopathic bacteria, while various environmental factors affect the severity of periodontitis. Previous epidemiological studies have shown positive correlations between aging and periodontitis. However, whether and how aging is linked to periodontal health and disease in biological processes is poorly understood. Aging induces pathological alterations in organs, which promotes systemic senescence associated with age-related disease. Recently, it has become evident that senescence at the cellular level, cellular senescence, is a cause of chronic diseases through production of various secretory factors including proinflammatory cytokines, chemokines, and matrix metalloproteinases (MMPs), which is referred to the senescence-associated secretory phenotype (SASP). In this study, we examined the pathological roles of cellular senescence in periodontitis. We found localization of senescent cells in periodontal tissue, particularly the periodontal ligament (PDL), in aged mice. Senescent human PDL (HPDL) cells showed irreversible cell cycle arrest and SASP-like phenotypes in vitro. Additionally, we observed age-dependent upregulation of microRNA (miR)-34a in HPDL cells. These results suggest that chronic periodontitis is mediated by senescent PDL cells that exacerbate inflammation and destruction of periodontal tissues through production of SASP proteins. Thus, miR-34a and senescent PDL cells might be promising therapeutic targets for periodontitis in elderly people. ## INTRODUCTION Periodontitis is a chronic inflammatory disease characterized by periodontal tissue destruction with loss of tooth-supportive bone. It is thought to be the most common infectious disease and affects more than $40\%$ of people aged over 30 years [1]. Colonization of dental biofilm involving periodontopathic bacteria triggers inflammation and excessive immune responses that exacerbate breakdown of periodontal tissue. In addition to bactericidal pathogens, various environmental factors affect the pathology and progression of periodontal disease. In particular, aging has been recognized as a major risk factor that affects the onset and severity of periodontitis [2]. Thus, understanding the biological mechanisms that regulate periodontal tissue and health by aging is an urgent issue to establish preventive protocols or specialized therapies for elderly persons in the field of periodontal medicine. In the process of aging, accumulated environmental stresses induce degeneration of multiple organs, which accelerates the morbidity and severity of lifestyle diseases [3]. Numerous epidemiological studies and diseased animal models have shown positive correlations between periodontitis and age-dependent, lifestyle-related diseases such as type 2 diabetes, obesity, rheumatoid arthritis, and heart infarction [4, 5]. Therefore, periodontitis and these lifestyle-related diseases would share common pathology for disease development. Previous attempts to clarify the effects of aging on homeostasis of periodontal tissue are limited [6–8] and periodontal diseases have not been fully elucidated at the molecular level. Cellular senescence is a major hallmark of senescence in organs and the whole body. Accumulated senescent cells in aged organs and tissues induce senescence of the body [9]. Cellular senescence is defined as a state of irreversible cell cycle arrest, but not apoptosis, in mammalian cells. Initially, cellular senescence was thought to be an intrinsic cellular mechanism to escape tumorigenesis. A large number of studies have indicated that senescent cells secrete various proteins such as proinflammatory cytokines, chemokines, growth factors, and metalloproteinases, termed SASP (senescence-associated secretory proteins) [10]. First, SASP was considered to be the prominent mechanism for recruitment of immunocompetent cells to eliminate tumor cells in aged organs. Recent studies have shown that senescent cells induce inflammation and impair wound healing in various chronic diseases, such as rheumatoid arthritis, atherosclerosis, and osteoporosis, through induction of SASP [5, 11]. Therefore, understanding cellular senescence is required to develop more effective therapies and prevention protocols for age-dependent, lifestyle-related diseases. However, whether and how cell types within periodontal tissue undergo cellular senescence with SASP have not yet been clarified. Periodontal ligament (PDL) is a soft fibrous tissue located between the tooth cementum and alveolar bone. PDL cells produce extracellular matrix (ECM) proteins, such as type I/III collagens and fibronectin, to maintain physiological elasticity and manage the mechanistic occlusal force in PDL [12]. PDL not only physically supports teeth, but also plays many biological roles in periodontal tissue. For example, PDL cells are responsible for local immune responses by producing various cytokines and adhesion molecules to act as a biological barrier in orchestration with other cells in periodontal tissue [13]. Additionally, multipotent mesenchymal stem cells in PDL tissue proliferate or differentiate for wound repair and tissue regeneration [14, 15]. Therefore, maintaining homeostasis in PDL is important for periodontal tissue. Impaired PDL is thought to be a risk to periodontal health, since defects in PDL at cellular levels can trigger the breakdown of periodontal tissue, especially in aged people. microRNAs (miRNAs), a class of small non-coding RNAs expressed in eukaryotes, are 20–22 nucleotide, endogenous single-stranded RNAs [16]. miRNAs suppress gene expression by inhibiting translation of their target genes and degrading target mRNAs through binding to complementary sites located in the 3′-untranslated regions of target mRNAs in a highly context-dependent manner. More than 2000 miRNAs have been identified and each miRNA can target hundreds of genes on basis of their short sequence. Thus, miRNAs are proposed to regulate various biological processes such as development, tumorigenesis, and organism aging through modulation of inflammation and cellular senescence [17–19]. In the past two decades, many kinds of miRNAs have been identified as causes of chronic inflammation [20] and many miRNAs target the NF-κB pathway, such as miR-146a/b, miR-155, and miR-21, which are thought to modulate the inflammatory response in various cell types [21–23]. Recent studies have indicated possible relationships between miRNAs and periodontitis [24]. Therefore, identification of senescent HPDL cells and elucidation of their molecular mechanisms, including miRNAs, are required to better understand periodontitis. In this study, we aimed to clarify the pathophysiological roles of cellular senescence in periodontal tissue and diseases. Identification of the functions of cellular senescence in periodontal tissues may facilitate gaining deep insights into inflammation and destruction of periodontal tissues in elderly people. Development of therapies targeting senescent cells in periodontal tissues may also be an effective strategy for elderly people. ## Analysis of periodontal tissue in aged mice Epidemiological studies have suggested a strong correlation between aging and periodontal diseases [25]. First, we compared alteration of the bone volume in alveolar bone at the maxilla between 6-week-old (young) and 68-week-old (aged) C57BL/6 male mice by μCT analysis. Aged mice showed a decreased level of the alveolar bone crest with a flattened shape of the tooth cusp (Figure 1A). Bone resorption in the horizontal direction was apparent in supportive alveolar bone surrounding molars in aged mice. Analysis of digital images indicated nearly two-fold bone resorption at molars of aged mice compared with young mice (Figure 1A, Box-and-whiskers plots, young mice; $$n = 7$$, aged mice; $$n = 11$$). To examine whether cellular senescence was involved in the pathology of periodontal diseases in aged mice, we performed β-galactosidase (β-gal) staining of periodontal tissues that were used for μCT analysis. Because enhanced β-gal activity in lysosomes at pH 6.0, SA (senescence-associated) β-gal is a general characteristic of cellular senescence in vitro and in vivo [26]. In aged mice, many SA β-gal-positive cells were found in periodontal tissue, but few cells were found in young mice. Interestingly, SA β-gal-positive cells were mainly localized in the periodontium, but not at gingival connective tissues in aged mice (Figure 1B, Right graph). High magnification images indicated that periodontal ligament cells and endothelial cells around blood vessels were positive for SA β-gal at PDL in aged mice. To confirm this, we performed immunohistochemistry for senescent markers p16 [27], lamin A/C [28], and sirtuin 1 (SIRT1) [29]. In mice aged over 100 weeks, p16 expression was increased ($p \leq 0.01$) and SIRT1 expression was significantly decreased in periodontium ($p \leq 0.01$) (Figure 1C). Lamin A/C expression was not changed. Consistent with these findings, p16 mRNA expression was increased in PDL tissue dissected from aged mice. Intriguingly, significant upregulation of the pro-inflammatory cytokine interleukin (IL)-6 was observed in PDL of aged mice ($p \leq 0.05$) (Figure 1D, young mice; $$n = 7$$, aged mice; $$n = 6$$). Thus, senescent cells had accumulated in the periodontium, which may induce inflammation to trigger the breakdown of alveolar bone in aged PDL tissue. **Figure 1:** *Analysis of periodontal tissue in aged mice. (A) Micro CT (μCT) analysis of alveolar bone in the upper jaw. Representative images of young (6-week-old) and aged (68~104 weeks-old) mice are shown. Scale bar = 1 mm. Quantification of the bone resorption rate in supportive alveolar bone was evaluated. Distance between the cement–enamel junction to the crest of alveolars bone at the a, mesial root at first molars, b, distal root at first molars, and c, mesial root at second molars. Box-and-whiskers plots shows median, 25th and 75th percentile with whiskers at the 5th and 95th percentile. Young mice; n = 7, aged mice; n = 11 Statistical analysis was completed using welch’s t-test, with p-values < 0.05 were considered statistically significant (*p < 0.01, #p < 0.05). (B) X-gal staining of frozen sections of periodontal tissue of the mesial root at first molars in the upper jaw. (×40) White arrows indicate SA-β-gal positive cells. Scale bar = 500 μm. Right panels show the enlarged image of the bold square in the left panel (×100). Scale bar = 100 μm. Abbreviations: D: Dentin; PD: Periodontal ligament; AB: Alveolar Bone. Representative data from three experiments are shown. Right graph shows the percentage of SA-β-gal positive cells in PDL of young or aged mice (*p < 0.01). (C) Representative immunohistochemistry images of p16, lamin A/C, and SIRT1 (40×). Scale bar = 200 μm. Abbreviations: D: dentin; PD: periodontal ligament; AB: alveolar Bone. Yellow arrow: antibody-positive cells (40×). Representative data from three experiments are shown. Lower graph shows percentages of antibody-positive cells in PDL area of young or aged mice (*p < 0.01). (D) Expression of p16, p21, and IL-6 in PDL tissue of young (6~13-weeks-old, n = 7) and aged (68~104-weeks-old, n = 6) mice. Expression of p16, p21, and IL-6 mRNA in PDL derived from freshly isolated mouse teeth was analyzed by qRT-PCR (#p < 0.05). Representative data from three experiments are shown.* ## Establishment of a senescent model of human PDL (HPDL) cells Molecular mechanisms underlying biological alterations and degeneration of tissue and organs with aging are gradually becoming clear. Among these, cellular senescence has emerged as a precise event in tissues and organs induced by aging, and has been recognized as a major cause of senescence in the body. To understand the characteristics of senescent cells in PDL, we applied replicative senescence to primary human PDL (HPDL) cells in vitro. Repetition of serial passaging is a general method to induce cellular senescence in cultured cells by cellular replication [30]. Somatic HPDL cells showed a maximum of around 40 population doublings (PDs) in vitro (Figure 2A). In accordance with the progression of passaging, broad, enlarged, and flattened morphological changes were apparent in HPDL cells during the cultivation period (Figure 2D). The growth rate of HPDL cells was reduced gradually and then the proliferative capacity almost reached its limit at around P35 that corresponded to 45 PDs (Figure 2A). HPDL cells at >P40 showed irreversible cell growth arrest defined as stable cell cycle exit, but not cell death. Attenuation of the growth rate in HPDL cells indicated induction of cellular senescence in a physiological manner (Figure 2A). **Figure 2:** *Establishment of senescent HPDL cells in vitro. (A) Long-term growth curve of primary human periodontal ligament (HPDL) cells. Cumulative population doublings (PDs) in each cell passage were estimated in long-term cultures. Final numbers of HPDL cells at the indicated passage are shown. P6, P30, and P40 represent early, premature, and late senescence of HPDL cells in vitro. Representative data from three experiments are shown. (B) SA β-gal staining of YPDLs and APDLs. Scale bar = 50 μm. White bar: YPDLs; black bar: APDLs (C) Quantification of SA β-gal-positive YPDLs and APDLs (*p < 0.01). Representative data from three experiments are shown. (D) Phalloidin staining of P7, P27, and P40 HPDL cells (×400). Scale bar = 200 μm. Representative data from three experiments are shown. (E) Quantification of the size of HPDL cells at P7, P20, and P35 HPDL cells. FSC and SSC of flow cytometric analysis are shown. Representative data from three experiments are shown. (F) Representative transmission electron microscopy images of mitochondria in YPDLs and APDLs. White arrows indicate lamellar shaped mitochondria. Black arrows indicate disorganized mitochondria (×31800). Scale bar = 500 nm (G) Transmission electron microscopy of induction of aggregated chromosomal DNA in YPDLs and APDLs. (×1760). Scale bar = 10 μm. (H) Analysis of SAHF in YPDLs and APDLs. DAPI staining of YPDLs and APDLs (×1000). Scale bar = 25 μm. (I) Confocal image of γH2AX staining in YPDLs and APDLs. Red: γH2AX; Green: Actin fiber; Blue: DAPI staining (×400). Scale bar = 25 μm. (J) Protein expression of cell cycle arrest-related factors p53, p21, p16, and Rb in P8, P17, P25, and P33 HPDL cells. β-actin was used as a loading control. Representative band images are shown, and the relative protein levels were quantified (*p < 0.01). (K) Increased expression of senescence-related biomarkers in YPDLs and APDLs. Relative mRNA expression of p16, p21, p53, and klotho to HPRT in HPDL cells quantified by qRT-PCR. Gray bar: YPDLs; Black bar: APDLs. Data are presented as the mean ± SE (*p < 0.01). Representative data from three experiments are shown.* To confirm cellular senescence of HPDL cells in vitro, we examined SA β-gal activity in HPDL cells (Figure 2B). Around $70\%$ of aged HPDL cells (APDLs, >P30) were positive for SA β-gal, where <$10\%$ of young HPDL cells (YPDLs, < P10) were positive for SA β-gal (Figure 2C). To characterize the morphological changes of HPDL cells, phalloidin staining was performed to visualize actin stress fibers. P7 HPDL cells clearly demonstrated a spindle-like cell shape with a compacted cell size, whereas P40 HPDL cells showed an enlarged cell shape with a spread shape (Figure 2D). FCM analysis confirmed the increase in cell size (FSC) and granularity (SSC) of APDLs compared with YPDLs at the single cell level (Figure 2E). ## Senescent HPDL cells produce ROS Senescent cells partly indicate metabolic changes such as impaired energy metabolism, autophagy, and glycolysis [31, 32]. To examine such changes, we observed the anatomical features of mitochondria, the major ATP/ADP power producers, by TEM analysis. Most mitochondria in APDLs showed a shortened bulge-like shape that was quite different from that of YPDLs, mitochondrial cristae in APDLs had a disrupted structure with ladder-like repeats in the short axis, whereas mitochondrial cristae in YPDLs had a normal structure with a longitudinal elongated shape (Figure 2F). Irregularly shaped mitochondria suggest damage, which produces excess ROS with failure of the redox balance. Consistent with the above observations, CM-H2CDFDA, which is a cell-permeable ROS indicator, stained accumulated cytosolic ROS in APDLs more strongly than in YPDLs (Supplementary Figure 1A, 1B). ## Dysregulation of the chromatin structure in senescent HPDL cells Robust compaction of chromosomes in the nucleus indicates heterochromatin formation [33]. Because alteration of the epigenetic landscape is a hallmark of cellular senescence in aging, we examined formation of senescence associated heterochromatin foci (SAHF) in HPDL cells. APDLs showed SAHF with apparent chromatin aggregation recognized by TEM analysis, which were hardly stained by DAPI (Figure 2G, 2H). Furthermore, DNA damage marker histone protein γH2AX was increased in APDLs (Figure 2I). ## Expression of cell cycle regulator proteins is enhanced in senescent HPDL cells To gain molecular insights into the irreversible cell cycle arrest of senescent HPDL cells, we evaluated the expression of major cell cycle regulators, namely p53, p16, p21, and Rb. Two major pathways – p53-p21 and p16Ink4a-RB effector pathways – contribute to cell cycle arrest at G1/S phase through inhibition of CDK2 or CDK$\frac{4}{6}$ [27]. Protein expression levels of p53, p16, p21, and Rb were increased with the progression of cell replication in HPDL cells (Figure 2J). APDLs showed increased mRNA expression of P21, P16 and p53, whereas Klotho expression was decreased compared with YPDLs (Figure 2K). These data suggest that APDLs, which was induced by >30 serial passages in vitro, satisfied the general characteristics of senescent cells. ## Senescent HPDL cells produce SASP proteins Recent studies have revealed a notable function of senescent cells, namely that they secrete various proinflammatory cytokines termed SASP, which affect their neighboring cells. SASP has been reported in various cell types such as fibroblasts [3], epithelial cells [3], vascular endothelial cells [34], and immunocompetent cells [35], and are considered to induce age-dependent inflammation and tissue degeneration in organs. To examine SASP in HPDL cells, we evaluated expression of IL-6 and IL-8 that are major constituents of SASP. Expression levels of IL-6 and IL-8 mRNAs in >P30 HPDL cells were higher than those in early passaged HPDL cells. Expression of IL-6 and IL-8 was increased with progression of cell replication in HPDL cells. Moreover, high production of IL-6 and IL-8 proteins in APDLs was confirmed by ELISAs of culture supernatants (Figure 3A, 3B). Additionally, we performed an antibody-captured cytokine array to monitor highly secreted cytokines other than IL-6 and Il-8 in senescent HPDL cells. As a result, in addition to IL-6 and IL-8, we found that CXCL1, GRO (growth-regulated oncogene)-α, MIF (macrophage migration inhibitory factor) and PAI-1 (plasminogen activator inhibitor-1) in APDLs (P30) were higher than those in YPDLs (P12) (Figure 3C). Density analysis of each spot indicated significant increases in APDLs (right panels, Figure 3C). Moreover, we examined expression of matrix metalloproteinases (MMPs) in senescent HPDL cells. Previous studies have demonstrated enhanced production of MMPs from senescent fibroblasts [36]. mRNA expression of MMP-1–3 and tissue inhibitor of MMPs (TIMP)-1 and 2 in APDLs was significantly higher than that in YPDLs (Figure 3D). Analyses of pro-MMP-1 and MMP-2 protein by ELISAs supported these results (Figure 3E). Furthermore, zymography confirmed enhanced enzymatic activity of pro-MMP-1, pro-MMP-3, pro-MMP-2, and MMP-2 in culture supernatants of APDLs (Figure 3F). Thus, senescent HPDL cells produced various SASP factors, including inflammatory cytokines, chemokines, and MMPs/TIMPs, which affect chronic inflammation. **Figure 3:** *Increased expression of IL-6 and IL-8 in senescent HPDL cells. (A) Relative mRNA expression of IL-6 and IL-8 in various passages of HPDL cells quantified by qRT-PCR (*p < 0.01 vs. P5). (B) IL-6 and IL-8 in conditioned medium in YPDLs and APDLs (*p < 0.01). (C) Enhanced production of SASP factors in senescent HPDL cells. Soluble factors secreted by P9, P18, and P30 HPDL cells were detected by an antibody dot blot array. In right panels, quantification of signal intensity of dots plots assay for conditioned medium of YPDLs and APDLs. Signal intensities of the major dot blots were normalized against control spots in each blot and shown as bar graphs (Groa, IL-1ra, IL-6, IL-8, MIF, PAI-1). Gray bars indicate YPDLs (P9) and black bars indicate APDLs (P30). Representative data from three experiments are shown. (D) Relative mRNA expression of MMP-1–3 and TIMP-1 and -2 in HPDL cells quantified by qRT-PCR (*p < 0.01). (E) Pro-MMP-1 and MMP-2 in conditioned medium of P10, P18, P26, and P32 HPDL cells (*p < 0.01 vs. P10). (F) Inverted images of zymography for conditioned medium of P10, P18, P26, and P32 HPDL cells. Dark spots indicate Pro-MMP-1–3 and MMP-2. Representative data from three experiments are shown.* ## Sterile inflammatory phenotype of senescent HPDL cells Inflammation in aged organs is partially characterized as sterile inflammation evoked without apparent infection by pathogens [37]. To determine whether this occurred in periodontitis of aged individuals, we examined the inflammatory response in senescent HPDL cells with or without bactericidal stimulation. APDLs showed upregulation of IL-6 mRNA expression in the steady state as described above. Notably, *Porphyromonas gingivalis* (P.g.) LPS, which is the major periodontopathic bactericidal pathogen, did not significantly enhance IL-6 mRNA expression; although, the inflammatory cytokine IL-1β induced IL-6 expression in HPDL cells in vitro. Moreover, P.g. LPS combined with IL-1β stimulation did not enhance IL-6 expression in our system (Figure 4A, Supplementary Figure 2). We confirmed this finding at the protein level by ELISA (Figure 4B). These results suggest that the intrinsic inflammation state of APDLs is higher than YPDLs and susceptibility to bactericidal pathogens but not inflammatory cytokine is low in APDLs. **Figure 4:** *IL-6 production induced by proinflammatory cytokines and bacterial pathogens in senescent HPDL cells. (A) Relative mRNA expression of IL-6 stimulated by IL-1β (1 ng/ml) and P.g LPS (1 μg/ml) in YPDLs and APDLs quantified by qRT-PCR (*p < 0.01). (B) IL-6 and IL-8 in conditioned medium of YPDLs and APDLs quantified by ELISA (*p < 0.01). Representative data from three experiments are shown.* ## Expression of microRNAs in senescent HPDL cells To clarify the molecular mechanism regulating cellular senescence of HPDL cells, we focused on miRNAs. To date, the roles of miRNAs in senescent characteristics of HPDL cells have not been clarified well. First, we performed comprehensive analysis of miRNAs in HPDL cells by comparing the miRNA expression profiles of HPDL cells at P5, 6, 7, 10, 15, 18, 32, and 34. We found that around 360 miRNAs among 2000 human miRNAs were significantly expressed in HPDL cells. Hierarchical analysis showed that HPDL cells with a close passage number shared similar miRNA expression profiles (Supplementary Figure 3A). This result indicated that replicative senescence of HPDL cells induced by our protocol was appropriate. To classify the variations in miRNA expression profiles during the process of cellular senescence, we performed K-means clustering analysis. Among the classified eight patterns, we focused on the miRNA group that increased along with passaging (Supplementary Figure 3B). In this group, seven miRNAs had >2-fold increased expression in P34 HPDL cells (Table 1). Ingenuity pathway analysis (IPA) revealed some miRNAs such as miR-146a and miR-34a that target the inflammation pathway in aged HPDL cells (Supplementary Figures 4 and 5). **Table 1** | miRNA | P34/P5 (log2) | | --- | --- | | miR-137 | 7.6 | | miR-146a | 7.2 | | miR-181a-5p | 2.0 | | miR-181a-1-3p | 1.5 | | miR-34a | 1.2 | | miR-2682 | 1.1 | | miR-127 | 1.1 | | miR-329 | 1.1 | ## Negative regulation of IL-6 by miR-146a in senescent HPDL cells First, we compared expression of miR-146a using of miRNA array datasets and then performed validation by qRT-PCR (Supplementary Figure 6A). The peak of miR-146a expression in each passage of HPDL cells was delayed compared with the peak of IL-6 expression (left graph, Supplementary Figure 6A). Consistent with a previous study, miR-146a may regulate the senescence phenotype of HPDL cells by silencing IL-6 expression through a negative regulatory mechanism in inflammation of senescent HPDL cells. To analyze the miR-146a function in HPDL cells, we introduced synthetic mimic or inhibitor oligos for miR-146a into HPDL cells (Supplementary Figure 6B). The miR-146a mimic inhibited protein but not mRNA expression of IL-6 in both YPDLs and APDLs. In contrast, effect of the miR-146a inhibitor was not concordant. ## Induction of IL-6 through suppression of SIRT1 by miR-34a in senescent HPDL cells The induction mechanism of robust IL-6 production in senescent HPDL cells could not be fully explained by miR-146a alone. Therefore, we examined the role of miR-34a in senescent HPDL cells. As shown in Figure 5A, the expression level of miR-34a was increased dramatically in P32 HPDL cells in the miRNA array dataset, which was confirmed by qRT-PCR analysis. Intriguingly, protein and mRNA expression levels of SIRT1 were sharply decreased in APDLs (Figure 5B). SIRT1 is a homolog of the Saccharomyces cerevisiae, Sir2 protein, a member of the sirtuin family [38], which promotes longevity in many organisms [29, 39]. It has been reported that miR-34a regulates expression of SIRT1 [40]. Introduction of the miR-34a mimic into HPDL cells attenuated expression of SIRT1 (Figure 5C). Additionally, the miR-34a inhibitor rescued SIRT1 expression and suppressed IL-6 expression in APDLs (Figure 5D). These results suggested that the high secretion of IL-6 from senescent HPDL cells was induced by miR-34a through suppression of SIRT1. **Figure 5:** *Increased expression of miR-34a in senescent HPDL cells. (A) Expression of miR-34a was increased depending on the passage of HPDL cells. Scores of the fluorescence intensity of labeled miR-34a in miRNA array analysis of P5, P6, P10, P18, P32, and P34 HPDL cells is displayed in the histogram. Right graph shows the expression of miR-34a in YPDLs (P6) and APDLs (P34) analyzed by qRT-PCR (*p < 0.01). (B) Decreased expression of SIRT1 in senescent HPDL cells. Expression of SIRT1 protein in P8, P17, P25, and P33 HPDL cells analyzed by western blotting. β-Actin was used as a loading control and the relative protein levels were quantified. (*p < 0.01 vs. P8) Right graph shows the expression of SIRT1 mRNA in YPDLs and APDLs measured by qRT-PCR (*p < 0.01). (C) Overexpression of miR-34a inhibited SIRT1 expression in HPDL cells. MiR-34a mimic and anti-miR-34a oligonucleotides were transfected into YPDLs and APDLs. Expression of SIRT1 protein was analyzed by western blotting. β-Actin was used as a loading control and the relative protein levels were quantified. Right graph shows the expression of SIRT1 measured by qRT-PCR. (*p < 0.01 vs. control) (D) Overexpression of miR-34a upregulated IL-6 expression in HPDL cells. Expression of IL-6 was measured by qRT-PCR (*p < 0.01, #p < 0.05 vs. control). (E) Overexpression of miR-34a upregulated NF-κB activity in HPDL cells. NF-κB transcription activity was analyzed by a luciferase reporter assay (*p < 0.01, #P < 0.05 vs. control). (F) TSA treatment induced IL-6 in YPDLs (P6). Expression of IL-6 and SIRT1 mRNA in YPDLs after TSA treatment (0, 50, 200, and 400 nM) was quantified by qRT-PCR (*p < 0.01 vs. none). (G) Expression of SIRT1 in YPDLs and APDLs after TSA treatment (0, 50, 200, and 400 nM). Expression of SIRT1 protein was analyzed by western blotting. β-Actin was used as a loading control and the relative protein levels were quantified. Representative data from three experiments are shown.* ## Enhanced NF-κB activity in senescent HPDL cells SIRT1 is a nicotinamide adenine dinucleotide-dependent class 3 histone deacetylase (HDAC). SIRTs deacetylate the lysine residue in the histone tail of various target genes and cellular proteins [41, 42]. Thus, we hypothesized that miR-34a may regulate IL-6 transcription in senescent HPDL cells via SIRT1 through its HDAC activity. To test this hypothesis, we measured transcriptional activity of endogenous nuclear factor-κB (NF-κB), which plays an important role in transcription of IL-6, in HPDL cells by a reporter assay. As shown in Figure 5E, APDLs had intrinsic high NF-κB activity. MiR-34a mimic treatment induced high NF-κB activity in YPDLs. In contrast, miR-34a inhibitor treatment significantly suppressed NF-κB activity in APDLs. These results suggested that miR-34a induced the transcription of IL-6 through regulation of NF-κB activity. Moreover, we found that trichostatin A (TSA) at 400 nM treatment induced IL-6 expression in YPDL cells without affecting protein expression of SIRT1 (Figure 5F, 5G). TSA is an anti-fungal antibiotic isolated from *Streptomyces hygroscopicus* and reported as a specific inhibitor of histone deacetylases (HDACs) in mammalian cells. TSA selectively inhibits enzymatic activities of class 1 and 2 HDACs, but not class 3 HDAC SIRT1 [43]. Therefore, our results suggest that IL-6 production in HPDL cells was induced by either SIRT1-dependent or -independent mechanisms of histone acetylation. ## SIRT1 suppresses IL-6 in an epigenetic manner in senescent HPDL cells To confirm our findings, we examined IL-6 production in SIRT1-deficient senescent HPDL cells. SIRT1 expression in YPDLs was higher compared with APDLs. Si-SIRT1 treatment clearly suppressed SIRT1 expression in HPDLs at protein level (Figure 6A). Consistent with this finding, si-SIRT1 treatment significantly enhanced IL-6 production in both of YPDLs and APDLs ($p \leq 0.01$) (Figure 6B). Treatment with resveratrol (RSV), a polyphenol and well-known activator of SIRT1, slightly increased SIRT1 protein expression (Figure 6C) and inhibited IL-6 and IL-8 production in both of YPDLs and APDLs ($p \leq 0.01$) (Figure 6D). These results suggest that a sufficient level of SIRT1 activity was required to maintain IL-6 at low levels in HPDL cells. **Figure 6:** *SIRT1 regulates IL-6 production in senescent HPDL cells. (A) Expression of SIRT1 after si-SIRT1 or si-control transfection into YPDLs and APDLs. Β-Actin was used as a loading control and the relative protein levels were quantified (*p < 0.01 vs. control). (B) Expression of IL-6 after si-SIRT1 or control transfection into YPDLs and APDLs. IL-6 production from YPDLs and APDLs was measured by an ELISA (*p < 0.01 vs. control). (C) Expression of SIRT1 after SIRT1 activator, resveratrol treatment, at the protein level measured by western blotting. Β-Actin was used as a loading control and the relative protein levels were quantified (*p < 0.01, #P < 0.05 vs. none). (D) IL-6 and IL-8 productions from YPDLs and APDLs after resveratrol treatment (0, 50 and 100 μM) (*p < 0.01 vs. none). Abbreviation: RSV: Resveratrol. Representative data from three experiments are shown.* ## DISCUSSION Physiological roles of cellular senescence in the maintenance of periodontal tissue homeostasis and pathogenesis of periodontitis have not yet been fully elucidated. In this study, we observed many senescent PDL cells in the periodontal tissue of aged mice with apparent alveolar bone loss. The ratio of SA β-gal-positive cells in the periodontium was greater than that in the bone and gingival connective tissue. In addition, we determined miR-34a partly regulated SASP trough the regulation of NF-κB by SIRT1 in HPDL cells. Previous studies reported a rejuvenating effect of rapamycin and a relationship with TLR9 in periodontal aging [6–8]. However, our findings suggest a novel mechanism for the pathology of chronic inflammation in periodontal tissue that is partly mediated by senescent PDL cells with SASP. To the best of our knowledge, this is the first study to identify: 1) the potential for senescent PDL cells to induce inflammation of periodontal tissue, and 2) a miRNA-dependent molecular mechanism of SASP in senescent PDL cells. Aged mice showed apparent alveolar bone resorption. Notably, bone resorption in aged mice was induced without artificial infection of bacteria such as P. gingivalis. SA β-gal-positive senescent cells were found in alveolar bones, periodontal ligament, tooth pulp, and gingival connective tissue. In particular, periodontal ligament showed a large number of SA β-gal-positive cells (Figure 1B). These results suggest that chronological aging accelerates organ senescence in periodontal ligament even without attachment of periodontopathic bacteria. Our results are consistent with these findings, namely the observance of hyposensitivity of APDLs against P.g. LPS (Figure 4). Additionally, endothelial cells and perivascular endothelial cells were positive for SA β-gal in periodontal ligament. Accumulation of senescent endothelial cells has been reported in the coronary artery wall of elderly people [34]. Therefore, we think the localization of SA β-gal-positive cells in periodontal ligament suggests similar pathologies of vascular defects, artery infarction, and periodontitis. Recently, cellular senescence was found in early embryogenesis and wound healing processes of mice [44]. We believe the requirement of cellular senescence for the development and maintenance of PDL under physiological conditions, but not the senescence of organs, requires further study. Taken together, our findings suggest fragility points in aged periodontal tissue and could be applied to new prevention methods or periodontal therapies by targeting senescent cells in aging periodontal ligament. To examine cellular senescence in periodontal ligament in vivo, we induced replicative senescence with telomere shortening in HPDL cells by continuous passaging. However, induction of SA β-gal was not achieved in all APDLs (Figure 2B). Previous studies have revealed that SASP proteins act on their originating or neighboring cells to promote cellular senescence in autocrine or paracrine manners [45, 46]. Our findings may suggest that cellular senescence spreads to other cells in an autocrine or paracrine manner via secreted factors or cell-to-cell interactions in aged HPDL cells, in addition to DNA damage. Therefore, we focused on SASP as the pathophysiological factor in chronic inflammation of aged periodontal tissue. Moreover, SASP has been reported to inhibit proliferation of other normal cells and even promote the progression of cancer [47]. These studies are consistent with our findings (Figure 2). Senescent HPDL cells showed production of representative SASP proteins and enzymes (Figure 3). SASP factors in senescent HPDL cells, including inflammatory cytokines, chemokines, and MMPs/TIMPs, may strongly contribute to the inflammation and destruction of aged periodontal tissue. In this study, we did not determine senescent HPDL-specific SASP factors in terms of types and their combinations. Even so, we believe our findings of regular SASP factors in senescent HPDL cells strongly indicate important roles of cellular senescence in the common pathology underlying periodontitis and age-related chronic diseases in aging populations. Age-dependent inflammation, known as “inflammaging”, is an emerging concept to explain age-dependent inflammation pathology, which is characterized by infiltrated immunocompetent cells and proinflammatory cytokine production at organ or systemic levels [48]. A general feature of aged tissue is low-level chronic inflammation without apparent bacterial infection, which is termed as sterile inflammation [48]. It has been reported that gingival fibroblasts from aged mice show lower IL-6 production after P.g. stimulation compared with those from young mice [49]. Consistent with this, our results indicated that chronic inflammation in senescent HPDL cells was acquired in the steady state, and IL-6 production in senescent HPDL cells was enhanced by proinflammatory cytokine stimulation, but not a bacterial pathogen (Figure 4). As one of the causes of the SASP phenotype with hypo-responsiveness to bacterial stimulation in senescent HPDL cells, DNA damage is thought to induce a proinflammatory cytokine signaling cascade via NF-κB and alteration of the IL-1R/Toll-like receptor signaling pathway [8]. To shed light on the molecular mechanisms regulating cellular senescence in HPDL cells, we focused on miRNAs. It has become clear that miRNAs regulate the onset and progression of diseases as well as their development by regulating cell proliferation, differentiation, apoptosis, metabolism and cellular senescence [19, 20]. Lin-4, which was the first identified miRNA in C. elegans, has been revealed to affect lifespan [50, 51]. To identify specific miRNAs in senescent HPDL, comprehensive miRNA analysis was performed in our study. MiR-146a, which is highly expressed in diseased sites of rheumatoid arthritis and other inflammatory diseases [52, 53], has been reported as a crucial factor for inflammation. IL-6 expression was altered by treatment with synthetic mimic oligos of miR-146a without stimulation by bacterial pathogens in HPDL cells (Supplementary Figure 6); however, endogenous expression of miR-146a was increased with the passage number and its peak expression was later than that of IL-6 and effect of inhibitor of miR-146a was not concordant (Figure 3 and Supplementary Figure 6). Therefore, miR-146a might play roles in terminating the inflammatory cytokine response to maintain the chronic inflammation of senescent HPDL cells. Tumor suppressor p53 induces miR-34a and miR-34a that affect cyclin-dependent kinases CDK$\frac{4}{6}$, anti-apoptotic BCL2, and longevity gene SIRT1 [54, 55]. Additionally, miR-34a has been reported to regulate reprogramming of somatic cells and expression of longevity gene SIRT1 [56]. In contrast to miR-146a, expression of miR-34a and IL-6 was increased with the number of cell passages in a coordinated manner and their peak expression was synchronized, while SIRT1 expression was decreased with the progression of HPDL cell passaging (Figures 5 and 6). In fact, miR-34a mimic treatment suppressed SIRT1 expression and upregulated NF-κB activity in HPDL cells (Figure 5C, 5E). The results of si-SIRT1 treatment strongly indicate a SIRT1-dependent IL-6 production mechanism in HPDL cells (Figure 6). Although RSV treatment slightly increased SIRT1 and strongly suppressed IL-6 in APDLs, we believe that RSV may suppress IL-6 production through the activation of other SIRT family proteins, mitochondria, and mTOR-dependent pathways, in addition to SIRT1. A recent study reported improvements in the reprogramming efficiency of miR-34a knockout mouse-derived somatic cells with Oct$\frac{3}{4}$, Klf4, Sox2, and c-Myc. Therefore, miR-34a is thought to play roles in maintenance of stemness, especially in cancer cells [57]. Accordingly, we believe that miR-34a might be important for the development and wound healing of periodontal tissue through effects on PDL stem cells. Class 1 and 2 HDAC-specific inhibitor TSA