text
string | predicted_class
string | confidence
float16 |
|---|---|---|
Right kidneys were used for preparation of total RNA using Trizol (Thermo Fisher Sci, Inc., Waltham, MA). qRT-PCR (Omniscript, QuantiTect, Qiagen, Valencia, CA) was performed in an ABI-Step One Plus Cycler (Fisher, Life Technologies, Waltham, MA) using the mouse beta actin forward primer: GGCTGTATTCCCCTCCATCG, and reverse primer: CCAGTTGGTAACAATGCCATGT, the mouse Enpp1 forward primer: CTGGTTTTGTCAGTATGTGTGCT and reverse primer: CTCACCGCACCTGAATTTGTT, the mouse Entpd5 forward primer: CCAAAGACTCGATCCCCAGAA and reverse primer: TGTTAGAAAGTTCACGGTAACCC, the mouse Ank forward primer: TACGGGCTGGCGTATTCTTTG and reverse primer: CACTGTAGGCTATCAGGGTGT, and the mouse Tnsalp forward primer CCAACTCTTTTGTGCCAGAGA and reverse primer: GGCTACATTGGTGTTGAGCTTTT.
|
study
| 97.2 |
Data are expressed as means±SEM and analyzed in Microsoft Excel 2010 or Graphpad Prism 6.0. Differences were considered significant if p-values, calculated using the unpaired, two-tailed Student’s t-test, linear regression analysis, or one-way ANOVA using Tukey’s adjustment for multiple comparisons, were smaller than 0.05.
|
other
| 63.53 |
Humans with loss-of-function of NPT2a [1–3] and NPT2c [4, 5] develop renal mineralization, which may manifest during early childhood prior to specific therapy or when inappropriately receiving active vitamin D analogs, but can also occur throughout life . To model these kidney abnormalities, we used 2 months old Npt2a-/- mice [39, 40] placed on a diet containing 0.6% calcium and 0.7% phosphorus (Harlan Teklad TD.2018S).
|
study
| 100.0 |
Urine pyrophosphate concentration (U-PPi, A) following an overnight fast and renal gene expression as indicated on the y-axis for ectonucleotide pyrophosphatase/phosphodiesterase 1 (Enpp1, B), progressive ankylosis (Ank, C), ectonucleoside triphosphate diphosphohydrolase 5 (Entpd5, D), tissue nonspecific alkaline phosphatase (Tnsalp, E) in mice fed regular chow for 10 weeks. The data represent mean±SEM of 4–19 mice, p-values shown above the lines of comparisons were calculated by one-way ANOVA using Tukey’s adjustment for multiple comparisons (A) and Student’s t-test (B-E).
|
study
| 100.0 |
Similarly, urine PPi excretion corrected for urine creatinine was increased in Npt2a-/- mice (3.0±0.53 micromole/mg, n = 19 vs. WT 1.3±0.42 micromole/mg, n = 9, p = 0.038) (Panel A in S1 Fig). Evaluation of whole kidney gene expression was unchanged for the PPi-generating enzyme Enpp1 (0.004±0.001, n = 9 vs. WT 0.005±0.001, n = 7, p = ns) and decreased for the PPi transporter Ank (0.00015±2.8e-5, n = 9 vs. WT 0.001±0.00014, n = 10, p = 0.007) (Fig 1B and 1C). Expression of the Pi-generating enzyme Entpd5 was decreased (0.06±0.01, n = 9 vs. WT 0.6±0.15, n = 10, p = 0.0073) and expression of Tnsalp, which hydrolyses PPi to Pi, was increased (0.07±0.02, n = 9 vs. WT 0.02±0.004, n = 10, p = 0.0043) (Fig 1D and 1E). Thus, the source of urine PPi in Npt2a-/- mice remains unclear and may be extrarenal, localized to a specific tubular segment inside the kidneys, or regulation may occur on the post-translational level.
|
study
| 100.0 |
To further evaluate the role of PPi in renal mineral deposit formation in the setting of renal phosphate wasting we next reduced endogenous PPi production using the hypomorphic murine Enpp1asj allele or administered sodium pyrophosphate by intraperitoneal injection as previously described to increase PPi.
|
study
| 100.0 |
Enpp1asj/asj mice develop renal mineralization on a “stone-forming” high phosphorus, low magnesium diet, while they develop no renal mineralization on regular chow [17, 42]. Presence of two hypomorphic asj alleles of Enpp1 blunted the increase of the urine PPi concentration of double-mutant mice when compared to Npt2a-/- mice on regular chow, albeit non-significantly (67±21, n = 4 vs. Npt2a-/- 1257±272 micromole/l, n = 19, p = 0.084, Fig 1A). Similarly, urine PPi excretion corrected for urine creatinine was decreased in double mutant mice (0.43±0.084 micromole/mg, n = 4 vs. Npt2a-/- 3.0±0.53 micromole/mg, n = 19, p = 0.044, panel A in S1 Fig). One or two hypomorphic asj alleles of Enpp1 furthermore increased the calcified area of double-mutant mice when compared to Npt2a-/- mice on regular chow in a gene dose-dependent fashion (0.3±0.07, n = 8 in Enpp1asj/+/Npt2a-/-, p = ns vs. 0.26±0.04% in Npt2a-/- and 0.69±0.15% in Enpp1asj/asj/Npt2a-/-, p<0.0001 vs. Npt2a-/-) while no mineral deposits were found in Enpp1asj/asj mice on regular chow (Fig 2A). Since increased calcified area in double mutants was due to an increase in number of calcifications, no difference was observed for mineralization size between Npt2a-/-, Enpp1asj/+/Npt2a-/-, and Enpp1asj/asj/Npt2a-/- mice (Fig 2B). Renal calcified area inversely correlated with spot urine PPi concentration (slope = -5.226e-005 ± 2.391e-005, R2 = 0.126, p = 0.036) (Fig 3A). No significant correlation was found for calcification area (Fig 3B) or when area and size were correlated with urine PPi corrected for urine creatinine (Panels C and D in S1 Fig).
|
study
| 100.0 |
Histomorphometric analysis of renal mineralization (%calcified area = 100*mineralization area/tissue area, A; calcification size = mineralization area/number of calcifications, um2, B) in 10 um sections of kidneys from mice fed regular chow for 10 weeks. The data represent individual animals (closed circles) with the means±SEM, p-values shown above the lines of comparisons were calculated by one-way ANOVA using Tukey’s adjustment for multiple comparisons, no significant differences were detected between groups in panel B.
|
study
| 100.0 |
All experimental WT and mutant mice from Fig 2 (n = 28) for which urine was available were evaluated using linear regression analysis to determine the association of renal mineralization with the urine pyrophosphate concentration (U-PPi) (% calcified area = 100*calcified area/total area A and calcification size = calcified area/number of mineralization B). Data points represent values of individual animals. Results of the linear regression analysis are shown as solid line with 95% confidence interval (stippled lines), R2 and p-values.
|
study
| 100.0 |
Intraperitoneal injection of sodium pyrophosphate was previously shown to reduce arterial calcification in an uremic mouse model . We used the dose of 160 micromole/Kg/day published by these authors and two weeks old Npt2a-/- pups for this experiment, because renal calcification is more pronounced when compared to older mice (Fig 4A and 4C). Size and body weight (BW) of mice in the treatment group were indistinguishable from vehicle and the animals appeared to be thriving well. Following sacrifice at four weeks of age we observed a reduction of renal mineral deposits by 33% in the treatment group (0.4±0.04, n = 9 vs. vehicle 0.7±0.06%, n = 12, p = 0.01) (Fig 4C and 4D) while mineralization size again was unaffected (Fig 4E). Plasma PPi levels at sacrifice were increased, albeit non-significantly (3.9±0.8, n = 9 vs. vehicle 2.0±0.4 micromole/l, n = 5, p = ns) (Fig 4F). Likewise, the U-PPi concentration was increased (244.9±33.2, n = 14 vs. vehicle 149.4 ± 28.8 micromole/l, n = 14, p = 0.039) (Fig 4G and panel B in S1 Fig).
|
study
| 100.0 |
Light micrographs of 10 um renal sections prepared from paraffin-embedded kidneys, obtained from mice with various genotypes fed regular chow for 10 weeks (A, upper panels: von Kossa, methylene green staining, 4X, and A, lower panels: von Kossa, hematoxylin and eosin staining, 40X); Transmission electron micrographs showing microspheres in double mutant mice on regular chow, inset with larger magnification shown to the right (B); Two weeks old Npt2a-/- pups treated with i.p. injections of vehicle or sodium pyrophosphate (160 micromole/Kg/day) for two weeks (C); Histomorphometric analysis of renal mineralization (%calcified area = 100*mineralization area/tissue area, (D); calcification size = mineralization area/number of calcifications, um2, (E), and plasma pyrophosphate levels (F) and urine pyrophosphate (U-PPi) (G) of two weeks old Npt2a-/- pups treated with i.p. injections of vehicle or sodium pyrophosphate (160 micromole/Kg/day) for two weeks, measured after overnight fast and 18–24 hrs. following the last treatment. The data represent individual animals (closed circles) with the means±SEM, p-values shown above the lines of comparisons were calculated by Student’s t-test.
|
study
| 100.0 |
Histological evaluation showed large interstitial mineral deposits that displaced the surrounding renal tubules. In addition, we observed small intraluminal mineral deposits in cortical and medullary tubular segments of the kidneys of Npt2a-/- and double-mutant mice (Fig 4A). Transmission electron images showed concentric spheres of similar morphology in Npt2a-/- and double-knockout mice (Fig 4B) as previously described for Npt2a-/- mice by us [13, 43] and others [33, 34]. No mineralization was observed in renal vasculature or in the renal pelvis of our mice.
|
study
| 100.0 |
Oral phosphate supplements are currently thought to be the primary intervention to reduce risk for renal mineralization in human carriers of NPT2a and NPT2c mutations. However, there is concern that oral phosphate therapy might contribute to the formation of renal mineralization despite reduced 1,25(OH)2D levels and reduced urinary calcium excretion under certain conditions, for example in patients with X-linked hypophosphatemia (XLH) treated with oral phosphate supplements given multiple times throughout the day [44, 45] and in otherwise healthy individuals following treatment with phosphate enema .
|
review
| 99.7 |
We recently reported that reduced urine levels of osteopontin (Opn), an extracellular matrix factor affecting binding of phosphate to hydroxyapatite crystals, contribute to the development of nephrocalcinosis in Npt2a-/- mice . The present report describes that the urine PPi concentration may be an additional modifier of renal calcifications in this mouse model.
|
study
| 100.0 |
Reduced Enpp1 activity increased the % calcified area in double mutant mice when compared to Npt2a-/- mice (Fig 4A), while the size of the calcium phosphate deposits was not affected. Similarly, intraperitoneal sodium PPi treatment reduced % calcified area, while calcification size was unchanged. Although further studies are required to define cause and effect, these data suggest that PPi inhibits nucleation (Figs 2A and 4A), which is different from the effect of Opn reported by us , that predominantly decreases mineralization size, consistent with the known role of Opn in calcium phosphate crystal growth. Interventions that increase both PPi and Opn would therefore be predicted to be additive.
