Split (1)

train · ~5.33M rows (showing the first 168k)

id string	text string	source string	added string	created string	metadata dict
PMC1180855	Introduction ============ Individuals may develop tolerance to adverse health effects elicited from repeated exposure to inhaled toxicants in several different occupational and environmental situations. Pulmonary tolerance can be defined as the lung\'s ability to withstand the detrimental effects of a toxic compound following multiple exposures. There are several implications of tolerance to human health, having both advantages and disadvantages with respect to the development of harmful health effects. Clinical investigations of zinc oxide (ZnO)- \[[@B1]\], endotoxin- \[[@B2]\], and ozone- \[[@B3]-[@B6]\] induced adverse respiratory effects have demonstrated inter-individual variability in the capacity to develop pulmonary tolerance following inhalation exposure. Because these clinical studies are more tightly controlled than epidemiologic studies, they suggest that genetic background and gene-environment interactions contribute to the development of pulmonary tolerance in humans. We initially characterized the pulmonary tolerance phenotype in an outbred mouse model by assessing levels of BAL protein and polymorphonuclear leukocytes (PMNs) following single (1X) and 5 daily repeated (5X) exposures to inhaled ZnO to begin the identification of the genes regulating the development of pulmonary tolerance to repeated toxicant exposure \[[@B7]\]. Significant genetic variability in the development of pulmonary tolerance to ZnO, endotoxin, and ozone was established in several inbred strains of mice in a subsequent study \[[@B8]\]. Of the strains tested, the BALB/cByJ (CBy) strain was tolerant and the DBA/2J (D2) strain was non-tolerant to BAL protein, PMNs, and macrophages following repeated ZnO exposure. Because inbred mouse strains are virtually identical at all loci throughout their genome, and also share several chromosomal regions of conserved synteny with humans, they are an ideal animal model in which to investigate genotype-environment interactions and identify genes controlling pulmonary responses to inhaled toxicants where no a priorievidence for their location exists \[[@B9]\]. Inbred strains of mice have been successfully utilized in quantitative trait locus (QTL) analyses to identify candidate genes that control susceptibility to adverse pulmonary responses induced by a variety of inhaled gases \[[@B10]-[@B13]\] and particulates \[[@B14],[@B15]\]. In the present study, a CByD2F2 mouse cohort was used to determine QTLs linked to the development of pulmonary tolerance to BAL protein, PMN, and macrophage phenotypes following repeated ZnO exposure. Materials and methods ===================== Mice ---- Inbred BALB/cByJ (CBy), DBA/2J (D2) and CByD2F1/J (F1) mice (6--7 weeks of age) were purchased from The Jackson Laboratory (Bar Harbor, ME). CByD2F1/J mice were crossed to produce F2 (intercross) offspring in our laboratory animal facility. All mice were acclimated for at least 1 week before exposure, housed in a positive pressure environment with a 12-hour light/dark cycle starting at 6:00 a.m., and provided with water and standard laboratory rodent chow (Purina, Indianapolis, IN) ad libitumexcept during exposure. All mice were handled in accordance with the standards established by the U.S. Animal Welfare Acts set forth in National Institutes of Health guidelines, and by the New York University School of Medicine Division of Laboratory Animal Resources. Zinc oxide generation, characterization, and exposure ----------------------------------------------------- Mice were exposed to ZnO (1.0 ± 0.2 mg/m^3^, mean ± SD) in stainless steel cages placed inside a 0.07 m^3^Plexiglas chamber. ZnO fumes were generated in a furnace as previously described \[[@B7],[@B16]\]. The ZnO particles had a mass median aerodynamic diameter of 0.3 μm and geometric standard deviation of 1.5. Samples of the chamber atmosphere were collected approximately every 40 minutes from the manifold of the exposure system with polytetrafluoroethylene filters (Type TX40HI20-WW, Pallflex Products Corp., Putnam, CT), and the ZnO concentration was determined gravimetrically using a microbalance (Model C-30, Cahn Instruments, Cerritos, CA). For linkage analyses, all F2 offspring (n= 299, 138 males and 161 females) were exposed at 7--12 weeks of age in a total of seven 5X exposure regimens. Development of pulmonary tolerance was assessed 24 hours after the fifth exposure, and exposure group sizes ranged from 36 to 45 animals. All ZnO exposures of F2 mice also contained at least two CBy, D2, and F1 mice for control purposes. Bronchoalveolar lavage (BAL) ---------------------------- Mice were euthanized by intraperitoneal injections of ketamine HCl (100 mg/kg, Vetalar, Fort Dodge Laboratories, Inc., Fort Dodge, IA) and sodium pentobarbital (175 mg/kg, Sleepaway, Fort Dodge Laboratories, Inc.), and the posterior abdominal aorta was severed. The lungs of each mouse were lavaged two times with 1.2 ml of Dulbecco\'s phosphate buffered saline without Ca^2+^or Mg^2+^(pH 7.2--7.4, 37°C, Invitrogen, Carlsbad, CA). The collected BAL was immediately placed on ice (4°C) following recovery. Measurement of BAL protein, total cell counts, and differential cell counts were performed as previously described \[[@B8]\]. DNA isolation and genotyping ---------------------------- Genomic DNA was isolated from kidney tissue of each phenotyped F2 animal and for CBy, D2, and F1 controls (Wizard Genomic DNA Purification Kit, Promega, Madison, WI). DNA concentration was determined using a Beckman DU-650 spectrophotometer, and each sample was diluted to 10 ng/μl for genotype analysis. PCRs were performed to genotype F2 offspring for SSLPs located throughout the mouse genome. Eighty-six unlabeled primer pairs for SSLPs that differed in length by at least 5% between the CBy and D2 progenitor strains were purchased from Research Genetics (ResGen/Invitrogen). PCR was performed in 20 μl reaction volumes in 96-well low profile plates (Fisherbrand, Fisher Scientific, Fairlawn, NJ) using a PTC-100 thermal cycler (MJ Research, Watertown, MA). The final concentration for each reaction was: 10 mM Tris-HCl (pH 8.3), 50 mM KCL, 2.5 mM MgCl~2~, 0.2 mM of each deoxynucleotide triphosphate (Promega), 1.1X Rediload (ResGen/Invitrogen), and 0.132 μM of each SSLP primer pair. This reaction mixture was added to 100 ng of genomic DNA and 0.45 U of TaqDNA polymerase (Roche Applied Science, Indianapolis, IN). Final reaction mixtures were initially denatured at 94°C for 3 min, followed by 36 amplification cycles (94°C for 30 seconds, 57°C for 45 seconds, and 72°C for 30 seconds + 1 second/cycle). A final extension step at 72°C for 7 min was followed by refrigeration (4°C). PCR products were differentiated on 3% agarose (Invitrogen) gels and all samples were visualized by ethidium bromide staining using a ChemiImager-4400 low light imaging system (Imgen Technologies, Alexandria, VA) and called by a single investigator. Any questionable calls in reading the genotype from the image were reviewed by a second investigator and if not resolved, that sample was rerun. Estimation of loci ------------------ The number of independently segregating loci was calculated using the following formula by Wright \[[@B17]\]: n= (P2 - F1)^2^/4(\\|σ^2^~F2~- σ^2^~F1~\\|), where nis an estimate of the number of independent loci; P2 and F1 are the mean BAL protein responses following 5X ZnO exposure in CBy and CByD2F1 mice, respectively; σ^2^~F2~and σ^2^~F1~are the computed variances of the F2 and CByD2F1 mice, respectively. Linkage analyses ---------------- A genome scan was performed to identify associations between genotypes and the BAL protein phenotype using a CByD2F2 mouse cohort. All phenotypic data were natural log normalized to generate a normal distribution to meet normality assumptions of the Map Manager QT computer program. Interval analyses were then performed by fitting a regression equation for the effect of a theoretical QTL at the position of each SSLP and at 1-centimorgan (cM) intervals between SSLPs using free, additive, recessive, and dominant regression models. The regressions and significance of each genotype/phenotype association (or likelihood χ^2^statistic) were calculated by Map Manager \[[@B18]\]. Permutation tests were performed on the phenotypic and genotypic data using Map Manager to generate empirical thresholds for significance following the methods of Churchill and Doerge \[[@B19],[@B20]\]. For the initial genome scan, the 15 most tolerant and 15 most non-tolerant F2 animals with respect to BAL protein, PMNs, and macrophages (i.e., the phenotypic extremes) were used for selective genotyping \[[@B9],[@B21]\]. Interval analyses were done as stated previously, and 10,000 permutations were performed to generate significant and suggestive likelihood χ^2^statistic thresholds for the BAL protein phenotype. Following the identification of a suggestive QTL for BAL protein on chromosome 1, the entire F2 cohort was examined for additional QTLs. For the BAL protein phenotype, three additional SSLPs on chromosome 1 were analyzed, and a permutation test (10,000 permutations) was performed with only chromosome 1. This method was similar to that used in previous linkage studies with inhaled particles and gases that utilized Map Manager QT \[[@B10],[@B14],[@B22]\]. All likelihood χ^2^statistic thresholds corresponded to those reported in the aforementioned linkage studies. Haplotype analysis ------------------ A haplotype analysis was carried out similar to that done previously by Prows and Leikauf to determine the contribution of each QTL and QTL combination to the overall BAL protein phenotype \[[@B15]\]. This method quantifies any difference in mean BAL protein levels that are linked with a particular haplotype. Haplotypes for the following SSLPs were used for this analysis: D1Mit291(101.5 cM), D4Mit254(82.5 cM), and D5Mit193(1.0 cM). Mean BAL protein concentrations for groups of F2 mice with the same haplotype at each QTL or QTL combinations were calculated and compared with the mean BAL protein of F2 mice with the other haplotypes to determine the contributions of these QTLs to the overall BAL protein phenotype. Results ======= Phenotypes of the CByD2F2 cohort -------------------------------- To further understand the role of genetic background in the development of pulmonary tolerance, an F2 (backcross) cohort derived from the CBy and D2 progenitors was phenotyped. The frequency distribution of the BAL protein, PMN, and macrophage phenotypic responses of the F2 cohort were within the ranges of similarly exposed CBy and D2 progenitor mice (Figure [1](#F1){ref-type="fig"}). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Frequency distribution of the number of BAL PMNs (×10^4^), macrophages (×10^4^), and protein (μg/ml) in BALB/cByJ, DBA/2J, CByD2F1/J, and CByD2F2 mice following 5 consecutive days of exposure to 1.0 mg/m^3^ZnO for 3 h/day. Vertical dashed lines represent approximate separation points between BALB/cByJ and DBA/2J phenotypes. ::: ![](1465-9921-6-73-1) ::: Selective genotyping -------------------- A genome-wide scan was performed using the 15 most tolerant and 15 most non-tolerant mice to initially identify possible QTLs influencing the development of pulmonary tolerance. Permutation tests on the BAL protein-extreme data set established a suggestive likelihood χ^2^statistic threshold of 9.9 and a significant likelihood χ^2^statistic threshold of 17.4. These values were consistent with the genome-wide probabilities projected by Lander and Kruglyak \[[@B23]\]. Interval mapping identified a suggestive QTL for the BAL protein phenotype on the distal end of chromosome 1 (Figure [2](#F2){ref-type="fig"}). No QTLs were identified for the PMN and macrophage phenotypes from selective genotyping. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Genome-wide scan for QTLs associated with the BAL protein phenotype by selective genotyping of the CByD2F2 cohort. For each plot, the x-axis is the length of the chromosome in centimorgans (cM), and the y-axis is the likelihood χ^2^statistic value as calculated by Map Manager. The upper and lower dashed lines represent significant (LRS = 17.4) and suggestive (LRS = 9.9) linkage thresholds, respectively. ::: ![](1465-9921-6-73-2) ::: Genotyping of the entire F2 cohort ---------------------------------- The entire F2 cohort was genotyped with three additional SSLPs on distal chromosome 1 to further analyze the suggestive BAL protein QTL on chromosome 1 identified from selective genotyping. Interval mapping of the entire F2 cohort confirmed the QTL on chromosome 1 between 101.0 cM (D1Mit426) and 109.6 cM (D1Mit293) (Figure [3](#F3){ref-type="fig"}). The peak likelihood χ^2^statistic value for this QTL exceeded the threshold value of 10.0 for significant linkage as determined by 10,000 permutations with all loci from chromosome 1 only. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Plot of a significant QTL on chromosome 1 that is associated with the BAL protein phenotype from analysis of the entire CByD2F2 cohort. The x-axis is the length of the chromosome in centimorgans (cM), and the y-axis is the likelihood χ^2^statistic value as calculated by Map Manager. The lower dashed line represents the suggestive linkage threshold (LRS = 3.8), the middle dashed line represents significant linkage threshold (LRS = 10.0), and the upper dashed line represents the highly significant linkage threshold (LRS = 18.8). ::: ![](1465-9921-6-73-3) ::: The entire F2 cohort was further analyzed with the initial 86 SSLPs for the BAL protein, PMN, and macrophage phenotypes. Suggestive QTLs for the BAL protein phenotype that were not previously characterized by selective genotyping were identified on chromosome 4 between 53.6 cM (D4Mit146) and 82.5 cM (D4Mit254), and on chromosome 5 between 1.0 cM (D5Mit193) and 18.0 cM (D5Mit148) (Figure [4](#F4){ref-type="fig"}). Suggestive QTLs for the BAL PMN and macrophages were identified on chromosomes 1 and 5, respectively (Figure [5](#F5){ref-type="fig"}). ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Plots of suggestive QTLs on chromosomes 4 and 5 associated with the BAL protein phenotype from analysis of the entire CByD2F2 cohort. The x-axis is the length of the chromosome in centimorgans (cM), and the y-axis is the likelihood χ^2^statistic as calculated by Map Manager. The upper and lower dashed lines in each plot represent significant (LRS = 15.8) and suggestive (LRS = 9.2) linkage thresholds, respectively. ::: ![](1465-9921-6-73-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Plots of suggestive QTLs on chromosomes 1 and 5 associated with the BAL PMN and macrophage phenotypes, respectively, from analysis of the entire CByD2F2 cohort. The x-axis is the length of the chromosome in centimorgans (cM), and the y-axis is the likelihood χ^2^statistic as calculated by Map Manager. The upper and lower dashed lines in each plot represent significant (LRS = 15.6) and suggestive (LRS = 9.2) linkage thresholds, respectively. ::: ![](1465-9921-6-73-5) ::: Haplotype analysis ------------------ Mean BAL protein levels of F2 mice with the same haplotype were calculated and compared with the mean BAL protein levels of F2 mice with the opposite haplotype in order to determine the contribution of each QTL and combinations of QTLs to the overall BAL protein phenotype (Figure [6](#F6){ref-type="fig"}). For each SSLP, F2 animals were genotyped as a homozygous CBy (CC), a homozygous D2 (DD) or heterozygous (H). For any individual SSLP, the greatest difference in mean BAL protein was found for D1Mit291. F2 mice that were DD at that locus had an average of 156 μg/ml more BAL protein than those mice that had CC or H haplotypes. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Differences in mean BAL protein levels of CByD2F2 mice with tolerant versus non-tolerant haplotypes at SSLPs representing the identified QTLs. Open bars, F2 mice with CC or CD genotypes (represented as H). Filled bars, F2 mice with a DD genotype (represented as D). Number within each baris the number of F2 mice with the given genotype or haplotype. Values are means ± SE. All comparisons of tolerant (open bars) haplotypes versus non-tolerant haplotypes (filled bars) were significant (P\< 0.05, ttest). ::: ![](1465-9921-6-73-6) ::: The combinatorial effect of the QTLs on chromosomes 1, 4, and 5 were also examined. For any combination of two QTLs, F2 mice that had a DD-DD haplotype for markers on chromosomes 1 and 5 (D1Mit291and D5Mit193) had an average of 310 μg/ml more BAL protein than those that were CC or H (CC/H) for those markers. The greatest difference in mean BAL protein levels were found in F2 mice that had a DD-CC/H-DD haplotype for the three QTLs on chromosomes 1, 4, and 5, (i.e., D1Mit291, D4Mit254, and D5Mit193), respectively. These mice had an average of 345 μg/ml more BAL protein than those that were CC/H at the markers across the three chromosomes. Identification of candidate genes --------------------------------- Within all of the QTLs, candidate genes discovered with potential roles in controlling the development of tolerance to inhaled ZnO are presented in Tables [1](#T1){ref-type="table"} and [2](#T2){ref-type="table"}. These genes were chosen as candidates from a thorough review of the existing literature and through the positional candidate gene approach, which combines knowledge of map position with the mouse transcript map \[[@B24]\]. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Positional candidate genes from linkage analysis of pulmonary tolerance to BAL protein following repeated ZnO exposure. ::: Chromosome QTL Interval Location Human Orthology Candidate Gene ------------ -------------- ---------- -------------------------- ------------------------------------------------------------------ Chr 1 92--112 cM 94.0 cM 1 (q21-q22) Dfy, Duffy blood group 98.0 cM 1 (q41-q42) Tlr5, toll-like receptor 5 98.6 cM 1 (q41-q42) Adprt1, ADP-ribosyltransferase 1 101.5 cM 1 (q41) Tgfb2, transforming growth factor-β2 105.0 cM 1 (q32) Traf5, TNF receptor-associated factor 5 106.0 cM 1 (q32-q41) Slc30a1, solute carrier family 30 (zinc transporter), member 1 Chr 4 62--82 cM 62.4 cM 1 (p35-p34.3) Ptafr, platelet-activating factor receptor 65.7 cM 1 (p35.3) Slc30a2, solute carrier family 30, member 2 68.0 cM 1 (p35), 1 (p34-p36) Pla2g2a, 2c, 2d-f, 5, phospholipase A~2~groups 75.5 1 (p36.3-p36.2), 1 (p36) Tnfrsf1b, 8, 9, TNF-receptor superfamily members 79.4 cM 1 (p36) Tnfrsf4, TNF-receptor superfamily, member 4 Chr 5 1--20 cM 14.0 cM 7 (q35-q36) Slc4a2, solute carrier family 4, member 2 17.0 cM 7 (p21) Il6, interleukin 6 18.0 cM 7 (pter-qter) Slc30a3, solute carrier family 30 (zinc transporter), member 3 ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Positional candidate genes from linkage analysis of pulmonary tolerance to BAL PMNs and macrophages following repeated ZnO exposure. ::: Chromosome QTL Interval Location Human Orthology Candidate Gene --------------------- -------------- ---------- ----------------- --------------------------------------------- Chr 1 (PMNs) 51--73 cM 52.0 cM 2 (q33-q37) Ccl20, chemokine (C-C motif) ligand 20 67.4 cM 2 (q21) Cxcr4chemokine (C-X-C motif) receptor 4 69.9 cM 1 (q31-q32) Il10, interleukin 10 Chr 5 (Macrophages) 43--58 cM 51.0 cM 4 (q13-q21) Areg, amphiregulin 51.0 cM 4 (q13-q21) Btc, betacellulin 51.0 cM 4 (q21) Cxcl1, chemokine (C-X-C motif) ligand 1 51.0 cM 4 (q21) Cxcl2, chemokine (C-X-C motif) ligand 2 53.0 cM 4 (q21) Cxcl5, chemokine (C-X-C motif) ligand 5 53.0 cM 4 (q21) Cxcl9, chemokine (C-X-C motif) ligand 9 53.0 cM 4 (q21) Cxcl10, chemokine (C-X-C motif) ligand 10 56.0 cM 4 (q21-q25) Spp1, secreted phosphoprotein 1 ::: Development of pulmonary tolerance in MOLF/Ei mice -------------------------------------------------- We identified toll-like receptor 5 (Tlr5) as an interesting candidate gene within the significant QTL for BAL protein on chromosome 1. A wild-derived inbred mouse strain called MOLF/Ei (M. m. molossinus) which has non-conservative mutations in Tlr5\[[@B25]\] was phenotyped for BAL protein following 1X and 5X ZnO exposure to determine whether Tlr5may function in the development of pulmonary tolerance (Figure [7](#F7){ref-type="fig"}). MOLF/Ei BAL protein was significantly increased above control values following 1X ZnO exposure (395 ± 38 μg/ml). However, MOLF/Ei mice exhibited a tolerant phenotype, as 5X BAL protein values (215 ± 22 μg/ml) were significantly decreased below that of the 1X exposure group. ::: {#F7 .fig} Figure 7 ::: {.caption} ###### BAL protein levels in wild-derived MOLF/Ei mice 24 h after single (1X) or repeated (5X) exposure to 1.0 mg/m^3^ZnO or air for 3 h. Protein levels of BALB/cByJ and DBA/2J mouse strains following ZnO exposure are also shown for comparison purposes. Values are means ± SE (n= 4--5 MOLF/Ei mice/exposure group). \* indicates significantly different from air-exposed MOLF/Ei controls, P\< 0.05 \[Student-Newman Keuls (SNK) test\]. + indicates significantly different from 1X MOLF/Ei exposure group, P\< 0.05 (SNK test). ::: ![](1465-9921-6-73-7) ::: Discussion ========== Clinical studies on the acquisition of tolerance to inhaled toxicants suggest that genetic background and gene-environment interactions contribute to the development of pulmonary tolerance in humans. We have previously determined that a genetic component exists in a mouse model of pulmonary tolerance to repeated ZnO exposure \[[@B8]\]. In the present study, we performed linkage analyses on an F2 mouse population derived from tolerant CBy and non-tolerant D2 strains to further ascertain the contribution of genetic background to the development of pulmonary tolerance, and identify candidate genes that may be important regulators in the acquisition of tolerance. Initial analysis using the most tolerant and non-tolerant F2 mice with respect to BAL protein generated a putative QTL (designated as the zinc-induced tolerance (ZIT~1~) locus) on the distal end of chromosome 1. Further assessment of the entire F2 cohort demonstrated that the QTL on chromosome 1 attained an LRS value for significant linkage, and also identified two suggestive QTLs located on chromosomes 4 and 5. Using a variation of the Wright equation \[[@B17]\], a minimum of three loci were estimated to be independently segregating with the BAL protein phenotype following 5X ZnO exposure, thus in agreement with the results of the linkage analysis. There are several approaches that can be pursued to focus in on the QTL intervals that were identified. For instance, increasing the number of F2 mice used in the QTL analysis is one alternative. The major disadvantage of this, however, is that segregating QTLs contribute a great deal of phenotypic \"noise,\" making it problematic when determining whether or not a given mouse has inherited a particular QTL \[[@B9]\]. Thus, in order to separate the effects of multiple loci, congenic mouse strains for each QTL could be generated, which could then be used to breed multicongenic lines to examine the existence of any epistatic effects. Additionally, future studies could employ the use of a backcross (CByD2F1 × CBy) mouse population to expand the evaluation of the QTL effects to the overall BAL protein phenotype. To measure the contribution of each individual QTL and each QTL combination to the overall BAL protein phenotype, the protein levels for F2 mice with each particular haplotype were compared. Mice with opposite allelic combinations for QTLs on chromosomes 1 and 5 had a difference of 310 μg/ml BAL protein, which accounts for approximately one-third of the total difference in mean BAL protein between the parental CBy and D2 strains. Additionally, the mean BAL protein level of F2 mice with a DD haplotype for QTLs on chromosomes 1, 4, and 5 was over half that of the non-tolerant D2 parental strain. These analyses suggest that although three QTLs were identified, the development of pulmonary tolerance to BAL protein is a decidedly complex phenotype that is regulated by a number of different genes, some of which were likely not identified by linkage analysis using an F2 cohort. Candidate genes within the significant QTL on chromosome 1 (ZIT~1~) that could play a role in controlling the development of tolerance to BAL protein following repeated ZnO exposure are presented in Table [1](#T1){ref-type="table"}. The Duffy blood group (Dfy) has been shown to modulate the intensity of inflammation following endotoxin exposure \[[@B26]\], and has a role in enhancing inflammatory cell recruitment to sites of inflammation by facilitating movement of chemokines across the endothelium \[[@B27]\]. ADP-ribosyltransferase 1 (Adprt1) and tumor necrosis factor (TNF) receptor-associated factor (Traf5) are functionally associated with nuclear factor (NF)-κB, a key transcription factor in the regulation of the inflammatory process \[[@B28],[@B29]\]. Additionally, activation of ADPRT1 plays a role in endotoxin-induced BAL protein increases \[[@B30],[@B31]\]. Activated transforming growth factor-β has been shown to be a mediator of bleomycin- and endotoxin-induced lung permeability (i.e., BAL protein) in mice \[[@B32]\]. Finally, solute carrier family 30 (zinc transporter), member 1 (Slc30a1) is a metal transporter on the plasma membrane that confers resistance to zinc and cadmium toxicity in vitrovia an efflux mechanism \[[@B33],[@B34]\]. Candidate genes for BAL protein tolerance identified within the suggestive QTL intervals on chromosomes 4 and 5 are also presented in Table [1](#T1){ref-type="table"}. Notable candidates include platelet-activating factor receptor (Ptafr) and phospholipase A~2~(Pla2) that have been implicated as potential mediators of toxicant-induced BAL protein increases \[[@B35]-[@B37]\]. Interleukin (IL)-6 protein was increased following ZnO exposure in humans \[[@B1],[@B38]\], and was hypothesized to be an anti-inflammatory suppressor of ZnO-induced lung injury. Interestingly, increased lung IL-6 levels have been shown to mediate pulmonary tolerance to ozone in rats \[[@B39]\]. Lastly, solute carrier family 30 (zinc transporter) members 2 and 3 (Slc30a2and Slc30a3) have been identified as zinc transporters that protect against zinc-induced toxicity in both cell culture and animal models \[[@B40],[@B41]\]. The development of pulmonary tolerance to BAL protein in our model could be regulated by any number of the aforementioned candidate genes. These candidates will be investigated in future studies of transcriptional and protein regulation to determine their roles in the development of tolerance. The suggestive QTLs for tolerance to BAL PMNs and macrophages (on chromosomes 1 and 5, respectively) were also examined for positional candidate genes (Table [2](#T2){ref-type="table"}). Chemokine (C-C motif) ligand 20 (Ccl20), chemokine (C-X-C motif) receptor 4 (Cxcr4), and interleukin-10 (Il10) were identified as candidates for BAL PMN tolerance. The movement of PMNs into inflammatory tissues is regulated by chemotactic factors (e.g., chemokines) that signal through numerous chemokine receptors \[[@B42]\]. PMNs are capable of producing several chemokines and proinflammatory cytokines, indicating that PMNs may be important in auto-direction of cell trafficking during inflammation. CXCR4 expression has been detected on human PMNs \[[@B43]\], and coordinated chemokine receptor gene expression may control the tissue-specific migration and activation status of PMNs into the lung. Human PMNs are also able to express CCL20 \[[@B44]\], which is able to recruit immature dendritic cells that play an important role in the initiation of the immune response as well as chronic inflammation \[[@B45]-[@B47]\]. Finally, IL-10 can be produced by T cells and is able to diminish PMN influx by inhibition of expression of proinflammatory chemokines \[[@B48]\], NF-κB (via I kappa kinase) \[[@B49]\], and TNF \[[@B50]\], as well as modulate cells and effector functions associated with an allergic response. With respect to the BAL macrophage phenotype, the epidermal growth factor receptor (EGFR) ligands amphiregulin (Areg) and betacellulin (Btc) were identified as candidate genes within the suggestive QTL on chromosome 5. Ligand-dependent activation of the EGFR by particles rich in metal content lead to activation of the MAP kinase signaling cascade and cytokine expression and secretion \[[@B51],[@B52]\]. Secreted phosphoprotein 1 (Spp1, also known as osteopontin) was also identified, and it can act as a chemoattractant for macrophages \[[@B53],[@B54]\]. Lastly, a host of chemokine (C-X-C motif) ligands were identified between 51 and 53 cM that could be involved in macrophage chemotaxis \[[@B55]\]. Again, tolerance to BAL PMNs and macrophages may be under the influence of any number of these candidate genes. In the present study, we identified toll-like receptor (Tlr5) as a candidate gene within the significant ZIT~1~QTL on chromosome 1 for tolerance to BAL protein. Toll-like receptors activate intra-cellular signaling that culminates in the induction of a multitude of effector genes \[[@B56]\]. Tlr5has been shown to be an important gene in the immune response to Gram-positive and Gram-negative bacterial flagellin \[[@B57],[@B58]\]. We utilized a wild-derived inbred mouse strain called MOLF/Ei which has non-conservative mutations in Tlr5that are associated with a lower level of expression \[[@B25]\] to determine whether Tlr5plays a role in the development of tolerance to BAL protein in our ZnO model. Because MOLF/Ei mice have a lower level of TLR5 mRNA expression compared to other strains \[[@B25]\], it was unclear whether we would observe tolerance after a single exposure to ZnO. Although the MOLF/Ei strain has no \"wild-type\" control strain per se, MOLF/Ei mice were tolerant to increased BAL protein following repeated ZnO exposure when compared to the non-tolerant D2 strain. These data support a role for Tlr5in the development of tolerance to BAL protein. Interestingly, Kleeberger and colleagues identified Tlr4as a candidate gene in a study of susceptibility to increased BAL protein in mice following a single exposure to 0.3 ppm ozone for 72 hours \[[@B22]\], which also suggests toll-like receptor signaling may be important in the regulation of inhaled toxicant-induced changes in BAL protein. The mechanism through which Tlr5signaling may regulate the development of tolerance to ZnO is unknown. While endotoxin \[[@B59]\] and bacterial flagellin \[[@B60]\] have been demonstrated as the primary ligands for Tlr4and Tlr5, respectively, the ligand(s) that is responsible for toll-like receptor signaling following exposure to inhaled toxicants such as ozone and ZnO is unknown. Although there have been no studies on endogenous ligands for Tlr5, several endogenous ligands for Tlr4have been identified. For example, fibronectin \[[@B61]\] and hyaluronic acid \[[@B62]\] are produced by lung cells during lung injury and are endogenous Tlr4ligands. The downstream pathway through which Tlr5may regulate the development of tolerance to ZnO-induced BAL protein is potentially via NF-κB, a transcription factor that is known to induce several cytokines involved in inhaled ZnO responsiveness such as IL-8 and IL-6 \[[@B1],[@B63]\]. Furthermore, NF-κB-dependent gene expression is decreased in Tlr5-mediated tolerance to flagellin in vitro\[[@B64]\]. Thus, it is plausible that a mutation in Tlr5, a regulatory element upstream of NF-κB signaling, could modulate tolerance to BAL protein from repeated ZnO exposure. Finally, Tlr5may function as a danger signal receptor in the development of ZnO tolerance, consistent with the \"danger model\" of innate immunity that explicates activation of the innate immune system by factors other than foreign antigens \[[@B65]\]. Whatever the case, much work is needed in understanding how toxicants function through toll-like receptor signaling mechanisms in the lung to regulate adverse responses such as BAL protein. In summary, linkage analysis of a large F2 mouse cohort identified significant linkage of a QTL (ZIT~1~) on chromosome 1 associated with tolerance to BAL protein following repeated exposure to inhaled ZnO. Suggestive QTLs were also identified on chromosomes 4 and 5 for BAL protein, on chromosome 1 for BAL PMNs, and on chromosome 5 for BAL macrophages. Haplotype analysis suggested that the combinatorial effects of these three loci contributed to the overall phenotype, which agrees with the calculated number of segregating loci. Tlr5was identified within the significant QTL for BAL protein on chromosome 1. Wild-derived Tlr5-mutant MOLF/Ei mice were determined to be tolerant to BAL protein following repeated ZnO exposure, suggesting a role for Tlr5in the development of pulmonary tolerance to inhaled toxicants. Conclusion ========== These data substantiate genetic background as an important influence in the acquisition of pulmonary tolerance following exposure to inhaled toxicants such as ZnO, and promising candidate genes exist within the identified QTL intervals that would be good targets for additional studies on the pathogenesis of tolerance. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= SCW: Participated in the design and coordination of the study, performed the study, and drafted the manuscript. LCC: Participated in the design of the study and helped draft the manuscript. TG: Conceived the study, participated in the design and coordination of the study, and helped draft the manuscript. Acknowledgements ================ The authors would like to thank the following people for their contributions to this study: Karen Galdanes, Margaret Krasinski, Kathy Baker, and Dr. Moon-shong Tang (New York University); Dr. George D. Leikauf (University of Cincinnati); Dr. Daniel R. Prows (Cincinnati Children\'s Hospital); Dr. Steven R. Kleeberger (National Institute of Environmental Health Sciences); Dr. Carrie L. Welch (Columbia University). This study was supported by USEPA Grant R-826244, USEPA Particulate Matter Center Grant R-827351, USEPA STAR Fellowship U-91578301, NIEHS Center Grant of Excellence ES-00260, and CDC/NIOSH (via Mt. Sinai School of Medicine, New York, NY) Grant T-42CCT210425-06-01.	PubMed Central	2024-06-05T03:55:59.717231	2005-7-18	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1180855/", "journal": "Respir Res. 2005 Jul 18; 6(1):73", "authors": [ { "first": "Scott C", "last": "Wesselkamper" }, { "first": "Lung Chi", "last": "Chen" }, { "first": "Terry", "last": "Gordon" } ] }
PMC1180856	Introduction ============ Severe pulmonary hypertension is a fatal disease with short life expectancy \[[@B1],[@B2]\]. Continuous intravenous administration of prostacyclin was documented to improve exercise capacity and survival in patients with idiopathic pulmonary arterial hypertension (IPAH, formerly primary pulmonary hypertension, PPH) \[[@B1],[@B3]\]. Possible disadvantages of this approach are catheter related septic events and systemic side effects including serious systemic hypotension. In patients with pulmonary hypertension associated with pulmonary fibrosis, systemic administration of vasodilators results in ventilation perfusion mismatch and impairment of arterial oxygenation. Inhalation of aerosolized iloprost, a long-acting prostacyclin analogue, has been shown to cause selective pulmonary vasodilatation in both primary and secondary pulmonary hypertension \[[@B4]-[@B6]\]. Long term use of nebulized iloprost was described to improve exercise capacity, event-free survival and hemodynamics in severe IPAH, and this finding was supported by a randomized, controlled phase III study in patients in NYHA class III and IV \[[@B7]\]. The use of aerosolized iloprost for acute pulmonary vasodilatation and putative long-term anti-remodeling effects in severe chronic pulmonary hypertension does, however, demand 6 -- 12 inhalations per day, as the vasodilatory effect levels off within \~60 min post nebulization. Against this background, recent studies addressed the impact of selective phosphodiesterase (PDE) inhibitors on prostacyclin-induced acute pulmonary vasodilatation, reporting a marked amplification and prolongation of the vasodilatory response to inhaled PGI~2~\[[@B8]\]. PDEs are enzymes that inactivate cyclic AMP and cyclic GMP, the second messengers of prostacyclin and NO \[[@B9],[@B10]\]. The characterization of the various PDEs currently known has largely profited from the employment of selective PDE inhibitors. Concerning the lung vasculature, the presence of the PDE isoenzymes 1, 3, 4 and 5 in the cytosolic and particulate phases (homogenized human pulmonary artery tissue) has been demonstrated \[[@B11]\]. Phosphodiesterase 1 is Ca^2+^/calmodulin dependant and hydrolyzes both cGMP and cAMP. PDE3 does possess high affinity for both cAMP and cGMP, with V~max~for cAMP usually greater than that for cGMP \[[@B9],[@B12]\]. PDE4 enzymes are characterized by their high affinity to cAMP, with cGMP representing a very poor substrate. In contrast, PDE 5 is cGMP-specific and was found to be highly expressed in lung tissue \[[@B13],[@B14]\]. Recent clinical data suggest that the PDE 1/5/6 inhibitor sildenafil (IC~50~values 280 nM, 3.5 nM and 37 nM, respectively \[[@B15]\]), which has been approved for the treatment of erectile dysfunction, is an effective pulmonary vasodilator in patients with pulmonary arterial hypertension \[[@B16]-[@B21]\]. Based on a very recent positive phase III study, sildenafil has been approved for the treatment of pulmonary hypertension in US. Interestingly, it has been shown that sildenafil synergizes with inhaled iloprost in patients with pulmonary hypertension \[[@B16],[@B22]\]. Hitherto no attempt was undertaken to clarify, which of the PDEs addressed by sildenafil is the most relevant for the effect of this agent in the pulmonary circulation, and whether combinations with further selective PDE inhibitors might even enhance the sildenafil effect. To address this issue, systemic application of per seineffective doses of specific PDE inhibitors in companion with inhalation of iloprost was undertaken in an experimental model of pulmonary hypertension in the present study. Methods ======= Materials --------- 8-Methoxymethyl-IBMX (8-Methoxymethyl-3-isobutyl-1-methylxanthine) and the thromboxane-A~2~mimetic U46619 were supplied by Sigma (Deishofen, Germany). Sildenafil was obtained from Pfizer (Sandwich, UK) and iloprost (Ilomedin^®^) was obtained from Schering A.G. (Berlin, Germany). All other chemicals and drug supplies were from standard commercial sources. Surgical Preparation -------------------- New zealand white rabbits weighing between 2.8 and 3.1 kg of either sex were anesthetized with a mixture of xylazine and ketamine and anticoagulated with 200 U/kg heparin \[[@B8]\]. Anaesthesia was maintained by a constant intravenous infusion of xylazine and ketamine through the right peripheral ear vein. Animals were tracheostomized and ventilated using a volume-controlled respirator (cat ventilator, Hugo Sachs Elektronik, March Hugstetten, Germany) with 8 ml/kg bodyweight and a frequency of 40 min^-1^. FiO~2~was set at 0.5 and a positive endexpiratory pressure of 0.5 mmHg was used throughout. The left A. carotis was cannulated for arterial pressure monitoring and a pulmonary artery catheter (4 Fr, Braun, Melsungen, Germany) was inserted into the pulmonary artery through the right external jugular vein. Hemodynamics and blood gases ---------------------------- Mean pulmonary artery pressure (P~PA~) and mean aortic pressure (P~SA~) were continuously recorded with fluid-filled force transducers (Braun, Combitrans, FRG). The level of the left atrium was set to zero. As described previously, pulmonary artery occlusion pressure was measured by gentle forwarding of the catheter to wedge position \[[@B23]\]. Pulmonary and systemic vascular resistances were calculated by standard formulas. As described previously, cardiac output (CO) was calculated by using the Fick principle \[[@B8]\]. Briefly, arterial and venous blood samples (1 ml) were stored on ice, and hemoglobin and oxygen saturation were measured using an OSM2 Hemoximeter (Radiometer-Copenhagen, Denmark). Oxygen uptake of the animals was measured online (Labotect O~2~-Controller, Goettingen, Germany). Nebulization ------------ Iloprost was nebulized by means of an ultrasonic nebulizer (Pulmo Sonic 5500, DeVilbiss Medizinische Produkte GmbH, Langen, Germany) which produces an aerosol with a mass median aerodynamic diameter (MMAD) of 4.5 μm and a geometric standard deviation (GSD) of 2.3. The nebulizer was placed in the inspiratory limb of the ventilation system as described previously \[[@B24]\]. Experimental protocols ---------------------- U46619 was continuously infused (dose range 0.5 to 2 μg/kg min) to increase pulmonary artery pressure from \~16 at baseline to \~26 mmHg within 20 min. As described previously, stable pulmonary hypertension is established by this approach \[[@B8]\]. Dose-effect curves of intravenous sildenafil, motapizone and 8MM-IBMX were established after reaching a stable pressure plateau, performing short-term infusions (10 min) with randomized doses of these agents. Hemodynamics and blood gases were measured at the end of the 10 min infusion period. A total dose of 0.4 ± 0.08 μg/kg iloprost, nebulized within a 10 min aerosolization maneuver, was used throughout all studies with iloprost inhalation. In the group with sole administration of this inhalative agent, the nebulization was performed after reaching a stable pressure plateau. In the combination experiments, the PDE inhibitors were administered intravenously at a dose which by itself did not reduce PAP significantly as short-term infusion (10 min), and iloprost was nebulized subsequently. Data analysis ------------- All data are given as means ± SEM. Differences between the different groups were assessed by use of analysis of variance and Student-Newman-Keuls test for multiple comparisons with a p value \< 0.05 regarded to be significant. Results ======= Baseline and U46619-induced pulmonary hypertension -------------------------------------------------- The continuous infusion of 1.3 ± 0.9 μg/kg min U46619 resulted in a significant increase of pulmonary artery pressure (P~PA~) to 26 mmHg as compared to 16 mmHg prior to U46619 (Table [1](#T1){ref-type="table"}). Cardiac output and mean systemic pressure (P~SA~) did not change significantly. The pulmonary vascular resistance increased from 275 to 592 dyne/s cm^-5^m^2^. No significant changes in blood gases were measured as compared to baseline values. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Summarized data of hemodynamics and blood gases in rabbits with U46619-induced pulmonary hypertension and inhalation of iloprost in the absence and presence of sub-threshold intravenous PDE inhibitors.^Δ^ ::: Control/U46619 U46619/Ilo U46619/Ilo/Sil U46619/Ilo/8MM-IBMX U46619/Ilo/Mota U46619/Ilo/8MM-IBMX/Sil U46619/Ilo/Mota/Sil ----------------------------- -------------------- ---------------- -------------------- ------------------------- --------------------- ----------------------------- ------------------------- -------------- ------------- -------------- ------------- -------------- ------------- -------------- P~SA~, mmHg 111 ± 4 105 ± 3 111 ± 3 111 ± 3 106 ± 3 105 ± 2 107 ± 2 104 ± 4 100 ± 3 96 ± 3 99 ± 3 97 ± 2 93 ± 4 92 ± 4 P~PA~, mmHg 15.9 ± 0.3 25.9\* ± 0.4 25.8 ± 0.6 22.7\* ± 0.4 25.7 ± 0.2 21.8\* ± 0.3 24.8 ± 0.7 21.1\* ± 0.8 27.3 ± 1,1 21.0\* ± 0.6 25.0 ± 0.4 20.4\* ± 0.3 24.7 ± 0.4 19.3\* ± 0.4 CO, ml/min 555 ± 23 544 ± 38 432 ± 27 452 ± 25 444 ± 19 488 ± 35 393 ± 24 437 ± 29 417 ± 21 446 ± 15 454 ± 27 505 ± 30 501 ± 35 555 ± 52 PAOP, mmHg 7.2 ± 1.2 7.6 ± 0.8 8.0 ± 1.1 7.3 ± 1.3 7.7 ± 0.9 7.4 ± 1.1 7.7 ± 0.8 7.6 ± 1.2 6.9 ± 1.3 7.3 ± 1.1 7.7 ± 1.2 7.7 ± 1.1 7.8 ± 0.8 7.2 ± 1.1 PVRI, dyne/s cm^-5^m^2^ 275 ± 34 592\* ± 49 725 ± 39 599\* ± 37 713 ± 58 519\* ± 44 765 ± 67 543\* ± 63 861 ± 64 540\* ± 71 670 ± 75 442\* ± 52 593 ± 56 383\* ± 61 P~a~O~2~, mmHg 226 ± 17 199 ± 12 183 ± 12 177 ± 10 191 ± 6 203 ± 4 228 ± 7 188 ± 12 201 ± 7 197 ± 11 171 ± 10 173 ± 11 179 ± 10 174 ± 14 PH~a~ 7.42 ± 0.02 7.35 ± 0.02 7.37 ± 0.03 7.32 ± 0.01 7.36 ± 0.02 7.32 ± 0.02 7.41 ± 0.02 7.38 ± 0.02 7.33 ± 0.01 7.31 ± 0.01 7.35 ± 0.02 7.34 ± 0.02 7.33 ± 0.02 7.29 ± 0.01 P~a~CO~2~mmHg 42.7 ± 1.3 43.0 ± 3.5 41.5 ± 2.0 40.7 ± 1.0 42.0 ± 4.1 45.2 ± 4.0 36.2 ± 1.6 37.7 ± 1.5 44.3 ± 1.2 46.2 ± 1.2 45.9 ± 1.7 47.6 ± 1.4 36.9 ± 2.0 38.0 ± 1.8 P~v~O~2~, mmHg 46.7 ± 1.3 44.1 ± 2.4 39.9 ± 1.7 42.6 ± 1.2 39.7 ± 2.3 44.9 ± 1.4 35.8 ± 1.5 39.7 ± 2.1 40.7 ± 1.64 42.7 ± 1.2 39.7 ± 2.5 44.9 ± 2.6 41.8 ± 1.2 44.0 ± 1.6 PH~v~ 7.30 ± 0.01 7.24 ± 0.02 7.34 ± 0.03 7.29 ± 0.01 7.23 ± 0.02 7.20 ± 0.02 7.37 ± 0.02 7.31 ± 0.03 7.28 ± 0.01 7.26 ± 0.01 7.29 ± 0.02 7.28 ± 0.02 7.28 ± 0.02 7.24 ± 0.01 P~v~CO~2~, mmHg 49.8 ± 0.9 54.9 ± 1.2 48.9 ± 2.3 49.7 ± 1.1 59.2 ± 1.6 60.6 ± 1.4 44.6 ± 2.5 44.9 ± 2.5 52.4 ± 1.4 54.2 ± 1.4 53.8 ± 2.0 52.9 ± 2.2 42.5 ± 2.3 45.9 ± 1.6 ^Δ^P~SA~, mean aortic pressure; P~PA~, mean pulmonary artery pressure; CO, cardiac output; PAOP, pulmonary artery occlusion pressure; PVR, pulmonary vascular resistance; P~a~O~2~, arterial PO~2~; pH~a~, arterial pH; P~a~CO~2~, arterial PCO~2~; P~v~O~2~, central venous PO~2~; pH~v~, central venous pH; P~v~CO~2~, central venous PCO~2~; Ilo, iloprost; Sil, sildenafil; 8MM-IBMX, 8-Methoxymethyl-3-isobutyl-1-methylxanthine; Mota, motapizone. The first two columns give summarized data for all groups. All other columns give pre- and post-iloprost nebulization data in animals undergoing preceding sub-threshold PDE-inhibitor infusion (n = 8 for each group). Asterisks indicate significant differences between pre and post-aerosolization values (\* = p \< 0.05). ::: Dose-effect curves of PDE-inhibitors ------------------------------------ Intravenous sildenafil, motapizone and 8MM-IBMX reduced P~PA~in a dose-dependent manner (Fig. [1A](#F1){ref-type="fig"}), with the dose-effect curves differing by \~two orders of magnitude between sildenafil and the two other compounds. As depicted in Fig. [1B](#F1){ref-type="fig"}, this pulmonary vasodilatation was accompanied by a significant systemic arterial pressure decrease in case of motapizone (dose range 6 -- 600 μg/kg × min) and 8MM-IBMX (dose range 70 -- 1500 μg/kg × min), but not in case of sildenafil (dose range 0.1 -- 10 μg/kg × min). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Dose effect curves of different PDE inhibitors on U46619-elicited pulmonary hypertension (a) and systemic arterial pressure (b). (a) Pulmonary artery pressure drop (ΔP~PA~, in mmHg) and systemic arterial pressure drop (ΔP~SA~, in mmHg) (b) are given (mean ± SEM of 6 independent experiments each). PDE-inhibitors were applied in different doses as short-term infusion. ::: ![](1465-9921-6-76-1) ::: Nebulization of iloprost ------------------------ Inhalation of 0.4 μg/kg aerosolized iloprost resulted in a significant decrease in U46619-induced pulmonary hypertension, from 25.8 ± 0.6 to 22.7 ± 0.4 mmHg P~PA~immediately after stop of nebulization (Table [1](#T1){ref-type="table"}, Fig. [2a,b](#F2){ref-type="fig"}). Pulmonary vascular resistance decreased in response to the prostanoid by 18% (Fig. [3](#F3){ref-type="fig"}). No significant changes in blood gases, cardiac output and systemic arterial pressure were noted (Fig. [4](#F4){ref-type="fig"}). Within \~18 min, 95% of the U46619-induced P~PA~plateau was reached again. The calculated area under the curve (AUC) was 470 ± 49 %ΔP~PA~× min (Fig. [5](#F5){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Influence of iloprost nebulization and its combination with sub-threshold doses of single (a) or combined (b) intravenous PDE inhibitors on U46619-elicted pulmonary hypertension. Pulmonary artery pressure (P~PA~, in % of U46619-induced increase) is given (mean ± SEM of 8 independent experiments each, SEM bars are missing when falling into symbol). Iloprost nebulization (Ilo aer; 0.4 μg/kg) is indicated by the horizontal bar. The PDE inhibitors were pre-applied as short-term infusion as follows: 200 μg/kg × min 8-Methoxymethyl-IBMX, 1 μg/kg × min sildenafil, 5 μg/kg × min motapizone. ::: ![](1465-9921-6-76-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Influence of iloprost nebulization and its combination with sub-threshold doses of intravenous PDE inhibitors on U46619-elicted pulmonary hypertension. Decrease in pulmonary vascular resistance (PVR, in %) at the end of the nebulization period is given (mean ± SEM of 8 independent experiments each). The PDE inhibitors were pre-applied as short-term infusion as follows: 200 μg/kg × min 8-Methoxymethyl-IBMX, 1 μg/kg × min sildenafil, 5 μg/kg × min motapizone. \, p \< 0.05 as compared to pre-nebulization value. ::: ![](1465-9921-6-76-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Influence of iloprost nebulization and its combination with sub-threshold doses of single or combined intravenous PDE inhibitors on systemic arterial pressure (P~SA~). The experiments correspond to those in Fig. 2; the decrease in P~SA~(in % of baseline) at the end of the iloprost nebulization (Ilo aer) period is given (mean ± SEM of 8 independent experiments each). ::: ![](1465-9921-6-76-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Influence of PDE inhibition on the area under the curve (AUC) of iloprost-induced decrease in pulmonary artery pressure (P~PA~). Measurements were performed from onset of iloprost nebulization (Ilo aer) until 68 min post aerosolization. AUC was calculated by standard techniques and is given as %ΔP~PA~× min. The maneuvers and dosages correspond to those depicted in Fig. 2. \, p \< 0.05 as compared to sole iloprost nebulization; †, p \< 0.05 as compared to all other groups. ::: ![](1465-9921-6-76-5) ::: Combined administration of a per se ineffective intravenous PDE-inhibitors and iloprost nebulization ---------------------------------------------------------------------------------------------------- In the presence of sub-threshold motapizone (5 μg/kg × min), the iloprost-induced vasodilatation was significantly increased, with a maximum P~PA~drop of 40.6 ± 4.6% of the U46619-induced pressor response, as compared to 23.8 ± 2.2% for sole iloprost nebulization (Fig. [2a](#F2){ref-type="fig"}). In addition, motapizone enhanced the iloprost induced PVR reduction (37% versus 18%) (Fig. [3](#F3){ref-type="fig"}). Moreover, the duration of the vasodilatory response, defined by P~PA~values below 95% of the U46619-induced pressure plateau, was significantly prolonged, from 18 min to 50 min. The AUC was markedly increased to 1351 ± 192 %ΔP~PA~× min (p \< 0.01; Fig. [5](#F5){ref-type="fig"}). Comparable efficacy was noted for intravenous sub-threshold sildenafil (1 μg/kg × min), which enhanced the maximum P~PA~drop to 36.3 ± 3.6% of the U46619-induced pressor response and PVR reduction to 27%, prolonged the post-nebulization vasodilatory effect to 50 min, and increased the AUC to 1183 ± 136 %ΔP~PA~× min. The combination of sub-threshold 8MM-IBMX with iloprost again enhanced the maximum P~PA~and PVR decrease in response to iloprost aerosolization and prolonged the vasodilatation to 50 min, with AUC values ranging at 1206 ± 177 %ΔP~PA~× min. No further amplification of the iloprost induced vasodilation was achieved by combination with sub-threshold doses of sildenafil plus 8MM-IBMX with an AUC of 1268 ± 115 %ΔP~PA~× min and an amplification of the iloprost effect on P~PA~decrease of 56.4 ± 4.0% (Fig. [2b](#F2){ref-type="fig"}). In contrast, further amplification was noted for the combination of inhaled iloprost with sub-threshold doses of sildenafil plus motapizone, which enhanced the P~PA~drop to 54.6 ± 3.7%, the PVR drop to 35% and increased the AUC to 1993 ± 166 %ΔP~PA~× min. In none of the groups, any significant decrease in systemic arterial pressure (Fig. [4](#F4){ref-type="fig"}) or deterioration of gas exchange (Table [1](#T1){ref-type="table"}) was noted. Discussion ========== The inhalation of nebulized prostanoids for treatment of pulmonary hypertension is an approach targeting selective vasodilatation in well ventilated and aerosol accessible lung regions. This strategy has been developed to circumvent side effects of the conventional intravenous therapy with prostacyclin, e.g. systemic hypotension and ventilation-perfusion mismatch. As anticipated from previous studies in patients with severe pulmonary hypertension \[[@B4],[@B5]\], aerosolization of iloprost was, indeed, effective in causing lung vasorelaxation without any decrease in systemic arterial pressure and any deterioration of gas exchange in the present rabbit model. After completion of the aerosolization maneuver, the iloprost effect levelled off within 20 -- 30 min, which is somewhat more rapid than in the human system (45--90 min). This difference is most likely due to some species variance in the kinetics of the iloprost catabolic pathway: being chemically stable -- in contrast to prostacyclin -- iloprost is converted to dinor-and tetranor-iloprost metabolites via beta-oxidation \[[@B25]\]. The liver is known to be a major site of this catabolic pathway, however, recent studies in isolated perfused rabbit lungs demonstrated that the conversion of iloprost to these beta-oxidation products also takes place in the lung tissue itself \[[@B26]\]. Phosphodiesterases represent the major route for cAMP and cGMP degradation in cells, thereby limiting the downstream effects of adenylate and guanylate cyclase activating agents such as prostacyclin or NO. Until now, 11 types of PDEs have been characterized, which differ in substrate specificity and regulatory properties \[[@B9],[@B27],[@B28]\]. Within the lung, PDE 3 and 4 represent the major cAMP hydrolyzing pathways \[[@B10]\], and monoselective inhibitors of PDE 3 and PDE 4 have been shown to possess pulmonary vasodilatory potency \[[@B8],[@B11],[@B29]\]. Motapizone is a highly selective PDE3 inhibitor, with an IC~50~value of 30 nM \[[@B30]\]. Thus, the finding that motapizone infusion causes dose-dependent pulmonary vasodilatation in intact rabbits with elevated pulmonary artery pressure is well in line with previous observations in this field. Notably, the motapizone effect was not pulmonary selective: in the dose range from 10 to 1000 μg/kg × min, both the pulmonary artery and the systemic artery pressure declined in a parallel fashion. PDE1 gene products are expressed in cardiac tissues from several species \[[@B31],[@B32]\]. A role of increased PDE1C expression in the cardio-protective effect of the stable prostacyclin derivative, 7-oxo-prostacyclin, indicates that PDE1C variants may be involved in tissue responses to cardiovascular stress \[[@B33]\]. In vascular smooth muscle cells derived from different species, several reports demonstrated PDE1 expression. PDE1C was shown to be present in proliferating smooth muscle cells \[[@B31],[@B34]\] and an increased expression of PDE1A1 in rat aorta was shown to contribute to the development of nitroglycerin tolerance \[[@B35]\]. Concerning the lung vasculature, only very limited data on PDE1 expression is available \[[@B11]\]. Our group recently observed that PDE1C is strongly upregulated in the pulmonary artery media of human lungs with severe pulmonary hypertension (R. Schermuly et al., non-published results). It is in line with this notion that the selective PDE1 inhibitor 8MM-IBMX induced dose-dependent pulmonary vasodilation in the presently investigated acute pulmonary hypertension model, however, without being specific for the pulmonary circulation, as evident from the parallel decline of systemic arterial pressure. The cGMP-specific phosphodiesterase PDE 5 is abundantly distributed in the lung tissue \[[@B13],[@B14],[@B36]\]. In a hypoxia-induced model of pulmonary hypertension in the rat, Cohen et al. demonstrated that the PDE 5 inhibitor E4021 selectively vasodilates the pulmonary circulation when being applied intravenously \[[@B36]\], and this observation was confirmed in a model of newborn lambs with persistent pulmonary hypertension \[[@B37]\]. Another specific PDE 5 inhibitor, E4010, has been shown to be a selective pulmonary vasodilatator in a hypoxic rat model of pulmonary hypertension \[[@B38]\]. Accordingly, the PDE 1/5/6 inhibitor sildenafil, which is approved for treatment of erectile dysfunction, was also recently shown to cause preferential pulmonary vasodilatation even when being systemically administered \[[@B39]\]. These data are well in line with the current finding in pulmonary hypertensive rabbits that intravenously infused sildenafil causes a doses-dependent decrease in pulmonary artery pressure, virtually without any decline in systemic arterial pressure. The rationale to combine cAMP-elevating agents, like prostacyclin or the stable prostacyclin analogue iloprost, with PDE3 inhibitors is obvious, and studies from our group already showed a marked amplification and prolongation of the pulmonary vasodilatory response to aerosolized prostacyclin in the presence of type 3 PDE inhibitors \[[@B8],[@B29]\]. The present investigation demonstrates that such synergistic effect also hold true for the PDE3 inhibitor motapizone and the longer acting agent iloprost: in the presence of minute doses of intravenously applied motapizone, the area under the curve of P~PA~decrease in response to the standard inhaled iloprost dose was nearly threefold increased, again without any decline in systemic arterial pressure or any deterioration of gas exchange being detectable. Interestingly, similar potency to increase the response to inhaled iloprost was also noted for subthreshold doses of the PDE 1 inhibitor 8MM-IBMX. This might be anticipated to some extent, as PDE1 causes degradation of both cAMP and cGMP (Fig. [6](#F6){ref-type="fig"}), thereby directly effecting downstream signalling of iloprost, and indirectly modifying this pathway via cGMP sensitive PDE inhibitors. The fact, however, that the area under the curve of PPA reduction was similarly augmented as in the presence of motapizone (\~threefold) suggests that PDE1 inhibitors are worth to be taken into consideration for further strategies to enhance beneficial prostanoid effects in the pulmonary circulation. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Schematic depiction of the crosstalk between different phosphodiesterases (PDE) and the effects of PDE inhibitors on cGMP and cAMP signaling. Different agonists, e.g. prostanoids or nitric oxide (NO) increase the intracellular concentrations of the second messengers cyclic adenosine monophosphate (cAMP) and guanosine monophosphate (cGMP). Phosphodiesterase (PDE) inhibitors stabilize the second messengers and amplify the efficacy of the agonists. 8MM-IBMX selectively blocks PDE1 which can hydrolyze both cyclic nucleotides. Motapizone inhibits PDE3, which hydrolyzes cAMP and is inhibitable by cGMP. Sildenafil blocks PDE5 and PDE1 and can therefore influence the cAMP and cGMP pathway. IP-R, prostacyclin receptor; AC, adenylate cyclase; sGC, soluble guanylate cyclase; smc, smooth muscle cell; PDE, phosphodiesterase. ::: ![](1465-9921-6-76-6) ::: Furthermore, the current study demonstrates that the PDE 1/5/6 inhibitor sildenafil also amplifies the pulmonary vasodilatory response to inhaled iloprost, and that in this respect subthreshold systemic doses of sildenafil are again virtually as effective as subthreshold doses of the PDE3 inhibitor motapizone. This amplification of the iloprost-induced P~PA~decrease again occurred without any decline in systemic arterial pressure and any gas exchange disturbances. The mechanisms underlying this sildenafil effect deserve further elucidation. Given the pharmacological profile of this agent \[[@B40],[@B41]\], it is unlikely that sildenafil caused relevant direct inhibition of lung PDE 3 and 4, thereby promoting prolongation of the half life of cAMP. Its effect may, however, well be explained by the known cross talk between the cAMP and the cGMP pathways: PDE 3 is inhibited by intracellular cGMP with an IC~50~of 0.1--1 μM \[[@B42]\]. A significant inhibition of PDE 5 by sildenafil may thus result in cGMP accumulation, given the fact that there is some permanent baseline stimulation of this pathway via endogenous NO, and next to causing per sesome vasodilatory effect, the sildenafil-induced cGMP may particularly be effective via PDE 3 inhibition and thereby enhanced sensitivity to inhaled iloprost. In addition to such mode of action, supported by previous studies in the cooperativity between the cGMP and the cAMP axis in the lung vasculature \[[@B8],[@B43]\], further (non-cGMP related) effects of sildenafil in the pulmonary circulation may involve inhibition of PDE1 (Fig. [5](#F5){ref-type="fig"}). This view is supported by the above discussed efficacy of selective PDE1 inhibition by 8-MM-IBMX. The IC~50~of sildenafil against PDE1 is about 280 nM \[[@B15]\], and although plasma levels of sildenafil are not addressed in this study, these levels could be achieved after sildenafil application. Thus, both indirect inhibition of PDE3 by increased cGMP and direct inhibition of PDE1 may explain the effects of sildenafil on the iloprost-induced vasodilation. Our studies in co-application of sildenafil with motapizone on the one hand and 8-MM-IBMX on the other hand do, however, favour the sildenafil-PDE1 axis: whereas the combination of sildenafil with the PDE1 inhibitor 8MM-IBMX did not further amplify the iloprost-induced vasodilation over the effect of each agent alone, the combination of sildenafil plus motapizone effected a further strong amplification of the prostanoid induced vasodilation. Besides being of interest as to the mode of action of sildenafil, this finding suggests that an optimum strategy to combine PDE inhibitors may result in even further augmentation of pulmonary vascular prostanoid efficacy as compared to the choice of one selective PDE inhibitor as partner for the prostanoid. In conclusion, intravenous administration of the PDE1 inhibitor 8MM-IBMX, the PDE 3 inhibitor motapizone and the PDE 1/5/6 inhibitor sildenafil causes dose-dependent pulmonary vasodilation in a rabbit model of pulmonary hypertension, with sildenafil possessing selectivity for the lung vasculature. Most interestingly, when applied in subthreshold doses, all PDE inhibitors enhanced and markedly prolonged the vasodilatory response to inhaled iloprost, without any systemic pressure decline or deterioration of gas exchange being detectable. Maximum efficacy was noted upon combination of sildenafil and motapizone. Combination of low dose systemic PDE inhibitors might thus be considered for enhancement and in particular prolongation of the lung vasorelaxant response to inhaled iloprost. Abbreviations ============= P~PA~, mean pulmonary artery pressure; P~SA~, mean aortic pressure; CO, cardiac output; U46619, stable thromboxane-A~2~analogue; PDE, phosphodiesterase; Mota, motapizone; Sil, sildenafil; Ilo, iloprost. Acknowledgements ================ This work was supported by the Deutsche Forschungsgemeinschaft (SFB 547), Project C6.	PubMed Central	2024-06-05T03:55:59.719856	2005-7-20	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1180856/", "journal": "Respir Res. 2005 Jul 20; 6(1):76", "authors": [ { "first": "Ralph Theo", "last": "Schermuly" }, { "first": "Christiane", "last": "Inholte" }, { "first": "Hossein Ardeschir", "last": "Ghofrani" }, { "first": "Henning", "last": "Gall" }, { "first": "Norbert", "last": "Weissmann" }, { "first": "Andreas", "last": "Weidenbach" }, { "first": "Werner", "last": "Seeger" }, { "first": "Friedrich", "last": "Grimminger" } ] }
PMC1180857	Background ========== Since the early 1950s, the concept of spatial conformationin general inorganic, organic and particularly biological chemistry has assumed a fundamental role in the study of the various properties of biological macromolecules (nucleic acids, proteins, carbohydrates, lipids) \[[@B1]\]. Because of the technologies of three-dimensional analysis, this concept is currently used in modern biology. The biological polymers that have been most widely studied in structural and functional terms are proteins and nucleic acids (DNA and RNA) \[[@B2]-[@B5]\]. It is now well established that the information needed to determine the three-dimensional structure of a protein is entirely contained in its linear amino acid sequence. It is likewise known that abrupt changes in environmental conditions (pH, temperature, pressure) may reversibly or irreversibly alter the tri-dimensional structure of a biological macromolecule, and thus change its specific function \[[@B6]\]. However, conformational changeis a still widely discussed concept. The definition of the spatial conformationof either a microscopic or a macroscopic anatomical structure(sub-cellular entity, cell, tissue, organ, apparatus, organism), and the definition of a change or modificationin its shape, are still unresolved problems, much debated by contemporary morphologists \[[@B7]-[@B12]\]. In its general sense, the term structuredenotes the property resulting from the configurations of the parts that form a Whole and their reciprocal relationships to each other and to the Whole itself. On the basis of this definition, two properties of all anatomical systems made up of organised biological mattercan be highlighted: a\. every anatomical structure is capable of expressing a particular function in a particular context; b\. the different configurations and functions of an anatomical entity emerge from structures organised in overlapping hierarchical levels. The term \'organised biological matter\' denotes anything that (1) has its own shapeand dimension, i.e. space-filling property, and (2) can reproduce or replicate itself in such a way as to give rise to \'entities\' that are similar in shape, dimension and functional properties to their progenitors. It is well known that human cells differ in their shapes, dimensions and sizes. All cells making up an adult organism derive from a single progenitor cell, from which arises an enormous number of cells with different shapes, dimensions, sizes, chemical compositions and physiological characteristics in a complex and dynamic process known as cell differentiation\[[@B1],[@B13]\]. Certain cells have specific, particular and consequently invariable characteristic shapes, regardless of whether they are isolated or grouped to form more complex anatomical entities known as tissues(Figure [1](#F1){ref-type="fig"}). However, other cells are subject to conformational changesthat depend particularly on the mechanical action exerted by their environment, the compression induced by contiguous cells, and either the complicated relationships between the cells and the extra-cellular matrixinvolved in the creation of tissue, or the surface tension of the biological fluid in which the cells are immersed \[[@B11],[@B12]\]. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Intra-cellular and/or extra-cellular stimuli determine the shape of an animal cell. In many cases intricate relationships between sub-cellular entities, such as the cytoskeleton, and environmental variables influence the cell\'s shape, dimensionand size. Liver parenchymal cells, called hepatocytes, are roughly polyhedral in situ (a)but when they are dissociated and immersed in a culture medium gradually take on a spherical shape. Tumoral liver cells may drastically change their morphological characteristics, as result of a high number of variables that influence the global behaviour of the cell (b). ::: ![](1742-4682-2-26-1) ::: Liver parenchymal cells (hepatocytes) are roughly polyhedral in situbut, when they are dissociated and immersed in a culture medium, gradually take on a spherical shape (Figure [1](#F1){ref-type="fig"}) \[[@B14],[@B15]\]. It has been widely demonstrated that grouped cells respect the laws of cytomorphogenesis(morphogenetic cell development) by maximally exploiting the space available to them \[[@B7]\]. The variabilityor constancyof cell shape also depends on the physical support provided by the internal cytoskeleton\[[@B16]-[@B20]\]. The fact that all living organisms can be classified on the basis of their appearance is an important indication that each has a specific form(i.e. one that is retained by every example of the same species). The morphological criterion is therefore of considerable importance in identifying and taxonomically classifying living organisms. Our aim here is to give meaning to the complex forms characterising anatomical entities in a similar way to that offered by spatial conformation in the chemical sciences. This attempt has led us to reflect upon and discuss the complex significance of the shape of an observed anatomical entity. Its changes have been defined in relation to variations in its status: from a normal (i.e.natural) to a pathological or altered state, introducing the concepts of kinematicsand dynamicsof anatomical forms, that of speedof their changes, and that of scaleof their observation. The complexity of living systems ================================ Unlike an anatomical entity, and despite the fact that it has a unique shape, a crystalhas no unequivocally defined size that can be used for classification; a small crystal of a given substance will always have the same general structure as a large crystal of the same type. Any fragmentof a crystal has the same physical and chemical characteristics as the whole crystal, but this is not true of any fragment of a living organism because the chemical compositions and physical properties of the individual parts do not correspond with the composition of the Whole. Furthermore, the various components of a living system are characterised by the integrationof precise functional criteria that form a Whole \[[@B21]\]. Returning once again to crystals, their macroscopic structures can easily be predicted on the basis of their microscopic structures; they lack what are called emergent properties: i.e.those that strictly depend on the level of organisation of the material being observed (Figure [2](#F2){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Human beings are complex hierarchical systems consisting of a number of hierarchical levels of anatomical organization (molecules, sub-cellular entities, cells, tissues, organs, apparatuses, and organism) that interrelate differently with each other to form networks of growing complexity. ::: ![](1742-4682-2-26-2) ::: The existence of different organisational levels governed by different laws was first indicated by systemist biologists, who stressed that a fundamental characteristic of the structural organisationof living organisms is their hierarchical nature(Figure [2](#F2){ref-type="fig"}). One of the pre-eminent characteristics of the entire living world is its tendency to form multi-level structures of \"systems within systems\", each of which forms a Whole in relation to its parts and is simultaneously part of a larger Whole. Systemismwas born in the first half of the twentieth century as a reaction to the previous mechanistic movement(also known as reductionism). It was based on an awareness that classical causal/deterministic schemata are not sufficient to explain the variety of interactions characterising living systems. Advances in the fields of cybernetics and biology led to the proposition of new interpretative models that were better suited to identifying and describing the complexity of phenomenathat could no longer be seen as abstractly isolated entities divisible into parts or explicable in terms of temporal causality, but needed to be studied in terms of the dynamic interactions of their parts. The word systemmeans \"putting together\". Systemic understanding literally means putting things in a context and establishing the nature of their relationships, and implies that the phenomena observed at each level of organisation (molecules, sub-cellular entities, cells, tissues, organs, apparatuses and organisms) have properties that do not apply lower or higher levels (Figure [2](#F2){ref-type="fig"}). As we have already said, according to systemic thought, the essential properties of a living being belong to the Whole and not to its component parts. This led to the fundamental discovery that, contrary to the belief of René Descartes, biological systems cannot be understood by means of reduction\[[@B21]-[@B24]\]. The properties of the individual component parts can only be understood in the context of the wider Whole. The biologist and epistemologist Ludwig von Bertalanffy provided the first theoretical construction of the complex organisation of living systems \[[@B25]\]. Like other organic biologists, he firmly believed that to understand biological phenomena, new modes of thought that went beyond the traditional methods of the physical sciences were required \[[@B26],[@B27]\]. According to Bertalanffy, living beings should be considered as complex systems with specific activities to which the principles of the thermodynamics of \"closed\" systems studied by physicists do not apply. Unlike closed systems(in which a state of equilibriumis established), open systems remain in a stationary state far from equilibriumand are characterised by the inputand outputof matter, energyand information\[[@B28]\]. James Grier Miller first introduced the Living System Theory(LST) about how living systems \'work\', how they maintain themselves and how they develop and change \[[@B29]\]. By definition, living systems are open, self-organizing systemsthat have the peculiar characteristics of life and interact with their environment. This takes place by means of information, matter and energy exchanges. The term self-organizationdefines an evolutionary process where the effect of the environment is minimal, i.e.where the generation of new, complex structures takes place fundamentally in and through the system itself \[[@B30],[@B31]\]. In open systems, it is the continuous flow of matter and energy that allows the system to self-organize and to exchange entropywith the environment. Supported by a plethora of scientific data, LST asserts that all the great variety of living entities that evolution has generated are complexly structured open systems \[[@B32]\]. They maintain thermodynamically improbable energy states within their boundaries by continuous interactions with their environments \[[@B32]-[@B34]\]. LST indicates that living systems exist at eight levels of increasing complexity: cells, organs, organisms, groups, organizations, communities, societies, and supranational systems\[[@B29],[@B32]-[@B34]\]. All living systems are organized into critical subsystems, each of which is a structure that performs an essential life process. A subsystem is thus identified by the process it carries out. LST is resulted an integrated approach to studying biological and social systems, the technology associated with them, and the ecological systems of which they are all parts \[[@B35],[@B36]\]. Exploration of the phenomena of life at increasingly microscopic levels (genome) showed that the characteristics of all living systems are encoded in their chromosomesby means of a single chemical substance that has a universal transcription code \[[@B1]\]. In this sense, biological research became largely reductionist (i.e.increasingly involved in the analysis of molecular details). Like its seventeenth-century mechanistic predecessor, it produced an enormous amount of significant data concerning the precise structure of individual genes without knowing how these communicateand cooperatewith each other in the development of an organism and its structural and functional modifications. Through continuing fundamental advances in molecular and cellular biology, molecular biologists discovered the basic building bricksof life, but this did not help them to understand the fundamental integrational processes of living beings \[[@B21]-[@B24]\]. As Sidney Brenner said: \"In one way, you could say all the genetic and biological work of the last sixty years could be considered a long interlude\....We have come full circle -- back to the problems left behind unsolved. How does a damaged organism regenerate with exactly the same structure it had before? How does the egg form the organism? \....In the next twenty-five years, we are going to have to teach biologists another language\....I do not know yet what its name is; nobody does\... \...It is probably wrong to believe that all logic lies at molecular level. It may be that we will need to go beyond the mechanisms of a clock\"\[[@B29]\]. In fact, a new language has emerged over the past few years that makes it possible to interpret and understand living organisms as highly integrated systems \[[@B26],[@B37]-[@B46]\]. Based on the concept of the complexityof the living, this language has given rise to several branches of study concerning the structure and organization of living organisms (such as the fractal geometry of Benoit Mandelbrot and other non-Euclidean geometries \[[@B47]\]) and the biological phenomena that take place within them (such as the Theory of Dynamic Systems, the Catastrophe Theory of René Thom, and the Chaos Theory \[[@B48]-[@B52]\]). The kinematics and dynamics of anatomical forms =============================================== It would therefore be desirable to introduce the concept of the complexity of forminto the anatomical sciences and encourage awareness that an anatomical structure observed at sub-microscopic level is governed by different laws when it is observed at microscopic or macroscopic level (Figure [3](#F3){ref-type="fig"}). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Complex dynamical changes in humans at different level of spatial organization. A.Examples of chromosomal alterations (mutations): a) deletionof a tract of DNA; b) duplicationof a tract of DNA sequence. B.The progressive changes occurring in the nucleus and cytoplasm that accompany the death of a cell. a)Normal cell; b)The nucleus becomes contracted and stains intensely. The cytoplasm is pinker, showing that it binds eosin(a common histochemical stain) more avidly. c)The nucleus disintegrates, appearing as a more or less central area of dispersed chromatin. This phase is called karyorrhexis. d)All nuclear material has now disappeared (kariolysis) and the cytoplasm stains an intense red colour. C.The final appearance of the liver (a)when it assumes the state of cirrhosis (b). Cirrhosis is the final stage of several pathogenic mechanisms operating either alone or in concert to produce a liver diffusely involved by fibrosis (abnormal extra-cellular matrix deposition) and the formation of structurally abnormal parenchymal nodules. D.Human life: from the embryonic stage of morula (a), through that of foetus (b), to the adult being (c). The times elapsing in the variousdynamical processes exemplified (A-D) are very different (simplified by green bars), ranging from nanosecondsto years. It is interesting to highlight the inverse relationshipbetween the level of anatomical complexityand timescale. ::: ![](1742-4682-2-26-3) ::: One of the fundamental problems facing the human mind is that of the succession of forms, introduced by René Thom in his book \"Stabilité Structurelle et Morphogenèse. Essai d\'une théorie générale des modèles\", first published in 1972 \[[@B48]\]. Whatever the ultimate nature of reality may be, it is undeniable that our Universe contains a variety of natural objects and living beings. These things and beings are forms: i.e.structures equipped with a certain morphological and functional stability that occupy a certain portion of space and last a certain length of time. It is a commonplace that the Universe is an incessant birth, development, and destructionof forms \[[@B48]\]. The succession of anatomical formsthus brings us to define: a\. The kinematics of anatomical forms, which studies temporal transformationsof an anatomical form without considering the nature of the entities to which it belongs or what causes changes (Figure [4a](#F4){ref-type="fig"}). When an anatomical form changes, one or more of its qualities is modified in comparison with analogous anatomical forms that are considered unchanged: e.g.a cell can change its shape or one of its associated qualities in a tissue in which other cells remain unchanged. The set of unchanged anatomical forms is called the reference system. A cell can therefore be said to be in a state of morphological stabilityor a phase of modificationin relation to a particular reference system, depending on whether its shape remains the same or varies over time in comparison with the other cells in the system (i.e.the tissue). ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Kinematics and dynamics of human dendritic cells and macrophage differentiation in vitro. Cultured in vitro, monocytes may change their shape, dimension and size when opportunely stimulated by specific growth factors. Kinematics studies these changes without considering the nature of the entities to which they belong or what causes the changes (A). Cultivation in vitrowith Granulocyte Macrophage-Colony Stimulating Factor (GM-CSF) alone or with Interleukin-4 (IL-4) selectively determines differentiation into macrophages or dendritic cells (B). In this case the study of the temporal transformations of primary monocytes in relation to the causes determining the changes, is defined as dynamics of the anatomical forms. ::: ![](1742-4682-2-26-4) ::: b\. The dynamics of anatomical form, which studies the temporal transformationsof an anatomical form in relation to the causes of the changes. An anatomical form in a state of morphological stability tends to preserve its shape in the surrounding space. However, if we apply any (internal or external) factor*u, it abandons this state of \'rest\' and enters a phase of modification(Figure [4b](#F4){ref-type="fig"}). This factor, which can be considered a true physical force, may act on the elements determining the shapeof the system (e.g.in the cell system: the plasmalemma or cytoskeleton) and/or those determining its function or its internal points(e.g.the nucleus, mitochondria, and the smooth and rough endoplasmic reticulum) \[[@B53]\]. The change in shape can be considered as a non-linear dynamic systemthat advances through statesthat are qualitativelydifferent (Figure [4](#F4){ref-type="fig"}). The word \'state\' denotes the pattern configurationof a system at a particular instant, which is specified by a large number of dynamic variables. A dynamic system can be characterised by a set of different states or possible pattern configurations (x) and a number of transitionsor steps (x) from one state to another during a certain time interval (t). When the transitions are caused by a generating element (u), the temporal behaviour of the system can be described by the general equation: x= f(x, u, t) where fis a non-linear functionand the dot denotes a differentiation with respect to time (t). c\. The speed of changeis the time necessary for a change in shape to occur or for the development of a perceptible difference between the modified entity and its unchanged reference system. In quantitative terms, it means the rapidity of the transformation of the anatomical form. However, the parameter timedepends on a large number of variables that are interconnected in a multitude of ways and in a non-linear manner \[[@B53]\]. This makes it extremely difficult to predict the exact time interval between two successive states. Although conformational changes are a continuum, differentiation into successive states is commonly based on differences in shape, dimensionsor functional activity(Figure [4](#F4){ref-type="fig"}). Modelling the complexity of living beings should take into account the 10--12 order-of-magnitude span of timescales for events in biological systems, whether molecular (ion channel gating: 10^-6^seconds), cellular (mitosis: 10^2^-10^3^seconds), or physiological (cancer progression, ageing: 10^8^seconds). d\. The scale of observation, by which is meant the level at which the interrelated parts of a complex structure is being studied. It must be emphasised that observed morphological patterns can often be conceptualised as macro-scalemanifestations of micro-scaleprocesses. However, observed patterns or system states are created or influenced by multiple processes and controls. Furthermore, those multiple processes operate at multiple spatialand temporalscales, both larger and smaller than the scale of observation. It is also necessary to highlight that there is no one \'true\' value for a measurement \[[@B52]\]. The measured value of any property of a biological object depends on the characteristics of the object. When these characteristics depend on the resolution of measurement, then the value measured depends on the measurement resolution. This dependence is called the scalingrelationship \[[@B47]\]. Self-similarityspecifies how the characteristics of an object depend on the resolution and hence determines how the value measured for a property depends on the resolution \[[@B47],[@B52]\]. Conclusive key points ===================== One of the basic problems in evaluating complex living forms and their changes is how to analyse them quantitatively. Although mathematical thought has not had the same impact on biology and medicine as on physics, the mathematician George Boole pointed out that the structure of living matter is subject to numerical relationshipsin all of its parts, and that all its dynamic actions are measurable and connected by defined numerical relationships. Boole saw human thought in mathematical terms and, given its nature, mathematics holds a fundamental place in human knowledge. The origins of the interest of mankind in the mathematics of formgo back to ancient times, when it coincided with the manifestation of specific practical needs and, more generally, the need to describe and represent the surrounding world. The use of geometry to describe and understand realityis essential insofar as it makes it possible to reconstruct the inherent rational order of things. According to Pythagoras, real knowledge was necessarily mathematical. This idea continued until the early years of the seventeenth century, when Galileo re-proposed the observations made by Pythagoras, with no substantial modification, by affirming that the Universe is written in the language of mathematics, whose letters are triangles, circles and other geometric figures. However, during the first half of the twentieth century, it was discovered that the geometric language of Euclid is not the only possible means of making axiomatic formulations, but that other geometries exist that are as self-consistent as classical geometry. This led to the flourishing of new geometrical languages capable of describing new spatial imaginations in rigorous terms. While successive generations of mathematicians were elaborating a large number of new non-Euclidean geometries, the beginning of the twentieth century saw the discovery of mathematical objects that seemed at first sight to be little more than curiosities devoid of practical interest (to the extent that they were even called \'pathological\'). However, in the mid-1970s, the mathematician Benoit Mandelbrot gave them new dignity by defining them as \"fractal objects\" and introducing with them a new language called \"fractal geometry\". Fractal geometry moves in a different developmental direction from the non-Euclidean geometries. Whereas the latter are based on the collocation of familiar objects in spaces other than Euclidean space, fractal geometry stresses the nature of geometric objects regardless of the ambient space. The novelty of fractal objects lies in their infinite morphological complexity, which contrasts with the harmony and simplicity of Euclidean forms but matches the variety and wealth of complex natural forms. In conclusion, we can highlight that the following points: a)Complexity is so pervasive in the anatomical world that it has come to be considered a basic characteristic of anatomical systems. b)Anatomical entities, viewed at microscopicand macroscopiclevel of observation, show different degrees of complexity. c)Complexity can reside in the structureof the system (having many diverse parts with varying interactions or an intricate architecture) or in its behaviour. Often, complexity in structure and behaviour go together. d)A complex system admits many descriptions (ways of looking at the system), each of which is only partially true. Each way of looking at a complex system requires its own description, its own mode of analysis and its own breakdown of the system into different parts; e)Almost all anatomical entities display hierarchicalforms: their component structures at different spatial scales, or their process at different time scales, are related to each other. Application of these concepts promises to be useful for analyzing and modelling the real significance of the shape, dimension and size of an observed anatomical system at a given scale of observation. Further, the changes of the system can be better defined in relation to variations in its status: from a normal (i.e.*natural) to a pathological or altered state.	PubMed Central	2024-06-05T03:55:59.722967	2005-7-19	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1180857/", "journal": "Theor Biol Med Model. 2005 Jul 19; 2:26", "authors": [ { "first": "Fabio", "last": "Grizzi" }, { "first": "Maurizio", "last": "Chiriva-Internati" } ] }
PMC1180858	Background ========== The viral pathogens that infect crop plants constrain food production and economic development throughout the world\'s agricultural regions. Viral diseases are difficult to prevent, and once established few means are available to counter their impact on yield. As a result, development and deployment of resistance crop varieties remains the most effective manner in which to combat the evolving threats presented by plant viral diseases. Underpinning such efforts is the need for robust diagnostic capacities to identify the species and strains of viral pathogens infecting crop plants and their related wild species, and to understand their distribution within a given geographical region. Access to simple, low cost tools for the molecular study of plant viral pathogens is central to generating the knowledge and improved germplasm required by scientists, breeders and farmers to combat these diseases and maximize crop yields. Effective methods for sampling, storage and retrieval of viral pathogens from infected plant tissues allows not only identification of the viral pathogens but also detailed molecular study of their genomes, generating increased understanding of their epidemiology, etiology and evolution. Diagnostic technologies are also required for virus indexing to facilitate certification of pathogen-free materials for the collection, maintenance and international exchange of the elite germplasm on which required plant improvement programs are based. Molecular characterization of the viruses that infect plant material is currently achieved by direct electrophoretical isolation from total nucleic acid, followed by cloning and subsequent analysis, or amplification of full or partial genomic sequences by polymerase chain reaction (PCR). PCR is the more powerful technique due to its ability to recover viral sequences and whole genome components from very low viral titres, and is now the preferred approach for most applications. Currently, total viral and genomic nucleic acids are isolated from infected tissues by methods such as Dellaporta et al. \[[@B1]\] which involve multi-step protocols for DNA or RNA extraction, precipitation and purification. A frequent limitation for studying viruses at the molecular level is the ability to reliably obtain high quality nucleic acids from putatively infected plant material. Plant tissues to be analyzed must be collected and preserved in order to maintain integrity of the nucleic acids until they can be processed. This poses challenges when sample numbers are large and when working in the field, most especially in the tropical and sub-tropical regions where plant viral pathogens are abundant. Field studies are thus constrained by the resources required for sample preservation and transportation, placing restrictions on the number of samples that can be collected in a given time and the size and remoteness of the regions that can be effectively surveyed. Timely processing and/or storage of the samples before they spoil can also be problematic in locations where access to well equipped laboratory facilities is limited. We report here the use of FTA technology for efficient sampling and recovery of viral pathogens from infected leaf tissues and their subsequent molecular analysis. Utilising the geminiviruses that infect maize (Zea mays), cassava (Manihot esculenta) and tomato (Lycopersicum esuclentum), in addition to Tobacco mosaic virus (TMV), Potato virus Y (PVY) and Tobacco etch virus (TEV), we provide evidence that diagnostic techniques can be applied to both DNA and RNA viruses eluted from FTA cards in a manner equivalent to conventional isolation methods, and that this cost-effective technology significantly simplifies the sampling and analysis of diseased plants in both the laboratory and field environments. FTA is a paper-based technology designed for the collection and archiving of nucleic acids, either in their purified form or within pressed samples of fresh tissue. Proprietary chemicals impregnated into the paper act to lyze cellular material and fix and preserve DNA and RNA within the fibre matrix \[[@B2]\]. After a short drying period, pressed samples can be stored at room temperature for extended periods and processed when required. Nucleic acids are recovered by removing small punches from the pressed area and washing with simple reagents. RNA and smaller DNA molecules, such as plasmids and viral genomic components, are eluted by a simple extraction buffer and used as template for amplification by PCR. Genomic DNA remains attached to the paper matrix but is available for amplification by PCR when the paper punch is included in the PCR reaction mix. Advantages of FTA technology have been realized for human DNA processing and forensic applications \[[@B3]\], for wildlife DNA samples \[[@B4]\] and applied to PCR-based genotyping \[[@B5],[@B6]\] but have not been well documented for use with plant pathogens. Recognizing the potential benefits this technology could bring to sampling and molecular study of viral crop diseases, we tested the efficacy of FTA for retrieval of viral pathogens from infected leaf tissues and for the detection of viral-derived transgene sequences in transgenic plants. Results ======= Use of FTA for sampling, retrieval and PCR-based analysis of DNA viruses ------------------------------------------------------------------------ Replicated samples from newly unfolded, symptomatic leaves of cassava, maize, tomato and Nicoticana benthamianaplants infected with geminiviruses were used to study the efficacy of FTA technology for sampling, retrieval and molecular analysis of these viruses. Geminiviruses are composed of monopartite or bipartite genomes of ssDNA, 2.7--2.8 kb in size. As such they can be eluted from the FTA paper matrix after appropriate washing steps and used as template for PCR diagnostic analysis. In order to compare efficacy of FTA technology compared to traditional methods, tissue from each sample was split in two, with one pressed onto an FTA card and the other used for DNA extraction via the Dellaporta method \[[@B1]\]. Universal primers UniF and UniR, designed to amplify the near full-length cassava geminivirus DNA-A component (Table [1](#T1){ref-type="table"}), were used to detect the presence of cassava mosaic geminiviruses (CMG) in infected cassava tissues. All plants sampled in this manner generated signals of the appropriate size. Signals were similar whether the DNA was eluted from FTA cards or extracted by the Dellaporta method (Fig. [2a](#F2){ref-type="fig"}), demonstrating that PCR amplification of sequences equivalent to the whole genomic component of a CMG was possible from samples preserved on FTA cards, in a manner equal to that from traditional DNA isolation techniques. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Oligonucleotides used for PCR amplification of viral and transgene sequences ::: Primer Sequence(5\'-3\') Target sequence ------------ ---------------------------------------------- ------------------------- EAB555/F (5\'-TACATCGGCCTTTGAGTCGCATGG-3\') EACMV DNA B EAB555/R (5\'-CTTATTAACGCCTATATAAACACC-3\') EACMV DNA B JSP001 (5\'-ATGTCGAAGCGACCAGGAGAT-3\') ACMV (AV1/CP) JSP002 (5\'-TGTTTATTAATTGCCAATACT-3\') ACMV (AV1/CP) UniF (5\'-KSGGGTCGACGTCATCAATGACGTTRTAC-3\') CMGs DNA A UniR (5\'-AARGAATTCATKGGGGCCCARARRGACTGGC-3\') CMGs DNA A MSVF (5\'-ATCCCTCCAAATTCCGACAC-3\') MSV MSVR (5\'-TCCATGTACAAAGCTCCTCT-3\') MSV C1F (5\'-GCAGATCTATGCCTCGTTTATTTAAAATATATGC-3\') TYLCV C1R (5\'-GCGGTACCTTACGCCTTATTGGTTTCTTCTTGGC-3\') TYLCV TMVF (5\'-GCGGTGGCGGCCGATCCATGGAACTTACAG-3\') TMV TMVR (5\'-GATTCGAACCCCTCGCTTTAT-3\') TMV POT1 (5\'-gacgaattcTGYGAYGCBGATGGYTC-3\') TEV & PVY POT2 (3\'-ACCACRTADCTBTTAcctaggtcag-5\') TEV & PVY AC1F (5\'-ATGAGAACTCCTCGTTTTAGAA-3\') ACMV-Kenya AC1 AC1R (5\'-ATGAGAACTCCT CGTTTTAGAA-3\') ACMV-Kenya AC1 MP141 (5\'-ATGATTGAACAAGATGGATTGCAC-3\') NptII coding sequence MP142 (5\'-TCAGAAGAACTCGTCAAGAAGGCG-3\') NptII coding sequence EACMV -- East African cassava mosaic virus; ACMV -- African cassava mosaic virus; MSV -- maize streak virus; TYLCV -- tomato yellow leaf curl virus; TMV -- tobacco mosaic virus; TEV -- tobacco etch virus; PVY -- potato virus Y EAB555/F&EAB555/R:PCR conditions consisted of 30 cycles of 94°C for 1 min, 58°C for 1 min. and 72°C for 2 mins. JSP001&JSP002: PCR conditions consisted of 30 cycles of 94°C for 1 min, 45°C for 1 min. and 72°C for 2 mins. Uni F&Uni R: K = G + T, R = A + G, S = G + C. PCR conditions consisted of 30 cycles of 94°C for 1 min., 58°C for 1 min. and 72°C for 2 min. MSV F&MSVR: PCR conditions consisted of 30 cycles of 94°C for 1 min, 59°C for 1 min and 72°C for 2 mins. C1F&C2R: PCR conditions consisted of 94°C for 10 mins followed by 35 cycles of 94°C for 30 secs., 60°C for 1 min and 72°C for 1.5 mins followed and extension of 7 mins at 72°C. AC1F&AC1R: PCR conditions consisted 94°C of 5 mins followed by 35 cycles of 94°C for 30 secs, 58°C for 1 min. and 72°C for 1 mins followed by extension of 10 mins at 72°C. MP141&MP142: PCR conditions consisted 5 mins. at 94°C followed by 35 cycles of 94°C for 30 secs, 58°C for 1 min. and 72°C for 1 mins. followed by extension time of 10 mins. at 72°C. ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### PCR amplification of geminiviruses components from symptomatic leaves of greenhouse-grown plants using traditional DNA isolation methods and FTA technology. (a) amplification of the 2.8 kb B component of cassava mosaic geminiviruses using universal primers UniF and UniR (Table 1) from independent infected plants. Template DNA was obtained either by extraction and purification of total DNA according to Dellaporta et al. \[1\] (0.2 μg template) or by elution of viral DNA from leaf tissue pressed onto FTA Classic Cards. (b) amplification of East African cassava Cameroon virus (EACMCV) (lanes 1--8) and African cassava mosaic virus (ACMV) (lanes 1--5) from diseased cassava plants isolated by Dellaporta-based methods (0.2 μg template) or from viral DNA isolated from leaf tissue pressed onto FTA cards. A 550 bp fragment of the B genomic component of EACMCV was amplified using primers EAB555/F and EAB555/R (Table 1) and a 500 bp fragment of the coat protein gene from the A genomic component of ACMV generated using primers JSP001 and JSP002 (Table 1). (c) amplification of the 1.07 kb C1 gene of the Egyptian strain of the monopartite tomato yellow leaf curl virus eluted from infected tomato (lanes 1,2,5 and 6) and N. benthamiana(lanes 3, 4, 7 and 8) leaves pressed onto FTA cards. Increasing the time which paper punches were soaked in elution buffer from 30 minutes (lanes 1--4) to 12 hours (lanes 5--8) increased the signal strength of the amplified viral sequence in both plant species. In all cases, M: marker, +C: positive control, -C: negative control, W: water control. CMG -- cassava mosaic geminiviruses; ACMV -- African cassava mosaic virus; EACMV -- East African cassava mosaic virus; TYLCV -- Tomato yellow leaf curl virus. ::: ![](1743-422X-2-45-2) ::: For FTA technology to be effectively used as a routine tool for PCR-based geminivirus diagnostics, it must allow for differentiation of the viral species infecting a given plant. Greenhouse-grown cassava plants infected with East African cassava mosaic Cameroon virus (EACMCV) or African cassava mosaic virus (ACMV) were tested for the presence of the specific geminivirus species by pressing symptomatic leaves onto FTA cards and by isolation of total DNA by the Dellaporta method. Two primer pairs, EAB 555F/EAB555R and JSP 001/ JSP 002 \[[@B7]\], designed to amplify all strains of EACMV-like and ACMV-like geminiviruses respectively (Table [1](#T1){ref-type="table"}), were employed to test for presence of the viral pathogens. Both ACMV and EACMV were detected from samples collected on FTA cards. PCR product characteristics were similar between the paper-based and traditional protocols in all thirteen plants analyzed in this manner (Fig. [2b](#F2){ref-type="fig"}). In some FTA derived samples, signal strength from the amplified products was lower than that generated from 0.2 μg of DNA used as template from the Dellaporta method but in all cases remained easily detectable. FTA technology was also used to sample tomato and Nicotiana bethamianaplants infected with an Egyptian strain of the monopartite geminivirus, Tomato yellow leaf curl virus (TYLCV) (GenBank:AY594174). Template DNA eluted from symptomatic leaves pressed onto FTA cards yielded expected bands in all plants tested with primer pair C1F and C1R designed to amplify the C1, replication associated gene (1.074 kb) from this virus (Fig. [2c](#F2){ref-type="fig"}). Investigations were carried out to further quantify the ability of FTA technology to fix, store and release geminivirus genomic components. Recombinant plasmid DNA carrying the B component of EACMCV \[[@B8]\] in quantities as follows: 0.8, 0.4, 0.2, 0.16, 0.08, 0.05, 0.04, and 0.001 μg, were mixed with 8 μl of sap extracted with distilled water from healthy cassava leaves. When these elutions were used for PCR, amplification using primers EAB 555F and EAB555R was successful in all cases except the lowest, where only 0.001 μg was loaded onto the card (Fig. [3](#F3){ref-type="fig"}). In our hands, therefore, FTA technology can be reliably employed to detect geminivirus loads within infected leaf tissues of cassava above the 40 picogram level. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### PCR amplification from serial dilutions of viral DNA elution from FTA cards. Serial dilutions of plasmid DNA carrying known amounts of the B genomic component of East African cassava Cameroon virus (EACMCV) were mixed with leaf sap extract from healthy cassava plants and spotted onto FTA cards. Viral DNA was eluted from FTA cards and used as template for PCR amplification of a 550 bp fragment using primers EAB555F and EAB555R guided experimental design (Table 1). 0.2 μg DNA was used as template in the positive control lane (C+). ::: ![](1743-422X-2-45-3) ::: Improvements to existing protocols ---------------------------------- Two important improvements were developed during the above studies and incorporated into the protocol supplied by Whatman International in order to increase the quality and yield of eluted virus from FTA cards. It was found that in some cases, paper punches removed from leaf tissues pressed into FTA cards retained green pigment after washes with TE buffer and FTA purification reagent. Release of these pigments during the final elution step inhibited subsequent PCR amplification of viral sequences. Addition of one or (if required) two five-minute washes with 70% ethanol prior to the FTA purification reagent step resulted in removal of most green pigmentation from the punches and ensured that the final elution was free from contaminates. It was also found that the yield of eluted viral DNA could be increased, and significantly enhanced amplification of the target sequence achieved, if the processed paper punches were soaked overnight in elution buffer at 4°C, compared to the 15--20 minutes at room temperature recommended by the manufacturer (Fig. [2c](#F2){ref-type="fig"}). These additional steps have become standard procedures in our laboratory. Cloning, sequencing and restriction analysis of viral elutes from FTA --------------------------------------------------------------------- Nucleic acid sequencing provides the highest level of viral diagnostic analysis available and facilitates development of additional tools for subsequent molecular-based studies of these pathogens. To determine whether viral DNA stored on FTA cards was suitable for downstream, high fidelity characterization and analysis, a 550 bp fragment from the DNA B component of EACMCV was PCR-amplified from greenhouse grown, CMG infected plants using primers EAB555/F and EAB55/R (Table [1](#T1){ref-type="table"}). Template DNA eluted from FTA cards and from conventional extraction methods were directly compared. PCR products were purified and cloned into the pGEM-T Easy vector. Two clones from each DNA recovery process were sequenced in both orientations and the DNA nucleotide sequences compared by multiple alignment using MegaAlign software of the DNASTAR computer package. No significant nucleotide sequence variation was observed when a corresponding EACMCV DNA-B sequence fragment (GenBank:AF112355) was compared to the clones sequenced in this study (results not shown). Clones from FTA-processed DNA were comparable with those from phenol extracted DNA, with nucleotide sequence comparison showing 99.8% identity between clones derived from FTA technology and the traditional method of viral DNA processing and phenol purification. These results indicate that viral DNA within plant tissues fixed on FTA card retains fidelity of its nucleotide sequences throughout the sampling, storage and recovery processes. When combined with the ability described above to clone and amplify whole genome sized sequences, we are confident that FTA technology can be employed to generate full genomic sequence data for geminiviruses and to produce infectious clones of these pathogens isolated from diseased plant tissues. Use of FTA to sample, recover and diagnose viruses from field-grown crop plants ------------------------------------------------------------------------------- Having demonstrated efficacy of FTA as a robust tool for sampling and recovery of high fidelity geminivirus DNA from greenhouse-grown material, the technology was tested in farmers\' fields in East Africa. Leaves from cassava and maize plants symptomatic for cassava mosaic disease (CMD) and Maize streak virus (MSV) (Fig. [1a](#F1){ref-type="fig"} and [1b](#F1){ref-type="fig"}) were pressed onto FTA cards in Malawi and Western Kenya. Samples were returned to the DDPSC and processed as described above. Strong signals were recovered in all seven maize samples tested (Fig. [4a](#F4){ref-type="fig"}) using primers designed to amplify a 500 bp fragment from the conserved region of this monopartite geminivirus (Table [1](#T1){ref-type="table"}). Likewise, all cassava samples collected in Malawi proved positive for the presence of EACMV-like geminivirus species (Fig. [4b](#F4){ref-type="fig"}). DNA eluted from FTA-pressed samples of symptomatic cassava leaves from Western Kenya was amplified using Universal primers UniF and UniR. The amplified 2.8 kb product was isolated from the agarose gel and cloned into pGEM-T Easy vector. Viral DNA was amplified by miniprep and subjected to restriction digestion with EcoRV. This enzyme is known to digest all ACMV-like and EACMV-like geminiviruses into unique polymorphic patterns, making it a useful tool for diagnostic analysis of CMD infections to the species and strain level \[[@B9]\]. Of the five plants analyzed in this manner, three were found to contain only EACMV-like viruses, one to be infected with ACMV and one to contain a dual infection with EACMV and ACMV (Fig. [4c](#F4){ref-type="fig"}). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Disease symptoms on field grown cassava and maize plants and FTA sampling method. (a) cassava plant in Western Kenya showing severe cassava mosaic disease symptoms (b) severe maize streak virus symptoms on maize plant in farmer\'s field in Malawi (c) symptomatic leaves are pressed into FTA Classic Cards (d) 2 mm diameter punches being removed from FTA Classic Card for subsequent viral DNA elution and molecular analysis ::: ![](1743-422X-2-45-1) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Analysis of geminivirus DNA eluted from FTA-preserved leaf tissues of infected maize and cassava plants growing in farmer\'s fields in Kenya and Malawi. (a) detection of maize streak virus from infected plants in Malawi. Primers MSV-F and MSV-R (Table 1) were used to amplify a 500 bp fragment from the conserved region of this monopartite geminivirus. (b) detection of East African cassava mosaic virus-like sequences from leaf tissues pressed onto FTA cards plants in Malawi. Primers EAB555F and EAB555R (Table 1) were used to amplify a 550 bp fragment of the B genomic component. (c) Restriction analysis of whole A genomic components (2.8 kb) of East African cassava mosaic virus (EACMV) and African cassava mosaic virus (ACMV) isolated from FTA leaf presses of diseased cassava leaves sampled in Western Kenya. The amplified PCR product was cloned into pGEM-T Easy vector (Promega), the DNA amplified by miniprep and digested with EcoRV for 1.5 hrs at 37°C. Unique bands generated by this restriction enzyme facilitate identification of single infections with EACMV and ACMV (lanes 1--3 and 5 respectively) and a plant co-infected with both geminivirus species (lane 4). M: marker, +C: positive control, W: water control. ::: ![](1743-422X-2-45-4) ::: Use of FTA technology for sampling, retrieval and analysis of RNA viruses ------------------------------------------------------------------------- Since RNA viruses are responsible for the majority of viral diseases in plants, we investigated the efficacy of FTA technology for sampling and retrieval of the commonly studied virus Tobamovirus; Tobacco mosaic virus (TMV) and two potyviruses; Potato virus Y (PVY) and Tobacco etch virus (TEV) potyviruses that infect a large number of plant species. Leaves of N. benthamianasymptomatic for these pathogens were pressed onto FTA cards. PCR amplification of cDNA generated from viral RNA eluted from FTA cards was compared to that isolated via standard methods (where 100--200 mg of leaf tissue was used to isolate total RNA) \[[@B10],[@B11]\]. In all cases PCR signals of the predicted sizes were obtained from both methods (Fig. [5](#F5){ref-type="fig"}). Signal strength generated from FTA derived samples was lower for all three viruses compared to that for RNA isolated by conventional methods. The lower signal strength from FTA samples reflected differences in the amount of RNA obtained by the two methods, and subsequently used as template for cDNA synthesis (40--60 ng/μl from FTA eluted samples compared to 0.4--1.0 μg/μl for conventional isolation). Nevertheless, signals obtained from the FTA cards were sharp and discrete, most especially for TMV (Fig. [5a](#F5){ref-type="fig"}), demonstrating that this technology is applicable as an efficient way of sampling, indexing, retrieving and detecting plant RNA viruses from infected plants. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### *RT-PCR amplification of three RNA viral pathogens recovered from diseased N. benthamianaleaves pressed on FTA cards. Traditional isolation methods (lanes 1--3) were compared to RNA eluted from leaf pressed onto FTA cards (lanes 4--6) (a) a 900 bp fragment of tobacco mosaic virus (b) 1.5 & 2.1 kb fragments of tobacco etch virus (c) 1.5 & 2.1 kb fragments of potato virus Y. In all cases, although generating signals of lower strength compared to traditional RNA isolation methods, RNA eluted from FTA cards proved suitable for detection by RT-PCR analysis. M: marker -C: positive control, W: water control. ::: ![](1743-422X-2-45-5) ::: Use of FTA for the detection of integrated transgenes ----------------------------------------------------- Simple and cost effective tools for monitoring transgenic plants in the field is becoming increasingly important as more developing countries initiate field trials of genetically modified crops. A major goal of the DDPSC is to employ pathogen derived resistance strategies to engineer cassava for elevated resistance to cassava mosaic disease \[[@B12]\] and to test these by carrying out field trials in Africa. In such circumstances it is necessary to assess not only infection of such plants by geminiviruses but also to have simple methods to sample and confirm the transgenic nature of plants within the field. We thus assessed the suitability of FTA technology to act as a reliable tool for PCR amplification of integrated transgene sequences. Leaf tissues from transgenic, virus-free cassava plants were pressed onto FTA cards. Single, 2 mm diameter punches were removed and processed with TE buffer and FTA Purification Reagent, but not subjected to an elution step. Instead, the punch was included in the PCR reaction mix, in addition to primers designed to amplify the AC1and nptIItransgenes (Table [1](#T1){ref-type="table"}) \[[@B12]\]. Amplification signals were successfully generated for both transgenes in all plants tested (Fig. [6](#F6){ref-type="fig"}), indicating that FTA technology is suitable for sampling and detecting both geminivirus-derived and non-geminivirus-derived genomic nucleotide sequences directly from plant tissues. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### PCR amplification of integrated transgene sequences from cassava. (a) amplification of a 800 bp fragment of the nptII selectable marker gene from transgenic plants of cassava cv. 60444 using primers MP141 and MP142 (b) amplification of the 1070 bp AC1transgene integrated into transgenic plants of cv. 60444 using primers AC1F and ACR (Chelleppan et al.,2004) (Table 1). ::: ![](1743-422X-2-45-6) ::: Conclusion ========== The above studies demonstrate that FTA technology is effective for sampling, storage and retrieval of viral pathogens from infected plant tissues growing under the greenhouse and field conditions. Storage and transport of purified nucleic acids on paper for subsequent elution has been common practice for many years. The important advantage brought by FTA technology is the ability to fix and reliably preserve nucleic acids within untreated host tissues. Benefits of this technology are realized at both the sampling and processing phases. Sampling plant material with FTA cards is reduced to simple, on-site hand pressing and is thus rapid and uncomplicated. The ability to store pressed and fixed samples at ambient temperatures also significantly reduces concerns regarding nucleic acid degradation during sampling and storage. Combined with the lack of bulk offered by paper-based collection, the potential number of samples that can be collected within a given time and location is significantly increased compared to conventional methods. Effective retrieval from FTA cards of RNA and DNA viral sequences, plus that of plant genomic DNA, has been demonstrated here. Such capacity eliminates the need for traditional multi-step extraction and purification procedures (some of which require the use of hazardous chemicals) and the requirement for refrigeration and centrifugation equipment. Processing of plant samples is reduced instead to a simple series of washes within a single Eppendorf tube. Importantly, all downstream analytical procedures remain unchanged from existing systems, meaning that no new investment in protocol development is associated with the application of FTA. The technology is also economically effective with samples costing less than \$0.75 each reach the nucleic acid elution stage. Nucleic acid elution and subsequent analysis requires removal of only a few punches from the tissue press, allowing the remainder to be archived for future reference. The length of time that plant viral genomes can be stored on FTA cards was not tested within this study, but research by the manufacturers provides evidence that under ambient conditions the integrity of DNA within pressed tissue samples is maintained for more than fourteen years \[[@B1]\]. If vacuum packed and placed within a fire-proof cabinet, FTA cards provide a long term, low cost, low risk archiving system for viral pathogen and plant genomic samples. The benefits described above have important implications for improving the efficiency of sampling plant tissues in the laboratory environment but increase greatly when working in the field, and most especially within remote areas and in developing countries where access to laboratory facilities, chemicals and equipment is limiting. Results obtained from FTA sampled material were effective and reproducible in our hands from the four plant species studies, whether collected from the greenhouse or returned to the USA from farmer\'s fields in Africa. Predicted PCR products were obtained in 100% and 80% of the cassava leaf samples collected from the greenhouse and field respectively, with all the MSV-infected field grown maize plants sampled yielding viral sequences. In all cases, FTA cards yielded viral nucleic acids of a quality equivalent to that obtained from tradition chemical extraction methods. A full range of diagnostic tools could be applied to viruses eluted from FTA cards, including PCR, RT-PCR, restriction analysis, cloning and nucleotide sequencing. The studies described here demonstrate that FTA offers a simple, sensitive and specific tool appropriate for the diagnosis and molecular characterization of plant viral pathogens isolated from plant tissues and transgene sequences integrated into the plant genome. We conclude that the application of this technology has the potential to significantly increase ability to bring modern analytical techniques to bear on the viral pathogens infecting crop plants. Methods ======= Sampling symptomatic leaves with FTA cards ------------------------------------------ Young, symptomatic leaves were removed from infected plants and placed on FTA^®^Classic Cards. A 16 cm^2^piece of parafilm was placed over the plant tissue and the rounded end of a plastic test tube used to apply moderate downward pressure with a slight twisting action until sap penetrated the reverse side of the FTA paper (Fig. [1c](#F1){ref-type="fig"}). Cards were dried at room temperature overnight and stored in paper bags until required for processing. Elution and amplification of DNA viruses from FTA cards ------------------------------------------------------- Young leaves of cassava, maize and tomato infected with geminivuruses were pressed onto FTA cards as described above. Three 2 mm diameter punches were removed from each chlorophyll-stained region using a Harris 2.0 mm punch (Fig. [1d](#F1){ref-type="fig"}) and placed in a sterile 1.5 ml Eppendorf tube. Paper discs were washed in 300 μl of TE buffer for five minutes followed by sequential five-minute washes with 300 μl of 70% ethanol and FTA^®^Purification Reagent. Punches were then transferred to a fresh 1.5 ml Eppendorf tube and allowed to dry for two hours at room temperature. Viral DNA was eluted by soaking the punches in 10--12 μl of elution buffer (10 mM Tris, 0.1 M EDTA, pH 8.5) for 20--30 minutes and stored at -20°C until required. Two microlitre aliquots of elute were used as template for PCR amplification in a 50 μl reaction volume. Primer sequences and PCR reaction conditions for the respective primers employed are provided in Table [1](#T1){ref-type="table"}. DNA was also extracted by traditional methods based on those of Dellaporta et al. \[[@B1]\]. For serial dilutions, recombinant plasmid DNA carrying the B component of EACMCV \[[@B8]\] in quantities; 0.8, 0.4, 0.2, 0.16, 0.08, 0.05, 0.04, and 0.001 μg was mixed with 8 μl of sap extracted with distilled water from healthy cassava leaves, loaded onto FTA cards and allowed to dry at room temperature for two hours. The stained region from each sample was cut from the FTA paper and processed as described above to elute viral the DNA. Cloning and nucleotide sequencing of viral DNA eluted from FTA cards -------------------------------------------------------------------- Template DNA eluted from FTA cards and obtained from Dellaporta-based conventional extraction methods \[[@B1]\] was amplified by PCR as described above. PCR products were purified and cloned into the pGEM-T Easy vector (Promega) and sequenced in both orientations. DNA nucleotide sequences were compared by multiple alignment using Mega Align option of DNASTAR computer package. A corresponding fragment sequence of EACMCV DNA B from (GeneBank: (AF112355) was used as a reference. Elution and RT-PCR amplification of RNA viruses from FTA cards -------------------------------------------------------------- Symptomatic leaves from tobacco plants infected with TMV, PVY and TEV were pressed onto FTA cards as described above. Eight disc, 2 mm in diameter were removed from the chlorophyll-stained region of each pressed sample and placed into a RNase-free/ DNase-free 1.5 ml Eppendorf tube. Five hundred μl of fresh RNA processing buffer (10 mM Tris-HCl, pH 8.0, 0.1 mM EDTA, 800 U/mL RNase Out™ {Invitrogen Life Technologies, Inc., USA} 200--250 μg/mL glycogen and 2 mM DTT) was added to each Eppendorf tube and incubated on ice with mixing every 5 min for a total of 30 mins. The paper discs were removed and eluted RNA precipitated using 1/10th volume of 3 M sodium acetate (pH 5.2), an equal volume of ice cold 100% isopropanol and incubated at -70°C for 30 mins. RNA was pelleted and washed with ice-cold 75% ethanol, dried and resuspended in 30 μl of DEPC-treated TE/ H~2~O. Isolation of RNA also performed by the above method directly from 100--200 mg of symptomatic fresh leaf tissue fresh. For TMV, total RNA was used directly for cDNA synthesis, but for PVY and TEV that possess polyA tails, messenger RNA was purified from total RNA using Oligotex-dT (Qiagen) according to the manufacturer\'s instructions. cDNA was synthesized in a 20 μl reaction using superscriptIII reversetranscriptase (Invitrogen). For TMV, 0.4--1.0 μg total RNA was used per reaction with TMV specific TMV-R reverse primer (Table [1](#T1){ref-type="table"}), and for PVY and TEV 30--100 ng mRNA was used with the potyvirus universal (POT1) reverse primer (3\'-ACCACRTADCTBTTAcctag gtcag 5\'). Subsequently, PCR was done on cDNA to amplify CPand MPsequences for TMV and the conserved region of CPand NiBfrom PVY and TEV potyviruses using specific primers (Table [1](#T1){ref-type="table"}). PCR amplification of genomic DNA sequences from FTA cards --------------------------------------------------------- Leaf tissues from transgenic, virus-free cassava plants were pressed onto FTA cards as described above. Single, 2 mm diameter punches were removed from chlorophyll-stained regions and placed in a sterile 1.5 ml Eppendorf tube. Punches were washed sequentially for five minutes each with 300 μl of TE buffer, 70% ethanol and FTA Purification Reagent, followed by transfer to a fresh 1.5 ml Eppendorf tube where they were allowed to dry for two hours. Single punches were added to 50 μl PCR reaction mixes containing primers MP141 and 142 for the amplification of the nptII coding sequence, and primers AC1F and AC1R to amplify the AC1*transgene sequence \[[@B12]\] (Table [1](#T1){ref-type="table"}). Competing interests =================== The authors declare that have no competing interests. Whatman International Inc. provided free samples of their FTA technology to facilitate early stages of the work reported here. None of the authors received payment from Whatman to undertake this research. Authors\' contributions ======================= JN developed FTA technology for isolation and detection of geminiviruses, carried out the cloning, sequencing and restriction analysis and produced the initial manuscript draft. NJT conceived the use to FTA for detection of DNA viruses, applied FTA for sampling cassava in the field in Africa, generated data on maize streak virus and use of FTA for PCR amplification of genomic sequences and prepared the final versions of the manuscript. JY carried out all RNA work described above. HA adapted FTA technology for detection of TYLCV and make discoveries to improve recovery of geminiviruses from FTA cards. JL conceived and contributed to the use of FTA in the field in Africa and provided critical input in drafting the manuscript. TA interpreted data, corrected the manuscript and provided supervision of JN. GT guided experimental design and corrected the manuscript. CMF applied FTA technology in the field to collect MSV from diseased maize plants, guided experimental design, conceived the use of FTA for detection of RNA viruses, provided overall supervision and financial support and corrected the manuscript. Acknowledgements ================ The authors thank personnel at Whatman International Inc. for technical advice and DDPSC Plant Growth Facility staff Ed Fischer and Jodie Nelson for plant care. This work was supported by the Donald Danforth Plant Science Center.	PubMed Central	2024-06-05T03:55:59.725540	2005-5-18	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1180858/", "journal": "Virol J. 2005 May 18; 2:45", "authors": [ { "first": "Joseph", "last": "Ndunguru" }, { "first": "Nigel J", "last": "Taylor" }, { "first": "Jitender", "last": "Yadav" }, { "first": "Haytham", "last": "Aly" }, { "first": "James P", "last": "Legg" }, { "first": "Terry", "last": "Aveling" }, { "first": "Graham", "last": "Thompson" }, { "first": "Claude M", "last": "Fauquet" } ] }