at 400 nM induced IL-6 expression in normal HPDL cells. However, TSA is generally effective at nanomolar levels in mammalian cells, and TSA at 200 nM showed no effects (Figure 5F). Therefore, we believe that HDACs other than the Sirtuin protein family, may not be involved in regulation of IL-6 by promoting the acetylation of lysine at the histone in senescent HPDL cells. Taken together, these findings suggest that accumulation of environmental stress triggers activation of the p53-miR-34a axis that suppresses SIRT1 and promotes NF-κB-dependent IL-6 production in senescent HPDL cells. This study had several limitations. It was designed and mainly carried out using primary HPDL cells in vitro. In our study, it was unclear whether and how the SASP of aged HPDL cells enhanced or resolved inflammation with senescent cell clearance by inducing immunocompetent cells in vivo. It is conceivable that experimental periodontitis can be induced in an aged animal, which should be examined to confirm the pathophysiological significance of cellular senescence in periodontal ligament in vivo. Additionally, to identify causes of senescent cells for the pathophysiology of age-related disease, their elimination has been approached [58–60]. This is an important issue for our study that should be examined in the future. We have been investigated the effects of NMN application in a mouse model of periodontitis and found it to effectively reduce oxidative stress in PDL tissue. To establish the clinical relevance of our study, we intend to perform epidemiological studies combined with analysis of clinical samples, such as extracted teeth with severe periodontitis. At present, we are identifying secreted proteins of gingival crevicular fluids to identify SASP proteins specific to senescent HPDL by proteomic analysis. In conclusion, cellular senescence may evoke inflammation and destruction of aged periodontal tissues through SASP in senescent PDL cells. Thus, elimination of senescent PDL cells or suppression of the miR-34a-dependent SIRT1-NF-κB axis represents an attractive therapeutic strategy to prevent periodontitis in elderly people. Moreover, our findings may lead to a new area in periodontal medicine-based etiology for age-related systemic diseases. Furthermore, we expect targeting of aging PDL cells to be a bidirectional treatment for complications such as diabetes, which is closely related to periodontal disease. ## Reagents IL-1β (R&D, MN, USA), P.g. LPS (WAKO, Tokyo, Japan), E. coli. LPS (WAKO) and Rapamycin (WAKO) were applied to cells at the indicated concentrations and periods. ## HPDL cell culture and induction of replicative Primary human PDL (HPDL) cells (ScienCell Research Laboratories Co., CA, USA) were used in this study. HPDL cells were maintained in α-MEM (WAKO) supplemented with $10\%$ fetal calf serum and antibiotics. To induce replicative senescence in HPDL cells in vitro, we used a modified NIH-3T3 cell protocol [61]. Briefly, 1 × 106 HPDL cells were plated on a 10-cm culture dish, cultured for 3 days, then harvested, and counted. Then, 1 × 106 HPDL cells were replated and this passaging cycle was repeated every 3 days. To establish senescent HPDL cells, cell growth of HPDL cells were monitored at each passage. Then, population doublings (PDs) of HPDL cells were calculated by the following formulas [62]: n is passage number, Npn-1 is number of plated cells in passage number n-1, and *Ncn is* the number of collected cells in passage number n. PDs of HPDL cells were used to draw a growth curve based on the number of proliferated cells at each passage of HPDL cells. PDs of HPDL cells were decreased gradually and the cell cycle was almost arrested at around passaged number (P) 30. In our experimental model, >P30 HPDL cells had decreased cell growth and showed near cell cycle arrest. P30 HPDL cells satisfied the various definitions of cellular senescence, such an enlarged cell shape, high SA β-gal activity, and SAHF formation. We judged that P30 HPDL cells had acquired senescent cell-like characteristics, namely premature senescence. Thus, we used >P30 HPDL cells as senescent or prematurely senescent HPDL cells (APDLs). ## Flow cytometry Analysis of the cell size and granularity of HPDL cells was performed by flow cytometry (FCM) using a FACSCalibur (BD Bioscience, CA, USA). FSC/SSC values were evaluated by CellQuest™ Pro software (BD Bioscience). ## ROS analysis To examine the level of intracellular reactive oxygen species (ROS), HPDL cells plated on glass bottom dishes (Matsunami, Tokyo, Japan) were incubated with 2.5 μM CM-H2DCFDA (Life Technologies, CA, USA) for ICC staining. FCM analysis was used to evaluate the peak intensity of FL1-H with CM-H2DCFDA staining by CellQuest™ Pro software. ## Micro-computed tomography Young (6-week-old) and aged (68–104-week-old) male C57BL/6 mice were obtained from Japan SLC Inc. (Shizuoka, Japan). The mice were maintained in the Animal Experiment Laboratory of Osaka University Graduate School of Dentistry until the indicated age in experiments. All animal experiments were approved by the Institutional Animal Care and Use Committee of Osaka University Graduate School of dentistry (permit number: 24-012-0) prior to the commencement of experiments. To examine alveolar bone loss in periodontal tissue of mice, microcomputed tomography (μCT) was conducted for quantification. Briefly, alveolar bones including maxillary molars were dissected and observed using an R_mCT2 3D micro X-ray CT system designed for use with laboratory animals (Rigaku, Tokyo, Japan) to evaluate alveolar bone loss. Alveolar bone resorption was measured in CT images using 3D image analysis software TRI/3D-BON (RATOC System Engineering, Tokyo, Japan). Alveolar bone loss was calculated by measuring the distance from the cement-enamel junction (CEJ) of the mesial root (a), the distal root at first molar (b), and the mesial root at the second molar to the alveolar bone crest (c) (Figure 1A–1C). ## SA β-gal staining Maxillae from mice were fixed in $4\%$ paraformaldehyde (PFA) in phosphate buffered saline (PBS) (Wako) overnight at 4°C and decalcified in 0.5 M EDTA (Wako) for 1 week. After decalcification, periodontal tissues were dehydrated using $15\%$, $20\%$, and $25\%$ sucrose in PBS. Then, periodontal tissues were embedded and in O.C.T. Compound (Sakura Finetek, Tokyo, Japan), frozen, and sectioned at 5 μm thicknesses in a mesio–distal orientation using a CM3050 S (Leica Microsystems, Wetzlar, Germany). Senescence-associated β-galactosidase (SA β-gal) activity in lysosomes at pH 5.5–6.0 was examined by a Senescence Detection Kit (Bio Vision, CA, USA) including SA β-gal staining solution, 5-bromo-4-chloroindol-3-yl β-D-galactopyranoside, 5 mM potassium ferrocyanide, 150 mM NaCl, and 2 mM MgCl2. For cultured HPDL cells, a slightly modified protocol was applied for SA β-gal staining. Briefly, HPDL cells were washed twice with PBS, fixed in $3\%$ PFA for 3 minutes, washed with PBS, and then incubated overnight in freshly prepared SA β-gal staining solution including 1 mg/ml X-Gal, 5 mM potassium ferrocyanide, 150 mM NaCl, and 2 mM MgCl2. ## Immunohistochemical staining After fixation and decalcification, maxillae were embedded in paraffin blocks (Sakura Finetek, Tokyo, Japan). Tissue samples were sliced from paraffin blocks (4-μm sections) using a REM 710 (Yamato, Saitama, Japan), deparaffinated three times in xylene for 5 min, and hydrated in a methanol gradient ($100\%$, $95\%$, $70\%$, and $50\%$). Blocking of unspecific peroxidase activity was performed for 30 min with $3\%$ H2O2 and $90\%$ methanol. Target Retrieval Solution [high-pH Citrate buffer (Agilent, CA, USA)] was used for antigen retrieval. The following antibodies were used: p16 (rat anti-CDKN2A/p16INK4a; dilution 1:200; Abcam, Catalog No. ab241543), SIRT1 (dilution 1:200; Abcam, Catalog No. ab189494), Lamin A + Lamin C (dilution 1:200, Abcam, Catalog No. ab133256). Incubation with primary antibody was performed overnight at 4°C. Subsequently, slides were washed with PBS for 10 min. A biotinylated secondary antibody was incubated initially for 30 min, followed by an avidin biotin complex kit (Vector Laboratories, catalog no. BA-4000) for an additional 30 min. VECTSTAIN Elite ABC Reagent (Vector Laboratories, Catalog No. PK-6100) was used for detection. Slides were counterstained with hematoxylin. Antibody-stained cells were counted in the PDL area. Quantification was performed using Image J (National Institutes of Health, Bethesda, MD, USA). ## Confocal fluorescence microscopy HPDL cells were plated on fibronectin-coated glass coverslips and cultured for 24 hours. Cell layers were fixed in $4\%$ PFA for 10 min, permeabilized with Triton X-100 for 10 min, and blocked with $1.5\%$ BSA for 1 hour. Actin fibers were stained with Anti stain 555 phalloidin or Anti stain 555 phalloidin (Cytoskeleton, CO, USA). Anti-SIRT1 (Cell Signaling Technology; CST, MA, USA) and γH2AX (CST) antibodies were used as primary antibodies and Alexa Fluor 594 goat anti-rabbit IgG (CST) was used as the secondary antibody for ICC staining. Nuclei were stained with VECTASHIELD Mounting Medium with 4'6-diamidino-2-phenylindole (DAPI) (Vector Lab., CA, USA). Immunofluorescence and quantitative image analysis were performed under a Leica SP8 microscope (Leica) using 63× or 100× oil immersion lenses with a numerical aperture (NA) of 1.4. After acquisition, images were processed with Airyscan (Zen software; Carl Zeiss, Oberkochen, Germany). ## Chromatin staining Nuclei were stained with VECTASHIELD Mounting Medium with DAPI. Formation of senescent-associated heterochromatin foci (SAHF) [63] in HPDL cells was observed by the Zeiss LSM 510 confocal microscope system (Carl Zeiss) with a 100× oil immersion lens with NA 1.4. ## Transmission electron microscopy (TEM) TEM analysis of HPDL cells was performed on the basis of a previous study [64]. Briefly, HPDL cells cultured on plates were fixed with $2\%$ glutaraldehyde in PBS at 4°C overnight, and then with $2\%$ tetraphosphate osmium at 4°C for 1 hour. Then, HPDL cells were dehydrated with $50\%$, $70\%$, $90\%$ and $100\%$ ethanol solutions, embedded in Quetol-812 (Nisshin EM, Tokyo, Japan), and polymerized at 60°C for 48 hours. Ultrathin sections cut enface at 70 nm thicknesses were collected on diamond knifes and placed on copper grids. They were stained with $2\%$ uranyl acetate at R/T for 15 min, and rinsed with distilled water, and then stained with Lead stain solution (Sigma-Aldrich Co., MO, USA) at R/T for 3 min. The grids were observed under a transmission electron microscope (JEM-1400 plus; JEOL Ltd., Tokyo, Japan) at an acceleration voltage of 80 kV. Digital images were captured with a CCD camera (Olympus Soft Imaging Solutions GmbH, Münster, Germany). Slice preparation and imaging analysis were performed in accordance with the protocols of Tokai Electron Microscopy, Inc. (Aichi, Japan). ## Luciferase assay HPDL cells were transfected with NF-κB reporter constructs (Promega, WI, USA) and microRNA mimic/inhibitor oligonucleotides using Lipofectamine 2000 (Life Technologies). For IL-1β treatment, HPDL cells were treated with IL-1β at 18 h after transfection. PRL-TK was cotransfected for normalization. Cell extracts were prepared at 48 h after transfection and the ratio of Renilla to firefly luciferase activity was measured using a Dual-Luciferase Reporter Assay System (Promega). ## microRNA array microRNA array analyses were conducted using the Agilent human miRNA microRNA array (8 × 60 K) miRBase ver 19.0 (Agilent Technologies, CA, USA). cDNA labeling, hybridization, and scanning were performed using the miRNA Microarray System with miRNA Complete Labeling and Hyb Kit and Agilent DNA Microarray scanner (CERI; an Agilent-certified service provider, Tokyo, Japan) in accordance with the manufacturer’s instructions. Hierarchical and K-means cluster analyses using GeneSpring GX 12.0 software (Agilent) were performed to evaluate the miRNA expression profiles of HPDL cells in each passage. ## qRT-PCR RT-qPCR was performed in accordance with previously described protocols [65]. Total RNA was isolated from cultured cells using a mirVana miRNA isolation kit (Thermo Fisher Scientific, MA, USA) and then converted to cDNA using a High Capacity RNA-to-cDNA Kit (Life Technologies). Semi-quantitative qRT-PCR was performed using the ABI 7300 Fast Real-Time PCR System with Power SYBR Green PCR Master Mix (Life Technologies) and gene-specific primers (TakaraBio, Shiga, Japan) in accordance with the manufacturers’ instructions. Relative expression was determined after normalization to HPRT expression. The expression level of mature microRNAs was determined using miScript II RT and miScript SYBR Green PCR Kits (Qiagen, Hilden, Germany) in accordance with the manufacturer’s protocols. U6 snRNA was used to evaluate mature miRNAs. The PCR primer sequences are listed in the supporting information (Supplementary Table 1). Total RNA was isolated from PDL tissue dissected from extracted molar teeth of mice under stereo microscopy. ## miRNA mimics and inhibitors miRNA mimics and inhibitors for miR-146a and -34a were purchased from Thermo Fisher Scientific and transduced into HPDL cells on the basis of the manufacturer’s protocols. A miRNA mimic, which is single-stranded locked oligonucleotide, acts as an endogenous miRNA that suppresses target mRNA expression and protein translation. A miRNA inhibitor, a double-stranded locked oligonucleotide, competes with endogenous miRNAs. We used mirVana™ miRNA Mimic Negative Control #1 as a control LNA in accordance with the manufacturer’s protocols, which suggest target gene expression from negative control-transfected samples as a baseline for evaluation of the effect of control and experimental miRNA mimics on target gene expression. Synthetic oligonucleotides were transduced into HPDL cells using Lipofectamine 2000. Then, total RNA, proteins, and culture supernatants of HPDL cells were harvested at 48 h after transfection. ## Western blot analysis Western blot analysis was performed on the basis of a previous report [65]. Briefly, HPDL cells were lysed in RIPA buffer (Millipore, MA, USA) containing a protease inhibitor cocktail (Roche, IN, USA), 1 mM sodium orthovanadate (Sigma-Aldrich), 1 mM Sodium fluoride, and 10 mM β-glycerophosphoric acid (Wako). Protein concentrations of the lysates were quantified by the Bradford Assay (Bio-Rad, CA, USA). Lysates were denatured in 5× Laemmli buffer containing 2-mercaptoethanol by boiling for 10 min at 95°C. After cooling, the proteins were separated by SDS-PAGE under reducing conditions and transferred to a PVDF membrane (GE Healthcare, IN, USA). Membranes were blocked with $5\%$ dry skim milk for 1 h and then probed with the following primary antibodies: mouse anti-human p53, rabbit anti-human p16, goat anti-human CTGF (Santa Cruz, TX, USA), rabbit anti-human p21, mouse anti-human Rb, rabbit anti-human SIRT1 (CST), and mouse anti-human β-actin (Sigma-Aldrich). After washing, membranes were incubated with secondary antibodies, horseradish peroxidase (HRP)-conjugated rabbit anti-mouse IgG or donkey anti-goat IgG (CST), and visualized with ECL prime Western Blotting Detection Reagents (GE Healthcare). ## Dot blot antibody array Screening of SASP factors produced by HPDL cells was carried out using a Human cytokine proteome array (R&D Systems, MN, USA) following the manufacturer’s instructions. HPDL cells were passaged every 3 days. Forty-eight hours after cell seeding, the cell culture medium was replaced with fresh medium and 72 hours later, the culture supernatants were harvested for the array. A list of cytokines and chemokines in the dot plot map is showed in the supporting information (Supplementary Table 2). ## ELISA IL-6 and IL-8 concentrations in HPDL cell culture supernatants were determined using Quantikine Human IL-6 and CXCL/IL-8 kits (R&D Systems) in accordance with the manufacturer’s protocol. ## Statistical analysis The presented data are representative of all results. All experiments were performed at least three times. Quantitative data are presented as the mean and standard deviation of three assays. Differences between two means were assessed using an unpaired students’ two-tailed t-test for two-sample comparisons or one-way analysis of variance for multiple comparisons with the Bonferroni post-hoc test. For statistical comparisons involving more than two groups, a one-way analysis of variance (ANOVA) with Bonferroni post-hoc test or a welch’s t-test (for non-parametric data) was performed to determine differences between groups in Box-and-whiskers plots. P values of less than 0.05 were considered to indicate significance. ## Data availability All data supporting the findings of this study are available within the article and its supplementary materials. ## References 1. Eke PI, Borgnakke WS, Genco RJ. **Recent epidemiologic trends in periodontitis in the USA.**. *Periodontol 2000* (2020) **82** 257-67. DOI: 10.1111/prd.12323 2. Reynolds MA. **Modifiable risk factors in periodontitis: at the intersection of aging and disease.**. *Periodontol 2000* (2014) **64** 7-19. DOI: 10.1111/prd.12047 3. Freund A, Orjalo AV, Desprez PY, Campisi J. **Inflammatory networks during cellular senescence: causes and consequences.**. *Trends Mol Med* (2010) **16** 238-46. DOI: 10.1016/j.molmed.2010.03.003 4. Singh T, Newman AB. **Inflammatory markers in population studies of aging.**. *Ageing Res Rev* (2011) **10** 319-29. DOI: 10.1016/j.arr.2010.11.002 5. Rea IM, Gibson DS, McGilligan V, McNerlan SE, Alexander HD, Ross OA. **Age and Age-Related Diseases: Role of Inflammation Triggers and Cytokines.**. *Front Immunol* (2018) **9** 586. DOI: 10.3389/fimmu.2018.00586 6. An JY, Quarles EK, Mekvanich S, Kang A, Liu A, Santos D, Miller RA, Rabinovitch PS, Cox TC, Kaeberlein M. **Rapamycin treatment attenuates age-associated periodontitis in mice.**. *Geroscience* (2017) **39** 457-63. DOI: 10.1007/s11357-017-9994-6 7. An JY, Kerns KA, Ouellette A, Robinson L, Morris HD, Kaczorowski C, Park SI, Mekvanich T, Kang A, McLean JS, Cox TC, Kaeberlein M. **Rapamycin rejuvenates oral health in aging mice.**. *Elife* (2020) **9** e54318. DOI: 10.7554/eLife.54318 8. Albuquerque-Souza E, Crump KE, Rattanaprukskul K, Li Y, Shelling B, Xia-Juan X, Jiang M, Sahingur SE. **TLR9 Mediates Periodontal Aging by Fostering Senescence and Inflammaging.**. *J Dent Res* (2022) **101** 1628-36. DOI: 10.1177/00220345221110108 9. Naylor RM, Baker DJ, van Deursen JM. **Senescent cells: a novel therapeutic target for aging and age-related diseases.**. *Clin Pharmacol Ther* (2013) **93** 105-16. DOI: 10.1038/clpt.2012.193 10. Coppé JP, Patil CK, Rodier F, Sun Y, Muñoz DP, Goldstein J, Nelson PS, Desprez PY, Campisi J. **Senescence-associated secretory phenotypes reveal cell-nonautonomous functions of oncogenic RAS and the p53 tumor suppressor.**. *PLoS Biol* (2008) **6** 2853-68. DOI: 10.1371/journal.pbio.0060301 11. Tchkonia T, Zhu Y, van Deursen J, Campisi J, Kirkland JL. **Cellular senescence and the senescent secretory phenotype: therapeutic opportunities.**. *J Clin Invest* (2013) **123** 966-72. DOI: 10.1172/JCI64098 12. Yamada S, Murakami S, Matoba R, Ozawa Y, Yokokoji T, Nakahira Y, Ikezawa K, Takayama S, Matsubara K, Okada H. **Expression profile of active genes in human periodontal ligament and isolation of PLAP-1, a novel SLRP family gene.**. *Gene* (2001) **275** 279-86. DOI: 10.1016/s0378-1119(01)00683-7 13. Okada H, Murakami S. **Cytokine expression in periodontal health and disease.**. *Crit Rev Oral Biol Med* (1998) **9** 248-66. DOI: 10.1177/10454411980090030101 14. Seo BM, Miura M, Gronthos S, Bartold PM, Batouli S, Brahim J, Young M, Robey PG, Wang CY, Shi S. **Investigation of multipotent postnatal stem cells from human periodontal ligament.**. *Lancet* (2004) **364** 149-55. DOI: 10.1016/S0140-6736(04)16627-0 15. Beertsen W, McCulloch CA, Sodek J. **The periodontal ligament: a unique, multifunctional connective tissue.**. *Periodontol 2000* (1997) **13** 20-40. DOI: 10.1111/j.1600-0757.1997.tb00094.x 16. Bartel DP. **MicroRNAs: genomics, biogenesis, mechanism, and function.**. *Cell* (2004) **116** 281-97. DOI: 10.1016/s0092-8674(04)00045-5 17. Olivieri F, Rippo MR, Monsurrò V, Salvioli S, Capri M, Procopio AD, Franceschi C. **MicroRNAs linking inflamm-aging, cellular senescence and cancer.**. *Ageing Res Rev* (2013) **12** 1056-68. DOI: 10.1016/j.arr.2013.05.001 18. Smith-Vikos T, Slack FJ. **MicroRNAs and their roles in aging.**. *J Cell Sci* (2012) **125** 7-17. DOI: 10.1242/jcs.099200 19. Gorospe M, Abdelmohsen K. **MicroRegulators come of age in senescence.**. *Trends Genet* (2011) **27** 233-41. DOI: 10.1016/j.tig.2011.03.005 20. Mendell JT, Olson EN. **MicroRNAs in stress signaling and human disease.**. *Cell* (2012) **148** 1172-87. DOI: 10.1016/j.cell.2012.02.005 21. Rodier F, Campisi J. **Four faces of cellular senescence.**. *J Cell Biol* (2011) **192** 547-56. DOI: 10.1083/jcb.201009094 22. Bhaumik D, Scott GK, Schokrpur S, Patil CK, Orjalo AV, Rodier F, Lithgow GJ, Campisi J. **MicroRNAs miR-146a/b negatively modulate the senescence-associated inflammatory mediators IL-6 and IL-8.**. *Aging (Albany NY)* (2009) **1** 402-11. DOI: 10.18632/aging.100042 23. Ma X, Becker Buscaglia LE, Barker JR, Li Y. **MicroRNAs in NF-kappaB signaling.**. *J Mol Cell Biol* (2011) **3** 159-66. DOI: 10.1093/jmcb/mjr007 24. Luan X, Zhou X, Trombetta-eSilva J, Francis M, Gaharwar AK, Atsawasuwan P, Diekwisch TGH. **MicroRNAs and Periodontal Homeostasis.**. *J Dent Res* (2017) **96** 491-500. DOI: 10.1177/0022034516685711 25. Huang L, Salmon B, Yin X, Helms JA. **From restoration to regeneration: periodontal aging and opportunities for therapeutic intervention.**. *Periodontol 2000* (2016) **72** 19-29. DOI: 10.1111/prd.12127 26. Dimri GP, Lee X, Basile G, Acosta M, Scott G, Roskelley C, Medrano EE, Linskens M, Rubelj I, Pereira-Smith O. **A biomarker that identifies senescent human cells in culture and in aging skin in vivo.**. *Proc Natl Acad Sci U S A* (1995) **92** 9363-7. DOI: 10.1073/pnas.92.20.9363 27. Alcorta DA, Xiong Y, Phelps D, Hannon G, Beach D, Barrett JC. **Involvement of the cyclin-dependent kinase inhibitor p16 (INK4a) in replicative senescence of normal human fibroblasts.**. *Proc Natl Acad Sci U S A* (1996) **93** 13742-7. DOI: 10.1073/pnas.93.24.13742 28. Yoon MH, Kang SM, Lee SJ, Woo TG, Oh AY, Park S, Ha NC, Park BJ. **p53 induces senescence through Lamin A/C stabilization-mediated nuclear deformation.**. *Cell Death Dis* (2019) **10** 107. DOI: 10.1038/s41419-019-1378-7 29. Giblin W, Skinner ME, Lombard DB. **Sirtuins: guardians of mammalian healthspan.**. *Trends Genet* (2014) **30** 271-86. DOI: 10.1016/j.tig.2014.04.007 30. Childs BG, Baker DJ, Kirkland JL, Campisi J, van Deursen JM. **Senescence and apoptosis: dueling or complementary cell fates?**. *EMBO Rep* (2014) **15** 1139-53. DOI: 10.15252/embr.201439245 31. Greer EL, Maures TJ, Ucar D, Hauswirth AG, Mancini E, Lim JP, Benayoun BA, Shi Y, Brunet A. **Transgenerational epigenetic inheritance of longevity in Caenorhabditis elegans.**. *Nature* (2011) **479** 365-71. DOI: 10.1038/nature10572 32. Balaban RS, Nemoto S, Finkel T. **Mitochondria, oxidants, and aging.**. *Cell* (2005) **120** 483-95. DOI: 10.1016/j.cell.2005.02.001 33. Narita M, Nũnez S, Heard E, Narita M, Lin AW, Hearn SA, Spector DL, Hannon GJ, Lowe SW. **Rb-mediated heterochromatin formation and silencing of E2F target genes during cellular senescence.**. *Cell* (2003) **113** 703-16. DOI: 10.1016/s0092-8674(03)00401-x 34. Minamino T, Yoshida T, Tateno K, Miyauchi H, Zou Y, Toko H, Komuro I. **Ras induces vascular smooth muscle cell senescence and inflammation in human atherosclerosis.**. *Circulation* (2003) **108** 2264-9. DOI: 10.1161/01.CIR.0000093274.82929.22 35. Kuwahara M, Suzuki J, Tofukuji S, Yamada T, Kanoh M, Matsumoto A, Maruyama S, Kometani K, Kurosaki T, Ohara O, Nakayama T, Yamashita M. **The Menin-Bach2 axis is critical for regulating CD4 T-cell senescence and cytokine homeostasis.**. *Nat Commun* (2014) **5** 3555. DOI: 10.1038/ncomms4555 36. Shin JW, Kwon SH, Choi JY, Na JI, Huh CH, Choi HR, Park KC. **Molecular Mechanisms of Dermal Aging and Antiaging Approaches.**. *Int J Mol Sci* (2019) **20** 2126. DOI: 10.3390/ijms20092126 37. Franceschi C, Garagnani P, Parini P, Giuliani C, Santoro A. **Inflammaging: a new immune-metabolic viewpoint for age-related diseases.**. *Nat Rev Endocrinol* (2018) **14** 576-90. DOI: 10.1038/s41574-018-0059-4 38. Frye RA. **Characterization of five human cDNAs with homology to the yeast SIR2 gene: Sir2-like proteins (sirtuins) metabolize NAD and may have protein ADP-ribosyltransferase activity.**. *Biochem Biophys Res Commun* (1999) **260** 273-9. DOI: 10.1006/bbrc.1999.0897 39. Imai S, Guarente L. **Ten years of NAD-dependent SIR2 family deacetylases: implications for metabolic diseases.**. *Trends Pharmacol Sci* (2010) **31** 212-20. DOI: 10.1016/j.tips.2010.02.003 40. Yamakuchi M, Ferlito M, Lowenstein CJ. **miR-34a repression of SIRT1 regulates apoptosis.**. *Proc Natl Acad Sci U S A* (2008) **105** 13421-6. DOI: 10.1073/pnas.0801613105 41. Nakahata Y, Kaluzova M, Grimaldi B, Sahar S, Hirayama J, Chen D, Guarente LP, Sassone-Corsi P. **The NAD+-dependent deacetylase SIRT1 modulates CLOCK-mediated chromatin remodeling and circadian control.**. *Cell* (2008) **134** 329-40. DOI: 10.1016/j.cell.2008.07.002 42. Yeung F, Hoberg JE, Ramsey CS, Keller MD, Jones DR, Frye RA, Mayo MW. **Modulation of NF-kappaB-dependent transcription and cell survival by the SIRT1 deacetylase.**. *EMBO J* (2004) **23** 2369-80. DOI: 10.1038/sj.emboj.7600244 43. Yoshida M, Matsuyama A, Komatsu Y, Nishino N. **From discovery to the coming generation of histone deacetylase inhibitors.**. *Curr Med Chem* (2003) **10** 2351-8. DOI: 10.2174/0929867033456602 44. Muñoz-Espín D, Cañamero M, Maraver A, Gómez-López G, Contreras J, Murillo-Cuesta S, Rodríguez-Baeza A, Varela-Nieto I, Ruberte J, Collado M, Serrano M. **Programmed cell senescence during mammalian embryonic development.**. *Cell* (2013) **155** 1104-18. DOI: 10.1016/j.cell.2013.10.019 45. Acosta JC, O'Loghlen A, Banito A, Guijarro MV, Augert A, Raguz S, Fumagalli M, Da Costa M, Brown C, Popov N, Takatsu Y, Melamed J, d'Adda di Fagagna F. **Chemokine signaling via the CXCR2 receptor reinforces senescence.**. *Cell* (2008) **133** 1006-18. DOI: 10.1016/j.cell.2008.03.038 46. Kuilman T, Michaloglou C, Vredeveld LC, Douma S, van Doorn R, Desmet CJ, Aarden LA, Mooi WJ, Peeper DS. **Oncogene-induced senescence relayed by an interleukin-dependent inflammatory network.**. *Cell* (2008) **133** 1019-31. DOI: 10.1016/j.cell.2008.03.039 47. Acosta JC, Banito A, Wuestefeld T, Georgilis A, Janich P, Morton JP, Athineos D, Kang TW, Lasitschka F, Andrulis M, Pascual G, Morris KJ, Khan S. **A complex secretory program orchestrated by the inflammasome controls paracrine senescence.**. *Nat Cell Biol* (2013) **15** 978-90. DOI: 10.1038/ncb2784 48. Xia S, Zhang X, Zheng S, Khanabdali R, Kalionis B, Wu J, Wan W, Tai X. **An Update on Inflamm-Aging: Mechanisms, Prevention, and Treatment.**. *J Immunol Res* (2016) **2016** 8426874. DOI: 10.1155/2016/8426874 49. Domon H, Tabeta K, Nakajima T, Yamazaki K. **Age-related alterations in gene expression of gingival fibroblasts stimulated with Porphyromonas gingivalis.**. *J Periodontal Res* (2014) **49** 536-43. DOI: 10.1111/jre.12134 50. Lee RC, Feinbaum RL, Ambros V. **The C. elegans heterochronic gene lin-4 encodes small RNAs with antisense complementarity to lin-14.**. *Cell* (1993) **75** 843-54. DOI: 10.1016/0092-8674(93)90529-y 51. Boehm M, Slack F. **A developmental timing microRNA and its target regulate life span in C. elegans.**. *Science* (2005) **310** 1954-7. DOI: 10.1126/science.1115596 52. Taganov KD, Boldin MP, Chang KJ, Baltimore D. **NF-kappaB-dependent induction of microRNA miR-146, an inhibitor targeted to signaling proteins of innate immune responses.**. *Proc Natl Acad Sci U S A* (2006) **103** 12481-6. DOI: 10.1073/pnas.0605298103 53. Kriegsmann M, Randau TM, Gravius S, Lisenko K, Altmann C, Arens N, Kriegsmann J. **Expression of miR-146a, miR-155, and miR-223 in formalin-fixed paraffin-embedded synovial tissues of patients with rheumatoid arthritis and osteoarthritis.**. *Virchows Arch* (2016) **469** 93-100. DOI: 10.1007/s00428-016-1939-4 54. Hayakawa T, Iwai M, Aoki S, Takimoto K, Maruyama M, Maruyama W, Motoyama N. **SIRT1 suppresses the senescence-associated secretory phenotype through epigenetic gene regulation.**. *PLoS One* (2015) **10** e0116480. DOI: 10.1371/journal.pone.0116480 55. Oberdoerffer P, Michan S, McVay M, Mostoslavsky R, Vann J, Park SK, Hartlerode A, Stegmuller J, Hafner A, Loerch P, Wright SM, Mills KD, Bonni A. **SIRT1 redistribution on chromatin promotes genomic stability but alters gene expression during aging.**. *Cell* (2008) **135** 907-18. DOI: 10.1016/j.cell.2008.10.025 56. Hermeking H. **The miR-34 family in cancer and apoptosis.**. *Cell Death Differ* (2010) **17** 193-9. DOI: 10.1038/cdd.2009.56 57. Jang E, Kim E, Son HY, Lim EK, Lee H, Choi Y, Park K, Han S, Suh JS, Huh YM, Haam S. **Nanovesicle-mediated systemic delivery of microRNA-34a for CD44 overexpressing gastric cancer stem cell therapy.**. *Biomaterials* (2016) **105** 12-24. DOI: 10.1016/j.biomaterials.2016.07.036 58. Baker DJ, Wijshake T, Tchkonia T, LeBrasseur NK, Childs BG, van de Sluis B, Kirkland JL, van Deursen JM. **Clearance of p16Ink4a-positive senescent cells delays ageing-associated disorders.**. *Nature* (2011) **479** 232-6. DOI: 10.1038/nature10600 59. Palmer AK, Xu M, Zhu Y, Pirtskhalava T, Weivoda MM, Hachfeld CM, Prata LG, van Dijk TH, Verkade E, Casaclang-Verzosa G, Johnson KO, Cubro H, Doornebal EJ. **Targeting senescent cells alleviates obesity-induced metabolic dysfunction.**. *Aging Cell* (2019) **18** e12950. DOI: 10.1111/acel.12950 60. Omori S, Wang TW, Johmura Y, Kanai T, Nakano Y, Kido T, Susaki EA, Nakajima T, Shichino S, Ueha S, Ozawa M, Yokote K, Kumamoto S. **Generation of a p16 Reporter Mouse and Its Use to Characterize and Target p16**. *Cell Metab* (2020) **32** 814-28.e6. DOI: 10.1016/j.cmet.2020.09.006 61. Jainchill JL, Aaronson SA, Todaro GJ. **Murine sarcoma and leukemia viruses: assay using clonal lines of contact-inhibited mouse cells.**. *J Virol* (1969) **4** 549-53. DOI: 10.1128/JVI.4.5.549-553.1969 62. Heo JY, Jing K, Song KS, Seo KS, Park JH, Kim JS, Jung YJ, Hur GM, Jo DY, Kweon GR, Yoon WH, Lim K, Hwang BD. **Downregulation of APE1/Ref-1 is involved in the senescence of mesenchymal stem cells.**. *Stem Cells* (2009) **27** 1455-62. DOI: 10.1002/stem.54 63. Aird KM, Zhang R. **Detection of senescence-associated heterochromatin foci (SAHF).**. *Methods Mol Biol* (2013) **965** 185-96. DOI: 10.1007/978-1-62703-239-1_12 64. Polishchuk EV, Polishchuk RS, Luini A. **Correlative light-electron microscopy as a tool to study in vivo dynamics and ultrastructure of intracellular structures.**. *Methods Mol Biol* (2013) **931** 413-22. DOI: 10.1007/978-1-62703-056-4_20 65. Kawahara T, Yamashita M, Ikegami K, Nakamura T, Yanagita M, Yamada S, Kitamura M, Murakami S. **TGF-Beta Negatively Regulates the BMP2-Dependent Early Commitment of Periodontal Ligament Cells into Hard Tissue Forming Cells.**. *PLoS One* (2015) **10** e0125590. DOI: 10.1371/journal.pone.0125590
--- title: Metabolically healthy and unhealthy obesity and the development of lung dysfunction authors: - Jae-Uk Song - Jonghoo Lee - Si-Young Lim - Hyun-Il Gil - Yoosoo Chang - Seungho Ryu journal: Scientific Reports year: 2023 pmcid: PMC10042802 doi: 10.1038/s41598-023-31960-7 license: CC BY 4.0 --- # Metabolically healthy and unhealthy obesity and the development of lung dysfunction ## Abstract We investigated the association of metabolically healthy (MH) and unhealthy (MU) obesity with incident lung dysfunction. This cohort study included 253,698 Korean lung disease-free adults (mean age, 37.4 years) at baseline. Spirometry-defined lung dysfunction was classified as a restrictive pattern (RP) or obstructive pattern (OP). We defined obesity as BMI ≥ 25 kg/m2 and MH as the absence of any metabolic syndrome components with a homeostasis model assessment of insulin resistance < 2.5: otherwise, participants were considered MU. During a median follow-up of 4.9 years, 10,775 RP cases and 7140 OP cases develped. Both MH and MU obesity showed a positive association with incident RP, with a stronger association in the MU than in the MH group (Pinteraction = 0.001). Multivariable-adjusted hazard ratios ($95\%$ CI) for incident RP comparing obesity to the normal-weight category was 1.15 (1.05–1.25) among the MH group and 1.38 (1.30–1.47) among MU group. Conversely, obesity was inversely associated with OP because of a greater decline in forced vital capacity than forced expiratory volume in 1 s. Both MH and MU obesity were positively associated with RP. However, the associations between obesity, metabolic health, and lung functions might vary depending on the type of lung disease. ## Introduction Lung function impairment, in both obstructive and restrictive patterns, is associated with chronic respiratory and non-respiratory disease development, including deaths from all causes and cardiovascular disease1–4, contributing to significant public health problems worldwide5,6. Furthermore, chronic obstructive lung disease commonly manifests after the age of 40; however, there is growing attention that lung dysfunction occurs much earlier than overt disease manifestation7. Early identifying modifiable risk factors for lung function impairment and understanding its pathophysiology is important to establish preventive measures to reduce chronic respiratory disease and other non-respiratory complications. Obesity and metabolic syndrome are associated with respiratory symptoms, lung disease, and spirometry lung function. Most studies have evaluated the effects of either obesity or metabolic syndrome on lung function separately with mixed results1,6,8–14. Obesity is often accompanied by metabolic abnormalities, such as hypertension, type 2 diabetes, insulin resistance, and dyslipidaemia. However, a subset of obese individuals do not always present with metabolic abnormalities despite having excessive body fat; this is referred to as metabolically healthy obesity (MHO), possibly contributing to favourable prognosis without adverse obesity-related outcomes15. Combined or isolated obesity phenotypes and metabolic health status may help elucidate whether obesity per se or the presence of co-existing metabolic abnormalities affects lung function impairment. However, to date, the longitudinal association between different metabolic health and obesity phenotypes and lung function impairment is generally unknown. We investigated the longitudinal relationship between body mass index (BMI), a proxy indicator of obesity, and metabolic health status with different lung function impairment types in a large cohort of apparently healthy young and middle-aged Korean adults, lung-disease free at baseline, who participated in a comprehensive screening examination, including repeated spirometry measures. ## Study population The Kangbuk Samsung Health *Study is* a cohort study involving Korean men and women who underwent comprehensive health examinations at the Total Healthcare Center of Kangbuk Samsung Hospital clinics in Seoul and Suwon, South Korea since January 1, 200216. More than $80\%$ of the participants were employees of various companies, local governmental organisations, or their spouses. In Korea, annual or biennial employee health screenings are required by the Industrial Safety and Health Law and are provided free of charge. The present cohort study included participants with at least one follow-up visit between January 1, 2011, and December 31, 2019 ($$n = 335$$,209). After the exclusion of 81,511 participants, 253,698 participants were ultimately included in the analysis (Fig. 1).Figure 1Flow chart for selecting the study population. Out of the 81,511 participants who were excluded, 71,801 met only one exclusion criterion, 9027 met two exclusion criteria, 663 met three exclusion criteria, and 20 participants met four exclusion criteria. This study was approved by the Institutional Review Board of Kangbuk Samsung Hospital (KBSMC 2022-03-055), which waived the requirement for written informed consent owing to the use of de-identified data obtained as part of routine health screening examinations. All procedures involved in this study of human participants were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. ## Data collection At baseline and follow-up visits, data on demographic characteristics, medical history, medication use, smoking status, physical activity level17, usual dietary intake, and other lifestyle habits were collected via standardised, self-administered questionnaires. Smoking status was categorised as never, former, or current smoker. Average alcohol consumption was calculated based on the frequency and amount consumed per drinking day and then categorised as none or < 20 and ≥ 20 g ethanol/day. Physical activity levels were classified as inactive, minimally active, and health-enhancing physical activity (HEPA)17. Dietary intake was assessed using a validated 106-item food frequency questionnaire, with portion sizes and consumption frequency recorded. Nutrient values were calculated using a Korean food composition table18. Body height and weight were measured by trained nurses, with participants wearing a hospital gown and no shoes. BMI was classified according to Asian-specific criteria19: underweight, < 18.5 kg/m2; normal weight, 18.5–23 kg/m2; overweight, 23–25 kg/m2; and obese, ≥ 25 kg/m2. Hypertension was defined as systolic blood pressure (BP) ≥ 140 mmHg, diastolic BP ≥ 90 mmHg, or current use of antihypertensive medication. Blood samples were obtained after participants had fasted for at least 10 h. Fasting blood measurements included glucose, glycated haemoglobin, lipid, insulin, and hsCRP levels. Insulin resistance was assessed using the following homeostasis model assessment of insulin resistance (HOMA-IR) equation: fasting blood insulin (µU/mL) × fasting blood glucose (mmol/L)/22.5. Diabetes was defined as fasting serum glucose ≥ 126 mg/dL, glycated haemoglobin ≥ $6.5\%$, or current insulin or antidiabetic medication use. Metabolically healthy (MH) persons were defined as having none of the following metabolic abnormalities, as previously applied20: [1] fasting glucose level ≥ 100 mg/dL or current glucose-lowering agent use, [2] BP ≥ $\frac{130}{85}$ mmHg or current BP-lowering agent use, [3] elevated triglyceride level (≥ 150 mg/dL) or current lipid-lowering agent use, [4] low high-density lipoprotein cholesterol (HDL-C) (< 40 mg/dL in men or < 50 mg/dL in women), or [5] insulin resistance, defined as HOMA-IR score ≥ 2.5. In contrast, metabolically unhealthy (MU) was defined as having one or more of these metabolic abnormalities. Spirometry was performed according to American Thoracic Society and European Respiratory Society recommendations21, using the Vmax22 system (Sensor-Medics, Yorba Linda, CA, USA)16. Forced expiratory volume in 1 s (FEV1) and forced vital capacity (FVC) were obtained in a