|
study
| 100.0 |
Enpp1 expression is positively regulated by phosphate in osteoblast cultures , and therefore we expected that expression is likewise increased in Npt2a-/- mice to explain the increased urine PPi levels. Instead, we found that Enpp1 expression is unchanged, possibly as a result of reduced Pi sensing in the absence of Npt2a. Furthermore, Ank expression was decreased and Tnsalp was increased, all predicted to reduce local PPi production. These findings suggest that PPi may be generated outside of the kidneys contrary to previous reports [25, 26], and elevate urine PPi despite unchanged or decreased local gene expression for Enpp1 and Ank, respectively. Consistent with this hypothesis is our finding that global reduction of Enpp1 activity in Enpp1asj/asj mutant mice decreased urine PPi levels and that intraperitoneal injection of sodium pyrophosphate increased urine PPi levels (Fig 4G). Alternatively, PPi production may be regulated locally by increased renal activities of Enpp1 and Ank on a post-transcriptional level.
|
study
| 100.0 |
Interestingly, urine PPi in 10 weeks old Npt2a-/- mice is higher than in 4 weeks old weanlings (1257±272 micromole/l vs. 149.4 ± 28.8 micromole/l). This may be a developmental change of urine PPi over the first 10 weeks of life and could be a contributing factor explaining the initial observation in Npt2a-/- mice reported by the Tenenhouse lab , that renal calcifications peak with weaning age and subsequently decrease during adult life in these mice.
|
study
| 100.0 |
Tissue specific ablation of Enpp1 (and possibly Ank) could help determine in future studies whether PPi is produced renally or extrarenally. Injection of recombinant Enpp1 may be able to reduce the renal calcifications in Npt2a-/- mice [26, 29] and provide further evidence of the causal relationship of this extracellular enzyme, urine PPi, and renal mineralization.
|
study
| 100.0 |
Also, separate evaluation of interstitial and luminal mineralization and PPi levels and/or activity of PPi generating enzymes may be of interest in future studies. Finally, determining how urinary pH, anion gap, citrate, oxalate, magnesium, and the expression of uromodulin (Tamm-Horsfall protein, THP) or Opn modify PPi action may help better understanding the pathogenesis of renal mineralization in Npt2a-/- mice.
|
study
| 99.94 |
In summary, we show here that urine PPi is increased in Npt2a-/- mice. Presence of one or two hypomorphic Enpp1asj alleles decreases urine PPi and increases renal mineral deposits in Npt2a-/- mice. Furthermore, the development of nephrocalcinosis and nephrolithiasis in these mice can be reduced by intraperitoneal administration of sodium pyrophosphate. These studies suggest that PPi may be a thus far unrecognized factor modulating the development of renal calcifications in Npt2a-/- mice which may be, if confirmed in humans, of diagnostic and therapeutic relevance for phosphaturic disorders.
|
study
| 100.0 |
Urine pyrophosphate excretion of mice fed regular chow for 10 weeks (U-PPi/U-crea, A) and urine pyrophosphate excretion (U-PPi/U-crea) of two weeks old Npt2a-/- pups treated with i.p. injections of vehicle or sodium pyrophosphate (160 micromole/Kg/day) for two weeks (B), measured after overnight fast and 18–24 hrs. following the last treatment. Linear regression analysis to determine the association of renal mineralization with the ratio of urine pyrophosphate/urine creatinine (U-PPi/U-crea) (% calcified area = 100*calcified area/total area C and calcification size = calcified area/number of mineralization D). The data represent individual animals (closed circles) or means±SEM, p-values shown above the lines of comparisons were calculated by one-way ANOVA using Tukey’s adjustment for multiple comparisons (A) and Student’s t-test (B-D).
|
study
| 100.0 |
Diffuse large B cell lymphoma (DLBCL) is the most common lymphoma in Western countries and, despite improvements obtained with chemoimmunotherapy, up to half of these patients cannot be cured [1, 2]. Currently, two main DLBCL subtypes are recognized based on their phenotypic homology with their putative cell of origin, the germinal center B-cell type and the activated B-cell like (ABC) type [2–5]. ABC-DLBCL is less responsive to standard regimens and is characterized by activation of B-cell receptor signaling and the nuclear factor kB pathway [2–5], providing therapeutic targets that are currently being explored in the clinic with compounds such as the Bruton's tyrosine kinase (BTK) inhibitor ibrutinib .
|
review
| 99.9 |
Analogous to the vast majority of human tumors and independent of their cell of origin, DLBCL cells also bear recurrent somatic mutations in genes coding for proteins involved in chromatin structure and remodeling that cause profound changes at the epigenetic level [3, 7]. Of clinical relevance, epigenetic changes can be at least partially reversed and epigenetic drugs can increase sensitivity to other anticancer agents [7–10]. Over the last few years, inhibitors of the bromodomain and extraterminal (BET) protein family have become the focus of extensive research as a novel class of epigenetic drugs . BET proteins are key epigenetic regulators of gene transcription and their inhibition has resulted in antitumor activity in different tumor models, including lymphomas [11–20]. OTX015 (MK-8628) is a thienotriazolodiazepine compound that potently inhibits the BET proteins BRD2, BRD3 and BRD4 . The compound competitively occupies the acetyl-binding pockets of BET bromodomains, leading to release of the BET protein from the chromatin . Importantly, in normal and cancer cells, more than half of all BRD4 proteins are bound to a small number of enhancers (super-enhancers) that control the expression of genes fundamental to the control and establishment of individual cell identities, such as PAX5, BCL6, CD79A, CD79B, FOXO1 in B-cells or PRDM1, IRF4 and MUM1 in plasma cells [22–24].
|
review
| 99.44 |
OTX015 has in vitro and in vivo antitumor activity as a single agent in different lymphoma models, including ABC-DLBCL . Clinical responses including complete remissions with single agent OTX015 have been recently reported in patients with relapsed or refractory lymphoma or acute leukemia enrolled in phase I studies, in the absence of major toxicities [25, 26]. Although the mechanism of action of BET inhibitors is likely pleiotropic, down-regulation of genes involved in B cell identity and germinal center formation, and, especially in the ABC-DLBCL setting in which such effects can lead to apoptosis, inhibition of the B-cell receptor and nuclear factor kB signaling pathways play an important role [12–14]. Since OTX015 presented in vitro synergism when combined with different agents in lymphoma models , we evaluated the in vivo activity of OTX015-containing combinations in an ABC-DLCBL xenograft model.
|
study
| 100.0 |
Based on the in vitro synergism observed for combinations of OTX015 with other compounds , we evaluated the activity of combinations of this bromodomain inhibitor in an in vivo model of ABC-DLBCL. Mice bearing xenografts of the ABC-DLBCL cell line SU-DHL-2 were treated with control or OTX015, BTK inhibitor ibrutinib, the mechanistic target of rapamycin (mTOR) inhibitor everolimus, the histone deacetylase inhibitor vorinostat, or the anti-CD20 monoclonal antibody rituximab as single agents or in OTX015-containing combinations. None of the mice showed any body weight loss during the treatment period. When given as single agents, OTX015 and all four other drugs caused tumor growth delay (Figure 1A). When given in combination, the antitumor activity was significantly greater, with an almost complete and immediate tumor eradication in mice receiving the OTX015-containing combinations, maintained throughout treatment (P<0.001) (Figure 1A). The degree of necrosis was also evaluated in three tumors per group. Tumors from mice treated with rituximab (P=0.0463), everolimus (P=0.0463) or ibrutinib (P=0.0431) as single agents, or with OTX015 combinations plus everolimus (P=0.0463), plus ibrutinib (P=0.0431), and plus vorinostat (P=0.0463) presented a higher percentage of necrotic cells than control mice (Figure 1B, 1C). Higher necrosis was observed in tumors from mice treated with the OTX015 and vorinostat combination compared to the single agent vorinostat group (P=0.0109). Together with our previous in vitro findings with OTX015 as a single agent and in combination , the OTX015 antitumor activity reported as single agent in the phase I hematologic study , and similar positive results of other combination regimens based on BET inhibitors [13, 16–18], these novel in vivo data confirm the combinability of OTX015 with classic cytotoxic and targeted therapies in lymphoma and provide supporting rationale for future clinical development strategies in lymphoma. Due to the direct effect of OTX015 and other BET inhibitors on MYC expression, independently of the presence of chromosomal translocations , also high-risk populations such the double-hit or double-expressor lymphomas could be targeted.
|
study
| 99.94 |
A. Changes in tumor volumes during treatment: Black, vehicle (control mice); Blue; single agent OTX015; Red, single agent targeted drug; Green, OTX015/targeted drug combination. B. Boxplots showing percentage of tumor necrosis at the end of treatment. In each boxplot, the line in the middle of the box represents the median and the box extends from the 25th to the 75th percentile (interquartile range). * P < 0.05 when compared with control (CTR) mice. C. Histopathological analysis revealed control mice or treated only with rituximab displayed vital cell with a diffuse growth pattern (upper and lower left); addition of OTX015 was associated with large areas of coagulative necrosis (Haematoxyln and Eosin, 200X).
|
study
| 100.0 |
Pharmacokinetics analyses showed similar OTX015 levels in plasma and tumor samples 4 h after the last treatment when administrated as a single agent, with values of ~750 ng/ml in plasma, which is equivalent to the 1.5 μM concentration that has strong in vitro activity , and ~750 ng/g of tissue for tumor samples (Figure 2A–2B). Terminal levels of the bromodomain inhibitor in all experimental groups treated with OTX015 in combination, differed from the group exposed to OTX015 as single agent. Co-treatment with ibrutinib or everolimus induced an increase in OTX015 concentrations both in plasma (Figure 2A) and tumor samples (Figure 2B). On the other hand, treatment with rituximab decreased OTX015 accumulation in the tumor tissue and OTX015 was not detected in plasma samples in mice concomitantly receiving vorinostat. These results are based on a limited number of mice but suggest that extended pharmacokinetic/pharmacodynamic studies should be mandatory in phase I combination studies to explore the behavior of OTX015 when administered with other agents.
|
study
| 100.0 |
In conclusion, OTX015 showed strong in vivo activity in a murine xenograft model of ABC-DLBCL when combined with ibrutinib, everolimus, rituximab, or vorinostat. Our results provide the rationale to explore OTX015-containing combinations in the clinical setting.