PMC1180859	Background ========== Most of gastric adenocarcinoma can be simply diagnosed by microscopic examination of biopsy specimen. Extremely well differentiated adenocarcinoma (EWDA) of the stomach is histologically too bland and too similar to benign foveolar epithelium to make a diagnosis as malignancy. Till date, several cases of EWDA of the stomach were reported by Japanese authors. But the cases reported as EWDA were heterogeneous groups histologically and phenotypically. Most of reported cases revealed well differentiated adenocarcinoma mimicking complete type intestinal metaplasia with intestinal mucin phenotype \[[@B1],[@B2]\]. Only few cases corresponded to EWDA of the stomach mimicking reactive foveolar epithelia with gastric mucin phenotype \[[@B3]\]. Recently I experienced a case of EWDA of the stomach which was very similar to benign foveolar epithelia histologically and phenotypically and reminiscent of gastric counterpart of adenoma malignum of the uterine cervix \[[@B4]\]. Case presentation ================= A 67-year-old male presented with a gastric mass incidentally found on the abdominal computed tomography (CT) for routine medical examination. Abdominal CT showed a fungating tumor at the gastric cardia and several lymph node enlargements at the left gastric and celiac axis. Gastric endoscopic examination revealed a huge fungating mass at the cardia, and subsequently mucosal biopsy was performed. Microscopically the biopsy specimen showed proliferations of bland looking hyperplastic foveolar epithelia with basally located small nuclei and fine nuclear chromatin in the heavy inflammatory background (Figure [1A](#F1){ref-type="fig"}). Some glands were destructed by inflammatory cell invasion and revealed mild epithelial atypia with mildly increased nuclei and loss of nuclear polarity reminiscent of reactive cellular atypia (Figure [1B](#F1){ref-type="fig"}). This biopsy specimen was diagnosed as foveolar epithelial hyperplasia. However, the clinical and endoscopic features of this patient were strongly suggestive of malignancy. The patient underwent radical total gastrectomy with Roux en Yanastomosis. The extent of lymph node dissection included first and second lymph node groups. The resected stomach revealed a huge fungating tumor (Borrmann type 1) at the cardia (Figure [2A](#F2){ref-type="fig"}). The tumor measured 7 cm in the greatest diameter. The cut surface of the tumor was whitish gelatinous and the tumor involved the whole layer of the gastric wall (Figure [2B](#F2){ref-type="fig"}). The remaining gastric mucosa is grossly unremarkable. Microscopic feature of the resected specimen was very similar to that of the biopsy specimen except an evidence of deep invasion. Microscopically the tumor was sharply demarcated from surrounding mucosa and composed of proliferations of deceptively bland glands lined by mucin-rich columnar cells with small basal nuclei (Figure [3A](#F3){ref-type="fig"}). Many glands are cystically dilated especially in deep portion and their glandular lumina were filled with abundant mucin (Figure [3B](#F3){ref-type="fig"}). Most of glands were too bland to discriminate from benign foveolar epithelial hyperplasia (Figure [4A](#F4){ref-type="fig"} and [4C](#F4){ref-type="fig"}), but some glands were more complicated or branched with mild to moderate cellular atypia revealing increased nuclei with loss of polarity and prominent nucleoli (Figure [4B](#F4){ref-type="fig"}). There was no evidence of individual cell invasion into lamina propria or solid growth of tumor cells. Chronic and acute inflammatory infiltrate was heavily associated within tumor. Well formed bland glands invaded to the serosa with focal desmoplastic reaction in adjacent stroma. Vascular and perineural involvements were associated. Tumor cells metastasized to 6 out of 76 regional lymph nodes. Metastatic tumor cells within regional lymph nodes were also very bland (Figure [4D](#F4){ref-type="fig"}). The pathological tumor stage corresponded to stage IIIA (T3N1M0). Immunohistochemically the tumor cells revealed no overexpression of p53 protein but high Ki-67 labelling index suggesting high proliferative activity (Figure [5D](#F5){ref-type="fig"}). The tumor cells and intraluminal mucin were diffusely expressed MUC1 (Figure [5A](#F5){ref-type="fig"}) and MUC5AC (Figure [5B](#F5){ref-type="fig"}) suggesting gastric foveolar phenotype. MUC2 expression was only focally detected (Figure [5C](#F5){ref-type="fig"}). The patient has been underwent adjuvant chemotherapy. Abdominal CT taken after 12 months suggested peritoneal carcinomatosis, multiple metastatic foci in the lung, and multiple retroperitoneal lymph node enlargements. The patient has survived with an evidence of multiple distant metastases for 18 months after operation. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Microscopic photograph of the biopsied specimen. A. Bland glands mimicking benign foveolar epithelial hyperplasia are noted in the heavy inflammatory backgrounds (hematoxylin and eosin, X400). B. Glands are destructed by inflammatory cell infiltrates and epithelial cells reveal mild atypia (hematoxylin and eosin, X400). ::: ![](1477-7819-3-28-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Macroscopic photograph of the tumor. A. Resected stomach reveals huge fungating tumor at the cardia. B. The cut surface of the tumor is whitish gelatinous and the tumor involves the whole layer of the gastric wall. ::: ![](1477-7819-3-28-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Low power view of microscopic photograph of the tumor. A. Tumor glands proliferate haphazardly with papillary configuration at the surface (hematoxylin and eosin, X40). B. Cystically dilated glands invade to proper muscle and subserosa (hematoxylin and eosin, X40). ::: ![](1477-7819-3-28-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### High power view of microscopic photograph of the tumor. A. Bland glands are lined by mucin-rich columnar cells with basal nuclei (hematoxylin and eosin, X400). B. More complicated glands are lined by more atypical nuclei with prominent nucleoli (hematoxylin and eosin, X400). C. Very benign looking glands are present within muscle (hematoxylin and eosin, X400). D. Metastatic tumor glands within lymph node show insignificant cellular atypia (hematoxylin and eosin, X400). ::: ![](1477-7819-3-28-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Immunohistochemical finding of the tumor. The tumor cells and intraluminal mucin are diffusely expressed MUC1 (A) and MUC5AC (B). MUC2 expression is focally noted in some tumor cells (C). Tumor cells reveal high Ki-67 labeling index (D). (Immunohistochemistry, X100). ::: ![](1477-7819-3-28-5) ::: Discussion ========== Traditionally, gastric carcinomas have been classified into two main types, the so-called intestinal and diffuse, based on the tendency toward gland-formation \[[@B1]\]. Recently using immunohistochemical staining techniques specific for gastric- and intestinal type mucins, the phenotype of each tumor was histologically reclassified \[[@B5],[@B6]\]. Of these, MUC1 and MUC5AC are expressed in the superficial foveolar epithelium of gastric mucosa, and MUC2 is expressed in goblet cells of intestinal mucosa or intestinal metaplastic cells of stomach. Up to date several cases of EWDA of the stomach have been reported in literatures \[[@B1]-[@B3]\]. The reports described histologically and phenotypically heterogeneous groups of EWDA and their results were not inconsistent with the present case. Niimi et al, described the usefulness of p53 and Ki-67 immunohistochemical analysis for preoperative diagnosis of EWDA of the stomach. However, in the present case as well as Nokubi et al\[[@B3]\]\'s case which was very similar to the present case in all respects, p53 overexpression was not observed in the EWDA. These two cases were equally associated with predominant gastric phenotype, which reflect the fact that gastric type adenocarcinoma of the stomach is less likely associated with p53 mutation pathway \[[@B6]\]. The immunostaining for high KI-67 was helpful to distinguish EWDA from benign foveolar epithelia in the present case. EWDA of the stomach is a rare highly differentiated adenocarcinoma in which most of the glands are impossible to distinguish from benign foveolar glands, particularly in biopsy specimen \[[@B3]\]. In the present case the resected specimen EWDA showed variable histologic features area by area. Although most of tumor glands were lined by deceptively bland, mucin-rich columnar cells with basal nuclei, more atypical areas were also detected at least focally. The more atypical areas were composed of complex or haphazard arrangement of the glands as well as increased nuclei with loss of polarity and prominent nucleoli. Because the most reliable diagnostic criteria of EWDA of the stomach like adenoma malignum of uterine cervix are deeper invasion and/or metastasis, it is difficult to make a correct diagnosis in biopsy specimen. However, multifocal and repeat biopsies and careful microscopic examination can elicit the recognition of more atypical areas suggesting malignancy. The present case was misinterpreted as benign foveolar epithelial hyperplasia for the biopsy specimen, but the radiological and endoscopic findings suggested malignancy strongly. The clinicopathologic correlation is also mandatory in cases of EWDA of the stomach. First of all, to keep in mind of the entity of EWDA is essential to reach to a correct diagnosis. Conclusion ========== The clinicopathologic profiles of gastric extremely well differentiated adenocarcinoma of gastric phenotype include cardiac location, fungating gross type, very similar histology to foveolar epithelial hyperplasia, foveolar mucin phenotype, lack of p53 over expressoin and high proliferative index. In gastric EWDA of gastric phenotype, the unique criterion of malignancy is an evidence of deeper invasion. It is good reason for considering this tumor as gastric counterpart of adenoma malignum of the uterine cervix. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= WAL performed pathologic examination literature search and preparation of manuscript. Acknowledgements ================ Permission of the patient was obtained for publication of his case records. This research was conducted by the research fund of Dankook University in 2003.	PubMed Central	2024-06-05T03:55:59.728513	2005-5-23	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1180859/", "journal": "World J Surg Oncol. 2005 May 23; 3:28", "authors": [ { "first": "Won", "last": "Ae Lee" } ] }
PMC1180860	Background ========== Inflammatory fibroid polyp (IFP) is a benign, nonmetastasing tumour of the digestive tract. It has been described to occur commonly in the stomach and intestines. Cases have been reported in rectum \[[@B1]\], duodenum \[[@B2]\] and oesophagus, though rare. It is a slowly growing tumour unlike this case which had rapidly grown in a period of five months. The usual presentation is either with dyspeptic or obstructive symptoms. The treatment is usually confined to local excision of the lesion either endoscopically or by an open procedure. The details of the clinical presentation, histopathological findings and therapeutic choices are discussed. Case presentation ================= A 76-year-old man presented with epigastric pain and dysphagia. Endoscopy revealed grade 1 oesophagitis with a normal stomach and duodenum. Five months later he was readmitted with anorexia, weight loss, dysphagia and anaemia. Endoscopy on this occasion revealed a large pedunculated polypoidal lesion arising from the gastro-oesophageal junction within a hiatus hernia. Pathological examination of biopsies taken from the polyp showed inflammatory changes in squamous mucosa with no evidence of malignancy. As the lesion was pedunculated it was decided to attempt endoscopic snare excision which proved unsuccessful, as the lesion was too large for the snare. At laparotomy a large polypoidal mass was found arising from the distal oesophagus. Through a gastrostomy the lesion was delivered into the abdomen, a linear stapler was applied to the base of the stalk and the lesion was excised. The patient made an uneventful recovery. The resected specimen (figure [1](#F1){ref-type="fig"}) measured 9 cm × 4 cm × 4 cm and consisted of soft and slimy tissue. Histology showed an intact squamous mucosa and a mass of myxoid fibroblastic tissue with plasma cells and a diffuse eosinophilic infiltrate (figure [2](#F2){ref-type="fig"}). There was no evidence of malignancy. These appearances were consistent with inflammatory fibroid polyp of the oesophagus. Immunohistochemical studies were negative for S 100 and smooth muscle actin excluding neural or smooth muscle origin. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Resected specimen of the polyp measuring 90 × 40 × 40 mm. ::: ![](1477-7819-3-30-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Thick and thin walled blood vessels, stellate fibroblasts, scattered lymphocytes and eosinophils. Shown clearly in the inset view H&E 100. ::: ![](1477-7819-3-30-2) ::: The patient underwent endoscopy 6 months after resection and there was no evidence of recurrence. He is currently asymptomatic 1 year after resection. Discussion ========== Inflammatory fibroid polyp (IFP) is a rare tumour of the digestive system. It was first described as a distinct pathological entity by J.Vanek in 1949 when he described 6 case reports of gastric submucosal granulomas with eosinophilic infiltration \[[@B3]\]. Since then there have been sporadic case reports but it is difficult to establish the true incidence. The most common site for this lesion in the intestinal tract is the stomach where an incidence of 4.5% of all gastric polyps has been reported \[[@B4],[@B5]\]. Our patient had an oesophageal lesion which is an unusual location although there have been 9 reported cases studies in the literature \[[@B6]-[@B12]\]. It usually occurs as a solitary lesion, the distal 1/3^rd^of the oesophagus being the commonest site for this tumour. Inflammatory fibroid polyp is a benign lesion of uncertain origin \[[@B7]\] some reports claiming it to be myofibroblastic \[[@B13]\] in origin and others suggest it arises from vascular or perivascular tissue \[[@B14]\]. The main characteristics of this lesion are eosinophilic infiltration and presence of characteristic connective tissue stroma. It is generally accepted that this is not a neoplasm, but is reactive process \[[@B8]\] to physical, chemical or microbiological stimuli. The presentation is dependant on the size, location or complications of the tumour. In those of the stomach, epigastric pain and bleeding are the common symptoms and those of the intestine \[[@B15]\] colicky abdominal pain is the most common symptom. In the oesophagus it can present with dysphagia, bleeding or gastro-oesophageal reflux symptoms \[[@B6],[@B7],[@B11],[@B16],[@B17]\]. The clinical presentation in our case was with progressive dysphagia and melaena. IFP has a submucous location and grows intraluminally. There is no reported evidence of metastatic spread in literature. It is difficult to differentiate IFP from the other lesions that occur in the oesophagus like leiomyoma, lipoma, pedunculated papilloma, fibrovascular polyp. The definitive diagnosis is histological. Histological differentiation is from fibrovascular polyp which shows mature fibrous tissue with scattered thin-walled blood vessels lined by flat non reactive endothelial cells with most of them containing fat. Differentiation from leiomyoma is based on the finding that bundles of nonneoplastic smooth muscle can be seen at the level of pre-existing muscularis propria in case of an IFP. The patient was initially evaluated with an endoscopy, which showed grade1 oesophagitis. Subsequent endoscopy localised the lesion as arising from the distal oesophagus near to the gastro-oesophageal junction, and had a benign appearance. The discovery of the lesion on endoscopy within five months of a previous normal endoscopy does give an indication that it may be rapidly growing \[[@B18]\] The previously reported studies showed that the usual presentation of IFPs varied from months to years, but we could not attribute any cause for the rapidity of growth of this lesion in our patient. Investigation modalities that are used for these lesions include radiographs, endoscopic ultrasound, CT scan and MRI. In our case in view of the acute presentation these were deferred. The treatment is confined to surgical excision of the lesion by endoscopic or open method, the decision based on the site and size of the tumour. There are cases which had to be resected by a lateral oesophagotomy or formal oesophagectomy \[[@B12]\]. There has also been reported use of Thermocautery and Nd YAG Laser in the treatment of small polyps. Our patient had an initial attempt of resection of the tumour endoscopically but in view of its size it had to be abandoned and subsequently resected after gastrostomy. Conclusion ========== The commonest site of IFP in the GI tract is stomach; IFP of the oesophagus is rare. Treatment is by excision that can be carried out endoscopically. LASER ablation and coagulation has been tried. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= SKG- wrote the original manuscript, review of literature, prepared requested revisions RTD- performed surgical resection, helped to draft the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ The consent of patient\'s next of kin was obtained for publication of this case.	PubMed Central	2024-06-05T03:55:59.729404	2005-5-30	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1180860/", "journal": "World J Surg Oncol. 2005 May 30; 3:30", "authors": [ { "first": "Shashi Kanth", "last": "Godey" }, { "first": "Robert T", "last": "Diggory" } ] }
PMC1181243	Background ========== Cardiovascular events represent a major health burden for modern societies and the acute coronary syndrome (ACS), defined as myocardial infarction or unstable angina, accounts for a large percentage of them \[[@B1]\]. Practice guidelines regarding ACS management have been widely published \[[@B2]-[@B4]\]. However, gaps have been observed between practice guidelines and how ACS is actually managed. Moreover, regional variations have been observed in the ACS management and outcomes \[[@B5]-[@B9]\]. This situation raises the need to better understand the link between treatment, care and outcomes in patients suffering from an ACS as well as to determine whether or not these relationships varied spatially. This article focuses on the link between invasive cardiac procedure (ICP), hospital death (HD), hospital length of stay (LoS) and early hospital readmission (EHR). ICP requires technical facilities and a professional expertise available only in specialized cardiology centers. For patients suffering from an ACS, exposure to ICP during the first hospitalization for myocardial infarction is considered to be protective of a readmission for cardiac reasons \[[@B10]\]. What is less known is the effect ICP may have on HD, LoS and EHR, and how this effect may vary from one region to another. Although the geographical localization of specialized centers for stroke treatment in Canada \[[@B11]\] has been described in terms of theoretical accessibility to specialized care for a specific population, a more comprehensive ACS model taking into account ICP, HD, LoS, EHR as well as demographic and geographic variables remains to be described. Using descriptive, comparative, and spatial analysis tools as well as cartographic representation, we assessed inter-regional disparities in the management of ACS. More specifically, our main objectives were to describe and compare the regional rates of ICP, HD, EHR, and also the average LoS after an ACS in 2000 in the province of Quebec. We also assessed whether there is a relationship between ICP and HD, LoS, and EHR. Secondary objectives intended to determine if the proximity of a specialized cardiology center influences the ICP, HD, LoS and EHR. Finally, we assessed the extent to which the relationships between these variables vary spatially. Methods ======= Design ------ We conducted a population-based cohort study using data from the Quebec\'s hospital discharge register. This register provides administrative data on patients hospitalized in acute care hospitals in the province of Quebec. Studies confirming the validity of the administrative hospital discharge data concerning myocardial infarction have previously been published \[[@B12],[@B13]\]. Studied population ------------------ For the studies on ICP, HD, and LoS, the studied population consisted of all patients 25 years and older living in the province of Quebec, who have been hospitalized in Quebec for ACS between January 1^st^and December 31^st^2000. For the study on EHR, we have selected those patients who survived the index hospitalization. The \"index hospitalization\" was the first hospitalization during the study period. We included patients who were hospitalized for acute myocardial infarction (code 410 of the International Disease Classification, 9^th^revision (IDC-9)) or other acute or subacute forms of ischemic cardiopathy (IDC-9 code 411) as the main diagnosis. Patients with an unknown geographic location code were excluded. Data sources ------------ Attributive and spatial data were used. Attributive data included all patient-data. This was obtained from the Quebec\'s hospital discharge register and the death register. Each patient was spatially referenced by his/her postal code of residence using data from ESRI \[[@B14]\], DMTI Spatial \[[@B15]\] and from the Quebec Ministry of Health and Social Services \[[@B16],[@B17]\]. The geographic coordinate system used for the cartographic presentation was GCS North American 1983. Studied variables ----------------- Patients were considered to have undergone an invasive cardiac procedure (ICP) if there was mention of an angiography, angioplasty or aortocoronary bypass as coded in the Quebec\'s hospital discharge register (Canadian Acts Codes beginning with 480 to 483 or 489) for the index hospitalization. The hospital death (HD) was defined as in-hospital death at the index hospitalization. The length of stay (LoS) was defined as the number of days between the patient\'s admission and discharge from the index hospitalization. If the care for ACS was delivered over several contiguous hospitalizations involving hospital transfer, the presence of an ICP, HD and the LoS were evaluated for the entire episode. The early hospital readmission (EHR) was defined as a hospital readmission for heart disease as the main diagnosis (ICD-9: 410 to 414) in the first 30 days following discharge from the index hospitalization. Distance to a specialized cardiology center, geographic location and the patient\'s age and gender were used as covariates. Specialized cardiology centers are hospitals that provide technical facilities and professional expertise in cardiology. These centers are the only hospitals that provide invasive cardiac procedures. The 16 specialized centers were identified via the Quebec tertiary cardiology network. The aerial distance to a tertiary cardiology center was categorized into a variable taking the value 1 if the residence was within an aerial distance of 32 km from the nearest specialized cardiology center (as defined by the geometrical centroid of the postal code area), the value 2 if the residence was between 32 and 64 km from the nearest cardiology center, the value 3 if the residence was between 64 and 105 km, and finally, the value 4 if the residence was farther than 105 km from the nearest cardiology center. We chose these cut points based on transportation time of respectively 60, 90 and 120 minutes to cover a distance of 32, 64 and 105 km \[[@B11],[@B18]\]. The localization of the residence was defined by the geometrical centroid of its postal code. The geographic grouping was based on the health administrative region of the patient\'s home location. Because of their geographic similarities and their small populations, the Nunavikand James Bay Cree Landswere merged with the Northern Quebec region. Statistical analyses -------------------- Descriptive analyses by age, gender and place of residence (administrative region) were done. Incidences of hospitalization for ACS were calculated in regard to the estimated population for 2000 \[[@B19]\]. We used the Pearson χ^2^test for comparisons between proportions and the Fisher F-statistics for comparisons of means (ANOVA after a logarithmic transformation) in the studied groups (age, gender, and region) \[[@B20]\]. For the 16 Quebec administrative regions, we calculated the standardized ICP, HD, and EHR ratios \[[@B21]\], as the ratios between observed and expected numbers given age and gender. We also calculated the standardized LoS as the age and gender weighted average of the LoS. Using a Hierarchical Cluster Analysis \[[@B22]\], we grouped in different classes the standardized ICP, HD, and EHR ratios as well as the LoS weighted average for the 16 regions. Hierarchical clusters analysis (HCA) is based on measures of similarities (defined by the squared Euclidean distance between the values) computed from values of one or several variables; at a first step, each case forms a cluster, then the two nearest or similar clusters are grouped to form a new cluster, and so on, until an appropriate number of clusters is reached or until all cases are grouped into a unique cluster. The centroid clusteringmethod was used. The choice of the number of clusters was based on visual inspection of the dendrograms produced by the method, the idea being to display a sufficient number of clusters in order to have homogeneous groups (within-group homogeneity) but dissimilar enough (inter-group heterogeneity) (see Figure [1](#F1){ref-type="fig"} for an example of dendrogram). In most of the methods of classification available in ArcGIS \[[@B23]\] (equally spaced, quintiles, natural breaks, mean and standard deviation method, etc), the number of groups must be a priorifixed, and the choice of the method often depends on the distribution of the data. We chose to use a HCA instead of these usual grouping methods because HCA can be used regardless of the data distribution and because we can choose the appropriate number of groups after a visual inspection of the dendrogram. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Dendrogram produced by a hierarchical cluster analysis of the standardized ICP ratios. ::: ![](1471-2261-5-21-1) ::: Multiple linear (log-normal) regression analysis was performed on LoS and multiple logistic regression analyses were performed on ICP, HD and EHR. Beyond age and gender, the potential predictors of the average log-transformation of LoS, the HD rate and the EHR rate were the presence of an ICP and the distance to a specialized cardiology center. All main effects were included in the models, but we allowed some interaction terms, gender× ICP, distance× ICP, and Gender× distance, to enter the models only if they were statistically significant at a 0.05 level. The parameter estimates between outcomes and predictors together with their 95% confidence intervals were calculated. The residuals derived from the global models were also calculated at a regional level and mapped. In the case of the log-normal model of LoS, we used the usual residuals measured as the difference between the observed mean and the mean of the expected (after a log-transformation of LoS). The residuals used in the logistic models for ICP, HD and EHR were, at a regional level, the signed deviance residuals, given by: ![](1471-2261-5-21-i1.gif) where yis the number of observed event (ICP, HD or EHR) in the region and μis the expected number of event (ICP, HD or EHR) in the region given by the global model. To see whether or not the association between outcomes and the potential predictors varied spatially (constant parameters or spatially varying parameters), we used a Geographically Weighted Regression(GWR) approach proposed by Fotheringham et al \[[@B24]\] on a random sample of 20% of the total cohort. A random sample was necessary because of limitations in the execution time of the GWR software. The GWR approach extends the traditional global regression framework ![](1471-2261-5-21-i2.gif) where regression parameters are constant over the whole study region, by allowing local rather than global parameters to be estimated so that the model is rewritten as: ![](1471-2261-5-21-i3.gif) where (u~i~,v~i~) denotes the coordinates in the ith point in space and the parameters now vary over the study region with geographic coordinates (spatial variability in parameters). In the case of the logistic regression, the left side of the equations above (y~i~) is replaced by logit \[Prob(Y~i~= 1\\|x~i~\], where logit (p) = log\[p/(1-p)\]. The difference between traditional regression and GWR is that, as opposed to traditional regression, GWR assumes implicitly that observed data near to location ihave more of an influence in the estimation of the local parameters than do data located farther from i. In essence, the GWR model measures the relationships inherent in the model around each location i. In the case of the log-transformation of LoS, a Monte Carlo significance test was used to infer on the spatial variability of the parameters, whereas in the logistic models, because of the unavailability of this test in the GWR software \[[@B25]\] for binary variables, we got a \"feel\" for the degree of variability in the regression coefficients by comparing the inter-quartile range of the local estimates with the standard error of the global estimate; a range between the upper and the lower quartiles (corresponding to 50% of all local estimates) much higher than two times the standard error (corresponding to 68% of the area of a normal distribution) indicates a spatial variability in the relationship \[[@B25]\]. The global regression analyses and residuals analyses were done using SAS Release 8.02 \[[@B26]\], the hierarchical cluster analyses were done using SPSS Release 11.0.1 \[[@B27]\] and the local estimates of the relationships between outcomes and predictors were estimated using GWR Release 3 \[[@B25]\]. Cartographic representations were done using ArcGIS Release 8.3 \[[@B23]\]. Ethical considerations ---------------------- This project was approved by the Sherbrooke University Hospital Ethics Board and the Commission d\'accès à l\'information du Québec(Quebec Commission to Information Access). Results ======= A total of 24,564 patients have been hospitalized for ACS in Quebec between January 1^st^, 2000 and December 31^st^, 2000. Of those, 20 patients were excluded because they were less than 25 years old or because there was an error in their administrative code of residence. Therefore, the study population totalled 24,544 patients, men accounting for 63% (n= 15,481) of it. The average age of this population was 66.7 years (± 13.0). The incidence of hospitalization for ACS varied greatly according to gender and age with the highest rates observed in men and very old people (Table [1](#T1){ref-type="table"}). A total of 1699 (6.9 %) individuals died during the index hospitalization and the HD rates varied according to gender and age (p\< 0.0001). An ICP rate in Quebec of 43.7% was much lower for women than for men (p\< 0.0001), and decreased with age, whereas the hospital LoS showed an opposite trend. Also, neither gender nor age seemed to influence the EHR rate. Finally, regional heterogeneity was observed in all outcomes considered in this study, namely, in incidence rates (range: 340 -- 827 per 100,000 inhabitants; p \< 0.0001), in ICP rates (range: 29.4 -- 51.6%; p \< 0.0001), in HD rates (range: 3.3 -- 8.2%; p \< 0.0001), in average LoS (range: 7.5 -- 14.4 days; p \< 0.0001), in median LoS (range: 5 -- 10 days), and in EHR rates (range: 4.7 -- 14.2%; p \< 0.0001). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Acute coronary syndrome (ACS) in Quebec in 2000, ACS hospitalization incidence (INC), invasive cardiac procedure (ICP) rate, hospital death (HD) rate, average (median) length of stay (LoS) (days), and early hospital readmission (EHR) rate by gender, age and region ::: Number INC\§\> ICP^§^n (%)* HD^§^n (%) LoS^§^Mean (Median) EHR\\n (%) --------------------------- ---------- ------------ ----------------- ---------------- ------------------------- ---------------- TOTAL 24,544 484 10,725 (43.7) 1699 (6.9) 11.5 (8) 1893 (8.3) Women 9063 347 3312 (36.5) 829 (9.2) 12.3 (8) 650 (7.9) Men 15,481 629 7413 (47.9) 870 (5.6) 11.0 (7) 1243 (8.5) Age \< 55 yrs 4955 146 2687 (54.2) 69 (1.4) 8.7 (6) 406 (8.3) Age 55--64 yrs 5315 721 2886 (54.3) 152 (2.9) 10.6 (7) 424 (8.2) Age 65--74 yrs 6678 1234 3329 (49.8) 389 (5.8) 12.7 (8) 504 (8.0) Age 75--84 yrs 5652 1833 1697 (30.0) 684 (12.1) 13.2 (8) 427 (8.6) Age ≥ 85 yrs 1944 2046 126 (6.5) 405 (20.8) 11.9 (8) 132 (8.6) 1 Lower St. Lawrence 801 566 260 (32.5) 57 (7.1) 11.7 (8) 61 (8.2) 2 Saguenay-Lac-St-Jean 835 438 361 (43.2) 50 (6.0) 8.7 (6) 74 (9.4) 3 Quebec City 2278 493 925 (40.6) 155 (6.8) 12.5 (8) 145 (6.8) 4 Mauricie, Central Qc 1870 558 860 (46.0) 153 (8.2) 14.4 (10) 136 (7.9) 5 Eastern Townships 1269 645 468 (36.9) 73 (5.8) 11.3 (8) 125 (10.4) 6 Montreal-Center 5282 406 2728 (51.6) 428 (8.1) 11.7 (7) 353 (7.3) 7 Outaouais 733 340 294 (40.1) 56 (7.6) 7.5 (5) 32 (4.7) 8 Abitibi-Témiscamingue 614 615 195 (31.8) 31 (5.0) 10.1 (7) 57 (9.8) 9 North Shore 449 664 154 (34.3) 15 (3.3) 11.8 (7) 60 (13.8) 10 Northern Quebec 93 447 36 (38.7) 4 (4.3) 9.8 (6) 9 (9.8) 11 Gaspé, Magdalen Is. 590 827 177 (30.0) 20 (3.4) 11.5 (8) 56 (9.8) 12 Chaudière-Appalaches 1530 583 449 (29.4) 106 (6.9) 10.2 (7) 202 (14.2) 13 Laval 1153 477 595 (51.6) 78 (6.8) 11.7 (7) 71 (7.4) 14 Lanaudière 1417 534 716 (50.5) 82 (5.8) 10.2 (7) 99 (7.4) 15 Laurentians 1611 509 667 (41.4) 88 (5.5) 10.5 (8) 126 (8.3) 16 Montérégie 4019 453 1840 (45.8) 303 (7.5) 12.0 (8) 287 (7.7) \* Per 100,000 inhabitants ^§^The difference between gender, age and regions is statistically significant (p \< 0.0001) \\ Early hospital readmission rates are calculated for ACS survivors only. The difference between gender, age is not statistically significant (p = 0.1065; 0.8357) but significant between regions (p \< 0.0001) ::: The map of the age/gender standardized ICP ratio (Figure [2a](#F2){ref-type="fig"}) highlights a decreasing gradient from Montreal metropolitan areas to peripheral regions. The cartographic representation of the age/gender standardized HD ratio (Figure [2b](#F2){ref-type="fig"}) shows that very low HD rates are observed in some remote regions. The trend in the age/gender standardized LoS (Figure [2c](#F2){ref-type="fig"}) shows patches of high average levels around South Center and very low levels at South Western. Finally, the cartographic representation of regional age/gender standardized EHR ratio (Figure [2d](#F2){ref-type="fig"}) shows an increasing gradient from peripheral regions to Montreal metropolitan areas. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Maps of the standardized ratios\* for invasive cardiac procedures (ICP), hospital death (HD), early hospital readmission (EHR), and of the standardized length of stay (LoS) after classification\\ ^§^. \* Standardized for age and gender \\Administrative regions were grouped in 4 homogeneous groups (hierarchical cluster analyses). The number beside each colour represents the overall standardized ratio (or the average) of the associated group ^§^Administrative regions: 1 : Lower St. Lawrence; 2 : Saguenay-Lac-St-Jean; 3: Quebec City; 4: Mauricie, Central Quebec; 5: Eastern Townships; 6: Montreal Center; 7: Outaouais; 8: Abitibi-Témiscamingue; 9: North Shore; 10: Northern Quebec; 11: Gaspé, Magdalen Islands; 12: Chaudière-Appalaches; 13: Laval; 14: Lanaudière; 15: Laurentians; 16: Montérégie ::: ![](1471-2261-5-21-2) ::: Multiple regression analyses (Table [2](#T2){ref-type="table"}) show that the presence of an ICP is correlated with an increased LoS, but with a decreased HD and EHR rates. Furthermore, patients living within 32 km to a specialized cardiology center had a higher likelihood of having an ICP at the index hospitalization, a shorter LoS, and a lesser likelihood of being readmitted within 30 days. The association between distance and HD is not statistically significant. Also, women received less ICP, stayed longer at index hospital, but had less early hospital readmissions. As shown in the interaction term between ICP and distance to a specialized cardiology center, we observe that the LoS is higher for patients with an ICP, but no clear trend is shown according to distance. On the other hand, it is clearly shown that for those patients with ICP, the closer they are from the nearest cardiology center, the lesser they stay at the hospital. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Results from multiple logistic regression models for Invasive cardiac procedure (ICP), hospital death (HD), and early hospital readmission (EHR), and from the log-normal regression model for the length of stay (LoS)^§^ ::: ICP β parameter (95% CI) HD β parameter (95% CI) LoS β parameter (95% CI) EHR β parameter (95% CI) ------------------------------ ---------------------------- --------------------------- ---------------------------- ---------------------------- Intercept 2.600 (2.455;2.744) \- 7.004 (-7.393; -6.616) 1.006 (0.937;1.075) \- 1.926 (-2.190;-1.662) Main effects Age \- 0.040 (-0.042;-0.037) 0.066 (0.061;0.071) 0.011 (0.010;0.012) \- 0.003 (-0.007;0.001) Women vs Men \- 0.200 (-0.257;-0.144) \- 0.103 (-0.222;0.015) 0.039 (0.013;0.064) \- 0.237 (-0.360;-0.113) ICP \- \- 0.980 (-1.152;-0.807) 0.470 (0.439;0.501) \- 0.838 (-0.964;-0.711) Distance cardiology Center 32--64 km vs \<32 km \- 0.247 (-0.323;-0.170) \- 0.132 (-0.285;0.022) 0.003 (-0.042;0.048) 0.138 (0.003;0.274) 64--105 km vs \<32 km \- 0.509 (-0.603;-0.416) 0.106 (-0.062;0.273) \- 0.083 (-0.135;-0.032) 0.356 (0.207;0.505) ≥105 km vs \<32 km \- 0.486 (-0.563;-0.409) \- 0.148 (-0.301;0.004) 0.040 (-0.003;0.083) 0.128 (-0.005;0.262) Interaction terms Women× ICP \- 0.4327 (0.181;0.685) \- 0.367 (0.151;0.583) ICP× Distance \- \- \- 32--64 km vs \<32 km 0.138 (0.069;0.207) 64--105 km vs \<32 km 0.276 (0.192;0.360) ≥105 km vs \<32 km 0.350 (0.280;0.419) ^§^CI: Confidence interval; A positive parameter estimate indicates an increased risk whereas a negative estimate indicates a reduced risk ::: A cartographic representation of the residuals associated to each global model is presented in Figure [3](#F3){ref-type="fig"}. These maps show that the heterogeneity observed between regions in the ICP, HD, and EHR rates, as well as in the log-transformation of the LoS, cannot be explained totally by age, gender, ICP (when applicable), and distance to a specialized cardiology center. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Maps of the residuals derived from the multivariate log-linear model for the length of stay (LoS) and the signed residuals\* derived from the multivariate logistic model for the invasive cardiac procedure (ICP), the hospital death (HD), and the early hospital readmission (EHR) \\^§^. \* The signed residuals are defined as: ![](1471-2261-5-21-i1.gif) \\Administrative regions were grouped in 4 homogeneous groups (hierarchical cluster analyses). The number beside each colour represents the overall standardized ratio (or the average) of the associated group ^§^Administrative regions: 1 : Lower St. Lawrence; 2 : Saguenay-Lac-St-Jean; 3: Quebec City; 4: Mauricie, Central Quebec; 5: Eastern Townships; 6: Montreal Center; 7: Outaouais; 8: Abitibi-Témiscamingue; 9: North Shore; 10: Northern Quebec; 11: Gaspé, Magdalen Islands; 12: Chaudière-Appalaches; 13: Laval; 14: Lanaudière; 15: Laurentians; 16: Montérégie ::: ![](1471-2261-5-21-3) ::: Some of the relationships β(u~i~, v~i~) between covariables and LoS are not uniform over the study surface. Indeed, when considering the log-normal model for LoS, the Monte Carlo significance test shows significant spatial variations in the GWR local estimates of the parameters associated to the covariables ICP (median estimate: 0.55, min: 0.26, lower quartile: 0.35, upper quartile: 0.73, max: 0.95, p\< 0.0001) and distance to a cardiology center (median estimate: 0.0015, min: -0.0014, lower quartile: 0.0004, upper quartile: 0.0031, max: 0.0107, p\< 0.0001), whereas it shows no significant spatial variability in the local estimates of the parameters associated to gender (p= 0.60) and age (p= 0.86), that is, in favor of constant regression parameters. Moreover, trend analyses of the parameter estimates show that an increased relationship between ICP and LoS is observed as we move away from Montrealand the city of Gatineauin Outaouais(Figure [4](#F4){ref-type="fig"}), but no clear trend is observed in the local relationships between distance to a specialized cardiology center and LoS. The local estimates of the relationship between ICP and HD varied from -0.94 to -0.58 with a median estimate of -0.84 (lower quartile: -0.84, upper quartile: -0.78), corresponding to a variation in the odds ratios from 0.39 to 0.56. Also the local estimates of the relationship between ICP and EHR varied from -0.79 to -0.36 with a median estimate of -0.39 (lower quartile: -0.48, upper quartile: -0.38). Nevertheless, no clear spatial variation was suspected in the GWR local estimates of the parameters associated to the covariables in the multiple logistic regressions for HD and EHR (Coefficients of variation: 7.9% and 12.5% respectively). ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Trend analyses of the local relationships β(u~i~, v~i~) between ICP and the log-transformation of the length of stay (LoS). ::: ![](1471-2261-5-21-4) ::: Discussion ========== Regional heterogeneity was observed in all outcomes. These spatial heterogeneities are not surprising since we are dealing with multivariate health processes evolving across social, economic and environmental explanatory related variables for which local conditions (physical environment and human activity) are very important \[[@B28]\]. Our results show an overall ICP rate of 43.7% in Quebec for the year 2000. This rate is higher than what has been reported in a previous study conducted by the CCORT group \[[@B29]\] in which the 30-day revascularization (angioplasty or aortocoronary bypass) rate in Quebec for the fiscal year 1999/2000 was 26%. This difference is probably due to the fact that, contrary to the CCORT study, we added the angiography without revascularization procedure in the definition of ICP. We did find relevant to include it in the definition of ICP as both this diagnostic procedure and revascularization are available only in a specialized cardiology center. By limiting our definition of ICP to revascularization procedures only, our observed provincial rate would be down at 34%. The remaining 8% difference between revascularization rates of 34% and 26% can be explained by the fact that the data used in this study is nearly one-year more recent than the data used by the CCORT group. One can also consider the fact that revascularization is more and more implemented in practice as a first intent procedure. We obtained an average LoS of 11.5 days (median LoS: 8 days), this is slightly higher than what other Canadian researchers have observed. Hall and Tu \[[@B6]\] have reported an average LoS for myocardial infarction in Quebec for the fiscal year 1999/2000 of 9.7 days after adjustment for age, gender and ICP. This difference is likely attributable to a difference in methodology as Hall and Tu excluded from their analysis patients with a LoS beyond the 97.5 percentile. We also observed gender variations in ICP rates, HD rates and LoS. Lower ICP rates and a longer LoS for older women have also been reported by other authors \[[@B30]-[@B33]\]. It has been argued that this difference between genders could however reflect a difference in the treatment indications rather than a difference based on gender \[[@B34]\]. After discharge from the index hospitalization, 8.3% of the population alive was readmitted within 30 days with a diagnosis of coronary heart disease. Some regions display extreme and opposite results. On one hand, the extremely low rate of EHR observed in the Outaouais region probably reflects an underestimation due to the close proximity to a specialized cardiology center located in the adjacent province. In fact, the geographical proximity of the Outaouais city of Gatineau located in the province of Quebec and the city of Ottawa located in the province of Ontario makes it difficult to interpret the data from this region. On the other hand, the high EHR rate observed in the Chaudière-Appalachesand the North Shoreregions cannot be explained with variables used in the study. Differences observed in early hospital readmission rates probably reflect more the difference in the managed cares than in health outcomes like morbidity. In fact, elevated rates in some regions may have been due to discharge after stabilization and then elective readmission at a specialized cardiology center for an ICP. An important finding of this study is the inverse relationship between ICP and HD, the inverse relationship between ICP and EHR, the increase of LoS with ICP, the positive relationship between the distance to a specialized cardiology center and EHR as well as LoS, and the negative relationship between the distance to a specialized cardiology center and ICP. The paradoxical finding of decreased LoS with proximity to specialized cardiology center while we find an increased LoS with ICP and an increased likelihood of ICP with proximity can be explained by the interaction between ICP and proximity to cardiology center in the evaluation of LoS. As seen in Table [2](#T2){ref-type="table"}, the interaction term included in the LoS model show that, for patients with ICP, the LoS is lower for patients near a cardiology center than for those farther, whereas, for patients not receiving ICP, those that are closer have not a lower LoS than the others. We can also argue that patients living far from a specialized cardiology center will stay longer at the hospital during the whole episode of care if they received an ICP because of hospital transfer from non-specialized to a specialized cardiology center. An explanation for the lack of association between distance and HD might be survival bias, that is, individuals in rural areas with ACS may be less likely to survive to hospitalization. Sicker patients in urban areas would be able to make it to the hospital. Moreover, the regional variability of the relationships between ICP and the distance to a specialized cardiology center with regards to the LoS is another interesting result. Accessibility to ICP facilities is a potential explanatory factor involved in the utilization of the ICP regularly put forward in the literature \[[@B35]\]. In our study, as shown by the maps of the residuals, variability in HD, LoS, and EHR can only be partly explained using a model taking into account ICP, accessibility to a specialized cardiology center, age and gender alone. A study by Scott \[[@B11]\] on the accessibility of specialized treatments for vascular cerebral accidents reports that the proximity of a specialized care center favours young and rich populations at the expense of older, native and underprivileged populations. According to Alter \[[@B36]\], geographic factors and accessibility to services do not explain the gradient of angiography use after a myocardial infarction in Ontario, and the author suggests that the observed regional differences in revascularization rates might actually reflect differences in regional socioeconomic factors such as age and socioeconomic status. However, these authors used a distance threshold of 50 km instead of 32 km which may reduce the strength of the link between distance and the use of specialized cardiology services. There is a need to better define, taking adjustment variables into account, what the ideal thresholds should be to appropriately describe access to ICP. To better understand geographical disparities in cares and health outcomes of ACS, there is a need to explore more comprehensively the contribution of socio-demographic variables. Rurality is one of these variables that are worth the effort to consider in subsequent studies \[[@B37]\]. Indeed, rural populations differ from urban ones not only because of the distance between them but also because they share different cultural and socioeconomic backgrounds. Among other variables that could also be important are social and material deprivation indices \[[@B38]\], in addition to medical care variables and professional attributes including academic affiliation and year of graduation for family physicians and cardiologists. Finally, one should also consider other major determinants of ACS outcomes, namely concomitant diseases like diabetes, hypertension, congestive heart failure, etc., and the use of secondary preventive drugs like angiotensin converting enzyme inhibitors, β-blockers, statins, and platelets inhibitors \[[@B39]\]. This study has some limitations. The nature of administrative data used in this study did not allow discriminating between planned readmission and readmission for a distinct ACS event. Even though the Quebec\'s hospital discharge register has been used for acute myocardial infarction \[[@B12],[@B13]\], the follow-up of the hospital\'s care episode spread out over different care institutions requires the construction of an algorithm, the accuracy of which needs to be further validated. Even if we observed that ICP increased the LoS but reduced the HD and EHR, we cannot argue that ICP increases the quality of care. In fact, recent work suggests that EHR might be an indicator of good medical care. However, it has been suggested that there is a complex network of factors influencing medical care, EHR and the association between them \[[@B40]\]. Linking the EHR measure to good medical care is still controversial. While some analyses \[[@B41]\] showed that the EHR rate was higher when care was less appropriate, other authors \[[@B42]\] argued that there was no statistically significant association and suggested that the slight association might reflect the difficulty in measuring quality of care. Moreover, how the hospital LoS relates to quality of care is controversial, this association seeming to vary according to how quality of care is defined. Some authors have observed a positive association between LoS, treatment and discharge scores \[[@B43]\] ; others have observed the opposite when defining quality of care based on physician judgments \[[@B44]\], whereas others have reported no significant relationship between LoS and quality of care as defined by readmission or mortality rates \[[@B45],[@B46]\]. In this study, since the hospital readmission could have been planned, we definitely cannot use the early hospital readmission as an indicator of good medical care. This study aimed at understanding medical care for ACS in order to model important administrative and health related outcomes. To our knowledge, this is the first to use advanced spatial statistical analysis with medico-administrative data from an entire province in Canada to build a model that could be later refined with the inclusion of other meaningful individual, socio-demographic and health care variables. Regional disparities in the province of Quebec, as highlighted by this study, may well represent an adaptation of the health care system for geographical disparities in order to deliver good quality of care, despite major limitations in terms of physical accessibility to specialized cardiology centers. To test this hypothesis, it would require a very complex research design, using qualitative and quantitative analyses. Such studies would have to take into account determinants of care at many levels such as: at the sociological, cultural, political, and economical level, as well as at the professional and geographical level \[[@B47],[@B48]\]. In terms of policy impact, it could also mean that regional based decision making may provide valuable contribution in the management of care for ACS, taking into account the limited physical accessibility of specialized cardiology centers. Conclusion ========== An important finding of this study is the inverse relationship between ICP and HD, the inverse relationship between ICP and EHR, the increase of LoS with ICP, the positive relationship between the distance to a specialized cardiology center and EHR as well as LoS, and the negative relationship between the distance to a specialized cardiology center and ICP. Moreover, the regional variability of the relationships between ICP and the distance to a specialized cardiology center with regards to the LoS is another interesting result. The EHR rates are clearly related to ICP and geographic patterns observed could reflect to some extent patients\' accessibility to revascularization specialized settings. Further studies are needed to clarify the nature of the link between geographical influence, ICP and EHR. Competing interests =================== This project has benefit from an unrestricted grant by Merck Frosst Canada Ltd. J.-P. Grégoire was employed by Merck Frosst Canada Ltd. Authors\' contributions ======================= AV, TN and JPG conceived the study, GB and AH participated to the geographical aspect of the study, JL performed the literature review on acute coronary syndrome, JC and TN performed the analyses. AV, TN, JPG and JC participated to the writing of the manuscript. AH produced the maps. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2261/5/21/prepub> Acknowledgements ================ We acknowledge the reviewers for their relevant comments and suggestions. This project was subsidized by the GEOIDE Network of Centers of Excellence and Merck Frosst Canada Ltd. The principal investigator was supported by the Department of Family Medicine, Université de Sherbrooke; the Clinical Research Center, Sherbrooke University Hospital; and the Fonds de Recherche en Santé du Québec.	PubMed Central	2024-06-05T03:55:59.730767	2005-7-11	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181243/", "journal": "BMC Cardiovasc Disord. 2005 Jul 11; 5:21", "authors": [ { "first": "Alain", "last": "Vanasse" }, { "first": "Théophile", "last": "Niyonsenga" }, { "first": "Josiane", "last": "Courteau" }, { "first": "Jean-Pierre", "last": "Grégoire" }, { "first": "Abbas", "last": "Hemiari" }, { "first": "Julie", "last": "Loslier" }, { "first": "Goze", "last": "Bénié" } ] }
PMC1181536	Brewster Kahle, creator of the Internet Archive ([www.archive.org](www.archive.org))---a digital library of Internet sites and other cultural artifacts in digital form---has been inspirational in discussing the Internet\'s potential to become a modern Library of Alexandria. He campaigns for a resource that makes all of humanity\'s knowledge available to all of humanity. The Internet certainly provides a number of resources for finding medical evidence. The Cochrane Collaboration ([www.cochrane.org](www.cochrane.org)), for example, posts freely available abstracts of systematic reviews of health interventions (access to the full text of the reviews requires a fee). PubMed ([www.ncbi.nlm.nih.gov/entrez/query.fcgi](www.ncbi.nlm.nih.gov/entrez/query.fcgi)), the United States National Library of Medicine\'s search service, provides access to abstracts of articles in MEDLINE, PreMEDLINE, and other related databases. PubMed\'s MyNCBI feature provides useful filters such as "free full-text," which shows papers for which the full text is available through the Internet, free of charge. The "HINARI" filter ([www.nlm.nih.gov/pubs/techbull/jf05/jf05\_myncbi.html\#filters](www.nlm.nih.gov/pubs/techbull/jf05/jf05_myncbi.html#filters)) shows papers for which the text is freely available to residents of a small number of developing world countries---those with a Gross National Product per capita below \$1,000---who are part of the HINARI agreement ([www.healthinternetwork.org](www.healthinternetwork.org)). PubMed Central ([www.pubmedcentral.nih.gov](www.pubmedcentral.nih.gov)) is the US National Institutes of Health\'s free digital archive of the full text of biomedical and life sciences journal articles. Yet, as many a doctor will point out, the bigger problem with medical knowledge today is not its paucity, but the difficulty of navigating what there is. Finding the right answer quickly for a patient is difficult, and perhaps nothing will replace a good medical librarian in finding that information. The rise of the search engine Google ([www.google.com](www.google.com)), along with other freely available search engines, has made it easier to find information, although the clinical uses of Google have not been as well documented as those of PubMed \[[@pmed-0020228-b1]\]. Google will not point to the answer to every question, and often the articles it finds in response to your question are not freely available. But for many clinical scenarios, Google and other search engines can provide, quickly enough, an answer that is good enough. This article aims to provide tips that will help with these clinical scenarios, saving time that can be used with a medical librarian to answer more difficult problems. Search Engine Basics {#s2} ==================== Google provides a Web search engine---a tool that constantly indexes the expanding World Wide Web and allows you to search the index. Google\'s Web site is deceptively simple, designed to give you results quickly ([Figure 1](#pmed-0020228-g001){ref-type="fig"}). Start by typing something into the text field and pressing the "Google Search" button. What you type in is the query, and what Google responds with is the results page. ::: {#pmed-0020228-g001 .fig} Figure 1 ::: {.caption} ###### Google\'s Home Page ::: ![](pmed.0020228.g001) ::: For example to learn about heart attacks, type "heart attack" as a query. Google\'s first page of results includes ten Web pages that cover heart attacks. The top right corner of [Figure 2](#pmed-0020228-g002){ref-type="fig"} shows that at the time of writing Google had found a total of about 20 million Web pages relevant to this query. Google ranks each of these Web pages by how many other Web pages provide links to them. This is the equivalent of the number of times a paper is cited; the more links a Web page gets, the greater the importance Google assigns to it, in the same way that the more citations authors receive, the greater the importance that academic institutions assign to their work. ::: {#pmed-0020228-g002 .fig} Figure 2 ::: {.caption} ###### Results of the Search Term "Heart Attack" ::: ![](pmed.0020228.g002) ::: Simply typing in the name of the medical condition is a good starting point, but it is a crude approach. For example, if your aim is to find information about thrombolysis for patients who have had a heart attack, then at least one of the 14.5 million pages that Google indexes in response to the query "heart attack" will be relevant. However, the first 20 pages Google produces say nothing about thrombolysis, and most of them are devoted to providing information for patients rather than clinicians. Rather than going through each of the millions of pages on heart attacks, it is faster to enter a slightly different query. To find Web pages that are appropriate for clinicians, the query should include words that clinicians use. "Myocardial infarction" provides around 2.1 million results from Google, and some of the sites listed on the first page are likely to be relevant to clinicians ([Figure 3](#pmed-0020228-g003){ref-type="fig"}). Being more specific with your search gives more specific results; the query "myocardial infarction thrombolysis" provides just 108,000 results, the first of which shows the guidelines on this topic \[[@pmed-0020228-b2]\] from the influential and well-respected National Institute for Clinical Excellence. ::: {#pmed-0020228-g003 .fig} Figure 3 ::: {.caption} ###### Results of the Search Term "Myocardial Infarction" ::: ![](pmed.0020228.g003) ::: Restricting the Web Sites Included in Your Search {#s3} ================================================= Google has hidden depths. For example, adding "site:" to the end of a query restricts the search to certain Web sites. To focus on guidelines from Web sites maintained by the US federal government, type "myocardial infarction site:gov." Using "site:nih.gov" focuses on the National Institutes of Health; "site:edu" restricts the search to American universities; "site:harvard.edu" to Harvard University; and "site:org" to nonprofit organizations. Using "site:fr" as a search term will restrict your search to French Web sites, although not all French Web site URLs end with "fr" (for example the French Web site of Médecins Sans Frontières is [www.paris.msf.org](www.paris.msf.org)). There are similar search terms that you can use to restrict your search to particular countries, national health systems, or government agencies. For example, "site:nhs.uk" restricts the search to the British National Health Service, while "site:gv.kr" focuses on South Korean government Web sites. Google also provides country-specific versions of its Web site. For example Google India ([www.google.co.in](www.google.co.in)) gives preferential ranking to Indian Web sites in its results and Google Kenya ([www.google.co.ke](www.google.co.ke)) provides a Kiswahili interface. The full list of country-specific Google sites is available at [www.google.com/language\_tools](www.google.com/language_tools). Other Google Features {#s4} ===================== At the top of the page (see [Figure 1](#pmed-0020228-g001){ref-type="fig"}) are some of Google\'s other tools. For example, to find images of hip prostheses, type "hip prosthesis" as your search term and click the "Google Search" button. Clicking on the "Images" link will show a series of relevant photographs and diagrams that have been reduced in size ([Figure 4](#pmed-0020228-g004){ref-type="fig"}). Clicking on any of these will display the image at full size. If the copyright owner of the image grants you permission, you can click on the image with the right-hand mouse button and choose to save it to your computer, then insert the image into your presentation or article. ::: {#pmed-0020228-g004 .fig} Figure 4 ::: {.caption} ###### Results of a Google Images Search Using the Search Term "Hip Prosthesis" ::: ![](pmed.0020228.g004) ::: The "News" link at the top of the page finds the latest news stories on a particular topic, and can be helpful for finding out what your patients have read in the lay press about a recent piece of medical research. The translation feature is useful for understanding content in languages that are not your own. On Google\'s English-language sites, the "Translate this page" link appears next to pages that are in languages other than English. Two books published by O\'Reilly---Google Hacks \[[@pmed-0020228-b3]\] and the shorter Google Pocket Guide \[[@pmed-0020228-b4]\]---provide useful additional tips and guidance. Google Scholar {#s5} ============== Perhaps the most clinically significant tool is Google Scholar ([scholar.google.com](scholar.google.com)), which is similar to PubMed in that it is a search engine that focuses on academic papers. In fact, many of the search results it returns are pages from the PubMed site. Google Scholar has a number of useful features that are not shared by PubMed. First, it is more comprehensive, indexing all academic fields, including non-biomedical ones. Second, and more importantly, the ranking mechanism is valuable. As with the rest of Google\'s technology, the pages are ranked based on the number of links that they receive. In the case of Google Scholar, "links" are citations from different papers. This means that review papers and seminal papers are most likely to top any list of results from a Google Scholar search. Google Scholar is not a replacement for PubMed, since it lacks PubMed\'s precision searching. Furthermore, finding newer papers with Google Scholar is difficult; newer papers will not have been cited as much and so will be at the bottom of the results, and sorting by publication date is not possible. Other Search Engines {#s6} ==================== Google is the most popular search engine, but it is by no means the only one. Other search engines have different approaches with their own advantages. For example, Microsoft Network\'s query builder ([search.msn.com](search.msn.com)) makes building complex queries easier. Yahoo\'s Creative Commons search feature ([search.yahoo.com/cc](search.yahoo.com/cc)) restricts searches to content (such as all of the content of the PLoS journals) that has been published under a Creative Commons license ([www.creativecommons.org](www.creativecommons.org)). These licenses are much less restrictive than the traditional "all rights reserved" copyright license. For example, if the content you have found (articles, photos, or images) is licensed under the Creative Commons Attribution License, you are legally entitled to reproduce it, distribute it, and make translations and derivative works, provided you cite the work properly. The search engine Teoma ([www.teoma.com](www.teoma.com)) clusters search results according to different meanings of the words in the query. This clustering is useful because the medical meaning of some words, such as "hip," is less commonly used than the non-medical meaning. Google lacks this clustering function. Finally, Vivisimo ([www.vivisimo.com](www.vivisimo.com)) can cluster results by subject ([Figure 5](#pmed-0020228-g005){ref-type="fig"}). Its ClusterMed ([www.clustermed.info](www.clustermed.info)) tool searches PubMed, while [www.biometacluster.com](www.biometacluster.com) simultaneously searches several relevant sources such as ChemBank and ClinicalTrials.gov. These are useful if you are searching for papers in a narrow specialty. ::: {#pmed-0020228-g005 .fig} Figure 5 ::: {.caption} ###### Vivisimo Searches the PubMed Database and Clusters the Results by Subject ::: ![](pmed.0020228.g005) ::: Conclusion {#s7} ========== All of these freely available search engines have their limitations, and they rarely give you the perfect answer to your clinical query. But they do at least help to reduce the obstacles to finding medical information online. Kahle would certainly approve. Citation: Al-Ubaydli M (2005) Using search engines to find online medical information. PLoS Med 2(9): e228. [^1]: Mohammad Al-Ubaydli is a physician and programmer. He is the author of the books Free Software for Busy People and Handheld Computers for Doctors ([www.handheldsfordoctors.com](www.handheldsfordoctors.com)), and is based in Bethesda, Maryland, United States of America. E-mail: <me@mo.md> [^2]: Competing Interests: Mohammad Al-Ubaydli wrote this article in his own time and without the support of federal funds. The views in the article are his alone and do not represent those of his employer, the National Institutes of Health.	PubMed Central	2024-06-05T03:55:59.734266	2005-8-2	{ "license": "Creative Commons Zero - Public Domain - https://creativecommons.org/publicdomain/zero/1.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181536/", "journal": "PLoS Med. 2005 Sep 2; 2(9):e228", "authors": [ { "first": "Mohammad", "last": "Al-Ubaydli" } ] }
PMC1181537	Britain\'s House of Commons Health Committee has recently recommended a fundamental realignment of the relationships between the pharmaceutical industry and government, regulators, doctors, the health service, and patients \[[@pmed-0020241-b1]\]. The committee said that the industry has interdigitated itself into every aspect of health care, and that government and others, including doctors, have taken the easy route of assuming that the interests of the industry and of the health services and patients are the same. The committee\'s report makes clear that reducing the influence of the industry would be good for everybody, including---paradoxically---the industry itself, which could concentrate on developing new drugs rather than on corrupting doctors, patient organisations, and others. "It is not in the long term interests of the industry for prescribers and the public to lose faith in it," says the report. "We need an industry which is led by the values of its scientists not those of its marketing force." Select Committees: Rationality before Realpolitik {#s2} ================================================= The Health Committee is one of many select committees of the House of Commons. The committees are comprised of members of parliament and politicians from all parties, and they can choose to examine any subject that raises matters of public importance. They receive written and oral evidence, including from government ministers, and produce reports and recommendations to which the government is required to respond. The 11-member Health Committee chose to examine the influence of the drug industry because of increasing public concern that this influence is excessive. The committee was particularly worried by the industry\'s role in promoting "medicalisation," the idea of a pill for every ill: "What has been described as the 'medicalisation' of society---the belief that every problem requires medical treatment---may also be attributed in part to the activities of the pharmaceutical industry" \[[@pmed-0020241-b1]\]. The committee, whose terms of reference are shown in [Box 1](#box1){ref-type="boxed-text"}, was also worried by the high prevalence of drug side effects. It heard from every interested party, including representatives of the drug companies, patients, doctors, medical journal editors, critics of the industry, and government ministers and officials. Box 1. Terms of Reference for the Health Committee Enquiry ---------------------------------------------------------- "The Health Committee is to undertake an inquiry into the influence of the pharmaceutical industry on health policies, health outcomes and future health priorities and needs. The inquiry will focus, in particular, on the impact of the industry on the following: drug innovationthe conduct of medical researchthe provision of drug information and promotionprofessional and patient educationregulatory review of drug safety and efficacyproduct evaluation, including assessments of value for money In doing so, the Committee will examine the influence of the pharmaceutical industry on the NHS; National Institute for Clinical Excellence (NICE); regulatory authorities and advisory and consultative bodies; prescribers, suppliers and providers of medicines; professional, academic and educational institutions; the (professional and lay) press and other media; and patients, consumers, the general public and representative bodies." (Information taken from \[[@pmed-0020241-b1]\]) The government does not have to accept the recommendations from select committees, and it recently rebuffed recommendations from the same Health Committee encouraging open access to scientific research \[[@pmed-0020241-b2]\]. Usually, the committees will be much bolder than the government, which is heavily lobbied and pays more attention to realpolitik than to rational argument. Just as the publishing industry pressured the government to ignore recommendations on open access \[[@pmed-0020241-b3]\], the pharmaceutical industry will be doing the same now---and the industry is powerful; it is Britain\'s third most profitable economic activity (after tourism and finance) and employs 83,000 people. The All-Pervasive and Persistent Influence of the Industry {#s3} ========================================================== Although the pharmaceutical industry is now perceived by the public as putting profits ahead of patients\' well-being \[[@pmed-0020241-b4]\], it is generally, as the committee makes clear, a force for good. Almost all of the drugs that have transformed medicine in the past half century have been developed and manufactured by the industry. "The discovery, development and effective use of drugs," says the committee, "have improved many people\'s quality of life, reduced the need for surgical intervention and the length of time spent in hospital and saved many lives" \[[@pmed-0020241-b1]\]. And making the industry into a scapegoat for failing to produce drugs for the diseases of the poor is in some ways no more sensible, I believe, than blaming washing machine manufacturers for poor hygiene standards in the developing world. The industry is part of the for-profit sector, and has what many philosophers might call a moral duty to maximise profits. Producing drugs for the poor requires imaginative public--private partnerships. It\'s also shallow thinking to view the industry as corrupters and doctors as the corrupted. As a doctor myself, I think that doctors are in many ways to blame for the debased relationship between themselves and the industry. The industry is (mostly) behaving in ways that are "normal" within the commercial sector. It is the doctors who depart from their ethical base when they insist on first-class fares and lavish entertainment from the industry so that they can attend an international conference. The fundamental problem, says the committee, is that the pharmaceutical industry\'s influence is too pervasive: "The industry affects every level of healthcare provision, from the drugs that are initially discovered and developed through clinical trials, to the promotion of drugs to the prescriber and the patient groups, to the prescription of medicines and the compilation of clinical guidelines." The committee goes into detail about each of these levels. Regulatory authorities, it says, are too close to the industry, meaning that they do not ensure that the industry works in the public interest. The clinical trials that are the essential evidence base for regulatory and clinical decisions are produced almost entirely by the industry, and the evidence that reaches authorities, doctors, and patients is biased. Guidelines for treating patients are distorted, not only because they must be based on biased evidence, but also because the organisations and people producing them are often in hock to the industry. The organisations may receive millions of British pounds for buildings and activities, while the individuals---particularly key opinion leaders (KOLs, as they are known in the trade)---may receive hundreds of thousands of pounds for consultancy, speaking fees, travel, research, and articles. "Drug companies are criticised for giving hospitality and recruiting 'key opinion leaders'," says the committee, "but the prescribers must be equally to blame for accepting the hospitality and some 'key opinion leaders' for lending their names to work they did not produce, often for very considerable sums." Next in the list of things that concerned the committee comes the industry\'s intensive marketing, which is becoming ever more important as the flow of drugs that offer major therapeutic advances (and so need much less marketing) dries up. Britain has some 8 000 drug company representatives, but the industry also spends millions on advertising, sponsorship, meetings, and increasingly, "medical education"---which often means a fine dinner and a lecture from a captive KOL. The report states: "Coupled with company-sponsored information from medical journals and supplements, 'medical education' materials, advertisements and sponsorship to attend conferences, workshops and other events, it is little wonder that prescribing practices are affected." Medical journals, as I\'ve argued in PLoS Medicine \[[@pmed-0020241-b3]\], are in some ways extensions of the marketing arm of the industry, while the free newspapers that overwhelm doctors in the developed world depend 100% on largesse from the industry. Individual journalists are also captured, the committee heard---and perhaps most troublesome is the way patient organisations have become so dependent on the industry. The committee concluded that "Measures to limit the influence of industry on patient groups are needed." Currently, in Britain, we see that the "patients" who are trying to convince the British government that it should ignore the advice of the London-based National Institute of Health and Clinical Excellence (NICE) (which says that drugs for Alzheimer\'s disease are not sufficiently cost effective) are in many ways agents of the companies that produce those drugs \[[@pmed-0020241-b6]\]. The consequences of all of these incestuous relationships, says the committee, are bad decisions on the regulation and prescription of drugs, over-reliance on drugs rather than on other interventions (such as dietary change, exercise, or counselling), and the "medicalisation" of life\'s problems, including baldness, shyness, unhappiness, grief, and sexual difficulties. Recommendations: "Let the Sun Shine In" {#s4} ======================================= The committee came up with 48 conclusions and recommendations, and I have listed some of the highlights in [Box 2](#box2){ref-type="boxed-text"}. The committee\'s main recommendation for the problems it identifies is transparency: "let the sun shine in." It begins by recommending that there be a clinical trials register, "maintained by an independent body" and containing full information. Companies should be required to put the information on the register "at launch as a condition of the marketing licence." The committee also wants regulatory authorities and ethics committees (the British equivalent of institutional review boards) to help with the design of trials to make sure that they are answering real questions. It didn\'t, however, recommend more public funding of trials. I believe that such funding is necessary in order to ensure that trials are addressing the most important questions---including head-to-head comparisons and trials of new drugs against older drugs and non-drug treatments. Advice to companies is unlikely to be effective. Box 2. Recommendations from the Health Committee Enquiry: Some Highlights ------------------------------------------------------------------------- The process of licensing drugs, and the medicines\' regulatory system, should both be more transparent.There should be an independent register of clinical trials.Clinical trials should focus on using health outcomes that are relevant to patients.More research should be undertaken into the adverse effects of drugs and the costs of drug-induced illness.The regulator should ensure greater restraint in medicines\' promotion.Tougher restriction should be placed on the prescribing activities of non-specialists.Doctors should be required to declare significant sums or gifts they receive as hospitality.The sponsorship of the drug industry should pass from the Department of Health to the Department of Trade and Industry---because the secretary of state for health cannot serve two masters (the public and the industry).(Information taken from \[[@pmed-0020241-b1]\]) There should be, says the committee, limits on the quantity of marketing materials, particularly in the first six months after launch, and stricter controls on marketing to junior doctors, nurses, and pharmacists. These proposals don\'t seem sufficiently thought through: it\'s hard to imagine how the proposals would be enforced, and they are patronising to junior doctors, nurses, and pharmacists---many of whom are much better, I suspect, at assessing evidence than burnt out, ageing, high-prescribing general practitioners. The Health Committee would also like to see an independent review of the Medicines and Healthcare Products Regulatory Agency (London, United Kingdom) plus a public inquiry every time a drug is withdrawn from the market on health grounds. It\'s hard to see the government implementing these recommendations, as inquiries are expensive and always create difficulties for government, but if bodies like the Medicines and Healthcare Products Regulatory Agency and NICE are to maintain public confidence they will have to distance themselves from the industry---and be seen to do so. Important first steps will be to make public more of the information they use to make their decisions and to exclude KOLs from their committees (which may be difficult, as KOLs include many prominent doctors and professors of pharmacology and therapeutics). Doctors\' organisations, says the committee, should produce publicly available registers of doctors\' links with industry. These registers---and this is my recommendation and not the committee\'s---should also include information on monetary amounts. Otherwise, it will not be possible to separate the KOLs from the vast numbers of doctors who receive pens, lunches, trips, and other gifts from the pharmaceutical industry. I doubt very much that doctors\' organisations will adopt these recommendations until forced to do so. In Britain, it\'s more embarrassing to ask people about money than sex. Plus, doctors might come to be seen as the villains rather than the good guys. The committee also wants patients\' organisations to declare their connections with industry and to make clear when ubiquitous "disease awareness" campaigns are funded by the industry, which is probably very common \[[@pmed-0020241-b6]\]. I agree with this support for transparency, and while recognising the penury of many patients\' organisations, I think they would do well to resist the lure of the industry\'s lucre as much as they can. Conclusion {#s5} ========== In the end, this report will probably be less remembered for its recommendations---most of which will probably be ignored---than for having brought the important debate over the excessive influence of the pharmaceutical industry to a broader public. We all stand to benefit from the reduction of that influence. Citation: Smith R (2005) Curbing the influence of the drug industry: A British view. PLoS Med 2(9): e241. KOL : key opinion leader NICE : National Institute of Health and Clinical Excellence [^1]: Richard Smith is Chief Executive of UnitedHealth Europe, London, United Kingdom. E-mail: <richardswsmith@yahoo.co.uk> [^2]: Competing Interests: RS was an editor for the BMJ for 25 years. For the last 13 of those years, he was the editor of the BMJ and chief executive of the BMJ Publishing Group, responsible for the profits of not only the BMJ but of the whole group, which published some 25 other journals. He stepped down in July 2004. He is now a member of the board of the Public Library of Science, a position for which he is not paid. The UnitedHealth Group, of which UnitedHealth Europe is a part, includes a company that performs clinical trials for the pharmaceutical industry. RS has no responsibility for or influence over this company.	PubMed Central	2024-06-05T03:55:59.735665	2005-8-2	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181537/", "journal": "PLoS Med. 2005 Sep 2; 2(9):e241", "authors": [ { "first": "Smith", "last": "Richard" } ] }
PMC1181538	Increasingly, the question of whether some forms of research, such as human cloning, infringe notions of human dignity has been proposed as a primary justification for the development of science policy. This is understandable. A commitment to human dignity is a widely shared value and the foundation for our understanding of human rights. The preamble to the Universal Declaration of Human Rights, adopted by the United Nations General Assembly in 1948, states that "recognition of the inherent dignity and of the equal and inalienable rights of all members of the human family is the foundation of freedom, justice, and peace in the world" \[[@pmed-0020244-b1]\]. However, in the context of emerging scientific advances, such as cloning technology and stem cell research, exactly how to judge whether human dignity is infringed upon or degraded is rarely explained. As an example, a 2002 report by the President\'s Council on Bioethics is titled "Human Cloning and Human Dignity: An Ethical Inquiry," but it fails to conceptualize human dignity or address the specific ways in which human cloning may impinge on human dignity \[[@pmed-0020244-b2]\]. This lack of clarity has the potential to hurt policymaking and, in the long run, degrade the possible substantive value of the principle of human dignity. The dilemma is that human dignity is a poorly conceptualized and vague concept. Further complicating matters, in a pluralistic society various groups and communities bring a diversity of worldviews, religious values, and cultural understandings that inform and shape their use of the concept of human dignity. The Concept of Dignity {#s2} ====================== There are numerous examples of policies that cite human dignity as a standard for dealing with controversial science issues. The UN Educational, Scientific, and Cultural Organization\'s "Universal Declaration on the Human Genome and Human Rights" recommends a ban on "practices which are contrary to human dignity, such as reproductive cloning" \[[@pmed-0020244-b3]\]. A 2003 World Health Organization report suggests that genetic databases create the need to balance "human dignity and human rights as against public health, scientific progress and commercial interests in a free market" \[[@pmed-0020244-b4]\]. Both Japan\'s 2001 stem cell research guidelines and Canada\'s recent legislation covering research involving human reproductive material claim the protection of human dignity as a primary objective of the regulatory regimes \[[@pmed-0020244-b5],[@pmed-0020244-b6]\]. And, of course, the concept of human dignity permeates research ethics policy (e.g., the Helsinki Declaration). Canada\'s primary research ethics document, the "Tri-Council Policy Statement," declares that the "cardinal principle of modern research ethics, as discussed above, is respect for human dignity" \[[@pmed-0020244-b7]\]. In these documents, the concept of human dignity is often used in the conventional legal and ethical manner to emphasize the right of individuals to make autonomous choices. This is most apparent in the context of research ethics documents and informed-consent policies. This conception treats human dignity as a means of empowerment. Some scholars have gone so far as to suggest that this is the only appropriate normative use of the idea of dignity \[[@pmed-0020244-b8]\]. Despite such claims, an alternative conception, dignity as a means of constraint, is increasingly common in the realm of science policy. Citations of human dignity in science policy discussions usually come in the context of concerns that some activities, such as human cloning and the commodification of human tissue, infringe some basic understanding of dignity \[[@pmed-0020244-b9]\]. Costa Rica\'s recent proposal to the UN for an international treaty banning cloning stands as a good example of this trend. The Costa Rican draft convention sought to "ensure respect for the dignity and basic rights of the human being" in the face of the "threat posed by experiments in the cloning of human beings" \[[@pmed-0020244-b10]\]. Likewise, in the area of stem cell research, opponents refer to the dignity implications as a rationale for limiting research on human embryos. In Europe, it is an underpinning of the "ordre public" (public policy) restriction of patent law, which has been used to deny patents on cloning technologies and human embryonic stem cells \[[@pmed-0020244-b11]\]. [](#pmed-0020244-g001){ref-type="fig"} ::: {#pmed-0020244-g001 .fig} ::: {.caption} ###### Some policymakers suggest that human cloning would infringe on dignity, but they rarely explain how to judge such infringement (Illustration: Margaret Shear, Public Library of Science) ::: ![](pmed.0020244.g001) ::: When used in this manner, dignity is meant to reflect a broad social or moral position that a particular type of activity is contrary to public morality or the collective good. It is not used as a source of individual rights, but as a justification for a policy response---usually a policy that is intended to curtail a given activity. While most would agree that human dignity is closely tied to the idea of the inherent worth of humans, policy documents and legal instruments rarely provide an explicit definition of dignity or how human worth might be degraded by a given technology or scientific activity. "\[Dignity\'s\] intrinsic meaning has been left to intuitive understanding, conditional, in large measure, by cultural factors" \[[@pmed-0020244-b12]\]. As such, its meaning will be very different depending on the values and background that an individual or community brings to the deliberations. Problems with "Dignity" {#s3} ======================= Why is this lack of clarity problematic? At worst, the concept of dignity appears to often be used as mere rhetorical dressing, adding little more to the policy debate than the weight or cachet of the concept. In such circumstances, dignity conveys a sense of general social unease, but with little explanation of how, exactly, human dignity is threatened. At times, reference to human dignity seems to have emerged, inappropriately, as a politically palatable articulation of the "yuck factor"---a way of "registering our concern" about activities that seem to threaten "those parts of the human condition that are familiar and reassuringly 'human,'" without the detailed explanation of why and how the activities are troubling \[[@pmed-0020244-b13]\]. The use of dignity as a policy justification can silence open debate, and may serve to blur an understanding of the real policy concerns behind a given technological innovation or scientific development. Moreover, without a clearer conception of human dignity and its requirements, it is not possible to evaluate technological innovation or scientific developments in the service of protecting human dignity. In addition, because human dignity is viewed as a foundational concept, its use may imply a degree of social consensus that simply does not exist. If something is said to infringe human dignity, one would expect a degree of agreement that this is so. However, modern societies are often pluralistic, and in pluralistic societies, consensus is often difficult to obtain, whether about human dignity or other complex social and ethical issues introduced by scientific innovations \[[@pmed-0020244-b14]\]. There is not even agreement about the foundation of human dignity---whether it is faith-based or secular---let alone what human dignity entails. In the debate on stem cell research, for example, not all agree on the moral status of the embryo and, as such, they cannot reach agreement on the degree to which embryonic stem cell research challenges human dignity. As a result, the use of dignity will not necessarily represent a broadly accepted social value, but, instead, it may express a particular worldview---a worldview that may not even reflect the majority opinion. In the context of legal instruments and policy documents, the use and content of human dignity may amount to a "political compromise informed by cultural, political, constitutional and other conditions, all of which can then evolve and change" \[[@pmed-0020244-b15]\]. In more extreme circumstances, it could involve "intolerant voices (whether of the majority or of an influential minority) expressing negative attitudes about certain practices, which attitudes are then translated into restrictions ostensibly in the interest of respect for human dignity" \[[@pmed-0020244-b9]\]. The Value of Human Dignity to Science Policy {#s4} ============================================ For the reasons cited above, we need to be cautious not to simply accept the concept of human dignity as a primary or sole justification for science policy. As noted by Hayry, "ethical controversies cannot be settled simply by stating that this or that solution respects or violates human dignity" \[[@pmed-0020244-b16]\]. Nevertheless, despite the concerns that have been articulated about the recent constraining and vague use of dignity in science policy, we believe that it remains a potentially useful normative concept. Indeed, the fact that the concept of human dignity "anchors different worldviews" may, paradoxically, be its greatest policy value. Most, if not all, agree that human dignity is tied to notions of human worth \[[@pmed-0020244-b15]\]. As such, a pronouncement that something infringes human dignity should be viewed as an opportunity to debate the values at play and the cultural underpinnings of the concern. Hayry concludes thus: "People can debate the merits of different notions of dignity, and they can argue the relative significance of opposing concepts in ethical disputes. Although it is probably true that dignity should never be used as a mere means, the concept of dignity could in this way be used as a means to further understanding between people and cultures" \[[@pmed-0020244-b16]\]. The key to its constructive use, however, is that dignity be used as a facilitator of policy debate, instead of a "door closer." This means the concept of human dignity should not be used as a slogan or as part of a mere assertion of harm. It also means that when human dignity is used as a grounding for science policy---be it as a justification for empowerment or technological constraint---as much specificity as possible should be provided. For example, how does cloning technology challenge the intrinsic worth of humans? Is it the mere replication of nucleic DNA or something else? Do the dignity concerns about stem cell research go beyond disputes about the moral status of the embryo? We believe that the close connection to the notion of human worth differentiates it from mere "intuitive" responses to a technology and, as such, should force an articulation of the way in which various communities view the foundations of human nature and worth. In a pluralistic society, this may require exploring the various approaches to dignity in different cultures and communities. Finally, communities will need to consider how our community-specific conceptions of dignity relate to broader regulatory policy. It may be difficult to develop a law or policy that can respect the diverse conceptions of human dignity, as highlighted by the divisive nature of the stem cell debate. But by using the concept of human dignity as an avenue for exploring differing philosophical approaches, we promote transparency, encourage dialogue, and help to avoid the simplistic application of dignity in science policy. Conclusion {#s5} ========== The issues being raised about the impact of scientific discoveries and new technologies on human dignity make it imperative to gain greater understanding about the meanings and requirements of this important but elusive concept. Otherwise, references to dignity will likely be ineffectual and potentially even a source of social division. Moreover, the one goal that all groups may agree upon, the importance of protecting human dignity, will never be achieved. We would like to thank all of the participants of the November 2003, March 2004, and September 2004 meetings of the Dignity and Science Policy Working Group. We would also like to thank the Stem Cell Network and the American Association for the Advancement of Science for the research support. Citation: Caulfield T, Chapman A (2005) Human dignity as a criterion for science policy. PLoS Med 2(8): e244. [^1]: Timothy Caulfield is Canada Research Chair in Health Law and Policy, Professor in the Faculty of Law and Faculty of Medicine and Dentistry, and Research Director of the Health Law Institute at the University of Alberta, Edmonton, Alberta, Canada. Audrey Chapman serves as Director of the Science and Human Rights Program, Co-Director of Science and Intellectual Property in the Public Interest, and Senior Associate for Ethics in the Program for Dialogue on Science, Ethics, and Religion, American Association for the Advancement of Science, Washington, D.C., United States of America. [^2]: Competing Interests: TC is on the editorial board of PLoS Medicine. AC declares that she has no competing interests.	PubMed Central	2024-06-05T03:55:59.737393	2005-8-2	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181538/", "journal": "PLoS Med. 2005 Aug 2; 2(8):e244", "authors": [ { "first": "Timothy", "last": "Caulfield" }, { "first": "Audrey", "last": "Chapman" } ] }
PMC1181539	Introduction {#s1} ============ The phenylethylamine derivative dopamine (DA) is critically involved in a wide variety of vital functions such as locomotion, feeding, emotion, and reward \[[@pbio-0030271-b01]--[@pbio-0030271-b03]\]. Major DA systems in the brain originate from brainstem DA neurons located in the substantia nigra pars compacta (SNc) and the ventral tegmental area (VTA). SNc neurons project mainly to the caudate/putamen or dorsal striatum (nigrostriatal system), whereas VTA neurons send their axons to the ventral striatum including the nucleus accumbens, as well as certain other limbic (mesolimbic system) and cortical areas (mesocortical system). Small DA-containing cell groups located primarily in the hypothalamus comprise the tuberoinfundibular DA system \[[@pbio-0030271-b04]--[@pbio-0030271-b06]\]. DA is synthesized from tyrosine by the rate-limiting enzyme tyrosine hydroxylase (TH), to produce L-DOPA which is quickly decarboxylated by L-aromatic acid decarboxylase (L-AADC) to DA \[[@pbio-0030271-b01],[@pbio-0030271-b03]\]. Intraneuronal DA is accumulated into synaptic vesicles by the vesicular monoamine transporter-2 (VMAT2) \[[@pbio-0030271-b07],[@pbio-0030271-b08]\]. DA released into the extracellular space exerts its physiological functions via activation of G protein-coupled D1-like and D2-like DA receptors \[[@pbio-0030271-b09]\]. Finally, DA in the extracellular space is subject to dilution by diffusion and metabolic degradation; however the major route of DA clearance from the extracellular space in the striatum/nucleus accumbens is the rapid recycling of the neurotransmitter back into dopaminergic terminals by the Na^+^/Cl^−^--dependent plasma membrane dopamine transporter (DAT) \[[@pbio-0030271-b10],[@pbio-0030271-b11]\]. Recycled DA in the dopaminergic terminals is then stored in the large intracellullar storage pool available for subsequent re-release \[[@pbio-0030271-b12],[@pbio-0030271-b13]\] . It is well established that DA neurotransmission in both dorsal and ventral striatum is essential for normal locomotor functions, and progressive degeneration of DA neurons in these areas is a known cause of Parkinson disease (PD). In most cases, PD becomes clinically apparent when the loss of dopaminergic neurons reaches 60%--70%, which leads to functional dysregulation of the related neuronal circuitry \[[@pbio-0030271-b14]--[@pbio-0030271-b17]\]. Major motor manifestations of DA deficiency in PD include, but are not limited to, resting tremor (tremor occurring in the absence of voluntary movement), rigidity (tonically increased muscle tone), bradykinesia/akinesia (slowness/difficulty in initiating movement), gait disturbance and postural instability, facial masking, and decreased eye-blinking \[[@pbio-0030271-b18]\]. Presently, there is no known cure for PD \[[@pbio-0030271-b19],[@pbio-0030271-b20]\], however its symptoms can be controlled by therapeutic interventions \[[@pbio-0030271-b21]\]. DA replacement therapy by administration of the DA precursor, L-DOPA, has been used for many years and remains the gold standard for treatment of PD \[[@pbio-0030271-b22],[@pbio-0030271-b23]\]. However, the efficacy of this treatment wanes with time, and fluctuations in motor performance as well as psychotic reactions and dyskinesias often develop. DA agonists, as well as several other classes of drugs directly or indirectly affecting DA function (monoamine oxidase \[MAO\] inhibitors, COMT \[catechol-o-methyl transferase\] inhibitors, and amantadine), have some beneficial effects in PD patients, but they are mostly used either at early stages of PD or are applied as adjunct medications to enhance the benefits of L-DOPA \[[@pbio-0030271-b21],[@pbio-0030271-b24],[@pbio-0030271-b25]\]. Due to these limitations of existing therapeutic approaches, the development of better anti-Parkinsonian drugs remains a major objective of PD research. Several lines of evidence suggest that development of novel non-dopaminergic approaches aimed at bypassing impaired dopaminergic transmission would be beneficial in PD, particularly at later stages \[[@pbio-0030271-b16],[@pbio-0030271-b26]--[@pbio-0030271-b28]\], however it is still unclear if these treatments would just potentiate action of residual DA or act completely independently of DA. A number of animal models of DA deficiency, based on pharmacologic, neurotoxic, or genetic approaches, have been developed to understand basic pathological processes leading to PD and/or to search for novel principles of therapy \[[@pbio-0030271-b29]--[@pbio-0030271-b36]\]. However, in rodents, the prolonged absence of DA is not compatible with life \[[@pbio-0030271-b03],[@pbio-0030271-b07],[@pbio-0030271-b08]\], and animals with chronic severe DA depletion are generally not available for routine experimentation. We have developed mice lacking a functional DAT (DAT-KO mice) \[[@pbio-0030271-b11]\] that display remarkable alterations in the compartmentalization of DA \[[@pbio-0030271-b12],[@pbio-0030271-b13],[@pbio-0030271-b37]\]. Lack of the DAT-mediated inward transport in these mice results in an elevated extracellular DA and at least 95% decreased intracellular DA stores. Unlike normal animals, these mice demonstrate remarkable dependence of the remaining DA on ongoing synthesis, and pharmacologic blockade of DA synthesis in DAT-KO mice provides an effective approach to eliminate DA acutely \[[@pbio-0030271-b12],[@pbio-0030271-b13]\]. Substituted phenylethylamine derivatives, amphetamines, that are structurally similar to DA and the endogenous trace amine β-phenylethylamine, represent a well-known group of compounds that potently affect psychomotor functions. Amphetamines are known to interact with plasma membrane monoamine transporters, including DAT, norepinephrine (NE) transporter (NET), and serotonin transporter. This complex interaction results in transporter-dependent efflux of monoamines into extracellular space from intraneuronal stores \[[@pbio-0030271-b10],[@pbio-0030271-b38],[@pbio-0030271-b39]\]. It is commonly believed that DAT-mediated efflux of DA is primarily responsible for the psychostimulant and hyperlocomotor actions of these drugs \[[@pbio-0030271-b38],[@pbio-0030271-b40],[@pbio-0030271-b41]\]. Intriguingly, recent studies have identified novel transporter-independent targets of amphetamines. It has been shown that amphetamines, as well as β-phenylethylamine, some monoamine metabolites, and several drugs affecting monoaminergic transmission, can directly activate specific G protein--coupled trace amine (trace amine 1 \[TA1\]) receptors \[[@pbio-0030271-b42]\] with currently unknown functional consequences \[[@pbio-0030271-b43],[@pbio-0030271-b44]\]. We report here that the pharmacologic inhibition of the rate-limiting enzyme of DA synthesis, TH, almost immediately depletes brain DA to undetectable levels in DAT-KO mice and induces a transient recapitulation of essentially all PD symptoms for up to 16 h. DA-deficient DAT-KO mice (DDD mice) thus represent an acute PD model that is useful for studying the efficacy of compounds that potentially can restore control of locomotion in the absence of any contribution of the dopaminergic system. By using this approach, we found that several amphetamine derivatives can counteract the behavioral manifestations of severe DA deficiency, suggesting that, in addition to well-known DA-mediated effects, amphetamine-like compounds can also affect motor functions in a DA- and DAT-independent manner. Results {#s2} ======= A Pharmacologic Approach for Provoking Selective DA Deficiency in DAT-KO Mice {#s2a} ----------------------------------------------------------------------------- The ability of α-methyl-p-tyrosine (αMT), a potent irreversible inhibitor of TH \[[@pbio-0030271-b29],[@pbio-0030271-b45],[@pbio-0030271-b46]\], to impede production of brain DA suggests a simple, but straightforward, strategy for producing an acute PD mouse model. However, numerous studies have documented that treatment of normal animals with αMT results only in a relatively slow and partial depletion of DA in brain tissues that is not sufficient for generation of PD-like symptoms \[[@pbio-0030271-b29],[@pbio-0030271-b45],[@pbio-0030271-b46]\]. This limited depletion is based upon how DA is stored. It is believed that the large intraneuronal DA storage pool that normally exists in striatal DA terminals provides sufficient DA to release and recycle back into releasing terminals up to the time when newly synthesized TH starts to regain its functional role \[[@pbio-0030271-b29],[@pbio-0030271-b45],[@pbio-0030271-b46]\]. Thus, in a normal animal, complete depletion of striatal DA is unachievable by TH inhibition alone, and additional depletion of vesicular DA by VMAT2 inhibitors, such as reserpine is required \[[@pbio-0030271-b33],[@pbio-0030271-b47]--[@pbio-0030271-b49]\]. Protocols designed for wild-type \[WT\] mice that use a dual inhibitor strategy (VMAT2 plus TH inhibitors) deplete DA to 1%--2% of control levels \[[@pbio-0030271-b33],[@pbio-0030271-b47]--[@pbio-0030271-b50]\], but the levels of other monoamine neurotransmitters that are substrates for VMAT2 are also severely affected. This nonselective targeting of monoaminergic signaling generally results in very complicated phenotypes that are not necessarily reflective of classic PD. In the absence of any pharmacologic treatment, the intraneuronal vesicular stores of DA in the striatum of DAT-KO mice are already profoundly depleted by at least 20-fold \[[@pbio-0030271-b12]\]. This selective depletion of DA in dopaminergic terminals of DAT-KO mice, as well as analogous depletion observed in mice lacking NET \[[@pbio-0030271-b51]\] or serotonin transporter \[[@pbio-0030271-b52]\] with NE and serotonin (5-HT), respectively, reflects the critical role of transporter-mediated recycling in the maintenance of intracellular storage pools \[[@pbio-0030271-b13]\]. With loss of the major intracellular storage pool of DA in DAT-KO mice, both the intracellular and extracellular levels of DA in the striatum become critically dependent upon ongoing DA synthesis. Therefore in DAT-KO mice, acute TH inhibition alone by αMT is sufficient to induce profound depletion of DA \[[@pbio-0030271-b12],[@pbio-0030271-b13],[@pbio-0030271-b37]\]. To explore this phenomenon in detail, we first measured the time-course of striatal DA depletion in DAT-KO and control mice following treatment with αMT ([Figure 1](#pbio-0030271-g001){ref-type="fig"}). In agreement with previous studies \[[@pbio-0030271-b13]\], we observed that in saline-treated DAT-KO mice, striatal tissue levels of DA were about 20-fold lower than in WT controls ([Figure 1](#pbio-0030271-g001){ref-type="fig"}A). The systemic administration of αMT (250 mg/kg IP) to DAT-KO mice produced rapid (15 min) and virtually complete (down to 5% of control levels in DAT-KO mice that is equivalent to less than 0.2% of WT control levels) depletion of striatal DA. In contrast, in WT mice the same treatment resulted in a relatively slow (4 h) depletion of only 60% of striatal tissue DA ([Figure 1](#pbio-0030271-g001){ref-type="fig"}B). The duration of the depletion in DAT-KO mice was extensive, lasting up to 16 h, until a recovery of DA, related to the de novo synthesis of TH, occurs \[[@pbio-0030271-b29],[@pbio-0030271-b45]\]. Notably, the rate of recovery of striatal DA levels was approximately the same in WT and DAT-KO mice. ::: {#pbio-0030271-g001 .fig} Figure 1 ::: {.caption} ###### αMT Induces Severe DA Depletion in the Striatum of DAT-KO Mice \(A) Tissue levels of DA in the striatum of saline-treated control WT and DAT-KO mice (n = 7 per group). Striatal levels of DA were significantly lower in DAT-KO versus WT mice (p \< 0.05, Student\'s t-test). \(B) Dynamics of the effect of αMT (250 mg/kg IP) on striatal tissue DA in WT and DAT-KO mice (n = 5--8 per group). DA levels were significantly lower versus control values at all the time points after αMT treatment in DAT-KO mice and 2--24 hours after treatment in WT mice (p \< 0.05, one-way ANOVA followed by Dunnet\'s multiple comparison test). The magnitude of the effect was significantly different between genotypes from 1 to 16 h after αMT injection (p \< 0.05, two-tailed Mann-Whitney U test). \(C) Tissue levels of NE in the frontal cortex of saline-treated WT and DAT-KO mice (n = 7 per group). \(D) Dynamics of the effect of αMT (250 mg/kg IP) on tissue levels of NE in the frontal cortex of WT and DAT-KO mice (n = 5--8 per group). NE levels were significantly lower versus control values at time points 2--16 after αMT treatment in DAT-KO mice and at 4--16 hours after treatment in WT mice (p \< 0.05, one-way ANOVA followed by Dunnet\'s multiple comparison test). The magnitude of the effect was not different between genotypes at any time point after αMT injection (p \> 0.05, two-tailed Mann-Whitney U test). \(E) Effect of αMT on extracellular DA levels in the striatum of WT mice, measured using in vivo microdialysis. Data are presented as a percentage of the average level of DA measured in at least three samples collected before the drug administration. (Saline, n = 5; αMT, n = 7). αMT significantly decreased DA levels 60--180 min after treatment (p \< 0.05, two-tailed Mann-Whitney U test versus respective time points in saline-treated controls). \(F) Effect of αMT on extracellular levels of DA in the striatum of DAT-KO mice, measured by using in vivo microdialysis in freely moving mice. Data are presented as a percentage of the average level of DA measured in at least three samples collected before drug administration. (Saline, n = 4; αMT, n = 6). αMT significantly decreased DA levels 20--180 min after treatment (p \< 0.05, two-tailed Mann-Whitney U test versus respective time points in saline-treated controls). Analysis of area under curve values for 120-min periods after drug administration revealed significant difference between DAT-KO and WT groups (p \< 0.05, two-tailed Mann-Whitney U test). Note also that the basal extracellular levels of DA in DAT-KO mice were significantly higher than in WT mice (predrug concentrations of DA in dialysates were: WT, 76 ± 17 fmol/20 μl; DAT-KO, 340 ± 63 fmol/20 μl). ::: ![](pbio.0030271.g001) ::: Because DA itself serves as a precursor for neuronal production of NE in NE neurons, the inhibition of TH should also impact NE production. To test the impact of TH inhibition on the NE system, the frontal cortex tissue NE concentrations were measured in WT and DAT-KO mice. As opposed to the DAT, NET expression is not altered in DAT-KO mice so that the storage pool, which is by far the predominant reservoir of NE in NE-enriched regions such as the frontal cortex, should not be significantly altered in these mutants. Accordingly, the levels of NE in the frontal cortex tissue of saline-treated DAT-KO mice did not vary from that of WT mice ([Figure 1](#pbio-0030271-g001){ref-type="fig"}C). Furthermore, αMT (250 mg/kg IP) treatment induced similar NE depletion in WT and DAT-KO mice by about 60% in 8 h after treatment. Importantly, the rates of partial NE depletion and recovery were almost identical between WT and DAT-KO mice ([Figure 1](#pbio-0030271-g001){ref-type="fig"}D). Thus, TH inhibition in DAT-KO mice induces rapid severe depletion of DA, but only partially and slowly affects NE, indicating selectivity of this marked depletion to neurons expressing the DAT. In order to demonstrate that targeting of TH by αMT depletes the functional extracellular pool of DA in living animals, we measured extracellular levels of striatal DA in freely moving mice by in vivo microdialysis. In agreement with total tissue DA data, αMT treatment essentially eliminated extracellular DA levels in DAT-KO mice ([Figure 1](#pbio-0030271-g001){ref-type="fig"}F), whereas only a partial decrease was observed in WT mice. ([Figure 1](#pbio-0030271-g001){ref-type="fig"}E). Thus, both intracellular and extracellular DA levels in the striatum of DAT-KO mice are critically dependent upon ongoing synthesis. DA Depletion in DAT-KO Mice Results in a Loss of Motor Control {#s2b} -------------------------------------------------------------- It is well known that DA plays a pivotal role in the control of various aspects of locomotor behaviors. Severe depletion of DA in αMT-treated DAT-KO mice results in a very specific akinetic phenotype ([Figure 2](#pbio-0030271-g002){ref-type="fig"}; see [Video S1](#sv001){ref-type="supplementary-material"}). The DA-depleted DAT-KO mice (DDD mice) become akinetic almost immediately after treatment, in contrast to the essentially normal motor function displayed by αMT-treated WT mice. Moreover, DDD mice develop extreme rigidity, body tremor, and ptosis (droopy eyelids). These behaviors are evident on several tests ([Figure 3](#pbio-0030271-g003){ref-type="fig"}). Akinesia was assessed by evaluating horizontal locomotor activity ([Figure 3](#pbio-0030271-g003){ref-type="fig"}A and [3](#pbio-0030271-g003){ref-type="fig"}B) and by an "akinesia" test ([Figure 3](#pbio-0030271-g003){ref-type="fig"}C); rigidity assessed by a catalepsy test ([Figure 3](#pbio-0030271-g003){ref-type="fig"}D), a "grasping" test ([Figure 3](#pbio-0030271-g003){ref-type="fig"}E), and a "bracing" test ([Figure 3](#pbio-0030271-g003){ref-type="fig"}F); whereas tremor ([Figure 3](#pbio-0030271-g003){ref-type="fig"}G) and ptosis ([Figure 3](#pbio-0030271-g003){ref-type="fig"}H, see also [Figure 2](#pbio-0030271-g002){ref-type="fig"}B) were visually determined \[[@pbio-0030271-b03],[@pbio-0030271-b53]--[@pbio-0030271-b58]\]. These behaviors were analyzed in WT and DAT-KO mice for 4 h after a αMT treatment when depletion of DA is most severe in DAT-KO mice but with relatively minor effect on NE levels (see [Figure 1](#pbio-0030271-g001){ref-type="fig"}). In all these measures DDD mice differed significantly from their WT littermates or saline-treated controls. Importantly, these abnormal behaviors in DDD mice, with the exception of ptosis, became maximal during the 30- to 60-min period following αMT exposure, thus correlating with the rate of DA depletion. Ptosis developed substantially later ([Figure 3](#pbio-0030271-g003){ref-type="fig"}H), suggesting an additional contribution of NE depletion to the full magnitude of this response \[[@pbio-0030271-b59]\]. Importantly, the righting reflex of DDD mice was normal at all time periods analyzed ([Figure 3](#pbio-0030271-g003){ref-type="fig"}I), indicating that this akinesia is not related to global sedation but rather to deficient movement control. It should be noted also that this global phenotype, which might be viewed as "freezing," can be on some occasions temporarily disrupted by an acoustic startle or other stressful stimulus. However, after manifesting a few movements, the animals return to an akinetic state (data not shown). Strikingly, DDD mice, when placed in water, were able to swim with periods of floating and active swimming for at least a 3-min period (see [Video S1](#sv001){ref-type="supplementary-material"}), indicating that under certain conditions, movement can occur essentially without DA. Finally, in agreement with neurochemical data (see [Figure 1](#pbio-0030271-g001){ref-type="fig"}B), the recovery from this profound akinetic phenotype in DDD mice occurs approximately 16--24 h following treatment ([Figure 3](#pbio-0030271-g003){ref-type="fig"}J). The full recovery of animals allows repeated treatment with αMT, and, in fact, DAT-KO mice chronically treated with αMT (100 mg/kg, IP, once every 3 d) for a period of 40 wk showed no negative consequences \[[@pbio-0030271-b60]\]. ::: {#pbio-0030271-g002 .fig} Figure 2 ::: {.caption} ###### Display of Behavioral Phenotypes of DDD Mice \(A) Akinesia and rigidity of DDD mice. Photos were taken 30 min after treatment of DAT-KO mice with αMT (250 mg/kg IP). \(B) Ptosis in DDD mice. Photos were taken 3 h after treatment of DAT-KO mice with αMT (250 mg/kg IP). Lower panel: Mouse on the left side is saline-treated DAT-KO mouse, whereas the mouse on the right side is αMT-treated DAT-KO mouse. ::: ![](pbio.0030271.g002) ::: ::: {#pbio-0030271-g003 .fig} Figure 3 ::: {.caption} ###### αMT-Induced Impairment in Motor Control in DAT-KO Mice Dynamics of locomotor activity following systemic administration of αMT (250 mg/kg IP) and saline (30 min after placement in the locomotor activity chamber) in WT (A) and DAT-KO (B) mice (n = 6--8 per group). Analysis of total distance traveled for 210 min after drug administration revealed significant effect of αMT treatment (p \< 0.05; Student\'s t-test) in DAT-KO but not WT mice (WT-saline, 516 ± 50 cm/210 min; WT-αMT, 505 ± 98 cm/210 min; DAT-KO--saline, 18,489 ± 4,795 cm/210 min; DAT-KO--αMT, 448 ± 75 cm/210 min). αMT (injected at time 0) induced profound alterations in the akinesia (C), catalepsy (D), grasping (E), bracing (F) tremor (G), and ptosis (H) tests, but did not affect the righting reflex (I) in DAT-KO mice. Behavioral tests were performed as described in [Materials and Methods](#s4){ref-type="sec"}. At all the time points, DAT-KO mice were significantly different versus respective values (data not shown) of saline-treated DAT-KO controls (p \< 0.05; Student\'s t-test n = 6 per group) in these tests with exception of 15-min time point for ptosis (H) and all time points for righting reflex test (I). In WT mice only the akinesia test (C) revealed minor, yet significant, effect (1.5--4 h after αMT treatment) versus values (data not shown) of the respective saline treated WT controls (p \< 0.05; Student\'s t-test; n = 6 per group). No significant alterations in any other test at any time point examined (D--I) was noted in αMT-treated versus saline treated (data not shown) WT mice. Locomotor activity is restored in DAT-KO mice 16--24h after αMT (250 mg/kg IP) treatment (J). ::: ![](pbio.0030271.g003) ::: L-DOPA and Nonselective DA Agonists Restore Motor Activity in DDD Mice {#s2c} ---------------------------------------------------------------------- The locomotor restoring effects exhibited by L-DOPA and DA agonists in various models of DA deficiency form one of the best-established paradigms in neuroscience \[[@pbio-0030271-b03],[@pbio-0030271-b15],[@pbio-0030271-b45],[@pbio-0030271-b61]\]. As expected, high doses of L-DOPA alone ([Figure 4](#pbio-0030271-g004){ref-type="fig"}A), or lower doses of L-DOPA given along with carbidopa ([Figure 4](#pbio-0030271-g004){ref-type="fig"}B--[4](#pbio-0030271-g004){ref-type="fig"}D) to reduce its peripheral metabolism via L-AADC inhibition, effectively restore locomotion in DDD mice. In fact, these treatments temporarily restore locomotion to the levels observed in untreated DAT-KO mice ([Figure 4](#pbio-0030271-g004){ref-type="fig"}A--[4](#pbio-0030271-g004){ref-type="fig"}D), which are normally at least 10 times more active than WT mice when placed into a novel environment \[[@pbio-0030271-b11],[@pbio-0030271-b13]\]. Other manifestations associated with DA deficiency as described in [Figure 3](#pbio-0030271-g003){ref-type="fig"} were also essentially completely reversed (data not shown). ::: {#pbio-0030271-g004 .fig} Figure 4 ::: {.caption} ###### L-DOPA and Nonselective DA Agonists Are Effective in Restoring Locomotion in DDD Mice DAT-KO mice were placed in the locomotor activity chamber and 30 min later were treated with αMT (250 mg/kg IP) and 1 h after αMT were challenged with single or multiple doses of a drug (interval between treatments is 1 h). L-DOPA itself (A) or in combination with carbidopa (B--D) effectively restored locomotion in DDD mice, as revealed by the significant effect of L-DOPA at doses 100 and 200 mg/kg IP, or combinations of L-DOPA/carbidopa at doses 20/20, 50/20, and 50/50 mg/kg, IP (analysis of total distance traveled for 1 h after each dose of the drug; p \< 0.05, two-tailed Mann-Whitney U test versus respective values in saline-treated DDD mice; data not shown). Nonselective DA receptor agonists, apomorphine (E) at doses 2 and 3 mg/kg SC, and pergolide (F) at doses 5, 10, and 20 mg/kg IP, induced locomotion in DDD mice (analysis of total distance traveled for 1 h after each dose of the drug; p \< 0.05, two-tailed Mann-Whitney U test, versus respective values in saline-treated DDD mice; data not shown). D2 DA receptor agonists bromocriptine (G), quinpirole (H), and D1 DA receptor agonist (+)-SKF81297 (I) were not effective, but the combinations of D1 and D2 DA agonists (+)-SKF81297 plus quinpirole at doses 5/1 and 10/5 mg/kg IP, induced significant locomotion in DDD mice (analysis of total distance traveled for 1 h after each treatment; p \< 0.05, two-tailed Mann-Whitney U test versus respective values in saline-treated DDD mice; data not shown). Experiments were performed in 6--12 mice per group. ::: ![](pbio.0030271.g004) ::: Efficacy of exogenous direct DA agonists was also tested in this model. Although the nonselective D1/D2 DA receptor agonists apomorphine and pergolide were somewhat effective in inducing forward locomotion ([Figure 4](#pbio-0030271-g004){ref-type="fig"}E and [4](#pbio-0030271-g004){ref-type="fig"}F), the activity levels of DDD mice following these treatments were substantially lower than those induced by L-DOPA. Strikingly, the selective D1 DA receptor agonist (+)-SKF81297 and D2 DA receptor agonists, bromocriptine and quinpirole, were ineffective in inducing forward locomotion when administered separately ([Figure 4](#pbio-0030271-g004){ref-type="fig"}G--[4](#pbio-0030271-g004){ref-type="fig"}I). However, the combined administration of the D1 and D2 agonists (+)-SKF81297 plus quinpirole restored movement and induced forward locomotion ([Figure 4](#pbio-0030271-g004){ref-type="fig"}J), supporting the well-established cooperative interaction of D1 and D2-like DA receptors in locomotor activity \[[@pbio-0030271-b62]\]. Movement-Restoring Actions of Amphetamine Derivatives in DDD Mice {#s2d} ----------------------------------------------------------------- The loss of DA signaling that creates the motor symptoms of PD occurs upstream of many nondopaminergic pathways. This suggests that activation or inhibition of some of these downstream neuronal circuits could potentially reverse the motor deficits independent of restoration of DA activity. We, therefore, tested several non-dopaminergic compounds that potentially could reverse the consequences of severe DA deficiency in DDD mice (see [Table 1](#pbio-0030271-t001){ref-type="table"}). Many of these compounds have been found to be effective in restoring some aspects of movement control in one or another experimental animal model of PD and/or in PD patients \[[@pbio-0030271-b21],[@pbio-0030271-b26],[@pbio-0030271-b27],[@pbio-0030271-b48],[@pbio-0030271-b49]\]. However, in DDD mice none of the drugs were effective in restoring the major aspects of movement control required for forward locomotion (distance traveled). Although it is likely that the lack of locomotor effects of these drugs in DDD mice is related to an unprecedented level of DA depletion in these mice, it should be emphasized that in our studies only a few doses or combinations of drugs were tested. Furthermore, several treatments, although not inducing forward locomotion per se, were, nevertheless, somewhat effective in reversing other manifestations of DA deficiency. For example, the NMDA receptor antagonist MK-801 was able to reduce rigidity and promote weak, disorganized movement that however did not result in a significant increase in forward locomotion ([Table 1](#pbio-0030271-t001){ref-type="table"}). Synthetic amino acid L-DOPS (L-threo-3,4-dihydroxyphenylserine), which is decarboxylated to NE by L-AADC, selectively reversed ptosis in DDD mice. Cumulative dosing experiments revealed ptosis scores (measured 1h after each treatment) of 2.50 ± 0.28 after 100 mg/kg, 0 after 200 mg/kg, and 0 after 400 mg/kg IP of L-DOPS (n = 4), whereas corresponding values for saline-treated controls (n = 6) were 3.3 ± 0.3, 3.7 ± 0.2, and 3.7 ± 0.2, respectively. Effects of 200 and 400 mg/kg of L-DOPS on ptosis in DDD mice were significantly different as compared to respective control values (p \< 0.05, Student\'s t-test) supporting an important role of NE in this behavioral manifestation \[[@pbio-0030271-b59]\]. Similarly, high doses of the trace amine β-phenylethylamine \[[@pbio-0030271-b44],[@pbio-0030271-b63]\] (with or without concomitant inhibition of MAO) did not induce forward locomotion, but did promote weak stereotypic reactions, such as head-weaving and sniffing (data not shown). Further investigations will be required to fully evaluate the efficacy of these drugs in DDD mice. ::: {#pbio-0030271-t001 .table-wrap} Table 1 ::: {.caption} ###### Treatments That Were Not Effective in Restoring Forward Locomotion in DDD Mice ::: ![](pbio.0030271.t001) ::: Unexpectedly, this initial screening revealed a potent effect of amphetamine derivatives on behavioral manifestations of DDD mice. High doses of d-amphetamine, d-methamphetamine, 4-chloro-amphetamine, phentermine, (±)-MDE ((±)-N-ethyl-3,4-methylenedioxyamphetamine HCl), (+)-MDE ((+)-N-ethyl-3,4-methylenedioxyamphetamine HCl), (−)-MDE ((−)-N-ethyl-3,4-methylenedioxyamphetamine HCl), (±)-MDA ((±)-3,4-methylenedioxyamphetamine HCl), (±)-6-OH-MDA ((±)-6-hydroxy-3,4-methylenedioxyamphetamine HCl), (±)-MDMA ((±)-3,4-methylenedioxymethamphetamine HCl), and (+)-MDMA ((+)-3,4-methylenedioxymethamphetamine HCl) were effective in reducing manifestations of akinesia and rigidity in DDD mice as detected in the catalepsy, grasping, and akinesia tests ([Figure 5](#pbio-0030271-g005){ref-type="fig"}A--[5](#pbio-0030271-g005){ref-type="fig"}C). However none of these drugs (with the exception of (+)-MDMA, see below) was effective in restoring movement control sufficiently to induce forward locomotion ([Table 1](#pbio-0030271-t001){ref-type="table"}). ::: {#pbio-0030271-g005 .fig} Figure 5 ::: {.caption} ###### Amphetamine Derivatives at High Doses Are Effective in Reversing Abnormal Motor Behaviors of DDD Mice DAT-KO mice were placed in the locomotor activity chamber and 30 min later were treated with αMT (250 mg/kg IP), and 1 h after αMT were challenged with single or multiple doses of drugs (in cumulative dosing experiments, the interval between treatments was 1 h). Grasping (A), catalepsy (B), and akinesia (C) tests were performed as described in [Materials and Methods](#s4){ref-type="sec"} 1 h after each dose (the only exception is (±)-MDMA at 80 mg/kg IP where measurements were performed 2 h after the drug administration). An asterisk indicates p \< 0.05 versus respective values of saline-treated DDD mice (one-way ANOVA followed by Dunnet\'s multiple comparison test). Experiments were performed in 6--16 mice per group. d-AMPH indicates d-amphetamine; METH, d-methamphetamine; and 4-chloro-AMPH, 4-chloro-amphetamine. ::: ![](pbio.0030271.g005) ::: DA-Independent Locomotor Effects of (+)-MDMA in DDD Mice {#s2e} -------------------------------------------------------- Among amphetamine derivatives, the most effective compound to counteract manifestations of akinesia and rigidity in DDD mice was (+)-MDMA ([Figure 5](#pbio-0030271-g005){ref-type="fig"}A--[5](#pbio-0030271-g005){ref-type="fig"}C). Thus, we tested (+)-MDMA in locomotor assay at even higher doses than those indicated in [Table 1](#pbio-0030271-t001){ref-type="table"}. As presented in [Figure 6](#pbio-0030271-g006){ref-type="fig"}A--[6](#pbio-0030271-g006){ref-type="fig"}C, (+)-MDMA at high doses was able to induce significant forward locomotion in DDD mice as measured by distance traveled in a locomotor activity test. This locomotor action of (+)-MDMA was observed in both cumulative ([Figure 6](#pbio-0030271-g006){ref-type="fig"}A) and single dose ([Figure 6](#pbio-0030271-g006){ref-type="fig"}B and [6](#pbio-0030271-g006){ref-type="fig"}C) treatments. In cumulative dosing experiments, a first treatment with 30 mg/kg of (+)-MDMA was not effective, but the subsequent