pre-bronchodilatory setting. The predicted FEV1 and FVC values were calculated using equations for a representative Korean population sample22. To calculate the predicted FVC% and predicted FEV$1\%$, we divided the measured value (L) by the predicted value (L) and converted the quotient into a percentage. The FEV1 to FVC ratio (FEV1/FVC) was calculated, and the actual measurements were used. Spirometry-defined lung function was classified as the restrictive pattern (RP, FEV1/FVC ≥ 0.7 and FVC < $80\%$ predicted) and obstructive pattern (OP, FEV1/FVC < 0.7)21. ## Statistical analyses Baseline characteristics of the participants were summarised by BMI category. The primary endpoints were RP and OP development. To assess for linear trends, we used the number of BMI categories as a continuous variable and tested it in each model. The follow-up time was calculated from the baseline examination to lung disease development or to the last health examination, whichever occurred first. Because the exact time of lung disease onset was unknown and occurred between the visit for lung disease diagnosis and the previous visit, we used flexible parametric proportional hazard models to account for this type of interval censoring23. This survival model parameterised log cumulative hazards as natural cubic splines of log time with three internal knots at the 25th, 50th, and 75th percentiles, respectively23,24. We estimated the hazard ratios (HRs) and $95\%$ confidence intervals (CIs) for incident lung disease comparing BMI category to the normal BMI category as the reference, overall and separately for MH and MU individuals. Interactions by metabolic health status were examined using likelihood ratio tests comparing models with and without multiplicative interaction terms. The models were first age- and sex-adjusted and then further adjusted for other potential confounding factors including study centre (Seoul or Suwon), examination year (1-year categories), smoking status (never, former, or current), alcohol consumption (none, < 20 g/day, ≥ 20 g/day, or unknown), physical activity level (inactive, minimally active, HEPA, or unknown), education level (high school graduate or less, college graduate or higher, or unknown), and total calorie intake (in quintiles or missing). We also fitted additional models adjusted for potential intermediate variables, including blood eosinophil count and hsCRP level, to evaluate potential mediators of the association between BMI and incident lung disease. Finally, to evaluate the effects of BMI changes and other covariates over time during follow-up, we conducted additional analyses with BMI and confounders as time-varying covariates in the models. The proportional hazards assumption was assessed by examining graphs of estimated log (–log) survival; ultimately, no violation of this assumption was found. We assessed collinearity among covariates using the variance inflation factor (VIF), and found no evidence of collinearity, with all VIF values less than 10. To conduct linear trend analyses, we included the median BMI values for each BMI category as continuous variables with linear terms in the regression models. To assess a quadratic trend, we centered the linear trend variable at the reference (normal-weight category) and then squared it. To assess the longitudinal associations between the obesity category and subsequent changes in FEV1 and FVC over time, we used linear mixed models using random intercepts and slopes, while adjusting for potential baseline confounders. We estimated the annual change in FEV1 ($95\%$ CIs) from baseline, as well as the mean difference in annual FEV1 change comparing each BMI subgroup category with the reference group (normal-weight group). The same analyses were repeated for FVC. All analyses were performed using Stata version 16 (StataCorp LP; College Station, TX, USA). P values < 0.05 were considered statistically significant. ## Results At baseline, the mean (standard deviation) age of the participants was 37.4 ± 7.7 years, and $57.3\%$ were male (Table 1; Fig. 1). The prevalence of underweight, normal weight, overweight, and obesity were $5.3\%$, $43.6\%$, $22.5\%$, and $28.7\%$, respectively, and the prevalence of MUH individuals with at least one metabolic syndrome component or insulin resistance was $46.4\%$. Increased BMI was positively associated with age; male sex; alcohol consumption; current smoking status; physical activity level; BP; fasting glucose, total cholesterol, triglyceride, LDL-C, HOMA-IR, and hsCRP levels; and eosinophil count, whereas BMI was inversely associated with HDL-C. Regarding spirometry values, increasing BMI categories were positively associated with FEV1 (L) and FVC (L), whereas FEV1/FVC decreased with increasing BMI categories. Table 1Baseline characteristics of study participants by body mass index category. CharacteristicOverallBMI category (kg/m2)p for trend < 18.518.5–22.923.0–24.9 ≥ 25Number253,69813,329110,58757,04872,734Age (years)a37.4 (7.7)33.9 (5.9)36.5 (7.4)38.6 (8.1)38.5 (7.8) < 0.001Sex (%)57.39.938.074.281.8 < 0.001Current smoker (%)21.76.414.226.332.1 < 0.001Alcohol intake (%)b23.87.415.828.335.0 < 0.001HEPA (%)c15.79.114.517.517.5 < 0.001Higher education (%)d84.384.084.385.183.80.159Energy intake (kcal/d)e,f1521 (1154–1925)1358 (999–1733)1448 (1085–1836)1569 (1210–1969)1632 (1262–2065) < 0.001Metabolic parameters Hypertension (%)9.61.24.110.219.0 < 0.001 Diabetes (%)3.10.41.33.36.2 < 0.001 Systolic BP (mmHg)a109.2 (12.9)98.9 (9.6)104.3 (11.4)111.5 (11.5)116.8 (12.1) < 0.001 Diastolic BP (mmHg)a69.9 (9.8)64.1 (7.6)66.8 (8.7)71.3 (9.3)74.7 (9.9) < 0.001 Glucose (mg/dl)a94.6 (13.8)88.7 (7.9)91.7 (10.8)95.7 (13.6)99.1 (17.1) < 0.001 Total cholesterol (mg/dl)a193.2 (34.0)178.2 (28.5)186.3 (31.5)196.9 (33.7)203.5 (35.4) < 0.001 LDL-C (mg/dl)a120.1 (31.9)98.8 (24.3)111.3 (29.0)125.8 (30.9)132.9 (32.0) < 0.001 HDL-C (mg/dl)a58.8 (15.3)71.5 (14.6)64.3 (14.9)55.8 (13.4)50.4 (12.1) < 0.001 Triglycerides (mg/dl)e90 (63–134)62 (50–78)73 (56–100)100 (72–143)130 (92–185) < 0.001 HOMA-IRe1.21 (0.80–1.80)0.83 (0.56–1.20)0.99 (0.67–1.40)1.25 (0.87–1.77)1.79 (1.23–2.60) < 0.001Inflammatory parameters hsCRP (mg/L)e,g0.40 (0.20–0.90)0.20 (0.20–0.40)0.30 (0.20–0.60)0.50 (0.30–0.90)0.80 (0.40–1.50) < 0.001 Eosinophil counta2.1 (1.2–3.4)1.8 (1.0–3.0)1.9 (1.1–3.2)2.2 (1.3–3.6)2.3 (1.4–3.6) < 0.001Pulmonary function test FEV1 (L)a3.41 (0.68)2.91 (0.46)3.23 (0.65)3.58 (0.66)3.64 (0.63) < 0.001 FEV1 (% pred)a99.4 (10.0)98.1 (9.5)100.0 (10.1)99.5 (9.9)98.6 (9.9) < 0.001 FVC (L)a4.05 (0.84)3.28 (0.52)3.78 (0.79)4.32 (0.80)4.40 (0.77) < 0.001 FVC (% pred)a98.1 (9.8)94.9 (8.9)98.3 (9.8)98.7 (9.7)97.9 (9.9) < 0.001 FEV1/FVCa84.5 (5.9)88.9 (6.1)85.7 (6.2)83.1 (5.4)82.7 (4.8) < 0.001Data are presented as ameans (standard deviations), emedians (interquartile ranges), or percentages.b ≥ 20 g of ethanol per day; c HEPA was defined as meeting either of the following criteria: [1] vigorous-intensity activity on three or more days per week accumulating ≥ 1,500 metabolic equivalent of task (MET) min/week or [2] seven days of any combination of walking, moderate-intensity activities, or vigorous-intensity activities, achieving at least 3000 MET min/week; d ≥ college graduate; famong 180,985 participants with plausible estimated energy intake levels (within three standard deviations from the log-transformed mean energy intake); gamong 123,194 participants without missing hsCRP values. BMI body mass index, BP blood pressure, FEV1 forced expiratory volume in 1 s, FVC forced vital capacity, HDL-C high-density lipoprotein cholesterol, HEPA health-enhancing physical activity, HOMA-IR homeostasis model assessment of insulin resistance, hsCRP high-sensitivity C-reactive protein, HOMA-IR homeostasis model assessment of insulin resistance, LDL-C low-density lipoprotein cholesterol, pred predicted. Table 2 presents the association of the BMI category with the overall incidence of RP in MH and MU participants. The median follow-up period was 4.9 years (interquartile range, 2.8–6.9; maximum, 8.8 years). The median frequency of follow-up visits, excluding the baseline visit, was 4 visits (interquartile range: 2–6 visits). During 1,221,414 person-years of follow-up, 10,775 new-onset cases of RP were identified, with an incidence rate of 8.8 cases per 103 person-years. The associations between BMI category and incident RP were reverse J-shaped (P for quadratic trend < 0.001) and obesity was associated with an increased risk of RP in MH and MU groups but this association was more evident in MU than in MH individuals (P for interaction = 0.001). After adjustment for age, sex, other confounders (model 1), the multivariable-adjusted HRs ($95\%$ CIs) for incident RP comparing underweight, overweight, and obesity with normal weight were 1.90 (1.74–2.08), 0.91 (0.83–0.98), and 1.15 (1.05–1.25), respectively, in the MH group, and 1.97 (1.65–2.35), 1.07 (1.00–1.15), and 1.38 (1.30–1.47), respectively, in the MU group. When changes in BMI and other confounders during follow-up were updated as time-varying covariates, similar associations were observed. Further adjustment for hsCRP level and blood eosinophil count did not qualitatively change the association between BMI category and RP.Table 2Development of restrictive lung diseases by body mass index category in metabolically healthy and unhealthy phenotypes. BMI category(kg/m2)Person-yearsIncident casesIncidence rate(cases per 103 PY)Age- and sex-adjusted HR ($95\%$ CI)Multivariable-adjusted HRa($95\%$ CI)HR ($95\%$ CI)b(in the model using time-dependent variables)Model 1Model 2Total ($$n = 270$$,190) < 18.562,646.073211.71.89 (1.75–2.05)1.90 (1.75–2.05)1.90 (1.75–2.06)2.17 (2.00–2.35) 18.5–22.9531,375.339697.51.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) 23.0–24.9277,652.623078.31.03 (0.97–1.08)1.03 (0.97–1.08)1.02 (0.97–1.08)0.94 (0.89–0.99) ≥ 25.0349,740.1376710.81.39 (1.32–1.46)1.38 (1.32–1.45)1.37 (1.31–1.44)1.34 (1.28–1.41) P for linear trend < 0.001 < 0.001 < 0.001 < 0.001 P for quadratic trend < 0.001 < 0.001 < 0.001 < 0.001Metabolically healthy ($$n = 128$$,548) < 18.554,008.059711.11.89 (1.73–2.07)1.90 (1.74–2.08)1.90 (1.74–2.08)2.19 (2.00–2.40) 18.5–22.9370,170.124266.61.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) 23.0–24.9128,735.97716.00.91 (0.84–0.99)0.91 (0.83–0.98)0.90 (0.83–0.98)0.83 (0.76–0.90) ≥ 25.092,426.46747.31.16 (1.06–1.27)1.15 (1.05–1.25)1.14 (1.04–1.24)1.11 (1.02–1.20) P for linear trend < 0.001 < 0.001 < 0.001 < 0.001 P for quadratic trend < 0.001 < 0.001 < 0.001 < 0.001Metabolically unhealthy ($$n = 141$$,642) < 18.58638.013515.61.96 (1.65–2.34)1.97 (1.65–2.35)1.98 (1.66–2.36)2.12 (1.77–2.54) 18.5–22.9161,205.215439.61.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) 23.0–24.9148,916.8153610.31.07 (1.00–1.15)1.07 (1.00–1.15)1.07 (1.00–1.15)1.00 (0.93–1.07) ≥ 25.0257,313.7309312.01.40 (1.32–1.49)1.38 (1.30–1.47)1.37 (1.29–1.46)1.37 (1.28–1.46) P for liner trend < 0.001 < 0.001 < 0.001 < 0.001 P for quadratic trend < 0.001 < 0.001 < 0.001 < 0.001P = 0.001 for the overall interaction between metabolic health status and BMI category for incident restrictive lung diseases (adjusted model 1).aEstimated from parametric proportional hazard models. Multivariable model 1 was adjusted for age, sex, centre, year of screening examination, education level, smoking status, alcohol intake, physical activity level, and total energy intake; model 2: model 1 plus adjustment for eosinophil count and high-sensitivity C-reactive protein level.bEstimated from parametric proportional hazard models with BMI category, smoking status, alcohol intake, physical activity level, and total energy intake as time-dependent categorical variables and baseline age, sex, centre, year of screening examination, and education level as time-fixed variables. BMI body mass index, CI confidence interval, HR hazard ratio, PY person years. Table 3 presents the association of BMI with the overall incidence of OP in MH and MU participants. During 1,226,550 person-years of follow-up, 7140 new-onset cases of OP lung disease were identified, with an incidence rate of 5.8 cases per 103 person-years. The BMI category was inversely associated with incident OP lung disease, and these associations were more pronounced in MU than in MH individuals (P for interaction = 0.001). The multivariable-adjusted HRs ($95\%$ CIs) for OP risk comparing underweight, overweight, and obesity with normal weight were 0.95 (0.80–1.12), 0.88 (0.81–0.96), and 0.72 (0.65–0.80), respectively, among MH individuals and 1.28 (0.96–1.71), 0.87 (0.80–0.94), and 0.59 (0.55–0.64), respectively, among MU individuals. These associations between BMI category and OP were similarly observed in time-dependent analyses and in analyses with further adjustments for hsCRP level and blood eosinophil count. Sensitivity analyses that excluded BMI outliers did not alter the results (Table S1). Additionally, the results were not qualitatively affected in sensitivity analyses that included average alcohol consumption and dietary intake as continuous variables (Table S2). In analyses using linear mixed models, we also examined the association between obesity category and serial change in absolute FVC and FEV1 values (Table S3). The annual decline in both FVC and FEV1 values was greater in the overweight and obesity categories compared with the normal weight category. Compared to FEV1, the mean difference in annual change was higher for FVC in the overweight and obesity categories compared to the normal-weight category. Table 3Development of obstructive lung diseases by body mass index category in metabolically healthy and unhealthy phenotypes. BMI category(kg/m2)Person-yearsIncident casesIncidence rate(cases per 103 PY)Age and sex-adjusted HR ($95\%$ CI)Multivariable-adjusted HRa($95\%$ CI)HR ($95\%$ CI)b(in the model using time-dependent variables)Model 1Model 2Total ($$n = 270$$,190) < 18.563,901.92003.11.02 (0.89–1.18)1.01 (0.88–1.17)1.01 (0.87–1.17)0.99 (0.85–1.16) 18.5–22.9533,215.929115.51.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) 23.0–24.9277,246.620567.40.88 (0.83–0.94)0.87 (0.82–0.92)0.87 (0.82–0.93)0.89 (0.84–0.95) ≥ 25.0352,185.619735.60.64 (0.60–0.68)0.62 (0.58–0.66)0.62 (0.58–0.65)0.61 (0.58–0.65) P for linear trend < 0.001 < 0.001 < 0.001 < 0.001 P for quadratic trend < 0.001 < 0.001 < 0.001 < 0.001Metabolically healthy ($$n = 128$$,548) < 18.555,017.41522.80.96 (0.81–1.14)0.95 (0.80–1.12)0.94 (0.80–1.12)0.92 (0.77–1.11) 18.5–22.9371,738.516874.51.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) 23.0–24.9128,470.87716.00.89 (0.81–0.97)0.88 (0.81–0.96)0.88 (0.81–0.96)0.91 (0.84–1.00) ≥ 25.092,550.44795.20.73 (0.66–0.81)0.72 (0.65–0.80)0.72 (0.65–0.80)0.69 (0.62–0.76) P for linear trend < 0.001 < 0.001 < 0.001 < 0.001 P for quadratic trend < 0.001 < 0.001 < 0.001 < 0.001Metabolically unhealthy ($$n = 141$$,642) < 18.58884.5485.41.29 (0.97–1.73)1.28 (0.96–1.71)1.28 (0.96–1.71)1.24 (0.91–1.69) 18.5–22.9161,477.412247.61.00 (reference)1.00 (reference)1.00 (reference)1.00 (reference) 23.0–24.9148,775.812858.60.88 (0.82–0.95)0.87 (0.80–0.94)0.87 (0.80–0.94)0.89 (0.82–0.96) ≥ 25.0259,635.214945.80.62 (0.57–0.67)0.59 (0.55–0.64)0.59 (0.55–0.64)0.59 (0.55–0.64) P for linear trend < 0.001 < 0.001 < 0.001 < 0.001 P for quadratic trend < 0.001 < 0.001 < 0.001 < 0.001P = 0.001 for the overall interaction between metabolic health status and BMI category for incident obstructive lung diseases (adjusted model 1).aEstimated from parametric proportional hazard models. Multivariable model 1 was adjusted for age, sex, centre, year of screening examination, education level, smoking status, alcohol intake, physical activity level, and total energy intake; model 2: model 1 plus adjustment for eosinophil count and high-sensitivity C-reactive protein level.bEstimated from parametric proportional hazard models with BMI category, smoking status, alcohol intake, physical activity level, and total energy intake as time-dependent categorical variables and baseline age, sex, centre, year of screening examination, and education level as time-fixed variables. BMI body mass index, CI confidence interval, HR hazard ratio, PY person years. Finally, we evaluated the effect of metabolic health status on lung function impairment within the same BMI category (Table 4). The MH group was considered the reference for each BMI stratum. The incident RP risk significantly increased in MU participants across all BMI categories compared to that in the MH group. Conversely, the incident OP risk did not significantly differ between the MH and MU groups in normal-weight and overweight individuals, although the incident OP risk was significantly lower in the MU compared with the MH group in the obese stratum (adjusted HR = 0.81, $95\%$ CI 0.73–0.89).Table 4Development of restrictive lung diseases by metabolically healthy status stratified by body mass index category. BMI category(kg/m2)Person-yearsIncident casesIncidence rate (per 103 PY)Age and sex-adjusted HR ($95\%$ CI)Multivariable-adjusted HR ($95\%$ CI)aHR ($95\%$ CI)b(in time-dependent model)Restrictive lung disease Normal weight Metabolically healthy370,170.124266.61.00 (reference)1.00 (reference)1.00 (reference) Metabolically unhealthy161,205.215439.61.11 (1.04–1.19)1.14 (1.07–1.22)1.16 (1.08–1.24) Overweight Metabolically healthy128,735.97716.01.00 (reference)1.00 (reference)1.00 (reference) Metabolically unhealthy148,916.8153610.31.31 (1.20–1.43)1.35 (1.24–1.47)1.39 (1.26–1.52) Obesity Metabolically healthy92,426.46747.31.00 (reference)1.00 (reference)1.00 (reference) Metabolically unhealthy257,313.7309312.01.34 (1.24–1.46)1.38 (1.27–1.50)1.49 (1.36–1.62)Obstructive lung disease Normal weight Metabolically healthy371,738.516874.51.00 (reference)1.00 (reference)1.00 (reference) Metabolically unhealthy161,477.412247.61.00 (0.93–1.08)0.98 (0.91–1.06)1.00 (0.92–1.08) Overweight Metabolically healthy128,470.87716.01.00 (reference)1.00 (reference)1.00 (reference) Metabolically unhealthy148,775.812858.61.00 (0.91–1.09)0.96 (0.88–1.05)0.92 (0.84–1.00) Obesity Metabolically healthy92,550.44795.21.00 (reference)1.00 (reference)1.00 (reference) Metabolically unhealthy259,635.214945.80.84 (0.76–0.94)0.81 (0.73–0.89)0.81 (0.73–0.90)aEstimated from parametric proportional hazard models. The multivariable model was adjusted for age, sex, centre, year of screening examination, education level, smoking status, alcohol intake, physical activity level, and total energy intake.bEstimated from parametric proportional hazard models with BMI category, smoking status, alcohol intake, physical activity level, and total energy intake as time-dependent categorical variables and baseline age, sex, centre, year of screening examination, and education level as time-fixed variables. BMI body mass index, CI confidence interval, HR hazard ratio, PY person years. ## Discussion In the current cohort study, both obesity and metabolic health status were independently associated with increased RP risk. Participants with obesity showed a higher incidence of RP in both MH and MU groups, although this association was stronger in the MU group. Likewise, a significantly increased RP risk was observed in MU individuals, across all BMI categories. Conversely, OP risk decreased as BMI increased in both MH and MU groups. Although both FVC and FEV1 declined more annually in obese participants when compared to normal-weight participants, the mean difference in annual decline was higher for FVC than for FEV1. Consequently, the inverse association between obesity and OP risk observed in our study was attributed to the greater decline in FVC among obese individuals, rather than FEV1. To our knowledge, this is the first longitudinal study demonstrating the differential impact of BMI and metabolic health status on lung function impairment risk, depending on spirometry parameters. Previous reports also showed the harmful effects of obesity6,10,12 and metabolic abnormalities1,8,9,14,25 on RP. Obesity mechanically causes RP by decreasing the diaphragm and compromising chest wall compliance, resulting in limited lung expansion and decreased lung volume. Moreover, the most conceivable factor for the association between RP and metabolic abnormalities is insulin resistance, which is significantly higher in patients with RP than in patients with OP; therefore, the major effect of insulin resistance may be on lung tissue, with a slight effect on airway diameter14,26. Insulin resistance reduces glucose utilisation and induces abnormal fat metabolism in skeletal muscles, possibly impairing mitochondrial ATP production and reducing skeletal muscle strength27. As forced respiration during spirometry requires respiratory skeletal muscle contraction, insulin resistance may mediate a decline in lung function, especially a greater decline in FVC than FEV128. Insulin resistance-related hyperglycaemia29 can also cause non-enzymatic glycosylation of collagen and elastin in the lung and chest wall, leading to consequent stiffening of the thorax and lung parenchyma30; increased RP risk may be closely associated with a combination of the mechanical effect of obesity and metabolic effect of insulin resistance. Accordingly, in our study, the association between obesity and RP risk was stronger in MU than in MH individuals. Interestingly, in contrast to a previous cross-sectional study31,32, we demonstrated that MHO was not harmless, especially for RP. In addition to temporal ambiguity owing to the cross-sectional design, this study defined MHO as having two or fewer metabolic syndrome components. Because impaired lung function risk is related to each metabolic parameter1,14,25,33 and the number of metabolic components33,34, a less strict definition for MH may not have provided a clear comparison of obesity per se with normal weight MH individuals. Overweight and underweight individuals were included in the reference group. This definition of comparison groups makes the findings difficult to interpret because of the harmful effect of being underweight on lung function35, which was also observed in our study. The longitudinal nature of our study, the strict definition of MH, choice of MH normal-weight participants as the reference group, and availability of repeated BMI measurements and metabolic health status and incorporation in the analysis possibly allowed us to reveal the effects of MHO on lung function. In our study, obesity and metabolic abnormalities were associated with decreased OP risk, especially in the obese stratum. Until now, the effects of obesity and metabolic syndrome on OP have been controversial, varying from a negative25,36 to positive association1,11,12, although generally no association has been observed concerning metabolic syndrome or its components1,9,14,37. The reasons for the mixed results and inverse association between obesity, metabolic unhealthiness, and OP risk in our cohort are unclear. Previous studies have shown an association of OP with systemic inflammation38, not metabolic syndrome9,14,26. However, systemic inflammation is largely dependent on the degree of obesity, especially abdominal obesity and obesity commonly coexists with metabolic abnormalities and systemic inflammation39; thus, it is difficult to evaluate the effect of obesity versus other accompanying metabolic and inflammatory factors given their interrelationship. In our study using linear mixed models, a greater decline in both FVC and FEV1 was observed in obese than normal-weight participants, despite the inverse association between obesity and OP risk based on FEV1/FVC ratio. Thus, a more pronounced decline in FVC than FEV1 by obesity mechanisms may be an explanation because this change can result in a higher FEV1/FVC ratio36, leading to positive correlations between obesity and FEV1/FVC ratios. Furthermore, the impact of obesity on OP may be underestimated when diagnosing OP with conventional screening spirometry, which is performed to measure lung function based on patient effort, including deep breaths and forced expiration. This measurement likely obliterates the impact of obesity, particularly on FEV1. In addition, functional airway debility may go undetected in screening spirometry for healthy subjects, because FEV1/FVC < 0.7 predominantly reflects large airway obstruction40. Therefore, careful consideration is required when assessing OP based on screening spirometry, especially in healthy young and middle-aged subjects. In the current study, both obesity and metabolic abnormalities appeared to be important risk factors for RP, even in apparently healthy individuals, although their effect on OP appears to be complicated to determine based on conventional spirometry. Our findings have several important clinical implications. RP is associated with increased mortality and cardio-metabolic diseases3,41. Therefore, this study has important strengths because it demonstrates the potential role of modifiable obesity and metabolic health status on impaired lung function, given the projected growing public health impact of lung function3,6,41 and the high prevalence of obesity and metabolic syndrome8,12. However, our study has several limitations. First, our results were obtained from young and middle-aged asymptomatic and relatively healthy Korean adults who participated in a regular health check-up program. Therefore, our findings cannot be generalised to other demographic populations. Second, BMI was used as a measure of obesity. However, its inability to distinguish between the composition and distribution of fat and muscle mass could cause individuals with similar BMI to have very different body compositions and metabolic profiles. Finally, we did not examine the long-term effects of obesity or metabolic abnormalities on lung function. Because an increased risk of adverse clinical outcomes by metabolic abnormalities may occur only after 8–10 years42, the follow-up duration (median 4.7 years) of the current study might have been relatively short to evaluate the apparent effect of obesity and metabolic abnormalities on lung function. In conclusion, both BMI and metabolic health status appear to affect lung dysfunction, and their association varies depending on lung disease type. Obesity, even MHO, is not a harmless condition, especially for RP. Our results suggest that an RP of lung function impairment may be regarded as a pulmonary manifestation of metabolic syndrome and obesity. Therefore, maintaining a healthy weight and remaining MH may help prevent chronic lung diseases. However, the pathophysiological mechanisms involved in different relationships according to lung function impairment patterns require further study. ## Supplementary Information Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-31960-7. ## References 1. Leone N. **Lung function impairment and metabolic syndrome: The critical role of abdominal obesity**. *Am. J. Respir. Crit. Care Med.* (2009.0) **179** 509-516. DOI: 10.1164/rccm.200807-1195OC 2. Sin DD, Wu L, Man SF. **The relationship between reduced lung function and cardiovascular mortality: A population-based study and a systematic review of the literature**. *Chest* (2005.0) **127** 1952-1959. DOI: 10.1378/chest.127.6.1952 3. Mannino DM, Buist AS, Petty TL, Enright PL, Redd SC. **Lung function and mortality in the United States: Data from the First National Health and Nutrition Examination Survey follow up study**. *Thorax* (2003.0) **58** 388-393. DOI: 10.1136/thorax.58.5.388 4. Pellegrino R. **Interpretative strategies for lung function tests**. *Eur. Respir. J.* (2005.0) **26** 948-968. DOI: 10.1183/09031936.05.00035205 5. Young RP, Hopkins JR. **Primary and secondary prevention of chronic obstructive pulmonary disease: Where to next?**. *Am. J. Respir. Crit. Care Med.* (2014.0) **190** 839-840. DOI: 10.1164/rccm.201405-0883LE 6. Guerra S. **Health-related quality of life and risk factors associated with spirometric restriction**. *Eur. Respir. J.* (2017.0). DOI: 10.1183/13993003.02096-2016 7. Colak Y, Afzal S, Nordestgaard BG, Lange P, Vestbo J. **Importance of early COPD in young adults for development of clinical COPD: Findings from the Copenhagen general population study**. *Am. J. Respir. Crit. Care Med.* (2021.0) **203** 1245-1256. DOI: 10.1164/rccm.202003-0532OC 8. Baffi CW. **Metabolic syndrome and the lung**. *Chest* (2016.0) **149** 1525-1534. DOI: 10.1016/j.chest.2015.12.034 9. Nakajima K. **A possible association between suspected restrictive pattern as assessed by ordinary pulmonary function test and the metabolic syndrome**. *Chest* (2008.0) **134** 712-718. DOI: 10.1378/chest.07-3003 10. King GG. **The effects of body weight on airway calibre**. *Eur. Respir. J.* (2005.0) **25** 896-901. DOI: 10.1183/09031936.05.00104504 11. Lam KB. **Airflow obstruction and metabolic syndrome: the Guangzhou Biobank Cohort Study**. *Eur. Respir. J.* (2010.0) **35** 317-323. DOI: 10.1183/09031936.00024709 12. Dixon AE, Peters U. **The effect of obesity on lung function**. *Expert Rev. Respir. Med.* (2018.0) **12** 755-767. DOI: 10.1080/17476348.2018.1506331 13. Paek YJ. **Association between low pulmonary function and metabolic risk factors in Korean adults: The Korean National Health and Nutrition Survey**. *Metab. Clin. Exp.* (2010.0) **59** 1300-1306. DOI: 10.1016/j.metabol.2009.12.005 14. Lin WY, Yao CA, Wang HC, Huang KC. **Impaired lung function is associated with obesity and metabolic syndrome in adults**. *Obesity* (2006.0) **14** 1654-1661. DOI: 10.1038/oby.2006.190 15. Stefan N, Haring HU, Hu FB, Schulze MB. **Metabolically healthy obesity: Epidemiology, mechanisms, and clinical implications**. *Lancet Diabetes Endocrinol.* (2013.0) **1** 152-162. DOI: 10.1016/S2213-8587(13)70062-7 16. Park HY. **Blood eosinophil counts and the development of obstructive lung disease: The Kangbuk Samsung Health Study**. *Eur. Respir. J.* (2021.0). DOI: 10.1183/13993003.03823-2020 17. Craig CL. **International physical activity questionnaire: 12-country reliability and validity**. *Med. Sci. Sports Exerc.* (2003.0) **35** 1381-1395. DOI: 10.1249/01.MSS.0000078924.61453.FB 18. Ahn Y. **Validation and reproducibility of food frequency questionnaire for Korean genome epidemiologic study**. *Eur. J. Clin. Nutr.* (2007.0) **61** 1435-1441. DOI: 10.1038/sj.ejcn.1602657 19. 19.World Health Organization & Regional Office for the Western Pacific. The Asia-Pacific perspective: redefining obesity and its treatment. (Health Communications Australia, 2000). 20. Chang Y. **Metabolically healthy obesity and development of chronic kidney disease: A cohort study**. *Ann. Intern. Med.* (2016.0) **164** 305-312. DOI: 10.7326/M15-1323 21. Miller MR. **Standardisation of spirometry**. *Eur. Respir. J.* (2005.0) **26** 319-338. DOI: 10.1183/09031936.05.00034805 22. Choi HS. **Validation of previous spirometric reference equations and new equations**. *J. Korean Med. Sci.* (2019.0) **34** e304. DOI: 10.3346/jkms.2019.34.e304 23. Royston P, Parmar MK. **Flexible parametric proportional-hazards and proportional-odds models for censored survival data, with application to prognostic modelling and estimation of treatment effects**. *Stat. Med.* (2002.0) **21** 2175-2197. DOI: 10.1002/sim.1203 24. Royston P. **Flexible parametric alternatives to the Cox model, and more**. *Stand. Genomic Sci.* (2001.0) **1** 1-28. DOI: 10.1177/1536867x0100100101 25. Ford ES, Cunningham TJ, Mercado CI. **Lung function and metabolic syndrome: Findings of National Health and Nutrition Examination Survey 2007–2010**. *J. Diabetes* (2014.0) **6** 603-613. DOI: 10.1111/1753-0407.12136 26. Fimognari FL. **The association between metabolic syndrome and restrictive ventilatory dysfunction in older persons**. *J. Gerontol. Ser. A Biol. Sci. Med. Sci.* (2007.0) **62** 760-765. DOI: 10.1093/gerona/62.7.760 27. Barzilay JI. **Insulin resistance is associated with decreased quadriceps muscle strength in nondiabetic adults aged >or=70 years**. *Diabetes Care* (2009.0) **32** 736-738. DOI: 10.2337/dc08-1781 28. Kim SK. **Decline in lung function rather than baseline lung function is associated with the development of metabolic syndrome: A six-year longitudinal study**. *PLoS ONE* (2017.0) **12** e0174228. DOI: 10.1371/journal.pone.0174228 29. Gallagher EJ, Leroith D, Karnieli E. **Insulin resistance in obesity as the underlying cause for the metabolic syndrome**. *Mt. Sinai J. Med. N. Y.* (2010.0) **77** 511-523. DOI: 10.1002/msj.20212 30. Ofulue AF, Thurlbeck WM. **Experimental diabetes and the lung. II. In vivo connective tissue metabolism**. *Am. Rev. Respir. Dis.* (1988.0) **138** 284-289. DOI: 10.1164/ajrccm/138.2.284 31. Lee HY. **Metabolic health is more closely associated with decrease in lung function than obesity**. *PLoS ONE* (2019.0) **14** e0209575. DOI: 10.1371/journal.pone.0209575 32. Lee J. **The effect of metabolic health and obesity on lung function: a cross sectional study of 114,143 participants from Kangbuk Samsung Health Study**. *PLoS ONE* (2022.0) **17** e0266885. DOI: 10.1371/journal.pone.0266885 33. Wang F. **High-density lipoprotein cholesterol: A component of the metabolic syndrome with a new role in lung function**. *Evid.-based Complem. Altern. Med.: eCAM* (2021.0) **2021** 6615595. DOI: 10.1155/2021/6615595 34. Park BH. **Chronic obstructive pulmonary disease and metabolic syndrome: A nationwide survey in Korea**. *Int. J. Tuberc. Lung Dis. Off. J. Int. Union Against Tuberc. Lung Dis.* (2012.0) **16** 694-700. DOI: 10.5588/ijtld.11.0180 35. Do JG, Park CH, Lee YT, Yoon KJ. **Association between underweight and pulmonary function in 282,135 healthy adults: A cross-sectional study in Korean population**. *Sci. Rep.* (2019.0) **9** 14308. DOI: 10.1038/s41598-019-50488-3 36. Colak Y, Marott JL, Vestbo J, Lange P. **Overweight and obesity may lead to under-diagnosis of airflow limitation: Findings from the Copenhagen City Heart Study**. *COPD* (2015.0) **12** 5-13. DOI: 10.3109/15412555.2014.933955 37. Chung JH. **Association between sarcopenia and metabolic syndrome in chronic obstructive pulmonary disease: The Korea National Health and Nutrition Examination Survey (KNHANES) from 2008 to 2011**. *COPD* (2015.0) **12** 82-89. DOI: 10.3109/15412555.2014.908835 38. Gan WQ, Man SF, Senthilselvan A, Sin DD. **Association between chronic obstructive pulmonary disease and systemic inflammation: A systematic review and a meta-analysis**. *Thorax* (2004.0) **59** 574-580. DOI: 10.1136/thx.2003.019588 39. Wu H, Ballantyne CM. **Metabolic inflammation and insulin resistance in obesity**. *Circ. Res.* (2020.0) **126** 1549-1564. DOI: 10.1161/CIRCRESAHA.119.315896 40. Tuomi T, Heliovaara M, Palosuo T, Aho K. **Smoking, lung function, and rheumatoid factors**. *Ann. Rheum. Dis.* (1990.0) **49** 753-756. DOI: 10.1136/ard.49.10.753 41. Engstrom G. **Lung function, insulin resistance and incidence of cardiovascular disease: A longitudinal cohort study**. *J. Intern. Med.* (2003.0) **253** 574-581. DOI: 10.1046/j.1365-2796.2003.01138.x 42. Kramer CK, Zinman B, Retnakaran R. **Are metabolically healthy overweight and obesity benign conditions? A systematic review and meta-analysis**. *Ann. Intern. Med.* (2013.0) **159** 758-769. DOI: 10.7326/0003-4819-159-11-201312030-00008