|
study
| 99.94 |
NOD-Scid (NOD.CB17-Prkdcscid/NCrHsd) mice (five weeks of age, approximately 20 g body weight; Harlan Laboratory, S. Pietro al Natisone, Udine, Italy) were subcutaneously engrafted with 15 x106 cells of the human ABC-DLBCL cell line SU-DHL-2, and randomly divided into 10 groups of six mice each. Treatment with OTX015 (50 mg/kg once daily, oral, qdx7/w x5w; Oncoethix GmbH, a wholly owned subsidiary of Merck Sharp & Dohme Corp, Lucerne, Switzerland; formerly Oncoethix SA) was initiated three days after the engraftment, while treatment with ibrutinib (5 mg/kg PO; qdx2/w x5w), everolimus (1 mg/kg PO; qdx2/w x5w), vorinostat (15 mg/kg PO; qdx2/w x5w; Selleckchem, Houston, TX, USA) or rituximab (3 mg/kg IV; qdx1/w x5w; Roche, Basel, Switzerland) was initiated when mice developed palpable tumors (100 mm3). Tumor size was measured twice weekly using a digital caliper. Tumor volumes were calculated as previously described . Mice maintenance and animal experiments were performed with study protocols approved by the local Swiss Cantonal Veterinary Authority (No. 10/2014). Tumor specimens were collected at the end of treatment. Necrosis was semi-quantitatively assessed on hematoxylin-eosin stained slides. The percentage of necrotic cells on the total amount of the neoplastic tissue was evaluated on the whole section in 3 mice per each group. Differences in tumor volumes and percentage of necrosis were calculated using the Wilcoxon rank-sum test (Stata/SE 12.1 for Mac, Stata Corporation).
|
study
| 99.94 |
For OTX015 plasma and tissue concentrations, samples were collected 4 h after the last OTX015 treatment. Mice were sacrificed and blood was collected from the heart in heparinized tubes, separated immediately by centrifugation (4000 rpm, 15 min, 4°C), and stored at −80°C. Plasma concentrations were measured using a validated Ultra Performance Liquid Chromatography with tandem Mass Spectrometry method, as previously described . For tissue measurements, frozen samples were weighed and then homogenized in 1 mL of water, and a 50 μL sample was prepared using the same extraction method as that used for plasma samples.
|
study
| 100.0 |
The transmission of infections can be sensitive to changes in population density, especially in the case of fluctuating wildlife populations [1–3]. When modelling disease transmission, the probability of encountering an infected individual is typically assumed to be either independent of (frequency-dependent) or linearly dependent on (density-dependent) population density . Sexually transmitted infections are generally described using frequency-dependent transmission because the number of sexual contacts is assumed to remain constant, regardless of population density . Infections that are transmitted through regular ‘every-day’ contacts are often assumed to be density-dependent .
|
study
| 99.9 |
The choice of which contact–density function to use in a model of disease transmission entails potentially significant consequences . Most importantly, the two functions differ in whether or not an infection is expected to persist below a critical density of individuals . The basic reproductive number (R0), defined as the number of secondary infections arising from the introduction of one infectious individual into a completely susceptible population, is a central epidemiological measure that characterizes the spread of an infection, and provides an immediate approximation as to how rapidly an infection can spread . In its simplest form, R0 = βN, where N is the population size and β is the transmission coefficient that consists of the rate p of becoming infected through contact with an infectious individual, multiplied by the contact–density function that equals cN/A (where A is area) when linear (density-dependent) and c when constant (frequency-dependent), and random homogeneous mixing is assumed . When transmission coefficient β changes with density, theory predicts a density below which the transmission rate is too low, causing R0 to fall below 1 and the disease to disappear, whereas no persistence threshold density exists when β remains constant, independent of density . Because changes in the transmission coefficient determine how quickly an infection can spread through a population, it can also be expected that the two contact–density functions will differently affect outbreak characteristics such as incidence, prevalence and outbreak size .
|
study
| 99.94 |
Because human populations are usually large and stable, and because some types of contacts do not appear to change with population density , many models of human disease transmission will not be significantly affected by the choice of the contact–density function. But when one is interested in modelling infection dynamics in populations of different sizes or periodically fluctuating densities, the shape of the assumed contact–density function may become highly important . It is therefore not surprising that studies on how transmission rates and contacts relate to density have mainly been conducted in wildlife [10,15–19]. The main approach in these studies has been to measure disease prevalence in a field or experimental setting in which densities are manipulated or vary naturally, followed by fitting models with different transmission–density functions to the data.
|
study
| 93.44 |
There has traditionally existed a focus on whether transmission is frequency- or density-dependent, and rarely are other, nonlinear transmission–density functions investigated . Nevertheless, it has been well established that the binary distinction between density independence and linearity can be inadequate, and a range of other possible nonlinear transmission functions has been suggested, often in the power law, asymptotic or logistic family [10,22–25]. Cowpox virus dynamics in a natural population of field voles (Microtus agrestis), for example, has been shown to be best described by a nonlinear power function intermediate between density independence and linearity . Intermediate density-dependence was also observed in Ambystoma tigrinum virus transmission in the tiger salamander .
|
study
| 99.9 |
Here, we want to investigate whether, and in which situations, implementation of the exact shape of the transmission response function is important. Although it would be possible to mathematically model the effects of different transmission–density functions for any hypothetical combination of demographic pattern and contact–density function, the sheer number of possible functions for each demographic situation would make it almost impossible to decide which functions are biologically relevant. To inform such models in a meaningful way, we need biological background data, i.e. a species for which population dynamics, a contact–density or transmission–density function, and disease dynamics have been quantified, but until recently no such data were available. Using a combination of data from recent experiments in which we quantified contact rates across a wide range of population densities in the rodent Mastomys natalensis and the infection parameters of Morogoro virus (MORV) in this rodent , we tested the effect of different transmission functions using a simple SIR (susceptible, infectious, recovered) transmission model in annually fluctuating host populations. By implementing a range of hypothetical combinations of infectious period, transmission rate and population size, we assessed what the effects of the contact–density function would be for infections with different characteristics.
|
study
| 100.0 |
Natal multimammate mice (M. natalensis) occur throughout sub-Saharan Africa, and are an important agricultural pest species and natural reservoir hosts for several microparasites that cause disease in humans, including Yersinia pestis (bubonic plague), leptospirosis and several arenaviruses including Lassa virus, which can cause severe haemorrhagic fever in humans [28–32]. Multimammate mouse demography has been studied thoroughly, which allows us to create a simple but accurate demographic model that will serve as a basis for transmission modelling (see below). In Tanzania, where most of the studies on its population ecology have been conducted, M. natalensis exhibits strong annual population fluctuations, with densities ranging from 10 ha−1 in the breeding season to greater than 300 ha−1 outside the breeding season . Importantly, we have recently quantified a contact–density relationship for this species, which provides us with a realistic biological background for fitting the transmission functions .
|
study
| 100.0 |
The simulated infection dynamics in this study are based on those of MORV, which naturally occurs in M. natalensis in Tanzania, and of which the transmission ecology and patterns of infectivity have been documented in detail; it has a latent period of about 3 days between infection and excretion (which we here ignored for simplicity), an infectious period of 30–40 days, and presumably lifelong immunity. MORV transmission can therefore be modelled using a simple SIR model (described below).
|
study
| 100.0 |
We investigated the effect of four different contact–density functions using stochastic MORV transmission models. While transmission in these models is stochastic, demography was implemented deterministically because this allows us to focus entirely on the effects of stochasticity in transmission, and because it reduces computation time.
|
study
| 99.94 |
The seasonally fluctuating densities of M. natalensis were modelled using a seasonal birth-pulse function, B(t)=k exp[−s cos2(πt−φ)], as described in Peel et al. . This is a flexible function in which a synchrony parameter (s) determines the length of the birth period, and another parameter (φ) determines the timing of the birth period. Parameter k ensures that the annual population size remains the same, by compensating the number of births for the (constant) mortality rate μ . Note that this means that while the annual population size is constant, population densities do fluctuate seasonally because all births happen within a few months, whereas deaths happen throughout the year. Function parameters were fitted visually to a 20-year dataset of monthly population densities of M. natalensis in Tanzania ( and more recent unpublished data (H Leirs, up to 2017); electronic supplementary material, figure S1–1). This deterministic demography was used as a basis for modelling demography in the stochastic SIR model (described below) in order to avoid the influence of stochastic changes in population density, and because this reduces the number of simulations that need to be performed.
|
study
| 100.0 |
A standard SIR (susceptible, infectious, recovered) model was used to simulate MORV transmission , described by the following set of coupled ordinary differential equations: dSdt=B(t) N−βSIN−μ S,dIdt=βSIN−γ I−μ IanddRdt=γ I−μ R, where B(t) is the time-dependent birth function described earlier, μ is the mortality rate, γ is the 1/infectious period and β = cp is the transmission coefficient, which consists of p (rate at which S becomes I when in contact with an I individual) and contact–density function c = f(N/A) that can acquire different shapes depending on population density (explained below).
|
study
| 100.0 |
In order to compare invasion and persistence probabilities between different contact–density functions, we used a stochastic discrete time version of this SIR model, where the transition rates between categories were modelled stochastically, resulting in two possible stochastic events: infection (decrease of S, increase of I) and recovery (decrease of I, increase of R). Events were assumed to occur continuously in time at a certain rate, and were modelled using the ‘adaptive tau-leap’ algorithm described in . Briefly, each short time-step δt, the number of events of each type (infection or recovery) that occurs is randomly drawn from a Poisson distribution with mean riδt, where ri is the rate of each type of event i. If the number of simulated events would cause any of the categories (S, I or R) to fall below 0, δt is halved and new events are drawn (=‘adaptive tau-leap’).
|
study
| 100.0 |
The model started at t = 0 and one infected individual was introduced after 1 year, at t0 = 1 (I → 1) in order to allow the initial population dynamics to stabilize. This infected individual was considered to be at the start of the infectious period upon introduction. Prior to performing the main analyses, the effect of introduction time was analysed. Different introduction times had an effect only on the linear and sigmoid functions, where they resulted in lower invasion probabilities between t0 values of 1.2 and 1.4, which was probably a result of the low population densities (electronic supplementary material, figure S2–1). There was no effect of introduction time on disease persistence (electronic supplementary material, figure S2–2).
|
study
| 100.0 |
The core of this study is the implementation of four different, biologically relevant contact–density functions c = f(N/A) (figure 1): (a) Constant function (or ‘frequency-dependence’) c=a1 (N/A)0, with fitting parameter a1. This function is independent of density, and typically (but not only, e.g. vector-borne infections ) used in the case of sexually transmitted infections where the number of sexual contacts is not expected to change with population density .(b) Linear function (or ‘density-dependence’) c=a2 (N/A), with fitting parameter a2. A linear function is typically used when assuming random mixing where (infectious) contacts increase linearly with population density .(c) Power function c=a3(N/A)0.5, with fitting parameter a3. This function has been used as an ‘intermediate’ between frequency- and density-dependence . A power of 0.5 was used in order to get the intermediate saturating shape, which was the most informative shape for the simulations. The power function contact rates increase at low densities, while the slope decreases at higher densities towards an asymptotic limit. This shape has been observed for contact rates in brushtail possums and elk [43–45], and has been shown to be a better predictor of cowpox transmission patterns than frequency- or density-dependence .(d) Sigmoid function c=a5/(1+ea6 ((N/A)−a7)) with fitting parameters a5, a6 and a7. This function has a minimum, constant number of contacts at low densities, after which contact rates increase with density until reaching a plateau when reaching a maximum number of contacts. This shape has been observed for multimammate mice contacts , and has been proposed previously as a biologically plausible function . Figure 1.Contact–density functions fitted to experimental data from Borremans et al. , showing mean degree (the number of individuals one focus individual contacted) for a range of population densities (number of animals per ha = N/A).