administration of 60 mg/kg induced significant forward locomotion ([Figure 6](#pbio-0030271-g006){ref-type="fig"}A) as well as reversal of other behavioral manifestations (see [Figure 5](#pbio-0030271-g005){ref-type="fig"}A--[5](#pbio-0030271-g005){ref-type="fig"}C) in DDD mice. Finally, testing of various single doses clearly showed a dose-dependence of the locomotor effect of (+)-MDMA in DDD mice ([Figure 6](#pbio-0030271-g006){ref-type="fig"}C). ::: {#pbio-0030271-g006 .fig} Figure 6 ::: {.caption} ###### (+)-MDMA Induces Forward Locomotion in DDD Mice (A--C) DAT-KO mice were placed in the locomotor activity chamber and 30 min later were treated with αMT (250 mg/kg IP) and 1 h after αMT were challenged with single (B and C) or multiple doses (A) of a drug (interval between treatments is 1 h) (n = 10--16 per group). Repeated treatment with (+)-MDMA (30 and 60 mg/kg IP) induces forward locomotion in DDD mice (A). Analysis of total distance traveled for 1 h after 60 mg/kg IP of (+)-MDMA reveals significant effect of treatment versus respective period in saline-treated controls (p\<0.05, two-tailed Mann-Whitney U test, data not shown). Dynamics (B) and dose-response (C) of locomotor effect of (+)-MDMA in DDD mice are shown. Pretreatment with D1 and D2 DA antagonists (SCH23390, 0.1 mg/kg SC plus raclopride, 2 mg/kg IP) 30 min before 100 mg/kg IP (+)-MDMA) did not affect locomotor action of (+)-MDMA (C). \(D) (+)-MDMA (100 mg/kg IP) fails to affect DA dynamics in the striatum of DDD mice as measured by in vivo microdialysis. Data are presented as a percentage of the average level of DA measured in at least three samples collected before αMT administration (n = 4). Analysis of area under curve values for 120-min periods after (+)-MDMA administration revealed no significant difference in comparison with respective values in control group ([Figure 1](#pbio-0030271-g001){ref-type="fig"}F; p \> 0.05, two-tailed Mann-Whitney U test). (E and F) (+)-MDMA (E) as well as d-amphetamine and d-methamphetamine (F) at moderate doses potentiate locomotor-stimulating effect of subthreshold dose of L-DOPA/carbidopa (10/10 mg/kg IP). DAT-KO mice were treated with αMT as described above (A--C) and 45 min after αMT were injected with amphetamines. L-DOPA/carbidopa was injected 15 min after amphetamines, and distance traveled for 2h was measured (n = 6--15 per group). Note, that no forward locomotion was observed after these doses of (+)-MDMA, d-amphetamine and d-methamphetamine without L-DOPA/carbidopa, whereas L-DOPA/carbidopa (presented as drug dose 0) induced only a modest but significant (p \< 0.05) increase in locomotion over saline-treated controls (data not shown). Single asterisk indicates p \< 0.05; double asterisks indicate p \< 0.01; and triple asterisks indicate p \< 0.001 versus saline-treated controls (C) or L-DOPA/carbidopa-treated (10/10 mg/kg IP) group (E and F) (two-tailed Mann-Whitney U test). d-AMPH, d-amphetamine; METH, d-methamphetamine. ::: ![](pbio.0030271.g006) ::: The locomotor stimulating effect of amphetamine and its derivatives are classically thought to result from the massive efflux of DA from presynaptic DA terminals via a mechanism including displacement of DA from vesicular storage and reversal of DAT-mediated DA transport \[[@pbio-0030271-b07],[@pbio-0030271-b38]--[@pbio-0030271-b40]\]. However, in DDD mice, there is only a minimal amount of DA remaining (\<0.2%) and the lack of the DAT precludes the possibility of amphetamine-mediated DA efflux. In fact, in vivo microdialysis studies confirmed that (+)-MDMA, at the effective dose necessary to induce significant locomotor activation in DDD mice, did not produce any detectable increase in striatal extracellular DA ([Figure 6](#pbio-0030271-g006){ref-type="fig"}D). Moreover, this locomotor stimulation by (+)-MDMA was not inhibited by simultaneous blockade of D1/D2 DA receptors when DDD mice were pretreated with a combination of the D1 and D2 DA receptor antagonists, SCH23390 and raclopride ([Figure 6](#pbio-0030271-g006){ref-type="fig"}C). Similarly, this pretreatment did not prevent the effects of amphetamine and phentermine on the akinesia and rigidity in DDD mice in grasping and akinesia tests (see [Figure S1](#sg001){ref-type="supplementary-material"}). In contrast, the same D1/D2 DA receptor blockade completely abolished the locomotor stimulating effects of L-DOPA/carbidopa (50/50 mg/kg IP) in DDD mice (see [Figure S2](#sg002){ref-type="supplementary-material"}). Taken together, these data indicate that (+)-MDMA can affect movement control in a DA-independent manner \[64\] and, most importantly, provide a proof-of-principle that pharmacologic activation of nondopaminergic neuronal pathways may be sufficient to restore movement even in the absence of DA neurotransmission. It should be noted that the locomotor-stimulating effect of (+)-MDMA in DDD mice was observed only after high doses of the drug, which may be potentially neurotoxic \[[@pbio-0030271-b64]\]. However, the lack of the DAT renders dopaminergic neurons in DAT-KO mice significantly less sensitive to the neurotoxic effects of amphetamines, such as methamphetamine \[[@pbio-0030271-b65]\], as well as to MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine) \[[@pbio-0030271-b66],[@pbio-0030271-b67]\]; thereby providing a unique opportunity to evaluate effects to large doses of amphetamines that would be impossible in normal animals \[[@pbio-0030271-b38]\]. It should be mentioned also that mice are generally less sensitive to MDMA neurotoxicity, particularly with regards to the serotonergic system \[[@pbio-0030271-b68]\]. Nevertheless, to directly evaluate the neurotoxic potential of MDMA in DAT-KO mice, we treated DAT-KO and WT mice with an established neurotoxic regimen of (±)-MDMA administration (4 injections of 20 mg/kg IP, every 2 h) \[[@pbio-0030271-b69]\] and assessed striatal tissue DA and 5-HT levels 7 d later. As might be expected, no significant differences in both DA and 5-HT levels were found between (±)-MDMA-treated and saline-treated DAT-KO mice (saline-treated DAT-KO mice (n = 6): DA, 0.53 ± 0.03 ng/mg tissue; 5-HT, 0.36 ± 0.03 ng/mg tissue; (±)-MDMA-treated DAT-KO mice (n = 7): DA, 0.58 ± 0.04 ng/mg tissue; 5-HT, 0.40 ± 0.02 ng/mg tissue), whereas the same regimen of treatment resulted in lethality of all treated WT mice (n = 7). Furthermore, to test whether the locomotor-stimulating effect of (+)-MDMA may be evident under certain conditions with lower (nonneurotoxic) doses of the drug, we co-administered (+)-MDMA with a minimally effective dose of L-DOPA/carbidopa (10/10 mg/kg, IP.). As shown in [Figure 6](#pbio-0030271-g006){ref-type="fig"}E, a potent synergistic effect of L-DOPA/carbidopa and (+)-MDMA was observed. Furthermore, similar effects were observed with relatively moderate doses of d-amphetamine and d-methamphetamine ([Figure 6](#pbio-0030271-g006){ref-type="fig"}F). Thus, a DA-independent locomotor effect of amphetamines can be markedly enhanced with additional dopaminergic stimulation. It is also important to note that in a similar experiment, MAO inhibitor deprenyl (5, 10, or 20 mg/kg IP) failed to potentiate the effects of L-DOPA/carbidopa (data not shown), indicating that this effect is not related to the well-known MAO-inhibiting action of amphetamines \[[@pbio-0030271-b38]\]. Nomifensine, but Not GBR12909 Affects Rigidity and Akinesia in DDD Mice {#s2f} ----------------------------------------------------------------------- Finally, to evaluate the potential of other TA1 receptor ligands for their ability to affect motor control in DDD mice, we elected to compare the effects of two potent DAT blockers that have been shown to be markedly different with regards to their activity at TA1 receptor. It has been recently reported that the mixed DAT and NET inhibitor nomifensine can also potently activate TA1 receptor whereas the selective DAT blocker GBR12909 completely lacks the ability to interact with TA1 receptor \[[@pbio-0030271-b42]\]. In DDD mice, both nomifensine and GBR12909 at doses tested (cumulative treatment with 10 and 30 mg/kg IP) were not effective in inducing forward locomotion or reversing catalepsy (data not shown). Nevertheless, nomifensine significantly reduced akinesia and rigidity in grasping and akinesia tests ([Figure 7](#pbio-0030271-g007){ref-type="fig"}A and [7](#pbio-0030271-g007){ref-type="fig"}B), whereas no such effects were observed with equivalent doses of GBR12909 ([Figure 7](#pbio-0030271-g007){ref-type="fig"}A and [7](#pbio-0030271-g007){ref-type="fig"}B). ::: {#pbio-0030271-g007 .fig} Figure 7 ::: {.caption} ###### Nomifensine, but Not GBR12909, Is Effective in Reversing Abnormal Motor Behaviors of DDD Mice DAT-KO mice were placed in the locomotor activity chamber and 30 min later were treated with αMT (250 mg/kg IP). Mice were challenged 1 h later with two doses (10 and 30 mg/kg IP) of each drug or saline with a 1 h interval between treatments (n = 9--15 per group). Grasping (A) and akinesia (B) tests were performed as described in [Materials and Methods](#s4){ref-type="sec"} 1 h after each dose. A single asterisk indicates p \< 0.05, double asterisks indicate p \< 0.01, and triple asterisks indicate p \< 0.001 versus respective values of saline-treated DDD mice (one-way ANOVA followed by Dunnet\'s multiple comparison test). Note that no significant differences between GBR12909-treated mice and saline-treated controls in both tests were found, whereas nomifensine-treated mice were significantly different from GBR12909-treated mice (p \< 0.05) in both experimental paradigms and doses tested. ::: ![](pbio.0030271.g007) ::: Discussion {#s3} ========== In this study we demonstrate that inhibition of DA synthesis in DAT-KO mice represents a straightforward approach for developing an acute model of severe DA deficiency exhibiting a characteristic behavioral phenotype that can be utilized for testing perspective anti-PD treatments. Furthermore, these observations provide functional evidence for an important role of DAT-mediated recycling mechanism in the maintenance of intraneuronal DA. Finally, the novel DAT- and DA-independent locomotor action of amphetamines identified in these mice directly demonstrates the possibility of movement in a DA-independent manner. Role of DAT-Mediated DA Recycling in the Maintenance of Intraneuronal DA Storage {#s3a} -------------------------------------------------------------------------------- DAT is commonly known as a major regulator of the duration and intensity of extracellular DA signaling. However the important role of DAT in the control and maintenance of the intraneuronal DA storage pool frequently remains overlooked. It is generally assumed that the intraneuronal storage of DA is replenished primarily from newly synthesized DA with some contribution from recycled DA. However, several lines of evidence support a predominant role of DAT-mediated recycling of DA for the maintenance of the large storage pool in DA terminals. First, mice lacking the DAT display dramatically decreased (20-fold) striatal tissue DA content, that mostly represents intraneuronal DA concentrations. Second, as we demonstrate in the present study, the remaining DA in all compartments is extremely sensitive to TH inhibition. Furthermore, pharmacologic studies have shown that significant DA depletion may occur after administration of DAT inhibitors, particularly after chronic drug treatment \[[@pbio-0030271-b13]\]. Importantly, in the frontal cortex, where DAT levels are normally low in comparison to the striatum, tissue DA concentration is also low and can be more significantly affected than in the striatum by αMT \[[@pbio-0030271-b70]\]. It is likely that the newly synthesized DA does not contribute directly to the large storage pool of DA in nigrostriatal terminals, but rather contributes to it indirectly via released and recycled DA. Thus, a cooperative function of both DA synthesis and transporter-mediated recycling processes is necessary for the maintenance of normal presynaptic monoamine concentrations. A Novel Acute Mouse Model of Severe DA Deficiency, DA-Depleted DAT-KO (DDD) Mice {#s3b} -------------------------------------------------------------------------------- By using a combination of genetic and pharmacologic approaches we have developed a novel acute mouse model of severe DA deficiency, DDD mice. The lack of an active recycling mechanism in DAT-KO mice results in a profound depletion of intraneuronal concentrations of DA leaving the remaining DA entirely dependent on ongoing synthesis. As a result, inhibition of DA synthesis essentially eliminates striatal DA in these mice leading to the extreme behavioral manifestations. In fact, DDD mice demonstrate a unique set of behaviors that reproduces symptoms of PD with high fidelity. Thus, the lack of DA combined with the striking and highly reproducible behavioral phenotype in these mice can be used as an excellent tool to evaluate the potential of drugs that can affect locomotion in a DA-independent manner. Furthermore, by adapting the dose of αMT to produce various degrees of DA depletion, these mice can also be employed to find novel approaches to restore movement under conditions of partially impaired DA transmission that might be more relevant to most PD cases. Several rodent models have been developed to understand pathological processes leading to PD and/or to screen for novel therapeutic strategies \[29,30,34−36,71\]. These models either recapitulate the loss of DA through pharmacologic or genetic manipulation, or recapitulate the neurodegenerative process through administration of selective neurotoxins and, recently, through mutations of specific proteins. However, in many of these models only incomplete and highly variable levels of DA depletion are achieved often precluding an accurate recapitulation of the neurological manifestations of PD. This poor behavioral expression of PD-related behaviors generally results in high level of false-positive results in drug screening tests in general, and particularly in those attempted to identify non-DA therapies \[[@pbio-0030271-b72]\]. Among several genetic mouse models of DA deficiency available today \[[@pbio-0030271-b73],[@pbio-0030271-b74]\], the most effective was developed by inactivation of TH in DA neurons (DA-deficient \[DD mice\]) \[[@pbio-0030271-b03],[@pbio-0030271-b75]--[@pbio-0030271-b80]\]. DD mice have provided important insights into the role of DA in movement control, feeding, and reward. This mutation results in severely impaired movement and feeding, which become apparent at 10 d and leads to death by 30 d. To maintain viable mice with the ability to move and feed requires daily treatment with L-DOPA, which results in an oscillation of striatal DA from about 1% to 10% over 24 h \[[@pbio-0030271-b77],[@pbio-0030271-b81]\]. Many behavioral manifestations observed in DDD mice in this study, such as rigidity and akinesia, were observed previously in DD mice \[[@pbio-0030271-b03],[@pbio-0030271-b76],[@pbio-0030271-b79]\]. Importantly, both of these models showed temporal locomotor reactivity to stress and demonstrated normal righting reflex and ability to swim, indicating that certain movements may occur in a DA-independent manner. Despite these similarities, some important differences were noted between these two genetic models of severe DA dysfunction. In DD mutant mice, a lack of TH resulting in permanently decreased DA signaling, as well as daily treatments with L-DOPA render these mice extremely supersensitive to DA stimulations \[[@pbio-0030271-b81]\], whereas excessive DA signaling in DAT-KO mice results in compensatory down-regulation (but nonuniform) of DA receptors \[[@pbio-0030271-b11],[@pbio-0030271-b13]\]. This may explain why certain behavioral manifestations of DA deficiency such as rigidity and akinesia may be more robust in DDD mice, whereas tremor was not observed in DD mutants \[[@pbio-0030271-b03],[@pbio-0030271-b76]\]. Furthermore, efficacy of L-DOPA and DA agonists are remarkably higher in DD in comparison to DDD mice \[[@pbio-0030271-b03],[@pbio-0030271-b76],[@pbio-0030271-b81]\]. Additionally, several other drugs, such as caffeine and N-methyl-D-aspartate receptor antagonist MK-801, that are able to induce locomotion in DD mutants \[[@pbio-0030271-b75],[@pbio-0030271-b80]\] are not effective in DDD mice ([Table 1](#pbio-0030271-t001){ref-type="table"}). In fact, down-regulation of DA receptor responsiveness combined with the extreme level of DA depletion in DDD mice may favor these mice as a very conservative approach for evaluating drugs that can affect locomotion in a DA-independent manner. Furthermore, rapid and effective elimination of DA in DDD mice may provide a simple in vivo approach to study DA receptor signaling \[[@pbio-0030271-b82]\] and/or to define neuronal circuitry involved in locomotor control \[[@pbio-0030271-b83]\]. DA-Independent Locomotor Action of Amphetamines {#s3c} ----------------------------------------------- Intriguingly, in both DD and DDD mice d-amphetamine was effective in restoring at least some aspects of locomotor behaviors. In DD mice, d-amphetamine (5 mg/kg IP) induced potent locomotor activation essentially up to the levels observed in WT controls. At the same time, a second treatment 2 h later by the same dose of the drug failed to induce locomotion in DD mice suggesting that this effect is dependent upon residual (after L-DOPA administration) DA which might be depleted by the first treatment with the drug \[[@pbio-0030271-b76]\]. In DDD mice, d-amphetamine itself was not able to induce forward locomotion at doses up to 60 mg/kg, but it produced significant effects on other manifestations of DA deficiency. Moreover, co-administration of relatively moderate doses of amphetamine (15 and 20 mg/kg) with a subthreshold dose of L-DOPA resulted in a marked locomotor activation of DDD mice. Thus, some DA tone seems to be necessary to express the full magnitude of locomotor activation by amphetamine, but it is evident that there is a DA-independent component of action that contributes to the overall effect of the drug. Further evidence for this idea relates to the fact that many other amphetamine derivatives are also active in reversing certain behavioral manifestations in DDD mice. Strikingly, both single and repeated treatment with (+)-MDMA was effective in inducing forward locomotion essentially without any contribution of DA. It is important to note that a potent anticataleptic effect of MDMA in haloperidol-treated rats \[[@pbio-0030271-b84]\] and antiakinetic effects in 6-OH-DA--lesioned rats \[[@pbio-0030271-b85]\] and MPTP-treated monkeys \[[@pbio-0030271-b64]\] have been recently reported. The present observations support these findings and suggest that these actions are not unique to MDMA but may be extended to other amphetamines. Further characterization of these unexpected effects of amphetamines may provide a novel framework in the search for potential anti-Parkinsonian drugs. Amphetamine derivatives are known mainly as indirect enhancers of monoaminergic (DA, NE, and 5-HT) transmission via complex interactions with the plasma membrane monoamine transporters and the vesicular storage of these monoamines \[[@pbio-0030271-b07],[@pbio-0030271-b10],[@pbio-0030271-b12],[@pbio-0030271-b38],[@pbio-0030271-b39]\]. It should be reiterated that a lack of DAT in DAT-KO mice excludes the possibility of major effects of amphetamines on DAT-mediated DA efflux from presynaptic DA stores \[[@pbio-0030271-b40]\]. Furthermore, a blockade of D1/D2 DA receptors was ineffective in preventing the locomotor stimulating action of (+)-MDMA. Thus, it is virtually impossible that the observed effects of MDMA and other amphetamines in DDD mice are directly related to DA transmission. Although it is possible that this effect may be due to transporter-mediated action of amphetamines on NE or 5-HT transmission \[[@pbio-0030271-b38],[@pbio-0030271-b40],[@pbio-0030271-b86]\], it should be noted that among several NE- and 5-HT--related drugs tested (desipramine, clonidine, the NE precursor DOPS, fluoxetine, 5-methoxy-N,N-dimethyltryptamine, 5-methyl-N,N-dimethyltryptamine, α-ethyltryptamine, and 5-HT1B agonist RU24969), none were effective in DDD mice in tests of forward locomotion or akinesia and rigidity (data not shown). Similarly, no locomotor effect of MAO-A or MAO-B inhibitors was observed in these mice, indicating that the locomotor effect of amphetamines may not be explained by MAO-inhibitory action \[[@pbio-0030271-b38]\]. Furthermore, it should be underlined that locomotor actions of amphetamines observed in DDD mice occur at doses that are much higher than necessary to induce classic transporter-mediated effects \[[@pbio-0030271-b10],[@pbio-0030271-b38],[@pbio-0030271-b83]\]. Amphetamines share close structural similarity with an endogenous trace amine of unknown function β-phenylethylamine \[[@pbio-0030271-b87]\]. Amphetamines and β-phenylethylamine similarly interact with the plasma membrane monoamine transporters to elevate extracellular monoamine concentrations \[[@pbio-0030271-b63]\]. Intriguingly, recent evidence indicates that many amphetamine derivatives, including MDMA, may also act directly as agonists of trace amine TA1 receptors, that are known to be activated by β-phenylethylamine \[[@pbio-0030271-b42],[@pbio-0030271-b88]\]. Several members of the family of trace amine receptors have been identified, however little is known about the pharmacology and functional role of these receptors in mammalian physiology \[[@pbio-0030271-b43],[@pbio-0030271-b44],[@pbio-0030271-b63]\]. It is reasonable to suggest that activation of TA1 receptors \[[@pbio-0030271-b42]\] or other trace amine receptors may provide a potential mechanism for DA-independent locomotor effect of MDMA and amphetamines in DDD mice. In line with this hypothesis, we observed that the DAT blocker nomifensine that can activate TA1 receptor, but not GBR12909 which is devoid this activity \[[@pbio-0030271-b42]\], is able to affect motor control in DDD mice. It should be noted, however, that in our initial exploration in DDD mice, we did not observe clear locomotor effects for any trace amine tested; but only a few doses, routes of administration, and combinations with enzyme inhibitors were investigated. Further detailed investigations will be needed to clarify the mechanism of locomotor action of amphetamines in DDD mice. Conclusions {#s3d} ----------- In summary, these results provide additional functional evidence for the critical role of DAT in the maintenance of DA storage in presynaptic terminals. Rapid and effective abolishment of DA by inhibition of DA synthesis in DAT-KO mice provides a novel approach to develop severe DA deficiency that might be used to identify neuronal mechanisms involved in motor control in the absence of DA. Amphetamines are capable of affecting neuronal systems involved in motor control through mechanisms independent of DAT, in particular, and DA in general. Materials and Methods {#s4} ===================== {#s4a} ### Animals {#s4a1} DAT-KO mice were generated as previously described \[[@pbio-0030271-b11]\]. Animal care was in accordance with the Guide for Care and Use of Laboratory Animals (National Institutes of Health publication \#865--23, Bethesda, Maryland, United States) with an approved protocol from the Duke University Institutional Animal Care and Use Committee. C57BL/6J × 129Sv/J hybrid WT and DAT-KO mice, 3--5 mo old, of both sexes were used. None of animals used in these studies had the neurodegenerative phenotype sporadically observed in DAT-KO mice \[[@pbio-0030271-b60]\]. ### Drugs {#s4a2} Drugs or saline (0.9% NaCl) were administered intraperitoneally (IP) or subcutaneously (SC) in a volume of 10 ml/kg. The drugs were either from Sigma (St. Louis, Missouri, United States) or supplied by the National Institute of Drug Abuse (NIDA). Drugs provided by the NIDA Drug Supply Program included: (±)-MDMA, (+)-MDMA, (±)-6-OH-MDA, (±)-MDA, (±)-MDE, (+)-MDE, (−)-MDE, and AET (α-ethyl-tryptamine acetate). ### Neurochemical assessments {#s4a3} Striatal tissue contents of DA and frontal cortical tissue levels of NE were assessed using HPLC-EC (high performance liquid chromatography with electrochemical detection) as described \[[@pbio-0030271-b08]\]. In vivo microdialysis measurements of striatal extracellular DA levels in freely moving mice were performed at least 24 h after implantation of a microdialysis probe as described previously \[[@pbio-0030271-b50]\]. Dialysate samples were assayed for DA using HPLC-EC. ### Behavioral methods {#s4a4} Locomotor activity of littermate WT and DAT-KO mice was measured in an Omnitech CCDigiscan (Accuscan Instruments, Columbus, Ohio United States) activity monitor under bright illumination \[[@pbio-0030271-b83]\]. All behavioral experiments were performed between 10:00 AM and 5:00 PM. Activity was measured at 5-min intervals. To evaluate the effects of drugs on motor behaviors, mice were placed into activity monitor chambers (20 × 20 cm) for 30 min and then treated with αMT (250 mg/kg IP). A drug or combination of drugs were injected 1 h after αMT administration, and various parameters of locomotor activity were monitored for up to 3 h. In cumulative dosing experiments, animals were treated with increasing doses of drugs after a 1-h interval. For the akinesia test, the mouse is held by the tail so that it is standing on forelimbs only and moving on its own. The number of steps taken with both forelimbs was recorded during a 30-s trial \[[@pbio-0030271-b57]\]. The presence of catalepsy was determined and measured by placing the animal\'s forepaws on a horizontal wooden bar (0.7 cm in diameter), 4 cm above the tabletop. The time until the mouse removed both forepaws from the bar was recorded, with a maximum cut-off time of 3 min \[[@pbio-0030271-b53]\]. In the grasping test of muscular rigidity, the mouse is suspended by its forelimbs on a metal rod (diameter: 0.25 cm) positioned approximately 20 cm above the table. The time the animal remains on the rod (maximum 1 min) was noted \[[@pbio-0030271-b58]\]. To assess rigidity in a bracing task, the number of steps taken with each forelimb when the mouse is pushed sideways over a distance of 50 cm was recorded \[[@pbio-0030271-b57]\]. Tremor was scored visually in mice using the rating scale \[[@pbio-0030271-b54]\]: 0, no tremor; 1, occasional isolated twitches; 2, moderate or intermittent tremor associated with short periods of calm; and 3, pronounced continuous tremor. Ptosis was scored as described \[[@pbio-0030271-b89]\]: 4, eyes completely closed; 2, half-open eyes; and 0, wide-open eyes; with 1 and 3 indicating intermediate values. The righting reflex was evaluated by turning the mouse onto its back five times. Normal mice immediately turn themselves over, to right themselves onto all four feet. Righting reflex was scored as follows: 0, no impairment; 1, on side one to two times; 2, on side three to four times; 3, on side five times; 4, on back one to two times; 5, on back three to four times; 6, on back five times; 7, sluggish when placed on back; and 8, righting response absent when on back and tail pinched \[[@pbio-0030271-b55]\]. ### Data analysis {#s4a5} The data are presented as mean ± SEM and analyzed using a two-tailed Student\'s t-test and one-way analysis of variance (ANOVA) followed by Dunnet\'s multiple comparison test or a two-tailed Mann-Whitney U test when appropriate. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### D1/D2 DA Receptor Blockade Does Not Prevent the Effects of Amphetamine and Phentermine on Rigidity and Akinesia in DDD Mice (4.8 MB TIF). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### D1/D2 DA Receptor Blockade Prevents the Locomotor Stimulating Effect of L-DOPA/Carbidopa in DDD Mice (4.4 MB TIF). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S1 ::: {.caption} ###### Behavioral Phenotype of DDD mice (10 MB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: This work was supported in part by grants from the National Institutes of Health NS-19576 and MH-40159. Competing interests. The authors have declared that no competing interests exist. Author contributions. TDS, MGC, and RRG conceived and designed the experiments. TDS and RRG performed the experiments. TDS, JMB, LSB, WCW, and RRG analyzed the data. WCW and MGC contributed reagents/materials/analysis tools. TDS, JMB, LSB, WCW, MGC, and RRG wrote the paper. Citation: Sotnikova TD, Beaulieu JM, Barak LS, Wetsel WC, Caron MG, et al. (2005) Dopamine-independent locomotor actions of amphetamines in a novel acute mouse model of Parkinson disease. PLoS Biol 3(8): e271. 5-HT : serotonin αMT : α-methyl-p-tyrosine DA : dopamine DAT : dopamine transporter DAT-KO mice : dopamine transporter knockout mice DD mice : dopamine-deficient mice DDD mice : dopamine-deficient DAT-KO mice IP : intraperitoneal L-AADC : L-aromatic acid decarboxylase MAO : monoamine oxidase NE : norepinephrine NET : norepinephrine transporter PD : Parkinson disease SC : subcutaneous SNc : Substantia Nigra Pars Compacta TA1 receptor : trace amine 1 receptor TH : tyrosine hydroxylase VMAT2 : vesicular monoamine transporter-2 VTA : ventral tegmental area WT : wild-type	PubMed Central	2024-06-05T03:55:59.739064	2005-8-2	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181539/", "journal": "PLoS Biol. 2005 Aug 2; 3(8):e271", "authors": [ { "first": "Tatyana D", "last": "Sotnikova" }, { "first": "Jean-Martin", "last": "Beaulieu" }, { "first": "Larry S", "last": "Barak" }, { "first": "William C", "last": "Wetsel" }, { "first": "Marc G", "last": "Caron" }, { "first": "Raul R", "last": "Gainetdinov" } ] }
PMC1181540	Introduction {#s1} ============ Despite its ubiquity and importance, aging remains a poorly understood process. This lack of understanding is due in part to the complexity of aging, which is characterized by the gradual and progressive decline of numerous physiological processes and homeostasis, eventually leading to death \[[@pbio-0030274-b01]--[@pbio-0030274-b06]\]. However, recent progress in aging research has made it clear that aging processes are amenable to biochemical and genetic dissection, in both humans and model organisms \[[@pbio-0030274-b03]--[@pbio-0030274-b06]\]. Both environmental and genetic alterations in model organisms have been found to have profound effects on aging and lifespan. In particular, dietary restriction has been found to dramatically increase the lifespan of organisms including yeast, flies, nematodes, and mammals \[[@pbio-0030274-b04],[@pbio-0030274-b05]\]; the mechanism by which this intervention reduces mortality rates is still under investigation. Additionally, inactivating myriad single genes has been found to be able to significantly increase the average lifespan of model organisms \[[@pbio-0030274-b04],[@pbio-0030274-b05]\], and the identification of the pathways to which these genes belong is beginning to shed light on their possible modes of action. For example, many genes implicated in regulating lifespan in the nematode Caenorhabditis elegans belong to the insulin/IGF (insulin-like growth factor) signaling pathway, indicating possible connections to metabolism and a state of arrested development known as "dauer" \[[@pbio-0030274-b04],[@pbio-0030274-b05]\]. Another class of genes found to be involved in the aging process is related to the production and scavenging of molecules known as reactive oxygen species (ROS), thus providing genetic evidence in support of a mechanistic theory of aging known as the free radical theory \[[@pbio-0030274-b02],[@pbio-0030274-b03]\]. The free radical theory of aging was first introduced by Harman almost half a century ago \[[@pbio-0030274-b02]\]. This theory, as well as the related "rate of living" theory proposed earlier by Pearl \[[@pbio-0030274-b01]\], holds that aging is at least in part due to deleterious side effects of aerobic respiration. Specifically, mitochondrial activity leads to the production of ROS that can damage many cellular components, including DNA, lipids, and proteins \[[@pbio-0030274-b03]\]. These ROS, such as the hydroxyl radical (OH•) and hydrogen peroxide (H~2~O~2~), are produced in large part by the mitochondrial electron transport chain. The free radical theory has garnered widespread support in recent years; in addition to the genetic evidence mentioned above, studies from a number of model organisms showing that decreasing ROS levels leads to an increase in lifespan indicate that ROS can strongly modulate the aging process \[[@pbio-0030274-b03]--[@pbio-0030274-b05]\]. Exactly how macromolecules damaged by ROS may lead to aging has been studied in detail in recent years, and the human brain has been intensively examined in this regard because of its overall importance in human senescence. For example, up to one-third of the proteins in the brains of elderly individuals may be oxidatively damaged, and these damaged proteins have been shown to sometimes have diminished catalytic function \[[@pbio-0030274-b03],[@pbio-0030274-b06]\]. One recent study of aging in the human brain demonstrated that oxidative damage to DNA can be caused by mitochondrial dysfunction, and tends to accumulate preferentially in some areas of the genome that include promoters, resulting in lower levels of transcription \[[@pbio-0030274-b07]\] (possibly due to loss of transcription factor or other protein binding \[[@pbio-0030274-b08]--[@pbio-0030274-b10]\]). In this same study, genome-wide patterns of aging-associated gene expression change in one region of the human brain cortex (the frontal pole; [Figure 1](#pbio-0030274-g001){ref-type="fig"}) were measured using DNA microarrays, and genes that had decreased transcription with age were shown to be the ones that are most susceptible to oxidative damage \[[@pbio-0030274-b07]\]. Since different regions of the human brain have been shown to accumulate DNA damage at different rates \[[@pbio-0030274-b11],[@pbio-0030274-b12]\], it is reasonable to suppose that these different regions may show different gene expression changes with age as a result. ::: {#pbio-0030274-g001 .fig} Figure 1 ::: {.caption} ###### The Seven Regions of the Human Brain Analyzed in This Work The seven regions---anterior cingulate cortex, Broca\'s area, caudate nucleus, cerebellum, frontal pole, prefrontal cortex, and primary visual cortex---are indicated in red. ::: ![](pbio.0030274.g001) ::: Complementing studies of aging differences in various tissues within a single species, research into the evolution of aging has begun to shed light on the similarities and differences between species, although the expectation for how well conserved the effects of aging will be on a gene-by-gene basis is still unclear. One the one hand, if aging is largely caused by the deleterious effects of many alleles late in life---as is often the interpretation of two widely held models for the evolution of aging, known as "antagonistic pleiotropy" and "mutation accumulation" \[[@pbio-0030274-b04],[@pbio-0030274-b05],[@pbio-0030274-b13]\]---then rapid evolutionary change of the aging process, at least at a mechanistic level, should be impossible. This reasoning is supported by empirical observations that many aging-related factors (such as ROS-induced damage) and pathways (such as insulin/IGF signaling) appear to be highly conserved \[[@pbio-0030274-b04],[@pbio-0030274-b05],[@pbio-0030274-b13]\]. However, even if the mechanistic underpinnings of aging are indeed relatively constant, the phenotypic effects may be subject to dramatic change. This is best demonstrated by the observation that artificial selection on model organisms in the lab can lead to dramatic changes in lifespan in a very small number of generations \[[@pbio-0030274-b05],[@pbio-0030274-b13]\], implying that the consequences of aging could be subject to rapid evolutionary change in the wild as well. Clearly, this issue cannot be resolved solely by laboratory evolution experiments or theoretical work. One promising approach to answering this question of evolutionary conservation lies at the level of gene expression: Do orthologous genes tend to undergo the same patterns of expression changes with age in diverse species, or can a common factor such as ROS lead to different gene expression patterns in different organisms? Using DNA microarrays, this question can now be addressed in a systematic, genome-wide manner. One such study found that a small but significant portion of aging-related gene expression changes are shared by the very distantly related nematode and fruit fly \[[@pbio-0030274-b14]\]; another study comparing aging patterns in muscle cells of two more closely related species, mouse and human, also found a great deal of divergence in aging patterns \[[@pbio-0030274-b15]\]. Although both of these studies are informative, neither addresses the questions of how quickly age-related gene expression patterns can evolve over short periods of time, and if humans in particular show unique patterns of aging not shared by closely related primates. The human brain is of particular interest for studying the divergence in phenotypes that have changed rapidly during evolution (such as aging). Brain-specific genes have undergone accelerated evolution in the lineage leading to human since the split with chimpanzee at the levels of both protein sequence and gene expression \[[@pbio-0030274-b16],[@pbio-0030274-b17]\], pointing to the numerous functional differences that have accumulated between these two species since their divergence only 5 to 7 million years ago. Aging in the human brain is also of interest because ROS-induced damage and age are both major risk factors in many neurodegenerative diseases (such as Alzheimer\'s, Parkinson\'s, and Amyotrophic Lateral Sclerosis \[[@pbio-0030274-b18]\]). In this study we addressed two questions about the relationship between gene expression and aging. First, using published data, we asked whether the pattern of gene expression change with age previously observed in the frontal pole \[[@pbio-0030274-b07]\] is representative of other regions of the human brain. Then, using data generated for this project, we asked how similar the aging-associated changes in gene expression observed in human brain \[[@pbio-0030274-b07]\] are to those observed in our closest living relative, the chimpanzee. Results {#s2} ======= In order to test whether different regions of the human brain show similar patterns of change with age, we utilized three independently published microarray expression datasets. These were: Lu et al. \[[@pbio-0030274-b07]\], mentioned above, in which the frontal pole regions of 30 individuals (aged 26--106 y) were used to identify hundreds of genes with clear up- or down-regulation associated with age; Khaitovich et al. \[[@pbio-0030274-b19]\], in which gene expression patterns of six brain regions ([Figure 1](#pbio-0030274-g001){ref-type="fig"}; prefrontal cortex, primary visual cortex, anterior cingulate cortex, Broca\'s area, caudate nucleus, and cerebellum) were studied in three individuals (aged 45, 45, and 70 y); and Evans et al. \[[@pbio-0030274-b20]\], in which three brain regions ([Figure 1](#pbio-0030274-g001){ref-type="fig"}: prefrontal cortex, anterior cingulate cortex, and cerebellum) from seven individuals (aged 18--70 y) were studied. The latter two studies were conducted to examine gene expression differences between regions of human brain; the data were not previously analyzed with respect to aging. All three studies used the same microarray platform (Affymetrix HG U95Av2), facilitating comparison between them. Aging Is Heterogeneous within the Human Brain {#s2a} --------------------------------------------- To achieve the most comprehensive picture of brain aging possible with these data, we first sought to study the patterns of aging in all six brain regions from Khaitovich et al. \[[@pbio-0030274-b19]\]. Because only three samples of two ages were available for each brain region in this dataset, only three general aging patterns were possible: up-regulation (the old sample is more highly expressed than either young sample), down-regulation (the old sample is more weakly expressed), or neither (the old sample is in between the young samples). Because thousands of genes would be expected to show each of these three patterns even in the absence of any genuine aging-related changes in gene expression, we were unable to use the three samples on their own to accurately identify genes changing expression with age. However, with the available data we could ask whether the genes whose expression changes with age in frontal pole \[[@pbio-0030274-b07]\] showed the same direction of change in each of six other brain regions \[[@pbio-0030274-b19]\]. In order to do this, we reanalyzed the data of Lu et al. \[[@pbio-0030274-b07]\], and identified 841 genes that showed a significant (p \< 0.01) Spearman rank correlation between age and expression level in frontal pole, irrespective of their fold change in expression; most of these were expected to be true positives, because only approximately 126 genes would be expected to pass this significance threshold by chance (corresponding to an estimated false discovery rate \[[@pbio-0030274-b21]\] of 126/841 = 15.0%). We classified these 841 genes as having either increasing or decreasing expression with age in frontal pole, and then as either increasing, decreasing, or constant in each of the six other brain regions. After discarding genes with no direction of change within each of the six brain regions, because these lack any information about aging changes, we tested how well the frontal pole data agree with the data from each of the six other regions. For example, comparing prefrontal cortex to frontal pole, we asked how many genes belong to each of four categories: (1) up-regulated in frontal pole and down-regulated in prefrontal cortex; (2) down-regulated in frontal pole and up-regulated in prefrontal cortex; (3) up-regulated in both regions; and (4) down-regulated in both regions. If the datasets showed similar aging patterns, we would expect an excess of genes in the latter two categories, whereas no such excess would be expected in the absence of a shared pattern. There are a number of statistical tests that can be used to quantify these patterns; we chose to use the nonparametric Spearman rank correlation coefficient (abbreviated as r). Values of r close to one indicate good agreement between aging patterns, whereas those close to zero indicate a lack of agreement. To assess the significance of these correlations, we randomly permuted the ages of the samples, and calculated the probability of observing a random correlation as strong as that found in the real data (see [Materials and Methods](#s4){ref-type="sec"}). Strikingly, all four regions of cerebral cortex for which we had expression data (prefrontal cortex, Broca\'s area, primary visual cortex, and anterior cingulate cortex) showed excellent agreement with the aging pattern in frontal pole ([Figure 2](#pbio-0030274-g002){ref-type="fig"}A; r \> 0.8 and p \< 0.02 for each). We note that the true similarity of aging patterns in these regions is likely to be even stronger than is indicated by the correlations because, as mentioned above, approximately 15% of our genes are expected to be false positives with no true aging-related changes. In sharp contrast to cortex, the cerebellum and caudate nucleus showed far less agreement with frontal pole ([Figure 2](#pbio-0030274-g002){ref-type="fig"}A; \\|r\\| \< 0.1 and p \> 0.4 for each). These results have several implications. First, the agreement between frontal pole and four regions of cortex indicates that we were able to accurately measure the direction of gene expression changes with age for most genes, even with only three samples from each region; thus the age range, number of samples, etc., are all sufficient to reflect the pattern of gene expression changes previously reported in frontal pole \[[@pbio-0030274-b07]\]. Second, we can have even greater confidence in the results from frontal pole \[[@pbio-0030274-b07]\], because they have been independently reproduced (albeit in different brain regions). Third, and most importantly, the human brain appears to have different aging patterns in cerebellum and caudate nucleus than in cortex. The fact that our four cortex samples all show strong correlations with frontal pole is akin to having a positive control, and it allows us to interpret the lack of correlation in cerebellum and caudate nucleus as evidence suggesting a difference in aging patterns, as opposed to several more trivial explanations (e.g., too few samples). ::: {#pbio-0030274-g002 .fig} Figure 2 ::: {.caption} ###### Aging in the Human Brain The abbreviations used are as follows: ACC, anterior cingulate cortex; BA, Broca\'s area; C, cerebellum; CN, caudate nucleus; PFC, prefrontal cortex; PVC, primary visual cortex. \(A) Correlations of aging gene expression patterns between human frontal pole \[[@pbio-0030274-b07]\] and each of the six regions of the human brain from \[[@pbio-0030274-b19]\] (from left to right, number of genes used are 656, 733, 684, 710, 690, and 603). The strong correlation for all four cerebral cortex samples indicates a reproducible aging pattern across all tested regions of cortex; this pattern does not hold for caudate nucleus or cerebellum. \(B) Correlations of aging gene expression patterns between human prefrontal cortex \[[@pbio-0030274-b20]\] and each of the six regions of the human brain from \[[@pbio-0030274-b19]\] (from left to right, number of genes used are 704, 832, 697, 784, 759, and 674). The strong correlations for all four cortex samples indicates a reproducible aging pattern across all tested regions of cortex but not caudate nucleus or cerebellum, confirming the result of (A). \(C) Correlations of aging gene expression patterns between cerebellum \[[@pbio-0030274-b20]\] and each of the six regions of the brain from \[[@pbio-0030274-b19]\] (from left to right, number of genes used are 213, 241, 204, 241, 244, and 204). The lack of any significant correlation, even when comparing the two cerebellum aging patterns to each other, suggests that human cerebellum lacks a reproducible aging pattern. ::: ![](pbio.0030274.g002) ::: In order to further test the similarity of aging patterns within the brain, we compared a third independent dataset to the data from Lu et al. \[[@pbio-0030274-b07]\] and Khaitovich et al. \[[@pbio-0030274-b19]\]. As described above, Evans et al. \[[@pbio-0030274-b20]\] sampled three brain regions from each of seven individuals. We first tested whether the aging patterns in the two cortex regions from Evans et al. \[[@pbio-0030274-b20]\] correlated more highly with the frontal pole aging changes \[[@pbio-0030274-b07]\] than did the cerebellum samples, as would be expected from [Figure 2](#pbio-0030274-g002){ref-type="fig"}A. Classifying the same 841 genes showing significant change with age in the frontal pole as either up-regulated or down-regulated with age in each brain region of this new dataset, we found the same general pattern of correlations as with the data from Khaitovich et al. \[[@pbio-0030274-b19]\]: Cerebellum showed a weaker correlation with frontal pole than did either cortex sample (prefrontal cortex, r = 0.70; anterior cingulate cortex, r = 0.61; cerebellum, r = 0.38). Although the cerebellum correlation is stronger here than in [Figure 2](#pbio-0030274-g002){ref-type="fig"}A, it is still not significantly different from zero (p = 0.17), even though the two cortex samples are both significant (p \< 0.01 each). This finding supports our conclusion that cerebellum ages differently than cortex. For a third test of aging patterns throughout the human brain, we determined the correlation of aging patterns between a single cortex region from Evans et al. \[[@pbio-0030274-b20]\] with all six of the brain regions from Khaitovich et al. \[[@pbio-0030274-b19]\]. We used the prefrontal cortex samples from Evans et al. \[[@pbio-0030274-b20]\] because, as mentioned above, this brain area shows a better agreement of aging patterns with frontal pole than does anterior cingulate cortex. To facilitate comparison with cerebellum (see below), we extended this analysis to all 12,558 probe sets present on the microarray; however in order to increase the signal/noise ratio, we then excluded genes with no apparent aging changes (age vs. expression \\|r\\| \< 0.5) in either dataset. This comparison showed the expected reproducibility of aging patterns across all four regions of the cortex: r \> 0.76 for all four ([Figure 2](#pbio-0030274-g002){ref-type="fig"}B; p \< 0.03 for each except for prefrontal cortex, for which p = 0.067). In contrast, neither cerebellum nor caudate nucleus showed a significant correlation ([Figure 2](#pbio-0030274-g002){ref-type="fig"}B; \\|r\\| \< 0.04 and p \> 0.4 for each), as expected from their lack of correlation with the frontal pole data shown in [Figure 2](#pbio-0030274-g002){ref-type="fig"}A. In addition to providing further support for our finding of an aging pattern common to all tested regions of cortex, this result demonstrated that even when comparing aging patterns from the two smaller microarray studies used here \[[@pbio-0030274-b19],[@pbio-0030274-b20]\], the age range, number of samples, etc., were sufficient to reveal a correlation when one exists. The lack of correlation between the aging pattern in cerebral cortex with those in cerebellum and caudate nucleus might arise if the quality of data in the cerebellum and caudate nucleus samples was lower than that of the cortex samples from both Khaitovich et al. \[[@pbio-0030274-b19]\] and Evans et al. \[[@pbio-0030274-b20]\], because lower quality of data would lead to weaker correlations. To address this possibility, we first compared the expression levels of the 841 genes used in [Figure 2](#pbio-0030274-g002){ref-type="fig"}A in the two 45-y-olds from Khaitovich et al. \[[@pbio-0030274-b19]\], because their equal age controls for the fact that we expect these genes not to have a very high correlation between sample of different ages (such as between the 45- and 70-y-olds). All six brain regions had highly reproducible expression levels; the lowest correlation among all six was for anterior cingulate cortex, with r = 0.952. The cerebellum data from Evans et al. \[[@pbio-0030274-b20]\] was of similarly high quality: Among five replicates of the same cerebellum samples analyzed in two different laboratories, the lowest correlation of expression levels among all genes was r = 0.964. Thus differing data quality could not explain the lack of correlation in cerebellum and caudate nucleus. Human Cerebellum Ages Less than Cortex {#s2b} -------------------------------------- There are two possible explanations for the difference in the aging patterns between cerebellum/caudate nucleus and cerebral cortex. One is that cerebellum and caudate nucleus have their own aging patterns distinct from that in cortex. The other possibility is that cerebellum and caudate nucleus are different from cortex because they each have far fewer genes changing expression with age than cortex does, and they thus lack a reproducible pattern of aging-associated gene expression changes altogether. To distinguish between these possibilities, one could attempt to calculate exactly how many genes change expression with age in each region; if cerebellum and/or caudate nucleus have aging-related changes in as many (but a different set of) genes than cortex, then the number of genes identified as changing in cerebellum and/or caudate nucleus should be comparable to any region of cortex. Unfortunately, as mentioned above, there is not enough statistical power to pursue this approach, given only three samples per region (or seven, as in Evans et al. \[[@pbio-0030274-b20]\]). Another way to differentiate between the two possibilities listed above would be to compare two datasets of cerebellum and/or caudate nucleus aging patterns to one another. If these regions have a reproducible pattern of many genes changing expression with age (as in the cortex samples of [Figure 2](#pbio-0030274-g002){ref-type="fig"}A and [2](#pbio-0030274-g002){ref-type="fig"}B), we should find a significant correlation. A comparison between the data from Evans et al. \[[@pbio-0030274-b20]\] and Khaitovich et al. \[[@pbio-0030274-b19]\] is suitable for this purpose because both datasets contain cerebellum samples and we already have a positive control that demonstrated our ability to find a correlation between aging patterns in these datasets when one exists ([Figure 2](#pbio-0030274-g002){ref-type="fig"}B). We thus expected to see a strong positive correlation between cerebellum aging patterns in our two datasets if and only if a large number of genes change expression with age in cerebellum. Because this analysis was carried out on all informative genes (age vs. expression \\|r\\| \> 0.5, as in [Figure 2](#pbio-0030274-g002){ref-type="fig"}B), instead of just the 841 with expression changes in the frontal pole, any reproducible changes in cerebellum should be found. Comparison of cerebellum aging from Evans et al. \[[@pbio-0030274-b20]\] with all six regions from Khaitovich et al. \[[@pbio-0030274-b19]\] gave an unambiguous result: Not a single region had a significant correlation ([Figure 2](#pbio-0030274-g002){ref-type="fig"}C; \\|r\\| \< 0.2 and p \> 0.4 for each), including the cerebellum--cerebellum comparison. From these results, we conclude that cerebellum has a different pattern of aging than cortex because significantly fewer genes appear to change expression with age in cerebellum. In order to further characterize the differences in aging patterns between cortex and cerebellum, we calculated the average expression levels in one representative region of cortex (prefrontal cortex) for the 841 genes changing strongly with age in frontal pole, in both young (45-y-old) and old (70-y-old) samples from Khaitovich et al. \[[@pbio-0030274-b19]\]. As expected, when separated into two groups by their direction of change with age, clear differences were seen between the young and old samples ([Figure 3](#pbio-0030274-g003){ref-type="fig"}). When the same genes were subjected to this analysis using their cerebellum expression levels, an interesting trend emerged: Although the genes that are up-regulated in cortex are also slightly up-regulated in cerebellum, those that are down-regulated in cortex show almost no change at all in cerebellum ([Figure 3](#pbio-0030274-g003){ref-type="fig"}). Thus, the difference in aging patterns between these two brain regions arises mainly from genes down-regulated in the cortex. The reason for this may be related to metabolic differences between cerebral cortex and cerebellum ([see Discussion](#s3){ref-type="sec"}). ::: {#pbio-0030274-g003 .fig} Figure 3 ::: {.caption} ###### Expression Levels in Human Cortex and Cerebellum Average expression levels (base two logarithm expression intensity; error bars indicate plus or minus one standard error) in prefrontal cortex were calculated for four sets of genes in both young (two 45-y-old) and old (one 70-y-old) human samples. Red indicates cortex expression levels; blue, cerebellum expression levels; solid lines, genes down-regulated in frontal pole; and dashed lines, genes up-regulated in frontal pole (connecting lines are not meant to imply linear changes in gene expression with age). The genes up-regulated with age in cortex are somewhat up-regulated in cerebellum, whereas those down-regulated in cortex do not change at all with age in cerebellum. ::: ![](pbio.0030274.g003) ::: Chimpanzee Cortex Ages Differently than Human Cortex {#s2c} ---------------------------------------------------- In order to study the relationship between brain aging patterns in humans and chimpanzees, we required gene expression data from the chimpanzee brain. Although four studies have already produced such data, three of these \[[@pbio-0030274-b16],[@pbio-0030274-b22],[@pbio-0030274-b23]\] examined only a single brain area, and the fourth \[[@pbio-0030274-b19]\] had an insufficient number of samples of appropriate age for our purposes. Therefore, we generated new data by measuring gene expression levels in three regions of the chimpanzee brain: prefrontal cortex, anterior cingulate cortex, and cerebellum (see [Materials and Methods](#s4){ref-type="sec"}). We had samples of all five regions from five individuals (aged 7 to approximately 45 y; see [Materials and Methods](#s4){ref-type="sec"}), as well as one additional cerebellum sample and two additional prefrontal cortex samples (although excluding the three extra samples made little difference in the analysis; see [Materials and Methods](#s4){ref-type="sec"}). Because of the very high sequence similarity between humans and chimpanzees \[[@pbio-0030274-b24]\], we were able to use microarrays designed for human sequences. Because we are only comparing chimpanzee samples directly with one another (comparisons with human are using only aging patterns, not actual expression levels), masking of microarray probes containing DNA sequence differences between human and chimpanzee was not necessary (and did not affect the analysis when tested). We compared the aging patterns of different regions within chimpanzee brains by applying the same methods as for comparison between aging patterns of different regions within the human brain. As in [Figure 2](#pbio-0030274-g002){ref-type="fig"}B and [2](#pbio-0030274-g002){ref-type="fig"}C, we used all informative genes (age vs. expression \\|r\\| \> 0.5) present on the microarray. Although we found no significant correlation when comparing cerebellum to either cortex region ([Figure 4](#pbio-0030274-g004){ref-type="fig"}A; \\|r\\| \< 0.07 and p \> 0.3 for both comparisons), we found a very strong agreement when comparing aging patterns of prefrontal cortex to anterior cingulate cortex ([Figure 4](#pbio-0030274-g004){ref-type="fig"}A; r = 0.894, p \< 0.005). This result is precisely analogous to our findings in human, where the entire cerebral cortex shares a single pattern of gene expression changes with age that is not found in the cerebellum (see [Figure 2](#pbio-0030274-g002){ref-type="fig"}A and [2](#pbio-0030274-g002){ref-type="fig"}B). Importantly, this also demonstrates that our chimpanzee samples are of sufficient number, quality, and age range to detect a correlation of aging patterns when one exists. ::: {#pbio-0030274-g004 .fig} Figure 4 ::: {.caption} ###### Aging in the Chimpanzee Brain The abbreviations used are as follows: ACC, anterior cingulate cortex; C, cerebellum; PFC, prefrontal cortex. \(A) Correlations of aging gene expression patterns between all three possible pairs of the three regions of the chimpanzee brain used in this work (from left to right, number of genes used are 1,343, 2,235, and 1,328). The strong correlation when comparing cortex regions indicates a reproducible pattern of aging in chimpanzee cortex. \(B) Correlations of aging gene expression patterns between human frontal pole \[[@pbio-0030274-b07]\] and each of the three regions of the chimpanzee brain used in this work (841 genes used in each comparison). The lack of any significant correlation suggests that human and chimpanzee brain aging patterns differ. ::: ![](pbio.0030274.g004) ::: We then tested whether brain aging in chimpanzee is similar to that of human. Using the 841 genes that change expression with age in human frontal pole \[[@pbio-0030274-b07]\], we tested the agreement between the aging-related changes in frontal pole and the changes in each of our three chimpanzee brain regions. As can be seen in [Figure 4](#pbio-0030274-g004){ref-type="fig"}B, none of the three regions showed any significant correlations with human frontal pole (\\|r\\| \< 0.13, p \> 0.4 for all three). Similar results were found when comparing chimpanzee aging changes in any brain region to the patterns from either of our other two human expression datasets \[[@pbio-0030274-b19],[@pbio-0030274-b20]\] for either the 841 genes or all aging-informative genes on the microarray (not shown). Therefore, we conclude that chimpanzee cortex has a reproducible pattern of aging-associated gene expression changes, but this pattern is completely different from that of human cortex. Alternative explanations such as lower chimpanzee data accuracy, insufficient chimpanzee age range, and very few age-related changes in chimpanzee cortex (as in human cerebellum) can all be eliminated because they would all preclude the strong correlation between aging patterns of the two chimpanzee cortex regions shown in [Figure 4](#pbio-0030274-g004){ref-type="fig"}A (other possible artifactual explanations for the difference are discussed in [Materials and Methods](#s4){ref-type="sec"}). Given this difference in aging patterns between humans and chimpanzees, we examined the expression levels of the chimpanzee orthologs of the 841 human genes that change expression with age in frontal pole in order to see if the expression levels of the chimpanzee orthologs of these genes resemble young humans, old humans, or neither. To test this, we first reanalyzed the expression data by masking all microarray probes with sequence differences between humans and chimpanzees \[[@pbio-0030274-b19]\]. We then calculated, for both human and chimpanzee prefrontal cortex, the average expression level for the set of genes that increase expression with age in frontal pole, as well as the average for the genes that decrease expression with age. The result is that in chimpanzee cortex, the orthologs of both sets of human genes (up-regulated and down-regulated) are expressed at the levels of their young human counterparts ([Figure 5](#pbio-0030274-g005){ref-type="fig"}). In other words, chimpanzee cortex expression levels strongly resemble expression levels in young but not old humans, at least among the set of genes tested here. Humans then diverge from these average expression levels as they age, whereas chimpanzee gene expression levels change in an almost entirely different set of genes. ::: {#pbio-0030274-g005 .fig} Figure 5 ::: {.caption} ###### Expression Levels in Human and Chimpanzee Cortex Average expression levels (base two logarithm expression intensity; error bars indicate plus or minus one standard error) in prefrontal cortex were calculated for four sets of genes in both young (two 45-y-old human, or five 7- to 12-y-old chimpanzee) and old (one 70-y-old human, or two older than 40-y-old chimpanzee) samples. Red indicates human genes; blue, chimpanzee genes; solid lines, genes (or orthologs of genes) down-regulated in human frontal pole; and dashed lines, genes (or orthologs) up-regulated in human frontal pole (connecting lines are not meant to imply linear changes in gene expression with age). The chimpanzee expression levels resemble young, but not old, human. ::: ![](pbio.0030274.g005) ::: The high correlation of gene expression aging patterns between the two regions of chimpanzee cortex implied that the genes for which both regions show the same direction of change (that passed our cutoff of age vs. expression \\|r\\| \> 0.5) are nearly all genuinely up- or down-regulated with age. Using this list of genes with a consistent aging pattern in the two cortex regions, there were 1,252 down-regulated and 700 up-regulated genes. Note that although the false-positive rate is likely to be low, we have no way to estimate the false-negative rate, so these numbers should not be interpreted as the total number of genes changing expression with age in chimpanzee cortex. Using this list of aging-associated genes, we tested for any significant enrichments of these genes in Gene Ontology annotation categories \[[@pbio-0030274-b19]\]. We did not find any enrichments for the set of genes down-regulated with age, although we found a number of significant enrichments for those up-regulated with age, including mitochondrial localization, protein degradation functions, and several metabolic processes (see [Table S1](#st001){ref-type="supplementary-material"}). Interestingly, and consistent with our finding of no similarity between human and chimpanzee aging patterns, there was little overlap between these enriched groups and those previously reported for human frontal pole \[[@pbio-0030274-b07]\]. Discussion {#s3} ========== In this study, we have made three main observations: First, aging-related gene expression changes are similar throughout all five tested regions of the human cerebral cortex. Second, this pattern of human cortex aging is not found in cerebellum or caudate nucleus, and at least in cerebellum this appears to be due to far fewer genes changing expression with age. Third, although chimpanzee cortex has a reproducible pattern of expression changes with age, it shares no detectable similarity with the aging pattern in human cortex. These conclusions raise a number of questions. For example, why does human cerebellum age differently than human cortex? We have shown that the majority of the difference is because genes down-regulated with age in cortex are not down-regulated in cerebellum (see [Figure 3](#pbio-0030274-g003){ref-type="fig"}); because fewer genes change with age in cerebellum than in cortex (see [Figure 2](#pbio-0030274-g002){ref-type="fig"}), it therefore follows that this is mostly due to fewer genes being down-regulated with age. What could cause fewer genes to have reduced transcription over time in cerebellum than in cerebral cortex? Cerebellum differs in many important respects from cortex; in particular, it has a lower metabolic rate than cortex in both human and rhesus macaque, regardless of age \[[@pbio-0030274-b25]--[@pbio-0030274-b27]\]. These observations of lower metabolic activity in cerebellum imply that if a consequence of aerobic respiration is ROS-induced DNA damage, then such damage should be greater in cortex than in cerebellum. Indeed, it has been shown that cerebellum has far fewer mtDNA (mitochondrial DNA) deletions than cortex, especially in old humans \[[@pbio-0030274-b11]\], and it accumulates less oxidative damage to both mtDNA and nuclear DNA than does cortex \[[@pbio-0030274-b12]\]. Therefore if the accumulation of DNA damage causes gene expression down-regulation \[[@pbio-0030274-b07]--[@pbio-0030274-b10],[@pbio-0030274-b28]\], then we would expect to see fewer aging-related gene expression level reductions in cerebellum than in cortex. Our confirmation of this prediction is quite consistent with the theory that ROS-induced damage is responsible for gene expression changes \[[@pbio-0030274-b07],[@pbio-0030274-b28]\], as well as the more general oxidative free radical theory of aging \[[@pbio-0030274-b02]\]. Similarly, one might ask why chimpanzee cortex ages differently than human cortex. If ROS-induced DNA damage is indeed a major cause of gene expression changes \[[@pbio-0030274-b07],[@pbio-0030274-b28]\], then the aging differences could be due to differential ROS susceptibilities of orthologous loci in the human and chimpanzee genomes (although ROS damage is unlikely to directly explain the difference in up-regulated genes as seen in [Figure 5](#pbio-0030274-g005){ref-type="fig"}, it may be indirectly responsible by down-regulating genes such as transcriptional repressors). It is difficult to know how plausible this is because we do not presently understand what factors lead to ROS damage susceptibility; regardless of the factors involved; however, it is quite possible that even the relatively few genetic (or epigenetic) differences between human and chimpanzee may be sufficient to cause drastic changes in ROS susceptibility, as is the case for other chromosomal properties such as DNA methylation \[[@pbio-0030274-b29]\] and recombination rate \[[@pbio-0030274-b30],[@pbio-0030274-b31]\]. One possible explanation for ROS susceptibility is that promoters driving high levels of transcription are more vulnerable to ROS, perhaps because of their more accessible chromatin structure and/or lower tolerance for oxidative damage; however, although highly expressed genes are indeed more likely to be down-regulated with age in human frontal pole, expression levels are far from explaining all of the variation in aging-related changes in either human or chimpanzee (not shown). It is likely that ROS susceptibility will have to be measured in a number of chimpanzee gene promoters, as has been done for human \[[@pbio-0030274-b07]\], in order to discover if differential oxidative damage can explain the human--chimpanzee divergence in aging patterns. Another implication of these results is related to the use of model organisms such as mouse, rat, and various primates as surrogates for human brain aging and neurodegeneration. The fact that even the chimpanzee, our closest living relative, has patterns of age-related gene expression changes almost entirely different than human implies that making specific inferences about human brain aging from model organisms may be difficult. This conclusion is supported by a study of brain aging in mice, in which in contrast to the results reported here for human, the cerebellum was found to contain more genes changing expression with age than cortex \[[@pbio-0030274-b32]\] (a difference that may be due to different relative metabolic rates of cortex and cerebellum in mouse compared to human). Model organisms are probably well suited for studying the mechanisms of aging (such as ROS-induced damage), which are likely to be conserved over great phylogenetic distances, but such conserved mechanisms may have species-specific outcomes at the level of individual genes. Thus, caution is warranted when trying to extrapolate the results of neurodegeneration research from model organisms to humans. Many other questions raised by this work are still unresolved. First, how diverse are aging patterns of gene expression change in human tissues outside the brain? A recent study finding similar aging profiles of human kidney cortex and medulla regions implies that the intra-organ variability in aging patterns observed in the present work may not be found in all organs \[[@pbio-0030274-b33]\]. Second, do human and chimpanzee differ in their aging patterns in tissues other than brain, or is the brain a special case because of its recent rapid morphological evolution in the human lineage? It will be interesting to test this for tissues that have not undergone any obvious rapid evolution (such as liver or kidney), as well as for tissues that are likely to have been under strong positive selection (such as testes). Third, does chimpanzee cerebellum have fewer gene expression changes with age than cortex, as is the case in human? More chimpanzee data will be needed to address this question, although it seems likely that the answer will be affirmative because the greater metabolic rate of cortex compared to cerebellum is conserved to rhesus macaque \[[@pbio-0030274-b27]\]. Fourth, does the human or chimpanzee cortex aging pattern represent the ancestral state of this pattern for these two species, or are they both highly diverged from that state? Examination of brain aging patterns in an outgroup species, such as rhesus macaque, may help to resolve this question. Fifth, is the rapid divergence of aging patterns along the human and/or chimpanzee lineage the result of selection on the aging process itself, or is the divergence an indirect consequence of selection on other aspects of the brain, or could it even be explained by random drift alone? And finally, can we use our understanding of the similarities and differences in brain aging of humans and chimpanzees to gain greater insight into the causes of, and possible treatments for, human neurodegeneration? We believe this will be possible because investigation of how a phenomenon such as neurodegeneration emerged during evolution might well point us towards its underlying causes Materials and Methods {#s4} ===================== Tissue samples and gene expression data {#s4a} --------------------------------------- Human microarray data was obtained from Pritzker Neuropsychiatric Disorders Research Consortium (<http://www.pritzkerneuropsych.org>) and the National Institute of Mental Health (NIMH) Silvio O. Conte Center, and from ArrayExpress and GEO databases. Only the seven "Type 1" control individuals \[[@pbio-0030274-b34],[@pbio-0030274-b35]\] were used from the Evans et al. \[[@pbio-0030274-b20]\] dataset. Chimpanzee postmortem samples were obtained from Yerkes Regional Primate Center, Biomedical Primate Research Centre and Anthropologisches Institut und Museum, Universität Zürich. All individuals suffered sudden death for reasons other than their participation in this study and without any relation to the tissues used; time between death and preservation of brain tissue (postmortem interval) did not correlate with expression of genes that changes with age. Age and sex for all individuals are listed in [Table S2](#st002){ref-type="supplementary-material"}. Total RNA was isolated from approximately 50 mg of frozen tissue using the TRIZol reagent according to manufacturer\'s instructions and purified with QIAGEN RNeasy kit (Qiagen, Valencia, California, United States) following the "RNA cleanup" protocol. RNAs were of high and comparable quality in all samples as gauged by the ratio of 28S to 18S ribosomal RNAs estimated using the Agilent 2100 Bionalyzer system (Agilent, Palo Alto, California, United States) and by the signal ratios between the probes for the 3′ and 5′ ends of the mRNAs of GAPDH and β-actin genes used as quality controls on Affymetrix (Affymetrix, Santa Clara, California, United States) microarrays. Labeling of 1.2 μg of total RNA, hybridization to Affymetrix HG U95v2 arrays, staining, washing, and array scanning were carried out following Affymetrix protocols. The samples were processed in random order with respect to age. All primary expression data generated for this study are publicly available at ArrayExpress database (<http://www.ebi.ac.uk/arrayexpress/>). Data were normalized using the Robust Multichip Average (RMA) method \[[@pbio-0030274-b36]\]. Among the chimpanzees used for this work, one was of indeterminate age, having been caught in the wild 40 y before its death ([Table S2](#st002){ref-type="supplementary-material"}). However, because we used nonparametric rank statistics for all analyses, the exact age was irrelevant; all that mattered was whether it was older or younger than our 44-y-old chimpanzee. For the results shown it was assumed to be older, although assuming it to be younger than 44 y strengthened the results of the analyses (the correlation of chimpanzee prefrontal cortex with anterior cingulate cortex increased from 0.894 to 0.927). Additionally, one pair of chimpanzees used were full siblings, and another pair were half siblings ([Table S2](#st002){ref-type="supplementary-material"}), which could be problematic if aging patterns are family-specific. However one member of each related pair could be excluded from the analysis without greatly affecting the results (correlation of chimpanzee prefrontal cortex with anterior cingulate cortex decreased from 0.894 to 0.858), indicating that relatedness among chimpanzees does not alter our conclusions. Excluding one member of each related pair also left the same five unrelated chimpanzees for each of the three brain regions tested, controlling for any possible effects of unequal sample sizes from each brain region. There are several possible artifactual explanations for our results not addressed in the main text. First, there is the possibility that gene expression changes correlated with age were caused by an unknown factor unrelated to the normal aging process. For example, if in the study by Khaitovich et al. \[[@pbio-0030274-b19]\] the 70-y-old had a disease that made his cerebellum and caudate nucleus appear "young" in their gene expression, but did not affect his cortex (because all four of his cortex regions showed the same reproducible pattern), then this could account for the results of [Figure 2](#pbio-0030274-g002){ref-type="fig"}A. However in order for this explanation to also account for the similar results of [Figure 2](#pbio-0030274-g002){ref-type="fig"}B, several elderly individuals from the Evans et al. \[[@pbio-0030274-b20]\] study would all have to be similarly afflicted in their cerebella as well. We found this to be extremely unlikely, because all ten individuals from these two studies were chosen in part for their lack of any known brain-related diseases \[[@pbio-0030274-b19],[@pbio-0030274-b20]\]. Similarly, one possible explanation for the results in [Figure 4](#pbio-0030274-g004){ref-type="fig"}A is that the cortexes (but not cerebella) of both of our old chimpanzees had a large number of gene expression differences compared to the young chimpanzees for a reason other than aging, such as a cortex-specific brain disease distinct from the normal aging process. As for the human subjects discussed above, this is extremely unlikely to be the case, because none of these chimpanzees had any apparent brain disease and both old chimpanzees (who are unrelated to each other) would have to be similarly afflicted to observe this effect. If gene expression changes in primates tend to be dependent on chronological age (so that, for example, gene expression levels in a 45-y-old chimpanzee are most similar to those in a 45-y-old human), then the different patterns of expression changes with age seen in humans and chimpanzees could be caused by the different age ranges of the human (18--106 y) and chimpanzee (7 to approximately 45 y) samples. To control for this possibility, we truncated the data of Lu et al. \[[@pbio-0030274-b07]\] to contain only the 11 individuals with age ≤ 45 y. Recalculating the age-expression correlations for each gene and comparing them to the chimpanzee data, we found no more similarity in which genes change expression with age than when using any of our three full human expression datasets \[[@pbio-0030274-b07],[@pbio-0030274-b19],[@pbio-0030274-b20]\]. Therefore, we find it unlikely that brain aging of chimpanzees is any more similar to that of humans when controlling for chronological age. Another possible explanation for the difference between human and chimpanzee aging patterns ([Figure 4](#pbio-0030274-g004){ref-type="fig"}B) is that the difference is actually due to the different environments experienced by the humans and chimpanzees during their lifetimes, and is not due to any intrinsic differences between these species. Because there is no way to possibly control for this---for both practical and ethical reasons a human cannot be raised in precisely the same environment as a chimpanzee (or even another human)---all that we can rigorously conclude is that the humans and chimpanzees used for the analyses herein did experience different patterns of gene expression change with age. We note that this general concern extends to all studies comparing any human\'s phenotype with that of another organism. One caveat concerning our interpretation of fewer genes having aging-associated changes in expression in cerebellum than in cortex is that the two human cerebellum datasets, although both consisting of grey matter of the cerebellum, were from different regions of the cerebellum (Khaitovich et al. \[[@pbio-0030274-b19]\] sampled the Vermis cerebelli, whereas Evans et al. \[[@pbio-0030274-b20]\] used the left lateral portion of the cerebellum); therefore it is technically possible that these regions of cerebellum each have their own reproducible aging pattern, which (if both shared no similarity either to each other or to cortex) would not be revealed by this analysis. We find this to be quite unlikely, given the very close proximity and functionally similar properties of these two regions, together with the finding that far more heterogeneous regions throughout the cerebral cortex share nearly identical aging patterns. And even if this improbable case were to be true, our conclusion of cerebellum grey matter as a whole lacking any reproducible aging pattern would still hold. Statistics {#s4b} ---------- All correlation coefficients reported here were calculated by Spearman rank correlation, a nonparametric method that is robust to the presence of any outliers. The correlation coefficients from comparisons of aging profiles between two tissues or species (as in [Figures 2](#pbio-0030274-g002){ref-type="fig"} and [4](#pbio-0030274-g004){ref-type="fig"}) were calculated on coordinates assigned to genes in each of the following categories: up-regulated in both regions (1,1), up-regulated in one region and down in another (\[1,−1\] or \[−1,1\]), or down-regulated in both regions (−1,−1). The correlation coefficients can be interpreted as scores directly proportional to the fraction of genes with the same direction of expression change with age. Probability values were calculated by randomization of ages, given the specific genes used in any particular comparison: the fraction of randomizations with a correlation coefficient greater than or equal to the observed value is the p-value given. Therefore this is a one-sided test, appropriate for the question of whether we could have agreement between aging patterns as strong as those observed, just by random chance. The only exceptions to our using this one-sided test were when we stated we were testing whether the correlation coefficient was significantly different from zero (as opposed to greater than zero); in these cases the test was two-sided. This randomization test is somewhat conservative; for example, analyzing the leftmost bar (prefrontal cortex) of [Figure 2](#pbio-0030274-g002){ref-type="fig"}A with Fisher\'s exact test yields a p-value of approximately 10^−99^, as opposed to approximately 0.015 from randomization. This large difference is due to the nonrandom structure of the expression data, which makes it more likely to observe strong correlations than would be expected in a set of random data. Finally, we note that our finding of significant agreement between aging patterns in datasets containing various numbers of microarrays does not imply that these numbers of microarrays will always yield sufficient power to find such a correlation if one exists. Supporting Information {#s5} ====================== Table S1 ::: {.caption} ###### Gene Ontology Groups Enriched in the List of Genes Up-Regulated with Age in Chimpanzee Cerebral Cortex "All detected" is the number of genes on the microarray that had an annotation in each of the three Gene Ontology (GO) categories. "Detected/group" is the number of genes on the microarray that belong to the specific GO group listed. "All selected" is the number of genes up-regulated with age in chimpanzee cortex that have annotations in each of the three GO categories. "Selected/group" is the number of genes up-regulated with age in chimpanzee cortex that belong to the specific GO group listed. The uncorrected p-value cutoff for each GO group is 0.005. (21 KB XLS). ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### Sample Information (18 KB XLS). ::: ::: {.caption} ###### Click here for additional data file. ::: We would like to thank the Pritzker Neuropsychiatric Disorders Research Consortium, the National Institute of Mental Health (NIMH) Silvio O. Conte Center, and J. Li personally for sharing the data from Evans et al. \[[@pbio-0030274-b20]\]; I. Hellmann for providing the sequence mask file for chimpanzee microarray data analysis; W. Enard and A. Chen for helpful discussions; and the Bundesministerium für Bildung und Forschung for financial support. JBP acknowledges support from the Burroughs Wellcome Fund and the William F. Milton Fund. MBE is a Pew Scholar in the Biomedical Sciences. HBF is an NSF predoctoral fellow. Competing interests. The authors have declared that no competing interests exist. Author contributions. HBF, PK, and JBP conceived and designed the experiments and analyses. PK performed the experiments. HBF analyzed the data. SP and MBE contributed reagents/materials/analysis tools. HBF, PK, JBP, SP, and MBE wrote the paper. Citation: Fraser HB, Khaitovich P, Plotkin JB, Pääbo S, Eisen MB (2005) Aging and gene expression in the primate brain PLoS Biol 3(9): e274. GO : Gene Ontology ROS : reactive oxygen species	PubMed Central	2024-06-05T03:55:59.743649	2005-8-2	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181540/", "journal": "PLoS Biol. 2005 Sep 2; 3(9):e274", "authors": [ { "first": "Hunter B", "last": "Fraser" }, { "first": "Philipp", "last": "Khaitovich" }, { "first": "Joshua B", "last": "Plotkin" }, { "first": "Svante", "last": "Pääbo" }, { "first": "Michael B", "last": "Eisen" } ] }
PMC1181541	The debilitating effects of Parkinson disease are well known: muscle rigidity, impaired movement, and the uncontrollable shaking that makes even the most mundane activity a challenge. The symptoms result from a progressive deterioration of the nervous system and degeneration of neurons that control motor function and coordinated movement. These neurons, found in the midbrain, produce dopamine, a chemical that transmits brain signals (called a neurotransmitter) that regulate motor control, energy levels, and motivation to achieve goals. In Parkinson disease, dopamine levels drop as neurons degenerate, producing the characteristic symptoms. With no cure on the horizon, the most common treatment involves administration of the dopamine precursor, L-DOPA, usually in pill form. Though symptoms subside at first, this treatment is rendered ineffective over time. To develop better therapies, scientists use animal models of human diseases to test the efficacy and toxicity of potential treatments before moving on to human trials. To be effective, animal models must recapitulate the disease symptoms as closely as possible. However, while previous models of Parkinson with incomplete dopamine depletion captured some of the manifestations of this condition, the dominance of the dopamine system on the control of locomotion has made it difficult to elucidate the contribution of other neurotransmitter systems. In a new study, Tatyana Sotnikova and colleagues from Duke University created such a model that recapitulates many of the symptoms of Parkinson. By eliminating the dopamine transporter---the protein responsible for recycling the chemical into neurons---in mice, the authors reduced dopamine levels in the midbrain by 20-fold. In addition, chemically inhibiting dopamine production in these mice resulted in essentially unmeasurable levels of the neurotransmitter, since it could now neither be produced at normal levels nor be recycled. Because these mice exhibited the symptoms of Parkinson disease remarkably well, the authors could test how well drugs that act independently of dopamine ameliorated symptoms of the disease. This approach allows the identification of drugs that may serve to improve treatment at later stages of the disease, when dopamine-producing neurons have been severely reduced in number and L-DOPA efficacy has been reduced. [](#pbio-0030303-g001){ref-type="fig"} ::: {#pbio-0030303-g001 .fig} ::: {.caption} ###### To model Parkinson disease, researchers bred mice with severe dopamine deficiencies that displayed rigidity, inhibited motion, and, as seen here, freezing behavior ::: ![](pbio.0030303.g001) ::: The authors tested a number of drugs at various doses and found that in addition to L-DOPA-related treatments, drugs related to amphetamine were effective in ameliorating muscle rigidity, tremor, and impaired movement. Most effective was methylenedioxymethamphetamine HCl (MDMA), commonly known as ecstasy. It has been shown that amphetamines can trigger release of neurotransmitters such as dopamine, serotonin, and norepinephrine and cause sudden bursts in neurotransmission, leading to a feeling of alertness, increased muscular activity, and reduced fatigue. This study, however, shows that treating mice with MDMA does not increase dopamine levels; furthermore, treating the mice with drugs related to serotonin or norepinephrine did not ameliorate the disease\'s symptoms. These results suggest that MDMA likely acts through a pathway unrelated to these common neurotransmitters. The authors tested the possibility that MDMA may be increasing transmission via receptors that respond to compounds that are normally present at very low levels, called trace amines. Activation of these receptors reduced rigidity and akinesia, as with MDMA, though to a much lower level. Thus, while it is possible that MDMA is acting through trace amine receptors, this may not be the only pathway used. Future studies will be required to elucidate how MDMA ameliorates Parkinson symptoms. The largest effects of MDMA on reducing symptoms was seen at levels that produce neurotoxic effects in wild-type mice, though administering non-neurotoxic doses of MDMA along with tiny amounts of L-DOPA that are normally ineffective proved as effective as higher doses of MDMA alone. Interestingly, the authors report an absence of significant side effects of even very high doses of MDMA administration on mice lacking the dopamine transporter; however, since patients with Parkinson disease do not necessarily lack the dopamine transporter, toxicity of MDMA and related compounds will need to be studied in greater detail in the future. This study opens the door to a search for compounds related to ecstasy, which may provide a more effective treatment for symptoms of Parkinson in the later stages of the disease---and hopefully allow patients to perform the simple functions of everyday life independently again.	PubMed Central	2024-06-05T03:55:59.748042	2005-8-2	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181541/", "journal": "PLoS Biol. 2005 Aug 2; 3(8):e303", "authors": [] }
PMC1181542	No matter how healthy a life one leads, no person has managed to live much longer than a century. Even though the advances of the modern age may have extended the average human life span, it is clear there are genetic limits to longevity. One prominent theory of aging lays the blame on the accumulation of damage done to DNA and proteins by "free radicals," highly reactive molecules produced by the metabolic activity of mitochondria. This damage is expected to reduce gene expression by damaging the DNA in which genes are encoded, and so the theory predicts that the most metabolically active tissues should show the greatest age-related reduction in gene expression. In this issue, Michael Eisen and colleagues show that the human brain follows this pattern. A similar pattern---which, surprisingly, involves different genes---is found in the brain of the aging chimpanzee. The authors compared results from three separate studies of age-related gene expression, each done on the same type of DNA microarray and each comparing brain regions in young versus old adult humans. In four different regions of the cortex (the brain region responsible for higher functions such as thinking), they found a similar pattern of age-related change, characterized by changes in expression of hundreds of genes. In contrast, expression in one non-cortical region, the cerebellum (whose principal functions include movement), was largely unchanged with age. In addition to confirming a prediction of the free-radical theory of aging (namely, that the more metabolically active cortex should have a greater reduction in gene activity), this is the first demonstration that age-related gene expression patterns can differ in different cells of a single organism. The authors found a similar difference in age-related patterns in the brain of the chimpanzee, with many genes down-regulated in the cortex that remained unchanged in the cerebellum. However, the set of affected cortical genes was entirely different between humans and chimps, whose lineages diverged about 5 million years ago. The explanation for this difference is unknown, but the finding highlights the fact that significant changes in gene expression patterns, and thus changes in many effects of the aging process, can accumulate over relatively short stretches of evolutionary time. [](#pbio-0030313-g001){ref-type="fig"} ::: {#pbio-0030313-g001 .fig} ::: {.caption} ###### Gene expression data from three microarray studies of primate brains were used to identify genetic changes associated with aging in humans and chimps. Different brain regions in both primates undergo distinct age-related changes in gene expression ::: ![](pbio.0030313.g001) ::: These results raise a number of questions about age-related gene expression changes, including whether metabolically active non-brain tissues display similar patterns of changes, and whether the divergence between human and chimp patterns was the direct result of selection, or was an inevitable consequence of some other difference in brain evolution. The patterns seen in this study also provide a starting point for understanding the network of genetic changes in aging, and may even reveal targets for treatment of neurodegenerative diseases.	PubMed Central	2024-06-05T03:55:59.748941	2005-8-2	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181542/", "journal": "PLoS Biol. 2005 Sep 2; 3(9):e313", "authors": [] }
PMC1181621	Depression in OSA ================= Definition and prevalence of OSA -------------------------------- OSA is by far the most common form of sleep disordered breathing and is defined by frequent episodes of obstructed breathing during sleep. Specifically, it is characterized by sleep-related decreases (hypopneas) or pauses (apneas) in respiration. An obstructive apnea is defined as at least 10 seconds interruption of oronasal airflow, corresponding to a complete obstruction of the upper airways, despite continuous chest and abdominal movements, and associated with a decrease in oxygen saturation and/or arousals from sleep. An obstructive hypopnea is defined as at least 10 seconds of partial obstruction of the upper airways, resulting in an at least 50% decrease in oronasal airflow. Clinically OSA is suspected when a patient presents with both snoring and excessive daytime sleepiness (EDS) \[[@B1],[@B2]\]. The diagnosis of OSA is confirmed when a polysomnography recording determines an Apnea-Hypopnea-Index (AHI) of \> 5 per hour of sleep \[[@B3]\]. Even if cutoff points have never been clearly defined, an AHI of less than 5 is generally considered being normal, 5--15 mild, 15--30 moderate and over 30 severe OSA. The prevalence of OSA is higher in men than in women. OSA is found in all age groups but its prevalence increases with age. In children, the prevalence of OSA is less well defined and has been estimated to be 2--8% \[[@B4]\]. In subjects between the ages of 30 to 65 years, 24% of men and 9% of women had OSA \[[@B5]\]. Among subjects over 55 years of age, 30--60% fulfil the criterion of an AHI \> 5 \[[@B6]-[@B8]\]. In a population of community-dwelling older adults, 70% of men and 56% of women between the ages of 65 to 99 years have evidence of OSA with a criterion of AHI \> 10 \[[@B9]\]. The abnormal respiratory events which are the hallmark of OSA are generally accompanied by heart rate variability and arousals from sleep, with frequent arousals being the most important factor resulting in EDS. With regards to sleep architecture, we find a significant increase in light sleep stage (mainly stage 1) at the expense of deep slow wave sleep (stages 3 and 4) and REM sleep. Slow wave sleep is sometimes even completely abolished. However clinically, patients are often not aware of this repetitive sleep interruption (with sometimes hundreds of arousals during one night), but simply do not feel restored in the morning. Other nocturnal symptoms can include restlessness, nocturia, excessive salivation and sweating, gastroesophageal reflux, as well as headache and dry mouth or throat in the morning on awakening. The extent to which daytime functioning is affected generally depends on the severity of OSA. Symptoms other than EDS which greatly impact daytime functioning are neuropsychological symptoms such as irritability, difficulty concentrating, cognitive impairment, depressive symptoms, and other psychological disturbances. Thus, OSA can easily mimic symptoms of a major depressive episode. Correlation studies of OSA and depression ----------------------------------------- Among the first studies investigating the relation between OSA and depression, Guilleminault et al. \[[@B10]\] reported that 24% of 25 male patients with OSA had previously seen a psychiatrist for anxiety or depression, and Reynolds et al. \[[@B11]\] showed that around 40% of 25 male OSA patients met the research diagnostic criteria for an affective disorder, with a higher risk of depression in those patients who were sleepier during the day. Similarly, Millmann et al. observed that 45% of his 55 OSA patients had depressive symptoms on the Zung Self-Rating Depression Scale, with the group scoring higher for depression also having a significantly higher AHI \[[@B12]\]. Whereas only 26% of OSA patients described themselves as currently depressed, 58% fulfilled DSM-III criteria for major depression of four or more depressive symptoms \[[@B13]\]. Others observed increased depression scores on the Minnesota Multiphasic Personality Inventory (MMPI) in patients with OSA \[[@B14],[@B15]\]. Indeed, Ramos Platon et al. found elevations in several MMPI scales in 23 OSA patients (moderate to high severity) compared to 17 controls \[[@B16]\]. Aikens et al. \[[@B17]\] showed that 32% of their OSA patients had elevated depression scores on the MMPI and in the same series of studies, there were twice as many OSA patients with elevated depression scores than age and sex matched primary snorers \[[@B18]\]. However, the percentage of depressive symptoms was not significantly different when compared to patients with other primary sleep disorders, such as periodic limb movements during sleep (PLMS) \[[@B19]\]. Most recently, in an epidemiological study of 18,980 subjects representative of the general population in their respective countries (UK, Germany, Italy, Portugal, and Spain) and assessed by cross-sectional telephone survey, Ohayon determined that 17.6% of subjects with a DSM-IV breathing-related sleep disorder diagnosis also presented with a major depressive disorder diagnosis, and vice versa \[[@B20]\]. This correlation persisted after controlling for obesity and hypertension. In contrast to the numerous studies observing a positive correlation between OSA and depression, some investigations found no association between both disorders. In a 5-year longitudinal study, Phillips et al. did not find any significant depressive symptoms in elderly patients with a relatively mild OSA (AHI\>5/h), when compared to a control group without OSA (AHI\<5/h) \[[@B21]\]. However, there are multiple limitations to this study, besides a relatively small sample size for group comparisons and a non-representative study population. OSA was only assessed at baseline, but not repeated at the five-year follow-up, i.e. neuropsychological data were compared between two groups based on OSA status five years earlier. Second, OSA severity was mild even in the OSA group. Third, the groups differed significantly by age, with the OSA group being older than the control group. Finally, the attrition rate over the five years was very high with only 42 out of the initial 95 subjects completing the follow-up assessment. In another large-scale study, Pillar and Lavie did not observe any association between respiratory disturbances and Symptom Check List 90 in 2,271 predominantly male patients assessed for OSA \[[@B22]\]. However, the SCL-90 questionnaire was developed as a screening tool for psychiatric patients, and not for a normal study population. Therefore, it might be a less sensitive tool with regards to milder forms of mood disturbances than other scales. Interestingly, Pillar and Lavie observed that among the minority of women in this study, those with severe OSA had higher depression scores than those with mild OSA. Bardwell found that other factors such as age, body mass index (BMI) and hypertension accounted for the correlation between sleep parameters and total mood disturbances in 72 OSA patients when compared to 40 controls \[[@B23]\]. However, the chosen cutoff point to distinguish between OSA and the control group in this study was relatively high (AHI of 15/h), thus subjects with a mild OSA were probably included in the control group. In sum, the majority of studies to date report an association between depression and OSA, but methodological considerations render the comparison between investigations difficult. Some of the mixed findings among studies can be explained by differences in sample size, study population, gender distribution, age and AHI cut-off in relation to age, as well as variability in terms of the questionnaires and scales used to assess depressive symptomatology. Given the heterogeneity of these data and considering the numerous confounding factors, future longitudinal studies of patient populations are required to better understand the relation between both disorders. Treatment Studies for OSA: reversibility of depressive symptoms? ---------------------------------------------------------------- The gold standard treatment for moderate to severe cases of OSAS is continuous or bilevel positive airway pressure (CPAP/BiPAP) which mechanically maintains the upper airways space open during sleep via the administration of ambient air with a certain pressure. The minimum necessary pressure level has to be titrated individually for each patient \[[@B24]\]. Other treatments, especially for mild cases of OSA, include weight loss, dental devices (which advance the tongue or mandible to increase posterior airway space) or upper airway surgery (e.g. combined tonsillectomy/ adenoidectomy, nasal reconstruction, and uvulopalatopharyngoplasty). Different upper airway surgical procedures can be used for particular cases with craniofacial abnormalities \[[@B25]\]. Overall, CPAP treatment studies for OSA and its effect on depressive symptoms have yielded controversial findings. Derderian et al. \[[@B26]\] compared results on the Profile of Moods Questionnaire before and after 2 months of CPAP treatment in an OSA group (n = 7) and showed a significant drop in Total Mood Disturbance. This improvement was correlated with an increase in slow-wave sleep. Those patients in the study of Millmann et al. who received CPAP displayed a significant decrease in their Zung Depression Scale scores \[[@B12]\]. Similarly, Engleman et al. reported an improvement in a comprehensive battery of mood and cognitive assessment scales after 4 weeks of CPAP treatment in 32 patients with moderate OSA \[[@B27]\] as well as in 16 patients with a mild OSA \[[@B28]\]. Means et al. \[[@B29]\] showed an improvement on Beck Depression Inventory (BDI) depression scores after 3 months of treatment in 39 OSA patients, and Sanchez et al. \[[@B30]\] confirmed lower BDI scores after 1 and 3 months of CPAP therapy in 51 OSA patients. Ramos Platon et al. \[[@B16]\] underscored the progressive improvement in depression scores on the MMPI scale over the first year of treatment. A systematic review on the influence of CPAP on neurobehavioral performance of patients with OSA also supported the clinical perspective that typically depressive symptoms remit together with EDS under CPAP therapy \[[@B31]\]. Among the negative studies on CPAP therapy and its effect on depression, Borak et al. \[[@B32]\] did not observe any improvement in emotional status after 3 and 12 months of CPAP therapy in 20 patients with severe OSA, similar to Munoz et al. \[[@B33]\] who also did not show improvement of BDI scores in 80 subjects with severe OSA after 12 months of CPAP. Using subtherapeutic CPAP as the placebo control, Yu et al. \[[@B34]\] and Henke et al. \[[@B35]\] found no difference in improvement on depression scores between the treatment and the control group, over a short treatment duration (1--3 weeks). However, whereas Borak, Munoz and Henke do not find any effect of CPAP therapy on mood, Yu observed a positive effect on mood of both CPAP therapy and the subtherapeutic CPAP control group. Intriguingly, there are no systematic differences with regards to the sample size, the initial severity of OSA or the duration of CPAP therapy which might explain the differences between studies observing an improvement after CPAP therapy and those who did not. Several issues have to be considered: First, it is difficult to design a good control (\"placebo\") condition for CPAP treatment. \"Sham-CPAP\" which uses insufficient positive airway pressure as a placebo condition (1 -- 2 cm H~2~0), is now used more frequently. Two of the negative studies employed this method for their control group, which raises the possibility that the previously observed positive effects of CPAP on mood may have been a placebo effect. Second, compliance to CPAP treatment is problematic, because patients have to wear a nasal or even an oranasal device during the entire night. The compliance may even be particularly decreased in depressed patients. Indeed, Edinger et al. \[[@B36]\] reported a positive correlation between lower depression scores on the MMPI prior to treatment and CPAP compliance at 6 months of treatment in 28 patients. However, Lewis et al. \[[@B37]\] did not find any association between baseline depression scores and subsequent CPAP use for the first month of treatment. The most important factor to explain the differences among these studies may be the variability in the severity of initial depressive symptoms. Whereas the severity of OSA itself does not seem to have a differential impact on mood improvement after CPAP therapy, the severity of depressive symptoms associated with OSA may impact response to CPAP treatment. As Millmann indicates, OSA patients with more severe mood symptoms responded better to CPAP treatment, whereas patients with less severe or no mood symptoms actually had less benefit from CPAP therapy \[[@B12]\]. However, all negative treatment studies either excluded subjects suffering from a major depressive disorder, or their depression scores were even at baseline in a normal range (baseline values: mean BDI of 7.5 in \[[@B32]\], mean depression score on POMS scale of 12.5 in \[[@B34]\], mean BDI of 8 in \[[@B33]\], and no information given on assessed GDS scores in \[[@B35]\]). Future studies should seek to include OSA patients with a broader range of depressive symptoms in treatment studies, to investigate whether CPAP might have a better effect on mood in more depressed OSA patients. OSA in depression ================= Compared to the large number of studies investigating depressive symptomatology in OSA patients, far fewer studies have focused on the screening for OSA in a primarily depressed study population. In one of the few investigations of the prevalence of OSA in a depressed cohort, Reynolds et al. found, in a small sample of 17 older patients with major depression, that 17.6% also had an OSA syndrome, compared to 4.3% of 23 healthy elderly controls \[[@B38]\]. This suggests that OSA might be an important confounding factor for studies on mood disorders in general, as its presence is not routinely determined in either research studies examining mood or clinical settings. However, many more studies are required to assess the prevalence of OSA in primarily depressed patients, particularly as it can be suspected from existing studies that OSA is greatly underdiagnosed in this patient population. Clinically, this is of particular concern, as sedative antidepressants and adjunct treatments for depression may actually exacerbate OSA. Notably hypnotics prescribed to treat depression-related insomnia might further decrease the muscle tone in the already functionally impaired upper airway dilatator muscles, blunt the arousal response to hypoxia and hypercapnia as well as increase the arousal threshold for the apneic event, therefore increasing the number and duration of apneas \[[@B39],[@B40]\]. These effects might differ depending on the patient population and the severity of OSA. Older depressive subjects are of primary concern: both, frequency of OSA and depressive symptoms increase with age, as do prescription and consumption of sedative psychotropic medication. Pharmacologic treatment of depression and depression-related insomnia in this age group should therefore routinely consider the potential presence of a concomitant OSA. Finally, as Baran and Richert point out, the diagnosis of a mood disorder in the presence of OSA has its very own challenges \[[@B41]\]. Considering the DSM-IV definitions \[[@B42]\], it could either be viewed as a mood disorder due to a general medical condition, or classified as an adjustment disorder with depressed mood, due in particular to EDS and its debilitating consequences on the patients\' daytime functioning. The identification of pathophysiological features that allow distinction between OSA and depression might assist with such diagnostic issues. Sleep architecture in depression and OSA ======================================== Both depression and OSA have been well characterized with regards to their sleep architecture. Typically, for major depression, polysomnography (PSG) findings confirm the patients\' complaints of insomnia, notably difficulties falling asleep (PSG: increase in sleep latency), frequent awakenings during the night and early morning awakenings (PSG: idem) as well as non-refreshing sleep (PSG: decrease in slow wave sleep). PSG furthermore reveals a shortened REM latency, i.e. the first episode of REM sleep appears earlier than usual, with an increase in total percentage of REM sleep during the night, as well as in its eye movement density (referred to as REM sleep disinhibition) \[[@B43]\]. On the other hand, the sleep of patients with OSA is fragmented, and contains a lot of transitional sleep stages (stage 1) at the expense of REM sleep and particularly of slow wave sleep (stages 3 and 4) \[[@B44],[@B45]\]. At least two studies have investigated sleep architecture at the interplay of OSA and depression or depressive symptoms. Reynolds et al. stated that, in contrast to the sleep EEG of depressed patients which characteristically shows a shorter latency of REM sleep, sleep apnea patients with depression displayed an increase in REM latency \[[@B11]\]. Bardwell et al. compared a group of 106 patients with and without OSA with regards to their sleep architecture. Depressed patients who also had OSA displayed a decrease in sleep latency when compared to the depressed group without OSA; and OSA subjects with depressive symptoms had a higher percentage of REM sleep than OSA subjects without depression \[[@B46]\]. Rather than distinguishing a primary depressive illness from an organic affective syndrome related to OSA \[[@B11]\], however, the aforementioned polysomnographic results underscore how both disorders interplay, thus confounding EEG findings characteristic for each disorder. Possible mechanisms underlying the association between depression and OSA ========================================================================= Sleep fragmentation and hypoxemia --------------------------------- The two main factors suspected to be responsible for depressive symptoms in OSA are sleep fragmentation and oxygen desaturation during sleep. Sleep fragmentation is a direct consequence of the recurrent microarousals associated with the apneas and hypopneas, and the nocturnal hypoxemia is due to the intermittent drops in oxygen saturation caused by the respiratory events \[[@B47]\]. Sleep fragmentation is the primary cause of EDS in OSA patients, and is suggested to result in the depressive symptomatology in OSA. This last perspective gains support from the finding that EDS as measured by the Epworth Sleepiness Scale (ESS) and the Maintenance of Wakefulness Test (MWT) was found to be correlated with higher depression scores on the Hospital Depression Scale (HAD-D) in 44 patients with OSA \[[@B48]\]. Furthermore, a Canadian study on 30 OSA patients showed a significant correlation between the severity of psychological symptoms on SCL-90 and less total sleep time, as well as percentage of wake time after sleep onset and ESS scores \[[@B49]\]. With respect to hypoxemia, Engleman et al. noted in a recent review that the effect size of cognitive impairment in OSA correlated highly with severity of hypoxic events, ranging from .3 standard deviations for milder levels of AHI to 2--3 standard deviations for higher levels of AHI \[[@B50]\]. Recently, preliminary imaging data suggests that hypoxemia related to OSA might also play a role in impacting mood. Cerebral metabolic impairment resulting from recurrent nocturnal hypoxemia in OSA have had previously been observed in several imaging investigations on OSA \[[@B51]-[@B53]\]; independently, white matter hyperintensities (WMH) have been linked to depressive symptomatology in studies on affective disorders \[[@B54]-[@B58]\]. Aloia et al. reported in a small sample of older patients with OSA more subcortical WMH in the brain MRI of patients with a severe OSA as compared to those with minimal OSA, and a tendency for a positive correlation between these subcortical hyperintensities and depression scores on the Hamilton Depression Scale \[[@B59]\]. Neurobiology of depression and upper airway control in OSA: the role of serotonin --------------------------------------------------------------------------------- The high comorbidity of OSA and depression also suggests that both disorders may share a common neurobiological risk factor. On the neurotransmitter level, the serotoninergic system has a central role as a neurobiological substrate underlying impairments in the regulations of mood, sleep-wakefulness cycle, and upper airway muscle tone control during sleep. Depression is associated with a functional decrease of serotoninergic neurotransmission, and is mostly responsible for the alterations in sleep as outlined above \[[@B60]\]. The physiopathology of OSA involves numerous factors, among whose the abnormal pharyngeal collapsibility during sleep is one of the most compelling. Serotonin delivery to upper airway dilatator motor neurons has been shown to be reduced in dependency of the vigilance state \[[@B61]\]. This leads to reductions in dilator muscle activity specifically during sleep, which may contribute to sleep apnea. However, whereas the role of serotonin in mood disorders has been largely documented, its involvement in the pathophysiology of sleep apnea remains to be clarified. Interestingly, molecules increasing 5-HT neurotransmission such as the Serotonin reuptake inhibitors (SSRI) are widely prescribed antidepressant molecules that are suggested to similarly improve the apnea hypopnea index in OSA. Serotoninergic drugs such as fluoxetine, protryptiline and paroxetine have already been tested for OSA, with limited success and numerous adverse effects \[[@B61]\]. Several 5-HT receptor ligands and bi-functional molecules are under development, which may in the future be able to target both, the depressive syndrome and OSA. Shared risk factors ------------------- OSA and depression share common risk factors, which may partly explain their high comorbidity in the general population. Very frequently in studies of the impact of OSA on cognitive and psychological functioning, a conglomerate of disorders is shown to contribute to the overall neuropsychological outcome. Therefore, the presence of a polypathology often associated with OSA, such as obesity, cardiovascular disease, hypertension and diabetes, should increase the suspicion of an underlying or coexisting OSA in a depressed patient. Both, depression and OSA, have independently been shown to be associated with metabolic syndrome, and also with the development of cardiovascular disease \[[@B62],[@B63]\]. The association between depression and metabolic syndrome has been suggested to be reciprocal \[[@B64]\], and a priori not attributable to genetic factors as twin studies revealed \[[@B65]\]. In particular, insulin resistance (IR) has been suggested to contribute to the pathophysiology of depressive disorder and has been proposed to subserve the association between depression and cardiovascular disease \[[@B66]\]. Similarly, OSA has been observed to be independently associated with the cardiovascular risk factors comprising metabolic syndrome \[[@B67]\], in particular IR \[[@B68]\]. The magnitude of this association has even led researchers to suggest that metabolic syndrome should encompass OSA \[[@B69]\]. Although OSA and depression share these common risk factors, there are currently no studies available which have investigated the issue of antecedent or consequence in the relationship between depression, OSA and metabolic syndrome, and if and how these three highly prevalent disorders may interact to exacerbate the risk for cardio -- and cerebrovascular morbidity and mortality. Clinical application ==================== As a consequence of the complex relationship between depression and OSA, the assessment of a patient\'s individual sleep history should be included in the standard psychiatric clinical interview, and specifically in the assessment of a depressive syndrome. A clinician should suspect OSA particularly in those depressed patients who present with its cardinal symptoms, namely, 1) loud snoring or intermittent pauses in respiration, as witnessed by a bed partner, associated with 2) excessive daytime sleepiness (EDS). Given that patients often deny the latter, standardized questionnaires such as the Epworth Sleepiness Scale (ESS) \[[@B70]\] or the Functional Outcome Sleep Questionnaire (FOSQ) \[[@B71]\] are useful tools to assess EDS. The ESS asks the patients to rate their chances to fall asleep during periods of relaxation or inactivity (such as reading, watching television), but also in more active settings (driving a car, sitting and talking to someone). EDS is by far the most frequent daytime symptom of OSA, whereas nocturnal symptoms include restlessness, nocturia, excessive salivation and sweating, gastroesophageal reflux, as well as headache and dry mouth or throat in the morning on awakening. Furthermore, the clinical picture frequently includes obesity and hypertension, and, in those patients who are not obese, special facial abnormalities which narrow the upper airway, such as retrognathia or micrognathia. However, it should be kept in mind that OSA may not be immediately apparent, but might present in an atypical fashion, with irritability, tiredness, disrupted sleep, difficulty concentrating, difficulties accomplishing tasks and generally decreased psychomotor performance \[[@B12]\]. Women are more likely to present with these symptoms \[[@B22],[@B72],[@B73]\], and have been suggested to be particularly underdiagnosed because of their atypical symptoms \[[@B74]\]. The importance of the sleep-wake complaints in a patient\'s depressive profile, and the onset of those complaints prior to the development of the depressive psychopathology should draw the clinician\'s attention to a potential underlying or coexisting OSA \[[@B75]\]. Third, particular attention should be paid to depressive patients who are resistant to treatment. In this case, OSA should be excluded as a major underlying contributing factor \[[@B76]\], as treatment of OSA could improve not only the compliance to pharmacological antidepressant treatment, but also the treatment response rate for depression \[[@B77]\]. Fourth, comorbid disorders of OSA may also catch the attention of the treating psychiatrist. In addition to the outlined association with the metabolic syndrome, Farney et al. observed that the likelihood of OSA increased significantly when either antihypertensive or antidepressant medications had been prescribed \[[@B78]\]. Depressed patients with a suspected OSA should be referred to a sleep disorders center for evaluation by nocturnal polysomnography, to confirm the diagnosis of OSA or the presence of other forms of sleep disordered breathing, such as the upper airway resistance syndrome \[[@B79]\]. This is of particular importance, as some of the adjunct treatments to the current pharmacological treatment of depression may actually exacerbate the condition. If the diagnosis of OSA has been established in a depressed patient, and treatment has been initiated, close follow-up of the improvement of the depressive symptoms might give some indications as to the extent to which the presence of OSA may have contributed to the depressive symptomatology. However, as Baran and Richert point out \[[@B41]\], the aforementioned diagnostic challenge of a depressive syndrome in the presence of OSA currently remains unresolved. On the other hand, systematic assessment of depressive symptoms with standardized clinical questionnaires in OSA patients is generally part of the evaluation process in all major sleep disorder centers. However, as these questionnaires have not been specifically designed to assess depression in OSA patients \[[@B80]\], they might be inappropriate to assess depression in this population, given that it is still unclear if OSA and depression display a true comorbidity or only share similar symptoms \[[@B41]\]. Typically, patients with severe depressive symptoms should be referred to a psychiatrist, particularly if such symptoms do not regress or if fatigue lingers after efficient treatment of OSA \[[@B81]\]. Conclusion ========== Recent studies underscore the existence of a complex relationship between depression and OSA in terms of clinical presentation, underlying pathophysiology and treatment. It should incite the treating psychiatrist to be highly aware of a possibly underlying or coexisting OSA in depressed patients. Up to 20% of all patients presenting with a diagnosed depressive syndrome may also have OSA, and vice versa. This relationship might vary widely, depending on age, gender, AHI cut-off and general demographic and health characteristics of the population under investigation. Future clinical research in this area should specifically examine depressed patient populations, taking into account the different sub-type of mood disorders, and investigate a broader range of depressive symptomatology in OSA patients. Basic research should further investigate the causal relationship between depression and OSA, as well as the potential mechanisms by which both disorders may interact. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= CS and ROH both reviewed the existing literature and drafted the manuscript. Both authors approved the final manuscript. Acknowledgements ================ This work was supported in part by National Institute of Health Grant AG 18784; by the Medical Research Service of the VA Palo Alto Health Care System; and by the Department of Veteran Affairs Sierra-Pacific Mental Illness Research, Education, and Clinical Center (MIRECC).	PubMed Central	2024-06-05T03:55:59.749580	2005-6-27	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181621/", "journal": "Ann Gen Psychiatry. 2005 Jun 27; 4:13", "authors": [ { "first": "Carmen M", "last": "Schröder" }, { "first": "Ruth", "last": "O'Hara" } ] }