--- title: Dual-signal readout paper-based wearable biosensor with a 3D origami structure for multiplexed analyte detection in sweat authors: - Yuemeng Cheng - Shaoqing Feng - Qihong Ning - Tangan Li - Hao Xu - Qingwen Sun - Daxiang Cui - Kan Wang journal: Microsystems & Nanoengineering year: 2023 pmcid: PMC10042807 doi: 10.1038/s41378-023-00514-2 license: CC BY 4.0 --- # Dual-signal readout paper-based wearable biosensor with a 3D origami structure for multiplexed analyte detection in sweat ## Abstract In this research, we design and implement a small, convenient, and noninvasive paper-based microfluidic sweat sensor that can simultaneously detect multiple key biomarkers in human sweat. The origami structure of the chip includes colorimetric and electrochemical sensing regions. Different colorimetric sensing regions are modified with specific chromogenic reagents to selectively identify glucose, lactate, uric acid, and magnesium ions in sweat, as well as the pH value. The regions of electrochemical sensing detect cortisol in sweat by molecular imprinting. The entire chip is composed of hydrophilically and hydrophobically treated filter paper, and 3D microfluidic channels are constructed by using folding paper. The thread-based channels formed after the hydrophilic and hydrophobic modifications are used to control the rate of sweat flow, which in turn can be used to control the sequence of reactions in the differently developing colored regions to ensure that signals of the best color can be captured simultaneously by the colorimetric sensing regions. Finally, the results of on-body experiments verify the reliability of the proposed sweat sensor and its potential for the noninvasive identification of a variety of sweat biomarkers. ## Introduction Point-of-care testing (POCT) has developed rapidly in recent years, as it has the advantage of not requiring pretreatment for the analysis of bodily fluids1–3. Among POCT devices, wearable sensors, a popular research topic, can be used to measure physicochemical parameters, such as body temperature4, electromyography5, heart rate6, and blood glucose7. They can also be simultaneously used for the noninvasive detection of biological fluids, such as human sweat8,9, saliva10,11, blood12,13, urine14,15, and tears16,17. Of these, sweat can be collected noninvasively and continuously as the subject exercises1,18 or through stimulation with pilocarpine19. Therefore, wearable sensors are suitable for sweat detection. Sweat contains many chemicals, including water, metabolites (glucose20,21, uric acid22,23, lactate24,25, cortisol19,26, and ethanol27), electrolytes28,29, and macromolecules30. These chemicals are closely associated with a wide range of diseases. Glucose detection can be a new method for the noninvasive detection of diabetes31,32. Lactate is an important biomarker of fatigue and anaerobic metabolism33. High concentrations of uric acid can lead to diseases such as diabetes, hypertension and gout22. The pH of sweat can be used to diagnose metabolic alkalosis34. Magnesium ions affect the “channels” through which potassium, sodium, and calcium ions move inside and outside cells. Cortisol is associated with stress, circadian rhythms, and major depression26,35. Of these, glucose, lactate and uric acid are associated with common physiological disorders, while H+, Mg2+ and cortisol concentrations are biochemical indicators for assessing mental stress. Current research on sweat detection can be divided into methods based on electrochemical sensing and optical sensing36–38. Colorimetric and electrochemical methods are the most commonly studied. Colorimetric methods for sweat detection39–41 are based on the relationship between the concentration of the analyte and the Red–Green–Blue (RGB) intensity of color. By designing microfluidic channels to collect and store the colorimetric sensing-based response of sweat markers, a smartphone can provide a convenient visual readout of the target biomarker42. Electrochemical sweat detection involves immobilizing biologically sensitive substances (antibodies26, enzymes1,43, etc.) on electrodes to generate specific reactions. With the development of flexible materials and electronics, wearable sensors have been further developed to greatly facilitate sweat detection6,9,29. Different materials (polymers8,43, paper-based materials40,44,45, and hydrogels18) and technologies (screen-printed technology46, 3D printing technology47) have been used to develop wearable sweat sensors. Wearable sweat sensors still have some shortcomings. First, the uniformity of the final color has a large impact on the results of color detection, and the colorimetric signals are easily influenced by light, distance, and angle. Second, the optimal response times of different analytical targets are different such that the optimal colorimetric signal cannot be simultaneously captured. Finally, different sweat analytes are detected in different ways and thus require different detection methods. In this study, we design and implement a low-cost, patch- and paper-based microfluidic sweat sensor that is easy to use. The sweat sensor simultaneously enables the combined noninvasive and on-body detection of multiple substances in human sweat, including glucose, lactate, uric acid, magnesium ions, and cortisol, as well as the corresponding pH value. The chip has a 3D origami structure to chemically construct hydrophobic microfluidic channels to selectively detect and quantify six analytes in sweat through different sensing strategies with respect to enzymatic reactions, pH indicators, complexes, and molecularly imprinted polymers (MIPs) on the surface of colorimetric and electrochemical electrodes. Wax dams are used to modify the hydrophilic properties of the thread-based channels, control the sequence of reactions of the colorimetric sensors to ensure optimal responses, and thus obtain the best results of detection. Finally, we perform on-body testing to evaluate the proposed sweat sensor, and the results verify its consistency and reliability. It has significant potential for use in the noninvasive detection of multiple biomarkers in sweat. ## Principle of reaction Sweat entered the colorimetric sensing area and the electrode layer, and the reaction ensued. The mechanism is shown in Fig. 1. The sensors for glucose, lactate, and uric acid (Fig. 1a, b) were dependent on the corresponding oxidase/horseradish peroxidase (HRP) cascade reaction. The measured indicators in sweat generated hydrogen peroxide (H2O2) under the catalysis of HRP. H2O2 oxidized 3,3’,5,5’-tetramethylbenzidine (TMB) and produced blue TMBox. The pH sensors (Fig. 1c) used pH indicators (litmus and bromophenol blue) that responded to changes in pH values ranging from 3 to 8 by changing color from yellow to purple. For sensors of the magnesium ion (Fig. 1d), a chromium black T indicator was used to interact with magnesium and change color from blue to purple. The cortisol electrochemical sensor (Fig. 1e) relied on the selective binding of cortisol to polypyrrole (PPy), thereby blocking electron transfer from the embedded Prussian blue (PB) redox probe. The binding of cortisol and PPy was determined by the cortisol concentration in sweat to allow the quantitative analysis of cortisol. Fig. 1Reaction principles of the six biomarkers in sweat. a Glucose, lactate and uric acid sensors respond to the oxidase/HRP cascade reaction. b Color change of the glucose, lactate and uric acid sensors at different concentrations. c pH sensor using a pH indicator (bromophenol blue as an example). d Magnesium ion sensor based on EBT. e Schematic diagram of the MIP-based cortisol sensing sensor ## Optimization and characterization of sweat chips The SEM results showed that as the duration in which wax was soaked in sweat increased, wax dams accumulated on the surface of the fiber in the cotton thread channels and thickened, as shown in Fig. 5g, h. With the gradual increase in the hydrophobicity of the cotton thread, the time needed for sweat to flow through increased. We used a bright blue solution to simulate sweat to investigate the effects of threads soaked in wax for different durations on the duration of liquid flow. As shown in Fig. 2a, the end of the thread was fixed in the circular paper-based zone of the reaction. Bright blue drops were added to the reaction zone at one end, and their real-time duration and process of flow were recorded. The results showed that the liquid could pass through the thread within 5 s without the latter being infiltrated by the wax solution. The tip of the cotton thread was briefly dipped in the wax solution, and the liquid could pass through ~25 s after diffusion. The tip of the cotton thread was then dipped in the wax solution until it had been completely infiltrated, and the liquid could then pass through in approximately 1 min. Following this, the cotton thread was completely infiltrated by the wax solution such that the liquid did not pass through it at all. Threads with different flow rates could thus be constructed for applications to control the sequence of the reaction in each area, according to the reaction rates of the target analytes in the five areas of colorimetric reaction, to achieve the best results. Fig. 2Effect of paper-based microfluidic channels investigated using a Brilliant Blue solution to simulate sweat. a Effects of threads soaked in wax for different durations on the duration of liquid flow. b Effect of the hydrophilic channel of the 3D paper-based microfluidic chip We prepared paper-based microfluidic chips by using the stamping method, which is easy, inexpensive, and fast. The chips had good hydrophobic properties, as needed, and this was confirmed by contact angle analysis. To verify the effect of the hydrophilic channel of the 3D paper-based microfluidic chip, a Brilliant Blue solution was added to its collection layer to observe the mobility of the liquid. Four chips were prepared, and 10 μL of the Brilliant Blue solution was added to them every 15 s. After dripping the solution on them 4, 8, 12, and 16 times from left to right, we unfolded the chips to observe the liquid flow inside. As shown in Fig. 2b, as the volume of the bright blue solution increased, the liquid gradually penetrated the collection layer and entered into the vertical channel, where it first came into contact with the electrode. The sweat then entered the horizontal channel, flowed through the thread into the colorimetric sensing layer, and eventually entered the volatile sweat layer to accumulate and evaporate when the volume of liquid was sufficiently large. The minimum amount of liquid required for the entire experiment was 160 μL. ## Performance of the sweat chip Images of the colorimetric sensor were acquired by a smartphone, and the intensity of the sensor was correlated with changes in the R, G, and B values to calculate the concentration of the analyte in sweat. The selection of the RGB data channel, the environment for photography, and the optimal response time all influenced the results. As shown in Supplementary Figs. S1–S5a in the ESM, the value of the R channel better fitted the change in color from white to blue caused by enzymatic reactions and that from blue to purple caused by the production of colored complexes. Therefore, we chose values of R for glucose, lactate, uric acid, and magnesium ions to quantify the intensity of color. For artificial sweat in the range of pH = 3–8, the (R + G + B)/3 value provided a significant difference among the pH indicators. The smartphone photo device was designed to ensure a dark environment while the flash of the smartphone served as the light source. Using glucose as an example, the calibration curves of the three were compared under natural light, incandescent light, and light from the photography device. The results (Supplementary Figs. S1–S5b) showed that the curves obtained using the photo device had the best linearity. To determine the optimum reaction time, artificial sweat containing 200 μM glucose, 20 mM lactate, 200 μM uric acid, and 5 mM magnesium chloride at pH = 3 was added to the colorimetric paper-based sensor. The intensities of the colors of glucose, uric acid, and pH were optimal after ~15 min, while those of lactate and magnesium were optimal after ~10 min (Supplementary Figs. S1–S5c). Owing to the fast reactions of lactate and magnesium, more hydrophobic thread channels were used for these biomarkers to delay the flow of sweat into the two sensing regions. Sweat flowed preferentially into the unwaxed thread during the reactions of glucose, uric acid, and pH. Thus, the more hydrophobic the thread used as the channel for lactate and magnesium was, the longer their response times, and this ultimately ensured optimal responses that yielded the best results for the five colorimetric reaction zones. In addition, when the volume of sweat was too large during the experiment, the pH indicator flowed back after completing the reaction and interfered with the detection of the other markers. We applied a hydrophobic thread to the pH channel to prevent this from occurring and obtained satisfactory results for the reaction. The experimental conditions (choice of RGB values, reaction times, reagent concentrations, etc.) were studied and optimized for each assay target by using artificial sweat samples, and details of the optimization and the results are provided in Supplementary Figs. S1–S5 in the ESM. The optimization of the concentrations of HRP, oxidase, TMB, and the indicator led to functionalized, modified filter paper. The dependence of the colorimetric signals on the concentrations of glucose, lactate, uric acid, pH, and magnesium ions under optimal conditions is shown in Fig. 3. The R2 values of glucose, lactate, uric acid, pH, and magnesium ions were 0.997, 0.991, 0.995, 0.994, and 0.992, respectively. Fig. 3Dependence of the analyzed signals on a glucose; b lactate; c uric acid; d pH; e magnesium ion; and f cortisol. Error bars indicate the standard deviation of the three sensors The MIP electrochemical cortisol sensor allowed quantitative measurement because different concentrations of cortisol molecules occupied the MIP cavity and impeded the charge transfer of Prussian blue. The regression equation for the range of concentrations from 1 × 10−9 M to 10 × 10−6 M was obtained by the concurrent method, with an R2 of 0.994 for 5 μL of cortisol added to the modified working electrode (Fig. 3f). In contrast, the curve did not change significantly after the dropwise addition of cortisol to the nonimprinted PPy electrode, demonstrating the lack of a cortisol-binding cavity within the PPy layer for detecting cortisol. Supplementary Fig. S6d shows the results of the incubation time optimization for cortisol. When the incubation time was less than 10 min, cortisol within the site was not fully bound. When the incubation time was longer than 10 min, the current response remained largely unchanged. Therefore, 10 min was used as the incubation time for cortisol. In addition, we investigated whether there was mutual interference among the six biomarkers. The responses of glucose, lactate, uric acid and cortisol were lowest in the corresponding sensor, while the responses of pH and magnesium ions were highest in the corresponding sensor. The results (in Supplementary Figs. S1–3g and S4–6e) showed good selectivity with no mutual interference. As the sweat chips were disposable, their reproducibility was assessed by using five sweat chips to measure the same samples under the same experimental conditions. The results (Supplementary Figs. S1–3h, S4–5f, and Supplementary Fig. S6c) showed that there was no significant difference in the color signals produced by the five biomarkers and that the results of the cortisol sensor were consistent as well. ## Assay of human sweat biomarkers Sweat samples from five adult volunteers in two states were analyzed to simultaneously identify the pH as well as the five biomarkers by using the proposed method, as shown in Fig. 4a. In this case, sweat from Subject 1 was collected while they were in a normal walking state, and that from Subjects 2–5 was collected as they were exercising. In the case of Subject 1, sweat was measured by fixing the chip to their arm for ~75 min. This sensing interval was very long. The performance of the sweat chip was assessed in terms of measuring the amounts of glucose, uric acid, and magnesium ions in the sweat of the subjects based on changes in their state before and after they had eaten (carbohydrates, animal offal high in purines, hazelnuts high in magnesium, and seafood). As shown in Fig. 4b, d, f, the concentrations of all three biomarkers increased after the subjects had eaten, with a consistent trend. However, owing to the different levels of digestion and metabolism of the subjects, the degree of improvement in these markers varied. The concentrations of the biomarkers in the sweat of Subject 1 were lower than those in the sweat of the other subjects, possibly owing to the lack of exercise that resulted in lower metabolism than that of the other subjects. Fig. 4Measurement of sweat biomarker levels in five adult volunteers using sweat sensors. a Assay of human sweat biomarkers. b Glucose; c lactate; d uric acid; e pH; f magnesium ion; g cortisol. Test subjects were told to perform low-intensity exercise (treadmill, 5 km/h) at 9 am on an empty stomach. At 6 pm, one hour after eating (carbohydrates, animal offal, hazelnuts, etc.), they performed high-intensity exercise (treadmill, 8 km/h). Test Subject 1 did not exercise, and resting sweat was collected. Error bars indicate the standard deviation of the three sensors The performance of the lactate sensor was assessed based on the intensity of the exercise performed by the subjects. Subject 1 had a lower resting lactate level (their R value increased) than that produced by exercise, demonstrating that exercise produces lactate. The other volunteers ran on a treadmill at 5 km/h at 9 am and 8–9 km/h at 6 pm. With the increase in the intensity of exercise, the lactate concentration increased to varying degrees (the R value decreased). Subject 2 had a high lactate concentration after exercise because they had a sweaty body type. The colorimetric sensor turned yellow, indicating an increase in the R value. The performance of the cortisol sensor was assessed by monitoring changes in the cortisol levels of the subjects in the morning and the afternoon. Figure 4g shows the signals obtained from all five subjects. They were higher in the afternoon than in the morning. The cortisol levels in the sweat of all subjects decreased consistently, demonstrating a high cortisol concentration in the morning and a low concentration in the afternoon. As cortisol levels are related to the level of stress, the extent of cortisol decline varied among the subjects. In addition, exercise also led to an increase in cortisol content, and thus, the resting subject had lower cortisol levels than the other subjects. ## Conclusion The paper-based microfluidic patch sweat sensor designed and implemented in this study enables the noninvasive and combined on-body detection of multiple indicators. Colorimetric and electrochemical detection was achieved by collecting sweat and analyzing the sweat indicators using different methods. The sensor has a 3D origami structure and is small, inexpensive, and portable. This makes it suitable for disposable use. We also constructed thread-based channels with different hydrophobic properties and used them to control the sequential response time of the colorimetric sensor for the optimal detection of multiple indicators during the same duration of sampling. A matching photo device for the chip ensured the consistency of the conditions of photography each time. The work here verified the potential and versatility of paper-based chips in the noninvasive identification of sweat biomarkers. ## Experimental materials and apparatus Uric acid (UA), magnesium chloride, lactate oxidase (LOD), uricase, 3,3’,5,5’-tetramethylbenzidine (TMB), and Eriochrome Black T Indicator (EBT) were purchased from Macklin Technologies Co., Ltd. (Shanghai, China). Cortisol and glucose oxidase (GOD) were purchased from Shenggong Bioengineering Co., Ltd. (Shanghai, China). L-Lactate was purchased from Bide Medical Technology Co., Ltd. (Shanghai, China), and carboxylated chitosan and peroxidase from horseradish (HRP) were purchased from Innochem Co., Ltd. (Beijing, China). Pyrrole, glucose, and hexane were purchased from Aladdin Co., Ltd. (Shanghai, China), and potassium ferricyanide (K3[Fe(CN)6]), litmus, bromophenol blue, sodium carbonate (Na2CO3), filter paper, and absolute ethanol were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). Brilliant Blue was purchased from Myrell Chemical Technology Co., Ltd. (Shanghai, China), alkylenone dimer (AKD) and triethanolamine were purchased from Shaoxin Biotechnology Co., Ltd. (Shanghai, China), phosphate buffer (PBS) was purchased from Jinxin BioReagent Studio, artificial sweat (ISO 3160-2) was purchased from Gao Xin Chemical Glass Instrument Co., Ltd., and cotton thread and wax were purchased from the market. Screen-printed electrodes were purchased from Zensor Electrochemical Cooperative and were modified with a CHI-660D ChenHua electrochemistry workstation. The cotton thread was characterized by scanning electron microscopy (SEM), and the hydrophilic and hydrophobic properties of the surface of the filter paper were characterized by a contact angle analyzer (DSA100). ## Preparation of sweat chips The wearable sweat chip consisted of two main parts: a microfluidic channel and a sensing area. It was square in shape, and each of its sides was 36 mm long. Its structure is shown in Fig. 5a, b. The chip formed a 3D channel from bottom to top that contained a collection layer (L1), a vertical channel (L2), an electrode layer, a horizontal channel (L3), a colorimetric sensing layer (L4), and a sweat evaporation layer (L5). Screen-printed electrodes were attached to the electrode layer to measure the cortisol in sweat. The colorimetric sensing layer consisted of five 10-mm-long cotton thread-based channels modified by using wax dams of different lengths and a circular sensing area with a diameter of 6 mm (filter paper modified with a specific chromogenic reagent). The remaining design of the layer (shown in Supplementary Fig. S7 in the ESM for paper-based channels) was patterned by using a chemical capping method48 to form the desired hydrophilic region. The entire process of sweat flow was as follows: sweat was absorbed through the collection layer, flowed up the vertical channel and the electrode layer through a chromatographic phenomenon, and reacted at the electrode layer. It flowed up again into the lateral channel, gathered at its center, and flowed into the colorimetric sensing layer. Finally, the excess sweat reached the evaporative layer. Fig. 5Diagram of the structure of the paper-based microfluidic noninvasive sweat sensor. a Diagram of the overall structure of the sweat sensor. b Overall solid view of the sweat sensor. c–e Contact angle tests of the hydrophobic and hydrophilic regions of the paper in the foldable sweat sensor. f–h SEM diagrams of cotton threads with different degrees of hydrophobicity (f: passes completely; g: passes for a certain time; h: cannot pass) The scheme for the preparation of each layer was as follows:i.Preparation of 3D paper-based microfluidic chips:The paper-based microfluidic pattern was designed using CAD. The hydrophilic pattern is represented by the white area in Fig. 5a, and the hydrophobic area formed by AKD is represented by the yellow area. The cut filter paper was first subjected to drying (37 °C in an oven), soaking (6 g/L AKD, the results of different AKD concentrations are shown in Supplementary Fig. S8), and drying again (37 °C oven) to form a clear hydrophobic barrier. Then, stamps of triethanolamine printing oil were used to stamp the corresponding areas, and they were placed in a 105 °C oven to dry to attain patterned hydrophobic areas. The hydrophobicity of the filter paper is shown in Fig. 5c–e.Cotton thread was used as the area of the channel. It was soaked in 10 mg/mL of an Na2CO3 solution (heated in a water bath for 10 min) and anhydrous ethanol (1 h) to remove surface wax and organic matter. The hydrophilic cotton threads were then placed in an oven at 37 °C to dry. The pretreated cotton threads were soaked in a wax solution for different times and then placed on a hot plate at 80 °C to allow the wax solution to diffuse within the cotton threads with different degrees of hydrophobicity. This was used to control the duration of passage of the liquid through the channel. SEM images of the cotton threads are shown in Fig. 5f–h.ii. Modification of the colorimetric sensing layer:We prepared sensors for glucose, lactate, and uric acid by the dropwise addition of chitosan, HRP, the corresponding oxidase, and TMB to the colorimetric sensing region. The volumes of HRP, oxidase, and TMB were modified and optimized. The results showed that when changes in their volumes for the glucose sensor were 0.2 mg/mL, 200 U/mL, and 15 mM, the sensor delivered the best performance. Similarly, the optimum changes in their volumes for the sensors of lactate and uric acid were 0.5 mg/mL, 20 U/mL, and 10 mM and 0.4 mg/mL, 100 U/mL, and 15 mM, respectively. The pH indicator (20:1 mixture of litmus and bromophenol blue) was modified for pH detection. The volume of the chromium black T indicator (1 g/L, 5 μL) was modified to detect magnesium ions. After drying the sensing and channel regions at room temperature, the colorimetric sensing layer was assembled using medical-grade tape and fixed to the 3D paper-based microfluidic chip. This was stored in a refrigerator at 4 °C.iii. Modification of screen-printed electrodes:The MIP films were fabricated by cyclic voltammetry (CV)-based electropolymerization in a PBS solution (0.1 M, pH = 7.4) containing 0.02 M pyrrole, 5 mM FeCl3, 5 mM K3[Fe (CN)6], 6 mM cortisol, and 0.1 M HCI at a potential range of −0.2 V to +0.9 V and a scan rate of 50 mV/s for ten cycles. After the electropolymerization process, the electrode was washed twice with deionized water to remove the remaining compounds. The MIP membrane was subjected to 20 cycles in PBS solution (0.1 M, pH = 7.4) by using CV over a potential range of –0.2 to +0.8 V at a scan rate of 50 mV/s. The embedded cortisol molecules were extracted from the MIP membrane to generate complementary cavities. The method used for the MIP above was applied to prepare the nonimprinted polymer (NIP), except that the cortisol molecule was not included as a template in the polymerization step. ## Detection of sensing signals The chip included areas for both colorimetric and electrochemical sensing. Image-related information was captured via a smartphone, and ImageJ was used to analyze and process the color signals. The electrochemical workstation was then used to detect and analyze the electrode signals (Scheme 1).Scheme 1Wearable sweat sensor based on a 3D paper structure for simultaneous analysis of multiple biomarkersi. Detection of colorimetric signals:A 30 μL sample of artificial sweat was dropped onto the chip. After 30 s, it rapidly entered the colorimetric sensing area via the hydrophobically modified thread-based channel to generate a colorimetric signal. For quantitative analysis, the color signal was collected by using a smartphone (HUAWEI P40 PRO), and the RGB color signal was analyzed by using the color analysis software ImageJ for the fast and accurate detection of the biomarkers. To ensure a consistent imaging environment (light, angle, and distance), a smartphone photo device was designed by using SolidWorks software and manufactured by using a 3D printer. The overall size of the device was 158 mm (L) × 72 mm (W) × 90 mm (H). It consisted of a smartphone, a black box, and a loading platform for the chip. The device was made of black resin to avoid interference from ambient light and unwanted reflections of light from inside the device during measurement. The size of the upper surface of the device was determined by the size of the smartphone. The loading platform was used to hold the paper-based chip, and the overall height of the device was determined by the focal length of the camera. It is shown in Supplementary Fig. S9 in the ESM.ii. Electrochemical signal detection:The screen-printed electrodes that were purchased formed a three-electrode system. The working electrode (3 mm in diameter) was a modified carbon electrode, the reference electrode was a Ag electrode, and the counter electrode was a carbon electrode. The electrode substrate was polypropylene, and the electrodes were adhered to the paper-based chip by medical tape. Chrono-current methods were implemented by using 0.1 V (relative to the Ag reference) to obtain calibration curves and implement on-body detection. ## Clinical samples This study was approved by the Medical Ethics Committee of Shanghai Jiao Tong University, China. All volunteers signed a document providing their informed consent, and all methods were performed in accordance with the relevant guidelines and regulations. The on-body assessment of the microchip was performed on five healthy volunteers (three males and two females). Before the test, the subjects were asked to perform low-intensity exercise (treadmill, 5 km/h) at 9 am on an empty stomach. At 6 pm, one hour after having eaten (carbohydrates, animal offal, hazelnuts, etc.), they performed high-intensity exercise (treadmill, 8–9 km/h). One of the subjects did not exercise, and their resting sweat was collected. The procedure was as follows: (i) The sweat chip was fixed to the subject’s upper-left arm with medical tape. ( ii) Half an hour of exercise caused their sweat to reach the sensing area through the microfluidic channel. ( iii) Once the response had been recorded, the chip was removed and used for data collection and analysis. ## Supplementary information SUPPLEMENTAL MATERIAL The online version contains supplementary material available at 10.1038/s41378-023-00514-2. ## References 1. Cao Q. **Three-dimensional paper-based microfluidic electrochemical integrated devices (3D-PMED) for wearable electrochemical glucose detection**. *RSC Adv.* (2019.0) **9** 5674-5681. DOI: 10.1039/C8RA09157A 2. 2.Ning, Q. H. et al. Point-of-care biochemical assays using electrochemical technologies: approaches, applications, and opportunities. Microchimica Acta189, 10.1007/s00604-022-05425-z (2022). 3. Wang K. **Smartphone-imaged multilayered paper-based analytical device for colorimetric analysis of carcinoembryonic antigen**. *Anal. Bioanal. Chem.* (2020.0) **412** 2517-2528. DOI: 10.1007/s00216-020-02475-1 4. 4.Hannah, S., Gleskova, H., Matuska, S. & Hudec, R. In 2018 ELEKTRO. 1–4 (IEEE, Peter Hockicko, Jozef Dubovan and Miroslav Markovic, 2018). 5. Hu Y. **Biomass polymer-assisted fabrication of aerogels from MXenes with ultrahigh compression elasticity and pressure sensitivity**. *J. Mater. Chem. A* (2019.0) **7** 10273-10281. DOI: 10.1039/C9TA01448A 6. Xia X, Zhang X, Serpe MJ, Zhang Q. **Microgel-based devices as wearable capacitive electronic skins for monitoring cardiovascular risks**. *Adv. Mater. Technol.* (2020.0) **5** 1900818. DOI: 10.1002/admt.201900818 7. Yao Y. **Integration of interstitial fluid extraction and glucose detection in one device for wearable non-invasive blood glucose sensors**. *Biosens. Bioelectron.* (2021.0) **179** 113078. DOI: 10.1016/j.bios.2021.113078 8. Nyein HYY. **A wearable patch for continuous analysis of thermoregulatory sweat at rest**. *Nat. Commun.* (2021.0) **12** 1823. DOI: 10.1038/s41467-021-22109-z 9. He J. **A thermoresponsive microfluidic system integrating a shape memory polymer-modified textile and a paper-based colorimetric sensor for the detection of glucose in human sweat**. *RSC Adv.* (2019.0) **9** 23957-23963. DOI: 10.1039/C9RA02831E 10. Calabria D. **Smartphone-based enzymatic biosensor for oral fluid L-lactate detection in one minute using confined multilayer paper reflectometry**. *Biosens. Bioelectron.* (2017.0) **94** 124-130. DOI: 10.1016/j.bios.2017.02.053 11. Manickam P, Pasha SK, Snipes SA, Bhansali S. **A reusable electrochemical biosensor for monitoring of small molecules (cortisol) using molecularly imprinted polymers**. *J. Electrochem. Soc.* (2016.0) **164** B54-B59. DOI: 10.1149/2.0781702jes 12. Gabriel EF. **Highly sensitive colorimetric detection of glucose and uric acid in biological fluids using chitosan-modified paper microfluidic devices**. *Analyst* (2016.0) **141** 4749-4756. DOI: 10.1039/C6AN00430J 13. Wang X. **Sensitive colorimetric assay for uric acid and glucose detection based on multilayer-modified paper with smartphone as signal readout**. *Anal. Bioanal. Chem.* (2018.0) **410** 2647-2655. DOI: 10.1007/s00216-018-0939-4 14. Jalal UM, Jin GJ, Shim JS. **Paper-plastic hybrid microfluidic device for smartphone-based colorimetric analysis of urine**. *Anal. Chem.* (2017.0) **89** 13160-13166. DOI: 10.1021/acs.analchem.7b02612 15. Lee W. **Thread/paper- and paper-based microfluidic devices for glucose assays employing artificial neural networks**. *Electrophoresis* (2018.0) **39** 1443-1451. DOI: 10.1002/elps.201800059 16. Moreddu R. **Integration of paper microfluidic sensors into contact lenses for tear fluid analysis**. *Lab. Chip* (2020.0) **20** 3970-3979. DOI: 10.1039/D0LC00438C 17. Xu J, Tao X, Liu X, Yang L. **Wearable eye patch biosensor for noninvasive and simultaneous detection of multiple biomarkers in human tears**. *Anal. Chem.* (2022.0) **94** 8659-8667. DOI: 10.1021/acs.analchem.2c00614 18. 18.Li, T. et al. An integrated and conductive hydrogel-paper patch for simultaneous sensing of chemical–electrophysiological signals. Biosens. Bioelectron.198, 10.1016/j.bios.2021.113855 (2022). 19. Tang W. **Touch-based stressless cortisol sensing**. *Adv. Mater.* (2021.0) **33** e2008465. DOI: 10.1002/adma.202008465 20. Ma D. **A novel photolithographic method for fabrication of flexible micro-patterned glucose sensors**. *J. Electroanal. Chem.* (2020.0) **876** 114720. DOI: 10.1016/j.jelechem.2020.114720 21. Promphet N. **Cotton thread-based wearable sensor for non-invasive simultaneous diagnosis of diabetes and kidney failure**. *Sens. Actuators B: Chem.* (2020.0) **321** 128549. DOI: 10.1016/j.snb.2020.128549 22. Yang Y. **A laser-engraved wearable sensor for sensitive detection of uric acid and tyrosine in sweat**. *Nat. Biotechnol.* (2020.0) **38** 217-224. DOI: 10.1038/s41587-019-0321-x 23. Dungchai W, Chailapakul O, Henry CS. **Use of multiple colorimetric indicators for paper-based microfluidic devices**. *Anal. Chim. Acta* (2010.0) **674** 227-233. DOI: 10.1016/j.aca.2010.06.019 24. Martin A. **Epidermal microfluidic electrochemical detection system: enhanced sweat sampling and metabolite detection**. *ACS Sens.* (2017.0) **2** 1860-1868. DOI: 10.1021/acssensors.7b00729 25. 25.Terse-Thakoor, T. et al. Thread-based multiplexed sensor patch for real-time sweat monitoring. NPJ Flexible Electron.4, 10.1038/s41528-020-00081-w (2020). 26. Torrente-Rodriguez RM. **Investigation of cortisol dynamics in human sweat using a graphene-based wireless mHealth system**. *Matter* (2020.0) **2** 921-937. DOI: 10.1016/j.matt.2020.01.021 27. Biscay J, Findlay E, Dennany L. **Electrochemical monitoring of alcohol in sweat**. *Talanta* (2021.0) **224** 121815. DOI: 10.1016/j.talanta.2020.121815 28. Nyein HYY. **A wearable microfluidic sensing patch for dynamic sweat secretion analysis**. *ACS Sens.* (2018.0) **3** 944-952. DOI: 10.1021/acssensors.7b00961 29. He W. **Integrated textile sensor patch for real-time and multiplex sweat analysis**. *Sci. Adv.* (2019.0) **5** eaax0649. DOI: 10.1126/sciadv.aax0649 30. Ji W. **Wearable sweat biosensors refresh personalized health/medical diagnostics**. *Research* (2021.0) **2021** 9757126. DOI: 10.34133/2021/9757126 31. Li G, Wen D. **Sensing nanomaterials of wearable glucose sensors**. *Chin. Chem. Lett.* (2021.0) **32** 221-228. DOI: 10.1016/j.cclet.2020.10.028 32. Chen Q, Zhao Y, Liu Y. **Current development in wearable glucose meters**. *Chin. Chem. Lett.* (2021.0) **32** 3705-3717. DOI: 10.1016/j.cclet.2021.05.043 33. Sakharov DA. **Relationship between lactate concentrations in active muscle sweat and whole blood**. *Bull. Exp. Biol. Med.* (2010.0) **150** 83-85. DOI: 10.1007/s10517-010-1075-0 34. Patterson MJ, Galloway SDR, Nimmo MA. **Effect of induced metabolic alkalosis on sweat composition in men**. *Acta Physiologica Scandinavica* (2002.0) **174** 41-46. DOI: 10.1046/j.1365-201x.2002.00927.x 35. 35.Sekar, M. et al. Review—towards wearable sensor platforms for the electrochemical detection of cortisol. J. Electrochem. Soc.167, 10.1149/1945-7111/ab7e24 (2020). 36. Yu H, Sun J. **Sweat detection theory and fluid driven methods: a review**. *Nanotechnol. Precis. Eng.* (2020.0) **3** 126-140. DOI: 10.1016/j.npe.2020.08.003 37. Luo M. **A paper-based plasma-assisted cataluminescence sensor for ethylene detection**. *Sens. Actuators B: Chem.* (2017.0) **240** 132-141. DOI: 10.1016/j.snb.2016.08.156 38. Liu X, Yang Y, Xing X, Wang Y. **Grey level replaces fluorescent intensity: fluorescent paper sensor based on ZnO nanoparticles for quantitative detection of Cu**. *Sens. Actuators B: Chem.* (2018.0) **255** 2356-2366. DOI: 10.1016/j.snb.2017.09.044 39. Vaquer A, Barón E, de la Rica R. **Detection of low glucose levels in sweat with colorimetric wearable biosensors**. *Analyst* (2021.0) **146** 3273-3279. DOI: 10.1039/D1AN00283J 40. Zhang Z. **A versatile, cost-effective, and flexible wearable biosensor for in situ and ex situ sweat analysis, and personalized nutrition assessment**. *Lab. Chip* (2019.0) **19** 3448-3460. DOI: 10.1039/C9LC00734B 41. Xiao G. **A wearable, cotton thread/paper-based microfluidic device coupled with smartphone for sweat glucose sensing**. *Cellulose* (2019.0) **26** 4553-4562. DOI: 10.1007/s10570-019-02396-y 42. Zheng J. **Microfluidic paper-based analytical device by using Pt nanoparticles as highly active peroxidase mimic for simultaneous detection of glucose and uric acid with use of a smartphone**. *Talanta* (2022.0) **237** 122954. DOI: 10.1016/j.talanta.2021.122954 43. Sempionatto JR. **Epidermal enzymatic biosensors for sweat vitamin C: toward personalized nutrition**. *ACS Sens.* (2020.0) **5** 1804-1813. DOI: 10.1021/acssensors.0c00604 44. Vaquer A, Baron E, de la Rica R. **Wearable analytical platform with enzyme-modulated dynamic range for the simultaneous colorimetric detection of sweat volume and sweat biomarkers**. *ACS Sens.* (2021.0) **6** 130-136. DOI: 10.1021/acssensors.0c01980 45. Yao S. **Programmed co-assembly of DNA-peptide hybrid microdroplets by phase separation**. *Chin. Chem. Lett.* (2022.0) **33** 1545-1549. DOI: 10.1016/j.cclet.2021.08.116 46. Ertuğrul Uygun HD, Uygun ZO, Canbay E, Gi̇rgi̇n Sağın F, Sezer E. **Non-invasive cortisol detection in saliva by using molecularly cortisol imprinted fullerene-acrylamide modified screen printed electrodes**. *Talanta* (2020.0) **206** 120225. DOI: 10.1016/j.talanta.2019.120225 47. 47.Calabria, D. et al. Smartphone-based chemiluminescent origami microPAD for the rapid assessment of glucose blood levels. Biosensors11, 10.3390/bios11100381 (2021). 48. Ning Q. **Rapid segmentation and sensitive analysis of CRP with paper-based microfluidic device using machine learning**. *Anal. Bioanal. Chem.* (2022.0) **414** 3959-3970. DOI: 10.1007/s00216-022-04039-x