|
study
| 100.0 |
(a) Constant function (or ‘frequency-dependence’) c=a1 (N/A)0, with fitting parameter a1. This function is independent of density, and typically (but not only, e.g. vector-borne infections ) used in the case of sexually transmitted infections where the number of sexual contacts is not expected to change with population density .
|
other
| 99.44 |
(c) Power function c=a3(N/A)0.5, with fitting parameter a3. This function has been used as an ‘intermediate’ between frequency- and density-dependence . A power of 0.5 was used in order to get the intermediate saturating shape, which was the most informative shape for the simulations. The power function contact rates increase at low densities, while the slope decreases at higher densities towards an asymptotic limit. This shape has been observed for contact rates in brushtail possums and elk [43–45], and has been shown to be a better predictor of cowpox transmission patterns than frequency- or density-dependence .
|
study
| 100.0 |
(d) Sigmoid function c=a5/(1+ea6 ((N/A)−a7)) with fitting parameters a5, a6 and a7. This function has a minimum, constant number of contacts at low densities, after which contact rates increase with density until reaching a plateau when reaching a maximum number of contacts. This shape has been observed for multimammate mice contacts , and has been proposed previously as a biologically plausible function .
|
study
| 99.94 |
Considering that β = cp, after fitting contact parameter c(N/A) to contact–density data for the four functions, a transmission rate p had to be determined before being able to compare the effect of the different transmission–density functions. This is exactly equivalent to the situation in which real data are known, but the underlying transmission–density function is not. In such a situation, different transmission–density functions are implemented in a model and fitted to the data. For each contact function c, a different transmission rate p must be fit in order for the transmission–density function to match the real data as closely as possible. If instead we were to use the same transmission rate p for each function, we would, in fact, be comparing entirely different transmission dynamics. Fitting p is typically done using a certain important characteristic of the real data, such as average outbreak size or annual cumulative incidence. For this purpose, we implemented a function-specific constant (qi) that was fitted for each function i to a common characteristic. Out of numerous possible transmission dynamic characteristics to choose for fitting, we opted for one that ensured that β, summed across the probability distribution of population densities occurring during 1 year in a simulated, deterministic model of demography, was the same for each contact function. Formally, this meant that: β=qi×Σj=1300fc(N/A)j×h(N/A)j, where fc is the contact–density function, j is the population density and h is the frequency distribution of densities in a year. We chose this method because it has the advantage of not selecting for certain outbreak characteristics such as prevalence or outbreak size. We nevertheless also examined the effect of using two alternative fitting methods. The first alternative method fits qi so that a deterministic transmission model resulted in a maximum annual prevalence of 40%. The second alternative method fits qi so that the final annual number of infections was 2N0, i.e. twice (an arbitrarily chosen number) the initial number of individuals. While these alternative fitting methods resulted in highly different values for the constant (density-independent) function, the three other functions were always very similar (electronic supplementary material, table S3–1 and figure S3–1). As we are mainly interested in the differences between the three non-constant functions, we did not further investigate the effects of these alternative β-fitting methods.
|
study
| 100.0 |
The effects of the contact–density functions were investigated through a number of meaningful epidemiological parameters: (i) SIR dynamics, including prevalence (I/N); (ii) invasion probability, defined as the proportion of stochastic simulations in which the infection manages to survive the first year after introduction, conditional on having started successfully (successful start = infection persistence time > t0 + infectious period). This definition was chosen for simplicity, biological relevance and consistency with a previous model of MORV transmission ; and (iii) persistence probability, defined as the proportion of simulations in which the infection is still present at t = 10 years, conditional on having survived the first year.
|
study
| 100.0 |
Invasion and persistence were estimated under a number of conditions of population size (N0), infectious period (1/γ) and transmission rate (p), where for each combination of conditions 1000 simulations were run. While we modelled changes in population density for each combination of parameters, we also assessed the effect of population size because this is expected to affect the probability for the infection to disappear from the population, independent of density. In order to ensure that we here implemented the effects of population size and not density, population density N/A was calculated assuming that the area occupied when initial population size N0 = 100 is 1 ha, and that area increases linearly with increasing values of N0 (i.e. when initial population size increases, area also increases). This, for example, means that while population density fluctuates seasonally in exactly the same way for N0 = 100 and N0 = 1000 (or any other N0), the former assumes an occupied area of 1 ha and the latter an area of 10 ha. Because transmission events are stochastic, the size of the occupied area can have important consequences for virus invasion and persistence.
|
study
| 100.0 |
The four fitted contact–density functions were: c=1.14(N/A)0 (constant), c = 0.0092(N/A) (linear), c=0.124(N/A)0.5 (power), c=2.13/(1+e−0.05 ((N/A)−101.2)) (sigmoid). As shown in figure 1, a sigmoid contact–density function resulted in the best fit to contact–density data measured for M. natalensis in a previous field experiment , while the constant function clearly did not fit the data well. AIC values for the functions were 59.7 (constant), 39.7 (linear), 42.6 (power), 34.3 (sigmoid). Akaike weights were 0.00 (constant), 0.06 (linear), 0.01 (power), 0.92 (sigmoid), providing additional support for the much better fit of the sigmoid function.
|
study
| 100.0 |
Because we are interested in the qualitative effects of the contact–density functions rather than in detailed differences that are specific to the model system, we here report the results with a focus on the qualitative aspects of SIR dynamics and prevalence. The constant function resulted in a relatively low epidemic peak during the breeding season, while the other three functions showed a clear epidemic peak where prevalence (proportion Infecteds; figure 2, red curves) increased sharply for the linear and sigmoid functions, but more gradually for the power function. Median peak prevalence for the four functions: 45% (constant), 68% (linear), 60% (power) and 72% (sigmoid). The dynamics of the proportion of Recovered (immune–antibody-positive; figure 2, blue curves) individuals were highly similar for the four functions, although they did differ in how low the Recovered proportion becomes at the end of the birth pulse: 15% (constant), 2% (linear), 14% (power) and 0.5% (sigmoid). There was a steeper build-up of Susceptibles (figure 2, green curves) for the linear and sigmoid than for the other two functions, due to the fact that transmission rates started increasing later in the year (at higher densities). Figure 2.SIR dynamics for the four different contact–density functions. For each simulation run in which there was successful persistence, the 8th year was retained (days 3030 to 3395). The figure shows all these outputs plotted on top of each other. The increase in Susceptibles corresponds with the birth pulse. Infectious period 1/γ = 60 days, transmission rate p = 50, initial population size N0 = 100 000.
|
study
| 100.0 |
SIR dynamics for the four different contact–density functions. For each simulation run in which there was successful persistence, the 8th year was retained (days 3030 to 3395). The figure shows all these outputs plotted on top of each other. The increase in Susceptibles corresponds with the birth pulse. Infectious period 1/γ = 60 days, transmission rate p = 50, initial population size N0 = 100 000.
|
study
| 99.94 |
Invasion and persistence probabilities were investigated for a range of population sizes (N0), infectious periods (1/γ) and transmission probabilities (p). Note that while infectious period results are reported in absolute days, they can also be interpreted in relation to the demographic timescale used in the simulations (e.g. annual breeding, brief recruitment period), which will aid comparison with other pathogen–host systems in which host densities fluctuate .
|
study
| 100.0 |
Successful invasion and persistence were more often observed for the constant function than for the other functions (figures 3 and 4). Even at low population sizes, successful invasion was almost certain for infectious periods of 30 days and longer, and was even observed for an infectious period of 7 days in sufficiently large populations. By contrast, for the other functions, successful invasion was never observed below infectious periods of 7 days, and even with an infectious period of 30 days invasion was rare for the sigmoid function. For the linear function, invasion probabilities were lower than for the power function, whereas persistence probabilities for these two functions were similar. The sigmoid function resulted in the lowest invasion or persistence success, where persistence was rarely observed for an infectious period of 30 days. Even when the infectious period was 60 days, persistence was observed only when population size was sufficiently large (e.g. 50% for a population size of 5000). Figure 3.Invasion probabilities for the different contact–density functions, for a range of infectious periods 1/γ and initial population sizes N0 (transmission rate p = 50). Simulations were conducted for all values indicated by tick marks on the axes, and results are interpolated between these values for illustration. Figure 4.Persistence probabilities for the different contact–density functions, for a range of infectious periods 1/γ and initial population sizes N0 (transmission rate p = 50). Persistence probabilities were calculated using only simulation runs in which there was successful invasion. Simulations were conducted for all values indicated by tick marks on the axes, and results are interpolated between these values for illustration.
|
study
| 100.0 |
Invasion probabilities for the different contact–density functions, for a range of infectious periods 1/γ and initial population sizes N0 (transmission rate p = 50). Simulations were conducted for all values indicated by tick marks on the axes, and results are interpolated between these values for illustration.
|
study
| 90.25 |
Persistence probabilities for the different contact–density functions, for a range of infectious periods 1/γ and initial population sizes N0 (transmission rate p = 50). Persistence probabilities were calculated using only simulation runs in which there was successful invasion. Simulations were conducted for all values indicated by tick marks on the axes, and results are interpolated between these values for illustration.
|
study
| 99.94 |
The effect of transmission rate on invasion and persistence was similar between the constant, linear and power functions, although the constant function still resulted in more successful invasion/persistence at lower transmission probabilities (electronic supplementary material, figures S4–1 and S4–2). The sigmoid function, however, was more affected by transmission probability than the other three functions, where persistence was only possible for a combination of high transmission rate and large population size (electronic supplementary material, figure S4–2).
|
study
| 100.0 |
Population densities often vary in space or time, and this can influence parasite transmission. Using data-driven models of virus transmission in a fluctuating population, we found that the way in which the transmission–density function is modelled can have important consequences for estimating invasion and persistence success.
|
study
| 99.94 |
While SIR dynamic patterns for the constant function were distinct from those of the three density-dependent functions, these three functions did not result in highly different SIR dynamics. Infection prevalence (the proportion of Infecteds) and seroprevalence (the proportion of Recovereds) patterns were very similar for the linear and sigmoid functions. For the power function, seroprevalence and infection prevalence patterns were slightly smoother, with less pronounced peaks, than for the linear and sigmoid functions. Invasion and persistence were clearly affected by the shape of the transmission–density function. They were always higher for the constant function than for the other functions. The sigmoid function resulted in the lowest invasion and persistence probabilities, and was most sensitive to population size, length of the infectious period and transmission rate. Invasion and persistence success for the linear and power functions were intermediate between the constant and sigmoid functions. Depending on the time at which the infection is introduced, the differences in invasion and persistence probability between the contact functions can become even more pronounced.