PMC1181622	Background ========== Eukaryotic genomes have many genes that fall within well-defined gene or super-gene families \[[@B1]\]. Both orthologuous and paraloguous genes within the same gene family may vary in functions at levels from subtle changes in regulation or catalytic efficiency to substantial evolution of new function. While functional divergence within a protein family is usually determined by changes in a few amino acid residues or domains. Identification of these has traditionally required considerable experimental effort. Developing computational tools for predicting these crucial residues or regions has become important in the field of current functional genomics. Many methods have been proposed, such as ancestral sequence inference \[[@B2]\], positive selection \[[@B3]\], and site-specific rate shifts \[[@B4],[@B5]\]. The new software SplitTesterreported here is focused on a special type of functional divergence that functionally associated amino acids do not have the same evolutionary relationship as the protein family. The software is designed to identify domains responsible for functional divergence by iteratively comparing split cluster to the functional classification. Identified inconsistence between the functional divergence and the phylogenetic relationship may provide valuable information for gene function prediction. For illustration, we applied SplitTesterto the reverse transcriptase family from a group of retrotransposons, Pseudoviridae. There are two subgroups of reverse transcriptase, according to the primer utilizations at the initial step of reverse transcription process. One subgroup binds full length tRNA molecule and another one binds tRNA fragment respectively as primer to initiate cDNA synthesis (reviewed in \[[@B6]\] also see \[[@B7]-[@B9]\]). Such difference in primer specificity can not be reflected from the inferred phylogenetic relationship of this protein family, that is, they are not monophyletic because of the parallel evolution for functional-related changes during the expansion of this protein family \[[@B10]\]. Thus, one may design a tree-based (clustering) algorithm that can define domains relevant to diverged function in a protein family with known functional subtypes. The software SplitTesterwe developed is to look for the local sequence alignments that display the clustering topology in agreement to a known functional split, using the evolutionary relationship of the gene family as the reference, which can be reconstructed by the conventional methods. Implementation ============== The algorithm implemented in the software SplitTesterbegins with a multiple sequence alignment of a protein family with known functional diversity (functional subgroups), defined here as a \'functional split\'. Usually, a functional split is based on a few but unknown diagnostic amino acid residues or regions that are expected to be in accordance with the functional split. If the functional subgroups are not consistent with the phylogenetic tree of the gene family, we may, in retrospect, identify the sequence region that may include amino acid residues crucial for the sought-after function, if the clustering analysis of this region shows the expected functional grouping. In the following we call this idea the split-clusteringfor simplicity. Figure [1](#F1){ref-type="fig"} illustrates the example for identifying reverse transcriptase amino acid sequences responsible for priming with full or half-tRNAs. Different reverse transcriptases are known to recognize either full length tRNA or tRNA fragment (half-tRNA) as primers, which forms the basis of the known functional split. However, the exact sequence region that is responsible for the primer choice remains unknown yet. The newly-developed method may be helpful to resolve this problem, using the split-clustering approach. That is, by identifying windows of amino acid residues that cluster reverse transcriptases to match the known functional split, one may identify candidate regions that are responsible for primer recognition diversity. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### The algorithm for the tree-based method to identify protein functional domains.Multiple amino acid sequences alignment is used as an input file. Phylogenetic trees from different windows of the alignment are generated by the neighbor-joining method. For each window, the program determines whether the tree from the local sequence (split cluster) matches a predefined functional split. If the split cluster is consistent with their functional split, the sequence window is a candidate for carrying out that function. The program is iterative and starts with very small windows (i.e. three amino acids), which gradually increase until the window size equals the length of the protein alignment. ::: ![](1471-2105-6-137-1) ::: We have developed a software package called SplitTester(Fig. [2](#F2){ref-type="fig"}). The input file is the protein sequence alignments, as generated by some conventional methods (e.g. ClustalX) \[[@B11]\]. The users should predefine the functional subtypes, usually based on functional differences of proteins that have been verified by the experimentation. In the current version, the partitioning of sequences is limited to two groups. In the case of multiple functional subtypes, one may start with crude partitions and progressively refining pairs of groups, or by comparing two groups pairwisely. The program uses the neighbor-joining (NJ) method for phylogeny inference or split-clustering \[[@B12]\]. The user can select one of several amino acid substitution distance matrices, including the mutation distance matrix, the hydrophobic distance matrix, PAM10-500 and BLOSUM 30--100 \[[@B13],[@B14]\]. Then, the split-clustering algorithm implemented in SplitTesterwill generate tree-like topologies for each sliding windows along the aligned sequences. The procedure is iterative and starts with very small windows (i.e. three amino acid residues), which slide along the length of the alignment. Window size gradually increases until it reaches the full length of the aligned proteins. All the examined windows are displayed in an plot as part of the output interface. The horizontal axis shows the position of sliding windows in the protein alignment and the vertical axis indicates the length of sliding windows. The tree-topology generated by the split-clustering from each window is compared to the predefined functional split. If it matches, this window is marked as a line on the output plot at the corresponding position and length. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### A snapshot of the Split-Tester software.A distance matrix was selected to compute phylogenetic relationships of the aligned input sequence data. The regions of the alignment that support the functional split are then plotted in the top window. The X-axis represents the length of the aligned sequences; the Y-axis represents increasing window size. After the computation is complete, the user can select a specific window for analysis by clicking on the left end of colored horizontal bars. The colors indicate the degree of confidence that a given window supports the predefined functional split (red = 100%; yellow = 75%; green = 50%; blue = 25%). The two panels on the lower right show all NJ equivalent trees generated from the selected window. The lower left window shows the actual sequences that support the predefined phylogenetic relationship within the selected window. ::: ![](1471-2105-6-137-2) ::: The distance method (e.g. NJ, UPGMA) implemented in most phylogenetic software normally uses the greedy algorithm for efficiency and simplicity, but only tracks a single locally best tree. Consequently, ignoring the alternative solutions can be misleading for phylogeny inference particularly when the sequence length is short \[[@B15],[@B16]\]. To improve the reliability of distance method, there are some efforts to track multiple partial solutions as it progresses \[[@B16]\]. Since split-clustering usually deals with the short sequence length, we modified the conventional algorithm by tracking all possible topologies equally best fitting the data. That is, at each step, the split-clustering searches for the minimal distance pair, creating a list of pairs with equally minimal distance, and then performs the neighbor joining on each pair in order following a depth-first search. Only the resulted unique topology trees are saved for subsequent analyses. In the following we call them NJ-equivalent tree topologies if it is inferred by the NJ algorithm. If the split-clustering analysis for a given window yields at least one (NJ-equivalent) tree whose topology matches the predefined functional split, we consider the window containing a potential functional signal. In the case of multiple NJ-equivalent trees, as explained above, the strength of the functional signal, or the degree of confidence, is measured by the percentage of NJ-equivalent trees that support the functional split. We used different colors to indicate the signal strength of the window: red means that all (100%) these trees derived from this window split the genes according to the specified function; yellow, green and blue indicate that 75%, 50% and 25% of the trees match the functional split, respectively. A color gradient is used to represent degrees of confidence between the above intervals. In the output plot, all the lines representing the candidate windows can be selected to display the sequences within this window and the corresponding phylogenetic trees in two separate panels. The panels are updated in real time with the progress of each window tested. While in theory our algorithm could be quite time consuming to run because of the potentially large number of individual NJ operations (time increases quadratically with the length of the alignment), in practice the calculation is very reasonable, because the sequence length is typically small. Our dataset (eight sequences each of 540 amino acid residues) required approximately six minutes for analysis of all possible windows using a Pentium 3 \~ 930 Mhz processor with 512MB RAM and the Microsoft Windows 2000 operating system. In practice, most domains will be identified when window sizes are smaller than 150 amino acids. Therefore, 2--3 minutes of running time is expected to be sufficient for most computers. SplitTesteris available as precompiled binary in a distribution package for Microsoft Windows from the following URL: <http://www.public.iastate.edu/~voytas/SplitTester>, and <http://xgu.zool.iastate.edu/>. Both a zip file of the installation files and a self-extracting installer are available. Documentation and example files are included in the distribution packages. Results and discussion ====================== We applied SplitTesterto understand functional diversity among retroelement proteins for primer utilization by retroelement reverse transcriptases. Reverse transcriptases of retroviruses (Retroviridae), Metaviruses (Metaviridae) as well as retrotransposons in the genusPseudovirus (member of Pseudoviridaefamily) use the 3\' acceptor stem of the host tRNA as a primer for DNA synthesis. This region pairs with the retroelement RNA template to start DNA synthesis from 3\'-OH of the tRNA. Retrotransposons of the genusHemiviruses (member of Pseudoviridaefamily) use a half-tRNA primer, and cDNA synthesis initiates from 3\'-OH of nucleotide 40, which resides within the anticodon stem-loop. It is likely that the primer template complexes for the two groups of elements have different structural conformations or properties, and that reverse transcriptases from the different groups have evolved the ability to recognize these differences. The split-clustering implemented by SplitTestercould explore candidate domains of reverse transcriptase related to primer selection. These identified candidates provide testable hypotheses that could be verified or falsified by the follow-up molecular genetic experimentation. We focused on members of the Pseudoviridae, which include both half-tRNA priming elements (namely the Hemiviruses: Ty5 (U19263), Osser(X69552), 1731(X07656) and copia(M11240)) and elements that use full length tRNAs (namely the Pseudoviruses: Ty1 (M18706), Opie-2(U68408), Tnt1 (X13777) and SIRE-1 (AF053008)) \[[@B17],[@B18]\]. The NJ phylogenetic tree from full length sequence alignment did not reflect the divergence of the two functional subtypes (Fig. [4A](#F4){ref-type="fig"}) \[[@B10]\]. For example, Osserand Tnt1 are the only two members of one cluster and the bootstrap value is 74. The genes seem to be clustered according to their hosts, e.g., 1731and copiafrom Drosophila, Ty5 and Ty1 from Saccharomyces cerevisiae, as well as Osser, Tnt1, SIRE-1and Opie-2from plants. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Functional divergence in reverse transcriptase.(A) The SplitTesteroutput for the reverse transcriptase dataset. Windows supporting the functional split are shown as colored lines in the plot. The X-axis represents the length of the aligned sequences; the Y-axis represents increasing window size (see legend to Fig. 2 for additional detail). (B) The X-ray structure of the HIV reverse transcriptase/primer/template complex (1RTD). The reverse transcriptase protein is represented by the yellow strand. The two green regions are domains identified by SplitTester. All residue numbers correspond to HIV sequence positions in 1RTD. Residues 166--215 and 280--311 in the aligned retrotransposon sequences correspond to 167--210 and 267--297 in the HIV 1RTD sequence, respectively. ::: ![](1471-2105-6-137-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Phylogenetic relationship from the full length multiple sequence alignment and the predicted region.(A) NJ phylogeny (MEGA3.0 \[32\]) from the reverse transcriptase full length sequence alignments clusters the genes from the same host: Osser, Tnt1, SIRE-1 and Opie-2 are from plant host, while Copiaand 1731are from Drosophila. Ty1 and Ty5 are from Saccharomyces cerevisiae. (B) Split cluster from the window length of 50 aa (position 166--215) in the predicted region 1 supports the functional subtype split. (C) Functional signal (measured by the bootstrap of node β in panel B), as well as the evolutionary background (measured by the bootstrap of node α in panel A), plotted against the window size. In the window length less than 90, the split-clustering supports the functional subtypes split and the bootstrap value reach the peak in window with length around 50 aa. A mixed topology is detected when window length is longer than 90 aa, measured by the bootstrap (γ) between two major subtrees. When more amino acid sites are included, the bootstrapping value converges to the node α in panel A. ::: ![](1471-2105-6-137-4) ::: Two signal regions were identified by SplitTesterwhen the mutation distance matrix was used to compute the cost of amino acid substitution and when gaps were considered as potentially informative characters. In the aligned amino acid sequence, the maximum windows showing functional split in each non-overlapping region cover positions 144--239 (region 1) and 269--314 (region 2) (Fig. [3A](#F3){ref-type="fig"}). Using the hydrophobicity matrix, SplitTesteralso identified these regions, as well as an extra region at the N-terminus (amino acids 1--29, region 3). Each of signal regions above is derived from the multiple overlapping signal windows with gradually increased window length. We chose 50 aa window (166--215) in region 1 and 32 aa window (280--311) in region 2. Both windows are close to the median window length with demonstrating high bootstrapping values (Fig. [4C](#F4){ref-type="fig"}). In general, one may use the median sized window for representing, because small windows are statistically unstable while a too broad window will demolish the functional signal of the window. We located the two windows identified by both matrices within the HIV reverse transcriptase sequence, based on the published protein sequence alignments of the reverse transcriptase family by Xiong and Eickbush \[[@B19]\], and mapped onto the crystal structure 1RTD \[[@B20]\] (Fig. [3B](#F3){ref-type="fig"}). HIV reverse transcriptase p66 has a \'right hand\' structure. The 50 aa window (166--215) in region 1 from the alignment correspond to the \"palm\" of the protein (residues 167--210 in HIV reverse transcriptase) that encompasses the polymerase active site wherein nucleotides are added to the 3\' end of the primer. The 32 aa window (280--311) in region 2 corresponds to one α-helix in the \'thumb\' region of HIV reverse transcriptase (residues 267--297), which directly contact the primer/template complex, as determined by cross-linking experiments \[[@B21],[@B22]\]. The region of β-sheets between the two identified domains (240--268 in the retroelement reverse transcriptase alignment) is called the primer grip and cross-links to the 3\'-OH group of the HIV primer, tRNA^lys^\[[@B23],[@B24]\]. We would speculate that primer grip is related to the primer binding. However, this region is not identified as determinant of primer specificity, indicating that the ability of the primer grip to interact with the 3\'-OH of the primer will be a common feature of both full-and half-tRNA priming retroelements. The two regions identified by SplitTestersurrounding the primer grip may play a role in distinguishing primer conformation or length. The results of SplitTester, therefore, can be well explained from the reverse transcriptase crystal structure and supported by experimental data. To further validate our findings, we randomly partitioned the gene members into two groups and reran the program. The candidate functional regions were not identified using any of the partitions (data not shown). We therefore conclude that the identified residues may separate the functional subtypes specifically. The strength of functional signal vs. the evolutionary background on the multiple windows of different length (from 8--200) in the signal region 1 can be well illustrated by the bootstrapping values (Fig. [4C](#F4){ref-type="fig"}). The right splitting of the two functional groups in this region (29-90aa), as predicted by SplitTester, showing a (local) maxima at the window length of 50 residues. Interestingly, when the window size is greater than 90, the topology by the split-clustering is a random mixture between functional split and evolutionary background; the bootstrapping value is for the deepest two subtrees. Rather, with the increase of the window length toward the window of full length sequence, the topology converges to the evolutionary tree as shown in Fig. [4A](#F4){ref-type="fig"}. Therefore, the region identified by SplitTesteris intuitively valid even though although the bootstrapping (re-sampling) value is not very high. We also observed the very similar pattern for region 2; the highest bootstrap value in window with size of 32 aa around the median length within the range of 8 to 46. Because the SplitTesteremploys a heuristic algorithm to decipher weak functional signal from the high level of evolutionary background, its power would be limited by several factors. If there are only a few amino acid residues related to a given functional split, or these residues are located in a broad sequence region, the statistical power for detection is low, as we expected. One improvement for SplitTesteris to divide the proteins into several small regions, for instance, based on the protein structure, or motifs. Then one can implement some algorithms to integrate weak signals from these non-overlapping small regions, which otherwise are indistinguishable from the background. There are several other factors that may affect the performance of SplitTester. The first one is the choice of substitution matrices of proteins. We have provided several options for the matrices. The user can select one of them based on their purpose. In general, BLUSOM matrices reflect the overall evolution history and hydrophobicity matrix can reflect more on chemistry properties of amino acids. Second, the inference power of this method might increase with the difference between the functional splitting and phylogenetic tree because the signals are more likely being distinguished from the background. Thus, increasing the sample size, for instance, sequencing more genes from different species is certainly helpful. Nevertheless, the example of reverse transcriptases we presented here has indicated that one may obtain valuable functional information even the difference between functional splitting and phylogenetic tree is weak as indicated by the bootstrap value. There have been many methods developed to analyze sequences involved in functional divergence between protein subtypes (e.g., \[[@B25],[@B26],[@B4]\]). Most of them are focused on identifying specific residues contributing to the functional differences of subtypes along the phylogeny. Hence, the prediction accuracy may rely on the quality of multi-alignments, the accuracy of phylogenetic inference, or the sufficient number of sequences. Since the SplitTesterfocus on a particular sequence region that supports the subtype functional split, the predicted results seem not to be strongly affected if a few positions are misaligned. We validated this claim by deleting the position 166 in the multiple sequence alignment, which contain conserved F in Pseudovirus subfamily and A, V, E, T in each gene of Hemevirus subfamily respectively. This position was selected because it can be easily identified manually as one of three \"seeds\" of the growing signal region (Fig. [3A](#F3){ref-type="fig"}). Fortunately, SplitTesterstill identified the almost identical region 1 as before only in overlapped shorter windows (145--221) and (167--238) (data not shown). Similar results were obtained when we deleted each of other two \"seeds\" at position 184 (conserved K only in Pseudovirus subfamily) and 207 (conserved A only in Hemivirus subfamily). Conclusion ========== SplitTestercan explore regions potentially responsible for functional divergence of proteins. The best scope of this software is to study the dataset with different functional clustering and phylogenetic tree. As shown by the case of reverse transcriptase, function-related signals may emerge when the functional split is inconsistent with the phylogenetic relationship of the protein family. Even if the functional clustering is consistent with its phylogeny, SplitTestermay also provide some useful information for amino acid residues important for functional divergence, e.g., the conserved Myb gene family. In spite that the phylogeny is the same as the known functional split of Myb genes, SplitTesterstill successfully identified 11 candidate residues that differentiate the two-and three-repeat Myb proteins, the major functional split of Myb gene family \[[@B27]\], because the sequence window including these residues shows a stronger functional signal. Indeed, these identified candidate residues are well supported by their locations on the NMR structure of the mouse c-Myb DNA binding domain 1MSF \[[@B28]\]. Moreover, these residues are called by type-II functional divergence by \[[@B29]\], which can also be predicted by the \"Evolution Trace\" method \[[@B30]\]. See [additional files 1](#S1){ref-type="supplementary-material"} and [2](#S2){ref-type="supplementary-material"} for the detail. Finally, we mention that, after combining SplitTesterwith other complementary methods, such as the method of \[[@B25]\], \"Evolution Trace\" method \[[@B30]\], \"Phylogenetic Inference\" by Sjolander \[[@B31]\], Diverge \[[@B5]\], we can develop a powerful analysis pipeline for predicting functional divergence from sequence domains to amino acid residues. In summary, we developed SplitTester -- a tool for exploring the functional domains in protein family. SplitTesterfocus on a specific type of functional divergence: the functional split is different from the evolutionary relationship. Using the split-clustering algorithm, SplitTesterscans all the possible local regions of protein sequence alignment to identify the domain that provide the same clustering topology as the functional split. In the analysis of retroelements reverse transcriptase family, we identified the regions splitting this family according to the primer specificity, implying function in the primer selection. The functional role can be well explained after we map the identified domain onto the structure of the reverse transcriptase protein. Availability and requirements ============================= • project name: SplitTester • Project home page:<http://www.public.iastate.edu/~voytas/SplitTester/> <http://xgu.zool.iastate.edu/software.html> • Operating system(s): windows 2000 and XP • Programming language: C Authors\' contributions ======================= XGao designed this project, analyzed the results and prepared the manuscript. KVV implemented the software. XGu and DFV supervised the project and helped to prepare the manuscript. All authors read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Display the result of identified residues in MyB protein. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Descript the detail of Myb protein family and the predicted key residues for functional divergence from SplitTester. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We thank Dr. Les Miller for helpful discussions. This work was supported by NIH grants GM61657 to D.F.V, and GM62118 to X.Gu.	PubMed Central	2024-06-05T03:55:59.752236	2005-6-1	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181622/", "journal": "BMC Bioinformatics. 2005 Jun 1; 6:137", "authors": [ { "first": "Xiang", "last": "Gao" }, { "first": "Kent A ", "last": "Vander Velden" }, { "first": "Daniel F", "last": "Voytas" }, { "first": "Xun", "last": "Gu" } ] }
PMC1181623	Background ========== Aneuploidies (chromosomal copy number changes) constitute a key mechanism in cancer progression \[[@B1],[@B2]\] and play important evolutionary roles in speciation \[[@B3]\] and adaptive mutation in yeast and microbial populations \[[@B4],[@B5]\]. Array-based comparative genomic hybridization (array CGH) has enabled fast genome-wide investigations of copy-number changes \[[@B6],[@B7]\]. However, once microarray experiments have been performed, accurate identification of amplifications and deletions requires a combination of manual discovery through data visualization and sophisticated statistical analysis. Computational methods can use additional data sources, such as gene expression, to facilitate the discovery and analysis of genomic aberrations. This is possible because the presence of amplifications or deletions of whole or partial chromosomes can have substantial effects on gene expression in the affected regions \[[@B8]-[@B10]\]. Gene expression microarray data can serve both as a second source of information for aneuploidy detection and perhaps as an indication of which genomic changes are most functionally relevant since mRNA transcript abundance more directly affects cellular phenotype than genomic DNA content. Therefore, an effective visualization and analysis system for aneuploidy detection should make use of both array CGH and gene expression data, and allow easy examination of overlaps in the corresponding data sets. Existing visualization tools include Caryoscope \[[@B11]\], CGHAnalyzer \[[@B12]\], Java Treeview\'s Karyoscope \[[@B13]\], and SeeCGH \[[@B14]\]. All of these were developed specifically for the analysis of array CGH data and with the exception of CGHAnalyzer, none allow convenient visualization of multiple experiments. Additionally, while they all offer a number of useful approaches to visualization, none include automatic statistical prediction to complement manual discovery of amplifications and deletions (see Table [1](#T1){ref-type="table"} for a detailed comparison of features of our software as compared to those of existing applications). To facilitate discovery of genomic aberrations from microarray data, novel methodology is required that integrates visualization with sophisticated statistical analysis and enables visualization of multiple experiments and data types simultaneously. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Comparison with existing visualization and analysis software. Comparison of ChARMView\'s functionality with existing visualization and analysis software including Caryoscope \[11\], CGHAnalyzer \[12\], Java Treeview\'s Karyoscope \[13\], SeeCGH \[14\], CGH-Explorer \[24\], CGH-PRO \[25\], and CGH-Miner \[26\]. ::: Feature ChARM View Caryo-scope TreeView (Karyosco pe) SeeGH CGH Analyzer CGH Explorer CGH PRO CGH Miner --------------------------------------------------------------------------- ---------------------------------------- ----------------------------- ----------------------------- ---------------------------------------- --------------------------------------- --------------------------------------- ---------------------- --------------------------------------- Platform most platforms (Java-based) most platforms (Java-based) most platforms (Java-based) Windows most platforms (Java-based) most platforms (Java-based) Linux, Windows Windows, Unix, Excel add-in Software availability freely downloadab le with registration freely downloadable freely downloadable freely downloadab le with registration freely downloadable with registration freely downloadable with registration freely downloadab le freely downloadable with registration Source-code license GNU GPL MIT license GNU GPL not available freely downloadable freely downloadable GNU GPL freely downloadable External software dependencies none none none requires MySQL database none none MySQL, R R Automatic statistical determination of single-array aberrations Yes No No No No Yes Yes Yes Statistical analysis of manually selected regions Yes No No No Yes Yes No No Simultaneous display of multiple experiments Yes No No No Yes Yes Yes Yes Statistical analysis of aberrations occurring in multiple experiments No No No No Yes No No No Aberration breakpoints/statistics export Yes No No No Yes Yes Yes Yes Image export Yes Yes Yes Yes Yes Yes Yes Yes Command-line statistical analysis feature Yes No No No No No No No Allows user-defined genomic feature annotation Yes Yes Yes Yes Yes Yes Yes Yes Web-linked genomic feature annotation No Yes Yes Yes Yes Yes Yes No ::: Here we describe ChARMView -- an integrated system that combines statistical analysis with effective visualization capabilities to enable interpretation of microarray data for aneuploidy discovery. Our system facilitates both manual and automated discovery of genomic aberrations from microarray data and can display multiple experiments and data types simultaneously. ChARM-View can be used to identify amplifications and deletions from array CGH or gene expression data independently or simultaneously, making it a powerful approach for identifying real and functionally relevant chromosomal changes. Implementation ============== ChARMView was implemented in Java using Swing set components to ensure cross-platform compatibility. Many of the visualization features were developed using the Open Source 2D graphics toolkit Piccolo developed at the University of Maryland \[[@B15]\]. Results and Discussion ====================== Methodology: statistical analysis --------------------------------- ChARMView computational analysis automatically detects regions of non-random spatial bias and is appropriate for any genomic data associated with chromosomal coordinates. Statistical analysis is based on our algorithm ChARM (Chromosomal Aberration Region Miner), described in detail in \[[@B16]\]. ChARM identifies potential breakpoints by a differential filter followed by an accurate expectation-maximization approach. The statistical significance of each identified region is evaluated with a one-sample sign test and a permutation-based mean test. By their formulation, the significance tests are valid for any size segment, but do lose power with decreasing segment size. ChARM has been evaluated on gene expression and array CGH data: it is robust and accurate for regions as small as 4--5 probes, and sensitive enough to detect aneuploidies even in mixed populations of cells \[[@B16]\]. As a system for dynamic and real-time data analysis and visualization, ChARMView requires very fast statistical algorithms. However, the permutation-based test as originally described in Myers et al. \[[@B16]\] requires non-trivial computation since it involves performing several thousand permutations of the chromosome order. To speed up the mean permutation test for the software system, we have developed an accurate approximation that requires many fewer permutations. The original version of the test requires computing the mean of the region of interest and comparing this with the means of similar-sized segments in randomly permuted data. We have verified that means of typical chromosomal segments in array CGH and gene expression data can generally be reasonably approximated with a normal distribution. This is a generally well-accepted claim even for small groups (\~10) unless the underlying population is extremely non-normal, which is typically not the case for log-transformed array CGH or gene expression data. The statistical significance of predicted aneuploidy region in ChARMView is obtained by computing means of 200 permutations of chromosome ordering of the actual data, estimating the parameters, and then integrating the tail of the underlying distribution beyond the observed value. Figure [1](#F1){ref-type="fig"} illustrates the correlation between p-values generated from 10,000 random permutations and p-values obtained from a normal approximation whose parameters were estimated with only 200 permutations. This approximation yields the precision of several thousand permutations based on significantly less computation. Completing a fully automated statistical analysis on a typical gene expression dataset (6000 genes over 16 chromosomes, measured in 16 experiments) requires approximately 7 seconds/experiment for a total of less than 2 minutes on a Pentium 4 3.2 GHz desktop. ChARMView also allows users to manually select regions to test for statistical significance. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Gaussian approximation of the permutation-based p-value.ChARMView performs 200 permutations to assess statistical significance of predictions, then uses Gaussian approximation to estimate the p-value. The p-values based on 200 permutations and Gaussian approximation show .9991 correlation to the p-values based on 10,000 permutations as computed by ChARM. ::: ![](1471-2105-6-146-1) ::: Methodology: visualization-based analysis ----------------------------------------- The most powerful aspect of ChARMView is integration of computational analysis with visualization. This combination of visualization and analysis enables users to view automated predictions of aneuploidies as well as analyze statistical significance of manually selected regions. Visualization is a critical complement to computational analysis as human perception can often identify subtle trends in the data that cannot be detected with purely computational methods. This is especially critical when comparing results of multiple experiments or experimental replicates, such as in cancer studies where researchers often search for recurring aneuploidies in a set of patients. ChARMView facilitates such discovery with visualization of multiple experimental replicates, experiments, and data types. The most common way to increase confidence in results of an experiment is to produce replicate microarray experiments. Data from such replicate experiments is usually averaged for computational analysis. However, viewing such replicates simultaneously is an effective approach to analysis, as people are often perceptive of subtle but repeated trends that are difficult to capture with a statistical test. This visualization-based approach does not make any assumptions, such as independence assumption of the typically used Fisher meta-analysis test \[[@B17]\]. Thus, aligning corresponding chromosomal data from several replicates of the same experiment typically allows the user to spot trends that might otherwise go unnoticed. Figure [2](#F2){ref-type="fig"} illustrates this phenomenon with two replicates of the same array CGH experiment. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Simultaneous visualization of replicate experiments.A set of replicate array CGH experiments from Dunham et al. \[5\] displayed with ChARMView (chromosome 4 of CP1AB, replicates 1 and 2 shown). The region identified by the arrows is hard to distinguish from noise in either of the replicates when viewed separately, but is clearly a region of positive bias when the replicates are viewed together. This is confirmed by statistical analysis. ::: ![](1471-2105-6-146-2) ::: The simultaneous display feature of ChARMView is also useful for visual analysis of computational prediction results for multiple experiments. This is an effective method for identifying common genomic aberrations in otherwise uncorrelated experiments or a characteristic aberration in a set of samples with a common phenotype. For example, a set of breast cancer samples \[[@B18],[@B19]\] can share the same bias in gene expression that corresponds to a predicted aneuploidy or a localized expression bias, as shown in Figure [3](#F3){ref-type="fig"}. Overlapping predictions serve as independent confirmations that the predicted aberration is real. Furthermore, results of such analysis of multiple samples can then be used to correlate specific chromosomal aberrations with phenotypic or clinical parameters. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Simultaneous visualization of multiple independent expression microarrays.Simultaneous visualization of overlapping significant expression biases in a set of four independent breast tumor samples from Sorlie et al. \[19\] (chromosome 6 of breast tumor expression profiles BC208A-BE, BC305A-BE, BC308B-BE, and BC111A-BE shown). Each red bar below the data indicates a predicted aberration identified independently on the corresponding experiment. ::: ![](1471-2105-6-146-3) ::: As array CGH techniques become more widely applied, the generation of copy number data is rarely the end goal of biological studies. Instead, a key challenge is deciphering which parts of a karyotypic profile are responsible for particular phenotypes. While sophisticated statistical and computational methods will certainly be required to answer these questions, the most effective approaches will also need to harness the power of human visual perception. To address this issue, ChARMView can display and analyze both array CGH and gene expression microarray data and display these diverse data and predictions for corresponding chromosomes simultaneously. Simultaneous display of array CGH and gene expression data enables researchers to observe the effect that amplification or deletion of particular sequences of genomic DNA has on the abundance of mRNA transcripts (Figure [4](#F4){ref-type="fig"}). We have noted a number of cases where large amplifications or deletions result in no detectable change in gene expression. These regions may be less likely to cause a particular phenotype than aneuploidies that result in drastic changes in gene expression. ChARMView facilitates convenient discovery of these changes, focusing further experimental investigation. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Identifying functionally relevant genomic aberrations.A small amplification evident in the array CGH data breast cancer data (top) \[18\], and its effect on mRNA expression (bottom) \[19\] (chromosome 15 of breast tumor sample 709B shown). ::: ![](1471-2105-6-146-4) ::: A final unique characteristic of ChARMView is that its visualization and statistical tools are developed for general use, independent of data type and organism. Any dataset with features that can be associated with chromosomal position can be imported and analyzed with ChARMView. For instance, the software has been particularly useful in identification of aneuploidies based on gene expression datasets although array CGH is the typical experimental approach for probing genomic amplifications or deletions. ChARMView has also been used to identify spatially-correlated biases in gene expression that are unrelated to altered chromosome structure. Generally, our tool can be used to identify any region of non-randomness with respect to position in genomic data with inherent ordering. In addition to its usefulness for a variety of data types, ChARMView can be applied to a variety of organisms. By default, the system provides chromosomal coordinates for Saccharomyces cerevisiaedata with ORF identifiers and human data with Unigene identifiers. However, any data that can be mapped to a set of linear chromosomes can be imported and analyzed by ChARMView. Illustration of application --------------------------- We have applied ChARMView to a number of array CGH and gene expression datasets, including data derived from both Saccharomyces cerevisiaeand human experiments. Here we present an example application of our software to array CGH data from experimental evolution experiments in which eight strains of budding yeast were analyzed for chromosomal copy number changes after 100--500 generations of growth in glucose-limited chemostats \[[@B5]\]. Dunham et al. confirmed aneuploidy regions identified by array CGH through pulsed-field gel electrophoresis, thus creating a standard for assessing our results. Our method identified all 12 of the confirmed aneuploidies and two additional regions of bias. The novel regions identified by our method correspond to biases smaller than the ones identified by Dunham et al. \[[@B5]\] and may reflect aneuploidy present in a subset of cells in the population or may be due to a hybridization artifact. Further laboratory experiments are required to further evaluate these predictions. Figure [5](#F5){ref-type="fig"} shows a screenshot of our application upon finishing automated statistical analysis of one of these experiments. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### *Screenshot of ChARMView applied to S. cerevisiaearray CGH data.Screenshot of ChARMView analysis of S. cerevisiaemolecular evolution experiments data from Dunham et al\[5\]. The right panel displays array CGH data arranged in the order of chromosomal position and amplification (red) and deletion (green) predictions. The left panel displays information for the selected region, including gene names and values and statistics for the selected amplification prediction. ::: ![](1471-2105-6-146-5) ::: We also present two specific instances from an array CGH breast cancer study where ChARMView can be used to visualize and accurately predict breakpoints of known amplifications. Figure [6A](#F6){ref-type="fig"} illustrates the results of ChARMView\'s automated statistical analysis on chromosome 1 array CGH profiles of three different breast tumor samples (110B, 112B, 122A) from \[[@B18]\]. The entire q arm of chromosome 1 is known to frequently amplified in breast cancer (typically observed in approximately 50--60% of tumors \[[@B20],[@B21]\]). Thus, we expect the amplications here to begin at or near the centromeric end of the q arm. ChARMView predicts breakpoints 3, 1, and 0 probes from the centromeric end of the q arm for samples 110B, 112B, and 122A respectively. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Predicted amplifications and deletions on breast cancer array CGH data.ChARMView automated predictions on three breast tumor array CGH profiles (110B, 112B, 122A) from chromosome 1 and three profiles (123B, 309A, and BC-A) from chromosome 17 of \[18\]. The predicted chromosome 1 breakpoints (identified by arrows in Figure 6A) are 3, 1, and 0 probes from the centromere. The predicted chromosome 17 amplification common to all three profiles (identified by arrows in Figure 6B) includes 7 genes known to be typically amplified with the ERBB2locus. All visible predictions have Bonferroni-corrected p-values less than .05 for both mean and sign significance tests. See Table 2 for a complete list of breakpoint predictions for each of the results pictured. ::: ![](1471-2105-6-146-6) ::: ChARMView can also be used to accurately find much smaller regions of amplification or deletion and the associated breakpoints. Figure [6B](#F6){ref-type="fig"} illustrates this capability on chromosome 17 array CGH profiles of three breast tumor samples (123B, 309A, and BC-A) from \[[@B18]\]. An amplicon frequently associated with breast tumors includes the ERBB2oncogene at 17q12. While breakpoints identified in individual tumors vary, recent studies have identified a group of 7 genes surrounding the ERBB2locus that are commonly amplified, including NEUROD2, MLN64, PNMT, ERBB2, GRB7, ZNFN1A3, and EST 48582 \[[@B22],[@B23]\]. ChARMView\'s amplification predictions for the three tumor profiles shown include 15, 18, and 13 probes respectively, all of which span the 7-gene region previously identified. All predictions shown in Figure [6](#F6){ref-type="fig"} have Bonferroni-corrected p-values less than .05 for both mean and sign significance tests. Complete lists of predicted breakpoints for both chromosome 1 and chromosome 17 amplicons are included in Table [2](#T2){ref-type="table"}. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Examples of predicted breakpoints in breast tumor aCGH case study. Listing of Unigene IDs corresponding to predicted breakpoints for ChARMView results pictured in Figures 6A and 6B. The Unigene ID and gene name are the first and last markers included in the predicted amplification. All results listed have Bonferroni-corrected less than p-values of .05 for both mean and sign significance tests. ::: Tumor sample* Chrom. Predicted start breakpoint Adjacent gene (in amplicon) Predicted end breakpoint Adjacent gene (in amplicon) ------------------ ------------ -------------------------------- --------------------------------- ------------------------------ --------------------------------- 110B 1 Hs.15871 ACP6 Hs.7395 (last marker) TFB2M 112B 1 Hs.59889 HMGCS2 Hs.7395 (last marker) TFB2M 122A 1 Hs.381235 SEC22L1 Hs.7395 (last marker) TFB2M 123B 17 Hs.97477 LYZL6 Hs.276916 NR1D1 309A 17 Hs.73817 CCL18 Hs.267871 PTRF1 BC-A 17 Hs.635 CACNB1 Hs.2340 HAP1 ::: Usage ----- ChARMView can be downloaded at <http://function.princeton.edu/ChARMView> and run on virtually any platform if the J2SE Java Runtime Environment version 1.4.2 or greater is present. A brief overview of the primary features of the software follows. ### Loading data ChARMView accepts all types of data from any organism provided that the features can be ordered on a set of linear chromosomes. Input files must be tab-delimited, specifically in the commonly-used .pcl format. Chromosome labels and position must be included in the input file unless the organism type is Saccharomyces cerevisiaeor human with ORF or Unigene identifiers, which ChARMView is able to order without coordinates. ### Viewing data Figure [5](#F5){ref-type="fig"} shows a typical ChARMView screenshot upon loading data and statistical analysis. The data display is zoomable and selectable with mouse-overs for identification of experiments and individual genes. Zoom features include standard single-click magnification, zoom to rectangle, and zoom reset (fit to screen) capabilities. When one or more gene or probe data points are selected, identifiers and associated annotation are displayed in the \"Results\" tab, which appears adjacent to the display panel. This allows users to select regions of interest on the display panel and retrieve lists of genes or probes within these regions. Additionally, any number of experiments may be viewed simultaneously by toggling the corresponding checkboxes in the \"Experiment Options\" tab, also adjacent to the display panel. ### Analyzing data ChARMView supports two different modes of analysis. The first employs the automated edge-finding algorithm discussed in Myers et al. \[[@B16]\] followed by statistical analysis. The second mode is for testing user-selected regions of data and only evaluates the statistical significance of the chosen region. Both methods of analysis rely on two tests of statistical significance: a mean-based permutation test, and a one-sample sign test. Details of both tests are discussed above and in Myers et al. \[[@B16]\]. P-values for these tests are reported for all regions found by the automated approach or selected by the user. Figure [5](#F5){ref-type="fig"} displays a typical view of statistical results for a single experiment. Note that the red and green rectangles below the data correspond to regions of predicted aberration. The p-value cutoff at which results of the statistical analysis appear in the display panel can be adjusted by applying p-value filters provided in the \"Prediction Options\" tab adjacent to the display panel. A p-value filter consists of a logical combination of the mean permutation test and/or the one-sample sign test and real-valued cutoffs for each test. These combinations specify how the selected p-value cutoffs will be used to deem statistical significance. For instance, one possible p-value filter is \"Sign AND Mean Tests\" with Sign p-value cutoff of 0.001 and Mean p-value cutoff of 0.01, which will result in only predictions with both Bonferroni corrected sign p-values of less than .001 and mean p-values of .01 being displayed. The Bonferroni corrected p-value is obtained by multiplying the raw p-value from both significance tests by the number of regions tested for that chromosome. Another possibility is to apply \"Sign OR Mean Tests\", which results in a prediction being displayed if at least one of these criteria is met at the specified significance level. While we recommend the \"Sign AND Mean Test\" option for general use, other combinations may be useful under certain circumstances. Users can select any displayed prediction, which results in the genes or probes and associated annotation in that particular region to be displayed in the \"Results\" tab adjacent to the display panel (Figure [5](#F5){ref-type="fig"}). ### Exporting results Publication quality images can be exported in multiple formats at any stage of the visualization. This includes images of exclusively raw data, results of statistical analysis, or combinations of these. In addition, predictions resulting from automated or manual statistical analysis can be exported in tab-delimited text form with the associated gene or probe identifiers and corresponding p-values. A p-value filter similar to that described in \"Analyzing data\" can be applied to all exported results to allow full user control over which predictions are included. Finally, lists of genes or probes for any object selected on the display panel can also be exported to text files to facilitate immediate analysis of regions identified by manual inspection. ### Command-line usage ChARMView can also be used in command-line mode to make automated predictions of amplification or deletions. This command-line feature can be used by invoking ChARMView as follows: java -Xmx300m -jar ChARM.jar -inputFile \<input-file\> -outputFile \<output-file\> -organismType \<organism-type\> -meanPvalCutoff \<mean-pvalue-cutoff\> -signPvalCutoff \<sign-pvalue-cutoff\> -sigTestType \<significance-test-type\> The possible organism types, which determine reference chromosomal coordinates, are: 1, Saccharomyces cerevisiae; 2, human; 3, other (user-provided coordinates). Possible significance test options include: 1, mean AND sign tests; 2-mean OR sign tests; 3, mean test only; 4, sign test only. When run in command-line mode, ChARMView outputs all predicted regions of amplification and deletion meeting the specified significance level. Conclusion ========== We have developed ChARMView, a statistical visualization system for analysis and discovery of genomic aberrations. Our system can analyze various types of genomic data, including gene expression and array CGH microarray data, for a variety of organisms, and has been developed to facilitate both manual discovery through powerful visualization as well as automated prediction through robust statistical analysis. ChARMView can identify and visualize even small copy number changes, and is sensitive enough to detect aneuploidies in mixed populations of cells. This combination makes ChARMView uniquely effective for identifying subtle trends, recurring aberrations in sets of experiments, and pinpointing functionally relevant copy number changes. Thus, this system is effective for identification of aneuploidies in cancer studies and molecular evolution experiments, as well as for routine analysis of microarray data for special biases. Availability and requirements ============================= Project name: ChARMView Project homepage: <http://function.princeton.edu/ChARMView> Operating system(s): Platform independent Programming language: Java Other requirements: J2SE Java Runtime Environment 1.4.2 or higher License: GNU GPL Any restrictions to use by non-academics: None Authors\' contributions ======================= CLM developed the methodology, the software components, and performed case studies on example datasets. XC, together with CLM, developed an early version of the software. CLM and OGT drafted the manuscript. OGT conceived of the idea for ChARM View and directed the development. All authors read and approved the final version of the manuscript. Acknowledgements ================ The authors would like to thank Matt Hibbs for valuable input about interface design and creative naming suggestions, Kai Li and David Botstein and their groups for valuable discussions, and Maitreya Dunham for providing biological data and input about functionality. CLM is supported by the Quantitative and Computational Biology Program T32 HG003284-01. OGT is partially supported by NSF grant NGS-0406415.	PubMed Central	2024-06-05T03:55:59.754871	2005-6-13	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181623/", "journal": "BMC Bioinformatics. 2005 Jun 13; 6:146", "authors": [ { "first": "Chad L", "last": "Myers" }, { "first": "Xing", "last": "Chen" }, { "first": "Olga G", "last": "Troyanskaya" } ] }
PMC1181624	Background ========== Recently the number of research groups generating sequence data such as expressed sequence tags (EST) and bacterial artificial chromosome (BAC) ends has increased dramatically due to dropping costs of obtaining DNA sequence. The size of groups involved in such projects ranges from members of sequencing centers with a large bioinformatic support staff to individual researchers with little data management expertise. As it is now typical for even the smallest projects to generate hundreds to thousands of sequences, tools that streamline annotation, data mining and data management have become increasingly important to biologists without large bioinformatic support staffs. Annotation of a large set of sequences typically involves attempting to subdivide the sequences into common functional categories. The usual course of action followed is to perform one or more BLAST searches against databases to reveal homologies with sequences previously functionally annotated. Using these homologies, categorical annotations such as Gene Ontology terms can be associated with the novel sequences and used to evaluate the sequence dataset. When dealing with the large numbers of sequence reads generated by these projects, storage and compilation of flatfile BLAST results related to this endeavor becomes rather cumbersome. Furthermore, comparative analysis of gene complement between related organisms can provide a number of insights into many areas of biology. Identification of a subset of genes which exist in a pathogen but not in a related non-pathogenic organism can provide targets for other functional analyses such as gene disruption experiments. While there exist other BLAST clustering packages, such as BeoBLAST or provided by NCBI, they do not provide a free open-source complete result management package with storage of results in a relational database (RDMS) \[[@B1],[@B7]\]. The ability to easily locate, data-mine and perform comparative analysis on BLAST results can be substantially simplified through the use of a RDMS. In order to facilitate sequence analysis and comparative genomics we have created NuclearBLAST. Implementation ============== The primary design goals for NuclearBLAST were to provide biologists a centralized system where BLAST results can be easily created and retrieved, a relational database storage system which can be easily mined for comparative analyses, and a program which would take advantage of clustered computing resources to increase the throughput of large BLAST jobs. A secondary objective was to design a solution which was open-source and freely available. This led us to preferentially utilizing a number of open-source software packages in the implementation of NuclearBLAST including BioPerl, Apache, PHP and MySQL as well as using Linux as the base operating system \[[@B2]-[@B5]\]. A web browser was chosen as the interface for NuclearBLAST since it provides a widely used, recognizable, globally available, platform independent interface. The Apache web server provides access control and encrypted connections if the user desires to secure their installation. BioPerl provides a parsing module for the returned BLAST results which are then loaded into the MySQL database. All information about requested BLAST searches are loaded into the database and the status of each query in a job is tracked in the database. This client server design of the program allows for the use of multiple computers performing the BLAST analyses in a clustered environment by using a job management software package such as PBS (Portable Batch System) \[[@B6]\]. NuclearBLAST can be clustered with as little as several lab computers used as worker nodes in their down time or as much as a dedicated compute farm. A minimal installation of NuclearBLAST requires a typical workstation machine acting as both the server and the worker. In order to keep the MySQL database to a manageable size and reduce redundant data in the system we opted to use BLAST\'s database format as the only store of sequence information in the program. Only the sequence names within each dataset and a minimal amount of metadata are stored in the MySQL database. We also minimized the database size by storing all result statistics in the database but excluding the actual alignments. When requested, the alignments are recreated on the fly by extracting the two sequences from the BLAST databases and blasting them against each other using bl2seq. Results ======= NuclearBLAST is a free open source batch BLAST analysis, storage and management system which provides a simple platform for performing BLAST analyses and mining of the results. NuclearBLAST is written in Perl which drives NCBI\'s blastallapplication to conduct its analyses and store all BLAST results in a MySQL database \[[@B7]\]. Users may import datasets, request analyses, and view results in the PHP web interface through their browser. Command-line programs support more complex or special-purpose queries. NuclearBLAST may be installed on a single machine, or it may be set up to distribute BLAST analyses to a cluster of machines by way of a batch queuing system. Sequence files are imported in FASTA format. NuclearBLAST uses NCBI\'s formatdbutility to parlay sequences into the format required for blastalltargets. When importing a sequence data set into NuclearBLAST, a user specifies whether the sequence may be used in subsequent searches as a query, a target, or both. Specifying \'both\' allows easy reciprocal BLASTs between smaller datasets. These parameters can be set to allow users access to large datasets as targets but not as queries. This guards against the casual mistake of querying very large datasets, such as the Genbank non-redundant protein database, against much smaller datasets (rather than vice versa). If unnoticed, such a mishap can divert a laboratory\'s analysis system for extended periods of time. Requesting batch BLAST jobs through the web interface is facilitated by a job request wizard which guides the user through the process. The user first chooses a query dataset (from datasets designated as allowable queries) (Fig [1a](#F1){ref-type="fig"}). The choice of query (specifically, whether it is a nucleotide or protein sequence) automatically limits the menu of programs of the BLAST \"family\" (BLASTN, TBLASTX, etc) to those appropriate for the type of query sequence (Fig [1b](#F1){ref-type="fig"}). The user\'s choice of program dictates the sequence type (nucleotide/protein) appropriate for targets; in the next stage of the wizard, when the user selects a target sequence dataset, only choices of the appropriate type appear on the menu (Fig [1c](#F1){ref-type="fig"}). The job request is completed by specifying an e-value threshold for returned results. (More advanced BLAST parameters can be stipulated by using NuclearBLAST\'s command-line request utility). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Screenshots of the NuclearBLAST Job Request Wizard. Panel ashows the set of possible query sets in the system. Upon choosing one of these, the appropriate BLAST sub-programs are made available in Panel b. After choosing one of the sub-programs, Panel carises which allows you to limit the e-value of stored results and gives choices of BLAST target databases in the system which are acceptable based upon prior decisions. ::: ![](1471-2105-6-147-1) ::: After a request is submitted, the status of the job can be continually monitored on the job\'s main page (Fig [2a](#F2){ref-type="fig"}). Invocations of blastallare scheduled by a simple first-in-first-out scheme. NuclearBLAST parses blastall\'soutput files using BioPerl and loads their constituent data into the database \[[@B8]\]. The work flow of a job from request to completion can be seen in Figure [3](#F3){ref-type="fig"}. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Screenshots of the NuclearBLAST web interface. Panel ashows the requested jobs and their progress. Clicking on a hyperlinked job number brings up a page, panel b, containing the results of multiple query sequences within that requested job. The link on this page showing the number of hits fetches the location and statistics of multiple hits to that query sequence, panel c. Examining a hit further provides a view of the HSP locations and statistics for a single hit, panel d. Hyperlink of a sequence name retrieves a page with information about that particular sequence. ::: ![](1471-2105-6-147-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Flow of a NuclearBLAST Job. Illustration of the program\'s work flow when progressing through a requested BLAST analysis ::: ![](1471-2105-6-147-3) ::: Results of a completed batch BLAST job can be browsed as ordered by e-value in a paginated fashion with the top hit available on the main result page (Fig [2b](#F2){ref-type="fig"}). Clicking on the query calls up a page which shows all hits to that sequence found, along with a graphical view of where each hit aligned to the sequence (Fig [2c](#F2){ref-type="fig"}). Each hit further contains a link to a page showing the alignment of the HSPs and other associated statistic (Fig [2d](#F2){ref-type="fig"}). In addition to ordered hierarchical browsing, NuclearBLAST offers a facility for searching for strings within text fields associated with the job (names or descriptions of queries or hits). This can help researchers quickly find certain selected targets amidst a large volume of search results. In addition to providing an easy web-based method of perusing BLAST results, NuclearBLAST supplies a number of command line scripts allowing expanded access to search results in the database. One such script exports a job\'s results to a tab delimited file for inclusion in a spreadsheet or publication. Another determines reciprocal best hits between two datasets that have been reciprocally BLASTed. A third script may be used to transfer Gene Ontology (GO) annotations to query sequences that have matches against a database annotated with these terms. As Compugen has made available GO annotations for much of the Genbank non-redundant protein database at the Gene Ontology website, one can easily annotate a large set of sequences with GO terms using NuclearBLAST \[[@B9],[@B10]\]. This provides the researcher an easy way to categorize their sequences into functional groupings. Future work on NuclearBLAST will extend the mining capabilities available on the command line as well as through the web interface and also expand integration with clustering software. Conclusion ========== NuclearBLAST provides a powerful tool to biologists for data mining and comparative genomic analysis of generated sequence. It has shown its utility in prior studies \[[@B11]\]. This program provides a simple interface for performing large batch BLAST searches, effectively manages a large number of search results, presents those results in an intelligibly browsable format, and provides an extensible platform for more thorough data mining of BLAST results. Availability and requirements ============================= The program, source and full documentation for installation and use are available at <http://www.alkahest.org> as well as in the additional file section \[see [Additional File 1](#S1){ref-type="supplementary-material"}\]. Installation of the software is walked through in the documentation and requires personnel with Unix/Linux installation experience. The software is licensed under the GNU GPL and requires PHP, Perl, Apache, and MySQL. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 The program, source and full documentation for installation are included. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We thank Michael Thon, Patsy Little, members of the Center for Integrated Fungal Research and the Plant Nematode Genetics Group for testing and feedback. This work is supported in part by an NSF IGERT fellowship to S.E.D. through the Functional Genomics graduate program, NSF Plant Genome program (\#0115642), UNC office of the President (UNC RA FY 2002), Genencor International, Phillip Morris, and NCAES.	PubMed Central	2024-06-05T03:55:59.757480	2005-6-15	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181624/", "journal": "BMC Bioinformatics. 2005 Jun 15; 6:147", "authors": [ { "first": "Stephen E", "last": "Diener" }, { "first": "Thomas D", "last": "Houfek" }, { "first": "Sam E", "last": "Kalat" }, { "first": "DE", "last": "Windham" }, { "first": "Mark", "last": "Burke" }, { "first": "Charles", "last": "Opperman" }, { "first": "Ralph A", "last": "Dean" } ] }
PMC1181625	Background ========== Microarray technology has provided biologists with the ability to measure the expression levels of thousands of genes in a single experiment. The vast amount of raw gene expression data leads to statistical and analytical challenges including the classification of the dataset into correct classes. The goal of classification is to identify the differentially expressed genes that may be used to predict class membership for new samples. The central diffculties in microarray classification are the availability of a very small number of samples in comparison with the number of genes in the sample, and the experimental variation in measured gene expression levels. While very effective methods for binary classification (i.e. classification into two classes) are known, these methods do not necessarily perform as well in the multi-class case \[[@B4]\]. This paper addresses the multi-class classification of microarray data, and the evaluation issues that arise in determining the validity of the performance measures. The classification of gene expression data samples involves feature selection and classifier design. Feature selection identifies the subset of differentially-expressed genes that are potentially relevant for distinguishing the classes of samples. The aim is to reduce the initial gene pool from 7,000--10,000 to 100--200. Several gene selection methods based on statistical analysis have been developed to select these predictive genes, they include t-statistics, information gain, twoing rule, the ratio of between-groups to within-groups sum of squares (BSS/WSS) and principal component analysis \[[@B4],[@B5]\]. In this study we explore the alternative methods provided by the RankGene software \[[@B3]\] for the initial feature selection task. Both supervised and unsupervised classifiers have been used to build classification models from microarray data. This study addresses the supervised classification task where data samples belong to a known class. Many classifiers have been used for this task such as Fisher Linear Discrimination Analysis, Maximum Likelihood Discriminant Rules, Classification Tree, Support Vector Machine (SVM), K Nearest Neighbour (KNN), and aggregated classifiers \[[@B4]\]. In this study we adopt the KNN classifier. KNN classification is based on a distance function such as the Euclidean distance or Pearson\'s correlation that is computed for pairs of samples in N-dimensional space. Each sample is classified according to the class memberships of its k nearest neighbours, as determined by the distance function. KNN has the advantages of simple calculation and the ability to perform well on data sets that are not linearly separable, often giving better performance than more complex methods in many applications (e.g. \[[@B4]\]). The aim of this study is to evaluate an evolutionary algorithm for multiclass classification of microarray samples by assessing its classification accuracy on microarray samples. We also investigate the feature selection step that is a necessary precursor to classification. These objectives require an appropriate evaluation method to determine the final figures for accuracy. Once the appropriate parameters for the evolutionary algorithm are determined, its performance is evaluated again using the .632 bootstrap estimation method to obtain a low-variance measure. Two published microarray datasets are used to test the performance of the algorithms, namely, the leukemia and NCI60 datasets. The contributions of this paper are: a comprehensive evaluation of an evolutionary classifier; an investigation of feature selection in learning classifiers; an analysis of frequently selected genes, and a comparison of gene rankings across several previous studies of the leukemia data. Systems and methodology ======================= Evolutionary algorithm ---------------------- Evolutionary algorithms have been applied to microarray classification in order to search for the optimal or near-optimal set of predictive genes on complex and large spaces of possible gene sets. Evolutionary algorithms are stochastic search and optimisation techniques that have been developed over the last 30 years. These algorithms are based on the same principles of evolution found in the biological world involving natural selection, and survival of the fittest. Evolutionary algorithms differ from other traditional optimisation techniques in that they involve a parallel search through a population of solutions. The evolutionary algorithm we employ maintains a population of predictors whose effectiveness as a classifier can be determined by using them as features in a KNN classifier. The size of the population and the number of features in a predictor are parameters that we shall explore experimentally in the following section. We assume that an initial gene pool of informative genes has been identified (GP). The initial predictors in the population are randomly constructed from the initial gene pool as indicated in Figure [1](#F1){ref-type="fig"}. A predictor contains between 10 and 50 genes and so defines a subset of the features (genes) that are identified in the initial feature selection step. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Evolutionary algorithm for multiclass classification. ::: ![](1471-2105-6-148-1) ::: Each predictor is scored according to its ability to classify the training data based on a leave-one-out cross validation (LOOCV) analysis. The scoring function Sin steps 3b and 3c in Figure [1](#F1){ref-type="fig"} is the sum of the training samples that are correctly classified in the LOOCV. That is, one of the samples is left out to be a pseudo test data, the classifier is built based on all but the left out sample, and the score incremented if the sample is correctly classified. The score is incremented by an additional amount proportional to the minimum separation of the clusters. The replication of a particular predictor depends on how the mutation effects the scoring function. The selection process uses a statistical replication algorithm. The predictors which have a higher score after mutation survive in the next generation. The termination condition is met when the all predictors give similar score over a specific number of generations. The most successful predictor is the one giving the fewest errors on the training samples. To measure the performance of the algorithm, the best predictor is evaluated again using test samples. The parameters of the evolutionary algorithm are fixed as follows: The probability that a predictor will be mutated is set to 0.7. The probability of adding a new gene to the predictor and the probability to delete a gene from any random position in predictor are set to 0.5. This makes adding a gene and deleting a gene equally probable. The KNN classifier uses the Euclidean distance measure. When surveying the performance for the population size and feature size parameters we report the average of the accuracy rate over 10 trials. The selection and evolution process is a modification of the statistical replication algorithm described in Deutsch \[[@B6]\]. The termination criteria is defined using both the maximum number of generations and the criteria of no improvement of maximum fitness value of the population. The predictor with highest fitness will be the one that contains the best subset of genes for the classification task. Genetic algorithm ----------------- In order to confirm the conclusions drawn from the experimental analysis, we repeat the classification task using a different search algorithm -- a standard genetic algorithm (GA) -- in combination with a k-nearest neighbour classifier. The GA encodes a weight for each feature in a chromosome, and maintains a population of chromosomes which are combined using crossover, and modified by mutation operators in a standard manner \[[@B7]\]. The GA explores the space of weights for each feature using a 1 bit representation, where the feature is assigned a weight of 0 or 1. In this configuration, the GA+KNN classifier performs feature selection and hence is directly comparable to the evolutionary algorithm. The GA was configured to have 20 chromosomes and was run for 100 generations in each trial. Methodology ----------- To investigate the initial feature selection step, and its effect on classifier learning, we apply six of the feature selection methods supported in the RankGene \[[@B3]\] software. These techniques have been widely used either in machine learning (e.g. information gain and Gini index) or in statistical learning theory (e.g. twoing rule, max minority, sum of variances). They quantify the effectiveness of a feature by evaluating the strength of class predictability when the prediction is made by splitting the full range of expression of a given gene into two regions, the high region and the low region. The split point is chosen to optimise the corresponding measure. Feature selection is performed using only the samples in the training data set. The evolutionary algorithm that we shall evaluate learns the optimal subset of features that classify the data, based on the initial set of selected features. In the trials reported here, the evolutionary algorithm will stop when the score of all predictors in the population give a standard deviation of less than 0.01 for ten consecutive generations, or the evolutionary algorithm reaches the maximum number of generations (200). Many evaluation methods has been investigated for small-sample error estimation. Typically, a microarray experiment provides a dataset of small size, and as a result the most commonly used method for error estimation is leave-one-out cross validation (LOOCV). The LOOCV error rate estimator is a straightforward technique and gives an almost unbiased estimator. A comparison of various error estimation methods is presented in \[[@B8]\], these include the resubstitution estimator, k-fold cross-validation, leave-one-out estimation, and bootstrap methodology. These methods were tested with many classifiers: linear discriminant analysis, 3-nearest-neighbour, and decision trees. For over all performance, the .632 bootstrap proved to be the best estimator in their simulations, but the drawback of this method is the computational cost in comparison to LOOCV. As LOOCV is almost unbiased and is fast, it is acceptable to use it for parameter analysis. We investigate the performance of the feature selection methods and evolutionary algorithm as follows: • The evolutionary classifier is tested on the leukemia dataset without the use of any feature selection method. • The RankGene methods are employed, and performance compared against the baseline and against each other for the leukemia data. • The GA-based classifier is evaluated on the leukemia data to validate these results. • The evolutionary classifier is tested on the leukemia and NCI60 datasets. • A ranking of frequently selected genes by Z-score is obtained, and the performance of the top ranked genes as a classifier is measured. • The .632 bootstrap error estimator is applied using the optimal parameters for all datasets. Results ======= We begin by demonstrating that the performance of the population of predictors improves on each iteration of the evolutionary algorithm. Figure [2](#F2){ref-type="fig"} shows the average scores in each iteration from several trials. The graph shows that the average score increases more rapidly over the first few generations in comparison with the final generations. The evolutionary algorithm typically converges and terminates in less than 50 generations. The runs terminate on different iterations depending upon when the termination condition is met. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### The average score in each iteration for several trails of the evolutionary algorithm on the leukemia dataset. ::: ![](1471-2105-6-148-2) ::: In the baseline test, the evolutionary algorithm is run on the entire set of genes (without applying any feature selection method) in order to obtain a baseline measure of performance. The baseline system is evaluated on the 7,070 genes of the leukemia dataset. The initial predictors in population are built by randomly selecting 10 genes to be an initial feature of the predictors. This means the evolutionary algorithm has to search for 10 optimal predictive genes set from the ![](1471-2105-6-148-i2.gif) possible subsets. Performance of the predictors is evaluated using KNN classifier to determine the LOOCV on training samples. After the best predictor is found in each generation, it will be tested again on test samples to give the performance based on an out-of-sample estimation. The KNN classifier classifies each sample in the test data using a database of all training samples. Table [1](#T1){ref-type="table"} reports the maximum and the average accuracy of the baseline system on 38 training samples and 34 test samples. The results show that the evolutionary algorithm gives predictors with perfect classification on the training samples but those predictors do not classify the test data well. The average accuracy on test data is 68% at best while the average accuracy on training data is up to 98%. Table [1](#T1){ref-type="table"} indicates that population size may be a more important factor than feature size for the baseline system. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### The accuracy of the baseline system built by randomly selecting genes from 7,070 genes in the leukemia dataset. ::: Population size Feature size Training data \[%\] Test data \[%\] ----------------- -------------- --------------------- ----------------- ------- ------- 10 30 89.47 85.79 79.41 64.41 50 89.47 84.21 76.47 63.53 30 30 94.74 98.42 85.29 68.82 50 97.37 96.05 73.53 67.35 50 30 100.00 98.42 85.29 70.29 50 100.00 98.42 85.29 72.64 ::: Building an informative initial gene pool by the RankGene software ------------------------------------------------------------------ We determine the best choice for the number of initial features, i.e. the initial gene pool, to be 100 by experiment (by exploring the performance of the classifier using the information gain ranking method for feature selection). Based on this result, the performance of the following six ranking methods for a range of population and feature sizes is investigated: R1. Information gain; R2. Twoing rule; R3. Gini index; R4. Sum minority; R5. Max minority; R6. Sum of variances. The details of each method can be found at <http://genomics10.bu.edu/yangsu/rankgene/>. Leukemia dataset ---------------- The parameters of the evolutionary classifier are evaluated on the set of 38 training samples and the 34 test samples of the leukemia data -- as described in the original work. We explore: population size {10, 30, 50} feature size {30, 50} and initial gene pool 100. The classification results are summarised in table [2](#T2){ref-type="table"}. When a predictor with 100% accuracy is learned in one or more of the test runs we indicate this by (\). Average accuracy over ten trials lies in the range 92--98%. Using feature selection, the accuracy on the test data is more than 19% greater than that of the baseline system. For a given feature selection method, the prediction accuracy on the test samples varies up to 3% across the range of parameter settings surveyed. The choice of feature selection method contributes up to 5% for a given set of algorithm parameters. Information gain consistently gives the best performance on the leukemia dataset. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### The average accuracy using out-of-sample prediction on the 34 leukemia test samples. The symbol (\) means that there is some perfect predictors found by the algorithm. The highest accuracy is written in bold. ::: Population size Feature size The accuracy of different rank methods on the Test data (out-of-sample) \[%\] ----------------- -------------- ------------------------------------------------------------------------------- ------- --------- --------- --------- --------- 10 30 97.35\* 95.29 93.82 92.94\* 93.53 94.12 50 98.24\* 95.59 94.71\* 93.82 93.53 95.00 30 30 96.74\* 92.65 94.41 95.00\* 94.71 93.82 50 97.06\* 95.00 95.00 93.82 95.30 93.82 50 30 97.35\* 93.82 94.71\* 92.06\* 94.12\* 93.82\* 50 96.17 93.82 92.65 92.35 94.12 94.71 Abbreviations: R1. Information gain; R2. Twoing rule; R3. Gini index; R4. Sum minority; R5. Max minority; R6. Sum of variances. ::: The results of the confirmation study are shown in table [3](#T3){ref-type="table"}, where feature selection improves accuracy by at least 27%: The accuracy on the training data in the baseline system (without feature selection) is 85.26% which is significantly greater than the test accuracy of 67.06%. A variation of 4.4% is observed in average accuracy across the RankGene methods. The best RankGene method is information gain. In comparison with feature selection by information gain, all other methods have a significantly lower accuracy on the testing data than information gain, while there are no differences in performance on the training data across the other RankGene methods. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### The average accuracy of the GA/KNN classifier using out-of-sample prediction on the 34 leukemia test samples. ::: The accuracy of different rank methods on the Test data (out-of-sample) \[%\] ------------------------------------------------------------------------------- ------- ------- ------- ------- ------- --------- Baseline R1 R2 R3 R4 R5 R6 67.06 98.24 95.29 93.82 95.00 94.71 95.00\* ::: NCI60 dataset ------------- The aim of this investigation is to find the best parameters and ranking method for the evolutionary classifier when applied to the NCI60 data. The same range of parameters is surveyed as for the leukemia data, and again the performance of six feature selection methods is evaluated. Due to the very small sample size of the NCI60 dataset, it is not possible to divide the data into training and testing sets. Thus, the accuracy of predictors in table [4](#T4){ref-type="table"} is given by using the LOOCV error rate estimation on the whole dataset. To get a more reliable performance of the evolutionary algorithm on the NCI60 dataset, the .632 bootstrap estimator will be used. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### The average accuracy on 5 sets of parameters and six ranking methods on the NCI60 data. No perfect predictors are found by the algorithm. The highest accuracy is written in bold. ::: Population size Feature size The accuracy of different rank methods on all dataset (LOOCV) \[%\] ----------------- -------------- --------------------------------------------------------------------- ------- ------- ------- ------- ------- 10 30 66.72 63.77 60.00 54.43 62.78 69.34 50 67.86 62.78 61.63 52.62 62.62 65.90 30 30 76.23 72.29 72.02 65.90 74.26 75.41 50 73.44 72.46 71.15 63.11 73.44 73.93 50 50 75.08 72.29 71.96 71.97 73.77 74.16 ::: The best classification score on the NCI60 data is 76.23%, and was obtained using information gain, with a population and feature size of 30. No predictor learned in any run of the system was able to classify all data 100% correctly. Discrimination method --------------------- The frequency of selection of the genes that are members of the best predictor across 100 independent trials is assessed in order to determine the reproducibility of the results. If a gene is consistently preferentially chosen as a member of a predictive set it would suggest that the gene selection operation is reproducible -- despite the random initialisation. Z-score analysis is one means to determine the significance of the observed frequency of an event against that which might have occurred by chance. This calculation normalises the frequency with which each of the initial genes was selected in all predictors that classify the training and test data perfectly \[[@B9]\]. The Z score can be calculated using equation (1). ![](1471-2105-6-148-i3.gif) Where S~i~denotes the number of times genes iwas selected, E(S~i~) is the expected number of times for gene ibeing selected, and σdenotes the square root of the variance. The calculation of E(S~i~) is as follows: let Anumber of perfect predictors found in the experiment, P~i~= (number of genes) / (number of genes in the initial gene pool). Then, E(S~i~) = P~i~\* A. The evolutionary classifier was run 100 times on the leukemia dataset using the best set of parameters: 100 initial genes constructed by the information gain and feature size = 50. There are 43 predictors (ignoring the duplication) that classify all training and test data correctly. Figure [3](#F3){ref-type="fig"} shows a plot of Z-score applied to the top ranked genes that are most frequently selected. The top 24 genes have a similar Z-score and the top 55 have a positive Z-score. In this case, it seems reasonable to choose the 55 top-ranked gene as the most discriminative genes. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### A plot of Z-scores for 100 ranked genes on the leukemia dataset. ::: ![](1471-2105-6-148-3) ::: To confirm the predictive power of the 55 top-ranked genes we constructed a classifier using these genes as features. As before, we measure the accuracy on the test set using the training set as a case-base. An accuracy of 100% is obtained. Additional file: [1](#S1){ref-type="supplementary-material"} shows the 55 top-ranked genes, their frequency of occurrence in the 43 best predictors and the Z-score value. This set of genes will be used in the further evaluation using .632 bootstrap estimator. The Z-score analysis was repeated on the NCI60 data and these results are shownin Figure [4](#F4){ref-type="fig"}. It can be seen that the Z-score of the 8th ranked gene is half that of the top ranked gene and that the top 40 genes have a positive Z-score. Additional File [2](#S2){ref-type="supplementary-material"} lists the top 40 genes. A classifier constructed using these genes has an accuracy of 73.8% by LOOCV, and 68.2% by a bootstrap estimate. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### A plot of Z-scores for 100 ranked genes on the NCI60 dataset. ::: ![](1471-2105-6-148-4) ::: .632 bootstrap estimates ------------------------ The .632 bootstrap method involves sampling the original dataset (with replacement) to obtain a new resampled set on which the classification error δis measured \[[@B8]\]. The accuracy εon the samples omitted from the resampled set is also determined. The bootstrap estimator a~b632~for a dataset of nsamples requires nresampled datasets to be constructed and the classification error calculated according to equation (2). The final bootstrap estimate is the average value of a~b632~over biterations of the procedure (we take b= 200). ![](1471-2105-6-148-i4.gif) On the leukemia data, the .632 bootstrap estimate for the accuracy of the evolutionary algorithm with population size 10 and feature size 50 is 96.40%. The .632 estimate for a classifier based on a fixed set of 55 top-ranked genes is 96.85%. On the NCI60 data, the .632 bootstrap estimate for the accuracy of the evolutionary algorithm with population size 30 and feature size 30, is 59.19%. The .632 estimate for a classifier based on 40 top-ranked genes is 68.18%. Discussion ========== On the data sets studied, we find that the evolutionary algorithm performs robustly over the space of parameters surveyed. Optimal values for population size and feature size are obtained, but the approach is not overly sensitive to these choices. This is a promising result as stochastic algorithms and techniques such as neural nets can be sensitive to parameter settings \[[@B10]\]. Despite showing approximately 97% accuracy on the leukemia data, the evolutionary algorithm does not perform as well on the NCI60 dataset. The .632 bootstrap estimate of 59% for the NCI60 data is significantly less than the 76% obtained by LOOCV. If we accept that the bootstrap method is appropriate, then we must conclude that the LOOCV estimates typically reported may be considerable overestimates. The LOOCV estimate we obtain is greater than the averages for the NCI60 data reported in \[[@B4],[@B5]\], but less than the best error rate reported in \[[@B11]\]. The bootstrap estimate is comparable with the results of \[[@B5]\]. Gene selection -------------- The value of the RankGene methods for feature selection is shown by the improvement in accuracy over the baseline results. Without feature selection, the evolutionary algorithm can discover predictors that give 100% accuracy on the training set, but perform poorly on the test set. A similar finding is obtained using a GA-based search with a KNN classifier. This indicates that the classes can be distinguished by any of a large set genes that are indicative of a category, but that these genes are not necessarily informative in the sense that they are activated in a comparable way across both the training and the testing sets. By allowing uninformative genes into the feature set we observe a degree of overfitting. The correct choice of RankGene method can improve classification by 5% for a given population size and feature size on the leukemia data (when using the evolutionary algorithm). The confirmation study reproduced this finding also. In comparison, the selection of population and feature sizes can influence performance by up to 3% (for a given RankGene method). Thus, feature selection has a significant influence on the classifier learning task. The optimal predictor --------------------- Several researchers have tried to find the optimal predictive gene set. It has been suggested that for the leukemia dataset the number of predictive genes included in a predictor should be less than 50 \[[@B12]\]. This conclusion was based on a Z-score analysis. Deutsch \[[@B6]\] found an average of 9 predictors was sufficient to classify the leukemia data perfectly, with some runs reducing the number to 2. We find 10 genes are not sufficient, and that that up to 55 genes are required, but have not attempted to identify a lower bound. Gene ranking ------------ For comparison, the rankings of selected genes found to be significant in previous analyses \[[@B13],[@B14]\] of the leukemia data are listed in Table [5](#T5){ref-type="table"} along with those ranked highly by our methods. The ranking of Ben-Dor \[[@B13]\] is of genes that discriminate between AML and ALL, while Thomas \[[@B14]\] ranks the 25 genes more highly differentially expressed genes in AML than in ALL, and provides a similar ranking for the genes differentially expressed in ALL. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Top ranking genes by Z-score (top 24 genes, this paper), by TNoM score (top 30 genes \[13\]^1^), and by differential expression in AML or in ALL (top 25 genes, \[14\]^2^). As the Z-score can give genes equal scores, a rank of 1 can be assigned to several genes. In our study 12 genes are ranked 1, in that of \[13\] 3 are ranked 1. The lowest place in a total ordering of the genes is indicated by figure in parenthesis, e.g. (≤12). Genes ranked in column 5 cannot appear in column 6, and vice versa, as indicated by (-). Otherwise, a blank entry indicates a ranking outside of the top 24, 25 or 30 genes respectively. ::: Gene Description Rank by Z-score Rank by TNoM^1^ Rank in AML^2^ Rank in ALL^2^ -------- ------------------------------------ ----------------- ----------------- ---------------- ---------------- M23197 Myeloid cell surface antigen CD33 1 (≤12) 1 (≤3) 22 (≤22) \- X04145 T-cell surface glycoprotein CD3 1 (≤12) M31211 Myosin light chain 1 (≤12) 6 (≤20) \- 4 (≤4) M31303 Leukemia-associated phosphoprotein 2 (≤13) 7 (≤32) \- 11 (≤13) U50136 Leukotriene C4 3 (≤24) 7 (≤32) 3 (≤3) \- M28170 B-lymphocyte surface antigen CD19 3 (≤24) J04132 T-cell surface glycoprotein CD3 3 (≤24) X95735 Zyxin 1 (≤3) 6 (≤6) \- M55150 Fumarylacetoacetate 6 (≤20) 1 (≤1) \- X59417 Proteasome iota chain 6 (≤20) \- 3 (≤3) U22376 C-myb \- 1 (≤1) ::: Table [5](#T5){ref-type="table"} indicates that the genes that rank highly by differential expression lie outside of the top 15 in the rankings that are derived to support classification, and some rank outside of the top 30 (as indicated by a blank entry in the table). For example, M23197, which is the human differentiation antigen CD33 and a known indicator of myeloid lineage \[[@B15]\], is ranked first the classification-based scores but is 22nd in terms of differential expression. Only four of the top 24 genes we identify are found in the 50 genes listed as having highest relative differential expression for AML/ALL respectively. In contrast, 14 of the top 30 genes of Ben-Dor \[[@B13]\] are in this list (only a selection are tabulated here). This difference can be accounted for by the fact that we address the 3-class problem while Ben-Dor solve the 2-class problem. Thus it appears that the differential expression of genes can be exploited in classifying samples but, even in the 2-class case will, not lead to optimal classification as it ignores the relationships between the expression levels in a predictive set of genes. Five of the genes we rank in the top 3 (range 1--24 in terms of a total ordering of genes) also lie in the equivalent range (rank 1--6, or range 1--20 in a total ordering) in \[[@B13]\]. Further, the human CD19 antigen gene M28170 and T-cell surface glycoproteins XO4145 and J04132 which are ranked in the top 24 in our study do not appear in the top 50 genes in Ben-Dor \[[@B13]\] or in Thomas \[[@B14]\]. CD19 is known to indicate B-lineage and CD3 to indicate T-lineage \[[@B15]\] This study confirms that significantly different sets of genes are found to be most discriminatory as the sample classes are refined. Conclusion ========== The evolutionary methods we have developed for microarray data classification perform robustly and accurately on the data sets examined. The results are in accord with clinical knowledge as demonstrated by a Z-score analysis of the genes most frequently selected for inclusion in a classifier: Genes known to discriminate between AML and Pre-T ALL leukemia are identified. This study also concludes that there are notable dependencies between the way in which the classification problem is formulated and the resulting rankings of discriminatory genes. Methods ======= The 3-class leukemia dataset \[[@B1]\] and the 9-class tumour NCI60 dataset \[[@B2]\] are used in this study. The leukemia dataset is available at <http://www.genome.wi.mit.edu/MPR>. In this dataset, gene expression levels were measured using Affymetrix high-density oligonucleotide arrays containing 6,817 genes. The dataset consists of 47 samples of acute lymphoblastic leukemia(ALL) and 25 samples of acute myeloblastic leukemia(AML). Originally, the dataset was built and analysed for binary classification: ALL and AML. However, it can be separated into three or four classes by using subtypes of ALL. To perform a multiclass classification task, 72 samples in the dataset are divided into three classes: ALL B-CELL(38), ALL T-CELL(9), and AML(25). The training set is composed of 27 samples of ALL and 11 samples of AML, and the test set is composed of 20 ALL and 14 AML samples. Golub et al.\[[@B1]\] have normalised the dataset by re-scaling intensity values to make the overall intensities for each chip equivalent and also fitted the data with a linear regression model. From 7,129 genes, the baseline genes were cut off before further analysis. The number of genes that are used in the multiclass classification task is 7,070. The NCI60 dataset is available at <http://genome-www.stanford.edu/sutech/download/nci60>. This dataset contains the gene expression profiles of 64 cancer cell lines measured by cDNA microarrays and is provided without normalisation. The gene expression data value is the relative intensity level of the mRNA samples and their references. The single unknown cell line and two prostate cell lines were excluded from analysis due to their small number. Nine classes of samples (61 cell lines) are identified: breast (7), central nervous system (5), colon (7), leukemia (6), melanoma (8), non-small-cell-lung-carcinoma or NSCLC (9), ovarian (6), renal (9) and reproductive (4). Excluding negative and missing values, the number of genes considered is reduced from 9,703 to 7,375 genes. Due to the noisy nature of the datasets resulting from microarray experiments, preprocessing is an important step. The NCI60 dataset has to be normalised to decrease the variation before feature selection or classification. Global normalisation is used to eliminate systematic variance. For NCI60 dataset, the analysis of relative gene expression level is done by using the log-ratios between a certain gene (labelled in red or Cy5) and a reference gene (labelled in green or Cy3) before the data is normalised. The normalised values Mare given by (3) where the constant cis estimated from the mean for the log-ratio log~2~(R/G). M= log~2~(R/G) - c (3) Authors\' contributions ======================= TJU designed and implemented the evolutionary classifier, and carried out the experiments. SA worked with TJU on the design of the study and helped to draft the manuscript. Both authors read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional file 1 The 55 top-ranked leukemia genes ordered by the frequency that the gene is selected. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional file 2 The 40 top-ranked NCI60 genes ordered by the frequency that the gene is selected. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ This work is supported by the Royal Thai government. The second author is supported by BBSRC grant BBSRC 15/BEP 17046, and under the Advanced Knowledge Technologies (AKT) Interdisciplinary Research Collaboration (IRC), which is sponsored by the UK Engineering and Physical Sciences Research Council under grant number GR/N15764/01. The AKT IRC comprises the Universities of Aberdeen, Edinburgh, Sheffield, Southampton and the Open University.	PubMed Central	2024-06-05T03:55:59.759427	2005-6-15	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181625/", "journal": "BMC Bioinformatics. 2005 Jun 15; 6:148", "authors": [ { "first": "Thanyaluk", "last": "Jirapech-Umpai" }, { "first": "Stuart", "last": "Aitken" } ] }
PMC1181626	Background ========== DNA microarrays have been applied with much success to study genomic patterns of gene expression across many organisms. It has become widely acknowledged that to extract hypotheses from these data, there are advantages to the integration of orthogonal sources of information, notably, molecular-interaction data \[[@B1]\]. Hypotheses derived from genomic-expression data typically involve pathways of metabolic and molecular information flow, and complex cellular processes and structures, formed by multiple interacting molecules. However, commonly these molecular interactions are gleaned ad hocfrom the literature. In model organisms such as Saccharomyces cerevisiae, integrative systems-biology approaches to genomic-expression analysis have developed and employed sophisticated methods for the computational extraction of biological knowledge. Examples include: biological module identification and abstraction \[[@B2]\]; discovery of regulatory networks \[[@B3],[@B4]\]; and identification of active pathways in networks \[[@B5]\]. A hallmark of these advanced methods is the integration of diverse genome-scale data sets, in particular, the combination of genomic-expression data and molecular-interaction data. Another common characteristic of these methods is the use of graphs (vertices and edges, or nodes and links) to represent such integrated data. Graphical methods are highly intuitive. Also, the formalism of the graph facilitates the development and application of graph algorithms and machine-learning techniques to extract information. In studies of human disease, a limited repertoire of computational techniques, including ANOVA, hierarchical clustering, and discriminant analysis, has been applied to extract information from genomic-expression data derived from human tissues. Until recently, a critical barrier has been a lack of large-scale machine-readable sources of high-quality human molecular interaction data. Using a combination of artificial-intelligence methods and expert human curation, several efforts have made substantial progress in amassing, from the literature, databases with large numbers (greater than 14000) of human molecular interactions. These include the Human Protein Reference Database (HPRD) \[[@B6],[@B7]\], the Biomolecular Interaction Network Database (BIND) \[[@B8],[@B9]\], the Database of Interacting Proteins (DIP) \[[@B10],[@B11]\], and the Transcription Factor Database (Transfac) \[[@B12]\]. Thus, the bottleneck has now shifted to the efficient integration of these data to enable the application of advanced network-based analysis and modelling methods. For this work, we have implemented solutions to this bottleneck and applied them to a set of genomic-expression data derived from biopsies of human liver tissue infected with Hepatitis C Virus (HCV) \[[@B13]\]. About 3% of all humans are infected with HCV \[[@B14]\], and currently no vaccine exists. Chronic viral hepatitis C results in liver fibrosis and cirrhosis in about 20% of those infected \[[@B15]\]. Liver transplant is often required. Specifically, we have developed two software tools, InteractionFetcherand CytoTalk, that function as plug-ins for Cytoscape, an open-source, platform-independent environment for the visualization and analysis of biological networks \[[@B16],[@B17]\]. InteractionFetcherand CytoTalksimplify the integration and analysis of interaction data (and other data types) with genomic-expression data. To demonstrate their utility, we applied them to generate and analyze a large network of human molecular-interaction pathways that are putatively active during the infection of human liver tissue with HCV. Implementation ============== InteractionFetcher, a Cytoscapeplug-in ------------------------------------------ InteractionFetcherdynamically retrieves remote biological information for selected nodes in the current network within Cytoscape. The plug-in requests biological data via the XML-RPC protocol \[[@B18]\] from a remote server, which retrieves the requested information from an SQL database and passes it back to the plug-in. The plug-in then adds the retrieved information to the current network as additional nodes, edges, and/or attributes. Currently implemented data types include: protein/gene synonyms, orthologs, sequences (gene/protein/upstream), and interactions/associations. Some of this information can be obtained via integrated queries. For example, retrieved gene/protein synonym information may be used to increase the number of molecular interactions that are found. Currently-available interaction-data sets include HPRD \[[@B6],[@B7]\], BIND \[[@B8],[@B9]\], DIP \[[@B10],[@B11]\], and several other predicted interaction and co-expression data sets \[[@B19]-[@B21]\]. Many options are available, including the ability to do cross-species queries, using ortholog information from Homologene \[[@B22]\] among species including H. sapiens, M. musculus, S. cerevisiae, C. elegans, and D. melanogaster. For example, if two proteins in H. sapienshave not been observed to interact, but both of their orthologs in S. cerevisiaeare known to interact, then an inferred interaction(also known as an interolog) can be added to the network. Moreover, the tool allows for easy viewing of the source database\'s web page or linked PubMed abstract(s) describing each fetched interaction. Because the source code for both the client and server of this plug-in are available, we hope that the capabilities of plug-ins such as these can be expanded by other researchers to include, for example, experimental data (such as mRNA expression levels), metabolic information, or functional annotations. Cytoscape, the InteractionFetcherand related plug-ins, plus all server-side software are open-source and may be obtained at our laboratory web site \[[@B23]\] or at the Cytoscapeweb site \[[@B17]\]. CytoTalk, a Cytoscapeplug-in -------------------------------- CytoTalkenables a Cytoscapeuser to dynamically interact with and manipulate the current network in a Cytoscapewindow from an external process. This plug-in runs an internal XML-RPC \[[@B18]\] server that enables the currently-displayed network and its various attributes to be manipulated from an external client that is XML-RPC-capable. Example clients may include Perl and Python scripts, scripts written in the Rstatistical language \[[@B24]\], UNIX shell scripts, C or C++ programs, or Java processes. It moreover expands the developmental possibilities of Cytoscapeplug-in developers by allowing other plug-ins to be written in these languages. The external process may be run on the same machine as Cytoscape, or anywhere else on an accessible network. The open-source CytoTalkand Cytoscapesoftware as well as example CytoTalkclients in Perl, Python, and Rmay be obtained at our laboratory web site \[[@B23]\] or at the Cytoscapeweb site \[[@B17]\]. Results and discussion ====================== Gene-expression data -------------------- For our study, we utilized expression data derived from 28 liver biopsies collected by \[[@B13]\] from 11 HCV-positive liver transplant patients, between 1 and 24 months post-transplant. Since roughly 50% of HCV+ liver transplant patients become re-infected during the two years after receiving their new livers \[[@B25]\], these biopsies provide a unique model for tracking the changes in gene expression during HCV infection \[[@B13]\]. To compare gene-expression patterns in liver tissue before and after infection with HCV, \[[@B13]\] collected 28 post-transplant liver biopsies, plus pre-transplant control biopsies, from 11 HCV+ liver-transplant patients. Liver biopsies were obtained at intervals of 3 to 6 months, between 1 and 24 months post-transplant \[[@B13]\]. These samples contain a mixture of cell types including hepatocytes, hepatic stellate cells, Kupffer cells (liver macrophages), in addition to various blood cells \[[@B13]\]. mRNA expression ratios of about 7000 genes were measured relative to a common reference pool of pre-transplant biopsies. Using Rosetta Resolver(R) software \[[@B26]\], the data were normalized and transformed to log~10~ratios, and p-values were computed for expression difference from the reference pool. The measurements showed a high degree of patient-to-patient variation. Most of the genes (5968) were significantly expressed (p\<10^-7^) in at least one of the 28 samples. The research of \[[@B13]\] involved genomic-expression data derived from human subjects. Construction of the molecular-interaction scaffold -------------------------------------------------- We sought to generate a network of molecular pathways that are active (either induced or repressed) in HCV-infected human liver cells. The effects of HCV infection are likely to be complex, and the presence of contaminating blood cells and mixtures of various cell types in the biopsy samples will add further complexity. In order to emphasize the network interface between viral molecules and human molecules, we initiated network construction with a small \"seed\" network of interactions among HCV-encoded molecules and between HCV-encoded and host-encoded proteins. Interaction data were curated from review articles (\[[@B27],[@B28]\], and references therein). The seed network also included the JAK-STAT interferon-response pathway that is known to play a role in the response to HCV infection \[[@B29]\]. This set comprised 106 interactions between 86 macromolecules (proteins and the viral RNA). The proteins were, when possible, cross-referenced to RefSeq protein identifiers \[[@B30],[@B31]\]. Figure [1](#F1){ref-type="fig"} shows the seed network visualized using Cytoscape\[[@B16],[@B17]\]. This network is available for exploration and analysis via Cytoscapeat our laboratory web site \[[@B23]\]. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Network of interactions among HCV-encoded molecules and host proteins.Triangular nodes represent HCV-encoded molecules. Host molecules are square nodes. Edges represent molecular interactions of several types: black for protein-protein, yellow for protein-DNA, light-green for phosphorylations, red for activations, dark-green for repressions, purple for covalent interactions, brown for methylations. Sources: \[14, 15 and references therein\]. ::: ![](1471-2105-6-154-1) ::: The seed network was expanded to a full \"scaffold\" network using the 5968 genes implicated by the genomic-expression data and large-scale molecular-interaction data sets in public databases by searching for interactions among the 5968 expressed genes and the molecules in the seed network. To automate the construction of the scaffold network, we implemented a Cytoscapeplug-in, InteractionFetcher, for dynamic retrieval of molecular interactions and binding partners via the Internet. InteractionFetcherrapidly adds interactions among molecules of interest in a network. In addition, it may be used to iteratively expand a network through \"in silico pull-down\" of molecules that are currently not present in the network but are known to interact with molecules that are present. Using this plug-in, we were able to integrate as many as 15,000 interactions among the proteins implicated by the HCV expression-data set and seed-network proteins (among the available interaction data sets, which include HPRD \[[@B6],[@B7]\], BIND \[[@B8],[@B9]\], DIP \[[@B10],[@B11]\], PreBIND \[[@B32],[@B33]\], and several other predicted and co-expression data sets; see Methods). However, for this paper, we restricted our search to individually curated human-only protein interactions from HPRD and BIND, resulting in a scaffold network of 4,592 unique interactions among 1,950 molecules (Figure [2](#F2){ref-type="fig"}). This network is available for exploration and analysis with Cytoscapeat our laboratory web site \[[@B23]\], which also allows for easy viewing of additional information provided by InteractionFetcher, such as each interaction\'s source database web page and PubMed abstract identifier(s). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Network of interactions among proteins implicated by genomic expression data.Genes were implicated by expression profiling of HCV-infected liver biopsy data \[13\]. The network of interactions was assembled from external databases HPRD \[6, 7\] and BIND \[8, 9\], and automatically integrated using InteractionFetcher. ::: ![](1471-2105-6-154-2) ::: Computational analysis of integrated gene-expression data and molecular-interaction data ---------------------------------------------------------------------------------------- A useful method of integrated analysis of expression data within an interaction network is the ActivePathsalgorithm \[[@B5]\]. This method identifies contiguous pathways or subnetworks that are active (induced or repressed relative to randomly selected subnetworks) in subsets of the expression data. We applied this algorithm, which is available as a Cytoscapeplug-in, to the scaffold network. Due to the high order (number of vertices) and size (number of edges) of the scaffold network, it was necessary to iteratively apply the algorithm, as suggested by the developers, to obtain increasingly smaller active subnetworks until they contained fewer than 100 nodes. The resulting four active subnetworks contained between 40 and 121 interactions. Because there were overlaps among these four highest-scoring active subnetworks, we combined them into a single fully connected active subnetwork. Additional analyses were performed by selecting scaffold subnetworks that are significantly active and/or co-regulated in temporal subsets of the microarray data. Because the scaffold network is not differentiated with regard to tissues, cell types, or cellular state, and the biopsy samples from which the expression data were derived likewise contain mixtures of cell types and other contaminants, the information in the active scaffold network does not, by itself, answer the questions we are addressing. To increase our chance of identifying pathways that might be modulated in response to HCV infection, we performed a differential analysis of the scaffold network, to identify subnetworks that become active more than eight months after transplant. This choice of cut-off was made to nearly-evenly divide the expression data into two halves (those from biopsies prior to, and after, eight months post-transplant), and by performing a differential analysis we can hope to subtract out some of the effects of the transplant and post-transplant immune response signals from those of HCV reinfection and progression. We used the Rstatistical environment \[[@B24]\] to perform this analysis. Because Ris external to Cytoscape, we developed a plug-in, called CytoTalk, that enables a user of R(or a wide variety of other environments or languages; see Methods) to interactively query and modify Cytoscapenetworks, thereby greatly expanding the analytical capabilities available to users of Cytoscape. We used Rwith CytoTalkto select the proteins and interactions implicated by specific statistical queries on the expression data. This enabled us to extract a subnetwork of genes that were significantly induced or repressed with \\|log~10~(ratio)\\| \> 0.4 in biopsies obtained more than 8 months after transplant. The proteins encoded by these genes form a \"late-active\" network. We similarly extracted an \"early-active\" network encoded by genes that were active in biopsies obtained earlier than 8 months after transplant. We compared these two networks and identified an \"only-late-active\" subnetwork that was not active prior to eight months, but was active afterward. The expectation is that this \"only-late-active\" subnetwork will contain pathways from the \"late-active\" network that are activated in response to Hepatitis C virus re-infection, while pathways from the \"early-active\" network that may contain pathways activated as a result of the transplant are removed. In Figure [3](#F3){ref-type="fig"}, we have integrated the seed network, the composite active-paths network, and the \"only-late-active\" network into one network. This network is available for exploration and analysis in Cytoscapeat our laboratory web site \[[@B23]\]. Genes that were induced on average after 8 months following transplant are indicated with a red colour. Genes that were repressed are green. We have highlighted the nodes and edges of the composite active-paths subnetwork in \"bold\". The network in Figure [3](#F3){ref-type="fig"} is significantly over-represented with genes of several biological processes, as annotated by the Gene Ontology Consortium Database \[[@B34],[@B35]\]; using the BioDataServertool in Cytoscape, and computed in Rvia CytoTalk, using the Bonferroni-corrected hypergeometric distribution. Among these include blood coagulation (p= 10^-11^), immune response (p= 10^-7^), proteolysis and peptidolysis (p= 10^-5^), lipid transport (p= 10^-3^), and complement activation (p= 10^-2^). In addition, nearly the entire JAK-STAT interferon-response signalling pathway is activated in this network. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Composite network of molecular pathways active in HCV-infected liver tissue.The network in Figure 1 was combined with active subnetworks from the network in Figure 2. The active subnetworks were identified by active-paths analysis (\[5\]; bold nodes and edges) and by identifying the subnetworks that changed most significantly in expression with time after transplant. Nodes (genes) colored red were induced in the expression data of biopsies from 8 months or more post-transplant; green nodes were repressed. Areas that contain differentially active pathways or subnetworks, as described in the text, are highlighted. ::: ![](1471-2105-6-154-3) ::: The visualization in Figure [3](#F3){ref-type="fig"} enables one to identify these pathways and see whether they are \"turning on\" (red) or \"turning off\" (green) in the expression data. For example, the blood coagulation pathway is active in the expression data, although (as is to be expected with large and complex pathways) not coherently induced or repressed. The interferon-response pathway and genes activated by ISGF are clearly induced, probably due in part to the immune response to viral infection and partly in response to standard treatment of HCV-positive patients with interferon-alpha. Also, genes encoding the Toll-like receptors TLR1 and 2, as well as the downstream signalling pathway connecting them, through MYD88, to the interferon-response pathway appear to be repressed. TLRs 1 and 2 are known viral detection receptors; it is known that TLR2 detects HCV \[[@B36]\]. The interleukin receptor IL1R1, upstream of MYD88, is also repressed along with other IL receptors, whereas IL1A and B are induced. Additionally, we see that many apoptosis-related genes encoding TNF, TNF receptors, and TNF-signalling factors, are activated, whereas growth factors (IGF and connected pathways), and cell cycle and translation-related pathways (e.g. CDKN and connected pathways) are repressed. Ignoring the observed responses that are likely due to by-products of the biopsy process (e.g. the blood coagulation pathway), the active pathways observed are jointly consistent with a large-scale response of complex molecular pathways to viral infection: hepatic cell reproduction is repressed and programmed cell death is induced. Finally we note that a visual inspection of the network suggests that many of the proteins that bind directly to HCV-encoded molecules (i.e., are their first neighbours in the network) appear on average to be down-regulated relative to the rest of the network. Statistical analysis of the data supports this suggestion. As computed via CytoTalkand R, about 80% of the first neighbours of the viral RNA and proteins are down-regulated in the network of Figure [3](#F3){ref-type="fig"}, compared to 35% of the remaining genes in the network (p= 0.0029). This finding suggests two non-exclusive possibilities: genes encoding HCV-neighbour proteins are targets of host regulatory mechanisms counteracting viral replication; or they are targets of virus-encoded regulatory mechanisms that sabotage anti-viral defences. Conclusion ========== The methods and software tools described here enable the efficient dynamic integrated analysis of diverse data in a major human-disease system. The results show the utility of integrating large-scale human molecular-interaction databases with genomic expression data. This approach is useful for the extraction of biological hypotheses, because it allows us to focus on groups of genes that are not only apparently active in the expression data, but are also functionally associated based on other data, such as molecular interactions. Thus, information that is not restricted to any one data type can be obtained. Moreover, our analyses suggest how various pathways act in concert, and serves as a large-scale window into the genomic response to HCV infection of liver cells. Because the tools and methods we have described are data-type-neutral, there is the prospect of further data integration for a more complete systems-biological approach to understanding viral infection and response mechanisms. The integration of additional, orthogonal sources of information such as detailed clinical data will enable quantitative associations of clinical variables with the activities of molecular pathways and processes. Availability and requirements ============================= • Project name: InteractionFetcher, SynonymFetcher, HomologFetcher, and CytoTalk: plug-ins for Cytoscape • Project home page: <http://labs.systemsbiology.net/galitski/hepc/> • Operating system(s): Platform independent • Programming language: Java • Other requirements: Java 1.4 or higher • License: GNU LGPL • Any restrictions to use by non-academics: license required for access to HPRD interactions (see \[[@B7]\]) Authors\' contributions ======================= DJR: Development of InteractionFetcher, CytoTalkand associated server-side software and databases, construction of seed and scaffold network, analyses of active pathways, functional analyses, manuscript preparation. IA: Expression data processing, ActivePathsand functional analysis. VT: Construction of seed and scaffold networks, functional analysis and biological interpretation. BS: Project conception and planning. TG: Guidance, construction of seed network, manuscript preparation. Acknowledgements ================ The authors would like to thank Maria Smith, Matt Fitzgibbon and Michael Katze for early access to the biopsy data. We would also like to gratefully acknowledge Paul Shannon and Rowan Christmas for their assistance with Cytoscape, Eric Deutsch for his support with data management and processing, and Akhilesh Pandey and Babylaksmi Muthusamy for their help in providing the HPRD interactions in a machine-readable format. We would also like to thank Wei Yan for helpful discussions. This project was funded by the National Institute on Drug Abuse, grant number 1P30DA01562501.	PubMed Central	2024-06-05T03:55:59.763156	2005-6-20	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181626/", "journal": "BMC Bioinformatics. 2005 Jun 20; 6:154", "authors": [ { "first": "David J", "last": "Reiss" }, { "first": "Iliana", "last": "Avila-Campillo" }, { "first": "Vesteinn", "last": "Thorsson" }, { "first": "Benno", "last": "Schwikowski" }, { "first": "Timothy", "last": "Galitski" } ] }