--- title: The glycoprotein CD147 defines miRNA‐enriched extracellular vesicles that derive from cancer cells authors: - Song Yi Ko - WonJae Lee - Melanie Weigert - Eric Jonasch - Ernst Lengyel - Honami Naora journal: Journal of Extracellular Vesicles year: 2023 pmcid: PMC10042814 doi: 10.1002/jev2.12318 license: CC BY 4.0 --- # The glycoprotein CD147 defines miRNA‐enriched extracellular vesicles that derive from cancer cells ## Abstract Extracellular vesicles (EVs) are ideal for liquid biopsy, but distinguishing cancer cell‐derived EVs and subpopulations of biomarker‐containing EVs in body fluids has been challenging. Here, we identified that the glycoproteins CD147 and CD98 define subpopulations of EVs that are distinct from classical tetraspanin+ EVs in their biogenesis. Notably, we identified that CD147+ EVs have substantially higher microRNA (miRNA) content than tetraspanin+ EVs and are selectively enriched in miRNA through the interaction of CD147 with heterogeneous nuclear ribonucleoprotein A2/B1. Studies using mouse xenograft models showed that CD147+ EVs predominantly derive from cancer cells, whereas the majority of tetraspanin+ EVs are not of cancer cell origin. Circulating CD147+ EVs, but not tetraspanin+ EVs, were significantly increased in prevalence in patients with ovarian and renal cancers as compared to healthy individuals and patients with benign conditions. Furthermore, we found that isolating miRNAs from body fluids by CD147 immunocapture increases the sensitivity of detecting cancer cell‐specific miRNAs, and that circulating miRNAs isolated by CD147 immunocapture more closely reflect the tumor miRNA signature than circulating miRNAs isolated by conventional methods. Collectively, our findings reveal that CD147 defines miRNA‐enriched, cancer cell‐derived EVs, and that CD147 immunocapture could be an effective approach to isolate cancer‐derived miRNAs for liquid biopsy. ## INTRODUCTION Extracellular vesicles (EVs) hold great potential for liquid biopsy because EVs can be detected in body fluids and protect their cargo from degradation (Maas et al., 2017; Xu et al., 2018). Because the EV cargo often reflects the genetic and biological status of the parental cell, constituents of EV cargo have been extensively investigated as potential cancer biomarkers (Hannafon et al., 2016; Shin et al., 2021; Wang et al., 2022; Xu et al., 2018; Zhou et al., 2014). However, almost all types of cells release EVs and it is unclear what proportion of EVs in body fluids of cancer patients derive from cancer cells. Circulating EVs are elevated in other pathologic conditions such as diabetes and hypertension (Li et al., 2016; Preston et al., 2003) that are common comorbidities of cancer patients (Roy et al., 2018). Cancer cell‐derived EVs might constitute only a minor fraction of EVs in body fluids of cancer patients who have comorbid conditions and/or small tumours. As such, detecting biomarkers in these EVs could be very difficult. Currently, there are no well‐defined methods that can identify cancer cell‐derived EVs in body fluids. The tetraspanins CD63, CD81 and CD9 are commonly used as surface markers of EVs, but are ubiquitously expressed (Maecker et al., 1997) and cannot differentiate EVs that are released by cancer cells and by normal cells. It has been reported that the cell surface proteoglycan glypican‐1 is enriched in circulating EVs of patients with pancreatic cancer (Melo et al., 2015). However, glypican‐1 is expressed in the stroma as well as in pancreatic cancer cells, and has been detected in EVs released by stromal fibroblasts (Nigri et al., 2022; Tsujii et al., 2021). Developing approaches that can distinguish cancer cell‐derived EVs in body fluids is therefore critical to improve detection of biomarkers contained in these EVs. Another important consideration for evaluating biomarkers contained in EVs is that an individual cell type, including cancer cells, can release several subpopulations of EVs that vary in their cargo (Kowal et al., 2016). EVs have been captured from body fluids by using antibodies to CD63, CD81 or CD9 (Campos‐Silva et al., 2019; Duijvesz et al., 2015; Logozzi et al., 2009). However, there is considerable evidence that these tetraspanins are unevenly distributed across EVs (Han et al., 2021; Mathieu et al., 2021; Tian et al., 2018). Furthermore, EVs that express at least one tetraspanin have been found to constitute <$60\%$ of total EVs in cancer cell‐conditioned media and body fluids (Mizenko et al., 2021; Tian et al., 2018). Tetraspanin‐negative EVs have not been well‐characterized. However, glioblastoma cells have been shown to release epidermal growth factor receptor variant III in large EVs that lack CD63 and CD81 (Yekula et al., 2019) and platelet‐derived growth factor receptor‐α in EVs that lack all three tetraspanins (Lee et al., 2018). These studies highlight the importance of investigating tetraspanin‐negative EVs as a source of cancer biomarkers. MicroRNAs (miRNAs) contained in EVs have attracted substantial attention as candidate biomarkers for cancer diagnosis, prognosis and recurrence because miRNA expression patterns are frequently dysregulated in cancer (Hannafon et al., 2016; Shin et al., 2021; Wang et al., 2022; Xu et al., 2018; Zhou et al., 2014). However, the miRNA content of EVs has been contentious. On one hand, there are currently more than 10,000 EV‐associated miRNA entries in the Vesiclepedia database, a compendium of biomolecules identified in EVs (Kalra et al., 2012). On the other hand, stoichiometric analysis of the miRNA content in EVs and functional studies have revealed that the majority of EVs do not contain miRNA copy numbers that are biologically significant (Albanese et al., 2021; Chevillet et al., 2014; Zhang et al., 2021). An implication of these studies is that there could exist a subpopulation of EVs that is selectively enriched in miRNA, but this subpopulation has hitherto not been identified. Here, we show that the glycoprotein CD147 defines a subpopulation of EVs that is distinct from tetraspanin+ EVs in its biogenesis, content and cellular origin. In contrast to tetraspanin+ EVs, CD147+ EVs are not generated by the Endosomal Sorting Complex Required for Transport (ESCRT) machinery, and are selectively enriched in miRNA through the interaction of CD147 with the miRNA‐binding protein heterogeneous nuclear ribonucleoprotein A2/B1 (hnRNP A2/B1). Furthermore, in contrast to tetraspanin+ EVs, CD147+ EVs predominantly derive from cancer cells and increase in prevalence in cancer patients from early stages of disease. Moreover, isolating circulating miRNAs by CD147 immunocapture, as compared to conventional methods, increases the sensitivity of detecting cancer cell‐specific miRNAs and yields miRNAs that more closely reflect the tumour miRNA signature. Because CD147 is overexpressed in many types of cancer, CD147 immunocapture could be used to detect cancer‐derived circulating miRNAs in multiple disease sites. ## Antibodies and plasmids Antibodies used in this study are described in Table S1. Plasmids for expressing CD63‐GFP, CD81‐GFP, CD9‐GFP, CD147‐GFP and CD98‐GFP fusion proteins were as follows: CD63‐pEGFP C2 (gift from Paul Luzio, University of Cambridge; Addgene #62964), mEmerald‐CD81‐10 (He et al., 2013), mEmerald‐CD9‐10 (gifts from Michael Davidson, Florida State University; Addgene #54031, #54029), pCMV3‐CD147‐GFPSpark and pCMV3‐CD98‐GFPSpark (purchased from Sino Biological Inc.; HG10186‐ACG, HG16415‐ACG). Other plasmids were as follows: tetracycline‐regulated miR‐302 cluster expression plasmid pCW57‐GFP‐miR‐302 (Peskova et al., 2019) (gift from Tomáš Bárta, Masaryk University; Addgene #132549), LSB‐hsa‐miR‐302a‐3p and LSB‐hsa‐miR‐302c‐3p reporter plasmids (Gam et al., 2018) (gifts from Ron Weiss, Massachusetts Institute of Technology; Addgene #103400, #103403), TRIPZ HGS shRNA, TRIPZ TSG101 shRNA and TRIPZ vector (purchased from Horizon Discovery Biosciences; RHS4740‐EG9146, RHS4740‐EG7251, RHS4750). ## Cell culture Parental 293T, HeLa, SKOV3 and 786‐O cell lines, normal primary human renal proximal tubule epithelial cells (RPTEC), and normal primary human umbilical vein endothelial cells (HUVEC) were purchased from American Type Culture Collection. The ID8 cell line was provided by Katherine Roby (University of Kansas). Sources of cell lines that stably express the tetracycline repressor protein were as follows: T‐Rex™−293T, T‐Rex™‐HeLa (Invitrogen) and T‐Rex™‐SKOV3 (Applied Biological Materials Inc.). The 293T cell line in which the HNRNPA2B1 gene was deleted by CRISPR/*Cas9* gene editing was purchased from Abcam. All cell stocks were confirmed to be free of mycoplasma contamination and authenticated by short tandem repeat analysis. Culture media was purchased from Corning. Cell lines were cultured in Dulbecco's Modified Eagle Medium (293T, 786‐O, ID8), RPMI 1640 (HeLa) or McCoy's 5A medium (SKOV3) supplemented with $10\%$ FBS (for parental lines) or Tet‐approved FBS (for T‐Rex™ lines), 100 units/mL penicillin and 100 μg/mL streptomycin. RPTEC were cultured in Dulbecco's Modified Eagle Medium/Ham's F‐12 medium (1:1) supplemented with $10\%$ FBS, 100 units/mL penicillin and 100 μg/mL streptomycin. HUVEC were cultured in Medium 199 supplemented with $20\%$ FBS, Endothelial Cell Growth Supplement (EMD Millipore), 100 units/mL penicillin and 100 μg/mL streptomycin. ## Cell transfection Stably transfected cell lines were generated by transfecting cells using Lipofectamine™ 3000 reagent (Invitrogen) as follows. Parental 293T cells were transfected with marker‐GFP fusion expression plasmids, followed by sorting of GFP‐expressing cells using a BD FACSAria™ II cell sorter. T‐Rex™−293T cells were transfected with TRIPZ vector, TRIPZ HGS shRNA or TRIPZ TSG101 shRNA expression plasmids, followed by selection with 0.5 μg/mL puromycin (Sigma‐Aldrich). Expression of shRNA was induced by adding 1 μg/mL doxycycline (Sigma‐Aldrich) to culture media. Donor cells for assaying EV‐mediated transfer of miR‐302 were generated by transfecting cells of T‐Rex™‐lines with miR‐302 cluster expression plasmid, followed by sorting of GFP‐expressing cells. miR‐302 expression was induced in donor cells by adding 1 μg/mL doxycycline to culture media. Recipient cells were generated by transfecting parental 293T cells with miR‐302a or miR‐302c reporter constructs, followed by sorting of mKate2‐expressing cells. ## Clinical specimens Studies using human tissue specimens were reviewed and approved by the Institutional Research Board of the University of Texas MD Anderson Cancer Center and Institutional Research Board of the University of Chicago. All tissue specimens were residual and received full informed consent for research use from all human subjects. Specimens of tumour tissue, ascites and plasma of patients with ovarian carcinoma, and plasma of patients with benign gynaecologic conditions, were obtained from the Ovarian Cancer Tumour Bank at the University of Chicago and from the Southern Division of the National Cancer Institute‐supported Cooperative Human Tissue Network (CHTN) at Duke University. CA125 levels in plasma samples were measured by using the CA125 Quantikine ELISA Kit (R&D Systems). Specimens of tumour tissue and plasma of patients with renal cell carcinoma were obtained from the Eckstein Tissue Acquisition Laboratory at the University of Texas MD Anderson Cancer Center and from CHTN. Clinicopathologic features of cases are described in Table S2. Plasma was isolated using EDTA‐treated tubes from independent batches of peripheral blood of healthy adult volunteers that were obtained from the University of Texas MD Anderson Cancer Center Blood Bank. Each batch contained pooled blood from different donors. Analysis of EVs in specimens of body fluids was performed blinded to clinical data. ## Animal studies Animal studies were reviewed and approved by the Institutional Animal Care and Use Committee of the University of Texas MD Anderson Cancer Center. Four‐week‐old female nude mice (purchased from Envigo) were inoculated s.c. with 2 × 106 T‐Rex™‐HeLa cells that express the miR‐302 cluster, or with 5 × 106 parental 786‐O cells. To induce miR‐302 in T‐Rex™‐HeLa cells, mice were fed pellets containing doxycycline (200 mg per kg of diet) (Bio‐Serv). Tumour diameters were measured with callipers twice a week, and tumour volume was calculated from two perpendicular measurements. Blood samples (200 μL) were collected retro‐orbitally using EDTA‐treated tubes prior to cancer cell injection, when tumours first become palpable, and every 2 weeks thereafter until tumours reached a volume of ∼1000 mm3. Animals were then euthanized by CO2 asphyxiation. Platelet‐depleted plasma samples were obtained by centrifugation for 15 min at 2000 x g. The only mice that were excluded from analysis were those that did not form tumours. ## Isolation of EVs EVs were isolated from conditioned media, plasma and ascites as previously reported (Ko et al., 2019). In compliance with current guidelines of the International Society for Extracellular Vesicles (Théry et al., 2018), full details are provided. Cells were cultured in media containing $2\%$ FBS for 48 h to generate conditioned media. Conditioned media and body fluids were centrifuged at 2400 x g at 4°C for 10 min to remove cells and cell debris, and then concentrated by using a Centricon® Plus‐70 centrifugal filter unit with an Ultracel® 100 kDa cutoff filter (Millipore) to remove soluble proteins and particulates of <100 kD in size. Each concentrated supernatant was mixed with 1.5 mL of Optiprep™ stock solution ($60\%$ (w/v) aqueous iodixanol, Axis‐Shield PoC) and placed on the bottom of a 14 × 95 mm polyallomer ultracentrifuge tube (Beckman Coulter). Iodixanol solutions were prepared for the discontinuous gradient by diluting OptiPrep™ stock solution in buffer containing 0.25 M sucrose, 10 mM Tris‐HCl (pH 7.4) and 1 mM EDTA. The gradient was formed by layering iodixanol solutions in the following order: 3.0 mL of $40\%$ solution, 2.5 mL of $20\%$ solution, 2.5 mL of $10\%$ solution, and 2.0 mL of $5\%$ solution. Centrifugation was performed at 200,000 x g in a SW 40 Ti rotor (Beckman Coulter) at 4°C for 18 h. Ten gradient fractions of 1.0 mL were collected from the top to bottom. The density of each fraction was determined from absorbance readings at 244 nm using a standard curve of serial dilutions of iodixanol solution (Schröder et al., 1997). Individual fractions were washed with phosphate‐buffered saline (PBS), concentrated by using an Amicon® Ultra‐4 centrifugal filter unit with an Ultracel® 100 kDa cutoff filter (Millipore), and then suspended in PBS for further analysis or purification. To purify EVs that express a given surface marker, 10 μg of biotinylated antibody to the marker was incubated with 100 μL of streptavidin‐conjugated magnetic beads (Invitrogen) at 4°C for 16 h, and washed with PBS. Antibody‐conjugated magnetic beads were then incubated with EVs (∼1 × 108) for 16 h at 4°C. Following magnetic separation, supernatants containing marker‐negative EVs were collected for further analysis. Pellets containing marker‐positive EVs were washed three times with PBS, and then processed as described below to isolate protein or RNA. ## Flow cytometry Acquisition and analysis of flow cytometry data were performed using a BD FACSCanto™ II cytometer equipped with FACS Diva™ software (BD Biosciences). Concentrations of antibodies used are listed in Table S1. To detect cell surface proteins, cells were suspended in PBS containing $1\%$ bovine serum albumin (BSA), and incubated with FITC‐conjugated antibody or isotype control at 4°C for 30 min. Thereafter, cells were washed with PBS containing $1\%$ BSA, fixed in $4\%$ paraformaldehyde, and acquired. Staining was evaluated in the gated population of viable singlet cells. A minimum of 10,000 events was analysed for each sample. Three independent experiments were performed to verify expression of a given surface protein in each cell type. Settings to detect EVs were optimized by using bead calibration kits (100, 200, 500 and 760 nm diameter beads, Bangs Laboratories). Unless indicated otherwise, the absolute number of EVs in a given sample was calculated by using 760 nm beads as counting beads and multiplying the ratio of EV events to beads events with the number of beads in the sample. To detect EV surface proteins, 100 μL of EV sample (∼2 × 106 EVs) was incubated with FITC‐conjugated antibody or isotype control at room temperature (RT) for 30 min. Following incubation, samples were diluted to a final volume of 500 μL in PBS and acquired. Staining was evaluated in the gated population of singlet EVs. A minimum of 10,000 events was analysed for each sample. Three independent experiments were performed to verify expression of a given surface protein in EVs derived from each cell type, where each experiment used a different batch of EVs. Contour and histogram plots were generated by using FlowJo™ software (FlowJo, LLC). ## Particle size analysis and immunogold labelling Particle size distribution of purified EVs was analysed by using a ZetaView® QUATT instrument (Particle Metrix) by Alpha Nano Tech LLC. For each batch purification of EVs, an average of 2 × 1011 vesicles was isolated and 10 replicate measurements were made. EV markers were detected by immunogold labelling as previously reported (Ko et al., 2019). Briefly, carbon‐coated, formvar‐coated nickel grids (200 mesh) were treated with poly‐L‐lysine for 30 min. EVs were fixed with $2\%$ paraformaldehyde, loaded onto grids, and allowed to absorb at RT for 1 h. Grids were then placed into PBS containing $2\%$ BSA and $0.1\%$ saponin for 20 min, followed by incubation with primary antibody at 4°C for 16 h. Control grids were incubated without primary antibody. Grids were then rinsed with PBS and floated on drops of secondary antibody conjugated to 10 nm gold particles at RT for 2 h. Concentrations of antibodies used are listed in Table S1. Following incubation, grids were washed with PBS, fixed in $1\%$ glutaraldehyde for 5 min, and washed in H2O. Grids were stained for contrast for 1 min with $1\%$ uranyl acetate and allowed to dry. Samples were evaluated under a JEM 1010 transmission electron microscope (JEOL USA, Inc.) at an accelerating voltage of 80 Kv. Images were captured using the AMT Imaging System (Advanced Microscopy Techniques Corp.). ## Immunoprecipitation Cells were lysed with immunoprecipitation buffer ($1\%$ Triton X‐100, 25 mM HEPES, 150 mM NaCl, 5 mM MgCl2) containing protease inhibitor cocktail (Thermo Fisher Scientific). Cell lysates (500 μg) were then incubated with 10 μg of antibody‐conjugated agarose beads at 4°C for 16 h. Thereafter, beads were washed two times with immunoprecipitation buffer and three times with PBS. The pellets were dissolved in 2X Tris‐glycine sodium dodecyl sulphate (SDS) sample buffer (Thermo Fisher Scientific), and analysed by immunoblot. ## Immunoblot analysis Proteins were extracted by lysing cells and EVs in M‐PER buffer (Thermo Fisher Scientific). Protein concentrations of lysates were determined by Bradford assay (BioRad). Lysates were electrophoresed on SDS‐polyacrylamide gels and then transferred to polyvinylidene difluoride membrane (GE Healthcare). Membranes were blocked with $5\%$ non‐fat milk in Tris‐buffered saline with $0.1\%$ Tween‐20 (TBS‐T) at RT for 1 h and then incubated with primary antibody at 4˚C for 16 h. Following washing with TBS‐T buffer, membranes were incubated with HRP‐conjugated secondary antibody at RT for 45 min, washed, and visualized with ECL detection reagent (Millipore). Concentrations of antibodies used are listed in Table S1. Immunoblot data was verified in three independent experiments. ## Immunocytochemistry Cells were plated in chamber slides at sub‐confluence and allowed to adhere. Thereafter, cells were fixed with $4\%$ formaldehyde on ice for 20 min, and then permeabilized with $0.1\%$ Triton X‐100 in PBS on ice for 20 min. Cells were rinsed three times with PBS, blocked with $1\%$ goat serum in PBS for 30 min, and then incubated for 16 h with FITC‐conjugated antibody. Concentrations of antibodies used are listed in Table S1. Following washing with PBS, cells were stained with 4,6‐diamidino‐2‐phenylindole (DAPI) (Sigma‐Aldrich). Cells were viewed and photographed under a LSM 710 confocal microscope (Zeiss) using ZEN® software (Zeiss). ## Analysis of EV‐mediated miRNA transfer EVs were isolated as described above from parental cell lines and donor cells that express the miR‐302 cluster and were treated with 0.2 μg/mL of RNase A (Thermo Fisher Scientific) for 30 min at 37°C to remove external RNA. Recipient 293T cells that express miR‐302a and miR‐302c reporter constructs were seeded at 2 × 104 cells per well in black 96‐well Optical‐Bottom plates (Thermo Fisher Scientific). Recipient cells were incubated without or with the addition of EVs (∼2 × 106) at 37°C for 48 h. As positive controls, recipient cells were transfected with miR‐302a or miR‐302c mimetics (Sigma‐Aldrich) (1 × 104 copy number) by using Lipofectamine 3000 reagent. Following incubation, recipient cells were washed with PBS. Fluorescence intensities of the far‐red fluorescent protein mKate2 and the blue fluorescent protein EBFP2 were measured in recipient cells using a Spark® microplate reader (Tecan). Activity of each miRNA was calculated as mKate2 intensity relative to EBFP2 intensity. Three independent experiments were performed for each assay, where each experiment used a different batch of EVs. ## Analysis of EV uptake EVs were isolated as described above from 293T cells that express marker‐GFP fusion proteins. Parental 293T cells were plated in 96‐well plates (2 × 104 cells/well) and incubated with EVs (∼2 × 106) at 37°C for 3, 6, 12, 18 and 24 h. Cells were then washed three times with PBS and evaluated for EV uptake by measuring GFP fluorescence intensity using a Spark® microplate reader. Untreated 293T cells were used as a blank. Uptake of a given set of marker‐positive EVs was assessed in terms of GFP fluorescence at each time‐point relative to GFP fluorescence at 24 h after EV addition. Four independent experiments were performed for each assay, where each experiment used a different batch of EVs. ## Isolation and quantification of miRNA Prior to isolating miRNA, all batches of EVs were treated with 0.2 μg/mL of RNase A for 30 min at 37°C to remove external RNA. miRNA was isolated using the two‐column PureLink™ miRNA isolation kit (Invitrogen) following manufacturer's instructions. Briefly, EVs or cells were lysed with Trizol™ reagent, followed by the addition of chloroform and centrifugation at 12,000 × g for 15 min at 4°C. The aqueous phase was collected, mixed with an equal volume of $100\%$ ethanol, and then loaded onto the first column to retain large RNA. Following centrifugation of the column at 12,000 × g for 1 min, the flow‐through was collected, mixed with a 2‐fold volume of $100\%$ ethanol and then loaded onto the second column to retain small RNA. The second column was centrifuged at 12,000 × g for 1 min and washed twice with washing buffer provided in the kit. Small RNA was eluted from the second column by the addition of RNase‐free water (50 μL) and centrifugation at 12,000 × g for 1 min. miRNA was isolated from equivalent volumes of body fluid samples (100 μL of mouse plasma, 200 μL of human plasma or ascites) by three methods. In the direct lysis method, 600 μL of Trizol™ reagent was directly added to the fluid sample, and miRNA was isolated as described above. In the precipitation method, the fluid sample was diluted by the addition of 300 μL of PBS, incubated with 126 μL of ExoQuick® reagent (System Biosciences) at 4°C for 16 h, and then centrifuged at 1,500 × g for 30 min. The pellet was lysed with Trizol™ reagent, and miRNA was isolated as described above. In the CD147 immunocapture method, the fluid sample was diluted by the addition of 800 μL of PBS, and incubated with CD147 antibody‐conjugated magnetic beads at 4°C for 16 h. Thereafter, beads were washed three times with PBS. Trizol™ reagent was directly added to the beads, and miRNA was isolated as described above. Concentration of miRNA in samples was determined by using the Quant‐iT™ microRNA assay kit (Invitrogen) according to manufacturer's instructions. Size distribution of RNA molecules was evaluated by using a 2100 Bioanalyzer™ with a small RNA chip (Agilent) according to manufacturer's instructions. ## RT‐qPCR of miRNA To normalize miRNA content in EVs and body fluids, cel‐miR‐39 spike‐in control (Qiagen) was added to Trizol™ reagent during miRNA isolation. RT‐qPCR of miRNA was performed by using a Taqman® MicroRNA Reverse Transcription kit (Applied Biosystems) and sequence‐specific stem‐loop primers for hsa‐miR‐302a‐3p, hsa‐miR‐302c‐3p, hsa‐miR‐1233‐3p, hsa‐miR‐210‐3p, cel‐miR‐39‐3p, and RNU‐48 (Applied Biosystems). Reactions were performed in a StepOne Plus™ Real‐Time PCR system with 1X Master Mix and 1X probes (TaqMan® microRNA Expression Assay, Applied Biosystems). Relative levels of miRNAs in EVs were calculated by using the comparative CT method (2‐∆∆CT) and the cel‐miR‐39 spike‐in control for normalization. RNU‐48 was used as an endogenous control for normalization of cellular miRNA levels. Absolute copy numbers of miRNAs in Figure 4b and Figure 8b, d and e were calculated from standard curves that were generated using miRNA mimetics. ## miRNA profiling Reverse transcription of miRNA, isolated from body fluids and tumour tissue, was performed by using the miRCURY LNA™ RT kit (Qiagen) with cel‐miR‐39 as a spike‐in control. qPCR was carried out by using the miRCURY LNA™ miRNA cancer focus PCR panel (Qiagen) that contains primer sets for 84 known cancer‐associated miRNAs. Reactions were performed in a StepOne Plus™ Real‐Time PCR system (Applied Biosystem) with miRCURY SYBR® Green Master Mix (Qiagen). Cycling conditions were as follows: 2 min at 95°C, followed by 40 amplification cycles of 10 s at 95°C and 60 s at 56°C, and by the melting curve. Ct values obtained from the different panels were adjusted by an Inter Plate Calibrator. For each case, only those miRNAs that were detected in at least one of the three fluid‐derived samples were considered for further analysis. The expression level of each miRNA was normalized using the ΔCt method (ΔCt = Ct of each miRNA—Ct of Cel‐miR‐39). For each case, correlations between ΔCt values in each fluid‐derived sample and ΔCt values in matching tumour tissue were assessed by Spearman test. ## Statistical analysis Statistical analysis was performed by using GraphPad Prism 9.0 software (GraphPad Software). Normality of data distribution in groups was assessed by Shapiro‐Wilk test. Significance of data in in vitro and in vivo assays was assessed, where indicated, by unpaired two‐tailed Student's t‐test for comparison of two groups, or by one‐way or two‐way ANOVA with Bonferroni's corrections for multiple comparisons. Unless otherwise indicated, multiple comparisons were made of unpaired samples. Data represent means ± SD unless otherwise indicated. P values of <0.05 were considered significant. ## CD147 and CD98 define subpopulations of EVs that are distinct from tetraspanin+ EVs We initially reviewed the two largest databases of proteins that have been identified in EVs derived from diverse cell types and body fluids (Kalra et al., 2012; Keerthikumar et al., 2016). The 100 most frequently identified EV proteins in the ExoCarta and Vesiclepedia databases contained 77 common proteins of which 10 are membranous (Figure 1a). These include three tetraspanins (CD63, CD81, CD9) and seven other less‐characterized proteins. Expression of the tetraspanins and five other common membrane proteins (CD147, CD98, CD71, CD29, CD49f) was confirmed in human cell lines of diverse origin, namely, 293T (embryonic kidney), HeLa (cervical carcinoma), SKOV3 (ovarian carcinoma, OVCA) and 786‐O (renal cell carcinoma, RCC) (Figure S1). EVs were isolated from media conditioned by these cell lines using a method we previously optimized that includes ultrafiltration to remove soluble non‐EV proteins and particulates, followed by fractionation based on buoyant density (Ko et al., 2019). Purified EVs were visualized by transmission electron microscopy to confirm their intact membranous structure (Figure S2A), and their size distribution was determined by nanoparticle tracking analysis (Figure S2B). The prevalence of surface proteins in EVs was evaluated by flow cytometry using an approach that we applied in a previous study (Ko et al., 2019). Settings were optimized for EV detection by acquiring microbeads of various diameters in the size range of EVs, and then used to detect surface staining of proteins in EVs (Figure S3A). About $8\%$–$47\%$ of EVs contained either CD63, CD81 or CD9 (Figures 1b and S3B). CD147 and CD98 were detected in $19\%$– $41\%$ and $13\%$–$33\%$ of EVs, respectively, whereas CD71, CD29 and CD49f were detected in only $0.1\%$–$16\%$ of EVs (Figures 1b and S3B). A study of protein topology in EVs has identified that several membrane proteins are displayed in EVs in an orientation that is reverse to that in the plasma membrane, that is ‘inside‐out’ (Cvjetkovic et al., 2016). This raises the possibility that the low detection of CD71, CD29 and CD49f in EVs might stem from masking of antibody epitopes. These markers were therefore excluded from further analysis. To confirm that CD147, CD98 and the three tetraspanins are associated with EVs, all density gradient fractions were evaluated for these markers by immunoblot. Full‐length forms of all five proteins, and not truncated forms corresponding to shed ectodomains, were detected in fractions within the buoyant density range of EVs (Figure 1c). **FIGURE 1:** *CD147 and CD98 are predominantly expressed in tetraspanin‐negative EVs. (a) Venn diagram of unique and common top 100 EV proteins in the ExoCarta and Vesiclepedia databases. (b) Percentages of EVs derived from 293T, HeLa, SKOV3 and 786‐O cells that express the indicated membrane proteins. Mean ± SD of n = 3 independent experiments are shown. Representative flow cytometric analysis of staining is shown in Figure S3B. (c) Fractions of the indicated buoyant densities were isolated by density gradient ultracentrifugation from media conditioned by 293T, HeLa, SKOV3 and 786‐O cells, and assayed for CD63, CD81, CD9, CD147 and CD98 by immunoblot. As a positive control, fractions were assayed for tumour susceptibility gene 101 (TSG101), an exosomal cargo protein. EV‐containing fractions of buoyant densities of 1.09–1.14 g/mL are indicated by asterisks. Immunoblot data was verified in three independent experiments. (d) Batches of total EVs were depleted of EVs that express a given surface marker (Marker A). The remaining pool of EVs (Marker A‐negative subpopulation) and the total EV population were assayed for other surface markers (refer Figure S4A, B). Shown are the percentages of EVs that express either CD63, CD81, CD9, CD147 or CD98 in the total EV population and in the indicated subpopulations of marker‐negative EVs derived from 293T cells. Mean ± SD of n = 3 independent experiments are shown. Representative flow cytometric analysis of staining is shown in Figure S5A. ns, not significant, **P < 0.01, ****P < 0.0001, by one‐way ANOVA with Bonferroni's corrections in d.* We next evaluated co‐expression of CD147, CD98 and the three tetraspanins in EVs. Tetraspanins act as membrane‐organizing scaffolds by engaging with one another and other membrane proteins and protrude only 4–5 nm from the membrane (Hemler, 2005; Kitadokoro et al., 2001). Because of steric hindrance, it is difficult to reliably detect these proteins by co‐staining with multiple antibodies. To overcome this limitation, we depleted EVs that express a given surface marker (Marker A), and then determined the proportion of EVs that express another surface marker (Marker B) in the remaining pool of Marker A‐negative EVs (post‐depletion) and in the original total EV population (pre‐depletion) (Figure S4A, B). This approach was initially used to analyse EVs derived from 293T cells. As compared to the total EV population, depletion of EVs that express a given tetraspanin significantly reduced the proportions of EVs that express either of the other two tetraspanins (Figures 1d and S5A). These findings indicate that the majority but not the entirety of tetraspanin+ EVs coexpress at least two tetraspanins, and are in keeping with several reports (Han et al., 2021; Mathieu et al., 2021; Tian et al., 2018). By contrast, depletion of either CD147+ EVs or CD98+ EVs did not significantly reduce the proportions of EVs that are CD63+, CD81+ or CD9+ (Figures 1d and S5A). Conversely, depletion of either CD63+ EVs, CD81+ EVs or CD9+ EVs did not reduce the proportions of EVs that are CD147+ or CD98+ (Figures 1d and S5A). Similar results were obtained in depletion experiments using EVs derived from HeLa, SKOV3 and 786‐O cells (Figure S5B–E). These findings strongly indicate that CD147 and CD98 are predominantly expressed in tetraspanin‐negative EVs. To confirm these findings, we performed triple depletion of CD63+ EVs, CD81+ EVs and CD9+ EVs, and observed that the remaining pool of tetraspanin‐negative EVs was significantly enriched in EVs that are CD147+ or CD98+ (Figure S5F). Furthermore, analysis of CD147 expression in CD98‐negative EVs, and of CD98 expression in CD147‐negative EVs, revealed that expression of CD147 and CD98 in EVs is almost mutually exclusive (Figures 1d and S5A–E). ## Biogenesis of CD147+ and CD98+ EVs is distinct from that of tetraspanin+ EVs Two broad types of EVs released by live cells have been described in terms of their subcellular origin. Exosomes are derived from multivesicular endosomes and range from 30 to 150 nm in diameter, whereas microvesicles (ectosomes) form through outward budding of the plasma membrane and range from 100 nm to 1 μm in diameter (Maas et al., 2017; Mathieu et al., 2019; Xu et al., 2018). To further investigate the possibility that CD147+ EVs and CD98+ EVs are distinct from tetraspanin+ EVs, we evaluated the sizes of these EVs. A limitation to determining the size of EVs that express a given surface marker is that binding of antibody to the marker alters EV size, and it is difficult to detach the antibody without damaging the integrity of EVs. To overcome this limitation, we generated 293T cell lines that express either CD63, CD81, CD9, CD147 or CD98 as GFP‐fusion proteins, and evaluated the size of GFP+ EVs derived from each of these lines by flow cytometry and by fluorescence nanoparticle tracking analysis. These two independent analyses showed that CD147+ EVs and CD98+ EVs are larger than tetraspanin+ EVs (Figure 2a, b). To confirm these findings, we performed immunogold labelling of endogenous surface proteins in EVs. Tetraspanins were mostly detected in smaller EVs (Figure 2c), in keeping with reports that these proteins are contained in exosomes though not exclusively (Escola et al., 1998; Kowal et al., 2016; Mathieu et al., 2021). By contrast, CD147 and CD98 were mostly detected in larger EVs (Figure 2c). **FIGURE 2:** *CD147+ and CD98+ EVs are larger than tetraspanin+ EVs. (a, b) Size distribution of EVs derived from 293T cells that express either CD63, CD81, CD9, CD147 or CD98 as GFP‐fusion proteins. In (a), representative forward scatter versus side scatter plots of gated GFP+ EVs, with estimates of size distribution based on size marker bead gates (refer Figure S3A). In (b), size distributions of GFP+ EVs evaluated by fluorescence nanoparticle tracking analysis. Each plot shows the combined result of 10 replicate measurements. (c) Immunogold labelling of markers in EVs derived from HeLa cells.* Confocal microscopy revealed that CD147 and CD98 predominantly localize to the plasma membrane (Figure 3a). By contrast, CD63 and CD81 mostly localize to the cytoplasm and CD9 is both cytoplasmic and membranous (Figure 3a). These differences in subcellular localization implicated divergence in the biogenesis of CD147+ and CD98+ EVs and of tetraspanin+ EVs. The most characterized pathway of exosome biogenesis is orchestrated by the ESCRT machinery that comprises four multi‐subunit complexes (ESCRT‐0, ‐I, ‐II, ‐III) and several accessory components (Henne et al., 2011). Hepatocyte growth factor‐regulated tyrosine kinase substrate (HGS, also known as HRS) and tumour susceptibility gene 101 (TSG101) are core components of ESCRT‐0 and ESCRT‐I, respectively (Henne et al., 2011), and knockdown of these components inhibits exosome secretion (Colombo et al., 2013). Expression levels of CD147, CD98 and the three tetraspanins were not altered by knockdown of HGS or TSG101 (Figure 3b). Notably, knockdown of either HGS or TSG101 significantly inhibited secretion of tetraspanin+ EVs, but did not affect secretion of CD147+ EVs or CD98+ EVs (Figure 3c). These findings indicate that the vast majority of CD147+ EVs and CD98+ EVs are not exosomes. Visualization of cells at high magnification revealed the presence of CD147 and CD98 in regions of the plasma membrane that bud outwardly and pinch off (Figure 3a). Taken together, these findings support the notion that CD147+ EVs and CD98+ EVs most likely are microvesicles and represent subpopulations of EVs that are distinct from tetraspanin+ EVs. **FIGURE 3:** *Biogenesis of CD147+ and CD98+ EVs is distinct from that of tetraspanin+ EVs. (a) Subcellular localization of CD63, CD81, CD9, CD147 and CD98 detected by immunofluorescence staining in HeLa cells and visualized by confocal microscopy. Nuclei were visualized by staining with DAPI. Enlarged insets show the presence of CD147 and CD98 in regions of the plasma membrane that bleb outwardly or have pinched off (denoted by arrows). (b, c) 293T cells that express the tetracycline repressor protein were stably transfected with tetracycline‐regulated HGS shRNA or TSG101 shRNA, or with TRIPZ vector. In (b), levels of HGS, TSG101 and the indicated surface markers detected by immunoblot in equivalent amounts of cell lysates (20 μg) of untreated and doxycycline‐treated cells. In (c), numbers of CD63+, CD81+, CD9+, CD147+ and CD98+ EVs produced by equivalent numbers of 293T cells (∼5 × 106) without and following doxycycline‐induced knockdown of HGS and TSG101. Mean ± SD of n = 5 independent experiments are shown. ns, not significant, ****P < 0.0001, by unpaired two‐tailed Student's t‐test in c.* ## CD147+ EVs transport biologically active miRNAs into recipient cells Of the EV cargo, miRNAs have attracted substantial interest but several elegant studies have disputed the content and significance of miRNAs in EVs (Albanese et al., 2021; Arroyo et al., 2011; Chevillet et al., 2014; Zhang et al., 2021). To investigate the possibility that miRNAs are enriched in only a subpopulation of EVs, we firstly established an assay system to evaluate the export of miRNA by EVs secreted by donor cells and the import of bioactive miRNA by EVs into recipient cells (Figure 4a). To ensure that miRNA bioactivity in recipient cells can be attributed to EV‐mediated miRNA transfer, we chose the miR‐302 cluster as a test miRNA because it is not endogenously expressed in mature cells (Suh et al., 2004). *We* generated donor cell lines in which miR‐302 expression is induced by doxycycline (Figure 4b), and recipient cell lines that express a dual‐reporter cassette containing a miR‐302 target sequence (Gam et al., 2018) (Figure 4a). Bioactivity of miR‐302, that is transported in donor cell‐derived EVs and taken up by recipient cells, was assayed by measuring mKate2 fluorescence (Figure 4a). Robustness of the system was confirmed by the inhibition of mKate2 fluorescence in recipient cells following stimulation with EVs derived from miR‐302‐expressing donor cells (Figure 4c). Donor cell‐derived EVs were then depleted of EVs that express a given marker, and the remaining marker‐negative EVs were used to stimulate recipient cells. As compared to unstimulated cells, miR‐302 bioactivity did not decrease following stimulation with CD147‐negative EVs, but was decreased by $27\%$–$41\%$ following stimulation with CD98‐negative EVs (Figure 4d). Stimulation with tetraspanin‐negative EVs more greatly decreased miR‐302 bioactivity (by $53\%$–$83\%$) and was as effective as stimulation with total (non‐depleted) EVs (Figure 4d). No significant differences were found in the rates of uptake of tetraspanin+ EVs, CD147+ EVs and CD98+ EVs by recipient cells (Figure 4e). These results indicate that the vast majority of EV‐associated miR‐302 is contained in tetraspanin‐negative EVs. Consistent with our findings that CD147 and CD98 are predominantly expressed in tetraspanin‐negative EVs, copy numbers of miR‐302 were significantly higher in CD98+ EVs than in tetraspanin+ EVs, and even higher in CD147+ EVs (Figure 4f). **FIGURE 4:** *CD147+ EVs transport biologically active miRNAs into recipient cells. (a) Schematic of assay system for EV‐mediated transfer of miR‐302. (b) Copy numbers of miR‐302a and miR‐302c in untreated and doxycycline‐treated 293T, HeLa and SKOV3 donor cells. Mean ± SD of n = 3 independent experiments are shown. (c, d) 293T recipient cells that express mKate2 with either a 3′ miR‐302a or miR‐302c target sequence were stimulated with EVs. mKate2 fluorescence was measured at 48 h thereafter and expressed relative to fluorescence intensity in unstimulated recipient cells. In (c), recipient cells were stimulated with comparable numbers of EVs (∼2 × 106) derived from parental and miR‐302‐overexpressing donor cells. As positive controls, recipient cells were transfected with miR‐302a or miR‐302c mimetics. In (d), comparable numbers of EVs (∼2 × 106) derived from miR‐302‐overexpressing HeLa donor cells were depleted of EVs that express either CD63, CD81, CD9, CD147 or CD98 or left undepleted, and thereafter used to stimulate recipient cells. Mean ± SD of n = 3 independent experiments are shown in c and d. (e) Uptake of EVs was evaluated in parental 293T cells following addition of comparable numbers of EVs (∼2 × 106) derived from 293T cells that stably express either CD63, CD81, CD9, CD147 or CD98 as GFP‐fusion proteins. Uptake of a given set of EVs is expressed in terms of GFP fluorescence intensity at each indicated time‐point relative to GFP fluorescence intensity at 24 h after EV addition. Mean ± SD of n = 4 independent experiments are shown. (f) Comparable numbers of CD63+, CD81+, CD9+, CD147+ and CD98+ donor cell‐derived EVs (∼2 × 107) were evaluated for relative copy numbers of miR‐302a and miR‐302c. Copy number of cel‐miR‐39 spike‐in control was used for normalization. Mean ± SD of n = 3 independent experiments are shown. ns, not significant, *P <  0.05, **P <  0.01, ***P <  0.001, ****P <  0.0001 by unpaired two‐tailed Student's t‐test in b; by one‐way ANOVA with Bonferroni's corrections in c, d and f.