|
study
| 100.0 |
The different consequences of the contact–density functions can probably be attributed to a number of key differences in their shapes. Considering that the infection is most sensitive to extinction during periods of low population density, the size of transmission coefficient β at low densities will be a highly influential factor. An important consequence of this is that larger population sizes are necessary for successful disease invasion/persistence when β is low during low-density periods. In our case, for example, a minimum population size of 10 000 (equivalent to a 100 ha area) was necessary for a 50% persistence success rate for the power function (30-day infectious period), while this was 50 000 for the linear function and larger than 100 000 (not tested) for the sigmoid function. Knowledge of contact rates at low population densities is therefore critical when estimating invasion and persistence thresholds. A second important factor is the rate at which β increases with density. The epidemic peak will be more pronounced when there is a strong increase of β with density (e.g. the sigmoid function at intermediate densities, but also the linear and power functions). The sigmoid function, for example, results in a steep increase in transmission rates during the juvenile recruitment season as soon as a threshold density of Susceptibles is reached (here around 80–100 ha−1).
|
study
| 100.0 |
When considering both the SIR dynamics and invasion/persistence of the three density-dependent functions, a contrast emerges than can have significant implications for fitting a β-density function to epidemiological data: while there was a clear effect of function shape on invasion and persistence, SIR dynamic patterns were less distinct. This means that it could be difficult to discern between different contact–density functions when fitting model parameters to real, inherently noisy data . Nevertheless, because the functions do introduce different invasion and persistence probabilities, it will, in some situations, be crucial to implement the correct function. Ideally, this choice is based on the quantified contact–density or transmission–density relationship of the host/infection system that is being studied, but such data are rarely available.
|
study
| 99.94 |
A possible solution for this problem could be to establish/use general links between certain biological traits and contact and transmission patterns . The shape of the transmission–density function is determined by a combination of infection and host characteristics, so based on these characteristics, it should theoretically be possible to a priori predict the shape of the function, at least roughly. Knowledge of density-dependent changes in home range size and overlap could, for example, be a useful proxy for the contact–density function. For male brushtail possums (Trichosurus vulpecula), it has been established that contacts increase with density according to a positive power function , which fits with the fact that this species is not territorial, and with the observation that home ranges are larger at low densities which may result in the maintenance of contacts. Such an inverse correlation between home range size and density was also observed for M. natalensis , and this may have similar results on contact rates at low densities, as the maintenance of contacts even at very low densities was also observed for this species . This pattern would be expected to be different for territorial species. In an enclosure experiment, movements of meadow voles (Microtus pennsylvanicus), which are strongly territorial, decrease significantly with density , and although the effect of density on contacts was not measured, it is not unlikely that this decrease in movement distance corresponds with a contact–density function that does not increase, or at least not linearly. As a final example, consider the experimental study of the transmission of the parasitic mite Coccipolipus hippodamiae in populations of the two-spot ladybird (Adalia bipunctata) . The mite is transmitted sexually, and although sexual contacts are typically assumed to be frequency-dependent, the authors observed that the transmission–density function was closer to linear density-dependence and therefore concluded that the common assumption that sexual transmission is frequency-dependent is not always true. Their study species (A. bipunctata), however, is known to be highly promiscuous, which means that sexual contacts are not limited to one or a few mates, but instead increase with density. A priori use of this knowledge about host and infection biology would have resulted in the more accurate prediction that sexual transmission of C. hippodamiae is density- rather than frequency-dependent.
|
review
| 96.94 |
We would like to note that the patterns observed in this modelling study may change with different transmission characteristics. A longer latent period, for example, might increase invasion and persistence probabilities. Likewise, if immunity is not lifelong, but individuals can instead return to the susceptible class, the infection might persist for longer periods. While such changes might cause the differences between contact–density functions to be more or less outspoken, we doubt that they would change the key patterns observed in this study.
|
study
| 99.94 |
Many wildlife species experience seasonal birth pulses and density fluctuations, and while it has been established that birth pulses can have strong effects on disease transmission , we now see that the shape of the transmission–density function can have further significant effects on disease invasion and persistence. The implementation of the transmission–density function should therefore be done with care, and as informed as possible. Although currently few studies have quantified the relationship between contacts and density, other relevant knowledge about host biology and behaviour can potentially be used for deciding on the best possible shape of the transmission–density function.
|
other
| 90.2 |
The Cucurbitaceae family is the second most large horticultural family in terms of economic importance after Solanaceae. It includes several important crops, such as melon (Cucumis melo), watermelon (Citrullus lanatus), cucumber (Cucumis sativus) and many Cucurbita species with edible fruits (Jeffrey, 1980). The genus Cucurbita (2x = 2n = 40), originated in the Americas, encompasses three economically important crop species such as Cucurbita pepo, Cucurbita moschata, and Cucurbita maxima, cultivated throughout temperate, sub-tropical, and tropical regions (Wang et al., 2011). Cucurbita pepo includes a wide assortment of varieties and cultivars, known for their unique fruit shape and color and appreciated for their culinary properties. Among different species of this genus, Cucurbita pepo have the greatest monetary value (Paris, 2008). Botanical classification based on allozyme variation recognized three subspecies in this species including: pepo, ovifera (syn. texana), and fraterna. Paris (1986) classified edible-fruited C. pepo into eight cultivar-groups: Acorn, Crookneck, Scallop, and Straightneck that belong to subsp. ovifera and Pumpkin, Zucchini, Cocozelle, and Vegetable Marrow that belong to subsp. pepo (Paris, 2010). The genome size of Cucurbita spp. is approximately 500 Mb (Arumuganathan and Earle, 1991). Recently, a high-quality draft of C. pepo (subsp. pepo cultivar-group Zucchini) genome with a sequences length of about 265 million base pairs (Mbp) was made available on CucurbiGene database as well as several C. pepo transcriptomes have been explored (Blanca et al., 2011; Wyatt et al., 2015; Vitiello et al., 2016; Xanthopoulou et al., 2016, 2017; Montero-Pau et al., 2017). However, still little is known about the genetic diversity of this noteworthy crop and even less has been done to explore its proteome. High-throughput sequencing of transcriptomes has opened the way to study the genetic and functional information stored within any organism at an unprecedented scale and speed.
|
review
| 99.9 |
Transcriptome generation through RNA sequencing (RNA-seq) is a technology that can be used in the high resolution and broad dynamic range gene expression studies and in the simultaneous understanding of the genes function (Wang et al., 2009). Basically, the protein-coding genes function is inferred by the analysis of structure, function and evolution of the proteins they encode (Guo, 2013). For the characterization of unannotated proteins, can result particularly useful to undertake orthology analysis. Proteome data are important resources for having an overall genome vision but at the same time achieving a high level of accuracy in comparative studies (Andolfo et al., 2014a). To this end, we sequenced and assembled the first transcriptome of zucchini cultivar “True French,” founder of important pathogen resistant commercial varieties and to harness the full potential of such data we performed also an high-quality proteome annotation. A total of 33,966 protein sequences were predicted, functionally annotated and compared to cucumber, melon, watermelon and Arabidopsis proteomes. In addition, disease resistance (R) gene family was finely characterized and several specie-specific R-genes expansion was detected in C. pepo.
|
study
| 100.0 |
Plants of Cucurbita pepo subsp. pepo cultivar-group Zucchini, variety True French, were grown in greenhouse facility at Department of Agricultural Science of University of Naples “Federico II” using standard horticultural practices. C. pepo cv. True French tissue samples were collected from young plants of about 10 cm high. Total RNA was isolated from ground, frozen leaf tissues using the SpectrumTM Plant Total RNA Kit (Sigma-Aldirch). A complete removal of traces of DNA was performed using On-Column DNase I Digest Set (Sigma-Aldirch). Quantity and integrity of the extracted total RNA were determined using NanoDrop ND-1000 Spectrophotometer (Thermo Fisher Scientific Inc., USA), on a denaturing formaldehyde gel and Agilent 2100 bioanalyzer (Agilent Technologies, USA) respectively, to be RIN > 8. Library preparation and sequencing were performed by the Genomix4Life S.r.l., spin-off of Salerno University. The sequencing library was prepared using the TruSeq RNA Sample Preparation Kit v2 (Illumina, San Diego, CA, USA) and paired-end reads of 100 bp were sequenced from the three independent samples on one lane of an illumine HiSeq 2000.
|
study
| 100.0 |
The quality control checks on raw sequence data (75,22 millions of paired reads totalling 15 e12 bp) from all the three data sets was performed using FastQC (Andrews, 2010). Raw reads were filtered to remove the adapter sequences and the poorer quality regions with sequence pre-processing tool, Trimmomatic (Bolger et al., 2014). Paired-end read duplicates from the PCR amplification step in the sequencing process were removed and only those reads with a mapping score ≥ 30 were kept in the alignments. The high quality reads were aligned against the C. pepo reference genome sequence version 3.2 (https://cucurbigene.upv.es/) with STAR aligner (version 2.4.0j). The resulting alignment was used as input to Cufflinks (version 2.2.1) for transcript assembly. PASA pipeline (version 2.0.2) was used to combine Cufflinks results with the public transcriptome version 3.0 (https://cucurbigene.upv.es/).
|
study
| 100.0 |
The proteome functional annotation was performed through a match search against four database (TAIR10, SWISS-PROT, TrEMBL and GenBank-NR) using DIAMOND in sensitive mode with a cut-off e-value of 1 e−5 (Buchfink et al., 2015). To add information about protein function to our proteome, a Blast2GO (Conesa et al., 2005) annotation, using default parameters, were conducted. Finally, the zucchini proteome was scanned with InterProScan v.5.13 (Jones et al., 2014) against the InterPro protein signature databases to identify and finely characterize plant resistance proteins.
|
study
| 100.0 |
To identify orthologous gene groups among C. pepo, C. melo, C. sativus, C. lanatus and A. thaliana we used OrthoMCL software with default settings. The association between reference R-genes (http://prgdb.crg.eu/) and relative orthologous group (OG) was detected using Best BLAST Hit method (BlastP, E < 1 e−5) and the output was filtered for a query coverage and identity percentage, both >50%.
|
study
| 100.0 |
The sequencing produced a total of 69,5 millions of clean paired reads, obtaining 13,9 e12 bp of RNA-Seq data for C. pepo (Supplementary Table 1). Transcriptome assembly yielded 68,720 transcripts, with mean length of 1,534 bp. The transcripts were translated and a high-quality proteome of 33,966 primary protein sequences, with mean length of 316 AA, were obtained. DIAMOND similarity-based searches were performed against the publically available databases (SWISS-PROT, TrEMBL, TAIR10, and GenBank-NR) to annotate C. pepo proteome (Supplementary Table 2). About 85% of proteins encoded by genes had homology with four principal databases and over 75% were functionally annotated (Table 1). In addition, a GO-annotation using Blast2GO were effected and a total of 256,138 GO-terms were assigned to about 65% (21480) of the predicted proteins (Supplementary Table 3).