PMC1181627	Background ========== Differentially expressed genes between two or more samples may be of interest to researchers for different reasons, for example, looking at causes of or treatments for diseases such as cancer. Given appropriately processed data, the researcher needs a methodology for assessing the genes in order to separate out ones of interest, i.e. genes with \"significantly\" different levels of expression in different samples. Widely used methods for single slide data include examining the ratio of expression levels for the gene in each of the two samples/channels (or the log ratio), which was the quantity examined in one of the first statistical analyses for differential expression in cDNA microarrays \[[@B2]\]. One of the earliest uses of this quantity for determining differential expression was the \"rule of two\", where if the gene\'s ratio of expression levels in the two channels/samples is greater than two or less than half, it is considered to be differentially expressed \[[@B3]\]. Methods for data with replicate slides include the standard ttest, which requires adjustment for the multiple comparisons being made. Modifications of this approach to account for multiple comparisons include the approach of Dudoit, Yang, Callow and Speed \[[@B1]\], which used a permutation analysis on Welsh\'s t-statistics, and the Significance Analysis of Microarrays (SAM) method, which modifies the t-statistic by adding a constant to the denominator \[[@B4]\]. A good summary of multiple testing adjustments is given by \[[@B5]\]. The idea of modeling the data as two groups of genes, one differentially expressed and one not, seems to be a natural and intuitive approach. This approach has been used in the context of a Bayesian analysis \[[@B6]\], EBarrays, assuming that the observed ratios had a gamma distribution the reciprocal of whose scale parameter itself had a gamma distribution, or, as an alternative assumption, that the observed log ratios were normally distributed and the prior for the mean was normal also. A two-component mixture model was used to model the two groups and the posterior probability was used to make inference about differential expression. This follows from work done for single slide data with a Gamma-Gamma hierarchical model \[[@B7]\]. Another approach using mixture models is given by Pan, Lin and Le \[[@B8]\]. This paper presents a very simple methodology based on mixture models called Normal Uniform Differential Gene Expression (NUDGE) detection. It is applicable to both single slide and replicated cDNA microarray datasets, produced by two of the more widely used experimental setups. After standardizing, the log ratio (or averaged across replicates log ratio) observations are modeled with a two-component mixture model; a normal component for those genes that are not differentially expressed and a uniform component for those that are. The mixture gives posterior probabilities of differential expression which do not need to be adjusted for multiple testing. This methodology is applied to three different experiments. The experiments include single replicate data (Like-Like), multiple replicate data (HIV and Apo Al), experiments with different samples being labeled with their own dyes (HIV) and experiments with all samples being labeled with one dye and compared to a reference sample (Apo Al). The results given by NUDGE are compared with those given by some other methodologies for these types of cDNA microarray experiments (different comparison methods used for different types of experiments). An R package called [nudge]{.underline} to implement the methods in this paper will be made available soon at \[[@B9]\] Results ======= HIV dataset ----------- The HIV dataset that we analyze consists of four replicate experiments comparing cDNA from CD4+ T cell lines at 1 hour after infection with HIV-1BRU with non-infected cell lines on each slide; see \[[@B10]\] for details. There were four slides in total with the same RNA preparations hybridized to each. This dataset is useful in testing the specificity and sensitivity of methods for identifying differentially expressed genes, since there are 13 genes known to be differentially expressed (spots containing PCR products from segments of the HIV-1 genome which the cDNA of the infected cells should hybridize to and the non-infected should not) called positive controls, and 29 genes known not to be (non-human genes which neither infected nor non-infected cDNA samples should hybridize to) called negative controls. There are 4608 gene expression levels recorded in each replicate. The four replicates have balanced dye swaps, so no mean normalization of the (averaged across replicates) log ratios was necessary provided we always used one sample (say the infected sample) in the numerator of the log ratio and the other (non-infected sample) in the denominator regardless of which dye was used to label which sample in each array/slide. NUDGE took a few seconds to run. All 13 positive controls, no negative controls and three other genes were found to be differentially expressed (with posterior probability greater than 0.5). It is clear from Figure [1](#F1){ref-type="fig"} that the rule of two under any normalization gave less than optimal results. In all cases the rule of two correctly found the positive control genes to be differentially expressed. However, in the unnormalized case it also incorrectly found 3 of the 29 negative controls to be differentially expressed, as well as 58 other genes (including the three found by NUDGE). In the variance-normalized case, it incorrectly found one of the 29 negative controls to be differentially expressed, as well as 27 other genes (including the three found by NUDGE). Even though the rule of two is suboptimal, its performance can be improved through the use of the normalization methods suggested here. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Different Normalizations of HIV Data. Different normalizations of HIV data: (a) raw data, (b) data normalized with respect to the variance. The bullets are the positive controls: NUDGE correctly found them all to be differentially expressed. Other genes found to be differentially expressed by NUDGE are indicated by a plus sign, and all genes found not to be differentially expressed by NUDGE are shown by small dots. Negative controls are indicated by a box. No negative controls were found to be differentially expressed by NUDGE. ::: ![](1471-2105-6-173-1) ::: Table [1](#T1){ref-type="table"} shows the results of different methods for the control genes. NUDGE had a perfect result for these genes, with no false positives and no false negatives. The Bonferroni-corrected ttest was the only method considered that recorded any false negatives. The rule of two (normalized or unnormalized), SAM and the EBarrays Lognormal-Normal model all had false positives. Only the EBarrays Gamma-Gamma model equaled NUDGE\'s performance on these control genes. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Summary of Results for HIV data for control genes ::: Method Number of False Negatives Number of False Positives ------------------------------------------- --------------------------- --------------------------- Rule of Two (on unnormalized data) 0 3 Rule of Two (on variance normalized data) 0 1 NUDGE 0 0 SAM 0 2 EBarrays (GG) 0 0 EBarrays (LNN) 0 1 ttest 0 1 Bonferroni corrected ttest 1 0 ::: In order to assess the stability of the different methods, the four replicates were split into two different subsets of two replicates each (still with balanced dye swaps), and the agreements and disagreements between the genes found to be differentially expressed in each of the two datasets was calculated for each of the methods. A summary of the results is given in Table [2](#T2){ref-type="table"}. The number of genes found to be differentially expressed in each of the datasets by each method is given in Table [3](#T3){ref-type="table"}. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Number of agreements and disagreements between the differentially expressed genes found in the two sets of two replicates for the HIV data ::: NUDGE SAM EBarrays GG EBarrays LNN ttest Bonferroni ttest --------------- ------- ----- ------------- -------------- --------- -------------------- Agreements 14 19 13 13 34 15 Disagreements 27 153 16 32 531 217 ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Number of genes declared to be differentially expressed by each method for the HIV data using 2 and 4 replicates ::: NUDGE SAM EBarrays GG EBarrays LNN ttest Bonferroni ttest ------------------ ------- ----- ------------- -------------- --------- -------------------- All 4 replicates 16 42 24 19 26 12 Replicates 1&3 30 49 23 27 193 83 Replicates 2&4 25 142 19 31 406 164 ::: Comparison of results depends on how one weights agreement (roughly indicating true positives) against disagreement (roughly indicating false positives). NUDGE had more agreement and less disagreement than EBarrays-LNN, and thus dominated it on both these criteria. The ttest, both raw and corrected, and SAM, had more agreement, but at the cost of a much higher level of disagreement than NUDGE. NUDGE had more agreement, but also significantly more disagreement, than EBarrays with a Gamma-Gamma model. Finally in order to check the empirical fit of the model to this data (where we know we have both differentially and non-differentially expressed genes) we plot the model\'s fitted density over a histogram of the normalized log ratios in Figure [2](#F2){ref-type="fig"}. The model seems to fit the normalized data fairly well. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Overlay of the model\'s fitted density on the normalized log ratios. Plot (a) shows a histogram of the normalized average log ratios for the HIV data along with a dashed line showing the model-fitted density. Plot (b) shows a close-up of the right-hand tail of the histogram (where the positive controls lie) with a dashed line showing the model-fitted density. ::: ![](1471-2105-6-173-2) ::: Like-like dataset ----------------- This dataset is from a microarray experiment where the same samples (with different dyes) were hybridized to an array with 7680 genes. The expression levels in the red and green dyes were extracted from the image using customized software written at the University of Washington (Spot-On Image, developed by R. E. Bumgarner and Erick Hammersmark). The genes should be equally highly expressed, as each sample is the same, so ideally we should find few differentially expressed genes. Figure [3 (a)](#F3){ref-type="fig"} shows the log ratios plotted against the log intensities. Here we see evidence of the dye effect, since if it were not present the data would fall with some variation about a zero-intercept horizontal line. Figure [3 (b)](#F3){ref-type="fig"} is a plot of the mean-normalized log ratio against the log intensity. In Figure [4](#F4){ref-type="fig"} we plot the absolute mean-normalized log ratio as a function of log intensity. We use a loess smoother of this as a robust estimate of how spread depends on log intensity. This is used to get the loess variance-normalized log ratios, which are plotted against the log intensities in Figure [3 (c)](#F3){ref-type="fig"}. The data now look much more normal and homoscedastic. The NUDGE method took less than 5 seconds to run with 10 iterations of the EM algorithm. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Different Normalizations of Like-like Data. Different normalizations of Like-like data: (a) raw data, (b) data normalized with respect to the mean, (c) data normalized with respect to both mean and variance. Diff. Exp. genes are genes found to be differentially expressed by NUDGE (with posterior probability greater than 0.5). ::: ![](1471-2105-6-173-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Absolute mean normalized log ratio versus log intensity for Like-like Data. Absolute mean normalized log ratio versus log intensity for Like-like Data. The loess line in this plot represents the estimate of the gene-specific Mean Absolute Deviation (MAD), a robust estimator of spread. ::: ![](1471-2105-6-173-4) ::: The results are summarized in Table [4](#T4){ref-type="table"}. NUDGE found 28 differentially expressed genes (with posterior probability greater than 0.5). This is a false positive rate of 0.4%. With no normalization, the rule of two declared 3233 genes to be differentially expressed, 42.1% of the total; clearly this is not appropriate. After the data had been mean-normalized, the rule of two found 281 differentially expressed genes, a false positive rate of 3.7%. When the data have been mean- and variance-normalized, the rule of two finds 105 genes, a false positive rate of 1.4%, still higher than NUDGE. Since there is only one replicate in this case, neither ttests, SAM nor EBarrays can be used to test for differential expression. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Results for the Like-like data ::: Method Estimated False Positive Rate ---------------------------------------------------------- ------------------------------- Rule of Two (on unnormalized data) 42.1% Rule of Two (on mean loess normalised data) 3.7% Rule of Two (on mean and variance loess normalised data) 1.4% NUDGE 0.4% SAM, EBarrays and t tests are not applicable to single slide data. ::: Apo AI dataset -------------- This dataset was analyzed in \[[@B1]\] and 8 genes were suggested to be differentially expressed. The data was obtained from 8 mice with the Apo AI gene knocked out and 8 normal mice. However the replicates were not created simply by comparing samples from control labeled with one dye versus knock-out mice labeled with the other. Instead, cDNA was created from samples from each of the 16 mice (both control and knock-out) and labeled with a red dye. The green dye was used in all cases on cDNA created by pooling all 8 control mice. The statistic used in \[[@B1]\] was ![](1471-2105-6-173-i1.gif) We used the numerator of this statistic, which is the same as Mdefined in equation (7) below, in place of ordinary average log ratios, as detailed in the Methods section. Again the method took only a few seconds to run. Figure [5](#F5){ref-type="fig"} shows the data at different stages of normalization along with the genes found to be differentially expressed in \[[@B1]\]. Table [5](#T5){ref-type="table"} shows the gene position numbers of those genes whose posterior probability of being differentially expressed was in the top sixteen found by NUDGE. All eight of the genes found by \[[@B1]\] to be differentially expressed were also found to be differentially expressed with high probability by our method. The lines in Figure [5](#F5){ref-type="fig"} indicating the rule of two cut-off appear either to miss genes that are differentially expressed (in the unnormalized and mean-normalized cases), or to give a large number of possible false positives (in the mean- and variance-normalized case). ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### NUDGE\'s Top 16 Genes from the Apo data ::: Top 16 genes in terms of NUDGE posterior probability of differential expression --------------------------------------------------------------------------------- ---------------------------------------- ------------------------------ Row numbers in data matrix Probability of differential expression Found by Dudoit et al \[1\]? 540 1.000 Yes 2149 1.000 Yes 5356 1.000 Yes 1739 0.999 Yes 4139 0.999 Yes 2537 0.998 Yes 4941 0.993 Yes 1496 0.829 Yes 5986 0.330 No 541 0.263 No 716 0.099 No 2538 0.087 No 1224 0.066 No 799 0.060 No 1204 0.057 No 3729 0.050 No ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Different Normalizations of Apo Data. Different Normalizations of Apo data: (a) raw data, (b) data normalized with respect to the mean, (c) data normalized with respect to both mean and variance. Diff. Exp. genes are genes found to be differentially expressed by NUDGE (with posterior probability greater than 0.5). ::: ![](1471-2105-6-173-5) ::: For application of SAM, the data were normalized in the standard way, by centering the log ratios across genes within a replicate about zero. Two different levels of the SAM control parameter delta gave reasonable answers when using SAM on this data set. The first level (0.61) found 15 genes to be differentially expressed, including the eight genes found in \[[@B1]\] and by NUDGE, and the False Discovery Rate was estimated to be 5.3%. If we assume that only these eight genes are correct, this would actually correspond to a False Positive Rate of 46.7%. The second level (3.53) found six genes to be differentially expressed, a subset of the eight genes found by \[[@B1]\], and the False Discovery Rate was estimated to be 13.3%. Assuming that only those eight genes are correct, this corresponds to a False Positive Rate of 0% but a False Negative Rate of 25%. These were the best results we obtained using SAM. For similarly normalized data, both the ttest and the Bonferroni adjusted ttest found the 8 genes identified by \[[@B1]\] to be differentially expressed. However, the ttest found an additional 852 genes to be differentially expressed at the 5% significance level (13.5% of all genes), and the Bonferroni adjusted ttest found an additional two genes to be differentially expressed. A summary of the results for the Apo data is given in Table [6](#T6){ref-type="table"}. ::: {#T6 .table-wrap} Table 6 ::: {.caption} ###### Results for the Apo data ::: Method Number of 8 Dudoit et al \[1\] genes found to be differentially expressed Number of other genes found to be differentially expressed ---------------------------------------------------- --------------------------------------------------------------------------- ------------------------------------------------------------ Rule of Two (on unnormalized data) 3 0 Rule of Two (on mean normalized data) 7 0 Rule of Two (on mean and variance normalized data) 8 134 NUDGE 8 0 SAM (delta = 0.61) 8 7 SAM (delta = 3.53) 6 0 ttest 8 852 Bonferroni corrected ttest 8 2 ::: Conclusion ========== We have proposed a simple method for detecting differentially expressed genes that is fast and can be applied to single-slide and multiple-replicate experiments, as well as to log ratio difference experiments. It accounts for the multiple comparisons involved, and produces a posterior probability of differential expression for each gene, rather than just a yes/no testing result. The posterior probabilities can be used either to declare which genes are differentially expressed, or to produce a ranked list of genes for further analysis. The method worked well for the three datasets that we analyzed. In terms of known false positives and false negatives, the method outperforms all multiple-replicate methods except for the Gamma-Gamma EBarrays method to which it offers comparable results with the added advantages of greater simplicity, speed, fewer assumptions and applicability to the single replicate case. Our method can be seen as a parametric alternative to adjustment of tests for multiple comparisons using false discovery rate ideas \[[@B11]\], or empirical Bayes formulations \[[@B12]\]. A similar idea was proposed in \[[@B13]\] for large numbers of tests, in which the distribution of the test statistic was modeled as a mixture of two normals, one corresponding to the null hypothesis being true, and the other to its being false. This differs from our approach in that we use a uniform distribution for the mixture component that corresponds to departures from the null, rather than a mean-shifted normal. Because of this, a method such as \[[@B13]\] could not find both over- and underexpressed genes. A similar idea with different distributional assumptions, using only normally distributed components is given in \[[@B8]\]. Instead of the average log ratios used in the method presented in this paper, \[[@B8]\] use a t-type statistic using the difference of average gene intensities. A more complex approach given by \[[@B14]\] involves modeling each level of differential expression with its own normal component. In our approach the important aspect of the mixture is the cutoff points where the weighted normal density falls below the height of the weighted uniform density. Points beyond the cutoff are declared to be differentially expressed (under a 0.5 posterior probability rule). These cut-off points are relatively unaffected by outliers which affect the range of the data and thus the range and height of the uniform component, because the normal density falls off very rapidly towards the tails, and also because the estimated mixture weights change accordingly. An important part of the method is normalization in terms of variance as well as mean. This extends the original lowess normalization in \[[@B1]\]. As a preprocessing step, it improves the performance not only of NUDGE, but also of other methods, including the simplest of all, the rule of two. Thus, this normalization method may be useful as a preprocessing tool for analysis of differential gene expression, regardless of which method is used to draw final inferences. Methods ======= Model for detecting differential expression ------------------------------------------- Our methods are applied to averages of normalized log ratios; we discuss the specification of these quantities in different experimental settings in the section on normalizations below. In this section we will refer to them simply as observed log ratios. We use logarithms to base 2. Our model is a normal-uniform mixture model \[[@B15],[@B16]\]. We begin by modeling the genes as two different groups: differentially expressed and non-differentially expressed. Each group is modeled by its own density, and so the data as a whole are modeled by a weighted mixture of these densities, where the weights correspond to the prior probabilities of being in each of the two groups. This results in a mixture model with two components. Since genes that are not differentially expressed have a true log ratio of zero, we model the observed log ratios for these genes, after an appropriate transformation, as a group with a Gaussian density. The differentially expressed genes have log ratios that are, for the most part, in some sense \"far\" from the other group. So these genes can be viewed as outliers from the main distribution of non-differentially expressed genes. These genes are modeled as uniformly distributed over an appropriately wide range. The model is ![](1471-2105-6-173-i2.gif) where x~i~is the observed log ratio for gene i, πis the prior probability that a gene is not differentially expressed, N(x\\|μ, σ^2^) denotes a Gaussian distribution with mean μand variance σ^2^, U~\[a,b\]~(x) denotes a uniform distribution on the interval \[a, b\], and Nis the number of genes. We estimate the model by maximum likelihood using the EM algorithm \[[@B17]\]. We define the unknown labels, z~i~, i =1,\..., N, where z~i~is 0 if gene iis not differentially expressed and 1 if it is. There are two steps in the algorithm: the Expectation, or E step, where the labels are estimated given the current parameter estimates, and the Maximization, or M step, where the model parameters, π, μand σ^2^, are estimated given the current estimates of the labels. The maximum likelihood estimates of aand bare ![](1471-2105-6-173-i3.gif) = min{x~i~: i= 1,\..., N}, and ![](1471-2105-6-173-i4.gif) = max{x~i~: i= 1,\..., N}; these do not change during the algorithm. The steps in the algorithm are as follows: ### Iteration k #### Expectation Step ![](1471-2105-6-173-i5.gif) #### Maximization Step ![](1471-2105-6-173-i6.gif) The likelihood for the model given parameter estimates at iteration kis ![](1471-2105-6-173-i7.gif) The above steps are iterated until convergence. Convergence can be checked by calculating the parameter estimates, the labels, and the logarithm of the likelihood at each step, given the current estimates of the parameters. Once the change in these quantities between steps gets small enough the algorithm is deemed to have converged. The increasing property of the EM algorithm guarantees that a local maximum is reached, but a global maximum cannot be guaranteed. This depends on the starting values. For the analyses in this paper, the starting value for the label z~i~was 1 if gene i\'s observed log ratio, minus the mean value for all genes and divided by the standard deviation of the values across all genes, was greater than 2 in absolute value, and 0 otherwise. This appeared to give good results. The final label estimate for gene i, ![](1471-2105-6-173-i8.gif), is the posterior probability that it is differentially expressed, given the parameter estimates. The posterior probabilities do not need to be adjusted for multiple comparisons. Normalizations -------------- There are two different types of experimental setup for which we will discuss normalization. The first is where the two different samples, say control and treatment, have each been labeled with a different color dye, say treatment with red (Cy5, R) and control with green (Cy3, G). In the second experimental setup, the treatment and control samples have replicates, with both control and treatment replicates being labeled with the same dye, say red (Cy5, R), and these are compared to a reference sample labeled with the other dye. Two of the data sets analyzed in this paper, the HIV and the Like-like datasets, are of the first type of setup. The other data set analyzed in this paper is the Apo AI mouse data \[[@B1]\] which is of the second type of setup, with pooled control mRNA used as its reference sample. Since there are slightly different normalizations and quantities of interest used for analysis in these two cases, we will discuss them separately below, referring to the first experiment type as the log ratio experiment (since the log ratios are the quantities of interest), and to the second as the log ratio difference experiment (since the differences of log ratios between control and treatment samples are the quantities of interest). ### Normalizations for the log ratio cDNA experiment The main problem in applying the Normal-Uniform mixture model is that the data need to be normalized in order for this model to be appropriate. In the basic type of cDNA experiment, the log ratio of expressions in the two samples is the quantity of interest. There are dye and other effects that add a bias, making the mean of the non-differentially expressed log ratios non-zero (see the Like-like example in the Results section). Also, the variance of the log ratios depends on the log of the total intensity, where the total intensity is defined as the product of the red and green intensities. We need to ensure that any normalization does not \"pull in\" the differentially expressed genes. ### Single slide normalizations The normalization of single slide log ratios is a two-step process. In the first step, the observed log ratios are regressed nonparametrically on the log intensities, using the lowess regression smoother \[[@B18]\], and the fitted value is subtracted from the observed log ratios. In our implementation, a modification, the loess smoother \[[@B19]\], is used in place of lowess. Specifically, ![](1471-2105-6-173-i9.gif) where Rand Gare the intensities in the red and green channels, and c(log(RG)) is the fitted value from loess regression of log(R/G) on log(RG), a situation we denote by ![](1471-2105-6-173-i10.gif). We got good results with a loess span in the range 60% to 80%. This generally did a good job of normalizing the mean but not the spread. The spread depends on the log intensity, log(RG), and we estimate a running mean absolute deviation by loess regression of the absolute mean-normalized log ratio on the log intensity. We then divide the mean-normalized log ratio by the loess-estimated mean absolute deviation in order to get our final estimate, ![](1471-2105-6-173-i11.gif) where ![](1471-2105-6-173-i12.gif). We got good results with a span between 10% and 20%. As can be seen from the figures in the Results section, this does a good job of making the log ratios for non-differentially expressed genes approximately normal and homoscedastic. ### Multiple slide normalizations with dye swap In dye swap experiments, there is an even number of replicates and they are divided into two groups with equal numbers of replicates. In the second group of replicates, the assignment of dyes to samples is the reverse of that in the first group. Log ratios in this case are taken with the different samples as numerator and denominator (since the assigned dyes will be different for the two groups and averaging must be done over the same ratio of samples not the same ratio of dyes). In that case, mean normalization is unnecessary, although normalization of the variance is still required. This is because we take the average of the log ratios across replicates, ensuring that the dye effect cancels out. ### Multiple slide normalizations without dye swap Here we take the average of log ratios and log intensities across replicates for each of the genes and apply the mean lowess normalization, given by equation (4), with average ratios and intensities in place of the single replicate log ratios and intensities. The variance normalization is not the same for multiple replicate slides as for a single slide. Because the average log ratios are not robust to outliers, even after mean normalization, we carry out a normalization based on variation across replicates rather than on variation depending on intensities, to downweight the influence of outlying observations. If the empirical standard deviation of the log ratios across replicates is greater than the absolute mean-normalized average log ratio for a gene, we divide its mean-normalized average log ratio by its standard deviation. If the empirical standard deviation of the log ratios across replicates is small, defined as smaller than the absolute mean-normalized average log ratio, we divide instead by a constant. The constant is chosen to be a high percentile (we use the 99th) of the distribution of the standard deviations of genes for which the absolute mean-normalized average log ratio is greater than the standard deviation. This avoids a gene being declared differentially expressed just because its empirical across-replicate standard deviation is small, as can easily happen by chance when there are few replicates. Thus the mean- and variance-normalized log ratio for a given gene is: ![](1471-2105-6-173-i13.gif) where mis the number of replicates, q~j~is the mean-normalized log ratio of replicate j, sis the standard deviation of log ratios across replicates, and kis the chosen percentile of the distribution of standard deviations of genes whose absolute mean-normalized average log ratio is greater than their standard deviation. ### Normalizations for the log ratio difference cDNA experiment Here the quantity of interest is the difference in log ratios between control and treatment replicates. We define ![](1471-2105-6-173-i14.gif) ![](1471-2105-6-173-i15.gif) where n~treatment~is the number of treatment replicates, n~control~is the number of control replicates, n= n~treatment~+ n~control~, q~treatment,i~is the log ratio of treatment replicate iand q~control,j~is the log ratio of control replicate j. With these definitions we give the multiple-replicates normalizations, defined analogously to those in the log ratio type experiment. ### Multiple slide normalizations We again use loess to allow dependence of the mean normalization of M on A in the following way: M~norm~= M- c(A) (9) where c(A) = loess(M\~ A), with the recommended span for the loess smoother being between 60% and 80%. For the variance normalization we again use the information about the variance contained in the replicates to get a robust estimator of the overall variance. We calculate the variance of log ratios across the n~control~replicates in the control dataset and call this V~control~. Similarly we calculate the variance of log ratios across the n~treatment~replicates in the treatment dataset and call this V~treatment~. Our estimate for the standard deviation s, in M for each gene is given by ![](1471-2105-6-173-i16.gif) We then develop the variance normalization similarly to the previous log ratio type experiment case. The variance normalization is given by ![](1471-2105-6-173-i17.gif) Summary of model and normalizations for different experiments ------------------------------------------------------------- A summary of the quantities of interest (used in the normalizations and normal uniform mixture model) and the normalizations is given in Table [7](#T7){ref-type="table"}. An R package called [nudge]{.underline} to implement the different normalizations and fit the model in this paper will be made available soon at \[[@B9]\]. ::: {#T7 .table-wrap} Table 7 ::: {.caption} ###### Summary of Methods ::: Type of Experiment Multiple Replicates? Dye Swap? Quantity of Interest Mean Normalization Variance Normalization ----------------------------------------------------------- ---------------------- ----------- ---------------------- -------------------- ------------------------ Sample 1 = Red, Sample 2 = Green No NA Equation (4) Equation (5) Sample 1 = Red, Sample 2 = Green Yes No Equation (4) Equation (6) Sample = Red, Sample = Green Sample = Green, Sample = Red Yes Yes NA Equation (6) Sample 1 & 2 = Red, Reference = Green Yes NA M, Equation (7) Equation (9) Equation (11) ::: Methods for comparison with NUDGE --------------------------------- We now give brief descriptions of the methods for finding differentially expressed genes that will be used for comparison with NUDGE in the datasets examined in the Results section. ### Rule of two This simple but popular method, mentioned in \[[@B3]\], involves examining the ratios or average ratios of the two channels for each gene, and calling those genes with a ratio or average ratio greater than two or less than half, differentially expressed. It requires some initial normalization and its performance can depend on the normalization. ### t test and adjusted t test One of the most obvious first approaches to try for this problem is the classical ttest, as used, for example, in \[[@B20]\]. A simple normalization consisting of centering the mean of the log ratios within each replicate is often used in this case. One needs to be able to estimate the standard deviations as well. Because of the large number of tests being run (thousands in the usual cDNA experiment setup), the standard ttest needs to be modified to account for the multiple testing. Traditionally the most popular adjustment has been the Bonferroni correction, as mentioned in \[[@B21]\]. For the Bonferroni correction with Ngenes/tests and significance level α, we instead call each test significant only if it is significant at the ![](1471-2105-6-173-i25.gif) level, controlling for the probability of one or more false positives. ### EBarrays This follows a hierarchical Bayes approach for modeling the gene expression levels as detailed in \[[@B22]\]. As in our approach, the data are assumed to be generated by a two-component mixture model, one component for differentially expressed and the other for non-differentially expressed genes, each with their own distribution. The parameters specifying these distributions are estimated from the data, whence the name Empirical Bayes. Results in this framework are given for two different parametric models in \[[@B22]\]. In the first model, the observed intensities for the replicates in each channel are assumed to be independently generated from a gamma distribution with a channel-specific scale parameter. The scale parameters are, in turn, assumed to have an inverse gamma distribution, whose parameters are estimated from the whole dataset. In the second model, the log ratios are assumed to be normally distributed, with gene-specific means that are themselves normally distributed. To normalize, the authors divided the log ratio for a given gene and replicate by the average log ratio across genes for that replicate. ### Significance analysis of microarrays (SAM) \[[@B4]\] The statistic used to test for differential expression is a regularized tstatistic, i.e. the mean value divided by the sum of the standard deviation and a constant. SAM controls the False Discovery Rate (FDR), i.e. the number of genes declared to be differentially expressed that are not in truth differentially expressed. A rejection region is fixed and SAM uses a permutation analysis to estimate the FDR. The user then decides on an acceptable rejection region based on their preferences for FDR. Acknowledgements ================ We thank Roger Bumgarner, Ka Yee Yeung and Raphael Gottardo for helpful comments and use of data. This research was supported by NIH grant 8 R01 EB002137-02.	PubMed Central	2024-06-05T03:55:59.765357	2005-7-12	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181627/", "journal": "BMC Bioinformatics. 2005 Jul 12; 6:173", "authors": [ { "first": "Nema", "last": "Dean" }, { "first": "Adrian E", "last": "Raftery" } ] }
PMC1181629	Background ========== The chiropractic educational system in North America is currently in a state of flux. Proposed programs such as the Florida State University School of Chiropractic and the conversion of some chiropractic schools into \"universities,\" suggests that we are searching for a better alternative to the status quo. Medical education in the early twentieth century underwent substantial change. Published in 1910, The Carnegie Foundation Bulletin 4, \"Medical Education in the United States and Canada\" is widely acknowledged as the study that resulted in the reformation and reconstruction of the entire medical educational system. The report renounced the plethora of private and proprietary medical schools of that era, and established scientific medicine and clinical teaching within a university system as the gold standard for teaching medicine. This report, submitted by Abraham Flexner, is more commonly known as the Flexner Report. Although criticism of the report has resulted in some alterations to the original post-Flexner system of medical education, the reliance on scientific and clinical knowledge remains the base of the professional education of medical doctors. Just before the publication of the Flexner Report, the Council on Medical Education had conducted a similar survey of all medical schools in North America. They essentially graded the existing medical schools at the time as A, B, or C according to a number of criteria including educational requirements, curriculum, and resources. To a large extent, Flexner\'s study of medical education at the turn of the century was an exercise in inescapable conclusions. It was less a pragmatic study aimed at unearthing problems within medical education than it was a fact-finding task dedicated to verifying the prevailing view of medical school academia and its insufficient base in science. Essentially, the answers to questions regarding the inadequacy of education were already well-known. Flexner simply graded the schools based on this knowledge. Perhaps the greatest mystique of the Flexner Report is how successfully the recommendations were followed. This report changed the educational system of an entire profession and many suggest that no other single study has been as visibly successful in accomplishing what the Flexner Report did. One reason for the enormous impact of the report was the huge financial resources that were allotted to those schools that followed the recommendations of the report. Monies from various philanthropists funded the growing expenses of the limited number of medical programs that met the standards. In the current period of inconsistency and controversy in chiropractic education, it is not surprising that we hear a call for a Flexner-style report in our educational system. Such a critical and comprehensive examination of all the existing programs, one would hope, would weed out ineffective practices and programs in our schools and result in a clear set of recommendations for the future of science-based chiropractic education. This idea is not new, having been proposed in a similar fashion by John J. Nugent, DC in the early to mid 20^th^century. This paper explores, through descriptive literature analysis and the author\'s experiences within the North American chiropractic educational system, the current status of chiropractic education in North America. Special consideration is given to the essential role of chiropractic governing bodies, the essential history of chiropractic education, including an overview of educational standards, curricula, externship and postgraduate training programs, along with evidence-based health care and the development of chiropractic researchers. Suggested changes to the North American chiropractic educational system are explored including higher admission standards, the need for a chiropractic college admissions test and admission interviews, creation of a research culture in chiropractic schools, support by chiropractic educational and governmental regulatory agencies and the qualifications of chiropractic school administrative staff. In addition, we explore the need for strong postgraduate residency-based educational programs to enhance the exposure of students to a larger volume and variety of patients. Discussion ========== The Current Status of US Chiropractic Education ----------------------------------------------- A profession is defined by a specialized body of knowledge requiring advanced training and by the dedication of its practitioners to the public good over their own enrichment. In exchange, professionals are granted considerable autonomy in setting standards and in the conduct of their work. Any professional level educational system must adopt the tenets of the academy: scientific thinking, rigor and critical analysis. Faculty in the academy have the dual duties of being teachers and scholars. Scholarship is the development of new knowledge, synthesis of the current state of knowledge, applications of that knowledge and teaching that incorporates that knowledge \[[@B1]\]. The commitment of chiropractic schools boards of regents/trustees and administrations to this paradigm of the academy and thus, promoting faculty scholarly activity, is vital to the effectiveness of the institutions. Without the support of the regents/trustees and administrators, the faculty is placed in a situation where it is difficult, at best, to provide the modern education necessary in an ever-changing evidence-based health care environment. Early chiropractic education included classes in some basic and clinical sciences along with philosophy of chiropractic. Performance of chiropractic students on basic science boards suffered as evidenced by a 23% pass rate for chiropractic students on these board exams. Medical students during this same period (1927--1953) had an 86% pass rate \[[@B2]\]. In North America C.O. Watkins, D.C., Joseph Janse, D.C., A.E. Homewood D.C. and others, sought to upgrade the profession by asking serious questions about the effects of spinal manipulation on human health and they recognized that a research base was vitally important to our future. Dr. Watkins was one of the pioneers in the National Chiropractic Association\'s efforts to raise educational standards in the 1930s and 1940s. He demonstrated sincere concern over the image of the profession and he surmised that the development of a scientific base for chiropractic care was critical to our acceptance. Dr. Janse was appointed dean of The National College of Chiropractic. He was also a key figure in the founding of chiropractic\'s three most prominent US regulatory bodies: the National Chiropractic Association\'s Council on Chiropractic Education (forerunner of the CCE), the National Board of Chiropractic Examiners (NBCE) and the Federation of Chiropractic Licensing Boards (FCLB). He was noteworthy for his research on spinal biomechanics, sacroiliac joint function, and the treatment of posture and gait abnormalities. In the USA in 1974, the Council on Chiropractic Education (CCE) was federally recognized as the agency for accreditation of programs and institutions offering the doctor of chiropractic degree. The current CCE \"seeks to insure the quality of chiropractic education in the United States by means of accreditation, educational improvement and public information. CCE develops accreditation criteria to assess how effectively programs or institutions plan, implement and evaluate their mission and goals, program objectives, inputs, resources and outcomes of their chiropractic programs\" \[[@B3]\]. Those schools that failed to meet the CCE standards no longer operated. All chiropractic colleges had achieved accreditation by 1995 and all also now hold accreditation with regional accrediting agencies for their baccalaureate programs and some for master\'s degree programs as well. While the standards for chiropractic education have advanced over the years, there remains much work to be done. Doxey and Phillips, in their paper on entrance requirements to the various professional health care disciplines demonstrated that chiropractic colleges have the least stringent matriculation requirements \[[@B4]\]. Currently, only one chiropractic college requires a baccalaureate degree as an admission requirement. Seven states currently require a baccalaureate degree before granting a chiropractic license and seven have it under consideration, but few of these require that the degree was acquired before entering chiropractic school \[[@B5]\]. There is currently no required chiropractic college admission test. Undergraduate training in chiropractic school consists of approximately 4,200 clock hours of didactic and practical education, with the last year spent treating patients, in some cases while still attending classes. There is only one chiropractic college in the U.S. that follows the academic standard of two semesters per year. Trimesters or quarter systems of education within chiropractic were used in an effort to reduce the time spent in school. In general, the first four to five academic terms are spent studying basic sciences while also learning the basics of spinal examination and treatment. Terms five through eight are spent in clinical classes such a diagnostic imaging, clinical neurology, physical examination, geriatrics, pediatrics, case management and the like. In addition, it is during these terms that students refine their diagnostic and treatment skills for the management of joint diseases, primarily of the spine. Currently, internship (more correctly externship) in the chiropractic profession is a one-year undergraduate endeavor, while it is a three to five year post-graduate program in medical and osteopathic training, including residency training. Some foreign chiropractic programs, such as Switzerland, mandate a one-year externship for recently graduated chiropractors before they are allowed to practice on their own. In addition, clerkships are routine in medical training, while they are not in chiropractic schools, although some chiropractic schools have had clerkship programs for students in lower terms. A number of chiropractic schools now offer hospital rotations to chiropractic externs. In these programs, externs spend a number of weeks working with MDs and DOs in specialty areas such as radiology, orthopedics, sports medicine, family practice, rheumatology and neurosurgery. Our cumulative observations suggest that the obvious contrast in numbers of patient encounters in a chiropractic externship, when compared to a medical/osteopathic internship, are sadly disconcerting from the perspective of the volume and variety of patient exposures. Post-graduate residencies are available to chiropractors, but residency-based training is not currently a requirement, or even commonplace, the exception being diagnostic radiology training leading to diplomate status. Chiropractic externs are currently required to complete 250 joint manipulations, 20 complete history and physical examinations, 20 radiology studies and 15 complete patient workups, from admission to discharge, during their last year in chiropractic school (externship) while treating outpatients. The CCE is mandating that these numbers increase incrementally over the next 6 years to a total of 35. Often these outpatients seen by chiropractic externs are friends and family members, some of whom are even paid by interns to attend the clinics for care. Nyiendo and Haldeman give credence to this finding in a study in 1986 where they concluded that \"patients \[in a chiropractic college teaching clinic\] are not truly representative of patients seen by chiropractors in the field; they are relatively young, with mild complaints.\" The study concludes by suggesting that these students\' clinical training may not reach the level that is necessary to manage patient problems in active practice after graduation \[[@B6]\]. Nyiendo confirmed these findings in 1990 \[[@B7]\]. Further investigation suggests that these patient types are consistent amongst chiropractic school clinics \[[@B8]\]. Instruction in evidence-based medicine (EBM) in American chiropractic schools also appears to be lacking. A search of the current literature finds only one study dedicated to teaching evidence-based health care in a chiropractic school \[[@B9]\]. One study on the use of EBM was performed in a community of chiropractors. The authors demonstrated substantial success in reducing radiography rates in patients with acute low back pain after educating the chiropractors about the current evidence for this intervention. The authors admit that the methods were quasi-experimental \[[@B10]\]. The Foundation for Chiropractic Education and Research (FCER) has helped to foster a research mentality and has developed a program that supports the training and development of chiropractic researchers. A number of chiropractic schools have received federal research grants but the number of researchers and grants appears to still be very small. Suggested Changes to the Chiropractic Educational System -------------------------------------------------------- In our opinion CCE needs to make the admissions standards more stringent, including the requirement for a baccalaureate degree prior to admission and the use of a chiropractic college admissions test. Some believe that increasing the difficulty of entry into chiropractic college would cause a dramatic decrease in enrollment. While we are certain that there would be a \"period of readjustment,\" every increase in standards to date has eventually resulted in a return to previous enrollment levels as the potential students now strive to reach an attainable, but obviously elevated, bar for admission. Mandatory interviews of applicants for chiropractic college admission would do much to help ascertain the background, breadth of knowledge, social skills and communication skills of applicants. Of course, this process will only work if it is used as a screening tool, where only the best applicants are accepted into the programs and those deserving rejection for valid reasons are actually rejected. The curricula of colleges need to be evidence-based, which probably will mean that certain unsupported beliefs and theories of the past will, of necessity, be abandoned. In particular, this means relegating much of the dogmatic, so-called, chiropractic philosophy, which was developed as nothing more than a legal tactic to prevent incarceration of chiropractors in the early twentieth century for practicing medicine without a license, to a class on the history of the profession. Students who perform poorly in chiropractic colleges should not be allowed to pass through the system essentially unabated, as happens currently in some institutions. We feel it unacceptable for chiropractic students to make any academic progress with grades of \'D\' or \'F\' on their transcripts. Such students should be given one chance at remediation and if unsatisfactory grades are achieved in the same class again or in other classes, these students should be expelled from the college. Some schools are moving to an \'A, B, C, F\' grading scale. While it may seem harsh, a learned and distinguished health care profession has little room for, nor should it tolerate, academic underachievement. While each college needs to have an active research department, all members of the faculty must accept their responsibilities as scholars. Our professional educational programs can no longer remain isolated from the academic community. Joining established research universities will help change the culture of the chiropractic professorate to one which values scholarship and models the joy of learning and discovery for their students. A \"publish or perish\" mentality for faculty, we suggest, would be a healthy and refreshing change. Administrative and board support for educational objectives is crucial for any substantive improvement in the training of new chiropractors. Often, chiropractic schools have hired administrators who have little or no formal training in education, providing more political, budgetary and marketing expertise than academic experience. High-level administrators with training in education, along with administrators who have political, budgetary and marketing experience, should become the norm in chiropractic programs. In addition to striving for university affiliation, our institutions must also endeavor to become less and less tuition-dependent. The current tuition-dependent system carries the burden of much of what is wrong with our current system. It fosters academic underachievement, admission of probably under-qualified, if not unqualified, students and under-funded research and faculty development programs. Probably of most critical importance in making positive change in our current educational programs is the establishment of mandatory post-graduate internships and residencies with hospital and interdisciplinary training. Exposure to a large volume and variety of patients is critical to our students training if the profession is to take a place at the center of our mainstream health care system. Interns and residents must be routinely exposed to patients with conditions that represent the full spectrum of potential diagnoses that are considered by chiropractors. This first hand, on-the-job experience by new chiropractors, not just via didactics or textbook exposure, is paramount to the best clinical experience available. Certainly hospital rounds would be a great advantage in this respect. Rigorous post-graduate residencies, such as is the case currently for radiology, need to be developed to train our brightest new doctors to be leaders Summary ======= The chiropractic profession must improve itself through higher educational standards, intellectual honesty and inter-disciplinary co-operation and research rather than continue to rely on patient testimonials and political friendships. We can only obtain cultural authority when we have brought our educational programs up to the level that the public expects of an expert, learned profession. Positive changes, including a chiropractic college admissions test, elevated chiropractic school entrance requirements and mandatory post-graduate residency-based training are suggested. Authors\' contributions ======================= LHW wrote the initial draft of this manuscript. All authors, thereafter, made substantial contributions to recomposing the manuscript as well as appraising it critically for its chief intellectual content. Each author has given approval of the final manuscript.	PubMed Central	2024-06-05T03:55:59.769571	2005-7-7	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181629/", "journal": "Chiropr Osteopat. 2005 Jul 7; 13:10", "authors": [ { "first": "Lawrence H", "last": "Wyatt" }, { "first": "Stephen M", "last": "Perle" }, { "first": "Donald R", "last": "Murphy" }, { "first": "Thomas E", "last": "Hyde" } ] }
PMC1181804	Background ========== The advent of the Internet has changed the way both professionals and consumers look for health information \[[@B1]\]. Abbott \[[@B2]\] found that the existing general public search engines have a high penetration into even restricted-access data repositories, yielding quality information alternative to traditional primary sources. Recently, Google has launched a beta-version of its Google Scholarsearch engine, Nature Publishing Group has changed its search engine to allow deep penetration, and Elsevier has created another specialised search engine for scientific literature, Scopus, which comes with a cost \[[@B3]\]. All of these widen the general public\'s access to high-quality health information. But Peterson \[[@B1]\] showed that the generally low skill level for search strategies that most customers have could lead to retrieval of inadequate information, which raises anxiety and decreases compliance. In response to this, Curro \[[@B4]\] has suggested a simple methodology to assess the quality of medical information retrieved on the Internet, but the impact of this strategy remains to be seen. In the meantime, the medical professional is certainly better advised to look for information that has appraised content. Such sources include online repositories of Critically Appraised Topics (CATs). CATs are short summaries of current medical literature addressing specific clinical questions and are frequently used by clinicians who try to implement principles of Evidence Based Medicine (EBM) \[[@B5]\]. Although some CAT libraries exist, a peer-to-peer sharing network as proposed by Castro \[[@B6]\] is not yet available. CAT Crawler \[[@B7]\], an online search engine, provides access to a number of public online CAT repositories and is the focus of the present study on retrieval quality. Two commonly used evaluation parameters are recall and precision \[[@B8]\]. The former measures the comprehensiveness of a search and the latter measures the accuracy of a search. Relevance is the key concept in the calculation of recall and precision but poses problems of multidimensionality and of dynamic quality. Schamber \[[@B9]\] has emphasized that relevance assessment differs between judges and for the same judge at different times or in different environments. Barry \[[@B10]\] and Schamber \[[@B11]\] have studied the factors affecting relevance assessments. Both studies have agreed that relevance assessments depend on evaluators\' perceptions of the problem situation and the information environment, and the perceptions encompass many other factors beyond information content when they make the relevance assessment \[[@B12]\]. Only a few studies have directly addressed the effect of the variation in relevance assessments on the evaluation of information retrieval systems \[[@B13]-[@B17]\]. All studies varied relevance assessments with evaluators from different domain knowledge background. All of them concluded that variation in relevance assessments among judges has no significant effect on measures of retrieval effectiveness. However, Harter \[[@B18]\] has questioned this conclusion because none of these studies employs real users who approach the system for information need, although some of them tried to simulate this condition. He also highlighted the need to develop measurement instruments that are sensitive to variations in relevance assessments. A common statistical method used in this context is the kappa score, which, in principle, is a contingency table based method that can eliminate chance concordance from the assessment. However, modern search engines usually have filter systems \[[@B3]\], which lead to a selection bias towards relevant documents. Feinstein et al \[[@B19]\] observed that in situations with high imbalance, the paradox of high agreement but low kappa scores can arise. Better filters create more bias, thus increasing the tendency to find such paradox results. In such a situation, a performance assessment based on kappa scores may become meaningless. The work presented here introduces a new parameter, Relevance Similarity, to address this problem. Based on this measurement parameter, the effect of the inter-evaluator variation of relevance assessment on the evaluation of the information retrieval performance was studied. The experiment was carried out on a collection of CATs. Two groups of evaluators participated in the relevance assessments on a set of retrieved topics from the medical meta-search engine, CAT Crawler. Methods ======= The retrieval system used in the study is the CAT Crawler meta-search engine. In a very brief summary, CAT Crawler can be described as a one-stop search engine for CATs stored over numerous online repositories. It has its own search engine, which allows the user to do a specific search rather than simply browse the repositories\' contents. The CAT Crawler\'s standard setting has been shown to yield search results of equal quantity and enhanced quality compared to the original search engines available at some of the repositories \[[@B20]\]. The detailed structural design of CAT Crawler \[[@B7]\] has been described previously. The workflow of the CAT Crawler\'s evaluation is summarized in Figure [1](#F1){ref-type="fig"}. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Workflow for analysing the effect of the inter-evaluator variation on CAT Crawler information retrieval system. ::: ![](1742-5581-2-6-1) ::: Relevance assessment of CATs in the test document set ----------------------------------------------------- Ten keywords (Table [1](#T1){ref-type="table"}) related to medicine were chosen as the test seed and submitted to the search engine. All together 132 CAT links were retrieved and then evaluated for their relevance by 13 people, who were categorized into three groups according to their level of training regarding medical knowledge. Among them, one physician represents medical professionals and is considered as the gold standard for the evaluation, the six evaluators in Group A were trained in biology or medicine, while the six evaluators in Group B had no medical or biological background. For the sake of this exercise, the physician\'s evaluation of the relevance of each topic was taken as the gold standard or \'true\' relevance of each retrieval result. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### CAT Link retrieval details. The numbers indicate how many documents were retrieved by the CAT Crawler meta-search engine. ::: Keyword Number of retrieved links -------------- ------------------------------- Appendicitis 8 Colic 9 Intubation 22 Ketoacidosis 2 Octreotide 3 Palsy 10 Prophylaxis 30 Sleep 16 Tape 3 Ultrasound 29 132 ::: Computation of RelevanceSimilarity ------------------------------------ For each retrieved CAT, the evaluation by every participant in Group A and B was compared with the gold standard set by the medical professional. The Relevance Similarityis defined as: ![](1742-5581-2-6-i1.gif) Relevance Similaritywas computed for each of 132 retrieved links. To compare the relevance assessment between Group A and B, a Chi-square test on the contingency table was carried out on all calculated Relevance Similarityvalues using the statistics software SPSS 11.0 (SPSS Inc., Chicago, IL, USA). In addition, kappa scores within evaluators of Group A and B were calculated respectively. Computation of recall and precision ----------------------------------- In this study, the retrieval system performance is qualified by recall and precision. CATs containing a particular keyword are defined as \"technically relevant\" documents for that keyword \[[@B20]\]. In the first step, for each keyword, technically relevant documents were identified from the experimental document set and individual recall was computed for every evaluator accordingly. In the next step, the recall was averaged over all evaluators in a single group. Finally, the recall was averaged over the ten keyword queries. Following a similar process, the average precision was calculated. Computation of kappa score -------------------------- To ensure the qualification of the physician as a gold standard, he re-evaluated the same document set a year after the initial assessment. A kappa score, observed agreement, positive and negative specific agreements between the two evaluations were calculated \[[@B21],[@B22]\]. The inter-evaluator kappa scores within each group were computed for comparison. Results ======= Analysis of the inter-evaluator variation ----------------------------------------- For each of the 132 retrieved links, Relevance Similaritywas calculated for both Group A and B (Table [2](#T2){ref-type="table"}). For instance, one CAT \"Plain Abdominal Radiographs of No Clinical Utility in Clinically Suspected Appendicitis\" was retrieved from <http://www.med.umich.edu/pediatrics/ebm/cats/radiographs.htm> upon querying the meta-search engine with the keyword Appendicitis. The gold standard rated it as relevant; all six evaluators in Group A rated it as relevant too; whereas, one out of six evaluators in Group B rated it as irrelevant. The corresponding similarity for this particular CAT is computed as: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Relevance Similarity for 132 retrieved CAT links. For each of the 132 documents retrieved by the CAT Crawler meta-search engine, Relevance Similarity (in %) was calculated for both Group A and B. Link S/Nattribute is the serial number to each document. ::: Link S/N Group A (%) Group B (%) Link S/N Group A (%) Group B (%) Link S/N Group A (%) Group B (%) -------------- ----------------- ----------------- -------------- ----------------- ----------------- -------------- ----------------- ----------------- 1 100 83.33 45 100 100 89 66.67 33.33 2 83.33 66.67 46 50 50 90 50 83.33 3 100 100 47 50 33.33 91 50 66.67 4 100 100 48 50 66.67 92 66.67 83.33 5 100 100 49 100 100 93 33.33 83.33 6 100 100 50 100 100 94 50 50 7 100 100 51 100 100 95 33.33 66.67 8 100 100 52 66.67 50 96 100 100 9 100 100 53 33.33 16.67 97 100 100 10 100 100 54 100 100 98 100 100 11 0 0 55 100 100 99 100 66.67 12 100 100 56 100 100 100 66.67 66.67 13 100 100 57 100 100 101 100 100 14 100 100 58 100 100 102 66.67 83.33 15 100 100 59 100 100 103 100 100 16 100 100 60 100 100 104 83.33 100 17 83.33 83.33 61 66.67 33.33 105 100 100 18 100 100 62 83.33 33.33 106 100 100 19 66.67 83.33 63 16.67 83.33 107 100 100 20 66.67 50 64 50 83.33 108 100 100 21 50 66.67 65 100 100 109 100 100 22 33.33 66.67 66 66.67 83.33 110 83.33 66.67 23 100 100 67 0 33.33 111 83.33 83.33 24 50 50 68 50 50 112 83.33 100 25 50 66.67 69 0 16.67 113 83.33 33.33 26 83.33 50 70 66.67 50 114 100 100 27 66.67 100 71 83.33 66.67 115 100 100 28 50 66.67 72 100 83.33 116 100 100 29 100 100 73 50 83.33 117 83.33 66.67 30 50 50 74 100 66.67 118 83.33 66.67 31 100 100 75 100 83.33 119 83.33 66.67 32 83.33 83.33 76 100 100 120 83.33 66.67 33 100 66.67 77 100 83.33 121 100 66.67 34 100 83.33 78 66.67 50 122 100 83.33 35 100 100 79 83.33 83.33 123 100 83.33 36 100 100 80 100 100 124 100 100 37 33.33 16.67 81 83.33 66.67 125 100 100 38 83.33 66.67 82 100 66.67 126 66.67 66.67 39 66.67 50 83 66.67 33.33 127 100 100 40 50 50 84 83.33 66.67 128 83.33 33.33 41 100 100 85 83.33 100 129 33.33 50 42 100 100 86 100 100 130 66.67 83.33 43 16.67 50 87 83.33 100 131 83.33 66.67 44 100 100 88 33.33 50 132 83.33 100 ::: ![](1742-5581-2-6-i2.gif) Figure [2](#F2){ref-type="fig"} shows the frequency analysis of Relevance Similarityfor every retrieved CAT. Both Group A and B have evaluated around 90% of retrieved CATs with more than 50% similarity to the gold standard. The gold standard and the other two groups have made exactly the same relevance assessment on about half of the retrieved CATs. As shown in the last two columns of Figure [2](#F2){ref-type="fig"}, participators in Group A have evaluated 65 CATs (49%) with the same relevance as the gold standard; those in Group B have evaluated 59 CATs (45%) with the same relevance as the gold standard. The Chi-square test performed using SPSS between these two categories resulted in a p-valueof 0.713. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Frequency analysis of evaluation similarity of Group A and B versus the gold standard for all 132 CATs. Compared to the gold standard, the blue bar indicates the number of CATs evaluated by Group A at a different similarity level; the red bar indicates the number of CATs evaluated by Group B at a different similarity level. ::: ![](1742-5581-2-6-2) ::: Evaluation of the retrieval system ---------------------------------- Average recall and precision was computed for each keyword query and all numerical data are listed in Tables [4](#T4){ref-type="table"} and [5](#T5){ref-type="table"} respectively, while Figure [3](#F3){ref-type="fig"} and [4](#F4){ref-type="fig"} provide a more intuitive view of the recall and precision evaluation of retrieval. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Average recall for the gold standard and the two groups of evaluators ::: Gold Standard Group A Group B ------------------ ------------------- ------------- ------------- Appendicitis 100.00 97.92 93.75 Colic 53.33 58.89 58.89 Intubation 37.84 41.44 40.09 Ketoacidosis 33.33 50.00 50.00 Octreotide 75.00 54.17 62.50 Palsy 54.55 65.15 65.15 Prophylaxis 64.86 69.82 56.76 Sleep 43.75 59.38 51.04 Tape 50.00 44.44 47.22 Ultrasound 36.17 38.30 39.36 Average 54.88 57.95 56.48 ::: ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Average precision for the gold standard and the two groups of evaluators ::: Gold Standard Group A Group B ------------------ ------------------- ------------- ------------- Appendicitis 100.00 97.92 93.75 Colic 88.89 98.15 98.15 Intubation 63.64 69.70 67.42 Ketoacidosis 50.00 75.00 75.00 Octreotide 100.00 72.22 83.33 Palsy 60.00 71.67 71.67 Prophylaxis 80.00 86.11 70.00 Sleep 43.75 59.38 51.04 Tape 100.00 88.89 94.44 Ultrasound 58.62 62.07 63.79 Average 74.49 78.11 76.86 ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Recall comparison. The bars indicate each of the three groups\' recall (in %) for the ten keywords. ::: ![](1742-5581-2-6-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Precision comparison. The bars indicate each of the three groups\' precision (in %) for the ten keywords. ::: ![](1742-5581-2-6-4) ::: Kappa scores ------------ The two evaluations of the document set carried out by the physician who served as the \"gold standard\" have a high concordance with a kappa score of 0.879. The inter-evaluator kappa scores ranged from 0.136 to 0.713 (0.387 ± 0.165) within Group A, and from -0.001 to 0.807 within Group B (0.357 ± 0.218) (Table [3](#T3){ref-type="table"}). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Kappa scores within Group A and Group B, de monstrating the paradoxically low kappa scores despite high agreement. ::: Group A Group B ----------- ------------- ------------- ------- ------- ------- ------- ------- ------- ------- -------- Evaluator 2 3 4 5 6 2 3 4 5 6 1 0.404 0.426 0.136 0.258 0.656 0.208 0.670 0.410 0.807 0.352 2 0.461 0.259 0.713 0.520 0.257 0.135 0.125 -0.001 3 0.180 0.438 0.439 0.440 0.643 0.353 4 0.241 0.270 0.370 0.250 5 0.404 0.330 ::: Discussion ========== Recall and precision remain standard evaluation parameters for the effectiveness evaluation of an information retrieval system. Both depend on the concept of relevance, i.e. the answer to the question whether the retrieved information is useful or not. A major problem lies in the fact that this answer may vary depending on multiple factors \[[@B9]-[@B11]\]. The perception of variance tempts one to assume that it must influence the assessment of retrieval efficiency, yet the small number of studies addressing this problem \[[@B13]-[@B17]\], including the one presented here, come to a different conclusion. This conclusion has been challenged \[[@B18]\], and the need to find measurement criteria for variance impact was recognized. Three decades ago, Saracevic \[[@B23]\] has suggested to conduct more experiments on various systems with differently obtained assessments in the research of relevance variation. In contrast to previous studies, the present one investigates the effect of relevance assessment on the performance of a specialized retrieval system, developed specifically for physicians trying to implement EBM into daily routine. The test collection is a set holding around 1000 CATs. The variance of evaluator behavior is directly addressed by measuring Relevance Similarity. The concept of Relevance Similarityis strongly dependent on the knowledge of \"true relevance\". It may be impossible to establish the true relevance of a given document. Whoever assesses a document may make an error. As soon as the document is assessed by another, the relevance may be attributed differently. For this reason, the \"true relevance\" is usually decided by expert committees, e.g. a group of specialists. Documents they assess in unison are assumed to be truly relevant or truly irrelevant; documents with variations in the assessment are either judged according to the majority\'s decision or following a brief decision rule. In the present study, this problem was solved differently. According to the domain knowledge disparity between the evaluators, they could be categorized as: one medical professional, six life scientists and six IT scientists. From the training point of view, the physician is most closely related to the medical field and his judgement was therefore used as the gold standard or \"true relevance\". While one may (or may not) doubt his qualification to assign true relevance, his re-assessment of the same document set one year after his initial evaluation shows a good correlation. Using kappa statistics, a kappa score of 0.879 indicated an \"excellent\" concordance \[[@B24]\]. Kappa statistics are a standard measure of inter-evaluator agreement. In the present study, kappa scores for Group A evaluators ranged from 0.136 to 0.713, and from -0.001 to 0.807 for Group B (Table [3](#T3){ref-type="table"}). Kappa statistics are based on the assumption that a \"true\" value is not known beforehand, and that a higher level of concordance signifies a higher probability to have a formed \"truth\". However, in situations where there is a strong bias towards either true or false positive, or true or false negative, high concordance can yield a low kappa score \[[@B19]\]. Positive and negative agreements have been suggested as an additional quality measurement in such cases. In the present study, we calculated positive and negative agreements \[[@B25]\] (P~pos~: 0.74--0.93; P~neg~: 0.15--0.82), but this does not give any additional information to that derived from kappa scores. While the calculation of kappa score does have its value, albeit not undisputed \[[@B19],[@B25]\], to rely on this calculation misses a philosophical point: human evaluators may assess as true or false a statement that is not so for reasons that depend on external factors (\"philosophies of life\", political, theological etc.) and err with high concordance because they have concordance on the external factors. By assessing the documents using a gold standard considered to stand for the \"true relevance\", the method of Relevance Similarityovercomes this problem. Internal concordance of the gold standard evaluator is demonstrated by his excellent kappa score, and his study subject of medicine as opposed to life sciences/computer sciences qualifies him for this position. With the physician as the gold standard, the Relevance Similarityfor Groups A and B was computed for the analysis of these groups\' agreement with the gold standard (Figure [2](#F2){ref-type="fig"}). For a high similarity level, Group A has more agreements with the gold standard than Group B. For example, for a relevance similarity level of 83.33%, Group A and the gold standard have evaluated 24 CATs with the same relevance. By comparison, Group B and the gold standard have an agreement over 21 CATs only. The same phenomenon occurs at a relevance similarity level of 100%. As the gold standard and Group A represent people with professional or some relevant medical domain knowledge, the result is consistent with what has been reported by Cuadra and Katter \[[@B26]\] and Rees and Schultz \[[@B27]\] that the agreement among evaluators with more subject knowledge is higher. On the other hand, a p-valueof 0.713 shows there is no significant difference between the mean relevance assessment of Group A and B as compared to the gold standard. Since the time of the Cranfield experiment \[[@B28]\], researchers have been aware of the difficulty of calculating the exact recall as this requires the true knowledge of the total number of relevant documents in the entire database. Even in the relatively small document repository used here that consists of around 1000 CATs in total, a visual control of all documents is unlikely to produce a reliable result in finding all files that contain the keywords, i.e. \"technically relevant\" documents. Using PERL scripts as described previously \[[@B20]\], this task is achieved reliably. The recall is computed accordingly. The average recall and precision over all queries (Table [4](#T4){ref-type="table"} and [5](#T5){ref-type="table"}) show that people with different domain knowledge have evaluated the retrieval system similarly. This supports the hypothesis of Lesk and Salton \[[@B13]\] that variations in relevance assessments do not cause substantial variations in retrieval performance. Their explanation is based on the fact that average recall and precision is obtained by averaging over many search requests. Concurring with this explanation, the average recall and precision for each keyword query in the present study (Table [4](#T4){ref-type="table"},[5](#T5){ref-type="table"} and Figure [3](#F3){ref-type="fig"},[4](#F4){ref-type="fig"}) does vary between the gold standard, Group A and Group B in response to variations in relevance assessments for each keyword by different evaluators. In this study, documents are judged for binary relevance, i.e. either relevant or irrelevant. Kekäläinen and Järvelin \[[@B29]\] have highlighted the multilevel phenomenon of relevance. The binary evaluation technique used in many studies is not able to represent the degree of relevance and hence leads to the difficulty of ranking a set of relevant documents. Recognizing the problem, many studies on information seeking and retrieval used multi-degree relevance assessments \[[@B30],[@B31]\]. It would be worthwhile to consider the effect of multi-level relevance rating scales on the performance evaluation of the retrieval system. Conclusion ========== The present study directly addresses the question whether variability of relevance assessment has an impact on the evaluation of efficiency of a given information retrieval system. In the present setting, using a highly specialized search program exclusively targeting Critical Appraised Topics \[[@B7]\], the answer to that question is a clear \"no\" -- the effectiveness of the CAT Crawler can be evaluated in an objective way. To what extent the subject knowledge of the end-user influences his perception of relevance of the retrieved information is certainly important from an economic view, as it will have an impact on his usage patterns of information retrieval systems. The results presented here demonstrate, however, that a safe evaluation of the retrieval quality of a given information retrieval system is indeed possible. While this does not allow for a qualitative control of the information contents on the plethora of websites dedicated to medical knowledge (or, in some cases, ignorance), the good news is that at least the technical quality of medical search engines can be evaluated. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= Author 1 (PD) participated in the design of the study, performed data analysis and drafted the manuscript. Author 2 (ML) has contributed on the statistical analysis of raw data. Author 3 (AM) participated in the design of the study and the drafting of the manuscript. Acknowledgements ================ The authors would like to thank the staff and students of the Bioinformatics Institute who have volunteered to evaluate the performance of the medical meta-search engine. We are also grateful for the help given by A. Ramasamy from University of Oxford, UK and A. L. Zhu from National University of Singapore on the statistical analysis of computed data. The manuscript was revised with the help of Dr. F. Tang and F. Mondry.	PubMed Central	2024-06-05T03:55:59.771909	2005-7-20	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181804/", "journal": "Biomed Digit Libr. 2005 Jul 20; 2:6", "authors": [ { "first": "Peng", "last": "Dong" }, { "first": "Marie", "last": "Loh" }, { "first": "Adrian", "last": "Mondry" } ] }
PMC1181805	Background ========== Anticipation is a medical observation that refers to the progressive worsening of a disease\'s symptoms and/or an earlier age of onset over successive generations of affected family members \[[@B1]\]. Although historically controversial, the concept gained widespread scientific acceptance with the identification in 1991 of unstable trinucleotide repeats associated with Fragile X syndrome \[[@B2],[@B3]\] and spinal and bulbar muscular atrophy (SBMA) \[[@B4]\]. Today, 35 human diseases, some of which also exhibit anticipation, have been associated with unstable repeats \[[@B5]\]. Diseases for which unstable microsatellites are the causative disease mechanism can be divided into those caused by coding or non-coding repeat expansions. The majority of disease-associated coding repeats identified to date are CAG-type repeats encoding an expanded poly-glutamine tract in affected individuals. CAG-type expansion disorders include spinal and bulbar muscular atrophy (SBMA) \[[@B4]\], dentatorubral-pallidoluysian atrophy (DRPLA) \[[@B6]\], Huntington disease (HD) \[[@B7]\] and a range of spinocerebellar ataxias (SCAs) including SCA1 \[[@B8]\], SCA2 \[[@B9]\], SCA3 \[[@B10]\], SCA6 \[[@B11]\], and SCA7 \[[@B12]\]. In these diseases, an expanded poly-glutamine tract results in a toxic gain of function causing either neuronal degeneration \[[@B13]\], or in mouse models of spinocerebellar ataxia (SCA), neuronal dysfunction due to Purkinje cell abnormalities \[[@B14]\]. The precise pathogenic disease mechanism is unknown but requires expression of the expanded polyglutamine tract. Neuronal inclusion bodies are observable on autopsy \[[@B14]\]. Untranslated repeats are diverse and include non-trinucleotide repeats. For example, progressive myoclonic epilepsy type 1 (EPM1) pathology results from an expansion of the dodecamer CCCCGCCCCGCG \[[@B15]\] and an ATTCT repeat expansion is the pathogenic agent in SCA10 \[[@B16]\]. In contrast to the coding repeat disorders, non-coding repeats can expand dramatically into the range of thousands of repeats \[[@B17]\]. Most non-coding repeat expansions are not associated with neuronal inclusion bodies on autopsy \[[@B14]\], with the exception of Fragile X-associated tremor ataxia syndrome \[[@B18]\], and nuclear foci observed in neurons of myotonic dystrophy patients \[[@B19]\]. Anticipation has been reported in a number of orphan diseases in which repeat expansion may have a role in etiology. These diseases include autosomal dominant limb-girdle muscular dystrophy \[[@B20]\], Crohn\'s disease \[[@B21]\], leukemia \[[@B22]\], nodal osteoarthritis \[[@B23]\], Parkinson\'s disease \[[@B24]\], rheumatoid arthritis \[[@B25]\], truncal heart defects \[[@B26]\], mood disorders \[[@B27]\], schizophrenia \[[@B28],[@B29]\], and anxiety disorders \[[@B30],[@B31]\]. Although no repeat expansions have been associated with any of these disorders, no comprehensive surveys have been undertaken. Historically if one suspected a polymorphic microsatellite repeat were associated with a disease, few bioinformatics resources were available to identify relevant repeats in the human genome. One approach now available is to browse the Tandem Repeats Finder (TRF) \[[@B32]\] track on the UCSC genome browser \[[@B33]\] within a genomic region of interest. TRF at UCSC was executed with liberal insertion and deletion (indel) and substitution penalties that allow the detection of larger, frequently impure repeats. Since pure repeat tracts are more likely to expand than impure repeat tracts following transmission \[[@B34]-[@B36]\] a large fraction of repeats presented at UCSC are probably not relevant for disease association studies. Furthermore, certain known disease-associated repeats, such as the GAA repeat in Friedreich\'s Ataxia (chr9:67,109,320-67,109,339) \[[@B37]\], are not detected at all at UCSC because they are too short to be detected by their TRF parameters. Other groups have created databases of all 2--16 repeat unit satellite repeats within human gene regions \[[@B38],[@B39]\] and of all 1--6 repeat unit microsatellites across prokaryotic and eukaryotic taxa \[[@B38]\]. Collins detected microsatellites with a novel algorithm and deposited this data in a relational database called GRID Short Tandem Repeats (STR) database \[[@B39]\]. This database included in silicopolymorphism detection of coding trinucleotide repeats by using the BLAST algorithm to detect each repeat\'s length polymorphisms within GenBank, but only for a subset of coding repeats \[[@B39]\]. These resources enrich the microsatellite repeat bioinformatics landscape but do not integrate these data with other published resources in a way relevant for repeat prioritization in disease-association studies. Also, these resources do not provide flexible interfaces for combining data in user-defined ways to allow dynamic generation of candidate repeat lists. For example, both the Microsatellites Repeat Database (MRD) \[[@B38]\] and the STR databases \[[@B39]\] provide static co-ordinates of candidate repeats for disease-association studies defined by the author\'s criteria, but lack the functionality to easily re-prioritize repeats based on user preferences. To address these deficiencies we created Satellog, a database that catalogs all pure 1--16 repeat unit satellite repeats in the human genome along with supplementary data we believe to be of use for the prioritization of satellite repeats in disease association studies. For each pure repeat Satellog can also calculate the percentile rank of its length relative to other repeats of the same class in the genome, its polymorphism within UniGene clusters \[[@B40]\], its location relative to known genes \[[@B41]\], and its expression profile in normal tissues according to the GeneNote database \[[@B42]\]. Repeats within Satellog can be prioritized based on any of their characteristics (i.e. repeat unit, class, period, length, length percentile rank, genomic co-ordinates), polymorphism profile within UniGene, proximity to or presence within gene regions (i.e. cds, UTR, 15 kb upstream etc), metadata of the genes they are detected within, and gene expression profiles within normal human tissues. Disease-associated repeats from 31 diseases were used as a test set to see what fraction could be detected independently within Satellog and what could be learned about polymorphic repeats in general. To showcase its utility, we used Satellog to prioritize repeats for disease-association studies in Huntington\'s disease and schizophrenia. Satellog is available as a web-queriable database along with all source code licensed under GNU General Public License at <http://satellog.bcgsc.ca>. Results ======= Summary statistics ------------------ A total of 8,357,425 pure repeats were detected by TRF in the human genome and were stored in Satellog. Of these, 5,398,328 or 64.6% were detected within an EnsEMBL-defined gene or within 60 kb flanking either side of an EnsEMBL gene. These repeats mapped to 7,260,625 genetic locations in or near EnsEMBL genes, reflecting the fact that some repeats were located within more than one gene. Of the genes in EnsEMBL, 92% (21,654 / 23,531) had at least one pure repeat within 60 kb of their gene boundaries. All repeats in Satellog clustered into 70,318 unique repeat classes. Overall, repeat counts correlated with decreasing chromosomal size, however chromosome 19 had the highest density of repeats in accordance with previously published reports \[[@B43]\] (Figure S1, Table S1 -- supplementary information available online at <http://satellog.bcgsc.ca/source.php>). Data summarizing repeat counts and density by repeat unit size and chromosome (Table S2), by specific repeat unit (Table S3) and by gene region (Table S4) are also available online as supplementary information. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Unstable coding repeats organized by descending standard deviation Sample output from Satellog. ::: unit length gene location pep name mean sd ---------- ------------ ------------------- ------------------------------ ----------- ----------- ---------- GCA 23 cds LQQQQQQQQQQQQQQQQQQQQQQQ AR 20.36 4.11 CAG 15 cds QQQQQQQQQQQQQQQH DRPLA 12.44 3.9 GGC 17 cds GGGGGGGGGGGGGGGGGE AR 15.1 3.54 CAG 19 cds QQQQQQQQQQQQQQQQQQQQ TBP 17.1 3.01 ACC 13 cds LPPPPPPPPPPPPP NULL 11.5 2.12 GGC 8 cds GGGGGGGGG GDF7 9 1.73 CTG 6 cds GSSSSSR PCDH12 7.2 1.55 CCG 9 cds PAAAAAAAAA NULL 6 1.41 GGGGCC 4 cds APAPAPAPAP CDKN1C 3.33 1.15 GGC 6 cds GGGGGG NULL 6.67 1.03 The ten most unstable coding repeats organized by descending standard deviation. Repeats highlighted in bold are known disease-associated repeats. (Note: trailing non-consensus amino acids are not artefactual output. Repeat units continue to be detected at the DNA level even it they do not completely achieve the consensus. For example in the second row above, the corresponding DNA sequence (CAG)~15~CA contains a trailing CA (of the subsequent CAT codon) that translates into histidine). ::: Characteristics of disease-associated repeats --------------------------------------------- Disease-associated repeats and their common properties were recently reviewed \[[@B5]\]. We queried the database with these sequences to observe any characteristic features of these repeats relative to all other repeats. We asked how many of these repeats could be identified as potentially unstable using only the bioinformatics resources within Satellog. The co-ordinates for 31 of the 35 disease-associated repeats were manually collected from the review and identified in Satellog. Repeats that were not analyzed either had a repeat period greater than 16 (thus not detected by our TRF parameters) or were polymorphic but not associated with any disease. For these disease-associated repeats, there is no record of their precise genomic co-ordinates. To address this, we used Satellog to probe for the probable repeat that corresponded to each disease by selecting all repeats of the expected class within each disease gene. All repeats were detected, except for the repeat responsible for blepharophimosis \[[@B44]\]. In 12 cases, more than one candidate was detected as the disease-associated repeat for a disease. These cases usually involve flanking repeats of the same class that are detected as two distinct repeats because of an interrupting unit, an established characteristic of some disease-associated repeats such as those responsible for SCA1 \[[@B35]\] and Fragile X syndrome \[[@B36]\]. In these cases, we simply retained both repeats and associated them with the disease. A total of 51 repeats were mapped for 31 diseases. Interestingly, these repeats were from only 6 repeat classes. Trinucleotide repeats are the most common repeat class implicated in disease \[[@B5]\], especially for disorders caused by coding repeat expansion. Of the disease-associated repeats we analyzed, 28 of the 31 were trinucleotide repeats with 16 being from the CAG repeat class, 11 from the GCG repeat class, and one each from the CCCCGCCCCGCG, CCTG, GAA, and ATTCT repeat classes respectively. These disease-associated repeat classes had dramatically different genomic distributions (Figure [1](#F1){ref-type="fig"}). For example, the CCCCGCCCCGCG dodecamer implicated in progressive myoclonic epilepsy type 1 (EPM1) \[[@B15]\] is the only pure repeat of its class detected in the human genome and therefore has a singleton as its distribution. The remaining repeat classes have broader distributions, particularly the GAA repeat class. GAA repeats have been reported to have a unique distribution relative to other trinucleotide repeats due to their evolutionary origin within Alurepeats \[[@B45]\]. Satellog recapitulated a distinct, expanded profile for GAA repeats relative to all other trinucleotide repeats (Figure [1](#F1){ref-type="fig"}). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Genome-wide repeat lengths of disease-associated repeat classes. Genomic distribution of repeat lengths of all repeat classes associated with disease. ::: ![](1471-2105-6-145-1) ::: We defined significant repeat length in the reference genome as any repeat with length within the top 5% of its class (corresponds to a percentile rank \< 0.05 in Satellog). Using this cut-off, we determined whether the reference genome repeat length is significant for any of the disease-associated repeats within their respective disease classes. Interestingly, 80% (24/30) of the disease-associated repeats in Figure [1](#F1){ref-type="fig"} were significantly long in the reference genome given their repeat class\' length distribution (percentile rank \< 0.05). In fact, 20 of 30 of all disease-associated repeats had a percentile rank of 0.01 or less indicating that these repeats were the extreme outliers within their class. Of the coding repeats, 12 of 17 had significant repeat lengths, including all the CAG-type repeats. Exceptions were the cleidocranial dysplasia (CCD), hand-foot-genital syndrome (HFGS), synpolydactyly, oculopharyngeal muscular dystrophy (OPMD), and holoprosencephaly coding GCG repeats. The CCCCGCCCCGCG dodecamer implicated in progressive myoclonic epilepsy type 1 (EPM1) is not included in this comparison because there were no other pure repeats of its class in the genome. Polymorphic repeats detected in UniGene clusters ------------------------------------------------ We used a bioinformatics approach to see if we could detect repeat polymorphisms within UniGene sequences. Of the 8,357,425 pure repeats detected by Satellog, 1.3% or 111,950 repeats were detected as transcribed by the EnsEMBL API (either in the UTR or coding sequence (cds) of the gene). Of these repeats, approximately half (57.4% or 64,116 repeats) were detected within UniGene cluster sequences. Finally, of these repeats, only 5,546 repeats were detected as polymorphic (defined as any repeat that had at least one sequence within a cluster with a different repeat length). A measure of repeat polymorphism was provided by calculating the standard deviation (sd) of all repeat lengths detected within a UniGene cluster. A total of 2,763, 541, and 4,244 polymorphic repeats were detected in coding, 5\'-UTR, and 3\'-UTR sequence respectively (Note, repeats may exist in more than one gene which is why the location break-down of the repeats is greater than the total number of distinct polymorphic repeats of 5,546). Our ability to generalize repeat polymorphism trends within genetic regions was confounded by increased sampling of the 3\' end of genes (Figure [2](#F2){ref-type="fig"}). To control for this, we compared the polymorphism profile of repeats in coding, 5\'UTR, and 3\'UTR regions that had equal sampling depth. By one-way ANOVA, we found a significant difference between coding (0.322 ± 0.134), 5\'-UTR (0.416 ± 0.207), and 3\'UTR (0.510 ± 0.184) repeats. There was significant repeat polymorphism in the 3\'-UTR sequence relative to coding sequence but not to 5\'-UTR sequence after controlling for sampling bias (Tukey-Kramer post-hoc multiple comparisons test, P\< 0.001). Next we evaluated the tolerance of repeat polymorphisms by various repeat periods in coding and UTR sequence. To observe if highly polymorphic repeats were restricted to certain repeat periods (defined as repeat unit length), the repeat period distribution was observed at progressively increasing sd values (Figure [3](#F3){ref-type="fig"} &[4](#F4){ref-type="fig"}). Untranslated repeats were well distributed across all repeat periods except for 16 mers at an sd cut-off of 1 (which roughly corresponded to repeat polymorphisms of 1 repeat unit). At increasing sd cut-offs, untranslated polymorphic repeats were detected as penta-, tri-and mainly di-nucleotide repeats (Figure [3](#F3){ref-type="fig"}). In contrast, while coding repeat polymorphisms were widely distributed at an sd of 1, they were mainly restricted to trinucleotide repeats at higher sd cut-offs (Figure [4](#F4){ref-type="fig"}). Although the untranslated repeats had higher sd values, their most polymorphic sd values were restricted to mono-and di-nucleotide repeats. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Boxplot comparison of polymorphic repeats from coding, 5\'-UTR and 3\'-UTR sequence. Median standard deviations (line through box) of all polymorphic repeats detected in coding, 5\'-UTR, and 3\'-UTR sequence. After controlling for sampling bias, coding and 5\'-UTR standard deviations did not significantly differ from each other, but did significantly differ from 3\'-UTR repeats implying that the 3\'-UTR tolerates larger, more expanded repeats (P\< 0.001). ::: ![](1471-2105-6-145-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Counts of unstable non-coding repeats at increasing instability cut-offs. Repeat period distribution of polymorphic non-coding repeats at increasing standard deviation (sd) cut-offs. ::: ![](1471-2105-6-145-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Counts of unstable coding repeats at increasing instability cut-offs. Repeat period distribution of polymorphic coding repeats at increasing standard deviation (sd) cut-offs. ::: ![](1471-2105-6-145-4) ::: Disease-associated repeats detected in UniGene clusters ------------------------------------------------------- To address whether known disease-associated repeats were polymorphic within UniGene clusters, we extracted the top ten most polymorphic coding and non-coding repeats, based on their sd value, and determined if any of the disease-associated repeats were also the most polymorphic. The repeats associated with SBMA (ARis the gene mutated in individuals affected with SBMA), DRPLA, and SCA17 (TBPis the gene mutated in individuals affected with SCA17) were detected as the first-, third-and fourth-most polymorphic coding repeats (Table [1](#T1){ref-type="table"}). The AIB-I repeat that confers increased risk of prostate cancer was also detected as polymorphic but not in the top ten. The repeat responsible for FRAXE was detected as polymorphic, but not as one of the top ten most polymorphic untranslated repeats (Table [2](#T2){ref-type="table"}). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Unstable untranslated repeats organized by descending standard deviation Sample output from Satellog. ::: unit length gene location name mean sd ---------- ------------ ------------------- ---------- ---------- -------- GT 9 3utr NULL 12.17 7.29 AT 25 3utr SPATA2 19 7.07 TA 10 3utr NULL 11.11 6.33 T 11 3utr LYZ 13.08 4.72 AC 23 3utr NAV1 17.71 4.39 AC 28 5utr NULL 25.4 3.71 GCA 16 5utr GLS 9.6 3.58 GCC 14 5utr DAZAP1 15 2.28 T 13 5utr NULL 15 2.24 T 19 5utr NULL 17.5 2.12 The ten most unstable untranslated repeats organized by descending standard deviation. No disease-associated repeats are present in this sample. ::: Of the 31 disease-associated repeats discussed previously, only 5 repeats were detected as polymorphic within UniGene clusters. We sought to understand why this occurred. Of the 31 disease-associated repeats, 4 failed to map within the genomic co-ordinates of any mapped UniGene cluster. The remaining 27 repeats mapped within a UniGene cluster\'s genomic co-ordinates. However, 16 of these failed to be detected within UniGene sequences even though they mapped within a UniGene cluster. This could be because of the 3\' bias of the UniGene sequences, the incomplete nature of the clusters \[[@B40]\], sequence errors in the representative UniGene cluster sequence we searched against for hits (Hs.seq.uniq -- see Methods for details), or the limitations of our mapping algorithm. Our approach enforces that the repeat must exist with at least 10 bp of flanking sequence, which leaves out repeats at the edge of UniGene clusters. The remaining 11 disease-associated repeats were detected within UniGene clusters, but only 5 of these repeats were polymorphic. On average, the repeats detected as polymorphic had more hits within UniGene clusters than those detected as stable (there were an average of 17.4 observations per repeat for the polymorphic repeats to 4.54 for stable repeats). This suggests that there is a greater chance of observing repeat polymorphism with deeper sampling. All of the polymorphic repeats were limited to one UniGene cluster and none of the lengths surpassed the disease pre-mutation threshold of 29, 25, 36, 42, and 39 pure repeats for the repeats responsible for increased prostate cancer risk (AIB-I), DRPLA, SBMA, SCA17, and FRAXE respectively \[[@B5]\]. Discussion ========== Although one might expect greater polymorphism in UTR sequence relative to coding sequence due to reduced evolutionary constraints, both 5\'-UTR and coding repeats had similar rates of polymorphism, whereas 3\'-UTR repeats had significantly greater polymorphism compared to these two groups. This may be due to the documented 3\'-UTR sequence over-representation in UniGene \[[@B40]\]. However, depending on whether the repeat is within coding or UTR sequence, there appears to be constraints regarding what repeat unit sizes can tolerate large polymorphisms. Of the more polymorphic UTR repeats (those with sd values greater than 3), there was a single trinucleotide repeat amongst mainly dinucleotide and mononucleotide repeats (Figure [2](#F2){ref-type="fig"}, Table [2](#T2){ref-type="table"}). On the other hand, the majority of coding repeat polymorphisms, although less pronounced, are almost entirely in factors of three (Figure [1](#F1){ref-type="fig"}, Table [1](#T1){ref-type="table"}). Our results support the observation that coding microsatellite polymorphisms are usually in-frame in order to avoid a deleterious phenotype resulting from frame-shift or to provide a rapid evolutionary response to a changing environment \[[@B46]\]. It is important to consider that larger repeat polymorphisms could cause a UniGene cluster to \"split\" into two distinct clusters. This could downplay a repeat\'s polymorphism because such repeats would not be evaluated as a single group, therefore decreasing the repeat\'s sd value. This issue was addressed by pre-mapping all UniGene clusters to the human genome. If the repeat co-ordinates were within 10 kb of the UniGene genomic co-ordinates, then the repeat length hits was retained and merged into a single sd value. In practical terms this was not an issue, since only one of our most polymorphic repeats (sd \> 2) mapped to two clusters. There are certain limitations in using the GeneNote database to establish expression of repeat-containing genes. Specifically, the GeneNote microarray experiments were conducted with whole tissues, not tissues from particular tissue sub-types \[[@B42]\]. For example, users limiting their search to repeats expressed in the brain must bear in mind the possibility that a transcript highly expressed in one anatomical region (i.e. hippocampus) may lack sufficient global expression to be detected in the whole brain tissue used by the GeneNote experiments. Users interested in expression in particular anatomical regions might benefit from integrating gene expression data from their anatomical region of interest with repeat data from Satellog. As an example of the utility of Satellog, we wished to see how it might have expedited research for groups in the past hunting for candidate unstable repeats. In 1992, haplotype analysis of linkage disequilibrium data in Huntington\'s disease patients had indicated a portion of 4p16.3 (chr4:1-4,600,000) as the likely location of the mutation \[[@B47]\]. We assumed that the investigators at the time were looking specifically for an unstable, brain-expressed, CAG repeat to explain the disease phenotype, similar to SBMA \[[@B4]\]. Using the Satellog database, we narrowed down our search for candidates repeats in this area from 13,804 to 13 (Figure [5](#F5){ref-type="fig"}). Three polyglutamine repeats are returned by the database, but the repeat implicated in Huntington\'s disease (chr4:3108016-3108074) stands out as a strong candidate due to its size. If we re-run this query and select only the top 5% of repeats relative to their class, chr4:3108016-3108074 is the only polyglutamine repeat. These repeat characteristics: CAG repeat type, brain expression and presence within the top 5% of its repeat class, plus the privilege of hindsight, easily allow us to distinguish this repeat as the lead candidate in this region. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Candidate repeats within Huntington\'s disease linkage region 4p16.3. Sample output from Satellog summarizing candidate repeats within the 4p16.3 Huntington\'s disease linkage region. Coding CAG-type repeats from chr4:1-4,600,000 were selected along with their peptide sequence, HUGO names and ensembl gene IDs. The repeat encoding 19 glutamines has been associated with Huntington disease progression. ::: ![](1471-2105-6-145-5) ::: Secondly, we sought to prioritize all repeats in disease in which unstable repeats might play a role but in which none have been successfully correlated with disease to date. Schizophrenia is one such disease with genetic linkage in region 22q \[[@B48]-[@B50]\] suggesting some role of chromosome 22 abberations in disease development. Microdeletions in this region in patients affected with Velocardial Facial Syndrome (VCFS) confers the most consistent genetic predisposition to developing schizophrenia \[[@B51]\]. First, we collected all repeats on chromosome 22 resulting in a total of 113,789 repeats. Next, since we only observed trinucleotide repeats and higher period repeats in our disease-associated set, we restricted our repeats to those with a period greater than 2 resulting in 91,918 repeats. Since the majority of the disease-associated repeats had a significantly longer reference genome length relative to other repeats of the same class, we selected the 2,934 repeats with a percentile rank less than 0.05. The cellular pathology associated with schizophrenia shows no evidence of nuclear inclusions mediated by polyglutamine expansions, therefore, the disease phenotype may be mediated by an expansion in the UTR region. We selected 27 repeats from our set that were located in either the 5\'-or 3\'-UTR. Assuming that genes relevant to schizophrenia are expressed in the brain, we limited our analysis to the 18 repeats that were within genes expressed in the brain. Of our final set of 18 repeats, 2 repeats in the 3\'-UTRs of CRKLand NIPSNAP1had evidence of repeat polymorphism in UniGene clusters (Table [3](#T3){ref-type="table"}). In this prioritization paradigm, we did not look at any intronic repeats which may mediate the neurological phenotype by a mechanism similar to that of Friedreich\'s ataxia \[[@B37]\]. The point is that the prioritization paradigm can be defined by the user to dynamically generate a list of candidate repeats based on feature preference within Satellog or the fluctuating biological interpretation of repeat instability. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Candidate repeats within the chromosome 22 schizophrenia linkage region. ::: chr start end unit length p-value gene location name tissue mean sd --------- ----------- ---------- ---------- ------------ ------------- ------------------- ---------- ------------ ---------- -------- 22 19632267 19632294 AAC 9 0.019894 3utr CRKL Brain 8.04 0.51 22 28276064 28276078 GGCCT 3 0.017437 3utr NIPSNAP1 Brain 2.97 0.17 Candidate repeats within the chromosome 22 linkage region implicated in schizophrenia along with the tissue expression call in the brain and UniGene cluster summary statistics indicating mean repeat length and polymorphism (standard deviation (sd) values \> 0). ::: Conclusion ========== Satellog enriches the current bioinformatics landscape in which repeats are viewed. For example, the GAA repeat in Friedreich\'s Ataxia \[[@B37]\] is not detected at all (chr9:67,109,320-67,109,339) in the UCSC genome browser \[[@B33]\] by the TRF \[[@B32]\] and Variable Number Tandem Repeats (VNTR) tracks. The VNTR feature in UCSC detects all perfect 2 to 10 repeat units with 10 or more copies. Repeats detected by this method may over-represent insignificant low period repeats and under-represent potentially interesting high period repeats. In Satellog, not only is the Friedreich\'s Ataxia GAA repeat detected, but its percentile rank also suggests that this size of GAA repeat is a relatively rare observation in the human genome (percentile rank = 0.045). Satellog integrates disparate data sources to give researchers an idea of how interesting certain repeats are based on their genetic location, tissue expression profile and polymorphism within UniGene. It should be noted that Satellog does not intend to be a de novodetection method for disease-associated repeats. Instead, it provides comprehensive, integrated bioinformatics platform to prioritize repeats in a convenient and efficient manner. Satellog also presents the first comprehensive identification and integration of disease-associated repeats with other genomic resources for use as bioinformatics reagents in other studies. Satellog should prove useful to investigators interested in prioritizing repeats for typing in diseases showing anticipation or in which repeat polymorphism is thought to play a role in etiology. In addition, given that all sequence information (i.e. the human genome sequence and UniGene sequences) is from presumed \"normal\" individuals lacking disease phenotypes; Satellog may also prove useful in extending our understanding of the normal role of repeats in genes and transcripts. Methods ======= Software dependencies --------------------- A perl script \"repeatalyzer.pl\" functions as a wrapper for a number of different programs to achieve the endpoints of Satellog. repeatalyzer.pl is run with perl v5.6.1 and used BioPerl v1.2 \[[@B52]\], the EnsEMBL Perl API (May 24^th^, 1999 release), MySQL v10.8 Distribution 3.23.21-beta (for pc-linux-gnu), BLAT v. 28 \[[@B53]\] and v. 34 of the human genome sequence \[[@B54]\]. This script was run in parallel on a 192 node linux cluster at the BCCA Genome Sciences Centre. More detailed methods information is available at <http://satellog.bcgsc.ca>. Detecting microsatellite repeats with Tandem Repeats Finder (TRF) ----------------------------------------------------------------- We chose to detect sequences repeated at least twice and secondly, we were interested in exclusively pure repeat tracts which are more likely to expand following transmission \[[@B34]-[@B36]\]. Command-line TRF has seven parameters that can be manually assigned at run-time which include matching weight, mismatch and indel penalties, match probability, indel probability, minimum alignment score to report, and maximum period size to report \[[@B32]\]. We found that matching weight, mismatch and indel penalties, minimum alignment score and maximum period size directly affected the length and purity of hits detected by TRF whereas changing the match and indel probability features was not useful. The match and indel probability features refer respectively to the percent identity and fraction of indels tolerated in each serial tandem unit detected as a hit. These features allow users to specify alternative expected matching and indel statistical distributions. Next we evaluated the ability of the matching weight and maximum period size parameters to detect short repeats. Period size refers to the length of the tandemly repeated DNA unit, for instance CAG repeats have a period of 3. Since TRF hits must be at least 10 bp, the smallest hit for each repeat class reported in Satellog is 10 divided by the repeat unit length. For example, for CAG repeats, the smallest hit detectable that satisfies the minimum hit length is a 3 1/3 repeat unit hit (i.e. CAG CAG CAG C). In short, only pentanucleotide and larger repeats have a minimum of two repeat units in Satellog. Lastly we investigated the utility of adjusting the mismatch and indel penalties. We found that setting the penalty for these parameters to 4090 produced no impure repeats as hits. TRF was run on whole chromosome FASTA files from v. 34 of the human genome downloaded from the UCSC genome browser. Hit purity was confirmed by visually inspecting the top high period hits (these hits have the highest probability of introducing indels due to the scoring scheme used by TRF \[[@B32]\]. Identifying unique repeat classes --------------------------------- A repeat can be represented in a number of ways in double-stranded DNA. TRF detects repeats by the first tandemly repeated unit, therefore, CAGCAGCAG, AGCAGCAGC, and GCAGCAGCA are detected as repeats of CAG, AGC, and GCA respectively. Furthermore, the reference human genome sequence is only presented as the positive strand. Repeats of GTC, TCG, and CGT on the positive strand represent 5\'-\>3\' CAG, AGC and GCA repeats respectively on the negative strand. Therefore, to identify all CAG repeats in the human genome it\'s necessary to detect all CAG, AGC, GCA, GTC, TCG, and CGT repeats on the positive strand. We developed an algorithm to generate all possible sequence varieties of a repeat unit on the positive and negative strands. Our repeat classification algorithm operates by taking an input repeat unit, i.e. CAG, removing the first letter (C in this case) and appending it to the end of the remainder (AG) to create the second repeat unit (AGC). This is then reverse complemented to generate the equivalent sequence on the negative strand (TCG). This procedure is repeated repeat unit length -- 1 times to generate a unique identifier henceforth referred to as the repeat class. Each repeat in Satellog is associated with a single unique repeat class. Preparing AffyMetrix expression data from the GeneNote database --------------------------------------------------------------- The GeneNote (Gene Normal Tissue Expression) database provides baseline normal expression data of human genes for use in disease studies \[[@B42]\]. GeneNote data is downloaded from the Gene Expression Omnibus (GEO). A total of twelve human tissue profiles are presented in GeneNote including bone marrow, brain, heart, kidney, liver, lung, pancreas, prostate, skeletal muscle, spinal cord, spleen, and thymus. These products were generated with the AffyMetrix HG-U95 A-E probe-set, covering 62,839 probe-sets. EnsEMBL genes have been mapped to AffyMetrix HG-U95 probes by the EnsEMBL project \[[@B41]\]. Once a repeat is detected either inside or within 60 kb of an EnsEMBL gene, that gene\'s normal expression profile is evaluated by cross-referencing its AffyMetrix tags to the GeneNote database within Satellog. Detecting repeat polymorphisms within UniGene clusters ------------------------------------------------------ UniGene contains the largest public repository of transcribed human sequence and represents an attempt to organize this wealth of expression data into discrete transcriptional loci \[[@B40]\]. All human UniGene sequences were processed for use with repeatalyzer.pl. For each repeat detected in UTR or coding sequence, the repeat plus 10 bp of flanking sequence was extracted from EnsEMBL and queried using the BLAT algorithm \[[@B53]\] against a BLAT-formatted database created from sequences representing the longest, highest quality stretch of DNA from each individual UniGene cluster (pre-selected by UniGene as the file Hs.seq.uniq). Polymorphism is evaluated only if BLAT analysis against all UniGene clusters resulted in 1) hits that achieved BLAT scores at least 85% of the theoretical maximum for a perfect hit 2) 90% of the query sequence matched identically within the cluster 3) the repeat mapped within 10 kb of the genomic co-ordinates of the UniGene cluster. If a hit to a UniGene cluster satisfied these criteria, the length of the repeat in the cluster is stored in Satellog. This feature allows investigators to query all repeats with polymorphisms in UniGene clusters from genomic regions of interest. repeatalyzer.pl overview ------------------------ Once the above software and data dependencies are configured, repeatalyzer.pl automatically populates Satellog (Figure [6](#F6){ref-type="fig"}). The script processes the flat files output by TRF. These files contain the repeat co-ordinates plus the repeat period (the size of the repeated unit), the sequence of the individual repeat unit, the entire repetitive sequence and the repeat length. Repeat co-ordinates are passed to the EnsEMBL API to confirm the authenticity of the co-ordinates generated by TRF. If the repeat is not detected within a gene with the EnsEMBL API, then progressively larger slices incrementing by 15 kb are taken in search of flanking genes. As soon as a gene is located in flanking sequence then no further flanking sequence is collected. However, if no genes are detected within 60 kb of the repeat co-ordinates then repeatalyzer.pl stops searching for genes. If a repeat is detected inside or within 60 kb adjacent to an EnsEMBL-defined gene then that gene\'s primary information (co-ordinates, HUGO name, EnsEMBL ID and description) are collected along with metadata stored in EnsEMBL such as Protein Data Bank (PDB) \[[@B55]\], Online Mendelian Inheritance in Man \[[@B40]\], Gene Ontology (GO) \[[@B56]\], and mappings to AffyMetrix probe sets. If the repeat is located in the 5\'-UTR, 3\'-UTR, or coding sequence of a gene then its polymorphism profile within UniGene clusters is evaluated. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### repeatalyzer.pl flowchart. Flowchart outlining how repeatalyzer.pl populates the Satellog database. ::: ![](1471-2105-6-145-6) ::: Generating a measure of repeat length significance -------------------------------------------------- After running the script to populate Satellog, each repeat\'s length is compared to the lengths of all repeats of the same repeat class. The majority of repeats associated with disease undergo expansions from already large reference genome lengths relative to other repeats of the same class \[[@B5]\]. Each repeat\'s percentile rank is calculated from the distribution of repeat lengths within each repeat\'s class. It reflects the proportion of repeats with the same or greater length from the repeat class\' genomic distribution. Authors\' Contributions ======================= PIM conceived of the study, wrote all analysis scripts, collected and input data into the database, analyzed the data, directed the Satellog website design, wrote all documentation and the tutorial accompanying the database and drafted the manuscript. CRM developed the online graphical user interface for the database, troubleshooted and re-indexed queries for the database and provided technical expertise for realizing the web version of Satellog. SLB participated in the design of the study and gave crucial intellectual direction to the final manuscript. BFFO participated in the design of the study and provided assistance with bioinformatics analysis. RSD provided key biological background to guide the design of the study. BRL participated in the design and strengthened the clinical perspective of the final manuscript. RAH participated in the study design, coordination, performed data analysis and gave critical direction to the final manuscript. All authors read and approved the final manuscript. Appendix ======== Figure S1 -- Repeat density (bp of repeat sequence / Mb) per human chromosome. Available online at: <http://satellog.bcgsc.ca/source.php>. Table S1 -- Total repeat count and density by chromosome Available online at: <http://satellog.bcgsc.ca/source.php>. Table S2 -- Repeat period count and density by chromosome Available online at: <http://satellog.bcgsc.ca/source.php>. Table S3 -- Repeat unit count and density by chromosome Available online at: <http://satellog.bcgsc.ca/source.php>. Table S4 -- Repeat unit count and density by gene region Available online at: <http://satellog.bcgsc.ca/source.php>. Acknowledgements ================ 1\) UBC/SFU CIHR Training Program for Bioinformatics in Health Research, Rooms 308/308A, 2206 East Mall, University of British Columbia, Vancouver, BC, V6T 1Z3, Canada 2\) Mark Mayo and Bernard Li at the BCCA Genome Sciences Centre for technical support with cluster computing. 3\) Martin Krzywinski for creating the Satellog logo.	PubMed Central	2024-06-05T03:55:59.775142	2005-6-10	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181805/", "journal": "BMC Bioinformatics. 2005 Jun 10; 6:145", "authors": [ { "first": "Perseus I", "last": "Missirlis" }, { "first": "Carri-Lyn R", "last": "Mead" }, { "first": "Stefanie L", "last": "Butland" }, { "first": "BF Francis", "last": "Ouellette" }, { "first": "Rebecca S", "last": "Devon" }, { "first": "Blair R", "last": "Leavitt" }, { "first": "Robert A", "last": "Holt" } ] }