* ## CD147+ EVs are enriched in miRNA through the interaction of CD147 with hnRNP A2/B1 In subsequent studies, we isolated tetraspanin+ EVs, CD147+ EVs and CD98+ EVs that are secreted by 293T, HeLa and SKOV3 cells, and measured the total miRNA content in comparable numbers of EVs of each subpopulation (Figure 5a). Tetraspanin+ EVs had the lowest miRNA content (Figure 5b). As compared to tetraspanin+ EVs, the total miRNA content was significantly though modestly higher (2–4‐fold) in CD98+ EVs derived from two of the cell lines, and substantially higher (8–17‐fold) in CD147+ EVs derived from all three cell lines (Figure 5b). Enrichment of miRNA in CD147+ EVs was confirmed by analysing the small RNA content in each subpopulation of EVs using an Agilent 2100 Bioanalyzer™ (Figures 5c and S6). To eliminate the possibility that miRNAs are associated with non‐EV components, purified EVs in all experiments were treated with RNase prior to miRNA analysis. Because non‐vesicular extracellular miRNAs form complexes with high‐density lipoprotein (HDL) (Vickers et al., 2011) and with Argonaute 2 (AGO2) (Arroyo et al., 2011), we confirmed that apolipoprotein A1 (APOA1), the major protein constituent of HDL, and AGO2 are not detectable in CD147+ EVs or CD98+ EVs (Figure 5d). To confirm our findings in clinical samples, we analysed EVs that were isolated from OVCA patient ascites and from plasma of patients with OVCA and RCC, and similarly found significant miRNA enrichment in CD147+ EVs (i.e. 9–26‐fold higher than tetraspanin+ EVs) (Figure 5e). **FIGURE 5:** *CD147+ EVs are enriched in miRNA through the interaction of CD147 with hnRNP A2/B1. (a, b) Comparable numbers of CD63+, CD81+, CD9+, CD147+ and CD98+ EVs derived from 293T, HeLa and SKOV3 cells were evaluated for miRNA content. In (a), numbers of marker‐positive EVs, calculated from the difference between EV counts in the initial input of total EVs and in the supernatant following immunocapture of EVs that express a given marker. In (b), total miRNA concentrations. Mean ± SD of n = 3 independent experiments are shown in a and b. (c) Agilent 2100 Bioanalyzer™ electropherogram profiles of small RNA isolated from HeLa cells and from comparable numbers of marker‐positive HeLa cell‐derived EVs (∼2 × 107). Small RNAs detected between the two dotted lines were considered as miRNAs. The peak at 4 nt corresponds to the loading control. (d) To confirm that CD147+ and CD98+ EVs are not contaminated with HDL or AGO2 complexes, total EVs were isolated from conditioned media of 293T, SKOV3 and HeLa cells, followed by immunocapture (IC) using antibodies to CD147, CD98 or Ig isotype control. Lysates of immunocaptured material were assayed by immunoblot for APOA1, the major protein constituent of HDL, and for AGO2. Conditioned media (CM) of 293T cells was included as a positive control for APOA1 and AGO2. (e) Total miRNA concentration in marker‐positive EVs isolated from ascites of OVCA patients (n = 5) and from plasma of patients with OVCA (n = 2) or RCC (n = 3). (f) Immunoblot of hnRNP A2/B1 in marker‐positive EVs derived from 293T cells. (g) Interaction of CD147 with hnRNP A2/B1 in 293T cells detected by immunoprecipitation (IP). (h, i) Comparable numbers of CD147+ EVs derived from parental and hnRNP A2/B1‐KO 293T cells were evaluated for total miRNA concentration (h) and small RNA content (i), as described in b and c. Mean ± SD of n = 3 independent experiments are shown in h. ns, not significant, *P < 0.05, **P < 0.01, ****P < 0.0001 by one‐way ANOVA with Bonferroni's corrections in a, b and e, where paired samples were analysed in e; by unpaired two‐tailed Student's t‐test in h.* Several RNA‐binding proteins control sorting of miRNA into EVs and have been detected in EVs (Lee et al., 2019; Santangelo et al., 2016; Temoche‐Diaz et al., 2019; Villarroya‐Beltri et al., 2013). One of these RNA‐binding proteins, hnRNP A2/B1, was detected in CD147+ EVs but not in tetraspanin+ EVs or CD98+ EVs (Figure 5f). To investigate the significance of hnRNP A2/B1 in CD147+ EVs, we used 293T cells in which the HNRNPA2B1 gene was deleted by CRISPR/*Cas9* gene editing (hnRNP A2/B1‐KO). Knockout of hnRNP A2/B1 did not alter cellular expression levels of CD147, CD98 and tetraspanins (Figure S7A), or the expression of these surface markers in EVs (Figure S7B, C). Notably, immunoprecipitation assays revealed that CD147 interacts with hnRNP A2/B1 (Figure 5g). Furthermore, the total miRNA content in CD147+ EVs derived from hnRNP A2/B1‐KO cells was significantly lower than the total miRNA content in CD147+ EVs derived from parental 293T cells (Figure 5h, i). These findings implicate that the selective enrichment of miRNA in CD147+ EVs occurs through the binding of hnRNP A2/B1 to miRNA and its interaction with CD147. ## CD147 is a candidate surface marker of cancer cell‐derived EVs Almost all types of cells release EVs, and there are no well‐defined surface markers that can distinguish EVs that are released by cancer cells into body fluids. CD147 and CD98 are overexpressed in a variety of solid tumours, but are also expressed in endothelial cells, leukocytes and some types of normal epithelium such as renal tubular epithelium (Cantor & Ginsburg 2012; Kosugi et al., 2015; Liao & Cantor, 2016; Xin et al., 2016; Xiong et al., 2014). To determine whether normal cells secrete CD147+ and CD98+ EVs, we evaluated renal proximal tubule epithelial cells (RPTEC) which are widely thought to be the cellular origin of clear cell RCC. Normal RPTEC expressed CD147 and CD98 at significantly lower levels than 786‐O RCC cells (Figure S8A), and secreted significantly fewer CD147+ and CD98+ EVs than equivalent numbers of 786‐O cells (Figure S8B). Similarly, normal endothelial cells expressed CD147 and CD98 at lower levels than 786‐O cells (Figure S8A), and secreted fewer CD147+ and CD98+ EVs (Figure S8B). By contrast, the numbers of tetraspanin+ EVs secreted by RPTEC and endothelial cells were not significantly different from numbers of tetraspanin+ EVs secreted by 786‐O cells (Figure S8B). These findings raise the possibility that CD147 and CD98 might enable identification of cancer cell‐derived EVs, and that CD147+ miRNA‐enriched EVs are predominantly secreted by cancer cells. To identify the cell‐of‐origin of CD147+ miRNA‐enriched EVs in body fluids, we analysed CD147+ EVs in longitudinally collected plasma samples of mice bearing human tumour xenografts. For comparison, we analysed mouse plasma EVs that express the tetraspanin CD9. Cancer cell‐derived EVs were distinguished from non‐cancerous host cell‐derived EVs by using antibodies that are specific to human and mouse surface markers, respectively. Validation of species‐specificity of antibodies is shown in Figure S9A. In mice with HeLa xenografts, the prevalence of CD9+ EVs and CD147+ EVs that are cancer cell‐derived progressively increased with tumour size (Figure 6a‐c). Notably, the significant increase in prevalence of CD147+ EVs was detected substantially earlier than for CD9+ EVs, that is at day 28 for CD147+ EVs when mean tumour size is 176 mm3 (Figure 6c) versus day 42 for CD9+ EVs when mean tumour size is 691 mm3 (Figure 6b). Similar results were obtained in mice with 786‐O xenografts (Figure 6a‐c). Furthermore, comparative analysis of EVs based on their cellular origin revealed that CD147+ EVs predominantly derive from cancer cells in both the HeLa and 786‐O models (Figures 6d and S9B). By contrast, the majority of CD9+ EVs were found to derive from non‐cancerous host cells (Figures 6d and S9B). These findings demonstrate that CD147+ EVs are predominantly released by cancer cells and from an early stage, and raise the possibility that increased levels of CD147+ EVs could be a useful indicator of early‐stage and/or low‐volume malignancy. **FIGURE 6:** *CD147+ EVs predominantly derive from cancer cells. Nude mice were inoculated s.c. with HeLa cells or with 786‐O cells. (a) Growth of tumours. Peripheral blood was collected at the indicated time‐points. (b‐d) Plasma EVs that derive from human cancer cells and from non‐cancerous mouse host cells were distinguished by staining with antibodies specific to human and mouse markers. Shown are percentages of CD9+ EVs (b) and CD147+ EVs (c) that derive from human cancer cells at each time‐point, and percentages of CD9+ EVs and CD147+ EVs that derive from human cancer cells (green columns) and from mouse host cells (purple columns) at the terminal time‐point (d). Mean ± SD of results in n = 5 mice are shown in a‐d. ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by one‐way ANOVA with Bonferroni's corrections in b and c; by two‐way ANOVA with Bonferroni's corrections for paired samples in d.* ## CD147+ EVs increase in prevalence in cancer patients from early stages of disease There is substantial evidence that EVs mediate cancer progression (Clement et al., 2020; Czystowska‐Kuzmicz et al., 2019; Ko et al., 2019; Xu et al., 2018; Zhou et al., 2014), but the heterogeneity of EVs in cancer is poorly understood. To gain insight, we analysed EVs in plasma samples of healthy adult volunteers and patients with either benign gynaecologic conditions or OVCA. Clinicopathologic features of cases are described in Table S2. As compared to healthy individuals, total numbers of EVs were elevated in patients with benign conditions and with early‐stage OVCA, and more so in patients with advanced‐stage OVCA (Figure 7a). Plasma levels of CA125, the most widely used OVCA biomarker, were elevated in OVCA patients, but alone could not differentiate patients with benign conditions and those with early‐stage OVCA (Figure S10A). The prevalence of tetraspanin+ EVs did not significantly differ between healthy individuals and patients with either benign conditions or OVCA (Figures 7b and S10B). Similar trends were found in plasma of patients with RCC. Total numbers of EVs were significantly elevated in patients with advanced‐stage RCC (Figure 7c), and there was no significant difference in the prevalence of tetraspanin+ EVs between healthy individuals and RCC patients irrespective of disease stage (Figures 7d and S11). **FIGURE 7:** *CD147+ EVs increase in prevalence in cancer patients from early stages of disease. Analysis of EVs isolated from equivalent volumes (200 μL) of plasma of (a, b) healthy adult volunteers and patients with either benign gynaecologic (gyn) conditions, Stage I/II OVCA or Stage III/IV OVCA, and (c, d) healthy adult volunteers and patients with either Stage I/II RCC or Stage IV RCC. Total numbers of EVs in plasma samples are shown in a and c. Percentages of EVs that express either CD63, CD81, CD9, CD147 or CD98 are shown in b and d. Representative flow cytometric analyses of staining are shown in Figures S10B and S11. Data of healthy volunteers is duplicated in a and c, and in b and d. Mean ± SD of n = 10 cases per group are shown in a‐d. ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by one‐way ANOVA with Bonferroni's corrections in a‐d.* In contrast to tetraspanin+ EVs, CD147+ EVs and CD98+ EVs each constituted only a small fraction (∼$1.6\%$) of the total EV population in healthy individuals (Figure 7b). The prevalence of CD98+ EVs was modestly (3‐fold) higher in patients with early‐stage and advanced‐stage OVCA (Figures 7b and S10B), and was 8‐fold higher in patients with advanced‐stage RCC but not significantly increased in those with early‐stage RCC (Figures 7d and S11). By contrast, the prevalence of CD147+ EVs was nearly 20‐fold higher in patients with early‐stage and advanced‐stage OVCA, and effectively differentiated OVCA patients, patients with benign conditions, and healthy individuals (Figures 7b and S10B). These findings raise the possibility that assaying CD147+ EVs could be more efficacious than CA125 for OVCA detection. Similarly, increased prevalence of CD147+ EVs was detected in RCC patients with early‐stage disease as well as those with advanced‐stage disease (Figures 7d and S11). ## CD147 immunocapture enhances detection of cancer‐derived circulating miRNAs Fractionation based on buoyant density has been widely regarded as the gold standard for isolating EVs at high purity (Théry et al., 2018), and was used to isolate all EVs in this study in conjunction with ultrafiltration. However, this isolation method is labour‐intensive and is impracticable for a clinical laboratory setting. Clinical studies of EV‐associated miRNAs have mostly used other methods to isolate EVs from body fluids rapidly but with less purity, such as precipitation with polymer‐based reagents (e.g. ExoQuick®) (Hannafon et al., 2016; Shin et al., 2021; Wang et al., 2022). EVs in body fluids have also been captured by using antibodies to tetraspanins (Campos‐Silva et al., 2019; Duijvesz et al., 2015; Logozzi et al., 2009). Our findings that CD147 is a candidate surface marker of cancer cell‐derived EVs and that miRNAs are enriched in CD147+ EVs raise the possibility that CD147 immunocapture could increase detection of cancer‐derived circulating miRNAs. To investigate this possibility, plasma samples of tumour‐bearing mice were divided into equivalent aliquots from which miRNA was isolated either by direct lysis of whole plasma, precipitation using ExoQuick® reagent, or immunocapture with CD147 antibody. Two groups of xenograft models were evaluated (HeLa, 786‐O). In both groups, the direct lysis and precipitation methods yielded 3–4‐fold more total miRNA than CD147 immunocapture (Figure 8a). To evaluate cancer cell‐specific miRNAs, we assayed miR‐302 in the HeLa model because tumours were established from miR‐302‐overexpressing HeLa cells and miR‐302 is not expressed in normal mature cells (Suh et al., 2004). In the 786‐O model, miR‐1233 was assayed because this miRNA is endogenously expressed in 786‐O cells (Dias et al., 2017) and has no mouse ortholog. Copy numbers of miR‐302a and miR‐1233 were 3–5‐fold higher in miRNA samples that were isolated by CD147 immunocapture than by the other two methods from plasma of mice with HeLa and 786‐O tumours, respectively (Figure 8b). These findings demonstrate that isolating circulating miRNAs by CD147 immunocapture increases the sensitivity of detecting cancer cell‐specific miRNAs. **FIGURE 8:** *CD147 immunocapture enhances detection of cancer‐derived circulating miRNAs. (a, b) Plasma was collected from mice with tumours derived from miR‐302‐expressing HeLa cells (at Day 50) or from parental 786‐O cells (at Day 49) (refer Figure 6a). miRNA was isolated from equivalent volumes of plasma (100 μL) either by direct lysis of whole plasma, precipitation using ExoQuick® reagent, or immunocapture (IC) with CD147 antibody. In (a), concentrations of total miRNA isolated by each of the three methods. In (b), copy numbers of miR‐302a (HeLa models) and miR‐1233 (786‐O models) detected in equivalent amounts of total miRNA. Mean ± SD of n = 5 independent samples are shown in a and b. (c) miRNA was isolated by each of the three methods from equivalent volumes of body fluids (ascites or plasma, 200 μL) of patients with OVCA or RCC. Expression levels of 84 cancer‐associated miRNAs in each of the three fluid‐derived samples of each case were evaluated by using a miRCURY LNA™ miRNA cancer focus PCR panel (Qiagen). For each case, only those miRNAs that were detected in at least one fluid‐derived sample were considered for correlation analysis. Shown are Spearman rank correlations between levels of miRNAs in each fluid‐derived sample and levels of miRNAs in matching tumour tissue of each case. Dotted lines indicate 95% confidence intervals. The number of miRNAs analysed for each case is indicated. (d, e) Copy numbers of miR‐210 detected in plasma of healthy volunteers and patients with either Stage I/II or Stage IV RCC where miRNA was isolated by direct lysis (d), and in the same plasma samples of healthy volunteers and patients with Stage I/II RCC where miRNA was isolated by direct lysis or CD147 immunocapture (e). Data of direct lysis samples is duplicated in d and e. Mean ± SD of n = 10 cases per group are shown. ns, not significant, *P  <  0.05, **P < 0.01, ***P  <  0.001, ****P  <  0.0001 by one‐way ANOVA with Bonferroni's corrections in a, b and d, where paired samples were analysed in a and b; by two‐way ANOVA with Bonferroni's corrections in e.* We next investigated whether miRNAs that are isolated from body fluids of cancer patients by CD147 immunocapture reflect the miRNA expression patterns of tumour tissues. Firstly, miRNA was isolated either by direct lysis, precipitation using ExoQuick® reagent or CD147 immunocapture from equivalent volumes of ascites of OVCA patients. For each case, expression levels of 84 cancer‐associated miRNAs were assessed in each of the three fluid‐derived miRNA samples, and then evaluated for correlations with expression levels of these miRNAs in matching tumour tissue. For all of the three cases analysed, the strongest correlation with the miRNA expression levels in the tumour was obtained with the fluid‐derived miRNA sample that was isolated by CD147 immunocapture (Figures 8c and S12). Similar results were obtained using plasma‐derived samples and matching tumour tissues of patients with either OVCA or RCC (Figure 8c). These findings indicate that circulating miRNAs that are isolated by CD147 immunocapture more closely reflect the tumour miRNA signature than circulating miRNAs that are isolated by conventional methods. Several independent reports have shown that miR‐210 is overexpressed in RCC and can be detected in the circulation of RCC patients (Dias et al., 2017; Iwamoto et al., 2014; Zhao et al., 2013). To investigate the possibility that isolating circulating miRNAs by CD147 immunocapture might improve the diagnostic performance of a cancer‐associated miRNA, we assayed levels of miR‐210 in plasma samples of the same cohorts of healthy individuals and RCC patients as in Figure 7c and d When circulating miRNA was isolated by direct lysis, a significant difference in miR‐210 copy numbers was detected between healthy individuals and patients with advanced‐stage RCC but not between healthy individuals and patients with early‐stage RCC (Figure 8d). By contrast, a significant difference in miR‐210 copy numbers was detected between healthy individuals and patients with early‐stage RCC when circulating miRNA was isolated by CD147 immunocapture (Figure 8e). Collectively, our findings indicate that CD147 immunocapture could be more effective than conventional methods for isolating cancer‐derived circulating miRNA for liquid biopsy. ## DISCUSSION The discovery that EVs convey biological information has reshaped our understanding of intercellular communication in homeostasis and disease. However, differentiating EVs in terms of their origin and content within a highly diverse population in body fluids has remained challenging. Several studies have interrogated the biochemical and biophysical heterogeneity among tetraspanin+ EVs released by cultured cells (Kowal et al., 2016; Mathieu et al., 2021; Temoche‐Diaz et al., 2019), but the characteristics of tetraspanin‐negative EVs are poorly defined. Here we report that CD147 and CD98 define subpopulations of EVs that are distinct from tetraspanin+ EVs. Although we cannot definitively delineate the subcellular origin of CD147+ EVs and CD98+ EVs, these EVs most likely represent microvesicles as CD147 and CD98 predominantly localize to the plasma membrane and notably in regions that bud outwardly and pinch off. In addition, CD147+ EVs and CD98+ EVs are produced in an ECSRT‐independent manner and are larger in size than tetraspanin+ EVs. We cannot rule out the possibility that CD147, CD98 and tetraspanins might be co‐expressed in a subpopulation of EVs. CD9 has been detected in microvesicles (Mathieu et al., 2021), and CD9+ EVs that contain CD147 have been detected in colorectal cancer patient sera (Yoshioka et al., 2014). Because microvesicles directly derive from the plasma membrane, it might be expected that the membrane protein repertoire of these EVs reflects that of the plasma membrane. Intriguingly, our study found that the distribution of CD147 and CD98 on EVs is almost mutually exclusive, whereas these proteins have been shown to interact on the cell surface (Cho et al., 2001). Taken together, these findings imply that clustering of CD147 and CD98 on the plasma membrane is dynamically reorganized during microvesicle formation. Along the same lines, Del Conde and colleagues found that activated monocytes release microvesicles that contain tissue factor but are deficient in CD45, whereas both proteins are expressed on the cell surface (Del Conde et al., 2005). EVs are often more highly secreted by cancer cells than by normal cells (Xu et al., 2018), but the origin of EVs in body fluids of cancer patients has not been established. Elevated levels of circulating CD147 have been detected in cancer patients (Łacina et al., 2022; Lee et al., 2016), and both the proteolytically cleaved extracellular domain of CD147 and its full‐length form are shed by cancer cells (Egawa et al., 2006). Our study supports a prior report that full‐length CD147 is shed, at least in part, in microvesicles (Sidhu et al., 2004). Our analysis of EVs secreted by normal cells indicate that the higher secretion of CD147+ EVs and CD98+ EVs by cancer cells is reflective of the higher cellular expression of CD147 and CD98 in cancer cells. Notably, our study shows that the prevalence of CD147+ EVs significantly increases from an early disease stage in OVCA and RCC patients and in xenograft models, and that CD147+ EVs predominantly derive from cancer cells. Other studies have detected elevated levels of CD147+ EVs in colorectal cancer patients, but the cell‐of‐origin of these EVs was not identified (Tian et al., 2018; Yoshioka et al., 2014). Furthermore, our study identified that the prevalence of tetraspanin+ EVs did not significantly differ between cancer patients and healthy individuals, and that the majority of CD9+ EVs, the most abundant subpopulation of tetraspanin+ EVs in body fluids, derive from non‐cancerous cells. Whereas the majority of previous functional studies of EVs in cancer have defined EVs by tetraspanin expression, our findings raise the possibility that CD147+ EVs could significantly account for the biological responses induced by cancer‐derived EVs. Moreover, our findings indicate that assaying CD147+ EVs instead of tetraspanin+ EVs or total EVs could be useful for detecting early‐stage or small tumours. Although a vast number of miRNAs have been detected in EVs, several studies have shown that the majority of EVs contain biologically insignificant amounts of miRNA (Albanese et al., 2021; Chevillet et al., 2014; Zhang et al., 2021). One explanation that could reconcile these findings is the existence of a subpopulation of EVs that is selectively enriched in miRNA, but this subpopulation has hitherto not been defined. A significant discovery in our study is that CD147+ EVs have a substantially (8–26‐fold) higher miRNA content than tetraspanin+ EVs. CD98+ EVs were found to have lower miRNA content than CD147+ EVs, indicating that enrichment of miRNA is not common to other microvesicles. Our findings indicate that the higher miRNA content in CD147+ EVs stems from the selective enrichment of hnRNP A2/B1 in CD147+ EVs through its interaction with CD147. It is unclear whether CD147 directly binds to hnRNP A2/B1. CD147 might interact with hnRNP A2/B1 through caveolin‐1. This possibility is supported by reports that hnRNP A2/B1 mediates sorting of miRNA into microvesicles by interacting with caveolin‐1 (Lee et al., 2019), and that caveolin‐1 interacts with CD147 (Tang & Hemler 2004). Although hnRNP A2/B1 has been reported to control sorting of miRNAs into exosomes (Villarroya‐Beltri et al., 2013), our findings support those of Jeppesen and colleagues who similarly did not detect hnRNP A2/B1 in tetraspanin+ EVs (Jeppesen et al., 2019). These investigators recently identified an intriguing class of miRNA‐enriched extracellular nanoparticles termed supermeres that contain hnRNP A2/B1 (Zhang et al., 2021). In contrast to EVs, supermeres lack an encompassing membrane and are substantially smaller than CD147+ EVs (i.e. <30 nm in diameter) (Zhang et al., 2021). Furthermore, supermeres are enriched in the RNA‐binding protein AGO2 (Zhang et al., 2021) which we did not detect in CD147+ EVs (Figure 5d). In addition, whereas supermeres contain shed ectodomains of membrane proteins (Zhang et al., 2021), only the full‐length form of CD147 was detected in EVs in the present study (Figures 1c and 5d). Collectively, these findings support the notion that CD147+ EVs are distinct from supermeres. Our findings that CD147 is a surface marker of cancer cell‐derived EVs, and that miRNAs are enriched in CD147+ EVs, are highly significant for liquid biopsy. Circulating miRNAs have attracted substantial interest as candidate biomarkers for cancer diagnosis, prognosis and recurrence (Dias et al., 2017; Hannafon et al., 2016; Shin et al., 2021; Wang et al., 2022; Zhou et al., 2014). However, trace amounts of miRNAs derived from small tumours may evade detection when isolated from body fluids by conventional methods of extracting all cell‐free miRNA or using polymer‐based reagents that precipitate EVs and also non‐EV material (Théry et al., 2018). Recently, small amounts of EV‐associated miRNAs in sera of breast cancer patients have been detected in situ by using nanoparticle probes called nanoflares (Zhao et al., 2020). However, this approach alone cannot distinguish EV‐associated miRNAs that derive from cancer cells from those that derive from non‐cancerous cells. Our study demonstrates that isolating circulating miRNAs by CD147 immunocapture increases the sensitivity of detecting cancer cell‐specific miRNAs, and that circulating miRNAs isolated by CD147 immunocapture more closely reflect the tumour miRNA signature than circulating miRNAs isolated by conventional methods. However, validation in large cohorts is needed. Because CD147 is overexpressed in ∼20 different types of cancer (Xin et al., 2016; Xiong et al., 2014), CD147 immunocapture could potentially be used to detect cancer‐derived circulating miRNAs in multiple disease sites. Furthermore, this approach only requires a small fluid sample (∼200 μL), does not require specialized equipment, and could be readily utilized in a clinical laboratory. ## AUTHOR CONTRIBUTIONS Song Yi Ko and Honami Naora developed the original hypothesis, conceived the study, and designed experiments. Song Yi Ko and WonJae Lee performed experiments. Song Yi Ko and Honami Naora analysed data. Melanie Weigert, Eric Jonasch and Ernst Lengyel provided clinical biospecimens. Song Yi Ko and Honami Naora wrote and edited the manuscript. WonJae Lee, Melanie Weigert, Eric Jonasch and Ernst Lengyel edited the manuscript. Honami Naora supervised the study. ## CONFLICT OF INTEREST STATEMENT The authors declare no competing interests. ## DATA AVAILABILITY STATEMENT All data associated with this study are present in the article and the Supplementary Information. ## References 1. Albanese M., Chen Y. A., Hüls C., Gärtner K., Tagawa T., Mejias‐Perez E., Keppler O. T., Göbel C., Zeidler R., Shein M., Schütz A. K., Hammerschmidt W.. **MicroRNAs are minor constituents of extracellular vesicles that are rarely delivered to target cells**. *PLoS Genetics* (2021) **17**. DOI: 10.1371/journal.pgen.1009951 2. Arroyo J. D., Chevillet J. R., Kroh E. M., Ruf I. K., Pritchard C. C., Gibson D. F., Mitchell P. S., Bennett C. F., Pogosova‐Agadjanyan E. L., Stirewalt D. L., Tait J. F., Tewari M.. **Argonaute2 complexes carry a population of circulating microRNAs independent of vesicles in human plasma**. *Proceedings of the National Academy of Sciences of the United States of America* (2011) **108** 5003-5008. DOI: 10.1073/pnas.1019055108 3. Campos‐Silva C., Suárez H., Jara‐Acevedo R., Linares‐Espinós E., Martinez‐Piñeiro L., Yáñez‐Mó M., Valés‐Gómez M.. **High sensitivity detection of extracellular vesicles immune‐captured from urine by conventional flow cytometry**. *Scientific Reports* (2019) **9** 2042. DOI: 10.1038/s41598-019-38516-8 4. Cantor J. M., Ginsberg M. H.. **CD98 at the crossroads of adaptive immunity and cancer**. *Journal of Cell Science* (2012) **125** 1373-1382. DOI: 10.1242/jcs.096040 5. Chevillet J. R., Kang Q., Ruf I. K., Briggs H. A., Vojtech L. N., Hughes S. M., Cheng H. H., Arroyo J. D., Meredith E. K., Gallichotte E. N., Pogosova‐Agadjanyan E. L., Morrissey C., Stirewalt D. L., Hladik F., Yu E. Y., Higano C. S., Tewari M.. **Quantitative and stoichiometric analysis of the microRNA content of exosomes**. *Proceedings of the National Academy of Sciences of the United States of America* (2014) **111** 14888-14893. DOI: 10.1073/pnas.1408301111 6. Cho J. Y., Fox D. A., Horejsi V., Sagawa K., Skubitz K. M., Katz D. R., Chain B.. **The functional interactions between CD98, beta1‐integrins, and CD147 in the induction of U937 homotypic aggregation**. *Blood* (2001) **98** 374-382. DOI: 10.1182/blood.V98.2.374 7. Clement E., Lazar I., Attané C., Carrié L., Dauvillier S., Ducoux‐Petit M., Esteve D., Menneteau T., Moutahir M., Le Gonidec S., Dalle S., Valet P., Burlet‐Schiltz O., Muller C., Nieto L.. **Adipocyte extracellular vesicles carry enzymes and fatty acids that stimulate mitochondrial metabolism and remodeling in tumor cells**. *The EMBO Journal* (2020) **39**. DOI: 10.15252/embj.2019102525 8. Colombo M., Moita C., van Niel G., Kowal J., Vigneron J., Benaroch P., Manel N., Moita L. F., Théry C., Raposo G.. **Analysis of ESCRT functions in exosome biogenesis, composition and secretion highlights the heterogeneity of extracellular vesicles**. *Journal of Cell Science* (2013) **126** 5553-5565. DOI: 10.1242/jcs.128868 9. Cvjetkovic A., Jang S. C., Konečná B., Höög J. L., Sihlbom C., Lässer C., Lötvall J.. **Detailed analysis of protein topology of extracellular vesicles ‐ Evidence of unconventional membrane protein orientation**. *Scientific Reports* (2016) **6**. DOI: 10.1038/srep36338 10. Czystowska‐Kuzmicz M., Sosnowska A., Nowis D., Ramji K., Szajnik M., Chlebowska‐Tuz J., Wolinska E., Gaj P., Grazul M., Pilch Z., Zerrouqi A., Graczyk‐Jarzynka A., Soroczynska K., Cierniak S., Koktysz R., Elishaev E., Gruca S., Stefanowicz A., Blaszczyk R., Golab J.. **Small extracellular vesicles containing arginase‐1 suppress T‐cell responses and promote tumor growth in ovarian carcinoma**. *Nature Communications* (2019) **10** 3000. DOI: 10.1038/s41467-019-10979-3 11. Del Conde I., Shrimpton C. N., Thiagarajan P., López J. A.. **Tissue‐factor‐bearing microvesicles arise from lipid rafts and fuse with activated platelets to initiate coagulation**. *Blood* (2005) **106** 1604-1611. DOI: 10.1182/blood-2004-03-1095 12. Dias F., Teixeira A. L., Ferreira M., Adem B., Bastos N., Vieira J., Fernandes M., Sequeira M. I., Maurício J., Lobo F., Morais A., Oliveira J., Kok K., Medeiros R.. **Plasmatic miR‐210, miR‐221 and miR‐1233 profile: potential liquid biopsies candidates for renal cell carcinoma**. *Oncotarget* (2017) **8** 103315-103326. DOI: 10.18632/oncotarget.21733 13. Duijvesz D., Versluis C. Y. L., van der Fels C. A. M., Berg V.‐v. d., M S., Leivo J., Peltola M. T., Bangma C. H., Pettersson K. S. I., Jenster G.. **Immuno‐based detection of extracellular vesicles in urine as diagnostic marker for prostate cancer**. *International Journal of Cancer* (2015) **137** 2869-2878. DOI: 10.1002/ijc.29664 14. Egawa N., Koshikawa N., Tomari T., Nabeshima K., Isobe T., Seiki M.. **Membrane type 1 matrix metalloproteinase (MT1‐MMP/MMP‐14) cleaves and releases a 22‐kDa extracellular matrix metalloproteinase inducer (EMMPRIN) fragment from tumor cells**. *Journal of Biological Chemistry* (2006) **281** 37576-37585. DOI: 10.1074/jbc.M606993200 15. Escola J.‐M., Kleijmeer M. J., Stoorvogel W., Griffith J. M., Yoshie O., Geuze H. J.. **Selective enrichment of tetraspan proteins on the internal vesicles of multivesicular endosomes and on exosomes secreted by human B‐lymphocytes**. *Journal of Biological Chemistry* (1998) **273** 20121-20127. DOI: 10.1074/jbc.273.32.20121 16. Gam J. J., Babb J., Weiss R.. **A mixed antagonistic/synergistic miRNA repression model enables accurate predictions of multi‐input miRNA sensor activity**. *Nature Communications* (2018) **9** 2430. DOI: 10.1038/s41467-018-04575-0 17. Han C., Kang H., Yi J., Kang M., Lee H., Kwon Y., Jung J., Lee J., Park J.. **Single‐vesicle imaging and co‐localization analysis for tetraspanin profiling of individual extracellular vesicles**. *Journal of Extracellular Vesicles* (2021) **10**. DOI: 10.1002/jev2.12047 18. Hannafon B. N., Trigoso Y. D., Calloway C. L., Zhao Y. D., Lum D. H., Welm A. L., Zhao Z. J., Blick K. E., Dooley W. C., Ding W. Q.. **Plasma exosome microRNAs are indicative of breast cancer**. *Breast Cancer Research* (2016) **18** 90. DOI: 10.1186/s13058-016-0753-x 19. He J., Sun E., Bujny M. V., Kim D., Davidson M. W., Zhuang X.. **Dual function of CD81 in influenza virus uncoating and budding**. *PLoS Pathogens* (2013) **9**. DOI: 10.1371/journal.ppat.1003701 20. Hemler M. E.. **Tetraspanin functions and associated microdomains**. *Nature Reviews Molecular Cell Biology* (2005) **6** 801-811. DOI: 10.1038/nrm1736 21. Henne W. M., Buchkovich N. J., Emr S. D.. **The ESCRT pathway**. *Developmental Cell* (2011) **21** 77-91. DOI: 10.1016/j.devcel.2011.05.015 22. Iwamoto H., Kanda Y., Sejima T., Osaki M., Okada F., Takenaka A.. **Serum miR‐210 as a potential biomarker of early clear cell renal cell carcinoma**. *International Journal of Oncology* (2014) **44** 53-58. DOI: 10.3892/ijo.2013.2169 23. Jeppesen D. K., Fenix A. M., Franklin J. L., Higginbotham J. N., Zhang Q., Zimmerman L. J., Liebler D. C., Ping J., Liu Q., Evans R., Fissell W. H., Patton J. G., Rome L. H., Burnette D. T., Coffey R. J.. **Reassessment of exosome composition**. *Cell* (2019) **177** 428-445. DOI: 10.1016/j.cell.2019.02.029 24. Kalra H., Simpson R. J., Ji H., Aikawa E., Altevogt P., Askenase P., Bond V. C., Borràs F. E., Breakefield X., Budnik V., Buzas E., Camussi G., Clayton A., Cocucci E., Falcon‐Perez J. M., Gabrielsson S., Gho Y. S., Gupta D., Harsha H. C., Mathivanan S.. **Vesiclepedia: A compendium for extracellular vesicles with continuous community annotation**. *PLoS Biology* (2012) **10**. DOI: 10.1371/journal.pbio.1001450 25. Keerthikumar S., Chisanga D., Ariyaratne D., Al Saffar H., Anand S., Zhao K., Samuel M., Pathan M., Jois M., Chilamkurti N., Gangoda L., Mathivanan S.. **ExoCarta: A web‐based compendium of exosomal cargo**. *Journal of Molecular Biology* (2016) **428** 688-692. DOI: 10.1016/j.jmb.2015.09.019 26. Kitadokoro K., Bordo D., Galli G., Petracca R., Falugi F., Abrignani S., Grandi G., Bolognesi M.. **CD81 extracellular domain 3D structure: Insight into the tetraspanin superfamily structural motifs**. *The EMBO Journal* (2001) **20** 12-18. DOI: 10.1093/emboj/20.1.12 27. Ko S. Y., Lee W., Kenny H. A., Dang L. H., Ellis L. M., Jonasch E., Lengyel E., Naora H.. **Cancer‐derived small extracellular vesicles promote angiogenesis by heparin‐bound, bevacizumab‐insensitive VEGF, independent of vesicle uptake**. *Communication Biology* (2019) **2** 386. DOI: 10.1038/s42003-019-0609-x 28. Kosugi T., Maeda K., Sato W., Maruyama S., Kadomatsu K.. **CD147 (EMMPRIN/Basigin) in kidney diseases: From an inflammation and immune system viewpoint**. *Nephrology Dialysis Transplantation* (2015) **30** 1097-1103. DOI: 10.1093/ndt/gfu302 29. Kowal J., Arras G., Colombo M., Jouve M., Morath J. P., Primdal‐Bengtson B., Dingli F., Loew D., Tkach M., Théry C.. **Proteomic comparison defines novel markers to characterize heterogeneous populations of extracellular vesicle subtypes**. *Proceedings of the National Academy of Sciences of the United States of America* (2016) **113** E968-E977. DOI: 10.1073/pnas.1521230113 30. Łacina P., Butrym A., Frontkiewicz D., Mazur G., Bogunia‐Kubik K.. **Soluble CD147 (BSG) as a prognostic marker in multiple myeloma**. *Current Issues in Molecular Biology* (2022) **44** 350-359. DOI: 10.3390/cimb44010026 31. Lee A., Rode A., Nicoll A., Maczurek A. E., Lim L., Lim S., Angus P., Kronborg I., Arachchi N., Gorelik A., Liew D., Warner F. J., McCaughan G. W., McLennan S. V., Shackel N. A.. **Circulating CD147 predicts mortality in advanced hepatocellular carcinoma**. *Journal of Gastroenterology and Hepatology* (2016) **31** 459-466. DOI: 10.1111/jgh.13148 32. Lee H., Li C., Zhang Y., Zhang D., Otterbein L. E., Jin Y.. **Caveolin‐1 selectively regulates microRNA sorting into microvesicles after noxious stimuli**. *Journal of Experimental Medicine* (2019) **216** 2202-2220. DOI: 10.1084/jem.20182313 33. Lee K., Fraser K., Ghaddar B., Yang K., Kim E., Balaj L., Chiocca E. A., Breakefield X. O., Lee H., Weissleder R.. **Multiplexed profiling of single extracellular vesicles**. *ACS Nano* (2018) **12** 494-503. DOI: 10.1021/acsnano.7b07060 34. Li S., Wei J., Zhang C., Li X., Meng W., Mo X., Zhang Q., Liu Q., Ren K., Du R., Tian H., Li J.. **Cell‐derived microparticles in patients with type 2 diabetes mellitus: A systematic review and meta‐analysis**. *Cellular Physiology and Biochemistry* (2016) **39** 2439-2450. DOI: 10.1159/000452512 35. Liao Z., Cantor J. M.. **Endothelial cells require CD98 for efficient angiogenesis**. *Arteriosclerosis, Thrombosis, and Vascular Biology* (2016) **36** 2163-2166. DOI: 10.1161/ATVBAHA.116.308335 36. Logozzi M., De Milito A., Lugini L., Borghi M., Calabrò L., Spada M., Perdicchio M., Marino M. L., Federici C., Iessi E., Brambilla D., Venturi G., Lozupone F., Santinami M., Huber V., Maio M., Rivoltini L., Fais S.. **High levels of exosomes expressing CD63 and caveolin‐1 in plasma of melanoma patients**. *PLoS One* (2009) **4**. DOI: 10.1371/journal.pone.0005219 37. Maas S. L. N., Breakefield X. O., Weaver A. M.. **Extracellular vesicles: Unique intercellular delivery vehicles**. *Trends in Cell Biology* (2017) **27** 172-188. DOI: 10.1016/j.tcb.2016.11.003 38. Maecker H. T., Todd S. C., Levy S.. **The tetraspanin superfamily: Molecular facilitators**. *The FASEB Journal* (1997) **11** 428-442. PMID: 9194523 39. Mathieu M., Martin‐Jaular L., Lavieu G., Théry C.. **Specificities of secretion and uptake of exosomes and other extracellular vesicles for cell‐to‐cell communication**. *Nature Cell Biology* (2019) **21** 9-17. DOI: 10.1038/s41556-018-0250-9 40. Mathieu M., Névo N., Jouve M., Valenzuela J. I., Maurin M., Verweij F. J., Palmulli R., Lankar D., Dingli F., Loew D., Rubinstein E., Boncompain G., Perez F., Théry C.. **Specificities of exosome versus small ectosome secretion revealed by live intracellular tracking of CD63 and CD9**. *Nature Communications* (2021) **12** 4389. DOI: 10.1038/s41467-021-24384-2 41. Melo S. A., Luecke L. B., Kahlert C., Fernandez A. F., Gammon S. T., Kaye J., LeBleu V. S., Mittendorf E. A., Weitz J., Rahbari N., Reissfelder C., Pilarsky C., Fraga M. F., Piwnica‐Worms D., Kalluri R.. **Glypican‐1 identifies cancer exosomes and detects early pancreatic cancer**. *Nature* (2015) **523** 177-182. DOI: 10.1038/nature14581 42. Mizenko R. R., Brostoff T., Rojalin T., Koster H. J., Swindell H. S., Leiserowitz G. S., Wang A., Carney R. P.. **Tetraspanins are unevenly distributed across single extracellular vesicles and bias sensitivity to multiplexed cancer biomarkers**. *Journal of Nanobiotechnology* (2021) **19** 250. DOI: 10.1186/s12951-021-00987-1 43. Nigri J., Leca J., Tubiana S.