|
study
| 100.0 |
A fine characterization of genes encoding domains similar to plant resistance (R) proteins, in C. pepo proteome was conducted. R-proteins can be categorized according to the presence and organization of protein domains, such as Toll/Interleukin-1 receptor (TIR), coiled coil (CC), the nucleotide-binding site (NBS), leucine-rich repeats (LRRs). A total of 64 R-proteins (also called NLR, NB-LRR, NBS-LRR, or NB-ARC-LRR proteins) were identified (Supplementary Table 4). The CNL (Coiled coil, Nucleotide-binding site, Leucine-rich repeats) class was divided into sub-classes based on sequence similarity with the canonical CNLs that contain an EDVID amino-acid motif, and the RPW8-like proteins (Andolfo et al., 2014b). Interestingly, an expansion of RPW8-NLR genes (11 out of 64) in C. pepo was discovered. Diversely, C. melo, C. sativus, and C. lanatus presented only three RPW8-NLRs for each species (Figure 1A). It is now well-known that RPW8-NLRs can function as helper NLRs for well-defined NLR–mediated resistance responses. Thus, they may enhance the C. pepo defense system to offset its reduced number of NRL receptors available (Sanseverino and Ercolano, 2012). In addition, C. pepo RPW8-NLRs showed a very high homology to ADR1 (activated disease resistance 1), R-gene that confer resistance again Erysiphe cichoracearumi, the causal agent of Powdery Mildew (PM) in A. thaliana (Micali et al., 2008). PM disease, caused by Podosphaera xanthii (syn. Sphaerotheca fuliginea) has an important economic impact on C. pepo varieties. ADR1-like proteins expansion, identified in C. pepo, could suggest an adaptive diversification induced by specie-specific pathogen pressure (Andolfo and Ercolano, 2015).
|
study
| 100.0 |
RPW8-NLR genes expansion and comparative analysis. (A) Evolutionary history of RPW8-NLR proteins annotated in Cucurbita pepo. Full NB domains (PF00931) of 26 RPW8-NLR proteins were used together with 3 well characterized reference genes (short gene name: ADR1, NRG1, and APAF1) to do a maximum likelihood analysis based on the Jones et al. w/freq. model. Model with the lowest BIC score (Bayesian Information Criterion) was considered to describe the substitution pattern the best. Sequence similarities were determined performing a ClustalW (default settings) multiple alignment. Evolutionary analyses were conducted in MEGA7. Labels show the bootstrap values higher than 50 (out of 100), are indicated above the branches. The tree is drawn to scale, with branch lengths proportional to the number of substitutions per site. Species to which belong sequences are indicated by colored spots. (B) Venn diagram of genes (Gs) clustered into orthogroups (OGs). Five species (Cucurbita pepo, Citrullus lanatus, Cucumis melo, Cucumis sativus, and A. thaliana) were used to generate the Venn diagram. In the graph were reported the number of specie-specific and common OGs.
|
study
| 100.0 |
A comparative analysis among C. pepo, C. melo, C. sativus, C. lanatus, and A. thaliana were performed to obtain functional information on our proteome. A total of 18,742 orthologous groups (OGs), which included 107,386 sequences, were identified (Supplementary Table 5). 9,453 OGs enclosed 69,982 sequences and were highly conserved in all analyzed genomes (Figure 1B). A core Cucurbitaceae proteins (9,465 sequences) clustered in 2099 OGs were detected. About 65% (22,214) of C. pepo predicted proteins were grouped in 13,953 OGs (Figure 1B). Several C. pepo gene family expansions associated to transmembrane R-genes were discovered. One hundred zucchini proteins, annotated as Receptor-like Kinase (RLK) and Receptor-like Protein (RLP) were clustered in five OGs (OG_00004, OG_00027, OG_00038, OG_00053, OG_01889, and OG_00579) and associated a well-characterized R-genes (Supplementary Table 5). Probably the expansion of cell surface receptors (RLKs and RLPs) and relative strengthening of first defence line represent adaptive dynamics to balance the limited number of cytoplasmic receptors (NRLs) (Andolfo and Ercolano, 2015). The C. pepo gene family expansions could be associated to the cucurbit-common tetraploidization recently identified by Wang et al. (2017). Furthermore, we identified a number of gene families related to important agronomical traits. Fourteen OGs related to OVATE gene family grouped 19 zucchini genes. OVATE is an important locus for fruit shape determination and plant development (Rodríguez et al., 2011). We identified three zucchini ortholog genes to PSY1 and PSY3 melon genes that putatively involved in carotenoid metabolism and fruit ripening (Garcia-Mas et al., 2012). The Cup000085g037789 is the ortholog gene of OR, a cloned gene that governs the fruit flesh colur in melon and in other important crops (Tzuri et al., 2015). Comparative analysis of C. pepo proteome can be used to identify orthologous genes for functional study. Our dataset represented a very important resource to reduce the plant breeding work for the identification of candidates for important agronomical traits (Supplementary Table 5).
|
study
| 99.94 |
The dataset submitted to NCBI include the raw read sequences of three biological replicates of Cucurbita pepo subsp. pepo cultivar-group Zucchini, variety True French, in FASTQ format. The raw reads of C. pepo can be accessed at NCBI with the following BioSample accession number: SAMN07426850 (www.ncbi.nlm.nih.gov/Traces/study/?acc=SRP114337). The C. pepo transcriptome annotation, in GTF format, and primary protein sequences in FASTA format can be accessed at FIGSHARE with the following link (https://figshare.com/s/8a083f60df238acdbc19). The Supplementary Material (Supplementary Tables 1–5) for this article can be found online at: (https://figshare.com/s/8a083f60df238acdbc19). Users can download and use the data freely for research purpose only with acknowledgment to us and quoting this paper as reference to the data.
|
other
| 99.7 |
GA was chiefly involved in data analysis, results interpretation and manuscript writing. AD was mainly involved in data analysis, results interpretation and manuscript writing. RD drafted the manuscript. AE provided a critical reading of the manuscript. RA assembled the transcriptome. ME coordinated the project and contributed to data analysis and results interpretation. All of the authors read and approved the final manuscript.
|
other
| 99.94 |
Leukocytes, or white blood cells (WBCs), can provide access to a veritable treasure-trove of information on the health status of each individual. For example, leukocytes can serve as indicators of pathogen infection1, especially for leukocyte-specific pathogens such as the human immunodeficiency virus (HIV). Additionally, during T-cell adoptive therapy isolated leukocytes can be expanded ex-vivo for re-targeting and re-infusion into an individual2. An important prerequisite for many leukocyte assays is the depletion of contaminating erythrocytes, or red blood cells (RBCs), as well as the isolation of specific phenotypes. Established methods are typically based on centrifugation, fluorescence-activated3, 4, and magnetic-activated cell sorting5–7. These methods are often costly, require relatively large samples, and expose leukocytes to harsh chemical and physical treatment, which may cause cell lysis, contribute to undesirable phenotypic changes8, and corrupt the information obtained from these cells9.
|
review
| 99.8 |
Recently, a number of microfluidic methods have been developed for phenotypic separation of leukocytes from small volumes of whole blood. Leukocytes may be separated based on size and deformability using hydrodynamic chromatography, such as deterministic lateral displacement (DLD) or margination. In DLD, cells navigate a 2D array of obstacles along different streamlines based on their distinct size and deformability10–12. Margination relies on preferential migration of leukocytes along a vessel wall and clustering of RBCs in the middle of a vessel, resulting from the hydrodynamic lift generated through the Fåhræus-Lindqvist effect13–16. Alternatively, leukocytes may be separated based on distinct dielectric properties, which arise from cell size as well as nuclear and membrane morphology17, 18, and magnetophoresis, which depletes deoxygenated RBCs based on their specific magnetic susceptibility19–21. A significant challenge for separating leukocytes from whole blood using these approaches is the lack of selectivity both in the depletion of RBCs, as well as in the selectivity for specific phenotypes.
|
review
| 99.75 |
A potentially simple and selective approach to isolate leukocytes from whole blood is microfiltration. Since RBCs can deform through openings (1–2 µm) much smaller than their diameter (~8 µm), while circulating leukocytes are spherical with diameters 6–8 μm22, microfilters consisting of pillars23, 24, weirs23, 25, 26, and membrane pores23, 27 have been developed to separate these leukocytes from RBCs. A key limitation in directly processing whole blood using microfiltration is the potential for cells to clog and foul the filter microstructures. Clogging of the filter matrix alters its hydrodynamic resistance and the filtration force on each cell, resulting in a significant reduction in selectivity within the filters. Cross flow filtration can be employed to divert trapped cells using a secondary flow tangential to the filter surface23, 28–31 but does not completely release the cells adsorbed within the filter microstructures and significantly reduces the overall selectivity as well as throughput.
|
study
| 98.1 |
In this study, we employed the microfluidic ratchet mechanism to perform perpetual microfiltration of leukocytes from whole blood. Building on results from cell deformability studies performed using AFM32, micropipette aspiration33, 34 and its microfluidic derivative35, we hypothesize that erythrocytes, leukocytes and leukocyte subpopulations can be sensitively separated based on their distinct cell deformability. Previously, the microfluidic ratchet mechanism has been shown to be able to effectively separate viable circulating tumor cells from whole blood36 with significant improvements in yield over conventional methods as well as to separate erythrocytes infected with Plasmodium falciparum 37 from uninfected erythrocytes to improve the sensitivity of malaria diagnosis. Using oscillating flow of cells through an array of funnel shaped microstructures, we show that whole blood can be processed without clogging or fouling the filter matrix. Consequently, this process enables highly selective separation of leukocytes and erythrocytes, as well as demonstrates the potential for phenotypic sorting of leukocytes.
|
study
| 100.0 |
The microfluidic ratchet mechanism filters cells through a funnel-shaped micro-pillar array with minimum gap widths (or pore sizes) that gradually decrease from the bottom row to the top row of the array (Fig. 1A). The sample cells are infused at the bottom-left corner of the matrix, and transported through the array using a vertical oscillatory flow, at the same time as a constant horizontal cross flow. These flows combine to propel the cells in a zigzag diagonal path through the constriction matrix with each oscillation cycle, until they reach a limiting pore size that prevents further upward transport (Fig. 1A). Upon reaching this point, the cells oscillate between two funnel rows, allowing the horizontal flow to divert them toward a specific outlet.Figure 1Principle and design of the microfluidic ratchet cell sorting device. (A) Schematic illustration of the cell sorting region consisting of a 2D array of funnel microstructures with progressively smaller pore sizes from the bottom to the top row. Unprocessed whole blood is introduced at the bottom-left corner of the array while cells flow along a diagonal path under a combined biased oscillatory flow and cross flow. Smaller and softer RBCs flow into the outlets lining the top of the sorting region while the flow of leukocytes is restricted based on their size and deformability, which divert each cell into specific outlets. Colored arrows indicate the ratchet motions of erythrocytes and different subsets of leukocytes. (B,C) The underlying principle of the microfluidic ratchet involves selectively transporting cells using a ratcheting effect. Smaller and more deformable cells flow across tapered constrictions during forward flow and are prevented from returning during reverse flow. Larger and less deformable cells are trapped by the constriction and released with each reverse flow. (D) Image of the microfluidic device infused with different food color dyes illustrating components of the microfluidic device including the sorting region, sample inlet (SI), cross-flow inlet (CFI), oscillatory flow inlets (OSC1 and OSC2), and nine outlets (O1–O9).