PMC1181806	Background ========== Eukaryotes need to distinguish their individual chromosomes in several essential processes. For example, homologous chromosomes must be aligned during meiosis and the nuclear positioning of specific chromosomes in interphase is conserved, controlled and important for correct gene expression \[[@B1],[@B2]\]. Therefore, there must be certain molecules, most likely proteins, that recognise chromosome-specific features. However, few protein complexes that recognise a specific chromosome and bind at multiple points along its entire length are known. One example is the MSL ribonucleoprotein complex of Drosophila melanogaster, a dosage compensation factor that equalises expression of sex-linked genes between males and females -- an essential process in animals with an XX-XY mode of sex determination \[[@B3]-[@B5]\]. In this species an autosome-specific protein has also been discovered; the Painting of fourth (POF) protein, which binds exclusively to the 4^th^chromosome \[[@B6]\]. The chromosome specificity of POF and MSL appears to have been conserved for a long period of evolutionary history, suggesting that chromosome-specific identifying features have also been conserved and have functional significance in Drosophila\[[@B7]\]. The fact that individual chromosomes can be uniquely targeted raises questions about how they are recognised. There are no known DNA targets that direct the binding of either the MSL complex or POF, and no sequences have previously been found that are chromosome-specific and distributed over its entire length. The development of appropriate computational methods is essential for linking functions to linear sequences. The toolbox for finding cryptic and complex sequence targets is still developing and most algorithms require extensive optimisation or a known motif \[for a review of current methods see \[[@B8]\]\]. Standard methods, e.g. BLAST and alignment approaches used in these kinds of analyses have several limitations, for instance they are neither exhaustive nor unbiased. To circumvent these limitations, Brāzma et al. \[[@B9]\] successfully combined a pattern discovery algorithm with clustering of expression data to predict gene regulatory elements in yeast \[see also \[[@B10]\]\]. We present here an alternative way to analyse large amounts of sequence data using multivariate statistics, combined with cytological observations and full genome annotations, to find sequence signatures composed of combinations of sequence motifs correlated to chromosomal regions without imposing any predefined assumptions. The multivariate approach is efficient in finding weak signals in large amounts of data. The method is neither biased nor heuristic, but still very fast. Abe et al. \[[@B11]\] used a related approach to study large-scale differences between distant genomes, and Bultrini et al. \[[@B12]\] used sequence motifs and multivariate statistics to find vocabularies defining intron regions in Drosophila melanogasterand Caenorhabditis elegans. In the study reported here, our aim was to use a multivariate approach to identify sequence signatures correlated to chromosome identity in Drosophila, and if possible link these signatures to function. The D. melanogastergenome sequence (release 3) \[[@B13]\] and its annotation (release 3.2) \[[@B14]\] have been thoroughly revised since their first releases and the cytology of the salivary gland polytene chromosomes (X, 2, 3, 4) provides a powerful tool for genome studies. Although the D. yakuba(2004-04-07 assembly) and D. pseudoobscura(Freeze 1) genome assemblies are not as complete as that of the D. melanogasterassembly, they provide valuable resources for attempts to identify conserved, potentially functional sequences. The three Drosophilaspecies examined in this study (D. yakuba, D. pseudoobscuraand D. melanogaster) all belong to the Sophophorasubgenus and are hereafter referred to as Dy, Dpand Dm, respectively. Dmand Dyboth belong to the melanogasterspecies group and Dpto the obscuraspecies group. Dmand Dyare separated by approximately 12.8 Myr and they are both separated from Dpby roughly 54.9 Myr \[[@B15]\]. Results ======= Whole chromosome analysis ------------------------- To construct data sets for a whole genome analysis, we scored all positions of all possible di-(16), tri-(64), tetra-(256), penta-(1024) and hexa-mers (4096) in the genome sequence of Dm, Dyand Dp. PCA (Principal Component Analysis) of the scores clearly separated the Muller\'s F-elements (the term F-element is used here because this chromosome is the 4^th^in Dm/Dyand the 5^th^in Dp\[[@B16]\]) from all other chromosomes along the first component (Figure [1A](#F1){ref-type="fig"}). The second component discriminate the non F-element chromosomes into two groups: one containing Dpchromosomes and the other containing the Dm/Dychromosomes (Figure [1A](#F1){ref-type="fig"}). The same pattern was observed when the di-, tri-, tetra-and penta-mers were analysed (data not shown). However, a large amount of the variation in the first component can be explained by differences in nucleotide composition between the chromosomes (Figure [1B](#F1){ref-type="fig"}, Table [1](#T1){ref-type="table"}). The sequence motifs that most strongly distinguish the F-elements contain only A/T nucleotides and are not very complex. To determine if more complex motifs can be used to separate the chromosomes, we need to remove most of the variation caused by the inequalities in their A/T contents. This was accomplished by dividing all scores by the expected scores, based on the chromosomal base composition. We then normalised the scores for all di-, tri-, tetra-, penta-and hexa-mer sequence motifs. After this normalisation the chromosomal separation was almost identical to the separation seen in the non-normalised PCA (Figure [1C](#F1){ref-type="fig"} shows results from the tetramer analysis) except when using penta-and hexamers. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Results of the PCA of whole chromosome sequences from Dm(○), Dy(△) and Dp(+). Chromosomes are colour-coded, as follows (according to the Dmnumbering: black = X, yellow = 2, blue = 3 and red = 4). L and R stand for the left and right arms of the metacentric chromosomes, respectively. (A) Score plot (R2 = 0.87) of the non-normalised hexamer analysis. (B) Loading plot of the analysis in (A). (C) Score plot (R2 = 0.84) of the normalised tetramer analysis. (D) Loading plot of the analysis in (C). The colouring of the hexamers in (B) and (D) is proportional to the A/T content. Pink is all A/T and blue is all G/C. ::: ![](1471-2105-6-158-1) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### The length, number of N and A/T content of all chromosomes used in this study. ::: Original sequence length \% N \% A/T \% removed by Tandem Repeats Finder \% A/T after Tandem Repeats Finder masking \% Removed by RepeatMasker \% A/T after RepeatMasker masking ------ -------------------------- ------- -------- ------------------------------------- -------------------------------------------- ---------------------------- ----------------------------------- Dm X 21780003 0.10 57.42 2.12 57.34 8.80 56.55 2L 22217931 0.01 58.08 0.81 58.07 6.59 57.54 2R 20302755 0.02 56.55 0.91 56.54 7.88 56.05 3L 23352213 0.05 57.92 0.87 57.93 6.77 57.36 3R 27890790 0.00 57.08 0.68 57.07 5.33 56.60 4(F) 1237870 0.08 64.71 1.17 64.53 26.70 64.58 Dy X 21591847 3.28 56.73 3.45 56.58 2L 22678881 1.31 57.19 1.56 57.18 2R 21288905 1.39 56.74 1.33 56.75 3L 24977971 2.19 57.57 1.57 57.57 3R 29717196 1.88 56.81 1.47 56.82 4(F) 1395135 2.16 64.53 3.38 64.51 Dp XL 24630256 4.10 54.22 2.86 54.33 4 26108043 3.64 55.99 2.31 56.02 3 19738113 3.41 53.52 1.50 53.61 XR 24186629 3.87 53.76 3.63 53.96 2 25998849 3.13 55.10 1.86 55.15 5(F) 849497 25.81 61.45 1.02 61.42 The percentages of base pairs removed by Tandem Repeats Finder and RepeatMasker as well as the A/T content of the remaining sequences are also given. The Dpchromosomes are listed in the same order as the corresponding Dmchromosomes. ::: Analysis of the sequence motifs shows that the F-element separation is no longer solely explained by A/T motifs (Figure [1D](#F1){ref-type="fig"}). In the analyses using penta-and hexa-mer motifs the DpF-element is more similar to the non F-element chromosomes and the Dm/DyF-elements separates more from each other (data not shown). The reason for this became clear when the different genomes were separately analysed. In all three species, the F-element separated from the other chromosomes along the first component, regardless of the motif length used (results of the hexamer analysis are shown in Figure [2](#F2){ref-type="fig"}). In Dm/Dythe X chromosome was separated from the other chromosomes by the second component, although less markedly than the F-element. Interestingly, the left arm of chromosome X in Dpseparates in the second component while the right arm clusters closer to the other chromosomes. This is in agreement with the hypothesis that the right arm of DpX is a later addition \[[@B15]\]. The left arms of DmX, DyX and the DpX are separated by the same hexamers. Many of the motifs causing the strong separation of the F-elements are the same in all three species. The top scoring penta-and hexamers can easily be aligned into longer motifs (Figure [3](#F3){ref-type="fig"} shows results from the Dmhexamer analysis), all of which are supported by hexamers in both sense and anti-sense orientation. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Results of the separate, normalised, whole chromosome PCA of the three genomes using hexamers. Chromosomes are colour-coded, as follows (according to the Dmnumbering: black = X, yellow = 2, blue = 3 and red = 4). L and R stand for the left and right arms of the metacentric chromosomes, respectively. (A) Score plot (R2 = 0.97) of the Dmanalysis. (B) Loading plot of the analysis in (A). (C) Score plot (R2 = 0.99) of the Dyanalysis. (D) Loading plot of the analysis in (C). (E) Score plot (R2 = 0.92) of the Dpanalysis. (F) Loading plot of the analysis in (E). The colouring of the hexamers in (B), (D) and (F) is proportional to the A/T content. Pink is all A/T and blue is all G/C. ::: ![](1471-2105-6-158-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Graph showing the 50 hexamers with the highest loadings in the normalised DmPCA. The combination of the eight hexamers into the nonamer is shown. The two hexamers not included in the motif are indicated by open boxes. ::: ![](1471-2105-6-158-3) ::: For further analysis, we excluded parts of the longer motifs supported by only sense or only anti-sense hexamers. It should be noted that after the hexamers included in the longer motifs, there was a clear drop in loading (Figure [3](#F3){ref-type="fig"}). This suggests that a longer motif causes overrepresentation of the top scoring hexamers in the F-element. To verify the existence and overrepresentation of these predicted longer motifs, we counted their numbers on all chromosomes. The longer motifs are clearly more common on the F-elements compared to the other chromosomes (Table [2](#T2){ref-type="table"}). In Dm/Dythe F-element motif is the same, but in Dpthe motif is different (Table [2](#T2){ref-type="table"}). We find the hexamers forming the longer motifs only in the normalised analysis. Our normalisation procedure assumes a random distribution of all nucleotides. If the nucleotide frequencies on, e.g. the DmF-element would be highly influenced by micro satellites or more complex repeats, it would make the normalisation assumption invalid. However, this is not the case. When we removed all simple sequence repeats using Tandem Repeats Finder \[[@B17]\] or both simple and complex repeats using RepeatMasker \[[@B18]\] the chromosomal nucleotide compositions stayed roughly the same (Table [1](#T1){ref-type="table"}). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### The number of longer motifs and pairs of motifs per Mbp on the different chromosomes. ::: F-element motifs / Mbp Pairs / Mbp Sum of included hexamers / Mbp ------ ------------------------ ------------- -------------------------------- Dm gtgggcgtg/cacgcccac X 55.8 4.5 2069.1 2L 42.4 2.6 1770.3 2R 44.1 3.1 1926.8 3L 39.5 2.3 1761.3 3R 36.7 1.3 1867.7 4(F) 154.4 39.6 1875.7 Dy gtgggcgtg/cacgcccac X 74.4 17.8 2247.3 2L 115.9 51.6 2319.5 2R 111.2 47.8 2337.3 3L 90.6 36.7 2132.7 3R 116.4 53.4 2399.1 4(F) 597.1 325.3 4810.2 Dp tacatatgta XL 102.5 16.4 3269.7 4 44.8 4.1 2319.1 3 66.2 6.8 2588.3 XR 56.2 5.5 2408.1 2 41.5 3.7 2313.3 5(F) 244.3 23.8 6890.8 The sum of the hexamers making up the longer motifs per Mbp are also given. ::: Since Dmis by far the most intensively studied of the three species we concentrated our efforts on the nine base pair long motif found in the DmF-element. Strikingly, in the F-element this nonamer is often found in pairs, i.e. two sense or two anti-sense nonamers are often situated close to each other. Furthermore, when we plotted the distances between the 192 motifs in the F-element we found that three different distances between them are overrepresented. Considering the 192 nonamers as 91 pairs, nine are separated by 17 (± 2) bp, 13 by 28 (± 1) bp and 10 by 79 (± 3) bp. In the subsequent analyses, we defined a pair as two nonamers separated by no more than 146 bp. According to this definition, 51% of the nonamers are organised in such pairs. The remaining nonamers seem to be randomly distributed in relation to each other. Only one pair on the entire F-element consists of one sense and one antisense nonamer. The nonamer pairs are even more enriched on the F-element than the nonamer (Table [2](#T2){ref-type="table"}). To assess whether this frequency of pairs is higher than expected by mere chance we randomised the positions of all 192 nonamers in a simulation repeated 10 million times, and calculated the number of pairs in each case. Since less than 51% of the nonamers were paired in every run, we conclude that the observed nonamer pair frequency significantly exceeds the expected frequency. This strongly suggests that the nonamers exist in pairs, are important for the separation of the F-element in Dmand might confer a selective advantage. The Dynonamers and Dpdecamers also occur in pairs (Table [2](#T2){ref-type="table"}). One of the atypical features of the DmF-element is the specific binding of the protein POF. To determine if the nonamers or nonamer pairs are correlated to the binding of POF to the F-element, we mapped POF binding sites on polytene chromosomes (Figure [4A,B](#F4){ref-type="fig"}). It is difficult to map polytene bands beyond cytological position 102E5 so we limited this analysis to the region 102A-102E5. Comparison of the sequence positions of the nonamer pairs (Figure [4C](#F4){ref-type="fig"}) with the staining pattern of POF protein on the polytene F-element (Figure [4B](#F4){ref-type="fig"}) showed that regions with few or no pairs correlate well with regions lacking POF binding. The genomic sequence corresponding to the cytological regions that do not bind POF comprises 59% of the sequence from positions 1 to 830,000. 79% of the nonamer pairs and 61% of the nonamers are located outside these regions. We tested the significance of these results in a simulation, repeated 10 million times, in which we randomised the positions of the nonamers and the nonamer pairs. In all of these simulations the number of nonamers or pairs was lower than the observed numbers in the POF-binding regions. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Localisation and mapping of POF on salivary gland F-element in Dm. (A) F-element stained with DAPI, showing cytological map positions. (B) F-element stained with anti-POF antibody. White lines indicate regions with weak POF staining. (C) Number of nonamers (grey) and nonamer pairs (black) per 10 kbp. The x-axis scale is in 100 kbp. White areas indicate regions of weak POF staining seen in (B). The sequence positions of the polytene chromosome bands are according to Dmannotation release 3.2. The genomic sequence 1 to 830 kpb shown in (C) corresponds to cytological region 102A to 102E5. ::: ![](1471-2105-6-158-4) ::: The separation of the X chromosome seen in Figure [2A,C,E](#F2){ref-type="fig"} is due to simple sequences such as A~n~, T~n~, C/A~n~and G/T~n~repeats in both the non-normalised and the normalised analysis. This finding is in agreement with in situhybridization data showing that C/A~n~and G/T~n~repeats are common on the X chromosome \[[@B19]\]. Positions of the hexamers that separate chromosome X show no clear correlation to the binding sites of the MSL complex defined by Demakova et al. \[[@B20]\] (data not shown). In an effort to determine the origin of the sequences causing the chromosomal separation in Dmseen in both the non-normalised and normalised PCA we repeated the analysis on three additional data sets. To evaluate the contribution of simple sequence repeats we masked the genome using Tandem Repeats Finder \[[@B17]\] and to evaluate the contribution of both simple and more complex repeats we used RepeatMasker \[[@B18]\]. We also merged all exon sequences of the different chromosomes. We then analysed the four datasets simultaneously, both with and without normalisation (Figure [5](#F5){ref-type="fig"}). The resulting plots show that the enrichment of simple A/T rich sequences on the F-element (seen in the non-normalised PCA, Figure [1B](#F1){ref-type="fig"}) cannot be explained by differences in repetitive elements. These sequence signatures were not removed by masking simple or more complex repetitive elements, implying that they are present in all non-exon sequences on the F-element (Figure [5A](#F5){ref-type="fig"}). Interestingly, the F-element exons do not share these sequences, but they still clearly separate from the exon sequences of the other chromosomes. Furthermore, the simple sequences that separate the X chromosome from the others distribute all over the non-exon sequences. In the PCA in which we accounted for differences in nucleotide composition, the separation was similar compared to the non-normalised analysis, except that the exons of the X chromosome separated from the exons of the other chromosomes (Figure [5B](#F5){ref-type="fig"}). It should be noted that the first component distinguishes between the exon sequences and the other sequences. The second component, however, separates all types of F-element sequences from the other chromosomal sequences. We conclude that the overrepresentation of some sequence signatures on the F-element cannot be attributed to either the high A/T content or the enrichment of repeated elements and that they are present in both exon and non-exon sequences. The general patterns we see are clearly not dependent on the type of sequence studied or differences in base composition. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### The combined PCA of four Dmdata sets using hexamers. The four data sets used are: the original sequence (□), the Tandem Repeats Finder masked sequence (▽), the RepeatMasker masked sequence (\) and the extracted exon sequences (◇). (A) Score plot (R2 = 0.93) of the non-normalised analysis. (B) Score plot (R2 = 0.82) of the normalised analysis. ::: ![](1471-2105-6-158-5) ::: In addition, we note that in the normalised PCA the RepeatMasked F-element separates more clearly from the original F-element sequence (Figure [5B](#F5){ref-type="fig"}) than in the non-normalised analysis. Many sequence signatures are shared by the F-elements in all four datasets. Examination of the top-scoring sequence motifs clearly shows that the RepeatMasked F-element lacks the nonamer motif described above (data not shown). We therefore studied the output file from RepeatMasker in further detail. According to RepeatMasker, 95.3% of the nonamer motifs reside within DINE-1 elements, and thus seem to be closely linked to them. The DINE-1 element has previously been shown by in situhybridisation to be enriched on the DmF-element \[[@B21]\]. We also note that in the DINE-1 sequence defined in the Repbase Update \[[@B22],[@B23]\] there is a duplication of approximately 60 base pairs, each of which contains a nonamer pair, and in both pairs the individual nonamers are separated by 29 base pairs. We also masked the genomes of Dyand Dpusing Tandem Repeats Finder. Since these genomes have not yet been annotated we could not use exons or RepeatMasker. The PCA results of the original sequences and the masked sequences in these species are virtually identical (data not shown). Fragment analysis ----------------- In the whole chromosome analysis we identified sequence signatures that are enriched on different chromosomes, but we did not investigate their linear organisation along the chromosomes. Therefore, to find sequence signatures evenly distributed over the chromosomes that are capable of distinguishing one chromosome from the others, we fragmented each of the Dm, Dyand Dpgenomes into 100 kb fragments. We then scored the positions of all possible di-, tri-, tetra-, penta-and hexa-mers in the 100 kb fragments of all chromosomes from each of the genomes. The first component of a PCA of these data mainly reflects differences in nucleotide composition between the fragments. Since the nucleotide composition can vary both between chromosomes and within single chromosomes we need to remove this variation in the dataset. One possibility would be to exclude the first component, but some of the variation caused by A/T skewing could still remain in the higher order components. To specifically remove the influence of variations in the base composition we created a Partial Least Squares (PLS) model using the non-normalised hexamer scores and the A/T content as a single response. We then used the residual matrix, after removing the variance described by the first component, for subsequent PCA analysis. The residual matrix is a normalised scoring matrix in which the variance in the data related to the base composition of the target sequence has been removed. The performance of the normalisation was evaluated by plotting the score values of the first component against the base composition of the fragments. As expected, the scores showed an almost perfect correlation with the base composition of the fragments (data not shown). PCA of the approximately 3600 fragments from all three species showed that the 33 F-element fragments cluster, and separate with minor overlaps from the other chromosomal fragments in the second component (Figure [6](#F6){ref-type="fig"} shows results from the hexamer analysis). In the tri-and tetra-mer analyses, the overlap with other chromosomes was more extensive than in the di-, penta-and hexa-mer analyses (data not shown). In the first component of the hexamer PCA, roughly a third of the Dpfragments cluster separately from other chromosomal fragments. The third component separates many of the Dm/DyX chromosomal fragments from the others, but only when using penta-and hexa-mers (data not shown). The sequence signatures responsible for the separation of the F-element are not the same as in the whole chromosome analysis and cannot easily be combined into longer motifs. For a full listing of the loadings for all 4096 hexamers for the first two components in the PCA see Additional file [1](#S1){ref-type="supplementary-material"}. In conclusion, the fragment analysis showed the existence of F-element-specific sequences that not only have been conserved for approximately 54.9 Myr, but also are linearly distributed along the sequenced part of the F-elements in Dm, Dyand Dp. Based on this conservation we speculate that there are sequence signatures that have a function for F-element identity. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### The hexamer PCA (R2 = 0.22) of 100 kb fragments (n = 3564) of the three genomes Dm(○), Dy(△) and Dp(+). Chromosomes are colour-coded, as follows (according to the Dmnumbering: black = X, yellow = 2, blue = 3 and red = 4). The loadings are presented in [Additional file 1](#S1){ref-type="supplementary-material"}. ::: ![](1471-2105-6-158-6) ::: When we plotted the scores from the second component (which separates the F-elements) against the chromosomal position we find that on average the Dpfragments are shifted towards the F-element fragments (Figure [7](#F7){ref-type="fig"} shows results from the hexamer analysis). The centromere proximal regions of the non F-element chromosomes in all species are shifted towards the F-element fragments and the distal regions in the opposite direction. This pattern is not as clear in Dpas in Dmand Dy. ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Scores from the second component in Figure 6 plotted against the linear order of the 100 kb fragments on the individual chromosomes from Dm(○), Dy(△) and Dp(+). Chromosomes are colour-coded according to the Dm/Dynumbering (black = X, yellow = 2, blue = 3 and red = 4). This should be noted when examining the Dpfragments. Proximal regions of Dmand Dychromosomes are indicated by arrows. ::: ![](1471-2105-6-158-7) ::: In the same way as for the whole chromosome study, we repeated the fragment analysis on chromosomes from the three species after masking them by Tandem Repeats Finder. The results from this masked dataset did not differ in any significant way from the prior analysis (data not shown). For Dm, we also masked the fragmented genome using RepeatMasker. A combined PCA with the original data, Tandem Repeats Finder masked data and RepeatMasker masked data showed that the F-element signatures distributed over the entire chromosome are not connected to either simple or complex sequence repeats (Figure [8](#F8){ref-type="fig"} shows results from the hexamer results). In this analysis many X chromosomal fragments separated from the other fragments. ::: {#F8 .fig} Figure 8 ::: {.caption} ###### The combined PCA of 100 kb fragments (n = 3399) of three Dmdata sets based on hexamers. The three data sets used are: the original sequence (□), the Tandem Repeats Finder masked sequence (▽) and the RepeatMasker masked sequence (\). Chromosomal origins of the fragments are indicated by colour (black = X, yellow = 2, blue = 3 and red = 4). ::: ![](1471-2105-6-158-8) ::: Interestingly, we note that in every PCA we performed most motifs had almost identical loading to their reverse complements. This was true for both the whole chromosome analysis and the fragment analysis, regardless of whether normalisation was applied and the motif length used. Baisnée et al. \[[@B24]\] have studied the reverse complement symmetry of DNA more thoroughly, but even though it seems to be universal, the underlying cause is not yet fully understood. Discussion ========== Sequence signature analysis --------------------------- In this work, we separately counted all di-(16), tri-(64), tetra-(256), penta-(1024) and hexa-mers (4096) and studied their distribution in the chromosomes of three Drosophilagenomes using PCA. Short motifs (up to tetramers) can be rapidly scored and analysed. However, the frequencies of such short motifs are strongly influenced by the abundance of simple sequence repeats. Motifs longer than tetramers are less affected by simple sequence repeats, but are computationally more demanding to analyse. Sometimes, when a group of sufficiently long sequences, e.g. hexamers, are found to be overrepresented in a genomic sequence, they overlap and form longer sequences with higher discriminative power, thus increasing the chance of identifying longer and more complex sequences than if shorter sequences, e.g. trimers, are used. The frequency of a sequence motif depends on both biological and stochastic factors. The expected frequency of a specific motif depends on the base composition of the chromosome. If the four nucleotides do not have equal frequencies in all chromosomes, the results from a non-normalised analysis will reflect the effects of a mixture of biological and stochastic factors. It is often difficult to isolate the effects of such factors, but a large part of the stochastic component can be removed by dividing all motif frequencies by the expected frequencies in a normalisation step. Otherwise, biologically interesting motifs may be masked by motifs that are common solely by chance. In this study, we used relatively basic normalisation procedures to account for differences in base composition. However, our multivariate approach could easily be extended to account for differences related to sequence complexity \[see e.g. \[[@B25]\]\] or any kind of prior knowledge about the target sequence. Whole chromosome analysis ------------------------- In many respects, the F-element in Dm(the 4^th^chromosome) is an atypical chromosome. It has an overall length of \~5 Mb, 3--4 Mb of which consists of simple satellite repeats and does not contain any known genes \[[@B26]\]. The remaining portion (1.23 Mb) has been sequenced and covers the cytogenetic bands 101E-102F on polytene salivary gland chromosomes. However, the banded portion appears to be a mosaic of unique DNA interspersed with moderate and low copy repetitive DNA \[[@B21],[@B27]-[@B30]\]. The F-element is largely heterochromatic in nature. The heterochromatic protein HP1 and the modified histone, methylated H3Lys9, have been found to be associated with most of the F-element \[[@B31],[@B32]\]. In accordance with its heterochromatic nature, the F-element has a higher A/T content compared to the other chromosomes. A high density of transposable elements (approximately six times higher than in the other chromosomes) is found in the DmF-element \[[@B33]\]. Another interesting feature of the F-element is that it is decorated by the chromosome-specific protein, POF (Painting of fourth), which specifically \"paints\" the entire chromosome \[[@B6]\]. The F-element is an atypical autosome and has been suggested to have a closer kinship with the X chromosome than with the other autosomes \[[@B16],[@B34]\]. The F-element has been suggested, partly on the basis of studies of the distant relative D. busckii, to originate from the X chromosome \[[@B35],[@B36]\]. The binding of POF to the F-element is reminiscent of the binding of the Drosophiladosage compensation complex to the male X chromosome, which mediates its hypertranscription \[reviewed by \[[@B4],[@B5]\]\]. In D. busckii, POF binds to the male X, further supporting the suggested relationship between the X chromosome and the F-element \[[@B6]\]. All chromosomes differ to some extent in nucleotide frequencies, with the F-element being extreme in this respect, having a high A/T content in all three species studied. When the raw data was analysed the F-elements in all three species separated collectively from the other chromosomes (Figure [1](#F1){ref-type="fig"}), due to differences in their contents of simple sequences containing only A and T. In Dmwe performed the analysis on four datasets, derived from the original sequence, and the sequences obtained after masking simple sequence repeats, both simple and more complex repeats and after removing everything except the exon sequences. The results show that the simple A/T sequences, which separate the F-element in the original data, are distributed throughout the non-exon F-element sequences and cannot be attributed to microsatellites and transposable elements. It should also be noted that the F-element exons separate equally well from the exons of other chromosomes. The X chromosome also separates from the other chromosomes, albeit to a lesser extent, due to differences in their simple sequences. The same chromosomal separation is seen regardless of the motif length used. As shown in Figure [1](#F1){ref-type="fig"}, all of the Dpchromosomes are shifted relative to the Dm/Dychromosomes, suggesting the presence of Dp-specific signatures in addition to the chromosome-specific signatures studied here. To detect more complex and potentially functional motifs hidden by the skewed base composition, we normalised our scores according to the base composition of each chromosome analysed. As shown in Figure [1C](#F1){ref-type="fig"}, the resulting separation was nearly identical to that seen in the non-normalised analysis (Figure [1A](#F1){ref-type="fig"}). The DmF-element was clearly separated even after removal of repeated elements from the genome (Figure [5B](#F5){ref-type="fig"}). It should be noted that the first component in this PCA (Figure [5B](#F5){ref-type="fig"}) distinguishes the exons from non exon sequences. In the second component, however, all F-element sequences including the exon sequences, cluster together. We conclude that the F-element exons also contain F-element signatures. The separate analysis of the three species showed that the pentamer and hexamer motifs that are most important for distinguishing the F-element can be aligned into longer sequences. Examination of the top scoring hexamers clearly shows that they are part of a nonamer in Dmand Dy, and of a decamer in Dp. These sequences are strongly enriched in the respective F-elements (Table [2](#T2){ref-type="table"}), although the individual hexamers in Dmare not enriched in the non-normalised analysis. Since Dmis the only annotated species, we concentrated our investigation on the Dm/Dynonamers. Plotting the positions of these nonamers in the DmF-element showed that they commonly occur in pairs, separated by no more than 146 bp, all but one of which consists of two sense or two anti sense nonamers. The individual nonamers are enriched roughly four-fold in the F-element, while the pairs are enriched about 15-fold. The nonamers and decamers are also organised in pairs in Dyand Dprespectively (Table [2](#T2){ref-type="table"}). We conclude that even though the method is based on relatively short sequence motifs, it still provides a potent means for finding longer and more complex sequence motifs. Since POF is a protein that specifically paints the DmF-element, we tested the possibility that the nonamer or nonamer pairs may be correlated to POF-binding sites. For this purpose, we stained polytene chromosome preparations using POF antibodies. After carefully mapping the banded regions, we compared the positions of nonamer pairs to the POF staining pattern. The genomic regions with few or no pairs correlate well with regions on the F-element that do not bind POF (Figure [3](#F3){ref-type="fig"}). We hypothesise that the nonamer pairs have a function and are directly or indirectly involved in POF binding to the F-element in Dm. However, this hypothesis needs to be verified experimentally. Since POF will not bind to a translocated DmF-element \[[@B6]\] the nonamer pairs are not sufficient by themselves for recruiting POF. If the pairs have a function, it is possible that some variation is allowed within the nonamer and that there are motifs of differing strength. According to our RepeatMasker analysis of the F-element, 95.3% of the nonamers are located within DINE-1 elements. As shown in Figure [2B](#F2){ref-type="fig"}, the hexamers forming the nonamer are important for the separation of the F-element. Nevertheless, after removing virtually all of the nonamers using RepeatMasker (Figure [5B](#F5){ref-type="fig"}) F-element separation was retained, indicating that other signatures, apart from the nonamer, help distinguish the DmF-element. In an extensive study of local deletions flanking a transpsoson reporter Sun et al. \[[@B37]\] showed that the genomic region 400000 to 440000 of the DmF-element is euchromatic. A nearby region induces gene silencing and is therefore considered to be heterochromatic. Sun et al. \[[@B37]\] attribute this to local induction of heterochromatin by the 1360 (hoppel) element. According to their model the 1360-induced heterochromatin can spread, but only \~10 kb or until encountering competition from euchromatic determinants. In this context, it should be stressed that this \"euchromatic\" region is enriched in DINE-1 fragments containing the nonamer pair region. We speculate that these nonamer-containing DINE-1 fragments act as euchromatic determinants. We have previously proposed that POF is involved in a chromosome-specific gene regulatory mechanism \[[@B7]\]. It should be noted that according to our cytological mapping POF binds within this euchromatic region (370000 to 430000). Fragment analysis ----------------- In the whole chromosome analysis we identified sequence signatures that are overrepresented in different chromosomes, but we did not study the linear organisation of the sequence signatures along the chromosomes. Instead, we divided each of the Dm, Dyand Dpchromosomes into 100 kb fragments to check for the presence of sequence signatures that can distinguish fragments of specific chromosomes from those of other chromosomes, especially signatures distributed over the whole chromosome. For such an analysis it is important to remove all variation connected to differences in nucleotide composition. Using a Partial Least Squares (PLS) model with A/T composition of every fragment as a single response we removed this bias. Strikingly, when the approximately 3600 fragments from all three species were analysed using PCA based on di-, penta-and hexa-mers the 33 F-element fragments clustered together (Figure [6](#F6){ref-type="fig"}). The motifs responsible for this separation were not the same as in the whole chromosome analysis. Nevertheless, this demonstrates the existence of sequence signatures that are capable of separating all F-element fragments from the three different species. Based on the relationship of these species we conclude that these signatures have been conserved for at least 54.9 Myr \[[@B15]\]. These conserved motifs are also linearly distributed along the sequenced part of the F-elements (Figure [6](#F6){ref-type="fig"}). The F-elements from the three species have high A/T contents and are probably all enriched in mobile and repeated elements. However, the motifs separating the F-element fragments are not connected to simple sequence repeats since masking such repeats did not alter the results. In addition, the DmF-element fragments clustered together when the original sequence was analysed together with sequences in which both simple and complex repeated elements had been masked (Figure [8](#F8){ref-type="fig"}). Therefore, the collective separation of F-element fragments in the three species cannot be attributed to any known repeated elements, and we speculate that the signatures we identified have a role in F-element identification. The X chromosomal fragments of Dm/Dy, but not Dp, can also be separated to some degree using penta-and hexa-mers. As shown in Figure [7](#F7){ref-type="fig"}, some non F-element fragments are more similar to the F-element fragments. These non F-element fragments are the centromere proximal regions of Dm/Dychromosomes 2 and 3. The heterochromatic nature of the F-element in Dmis well established, e.g. by its enrichment of HP1 and H3K9 methylation \[[@B31]\]. In our analysis, the proximal regions of chromosomes 2 and 3 in Dm/Dyshowed similarity to the F-element. It is interesting that an anti-metH3K9 antibody decorates the proximal regions of chromosomes 2 and 3 as well as the F-element in Dm. The proximal region of X is also stained, but to a much lesser extent using this antibody (JL unpublished results). We note that the same pattern is present in Figure [7](#F7){ref-type="fig"}. We must consider the possibility that chromatin similarities cause the partial overlap of the F-element and the proximal regions of chromosomes 2 and 3 (and that the heterochromatic nature of the F-element caused its observed separation from the other chromosomes). It is difficult to fully separate chromosome-and chromatin-specific effects. Sequences that have high A/T contents and are enriched in repetitive elements tend to be heterochromatic. As shown in Figures [5](#F5){ref-type="fig"} and [8](#F8){ref-type="fig"}, the F-element separation was retained after normalising for differences in A/T content. Furthermore, the results were not significantly different when simple sequence repeats were removed using Tandem Repeats Finder, or when simple sequence repeats and repetitive elements were removed using RepeatMasker. The findings even apply to the exon sequences. Thus, we conclude that our methodology is capable of detecting chromosome-specific sequences. Conclusion ========== We have shown that the F-elements of three species that separated roughly 55 Myr ago share sequences that are distributed over the entire chromosomes. These sequences are not related to their unusually high A/T contents or any known repeated elements. In conclusion, our results support the existence of sequence signatures that confer chromosome specific integrity in Drosophila. Methods ======= Hexamer scoring --------------- We scored all positions of all possible di-(16), tri-(64), tetra-(256), penta-(1024) and hexamers (4096) in the genome sequence of Dm, Dyand Dp. Every motif was counted in each target sequence. Full-length chromosomes and 100 kb fragments were used as targets. Scoring was done by a sliding window approach, sliding one nucleotide at a time. The scoring function gives a two dimensional data-matrix with target sequences as objects (rows) and the total score for each motif as variables (columns). By dividing each element in the matrix by the length of its target sequence a relative score is obtained. Prior to analysis all data were mean-centred, i.e. each value was adjusted by subtracting the average value for the corresponding variable. All scoring and data normalisation procedures were performed using custom software developed in C, Java and Perl. The software can be obtained, on request, from the corresponding author. Multivariate analysis --------------------- ### Principal Component Analysis -- PCA The central idea of PCA is to extract a few, so-called, principal components describing most of the variation present in the data. The principal components are linear combinations of the original variables and uncorrelated to each other. ![](1471-2105-6-158-i1.gif) where tare the scores, pthe loadings, Ais the number of principal components and Eis the residual matrix. The principal components can be determined using the NIPALS algorithm \[[@B38]\] or by Singular Value Decomposition (SVD) \[[@B39]\]. The scores (t) show how the objects and experiments relate to each other. The loadings (p) reveal variables that have an important influence on the patterns seen in the score plot. ### Partial Least Squares -- PLS PLS is a multivariate regression method that relates the data matrix (X, the scoring data) to single (y) or multiple (Y) response(s). PLS has proved to be a powerful tool for finding relationships between descriptor matrices and responses, especially when there are more variables than observations and the variables are co-linear to each other and noisy. In our study, PLS was used to normalise the data by removing the variance in the scoring data that was correlated to the A/T content of the chromosome fragments. The PLS theory and methods discussed here concern single y-responses. As in PCA, principal components are constructed to reduce the dimensions of X. In order to obtain the principal components, PLS maximizes the covariance between the response variable yand a linear combination of the original variables t= Xw, where tis the score vector, Xis the data matrix and wis the weight vector. For a more in-depth description of PLS, see \[[@B40]-[@B42]\] and references therein. ![](1471-2105-6-158-i2.gif) where tis the score vector for X, Ais the number of PLS components, pis the loading vector for X, cis the loading vector for Y, Eis the residual matrix for Xand Fis the residual matrix for Y. All multivariate analyses and visualisations were performed using the Evince software package <http://www.umbio.com>. Data normalisation ------------------ ### Probability normalisation The probability of successfully aligning a motif to a target depends on the base composition of the motif sequence and the target sequence. For example, the chance of finding a given A/T-rich motif is relatively high in an A/T-rich target due to their similarity in base composition. Probability normalisation removes this systematic bias from the data. Each value is normalised by dividing the observed number of hits by the expected number of hits. The initial scoring is performed as described above, except that the scores are not divided by the target sequence length. The number of expected hits was calculated as follows: ![](1471-2105-6-158-i3.gif) where Nis the target sequence length, i= {G,A,T,C}, f(i) = frequency of base iin the target sequence and n~i~= count of base iin the hexamer. ### Fragment normalisation To remove all variance in the scoring matrix obtained from the 100 kb fragment analysis that was solely related to the base composition of the target sequences, a different normalisation was applied, in which we created a PLS model with the base composition of every fragment as a single y-response and the scoring matrix as an x-matrix. By removing the variance explained by the first component a residual matrix was obtained, in which all variation caused by differences in base composition amongst the fragments had been removed. The residual matrix Ewas calculated as follows: E= x - tp\' Where xis the hexamer scoring matrix, t= PLS-scores for the 1^st^component and p\'= PLS-loadings for the 1^st^component. The normalised data were then used for PCA analysis of the fragmented genome. Repeat masking -------------- RepeatMasker \[[@B18]\] was run using default parameters, MaskerAid \[[@B43]\] and the Drosophilalibrary file from Repbase \[[@B22],[@B23]\]. Tandem Repeats Finder \[[@B17]\] was run using default parameters and a maximum period size of 500. Polytene chromosome staining ---------------------------- Polytene chromosomes from 3^rd^instar larvae of wild type Dmwere prepared and stained essentially as previously described \[[@B44]\]. Salivary glands were fixed in 2% formaldehyde in PBS, 0.1% Triton X-100, 0.2% NP-40 for 30 seconds followed by 2 minutes in 50% acetic acid, 1% formaldehyde. Polytene chromosomes were squashed as previously described \[[@B44]\]. The slides were washed for 30 minutes in 1 × PBS, 0.1% Triton X-100, transferred to blocking solution (0.1 M maleic acid, 0.15 M NaCl, 1% Boehringer blocking reagent) and incubated for 30 minutes at room temperature. The slides were then incubated overnight at 4°C with a rabbit polyclonal anti-POF primary antibody \[[@B6]\]. The slides were washed for 2 × 10 minutes in 0.1 M maleic acid, 0.15 M NaCl, 0.3% Tween 20 and blocked for 30 minutes. As a secondary antibody, a donkey anti-rabbit conjugated with Cy3 (Jackson Laboratories) was used, diluted 1:400 and incubated at room temperature for 2 hours. The squashes were counterstained with DAPI (1 μg/ml) and washed for 2 × 10 minutes before mounting with Vectashield (Vector). Chromosomes were analysed using a Zeiss Axiophot microscope equipped with a KAPPA DX20HC CCD camera. Images were assembled, contrasted and merged electronically using Adobe Photoshop. Well spread F-elements were mapped according to Saura et al. \[[@B45]\] and POF-binding regions were defined. To correlate cytological positions to sequences we used the Dmgenome release 3.2. Since all sequences are annotated to cytological bands and POF binds preferentially to interbands we used regions with no POF binding for comparison. This is the reason why regions lacking POF binding are used for the correlation study. Authors\' contributions ======================= All authors were involved in the initial project discussion. PS and FP did the detailed planning, carried out all computational work, performed the analysis and wrote the draft manuscript. JL carried out the chromosome staining and participated in the final analysis. AOS carried out the mapping. All authors contributed to, read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 The loadings of the PCA in Figure [6](#F6){ref-type="fig"}. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We thank Anssi Saura, Petter Lindgren and four anonymous reviewers for valuable comments. This work was supported by grants from the Swedish Research Council, the Åke Wiberg Foundation (J.L) and the Nilsson-Ehle Foundation (P.S).	PubMed Central	2024-06-05T03:55:59.779626	2005-6-23	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181806/", "journal": "BMC Bioinformatics. 2005 Jun 23; 6:158", "authors": [ { "first": "Per", "last": "Stenberg" }, { "first": "Fredrik", "last": "Pettersson" }, { "first": "Anja O", "last": "Saura" }, { "first": "Anders", "last": "Berglund" }, { "first": "Jan", "last": "Larsson" } ] }
PMC1181807	Background ========== The production of recombinant proteins in eukaryotic cells requires the fusion of the target genes to promoter sequences functional in the eukaryotic cells and exhibiting the desired expression pattern. The molecular cloning procedures necessary for the construction of these fusions are usually performed in bacteria, mostly in Escherichia coliK12. A heterologous expression of powerful toxins like for example botulinum toxin, tetanus toxin or diphtheria toxin may pose a risk for the persons carrying out the experiments. A careful risk analysis considering an exposition to the toxin produced in the eukaryotic cells as well as a potential danger originating from toxin production by the recombinant E. colihas therefore to be carried out. Since the nucleic acid sequences characterising eukaryotic-type and bacterial promoters are different, it is usually regarded improbable that promoter sequences functional in eukaryotic cells are able to direct a considerable transcription initiation in bacteria \[[@B1],[@B2]\]. There are numerous differences between the transcription machinery of eukaryotes and eubacteria represented by different structures of promoters and the complexity of RNA polymerases and transcription factors \[[@B3]-[@B6]\]. On the other hand, the RNA polymerases are evolutionary conserved and belong to a protein family termed \"multisubunit RNAP family\" \[[@B4],[@B7],[@B8]\]. The subunits of the eubacterial RNA polymerase all have counterparts in the subunits of the RNA polymerase from Archaea and the three RNA polymerases from eukaryotes \[[@B4],[@B8]-[@B11]\]. There is also an obvious similarity between the TATA box from eukaryotes and Archaea to the -10 consensus sequence from eubacterial promoters. This discrepancy reflected by differences versus similarities of the transcription machinery of organisms belonging to different kingdoms motivated us to analyse the capacity of promoter sequences to direct gene expression in foreign hosts. In previous studies we have shown that a high percentage of eukaryotic-type promoters specific for plants can direct gene expression in eubacteria. By testing promoter activity of ten plant-specific promoters in five eubacterial species we could show that in 50 % of the tested combinations the promoter sequences could be used in the bacterial host \[[@B12],[@B13]\]. In a subsequent study we demonstrated that not only eukaryotic promoter sequences but any type of eukaryotic DNA has a high probability to initiate transcription after transfer into bacteria. This was shown by transfer of random DNA fragments from the yeast Saccharomyces cerevisiaeinto E. coliK12 and detection of initiation of significant gene expression in about 80 % of the transformed E. coli\[[@B14]\]. In this study we addressed the question if heterologous gene expression in E. coliby viral promoters has to be considered when specifying safety measures for correspondent experiments. We selected five frequently used viral promoter sequences: (i) the polyhedrin promoter from baculovirus (P PH), (ii) the enhancer and immediate early promoter from CMV (P CMV), (iii) the early promoter from SV40 (P SV40), (iv) the thymidine kinase promoter from HSV1 (P TK), and (v) the 5\'LTR promoter from HIV (P LTR). The heterologous gene expression of these promoters in E. coliK12 was tested using the luxABgenes from Vibrio harveyias reporter system. Two of the promoters (P PH and P CMV) were analysed in more detail to determine the transcription start sites in E. coli. By additionally fusing these two promoters to a gene encoding a cytotoxin (hemolysin gene vllyfrom Vibrio vulnificus) the relevance of the heterologous gene expression by viral promoters for safety considerations was further supported. Results and discussion ====================== Lux gene expression in E. colidirected by viral promoters ----------------------------------------------------------- To investigate whether viral promoters can initiate gene expression in bacteria, we constructed fusions between various viral promoters and the promoter-less luxABgenes from V. harveyi, transformed the fusions into E. coliK12 and measured the luminescence of the transformants. The viral promoters analysed were the polyhedrin promoter from baculovirus (P PH), the enhancer and immediate early promoter from CMV (P CMV), the early promoter from SV40 (P SV40), the thymidine kinase promoter from HSV1 (P TK), and the 5\'LTR promoter from HIV (P LTR). As positive controls gene expression levels caused by homologous bacterial promoters in E. coliwere determined by fusing the promoter from the nptIIIgene (P NPTIII) and the promoter from the TEM-1 β-lactamase gene (P BLA) to the luxABgenes. The vector pKKlux without any promoter sequence was included as negative control. The results of the luminescence measurements are shown in Fig. [1A](#F1){ref-type="fig"}. The vector pKKlux generated a background luminescence (212 RLU/10 s), which was identical to the luminescence caused by the P TK construct (210 RLU/10 S). The P PH (824 RLU/10 s) initiated a weak gene expression in E. coli, while the P LTR (2382 RLU/10 s) caused a luminescence 10 times above background. The P CMV (11981 RLU/10 s) and the P SV40 (12301 RLU/10 s) initiated a strong gene expression in E. coli, which was comparable to the gene expression of the bacterial P BLA (16350 RLU/10s). The bacterial P NPTIII caused the highest luminescence values with 36011 RLU/10 s. In summary, four of the five viral promoters tested generated a luminescence at least four times above background (P PH, P CMV, P SV40, P LTR). The luminescence caused by the P CMV and P SV40 was even similar to the luminescence caused by the bacterial P BLA supporting the relevance of the heterologous gene expression with respect to the phenotypic expression of transferred traits. Confirmation that our reporter assay reliably reflects the strength of the tested promoters was achieved by performing quantitative RT-PCR assays using primers and probes specific for the luxBgene (Fig. [1B](#F1){ref-type="fig"}). Mapping of DNA sequences in viral promoters serving as transcription start sites in E. coli --------------------------------------------------------------------------------------------- We chose two different approaches to determine the location of DNA sequences in the viral promoters which may be used as transcription initiation sites by the RNA polymerase from E. coli. The first approach consisted in a statistical analysis of the nucleotide sequences of the promoters with a neural network promoter prediction program \[[@B15]\]. The number of predicted prokaryotic promoters in the five sequences analysed varied from zero to six. Surprisingly, the promoter prediction program identified no potential promoter sequence in the P SV40, which was the promoter causing the strongest luminescence in our reporter assays (Fig. [1A](#F1){ref-type="fig"}). The construct with the P CMV, which was almost as strong as the construct with the P SV40 in the luminescence assay, was shown by statistical analysis to have six potential start sites. In each of the remaining three promoter sequences (P PH, P TK, P LTR), the promoter prediction program predicted one potential start site. Interestingly, a potential promoter was also identified in the P TK sequence, although the P TK luxABfusion pointed out to be negative in the luminescence assay. In summary, it can be concluded that a prediction of the gene expression activity of a DNA sequence solely on the basis of a statistical analysis of the sequence is not reliable, and that only an experimental approach using a reporter system can provide this information. This result is in good agreement with the outcome of our previous studies \[[@B12]-[@B14]\] on the functionality of plant promoters in bacteria and on the promoter activity of yeast DNA in E. coli. There we also demonstrated that it was not possible to reliably predict the occurrence of a bacterial promoter in a DNA sequence by statistical analysis. Our second approach consisted in the experimental determination of transcription start sites by the 5\' RACE-method and sequencing of the 5\' RACE products. For this purpose, we selected two viral promoters which had both been shown to direct gene expression according to the results of the luminescence measurements. The construct with the P CMV belonged to the viral promoters initiating a very strong heterologous gene expression, while the P PH caused a weak heterologous gene expression in E. coli. Constructs containing the bacterial promotersP BLA and P NPIII were included into the experiments as positive controls. The results summarised in Fig. [2](#F2){ref-type="fig"} show the transcription start sites and the location of the putative -10 and -35-regions. The transcription start sites determined in the P BLA and P NPTIII promoter regions by the 5\' RACE method were identical with the start sites described by other authors \[[@B16],[@B17]\] confirming our experimental approach. We could identify specific start sites in the two tested constructs containing the viral promoters P PH and P CMV confirming that the luminescence resulted from specific transcription initiation events. One start site was identified in the P PH, whereas two start sites were found in the P CMV (Fig. [2](#F2){ref-type="fig"}). In the upstream region of all transcription start sites we found at a distance of 5 to 7 bp a region with similarity to the -10 (TATAAT) consensus sequence which was separated by 14 to 18 bp from a region with similarity to the -35 (TTGACA) consensus sequence. It was very astonishing, that none of the transcription start sites within the viral promoters determined by the 5\' RACE method corresponded to the transcription initiation sites predicted by the promoter prediction programme (data not shown). The difficulties encountered in the efforts to predict promoter activities in bacteria by statistical analysis can at least in part be explained by the importance of the nucleotide sequences surrounding the -10 and -35 regions which determine the physical-chemical and structural characteristics of the DNA and the promoter strength \[[@B18],[@B19]\]. Jacquet and Reiss \[[@B18]\], for example, analysed the influence of the context of the -10 and -35 regions on transcription efficiencies and found that transcription efficiencies varied by a factor of ten depending on the sequences surrounding the consensus sequences. Progress in bacterial promoter prediction has been made using neural network programs like the one we have employed for our statistical analysis. Demeler and Zhou \[[@B20]\] reported a prediction accuracy of 98.4 % using a neural network for the prediction of E. colipromoters. Further problems of promoter predictions consist in the existence of different sigma factors binding to different recognition sequences and in the variation between the consensus sequences of different bacterial species. Expression of the hemolysin gene vllyfrom V. vulnificusin E. colidirected by viral promoters -------------------------------------------------------------------------------------------------- As heterologous gene expression was observed using luciferase as reporter system, we were interested, if the promoter activity of the viral promoter sequences in E. coliwas sufficiently strong to significantly express virulence factors. This would be of importance for risk assessments of corresponding genetically modified organisms. For instance it has been shown that the transfer of the inv locus necessary for the invasion of Yersinia pseudotuberculosisinto host cells enabled a non-invasive strain of E. colito penetrate cultured cells \[[@B21]\]. Similar observations have been made with the invasion locus of Mycobacterium avium. After transfer of this locus into non-invasive E. coliand Mycobacterium smegmatisthe recipients could invade epithelial cells \[[@B22]\]. A worst case scenario would be the heterologous expression of a toxin able to operate without being dependent on other virulence factors. Many hemolysin genes fulfil this condition. The phenotypic effect of hemolysins can be relatively easily monitored and hemolysin genes are prevalent in many pathogenic bacteria. We selected the vllygene from V. vulnificusas reporter gene \[[@B23]\]. An advantage of vllyfor our purposes is its small size (1071 bp) facilitating the amplification and cloning procedures. Furthermore, a 1.3 kb fragment of V. vulnificusDNA carrying vllyhas been shown to confer a hemolytic phenotype onto E. coli\[[@B23]\]. We inserted the promoter-less vlly(including the Shine-Dalgarno-Sequence) downstream from the viral promoters P PH and P CMV present in the pKKlux constructs and transformed the promoter-vllyfusions into E. coliK12. As negative control, vllywas inserted into pKKlux without any promoter sequences. We first monitored the vllyexpression by plating the transformants onto blood agar plates and observing hemolysis zones around the colonies. E. colicontaining P PH as well as E. colicontaining P CMV were hemolytic (Fig. [3B,C](#F3){ref-type="fig"}). However, a quantification to what extent a lysis of erythrocytes had occurred, was not possible with this method. We therefore also measured release of haemoglobin from lysed erythrocytes in a liquid hemolysis assay. Since Vlly had been shown before to be located in the periplasma and cytoplasma of recombinant E. coli\[[@B23]\], our assay involved sonication of the E. colicells to liberate periplasmic and cytoplasmic Vlly. The outcome of a typical hemolysis assay is shown in Fig. [3D](#F3){ref-type="fig"}. Both viral promoters initiated an expression of vlly. In accordance with the luminescence assays, the P CMV directed a very strong vllyexpression resulting in lysis of almost half the erythrocytes present in the assay, while the P PH was much weaker causing a lysis of about 20% of erythrocytes. Conclusion ========== We showed that four (polyhedrin gene promoter from baculovirus, immediate early promoter from CMV, early promoter from SV40 and LTR promoter from HIV 1) out of five tested viral promoters carry structural features required by the bacterial RNA polymerase to initiate transcription. Two promoters (P CMV and P SV40) were as strong as the promoter from the bacterial TEM1-β-lactamase gene. The determination of the transcription start sites in selected viral promoters confirmed the presence of sequences with homology to the bacterial -10 and -35 promoter consensus sequences. Two promoters (P CMV and P PH) were shown to be able to direct expression of a bacterial cytotoxin, which illustrated the relevance of this type of heterologous gene expression for the specification of safety measures for the handling of genetically modified organisms. A strong expression of a foreign gene by a heterologous promoter must either be taken into consideration when defining safety measures or if necessary, it must be excluded by performing appropriate experiments. If there is the need for exclusion of a heterologous gene expression, care should be taken in choosing convenient promoter sequences. Alternatively, site-directed mutagenesis can be employed to design promoter sequences according to the experimental requirements. By exchanging nucleotides, which are required for binding of the bacterial RNA-polymerase, but which are of no or little importance for the transcription initiation in the final eukaryotic recipient, it is possible to minimise the heterologous gene expression while maintaining the desired characteristics of the promoter with regard to its function in eukaryotic cells or tissues \[[@B13]\]. Finally, the generation of a functional gene product not only depends on the presence of functional promoter sequences but can be also influenced by factors like the presence or absence of introns, the codon usage, post-translational modifications or the necessity of protein secretion. Methods ======= Bacteria and growth conditions ------------------------------ Escherichia coliK12 strain DH5α \[[@B24]\] and Vibrio vulnificusstrain CH1603 (O:8, isolated from the Baltic Sea, Germany) were grown at 37°C overnight in LB medium \[[@B25]\]. Measurement of luminescence --------------------------- The measurement of luciferase activity was performed as described before \[[@B12]\]. Bacterial cultures were grown at 28°C up to an optical density (λ = 600 nm) of 1.0 to 1.3 and diluted to contain 10^6^cells per ml. After transfer of 100 μl (microliter) of the diluted cultures into microtiter plates, 50 μl of 2 % decanal in 50 mM sodium phosphate buffer, pH 7.0, were added and the luminescence (RLU: relative light units) was measured in triplicate at 28°C for 10 s in the Microlumat LB96P from EG&G Berthold (Bad Wildbad, Germany). Detection of hemolysis ---------------------- To detect hemolysis on blood agar plates, a broth culture of the bacteria was either streaked or plated onto enterohemolysin agar plates (Oxoid, Wesel, Germany) and incubated at 37°C. The plates were evaluated visually after 40 hours. A quantification of the hemolytic activity of the bacteria was achieved using defibrinated sheep blood (Oxoid, Wesel, Germany) that was washed three times with PBS buffer (137 mM NaCl, 2.7 mM KCl, 10 mM Na~2~HPO~4~, 2 mM KH~2~PO~4~, pH 7.4). The bacteria were incubated at 37°C up to an optical densitiy (λ = 588 nm) of 1.3. To liberate intracellular hemolysin, the cells were disrupted using ultrasound for 1 min at 50 watt in the Branson Sonifier 450 (Branson Ultrasonics Corporation, Danbury, CT, USA). The disrupted cell lysate was at once placed on ice and centrifuged for 10 min at 9300 g. 970 μl of the supernatant were spiked with 20 μl washed blood and 10 μl 1 M CaCl~2~and incubated at 37°C for 45 min. Every 5 min the suspension was carefully mixed. The mixture was then centrifuged at 1500 g for 10 min and the haemoglobin in the supernatant was quantified by measuring the OD at 540 nm. The value for complete lysis of erythrocytes (OD\>2) was obtained by using distilled water instead of culture supernatant. All cultures were measured in duplicate. Molecular biology techniques ---------------------------- Common molecular biology techniques (DNA isolation, restriction digestion, ligation, electrophoresis) were carried out according to standard protocols \[[@B25]\] or according to the recommendations of the manufacturers of kits and enzymes. Sequencing reactions were performed by using the Prism Big Dye™ FS Terminator Cycle Sequencing Ready Reaction Kit from PE Applied Biosystems, Weitersheim, Germany. Transformation of E. coliwas performed according to the method of Hanahan \[[@B24]\]. Construction of recombinant plasmids ------------------------------------ Viral promoters as well as bacterial promoters were inserted into the vector pKKlux \[[@B12]\]. pKKlux (7.7 kb) is a derivative of the promoter probe vector pKK232-8 \[[@B26]\], which carries a promoter-less cat(chloramphenicol-acetyltransferase) gene in front of the multiple cloning site. Read-through into the catgene is prevented by the transcription terminator of the rrnb gene from E. coli.pKK232-8 contains a beta-lactamase gene to allow selection by adding ampicillin (100 μg/ml) into the medium. pKKlux was generated by inserting in front of the catgene the promoterless luciferase genes luxABfrom Vibrio harveyipresent in plasmid pUT/mini-Tn5 luxAB \[[@B27]\], allowing the determination of promoter activities by measuring luminescence. The promoter fragments were amplified by PCR using primers provided with restriction sites for the restriction enzymes SmaI and XbaI allowing ligation into the SmaI / XbaI sites of pKKlux. This cloning strategy guaranteed correct orientation of the promoters upstream from the promoter-less luxABgenes. Only the PTK, which contains a SmaI site, was not digested with SmaI prior to ligation of the PCR fragment into pKKlux. Table [1](#T1){ref-type="table"} describes the generated promoter fragments and the templates and primers used for their amplification. For cloning of the hemolysin gene vllyfrom V. vulnificus\[[@B23],[@B28]\], a 1118 bp fragment carrying a promoterless vllywas amplified using the primers vlly-S (GGTCTAGAGCAGTCTAAAAGGAGAAAGTCATGGTGG) and vlly-AS (GGTCTAGACCCGATGAGGAAAGGTGATCC). The primers had been provided with XbaI restriction sites allowing insertion of vllyin the XbaI site of the pKK232-8-derivatives containing the promoters described in Table [1](#T1){ref-type="table"}. Correct orientation of vllydownstream of the analysed promoters was confirmed by sequencing. Quantification of lux mRNA -------------------------- Bacterial cultures were grown over night in LB medium at 28°C. 200 μl of these cultures were inoculated into 5 ml of LB medium and grown for 4 to 6 hours to an optical density (λ = 600 nm) of 1.0. Aliquots of these cultures were taken for RNA isolation. RNA was isolated with the SV Total RNA Isolation System Kit from Promega, Madison, USA, which includes a DNase digestion step. The RT-PCR was performed with the Titan™ One Tube RT-PCR System from Boehringer Mannheim (Indianapolis, Ind., USA) in the ABI Prism 7700 Sequence Detection system (Applied BioSystems Division, Perkin Elmer, Foster City, Calif., USA). The primer pair used for the quantitative RT-PCR \[Lux2270F (CCGTTAACCCACACGCGT) and Lux2329R (TGCTCGTCGCATTCACAAA)\] amplified a 60 bp fragment from the luxBgene. The dually labelled detector probe hybridising within the sequence between these primers had the sequence (FAM)-CACTGAAGGCGGTCCTGCGCA-(TAMRA). The RT-PCR was performed with 0.1 μg RNA in 50 μl reaction mix according to the recommendations of the manufacturer (with RT-reaction buffer, 1.5 mM MgCl~2~, 0.2 mM of each dNTP, 1 μM downstream primer, 1 μM upstream primer, 5 mM DTT, 5 U RNase inhibitor, 1 μl of enzyme mix containing AMV, Taq DNA polymerase and Pwo DNA polymerase). The quantitative TaqMan RT-PCR furthermore required addition of 180 nM dually labelled probe and 1 μM ROX (6-Carboxy-X-rhodamin) as passive reference dye. The reverse transcription was carried out at 50°C for 30 min and terminated by heating at 95°C for 10 min. Amplification was carried out by running 35 cycles with 30 sec at 95°C and 1 min at 60°C. Samples were measured in duplicate. The amount of the lux mRNA was determined by the help of a standard established with known amounts (0.073 pg to 730 pg) of the plasmid pKKlux. Mapping of transcription start sites with the 5\' RACE method ------------------------------------------------------------- Total RNA from broth cultures was isolated with the SV Total RNA Isolation System Kit from Promega, Madison, USA. The 5\' ends of the lux transcripts were mapped with the 5\' RACE (Rapid Amplification of cDNA Ends) System Kit from Life Technologies, Inc., Rockville, USA. 1 to 5 μg of total RNA was used to synthesise cDNA from the 5\' end of the lux mRNA with the gene-specific primer GSP1/luxA (CAACATAAGGATTCCC). The reaction was performed with the SuperScript II RT at 42°C for 50 min. A homopolymeric C-tail was added to the cDNA with the terminal deoxynucleotidyl transferase. The RACE products were synthesised using the abridged anchor primer GGCCACGCGTCGACTATACGGGIIGGGIIGGGIIG and the gene-specific primer GSP2/luxA (GCGTACTAGTCAGTGAAGTGGTGCTCTAGCAACC). The PCR program was run with 35 cycles composed of a denaturation step of 1 min at 94°C, an annealing step of 30 sec at 65°C and an elongation step of 2 min at 68°C. The RACE products were directly sequenced to identify transcription start sites. Authors\' contributions ======================= AL: design of project, preparation of the manuscript, statistical analysis of nucleotide sequences for promoter search. MM: construction of recombinant plasmids, luminescence measurements, measurement of hemolysis, real-time PCR, determination of transcription start sites. JC: construction of recombinant plasmids, luminescence measurements, real-time PCR, determination of transcription start sites. DJ: preparation of and instructions to 5\'RACE. BA: conception of project and follow-up discussions of results. All authors read and approved the final manuscript. Acknowledgements ================ We thank Barbara Freytag and Beate Meister for excellent technical assistance and are grateful to Dr. Sybille Somogyi for providing the 5\' LTR region. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### *Gene expression caused by viral promoters in E. coli**. Luminescence (Panel A) and amount of lux AB mRNA transcripts determined by Real-time RT-PCR (Panel B) of cultures of E. colicontaining fusions between the luxABgenes in the vector pKKlux and the viral promoters P TK (thymidine kinase promoter from HSV1), P PH (polyhedrin promoter from baculovirus), P CMV (immediate early promoter from CMV), P SV40 (early promoter from SV40) and P LTR (5\' LTR promoter from HIV1) or the bacterial promoters P BLA (promoter from the TEM-1 β-lactamase gene) and P NPTIII (promoter from the neomycinphosphotransferase III gene). E. colicontaining pKKlux was included in both experiments as negative control. The luminescence (Panel A) was measured in triplicate and the columns represent the average of the three measurements with the standard deviation. The lux AB mRNA (Panel B) was measured in duplicate and the columns represent the average of the two measurements. ::: ![](1472-6750-5-19-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Identification of transcription start sites. Transcription start sites were identified by the 5\' RACE method in E. colicontaining pKKlux derivatives with fusions between the promoterless luxAB genes and the promoter from the nptIIIgene (P NPTIII), the promoter from the TEM-1 gene (P BLA), the polyhedrin promoter (P PH) and the immediate early CMV promoter (P CMV). The sequences shown cover the transcription initiation sites identified in E. coli(indicated by arrows) and their upstream regions. The putative -10 and -35 regions are underlined. Two transcription start sites \[a), b)\] were identified in the P CMV. ::: ![](1472-6750-5-19-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Expression of hemolysin genes in E. colidirected by viral promoters. Hemolysis of sheep blood erythrocytes by E. coliDH5α containing the vllycoding sequence without promoter (panels A and D), with the promoter from the polyhedrin gene (P PH: panels B and D) or with the immediate early promoter from CMV (P CMV: panels C and D). Panels A to C show the hemolysis in blood agar plates visible as cleared zones around the colonies. Panel D indicates the percentage of erythrocytes lysed in a liquid blood assay. The columns show the average of two measurements. ::: ![](1472-6750-5-19-3) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Promoters analysed in this study. The promoters (column 1 and 4) were amplified by PCR using the template DNA (column 2) and the primers (column 3) listed in the table. ::: ---------------------------------------------------------------------------------------------------------------------------------------------------------------- Promoter fragment Template DNA^a^ Primer sequence^b^ Size of promoter-fragment ----------------------------------------------------------- --------------------- -------------------------------------------------- --------------------------- P PH: Polyhedrin promoter from baculovirus pBacPAK8 \[29\] PH-1: GCCCGGGCCATCTCGCAAATAAATAAG\ 78 bp PH-2: GTCTAGACAGGGATCCGTATTTATAGG P CMV: Enhancer and immediate early promoter from CMV pRL-CMV \[30\] CMV-1: GCCCGGGGATCTTCAATATTGGCCATTAGCC\ 802 bp CMV-2: GTCTAGACACTGACTGCGTTAGCAATTTAAC P SV40: Early promoter from SV40 pGL3 \[31\] SV40-1: GCCCGGGCTAGCCCGGGCTCGAGATCTG\ 223 bp SV40-2: GTCTAGATTTGCAAAAGCCTAGGCCTCC P TKThymidine kinase promoter from HSV1 pRL-TK \[32\] TK-1: GCCCGGGATCTAAATGAGTCTTCGGACCTCGC\ 788 bp TK-2: TCTAGACGAGACTGTTGTGTCAGAAGAATCAAGC P LTRHIV-1 subtype D 5\'LTR promoter isolate 92UG021 LTR-1: GGCCCGGGATTAGATATCCACTGACCTTTGGATGGTGC\ 412 bp LTR-2: GGTCTAGATTAAGCAGTGGGTTCCCTAGCTAGCC P NPTIII: nptIIIpromoter pBin19 \[33, 34\] NPTIII-S: GCGCCCGGGCATAATTGTGGTTTCAAAATCGGC\ 193 bp NPTIII-AS: GGGTCTAGATTATTATTTCCTTCCTCTTTTC P BLA: TEM-1 promoter pKK232-8 \[26, 35\] Amp-S: GCGCCCGGGCGTCAGGTGGCACTTTTCG\ 134 bp Amp-AS: GGGTCTAG*AACTCTTCCTTTTTCAATATTATTG ---------------------------------------------------------------------------------------------------------------------------------------------------------------- ^a^pBacPAK8 was from Clontech (Palo Alto, CA, USA), pRL-CMV, pGL3 and pRL-TK were from Promega (Madison, WI, USA). The 5\'LTR promoter sequence was kindly provided by S. Somogyi. ^b^Restriction sites added to the primers for cloning purposes are italicised. :::	PubMed Central	2024-06-05T03:55:59.784372	2005-6-20	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181807/", "journal": "BMC Biotechnol. 2005 Jun 20; 5:19", "authors": [ { "first": "Astrid", "last": "Lewin" }, { "first": "Martin", "last": "Mayer" }, { "first": "Janet", "last": "Chusainow" }, { "first": "Daniela", "last": "Jacob" }, { "first": "Bernd", "last": "Appel" } ] }
PMC1181808	Background ========== Success in the management of soft tissue sarcomas (STS) is often limited by the extension of lesions to neurovascular structures, because of the difficulty in dissecting the neurovascular bundle from the tumor without compromising the function and local recurrence of residual lesions. Patients with STS involving or extending to neurovascular structures may be sometimes advised to undergo an amputation. In an effort to preserve limbs, conservative surgery with various modalities such as nerve resection with or without nerve grafts, irradiation, and hyperthermia for preservation of peripheral nerves has been reported \[[@B1]-[@B4]\]. Brachytherapy, interstitial tumor bed irradiation, following conservative surgery has been shown to provide excellent local control and limb preservation in patients with STS, whereas little is known about the tolerance of peripheral nerves to brachytherapy. The placement of afterloading catheters directly upon neurovascular structures to irradiate such lesions appropriately subjects them to high dose radiation. The American Brachytherapy Society suggested that brachytherapy, especially high-dose-rate (HDR) brachytherapy, should be used with caution in such situations \[[@B5]\]. We treated 7 patients with conservative surgical resection and placement of afterloading catheters directly upon these critical structures followed by fractionated HDR brachytherapy. The purpose of this study was to evaluate the efficacy and radiation neurotoxicity of HDR brachytherapy in patients with STS in contact with neurovascular bundles. Methods ======= Patients -------- Between 1995 and 2000, thirty patients with STS were treated in our institute with limb salvage surgery followed by fractionated HDR brachytherapy, after obtaining informed consent from the patients or their guardians. Among them, 7 patients with STS involving the neurovascular bundle were enrolled in this study. The median age at the time of the surgery was 53 years (range, 15--84). There were 4 males and 3 females. All of the 7 patients presented with primary tumors without distant metastasis. Three patients had malignant fibrous histiocytoma, two patients had myxoid liposarcoma, one patient had synovial sarcoma, and one had extraskeletal chondrosarcoma. The median tumor size, defined as the maximum diameter of the tumor at pathological analysis, was 9.4 cm (range, 5.5--13.0 cm). Three patients received perioperative systemic therapy, which mainly consisted of ifosfamide and doxorubicin. The clinical and pathological characteristics of the patients are listed in Table [1](#T1){ref-type="table"}. Surgery and HDR brachytherapy ----------------------------- All of the STS involving or adjacent to neurovascular structures were able to be marginally resected with careful dissection of these critical structures. The pathological examination of surgical specimens demonstrated that there were no gross residual lesions and one microscopic positive margin, which was defined as tumor cells present at the resection margins. A microscopically negative margin was defined as no tumor cells at the margins. In all 6 lesions with negative margins, tumor cells were seen within 1 mm from the margins at the neurovascular bundle (Table [2](#T2){ref-type="table"}). The brachytherapy technique has been previously described \[[@B5],[@B6]\]. In brief, the technique of brachytherapy used afterloading catheters placed intraoperatively in the tumor bed as a single plane implant. The afterloading catheters with a 1 cm spacing in between were placed within the surgical bed directly upon the preserved neurovascular bundles. The catheters were located either parallel or vertical to the limb axes. The afterloading catheters were fixed in position in the target region using absorbable sutures and secured to the skin at the catheter exit site with buttons. Postoperative localization films, along with opaque dummy sources inserted into each individual catheter, were taken to confirm the position of the catheters and for contouring and HDR brachytherapy planning. The three-dimensional treatment planning was performed by the Nucletron PLATO planning station (Nucletron Corp., Columbia, MD) according to the International Commission on Radiation Units and Measurements (ICRU) Repost 58 \[[@B7]\]. The clinical target volume (CTV) was defined by expanding the resected tumor bed by 0.5 cm margins. The planning target volume (PTV) was defined by expanding the CTV by 1.0 cm margins on the catheter plane. The PTV ranged from 52 to 134 cm^3^(mean, 76). Multiple dose prescription points (5 mm from the source axis) were defined in accordance with the each dwell position. The prescribed dose at the periphery of the PTV was 5 Gy per fraction. The PTV was expected to cover the whole dissected portion of the nerve because the width of the nerve was less than 5 mm in all patients. Seven to ten days after the implantation, the catheters were loaded with nominal 10 Curie Iridium-192 using Microselectron-HDR (Nucletron Corp., Columbia, MD). The total dose was 50 Gy, administered as 5 Gy per fraction. Treatments were given twice a day over 5 days with a minimum of 6 hours between fractions. However, Patient 5 with low-grade extraskeletal chondrosarcoma received 30 Gy of HDR brachytherapy as a boost combined with 20 Gy of adjuvant external beam radiotherapy (EBRT) (Table [2](#T2){ref-type="table"}). Electrophysiological study -------------------------- Motor nerve conduction velocity (MCV) of peripheral nerves was measured by standardized techniques using Nicolet VikingQuest (Nicolet Biomedical Inc., Madison, WI) at the latest follow-up. In brief, for MCV of the median nerve, supramaximal stimuli were applied to the elbow, and the compound muscle action potential was recorded from surface electrodes placed over the abductor pollicis brevis muscle. To evaluate the MCV of the sciatic nerve, F-wave conduction velocity (FWCV) of the tibial nerve was used, according to Kimura et al \[[@B8]\]. F responses were recorded with surface electrodes over the abductor hallucis, following supramaximal stimulation at the ankle. A total of 16 F-waves were recorded and minimal F-wave latency was determined. Follow-up and statistical analysis ---------------------------------- The time of follow-up was calculated from the date of the operation. All survival analyses were evaluated with the Kaplan-Meier product-limit method. The median follow-up period was 4 years (range, 13--77 months). Results ======= Oncological outcomes -------------------- There was one local failure and one distant failure during patients\' clinical courses (Table [2](#T2){ref-type="table"}). Patient 2 developed a local recurrence outside the PTV of brachytherapy 25 months postoperatively. This failure was successfully salvaged with complete resection of the recurrent lesion and 50 Gy of adjuvant EBRT, but the patient experienced a femoral shaft fracture 32 months later. Patient 6 died of progressive pulmonary metastases but with local control maintained. Patient 7 died of heart problems. The other four patients survived and continued to be disease-free. The 5-year actuarial overall survival, disease-free survival, and local control rates were 83.3, 68.6, and 83.3%, respectively. Complications ------------- Two patients had nerve-associated complications at the time of follow-up (Table [2](#T2){ref-type="table"}). Patient 1 developed an immediate postoperative palsy of the anterior interosseous nerve prior to brachytherapy. Patient 5 had experienced sensory loss of the tibial nerve since before treatment, possibly caused by the lesion compressing the tibial nerve. In both cases, these complications have been improving. Therefore, there was no practical evidence of HDR brachytherapy-induced neurotoxicity. Complications other than nerve-associated problems included the above-mentioned femoral shaft fracture. To further investigate the subclinical nerve damage by HDR brachytherapy, MCV studies were carried out. Of five survivors, two patients who had radiation-unrelated neuropathy were excluded. The MCV value of the median nerve and FWCV values of the tibial nerves (for MCV of the sciatic nerves) were in the normal range \[[@B8],[@B9]\], consistent with our clinical findings (Table [3](#T3){ref-type="table"}). Discussion ========== Brachytherapy has many theoretical and practical advantages, compared to EBRT. The shorter overall treatment time offers the patient conveniences and reduces the financial cost of treatment \[[@B10]\], in contrast to the standard 7--8 week course of EBRT. The rapid dose fall-off of brachytherapy spares more surrounding normal tissues \[[@B5]\]. Several large clinical studies have demonstrated the efficacy of conventional low-dose-rate (LDR) brachytherapy as an adjuvant therapy for STS. LDR brachytherapy provided adequate local control and acceptable morbidity compared favorably with those of EBRT \[[@B11],[@B12]\]. According to prospective randomized trials, adjuvant brachytherapy improves local control after complete resection of soft tissue sarcomas. This improvement is limited to patients with high-grade histopathology \[[@B6],[@B13]\]. Following these reports, we have been treating only high-grade soft tissue sarcoma with adjuvant brachytherapy after 1996. In this study, Patient 5 with a low-grade lesion received adjuvant brachytherapy in 1995. Regarding the direct effect of LDR brachytherapy on peripheral nerves, it has been reported that none of the 38 patients with STS involving the neurovascular bundle developed radiation neuropathy after receiving conservative tumor resection and cumulative doses less than 9,000 cGy of LDR brachytherapy combined with or without EBRT \[[@B14]\]. A histological and electrophysiological study using rabbits observed that irradiation by iridium-192 LDR brachytherapy to doses up to 13,000 cGy on the carotid-sheath contents including the vagus nerve was well tolerated \[[@B15]\]. Compared to LDR brachytherapy the use of HDR is an attractive alternative, because this technique allows treatment to be given in minutes instead of days, eliminating the radiation hazards and prolonged hospital stays associated with LDR. HDR brachytherapy is expected to replace traditional LDR brachytherapy, although there is limited experience in the use of HDR, both in terms of the duration and the number of cases, compared to LDR \[[@B16],[@B17]\]. Further clinical data are needed to determine the specific role of HDR in the management of STS. There are no clinical reports, which properly evaluate HDR brachytherapy-inducing neuropathy, and no experimental studies on nerve tolerance to HDR brachytherapy. In general, HDR brachytherapy is believed to carry a large risk of nerve damage. It has been proposed that a layer of gel-foam or muscle should be interposed between the catheters and the neurovascular bundles \[[@B18],[@B19]\]. Nevertheless, the results of our study showed that there was no practical and electrophysiological finding of neurotoxicity of HDR brachytherapy. Based on their clinical study \[[@B14]\], Zelefsky et al. speculated that the threshold tolerance of peripheral nerves to LDR brachytherapy might be higher than to EBRT. Similarly, HDR brachytherapy may have a higher threshold tolerance than EBRT, attributable to different radioactive sources. Despite the small number of patients, our findings suggest that HDR brachytherapy may not adversely affect peripheral nerve function in the treatment of STS with neurovascular involvement. Latency also needs to be considered in evaluation of radiation neuropathy. A long-term follow-up study of EBRT reported that the incidence of complications involving nerves increased with time after radiation and not all of the cases were detected at 5, or even 10 years after the treatment \[[@B20]\]. More clinical data with a large number of patients treated with HDR brachytherapy and longer follow-up periods are required to detect further long-term morbidity. Conclusion ========== In this study, there was no practical and electrophysiological finding of neurotoxicity of HDR brachytherapy. Despite the small number of patients, our encouraging results are valuable for limb-preserving surgery of unmanageable STS involving critical neurovascular structures. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= TK conceived of the study, collected and analysed the data, and drafted the manuscript. TS, SS, TM, KH, and MO participated in the study design and manuscript preparation. HK carried out the electrophysiological study and a critical review of the manuscript. MK performed brachytherapy and a critical review of the manuscript. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2407/5/79/prepub> Figures and Tables ================== ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Patients\' clinicopathological characteristics ::: No Sex Age (years) Histology Site Preserved nerve Tumor grade Size (cm) ---- -------- ------------- -------------------------------- --------- ----------------------- ------------- ----------- 1 Female 78 Myxoid liposarcoma Forearm Median, Radian, Ulnar High 13.0 2 Male 48 Malignant fibrous histiocytoma Thigh Sciatic High 5.5 3 Female 48 Myxoid liposarcoma Forearm Median High 5.5 4 Male 35 Malignant fibrous histiocytoma Thigh Sciatic High 9.0 5 Male 61 Extraskeletal chondrosarcoma Pople Popliteal Low 10.0 6 Female 15 Synovial sarcoma Buttock Sciatic High 10.0 7 Male 84 Malignant fibrous histiocytoma Buttock Sciatic High 13.0 ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Treatment results ::: No. Margin PTV^a^(cm^3^) Brachytherapy (Gy) EBRT^b^(Gy) Failure Complication Follow-up (months) ----- ---------- --------------- -------------------- ------------- --------- --------------- -------------------- 1 Negative 65 50 \- \- Motor paresis AWD^c^(77) 2 Negative 63 50 (50) Local Fracture AWD (75) 3 Negative 52 50 \- \- \- AWD (62) 4 Negative 93 50 \- \- \- AWD (43) 5 Negative 64 30 20 \- Sensory loss AWD (34) 6 Negative 60 50 \- Lung \- DOD^d^(32) 7 Positive 134 50 \- \- \- DOC^e^(13) PTV^a^= Planning target volume, EBRT^b^= External beam radiation therapy, AWD^c^= Alive without disease, DOD^d^= Dead of disease, DOC^e^= Dead from other cause ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Electrophysiological results ::: No. Preserved nerve Motor nerve or F-wave conduction velocity (m/s) Months after brachytherapy ----- ----------------- ------------------------------------------------- ---------------------------- ---- 2 Sciatic nerve 50.3 52.3 75 3 Median nerve 55.0 56.0 62 4 Sciatic nerve 58.5 50.1 43 :::	PubMed Central	2024-06-05T03:55:59.786891	2005-7-19	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181808/", "journal": "BMC Cancer. 2005 Jul 19; 5:79", "authors": [ { "first": "Tadahiko", "last": "Kubo" }, { "first": "Takashi", "last": "Sugita" }, { "first": "Shoji", "last": "Shimose" }, { "first": "Toshihiro", "last": "Matsuo" }, { "first": "Ken", "last": "Hirao" }, { "first": "Hiroaki", "last": "Kimura" }, { "first": "Masahiro", "last": "Kenjo" }, { "first": "Mitsuo", "last": "Ochi" } ] }
PMC1181809	Background ========== Data concerning incidence and prevalence of cancer in the Soviet Union (USSR) and Russia have traditionally not been provided on a basis that facilitated comparison with data from countries in western parts of Europe \[[@B1]\]. This was due to differences in routines and procedures for reporting, registration and diagnostics \[[@B2]\], as well as to a lack of information that described the procedures. Today, the cancer registry in St Petersburg is the only regional cancer registration in Russia that has been recognised by the European Network of Cancer Registries to fulfil international standards, but efforts have been made to standardise the system within Russia \[[@B2]\]. In 1997, the oncological hospital (OD) in Arkhangelsk and the Universitetet i Tromsø (Norway) started the planning of a population based cancer registry for Arkhangelskaja Oblast (AO). The systematic registration started in 1999, and the objective was to register all cancer cases in AO from 1993 and onwards. A verifiable cancer registry that covers the population of AO will not only give valuable insight into the cancer incidence in Russia, but also into the incidence in a northern population that has been relatively homogenous both ethnically and socio-economically. The aim of this investigation was to assess the content and quality of the AO cancer registry (AKR), and to present the site-specific cancer-incidence rates in AO in the period 1993--2001. Methods ======= Context ------- AO is an administrative unit with 1.3 million inhabitants. It covers 587 000 km^2^(larger than France) north of the 60^th^latitude in North-West Russia (Figure [1](#F1){ref-type="fig"}). The unit consists of 26 boroughs, whereof 6 are cities. In addition, there is an autonomous region, Nenets, in the north. Until the late 1990s, twelve percent of the land area was for use by military units only (Nenets not included) \[[@B3]\]. The administrative centre is the city Arkhangelsk, which houses 28 percent of the population. The main industries are pulp and paper mills, and heat and power plants. Gas and oil industry predominate in Nenets. The population of AO grew during the 1980s and reached a historical high of 1 576 000 inhabitants in 1991. In the years that followed, the mortality and migration increased, and the birth rate went down \[[@B4],[@B5]\]. In year 2000, the largest ethnic groups were Russians (92 percent); Ukrainians (3.4); Byelorussians (1.3); Komi (0.5); and Nenets (0.5), and 74 percent of the population resided in urban areas \[[@B5]\]. In 1999, the life expectancy in AO was 71.1 years for women and 58.0 years for men, which was a decrease from 74.7 and 64.0, respectively, ten years earlier \[[@B5]\]. Life expectancy was higher in urban than in rural areas, but lower than in the neighbouring regions in North-West Russia \[[@B5],[@B6]\]. The most frequent causes of death were connected to the circulatory system (54%); accidents (16%); and tumours (14%) \[[@B4]\]. The only oncological hospital (OD) in AO is located in the city Arkhangelsk, where 35 of the about 50 practising oncologists in 2002 worked. About one-half of the boroughs had at least one oncologist. The boroughs without an oncologist did instead have a medical worker who was responsible for cancer recording and reporting, but this person had, in general, no contact with the cancer patients. Every Russian citizen has the right to receive medical care without service fees. When a patient is referred to a district or central hospital, the local hospital might cover the travel expenses if financial resources are available. But more often than not, the patients must cover these expenses themselves. The latter is also true for persons seeking a doctor or a hospital within his/her home borough. All forms of cancer treatment are available in the city Arkhangelsk. Radiotherapy is only available there, while surgery is carried out in every hospital where there is a surgeon qualified to perform the appropriate surgery. The same is true for chemotherapy, but this treatment can only be given after the patient has been examined at the OD, and under supervision by them. The pathology department at the OD, which is the only one in AO, has had separate laboratories for cytology and histopathology. The latter had 3 pathologists, and processed 35 000 surgical pathology specimens annually. The department had available to them a complete version of the WHO International Histological Classification of Tumours in Russian. Immunohistochemistry was established at the AO Main Hospital in 2001, in co-operation with the hospital in Tromsø. The pathology department and the main hospital co-operate closely, and frequently consult each other when faced with difficult cases. Study population ---------------- The population in this study consists of all individuals registered as residents of AO. All new cancer cases in the period 1993 -- 2001 registered in the AKR were included in the study (International Classification of Disease, version 10 (ICD-10): C00-C95, except for C77-78 (secondary malignant neoplasm)), and the data obtained were anonymous. The estimated incidence rates of cancer were adjusted for age using the world standard (5-year age-group intervals) \[[@B7]\]. In instances where an individual was diagnosed with cancer more than once during a six-month period, only the first diagnosis was included in the rate estimates. When a registered diagnosis of cancer subsequently showed not to be cancer, the case was not included. The annual gender and age-group-specific population figures for AO were obtained from the federal statistics office of AO. In the calculations of cancer incidence for year 2001, population figures for year 2000 were used for the age groups older than 69 years. Cancer reporting and the Arkhangelsk Cancer Registry ---------------------------------------------------- The cancer reporting and registration is founded in legislation. The oncologist/doctor who determines a cancer by diagnosis fills in a report form, and is required to submit it to the OD within three days. Notification has been obligatory since the 1960s. The form contains guidelines for proper fill-out. For instance, the diagnosis is to be spelled out in words. If the local oncologist/doctor is uncertain about the diagnosis, the patient is to be referred to a larger hospital or the OD. For patients who are treated locally by a doctor who is not an oncologist, a medical report is submitted when a patient is discharged. The medical-report form also includes a field for description of the treatment and course of the disease. Both ICD-codes and words are used in the description. When the OD receives a medical report with incoherence between the two, the OD systematically contacts the local hospital for clarification. In the submitted report, it is noted whether the patient is a resident of another borough than where he/she was diagnosed/treated. When a resident of AO is diagnosed elsewhere in Russia, the diagnosing hospital forwards the report to the OD in AO, and vice versa when non-AO residents are diagnosed in AO. Thus, reports of cancer cases among students and conscripts are forwarded to the regions in Russia where they have permanent residence. When a person dies, a doctor submits a medical document to the local administrative office, which issues a death certificate and a certificate about identity and citizenship (no burial can take place before a death certificate is issued). The certificates are forwarded to the AO statistical office. The cause of death is established either on the basis of the deceased\'s anamnesis, or by pathological or forensic examination. Only one cause of death is recorded on the certificate, and the statistical office defines the ICD-code. Should the recorded cause of death be changed later, a notification is submitted to both the administrative and statistical offices. The family of the deceased can object to, or demand, a post-mortem examination. If a pathological examination reveals cancer, then both the stage and metastasis are made note of on the certificate. Medical employees from the OD visit the statistical office monthly to pick up information about cancer deaths. The OD then requests information about the revealed individuals from the registrar\'s office. This information is checked against the AKR to find out whether the specific case already has been registered. The information from the registrar also reveals whether an individual has changed surname between the time of diagnosis and time of death. The OD sends an annual report to the ministry in Moscow about cancer incidence and mortality in AO. The report also includes the number of cases that was verified histologically and clinically. The OD is required to verify unverified cases within two years. Cancer registration ------------------- In 1996, a computer programme was put into use in the clinical work due to a directive from the health ministry. Data about cancer cases had until then only existed on paper-based patient records that were organised in an archive. The work to set up the AKR started in 1998. A data-entry programme was made using Access^®^, and all reported cases of cancer since 1993 were registered retrospectively. From 2001, the cancer cases were registered as soon as reports were received. The computer programme for the registration has been revised and improved twice since the registration started. Simultaneously, the information registered and the number of database fields included was modified (see Table [1](#T1){ref-type="table"}). The electronic cancer registry has been administered and run by the OD. The registration in the AKR was based on the information obtained on submitted form(s), and was updated if revised information was reported later. The oncologist in charge of the AKR routinely checked that reported histology and diagnosis were in accordance, and set the ICD-code if it was missing. The diagnosis was registered both in words and with a three character ICD-code. ICD-9 codes were used in registration of cases from the period 1993 -- 99, and ICD-10 from year 2000. Thus, in the registered ICD-9 codes it was not possible to separate cancers in the sigmoid junction and rectum, so in the present study sigmoid junction was included in the rate estimate of rectum cancer -- and not of colon cancer. The data registered about each case are presented in Table [1](#T1){ref-type="table"}. Not all of the data items were considered a must. But the following information has been obligatory (and inquired for if missing on the submitted form): name, patronymic, surname, gender, date of birth, place of residence, diagnosis, date of diagnosis, histology, stage according to the TNM-classification, method of diagnosis and treatment, place of treatment, and whether the patient is alive or dead. If dead, the following information was required: date of death, cause of death, and whether a pathological examination had been conducted. The identification of an individual was based on name, surname, patronymic and birth date. The AKR registered whether a tumour was first revealed through autopsy, or if a case was first reported through a death certificate. The outcome of the cancer disease was also registered. The patients\' last known status was updated annually. The updating was based on a final-treatment report, or on the patient\'s registration card if treated at the OD. One of the following eight possible outcomes was noted: alive; dead due to the treatment, the cancer, or another cause; treated basal cell carcinoma (and 5 years have passed without a relapse); moved to another region of Russia or abroad; the diagnosis showed to not be cancer; or unknown. The method of diagnosis was registered as one of the following: only clinical; histological; cytological; isotopic; endoscopic; x-ray; electronic tomography; ultra sound; surgery; or unknown. There were tick-off boxes on the report form for six of the methods, while the others usually were noted if used. The registration procedures in the AKR included some internal quality control systems. After entry; the name, age, borough and diagnosis were checked visually against the received information to ensure correctness. The database programme used since year 2000 automatically recognises whether an individual has been registered before if the following data are identical: name; patronymic; surname; gender and birth date. In addition, the following logical checks are embedded in the programme: 1) that the birth date precedes the date of diagnosis in terms of real time; 2) that the date of diagnosis precedes the date of treatment/surgery; 3) that gender-specific diagnoses are logical in terms of registered gender. The latter was previously done by visual checks. Visual checks were also done to look for contradictions between diagnosis and histological verification and stage, respectively. Furthermore, most data-entry fields are formatted, which ensures that impossible or undefined values cannot be entered. In fields where a numerical value represents some defined information, the field is linked to a drop-down window that depicts and explains the range of possible values that can be entered. The fields that are formatted are depicted in Table [1](#T1){ref-type="table"}. Quality and content assessment ------------------------------ To assess the completeness and correctness of the AKR, two of the authors (JAL and AV) visited two district hospitals (in Severodvinsk and Kholmogory) to check whether the living cancer patients in their files had been registered. The two hospitals were given a one-day notice before the control. The researchers were given full access to the cancer-patient files. In Severodvinsk 42 out of 723 and in Kholmogory 61 of 396 journals of living cancer patients were randomly selected and looked for in the AKR database. In addition, in Severodvinsk the following specific items were checked against the registry for each of the selected journals: name; year of birth; diagnosis; and date of diagnosis. These data were read aloud from the hospital files by JAL and checked in the database records by AV. The assessment revealed that six of the cancer patients residing in Severodvinsk were not registered. One of the 42 could not be checked because the person was not a resident of the borough. Out of the 59 who were residents of Kholmogory, none were missing in the AKR. Thus, overall 6 out of 100 (6 %) cancer cases belonging to the population were not registered (95% CI: 1.3--10.7%). The data obtained in Severodvinsk were also used to assess the data-entry error in the AKR. In two of the checked files (5.7%), the diagnosis registered differed from the diagnosis in the journal. Both were recorded as ICD-10 C19 in the files, but registered as C20 and C50 in the AKR. One surname was incorrectly registered (1%), while for three cases there was a discrepancy in year of birth (3%). The registered year of diagnosis was in accordance with the data in all the checked journals. One record was missing the year of diagnosis and one other was missing the birth year, but the data were recorded in the patients\' files. The data-entry error rate was also assessed by checking the consistency between registered gender and gender-specific diagnoses (ICD-9: 174, 175, 180, 182, 183, 185 and 186) (this check was embedded in the computer programme from year 2000). There was a discrepancy between registered gender and diagnosis in one out of 4294 records (one case with the diagnosis ICD 174, female breast cancer, was registered as a male). The quality of the AKR was further assessed by looking at distributions of data; the proportions of missing values; the presence of non-logical values; and the proportion of registered cases whose case information was obtained from a death certificate. In September 2001, Helge Stalsberg, a professor in morphology at the University of Tromsø and head doctor of the Department of Pathology at the University Hospital, was invited to Arkhangelsk to review the histological-diagnostic practise. Stalsberg was given full access to the material, including protocols concerning the diagnoses given for each case. He selected a priorito review the slides of the 20 most recent cases of each of the following cancer types: lung; stomach; colon; breast; and testes. Without consulting the protocol, he set a diagnosis one by one. His diagnoses were subsequently compared with the protocolled diagnoses. For testis cancer, only six cases were reviewed, as only six cases had been diagnosed so far that year. Thus, in total 86 cases were reviewed -- all from 2001. The quality of the histological sections was considered adequate. There was full agreement concerning presence of invasive carcinoma in all cases of stomach and testis cancer. For both breast and colon cancer there were agreements for 19 of 20 slides. One of each was in disagreement on in situversus invasive. For lung cancer, 2 of 20 were in disagreement on in situdysplasia versus no dysplasia. The microtomes and microscopes possessed by the laboratory were found fully adequate for histological-diagnostic work. Results ======= A total of 35 224 cases of cancer, diagnosed in the period 1993 -- 2001, were registered in the AKR. For 13 of the cases, the diagnosis had later been changed to non-cancer, while 42 cases were benign, in-situ carcinoma or secondary cancers. In addition, 472 registered cases had been revealed through autopsy. Thus, 34 697 cases were included in the calculations of cancer incidence. Of these, 95.2 percent had never had cancer before, and men constituted 51.3 percent. The number of new cases varied from 3559 in 1996 to 4153 in year 2000. In terms of age, the mode was 67 years, the median 63 years, and the range 0 -- 99 years. The annual proportion of diagnoses that were verified histologically was around 60 percent, while the proportion verified cytologically increased from 15 to around 20 percent in the studied period. The highest proportion verified histologically was for rectal cancer (81.4 %), while breast cancer was the cancer type most likely to be verified cytologically (36.8 %). (These figures include cases that were registered based on death certificates). The highest proportion verified by one of these two methods was in the age group 20 -- 29, and the proportion was lower the older the age group. The age-adjusted incidence rate for all sites combined was 164/100 000 for women and 281/100 000 for men. The highest incidence was cancer of the trachea, bronchus and lung (16.3% of all cases), whereof 88.6 % of the cases were among men. The second most frequent was stomach cancer (14.7%). Among women, cancer of the breast constituted 15.9 percent of all cases. The age-adjusted gender-and site-specific incidence rates are presented in Table [2](#T2){ref-type="table"}. The cancer sites with the highest rates among men were: lung (77.4/100 000); stomach (45.9); rectum (13.4); oesophagus (13.0); and colon (12.2). Among women they were: breast (28.5); stomach (19.7); and colon (12.2). Some trends of cancer incidence are presented in Figures [2](#F2){ref-type="fig"}, [3](#F3){ref-type="fig"}, [4](#F4){ref-type="fig"}, [5](#F5){ref-type="fig"}, 6. Discussion ========== According to MacLennan (IARC), the following information is considered basic in a population-based cancer registry: date of diagnosis; basis of diagnosis; site of primary tumour; morphology; behaviour of tumour; source of information; place of residence; ethnic group; and personal information sufficient to ensure the recognition of an individual when the same person is reported more than once to the registry \[[@B8]\]. All these items were registered in the AKR, except for ethnicity. Ethnicity could be of interest to register in AO due to the Nenets indigenous population (0.5% of total population), but this was made unattainable when the Russian Federation implemented the use of \"Russian\" as nationality for all in identity documents. No unique identification number existed in Russia, but the identification of the individual was fulfilled in the AKR by registering the full name, gender and birth date. Quality ------- The AKR is a comprehensive registry in terms of important information about each cancer case, and had few data-entry errors. The pathology department at the OD works according to the classifications of the WHO International Histological Classification of Tumours, and the diagnostic procedures, equipment and skills were found to be fully adequate. The correctness of incidence rates is influenced by the completeness of cover. The numerator in the rate estimates is deflated if not all cases of cancer in the official resident population were captured by the health system, and/or if the AKR did not receive reports about all diagnosed cases. The denominator, on the other hand, is incorrect if the health system of AO covered more or fewer people than the official population figures. The results of the controls carried out in Severodvinsk and Kholmogory indicate that the submission of cancer reports to the AKR was not fully complete, and that the reporting varied from district to district. There was no apparent explanation for the difference in reporting between the two clinics, which both had an oncologist. Incomplete reporting has also been found to hamper the estimations of cancer incidence in established systems with a long tradition of population based cancer registration, such as in Norway \[[@B9]\]. On the other hand, a fully complete cancer registry is inconceivable. The AKR is a population-based cancer registry that is meant to cover the population of AO. Conscripts and students from elsewhere were not included in the official population figures, and cancer cases among them were not registered. There were two sub-groups of the official population of which some members were likely to survive, or die, with an undiagnosed cancer, and thereby contribute to an underestimation of incidence rates; namely, the indigenous Nenets and the poor elderly. The former mainly live in the north, and those among them who lived traditionally on the tundra as herders and hunters were less likely to seek or receive medical care from the public health care system. Poverty in Russia increased markedly in the 1990s -- especially among elderly. Although health services in Russia were free and without service fees, the transport to the doctor or hospital usually had to be paid by the individual. Distances in rural areas are vast, and travelling was often uncomfortable and relatively expensive. Persons of age 80 years or older had the opportunity to transport without paying by inquiring for a requisition in advance. However, it is not unlikely that an increasing proportion of the elderly died from, or with, an undiagnosed cancer. Nevertheless, the age-specific incidence rates, all sites combined, in the older age groups did not decline within the study period. Other sub-groups that were included in the official population figures, but possibly not fully reported to the AKR, were professional military personnel and railway and shipyard workers. These groups had their own health care privileges or systems that did not sort under the health authorities in AO, but directly under the state. This meant that patients who could not be treated within their own health system were sent to Moscow for treatment. Reporting was also done directly to Moscow, but this ended in the late 1990s for shipyard and railway workers. The OD had contact with the patients from these groups only if referred to them, and subsequently these patients would also be included in the AKR. The immediate families of these workers could also use the same health care system. The shipyard workers were located the city of Severodvinsk (population 230 000), which was a closed city due to the construction of nuclear and naval vessels at the shipyard. Figures about the number of workers in each of these groups were not publicly available, so the number of missing cases in the AKR from these two groups of workers is not readily estimated. But presumably the reduced allocation of funds after the disintegration of the USSR had the effect that fewer workers were sent to Moscow for cancer treatment. Furthermore, the retirement age for shipyard workers was as low as 45 years for women and 50 years for men, meaning that the workers were retired before they were in the highest risk groups in terms of age for most cancer types. The retirement age for male military personnel was 55 years. Another group of concern in terms of population coverage was emigrants. People moving from AO to somewhere else in Russia were likely to register change of address, and if not, the report forms, if diagnosed elsewhere, would have been forwarded to AO. But quite a few people have emigrated to other former Soviet republics or other countries, and these people had no incentive or requirement to register. According to official figures there was a net emigration from AO of about 6000 people per year on average in the study period \[[@B10]\]. However, the first census since the disintegration of the USSR, which was held during the autumn of 2002, revealed that the actual population of AO was about 90 000 people lower than the official population figures, and that 2/3 of the deficit was among men \[[@B11]\]. Thus, the net emigration had been more than twice the official figures -- on average. Based on age in 2002, the largest population discrepancies for men were in the age groups 35--39 and 60--64, and in the age groups 10--14 and 60--64 for women. The underestimation in the reported overall incidence rates for the whole study period should then be less than four percent due to this population factor. But since the population discrepancy varied with age group and gender, the underestimation will be of a higher magnitude for some cancer types and lower for others. The discrepancy between the official population-figures used in the present rate estimations and the real population size was presumably accumulative over time since the disintegration of the USSR. Hence, the influence of an inflated denominator on the estimated incidence rates was likely small in the early years of the study period, and more profound in the rates for 2000 and 2001. The findings and notions discussed above indicate that the reported rates are underestimations, especially among men. However, the relative magnitudes of the site-specific cancer incidences within a gender would only have been affected by underreporting of cases if the population sub-groups discussed above were unusually prone, or less disposed, to certain types of cancer. Cancers related to the elderly were presumably of most concern in this perspective, but in the shipyard there might have been exposure to asbestos. The proportion of cancer cases that was not captured by the health system, or diagnosed cases not caught by the AKR, can be estimated by calculating the annual proportion of cases among not-previously-registered individuals that were obtained from death certificates. The completeness of cover may also be evaluated by comparing the incidence rates of the different cancer types. The number of cases that were registered based on death certificate was 419 (2.4%) among men and 538 (3.0%) among women. The proportion varied from \<0.6 percent in the years 1993--94 and 2000--01 to 5 percent in the period 1995--99. The actual proportion-level is an indication of how well the system works, and if the system worked consistently the proportion should have been fairly constant. Thus, the ability of diseased people to seek examination and treatment, and/or the function of the system, appears to have been impaired during the years of severe economic hardship. A study by Shkolnikov et al found that cancer deaths in the older age groups were under-recorded in Russia -- especially in rural areas \[[@B12]\]. In Norway, the proportion of registered cases obtained from death certificates was 2.6 percent for both genders in 1993 \[[@B7]\]. The analysis revealed that almost 80 percent of all cases were verified histologically or cytologically. In the former USSR as a whole, 69.8 percent of the diagnoses were verified microscopically in 1989, and the proportion of cases verified varied from 62.4 to 81.3 percent between the different republics or oblasts within the union \[[@B1]\]. In the neighbouring country Norway, 86 percent of all cases since 1953 have been verified histologically \[[@B13]\]. Cancer incidence ---------------- The age-adjusted cancer incidence among men, all-sites combined, was similar to the incidence in Norway. But among women the incidence was more than 30 percent lower in AO \[[@B13]\]. However, one should be careful in comparing the incidence rates in Russia with western countries since the competing risks of disease and death were different. The magnitude of the large difference in cancer incidence between men and women in AO was likely not due to an underestimation of rates, as the mentioned possible causes of underestimation mainly concerned men. For most cancers, the site-specific incidences found in AO were comparable both in magnitude and relative magnitude to rates for Russia in 1990, as reported by the International Agency for Research of Cancer (IARC) \[[@B14]\]. The total age-adjusted incidence among men was 284/100 000 in Russia vs. 267 in AO, while 170 vs. 151/100 000 among women (ICD-9: 173 was not included in these calculations). In comparison with the IARC-reported incidence, the site-specific incidence in AO among men was higher for oesophagus, and lower for larynx and testis cancer. Also women had a higher incidence of oesophagus cancer in AO, but the incidences of cancer of the lung, breast and cervix uteri were lower \[[@B14]\]. Compared to the incidence of female genital cancers in St. Petersburg, the incidences of corpus uteri and ovary cancer were lower in AO, while the incidence of cervix uteri cancer was of similar magnitude \[[@B15]\]. Interestingly, the rate of stomach cancer was relatively high and colon cancer low, just as in Norway 30--40 years ago when the rates were quite different from today \[[@B16]\]. The age distribution in AO, as in Russia, is different than the world standard population. Compared to the world standard (and most Western-European populations), there were relatively few older men and children below 10 years of age, as well as a small World-War II generation (age 50 -- 59 in the study period). On the other hand, the post-war generation was relatively large (age 35 -- 49). Thus, the distribution in AO was pear-shaped, in contrast to the pyramid-shape of the world standard population and the fairly evenly distribution of most Western-European populations. This means that the age adjustment of the incidence rates in AO contributed to an over-estimation of the cancer types that were most likely to develop in the age groups that were relatively small, and vice versa, compared to the actual burden of those cancers in the society. The reported incidence rates in AO contribute to the discussion and generation of hypotheses concerning the role of different life-style and environmental factors in the aetiology of different cancer types. The AKR provides data for estimations and interesting insight to the cancer incidence in a northern Russian population. Using the registry for further investigations into district and age-group-specific differences in cancer incidence would shed additional light on the concerns and questions surrounding environmental and life-style aspects and cancer. Conclusion ========== We believe that the reported incidence rates reflect the cancer situation in AO well. Our findings then confirm and strengthen the indication that the incidence of melanoma of the skin and cancers of the stomach, larynx, liver, pancreas, prostate, colon, and bladder are quite different in male populations in Russia than, for instance, in Norway. Among women, most major cancer types, except stomach, appear to be relatively low in Russian populations. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= AV did the data analysis, quality assessment of the data and was the senior author. JAL was responsible for collecting the data, the data entry and the information about the health system and sub-populations. DSK designed the data entry programmes and extracted the database used in this study. AVT initiated the AKR, supervised the research group, and gave approval for the final version to be published. Sadly, AVT passed away not long thereafter. TSP made the AKR possible administratively. EL was the principal investigator, revised the article critically and gave approval of the final version. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2407/5/82/prepub> Acknowledgements ================ This investigation was sponsored by the Norwegian Cancer Association, which also sponsored the set up of the AKR. The authors thank Helge Stalsberg for the review of the histological diagnostic practise in Arkhangelsk, and the staff in the AKR for their work and co-operative efforts. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### North-western Russia \[17\]. ::: ![](1471-2407-5-82-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Age-adjusted, gender-specific incidence of cancer, all sites combined, in Arkhangelskaja Oblast. ::: ![](1471-2407-5-82-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Trend of age-adjusted colon and stomach cancer among men and women in Arkhangelskaja Oblast. ::: ![](1471-2407-5-82-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Trend in age-group specific cancer incidence among women in Arkhangelskaja Oblast, all sites combined. ::: ![](1471-2407-5-82-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Trend in age-group specific cancer incidence among men in Arkhangelskaja Oblast, all sites combined. ::: ![](1471-2407-5-82-5) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Data items registered in the Arkhangelsk Cancer Registry ::: Data registered Comments --------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------------- --- Number at registration card Surname, given name, patronymic In separate fields from 1995 Gender C Residence (urban or rural) C Date of birth For the period 1995--99 the year was registered Rajon (borough) of residence C Home address From 2000 also home oblast included Occupation C Date when registered as sick (by general health system) Result of treatment / patient\'s status Last known, at the end of the year. C Date of last status Clinical health group at the time of registration e.g. Will be cured, etc. From 1995 C Clinical health group at the end of year From 1995 C Descriptive diagnosis Cancer diagnosis ^1^ Only for the years 93--94 C Diagnosis by ICD-9 code ^1^ Only for the years 1993--99 C Diagnosis by ICD-10 code ^1^ Only for the years 1993--94, 2000- C How the tumour was detected Registered if method was autopsy or only obtained from death certificate C Stage C TNM From 1995 Method of verification of diagnosis Morphologic, cytologic, X-rays, endoscopy, isotopic, ultra sound, computer tomography, operation, only clinical, or unknown C Date of diagnosis From 1995 Reason for late diagnosis From 2000 C Morphological code C Main cancer site (when more than one site) For internal use clinically Histological structure of tumour Classification for internal use. From 1995 C Synchronic tumour Yes/no. Only in period 1995--99 Localisation of cancer (side of body) Right/left C Information about treatment For the period 1995--s99 Method of treatment, if radical From 2000 C Institution where diagnose was determined From 2000 C Institution where patient is under observation Medical institution observing the patient C Metastasis? Yes/no. Only in period 1995--99 Number of tumours From 2000 C Date of last contact with patient From 2000 Cause of death From 1995. Including code of cause from 2000 C Date of death From 1995 Pathological examination Yes/no. From 2000 C Nationality From 1995 C Date cancer report received From year 2000 C = The field in the AKR database was formatted. I.e. data entry was restricted to possible or certain values. ^1^Three fields in the database, but in the period 1993--94 the values appeared in all three fields when data were entered in one of them. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Cancer in Arkhangelskaja Oblast 1993--2001. Number of new cases and age-adjusted incidence rates per 100 000, by primary site and gender ::: Males Females -------------------------------------------- --------- ------------- ----------- ------------- ------ ------ Lip 140 C00 266 4.2 95 0.7 Tongue 141 C01-02 109 1.6 27 0.2 Gum 143 C03 35 0.5 20 0.2 Floor of mouth 144 C04 88 1.3 5 0.1 Palate 145 C05-06 66 1.0 24 0.2 Salivary glands 142 C07-08 29 0.4 39 0.4 Tonsil / oropharynx 146 C09-10 168 2.4 42 0.4 Nasopharynx 147 C11 24 0.3 17 0.2 Periform sinus and other unspecified 149 C12, C14 58 0.8 4 0.0 Hypopharynx 148 C13 117 1.7 7 0.1 Oesophagus 150 C15 814 13.0 414 3.3 Stomach 151 C16 2880 45.9 2232 19.7 Small intestine 152 C17 32 0.5 27 0.3 Colon 153 C18 753 12.2 1346 12.2 Rectosigmoid junction, rectum and anus 154 C19-21 805 13.4 858 7.8 Liver 155 C22 325 5.1 221 2.0 Gallbladder and other unspecified 156 C23-24 82 1.4 183 1.5 Pancreas 157 C25 590 9.2 566 4.9 Other digestive organs 159 C26 2 0.0 4 0.0 Nasal cavity and paranasal sinuses 160 C30-31 58 0.8 49 0.5 Larynx 161 C32 472 6.9 33 0.4 Trachea, bronchus and lung 162 C33-34 5011 77.4 646 5.8 Pleura, thymus, mediastinum and heart 163,164 C37-38,C45 53 0.8 43 0.4 Other respiratory 165 C39 1 0.0 0 0.0 Bone, connective tissue, Kaposi\'s sarcoma 170 C40-41,C46 120 1.9 78 0.8 Melanoma of skin 172 C43 169 2.5 291 3.0 Other skin 173 C44 856 14.4 1478 12.9 Nervous system 192 C47,C70,C72 19 0.3 21 0.3 Retroperitoneum and peritoneum 158 C48 86 1.4 117 1.1 Connective and other soft tissue 171 C49 102 1.5 115 1.2 Breast 174,175 C50 20 0.3 2689 28.5 Vulva, vagina. other female genital organ 184 C51-52, C57 0.0 161 1.4 Cervix uteri 180 C53 0.0 790 8.4 Corpus uteri 182 C54 0.0 779 8.1 Uterus. unspecified 179 C55 0.0 3 0.0 Ovary 183 C56 0.0 854 9.0 Placenta 181 C58 0.0 11 0.2 Penis and other male genital organs 187 C60, C63 31 0.5 0.0 Prostate 185 C61 611 11.1 0.0 Testis 186 C62 110 1.5 0.0 Urinary tract, except bladder 189 C64-66,C68 592 9.1 574 5.7 Bladder 188 C67 737 11.6 223 1.8 Eye 190 C69 33 0.5 46 0.6 Brain 191 C71 303 4.4 289 3.7 Thyroid gland 193 C73 98 1.4 457 5.1 Adrenal and other endocrine glands 194 C74-75 27 0.4 18 0.2 Other or unspecified site 195,199 C76,C80 257 4.0 188 1.8 Hodgkin\'s disease 201 C81 184 2.7 158 2.1 Non-hodgkin\'s lymphoma 200 C82-83,C85 198 3.3 157 1.7 Other lyphoid and histiocytic tissue 202 C84,C96 18 0.3 10 0.1 Multiple myeloma 203 C88-90 70 1.1 96 0.9 Lymphoid leukaemia 204 C91 213 3.5 187 2.0 Myeloid leukaemia 205 C92 94 1.4 138 1.5 Monocytic leukaemia 206 C93 5 0.1 8 0.1 Other specified leukaemias 207 C94 21 0.3 25 0.3 Unspecified leukaemias 208 C95 26 0.4 39 0.4 :::	PubMed Central	2024-06-05T03:55:59.788502	2005-7-19	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181809/", "journal": "BMC Cancer. 2005 Jul 19; 5:82", "authors": [ { "first": "Arild", "last": "Vaktskjold" }, { "first": "Jelena A", "last": "Lebedintseva" }, { "first": "Dmitrij S", "last": "Korotov" }, { "first": "Anatolij V", "last": "Tkatsjov" }, { "first": "Tatjana S", "last": "Podjakova" }, { "first": "Eiliv", "last": "Lund" } ] }
PMC1181810	Background ========== Neurofibromatosis type 1 (NF1) is an autosomal dominant neurocutaneous disorder with an estimated prevalence of 2 to 3 cases per 10,000 population \[[@B1]\]. NF1 is characterized by neurocutaneous signs such as café-au-lait spots, axillary freckling, and cutaneous neurofibromas. Cognitive deficits and academic learning difficulties are the most common neurological complications of NF1 in childhood. Moreover, patients with NF1 are at an increased risk of developing nervous system neoplasms. Malignancy other than of nervous system origin can develop, with the prevalence of such malignant tumors reportedly being four times higher in NF1 patients than in the general population \[[@B2]\]. Gastrointestinal malignancy was reported to be associated with NF1 patients, with neuroendocrine tumors being the most prevalent \[[@B3]\]. However, there have been only one report about gastrinoma associated with NF1. Zollinger-Ellison syndrome (ZES) is characterized by hypersecretion of gastrin from a gastrinoma that leads to gastric acid hypersecretion and, most notably, clinical symptoms of refractory peptic ulcer disease. Zollinger and Ellison were the first to report recurrent peptic ulcers of the jejunum associated with nonbeta islet cell tumors of the pancreas \[[@B4]\]. The diagnosis is based on a combination of criteria, including clinical presentation, serum gastrin level, acid-secretion test, and diagnostic imaging studies. Improvements in medical management using proton-pump inhibitors mean that gastric acid hypersecretion can be effectively controlled in most cases. Determining the presence of metastasis to regional lymph nodes or the liver to monitor malignant progression is important before definitive treatment. ZES in the form of multiple endocrine neoplasia type 1 (MEN1) needs special consideration, since the disease progression, presentation, and treatment differ from those in the sporadic form \[[@B5]\]. In this report, we describe a patient with NF1 who was diagnosed as ZES and successfully treated by curative surgery. Case presentation ================= A 41-year-old female NF1 patient was admitted for the evaluation of a 3-year history of recurrent epigastric soreness, heartburn, and diarrhea. Repeated endoscopic examinations revealed recurrent duodenal ulcers. The symptoms were relieved by proton-pump inhibitors, but recurred when the medication was withdrawn. She was previously diagnosed with NF1 based on clinical features and family history. Her first-degree relatives (i.e., brother, sister, and second son) were also affected by fully developed features of NF1. She had no cognitive dysfunction or learning disabilities, and showed normal intellectual development. A physical examination revealed several café-au-lait spots and multiple small nodules on the anterior chest and areolar area, and also multiple axillary freckles. A skin-nodule biopsy demonstrated characteristic neurofibromas. On admission, upper gastrointestinal endoscopy revealed multiple shallow ulcers in the descending duodenum. The rapid urease test and urea breath test for Helicobacter pyloriwere both negative. The serum fasting gastrin level was \>1,000 pg/mL and 837 pg/mL in two consecutive measurements. A secretin stimulation test was not performed due to the unavailability of secretin. As a diagnosis of gastrinoma was strongly suggested, radiologic evaluations were performed to locate the lesion. Abdominal computed tomography (CT) and magnetic resonance imaging (MRI) scans showed a 3 × 2-cm, clearly defined, well-enhanced mass adjacent to the duodenal loop in the subhepatic space (Figure [1](#F1){ref-type="fig"}). No metastatic lesions were observed in the liver or regional lymph nodes. The association with MEN1 status was evaluated by hormonal and radiologic investigations of the parathyroid, pituitary, pancreas, and adrenal gland, which proved to be negative. As a definitive treatment, an intraabdominal mass of about 2.5 × 2 cm with a thick fibrous capsular outer layer was surgically isolated and completely resected from the right lateral border of the descending duodenum. Gastrinoma was finally diagnosed by immunohistochemical staining (Figures [2](#F2){ref-type="fig"}, [3](#F3){ref-type="fig"}). An incidental finding during the operation was numerous, small (\<1.0 cm), whitish nodular masses on the serosal surface of the small bowel, many of which were enucleated. The pathologic diagnosis was gastrointestinal stromal tumor (GIST) with a low mitotic index as evidenced by C-kit positivity (Figures [4](#F4){ref-type="fig"}, [5](#F5){ref-type="fig"}). The serum gastrin level decreased markedly to 99.1 pg/mL at 1 month after the operation, and to 53.2 pg/mL 3 months later. There were no symptomatic or radiologic recurrences during a 14-month follow up. Discussion ========== NF1 is an autosomal dominant single-gene disorder with an estimated prevalence of 2 to 3 cases per 10,000 population \[[@B1]\]. NF1 is characterized by neurocutaneous signs, such as café-au-lait spots, axillary freckling, and cutaneous neurofibromatosis, and iris hamartomas \[[@B6]\]. Neurofibromas are benign peripheral nerve tumors composed of proliferating Schwann cells and fibroblasts. They present as multiple palpable, rubbery cutaneous tumors, and are generally asymptomatic. In our case, the diagnosis of NF1 was established based on characteristic clinical features including dermatologic findings, pathological diagnosis of neurofibroma, and familial history of NF1 in first-degree relatives. These fulfilled the diagnostic criteria of NF1, and so further genetic testing was considered unnecessary \[[@B6]\]. There was a possibility that the skin lesions were associated with the MEN1 and that the gastrinoma occurred in the context of MEN1. However, meticulous laboratory and radiologic investigations indicated that the patient was not affected by MEN1. NF1 patients have mutations of the NF1 gene on chromosome 17 \[[@B7]\]. The NF1 gene is a tumor-suppressor gene encoding a protein, neurofibromin, which functions in normal cells as a suppressor of the ras signaling cascades \[[@B8]\]. The development of malignancy in NF1 is frequently encountered, and can influence the outcome \[[@B9]\]. The prevalence of malignant tumors is reportedly four times higher in NF1 patients than in the general population \[[@B2]\]. Patients with NF1 are at an increased risk of developing central nervous system tumors, including optic pathway gliomas and brainstem and cerebellar tumors \[[@B10],[@B11]\]. Gastrointestinal involvement in NF1 occurs at diverse locations, and its clinical features encompass variable manifestations including intestinal bleeding, ischemia, angina, infarction, and bowel obstruction \[[@B12]-[@B16]\]. A previous report suggested that gastrointestinal involvement in NF1 occurs as intestinal ganglioneuromas, GISTs, and carcinoids of the periampullary region of the duodenum that may be associated with pheochromocytoma \[[@B3]\]. Several cases of gastrointestinal adenocarcinomas have been reported in literature \[[@B17],[@B18]\], but the most common gastrointestinal malignancy in NF1 is neuroendocrine tumors in the upper gastrointestinal tract, characteristically in the periampullary region \[[@B19]-[@B21]\]. Many are classified as somatostatinomas \[[@B22],[@B23]\], and NF1 has rarely been associated with insulinomas and gastrinomas \[[@B24],[@B25]\]. Our literature review revealed only one reported case of ZES associated with NF1, and hence the present case represents only the second reported case. The multiple small nodular tumors attached to the serosal surface of the small bowel observed during laparotomy in our patient were found to be stromal tumors with a low malignant potential. We assumed that these stromal tumors were an incidental finding that carried no current potential risk of developing into malignant GIST. Most patients with ZES initially present with abdominal pain, heartburn, diarrhea, and weight loss. These patients are often misdiagnosed as idiopathic peptic ulcer disease, resulting in a delayed true diagnosis that leads to a more advanced stage of the disease. The diagnosis of ZES is based on previously described criteria \[[@B26]\]. A fasting serum gastrin level of \>200 pg/mL should raise the suspicion of the disease, and a level of \>1000 pg/mL without the presence of achlorhydria secondary to atrophic gastritis is virtually indicative of ZES. Gastric secretory testing usually demonstrates the presence of a basal acid output of greater than 15 mEq in unoperated patients. The secretin stimulation test is a useful way to confirm the diagnosis of ZES (where the increase is \>200 pg/mL), when the fasting serum gastrin level is only modestly elevated and the diagnosis is in doubt \[[@B27]\]. Finally, a positive histologic confirmation of gastrinoma provides definitive evidence of the presence of ZES. In our case, an elevated level of gastrin and characteristic pathologic findings indicated the correct diagnosis of ZES. More than 80% of the tumors in ZES are located in the so-called gastrinoma triangle \[[@B28]\]. In addition, more than 90% of duodenal gastrinomas occur in the first or second portion of the duodenum, and about 50% of them are solitary \[[@B29]\]. In our case, the tumor was located in the extraduodenal space within the gastrinoma triangle. Various imaging methods are available for localizing a gastrinoma. In our patient, the tumor was detected by conventional CT and MRI scans as a isolated, well-contained mass without metastasis to the liver or lymph nodes, which enabled curative resection. Conclusion ========== Most patients with NF1 never develop major complications, but the mainstay of care remains anticipatory guidance and surveillance for treatable complications. As in our case, NF1 can be related to neuroendocrine tumors that will progress if not managed early in the course of the disease. As accurate diagnosis of gastrinoma can be difficult, a high level of suspicion and timely evaluation should be applied to NF1 patients manifesting characteristic symptoms attributable to ZES. List of abbreviations used ========================== CT: computed tomography GIST: gastrointestinal stromal tumor MEN1: multiple endocrine neoplasia type 1 MRI: magnetic resonance imaging NF1: neurofibromatosis type 1 ZES: Zollinger-Ellison syndrome Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= CHP, YEJ, HSK, SKC, JSR, and SJK made substantial contributions to the conception and interpretation of clinical data and case-related studies, and clinical decisions. YSK, JCK, and CKC performed the surgery and provided final diagnosis of the disease. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2407/5/85/prepub> Acknowledgements ================ Written consent was obtained from the patient for publication of the patient\'s details. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Abdominal CT scan showing a 3 × 2 cm, well-defined, uniformly enhanced mass adjacent to the duodenal loop. ::: ![](1471-2407-5-85-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### The pathologic findings were of a well-encapsulated solid mass containing cells with a diffuse, trabecular growth pattern. (×40, H&E). ::: ![](1471-2407-5-85-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### The tumor cells were strongly immunopositive for gastrin. (×250, avidin-biotin-peroxidase). ::: ![](1471-2407-5-85-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Biopsy of the multiple small nodules revealed the haphazard arrangement of uniform spindle cells in a collagenous stroma. (×100, H &E). ::: ![](1471-2407-5-85-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### The tumor cells were immunopositive for C-kit (CD117). (× 100, H &E). ::: ![](1471-2407-5-85-5) :::	PubMed Central	2024-06-05T03:55:59.796881	2005-7-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181810/", "journal": "BMC Cancer. 2005 Jul 21; 5:85", "authors": [ { "first": "Wan-Sik", "last": "Lee" }, { "first": "Yang-Seok", "last": "Koh" }, { "first": "Jung-Chul", "last": "Kim" }, { "first": "Chang-Hwan", "last": "Park" }, { "first": "Young-Eun", "last": "Joo" }, { "first": "Hyun-Soo", "last": "Kim" }, { "first": "Chol-Kyoon", "last": "Cho" }, { "first": "Sung-Kyu", "last": "Choi" }, { "first": "Jong-Sun", "last": "Rew" }, { "first": "Sei-Jong", "last": "Kim" } ] }