‐S., Finetti P., Guillaumond F., Martinez S., Lac S., Iovanna J. L., Audebert S., Camoin L., Vasseur S., Bertucci F., Tomasini R.. **CD9 mediates the uptake of extracellular vesicles from cancer‐associated fibroblasts that promote pancreatic cancer cell aggressiveness**. *Science Signaling* (2022) **15**. DOI: 10.1126/scisignal.abg8191 44. Peskova L., Cerna K., Oppelt J., Mraz M., Barta T.. **Oct4‐mediated reprogramming induces embryonic‐like microRNA expression signatures in human fibroblasts**. *Scientific Reports* (2019) **9**. DOI: 10.1038/s41598-019-52294-3 45. Preston R. A., Jy W., Jimenez J. J., Mauro L. M., Horstman L. L., Valle M., Aime G., Ahn Y. S.. **Effects of severe hypertension on endothelial and platelet microparticles**. *Hypertension* (2003) **41** 211-217. DOI: 10.1161/01.HYP.0000049760.15764.2D 46. Roy S., Vallepu S., Barrios C., Hunter K.. **Comparison of comorbid conditions between cancer survivors and age‐matched patients without cancer**. *Journal of Clinical Medicine Research* (2018) **10** 911-919. DOI: 10.14740/jocmr3617w 47. Santangelo L., Giurato G., Cicchini C., Montaldo C., Mancone C., Tarallo R., Battistelli C., Alonzi T., Weisz A., Tripodi M.. **The RNA‐binding protein SYNCRIP is a component of the hepatocyte exosomal machinery controlling microRNA sorting**. *Cell Reports* (2016) **17** 799-808. DOI: 10.1016/j.celrep.2016.09.031 48. Schröder M., Schäfer R., Friedl P.. **Spectrophotometric determination of iodixanol in subcellular fractions of mammalian cells**. *Analytical Biochemistry* (1997) **244** 174-176. DOI: 10.1006/abio.1996.9861 49. Shin S., Park Y. H., Jung S. H., Jang S. H., Kim M. Y., Lee J. Y., Chung Y. J.. **Urinary exosome microRNA signatures as a noninvasive prognostic biomarker for prostate cancer**. *NPJ Genomic Medicine* (2021) **6** 45. DOI: 10.1038/s41525-021-00212-w 50. Sidhu S. S., Mengistab A. T., Tauscher A. N., LaVail J., Basbaum C.. **The microvesicle as a vehicle for EMMPRIN in tumor–stromal interactions**. *Oncogene* (2004) **23** 956-963. DOI: 10.1038/sj.onc.1207070 51. Suh M. R., Lee Y., Kim J. Y., Kim S. K., Moon S. H., Lee J. Y., Cha K.‐Y., Chung H. M., Yoon H. S., Moon S. Y., Kim V. N., Kim K. S.. **Human embryonic stem cells express a unique set of microRNAs**. *Developmental Biology* (2004) **270** 488-498. DOI: 10.1016/j.ydbio.2004.02.019 52. Tang W., Hemler M. E.. **Caveolin‐1 regulates matrix metalloproteinases‐1 induction and CD147/EMMPRIN cell surface clustering**. *Journal of Biological Chemistry* (2004) **279** 11112-11118. DOI: 10.1074/jbc.m312947200 53. Temoche‐Diaz M. M., Shurtleff M. J., Nottingham R. M., Yao J., Fadadu R. P., Lambowitz A. M., Schekman R.. **Distinct mechanisms of microRNA sorting into cancer cell‐derived extracellular vesicle subtypes**. *Elife* (2019) **8**. DOI: 10.7554/eLife.47544 54. Théry C., Witwer K. W., Aikawa E., Alcaraz M. J., Anderson J. D., Andriantsitohaina R., Antoniou A., Arab T., Archer F., Atkin‐Smith G. K., Ayre D. C., Bach J.‐M., Bachurski D., Baharvand H., Balaj L., Baldacchino S., Bauer N. N., Baxter A. A., Bebawy M., Zuba‐Surma E. K.. **Minimal information for studies of extracellular vesicles 2018 (MISEV2018): A position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines**. *Journal of Extracellular Vesicles* (2018) **7**. DOI: 10.1080/20013078.2018.1535750 55. Tian Y., Ma L., Gong M., Su G., Zhu S., Zhang W., Wang S., Li Z., Chen C., Li L., Wu L., Yan X.. **Protein profiling and sizing of extracellular vesicles from colorectal cancer patients via flow cytometry**. *ACS Nano* (2018) **12** 671-680. DOI: 10.1021/acsnano.7b07782 56. Tsujii S., Serada S., Fujimoto M., Uemura S., Namikawa T., Nomura T., Murakami I., Hanazaki K., Naka T.. **Glypican‐1 is a novel target for stroma and tumor cell dual‐targeting antibody–drug conjugates in pancreatic cancer**. *Molecular Cancer Therapeutics* (2021) **20** 2495-2505. DOI: 10.1158/1535-7163.mct-21-0335 57. Vickers K. C., Palmisano B. T., Shoucri B. M., Shamburek R. D., Remaley A. T.. **MicroRNAs are transported in plasma and delivered to recipient cells by high‐density lipoproteins**. *Nature Cell Biology* (2011) **13** 423-433. DOI: 10.1038/ncb2210 58. Villarroya‐Beltri C., Gutiérrez‐Vázquez C., Sánchez‐Cabo F., Pérez‐Hernández D., Vázquez J., Martin‐Cofreces N., Martinez‐Herrera D. J., Pascual‐Montano A., Mittelbrunn M., Sánchez‐Madrid F.. **Sumoylated hnRNPA2B1 controls the sorting of miRNAs into exosomes through binding to specific motifs**. *Nature Communications* (2013) **4** 2980. DOI: 10.1038/ncomms3980 59. Wang W., Jo H., Park S., Kim H., Kim S. I., Han Y., Lee J., Seol A., Kim J., Lee M., Lee C., Dhanasekaran D. N., Ahn T., Song Y. S.. **Integrated analysis of ascites and plasma extracellular vesicles identifies a miRNA‐based diagnostic signature in ovarian cancer**. *Cancer Letters* (2022) **542**. DOI: 10.1016/j.canlet.2022.215735 60. Xin X., Zeng X., Gu H., Li M., Tan H., Jin Z., Hua T., Shi R., Wang H.. **CD147/EMMPRIN overexpression and prognosis in cancer: A systematic review and meta‐analysis**. *Scientific Reports* (2016) **6**. DOI: 10.1038/srep32804 61. Xiong L., Edwards C. K., Zhou L.. **The biological function and clinical utilization of CD147 in human diseases: A review of the current scientific literature**. *International Journal of Molecular Sciences* (2014) **15** 17411-17441. DOI: 10.3390/ijms151017411 62. Xu R., Rai A., Chen M., Suwakulsiri W., Greening D. W., Simpson R. J.. **Extracellular vesicles in cancer – Implications for future improvements in cancer care**. *Nature Reviews in Clinical Oncology* (2018) **15** 617-638. DOI: 10.1038/s41571-018-0036-9 63. Yekula A., Minciacchi V. R., Morello M., Shao H., Park Y., Zhang X., Muralidharan K., Freeman M. R., Weissleder R., Lee H., Carter B., Breakefield X. O., Di Vizio D., Balaj L.. **Large and small extracellular vesicles released by glioma cells in vitro and in vivo**. *Journal of Extracellular Vesicles* (2019) **9**. DOI: 10.1080/20013078.2019.1689784 64. Yoshioka Y., Kosaka N., Konishi Y., Ohta H., Okamoto H., Sonoda H., Nonaka R., Yamamoto H., Ishii H., Mori M., Furuta K., Nakajima T., Hayashi H., Sugisaki H., Higashimoto H., Kato T., Takeshita F., Ochiya T.. **Ultra‐sensitive liquid biopsy of circulating extracellular vesicles using ExoScreen**. *Nature Communications* (2014) **5** 3591. DOI: 10.1038/ncomms4591 65. Zhang Q., Jeppesen D. K., Higginbotham J. N., Graves‐Deal R., Trinh V. Q., Ramirez M. A., Sohn Y., Neininger A. C., Taneja N., McKinley E. T., Niitsu H., Cao Z., Evans R., Glass S. E., Ray K. C., Fissell W. H., Hill S., Rose K. L., Huh W. J., Coffey R. J.. **Supermeres are functional extracellular nanoparticles replete with disease biomarkers and therapeutic targets**. *Nature Cell Biology* (2021) **23** 1240-1254. DOI: 10.1038/s41556-021-00805-8 66. Zhao A., Li G., Péoc'h M., Genin C., Gigante M.. **Serum miR‐210 as a novel biomarker for molecular diagnosis of clear cell renal cell carcinoma**. *Experimental and Molecular Pathology* (2013) **94** 115-120. DOI: 10.1016/j.yexmp.2012.10.005 67. Zhao J., Lu C., Li Y., Ma Y., Deng J., Li L., Sun J.. **Thermophoretic detection of exosomal microRNAs by nanoflares**. *Journal of the American Chemical Society* (2020) **142** 4996-5001. DOI: 10.1021/jacs.9b13960 68. Zhou W., Fong M. Y., Min Y., Somlo G., Liu L., Palomares M. R., Yu Y., Chow A., O'Connor S. T. F., Chin A. R., Yen Y., Wang Y., Marcusson E. G., Chu P., Wu J., Wu X., Li A. X., Li Z., Gao H., Wang S. E.. **Cancer‐secreted miR‐105 destroys vascular endothelial barriers to promote metastasis**. *Cancer Cell* (2014) **25** 501-515. DOI: 10.1016/j.ccr.2014.03.007
--- title: 'Development of a risk prediction score and equation for chronic kidney disease: a retrospective cohort study' authors: - Shin Kawasoe - Takuro Kubozono - Anwar Ahmed Salim - Haruhito Yoshimine - Seiichi Mawatari - Satoko Ojima - Takeko Kawabata - Yoshiyuki Ikeda - Hironori Miyahara - Koichi Tokushige - Akio Ido - Mitsuru Ohishi journal: Scientific Reports year: 2023 pmcid: PMC10042816 doi: 10.1038/s41598-023-32279-z license: CC BY 4.0 --- # Development of a risk prediction score and equation for chronic kidney disease: a retrospective cohort study ## Abstract Chronic kidney disease (CKD) is a risk factor for end-stage renal disease and contributes to increased risk of cardiovascular disease morbidity and mortality. We aimed to develop a risk prediction score and equation for future CKD using health checkup data. This study included 58,423 Japanese participants aged 30–69 years, who were randomly assigned to derivation and validation cohorts at a ratio of 2:1. The predictors were anthropometric indices, life style, and blood sampling data. In derivation cohort, we performed multivariable logistic regression analysis and obtained the standardized beta coefficient of each factor that was significantly associated with new-onset CKD and assigned scores to each factor. We created a score and an equation to predict CKD after 5 years and applied them to validation cohort to assess their reproducibility. The risk score ranged 0–16, consisting of age, sex, hypertension, dyslipidemia, diabetes, hyperuricemia, and estimated glomerular filtration rate (eGFR), with area under the curve (AUC) of 0.78 for the derivation cohort and 0.79 for the validation cohort. The CKD incidence gradually and constantly increased as the score increased from ≤ 6 to ≥ 14. The equation consisted of the seven indices described above, with AUC of 0.88 for the derivation cohort and 0.89 for the validation cohort. We developed a risk score and equation to predict CKD incidence after 5 years in Japanese population under 70 years of age. These models had reasonably high predictivity, and their reproducibility was confirmed through internal validation. ## Introduction Progressive kidney dysfunction leads to end-stage kidney disease (ESKD), requiring dialysis or transplantation. The number of patients requiring dialysis in Japan continues to increase, exceeding 330,0001. Patients with chronic kidney disease (CKD) have an increased relative risk of coronary heart disease, heart failure, and stroke compared to those without CKD2–5. ESKD and cardiovascular diseases secondary to the renal impairment have become important medical problems that lead to a decline in the quality of life and increased national health care costs. A changing lifestyle and an aging population have dramatically altered each pathogenetic mechanism contributing to renal impairment. According to data from the Japanese population, diabetic nephropathy was the most common primary disease among patients requiring dialysis ($39.0\%$), followed by chronic glomerulonephritis ($27.8\%$), and nephrosclerosis ($10.3\%$)1. The incidence of diabetic nephropathy and nephrosclerosis is increasing annually, while the incidence of chronic glomerulonephritis is declining1. Lifestyle and the coexistence of other lifestyle-related diseases play a major role in renal dysfunction. It is very important to focus on risk factors to identify patients who may develop renal dysfunction in the future, because improving renal function is difficult after deterioration. Prophylactic intervention in a high-risk population with renal dysfunction may reduce progression and ESKD and cardiovascular mortality6. In developed countries, CKD is generally associated with old age, diabetes, hypertension, obesity, and cardiovascular disease7. Some of these can be improved by promoting lifestyle and behavioral changes. It has been reported that smoking cessation inhibited the progression of CKD. In diabetic patients with normal renal function and albuminuria, smoking cessation has been reported to decrease albuminuria8. Metabolic syndrome control were independently associated with a lesser progression of diabetic nephropathy9. In Japan, since 2008, people aged ≥ 40 years are recommended to undergo regular health checkups to prevent lifestyle-related diseases. It would be clinically useful to find a way to predict CKD using variables measured during these checkups. Several studies have developed predictive scores for CKD, but they differ in the target population’s ethnicity and score complexity, determined by the number and classification of factors10,11. Since the prevalence of renal dysfunction and mechanisms associated with its development vary by race, a unique Japanese scoring system is needed to predict CKD in the Japanese population. Although Nelson’s model included four cohorts of Japanese, they represented only $1.6\%$ of the total participants11. Furthermore, although a complex scoring system may provide accurate prediction, it is inconvenient in daily clinical practice. Therefore, we created a simple risk score and an equation for incident CKD after 5 years using Japanese large-scale health checkup data, and evaluated its internal validity. ## Data and study population This study was conducted in the same population as the participants for whom we previously created a risk model for developing hypertension12. Data were collected from participants aged 30–69 years after annual health checkups at Kagoshima Kouseiren Hospital between April 2008 and March 2016. We selected participants whose data were available at baseline and after 5 years (range, 3–7 years). Participants were excluded if they had CKD, defined as an estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73 m2; were undergoing dialysis therapy; were post-renal transplantation at baseline; and those with missing data. Two-thirds of the participants were randomly assigned to the derivation cohort to generate a score and an equation to predict CKD. The remaining one-third was used as a validation cohort to assess the validity of the score and equation obtained in the derivation cohort. This study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Ethics Committees of the Graduate School of Medical and Dental Sciences, Kagoshima University. The Ethics Committee has approved that since only existing anonymized data were used in this study, it is not necessary to obtain the informed consent of each individual. ## Risk factors Age was categorized into four groups: 30–39, 40–49, 50–59, and 60–69 years. Height and weight were measured using standard anthropometric methods. Body mass index (BMI) was calculated as weight (kg) divided by height squared (m2) and categorized into ≤ 24.9 kg/m2 and ≥ 25.0 kg/m2 groups. The following categories were determined by a self-administered questionnaire: smoker (currently smoker) or non-smoker (have never smoked or smoked in the past), non-frequent drinker (drinking less than 10 days per month) or frequent drinker (drinking more than 10 days per month), habit exerciser (30 min or more a day). Data on drugs for hypertension, diabetes, dyslipidemia, and hyperuricemia were collected using a self-administered questionnaire. Blood samples were obtained after an overnight fast. Serum lipids, glucose, uric acid, and creatinine levels were measured using standard laboratory procedures. Underlying diseases were defined as follows: hypertension (under treatment with antihypertensive agents or blood pressure ≥ $\frac{140}{90}$ mmHg), diabetes (under treatment with oral hypoglycemic agents or insulin, or fasting blood glucose ≥ 126 mg/dL), dyslipidemia (under treatment with lipid-lowering agents, serum triglycerides ≥ 150 mg/dL, serum low-density lipoprotein-cholesterol ≥ 140 mg/dL, or serum high-density lipoprotein-cholesterol < 40 mg/dL), and hyperuricemia (under treatment with uric acid lowering agents or serum uric acid level > 7.0 mg/dL). The eGFR was determined according to the new Japanese coefficient for the modified isotope dilution mass spectrometry-traceable Modification of Diet in Renal Disease study equation:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{eGFR}} = {194} \times {\text{SCr}}^{{ - {1}.0{94}}} \times {\text{Age}}^{{ - 0.{287}}}. $$\end{document}eGFR=194×SCr-1.094×Age-0.287. For women, eGFR was multiplied by a correction factor of 0.73913. The baseline renal function was categorized into four groups by baseline eGFR: 60.0–69.9, 70.0–79.9, 80.0–89.9, and ≥ 90.0 mL/min/1.73 m2. The outcome was CKD at the 5-year follow-up, defined as eGFR < 60.0 mL/min/1.73 m2. ## Statistical analysis As in our previous risk model for developing hypertension, we used a similar analytical method12. Continuous variables (age, BMI, blood pressure, and hematological parameters) are expressed as mean ± standard deviation, except for skewed distributed indices including triglyceride and blood glucose levels, which are expressed as median (1st quartile, 3rd quartile). Categorical variables, including underlying diseases and lifestyle variables, are expressed as proportions (percentages). Differences between the derivation and validation cohorts for normally distributed continuous variables, skewed-distribution continuous variables, and categorical variables were analyzed using the Student’s unpaired t-test, Wilcoxon test, and χ2 test, respectively. Univariable and multivariable logistic regression analyses were performed for each variable to estimate the odds ratio and $95\%$ confidence interval for CKD incidence. To create a risk score that predicts 5-year incidence of CKD, the following scores corresponding to standardized beta coefficients were assigned to each risk factor category for items that were significant in the multivariable logistic regression analysis, based on the methodology used in the Japan Epidemiology Collaboration on Occupational Health Study Group’s study: 1, β = 0.01–0.20; 2, β = 0.21–0.80; 3, β = 0.81–1.20; 4, β = 1.21–2.20; and 5, β > 2.2014–16. The reference category for each variable was given a point of 0, and the risk score for developing CKD was calculated as the sum of the individual points. The discriminative performance of the score was assessed using the area under the curve (AUC) from the receiver operating characteristic (ROC) analysis. The performance of each score was assessed using sensitivity, specificity, positive predictive value, negative predictive value, and the Youden index16. The consistency of association between the scores and CKD incidence was evaluated using the Cochran–Armitage trend test. Then, the score was applied to the validation cohort and ROC analysis was performed. To compensate for the weakness of the study being derived from data from a single institution, a sensitivity analysis was performed using the bootstrap resample method. The $95\%$ bootstrap confidence interval of the odds ratios in multivariable logistic regression analysis were calculated based on 2500 bootstrap resamples. We then used the β coefficients for the risk factors that were significant in the logistic regression analysis to create an equation that directly calculates the proportion of CKD after 5 years, and assessed its predictive ability and reproducibility in the same way as for the score. In developing the equation, continuous variables such as age and eGFR were used as continuous values without categorization. We evaluated the calibration using calibration plots. All statistical analyses were performed using JMP Pro version 14 (SAS Institute, Cary, NC, USA) for Windows. Statistical significance was set at $P \leq 0.05.$ ## Baseline characteristics Overall, 167,706 participants aged 30–69 years underwent medical examinations at least once during the study period. Among these, 71,002 individuals had available 5-year follow-up data. A total of 12,579 individuals were excluded (baseline CKD, 7,871; missing variables, 4,708). Finally, the data of 58,423 participants (age, 53.8 ± 10.2 years; male, $50.0\%$) was analyzed. Table 1 shows the baseline characteristics of the derivation and validation cohorts. The age and proportion of men were 53.8 ± 10.2 years and $49.8\%$ in the derivation cohort and 53.8 ± 10.2 years and $50.1\%$ in the validation cohort, respectively. After the follow-up period (median, 5.0 years; 1st quartile, 4.7 years; 3rd quartile, 5.1 years), we identified 2,679 ($6.9\%$) and 1,319 ($6.8\%$) cases of new-onset CKD in the derivation and validation cohorts, respectively. Table 1Baseline characteristics of study population in the derivation and validation cohorts. Derivation cohortValidation cohortN = 38,948N = 19,475Age, years53.8 ± 10.253.8 ± 10.2Men, %49.850.1BMI, kg/m223.3 ± 3.423.3 ± 3.4SBP, mmHg123.8 ± 18.4124.0 ± 18.1DBP, mmHg76.6 ± 11.376.6 ± 11.2Hypertension, %32.532.5Diabetes, %8.58.3Dyslipidemia, %45.945.4Hyperuricemia, %11.211.2Current smoking, %20.921Frequent drinking, %13.213.1Creatinine, mg/dL0.71 ± 0.140.71 ± 0.14eGFR79.8 ± 12.479.7 ± 12.5Uric acid, mg/dL5.1 ± 1.45.1 ± 1.4Triglyceride, mg/dL89 [63, 129]89 [64, 130]LDL-C, mg/dL122.8 ± 31.2122.6 ± 31.0HDL-C, mg/dL60.4 ± 14.960.4 ± 14.8BG, mg/dL95 [89, 103]95 [89, 103]Baseline characteristics of study population in the derivation and validation cohorts. Continuous variables are expressed as mean ± standard deviation, except for triglyceride and blood glucose levels, which are expressed as median [1st quartile, 3rd quartile]. Categorical variables, including cardiovascular risk factors and lifestyle variables, are expressed as number of subjects and proportions (percentages).BMI body mass index, SBP systolic blood pressure, DBP diastolic blood pressure, eGFR estimated glomerular filtration rate, LDL-C low-density lipoprotein cholesterol, HDL-C high-density lipoprotein cholesterol, BG blood pressure. ## Association between risk factors and incident CKD The association between incident CKD and risk factor candidates is shown in Table 2. In the univariable model, older age, male sex, higher BMI, non-current smoking, non-frequent alcohol drinking, hypertension, diabetes, dyslipidemia, hyperuricemia, and lower eGFR were associated with an increased risk of CKD. In the multivariable model, older age, male sex, non-frequent drinking, hypertension, diabetes, dyslipidemia, hyperuricemia, and lower eGFR were significantly associated with an increased CKD risk. Our risk models included age, gender, hypertension, diabetes, dyslipidemia, hyperuricemia, and eGFR. Considering the lack of sufficient information on alcohol consumption, it was not included in the risk model components. Table 2Odds ratios and $95\%$ confidence interval of 5-year incidence of CKD for each risk factor. Risk factorsNo. of subjectsNo. of cases (%)UnivariableMultivariableOR ($95\%$ CI)OR ($95\%$ CI)Age, yars30–39453264 (1.4)RefRef40–498405280 (3.3)2.41 (1.83–3.16)1.44 (1.08–1.90)50–5911,805736 (6.2)4.64 (3.59–6.01)2.03 (1.55–2.65)60–6914,2061599 (11.3)8.85 (6.88–11.39)3.09 (2.37–4.03)SexMen19,4031433 (7.4)1.17 (1.08–1.27)1.19 (1.08–1.32)Women19,5451246 (6.4)RefRefBMI, kg/m2 < 24.928,3151843 (6.5)RefRef > 25.010,633836 (7.9)1.23 (1.13–1.33)1.04 (0.95–1.15)Current smokingNo30,8092248 (7.3)1.41 (1.27–1.56)RefYes8139431 (5.3)Ref1.05 (0.93–1.19)Frequent drinkingNo33,8152391 (7.1)1.28 (1.13–1.45)1.19 (1.03–1.36)Yes5133288 (5.6)RefRefHypertensionNo26,2911371 (5.2)RefRefYes12,6571308 (10.3)2.09 (1.94–2.27)1.45 (1.33–1.58)DiabetesNo35,6142358 (6.6)RefRefYes3334321 (9.6)1.50 (1.33–1.70)1.30 (1.14–1.49)DyslipidemiaNo21,0841204 (5.7)RefRefYes17,8641475 (8.3)1.49 (1.37–1.61)1.12 (1.03–1.22)HyperuricemiaNo34,5862225 (6.4)RefRefYes4362454 (10.4)1.69 (1.52–1.88)1.30 (1.15–1.47)eGFR > 90.0720328 (0.4)RefRef75.0–89.916,298211 (1.3)3.36 (2.26–4.99)2.94 (1.98–4.37)60.0–74.915,4472440 (15.8)48.1 (33.1–69.8)36.8 (25.3–53.6)The odds ratios and $95\%$ confidence intervals of 5-year incidence of CKD for each risk factors using logistic regression analysis. In multivariable model, the odds ratios were adjusted for the following variables: age categories, sex, body mass index categories, current smoking, frequent drinking, hypertension, diabetes, dyslipidemia, hyperuricemia, and eGFR categories. BMI body mass index, eGFR estimated glomerular filtration ratio, OR odds ratio, CI confidence interval. ## Risk prediction score for 5-year incidence of CKD The risk factors and points derived for each category are shown in Table 3. The risk score, sum of all points, could range from 0 to 16. From the ROC curve predicting CKD incidence, the AUC was 0.78 (Fig. 1a). Table 4 shows the predictive performance for a range of cutoff points. A score of ≥ 8 showed the highest Youden index in the derivation cohort, with a sensitivity of 0.90 and specificity of 0.52. The observed incidence of CKD for each score is shown in Fig. 2. At a score of 0 to 6, the observed risk of 5-year CKD was less than $1\%$. As the score increased, the risk gradually increased ($P \leq 0.001$). At a score of 14 to 16, more than $25\%$ of the participants had developed CKD after 5 years. Table 3Points assigned to predict 5-year incidence of CKD.Risk factorsβPointAge, years30–39040–490.36250–590.71260–691.133SexWomen0Men0.181HypertensionNo0Yes0.372DiabetesNo0Yes0.272DyslipidemiaNo0Yes0.111HyperuricemiaNo0Yes0.262eGFR, > 90.0075.0–89.91.08360.0–74.93.615We assigned each category of risk factor with one of the following point scores, corresponding to the standardized β coefficients of multivariate logistic regression,: 1, β = 0.01–0.20; 2, β = 0.21–0.80; 3, β = 0.81–1.20; 4, β = 1.21–2.20; and 5, β > 2.20. The reference category for each variable was given a score of 0. β, standardized regression coefficient. Figure 1ROC curves in the derivation and validation cohorts for the risk score for predicting the 5-year incidence of CKD. Receiver operating characteristic curves of the derivation cohort (a) and validation cohort (b) for the risk score to predict the 5-year incidence of CKD. AUC area under the curve, CKD chronic kidney disease, ROC receiver operating characteristic. Table 4Predictive performance of the developed 5-year CKD risk score. Derivation cohortRisk scoresSensitivitySpecificityPPVNPVYouden index > 01.000.000.071.000.00 > 11.000.020.071.000.02 > 21.000.030.071.000.03 > 31.000.070.071.000.07 > 41.000.130.081.000.12 > 51.000.170.081.000.17 > 60.990.270.090.990.25 > 70.970.390.110.990.36 > 80.900.520.120.970.43 > 90.760.660.140.960.42 > 100.560.790.170.950.35 > 110.410.870.180.940.28 > 120.210.940.220.940.15 > 130.120.970.230.930.09 > 140.050.990.270.930.04 > 150.011.000.270.930.01 > 160.011.000.350.930.01Validation cohortRisk scoresSensitivitySpecificityPPVNPVYouden index > 01.000.000.071.000.00 > 11.000.020.071.000.02 > 21.000.030.071.000.03 > 31.000.070.071.000.07 > 41.000.120.081.000.12 > 50.990.170.081.000.17 > 60.990.270.090.990.26 > 70.970.390.100.980.36 > 80.890.530.120.970.42 > 90.750.660.140.960.41 > 100.560.790.170.950.35 > 110.420.860.190.940.29 > 120.220.940.220.940.16 > 130.130.970.230.930.10 > 140.050.990.270.930.04 > 150.011.000.320.930.01 > 160.011.000.410.930.00Predictive performance of the developed 5-year CKD risk score in derivation and validation. PPV positive predictive value, NPV negative predictive value. Figure 2Proportions of developing CKD after 5 years at each score in the derivation and validation cohort. Proportion of CKD development after 5 years for each score in the derivation and validation cohort. The lined bars represent the proportion of 5-year incidence of CKD in the derivation cohort, and the black-shaded bars represent the proportion of 5-year incidence of CKD in the validation cohort. CKD chronic kidney disease. When the risk score was applied to the validation cohort, the AUC was 0.79, similar to the derivation cohort (Fig. 1b). At the cutoff point of 8 ≥ points, the score had a sensitivity of 0.89 and specificity of 0.53, which was similar to the derivation cohort (Table 4). As the score increased, the risk gradually increased ($P \leq 0.001$, Fig. 2). The results of the sensitivity analysis using the bootstrap method are shown in Table 5. The odds ratios and $95\%$ confidence intervals from the bootstrap method were similar to those from the derivation cohort in the main analysis. Table 5Mean odds ratios and $95\%$ confidence intervals of multivariable logistic regressin analysis from bootstrap resample method. Risk factorsOR ($95\%$ CI)Age, years30–39Ref40–491.49 (1.20–1.91)50–592.08 (1.68–2.63)60–693.16 (2.57–4.04)SexMen1.19 (1.08–1.32)WomenRefBMI, kg/m2 < 24.9Ref > 25.01.03 (0.96–1.12)Current smokingNoRefYes1.02 (0.93–1.12)Frequent drinkingNo1.18 (1.04–1.31)YesRefHypertensionNoRefYes1.38 (1.28–1.48)DiabetesNoRefYes1.38 (1.23–1.53)DyslipidemiaNoRefYes1.11 (1.03–1.20)HyperuricemiaNoRefYes1.32 (1.20–1.46)eGFR > 90.0Ref75.0–89.92.76 (2.06–3.90)60.0–74.937.6 (28.5–52.3)The odds ratios and $95\%$ confidence intervals of 5-year incidence of CKD for each risk factors using logistic regression analysis with 2500 bootstrap resampling. In multivariable model, the odds ratios were adjusted for the following variables: age categories, sex, body mass index categories, current smoking, frequent drinking, hypertension, diabetes, dyslipidemia, hyperuricemia, and eGFR categories. BMI body mass index, OR odds ratio, CI confidence interval. ## Risk prediction equation for 5-year incidence of CKD We developed an equation to predict CKD probability after 5 years, using age (years old), sex (female, 0; male, 1), hypertension (no, 0; yes, 1), dyslipidemia (no, 0; yes, 1), diabetes (no, 0; yes, 1), hyperuricemia (no, 0; yes, 1), and eGFR (mL/min/1.73 m2). ## Probability of 5-year incidence of CKD Probability = 1/ (1 + exp[−{9.4876 + 0.0311 × age + 0.2400 × sex + 0.3470 × hypertension + 0.0893 × dyslipidemia + 0.3444 × diabetes + 0.0832 × hyperuricemia + -0.1980 × eGFR}]). The median probability obtained from the derivation cohort was 0.018 (interquartile range 0.002–0.084), and the AUC value of the ROC curve for the development of CKD after 5 years was 0.88, with a sensitivity of 0.84 and a specificity of 0.78 at a cutoff value of 0.077 calculated from the Youden index (Fig. 3a). When the risk equation was applied to the validation cohort, the AUC was 0.89, with a sensitivity of 0.87 and a specificity of 0.77 with a cutoff value of 0.075 (Fig. 3b). The results of the calibration are shown in Fig. 4. Good visual calibration is achieved for both the derivation and validation cohorts. Figure 3ROC curves in the derivation and validation cohorts for the risk equation for predicting the 5-year incidence of CKD. Receiver operating characteristic curves of the derivation cohort (a) and validation cohort (b) for the risk equation to predict the 5-year incidence of CKD. AUC area under the curve, CKD chronic kidney disease, ROC receiver operating characteristic. Figure 4Calibration plots for the equation model in derivation and validation cohorts. The visual agreement between the CKD predictions (Predicted probability) and observations (Actual probability) for the equation model in the derivation and validation cohorts. ( a) derivation cohort (b) validation cohort. ## Discussion We developed a score and an equation and to predict the 5-year risk of incident CKD defined as eGFR < 60 mL/min/1.73 m2 based on seven indicators (age, sex, hypertension, diabetes, dyslipidemia, hyperuricemia, and eGFR) from large-scale health checkup data. CKD incidence increased with increasing risk scores, and the same predictive ability was validated when applied to population not included in the score development. We have achieved our goal of creating risk models based on simple indexes that is applicable to the Japanese population under 70 years of age and can be easily used in clinical practice. Several studies have reported risk scores for predicting future renal dysfunction. O'Seaghdha et al. studied the Framingham Offspring Study cohort of Caucasians and developed a score (0–15 points) to predict CKD incidence (eGFR < 60 mL/min/1.73 m2) at 10 years based on age, hypertension, diabetes, baseline eGFR category, and the presence of proteinuria using a test paper method. The c-statistic for the CKD risk score was 0.74 when validated using an external cohort of whites and blacks, and the predictive ability was good even when for only blacks10. Predictive equations for CKD using data from over 5 million people in 34 cohort studies from 28 countries have also been reported. In addition to age and sex, race/ethnicity, eGFR, history of cardiovascular disease and hypertension, smoking history, BMI, and urinary albumin are used for calculation. For diabetic patients, glycated hemoglobin and the use of diabetic medications were added. The C statistic was 0.845 in the absence of diabetes and 0.801 in the presence of diabetes11. We should apply these prior risk models to our study population and compare the predictive abilities. However, these two risk models could not be applied to our study population because they require urinary protein and albumin information. The AUC and c-statistic of the ROC curve are commonly used as indicators of the discriminative ability of risk scores. The model developed in this study had an AUC of 0.79 for the score and 0.89 for the equation. Although direct comparison with previous reports was not possible, the discriminative ability of our model is somewhat superior. As described above, several risk prediction models have been reported to predict the development of CKD in various ethnic groups in different countries. Since CKD prevalence and its risk factors varies by race, it is preferable to use the prediction model determined from *Japanese data* to predict CKD risk in Japanese people. Although Nelson et al. included four cohorts of Japanese, they represented only $1.6\%$ of the total participants and included data as old as the 1970s, whereas the present study included only Japanese and used relatively recent data from 2008 to 2016 to develop the score and equation. These factors may have contributed to the improvement in discrimination. In Japan, all men and women aged ≥ 40 years are recommended to undergo annual specific health checkups and receive guidance focusing on visceral obesity if needed. We aimed to create a prediction model based on anthropometric measurements, blood tests, questionnaires, and other data obtained from these checkups. We achieved a good prediction accuracy of AUC 0.78 for the risk score and 0.88 for the risk equation. The risk score obtained in this study can be easily calculated by healthcare workers in daily clinical practice, enabling patients to understand the extent of their risk and contribute to decision-making for treatment and lifestyle improvement. The risk prediction equation has the advantage of more accurately predicting CKD risk, but they are not easy to calculate and should be considered for practical application in the form of online calculators or others. In addition, development of an application that can calculate this risk score would be more convenient and clinically useful for assessing risk. Age-associated loss of kidney function has been recognized for decades. With aging, many participants exhibit a progressive decrease in the glomerular filtration rate and renal blood flow. The decrease in glomerular filtration rate is due to a reduction in the glomerular capillary plasma flow rate and glomerular capillary ultrafiltration coefficient17. The proportion of participants over 70 years who had a medical checkup after 5 years was low. In addition, $40\%$ of the participants had already developed CKD, making the number of cases available for analysis small. Therefore, participants over 70 years were excluded from this study. Previous epidemiologic studies indicate that the incidence of ESKD is higher in men than in women18. A recent study found that female mice were more tolerant to ischemic-reperfusion injury than male mice and that female mice receiving supplemental estrogen before ischemia were further protected19. The possible mechanisms underlying the reno-protective role in females seem to be related to estrogen. This is supported by clinical studies demonstrating that premenopausal women are better protected from renal and cardiovascular disease compared to age-matched men; this protective effect seems to be lost with aging and menopause20. However, the mechanisms by which estrogen confers protective renal effects are not well understood. Hypertension has been reported to be a risk factor for both CKD and ESKD21,22. Systemic hypertension causes an increase in intraglomerular capillary pressure, leading to glomerulosclerosis and loss of renal function22. CKD is associated with specific qualitative and quantitative lipid abnormalities, resulting in specific dyslipidemias23–25. Specific abnormalities in lipoprotein metabolism, caused by the inappropriate activity of some key enzymes and metabolic pathways, develop in the early stages of renal failure and result in dyslipidemia, which is a risk factor for the development of atherosclerosis. As CKD progresses, dyslipidemia worsens and may adversely affect renal function through the promotion of atherosclerosis23,26,27. Iseki et al. found a significant association between elevated serum uric acid and serum creatinine levels in a Japanese cohort study28. Further, high uric acid level (≥ 6.0 mg/dL) was a risk factor for ESKD development in women29. It is suggested that hyperuricemia may damage vascular endothelial cells and contribute to renal dysfunction through the production of reactive oxygen species by xanthin oxide reductase, activation of the renin-angiotensin system, and induction of inflammation by inflammasome activation30,31. Although alcohol consumption was a significant factor in the multivariate analysis, we did not include it as a risk factor in this study because we do not have sufficient information on it. In the questionnaires, the frequency per month and daily intake regarding alcohol consumption were asked as separate items. Therefore, we were unable to find an appropriate way to integrate frequency and quantity of intake into a single variable, "alcohol intake“. The association between alcohol intake and CKD is not simple. Moderate alcohol intake (20–40 g ethanol/day) is not a risk factor for CKD32, but rather inhibits its progression33. On the other hand, heavy alcohol intake (> 60 g/day of ethanol) has been reported to be a risk factor for CKD34. For these reasons, we decided that it would be difficult to use the data with alcohol intake not clearly defined. In addition to the above, this study had several limitations. First, the data were not collected prospectively, and the results should be further validated in a prospective observational study. Second, the participants were limited to those who underwent health examinations at a single institution in Japan. Furthermore, participants tend to be highly interested in their own health, so selection bias cannot be avoided. Third, since we do not have information on the etiology of nephropathy, we do not know if these models work better in predicting nephrosclerosis or diabetic nephropathy compared to glomerulonephritis. Fourth, although proteinuria is itself an important risk factor for the development of CKD, urinary protein was not addressed in this study because of the low rate of obtaining urinalysis results in our participants' data. Finally, due to the large population that was randomly divided into derivation and the validation cohort, both cohorts had nearly identical characteristics. The internal validation therefore has limited informative value and external validation is necessary to assess the discriminative and predictive abilities of scores in real life. In conclusion, we have developed a clinical risk score and equation to predict the occurrence of CKD after 5 years in a Japanese population under 70 years of age, using age, sex, hypertension, diabetes, dyslipidemia, hyperuricemia, and eGFR levels. Predictive ability was comparable with previous reports, and reproducibility was confirmed through internal validation. These prediction methods could be simple and useful tools for detecting and recognizing high-risk CKD patients in Japan and for providing appropriate counseling. ## References 1. Nitta K. **Annual dialysis data report 2017, JSDT Renal Data Registry**. *Ren. Replace. Ther.