|
study
| 99.94 |
Principle and design of the microfluidic ratchet cell sorting device. (A) Schematic illustration of the cell sorting region consisting of a 2D array of funnel microstructures with progressively smaller pore sizes from the bottom to the top row. Unprocessed whole blood is introduced at the bottom-left corner of the array while cells flow along a diagonal path under a combined biased oscillatory flow and cross flow. Smaller and softer RBCs flow into the outlets lining the top of the sorting region while the flow of leukocytes is restricted based on their size and deformability, which divert each cell into specific outlets. Colored arrows indicate the ratchet motions of erythrocytes and different subsets of leukocytes. (B,C) The underlying principle of the microfluidic ratchet involves selectively transporting cells using a ratcheting effect. Smaller and more deformable cells flow across tapered constrictions during forward flow and are prevented from returning during reverse flow. Larger and less deformable cells are trapped by the constriction and released with each reverse flow. (D) Image of the microfluidic device infused with different food color dyes illustrating components of the microfluidic device including the sorting region, sample inlet (SI), cross-flow inlet (CFI), oscillatory flow inlets (OSC1 and OSC2), and nine outlets (O1–O9).
|
study
| 99.9 |
The operation of this cell sorting mechanism relies on two key design principles: First, cells are deformed through an asymmetrical, funnel-shaped constriction, which we previously showed to require less pressure to transit cells in the direction of the taper (Fig. 1B) than against the taper (Fig. 1C)38. Second, cells are transported using a biased oscillatory flow that results in a net upward migration. The coupling of these two effects enables selective transport of cells based on their ability to deform through microscopic constrictions in a ratcheting manner. The ability to filter cells using oscillatory flow is essential to the perpetuation of the filtration process, as each reverse flow clears the funnel filters of non-specific adherent cells to eliminate clogging, which enables the processing of high-density samples like whole blood. Moreover, since cells come into contact with the funnel filters only momentarily during oscillatory flow, the hydrodynamic resistance of the filter, and consequently the filtration force experienced by each incoming cell, remains consistent during the entire process (supplemental information, Figure S1, Table S1). Finally, unlike traditional binary separation, the array-based sorting process allows a heterogeneous sample to be sorted into multiple fractions in a single run.
|
study
| 100.0 |
The fractionation of the cell sample into the outlets after sorting is determined by the dimensions and arrangement of the funnel pores within the sorting matrix. The funnel array consists of 35 rows and 630 columns, occupying an overall area of 1875 × 9525 µm. The funnel pore size is kept constant for each set of four rows and progressively decreases with each row from 8 µm at the bottom to 2 µm at the top of the sorting region. The gap between each row of funnels is 50 µm. As a result, the 35 rows of funnel array sort the input whole blood into 9 fractions in outlet 1 to 9 (O1–O9), with O1 corresponding to the most deformable fraction and O9 corresponding to the least deformable fraction. This geometry of the funnel constrictions is designed to deform cells laterally through 8, 6.5, 5.5, 4.5, 4, 3.5, 3, 2.5 and 2 µm pore sizes, while providing stress relief in the orthogonal direction, which is kept constant at a thickness of 11–15 µm. Fluid flow in the sorting matrix is manipulated through supporting microchannel networks, including a blood sample inlet (SI), cross-flow inlet (CFI) and oscillation flow inlets (OSC1 and OSC2), as illustrated in Fig. 1D. The design of the microchannel networks follows a hydrodynamic resistance model, which determines the filtration forces applied across the funnel filter and the flow rate inside the sorting matrix. A detailed hydrodynamic model of fluid flow in this microstructure is presented in the supplemental materials. Based on this prototype design, the microfluidic ratchets are able to process whole blood at a rate of 5 µl per hour. While this throughput is low compared to other biophysical cell separation methods10–12, 14, 15, the throughput of this device could ultimately be improved by increasing the filtration area, as well as by paralleling the mechanism.
|
study
| 100.0 |
Initially, the device was evaluated of its ability to separate leukocytes from the highly abundant erythrocytes. To distinguish leukocytes from erythrocytes, blood samples were pre-stained with Hoechst 33342 DNA stain and infused into the sorting region (Fig. 2A). As shown in Fig. 2B–E, the sample formed the characteristic diagonal trajectory. Individual leukocytes (blue) that have reached their limiting funnel constrictions were found to transit horizontally as expected (Fig. 2E and supplemental video). As a result, RBCs were collected in fractions O1-O3 (Fig. 2F; 2, 2.5 and 3 µm pores) while leukocytes were preferentially segregated among outlets O4-O9 (Fig. 2G; 3.5, 4, 4.5, 5.5, 6.5 and 8 µm pores). These results indicated that 3 µm is a suitable cut-off pore size for the separation of erythrocytes and leukocytes, which is similar to 3.5 µm pores used by other microfilters with pillar, weir, and membrane configurations23. It is important to note that staining the leukocytes required washing the cells, which removed the plasma component of blood. In this case, the blood cells were re-suspended at ~45% hematocrit. Both stained blood cells and unstained whole blood were tested and observed to behave equivalently.Figure 2Microfluidic ratchets to separate leukocytes directly from whole blood. (A) Image of the sorting region of microfluidic ratchets infused with food color dyes illustrating the diagonal trajectory of the sample inlet flow. (B–E) Images of the leukocytes stained with Hoechst 33342 being separated from the whole blood and traversing diagonally through the funnel matrix. (F,G) Bright field and fluorescence images of the outlets showing the separate reservoirs into which RBCs and leukocytes are collected.
|
study
| 100.0 |
Microfluidic ratchets to separate leukocytes directly from whole blood. (A) Image of the sorting region of microfluidic ratchets infused with food color dyes illustrating the diagonal trajectory of the sample inlet flow. (B–E) Images of the leukocytes stained with Hoechst 33342 being separated from the whole blood and traversing diagonally through the funnel matrix. (F,G) Bright field and fluorescence images of the outlets showing the separate reservoirs into which RBCs and leukocytes are collected.
|
study
| 94.06 |
The separation of leukocyte from erythrocytes was investigated as a function of oscillation pressure and funnel thickness, which varied from 14–20 kPa and 11–15 µm respectively. This operating range was established using previous experiments of acceptable operating conditions (see supplemental information). As shown in Fig. 3A–C, higher filtration pressures increased the frequency for cells to transit narrower openings, represented by a slight leftward shift of leukocytes’ outlet distributions for three different donors’ blood samples. Increasing the funnel thickness from 11 to 15 µm allows each cell to have greater vertical extension during deformation. However, as show in Fig. 3D–F, this change has minimal effect on the leukocyte distributions. RBCs were found exclusively in O1–3 in all scenarios except when the filtration pressure was decreased to 14 kPa, in which case, a small fraction of RBCs escaped into O4, leading to decreased leukocyte separation efficiency. Here, separation efficiency is defined as the pure fraction of leukocytes captured in outlets over the total number of leukocytes entering the sorting region (Fig. 4A). Since device thickness (11–15 µm) does not seem to have a strong influence on the efficiency of leukocytes separation from RBCs (Fig. 4B), we select a modest oscillation pressure of 17 kPa, as well as a device thickness of 11 µm to make the device robust to potential debris in the sample. Based on this configuration the microfluidic ratchet device is able to obtain a leukocyte separation efficiency of 98–99% (Fig. 4).Figure 3Leukocyte distribution at outlets as a function of filtration pressure applied at OSC2 and funnel thickness. (A–C) Leukocyte deformability profiles of three different donors at various filtration pressure with fixed funnel thickness of 11 µm. (D–F) Leukocyte deformability profiles of three donors at two different funnel thickness at a fixed filtration pressure of 17 kPa. Figure 4Performance of deformability-based separation of leukocytes from whole blood as a function of (A) filtration pressure and (B) funnel thickness. (P values determined using Student’s T-test).
|
study
| 100.0 |
Leukocyte distribution at outlets as a function of filtration pressure applied at OSC2 and funnel thickness. (A–C) Leukocyte deformability profiles of three different donors at various filtration pressure with fixed funnel thickness of 11 µm. (D–F) Leukocyte deformability profiles of three donors at two different funnel thickness at a fixed filtration pressure of 17 kPa.
|
study
| 100.0 |
To investigate whether deformability based sorting could separate granulocyte and lymphocyte subpopulations directly from whole blood, we pre-stained donor blood for CD3 and CD19 in order to identify the majority of lymphocytes (T cells and B cells) as well as for CD66b to identify granulocytes. Following microfluidic ratchet enrichment, leukocytes were distributed among O4–O9. Lymphocytes were preferentially distributed around O6 (retained between 4.5 to 5.5 µm pore size), where such cells can be obtained at 62–68% purity (Fig. 5A–F). Granulocytes were distributed around outlets O8 or O9 (retained between 6.5 to 8 µm funnel opening), where such cells can be obtained at 88–95% purity in these outlets (Fig. 5A–F). The images in Fig. 5G and H demonstrates that lymphocytes and granulocytes comprise the majority of the cells in O6 and O8 respectively. Based on their distribution into outlets, it was observed that leukocytes experience significantly less deformation than would be expected within smaller capillaries, such as pulmonary capillaries, which has circular constrictions that are ~5 µm in diameter39, 40. Specifically, lymphocytes (average diameter 7 µm) are typically blocked by the 4.5 × 11 µm constriction while neutrophils (average diameter 7.5 µm) are blocked by the 6.5 × 11 µm constriction.Figure 5Separation of lymphocytes and granulocytes from whole blood. (A–C) Deformability profiles of lymphocytes and granulocytes enriched from blood of three different donors. (D–F) Relative abundance of lymphocytes and granulocytes in each outlet. (G,H) Images of the immunostained lymphocytes (green) in O6 and granulocytes (red) in O8. (I) Comparison of distributions of leukocytes in O4–9 of three different samples: (1) whole blood without stain, (2) with Hoechst 33342 and (3) with anti-CD markers conjugated with fluorescence tags.
|
study
| 100.0 |
Separation of lymphocytes and granulocytes from whole blood. (A–C) Deformability profiles of lymphocytes and granulocytes enriched from blood of three different donors. (D–F) Relative abundance of lymphocytes and granulocytes in each outlet. (G,H) Images of the immunostained lymphocytes (green) in O6 and granulocytes (red) in O8. (I) Comparison of distributions of leukocytes in O4–9 of three different samples: (1) whole blood without stain, (2) with Hoechst 33342 and (3) with anti-CD markers conjugated with fluorescence tags.