PMC1181811	Background ========== Animals use the gustatory system to provide information about food quality. For example, sweet-tasting foods may have a high caloric content and are preferred, while bitter-tasting foods often contain toxic substances, and are generally avoided. Two families of G protein-coupled receptors (GPCRs) expressed in subpopulations of taste receptor cells (TRCs) of the gustatory epithelium have been implicated in the detection and transduction of sweet, bitter and umami (i.e., glutamate) taste: T1Rs for sweet-and umami-tasting stimuli \[[@B1]-[@B8]\], and T2Rs for bitter-tasting compounds \[[@B9]-[@B11]\]. The genes that encode T2Rs, the Tas2rs, were first identified by database mining of mammalian genomes near chromosomal markers previously linked to differences in bitter taste sensitivity \[[@B9],[@B11]-[@B18]\]. In mice, the majority of Tas2rslie within a single cluster on distal chromosome 6. Thirty-three human Tas2rs(including 8 pseudogenes) and thirty-six mouse Tas2rs(including 3 pseudogenes in C57BL/6J mice) have been identified \[[@B9],[@B11],[@B19]\], and several of these respond to particular bitter stimuli in heterologous expression assays \[[@B10],[@B20]-[@B23]\], or represent a strong candidate gene for a specific bitter taste quantitative trait \[[@B18],[@B24],[@B25]\]. Several quantitative trait loci (QTL) have been identified that influence two-bottle intake of bitter stimuli in the mouse, including loci for quinine (Qui) \[[@B12],[@B16],[@B26]\], cyclohexamide (Cyx) \[[@B13]\] and sucrose octaacetate (Soa) \[[@B14],[@B15],[@B17]\] sensitivity. Each of these QTL map to mouse distal chromosome 6 and are linked to the marker D6Mit13, which lies within a cluster of 24 intact Tas2rsin the C57BL/6 genome (e.g., \[[@B16],[@B27],[@B28]\]). However, the interpretation of these studies remains problematic for two reasons. First, the density of chromosomal markers and number of recombinant inbred (RI) strains used in these earlier studies did not permit the physical definition of the intervals containing each QTL. Second, these previous attempts to map bitter taste QTLs relied on behavioral assays that measured consumption, and were thus susceptible to contributions of post-ingestive effects such as toxicity. As we have shown previously, such effects can confound the quantification of bitter taste behaviors \[[@B29]\]. Therefore, the relevance and/or contribution of the aforementioned QTLs to bitter taste remain unclear. Furthermore, a number of physiological studies have suggested that the transduction of some amphiphilic bitter compounds, such as quinine and denatonium benzoate, may stimulate taste receptor cells independently of GPCRs (e.g. \[[@B30]\]). Quinine may directly activate G proteins, and both quinine and denatonium can block K^+^channels \[[@B31]-[@B36]\] ; caffeine, another bitter-tasting substance, directly inhibits intracellular phosphodiesterase \[[@B33]\]. However, the relative contributions of T2R-dependent and T2R-independent mechanisms to the detection of these bitter stimuli are unknown. Here we use a taste-salient brief-access lick test \[[@B29],[@B37]\] to measure taste sensitivities in C57BL/6J (B6), DBA/2J (D2) and BXD/Ty (BXD) recombinant inbred (RI) mice to two bitter stimuli, quinine hydrochloride (QHCl) and denatonium benzoate (DB). Using 17 BXD lines that were genotyped at 762 informative chromosomal markers, we mapped a major QTL for QHCl taste to a \~5 Mb interval on distal chromosome 6 that contains all 24 of the Tas2rgenes in the distal cluster. We analyzed the sequence of each Tas2rallele in the parental strains (B6 and D2) and 29 RI lines. This analysis revealed that all 24 genes are polymorphic between the two strains, and that these 24 Tas2rscomprise a single haplotype that correlates with QHCl taste sensitivity. Results ======= Taste testing ------------- Previous efforts to map QTL for bitter taste have utilized consumption tests that may be confounded by the contributions of post-ingestive effects \[[@B29]\]. We used a modified brief-access lick test, which minimizes the contribution of such effects \[[@B29],[@B37]\] to determine whether B6 and D2 mice display differences in taste sensitivity to the taste stimuli QHCl and DB. After initially screening B6 and D2 mice to determine stimulus concentrations that were aversive but not saturating \[[@B47]\], we selected two ligand concentrations for each compound that best differentiated the two strains. Subsequent taste testing of BXD RI lines was restricted to these two concentrations (1 and 3 mM for both QHCl and DB). Avoidance by male and female B6 and D2 mice increased (as indicated by the decreased lick ratio) in a concentration-dependent manner for both compounds (Figure [1A](#F1){ref-type="fig"}; Table [1](#T1){ref-type="table"}). There was a significant strain difference for both 1 and 3 mM QHCl (F\[1,25\] \> 24.6; p \< 0.0001). D2 mice displayed decreased aversion relative to B6 mice at both concentrations. On the other hand, the strains did not significantly differ in taste sensitivity to DB (Figure [1A](#F1){ref-type="fig"}). There were no significant effects of gender. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Mean lick ratios for B6, D2 and BXD mice. ::: Strain n Water licks/5s 1 mM DB 3 mM DB 3 mM PR 10 mM PR 1 mM QH 3 mM QH ------------ ------- -------------------- ------------- ------------- ------------- -------------- ------------- ------------- B6 16 29.02 ± 1.4 0.606 0.318 0.311 0.164 0.241 0.144 D2 12 33.91 ± 2.4 0.465 0.286 0.616 0.429 0.758 0.305 BXD1 5 32.48 ± 1.5 0.372 0.202 0.314 0.252 0.746 0.422 BXD2 4 36.50 ± 0.9 1.025 0.605 0.208 0.230 0.320 0.215 BXD5 5 37.23 ± 2.7 0.628 0.354 0.502 0.636 0.836 0.452 BXD6 5 22.43 ± 3.5 0.442 0.132 0.200 0.124 0.130 0.080 BXD11 7 34.48 ± 3.9 0.310 0.223 0.387 0.260 0.479 0.260 BXD13 5 27.74 ± 3.4 0.360 0.290 0.472 0.168 0.692 0.330 BXD14 7 31.09 ± 3.5 0.741 0.304 0.293 0.174 0.206 0.126 BXD15 5 37.36 ± 1.5 0.178 0.120 0.468 0.198 0.362 0.222 BXD20 5 31.26 ± 2.7 0.252 0.200 0.162 0.128 0.150 0.116 BXD21 6 19.77 ± 1.5 0.390 0.262 0.313 0.133 0.192 0.217 BXD24 5 33.03 ± 2.4 0.318 0.148 0.490 0.238 0.518 0.370 BXD27 5 33.46 ± 2.5 0.364 0.434 0.216 0.126 0.146 0.142 BXD29 3 40.19 ± 1.5 0.257 0.187 0.657 0.173 0.633 0.333 BXD31 5 19.45 ± 0.8 0.302 0.238 0.152 0.158 0.218 0.194 BXD32 6 27.22 ± 2.3 0.105 0.107 0.385 0.203 0.405 0.245 BXD33 6 29.03 ± 2.2 0.757 0.387 0.377 0.237 0.342 0.188 BXD34 6 23.01 ± 3.4 0.310 0.217 0.843 0.257 0.693 0.355 The number of individual mice tested for each strain (n) is listed in the second column. Subsequent columns show the mean lick rate to water during testing (± SEM), mean lick ratio for denatonium benzoate (DB), PROP (6-n-propylthiouracil; PR) and quinine hydrochloride (QH) at each of two concentrations (see Methods for details). ::: ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Lick ratios (mean ± SE) for B6, D2 and BXD strains. (A)Mean lick ratios for B6 (filled circles) and D2 (open circles) mice at two concentrations of QHCl and DB. In all panels, a lower mean lick ratio indicates a greater aversion, and therefore greater taste sensitivity, to the stimulus. For panels B, C, and D, each BXD strain is represented by a different color, and listed in order from least sensitive to most sensitive to 1 mM QHCL. (B)Mean lick ratios for the six BXD strains that are most sensitive to QHCl in this assay. (C)Mean lick ratios for the five BXD strains that are least sensitive to QHCl in this assay. (D)Mean lick ratios for the six BXD strains intermediate in QHCl taste sensitivity to those in (B) and (C). Cutoffs for the three QHCl taster groups were arbitrarily set, as there was a continuity of the phenotype at 1 mM QHCl: sensitive strains exhibited a lick ratio for 1 mM QHCl of ≤ 0.3, intermediate strains from 0.31-0.6, and insensitive strains \> 0.6. The absence of two distinct phenotypic classes suggests that sensitivity to QHCL is under polygenic control. ::: ![](1471-2156-6-32-1) ::: We next tested mice from 17 BXD lines in the same manner. BXD mice also typically avoided both stimuli in a concentration dependent manner (Figures [1B--1D](#F1){ref-type="fig"}; Table [1](#T1){ref-type="table"}). However, QHCl and DB taste sensitivity vary independently across these RI strains: some strains highly sensitive to QHCl are relatively insensitive to DB, and vice versa (Figures [1B--1D](#F1){ref-type="fig"}). QTL mapping ----------- Linkage analysis was conducted using Map Manager QTX (version 0.30\[[@B38]\]). No significant QTLs were identified for DB taste sensitivity, although several associations with markers on chromosomes 2,8 and 12 were \"suggestive\" (LRS \> 9.4, genome-wide p = 0.65; see Additional File [1](#S1){ref-type="supplementary-material"}). A significant (LRS \> 20.5; genome-wide p = 0.05) QTL for sensitivity to 1 mM QHCl was indicated on chromosome 6, with a second, suggestive (LRS \> 11.4; genome-wide p = 0.65) QTL on chromosome 8 (Figure [2A](#F2){ref-type="fig"}); at 3 mM QHCl, both of these QTL were suggestive (LRS \> 10.9) but did not reach genome-wide significance (Figure [2B](#F2){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### A major QTL for QHCl taste on mouse chromosome 6. (Top panel)The interval map (see Methods) shows a significant QTL on chromosome 6 (green) and a suggestive QTL on chromosome 8 (yellow) affecting taste responses to 1 mM QHCl. (Bottom panel)For 3 mM QHCl, both QTL were suggestive (yellow). The dashed line indicates genome-wide significance. ::: ![](1471-2156-6-32-2) ::: The chromosome 6 QTL was linked to a single marker, D6Mit13 (Table [2](#T2){ref-type="table"}, Figure [3](#F3){ref-type="fig"}). Adjacent proximal markers D6Mit254 and D6Mit194 are unlinked to the QHCl QTL, as is distal marker D6Mit374. Across the 17 RI lines tested there is at least one recombination event between D6Mit13 and either D6Mit254 (and D6Mit194, the physical position of which is not well defined) or D6Mit374. An additional proximal marker, D6Mit61, which lies between D6Mit194 and D6Mit13, was identified from genotypes of the BXD lines reported by the Jackson Laboratories. BXD/Ty-34 RI mice display a clear D2 phenotype for QHCl taste (Figure [1C](#F1){ref-type="fig"}) and D2 genotype for D6Mit13, but have a B6 genotype for D6Mit61 \[[@B39],[@B40]\], indicating that D6Mit61 is unlinked to the QHCl QTL. Therefore, this QTL interval can be conservatively defined as that portion of mouse chromosome 6 that lies between D6Mit254 and D6Mit374, but is most likely restricted to the region between D6Mit61 and D6Mit374. Physical mapping of the single linked marker, D6Mit13, and the two closest unlinked markers, D6Mit61and D6Mit374, was performed in silicobased on the May, 2004 build of the public B6 genome. Based on these marker positions, the size of the QHCl chromosome 6 QTL is less than 5.0 Mb (Figure [3](#F3){ref-type="fig"}). This region contains a number of known genes, all but eleven of which encode members of two large receptor families: natural killer cell lectin-like receptors, and T2R-type taste receptors. The Tas2rgenes (which encode the T2Rs) are found clustered within a 1.2 Mb interval on either side of D6Mit13 (Figure [3](#F3){ref-type="fig"}, Figure [4](#F4){ref-type="fig"}). Because of their proximity to the linked marker, their demonstrated expression in taste receptor cells, and their role in the detection of at least some bitter-tasting compounds, we hypothesized that one or more of the 24 Tas2rsat this locus were responsible for the major QHCl taste sensitivity QTL. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### The QHCl QTL is linked to a single marker on chromosome 6. (A)As shown in the interval map for chromosome 6, the trait value (lick ratio for 1 mM QHCl) correlates strongly across BXD RI strains with the polymorphic marker D6Mit13 (bold). The dashed line indicates genome-wide significance. (B) The QHCl QTL (which lies between unlinked markers D6Mit61 and D6Mit 374) contains a cluster of putative bitter taste receptor genes, the Tas2rs(gray box). Physical positions of the polymorphic markers are given in Mb, and are based on the May, 2004 build of the B6 mouse genome. The physical position of D6Mit194 (\) is tentative. ::: ![](1471-2156-6-32-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### A map of the distal chromosome 6 Tas2rcluster. Twenty-four intact Tas2rgenes map to distal chromosome 6 (black). The Tas2rs are found in two subclusters on either side of the polymorphic marker D6Mit13(red) and two genes encoding proline-rich salivary proteins (Prp2and Prh1; red). Map positions, in Mb, represent chromosome 6 positions in the May, 2004 assembly of the B6 genome. ::: ![](1471-2156-6-32-4) ::: T2R alleles ----------- If one (or more) Tas2rsunderlie the chromosome 6 QHCl taste sensitivity QTL, we would predict that one (or more) Tas2rgenes would exhibit one of three likely characteristics: (1) A Tas2rallele is a pseudogene, or is deleted, in D2 (QHCL-insensitive), but not B6 (QHCl-sensitive), mice; (2) Missense mutations in the single coding exon of a D2 Tas2rallele impact protein functions such as ligand binding or receptor coupling to downstream signaling cascades; (3) Mutations in noncoding or regulatory regions of a D2 Tas2rallele affects expression of the protein product. Though we considered all three of these to be valid possibilities, we initially focused on the likelihood that deletion or mutation within the coding sequence of a single D2 Tas2rwould correlate with the QHCl taste insensitivity phenotype. Twenty-four intact Tas2rs, along with three apparent Tas2rpseudogenes, have been identified in the distal chromosome 6 cluster of B6 mice \[[@B19]\] (Figure [4](#F4){ref-type="fig"}). We designed oligonucleotides to non-coding regions flanking the coding sequence of each intact Tas2r\[see [Additional file 2](#S2){ref-type="supplementary-material"}\]. Using these oligos, we amplified each Tas2rcoding sequence from D2 genomic DNA. PCR products were subcloned into cloning vectors and sequenced. Comparisons of the sequences of B6 and D2 orthologues revealed that only two of the twenty-four Tas2ralleles examined, Tas2r106and Tas2r124, were identical across strains at the amino acid level (data not shown). A third, Tas2r120, could not be amplified from D2 genomic DNA (Figure [5](#F5){ref-type="fig"}) using either of two pairs of oligonucleotides (Additional file [2](#S2){ref-type="supplementary-material"}), suggesting that this Tas2ris deleted in D2 mice. Two D2 alleles, Tas2r103and Tas2r117, contained numerous missense mutations and small deletions that create frame shifts and premature termination; these two genes may be pseudogenes in this strain. The remaining 19 Tas2rs contained between one and 16 missense mutations. All 24 Tas2rsexamined have different alleles in B6 and D2 mice, and 307 single nucleotide polymorphisms are present within coding exons (data not shown). Although polymorphic residues between B6 and D2 Tas2rsare found in all regions of the receptors, 23% of the amino acid changes seen are within the first two extracellular loops of the T2Rs (data not shown). ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Allelic variation across strains for four Tas2rs. B6 and D2 alleles of four Tas2rscan be differentiated based on diagnostic restriction digests of amplified PCR products (Tas2r105, Tas2r116and Tas2r131) or on the presence or absence of a PCR product (Tas2r120). In each of the 17 BXD strains tested, Tas2r* genotype was always correlated with QHCl taster phenotype (blue = B6 taster phenotype, red = D2 taster phenotype). See [additional file 1](#S1){ref-type="supplementary-material"}: Table 3 for restriction enzymes and oligonucleotides. ::: ![](1471-2156-6-32-5) ::: The variability between orthologous receptors in these two inbred strains suggested that it might be possible to narrow the physical boundaries of the QHCl taste QTL by determining which Tas2ralleles are correlated with QHCl taste sensitivity. Therefore, we proceeded to screen genomic DNA from 29 available BXD RI lines, including the 17 that we had used in taste testing, for the Tas2ralleles they contained. In most cases, we were able to identify diagnostic restriction endonuclease digests that would allow us to quickly identify whether a particular Tas2rPCR product was amplified from a B6 or D2 allele. We did not analyze three genes (Tas2r104, Tas2r114and Tas2r110) where no diagnostic restriction endonuclease could be identified. For Tas2r120, which is likely deleted in D2 mice, the absence of a PCR product was diagnostic of the D2 genotype for this gene. Surprisingly, we discovered that there have been no apparent recombination events within the distal chromosome 6 cluster during the generation of the BXD RI lines. For all RI lines tested, every Tas2rwithin an individual RI line originated from the same parental strain (Figures [5](#F5){ref-type="fig"}, [6](#F6){ref-type="fig"}). Furthermore, the genotype of each Tas2rgene always correlated with the QHCl taste phenotype (Figures [6](#F6){ref-type="fig"}, [7](#F7){ref-type="fig"}), suggesting that the entire Tas2rcluster is a single haplotype that varies with QHCl taster status. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### The Tas2r cluster is a single haplotype in BXD/Ty RI mice. The coding exon of each of 21 Tas2rsin the distal chromosome 6 cluster was amplified genomic DNA from 29 BXD/Ty RI strains. Each Tas2rwithin an individual BXD strain originated from the same parental strain (B6 allele = gray, D2 allele = white). The 17 BXD strains that were behaviorally tested in this study are indicated (\). ::: ![](1471-2156-6-32-6) ::: ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Tas2rgenotype correlates with QHCl taste phenotype. Mean lick ratios of B6, D2 and BXD strains reported in Figure 1 are grouped based on Tas2rhaplotype (B6 haplotype = blue, D2 haplotype = red). B6 mice (blue line on left panel) are more sensitive to 1 mM and 3 mM QHCl than are D2 mice (red line on left panel) in brief access taste tests. Similarly, BXD strains with the B6 Tas2rhaplotype (blue lines, right panel) are more sensitive to QHCl than are BXD strains with the D2 Tas2rhaplotype (red lines, right panel). The BXD strains are listed in order from least to most sensitive to 1 mM QHCL. ::: ![](1471-2156-6-32-7) ::: Discussion ========== The gustatory system of mammals is thought to detect thousands of chemically-diverse bitter-tasting substances \[[@B41]\]. Although specific receptors, enzymes and channels have been implicated in the transduction of bitter stimuli, how interactions of bitter stimuli with taste receptor cells lead to cellular activation and signaling to the central nervous system is still poorly understood. We have found that a single QTL on distal chromosome 6 accounts for most of the variation in QHCL taste sensitivity between B6 and D2 mice. This QTL maps to the same chromosomal position as a previously identified QTL for quinine intake, Qui\[[@B16],[@B28]\], indicating that taste is the major factor in regulating quinine aversion. This is an important distinction, as the consumption of bitter-tasting stimuli can be dependent on factors independent of taste, such as toxicity \[[@B29]\]. Using 17 RI lines and 762 chromosomal markers, we have restricted the quinine taste QTL to a \< 5 Mb region on distal chromosome 6 that contains 24 Tas2rgenes. At least 60 other genes also lie within this interval, including two genes that encode proline-rich salivary proteins, Prp2and Prh1; these proteins appear to play no direct role in bitter taste \[[@B48]\]. Tas2rsare the most likely candidates for the QHCl quantitative trait gene(s) due to: (1) their expression in taste receptor cells and (2) genetic and functional evidence linking them to the detection of a number of bitter taste stimuli. As of yet there is no evidence for quinine activation of T2Rs from functional assays of these receptors in heterologous cells or membrane preparations, likely due to the lipophilic nature of quinine \[[@B23]\]. However, several physiological studies have suggested that quinine can directly activate G proteins or cationic conductances, or can block K^+^channels in taste receptor cells \[[@B34]-[@B36]\]. While our data indicates that quinine taste is largely T2R-dependent, it is not exclusively so. For example, the BXD RI lines exhibited a range of quinine sensitivity, with several strains having similar sensitivities to that of B6, some strains with sensitivities similar to that of D2, and a third group with a more intermediate phenotype (Figures [1](#F1){ref-type="fig"}, [7](#F7){ref-type="fig"}). This observation is consistent with a polygenic basis for quinine taste \[[@B16],[@B26]\]. Also, a suggestive QTL on chromosome 8 does not contain any Tas2rgenes, but does contain a number of genes encoding ion channels, enzymes and members of other receptor families (our unpublished data). It will be interesting to determine whether this suggestive QTL is linked to quinine taste and, if so, whether it is specific for this single bitter stimulus or more broadly related to all bitter taste. Of the 29 BXD RI lines examined, there was no apparent recombination event within the chromosome 6 Tas2rcluster. While increasing the number of BXD RI lines or the number of markers used for genotyping them would facilitate the definition of smaller QTL intervals, in this case such an effort is unlikely to permit the identification of one or a few Tas2rsinvolved in quinine taste. For example, we examined six lines of AXB and BXA RIs with reported recombinations around D6Mit13; a small sampling of the Tas2rsin these RI lines again indicated no recombinations within the Tas2rcluster (data not shown). Behavioral genetic approaches have been invaluable for identifying genes involved in taste function, such as the Tas1r3gene that encodes a receptor important for sweet and umami taste \[[@B42]\]. Positional cloning also permitted the identification of the Tas2rresponsible for the majority of variance of phenylthiocarbamide (PTC) taste sensitivity in humans \[[@B18]\]. In both of these cases, however, the genes linked to saccharin or PTC taste were not tightly clustered with paralogues. For bitter taste, behavioral genetic approaches may be more useful for identifying genes encoding downstream signaling molecules or components of T2R-independent transduction mechanisms. For example, a QTL for PROP avoidance has been suggested on chromosome 7 \[[@B16]\], and we observe a suggestive QTL for quinine taste on chromosome 8 (Figure [2](#F2){ref-type="fig"}); in neither case are Tas2rsfound at these loci (data not shown). It is somewhat puzzling that 22 of the 24 Tas2rsexamined encode variant proteins in B6 and D2 mice even though these strains exhibit similar taste responses to bitter compounds such as DB or cyclohexamide \[[@B47]\]. Taken together, these observations suggest that Tas2rsare quite tolerant of variation, and that perhaps most of the differences observed do not affect domains important for ligand interactions or receptor-mediated signaling mechanisms. Interestingly, 23% of missense mutations in D2 Tas2rsaffect the first two extracellular loops of the receptors. These two loops have been recently shown to affect the ligand response profiles of some T2Rs \[[@B23]\]. More systematic analyses of structure-function relationships between these T2R variants and an array of bitter stimuli are necessary to determine which changes may impact ligand binding, interactions with other proteins, or overall receptor structure. Such large numbers of nonsynonymous substitutions between orthologues is suggestive of adaptive selection. Analysis of sequence diversity of Tas2rsin humans, great apes and old world monkeys suggest that Tas2rsare subject to some degree of positive selection \[[@B43],[@B44]\]. However, the fact that these two mouse strains, members of the same species, are so closely related makes this explanation problematic. It is possible that B6 and D2 mice, which have a similar origin in the early 20^th^century, inherited different Tas2rhaplotypes present in wild mouse populations prior to inbreeding. Characterization of Tas2rsequences of several wild mouse species or subspecies, or in other inbred lines, would shed light on this issue. Conclusion ========== In conclusion, we have found that sensitivity to the bitter-tasting substance quinine, as assayed by a taste specific brief-access test, is a polygenic trait in mice. However, the major mechanism for quinine taste transduction is likely dependent on one or more T2R receptors. Most Tas2rgenes in the distal chromosome 6 cluster are polymorphic across inbred strains of mice, and this cluster forms a single haplotype that correlates with quinine taste sensitivity. The numerous differences in T2R protein sequence between these two mouse strains suggests that T2Rs are broadly tuned receptors quite tolerant to sequence variation. This tolerance may help to preserve the ability of T2R-expressing, bitter-sensitive taste cells to respond to a wide array of potentially toxic stimuli. Methods ======= Mice and solutions ------------------ A total of 188 adult male and female mice were behaviorally tested in these experiments: 16 C57BL/6J (B6; 9 males, 7 females), 12 DBA/2J (6 females, 6 males), and 90 BXD/Ty recombinant inbred mice (average = 5 / line; 64 males, 26 females) from 17 unique lines (1, 2, 5, 6, 11, 13, 14, 15, 20, 21, 24, 27, 29, 31, 32, 33, 34). All mice were either obtained from Jackson Laboratories (Bar Harbor, ME), or were bred from mating pairs at UTHSC. At time of testing, mice were individually housed in standard shoebox cages with woodchip bedding and ad libitum food (Teklad 8640 rodent diet). Taste stimuli used in this experiment were made from reagent-grade chemicals: Sucrose, denatonium benzoate, 6-n-propylthiouracil and quinine hydrochloride (Sigma Aldrich Corp.; St. Louis, MO). Concentrations of each solution were made fresh daily using distilled water, and all taste stimuli were presented at room temperature. All animal protocols were approved by the UTHSC Institutional Animal care and Use Committee. Brief-access tests ------------------ All behavioral tests were conducted in the commercially available Davis MS-160 gustometer (DiLog Instruments, Inc., Tallahassee FL). Testing procedures were similar to those described earlier \[[@B29],[@B37]\]. Briefly, after 24 hours of water deprivation, naïve mice are given a single 20-minute trial consisting of access to a single bottle of distilled water (sipper tube training). On day 2, mice could initiate up to sixteen 5 s trials with a single lick to one of four bottles containing distilled water (trial training). Testing occurred in sessions 3 and 4, with one test session per day per mouse. Trials were 5 s in length with an inter-trial interval of 10 s, and mice had up to 120 s to initiate a trial; if a trial was not initiated during this interval, the shutter closed for 10 s and the next trial was presented. Mice were tested with 2 concentrations each of 4 different taste stimuli \[1 and 3 mM QHCl, 1 and 3 mM DB, 3 and 10 mM PROP (unpublished data), and 0.01 and 0.1 M sucrose\]. Stimulus trials were presented in 3 blocks of 8 trials, for a total of 24 possible trials per test session. Each block consisted of each concentration of stimulus plus four presentations of distilled water in random order. Individual mice were also tested in random order. The dependent measure for each computed for each mouse was the lick ratio(average number of licks to stimulus~x~/average number of licks to water) where xis a given concentration of stimulus and the average number of licks to water is derived from the water trials during both test sessions. Lick ratio data for each stimulus were compiled for all individual mice, and means were prepared for each strain. B6 vs. D2 comparisons (Fig. [1A](#F1){ref-type="fig"}) were made using main effects ANOVA. Lick ratios for individual mice to sucrose were generally \~1.0 (data not shown), indicating that either concentration of this compound was licked at a similar rate to water by these thirsty animals. This stimulus was intended as a \"neutral\" stimulus, albeit one that has different sensory properties than water, and therefore not analyzed further. This was done to encourage sampling on \"non-water\" trials, as there is some evidence that mice detect distilled vs. adulterated water in brief-access taste tests based on olfactory clues; there is no evidence that mice can detect or distinguish among concentrations of a particular stimulus \[[@B37]\]. QTL mapping ----------- Linkage analysis was conducted on BXD mice using freely available software (Map Manager QTX \[[@B38]\]), and BXD genotype data shared by Robert W. Williams, University of Tennessee Health Science Center \[[@B45]\]. Simple interval mapping was conducted. This method evaluates the association between trait values (lick ratios) and expected genotype of a hypothetical quantitative trait locus (QTL) at multiple analysis points between each pair of adjacent marker loci. The significance of each potential association is measured by the likelihood ratio statistic (LRS; e.g. \[[@B46]\]). Permutation analysis (x2000) was used to determine genome-wide significance criteria for LRS scores. Significance was set at p \< 0.05 and suggestive refers to p \< 0.63. Additional markers used to refine the QTL on chromosome 6 were identified from the Jackson Laboratories online resources for the BXD RI strains \[[@B40]\]. Identification of T2R alleles ----------------------------- Oligonucleotides were based on published mTas2rB6 or 129/SvJ cDNA sequences or on the public B6 genome. Entire coding regions plus \~50 kb of flanking sequence of each single-exon Tas2rwas amplified from D2 or BXD RI genomic DNA (Jackson Laboratories, Bar Harbor, ME) by polymerase chain reaction (PCR) using a high-fidelity polymerase TaqPro Complete (Denville Scientific, South Plainfield, NJ). PCR products were subcloned into pGemT-Easy (Promega, Madison, WI) and sequenced at the University of Maryland School of Medicine Biopolymer Core. The sequences of D2 products were compared to B6 sequences available in Genbank (see Additional file [2](#S2){ref-type="supplementary-material"}), and polymorphisms identified. When possible, unique restriction sites were identified that differentiated B6 and D2 alleles, and the corresponding restriction endonucleases were used in diagnostic digests of Tas2rcDNAs amplified from genomic DNA of each BXD/Ty RI strain. For Tas2r120, the absence of a PCR product was considered diagnostic of the D2 allele. Authors\' Contributions ======================= TN conducted the in silicoanalysis of the quinine taste QTL, analyzed the Tas2rgenes, participated in the design of the study and drafted the manuscript. JB conducted the behavioral studies and QTL analysis. SM assisted with the in silicoanalysis of the quinine taste QTL and the Tas2rs, and with the QTL analysis. JB and SM conceived of the study, participated in its design, and edited the manuscript. All authors read and approved the final manuscript. Comments and requests should be addressed to JB or SM. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Linkage of a QHCl QTL to D6Mit13 on chromosome 6. ::: 1 mM qhcl* 3 mM qhcl --------- --------------- ------------------- --------------- --------------- ------------ --------- --------- ----------- --------- --------- Chr Marker Position (Mb) LRS % Var *p* Add LRS % Var *p* Add 6 D6Mit150 116.543 12.1 51 0.0005 0.16 11.9 50 0.00057 0.08 6 D6Mit254 125.974 11.4 49 0.00075 0.16 11.1 48 0.00087 0.07 6 D6Mit194 126.895 11.4 49 0.00075 0.16 11.1 48 0.00087 0.07 6 D6Mit61 129.173 nd nd nd nd nd nd nd nd 6 D6Mit13 132.672 21.0 71 0.000001\* 0.19 17.9 65 0.00002 0.09 6 D6Mit374 134.172 14.0 56 0.00018 0.17 14.3 57 0.00015 0.08 6 D6Mit301 136.104 14.0 56 0.00018 0.17 14.3 57 0.00015 0.08 6 SO6Gnf140.060 140.060 11.9 50 0.00056 0.17 12.7 53 0.00037 0.08 8 SO8Gnf046.785 46.785 13.4 55 0.00025 0.17 12.2 51 0.00047 0.08 Markers are listed from proximal (D6Mit150) to distal (SO8Gnf046.785), with physical position indicated in Mb. The physical position of D6Mit194 (italics) should be considered tentative. The LRS (likelihood ratio statistic) is listed for each locus, signifying the level of association of the trait (QHCl taste sensitivity) with each locus. Variance (Var) refers to the amount of the total trait variance explained by a QTL at this locus, as a percentage. Additive regression coefficients (Add) are listed for each association; in each case the coefficient is positive, indicating that D2 alleles increase the trait value (i.e. higher lick ratios). In simple marker regression analysis, all of these loci are associated with QHCl sensitivity at p\< 0.001; only the association of sensitivity to 1 mM QHCl with D6Mit13 reaches genome-wide significance (asterisk). ::: Supplementary Material ====================== ::: {.caption} ###### Additional File 2 Table 4: Molecular biological methods for the analysis of Tas2rs. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 1 Table 3: Marker regression results for DB taste sensitivity. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ The authors thank Sandeep Raghow and Aldan Shank for technical assistance with the behavioral studies and Robert Lane for useful discussions. This study was supported by grants from the National Institute on Deafness and Other Communication Disorders (DC05786 to SM, DC04935 to JB), a Program Enrichment Fellowship from the University of Maryland, Baltimore (TN) and by the University of Tennessee Health Sciences Center (JB).	PubMed Central	2024-06-05T03:55:59.798026	2005-6-6	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181811/", "journal": "BMC Genet. 2005 Jun 6; 6:32", "authors": [ { "first": "Theodore M", "last": "Nelson" }, { "first": "Steven D", "last": "Munger" }, { "first": "John D", "last": "Boughter" } ] }
PMC1181812	Background ========== The modern day cultivars of rice, in spite of all their high yielding potential and other desirable features are handicapped with narrow genetic base for most of the agronomically important traits including the dwarf habit, which is the major yield enhancing trait. Recent study of high yielding Indian rice varieties for their ancestry revealed that hardly 5 to 6 accessions accounted for more than 90% of their genetic constitution, confirming that the cultivar gene pool being depended on now for improvement represent hardly 15% of the total genetic variability available in rice germplasm (E A Siddiq, personal communication). Rice is endowed with very rich genetic diversity. Wild/weedy species along with very large number of primitive cultivars and landraces constitute an important reservoir of useful genes. The size of additional variability they can provide would be of great value to the ongoing crop improvement endeavor. Large genetic variability still remains untapped in the wild relatives and primitive cultivars of rice \[[@B1]\]. Considering the large hidden variability and very rare and agronomically important genes they possibly possess, utilization of the wild species is critical to future crop improvement \[[@B2]\]. Utilization of these exotic species as donors in interspecific crosses is one of the strategies to harness their hidden potential and broaden the genetic diversity of the existing gene pool. Over the last decade, wild species in rice have been successfully utilized for introgression of diverse traits such cytoplasmic male sterility (cms) \[[@B3]-[@B6]\], abiotic and biotic stress \[[@B7]-[@B11]\], yield and its components \[[@B12]-[@B18]\] and grain quality \[[@B19]-[@B21]\] into the cultivars. A great deal of work in the recent past, on the wild species of rice, concentrated on the utilization of these species for quantitative traits such as yield and its components long with grain quality. In the first ever report on the use of wild species for introgression of quantitative characters, two yield QTLs, yld1.1 and yld 2.1, each of which is capable of increasing yield by about 18% have been identified in a Malaysian accession of O. rufipogon\[[@B12],[@B13]\]. This was a precursor to many studies resulting in the identification of numerous QTLs pertaining to yield and grain quality \[[@B12]-[@B21]\]. Keeping in view the unlimited potential of wild/weedy species of rice for yield genes as evident from the foregoing research, the present study reports the identification and mapping of molecular marker-associated yield QTLs in an Indian accession of O. rufipogon(IC 22015). An interspecific testcross population, derived using an advanced backcross QTL strategy (AB-QTL) \[[@B22]\], between O. rufipogonand IR 58025A, a widely used cms line in India, was used to map QTLs related to yield and it\'s components. The AB-QTL method has been successfully employed earlier in tomato and rice to transfer positive alleles from phenotypically inferior wild and weedy species into elite cultivars \[[@B23]-[@B25],[@B14]-[@B19]\]. In addition to identifying potential novel QTLs for yield and it\'s components, the results from the current study will provide additional data for comparison with QTLs that are previously documented in rice. Comparisons across different genetic backgrounds will provide information about the conservation of QTLs and help us to understand the interactions of QTL alleles across multiple backgrounds and environments. Results ======= Trait analysis and field performance ------------------------------------ The phenotypic analysis of the 251 testcross families showed that the frequency Distribution of all traits approximately fit normal distribution (Figure [1](#F1){ref-type="fig"}). As expected in an interpsecific cross, character wise frequency distribution of testcross families showed transgressive segregants for all the traits. For a depiction of variation in tiller number and panicle length in the testcross families, see [additional file 1](#S1){ref-type="supplementary-material"}. The average grain yield of the testcross families was 6.08 t/ha, with the range varying from 3.90 to 9.45 t/ha, while yield per plant ranged from 7.5 to 36.0 g with an average of 19.5 g. Thirteen testcross families outperformed the hybrid check, KRH2, by more than 20% for plot yield and as many as 39 families showed more than 20% increase in yield per plant as compared to KRH2 (Table [1](#T1){ref-type="table"}). Of the 251 testcross families studied in all, 75 showed at least 20% increase over KRH2 for three or more yield components. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Mean phenotypic traits for 13 yield components across 251 testcross families as compared to IR 58025A, IC22015 (wild) and KRH2 ::: Trait IR 58025A^a^ IC 22015^b^ KRH2^c^ Range in Testcross families No. of families showing \>20% increase over KRH2 ------------------------- ------------------ ----------------- ------------- --------------------------------- ------------------------------------------------------ Plant height (cm) 80 119 118 93 -- 177 26 Number of tillers 9 32 11.2 7 -- 16 18 Number of panicles 7 28 10 6 -- 14 20 Panicle length (cm) 24 29 23.5 20.5 -- 34.5 1 Spikelet number/panicle 175 150 167 67 -- 265 88 Spikelet number/plant 1350 3500 1880 737 -- 3074 74 Grain number/panicle 152\* 35 117 30 -- 185 101 Grain number/plant 1064\* 700 1187 322 -- 2310 63 Spikelet fertility (%) 0 15--20 68 42 -- 91 42 1000 grain weight (g) 20 11.5 22.5 17.5 -- 31.3 1 Yield / plant (g) 16\* 9 19 7.5 -- 36 39 Harvest index 32\* \- 45 27 -- 56 7 Yield (t/ha) 4.2 \- 7 3.9 -- 9.45 13 cm: Centimeters, g: Grams, t/ha: Tonnes / hectare a -- O. sativaparent, b -- O. rufipogonparent, c- hybrid check \* for the traits where IR 58025A has no corresponding value, the value from IR 58025B, an isogenic line of IR 58025A is used. ::: ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Frequency distribution of the 251 testcross families for yield and its components. Arrow indicates the value of the hybrid check, KRH2. Y-axis: Number of individuals. ::: ![](1471-2156-6-33-1) ::: Trait correlations ------------------ The trait correlations confirmed to the expected results. Significant positive correlations (P \< 0.01) included SPY (Single plant yield) × SNP (Spikelet number per plant) (0.552), SPY × GNP (Grain number per plant) (0.581) and HI (Harvest index) × SPY (0.539) whereas, the significant negatively correlated traits included PH (Plant height) × SF (Spikelet fertility) (-0.362), GW (Grain weight) × NT (Number of tillers) (-0.255), GW × NP (Number of panicles) (-0.284) and HI × PH (-0.298). Interestingly GW had no significant effect on PY (Plot yield), but showed negative correlation with NP. For detailed character pair correlations among the traits see [additional file 2](#S2){ref-type="supplementary-material"}. Marker polymorphism ------------------- Two hundred and ten microsatellite markers were used to screen the parents for identifying polymorphic markers. Eighty markers (38%) detected polymorphism. The polymorphism is lower compared to earlier studies involving O. rufipogon, where the polymorphism ranged from 60--90% \[[@B13],[@B15],[@B17]\]. Polymorphism is a measure of genetic diversity and varies with the parental combinations used. Earlier studies using a Malaysian accession of O. rufipogon(IRGC 105491) have indicated varying frequencies of SSR polymorphism with indica(\~60%) \[[@B13],[@B17]\] and japonica(90%) \[[@B15]\] recurrent parents. The lower percentage polymorphism may be due to a higher degree of genetic similarity between O. rufipogonand O. sativaused in this study compared to those used earlier. Marker segregation ------------------ The expected genotypic ratio in the BC~2~ population would be 3:1 for homozygous IR 58025A : heterozygous IR 58025A/O. rufipogon(87.5% IR 58025A alleles to 12.5 O. rufipogonalleles). Out of the 80 marker loci, 28.75% (23 markers) were skewed towards one or the other parent resulting in an allele frequency of 83.26% IR 58025A alleles to 16.74% O. rufipogonalleles (Table [2](#T2){ref-type="table"}). While 12.5% (10 markers) were skewed towards O. sativaparent, 16.25% (13 markers) were skewed (X~2~\> 6.6, p \< 0.01) towards O. rufipogon.. The skewed markers were distributed on chromosomes 1, 2, 3, 5 and 8 with most of the markers on chromosome 2. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Chi square values of the markers showing segregation distortion in the test cross progeny ::: Marker Chromosome X^2^ Chr. Position^a^ Skewness^b^ ------------ ---------------- ----------- ---------------------- ----------------- RM84 1 24.88\\ 0.0 IR 58025A RM243 1 8.28\* 98.1 IC 22015 RM212 1 19.47\\ 209.8 IR 58025A RM14 1 11.92\\ 268.4 IR 58025A RM262 2 42.54\\ 98.6 IC 22015 RM207 2 10.64\* 227.5 IR 58025A RM211 2 47.40\\ 0.0 IC 22015 RM53 2 19.47\\ 23.2 IC 22015 RM8 2 28.83\\ 42.3 IC 22015 RM240 2 24.88\\ 182.3 IC 22015 RM183 2 17.81\\ 102.4 IC 22015 RM55 3 8.28\* 193.7 IC 22015 RM60 3 23.00\\ 0.0 IC 22015 RM7 3 17.81\\ 91.6 IR 58025A RM251 3 30.95\\ 108.6 IC 22015 RM203 3 11.92\\ 173.1 IR 58025A RM22 3 14.72\\ 27.5 IR 58025A RM13 5 14.23\\ 24.8 IC 22015 RM169 5 28.85\\ 62.4 IC 22015 RM249 5 21.19\\ 71.1 IR 58025A RM164 5 8.28\* 87.0 IR 58025A RM230 8 35.37\\ 144.2 IC 22015 RM44 8 10.64\* 47.4 IR 58025A ^a^Location of the marker on the chromosome in centiMorgans ^b^Skewed marker segregation towards the O. sativa(IR 58025A) or O. rufipogon(IC 22015) parent \Significant at p \< 0.01 \\Significant at p \< 0.001 ::: QTL analysis ------------ A total of 39 QTLs were identified using composite interval mapping (CIM) and interval mapping (IM). CIM analysis detected fewer QTLs (25 QTLs) than IM (31 QTLs). While 17 QTLs (43.58%) were detected by both the methods, IM identified 14 QTLs (35.89%) exclusively and 8 QTLs (20.51%) were only detected by CIM (Table [3](#T3){ref-type="table"}). Single marker analysis identified a total of 45 QTLs for the 13 traits studied \[See [additional file 3](#S3){ref-type="supplementary-material"}\]. Forty two out of the 45 QTLs identified by single marker analysis were either identified by CIM or IM, so these will not be discussed separately. Three QTLs, sf1.1, spp1.1and hi1.1were only identified by SMA. The variation in the number of QTLs detected by different methods has been previously reported for interspecific crosses involving O. rufipogon\[[@B15],[@B17]\]. The 39 QTLs were distributed on chromosomes 1, 2, 3, 5, 8 and 9 (Figure [2](#F2){ref-type="fig"}). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### QTLs related to yield and yield components detected in an IR58025A × O. rufipogon(IC 22015) population ::: Trait* Chromosome Marker Interval CIM IM ---------------------------------- ---------------- --------------------- ----------- --------- -------- --------- ------- ------- --------- Plant Height ph1.1 1 RM220 -- RM272 IC 22015 5.32 17.48 -17.33 5.46 21.12 -19.09 ph1.2 1 RM272 -- RM259 IC 22015 4.2 6.82 -2.45 3.82 5.87 -2.03 ph9.1 9 RM257 -- RM242 IR 58025A 3.7 5.63 9.53 4.21 7.408 10.83 Number of tillers per plant nt2.1 2 RM262 -- RM183 IC 22015 2.99 11.11 -1.28 nt5.1 5 RM169 -- RM249 IC 22015 2.53 5.9 -1.06 Number of panicles per plant np2.1 2 RM262 -- RM183 IC 22015 2.5 6.8 -1.023 np2.2 2 RM324 -- RM262 IC 22015 3.23 10.81 -1.28 Panicle length pl2.1 2 RM250 -- RM208 IC 22015 10.9 19.28 -9.53 10.76 20.85 -9.16 pl5.1 5 RM249 -- RM164 IC 22015 6.61 18.93 -9.04 6.66 20.85 -9.16 pl9.1 9 RM242 -- RM205 IC 22015 8.11 17.3 -9.24 9.48 20.85 -9.16 Spikelet number per panicle sn2.1 2 RM250 -- RM208 IR 58025A 4.20 19.13 103.53 4.57 19.71 106.56 Spikelet number per plant snp2.1 2 RM262 -- RM183 IC 22015 3.13 12.27 -321.31 snp5.1 5 RM194 -- RM249 IC 22015 3.98 11.79 -413.25 snp5.2 5 RM169 -- RM249 IC 22015 2.52 10.29 -386.74 snp8.1 8 RM44 -- RM350 IR 58025A 3.11 6 383.53 snp8.2 8 RM44 -- RM223 IR 58025A \- \- \- 2.83 6.1 381.67 Grain number per panicle gn2.1 2 RM250 -- RM208 IR 58025A 3.32 16.65 72.52 gn5.1 5 RM194 -- RM169 IC 22015 2.98 6.12 -5.67 3.45 7.65 -6.82 Grain number per plant gnp2.1 2 RM262 -- RM183 IC 22015 2.79 5.42 -171.64 3.68 12.5 -256.41 gnp2.2 2 RM183 -- RM263 IC 22015 3.31 6.5 -186.41 gnp3.1 3 RM16 -- RM203 IR 58025A 3.14 9.8 248.31 2.71 9.98 241.4 gnp5.1 5 RM194 -- RM249 IC 22015 5.73 12 -285.18 3.35 6.5 -193.2 Spikelet fertility sf1.1 1 RM212 -- RM315 IR 58025A 3.26 5.2 2.45 sf3.1 3 RM251 -- RM36 IR 58025A 4.35 6.7 1.45 3.72 5.78 1.27 Grain weight gw2.1 2 RM250 -- RM208 IC 22015 3.25 10.4 -5.21 gw2.2 2 RM324 -- RM262 IR 58025A 2.91 7.16 82.41 gw2.3 2 RM262 -- RM183 IR 58025A 3.17 10.8 0.98 gw9.2 9 RM242 -- RM205 IC 22015 3.21 13.95 -4.73 Yield per plant yldp2.1 2 RM262 -- RM183 IC 22015 4.38 12.21 -3.54 4.35 14.2 -3.79 yldp2.2 2 RM183 -- RM263 IC 22015 3.59 7.05 -2.70 yldp9.1 9 RM242 -- RM205 IC 22015 4.18 23.2 -13.84 Harvest index hi2.1 2 RM183 -- RM263 IC 22015 3.12 5.83 -2.85 2.76 5.64 -2.8 Plot yield yld1.1 1 RM243 -- RM81A IC 22015 4.23 6.98 -3.98 3.87 5.86 -3.09 yld2.1 2 RM262 -- RM263 IC 22015 31.92 38.46 -216.04 35.33 50.47 -238.51 yld8.1 8 RM350 -- RM210 IC 22015 3.86 4.67 -67.45 yld8.2 8 RM210 -- RM256 IC 22015 3.35 3.98 -62.92 4 10.88 -103.54 yld8.3 8 RM38 -- RM25 IC 22015 4.56 7.99 -89.13 yld8.4 8 RM223 -- RM210 IC 22015 7.02 20.24 -138.96 yld8.5 8 RM256 -- RM230 IC 22015 6.42 15.35 -134.53 ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Distribution of QTLs on the molecular linkage maps constructed based on BC~2~testcross population of IR 58025A (O. sativa) × IC 22015 (O. rufipogon) ph: Plant height, nt: Number of tillers per plant, np: Number of panicles per plant, pl: Panicle length, sn: Spikelet number per panicle, snp: spikelet number per plant, gn: Grain number per panicle, gnp: Grain number per plant, sf: Spikelet fertility, gw: Grain weight, yldp: Yield per plant, hi: Harvest index, yld: Plot yield. ::: ![](1471-2156-6-33-2) ::: Interaction among QTLs ---------------------- A two-way test to detect epistatic interactions between marker loci was performed using the EPISTAT software \[[@B26]\]. The analysis identified a total of 15 interactions consisting of 20 markers spread across 8 different chromosomes (Table [4](#T4){ref-type="table"}). These markers did not contribute to the phenotype singly but had a significant effect on the phenotype in combination with another marker indicating strong G × G interactions. This may be one of the reasons for the transgressive segregants obtained. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Significant two-way interactions between marker loci determined using EPISTAT program ::: Trait Marker 1 Marker 2 MC-test^\#^ ----------- -------------- -------------- ----------------- --- ------- NT RM282 3 RM210 8 0.002 PL RM243 1 RM8 2 0.001 SN RM22 3 RM169 5 0.001 SNP RM282 3 RM210 8 0.001 GN RM211 2 RM263 2 0.000 GNP RM262 2 RM169 5 0.002 RM169 5 RM214 7 0.000 SF RM36 3 RM340 6 0.000 HI RM5 1 RM240 2 0.001 RM272 1 RM211 2 0.002 RM272 1 RM50 6 0.000 RM60 3 RM257 9 0.005 PY RM263 2 RM183 2 0.000 RM263 2 RM264 8 0.077 RM183 2 RM38 8 0.073 ^\#^Monte Carlo simulation using EPISTAT program to evaluate significance of interactions ::: Discussion ========== Marker segregation ------------------ The allele frequency in a BC~2~population without selection would be 87.5% IR 58025A alleles to 12.5% O. rufipogonalleles. Twenty three markers (28.75%) were skewed towards one or the other parent resulting in an allele frequency of 83.26% IR 58025A alleles to 16.74% O. rufipogonalleles. Ten marker loci (12.5%) were skewed towards O. sativaparent, whereas, 13 markers (16.25%) had over representation of O. rufipogonalleles. Skewness of markers towards one of the parents has been documented for interspecific as well as intersubspecifc crosses in rice \[[@B14]-[@B19]\], \[[@B27]-[@B29]\]. A comparison of the results with earlier studies involving O. rufipogonrevealed that the percentage of skewed markers was lower compared to that reported by Moncada et al\[[@B15]\] (37.6%) and Thompson et al\[[@B18]\] (42.5%) and higher compared to Septiningsih et al(21.4%) \[[@B17]\]. All the three previous studies used same accession of O. rufipogon(IRGC 105491)) but different recurrent parents. This suggests that the polymorphism percentage is relative and depends on parental combination. Skewness towards the elite parent could have been due to the intensity of selection imposed in the BC1 generation, while, skewness towards O. rufipogonmay be due to reduced recombination and linkage drag in some regions of an interspecific population \[[@B30],[@B31]\]. While, segregation distortion of RM251 and RM7 on chromosome 3 may be due to their proximity to the gamete abortive gene, ga3\[[@B32]\], the deviation from the mendelian ratio of RM249 and RM44 might be due to the presence of these markers close to the centromeres of chromosome 5 and 8 respectively. Some of the chromosomal regions may have been distorted due to the selection imposed in BC1. The segregation distortion, towards IR 58025A, of marker loci RM84 on chromosome 1, RM207 on chromosome 2 and RM22 on chromosome 3, which exist in close proximity to dwarfing genes d18, d29and d50respectively \[[@B32]\] may be due to the selection for semi-dwarf plant height in BC1 generation. The distortion of RM13, RM169 and RM262 (markers associated with QTLs for number of tillers) towards O. rufipogon, may be due to the selection imposed on BC1 for high tiller number, a trait that was superior in O. rufipogon. RM13 was also shown to exhibit segregation distortion towards O. rufipogonin an earlier study \[[@B17]\]. Trait correlations ------------------ The present study confirms that major components follow significant positive relationship with yield. Most of the trait correlations confirm with those reported earlier for studies involving O. rufipogon. Grain weight was negatively associated with both spikelet number per panicle and grain number per panicle \[[@B13],[@B17],[@B18]\]. In the present study, the correlation between grain weight and yield was non significant as also reported earlier for an IR64/O. rufipogoncross \[[@B17]\], However other studies on O. rufipogonreport a positive correlation between yield and grain weight \[[@B13],[@B18]\]. There is no correlation detected in the present study between panicle number and yield, however, a positive correlation between these two traits was reported in an IR64/O. rufipogonderived cross \[[@B17]\]. O. rufipogonderived QTLs for yield improvement ------------------------------------------------ Oryza rufipogonalleles had a beneficial effect on 74% of the QTLs obtained for yield and yield components in the present study. This is a higher percentage than documented for interspecific crosses in rice. In previous studies involving O. rufipogon, alleles from wild species had beneficial effect in 35--58% of the QTLs \[[@B13],[@B15],[@B17],[@B18]\]. The higher percentage reported here might indicate the presence of a larger number of favorable alleles in this accession of O. rufipogoncompared to the one used in the previous studies. Alternatively, IR 58025A might have inferior alleles at many of the loci compared to the O. rufipogonalleles or the alleles introgressed from the wild species may interact better with the IR 58025A background compared to the O. sativaaccessions used earlier. The intensive selection in the BC1 for higher tiller number, a superior trait in O. rufipogon, may be another reason for the increased contribution of wild alleles. The O. rufipogonalleles contributed to an increase in panicle length (pl2.1, pl5.1, pl9.1), number of tillers (nt2.1, nt5.1), number of panicles (np2.1, np2.2), spikelet number per plant (snp2.1, snp5.1, snp5.2), grain number per panicle (gn5.1), grain number per plant (gnp2.1, gnp2.2, gnp5.1), grain weight (gw2.1, gw9.1), yield per plant (yldp2.1, yldp2.2, yldp9.1), harvest index (hi2.1) and plot yield (yld1.1, yld2.1, yld8.1, yld8.2, yld8.3, yld8.4, yld8.5). The O. rufipogonalleles also resulted in an increase in plant height (ph1.1, ph1.2), however, they did not enhance spikelet fertility and spikelet number per panicle. It is interesting because, the population had transgressive segregants for both the traits, in spite of the inferiority of O. rufipogonfor these traits. The alleles in IR 58025A may be dominant at this loci compared to the alleles from O. rufipogon. However, despite its inferiority for the trait, the alleles from the wild species had beneficial effect on grain weight indicating that the alleles contributing to grain weight might interact positively with the genetic background of IR 58025A. Interaction among QTLs ---------------------- An analysis to identify the potential epistatic interactions between marker loci, using EPISTAT software \[[@B26]\], identified 20 markers resulting in 15 two-way interactions (Table [4](#T4){ref-type="table"}). All these markers had no effect on the trait singly but resulted in an enhanced effect when combined with another marker. The resulting G × G interactions between these markers may be one of the reasons for the appearance of transgressive segregants in the population. Several chromosomal regions were associated with more than one trait, indicating linkage or pleiotropic effects. For example, the QTLs gnp2.2and yldp2.2, associated with an increase in grain number per plant and yield per plant respectively were located in the same region on chromosome 2. Similarly, the region associated with nt2.1which controlled an increase in number of tillers was linked to np2.1, gnp2.1, yldp2.1, hi2.1and yld2.1controlling an increase in number of panicles, grain number per plant, yield per plant, harvest index and plot yield respectively. The O. rufipogonalleles had beneficial effect on all these traits. However, the same region is associated with a negative QTL from O. rufipogon, gw2.3, resulting in decreased seed weight. At a different chromosomal region, O. rufipogonallele associated with a QTL gw2.1, leading to an increase in grain weight is linked to two negative QTLs, sn2.1and gn2.1, which result in decrease in spikelet number per panicle and grain number per panicle. The reverse is true for the region associated with another QTL for grain weight, gw2.2. This negative QTL from O. rufipogonis linked with two QTLs corresponding to grain number per plant, gnp2.1and yield per plant, yldp2.1, where the O. rufipogonalleles had positive effect. It is very interesting that the same chromosomal region associated with a positive QTL for grain weight coincides with negative QTLs for spikelet number and grain number and vice versa. As grain weight is negatively correlated with spikelet number and grain number, it is tempting to speculate that the same QTL might contribute to both the phenotypes. Further characterization of this region by fine mapping and identification of genes underlying it will throw more light on whether the same set of genes, regulated differentially, or an entirely different set of genes govern these phenotypes. The association of positive and negative QTLs to the same chromosomal regions was earlier reported for studies involving O. rufipogonwhere the positive traits for grain weight and panicle length together and panicle length alone were linked with negative QTLs for plant height and broken rice respectively \[[@B17],[@B19]\]. In lieuwith the association of the positive and negative QTLs to same chromosomal regions, a careful selection will be needed to avoid negative characteristics in the crop improvement process. Comparison with QTLs from other wild species -------------------------------------------- A comparison of the QTLs obtained with the other wild species including O. rufipogon, O. glaberrimaand O. glumaepatularevealed that 27 out of 39 QTLs obtained in this study had congruent occurrences with QTLs reported earlier (Table [5](#T5){ref-type="table"}). The QTLs that overlap with other studies fall into two categories: i) QTLs that share similar map position and mapped to same trait and ii) QTLs that share a similar map position, but mapped to a different trait. The QTL for panicle length, pl2.1, is mapped to the same region under same name in a study involving O. rufipogon/Jefferson \[[@B18]\], while it is associated with yield components like grains/panicle and yield in a cross involving V20A/O. rufipogon\[[@B13]\]. However, the O. rufipogonalleles contributed to positive effect in both the cases. In case of pl9.1, QTLs under same name and associated with same chromosomal regions were reported previously in three separate studies involving O. rufipogon(IRGC 105491) \[[@B13],[@B17],[@B18]\]. However, the same region is associated with a negative QTL for panicle length (pnl) in a study involving another accession (P 16) of O. rufipogon\[[@B14]\]. This indicates an accession based variation in the alleles at this locus, with alleles derived from accessions IRGC 105491 and IC 22015 superior to the alleles from O. rufipogonaccession, P 16. The alleles from O. rufipogonincreased yield at yld1.1, yld2.1, yldp2.1, yldp2.2and yld8.1. While, QTLs with same names as yld1.1and yld8.1were reported in similar regions in a cross V20A/O. rufipogon, the beneficial effect of yld2.1, yldp2.1and yldp2.2coincided with an increase in panicle length \[[@B13]\]. The position of yld8.1overlaps with another yield component, grains per panicle, gpp8.1, in a cross involving Jefferson/O. rufipogon\[[@B18]\]. The QTLs for grain weight, gw2.1and gw9.1shared same names and had orthologous regions with the QTLs identified in V20A/O. rufipogoncross \[[@B13]\]. While, O. rufipogonhad beneficial effect on grain weight in the present study, they had a negative effect in the earlier study, suggesting that the alleles at this locus might be superior to those reported earlier. However, a negative QTL from grain weight in this study, gw2.3, is associated with panicle length, pl2.1, in case of \[[@B13]\] where the O. rufipogonalleles had beneficial effect. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Comparison of QTLs with other studies involving wild rice species ::: Chr./marker QTLs in this study QTLs identified in other wild species Ref -------------------------- ---------------------------- ------------------------------------------- ---------- ---------- ---------- ---------- --------- ---- Chromosome 1 RM220 -- RM272 ph1.1 spp1 gyp1 fgp1 gypa1 16 sp1 al1 14 RM272 -- RM259 ph1.2 dth1.1 18 RM243 -- RM81A yld1.1 yld1.1 13 BR 19 RM212 -- RM315 sf1.1 BR 19 ph1.1 pl1.1 17 pss1.1 pth1.2 ph1.2 pl1.1 18 Chromosome 2 RM250 -- RM208 pl2.1,sn2.1,gn2.1,gw2.1 yld2.1 gpp2.1 17 gl2.1 gw2.1 18 ph2.1 gpl2.1 15 ppl2.1 gpl2.1 yld2.1 13 RM262 -- RM183 nt2.1,np2.1,yld2.1,gw2.3 dtf2 16 gnp2.1,yldp2.1, snp2.1 pl2.1 13 RM262 -- RM263 gnp2.1,yldp2.1,hi2.2 dtf2 spp2 fgp2 16 pl2.1 13 al2 14 HR CR BR 19 Chromosome 3 RM36 -- RM251 sf3.1 CR 19 dth3.2 sh3.2 18 kl3.1 20 amy3 21 Chromosome 5 RM194 -- RM249 gn2.1,snp5.1,gnp5.1 gw5.1 13 Chromosome 8 RM44 -- RM350 snp8.1 tnr8 dnr8 plh8 16 sh8.1 18 cp8.1 20 RM350 -- RM210 yld8.1 GD 19 gpp8.1 ph8.1 18 ph8.1 pl8.1 gpl8.1 yld8.1 13 Chromosome 9 RM257 -- RM242 ph9.1 gw9.1 13 RM242 -- RM205 pl9.1,gw9.1 pnl 14 pl9.1 spp9.1 13 dyg 19 pl9.1 17 gw9.1 gpp9.1 spp9.1 pl9.1 yld9.1 tt9.1 18 ::: Comparison of QTLs across Oryza species --------------------------------------- The present study identified a total of 39 QTLs. Thirty QTLs have corresponding occurrences with QTLs reported earlier, while, 9 QTLs (nt2.1, nt5.1, snp5.1, hi2.1, yldp2.1, yldp9.1, yld2.1, yld8.1, yld8.5) are novel and reported for the first time. The results are on the expected lines, as new parental combinations especially involving exotic species are likely to unfold novel variability. Of the three QTLs detected for plant height, the O. rufipogonalleles increased plant height at two loci while another QTL decreased plant height. All the three QTLs have been reported previously (Table [6](#T6){ref-type="table"}). The similarity of the regions associated with QTLs for plant height with other studies involving O. rufipogon, indicaand japonicacultivars indicates that the location of alleles for plant height are conserved across different genetic and environmental backgrounds. The two QTLs for tiller number, nt2.1and nt5.1, identified in this study are novel and have no correspondences with QTLs reported earlier for this trait. This indicates that these may be a potentially new set of alleles specific for this accession of O. rufipogon. All the QTLs for number of panicles have been reported earlier (Table [6](#T6){ref-type="table"}). ::: {#T6 .table-wrap} Table 6 ::: {.caption} ###### Comparison of QTLs across Oryzaspecies ::: QTL QTLs reported in earlier studies ----------- ------------------------------------------------------------ ph1.1 PTHT \[33,34\], ph1\[35\], qPH-1\[36\], ph1-1\[37\] ph1.2 PTHT \[33,34,36\], ph1\[35\], ph1-1(10), ph1\[38\] ph9.1 ph9\[37,38\] np2.1 tns2\[35\] np2.2 PNNB \[35,40,57\] pl2.1 PNL \[53\], pl2.1 \[18\] pl5.1 PNL \[40\], QPI5 \[53\] pl9.1 qPI9b\[53\], pl9.1\[13\], pl9.1\[17\], pl9.1\[18\] sn2.1 SPKNB \[53\] snp2.1 tns5\[35\] snp5.1 SPKNB \[34,40,53\] snp8.1 SPKNB \[40\], tns5\[35\] snp8.2 SPKNB \[40\], tns5\[35\] gn2.1 gpp2.1\[17\] gn5.1 FGRNB \[41\] gnp2.1 FGNRB \[39, 55\] nfg\[35\] gnp2.2 fgp2\[16\] gnp3.1 gp3\[41\] gnp5.1 qNFGP-5-1\[56\] sf1.1 SF \[51\], pss1.1\[18\] sf3.1 SF \[51\], S3b\[52\], SF \[11\] gw2.1 GW \[39\], gw2.1\[41\] gw2.2 GW \[39\], tgwt\[35\], GW \[54\] gw2.3 kw2-2\[55\] gw9.1 gw9.1\[18\] yldp2.2 GRYLDPPL \[35,39,40\] yld1.1 GRYLD \[39\], yld1.1\[13\] yld8.1 GRYLD \[39\], yd8\[41\], yld8.1\[13\] yld8.3 GRYLD \[40\] yld8.4 GRYLD \[35,39\], yd8\[41\] ::: All the three QTLs for panicle length, pl2.1, pl5.1and pl9.1, were trait enhancing and overlapped with the regions identified earlier for the same trait (Table [6](#T6){ref-type="table"}). This is in agreement with earlier studies where O. rufipogonalleles had a positive effect \[[@B13],[@B16]-[@B18]\]. The large number of studies implicating a similar region as pl9.1indicate that this region has a similar predictable effect on the phenotype irrespective of the genetic background. Two QTLs were identified for grain number per panicle (gn2.1, gn5.1) and one for spikelet number per panicle (sn2.1) in the present study. The alleles from O. rufipogonhad a negative effect on the QTLs on chromosome 2, whereas, they enhanced the number of grains per panicle on chromosome 5. A similar negative effect of O. rufipogonalleles had been reported for grain number per panicle, on chromosome 2, in a cross involving IR64/O. rufipogon\[[@B17]\]. The negative effect across O. rufipogonaccessions indicates the superiority of the O. sativaalleles at this locus. Five QTLs were identified for spikelet number per plant and 4 QTLs for grain number per plant. All the four QTLs for grain number per plant were reported earlier, while only four of the five QTLs for spikelet number per plant are documented (Table [6](#T6){ref-type="table"}). The O. rufipogonalleles had negative effect on snp8.1, snp8.2and gnp3.1whereas, they had beneficial effect on all the other QTLs. The QTL for grain number per plant, gnp2.2, coincided with a QTL reported for the same trait (fgp) in a cross involving O. glumaepatula\[[@B16]\]. While, the O. rufipogonalleles had a beneficial effect in the present study, the O. glumaepatulaalleles had a negative effect, indicating that this accession of O. rufipogonhas a novel set of alleles at this locus that are superior to O. glumaepatula. The QTL snp5.2is novel and is reported for the first time. Again, this indicates the possibility of the presence of novel alleles in this accession of O. rufipogon. Two QTLs, sf1.1and sf3.1, both conferring negative effect, were identified for spikelet fertility. Both the QTLs have been reported earlier and they also are in agreement with earlier study indicating the negative effect of the O. rufipogonalleles on this trait \[[@B18]\]. All the four QTLs for grain weight have been reported earlier indicating that the allele set may be common across most of the genetic backgrounds. The O. rufipogonalleles contribute to positive effect for two of these QTLs (gw2.1, gw9.1), while the other two derive negative effect from the wild alleles. The beneficial effect at gw2.1and gw9.1is in contrast to what has been previously reported for this trait in a Jefferson/O. rufipogoncross, where the O. rufipogonalleles have a deprecating effect on both these QTLs \[[@B18]\]. This indicates that alleles at these loci may be superior in this accession of O. rufipogonor the same set of alleles might perform better in the IR 58025A background compared to the Jefferson background or the G × E interactions might be at play. Six of the eight QTLs identified for yield have been reported earlier \[[@B35],[@B39]-[@B41]\] suggesting that the QTLs for yield are conserved across different genetic backgrounds. Two QTLs, yld2.1 and yld8.5are reported for the first time. The O. rufipogonalleles had beneficial effect on all the eight QTLs. Of the two QTLs identified for yield per plant, the O. rufipogonalleles were responsible for increase in yield in both the cases. While, yldp2.1is novel and reported for the first time, yldp2.2coincides with the regions reported earlier for this trait (Table [6](#T6){ref-type="table"}). The QTL for harvest index, hi2.1, is novel and is reported for the first time in this study. Conclusion ========== The study while confirming the view that the progenitor species constitute the largest source of still unfolded variability for traits of complex inheritance like yield and its components has helped identify additional novel variability for yield improvement. The novel QTLs identified are good candidates for fine mapping and positional cloning studies, while, the QTLs that are mapped to regions consistent with other studies can be useful for marker-assisted transfer of these QTLs. The availability of the complete rice genome sequence and rapid advances being made in the area of genomics will help dissect and characterize yield related QTLs further. Considering the potential of yield influencing new QTLs, more research is warranted to unearth and use more and more novel yield related gene blocks hidden in closely related wild/weedy species and primitive cultivars, if the rice dependent world is to truly attain and sustain food security. Methods ======= Choice of parents ----------------- IR 58025A, a commercial cms line developed by Directorate of Rice Research (DRR), India, <http://www.drrindia.org/> was used as a recurrent parent. IR 58025A grows to a height of 80 cm and is characterized by having long grain type and early maturity along with good milling and eating qualities. The O. rufipogonaccession, IC22015, collected from Kerala, India, and maintained at DRR was used as a donor parent. Development of mapping population --------------------------------- An advanced backcross strategy as described in \[[@B13]\] was followed to develop the mapping population. A single plant of O. rufipogon(IC 22015) was used as a male parent and crossed to IR 58025A to generate F~1~plants. Fourteen F~1~plants, whose hybrid nature was confirmed with microsatellite markers were backcrossed to IR 58025B (an isogenic line of IR 58025A) used as male to produce BC~1~. Fifty BC~1~plants, looking morphologically like IR 58025A were backcrossed to IR 58025B to produce BC~2~. Out of a population of 3000 BC~2~plants obtained, 251 male sterile plants were randomly selected and testcrossed to KMR3, the restorer of IR 58025A to produce 251 testcross families. The 251 BC~2~testcross families constituted the mapping population. Simultaneously, under similar conditions, a cross was made between IR 58025A and KMR3 to obtain the hybrid, KRH2, to be used as the check. Phenotypic evaluation of mapping population ------------------------------------------- The 251 testcross families, two parents and checks viz., KRH2, Jaya and IR64 were grown under irrigated conditions at DRR in an augmented block design in two replications with checks repeated after every 10 families. Each of the testcrosses and the check consisted of 40 plants planted in 4 rows of 10 plants each adopting a uniform spacing of 20 cm × 20 cm. Six plants in the middle of each of these families were evaluated for the following yield related traits: Plant height (PH)-- length of the tallest tiller (cm) from soil surface to the tip of the panicle. Tiller number per plant (NT)-- Total number of tillers per plant. Panicles per plant (NP)-- Panicles with seed set exceeding 15%. Panicle length (PL)-- length (cm) from neck to tip of the panicle. Spikelet number per panicle (SN)-- number of spikelets including empty and filled ones averaged over five randomly chosen panicles in each plant. Spikelet number per plant (SNP)-- total number of spikelets including empty and filled ones in each plant computed as average number of spikelets per panicle × number of productive tillers. Grain number per panicle (GN)-- number of filled spikelets per panicle averaged over five randomly chosen panicles in each plant. Grain number per plant (GNP)-- number of filled spikelets in a plant computed as average number of filled spikelets per panicle × number of productive tillers per plant. Spikelet fertility (SF)-- ratio of filled spikelets to the total number of filled and unfilled spikelets per panicle, expressed in percentage. Grain weight (GW) -- weight (g) of 1000 filled spikelets, averaged over six samples taken from the bulk-harvested grain from each plant. Harvest index (HI)-- ratio of filled grains to biomass (filled grains, unfilled grains and straw of the plant) in terms of weight (g) expressed in percentage. Grain yield per plant (Yldp)-- weight (g) of filled grains per plant. Grain yield (Yld)-- weight (g/kg) of filled grains harvested from each testcross family (40 plants) extrapolated to tonnes per hectare. Trait correlations ------------------ Correlations between character pairs were computed at p \< 0.05 and p \< 0.01 in Excel using trait averages. DNA extraction -------------- DNA was extracted from two months old leaf tissue using the protocol of Dellaporta \[[@B42]\]. Parental polymorphism and linkage map construction -------------------------------------------------- A set of 210 randomly selected microsatellite markers (Donated by Rockefeller Foundation to EAS) spanning all the 12 chromosomes were screened among the O. sativaand O. rufipogonparents. A total of 80 polymorphic microsatellite markers separated by an average distance of 15.37 cM were used to analyze the 251 testcross progeny. Linkage maps were constructed using the Mapmaker version 3.3 \[[@B43]\] following Kosambi Function \[[@B44]\]. Linkage groups were determined using \'group\' command with an LOD score of 3.0 and a recombination fraction of 0.5. Order of the markers for each group was determined using \'order\' and \'ripple\' commands. Assignment of linkage groups to the respective chromosomes was done based on the rice maps developed at Cornell University \[[@B18],[@B45]\]. QTL analysis ------------ QTLs were analyzed using single marker analysis (SMA), interval mapping (IM) and composite interval mapping (CIM). Single marker analysis wasperformed by regression of field performance on marker genotypes using standard analysis of variance (ANOVA) procedure at a statistical threshold of p \< 0.01 and assuming regular segregation of wild and cultivated alleles in the testcross families. The proportion of observed phenotypic variance attributable to a particular QTL was estimated as the difference between the mean of the segregants having the O. rufipogonallele and the mean of the segregants that did not have the O. rufipogonallele. The phenotypic variance over the check KRH2 was also calculated in a similar manner. QTL analysis by interval mapping (IM) and Composite interval mapping (CIM) \[[@B46]\] was done using QTL Cartographer 3.0 \[[@B47]\]. The significant threshold value for identification of a QTL (both for IM and CIM) was determined based on permutation tests at a significance level p \<0.01 \[[@B48]\]. Based on 1000 permutations for each trait, the threshold for IM and CIM corresponded to minimum LOD score value of 2.5. The proportion of phenotypic variance (R2) and additive effect were determined for each trait. The deviations from the expected mendelian ratio was calculated using MapDisto software \[[@B49]\] and the digenic interactions between marker loci were determined using EPISTAT software \[[@B26]\]. The QTL nomenclature followed was as reported in \[[@B50]\]. Note ---- The material used in this study can be obtained from Prof E.A. Siddiq, Honorary Professor, Center for DNA Fingerprinting and Diagnostics, Nacharam, Hyderabad, 500 076, India. The raw data used for analysis can be obtained from Dr. M. Pradeep Reddy, Department of Biology, McMaster University, Hamilton, ON, Canada. E-mail: <reddyp@mcmaster.ca> Authors\' contributions ======================= MPRparticipated in the design, carried out the field studies and marker studies, analyzed the data and darafted the manuscript. NSparticipated in the coordination of the study and helped in the design of the study, analysis of data and drafting the manuscript. VLNRcarried out the field studies, marker studies and helped in analysis of data. EASconceived of the study and participated in its design and coordination. All authors read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Transgressive Segregants ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Correlation Data ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 3 Single Marker QTLs ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We thank Indian Council for Agricultural Research for the financial support to carry out the project and Directorate of Rice Research for providing the necessary facilities. We thank the Rockefeller Foundation for donating SSR primers to EAS. We wish to thank the two anonymous reviewers whose critical comments enabled us to improve the manuscript considerably.	PubMed Central	2024-06-05T03:55:59.801943	2005-6-13	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181812/", "journal": "BMC Genet. 2005 Jun 13; 6:33", "authors": [ { "first": "Pradeep Reddy", "last": "Marri" }, { "first": "Sarla", "last": "N" }, { "first": "Laxminarayana V", "last": "Reddy" }, { "first": "EA", "last": "Siddiq" } ] }
PMC1181813	Background ========== Recent systematic searches for transcribed regions have yielded growing evidence suggesting that the fraction of the genome being transcribed is much larger than previously thought \[[@B1]-[@B4]\]. In particular, it is becoming clear that a significant fraction of the expressed DNA comes from non-coding regions \[[@B2]\]. The study of the role and function of this large amount of non-coding RNAs attracts much attention and effort. The yeast Saccharomyces cerevisiaeis one of the most studied model organisms, and the first eukaryote to have its genome sequenced \[[@B5]\]. Its genomic structure is by far simpler than that of mammalian cells, as splicing and alternative splicing play only a minor role. The early availability and simple structure of the yeast genome made it unnecessary to apply transcriptome-based methods such as large-scale sequencing of ESTs. Instead, it was first assumed that as a good first approximation, its transcriptome was well represented by the set of all open reading frames (ORFs) longer than 100 amino acids. This set of ORFs accounts for most, if not all, coding genes, but ignores the non-coding expressed regions. As a result, although much high-throughput expression data have been accumulated for the yeast ORFs, far less is known on the expression of non-coding regions. Recent studies compared several yeast genomes and showed that many of the yeast ORFs believed to be protein-coding genes are actually not conserved even in closely related species \[[@B6]-[@B9]\]. These studies suggested that the non-conserved ORFs do not actually code for proteins. These ORFs (825 out of 6703 yeast ORFs; 12%) were thus called \"dubious ORFs\" (DOs) \[[@B10]\]. In this work, we use the expression data collected for these DOs to study the expression pattern of non-coding regions in the yeast genome. We rely entirely on the current DO annotation and do not attempt to find new such ORFs. The probes in commercial yeast DNA microarrays (such as those developed by Affymetrix) were chosen to represent all the putative ORFs larger than 100 amino acids, and thus include all the ORFs now classified as dubious. Since the dubious ORFs are now known to be non-conserved and thus presumably non-coding, it follows that one can treat the expression measured by the corresponding probes as representing the whole non-coding parts of the yeast genome. While Affymetrix probes are biased to the 3\' end of the ORF, this bias is irrelevant, as we know that no protein is encoded in this region. Our results on the expression levels of dubious ORFs are thus not limited to these ORFs only. One should consider these genomic regions as a random sampling of the whole non-coding part of the yeast genome. A surprising abundance of antisense transcripts, RNAs transcribed from opposing DNA strands at the same genomic locus, was recently observed in several eukaryotic genomes \[[@B4],[@B11]-[@B13]\]. Some antisense transcripts have been shown to regulate gene expression \[[@B14],[@B15]\] but in general not much is known about how antisense transcription regulates gene expression in mammalian cells. A large fraction of the DOs show partial overlap with other, usually conserved, ORFs encoded on the opposite strand \[[@B16]\]. In fact, almost all of the sense/antisense (S/AS) ORF pairs in yeast are non-DO/DO pairs: 501 of the 538 pairs (93%) are DO/non-DO, while only 22 pairs are DO/DO, and 15 are non-DO/non-DO respectively. Thus, the extent of the antisense phenomenon in yeast critically depends on the prevalence of DO transcription. In the following, we show that a large fraction of the DOs are actually expressed. The expression pattern of the S/AS pairs is analyzed, showing that the expression profiles of two antisense transcripts are correlated. Results ======= Expression of dubious ORFs -------------------------- We started by accumulating a large body of Affymetrix gene expression experiments, which includes 154 experiments performed in 12 different studies. Background subtraction and normalization procedure for this data set are described in Methods. The dataset includes recorded expression data for 6437 and 589 non-DOs and DOs, respectively. In the following we use this data to show that many of the DOs are actually expressed. Moreover, we find that the expression levels of S/AS DO/non-DO pairs are correlated. This latter result further supports the idea that the expression of the dubious genes is a true biological phenomenon rather than an artifact or a random experimental error. First, we analyzed the expression profiles of the DOs, in order to see whether some of them are expressed after all. The microarray expression profiles were searched, looking for DOs that were expressed at high levels in particular conditions. We define a DO as a \"strongly expressed DO\" (SDO) if its level of expression exceeds the 70^th^percentile in at least one condition in our data set, and it does not overlap any non-DO. This expression criterion depends only on the expression intensity of the ORFs relative to other probes in the same experiment, and hence does not depend on the background subtraction and normalization of different data sets. The 70th percentile threshold roughly corresponds to a threshold of 200 in normalized Affymetrix average difference (PM-MM) units, which is a conservative cutoff that minimizes the false positives rate \[[@B17],[@B18]\]. We find 164 such SDOs (28% of all DOs with recorded expression), which are listed in Table 1 of Additional file [1](#S1){ref-type="supplementary-material"} along with their 5 highest expression values. These 164 SDOs contain 88 antisense DOs and 76 non-antisense DOs, which are 26% and 31% of the AS and non-AS DO with recorded expression respectively. Hence, the strong expression of the SDOs cannot be accounted for by the antisense phenomenon, nor attributed to an artificial connection between the strands, like in cDNA expression experiments. One thus may conclude that at least 28% of the DOs are actually expressed into RNA under specific biological conditions. Antisense expression correlations --------------------------------- We then studied the relation between the expression profiles of S/AS pairs. 495 DOs participate in DO/non-DO S/AS pairs (6 DOs have two overlapping non-DOs), of which 341 have recorded expression and 333 exceed the background level (see Methods). We compared the distribution of Pearson\'s correlations (PC) \[[@B19]\] between the expression profiles of these DO/non-DO S/AS pairs with that of three control sets: (i) Randomly picked non-DOs from the S/AS set, paired with randomly picked DOs (see Figure [1](#F1){ref-type="fig"}) (ii) Randomly picked DOs from the S/AS set, paired with randomly picked non-DOs (iii) Randomly picked DOs paired with randomly picked non-DOs (It is important to use random non-DO/DO pairs as control since the expression level of DOs is, on average, three times lower as that of non-DOs). If the observed expression of DO is a result of random experimental error, or caused in any way by our normalization procedure, we would expect no significant difference between the PC distribution of S/AS DO/non-DO pairs and that of one of the random sets of DO/non-DO. The results are now summarized in the first row of Table 3 of Additional file [1](#S1){ref-type="supplementary-material"}. A significant difference is observed (P \< 2 × 10^-10^;χ^2^test) between the S/AS PC distribution and the PC distributions for each one of the other random sets, the S/AS pairs being more correlated. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Distributions of PC for S/AS DO/non-DO pairs (blue) and randomly reshuffled DO/non-DO pairs from the S/AS set (green). 103,479 pairs used to calculate the second distribution. There is a significant difference between the S/AS and random distribution (χ^2^= 124, DF = 9 and P= 10^-22^). S/AS pairs are significantly more correlated than random DO/non-DO pairs. For the S/AS distribution average = 0.076 and SD = 0.22. For the reshuffled distribution average = -0.018 and SD = 0.18. The significance of the difference in the averages is P~t~= 10^-20^(student t-test). ::: ![](1471-2164-6-93-1) ::: Employing the \"leave one out\" method, we further checked that the results do not follow from one single study, which is either faulty, use inadequate experimental technique, or not compatible with our processing methods. We carried out statistical tests leaving each time one of the data sets out of the full data set. Table 3 of Additional file [1](#S1){ref-type="supplementary-material"} shows that regardless of the study left out or the control set of random pairs chosen, there is a significant difference between S/AS DO/non-DO and random pairs PC distribution, S/AS pairs being on average more correlated than random pairs. Thus, our results show an abundant transcription of the strand opposite to coding genes in yeast, where the opposite strand expresses a DO. The expression of the two strands is correlated. Experimental confirmation ------------------------- In order to experimentally confirm the expression of DOs, we carried out an RT-PCR analysis. First, the expression of 4 SDOs was analyzed. Oligonucleotides specific for ORFs YER121w, YMR245w, YDR525wand YJL199cwere synthesized, and RT-PCR analysis was carried out. In this assay RNA products created by transcription of specific genes are first amplified by reverse-transcription, and the resulting products are further amplified by PCR. An RT- PCR product of the expected size was obtained for each of the four analyzed genes (Figure [2A](#F2){ref-type="fig"} and data not shown). No RT-PCR product was obtained with samples in which the reverse-transcription step was omitted, indicating that the results obtained represent true amplification of mRNA molecules. The results obtained confirm that these ORFs are transcribed. Direct DNA sequencing of the RT-PCR products confirmed the identity of the amplified sequences. Next, we analyzed S/AS non-DO/DO pairs. These include the ORFs YBR112c/YBR113w, YGR181w/YGR182cand YPL181w/YPL182c. All the DOs present in these pairs are expressed, as demonstrated by the RT-PCR product obtained individually with primers specific for each individual ORF (Figure [2B](#F2){ref-type="fig"} and data not shown). As before, we observed no RT-PCR product when the reverse-transcription step was omitted, again demonstrating the presence of a transcribed RNA molecule. To rule out potential artifacts related to the presence of additional transcripts emanating from adjacent ORFs, we also performed RT-PCR reactions with primers encompassing both ORFs (p1 and p4 in Figure [2B](#F2){ref-type="fig"}). No amplification was detected, ruling out the possibility that the detection was due to the presence of a single, joint, transcript. In order to reject the possibility of the DO YGR182cbeing part of a long UTR of its adjacent ORF YGR183C, we calculated the expression Pearson correlation between these two ORFs and found it to be rather weak, only 0.26. For comparison, the expression correlation between YGR183Cand its antisense YGR181wis 0.39. In addition, for the other pairs tested the transcripts are divergent, so there cannot be a transcript coming from a nearby gene. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### RT-PCR analysis of dubious ORFs. A) Reverse-transcription was carried out using oligo dT and specific primers for each ORF. RT: reverse transcription followed by PCR amplification. N: No reverse transcriptase added. G: Genomic DNA (positive control). NS: No RNA or DNA substrate added (negative control). B) A similar RT-PCR analysis was carried out for sense-antisense pairs. Shown here are YGR181w/TIM13and the DO YGR182c. No amplification was detected with primers p1 and p4. If a transcript was present that encompasses both ORFs, amplification was expected, of the size observed in the reaction carried out with genomic DNA ::: ![](1471-2164-6-93-2) ::: We thus conclude that many of the dubious ORFs are transcribed, some of which as sense/antisense pairs. As explained above, genomic comparisons have demonstrated that these dubious ORFs are not conserved among closely related yeast species, and thus probably do not translate into proteins. If indeed these genomic loci do not code for proteins, nothing distinguishes them from any other genomic region. However, we clearly show here that many of these genomic loci are transcribed, sometimes at very high levels. Therefore, our results suggest that a large fraction of the yeast genome is transcribed, even if it does not encode for conserved proteins. In addition, we have shown that yeast cells express numerous sense/antisesnse transcripts that may play a role in controlling gene expression, in accordance with previous reports for other species genomes \[[@B4],[@B11]-[@B13],[@B20]\]. Hurowitz and Brown \[[@B16]\] have recently analyzed the lengths of yeast mRNAs using the Virtual Northern method. A categorized list of 820 DOs was included in their study; each DO was classified according to the agreement between its observed and expected mRNA transcript length. Transcript measurements were available for 243 of the DOs. Of these, 192 (79%), exhibited good agreement. We find a high degree of association between the SDO set and the transcribed lists: 163 out of 164 SDOs were categorized; 60 of them were classified as exhibiting good agreement, and 77 had some agreement (P = 10^-5^and 7× 10^-8^accordingly; a hypergeometric probability function -- see Methods). These experiments thus further support our claim that the SDOs are actually expressed. It should be noted that the definition of SDOs is cut-off and dataset dependent. Reducing the strong DO expression threshold from 70 percentile to 50 percentile results in an increase in both SDOs and S/AS pairs: 336 SDOs and 214 S/AS SDO/non-DO pairs (compared with 164 and 88, respectively). The addition of the expression data set from Roth et al. (1998) \[[@B21]\] (4 additional conditions) results in a total of 251 SDOs, of which 149 are paired with a non-DO antisense. We chose to ignore this data set due to the sharp increase in the number of SDOs, which is not reflected in any of the other studies used. Discussion ========== We demonstrated that a significant portion of yeast\'s DOs is expressed. While we cannot rule out the possibility that some of the DOs do actually code for proteins, there is strong evidence against this possibility \[[@B6]\]. In addition, we compared the list of 825 yeast\'s DOs with a list of homologous pairs of Ashbya gossypiiand yeast ORFs \[[@B7]\]. Only one out of 825 members of our DOs list (YJR012C) had a homolog among Ashbya gossypiigenes. This further supports the view that the DOs do not code for proteins. Combining this finding with our strong evidence for expression of the DOs, it follows that a large fraction of the yeast genome is transcribed but not translated. This finding is in accordance with similar results for a number of other organisms studied \[[@B1]-[@B4]\]. The question thus arises: what is the role and function of this transcription? Early models of differential gene expression \[[@B22]\] assumed that most of the genome is transcribed, and that untranslated regions, such as the 3\' and 5\' ends of the genes, could play a central role in regulation of expression. In accordance, a recent study has studied full-length yeast transcripts, and found that untranslated regions (UTRs) encompass about 300 nucleotides per gene, much longer than previously expected. According to this study, 15% of the yeast genome is transcribed as UTRs \[[@B16]\]. UTRs have important roles in regulating mRNA localization and translation \[[@B23]\]. Therefore, some of the expressed DOs could be actually within the 3\' or 5\' UTR of an adjacent coding ORF. However, this explanation can account for only part of the expressed dubious ORFs we observed, as the distance of many of the expressed ORFs to their adjacent coding ORF is larger than 300 bp. It thus appears that these represent non-coding RNA genes. A growing number of non-coding genes, such as small nuclear RNAs and microRNAs, have been recently observed in many different organisms, in addition to the known tRNAs and rRNAs. These RNA genes are transcribed in a regulated way and, although not coding for proteins, play an important role in regulating the expression of other genes. For example, the yeast SER3gene has been recently shown to be regulated by the noncoding intergenic transcript SRG1\[[@B24]\]. Computational effort has been made to find such genes employing comparative genomics and domain analysis methods \[[@B25]\]. However, genes of this type have been largely over-looked in all large-scale analyses of the yeast genome that were based on the search for ORFs. It is thus possible that non-coding genes cover a significant part of the yeast genome, and account for the expression of part of the dubious ORFs. Moreover, an important subset of the non-coding DOs is the set of many antisense RNAs -- RNAs expressed from the strand opposite to a coding gene -- that may act to regulate the translation of the coding gene on the other strand. Our results show that, similar to other eukaryotes genomes \[[@B4],[@B11]-[@B13],[@B20]\], yeast cells express an abundance of antisense transcripts, and virtually all of these transcripts are non-coding. In addition, we found for the first time an association between the expression levels of sense and antisense genes. These results open the way to further understanding of the antisense phenomena in yeast, which is the most studied eukaryotic model organism. Conclusion ========== We have provided evidence that a significant portion of the non-coding regions in the yeast genome is transcribed, in accordance with similar results recently obtained for many other organisms. We have shown that the antisense phenomenon in yeast is almost solely limited to non-coding regions, and that correlated transcription of antisense non-coding genes is abundant in the yeast genome. A number of possible regulatory mechanisms based on non-coding transcripts have been suggested, but the overall role of this phenomenon is yet elusive. Methods ======= Yeast genes ----------- A list of yeast genes was retrieved from Sacchromyces Genome Database \[[@B10]\]. There are 7156 genes on the list; of them 6703 are ORFs, from which 825 are DO. All yeast data is correct to November 2003. We will limit our discussion to the ORFs. Expression data set ------------------- Affymetrix expression data sets were collected from 12 sources \[[@B26]-[@B37]\]. Altogether 154 different conditions were assembled. 6437 and 589 non-DO and DO respectively have recorded expression data in our data set. Background subtraction and normalization ---------------------------------------- The average background (BG) level for each condition was presumed to be at the conservative level of the twentieth percentile. For comparison, the parallel value used in \[[@B38]\] is five percentile. To avoid BG fluctuations effects, the background level was subtracted from the expression measure, and any expression level below this cutoff was set to 0. This subtraction considerably reduces artificial correlation between weakly expressed ORFs. After BG subtraction, expression levels were converted into relative RNA abundance. Hypergeometric distribution function \[[@B39]\] ----------------------------------------------- The probability of drawing m+n DOs, containing at least m transcribed DO, out of a categorized list containing M+N transcribed and untranscribed DO respectively is: ![](1471-2164-6-93-i1.gif) RT-PCR ------ RNA was extracted from 5 × 10^7^cells using the RNeasy kit (QIAGEN Inc.) according to the manufacturer instructions. Prior to the RT-PCR step, genomic DNA was degraded by RQ1 RNase-free DNase (Promega Inc.). Complete removal of contaminating DNA was verified by negative control PCR reactions with a specific set of primers. 1 μg of total RNA was used as a template for cDNA synthesized using the Expand™ Reverse Transcriptase Kit (Boehringer Mannheim) and 500 ng of oligo (dT)~15~(Boehringer Mannheim) as a primer. One quarter of each cDNA preparation was used as a template in a PCR reaction using specific primers. The products were subjected to agarose electrophoresis. Primers ------- The following primers were used (F; forward, R: reverse): YER121w: CAAGGCCAGCAGAGGAAAAG (F) and ATGTGCGTATGAAGCGGTTG (R). YMR245w: TCTGTATATTCTGTATCTATGTTCCTGC (F) and AAATGGCCTATTGTATTGTCAGGTC (R). YDR525w: CAAGAATTTCTCGAGTTCCTTATATATGAG (F) and AGTTTATTTCCAAAATAGCGAAGACC (R). YJL199c: CGACTGCCGCTGTTCATTCT (F) and CTTCTTGTTGCCGGCCTG (R). Sense/antisense pairs: YGR181w: P1: CTATCATCTATCTTTGGCGGCG, P2: GGTTGTCTTTGCAGTAGTGGCTG. YGR182c: P3: CAAAGGATACCAAGAAAATGCTATTACG, P4: TGGAAATAGACAGAACGAGCC YBR113w: P1: GTTCACCGCCCGGATTC, P2: TTTTACAAACCACGTCAGGAGTTC YBR112c: P3: ACTGAAGAGGCGGAGCCAG, P4: CAAAGTAGGTTTGATTACAGTTATCGTTG YPL181w: P1: AGTCTGATCGAGAGGAATTTGTACG, P2: CTCTAGTCAGGTCGTCCATCATTG YPL182c: P3: GACCATCAATAGTTTGTTTCCTTCG, P4: CTCTAGTCAGGTCGTCCATCATTG Authors\' contributions ======================= MH and EE conceived and designed the research plan and participated in all aspects of data collection and analysis. EYL participated in data analysis and interpretation. GL carried out the molecular studies. MK participated in the design of the study. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Additional material. ::: ::: {.caption} ###### Click here for file :::	PubMed Central	2024-06-05T03:55:59.807902	2005-6-16	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181813/", "journal": "BMC Genomics. 2005 Jun 16; 6:93", "authors": [ { "first": "Moshe", "last": "Havilio" }, { "first": "Erez Y", "last": "Levanon" }, { "first": "Galia", "last": "Lerman" }, { "first": "Martin", "last": "Kupiec" }, { "first": "Eli", "last": "Eisenberg" } ] }
PMC1181814	Background ========== After publication of the new CDC guideline on hand hygiene, hospital hygiene practices are currently changing in many countries \[[@B1]\]. For the post-contamination treatment of healthcare workers\' hands, it has been suggested to use alcohol-based hand rubs instead of soap and water \[[@B1]\]. This recommendation is based on a superior efficacy and dermal tolerance of the majority of alcohol-based hand rubs \[[@B1],[@B2]\]. The efficacy of hand hygiene preparations is usually tested in suspension tests (e.g. prEN 12054) and under in vivoconditions (e.g. EN 1500). Both norms have been used in many studies to determine the efficacy of preparations for hand hygiene and to determine differences among them \[[@B3]-[@B5]\]. Neutralization of residual active agents must be achieved immediately after the end of exposure and is done by a 1:10 dilution, and in addition, chemically by supplementing neutralizing agents to the sampling fluids \[[@B6]\]. The rationale for neutralisation is that, especially at short exposure times like 30 s for hygienic hand disinfection, any remaining bacteriostatic or bactericidal activity of the active agents may lead to a false positive efficacy assessment. Difficulties in neutralization of chlorhexidine gluconate have been reported before. In 1978 it was already shown that detection of surviving Staphylococcus aureuscan be completely masked by residual chlorhexidine gluconate which inhibits multiplication in standard efficacy tests without valid neutralization \[[@B7]\]. Three years later it was reported for a hand wash preparation based on 4% chlorhexidine in a dilution of 1:10, that a contamination of S. aureuscan not be recovered in the absence of neutralizing agents which is indicative of bacteriostatic carry-over effects of the antiseptic \[[@B8]\]. A similar observation was made many years later with Enterococcusspp. in a similar test design \[[@B9]\]. If a valid system for neutralization of chlorhexidine was used in suspension tests, 4% chlorhexidine revealed only little activity within 5 min against different strains of vancomycin-susceptible and-resistant Enterococcusspp. \[[@B10]\]. In 2002, it was shown that even some commonly used combinations of neutralizing agents (e.g. 3.5% Tween, 0.5% lecithin and 0.5% sodium thiosulfate) may not enable recovery of Escherichia colior S. aureusif they are exposed to chlorhexidine gluconate (0.5%). This effect was only seen with chlorhexidine gluconate but not with other non-volatile antimicrobial agents such as povidone iodine (10%), benzalkonium chloride (0.5%) or alkyldiaminoethylglcine hydrochloride (0.5%) \[[@B11]\]. These results have raised concerns on the validity of many studies on the in vitro and in vivo efficacy of chlorhexidine gluconate which have been published over the last decades \[[@B2]\]. So far, the influence of effective neutralization on the results of efficacy assessments was only studied in suspension tests. But in tests under practical conditions such as described by EN 1500, its effect has never been addressed. Therefore, we wanted to verify the effect of successful neutralization in a test under practical conditions (in vivo) with two different ethanol-based hand rubs, one of them containing 1% chlorhexidine gluconate. Methods ======= Products and application ------------------------ The following preparations were used: Iso-propanol (60%, v/v) as the reference alcohol of EN 1500; product A (3M, St. Paul, USA), based on ethanol (61%, w/w) and chlorhexidine gluconate (1%, w/w); and product B (Bode Chemie GmbH & Co., Hamburg, Germany), based on ethanol (85%, w/w). Product A was chosen as an ethanol-based hand rub with a high concentration of chlorhexidine gluconate, product B was chosen as an ethanol-based hand rub with a higher ethanol concentration but without chlorhexidine gluconate. The reference alcohol was applied according to EN 1500: 2 × 3 ml for a total of 60 s which has been shown to have an excellent reproducibility \[[@B12],[@B13]\]. The two ethanol-based products were applied as commonly done in clinical practice: 1 × 3 ml for a total of 30 s. Neutralization -------------- One set of experiments was carried out without any neutralizing agents in the sampling fluids but with 0.5% lecithin and 4% polysorbate 20 in the dilution fluid. The other set of experiments under identical test conditions was carried out with the same combination but in the sampling and the dilution fluids. These agents were chosen as they represent commonly used neutralizing agents in testing chemical disinfectants \[[@B5],[@B9]\]. Evaluation of neutralizing agents --------------------------------- The neutralizing agents were evaluated according to the \"standard test methods for evaluation of inactivators of antimicrobial agents\", ASTM E 1054-02 \[[@B14]\]. The test consists of four parts: ### Test organism viability Butterfields phosphate buffer was inoculated with a volume of test organism to yield 30 -- 100 colony-forming units (CFU) per milliliter (ml). The dilution was vortexed. After 1 min and 1 h, the dilutions were vortexed and 1 ml was pipetted into 2 petri dishes. Approximately 15 -- 20 ml of trypticase soy agar (TSA) without neutralizers was poured into the petri dishes. ### Neutralizer effectiveness A 1:10 dilution of each hand rub was produced by adding 1 gram of a hand rub to 9 ml of Butterfields buffer supplemented with the neutralizing agents. The dilution was vortexed and 1 ml of the dilution was discarded. Within 5 s, 300 -- 1000 CFU of the test strain was added to the dilution to yield 30 -- 100 CFU/ml. After 1 min and 1 h, the dilutions were vortexed and 1 ml of each was pipetted into two petri dishes. Approximately 15 -- 20 ml of TSA with neutralizing agents were poured into the petri dishes. ### Neutralizer toxicity A 1:10 dilution was produced by adding 1 ml of Butterfields phosphate buffer to 9 ml of Butterfields buffer supplemented with the neutralizing agents. The dilution was vortexed and 1 ml of the dilution was discarded. Within 5 s, 300 -- 1000 CFU of the test strain was added to the dilution to yield 30 -- 100 CFU/ml. After 1 min and 1 h, the dilutions were vortexed and 1 ml of each was pipetted into two petri dishes. Approximately 15 -- 20 ml of TSA containing neutralizing agents was poured into the petri dishes. ### Test material control A 1:10 dilution of each hand rub was produced by adding 1 ml of the challenge organism suspension to 9 g of the hand rub. The dilution was vortexed. After 1 min and 1 h, the dilutions were vortexed and 1 ml of each was pipetted into two petri dishes. Approximately 15 -- 20 ml of TSA without neutralizers was poured into the petri dishes. ### Culture conditions and incubation The plates were incubated at 36°C for 24 -- 48 h. All test were repeated an additional 2 times, for a total of three replicates. Efficacy test of hand rubs -------------------------- Fifteen volunteers participated in each experiment. Hands were washed for 1 minute with soft soap and dried with paper towels. A contamination fluid was prepared with E. coliATCC 11229 by growing the test organism in two test tubes containing 5 ml of tryptic soy broth (TSB) for 18 -- 24 h at 36 ± 1°C. These cultures were inoculated in two bottles with 1 l TSB, incubated at 36 ± 1°C for 18 -- 24 h and pooled. It contained between 2 × 10^8^and 2 × 10^9^CFU ml^-1^. The hands of the volunteers were immersed in the contamination fluid up to the mid-metacarpals for 5 s with fingers spread, and then allowed to air dry for 3 min which resembles a contamination of hands after patient care. Fingertips were rubbed for 1 min in a petri dish containing 10 ml of TSB (pre-values). Either 3 ml of a product or 2 × 3 ml of the reference alcohol (2-propanol, 60% v/v) were applied to the hands (cross-over for each volunteer). The rub-in period was 30 s for each product and 60 s for the reference alcohol. Immediately after the rub-in period, fingertips were rubbed again for 1 min in a petri dish containing TSB with and without neutralisers. For both reference and test procedures, the log counts from the left and right hands of each subject were averaged separately for pre-values and post-values. The arithmetic means of all individual log reduction factors (RF) were calculated. A product shall not be significantly less effective in comparison to the reference disinfection. Design and statistics --------------------- All three agents were tested in individual experiments. Test subjects were treated as paired groups, the same panel of subjects was used for all three experiments (hence paired). This design was justified with the knowledge that hands are artifically contaminated in each experiment and that the artificial contamination leads to highly reproducible pre-values \[[@B12],[@B13]\]. It is therefore very unlikely that the pre-value is influenced by a treatment of hands with a hand rub prior to the artificial contamination. In a first step, means were compared in each group (with and without neutralizing agents in the sampling fluid) in an analysis of variance (ANOVA; SPSS). In a second step, means were analyzed with the Wilcoxon-Wilcox test (Friedman analysis). A significance level of P = 0.1 is set in EN 1500 \[[@B6]\]. Results ======= Evaluation of neutralizing agents --------------------------------- On average, the viability of the test organism after 1 min and 1 h is characterized by a viable count of 88 ± 9 CFU/ml which is equivalent to a mean of 1.95 ± 0.04 log~10~CFU/ml. Exposure to the neutralized active agents brought about, on average, 88 ± 5 CFU/ml (1.94 ± 0.03 log~10~). There was no significant difference between the two (P = 0.97; two-sided t-test for paired samples). Exposure of the test organism to the neutralizing agents alone resulted in an average of 90 ± 7 CFU/ml (1.94 ± 0.03 log~10~). There was no significant difference between the two (P = 0.66). The combination of 0.5% lecithin and 4% polysorbate 20 was found to be not toxic to the test organism and to effectively neutralize ethanol and chlorhexidine gluconate. Efficacy of hand rubs --------------------- Without neutralizing agents in the sampling fluid, the reference treatment reduced the artificial contamination of hands by 3.7 log~10~, product A by 4.8 log~10~and product B by 3.3 log~10~(Table [1](#T1){ref-type="table"}). With neutralizing agents in the sampling fluid, the reference treatment reduced the artificial contamination of hands by 3.5 log~10~, product A by 2.7 log~10~and product B by 3.3 log~10~. A significant difference was found between the means of all three treatments in the group with neutralizing agents in the sampling fluid (P = 0.001; ANOVA) and in the group without neutralizing agents in the sampling fluid (P = 0.011). In comparison to the reference treatment, product B lead to a lower mean reduction of 3.3 log~10~(no neutralization in sampling fluids) and 3.3 log~10~(neutralization in sampling fluids), the differences, however, were not significant (P \> 0.1; Wilcoxon-Wilcox test). Without neutralizing agents in the sampling fluids, product A yielded a mean reduction of 4.8 log~10~which was significantly higher as compared to the reference treatment (P = 0.02). With neutralizing agents the mean reduction amounted to 2.7 log~10~which was significantly lower as compared to the reference treatment (P = 0.033). The total number of samples without detectable bacteria after treatment with product A was 9 out of 15 when tested without neutralizing agents in the sampling fluid, and 0 out of 15 with neutralizing agents (Table [1](#T1){ref-type="table"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Mean reduction ± SD of Escherichia colifrom artificially contaminated hands of 15 volunteers by reference treatment (2 × 3 ml of 60% iso-propanol for 60 s) or treatment with product A or product B (each 3 ml for 30 s); experiments differed only by the presence of valid neutralizers in the sampling fluid. ::: Preparation Active agent(s) and concentration(s) No neutralizers in sampling fluid P-value (vs. reference)\* Neutralizers in sampling fluid P-value (vs. reference)\* ------------------- ------------------------------------------------------ ----------------------------------- --------------------------- -------------------------------- --------------------------- ----------- ------- Product A Ethanol (61%, w/w) chlorhexidine gluconate (1%, w/w) 9 / 15 4.8 ± 1.5 0.02 0 / 15 2.7 ± 0.4 0.033 Reference alcohol Iso-propanol (60%, v/v) 0 / 15 3.7 ± 0.8 n.a. 0 / 15 3.5 ± 0.9 n.a. Product B Ethanol (85%, w/w) 0 / 15 3.3 ± 0.7 0.44 0 / 15 3.3 ± 0.7 0.11 ::: Discussion ========== For the first time we were able to show that lack of effective neutralizing agents in the sampling fluids leads to false positive efficacy assessments of an ethanol-based hand rub containing 1% chlorhexidine. This effect is explained by residual bacteriostatic or bactericidal activity of chlorhexidine gluconate because the other hand rub based on 85% ethanol without chlorhexidine gluconate was effectively neutralized by a 1:10 dilution alone. In addition, the experiments without neutralization in the sampling fluid showed that in 9 out of 15 samples with product A there was no bacterial growth pointing to residual bacteriostatic or bactericidal activity on the agar plates. With neutralizers in the sampling fluids, however, all 15 samples per experiment revealed countable numbers of colonies. With the reference alcohol alone and with product B not containing chlorhexidine gluconate, countable numbers were found in all 15 experiments, both with and without neutralizing agents in the sampling fluid. Only few studies with alcohol-based hand rubs containing chlorhexidine gluconate have been carried out with valid neutralization of the active agents. The efficacy of product A for surgical hand disinfection was recently evaluated according to prEN 12791. It was found to be significantly less effective compared with the reference treatment, both in the immediate and the sustained effect \[[@B15]\]. It was ensured that neutralization was effective according to the European test method \[[@B15]\]. Even after 3 hours, product A containing 1% chlorhexidine gluconate did not match with the efficacy of the reference alcohol n-propanol (60%, v/v) alone, indicating that 1% chlorhexidine did not really contribute to the overall efficacy. Another study was recently finished using the same test design and assuring effective neutralization. One preparation contained 70% iso-propanol and 0.5% chlorhexidine gluconate. It was less effective as compared to the reference alcohol n-propanol (60%, v/v) indicating once again that 0.5% chlorhexidine did not really contribute to the overall efficacy (Rotter ML et al. 2005; unpublished data). And in 1990, it was shown that any residual activity of chlorhexidine gluconate on hands can be destroyed simply by using an anionic soap which apparently neutralizes residual chlorhexidine gluconate \[[@B16]\]. These results further support doubts on the real benefit of chlorhexidine gluconate as a non-volatile active agent in hand hygiene. Neutralization must be effective in order to stop any residual bacteriostatic or bactericidal activity and at the same time it must be non-toxic to the test organism. An ASTM standard \[[@B14]\] and a European norm \[[@B17]\] have been published considering both elements. The test system of the European norm prEN 12054 has been shown previously to be suitable to determine the effectiveness and non-toxicity for standard neutralizing agents \[[@B5],[@B18]\]. We are not aware of any comparative studies for the European and US test methods to determine if one of the two methods may be superior to the other. Both methods include relevant test criteria and may be used for the assessment of effectiveness and non-toxicity for neutralizing agents. Conclusion ========== Lack of effective neutralization may yield false positive efficacy data for alcohol-based hand rubs containing chlorhexidine. The crucial step of neutralization seems to occur in the sampling fluid itself where any residual bacteriostatic or bactericidal activity should be stopped immediately after the preset application time. This is particularly important at short application times such as 30 s. Competing interests =================== The first author is paid employee of Bode Chemie GmbH & Co., Hamburg, Germany. Authors\' contributions ======================= GK designed the study, analysed the data and wrote the manuscript. MS and CH performed the experiments and participated in writing the manuscript. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2334/5/48/prepub>	PubMed Central	2024-06-05T03:55:59.809634	2005-6-20	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181814/", "journal": "BMC Infect Dis. 2005 Jun 20; 5:48", "authors": [ { "first": "Günter", "last": "Kampf" }, { "first": "Marc", "last": "Shaffer" }, { "first": "Corrine", "last": "Hunte" } ] }
PMC1181815	Background ========== Kocuriais a member of the Micrococcaceaefamily and consists of nine species. It was previously classified into the genus of Micrococcus, but was dissected from Micrococcusbased on phylogenetic and chemotaxonomic analysis \[[@B1]\]. The organism is widespread in nature and is frequently found as normal skin flora in humans and other mammals. Documented cases of infections due to Kocuria spp.are limited. The type species K. roseahas been reported to cause catheter-related bacteremia \[[@B2]\]. Another member of the genus, K. kristinae(previously known as Micrococcus kristinae), was first described in 1974 \[[@B3]\]. This organism is an aerobic, gram-positive coccus occurring in tetrads, and the majority of strains are non-pathogenic. Clinically similar to K. rosea, a single case of catheter-related bacteremia due to K. kristinaehas been reported in a patient with ovarian cancer \[[@B4]\]. Here we report the first case of K. kristinaeisolated from bile in a patient with acute cholecystitis. Case presentation ================= A 56-year old Chinese man, who had a known history of asymptomatic gallstones, presented with right upper quadrant abdominal pain for five days associated with fever. Laboratory investigations showed neutrophilia, but the liver function test was normal. Ultrasound examination of the abdomen revealed distended gallbladder associated with multiple gallstones, prominent intrahepatic ducts and enlarged lymph nodes at the porta hepatitis region. Laparoscopic cholecystectomy performed for a diagnosis of acute cholecystitis showed distended and gross thickened gallbladder and omental adhesions. The bile was turbid and two stones were found impacted at the Hartmann\'s pouch. The cystic duct was normal. The patient developed post-operative fever and intravenous levofloxacin at a dosage of 500 mg daily was started as empirical treatment. Bile culture subsequently yielded a pure growth of K. kristinae(see microbiology diagnosis). Fever resolved readily after levofloxacin therapy, which was continued orally at the same dosage for a total duration of 14 days. He made an uneventful recovery. Microbiological diagnosis ========================= Culture of bile from gall bladder was performed with sheep blood agar, MacConkey agar and chocolate agar. The plates were incubated at 35°C for 48 hours. Anaerobic culture was performed using Schaedler blood agar and incubated at 35°C for 48 hours. Gram-positive cocci arranged in tetrads were isolated from pale cream colonies after two days incubation. The organism was non-hemolytic, catalase positive, coagulase negative and non-motile. Identification was performed using Biomerieux ID32 Staph ATB system and BD Phoenix PMC/ID-13 system. The isolate was identified as Kocuria kristinaewith a probability of identification of 99.9 % and confidence value of 99% for the ATB system and Phoenix system respectively. Identification of the isolate was confirmed using 16S rRNA sequencing (MicroSeq™, Applied Biosystems, USA), as misidentification of coagulase-negative staphylococci as Kocuriaspecies has been described \[[@B5]\]. Analysis of nucleotide sequence with BLAST programs showed 100% DNA sequence homology with K. kristinae. Antibiotic sensitivity test was performed using the disc diffusion method according to Clinical and Laboratory Standards Institute (formerly NCCLS) guidelines for Staphylococcus. The isolate was sensitive to penicillin, cloxacillin, erythromycin, clindamycin, linezolid, trimethoprim/sulfamethoxazole, vancomycin and levofloxacin. Discussion ========== Members of the genus Micrococcusare found as normal flora of the skin and mucosa. Infections related to Micrococcus spp.are uncommon but are recognized, especially in immunocompromised patients with underlying diseases. The organism M. luteushas been described as the causative agent in meningitis \[[@B6]\], intracranial abscess \[[@B7]\], arthritis \[[@B8]\], pneumonia \[[@B9]\] and catheter-related sepsis in patients undergoing hemodialysis \[[@B10]\] or leukaemia treatment \[[@B11]\]. Other infections associated with Micrococcusand related organisms include continuous ambulatory dialysis peritonitis \[[@B12]\], endocarditis \[[@B13]\] and infection of cerebrospinal fluid shunts \[[@B14]\]. More recently, Micrococcus spp.is implicated in central venous catheter infection in patients with pulmonary hypertension receiving continuous epoprostenol infusion \[[@B15],[@B16]\]. Kocuriais previously classified as Micrococcusand, being inhabitants of the skin, it is not surprising that K. roseaand K. kristinaehave been incriminated as pathogens causing catheter-related bacteremia \[[@B2],[@B4]\]. Misidentification of coagulase negative staphylococcus as Kocuriausing standard biochemical analysis is not uncommon due to phenotypic variability \[[@B5]\]. The utilization of genotypic assay such as 16s rRNA is required to confirm species identity as in the present case may be required, particularly for unusual clinical scenarios. The K. kristinaeorganism isolated in our patient was sensitive to most of the commonly used antibiotics. A report in the literature on 219 strains of Kocuriaand Micrococcusshows that most strains are sensitive to doxycycline, cetriaxone, cefuroxime, amikacin, and amoxicillin with clavulanic acid, but most are resistant to ampicillin and erythromycin \[[@B17]\]. The duration of therapy in general depends on site and severity of infection. If bacteremia is present or likely, duration of 10 -- 14 days is commonly employed. Bile cultures are sterile in 25 -- 50% of acutely inflamed gallbladders. Bacterial infection in acute cholecystitis is usually a secondary event, and is most commonly due to enteric bacteria. A recent study from the Netherlands on microbes isolated from bile after cholecystectomy \[[@B18]\] showed a predominance of Escherichia coli, followed by Klebsiella spp.and Streptococcus spp.Significantly, two studies on infective complications after open \[[@B18],[@B19]\] and laparoscopic cholecystectomy showed no correlation between positive bile culture and post-operative infection. These findings, together with the lower incidence of wound infections after laparoscopic cholecystectomy, would cast doubt on the use of routine antibiotics prophylaxis as recommended for biliary surgery. However, the development of post-operative fever in our patient necessitated the use of empirical antibiotic cover. Levofloxacin used in the present case is a third generation fluoroquinolone with a broad spectrum of antibacterial activity, which has been shown to give adequate serum and gallbladder tissue concentrations in biliary tract surgery \[[@B20]\]. We describe the first case of K. kristinaeinfection associated with acute cholecystitis. Interestingly, a related skin commensal Staphylococcus aureushas been recognized as the primary pathogen in unusual cases of acute cholecystitis \[[@B21]\]. S. aureusassociated acute cholecystitis might be encountered in the clinical setting of bacteremia due to infective endocarditis or nosocomial acquisition in patients with chronic medical conditions \[[@B21]\]. While unfortunately a blood culture was not taken in our patient, the presence of gallstone, good pre-morbid status and prompt resolution of fever after antibiotics would point against a possible endovascular focus of infection. Conclusion ========== Although previously regarded as an innocuous microorganism, there have been a number of recent reports describing the association between Kocuria spp.and infectious diseases. The complete clinical spectrum of infections caused by this group of bacteria will be more apparent after the report of more cases. The physician should not therefore underestimate the importance of K. kristinaewhen isolated from clinical specimens. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= ESKM, CLPW and KTWL carried out the laboratory studies of the patient. WCY performed the 16s rRNA sequencing. ACWC performed the operation and provided clinical details. ECHC followed up the patient and obtained consent from the patient to publish this case report. ESKM and CLPW drafted the manuscript. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2334/5/60/prepub> Acknowledgements ================ Written consent was obtained from the patient for publication of this case report.	PubMed Central	2024-06-05T03:55:59.811039	2005-7-19	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181815/", "journal": "BMC Infect Dis. 2005 Jul 19; 5:60", "authors": [ { "first": "Edmond SK", "last": "Ma" }, { "first": "Chris LP", "last": "Wong" }, { "first": "Kristi TW", "last": "Lai" }, { "first": "Edmond CH", "last": "Chan" }, { "first": "WC", "last": "Yam" }, { "first": "Angus CW", "last": "Chan" } ] }