* (2019) **5** 53. DOI: 10.1186/s41100-019-0248-1 2. Meisinger C, Döring A, Löwel H. **Chronic kidney disease and risk of incident myocardial infarction and all-cause and cardiovascular disease mortality in middle-aged men and women from the general population**. *Eur. Heart J.* (2006) **27** 1245-1250. DOI: 10.1093/eurheartj/ehi880 3. Muntner P. **Incidence of stroke symptoms among adults with chronic kidney disease: Results from the REasons for Geographic and Racial Differences in Stroke (REGARDS) study**. *Nephrol. Dial. Transplant.* (2012) **27** 166-173. DOI: 10.1093/ndt/gfr218 4. Hemmelgarn BR. **Relation between kidney function, proteinuria, and adverse outcomes**. *JAMA* (2010) **303** 423-429. DOI: 10.1001/jama.2010.39 5. Kottgen A. **Reduced kidney function as a risk factor for incident heart failure: The atherosclerosis risk in communities (ARIC) study**. *J. Am. Soc. Nephrol.* (2007) **18** 1307-1315. DOI: 10.1681/ASN.2006101159 6. James MT, Hemmelgarn BR, Tonelli M. **Early recognition and prevention of chronic kidney disease**. *Lancet* (2010) **375** 1296-1309. DOI: 10.1016/S0140-6736(09)62004-3 7. Levey AS, Coresh J. **Chronic kidney disease**. *Lancet* (2012) **379** 165-180. DOI: 10.1016/S0140-6736(11)60178-5 8. Chase HP. **Cigarette smoking increases the risk of albuminuria among subjects with type I diabetes**. *JAMA* (1991) **265** 614-617. DOI: 10.1001/jama.1991.03460050068022 9. Duran-Perez EG. **Treatment of metabolic syndrome slows progression of diabetic nephropathy**. *Metab. Syndr. Relat. Disord.* (2011) **9** 483-489. DOI: 10.1089/met.2011.0056 10. O'Seaghdha CM. **A risk score for chronic kidney disease in the general population**. *Am. J. Med.* (2012) **125** 270-277. DOI: 10.1016/j.amjmed.2011.09.009 11. Nelson RG. **Development of risk prediction equations for incident chronic kidney disease**. *JAMA* (2019) **322** 2104-2114. DOI: 10.1001/jama.2019.17379 12. Kawasoe M. **Development of a risk prediction score for hypertension incidence using Japanese health checkup data**. *Hypertens. Res.* (2022) **45** 730-740. DOI: 10.1038/s41440-021-00831-x 13. Matsuo S. **Revised equations for estimated GFR from serum creatinine in Japan**. *Am. J. Kidney Dis.* (2009) **53** 982-992. DOI: 10.1053/j.ajkd.2008.12.034 14. Guasch-Ferré M. **A risk score to predict type 2 diabetes mellitus in an elderly Spanish Mediterranean population at high cardiovascular risk**. *PLoS One* (2012) **7** e33437. DOI: 10.1371/journal.pone.0033437 15. Lindström J, Tuomilehto J. **The diabetes risk score: A practical tool to predict type 2 diabetes risk**. *Diabetes Care* (2003) **26** 725-731. DOI: 10.2337/diacare.26.3.725 16. Nanri A. **Development of risk score for predicting 3-year incidence of type 2 diabetes: Japan epidemiology collaboration on occupational health study**. *PLoS One* (2015) **10** e0142779. DOI: 10.1371/journal.pone.0142779 17. Weinstein JR, Anderson S. **The aging kidney: Physiological changes**. *Adv. Chronic Kidney Dis.* (2010) **17** 302-307. DOI: 10.1053/j.ackd.2010.05.002 18. Ricardo AC. **Sex-related disparities in CKD progression**. *J. Am. Soc. Nephrol.* (2019) **30** 137-146. DOI: 10.1681/ASN.2018030296 19. Aufhauser DD. **Improved renal ischemia tolerance in females’ influences kidney transplantation outcomes**. *J. Clin. Invest.* (2016) **126** 1968-1977. DOI: 10.1172/JCI84712 20. Coggins CH. **Differences between women and men with chronic renal disease**. *Nephrol. Dial. Transplant.* (1998) **13** 1430-1437. DOI: 10.1093/ndt/13.6.1430 21. Hamrahian SM, Falkner B. **Hypertension in chronic kidney disease**. *Adv. Exp. Med. Biol.* (2017) **956** 307-325. DOI: 10.1007/5584_2016_84 22. Lea JP, Nicholas SB. **Diabetes mellitus and hypertension: Key risk factors for kidney disease**. *J. Natl. Med. Assoc.* (2002) **94** 7S-15S. PMID: 12152917 23. Parmar JA, Joshi AG, Chakrabarti M. **Dyslipidemia and chronic kidney disease**. *ISRJ* (2014) **3** 396-397 24. Ahmed MH, Khalil AA. **Ezetimibe as a potential treatment for dyslipidemia associated with chronic renal failure and renal transplant**. *Saudi J. Kidney Dis. Transpl.* (2010) **21** 1021-1029. PMID: 21060168 25. Kwan BC, Kronenberg F, Beddhu S, Cheung AK. **Lipoprotein metabolism and lipid management in chronic kidney disease**. *J. Am. Soc. Nephrol.* (2007) **18** 1246-1261. DOI: 10.1681/ASN.2006091006 26. Balode AA, Khan ZH. **Serum lipid profile in chronic kidney disease patients on haemodialysis**. *IJAR* (2013) **3** 20-22. DOI: 10.15373/2249555X/AUG2013/7 27. Chen SC. **Association of dyslipidemia with renal outcomes in chronic kidney disease**. *PLoS One* (2013) **8** e55643. DOI: 10.1371/journal.pone.0055643 28. Iseki K. **Significance of hyperuricemia on the early detection of renal failure in a cohort of screened subjects**. *Hypertens. Res.* (2001) **24** 691-697. DOI: 10.1291/hypres.24.691 29. Iseki K. **Significance of hyperuricemia as a risk factor for developing ESRD in a screened cohort**. *Am. J. Kidney Dis.* (2004) **44** 642-650. DOI: 10.1016/S0272-6386(04)00934-5 30. Yu MA, Sánchez-Lozada LG, Johnson RJ, Kang DH. **Oxidative stress with an activation of the renin-angiotensin system in human vascular endothelial cells as a novel mechanism of uric acid-induced endothelial dysfunction**. *J. Hypertens.* (2010) **28** 1234-1242. DOI: 10.1097/HJH.0b013e328337da1d 31. Maejima I. **Autophagy sequesters damaged lysosomes to control lysosomal biogenesis and kidney injury**. *EMBO J.* (2013) **32** 2336-2347. DOI: 10.1038/emboj.2013.171 32. Yamagata K. **Risk factors for chronic kidney disease in a community-based population: A 10-year follow-up study**. *Kidney Int.* (2007) **71** 159-166. DOI: 10.1038/sj.ki.5002017 33. Shlipak MG. **Cardiovascular mortality risk in chronic kidney disease: Comparison of traditional and novel risk factors**. *JAMA* (2005) **293** 1737-1745. DOI: 10.1001/jama.293.14.1737 34. Shankar A, Klein R, Klein BE. **The association among smoking, heavy drinking, and chronic kidney disease**. *Am. J. Epidemiol.* (2006) **164** 263-271. DOI: 10.1093/aje/kwj173
--- title: Lifestyle modification and medication use among diabetes mellitus patients attending Jimma University Medical Center, Jimma zone, south west Ethiopia authors: - Aster Wakjira Garedow - Tsiyon Mekoya Jemaneh - Addisalem Gebresilase Hailemariam - Gorfineh Teshome Tesfaye journal: Scientific Reports year: 2023 pmcid: PMC10042818 doi: 10.1038/s41598-023-32145-y license: CC BY 4.0 --- # Lifestyle modification and medication use among diabetes mellitus patients attending Jimma University Medical Center, Jimma zone, south west Ethiopia ## Abstract Diabetes, a non-communicable metabolic disease, causes multiple complications and deaths worldwide. It is a complex, chronic disease that requires continuous medical care with multifactorial risk reduction strategies beyond glycemic control. Ongoing patient education and self-management support are critical for preventing acute complications and reducing the risk of long-term complications. There is ample evidence that healthy lifestyle choices, such as a healthy diet, moderate weight loss, and regular exercise, can maintain normal blood sugar levels and minimize diabetes-related complications. In addition, this lifestyle change has a major impact on controlling hyperglycemia and can help to maintain normal blood sugar levels. This study aimed to assess lifestyle modification and medication use in patients with diabetes mellitus at Jimma University Medical Center. Hospital-based prospective cross-sectional study was conducted from April 1 to September 30, 2021 among DM patients who have follow-up at diabetic clinic of Jimma University Medical Center. Consecutive sampling was used until the required sample size was achieved. Data were checked for completeness, then entered into Epidata version 4.2 software and exported to SPSS version 21.0. Pearson’s chi-square test was performed to determine the association between KAP and independent factors. Variables with a p value < 0.05 were considered significant. A total of 190 participants took part in this study with a response rate of $100\%$. In this study, 69 ($36.3\%$) participants had good knowledge, 82 ($43.2\%$) moderate knowledge and 39 ($20.5\%$) poor knowledge, 153 ($85.8\%$) had positive attitudes, 141 ($74.2\%$) had good practice. Marital status, *Occupational status* and educational status were significantly associated with knowledge and attitude towards LSM and medication use. Marital status was the only variable that remained significantly associated with knowledge, attitude and practice towards LSM and medication use. The result of this study showed that more than $20\%$ of the participants had poor knowledge, attitude, and practice towards medication use and LSM. Marital status was the only variable which remained to be significantly associated with KAP towards LSM and medication use. ## Introduction Diabetes mellitus (DM) is a group of metabolic disorders characterized by hyperglycemia *Diabetes mellitus* (DM) is a group of metabolic disorders characterized by hyperglycemia resulting from defects in insulin secretion, action or both”. Diabetes mellitus is classified into two types namely; type 1 (T1DM) and type 2 diabetes mellitus (T2DM)1. Diabetes is a complex, chronic illness that requires continuous medical care with multifactorial risk-reduction strategies beyond glycemic control. Ongoing patient self-management education and support are critical for preventing acute complications and reducing the risk of long-term complications2. The prevalence of DM is rapidly rising globally at a threatening rate. According to world health organization, 346 million individuals globally were diagnosed to have diabetes, among which $90\%$ of individuals have T2DM. This value is expected to rise to 380 million individuals by the year 20253, Globally; an estimated 422 million adults are living with DM, type 2DM makes up about 90–$95\%$ of all cases and the remaining 2–$5\%$ is type 14. Throughout the last 20 years, the incidence of diabetes has been raised intensively in many parts of the world5. More than twelve million people were estimated to be living with diabetes in Africa which is projected to increase to 23.9 million by 20306,7. The reasons for DM are urbanization, lack of physical activity, sedentary life style and obesity8. If not all but most of the factors are modifiable through proper lifestyle practices, drug taking behavior and screening for complications9. In Ethiopia, the prevalence of diabetes was $3.5\%$ in 2011, and the extrapolated prevalence in 2013 was $4.36\%$. It is also known that a large number of people remain undiagnosed, with an estimated number of undiagnosed cases reported to be 1.39 million people in 20136 Inadequate knowledge about the disease, prognosis, complications and treatment results in poor glycemic control which in turn leads to increase in morbidity10. This increases the need for proper education regarding alterations in lifestyle (exercise and food), medication adherence, regular screening in patients with DM11–13. The increase in the incidence of diabetes in developing countries follows the trend of urbanization and life style changes14. There was no attention given to diabetic education in Ethiopia, there were no diabetes nurse educators and diabetes dieticians in the country and those who provided health services for diabetes had no special training for diabetes care15. Control of the DM through a tight schedule of blood glucose and urine sugar monitoring, medication and adjustment to dietary condition need patient’s regular attention and discipline16,17. Proper metabolic conditions depend on several factors such as patient awareness on the pathophysiological aspect and those concerned with disease treatment, nutritional reduction, increased physical activity, regular foot inspection, signs and symptoms of hypoglycemia and prevention of chronic diseases, disease management in special situations and family support18. The present study assessed the knowledge, attitude and practice of DM patients regarding medication use and life style modification. ## Study setting and population The study was conducted at Jimma University Medical Center (JUMC) from April 1 to September 30, 2021 among DM patients who have follow-up in the diabetic clinic of JUMC which is the only referral hospital, serves 20 million catchment areas. Patients were eligible for inclusion if they were greater than 18 of age and were willing to provide informed consent and excluded if they were unconscious and not willing to give consent. The dependent variable was KAP and independent variables included patients-related factors: age, sex, BMI, educational status, residence, monthly income, marital status and occupation. ## Study design, sample size determination and sampling technique A hospital-based prospective cross sectional study was conducted. The sample size was calculated by using simple proportion formula and consecutive Sampling Technique was used until required sample size was obtained.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\text{n}} = \frac{{\left({z \propto /2} \right)2p\left({1 - p} \right)}}{d2}$$\end{document}n=z∝/22p1-pd2where: n—required sample size, Z—standard deviation normal value at $95\%$ CI which is 1.96, p—prevalence of KAP, $$p \leq 0.5$$, d—possible margin of error that can be tolerated $5\%$ (0.05), $q = 1$ − p. Thus; n = [1.96]2 * 0.5(0.5)/(0.05)2 = 384. The source population of DM on follow up during 2020 in JUMC = $$n = 3052$$ which were < 10,000, we can use correction formula to calculate final sample size.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\text{n}}_{f} = \frac{n}{1 + n - 1/N}$$\end{document}nf=n1+n-1/Nnf = = $\frac{384}{1}$ + 384 − $\frac{1}{3052}$ = 190 ## Data collection procedures Data were collected using a semi-structured questionnaire through interview and medical chart review. It includes socio-demographic, medication-related, and laboratory as well as clinical and diseases-related characteristics of the patient. Data was collected by four data collectors by following DM patients monthly and supervised after training was given on the detail of data collection tools/checklists for 2 days before data collection started. For KAP analysis towards DM the questionnaires have four major parts: socio-demographic information part, Health Seeking Behaviour, Health Status/occupational characteristics part and three domains (Knowledge, Attitude and practice). The knowledge aimed to assess causes, identification and management of DM. The item in the domain had two or four options (Yes, No and/or ‘Do Not Know’ and ‘Unsure’)). To assess the attitude and practice the questionnaires, we used items with yes or no options or Strongly agree, Agree, Not sure, Disagree, Strongly disagree. ## Data quality control and assurance The International Council on Harmonization Guidelines for Good Clinical Practice and the Declaration of Helsinki's guiding principles will both be followed during the study's execution. The following specific ethical concerns were taken into account during the procedure. These include the fact that participation is voluntary, getting written agreement, after a necessary description of the goal, benefit, and risk of the study, as well as the subject's right to decide whether or not to participate. preserving privacy by omitting participant names in favor of a code. An independent party developed the surveys in English, had them translated into Amharic and the native language Afan Oromo, and had those translations returned to English. In ensure uniformity, the surveys were written in English, translated into Amharic and the regional language Afan Oromo, and then back into English. Prior to the actual data collection, a pre-test was carried out on $5\%$ of the study participants by randomly chosen patients to assess the validity and consistency of the structured data collection format. Data were compiled, cleaned up, programmed, and consistency-checked. The supervisor attentively observed each step of data collecting and recording, and any gaps were immediately shared with the data collectors. ## Data processing and statistical analysis Data was checked for completeness, coded, cleaned compiled then entered to Epidata version 4.2.0.0 software and exported to SPSS version 21.0 for analysis. Descriptive statistics were computed and presented in text and tables. Pearson chi-square was done to see the association between KAP and independent factors. Variables with a p value < 0.05 were considered as significant. ## Ethical consideration Institutional review board approval was obtained from Jimma University and written informed consent was obtained from each study participant and guardians of study participants who were illiterate or uneducated. ## Definition of terms Knowledge: understanding of subjects related to diabetes or level of information towards diabetes mellitus. Good knowledge: Those who scored above 15 knowledge questions [19,20] [30]. Moderate knowledge: Those who scored 11–14 knowledge questions20 Poor knowledge: Those who scored 0–10 knowledge questions20,21 Attitude: The way of Diabetes Mellitus patients feels about the disease and its management. Positive attitude: Score of 18-25attitudes point19 Negative attitude: Score of 0–17 attitude point22 Practice: The habit which the diabetic patients exhibits Good practice: Score of 5–9 practice point22 Poor practice: Score of 0–4 practice point17 Life style modification: Refers to the change in living pattern of diabetes to reduce complication of disease and for better outcome of their medication, include non-pharmacological management such as diet modification, and exercise design to treat problem of DM patients23. ## Sociodemographic characteristics of the study participants During 6 month study period, 190 DM patients were included. Most of study participants were in age group of 41–50 years with mean age of 45.83 ± 17.7. Majority of the study participants 115 ($60.5\%$) were males, 65 ($63.5\%$) had no regular income, 108 ($56.8\%$) were living in rural area and 125 ($65.8.\%$) were married. Most of study participants were 53 ($51.5\%$) farmers and 73 ($38.4\%$) had primary educations (Table 1).Table 1Distribution of study participants by socio-demographic characteristic at chronic follow up at JMC Jan 3–20 2021.VariableNumber (N) = 190Percent (%)SexMale11560.5Female7539.5ResidenceRural10856.8Urban8243.2Age (in year)18–302814.731–403618.941–505126.851–603820.0 > 603719.5ReligionCatholic199.1Muslim8339.7Orthodox6430.8Protestant4320.6OccupationGovernment employee2714.2Farmer4624.2Student2814.7Merchant3015.8House wife5931.1Marital statusSingle3216.8Married12565.8Divorced115.8Widowed2211.6Educational statusNo formal education5830.5Primary7338.4Secondary2412.6College and above3518.4Monthly in come in birr < 10008041.81001–19993518.4 > 30007539.5 ## Clinical characteristics of respondents According to background information obtained from patient profile 58 (30.5)% had Type-I DM and 132 ($69.5\%$) had Type-II DM. Duration of disease since diagnosed 0–4 years 93 ($41\%$), 5–9 years 77 ($40.5\%$), 10–14 years 24 ($12.6\%$) and above 15 years 11 (5.8)% (Table 2).Table 2Distribution of diabetics by type of DM, and duration since diagnosis from Health background at JMC Jan 3–20-/2021.VariablesMale (N)Female (N)Total (N)Type of DMI322658II8349132Total11575190Duration of DM in years0–44533785–943347710–1419524 > 158311Total11575190 ## Knowledge regarding diabetes and management Regarding knowledge of respondent about Hyperglycemic symptoms, 149 knows that hyperglycemia causes increased thirst, 57 about increased urination, 51 about loss of consciousness, and 11 increased appetites, 35 about loss of weight and about 9 of them don’t know the symptoms at all. Concerning symptoms of hypoglycemia 66 know that sweating is caused by hypoglycemia, 76 chills/rigors, 110 loss of consciousness 73 palpitation and 7 don’t know any symptom (Table 3).Table 3Distribution of diabetics towards the knowledge of hypoglycemia and hyperglycemia at JMC DM clinic from Jan 3–20 2021.Symptom of hypoglycemiaMale (N)Female (N)Total (N)Symptom of hyperglycemiaMale (N)Female (N)Total (N)SweatingYes453166Increased thirstYes8960150No7044114No261440PalpitationYes462773Increased urinationYes8050130No6948117No352560Loss of consciousnessYes6941110Loss of consciousnessYes211738No463480No9458152Chills (rigors)Yes463076Loss of weightYes241135No6945115No9164155Don’t know347Increased appetiteYes422264No7353126Don’t know459 ## Knowledge regarding life style modification and management Concerning alcohol intake $77.4\%$ know that it’s important to reduce alcohol intake, $75.3\%$ know that regular exercise were important, $42.1\%$ knows importance of smoking cessation, $65.5\%$ importance of alcohol cessation and $52.6\%$ about importance of diet control, $69.09\%$ of study participants know that DM patient should avoid sugar and other sweet diets but $30.91\%$ know it is important to use it. Concerning exercise $60.91\%$ have negative response, the rest $39.09\%$ of them have positive response (Table 4).Table 4Knowledge of DM patients on life style modification, at JMC DM clinic from Jan 3–20 2021.Male N (%)Female N (%)Total (%)Alcohol intakeYes85 ($44.7\%$)62 ($32.6\%$)147 ($77.4\%$)No30 ($15.7\%$)13 ($10\%$)43 ($22.6\%$)Lifestyle modification Weight reduction (regular exercise)Yes87 ($45.8\%$)56 ($29.5\%$)143 ($75.3\%$)No28 ($14.7\%$)19 ($10\%$)47 ($24.7\%$) Smoking cessationYes53 ($27.9\%$)27 ($14.2\%$)80 ($42.1\%$)No62 ($32.6\%$)48 ($25.3\%$)110 ($57.9\%$) Cessation alcohol intakeYes75 ($39.5\%$)50 ($26.3\%$)125 ($65.8\%$)No40 ($21.1\%$)25 ($13.2\%$)65 ($34.2\%$) Diet controlYes61 ($32.1\%$)39 ($20.5\%$)100 ($52.6\%$)No54 ($28.4\%$)36 ($18.9\%$)90 ($47.3\%$)Don’t know11 ($5.8\%$)9 ($4.7\%$)4 ($10.5\%$) Is it important to use sugar and sweet dietsYes37 ($19.5\%$)26 ($13.6\%$)63 ($33.1\%$)No72 ($37.9\%$)45 ($23.6\%$)117 ($61.5\%$)I don’t know6 ($3.1\%$)4 ($2.1\%$)10 ($5.2\%$) Doing exerciseYes78 ($22.73\%$)47 ($16.36\%$)125 ($65.8\%$)No19 ($11.2\%$)18 ($6.37\%$)37 ($19.4\%$)Don’t know18 ($27.27\%$)10 ($16.07\%$)28 ($14.8\%$) ## Knowledge regarding medication, side effects and management Regarding knowledge of diabetes on their type of medications 40 respondents uses insulin with oral hypoglycemic agents, 86 use oral hypoglycemic agents, 57 use insulin only and 7 uses other. Concerning knowledge of study participants on medication side effects $27.4\%$ know that medications of diabetes mellitus can cause hypoglycemia, $12.1\%$ allergic reaction, $18.4\%$ anemia and $16.3\%$ others (GI upset, headache and nausea). Regarding the way in which they manage the side effects, $43.7\%$ manage by visiting hospital, $21.6\%$ by taking another medications, $18.9\%$ stop taking medication, $15.8\%$ by other ways (changing eating style and taking rest) (Table 5).Table 5Knowledge of patients on medication side effects and management at JMC, DM clinic from Jan 3–20-/2021.Male N (%)Female N (%)Total (%)Observed side effects Allergic reaction13 ($6.8\%$)10 ($7.7\%$)23 ($12.1\%$) Anemia15 ($7.9\%$)20 ($10.5\%$)35 ($18.4\%$) Hypoglycemia29 ($15.2\%$)23 ($12.2\%$)52 ($27.4\%$) Weight loss34 ($17.9\%$)15 ($7.9\%$)49 ($25.8\%$) Others weight (gain, GIT upset)24 ($12.6\%$)7 ($3.7\%$)31 ($16.3\%$) Total115 ($61.2\%$)75 ($38.8\%$)190 ($100\%$)Management of side effects Go to hospital46 ($24.2\%$)37 ($19.5\%$)83 ($43.7\%$) Taking another medication25 ($13.2\%$)16 ($8.4\%$)41 ($21.6\%$) Stop medication20 ($10.5\%$)16 ($8.4\%$)36 ($18.9\%$) Others (change eating style and taking a rest)24 ($12.6\%$)6 ($3.2\%$)30 ($15.8\%$) ## Attitude and practice of respondents Regarding the attitude of patients, 142 ($74.7\%$) of them had positive attitude on life style modification, 185 ($97.4\%$) of them had positive attitude on regular visit of DM clinic, 156 ($82.1\%$) had positive attitude on diabetes care and treatment, $95.78\%$ had positive attitude on lifelong treatment of their DM, $51.1\%$ of respondents had positive attitude on continues use of medication; even if their glucose is controlled (Table 6).Table 6Self- care attitude of diabetes patient to control blood glucose and disease complication at JMC, DM clinic from Jan 3–20-/2021.Assessment of attitude perceptionSAANDSDDoes DM can be treated615021121DM requires long term treatment19163620Could you stop your medication whenever your blood glucose is controlled33159970Could life style modification such as dietary modifications, moderation of alcohol intake and cigarette smoking cessationPlay role in treatment of DM31394440Regular visit of clinic has benefit for DM patients50135500 Regarding practice of study participants 152 ($80\%$) participants have positive response for storage and use of medication, and 172 ($90.5\%$) for regular blood taste. Majority have positive response to avoid alcohol drinking, stop smoking, inspection of foot daily for any color Change or any wound and go to clinic as appointed (Table 7).Table 7Self- care practice of diabetes patient to control blood glucose and disease complication at JMC, DM clinic from Jan 3–20-/2021.VariableMaleFemaleTotalRegular exerciseYes6943112No463278DrinkerYes311142No8464148Diet controlYes7547122No402868Yes231134No7776153Clinic visitYes10567172No10818Regular urine testYes6738105No483785Regular blood testYes10666172No9918Inspection of footYes5942101No563389Medication use and storageYes9656152No191938 ## Association between KAP and dependent variables Pearson chi-square association shows that occupation ($$p \leq 0.014$$), marital status (p = < 0.001) and educational status ($$p \leq 0.013$$) were significantly associated with knowledge towards LSM and medication use (Table 8).Table 8Association of knowledge with socio demographic variables towards LSM and medication use among DM patients at JMC, January 3–20-2021.VariablesStatus of knowledgep valueGood knowledgeModerate knowledgePoor knowledgeSexMale4350220.829Female263217Age18–30131320.09231–401314941–502321751–6014159 > 6061912ReligionMuslim1445250.26Orthodox15309Protestant844Catholic531OccupationGovernment employee91530.014Merchant11154Farmer171415Student1792House wife152915Marital statusMarried455525 < 0.001Single20111Divorce353Widowed11110Educational statusNo formal education1325200.013Primary312814Secondary12102College and above13193ResidenceUrban area3134170.911Rural area384822Monthly income < 10002930210.4001001–299914174 > 3000263514 ## Factors associated with attitudes Pearson chi-square association shows, age with p value of (p = < 0.001), occupational status ($$p \leq 0.04$$), marital status (p = < 0.001), and educational status (p = < 0.001) were significantly associated with the attitude of participants towards LSM and medication use of diabetes patient (Table 9).Table 9Pearson chi-square association of attitude with socio demographic variables towards LSM and medication use among DM patients at JMC, January 3–20-2021.VariablesStatus of attitudep valuePositive attitudeNegative attitudeSexMale96190.259Female976Age18–30271 < 0.00131–4034241–5031351–602374 > 601215ReligionMuslim91200.269Orthodox495Protestant142Catholic30OccupationGovernment employee2700.04Merchant246Farmer3313Student271House wife527Marital statusMarried11312 < 0.001Single311Divorce74Widowed1210Educational statusNo formal education3820 < 0.001Primary685Secondary231College and above341ResidenceUrban area7570.051Rural area8820Monthly income < 100069110.0711001–2999269 > 3000687 ## Factors associated with practices Pearson chi-square association shows, age with p value of ($$p \leq 0.03$$), and marital Status ($$p \leq 0.049$$) of the participants were significantly associated with practice toward LSM and medication use of diabetes study participants (Table 10).Table 10Pearson chi-square association of practice with socio demographic variables towards LSM and medication use among DM patient at JMC, January 3–20 2021.VariablesStatus of practicep valueGood practicePoor practiceSexMale90250.114Female5124Age18–302440.0331–4032441–50361551–602711 > 602215ReligionMuslim81300.132Orthodox377Protestant143Catholic90OccupationGovernment employee2340.08Merchant237Farmer3016Student253House wife4019Marital statusMarried95300.049Single284Divorce83Widowed1012Educational statusNo formal education38200.108Primary school5419Secondary school186College and above314ResidenceUrban area62200.701Rural area7929Monthly income < 100057530.7231001–2999278 > 30005718 ## Discussion Majority of respondents in this study came from the age groups 41–50 years ($26.8\%$). Which is similar with study conducted at Adama hospital21. This is reflective of the fact that the etiology of diabetes mellitus usually at old age24,25–27. In this study respondents with no formal education consists 58 ($30.5\%$) and only ($18.4\%$) respondents with college and above education. This result may be the direct consequence of scarcity of higher education system in Ethiopia in the past23. This shows that there is some improvement in educational status of DM patient compared with the study conducted in 2011 at JUSH which reported as there was no attention given to diabetic education in Ethiopia22. Majority of respondents in this study had income less than 1000 birr 80 ($42.1\%$). This low income among respondents could limit their accessibility and affordability of a well-balanced diet and healthy food and it was considered as the main factors (barrier) to their practice of life style modification and proper use of their medications. This finding was in keeping with study conducted in Gondar; in which majority of the study participants 139 ($43\%$) had very low monthly income28. In terms of Knowledge Assessment, 69 ($36.3\%$) had good knowledge, 82 ($43.2\%$) moderate knowledge and 39 ($20.5\%$) poor knowledge of LSM and medication use of the DM patient. In contrast to this finding study done in Western Nepal shows that knowledge, attitude and practice among diabetes patients were low15. This discrepancy might be due to study setting, study design, and geographical area. Regarding Knowledge toward life style modification, $75.3\%$ respond that regular exercise where important for DM patient, and $42.1\%$ knows importance of smoking cessation, $65.5\%$ cessation of alcohol importance and $52.6\%$ about importance of diet control. In line with this Study done in Malaysia identified a good KAP score of life style modification required for diabetes patients10. Regarding the attitude of respondent, 142 ($74.7\%$) of them have positive attitude on life style modification, 185 ($97.4\%$) of them had positive attitude on regular visit of DM clinic, 156 ($82.1\%$) had positive attitude on diabetes care and treatment, $95.78\%$ lifelong treatment of DM, $51.1\%$ of respondents have positive attitude on continues use of medication; even if their glucose is controlled. This finding is similar to study done in South Africa in which the majority of respondents $92.7\%$ and $51.6\%$ had positive attitude on lifestyle modification and medication use respectively28. Among study respondent about 141 ($74.2\%$) had good practice of LSM and medication use, while 49 ($24.8\%$) had poor practices which is comparative with study done in Gondar, Ethiopia, that were $74.4\%$ showed good practice21. However, it was higher than the study done in Harar, Ethiopia, in which $39.2\%$ had good self-care practice [30]. This might be due to difference in socio demographic, study design, study setting and access to health education programs. In this study, occupation status and educational status showed a significant association with knowledge, whereas age, occupation status, marital status and educational status showed a significant association with attitude, age and marital Status showed a significant association with practice towards LSM and medication use. This finding was similar to the studies conducted in different settings15,20,22.The difference among different study findings may be difference literacy of patients, training given to the patients, availability of information on different facilities. ## Limitations of study The current study is only focused on patients aged 18 years and above, conducted at single setting, JMC and it did not consider DM patients who did not visit the health institutions during the study period. ## Conclusion The result of this study showed that more than $20\%$ of the participants had poor knowledge, attitude, and practice towards medication use and LSM. Marital status was the only variable which remained to be significantly associated with all three: knowledge, attitude and practice towards LSM and medication use. ## Recommendation Lifestyle modification has a great role in the prevention and control of blood glucose raised. But, there has to be much to be done on this, since its prevalence is increasing rapidly in worldwide especially in developing nations on the top of changed lifestyle, the deficit in the knowledge, attitude, and practice towards lifestyle modification. On the basis of the study findings, the following recommendations were forwarded Health education about diabetes and types of lifestyle modification has to be strengthened. Better to create awareness to the community through health education about the importance of physical exercise, weight loss, feet care, healthy dietary habit and cessation of smoking. ## References 1. 1.Chawla, R. Diabetes: An Introduction, Manual of Diabetic Care (2014). 2. 2.American Diabetes Association (ADA). Standards of medical care in diabetes. J. Clin. Appl. Res. Educ.43(1), S14–S31. 10.2337/dc20-S002 (2020). 3. Jasper US, Opara MC, Pyiki EB. **Knowledge of insulin use and its determinants among Nigerian insulin requiring diabetes patients**. *Diabetes Metab. Disord.* (2014.0) **13** 1-8 4. Weinger K, Beverly EA, Smaldone A. **Diabetes self-care and the older adult**. *West. J. Nurs. Res.* (2014.0) **36** 1272-1298. DOI: 10.1177/0193945914521696 5. Mohammadi S, Karim NA, Talib RA, Amani R. **Knowledge, attitude and practices on diabetes among type 2 diabetic patients in Iran: A cross-sectional study**. *Sci. J. Public Health.* (2015.0) **3** 520-524. DOI: 10.11648/j.sjph.20150304.20 6. Kassahun T, Gesesew H, Mwanri L, Eshetie T. **Diabetes related knowledge, self-care behaviours and adherence to medications among diabetic patients in Southwest Ethiopia: A cross-sectional survey**. *BMC Endocr. Disord.* (2016.0) **16** 28. DOI: 10.1186/s12902-016-0114-x 7. 7.World Health Organisation Diabetes, Retrived April (2016). 8. 8.Stuckey, H. & Benedetti, M. M. Enhancing diabetes awareness a prevention: The second diabetes attitudes, wishes and needs (DAWN2) study (2011). 9. Hassali MA, Shafie AA, Sulaiman SAS. **The 14-item Michigan diabetes knowledge test: Translation and validation study of the Malaysian version**. *Pract. Diabetes Int.* (2010.0) **27** 1-5 10. Norhafizah S, Siti ZM, Riyanti S. **Medication labeling literacy among Malaysian with diabetes: A cross sectional study**. *J. Diabetes Res. Clin. Metab.* (2012.0) **2050–0866** 1-23 11. Surendranath A, Nagaraju A. **Study to assess the knowledge and practice of insulin self-administration among patients with diabetes mellitus**. *Asian J. Pharm. Clin. Res.* (2011.0) **5** 63-66 12. Adepu R, Somashekar M. **Influence of structured patient education on therapeutic outcomes in diabetes and hypertensive patients**. *Asian J. Pharm. Clin. Res.* (2010.0) **16** 174-178 13. 13.International Diabetic Federation. IDF Diabetes Atlas, 6th ed (IDF, 2014). 14. Urquhart J. **The odds of three nons when appropriately prescribed medicine is not worcininon compliance, non-absorption and non-response**. *Br. J. Clin. Pharmacol. Diabetes Educ.* (2002.0) **2** 96-97 15. Gudina EK, Amade ST, Tesfamichael FA, Ram R. **Assessment of quality of care given to diabetic patients at Jimma University Specialized Hospital diabetes follow-up clinic, Jimma, Ethiopia**. *BMC Endocr. Disord.* (2011.0) **11** 19. DOI: 10.1186/1472-6823-11-19 16. Palaian S, Acharya LD, Rao PG, Shankar PR, Nair NM, Nair NP. **Knowledge, attitude, and practice outcomes: Evaluating the impact of counseling in hospitalized diabetic patients in India**. *P AND T.* (2006.0) **31** 383-400 17. Wild S, Roglic G, Green A, Scree R, King H. **Global prevalence of diabetes Estimates for the year 2000 and projection for 2030**. *Diabetes Care* (2004.0) **27** 1047-1053. DOI: 10.2337/diacare.27.5.1047 18. Upadhyay DK, Palaian S, Shankar PR, Mishra P, Pokhara N. **Knowledge, Attitude and Practice about Diabetes among Diabetes Patients in Western Nepal**. *Rawal Med. J.* (2008.0) **33** 8-11 19. Badruddin N, Basit A, Hydrie IZ, Hakeem R. **Knowledge, attitude and practices of patient visiting diabetes care unit**. *Pak. J. Nutr.* (2002.0) **11** 99-102 20. Adem AM, Gebremariam ET, Gelaw BK, Ahmed M, Fromsaseifu M, Thirumurugan DG. **Assessment of knowledge, attitude and practices regarding life style modification among type 2 diabetic mellitus patients attending Adama Hospital Medical College, Oromia Region, Ethiopia**. *Glob. J. Med. Res.* (2014.0) **14** 6 21. Asmelash D, Abdu N, Tefera S, Baynes HW, Derbew C. **Knowledge, attitude, and practice towards glycemic control and its associated factors among diabetes mellitus patients**. *J. Diabetes Res.* (2019.0) **9** 3-6 22. Abyot E, Fitsum S. **Knowledge, attitude and practice among diabetic patients on insulin therapy towards the disease and their medication at a university hospital in Northwestern Ethiopia: A cross-sectional study**. *IJPSR.* (2014.0) **5** 685-692 23. Okonta HI, Ikombele JB, Ogunbanjo GA. **Knowledge, attitude and practice regarding lifestyle modification in type 2 diabetic patients**. *Afr. J. Prim. Health Care Fam. Med.* (2014.0) **6** 1-6. DOI: 10.4102/phcfm.v6i1.655 24. Al Bimani ZS. **Evaluation of T2DM related knowledge and practices of Omani patients**. *Saudi Pharm. J.* (2014.0). DOI: 10.1016/j.jsps.2013.12.006C 25. Fauci AS, Braunwald E, Kasper DL, Hauser SL, Longo DL, Jameson JL. *Harrison’s Principles of Internal Medicine* (2012.0) 26. Ismaeil FMR, Ali N. **Diabetic patients knowledge, attitude and practice toward oral health**. *J. Educ. Pract.* (2013.0) **4** 19-25 27. Berhe KK, Kahsay BA, Gebru BH. **Adherence to diabetes self-management practices among type II diabetic patients in Ethiopia; A cross sectional study**. *Green J. Med. Sci.* (2013.0) **3** 21-23 28. Mekonnen CK, Abate HK, Tegegne ET. **Knowledge, attitude, and practice toward lifestyle modification among diabetes mellitus patients attending the University of Gondar Comprehensive Specialized Hospital Northwest, Ethiopia**. *Diabetes Metab. Syndr. Obes. Targets Ther.* (2020.0) **13** 1969-1977. DOI: 10.2147/DMSO.S250787