|
study
| 100.0 |
To investigate whether the fluorescence stains or the staining process alters leukocyte deformability to affect their distribution after sorting, experiments were performed to compare the outlet distributions of whole blood samples without stain, with Hoechst 33342 stain, as well as with fluorescently conjugated markers using samples from the same donor. As shown in Fig. 5I, the leukocyte distribution is very consistent, which confirms that fluorescence staining has no effect on leukocyte deformability.
|
study
| 100.0 |
We investigated the intra- and inter-individual variability of granulocyte and lymphocyte distributions after sorting. The intra-individual variability for lymphocytes and granulocytes were established through four independent experiments on blood from a single donor, while the inter-individual variability was assessed using experiments on blood from three different donors. Both intra- and inter-individual distributions for lymphocytes were remarkably consistent, with ~60% retention in O6 (Fig. 6A–B). In contrast, granulocyte distribution varied significantly, both intra- and inter-individually (Fig. 6C–D). This result is not surprising as granulocytes are dominated by neutrophils, which represent more than 90% of these cells, and neutrophils are known to adopt different biophysical characteristics upon activation by chemical40 or mechanical stresses41.Figure 6Outlet distributions of lymphocytes and granulocytes. (A,B) Outlet distributions of lymphocytes from three donors. (C,D) Outlet distributions of granulocytes from three donors. Each inter-individual distribution profile data point is a mean of triplicate experiments. The error bars in Fig. 6B and D represent the standard deviation of results from 3 trials.
|
study
| 100.0 |
Outlet distributions of lymphocytes and granulocytes. (A,B) Outlet distributions of lymphocytes from three donors. (C,D) Outlet distributions of granulocytes from three donors. Each inter-individual distribution profile data point is a mean of triplicate experiments. The error bars in Fig. 6B and D represent the standard deviation of results from 3 trials.
|
study
| 99.94 |
Lastly, the distribution of monocytes was evaluated to determine whether they could be separated from lymphocytes and granulocytes, based on deformability. Monocytes make up 2–8% of the total leukocyte population. They average 7.5 µm in diameter22, making them slightly larger than either lymphocytes (6.2 µm) or granulocytes (7.0–7.3 µm). We sorted monocytes from whole blood after pre-staining with anti-CD14 marker and observed that they were retained primarily in outlets O8–O9, by the 6.5 µm constrictions (Fig. 7). This result suggests that these cells could be efficiently enriched from lymphocytes but not from granulocytes using the current version of the microfluidic ratchets device. The purity of monocytes in O8 range 4.5–20%, relative to granulocytes, which represents only a modest increase from the standard frequency of these cells in peripheral blood leukocyte population. However, the nearly exclusive accumulation of monocytes within these outlets suggests that future iterations of this device could resolve monocytes from granulocytes by optimization of the pore size (in the range between 6.5 and 8 µm) and funnel thickness.Figure 7Distribution of monocytes sorted from whole blood from the same donor in three trials. Inset graphs show the composition of leukocytes subsets in O8 and 9.
|
study
| 100.0 |
The microfluidic ratchet mechanism employed in this study overcomes some key limitations of both conventional microfiltration and microfluidic biophysical cell sorting systems to enable leukocyte separation from whole blood with 100% purity, as well as phenotypic sorting of leukocyte subpopulations. Microfiltration methods developed previously using pillars23, 24, weir23, 25, 26 and membrane pores23, 27 have all shown a tendency to clog, which increases hydrodynamic resistance, reduces selectivity and necessitates increased filtration pressure or periodic washing of the filter region. While secondary tangential flow partly alleviates device clogging23, 28–31, the oscillating flow used in microfluidic ratchet sorting enables clog-free and nearly perpetual processing of high-density whole blood samples. Since the deformation of cells is less than that which leukocytes normally experience when traversing through micro-capillaries in the body39, 40, this process does not adversely affect the viability of the leukocytes.
|
study
| 100.0 |
The performance of the microfluidic ratchet mechanism can be compared to other filtration methods using several metrics, including separation efficiency (the fraction of input leukocytes captured), leukocyte purity (the proportion of leukocytes in the output cell suspension), RBC depletion rate and overall sample throughput. In comparison with microfiltration, the key advantages of microfluidic ratchet sorting are the separation efficiency and leukocyte purity. While previous microfiltration methods typically report poor separation efficiency or fail to report this metric (Table 1), microfluidic ratchet consistently generates pure leukocyte isolates from whole blood with only 1–2% loss. The microfluidic ratchet also compares favorably with other biophysical separation (Table 2), including magnetophoresis, dielectrophoresis, and leukocyte margination. Magnetophoresis uses paramagnetic properties of RBCs to deplete these cells from the leukocyte suspension19–21. While RBC depletion using this approach exceeds 95%, the separated leukocytes remain at low relative purity because of the large number of starting RBCs (106/µl) in whole blood. Conversely, size-based sorting strategies, such as dielectrophoresis17, DLD10–12 and leukocyte margination14–16 can achieve superior leukocyte separation efficiency. However, a key advantage of the microfluidic ratchet mechanism is that it combines both size and deformability based selection to not only generate pure leukocyte suspensions but also to segregate leukocyte subpopulations, despite significant overlap in the sizes of different leukocyte subpopulations22.Table 1Performances of recent research in leukocytes separation from whole blood using microfiltration.Filtration Strategies and Filter Types*Performance MetricsLeukocyte Separation EfficiencyLeukocyte PurityRBC Depletion RateThroughputRequirement for Blood dilutionMicrofluidic Ratchet Current Study 98–99%100%100%5 µL hour−1 NoDead end; Pillar46 18–25%—(<1%)84–89%15–50 µL min−1 Yes (1:2 in PBS)Dead end; weir26 71%~10% (210-fold enrichment)n.d.4000 cells s−1 Yes (1:10 in PBS)Dead end; weir47 60%n.d.n.d.3–15 µL hour−1 NoDead end; pore27 >90%n.d.n.d.Only process 1.5 µL whole bloodNoDead end; pore23 72–85%n.d.n.d.Only process 200 µL whole bloodNoDead end; weir23 ~70%n.d.n.d.Only process <50 µL whole bloodNoDead end; pillar23 70–95%n.d.n.d.Only process 300 µL whole bloodNoCross flow; weir29 98%72%n.d.3.6 µL hour−1 YesCross flow; pore31 27.4 ± 4.9%93.5 ± 0.5%n.d.~17 µL min−1 NoCross flow; pillar 30 90%<1%90%10 µL min−1 NoCross flow; weir30 20%<1%40%10 µL min−1 Yes (1:100 in PBS)Cross flow; pillar23 70–95%n.d.n.d.20 µL min−1 NoCross flow; pillar28 97%<1%50%5 µL min−1 n.d.n.d. Data not available.*References in uppercase. Table 2Performances specifications of recent microfluidic research in leukocytes separation from whole blood using biophysical methods.Method*Performance MetricsLeukocyte Separation EfficiencyLeukocyte PurityRBC Depletion RateThroughputRequirement for Blood DilutionMicrofluidic Ratchet Current Study 98–99%100%100%5 µL hour−1 NoMagnetophoresis19 n.d.n.d93.5%5 µL hour−1 Yes (1:10 in PBS)Magnetophoresis20 n.d.n.d95%0.5–0.7 mL hour−1 Yes (1:20 in PBS)Magnetophoresis21 n.d.n.d93.7%0.12–0.92 µL min−1 Yes (1:40 in PBS)Dielectrophoresis (DEP)17 76.9–92.1%n.d.n.d.50 µL hour−1 Yes (1:5 in PBS)DLD10 99%n.dn.d.1 µL min−1 NoLeukocyte margination14 80%90%n.d.600 µL hour−1 Yes (0.5–2% hematocrit)Leukocyte margination15 97%100%n.d.1 µL min−1 Yes (1:1000 in PBS)Leukocyte margination16 67%1%n.d.1 µL hour−1 Non.d. Data not available.*References in uppercase.
|
study
| 99.75 |
A key disadvantage of the microfluidic ratchet mechanism is sample throughput, which at 5 µl per hour for the prototype studied here is significantly less than other methods, such as DLD, which is capable of inhibiting clot formation42 and processing 10 ml of whole blood per minute12. This limited throughput results from the ratchet transport process requiring oscillatory flow, which significantly improved selectivity, but at the cost of throughput. The throughput of the prototype device studied here is sufficient for many leukocyte assays, including flow cytometry, ELISA, and chemotaxis. However, we also recognize there are many applications where greater throughput is desired. In these situations, throughput can be increased by parallelizing the mechanism, as well as optimizing the device design for the specific application, as we have done for a device to enrich for circulating tumor cells, which has a sample throughput of 1 ml/hour36.
|
study
| 100.0 |
In summary, the microfluidic ratchet mechanism enables highly selective sorting of leukocytes directly from whole blood. We showed this mechanism can achieve clear separation of lymphocytes and granulocytes, and we believe with further refinement, the separation of other leukocyte phenotypes can be achieved. Another important application of this technology is the interrogation of the phenotypic plasticity of leukocytes to identify biomechanical changes associated with leukocyte activation, such as neutrophil activation during cancer progression43 and viral infection44. This approach represents a potential alternative to flow cytometry to enrich for leukocytes that adopt specific activation states. After sorting, the fractionated cells are immediately available for downstream molecular characterization, such as immunofluorescence, transcriptome analysis, or cytokine secretion assays. Together, the microfluidic ratchet mechanism represents a simple and compelling approach for biophysical separation and characterization of leukocytes as part of biological assays.
|
study
| 99.94 |
The microfluidic ratchet device consists of a single fluidic layer fabricated using soft-lithography of polydimethylsiloxane (PDMS) silicone. The mold for the microstructures consists of two photo-lithographically defined layers fabricated on a silicon wafer. The sorting region containing a matrix of funnels was fabricated using SU-8 3010 photoresist (MicroChem, Newton, MA, USA) producing features with thickness of 11–15 µm. The supporting microfluidic channels (SI, OSC1, OSC2, CFI and Outlets) were made from SU-8 3025 photoresist with a thickness of ~25 µm. The patterns for both masks were drawn using AutoCAD software.
|
study
| 95.44 |
The SU-8 3010 microstructures were fabricated on a cleaned 100 mm silicon wafer. After dehydration baking at 200 °C for 5 minutes, photoresist SU-8 3010 was spread onto the wafer at 1500 to 1800 rpm for 30 seconds to create a thickness of 11 µm to 15 µm, as measured using a profilometer (Alpha step 200). The wafer was then soft baked at 95 °C for 4 minutes before being exposed to UV light in a mask aligner with uniform exposure for 40 seconds. The exposed wafer was given a post exposure bake at 65 °C for 1 minute, 95 °C for 3 minutes and then 65 °C for 1 minute. Finally, the wafer was developed using SU-8 developer (MicroChem). The geometry of the SU-8 3010 photoresist was stabilized by further baking with ramped temperature at the acceleration of 100 °C hour−1 from 40 °C to 200 °C, held at 200 °C for one hour, and then gradually cooled to 40 °C.
|
study
| 99.8 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.