PMC1181816	Background ========== Enteric bacteria classified as Salmonellaare responsible for a wide variety of illnesses, including typhoid fever, food poisoning, gastroenteritis and septicaemia. The genus Salmonellacomprises two species, namely Salmonella entericawhich can be subdivided into more than 2 300 serovars, and Salmonella bongori. It has been proposed that the evolution of the genus has progressed in three major phases \[[@B1]\]. In the first phase Salmonelladiverged from E. coliby acquiring Salmonellapathogenicity island 1 (SPI 1) which encodes virulence factors required by Salmonellafor the intestinal phase of the infection \[[@B2]\]. The formation of the two species S. entericaand S. bongoriis considered the second phase in the evolution in which Salmonellapathogenicity island 2 (SPI 2) was acquired. Its role in pathogenicity is yet to be established but its relevance has been demonstrated for the development of systemic infection \[[@B3]\]. The last stage of the evolution of Salmonellais considered to be the formation of S. entericasubspecies I, which enabled a dramatic expansion of host specificity of the species. While the rest of the subspecies of S. entericaare adapted to heterothermic vertebrates, subspecies I strains are capable of colonising mammalian and avian hosts \[[@B4]\]. S. entericasubspecies I is routinely subdivided into serovars on the basis of the expression of three surface antigens (Ag\'s), the somatic O Ag, the flagella H1 and H2 Ags, and the capsular Vi Ag, according to the Kauffmann-White scheme \[[@B5]\]. Although the serovars are very closely related, they have different host ranges and cause different disease signs. Serovar Typhimurium is a generalist that infects a wide range of animals (humans, wild rodents, poultry, pigs, cattle). Some serovars are host specific, infecting only one animal host e.g. serovar Pullorum infects only poultry and serovar Typhi infects only humans \[[@B4]\]. The genetic and molecular basis for the different host ranges and host specificities of the serovars of S. entericasubspecies I are not clear. The evolution and acquisition of the pathogenicity islands of Salmonellaleading to the formation of subspecies I and allowing for the use of homeothermic animals as a host is extensively studied. However, it is not clear how host-restricted serovars appeared and whether they acquired different virulence determinants compared to the host generalists. For serovar Pullorum the process of host adaptation has been accompanied by point mutations resulting in loss of the ability to mediate mannose sensitive agglutination (MHSA) and to express flagella \[[@B6],[@B7]\]. Strains of serovars Typhimurium isolated from avian hosts also lack mobility and MHSA \[[@B4]\], and these changes also result in a 100-fold reduction in the virulence of serovar Typhimurium in mice \[[@B7]\]. However, the highly host-specific phenotype of serovar Pullorum cannot be explained by these point mutations alone. For instance, serovar Typhimurium initiates disease by entering the Peyer\'s patches, where it invades the circulating lymphoid cells \[[@B8]\]. In contrast, serovar Pullorum is incapable of entering the Peyer\'s patches, cannot survive and multiply in the cells of the mouse reticuloendothelial system, and is internalised by the murine macrophages by a mechanism different to the one of serovar Typhimurium \[[@B9]-[@B11]\]. Initial hybridisation studies showed that the serovars of S. entericashare \>90 % of their DNA content \[[@B12]\]. The comparison of their genomes revealed that despite their similarity each serovar has many insertions and deletions relative to the other serovars, which vary in size from 1 to 50 kB \[[@B13]\]. However, the differences observed at the DNA level have so far not been related to protein expression. It is of great importance to determine if the differences observed at the genomic level are in fact translated into proteins as has been reported by Taoka and co-workers \[[@B14]\], who found that the majority of the horizontally transferred genes in the genome of E. coliare not translated into proteins, presumably because they are inadequate for the translational machinery of the cell. Another approach for the identification of serovar-specific factors involved the introduction of virulence-associated DNA regions of host generalists into host-specific serovars, to expand their host range. This approach was unsuccessful, suggesting that multiple genes are responsible for the host-restricted phenotypes \[[@B15]\]. Therefore, scientific approaches which enable global measurements of gene expression on a genome-wide scale would give a better understanding of the differences in gene-regulation patterns between serovars with host restricted and host-generalist phenotypes. In this study we used a standard proteomic approach combining two dimensional gel electrophoresis (2D GE) and tandem mass spectrometry (LC/MS/MS) to compare the expression patterns of isolates of the highly host-restricted serovar Pullorum and the host generalist serovar Typhimurium. The protein expression patterns of S. entericaserovar Typhimurium have been extensively studied and an annotated reference map of the cell envelope proteins of serovar Typhimurium has been published \[[@B16],[@B17]\]. Over 800 proteins expressed by serovar Typhimurium have been recently identified using two-dimensional HPLC-MS, including some potentially associated with multiple antibiotic resistance \[[@B18]\]. Changes in the expression pattern of serovar Typhimurium have been monitored when grown in media with low pH \[[@B29],[@B30]\] and when exposed to bile \[[@B21]\], and an attempt has been made to identify proteins with increased levels of expression when grown intracellularly \[[@B22]\]. However, the protein expression pattern of serovar Pullorum is yet to be determined and, more importantly, comparative analysis of strains with different sero-specificity have not been performed to date. A comparison of the proteome of a serovar that is not host-specific to the serovar that is highly host-specific may help reveal what factors enabled Salmonellato overcome species barriers and adapt to new hosts and, ultimately, should give an insight into the process of host adaptation and the emergence of new pathogens. Results and discussion ====================== An annotated reference map of S. entericacytosolic proteins was created using strain 74 of serovar Typhimurium obtained from the National Collection of Type Cultures (London, UK). Seven hundred and seventy-one spots were detectable after staining with SYPRO Ruby, of which 233 were identified by LC/MS/MS. The proteins identified represented 200 open reading frames (ORF\'s), which constitute 4.4 % of the 4558 protein coding sequences predicted in the genome of Salmonella Typhimurium LT2 \[[@B23]\]. Each protein was represented, on average, by 1.165 spots on the gel. Twenty-four proteins were detected in more than one isoelectric form. Aldehyde dehydrogenase B, glyceraldehyde-3-phosphate dehydrogenase and an oligopeptide-binding protein precursor were each present in four isoforms. For most of these the pI difference was relatively small, but differences of around two to three pI units were also observed. In the case of transaldolase A and thiol disulphate, interchange protein forms were observed which differed in molecular mass as well as pI. In the reference map of Salmonellamany of the substrate-binding proteins identified appeared as a series of isoelectric forms, which was indicative of posttranslational modifications or amino acid substitutions. Such diversity of isoforms has been described for other bacterial proteomes. The ratio of the number of spots to the number of ORFs has been reported to be 1.4--2 in E. coli\[[@B24],[@B25]\], 1.6 for Chlamydia pneumoniae\[[@B26]\] and 1.42 for Staphylococcus aureus\[[@B27]\]. It should be noted that, in most cases, the multiplicity of the pI exhibited by a protein was not accompanied by a significant change in the molecular weight, suggesting that the generation of isoelectric forms may involve modifications such as phosphorylation \[[@B28]\], methylation \[[@B29],[@B30]\] or deamidation \[[@B31]\], rather than an introduction of high molecular weight groups such as long glycan chains. Some of the isoforms observed appeared to be serovar-specific. The isoforms of the enzyme superoxide dismutase A differed in their pI point but had identical molecular weights. One of the isoforms was expressed by all isolates of serovar Typhimurium and the second isoform was characteristic for serovar Pullorum. Similarly, serovar Pullorum expressed three different forms of the lysine, arginine, ornithine-binding periplasmic protein precursor while the expression maps of the isolates of Typhimurium contained only two isoforms. However, the exact nature of these putative posttranslational modifications and their physiological relevance remains to be elucidated, and the possibility that some isoforms occur during sample preparation cannot be ruled out. Some of the most intense spots on the gels were attributable to metabolic proteins, including the full set of Krebs cycle enzymes and all the enzymes of glycolysis with the exception of glucokinase, which catalyses the first step of the glycolytic pathway. The gene for glucokinase is present in the genome of SalmonellaTyphimurium and as a glycolytic enzyme it is localised in the cytosol. The predicted molecular weight and pI point from the genome sequence is 34 564 Da and 5.83 respectively. Therefore theoretically the enzyme should be detected using the 2D GE protocol described. The failure to detect this enzyme may be due to its low expression or, possibly, because S. entericaobtains glucose 6-phosphate via a different route. A membrane bound enzyme complex, namely phosphoenol pyruvate phosphotransferase system (PTS), which couples the transport of sugars through the cell membrane with their phosphorylation, has been extensively studied in E. coliand Salmonella\[[@B32]\]. Two of the components of this system, enzyme I and a glucose-specific component IIA, were identified in the profile of serovar Typhimurium (spot 123 and 46, respectively). It is also possible that posttranslational modifications could have altered the pI of the glucokinase to a value outside the pH range of the IPG strip used. Several of the enzymes of the pentose phosphate pathway were also identified, including glucose 6-phosphate dehydrogenase, transaldolase, transketolase and phosphogluconate dehydrogenase, which suggests that this pathway is also active under these growth conditions. Glucose 6-phosphate dehydrogenase also participates in the Entner Duodoroff pathway. However, since no other components of this pathway were detected it was likely that the pathway was repressed while glycolysis via the Embden Meyerhof pathway was active. Many of the proteins identified were associated with protein biosynthesis, including elongation factors Tu, Ts, G and P, and tRNA synthetases for arginine, asparagine, aspartic acid, proline, glutamine, glycine, histidine, lysine and tyrosine. Chaperone proteins identified included dnaK, GroL, GroS and HtpG. Five ribosomal proteins (30S ribosomal protein S6, S1, S2 and 50S ribosomal proteins L7/L2, L9/L12) were detected, but others remained unobserved, presumably because of their high basicity \[[@B27]\]. An important subset of the proteins detected are involved in defence against oxidative damage, including two superoxide dismutases (SodA and SodF), alkyl hydroperoxide reductase (22 kDa subunit), a putative peroxidase, an oxidoreductase (ucpA), a putative catalase and catalase HPI. Under anaerobic conditions expression levels of SodA and the alkyl hydroperoxide reductase were reduced, whereas SodF was upregulated (data not shown). We also identified thirteen hypothetical proteins for which no function has yet been assigned. Comparative analysis of the expression patterns of serovars Typhimurium and Pullorum ------------------------------------------------------------------------------------ Comparison of the expression maps of serovars Typhimurium and Pullorum revealed that, despite the similarities in the expression patterns there was a high degree of variation amongst clinical and laboratory isolates, even when they were from the same serovar. This finding corresponds to the comparative analysis of the serovars of Salmonellaperformed using a DNA microarray \[[@B33]\], which indicated that classification of Salmonellastrains into genomovars based on the similarity in their genome sequences is more appropriate but does not always correspond to serovar. A similar selection of proteins were expressed by all isolates and differences observed were mainly in the level of expression but a few characteristic proteins were also present. The comparison of the expression patterns revealed that there was no variation in expression of enzymes involved in glycolysis, Krebs cycle and the pentose phosphate pathway, as well as the chaperones and the proteins involved in protein biosynthesis. The majority of the differences observed were not serovar-specific e.gthe laboratory reference strain of serovar Typhimurium overexpressed several substrate-binding periplasmic proteins including maltose binding protein precursor, oligopeptide binding protein precursor and ABC superfamily dipeptide transport protein, while the clinical isolates of the same serovar expressed these proteins at very low levels. Such differences may be due to the different conditions that the laboratory strains and clinical isolates are subjected to, with the former gradually adapting to the laboratory while undergoing continuous subculturing. These may have favoured changes which otherwise would not have happened in non-laboratory conditions. A similar observation was reported for Helicobacter pyloriby Hynes and co-workers \[[@B34]\]. Therefore it is becoming increasingly apparent that the expression profiles of both laboratory reference strains and clinical isolates should be considered when characterising the proteome of any microbial pathogen \[[@B35]\]. Despite the high level of variation amongst the isolates characterised, several serovar-specific factors were also observed. Two transport proteins, sulphate (sbp) and thiosulphate (cysP) binding proteins, were detected as prominent spots in the profile of serovar Pullorum but were not expressed by any of the Typhimurium isolates (Figure [3](#F3){ref-type="fig"}). Several strains representing other serovars (Enteritidis, Choleraesius and Dublin) were also tested and none of them expressed these two transport proteins, suggesting that under the growth conditions utilised they are characteristic only for serovar Pullorum. Furthermore, serovar Pullorum showed a significantly higher (p = 1.618 × 10^-4^) level of expression of the enzyme cysteine synthase (cysK) in comparison to serovar Typhimurium (Figure [3](#F3){ref-type="fig"}). It has been reported that E. coliand S. entericaserovar Typhimurium harbour a sulphate transport system which is part of the cysteine regulon and is controlled in parallel with cysteine-biosynthetic enzymes \[[@B36]\]. CysP is part of this system and its expression in serovar Typhimurium is induced when grown under sulphur limitation \[[@B37]\]. However, it is possible that in serovar Pullorum this protein is constitutively expressed. Although the sulphate binding protein (Sbp) has been extensively studied \[[@B37]-[@B39]\] its functional relationship with cysP is yet to be elucidated. Their increased level of expression correlated with the increased level of cysteine synthase (CysK), which is controlled by the same operon. It is likely therefore that there is a functional association between these three proteins which may be relevant to the host adaptation of serovar Pullorum. Another difference in the expression patterns was the absence of the protein yghA in the expression map of serovar Pullorum. YghA has been annotated as a putative oxidoreductase but its function is not clear. The characterisation of the genome of serovar Pullorum revealed a high level of genomic plasticity caused by a large insertion disrupting the balance of the genome \[[@B40]\]. Further studies should be aimed at determining if the gene for this hypothetical protein is present in the genome of serovar Pullorum or if the difference in the expression of yghA is a result of differential regulation on translational or post-translational level. Conclusion ========== Microorganisms vary in their mechanisms of survival, some remaining at a particular site where nutrients are more favourable, while others are metabolically more versatile and disseminate more readily and consequently manifest different disease signs. Such differences in the capacity to spread and adapt to different conditions can be observed amongst the serovars of Salmonella entericasubspecies I \[[@B41],[@B42]\]. The existence of host generalist and host-adapted variants of this species presents a unique opportunity to study the mechanisms defining the process of host adaptation. During adaptation to a new environment the metabolic activity of the cell can be expected to change. This was confirmed by the differential expression of the two substrate transport proteins (sulphate and thiosulphate binding protein) which were characteristic for the host-adapted serovar Pullorum. It can be speculated that the high sulphurous content of the egg, has favoured the increased expression of sulphate transporting proteins in serovar Pullorum. The sulphate ions may subsequently be reduced and used for the synthesis of cysteine \[[@B43]\] which corresponds to the elevated level of expression of the enzyme cysteine synthase reported in this study. The differences observed between serovars Typhimurium and Pullorum suggest that there are variations in the expression patterns even between closely related bacteria. In this case 2D GE combined with MS proved a useful tool for identifying proteins differentially expressed in serovars with different host specificity and pathogenic potential. Future studies are now in progress to compare the proteome of a large number of serovars that specifically affect man. Methods ======= Bacterial strains ----------------- The profiles of seven independent isolates of serovar Typhimurium and two of serovar Pullorum were used for the compative analysis. Strain 74 of serovar Typhimurium and 10704 of Pullorum were obtained from the National Collection of Type Cultures, (Health Protection Agency, London, UK). Strains A01, C01 of serovar Typhimurium and B52 of serovar Pullorum were obtained from the SalmonellaReference Collection at the University of Calgary, Canada. Four additional clinical isolates of serovar Typhimurium (strains 23, 56, 204 and 227) were supplied by the Department of Risk Research in VLA-Weybridge, UK. Protein extraction ------------------ Bacteria were cultured on Columbia Blood Agar (Oxoid, Basingstoke, UK) for 18 h at 37°C under standardised conditions. Cells harvested from three plates were lysed in 100 μl of 0.3% w/v SDS, 200 mM DTT and 50 mM Tris (pH = 8.0), then digested with DNase I and RNase A. The samples were vortexed with 0.3 g of glass beads (\< 105 μm diameter) and homogenised five times for 1 min using a Mickle cell disintegrator (Mickle Laboratory Engineering Co. Ltd, UK) with 1 min cooling on ice after each homogenisation period. Cell debris was removed by centrifugation (21 000 × g for 30 min at 4°C) and proteins were mixed with urea/thiourea based rehydration solution \[[@B44]\]. 2D GE ----- Samples containing a total of 150 μg of protein were loaded into 18 cm IPG strips (pH 3--10 NL, Amersham Biosciences, UK) by in-gel rehydration \[[@B45]\]. IEF was performed using an Investigator 5000 apparatus (Genomic Solutions, USA) for 24 h, (85 000 Vh at a maximum voltage of 5000 V and a maximum current of 110 μA.) The focused IPG strips were equilibrated twice (2 × 30 min) in DTT and iodoacetamide as described by Görg et al., \[[@B46]\]. Second dimension electrophoresis was performed on 10 % Duracryl^®^gels (Proteomic Solutions, France) using Tris/Tricine buffer chemistry as recommended by Fountoulakis et al., \[[@B47]\], for 5 h, at a maximum voltage of 500 V and maximum power of 20 000 mW per gel. Spots were visualized with SYPRO Ruby^®^(Molecular Probes, UK) then counter-stained with colloidal Coomassie G \[[@B48]\] prior to manual spot excision. Reproducibility and data analysis --------------------------------- To perform the comparative analysis a total of twelve profiles of serovar Typhimurium were analysed, including four replicates of the profile of strain 74 and duplicate profiles of strains 204 and 227 and additional profiles of strains 23, 56, A01 and C01. Five profiles of serovar Pullorum were used for the comparison, three of strain 10704 and two of strain B52. The intensity values of the protein of interest were estimated using ProteomWeaver software (Definiens, Germany) and the values were subjected to the Students T-test. Differential expression was reported only when the intensity values were found significantly different (p \< 0.05). Trypsin digestion and LC/MS/MS ------------------------------ Excised spots were digested in situwith trypsin, using an Investigator ProGest robotic digestion system (Genomic Solutions, Huntington, UK) as previously described \[[@B49]\]. Tandem electrospray mass spectra were recorded using a Q-TOF hybrid quadrupole / orthogonal acceleration time of flight spectrometer (Micromass, Manchester, UK) interfaced to a Micromass CapLC capillary chromatograph. Samples were dissolved in 0.1% v/v formic acid, and injected onto a Pepmap C18 column (300 μm × 0.5 cm; LC Packings, Netherlands), The capillary voltage was set to 3,500 V, and data-dependent MS/MS acquisitions were performed on precursors with charge states of 2, 3 or 4 over a survey mass range of 500--1300. The collision voltage was varied between 18 and 45 V depending on the charge and mass of the precursor. Database searching parameters ----------------------------- Proteins were identified by correlation of uninterpreted tandem mass spectra to entries in SwissProt/TREMBL, using ProteinLynx Global Server (Versions 1.1, Micromass). No taxonomic, mass or pI constraints were applied. One missed cleavage per peptide was allowed, and the fragment ion tolerance window was set to 100 ppm. Carbamidomethylation of cysteine was assumed, but other potential modifications were not considered in the first pass search. All matching spectra were reviewed manually, and in cases were the score reported by ProteinLynx global server was less than 100, additional searches were performed against the NCBI nr database using MASCOT, which utilizes a robust probalistic scoring algorithm \[[@B50]\]. Where identifications were based on a single matching peptide the sequences were confirmed by manual sequencing using the MassLynx program Pepseq. Measured parent and fragment masses were typically within 0.03 Da of their calculated values. Abbreviations ============= Ag antigen DTT dithiothreitol HPLC high performance liquid chromatography IEF isoelectric focusing IPG immobilised pH gradient LC liquid chromatography MS/MS tandem mass spectrometry SDS sodium dodecyl sulphate SPI Salmonellapathogenicity island ORF open reading frame Tris tris (hydroxymethyl) aminomethane 2D GE two dimensional gel electrophoresis Authors\' contributions ======================= VE carried out the protein work, analysed the 2D GE profiles and drafted the manuscript. RW and SB performed the mass spectrometry identification of the proteins. VE, HNS and SEG participated in the design of the study. HNS conceived and coordinated the study. All authors read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Protein identities in the reference map of Salmonella Typhimurium obtained using LC/MS/MS. The table contains the list of proteins identified in the reference map of Salmonella Typhimurium accompanied by their spot number on the gel image (Figure [1](#F1){ref-type="fig"}) and by their unique SWISS Prot identifier. The amino acid sequences of the matching peptides determined using LC/MS/MS are also included. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ V. Encheva and H.N. Shah wish to acknowledge a grant from Dr Hadwen Trust for Humane Research while R. Wait thanks the arthritis research campaign and the Medical Research Council for financial support. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Annotated reference map of the cytosolic proteins of S. entericaserovar Typhimurium. Separation was performed with 18 cm pH 3--10 NL IPG strips and 10 % Duracryl gels. Spots were visualised with SYPRO Ruby and imaged with a Typhoon Scanner (Amersham Biosciences, UK). The labelled spots were excised and analysed by LC/MS/MS, ([Additional file 1](#S1){ref-type="supplementary-material"}). The empty circles represent spots detected only in the profile of serovar Pullorum. ::: ![](1471-2180-5-42-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Annotated map of the cytosolic proteins of S. entericaexpressed in more than one mobility form (isoform). A total of 21 proteins were detected as more than one spot. The identities of the proteins are presented in ([Additional file 1](#S1){ref-type="supplementary-material"}). Each protein spot is outlined with a red circle and the isoforms are connected with red arrows. The higher number of isoforms (4), were detected for aldehyde dehydrogenase B (spot 10), oligopeptide binding protein precursor (spot 20) and glyceraldehyde 3-phosphate dehydrogenase (spot 6). The majority of the isoforms differed slightly in pI point but in some cases (e.g spot 13 and 11) differences in the molecular weight were also observed. ::: ![](1471-2180-5-42-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Comparison of the protein expression of serovars Typhimurium (A and B) and Pullorum (C and D). The strains of serovar Pullorum expressed sulphate (Sbp) and thiosulphate binding protein (cysP) which were absent from the profile of Typhimurium. The enzyme cysteine synthase (cysK) was detected in the profile of both serovars but showed two fold higher expression in serovar Pullorum. The hypothetical oxidoreductase yghA was present in the profiles of serovar Typhimurium but absent from Pullorum. Mdh and gapA were present in all profiles and were used as reference spots in this comparison. The (x) symbols correspond to protein spots missing from the corresponding area of the gel. ::: ![](1471-2180-5-42-3) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Cytosolic proteins of serovar Typhimurium detected in more than one isoform. ::: Spot number Protein Number of isoforms Gene name ------------- ----------------------------------------------------------- -------------------- ----------- 1 2,3 bisphosphoglycerate dependent phosphoglycerate mutase 3 gpmA 2 DnaK suppressor protein 2 dksA 3 Enolase 2 eno 4 Glutamate dehydrogenase 2 gdhA 5 Glutamine binding periplasmic protein precursore 2 galU 6 Glyceraldehyde 3-phosphate dehydrogenase 4 gapA 7 Lysil tRNA synthetase 2 lysS 8 Lysil arginine ornithine binding periplasmic protein 3 argT 9 Maltose binding protein precursor 2 malE 10 Aldehyde dehydrogenase B 4 aldB 11 Superoxide dismutase F 2 sodF 12 Superoxide dismutase A 2 sodA 13 Thiol:disulphate interchange protein 2 dsbA 14 Transketolase 2 STY3236 15 Phosphoglycerate kinase 2 pgk 16 Putative NAD dependent aldehyde dehydrogenase 2 STM 1627 17 Dihydrolipoamide dehydrogenase 2 lpdA 18 ABC superfamily dipeptide transport protein 2 dppA 19 Silent usher protein precursor 2 ushA 20 Periplasmic oligopeptide binding protein precursor 4 oppA 21 Acetyl CoA synthase 2 acs 22 Glutamate aspartate binding periplasmic protein precursor 3 gltI :::	PubMed Central	2024-06-05T03:55:59.812266	2005-7-18	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181816/", "journal": "BMC Microbiol. 2005 Jul 18; 5:42", "authors": [ { "first": "Vesela", "last": "Encheva" }, { "first": "Robin", "last": "Wait" }, { "first": "Saheer E", "last": "Gharbia" }, { "first": "Shajna", "last": "Begum" }, { "first": "Haroun N", "last": "Shah" } ] }
PMC1181817	Background ========== Osteoarthritis (OA) is a slowly developing articular disease, characterized mainly by cartilage degeneration, which is reflected clinically by a gradual development of joint pain, stiffness, and loss of full range of movement. OA is the most common disease to affect synovial joints, being a major cause of musculoskeletal pain, reduced quality of life and disability. About 40--60% of patients with radiological osteoarthritic changes suffer from clinical symptoms of pain, stiffness and loss of mobility \[[@B1]\], and around 55% of patients with OA report pain as the worst aspect of the disease \[[@B2]\]. OA is strongly associated with ageing and, with an increasing elderly population, of major socioeconomic importance. Current treatments for OA include a wide range of non-pharmacological, pharmacological and surgical options, although evidence to support their effectiveness is variable and there are no curative treatments. Therapies focus on reducing symptoms such as pain and stiffness, and minimalizing functional limitation and disability \[[@B3]\]. However, pain control is the primary aim of treating patients with OA and, in evaluating symptoms, pain should be the primary outcome variable \[[@B4]\]. Non-steroidal anti-inflammatory drugs play an important role in the pharmacological management of OA \[[@B5]\]. However, their lack of efficacy or potential toxicity may limit their use, in particular, the withdrawal of certain Cox-2 inhibitors has restricted the choice of therapies \[[@B3],[@B5]-[@B7]\], and problems of persisting pain remain. Pain from OA may be caused by factors other than inflammation \[[@B8]\], therefore the logical next step in the treatment of OA-related pain is the use of strong opioids. Within a management programme aimed at improving physical and social function, guidelines recommend their use when other appropriate therapies have failed to provide adequate pain relief over a reasonable period of time \[[@B9]-[@B13]\]. Transdermal fentanyl (TDF), providing systemic delivery of fentanyl at a constant rate for 72 hours \[[@B14]\], has been shown to be effective in controlling pain and improving some quality of life parameters for people with chronic non-malignant pain \[[@B15]-[@B18]\]. The efficacy of opioids in controlling pain in patients with OA has been demonstrated in three randomized controlled trials \[[@B19]-[@B21]\]. Moreover, a prospective study to investigate the efficacy and tolerability of TDF in 243 patients with severe OA pain of the knee and/or hip demonstrated significant reductions in pain at rest and on movement, and provided evidence of functional improvement \[[@B22]\]. Very few patients needed doses higher than the 25 μg/hr starting dose after 30 days of treatment \[[@B22]\]. The present study was undertaken to evaluate the utility and safety of TDF for the treatment of pain associated with RA or with OA of the knee or hip, which was not adequately controlled by NSAIDs, Cox-2 inhibitors, paracetamol or weak opioids at optimal doses. As it was an open-label study, it was not designed to prove efficacy of the treatment but to investigate practical aspects of opioid therapy, such as the usefulness of a test-period or concomitant use of antiemetic therapy and to serve as a pilot for a double-blind trial in this patient population. A test period for evaluation of a patient\'s response to opioid treatment is recommended \[[@B10]\] and in this study lasted 4 weeks. Results from the total study population and from patients with RA will be reported elsewhere. This paper reports the effects of TDF on pain and functioning in patients with OA of the knee or hip. Methods ======= Patient selection ----------------- All participants were outpatients requiring supplementary analgesia because of moderate or severe pain, which was not adequately controlled with paracetamol, NSAIDs, Cox-2 inhibitors or weak opioids (e.g. tramadol or codeine). Because non-opioid analgesia is not always taken at sufficient doses to achieve pain control \[[@B24]\] the trial employed a run-in period during which analgesia was optimised. Patients who still had moderate or severe pain at the end of this period could enter the main treatment phase of the study. Patients had to be over 50 years old, have OA of the knee or hip and meet the OA criteria of the ACR \[[@B23]\]. They had to have radiographic evidence of OA, and be waiting for hip or knee replacement as indicated by an orthopaedic surgeon. If participants were taking corticosteroids and/or NSAIDs they had to have received a stable dose for at least three months before screening and expect to remain on a stable dose for the duration of the trial. Patients were excluded from the study if they had received regular treatment with a strong opioid (e.g. morphine) or had received more than the maximum recommended dose of weak opioids or other analgesics in the four weeks before the study. Strong opioids other than fentanyl, supplementary weak opioids or other treatments that might alter the degree or nature of pain could not be started during the study. Patients were excluded if they had continuous pain of non-arthritis origin, or had undergone surgery/arthroscopy within 3 months, intra/peri articular injections for arthritis pain (e.g. steroid injection) within 6 weeks, or arthrocentesis within 4 weeks of the study start. Study design ------------ Screened patients satisfying the selection criteria each gave written informed consent before inclusion in this international, open, prospective trial. The study was carried out in accordance with the latest revision of the Declaration of Helsinki and Good Clinical Practice and was approved by independent local ethics committees. During the one-week run-in period, non-opioid analgesic treatment was increased to the maximum tolerated or maximum recommended dose, while weak opioids were kept stable. All patients with pain control rated as poor or very poor on a 5-point scale at the end of the run-in period started treatment for 28 days with transdermal fentanyl (TDF, Durogesic^®^) at a dose of 25 μg/h. Patches were replaced every 72 hours (3 days). Previous non-opioid analgesia was continued and kept stable, but weak opioids were discontinued. The dose of TDF could be increased in steps of 25 μg/h every 72 hours (days 3,6 and 9) if required (no maximum dose specified) until adequate pain control was achieved. After 28 days or if necessary for other reasons, e.g. if side effects occurred or the treatment was not effective, a similar downward titration regimen was employed. Paracetamol 500 mg tablets were provided for supplementary analgesia and could be used in doses of up to 4 g/day. With the exception of paracetamol, other non-opioid analgesics were kept stable and no short acting opioids were added during down titration. Metoclopramide 10 mg three times a day was given to all patients for the prevention of nausea and vomiting during the first week of treatment, after which it was taken as needed. Assessments ----------- Patients were evaluated in the clinic at screening (day -7), at baseline (day 0), on days 7, 14 (optional because some clinicians see their patients only every 4 weeks) and on day 28 (trial end), on other days if dose adjustment was necessary, and at the end of a one or two week tapering-off period. The primary efficacy variable was pain control, evaluated weekly on a five-point assessment scale ranging from very poor to excellent. For this assessment, the investigator presented the question \"would you rate your pain control today as being excellent, good, moderate, poor or very poor?\" Patients also completed a pain assessment questionnaire (shortened version of the Wisconsin Brief Pain Inventory (WBPI) \[[@B25]\]) (10-point rating scale: 0, best to 10, worst). The degree of pain after 24 hours of treatment was assessed by asking patients about the amount of pain that they had \'right now\'. Patients also recorded pain intensity in a diary on a five-point scale. Patients completed a treatment assessment questionnaire consisting of 10 items scored on a Likert scale. The acute version of the SF-36 quality of life questionnaire \[[@B26]\] including eight quality of life domains (physical functioning, physical role limitations, emotional role limitations, social functioning, body pain, general mental health, vitality and general health perceptions) was completed after the run-in phase and at day 28. Functionality of patients was assessed at the same time points using the Western Ontario and McMaster Universities Osteoarthritis Index LK 3.1 (WOMAC questionnaire) \[[@B27]\]. This has 24 questions that are each evaluated on a 5-point severity scale (0, none to 4, extreme), it assesses three areas: pain (5 questions), stiffness (2 questions) and functional impairment (17 questions). Maximum scores for each section differ and, therefore, scores are normalized by weighting severity to assist interpretation. Statistical analysis -------------------- There was no formal sample size calculation. A safety analysis was performed using data from all patients entering the trial. An intent-to-treat analysis, comprising all OA patients with at least one post-baseline measurement of the primary endpoint (pain control) and treated at least once with trial medication, was also performed \[[@B28]\]. The ANCOVA model was used to analyse the change from baseline to endpoint, and the influence of baseline values. The ANCOVA model was also used to determine differences by centre. The Wilcoxon signed-rank test was used to compare intragroup results and results at each time point or endpoint with baseline, where applicable. Statistical tests were interpreted at the 5% significance level (two-sided). Results ======= Patient characteristics ----------------------- A total of 159 patients, 102 with OA of the knee and 57 of the hip, were recruited into the study from 47 centres in 11 countries and started treatment with TDF. Patient characteristics are shown in Table [1](#T1){ref-type="table"}. All but three patients used analgesic treatment in the month before screening and the treatments were similar for hip and knee groups. The most commonly used non-opioid analgesics were paracetamol 21%, diclofenac 20% and rofecoxib 10%. The most commonly used weak opioid was tramadol (33%). During this time, half the patients (50%) used a non-opioid only; 28% used a combination of a non-opioid and weak opioid and 18% a weak opioid only. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Patient characteristics ::: OA knee OA hip -------------------------------------- ------------- ------------ N (baseline) 102 57 ITT analysis population 91 52 Mean age (years) ± SE (range) 68 ± 0.9 66 ± 1.2 (49--88) (47--87) Previous medication (% of patients): Non-opioids 81 79 Weak opioids 50 40 ::: Concomitant medication with possible analgesic effects during the treatment phase included paracetamol (69%), NSAIDs (45%), weak opioids (33%), Cox-2 inhibitors (13%), other analgesics (8%), steroids (6%), immunomodulating drugs (1%), and strong opioids (1%). Of those taking weak opioids, most took them on the first day of treatment only. Only two patients used opioid rescue medication and had to be considered as protocol violators. There was a reduction in the use of NSAIDs (from 45% to 18%) and Cox-2 inhibitors (9%) and paracetamol (15%) after the first 24 hours of treatment. Rescue medication was required by 59% of participants, of whom almost all used non-opioids, particularly paracetamol. Most patients suffered from concomitant diseases: 59% of the patients had currently active cardiovascular diseases, 50% musculo-skeletal, 48% genito-urinary, 47% endocrine, and 19% gastrointestinal disease. The most frequently used non-analgesic concomitant therapies during treatment were metoclopramide 49% and omeprazole 12%. Discontinuations ---------------- Of the 159 participants recruited, 25% withdrew from the trial during the treatment phase. Reasons for withdrawal were adverse events (35), insufficent response (1) and other reasons (4). Around half of the drop-outs (14%, 21 of the 35 due to adverse events) occurred in the first week of TDF treatment (Figure [1](#F1){ref-type="fig"}). Of the 48 patients who started the optional tapering-off phase, 11 (23%) discontinued prematurely (10 because of adverse events and one was lost to follow-up/surgery). Patients were treated for an average of 22.3 ± 0.92 days. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Patient disposition. ::: ![](1471-2474-6-31-1) ::: Study medication ---------------- All patients started at a dose of 25 μg fentanyl/h and daily doses ranged from 25 to 125 μg/h. The maximum dose was used by only one patient. The mean daily dose for week 1 was 26 μg/h, which increased slightly to 37 μg/h in week 4. Over half (54%) of all patients used 25 μg/h as a maximum dose during the study (61% OA knee, 42% OA hip). Evaluation of efficacy ---------------------- ### Primary efficacy variable Adequate pain control was defined as a score of \'moderate\', \'good\' or \'excellent\' on the 5-point pain control assessment scale. The proportion of patients with adequate pain control increased during the one-week run-in period (during which doses of non-opioid analgesia were optimised) from 4% to 27%. These patients continued in the study and accounted for the majority of protocol violators. At baseline, 25% of patients reported very poor pain control, 48% poor and 25% moderate pain control, and there was no notable difference between those with OA knee and OA hip (Table [2](#T2){ref-type="table"}). A further increase in the proportion of patients with adequate pain control was observed after TDF treatment, particularly in the first week of treatment (to 74%) when 37% patients reported moderate, 29% good, and 8% excellent pain control. Adequate pain control was reported by 80% and 88% patients on days 14 and 28, respectively. At endpoint, 83% of patients considered their pain controlled, with 37% reporting moderate, 38% good, and 8% excellent pain control (Table [2](#T2){ref-type="table"} & Figure [2](#F2){ref-type="fig"}). About 10% more patients with OA hip reported good or excellent pain control at endpoint than those with OA knee. However, of the patients who already experienced adequate pain control after the run-in phase (OA knee, 24 moderate and one good pain control; OA hip, 12 moderate and one good pain control), 50% of both groups improved further during TDF treatment. Overall, 81% of participants with OA hip and 75% with OA knee improved from baseline to endpoint by at least one pain category. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Pain control assessments (number of patients and percentage in each category) ::: ScreeningN (%) BaselineN (%) EndpointN (%) ----------- -------------------- ------------------- ------------------- Very poor 27 (19) 36 (25) 4 (3) Poor 110 (77) 68 (48) 21 (14) Moderate 5 (3) 36 (25) 53 (37) Good 1 (1) 2 (2) 54 (38) Excellent 0 0 11 (8) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Pain control assessment. ::: ![](1471-2474-6-31-2) ::: ### Effect size An effect size was calculated by assigning numerical values to the pain categories (1 = very poor, 2 = poor, 3 = moderate, 4 = good, 5 = very good). This gives a mean pain control score at baseline of 2.03 (95% CI 1.9, 2.2), rising to 3.33 (3.2, 3.5) at endpoint and a mean change from baseline of 1.3 (SD ± 1.14). This gives an absolute effect size of 1.14. ### Secondary efficacy variables #### Pain control Patients reported a significant reduction in pain from baseline to endpoint for each WBPI item at every time point (p \< 0.001). The mean reduction in \'pain at its worst\' was 1.8 points (from 8.1 to 6.3), \'pain at its least\' was 1.6 points (from 4.4 to 2.8), and \'pain on average\' was 2.0 points (from 6.4 to 4.4). The amount of trouble or bother the pain was causing also decreased by 2.7 points on average (from 7.2 to 4.5). The mean reduction in \'pain right now\' was 2.6 points (from 6.1 to 3.5) from baseline to endpoint. A significant reduction in \'pain right now\' was reported as early as 24 hours after baseline (1.3 points, from 6.0 to 4.7). From patients\' diaries, the mean score for degree of pain was significantly decreased at each time point, and from severe pain (score 3) to moderate pain (score 2) from the run-in period to endpoint (p \< 0.001). Results were similar for the patients\' highest score for their degree of pain. Thus, while at baseline 58% (79/137) reported severe/extreme pain, 4% (5) mild, and only two patients were without pain, by endpoint 41% (56/138) reported moderate pain, 30% (41) mild and 7% (9) no pain. There was little difference between the OA knee and OA hip group. #### Treatment assessment In their assessment of treatment (total n = 125, OA knee 82, OA hip 43), 63% of patients rated TDF positively with respect to pain control and 84% would recommend TDF for their type of pain. Most patients were satisfied with its convenience of use (93% thought it easy/extremely easy to use; 85% were very/somewhat pleased by the way it\'s used), and 53% considered side effects were not an issue. In general, there was a difference of less than 10% between the OA knee and OA hip groups. In assessing how they had felt over the past week, the percentage of all patients who answered good or very good increased during the study from 7% (10/142) during the run-in period to 32% (31/97) in week 4, and their scores at all time points were significantly better than before treatment (p \< 0.001). By the end of the study, help with basic activities was required by only 28% of patients, with 49% relying less on their helper. Quality of life --------------- For the 122 patients who completed the SF-36 quality of life questionnaire, there were statistically significant improvements in all domains from baseline to endpoint, including overall physical health (p \< 0.001) and mental health (p \< 0.05) (Table [3](#T3){ref-type="table"}). Despite optimization of previous treatment, quality of life scores were low at baseline, with the patients\' underlying osteoarthritis particularly affecting role physical and bodily pain. It was in these two areas that patients showed greatest absolute improvement with TDF treatment. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Quality of life (SF-36 scores) ::: Domain n Score at baseline(Mean ± SE) Mean change between baseline and endpoint(95% CI) **p-value** ------------------------ ------- ---------------------------------- ------------------------------------------------------- --------------- Physical functioning Physical functioning 120 30.0 ± 2.15 4.0 (-0.09, 7.99) \<0.05 Role physical 119 11.8 ± 2.38 15.8 (8.89, 22.62) \<0.001 Bodily pain 122 23.7 ± 1.54 17.1 (13.21, 20.97) \<0.001 General health 119 44.1 ± 2.09 3.6 (0.09, 7.05) \<0.05 Mental health Vitality 119 34.6 ± 1.74 6.3 (3.24, 9.45) \<0.001 Social functioning 122 50.3 ± 2.55 7.9 92.99, 12.79( \<0.05 Role emotional 117 34.9 ± 3.97 10.8 (1.85, 19.81) \<0.05 Mental health 118 53.0 ± 2.06 4.7(1.28, 8.03) \<0.05 Summary measures Physical health 111 27.0 ± 0.69 4.1 (2.73, 5.54) \<0.001 Mental health 111 41.7 ± 1.17 2.5 (0.46, 4.64) \<0.05 ::: WOMAC ----- The mean score for all 24 questions from the three summary parameters (pain, stiffness and physical functioning) improved significantly from baseline to endpoint for all groups (total population p \< 0.001, knee p \< 0.001, and hip p \< 0.05 for all questions) (Table [4](#T4){ref-type="table"}). The percentage of patients in the combined group and in the knee sub-group who reported no pain, stiffness or physical difficulties increased for all items. A similar increase occurred in the OA hip group except for \'stiffness after first awakening\', \'rising from bed\' and \'getting in/out of the bath\' which showed little change. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### WOMAC scores (mean normalised score ± SE, all changes from baseline are statistically significant, p \< 0.001 in all cases) ::: Baseline Endpoint Change from baseline to endpoint(95% CI) ---------------------- -------------- -------------- ---------------------------------------------- Pain 6.3 ± 0.15 4.6 ± 0.21 -1.7 (-2.05, -1.26) Stiffness 6.1 ± 0.19 4.7 ± 0.23 -1.4 (-1.89, -1.0) Physical functioning 6.6 ± 0.15 5.1 ± 0.19 -1.5 (-1.83, -1.14) Overall 19.0 ± 0.4 14.3 ± 0.57 -4.7 (-5.69, -3.62) ::: The majority of participants showed an improvement in score for the three summary measurements of pain, stiffness and physical functioning, and for the overall WOMAC score. Mean overall WOMAC score improved significantly (p \< 0.001) from baseline to endpoint (Table [4](#T4){ref-type="table"}). In addition, the change from baseline to endpoint in score for \'pain right now\' (from the pain assessment questionnaire) showed a weak positive correlation with the change in overall WOMAC score (Spearman correlation coefficient: non-normalized 0.344 and normalized 0.384). Evaluation of safety -------------------- Adverse events occurring during the treatment phase and tapering off phase were those associated with strong opioid treatment (Table [5](#T5){ref-type="table"}). Adverse events were reported by 6% (9/159) of patients during the run-in period, 65% (68/104) of patients during the treatment period and 25% (12/48) during the optional tapering off period. The study medication was permanently stopped in 25% (39) of cases, particularly because of nausea (53%), vomiting (47%) and dizziness (18%). (No falls or fractures were reported.) Withdrawal syndrome was reported in two cases (OA hip) during tapering off -- one was mild, the other moderate, and both resolved without specific treatment. No deaths occurred. Two patients reported at least one serious adverse event during the treatment phase (severe asthenia and anorexia in one case, bronchitis in the other) and one experienced a serious adverse event during the tapering-off phase (hospitalization due to chest pain and arrythmia), but these were considered unrelated to the study drug. There were no clinically significant changes in vital signs during the study. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Adverse events (AEs) reported during the treatment phase and tapering off phase by \>5% of participants ::: Preferred term Total AEs n (%) ----------------------------- --------------------- Treatment phase (N = 104) Nausea 51 (32%) Vomiting 41 (26%) Somnolence 25 (16%) Dizziness 14 (9%) Constipation 10 (6%) Asthenia 9 (6%) Pruritus 8 (5%) Tapering off phase (N = 42) Nausea 5 (10%) Vomiting 3 (6%) ::: Discussion ========== The study was intended to evaluate the utility of TDF under routine conditions and to investigate different practical issues, such as the usefulness of a test-period and use of concomitant antiemetic treatment during the first weeks. It was not designed as a primary efficacy study, since the efficacy of TDF in providing pain relief has already been demonstrated \[[@B14]-[@B18]\]. For this reason it was considered unethical to include a placebo control group. A one-month test period \[[@B10]\] is sufficient to show a role for TDF in the treatment of pain caused by OA of the knee or hip that is not adequately controlled by NSAIDs, Cox-2 inhibitors, paracetamol or weak opioids at optimal doses. Findings support previous conclusions that OA-induced pain can be successfully treated with TDF, and that this may result in improved functioning \[[@B22]\]. Not all patients with OA of the knee or hip receive adequate doses of their current non-opioid analgesia because pain control can be improved in some when doses of these medications are optimized. Undertreatment of OA-related pain has been reported previously \[[@B24]\]. Adequate pain control was achieved for most patients after one week of TDF, at the starting dose of 25 μg/hr in over half of all patients, and relief was maintained over the treatment period. Pain was controlled in 88% of patients after one month with nearly 40% reporting mild or no pain. A few patients started treatment with TDF, despite adequate pain control during the run-in phase. These patients were considered to be protocol violators. However, their inclusion provided the opportunity to determine whether TDF could give additional benefits to these patients above those already gained by their current medications. About half of these patients achieved further pain reduction while treated with TDF. Over the treatment period, the numbers of all patients using other analgesics, especially NSAIDs decreased substantially (from 45% to 19%). This may be beneficial in reducing side effects such as gastrointestinal bleeding. Nearly all patients taking weak opioids took them on the first day of treatment only when serum levels of fentanyl had not yet achieved steady state. Paracetamol may be useful for breakthrough pain in some patients but need for rescue medication should be evaluated on an individual basis. Metoclopramide did not appear to prevent nausea or vomiting. Treatment was considered favourable in terms of efficacy, side effects and convenience. A preference for treatment with TDF over sustained release morphine has previously been shown by patients with chronic non-cancer pain \[[@B16]\]. The general health measure, SF-36, indicated that control of pain significantly improved both physical and mental components of quality of life. The improvement in mental health may be related to improved functioning which permits greater social activity. In spite of similar pain control, patients with OA of the hip appeared to have slightly more difficulty with stiffness and general mobility than those with OA of the knee, such as getting in/out of the bath or bed, probably due to the location and function of these joints. WOMAC is a reliable, valid and responsive multidimensional, self-administrated outcome measure designed specifically to evaluate patients with OA of the knee or hip \[[@B27]\]. Overall pain, stiffness and function significantly improved after one month of TDF treatment for both OA knee and OA hip patients, with improvements in nearly all items of WOMAC summary categories. Quality of life would be expected to be improved with pain relief, although significant pain relief would not necessarily be associated with reduced stiffness and increased physical function \[[@B29]\]. Increased functioning was also indicated by the fact that half of all patients required less help with daily activities of living. Return of a full range of motion is unlikely to occur when marked structural damage of the joint has occurred. Overall, the spectrum of reported side effects was consistent with those commonly associated with opioid therapy and with previous experience with TDF. Constipation was not a problem for this patient population as with other patients receiving TDF for non-cancer pain or cancer pain \[[@B16],[@B29],[@B30]\], and tolerance to other side effects is likely to develop with continued treatment \[[@B31]\]. In addition to nausea and vomiting, patients should be warned of the possibility of dizziness when starting strong opioid treatment. This is especially important for the elderly OA population in order to prevent falls. In the present study, 9% of patients reported dizziness. As might be expected in an elderly population, many patients had co-existing diseases. For example, 59% had cardiovascular disease on entry, which is of particular interest given the concern about the cardiovascular safety of some Cox-2 inhibitors which, since this trial was undertaken, has led to the withdrawal of rofecoxib. This trial demonstrates that patients with OA of the knee or hip continue to experience pain even at optimal doses of their non-opioid treatment, providing a major reason why patients and clinicians alike are often dissatisfied with current therapies. The study also shows that patients with OA benefit from additional pain control provided by TDF \[[@B32]\]. Clinicians are beginning to accept that strong opioids are well tolerated and effective and should be made available when non-opioids have failed to control pain \[[@B33]\]. Our findings support this reasoning and suggest that opioids should be made more widely available where appropriate. Statement of competing interest =============================== Dr Richarz is an employee of Janssen-Cilag and the other authors have both received research funding from Janssen-Cilag, which funded this study. Authors\' contributions ======================= UR designed the study and coordinated it. XLL and KP recruited patients to the study and contributed to the interpretation of the findings. All authors contributed to developing the manuscript for publication. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2474/6/31/prepub> Acknowledgements ================ This study was funded by Janssen-Cilag. We would like to acknowledge the FEN-INT-30 Study Group for their participation in this trial and to thank Dr Susan Libretto for help in preparing the manuscript.	PubMed Central	2024-06-05T03:55:59.814812	2005-6-15	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181817/", "journal": "BMC Musculoskelet Disord. 2005 Jun 15; 6:31", "authors": [ { "first": "Xavier", "last": "Le Loët" }, { "first": "Karel", "last": "Pavelka" }, { "first": "Ute", "last": "Richarz" } ] }
PMC1181818	Background ========== Obesity is a major health problem in the western world. Nearly two-thirds of the U.S. adults are overweight (BMI \> 25) and of these one-half are obese (BMI \>30) \[[@B1]\]. Obesity is not only associated with increase in morbidity, mortality and reduction in life expectancy \[[@B2]\], but also leads to increase in the incidence of diabetes \[[@B3]\], hypertension \[[@B4]\], dyslipidemia and coronary artery disease \[[@B5]\]. Both diabetes and hypertension together account for approximately 70% of end-stage renal disease (ESRD). Approximately 300,000 adult deaths in the United States each year are attributable to unhealthy dietary habits and physical inactivity or sedentary behavior, with obese individuals having a 50 to 100 percent increased risk of death from all causes; most of the increased risk is due to cardiovascular causes \[[@B6],[@B7]\]. Obesity has also resulted in an increase in the cluster of disorders often referred to as the \"metabolic syndrome\". Although kidney disease has not yet been recognized as a major component of this metabolic syndrome, accumulating evidence suggests that even in non-diabetic obese patients, there is some degree of renal dysfunction that can lead to more serious injury to the kidneys as metabolic and hemodynamic disturbances worsen with prolonged obesity \[[@B8],[@B9]\]. We report a case that illustrates the stabilization of renal function with obesity directed therapy. Case Report =========== A 43-year-old Caucasian male was referred to the nephrology clinic at Overton Brooks VAMC by his primary care practitioner, in November of 2002 for management of his chronic kidney disease. He was asymptomatic. His BP was well controlled at 115/83 mmHg. He was morbidly obese with a body mass index (BMI) of 46, chronic kidney disease stage 4 (MDRD GFR of 16 ml/min), non insulin dependant diabetes mellitus, hypertension, coronary artery disease status post stent placement and hyperlipidemia. His medications included nifedipine, fosinopril, atenolol, rosiglitazone, furosemide, simvastatin, aspirin, glyburide and calcium carbonate. Laboratory results: Serum creatinine 4.3 mg/dl, BUN 54 mg/dl, normal electrolytes, serum calcium 8.8 mg/dl, serum phosphorus 4.9 mg/dl, random urine protein 292 mg/dl, random urine creatinine 49 mg/dl, urine protein/creatinine ratio of 5.9, hemoglobin A1c 7% and hemoglobin 13.9 g/dl. Patient was educated about the course and prognosis of his kidney disease and advised diet and exercise for weight loss. He was referred for arterial-venous fistula placement for providing renal replacement therapy in future. Over the next 6 months the patient failed all conservative methods of weight loss including the use of orlistat. His morbid obesity posed a major contraindication for enrolling him for kidney transplantation. He agreed to the surgical therapy option for treating his obesity. He was referred for bariatric surgery in June 2003. After the bariatric surgery in September 2003, he had lost 60 pounds at 6 months (BMI 37). He was able to discontinue all his oral hypoglycemic agents maintaining a hemoglobin A1c of 6.2% and required only one anti-hypertensive medication to achieve the recommended target blood pressure reading. His BUN and creatinine has remained at 22 mg/dl and 4.6 mg/dl respectively. The patient is being followed at regular intervals and over the course of the next eight months has lost an additional 30 pounds (BMI 32), a total weight loss of 90 pounds since the bariatric surgery. His serum creatinine has remained stable at 4 mg/dl, BUN 37 mg/dl, random urinary protein 99 mg/dl, random urinary creatinine 121 mg/dl, urine protein/creatinine ratio of 0.8 and hemoglobin A1c 5.1%. The inverse creatinine to time plot as shown in figure [1](#F1){ref-type="fig"} clearly demonstrates the stabilization of the renal function 15 months following his weight loss surgery. The patient was being evaluated for pre-emptive renal transplantation and because of his previous history of coronary artery disease he underwent a left heart catheterization study in March 2005. Unfortunately, despite all precautions he developed radio-contrast induced nephropathy and had to be initiated on renal replacement therapy. He currently remains on dialysis and is awaiting a renal transplantation, which would not have been possible without his weight loss. Discussion ========== The case report presented here illustrates the benefits of weight reduction on the progression of kidney disease. There are few studies investigating the pathophysiology of obesity and its early effects on kidney structure and function. Clinical as well as laboratory animal studies have suggested the role of glomerular hypertension due to renal vasodilatation and increase in hydrostatic pressure leading to increased glomerular wall stress and increased tubular sodium absorption \[[@B10],[@B11]\]. The other proposed mechanism of excessive tubular sodium re-absorption include increased intra-renal pressures caused by the excess accumulation of adipose tissue in the viscera with compression of the loop of Henle and vasa recta leading to sluggish flow in the renal tubules and vasa recta and thus causing an increase in the tubular sodium re-absorption \[[@B9],[@B12]\]. Increased sodium re-absorption in the loop of Henle initially reduces the macula densa sodium chloride delivery thereby initiating a macula densa feedback that causes vasodilatation of afferent arterioles; this increases renin secretion despite sodium retention and volume expansion. The compensatory vasodilatation of afferent arterioles resulting in an initial rise in glomerular filtration rate (GFR) and increase in the glomerular wall stress leads to increased extra cellular matrix formation and fibrosis along with injury to the kidneys and nephron loss with a resultant decrease in GFR over a prolonged period of time. There is also a growing body of evidence that obesity per se is a pro inflammatory state. Obesity is associated with increased levels of acute phase reactants, and cytokines as well as reactive oxygen species \[[@B13],[@B14]\]. Glomerular hyperfiltration also causes loss of protein in urine, which promotes glomerular inflammatory responses thus leading to progression of chronic kidney disease. In the already pre-existing pro-inflammatory state of obesity this could have an additive effect. Proteinuria seen in obese patients is often considered to be secondary to focal and segmental glomerulosclerosis. However, Kambham et al have reported a distinct obesity related histopathological change in the glomeruli, referred to as obesity-related glomerulopathy and was characterized by glomerulomegaly and focal segmental glomerulosclerosis. This entity defers from idiopathic focal segmental sclerosis with a lower incidence of nephrotic syndrome, more indolent course, consistent presence of glomerulomegaly, and milder foot process fusion \[[@B15]\]. We did not perform a renal biopsy hence we do not know whether proteinuria was secondary to obesity related glomerulopathy or idiopathic focal segmental sclerosis. Adequate treatment of obesity reduces proteinuria and decreases the need for medications such as angiotensin converting enzyme inhibitors or angiotensin-receptor blockers, which are known to further reduce the glomerular filtration rate. Hence targeting obesity should benefit in maintaining and preserving kidney function, regardless of whether weight reduction is achieved by diet and exercise or by bariatric surgery. Even though there have been no large studies directly comparing the effectiveness of different methods of weight loss on progression of kidney dysfunction, there is little doubt that weight loss reduces hypertension and type 2 diabetes, the two main risk factors for development of end stage renal disease. Bariatric surgery is a rapidly evolving branch of surgical science with the aim to induce major weight loss in those whose obesity places them at high risk of serious health problems. Bariatric surgery leads to withdrawal of anti-diabetic therapy in about 60% of patients, and reduction of therapy for many other diabetic patients. Eighty-two percent of the 165 type-2 diabetes mellitus patients in the uncontrolled observational study by Greenville achieved target glycemic control without any medications after an average of 14-years follow up \[[@B16]\]. The SOS study was a well-designed prospective study on obese patients in which, bariatric surgery was compared with a non-surgical group. The glycemic control without medication after 2-years in the surgical group was achieved in twice the number of patients compared to the control group \[[@B17]\]. Alexander et al studied 30 morbidly obese patients; 19 with chronic kidney disease and 11 with renal transplantation; and reported gastric bypass surgery to be an effective means for achieving significant long-term weight loss and relief of co-morbid conditions in patients with renal failure on dialysis, in preparation for transplantation, or after transplantation \[[@B18]\]. The above observations that bariatric surgery leads to reduction in the risk factors associated with development of ESRD and our case report showing that the progression of CKD was retarded with postponement of dialysis raises the question: Should bariatric surgery be recommended in the morbidly obese who fail to achieve sufficient weight loss using non-surgical approaches especially those who are young and have other metabolic syndrome risk factors and are at a favorable anesthetic/surgical risk? Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= SA: Collected the data and drafted the manuscript; VTV: Participated in the coordination and drafting of manuscript; RG: participated in the data collection and formatting of the manuscript NKA: participated in the clinical management of the case and drafting and finalizing the manuscript and TJV: Conceived the idea, coordinated the data and helped in drafting and finalizing the manuscript. All authors have read and approved the manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2369/6/7/prepub> Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Relationship between renal function as 1/creatinine and BMI. ::: ![](1471-2369-6-7-1) :::	PubMed Central	2024-06-05T03:55:59.818031	2005-6-15	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181818/", "journal": "BMC Nephrol. 2005 Jun 15; 6:7", "authors": [ { "first": "Sujata", "last": "Agnani" }, { "first": "Vidula T", "last": "Vachharajani" }, { "first": "Rohit", "last": "Gupta" }, { "first": "Naveen K", "last": "Atray" }, { "first": "Tushar J", "last": "Vachharajani" } ] }
PMC1181819	Background ========== The conserved Notch signaling pathway has well established roles in cell fate determination during development. Transmembrane Notch receptors are activated by transmembrane DSL ([D]{.underline}elta/[S]{.underline}errate/[L]{.underline}AG-2) family ligands \[[@B1]-[@B6]\]. The intracellular (IC) domain of Notch is proteolytically released by presenilins and translocates to the nucleus \[[@B7]-[@B11]\], where it acts as a transcriptional activator abetted by CSL ([C]{.underline}BF1/[S]{.underline}u(H)/[L]{.underline}AG-1) proteins \[[@B12]-[@B15]\]. In C. elegans, the LIN-12 Notch receptor is activated by LAG-2 and related DSL ligands \[[@B16]-[@B19]\], and proteolytically processed by the presenilins SEL-12 and HOP-1 \[[@B20]-[@B22]\]. The CSL protein LAG-1 interacts with LIN-12IC to activate transcription of target genes \[[@B23]\]. Notch receptors and ligands are expressed in adult vertebrate neurons \[[@B24],[@B25]\]; recent studies in Drosophilaand mice suggest that altering Notch signaling results in defective neuronal function \[Costa, 2003 \#50; Ge, 2004 \#46; Presente, 2004 \#31; Saura, 2004 \#32; Wang, 2004 \#51;Yoon, 2005 \#69\]. The importance of these findings is underscored by the fact that several genetic diseases associated with neuronal defects and/or late onset symptoms map to mutations in Notch pathway genes \[[@B32]-[@B36]\]. However, it remains unclear from these studies whether Notch signaling is acutely affecting neuronal physiology or if it is causing permanent changes in cell fate and/or structure due to developmental defects or aberrant growth. Here, we report a new role for lin-12signaling in the adult C. elegansnervous system, using behavior as an indicator of neuronal activity. C. eleganspredominantly move forward, but they spontaneously initiate backward locomotion. Genetically modulating lin-12activity alters the rate of initiation of spontaneous reversals. Using inducible RNAi and a conditional, gain-of-function allele of lin-12, we show that this behavioral change can occur within a few hours of altering lin-12activity in post-developmental adults. We also show that these inducible behavioral changes are rapidly reversible, strongly suggesting that lin-12mediated behavioral changes are unlikely due to changes in cell fate. Altering lin-12activity in a subset of interneurons is sufficient to alter behavior. glr-1, an AMPA/kainate receptor homolog gene expressed in these interneurons, genetically interacts with lin-12. Our results demonstrate a novel, post-developmental role for lin-12signaling that is clearly distinct from its role in cell fate determination. Results ======= Altering lin-12activity increases spontaneous reversals during locomotion --------------------------------------------------------------------------- To assess a role for lin-12Notch signaling in behavior, we first examined spontaneous reversal rates during locomotion in lin-12mutant animals (Fig. [1](#F1){ref-type="fig"}). Normal animals moving forward consistently initiate backward locomotion approximately 10 times per 3 minutes. Reversal rates were significantly increased in lin-12(n941) loss of function (lf) animals, which completely lack lin-12gene function. The behavioral defect of lin-12(lf) animals was rescued by a previously described transgene containing a lin-12cDNA driven by the lin-12promoter \[[@B37]\]. Furthermore, lin-12(lf) behavioral defects could be recapitulated by RNAi (see below). These results indicated that loss of function in lin-12caused increased reversal rates. The effect of increased lin-12activity on reversal rates was then assessed using lin-12(n137n460) (Fig. [1](#F1){ref-type="fig"}), a gain of function, cold sensitive (gfcs) allele \[[@B38],[@B39]\]. Reversals were not significantly increased in lin-12(gfcs) animals raised at the permissive temperature (25°C), but were dramatically increased in animals raised at the restrictive temperature (15°C). Cultivation temperature had no effect on reversal rate in wild type animals. The increased reversal rate of lin-12(gfcs) animals was due to increased lin-12activity, as transgenic animals that overexpress LIN-12 (lin-12p::lin-12(OE)) also had increased reversals. Thus, both gain and loss of function in lin-12causes increased reversal rates. An allelic series of lin-12mutants reveals complex regulation of behavior --------------------------------------------------------------------------- To further characterize the relationship between lin-12activity and reversal rates, we assessed behavior across the lin-12allelic series ordered based on the severity of previously determined vulval defects (Table [1](#T1){ref-type="table"}). lin-12alleles can be grouped into 4 classes: strong loss of function, weak gain of function, moderate gain of function, and strong gain of function. lin-12(n941)null animals are sterile and display protruding vulva that usually burst in adult animals; these animals had increased reversals. The vulval and reversal phenotypes of lin-12(n941)/+animals were normal, indicating that the lin-12(n941)mutation is recessive. lin-12(n302), lin-12(n379), and lin-12(n676)are weak gain of function alleles \[[@B39]\]; these animals were fertile, vulvaless, and had slightly decreased reversal rates. lin-12(n137n460)acts as a moderate gain of function allele; these animals are cold-sensitive, display multiple pseudovulvae at the restrictive temperature, and had increased reversals. Finally, lin-12(n137)and lin-12(n427)are strong gain of function alleles that display multiple pseudovulvae and had strongly decreased reversals. We note that lin-12(n137)/lin-12(n941)hemizygote animals have multiple pseudovulvae and have high reversal rates consistent with lin-12(n137n460)phenotypes. Since lin-12p::lin-12(OE) (Fig. [1](#F1){ref-type="fig"}) and other transgenic animals overexpressing lin-12(see Fig. [5B](#F5){ref-type="fig"}) recapitulate the moderate lin-12gain of function allele, we expected that injection of the lin-12p::lin-12construct at a higher concentration would lead to transgenic animals that recapitulate the strong lin-12gain of function alleles. However, we were unable to generate viable transgenic lines using higher concentrations of lin-12p::lin-12(data not shown; see Methods for details). We conclude that altering lin-12activity results in complex changes in the pattern of reversal behavior; the implications of this allelic series are discussed below. Altering lin-12activity in adult animals is sufficient to increase reversal rates ----------------------------------------------------------------------------------- lin-12Notch plays well established roles in development. Therefore, we asked whether increased reversals in lin-12mutant animals depended on lin-12activity during development or in adults. lin-12loss of function was induced by expressing an inverted repeat of a lin-12cDNA fragment under the control of a heat shock promoter to knock down lin-12activity by RNAi (hsp::lin-12(RNAi)) in otherwise normal adult animals (Figure [2A](#F2){ref-type="fig"}). Uninduced hsp::lin-12(RNAi) adult animals (filled triangles) raised at 25°C had normal reversal rates, while heat shock induction resulted in dramatically increased reversal rates within 4 hours. Reversals returned to near basal levels after overnight recovery (approx. 14 hours). Heat shock had no effect on wild type control animals (filled circles). As a control, we generated transgenic animals containing an inverted repeat of a cDNA fragment from the [G]{.underline}protein coupled [r]{.underline}eceptor [k]{.underline}inase-2 grk-2)gene under control of the heat shock promoter (hsp::grk-2(RNAi)). grk-2loss of function causes sensory defects \[[@B40]\], but had no effect on reversal rates (data not shown). Heat shock induction of hsp::grk-2(RNAi) did not alter reversal rates (open squares), indicating that neither the presence of the heat shock vector nor overexpression of an unrelated dsRNA influenced reversal rates. We conclude that loss of function of lin-12in adult animals is sufficient to alter behavior. We examined lin-12(gfcs) animals in temperature shift experiments (Figure [2B](#F2){ref-type="fig"}). lin-12(gfcs) adults raised at the restrictive temperature 15°C (filled square, t = 0) initially had increased reversal rates. When these animals were moved to the permissive temperature of 25°C (dotted line with filled squares), reversal rates gradually decreased, and after 3 hours reversals decreased to wild type levels. In reciprocal experiments, lin-12(gfcs) adults raised at 25°C (open square, t = 0) initially had almost normal reversal rates. When they were moved to 15°C (solid line with open squares), reversal rates gradually increased until they reached levels comparable to those of lin-12(gfcs) animals raised at 15°C. When these animals were moved back to 25°C (dotted line with open square), reversal rates decreased to original levels within 2 hours. Temperature shifts and cultivation temperature had only minimal effects on control wild type animals (open and filled circles). Taken together, these data demonstrate that altering lin-12Notch activity for a few hours in post-developmental adult animals is sufficient to change behavior and suggests that lin-12activity is regulating a physiological, not a developmental, process. lin-12is not required in the vulval lineage to regulate reversals ------------------------------------------------------------------- Where does lin-12function to regulate reversal rates? LIN-12 is expressed in the somatic gonad and vulval lineages, based on previous studies using a functional lin-12::gfptransgene \[[@B37]\]. To test if lin-12activity in these tissues regulated reversal rates, we eliminated the somatic gonad and the vulva by killing the progenitor cells of these lineages using a laser and then determining the reversal rates of the operated animals. Vulval development depends on cell-cell signaling from the anchor cell to the vulval precursor cells \[[@B41]\]. The anchor cell (and somatic gonad) is derived from one of two equipotent cells called Z1 and Z4 in L1 larvae; thus, killing Z1 and Z4 results in animals that lack gonads and vulvae. Wild type Z1-Z4-killed adult animals had normal reversal rates, indicating that these tissues play no role in regulating the reversal rate in wild type animals (Fig. [3](#F3){ref-type="fig"}). We tested the role of the vulval and gonadal lineages in regulating reversal rates in lin-12(RNAi) (see Methods for details) and lin-12p::lin-12(OE) animals. Z1-Z4-killed lin-12(RNAi) and lin-12p::lin-12(OE) animals maintained high reversal rates comparable to mock treated animals. These data indicate that lin-12must function outside of the somatic gonad and vulva to regulate reversal rates. We considered the possibility that the gross morphological vulval defects in lin-12mutant animals, but not lin-12signaling per se, might account for changes in reversal rates. lin-12(lf) animals have a large protruding vulva, while lin-12(gfcs) animals raised at the restrictive temperature 15°C have multiple pseudovulvae. We measured basal locomotion rates in lin-12(lf) and lin-12(gfcs) animals and found that they had slight but significant decreases in basal movement rate (Table [2](#T2){ref-type="table"}). However, lin-12(RNAi) and lin-12p::lin-12(OE) animals, which phenocopy the reversal phenotypes but not the vulval phenotypes of the mutant animals, had normal basal movement rates. These data indicate that morphological defects of the vulva in lin-12mutant animals cannot account for the altered behavior. Taken together, we conclude that lin-12expressed in the vulva and somatic gonad does not contribute to the regulation of reversal rates during locomotion. lin-12acts in a subset of glr-1expressing neurons to regulate reversals --------------------------------------------------------------------------- lin-12might act in or upon neurons to control C. elegansbehavior. Previous studies have implicated the command interneurons AVA, AVB, AVD, AVE and PVC in regulating normal forward and backward locomotion \[[@B42]\]. The intrinsic activation of these interneurons affects reversal rates \[[@B43]\]. Other neurons presynaptic to the command interneurons, such as ASH \[[@B43]\] and AIY \[[@B44]-[@B46]\], can affect reversal rates as well. We did not detect any overt cell fate changes or morphological defects in any of these or other neurons in lin-12(lf), lin-12(gfcs) or lin-12(n137gf) mutant animals (data not shown), consistent with a previous report \[[@B47]\] and supporting our conclusion that lin-12mediated behavioral changes were not due to developmental defects. LIN-12 expression was not detected in any of these neurons either by immunohistochemical analysis or by GFP fluorescence (data not shown). However, occasional LIN-12::GFP expression was observed in the RIG neurons of young larvae (Fig. [4](#F4){ref-type="fig"}). These observations and the preceding temperature shift experiments suggested that LIN-12 may be expressed in adult neurons at levels too low to detect. Increasing LIN-12::GFP levels further caused lethality (data not shown); therefore, a functional approach was taken to determine whether lin-12acts in the nervous system. lin-12activity was knocked down by RNAi in a subset of neurons by expressing lin-12dsRNA under the control of the glr-1promoter, which drives expression in the aforementioned command interneurons and twelve other classes of neurons including RIG \[[@B48],[@B49]\] (glr-1p:::lin-12(RNAi)). Because RNAi effects can spread systemically \[[@B50]\], we first validated the cellular specificity of this approach. The lin-12cDNA fragment used to generate the glr-1::lin-12(RNAi) constructs was derived from a lin-12::gfpfusion; thus, the dsRNA expressed in these transgenic animals contains both lin-12and gfpsequences. When glr-1p::lin-12(RNAi) constructs were injected into strains that express GFP in the intestine or in ASH sensory neurons (which are physically close to glr-1expressing neurons), no decreases in GFP fluorescence were observed (Figure [5A](#F5){ref-type="fig"}). Furthermore, these transgenic animals had grossly normal fertility and vulval morphology (data not shown). Thus, RNAi effects did not appear to spread from glr-1expressing neurons to nearby neurons, the intestine, or to the vulva. Also, we found that transgenic animals injected with the glr-1promoter fragment alone (glr-1p::(0)) or constructs expressing gfponly dsRNA under control of the glr-1promoter had no effect on reversal rates (Fig. [5B](#F5){ref-type="fig"}). When we examined the behavior of glr-1p::lin-12(RNAi) animals, we found that reversal rates increased significantly. Thus, knocking down lin-12activity in glr-1expressing neurons was sufficient to recapitulate lin-12(lf)behavioral defects. The requirement for lin-12activity in the nervous system was also tested by driving lin-12cDNA expression using the glr-1promoter (Fig. [5B](#F5){ref-type="fig"}). Increasing lin-12activity by overexpressing either a full length lin-12cDNA or a truncated, activated form of lin-12under the control of the glr-1promoter (glr-1p::lin-12(OE) and glr-1p::lin-12IC, respectively) also increased reversal rates. Expression of GFP using the glr-1promoter (glr-1p::gfp) as a control had no effect (Fig. [5B](#F5){ref-type="fig"}). Finally and most significantly, expression of the lin-12cDNA under the control of the glr-1promoter (glr-1p::lin-12(+)) rescued the behavioral defects of lin-12(lf) animals, restoring reversal rates to wild type levels (Fig. [5C](#F5){ref-type="fig"}). These results demonstrate that lin-12activity in glr-1expressing neurons is sufficient to regulate reversal rates. Increased lin-12activity affects reversal rates via RIG neurons ----------------------------------------------------------------- The RIG neurons, in which we observed weak LIN-12::GFP expression, express glr-1. Interestingly, the RIG neurons are presynaptic to command interneurons. The role of RIG neurons in spontaneous reversal rates was tested by laser ablation. Laser ablation of the RIG neurons did not dramatically affect reversal rates in wild type or lin-12(RNAi) animals. However, eliminating the RIG neurons of lin-12p::lin-12(OE) animals ameliorated reversal rate increases (Fig. [6](#F6){ref-type="fig"}), Thus, increased lin-12function in the RIG neurons is likely responsible for increased reversal rates in lin-12p::lin-12(OE) animals. Despite the fact that lin-12overexpression recapitulated lin-12(gfcs) behavioral defects, we considered the possibility that the lin-12p::lin-12(OE) transgene might act ectopically or during development to alter reversal rates. Increasing lin-12activity in adult lin-12(gfcs) animals in temperature shift experiments was sufficient to increase reversal rates (Fig. [2B](#F2){ref-type="fig"}). Therefore, we carried out RIG laser ablations in lin-12(gfcs) animals, using the same temperature shift paradigm described above (Fig. [2B](#F2){ref-type="fig"}). RIG killed, temperature shifted lin-12(gfcs) animals had normal reversal rates, while mock treated lin-12(gfcs) control animals retained high reversal rates. We conclude that increased lin-12activity in the RIG neurons of adult animals increases reversal rates. Genes that interact with lin-12to regulate reversals ------------------------------------------------------ To determine if the canonical lin-12signaling pathway regulates spontaneous reversal rates, we examined the reversal rates of animals that are defective in genes of the lin-12pathway, specifically lag-1(which encodes a transcriptional cofactor that is the major effector for lin-12signaling) and lag-2(which codes for a Notch ligand) (Fig. [7A](#F7){ref-type="fig"}). The reversal rates of partial loss-of-function lag-1and lag-2mutant animals were relatively normal. However, partial loss of lag-1 function suppressed increased reversal rates in glr-1p::lin-12ICanimals that had constitutively activated lin-12signaling, consistent with lag-1functioning downstream of lin-12regulating reversal rates. Also, a semidominant allele of lag-2that suppresses the lin-12gain-of-function multivulval phenotype caused increased reversal rates. Although strong loss of function alleles could not be tested due to embryonic lethality, our results suggest that lin-12, lag-2, and lag-1likely act together in the nervous system to regulate reversals. Finally, given the previously described role of the glr-1AMPA receptor in the command interneurons \[[@B43],[@B48],[@B49],[@B51]\], we examined more closely the role of glr-1in spontaneous reversals and lin-12mediated behavioral changes (Fig. [7B](#F7){ref-type="fig"}). Consistent with a previous report, complete loss of glr-1function (glr-1(lf)) alone had no effect on reversal rates \[[@B43]\]. However, we found that overexpression of glr-1(glr-1p::glr-1(OE)) increased spontaneous reversal rates. We note that different constructs are used here than previous studies \[[@B43]\] (see Methods for details) and that reversal rates can be dependent on assay conditions. Both glr-1(lf);lin-12p::lin-12(OE) and glr-1(lf);glr-1p::lin-12(RNAi) animals had dramatically decreased reversal rates (below wild type levels). Yet, there were no dramatic changes in the expression of a glr-1p::gfptranscriptional reporter in lin-12(gfcs) or (lf) animals (data not shown). Our results suggest that glr-1AMPA receptor activity, but not levels, are modulated by lin-12signaling to regulate reversals. Discussion ========== In this study we demonstrate a non-developmental role for lin-12Notch in the adult nervous system regulating C. elegansbehavior. lin-12mediated behavioral changes can be rapidly induced within a few hours in adult animals and are reversible. Knocking down lin-12activity by RNAi or by activating lin-12in glr-1expressing neurons is sufficient to reproduce the behavioral defects of lin-12mutant animals. The rapidity with which behavioral changes can be induced in post-developmental adult animals argues that neither lin-12mediated cell fate changes nor de novoneurite outgrowth are the likely mechanisms for altering behavior. Rather, our results are consistent with a novel role for lin-12signaling acutely regulating neuronal physiology via transcriptional activation, clearly distinct from previously described roles in cell fate specification. Signaling pathways used to pattern the developing nervous system can also play important roles in the adult nervous system. For example, ephrins and Eph receptors function both in nervous system patterning during development and in synaptic plasticity in the adult nervous system (reviewed in \[[@B52]\]). Recent studies suggest that Notch signaling may also play a role in adult neurons. In Drosophila, adult animals harboring temperature sensitive, loss-of-function Notch alleles are defective for long term memory formation after one to two days at the restrictive temperature \[[@B27],[@B28]\]. In mice, Notch1 and CBF1 heterozygous adult animals have specific defects in spatial learning and memory \[[@B26]\]. Similarly, adult mice in which Notch protein levels have been partially depleted by antisense RNA are defective in long term potentiation (LTP) \[[@B30]\]. Conditional knockout of both presenilin genes in the postnatal forebrain in mice results in defects in long-term contextual memory and LTP, when assayed in two month old animals \[[@B29]\]. Our heat shock and temperature shift experiments indicate that behavioral defects appear within hours, suggesting that Notch mediated alterations in neuronal function can occur on a much shorter timescale than days \[[@B27],[@B28]\] or months \[[@B29]\] as previously reported. The lin-12allelic series for reversal rates is complex. In particular, lin-12(n137n460)gain-of-function hemizygotes, heterozygotes, and homozygotes all have high reversal rates, while stronger gain-of-function alleles (n427and n137) have decreased reversals, raising the possibility that lin-12(n137n460)could be a neomorphic allele. Several lines of evidence argue against this hypothesis. First, based on vulval phenotypes, there is no evidence of any neomorphic activity. lin-12(n137n460), which is a recessive hypermorphic allele, is a revertant of lin-12(n137), a dominant hypermorphic allele; the n460mutation confers a temperature sensitive, partial loss of function onto n137\[[@B38],[@B39]\]. Both the n137and the n137n460alleles cause multiple pseudovulvae, indicating that lin-12(n137n460)is simply a weaker hypermorph than lin-12(n137). Second, modestly increasing lin-12activity through several other independent means also caused increased reversals. These include moderate overexpression of lin-12(lin-12p::lin-12(OE)) at levels that do not affect fertility and vulval development, and placing the strong hypermorphic allele lin-12(n137)over the null allele (i.e., lin-12(n137/lin-12(n941)animals). We favor the hypothesis that the unconventional lin-12allelic series for reversal rates reflects the underlying complexity of Notch signaling and the neuronal signaling pathways that regulate behavior. lin-12acts at multiple places during vulval cell fate specification, specifically the AC/VU decision and VPC lateral inhibition, resulting in a complex allelic series for vulval phenotypes. Similarly, lin-12gain and loss of function may have different cellular foci for action in the nervous system, making it difficult to predict the behavioral output based on simple genetic rules. This is partially supported by the RIG ablation studies, wherein killing RIG neurons in lin-12gain of function animals ameliorated reversal increases, but had no effect in lin-12loss of function animals. Alternatively, lin-12may act coordinately with other genes to regulate reversals. Further genetic studies may lead to a clearer picture. Consistent with this hypothesis, we have found that glp-1, another C. elegansNotch homolog, modulates reversal rates (in preparation). Our data suggest that lin-12regulates reversal rates in a complex fashion. The behavioral changes observed in lin-12animals are dramatically dependent on GLR-1 AMPA receptor function. Taken together with our finding that lin-12acts in glr-1expressing neurons to regulate reversals, it suggests a possible relationship between AMPA receptors and Notch receptors in post-developmental synaptic plasticity. This is consistent with a recent study that demonstrated that altering Notch signaling caused defects in LTP in mice \[[@B29],[@B30]\]. Based on our genetic analysis, glr-1may be a target of lin-12signaling or lin-12signaling may act in parallel with glr-1. For example, lin-12signaling may modulate other glutamate-gated currents to influence membrane excitability. Consistent with this hypothesis, loss of function in avr-15, one of several semi-redundant C. elegansgenes encoding conserved glutamate-gated chloride channel subunits \[[@B53]\], results in increased reversals. avr-15is expressed in the AVA command interneurons (data not shown) and chloride currents have been observed in these interneurons \[[@B51]\], making AVR-15 a candidate target for regulation by LIN-12 signaling. Similarly, loss of function of nmr-1, which encodes an NMDA glutamate receptor subunit, results in decreased spontaneous reversals \[[@B54]\], suggesting that nmr-1activity could be influenced by lin-12. Additional behavioral and genetic analysis will be required to further delineate the targets of lin-12signaling in adult neurons. It should be noted that defects in Notch signaling can result in pleiotropic developmental disorders and nervous system dysfunction. CADASIL syndrome is associated with mutations in human Notch3 and is characterized by seizures, late onset neurodegeneration and vascular defects \[[@B32]\]. Mutations in Jagged1 (a DSL protein family member) are implicated in Alagille syndrome, which is characterized by defects in liver, cardiac, and skeletal tissues, and less frequently, neurovascular defects and mental retardation \[[@B34],[@B35]\]. Familial, early onset Alzheimer\'s disease is often caused by mutations in presenilin 1 or presenilin 2 \[[@B33],[@B36]\]. The developmental defects associated with CADASIL and Alagille syndromes make it difficult to establish a role for Notch signaling in neurons, but it may play a role in the defects observed in some of the late-onset symptoms. Given the emerging role for Notch signaling in the adult nervous system, a role for defective Notch signaling in these and other neurological disorders warrants further investigation. Conclusion ========== We have demonstrated a novel role for lin-12Notch in Caenorhabditis elegansin the adult nervous system. Changing lin-12activity postdevelopmentally in adult animals alters the spontaneous reversal rates during locomotion. lin-12activity in the vulva and somatic gonad, where lin-12expression was previously reported, is not required to control reversal rates. In contrast, altering lin-12activity in specific neurons is sufficient to alter behavior. lin-12likely acts through the canonical Notch signaling pathway that includes the ligand lag-2and the downstream effector lag-1. The neuronal function of lin-12is clearly independent from cell fate specification during development. Methods ======= Behavioral assays ----------------- Spontaneous reversals are modulated by sensory input, environmental conditions and feeding status \[[@B43],[@B55]\]. To control these variables, animals were cultured on NGM agar plates containing OP50 E. coliat 25°C, except in temperature shift experiments, in which animals were cultured at 15°C and moved to room temperature 30 minutes prior to assays. Young adults (containing at least 4 eggs) were moved from the bacterial lawn of an uncrowded plate to an NGM plate lacking food, allowed to crawl around briefly to remove bacterial residue, then quickly transferred to another NGM plate lacking food for assays. Spontaneous initiation of backward locomotion was recorded over three minutes during the next 1.5 to 10.5 minutes with the lid on. Up to three animals per assay plate per trial were used; no effect on reversal rates was observed for up to three animals per plate. Freshly poured NGM agar plates were dried in a laminar flow hood for approx. 2 hours, sealed with Parafilm, then stored at 4°C at least overnight. Plates were allowed to warm up to room temperature for at least 30 minutes prior to use. Several assay plates were tested until a plate that resulted in an average of 10 reversals in 3 minutes was observed for N2 control animals; this plate was then used for all subsequent assays on that day. Each initiation of backward locomotion was scored as one reversal; omega turns without reversals were not scored. A subset of animals was scored blind as to genotype and/or transgene to confirm results. lin-12mutants have defective vulvae, which are visually obvious; therefore, lin-12mutant animals were scored independently by two observers. Statistical analysis was performed using the two tailed Student\'s t test. Laser ablations --------------- Laser ablations were performed as previously described \[[@B56]\] using a Micropoint ablation system (Photonic Instruments, St. Charles, IL). RIG ablations were undertaken in nyIs60animals expressing flp-18p::GFP\[[@B57]\]. These animals are uncoordinated but have normal spontaneous reversal rates. lin-12(lf) mutant animals were not subjected to laser microsurgery because they rarely survived the procedure. lin-12(gfcs) mutant animals did not survive laser microsurgery as L1 larvae, but most survived when operated on as L2-L3 larvae. After laser microsurgery, lin-12(gfcs) animals were allowed to recover at the permissive temperature 25°C for 1--2 days, then were shifted to the restrictive temperature 15°C for 4 hours prior to behavioral assays. The flp-18p::gfptransgene did not affect the temperature dependence of lin-12(gfcs) phenotypes (data not shown). After behavioral assays were completed, successful ablation of the RIG neurons was scored by the lack of GFP labeled neuronal cell bodies in the retrovesicular ganglion. In nearly all laser ablation experiments, mock treated animals with altered lin-12activity had slightly lower reversal rates than untreated animals. However, they still had significantly higher reversal rates than wild type mock treated animals. Molecular biology ----------------- Plasmids used for transgenes are as follows: lin-12p::lin-12(OE), plin-12::gfp; hsp::lin-12(RNAi) and lin-12(RNAi), pHA\#394; hsp::grk-2(RNAi), pHA\#327; glr-1p::(0), pHA\#421; glr-1p::glr-1(OE), pCR\#3; glr-1p::gfp(RNAi), pKP\#6 and pHA\#424; glr-1p::lin-12(OE) and glr-1p::lin-12(+), pHA\#444; glr-1p::lin-12IC, pHA\#382; glr-1p::lin-12(RNAi), pHA\#380 and pHA\#381. Plasmid details are available upon request. Genetics and strains -------------------- Strains used in this study: N2 Bristol wild type isolate, lin-12(n137n460gfcs), lin-12(n941lf)/unc-32, lin-12(n941lf)/eT1, lin-12(n941lf)/qC1, lin-12(n137)/unc-32, lin-12(n302), lin-12(n379), lin-12(n427), lin-12(n676), lag-1(om13), lag-2(sa37), lag-2(q420), glr-1(n2461) ncl-1(e1865), pha-1(e2123ts), nyIs60 \[lin-15(+) flp-18p::gfp\], mgIs18 \[lin-15(+) ttx-3p::gfp\], nuIs25 \[lin-15(+) glr-1p::glr-1::gfp\], nuIs1 \[lin-15(+) glr-1p:::gfp\], rtIs11 \[osm-10p::gfp\], and rtIs18 \[elt-2p::gfp\]. Transgenes were co-injected using pha-1(+) (pBX1), myo-2p::gfp(pPD48.33), and/or elt-2p::gfp(pJM67) as markers; details upon request. Heat shock induction occurred at 33°C for 2 hours. hsp::lin-12(RNAi) introduced at 8 ng/μl yielded inducible transgenic lines (hsp::lin-12(RNAi)); introduction at 50 ng/μl resulted in lines with increased reversal rates even in the absence of heat shock (26.8 ± 1.9 reversals/3 min., n = 11; see also Fig. [3](#F3){ref-type="fig"}); these lines are designated lin-12(RNAi) in the text to distinguish them from the inducible hsp::lin-12(RNAi) lines. Transgenic lines overexpressing lin-12p::lin-12at very high levels (100 ng/μl) often had extra vulvae and were difficult to generate and maintain; these animals were used only for expression analysis. Moderate overexpression (50 ng/μl) of lin-12p::lin-12was not overtly deleterious and vulval perturbations were infrequent; these animals were used for behavioral analysis. The integrated transgene nuIs25that overexpresses a GFP tagged glr-1rescue construct \[58\] increased reversals (shown in Fig. [7B](#F7){ref-type="fig"}). We also generated extrachromosomal arrays marked by pha-1that overexpress a glr-1rescue construct lacking GFP; animals carrying these arrays also had increased reversals (15.7 ± 0.8 reversals/3 min., n = 26, p\<10^-4^vs. wild type). Authors\' contributions ======================= M.Y.C, J.L.-F., T.T., and A.C.H. all contributed to the genetic, molecular, and behavioral experiments. M.Y.C. and A.C.H. drafted the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ We wish to thank Bob Horvitz, Iva Greenwald, Paul Sternberg, Stuart Kim, Villu Maricq, Josh Kaplan, Chris Rongo, Oliver Hobert, Andy Fire, Chris Li, Calum Macrae, Robert Nowak and Diane Levitan for strains, plasmids, and use of equipment, and members of the Hart, van den Heuvel, and Artavanis-Tsakonas laboratories and the C. elegansresearch community for helpful discussions. We acknowledge the assistance of CaenorhabditisGenetics Center for providing numerous strains and the help of Enrico Montana and Alex Ihring during the MBL Neurobiology course, 2002. This work was supported by an NIH NIGMS grant to A.C.H. and a MBRC Tosteson fellowship to M.Y.C. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### *Altering lin-12activity increases spontaneous reversal rates.Animals of the genotypes indicated were tested for mean number of reversals per 3 minutes on NGM agar plates (see Methods for details). lin-12(lf) is lin-12(n941), a complete loss of function allele. lin-12(gfcs) is lin-12(n137n460), a cold sensitive, gain of function allele. lin-12p::lin-12(+) and lin-12p::lin-12(OE) are transgenic animals that have been injected (at different concentrations, see Methods) with plin-12::gfp, a plasmid that expresses a functional lin-12cDNA fused to gfpunder the control of the lin-12promoter 37. \\\ p\<0.0001 vs. wild type. ::: ![](1471-2202-6-45-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### *Changing lin-12activity in adults alters spontaneous reversal rates.Changing lin-12activity in adults alters spontaneous reversal rates. A.Reversal rate changes in hsp::lin-12(RNAi) animals. Filled triangles indicate hsp::lin-12(RNAi) animals; filled circles indicate wild type control animals; and open squares indicate hsp::grk-2(RNAi) control animals. Animals were heat shocked at 33°C for 2 hours, allowed to recover at 25°C then tested 4 hours later. After recovery at 25°C overnight (approx. 16 hours), animals were tested again. \\\ p\<0.001 vs. wild type. B.Reversal rate changes in lin-12(gfcs) animals. lin-12(gfcs) animals raised at the permissive temperature (25°C) are indicated by open squares, and those raised at the restrictive temperature (15°C) by filled squares. Open and filled circles indicate wild type animals raised at 25 and 15°C, respectively. Temperature shifts from 25 to 15°C are indicated by solid lines, and those from 15 to 25°C are indicated by dotted lines. \* p\<0.01 vs. t = 0 hrs.; \\ p\<0.01 vs. t = 4 hrs.; \\\* p\<0.001 vs. t = 0 hrs. ::: ![](1471-2202-6-45-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### **lin-12is not required in the gonadal or vulval lineages to regulate reversal rates.*The somatic gonads and vulval tissues were eliminated by killing the progenitor Z1 and Z4 cells in L1 larvae with a laser microbeam (see text for details). Successful ablation of Z1 and Z4 was scored visually as follows: animals in which both Z1 and Z4 were ablated lacked gonads and vulvae; animals in which only one of the two cells were killed results in a protruding vulva; and animals in which neither cell was killed resulted in fertile animals with normal vulvae. Only animals in which both Z1 and Z4 were killed were scored for behavior. The reversal rates of mock treated lin-12(RNAi) and lin-12p::lin-12(OE) animals were slightly lower compared to untreated animals, but they were still significantly higher than that of wild type. \\* p\<0.01, \\\* p\<0.001. Statistical comparisons are to wild type. ::: ![](1471-2202-6-45-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### LIN-12::GFP is expressed in RIG neurons.RIG neurons are indicated by arrowheads. Expression of LIN-12::GFP was detected in approximately 25% of L1/L2 animals. The identity of RIG neurons was confirmed by using a nmr-1::dsRedreporter gene that labels the AVG neuron, which is located in between the RIG neurons (data not shown). Scale bar = 10 μm. ::: ![](1471-2202-6-45-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### **lin-12likely acts in a subset of glr-1 expressing neurons to regulate reversals.A.*RNAi driven by the glr-1promoter expressing dsRNA does not spread to nearby tissues. The glr-1promoter drives expression in the command interneurons AVA, AVB, AVD, AVE, and PVC, as well as AIB, AVG, AVJ, DVC, PVQ, RIG, RIM, RIS, RMD, RMEL/R, SMD, and URY, all of which (except DVC, PVQ, and PVC) are located in the head 48, 49. The lin-12cDNA fragment used to express lin-12dsRNA also contains GFP sequences in cis. The dsRNA expressing construct was introduced into strains expressing osm-10p::gfpin ASH neurons or elt-2p::gfpin intestinal cells, and compared to control strains for GFP expression levels. Adult animals from multiple transgenic lines were scored; representative images are shown. B.Effect of altering lin-12activity in glr-1expressing neurons. glr-1pindicates the glr-1promoter used to drive expression of various transgenes, and glr-1p::(0)indicates the promoter only control. lin-12(RNAi) and gfp(RNAi) indicate lin-12and gfpdsRNA, respectively. lin-12(OE) indicates transgenic animals injected with a rescuing lin-12cDNA construct at a high concentration (see Methods). lin-12ICindicates a truncated, activated lin-12allele that lacks the extracellular domain.\\*p\<0.01 and \\\*p\<0.001 vs. wild type, respectively. C.Expressing lin-12cDNA in glr-1expressing neurons rescues the lin-12(lf) reversal defect. \\\* p\<0.001. ::: ![](1471-2202-6-45-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### *The RIG neurons are a likely site of lin-12gain of function action.To facilitate RIG neuron identification, these experiments were carried out in a flp-18p::gfpbackground (see Methods). lin-12(gfcs) animals were raised at 25°C, then were shifted as young adults to 15°C 4 hours prior to behavioral assays. \ p\<0.05, \\\* p\<0.0001. ::: ![](1471-2202-6-45-6) ::: ::: {#F7 .fig} Figure 7 ::: {.caption} ###### *Genes that interact genetically with lin-12to regulate reversals.A.*lag-1and lag-2likely function with lin-12to regulate reversals. Complete loss of function in lag-1and lag-2cause lethality; therefore, partial loss of function alleles were used (om13and q420, respectively). lag-2(sd)is lag-2(sa37), a semidominant suppressor of lin-12gain of function. B.The AMPA/kainate glutamate receptor glr-1genetically interacts with lin-12. glr-1(lf) is glr-1(n2461), a genetic null allele. glr-1p::glr-1(OE) is nuIs25(see Methods for details). \\\* p\<0.001 vs. wild type. ::: ![](1471-2202-6-45-7) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Allelic series of lin-12mutants. Vulval phenotype abbreviations are as follows: Pvl, protruding vulva; WT, wild type; Vul, vulvaless; and Muv, multiple pseudovulvae. All animals were tested at 25°C except animals carrying the cold-sensitive lin-12(n137n460)allele were raised at the non-permissive temperature 15°C. Control animals raised at 15 or 25°C had wild type reversal rates (see Figs. 1 and 2). ::: strain vulval phenotype reversals/3 min. ± S.E.M. n --------------------------------- ------------------ --------------------------- ---- lin-12(n941)(null) Pvl 17.3 ± 1.7 15 lin-12(n941)/+ WT 10.8 ± 1.5 15 lin-12(+) WT 10.5 ± 0.7 45 lin-12(n302)/lin-12(n941) Vul 9.3 ± 1.3 15 lin-12(n302) Vul 8.8 ± 1.1 13 lin-12(n379) Vul 7.7 ± 0.5 15 lin-12(n676) Vul 6.0 ± 0.3 5 lin-12(n137n460)/lin-12(n941) WT 22.9 ± 1.5 12 lin-12(n137n460)/+ WT 20.1 ± 1.8 9 lin-12(n137n460) Muv 23.7 ± 1.8 12 lin-12(n137)/lin-12(n941) Muv 17.6 ± 1.2 5 lin-12(n137)/+ Muv 6.0 ± 1.6 7 lin-12(n427) Muv 3.9 ± 0.2 15 lin-12(n137) Muv 3.6 ± 1.1 10 ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Basal locomotion rates of animals with altered lin-12activity. Animals were tested under identical conditions as reversal assays in 10 second bins. A single body bend was scored as a complete dorsal to ventral oscillation. Only forward moving animals were scored; if an animal reversed direction during the assay, that data point was discarded. ::: strain body bends/10 sec. ± S.E.M. n pvalue ----------------------- ----------------------------- ---- ----------------------- wild type 5.2 ± 0.2 26 lin-12(lf) 3.9 ± 0.2 43 \<10^-5^vs. wild type lin-12(gfcs) 25°C 5.1 ± 0.2 20 lin-12(gfcs) 15°C 3.8 ± 0.2 24 \<10^-4^vs. wild type lin-12(RNAi) 5.3 ± 0.3 24 lin-12p::lin-12(OE) 5.2 ± 0.2 24 :::	PubMed Central	2024-06-05T03:55:59.819447	2005-7-12	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181819/", "journal": "BMC Neurosci. 2005 Jul 12; 6:45", "authors": [ { "first": "Michael Y", "last": "Chao" }, { "first": "Jonah", "last": "Larkins-Ford" }, { "first": "Tim M", "last": "Tucey" }, { "first": "Anne C", "last": "Hart" } ] }
PMC1181820	Background ========== In 1937, Hulusi Behçet, a Turkish dermatologist, first described a chronic autoimmune disease bearing his name with characteristic orogenital aphtous ulceration and uveitis \[[@B1]\]. The etiology of Behçet\'s disease, which is most commonly seen in Asia and Mediterranean area, is still unknown. The primary pathology is a vasculitis affecting skin, joints, pulmonary, gastrointestinal, urinary, and nervous systems \[[@B2]\]. Its vascular complications are most frequently manifested as thromboembolism in veins and pseudoaneurysm in arteries \[[@B3]\]. Although pseudoaneurysms are the most common form of arterial involvement in Behçet\'s disease, we could only find one case reported by Rolland et al.\[[@B4]\] with Behçet\'s disease and cardiac pseudoaneurysm. Occasional cases of cardiac pseudoaneurysms have been reported in association with rheumatoid arthritis \[[@B5]\] and Kawasaki\'s disease \[[@B6]\]; however large cardiac pseudoaneurysms are mostly complications of cardiac surgery, myocardial infarction, endocarditis, and chest trauma \[[@B7]\]. In this report we present a patient with Behçet\'s disease and a huge left ventricular pseudoaneurysm. Case presentation ================= A 13 years old boy with Behçet\'s disease was referred to our hospital with chills, fever, cough, and chest pain of one month duration in June 2001. The diagnosis of Behçet\'s disease was established 4 years prior to this admission, that had presented with oral aphtae, orchitis, right eye uveitis leading to blindness, recurrent pseudofolliculitis, knee arthritis, and lower extremity deep vein thrombosis, all attributable to this autoimmune disorder. He had been treated by Prednisolone (15 mg/day) and Methotrexate (7.5 mg/week) for last 7 month. Physical examination revealed III/VI to-and-fro murmur along the left sternal border and an S3 gallop. The posterior-anterior chest X-ray (Fig. [1](#F1){ref-type="fig"}), computed tomography scan with intravenous contrast media at the level of T5--T8 (Fig. [2](#F2){ref-type="fig"}), magnetic resonance imaging (MRI) (Fig. [3](#F3){ref-type="fig"}), coronary angiography and echocardiography all revealed a 10.1 × 14.8 cm left ventricular pseudoaneurysm in the anterior wall of left ventricle. There was no narrowing suggesting coronary artery disease in his coronary angiography. In two-dimensional echocardiography, the typical features of pseudoaneurysm was noted including a relatively narrow neck in comparison with the diameter of the aneurysm, sharp discontinuity of the endocardium at the site at which the aneurysm communicates with the left ventricle, no noticeable valvular dysfunction, and left ventricular wall motion abnormalities. The orifice to radius ratio was not measurable due to the large size of pseudoaneurysm. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### The PA chest X-ray: left-sided pleural effusion and a large mass in anterolateral part of left lung which had overshadowed the left border of the heart. Arrowhead indicates shift of the heart to the right side. ::: ![](1471-2482-5-13-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Chest computed tomography scan with contrast at the level of T7 showing the large pseudoaneurysm. The lesion was a well-defined partially calcified mass with tubular density adjacent to the heart. Arrowheads indicate the calcifications. It was filled with contrast medium concurrently with the heart. This lesion, which was mostly occupied by thrombosis, had a mass effect on heart. ::: ![](1471-2482-5-13-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### The MRI indicated a mass with inhomogeneous signals implying the presence of blood and clots in addition to calcification. (SVC = Superior Vena Cava; IVC = Inferior Vena Cava; LV = Left Ventricle; P = Pseudoaneurysm) ::: ![](1471-2482-5-13-3) ::: At surgery, the ostium (arrowhead in Fig. [4](#F4){ref-type="fig"}) was measured 2.5 × 3.0 cm. The above mentioned results warranted a surgery with median sternotomy approach. After pericardotomy, a pulsatile mass appeared at the tip of left ventricle with a fistula to the heart. The pseudoaneurysm and large amounts of thrombus within it were resected, and the defect in the left ventricular wall was repaired by Teflon Plegeted Prolen 4/0. The patient made an uneventful recovery. The pathologic examination revealed a fibrous pseudoaneurysm including areas of old hemorrhage and thrombosis and chronic inflammation. There was no complication for next 24 months follow up period after the operation. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Chest computed tomography scan with contrast at T7 level. Arrowhead marks the ostium. ::: ![](1471-2482-5-13-4) ::: Conclusion ========== We have presented an unusual patient with Behçet\'s disease and a large (10.1 × 14.8 cm) left ventricular pseudoaneurysm. Behçet\'s disease is a systemic disorder with mucocutaneous, ophthalmic, neurological, cardiovascular, pulmonary, gastrointestinal, urogenital and musculoskeletal involvement. Its vascular manifestations are thrombophlebitis and, less frequently, arterial lesions such as pseudoaneurysms, occlusions and stenoses \[[@B8]\]. About 8% of the patients with Behçet\'s disease have severe vascular complications such as arterial pseudoaneurysms and occlusions \[[@B1]\]. Pseudoaneurysms are the most common form of arterial involvement in Behçet\'s disease \[[@B3]\]. Cardiac involvement is rare in Behçet\'s disease \[[@B9]\] and engages only about 6% of patients \[[@B4]\]. Cardiac involvement in this disorder is a diffuse process that involves both cardiac structure and vascular elements. Higher incidences of interatrial septum aneurysm (6% to 31%), mitral valve prolapse (3% to 25%), mitral regurgitation (6% to 40%), and aneurysmal dilatations of valsalva sinus and ascending aorta were observed in the Behçet\'s disease patients than in the normal subjects \[[@B10]\]. Pericarditis, myocarditis, endocardial fibrosis, conduction defects, and aortic regurgitation were also observed \[[@B11]\]. Although left ventricular aneurysms with Behçet\'s disease have occasionally been reported, we found only one case of cardiac pseudoaneurysm in these patients reported in literature. Rolland et al.\[[@B4]\] reported a 29 years old patient with Behçet\'s syndrome and a false left ventricular aneurysm and coronary artery aneurysm, which were repaired under cardiopulmonary bypass with no postoperative complications. Cardiac pseudoaneurysm is defined as a rupture of the myocardium that is contained by pericardial adhesions or the epicardial wall. This phenomenon can be explained by myocardial fragility induced by ischemia due to vasculitis process of Behçet\'s disease \[[@B4]\]. Because there was no evidence suggesting coronary artery disease in his coronary angiography of our patient, it could be assumed that this pseudoaneurysm was resulted from rupture of the left ventricle due to angiitis and the myocardial fragility induced by ischemia. In contrast to a true ventricular aneurysm, in which the wall is composed of myocardial scar tissue, the wall of a pseudoaneurysm is composed of thick fibrous tissue and pericardium \[[@B12]\]. In our case, pseudoaneurysm was consisted of profuse fibrous tissues and thromboses in various stages. Cardiac pseudoaneurysms have the potential to leak or rupture and can be the source of peripheral emboli \[[@B10]\]. Different reports have discussed that such a contained rupture has a greater propensity for rupture than a true aneurysm, whose wall contains myocardium. Rupture of a left ventricular pseudoaneurysm is usually fatal; hence appropriate recognition and treatment (early surgery) even for asymptomatic patients is strongly recommended \[[@B7]\]. The diagnosis of pseudoaneurysm is not straightforward and is rarely suggested by clinical signs and symptoms \[[@B7]\]. In our patient the pseudoaneurysm presented with nonspecific symptoms and signs. Thus, such a diagnosis was highly unlikely before getting the results of imaging. Various imaging methods have been used to diagnose pseudoaneurysm, including two-dimensional and contrast echocardiography \[[@B13]\], computed tomography, magnetic resonance imaging, and left ventricular angiography. Each has its advantages and disadvantages but echocardiography has become the most common examination used for first diagnosis because it can evaluate other associations such as valvular regurgitation, thrombus formation, and ventricular function, are often important supplements to clinical management \[[@B7]\]. Like our patient, chest radiography sometimes shows a localized bulge on the cardiac silhouette. On computed tomography, pseudoaneurysms are characterized by an abrupt disappearance of the myocardial wall at the border of the pseudoaneurysm. Magnetic resonance imaging shows the low signal of the pericardium, which constitutes the only wall of the pseudoaneurysm \[[@B7]\]. Surgical repair is usually recommended when a left ventricular pseudoaneurysm is detected \[[@B14],[@B15]\]. In this case surgical intervention was mandatory, partly due to the young age of the patient. However, in cases of post-infarction left ventricular pseudoaneurysm, which is one of the most common causes, routine surgical repair regardless of other clinical characteristics of the patient remains as a matter of discussion. Some authors believe that the necessity of surgical repair in these cases should be individualized for each patient \[[@B7],[@B16]\]. In patients with post-infarction left ventricular pseudoaneurysm, surgical repair of pseudoaneurysm was associated with an acceptable surgical mortality rate \[[@B7]\] and the long term outcome appears relatively benign \[[@B13]\] and late death was related primarily to the underlying disease or cardiac dysfunction \[[@B7]\]. There is no data on the long term prognosis of patients with cardiac pseudoaneurysms in Behçet\'s disease; however, long-term survival could not be expected given the diffuse involvement of cardiac structure and vascular elements \[[@B10]\]. Considering its fatality and nonspecific manifestations, one should consider cardiac pseudoaneurysms as a potential risk in any patient with Behçet\'s disease. Thanks to early diagnosis and surgery, our patient was treated successfully and had no complications in a follow-up period of 24 months. Abbreviations ============= SVC = Superior Vena Cava IVC = Inferior Vena Cava LV = Left Ventricle Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= SMM: Data collection, participated in the design of the study, critical review of the manuscript. PE: Conceived the study and wrote the manuscript, participated in the design of the study. MHM: Supervisor and conductor of treatment team and critical reviewer of the manuscript. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2482/5/13/prepub> Acknowledgements ================ Authors wish to thank Dr. Mohammad Kazazi, Dr. Abbas Salehi-Omran, Dr. Vafi Salmasi, Dr. Zahra Alizadeh, Dr. Golnar Mortezaie and Dr. Alireza Moayeri for their invaluable help.	PubMed Central	2024-06-05T03:55:59.822895	2005-6-14	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181820/", "journal": "BMC Surg. 2005 Jun 14; 5:13", "authors": [ { "first": "Seyed Mojtaba", "last": "Marashi" }, { "first": "Payam", "last": "Eghtesadi-Araghi" }, { "first": "Mohammad Hussein", "last": "Mandegar" } ] }
PMC1181821	Background ========== In recent years, isoflavones have increased in popularity as an alternative to conventional hormone replacement therapy for the relief of hot flashes and other symptoms associated with menopause. Currently, isoflavones are available as tablets, capsules, powders (particularly soy protein powders), drinks and bars \[[@B1]\] as well as a component of traditional soy foods. Typically, supplements provide 25--100 mg total isoflavones if consumed according to package directions \[[@B1]\]. Yet, due to the increasing variety of soy foods in the marketplace, consumers can easily consume 100 mg or more of total isoflavones each day from the diet alone. Although soy foods have been available for millennia, isoflavone supplements are relatively new and few drug/supplement or nutrient/supplement interactions have been identified \[[@B1]\]. However, consumers should be advised to use caution when taking isoflavone supplements because the potential for unidentified interactions does exist. This case study presents a postmenopausal woman who experienced an isoflavone/nutrient interaction, which resulted in a hypertensive crisis requiring medical intervention. Case presentation ================= A 51-year-old postmenopausal non-Hispanic white woman was treated for a hypertensive crisis at a regional medical center in eastern Arizona. She had complained of symptoms for one week prior to admission, including light-headedness, headaches, and high blood pressure by self-measurement. Ten days prior to admission, the patient had been enrolled in a university-sponsored research trial designed to investigate the extent to which vitamin C and soy isoflavones, as supplements to a habitual diet, could provide antioxidant effects by reducing in vivooxidative damage to cells, either alone or synergistically. During trial screening the patient reported typically consuming soy or soy products twice a week; no regular alcohol consumption; no history of hypertension or cardiovascular disease (although there was a family history of mild hypertension); no current medical supervision or care for any chronic health problems; no current use of over-the-counter or prescription medications and a routine exercise pattern of three times a week for 30--60 minutes. The participant weighed 175 pounds (79.5 kg), stood 5\'8\" (1.73 m), with a body mass index of 26.7 kg/m^2^. Early in the research trial, the patient was randomized to receive 500 mg vitamin C plus 5 mg/kg body weight soy isoflavones. On trial day 3, the patient reported to the investigators that she felt \"odd\" and \"light-headed.\" At the time, this was not attributed to the study-related supplements because the participant reported experiencing infrequent headaches for the past 20 years. On trial days 6 and 7 of the treatment period, the participant had her blood pressure checked by an automated machine; the readings were in the range of 140--150/92--98 mmHg vs. her usual BP of 120/82 mmHg. Due to this unexpected occurrence, the investigators requested that she stop consuming the supplements and drop out of the study. The incident was reported the university\'s Institutional Review Board Research Compliance Office, and the research trial was allowed to continue. Unbeknownst to the investigators, the participant chose to ignore the request to discontinue the supplements and continued to take the supplements on trial days 8 and 9. On trial day 9 she found her BP to be 159/110 mmHg. That night, she experienced an intense headache, a feeling of anxiety, and difficulty sleeping. Around midday on trial day 10, she stopped by a regional medical center to have her BP checked by a medical professional before going hiking. At that time, her BP was 226/117 mmHg; she reported that \"my head feels like it is going to explode\" and she was admitted to the emergency room. Laboratory analyses, including a complete blood count, metabolic panel and thyroid stimulating hormone test, were all within normal limits. A CT scan of the head showed no abnormalities or intracranial hemorrhages and a 12 lead EKG showed a normal sinus rhythm. At this time, the participant reported to the physician a 20-year history of chronic headaches that had resolved with better sleep habits and a higher fluid intake. The participant was then given 20 mg of the alpha~1~and beta-blocker labetalol HCl via intravenous infusion (see Table [1](#T1){ref-type="table"}). Subsequent to administration of the medication, the participant\'s blood pressure slowly dropped below critical levels, but did not reach normal limits. She was dismissed from the emergency room after 3 1/2 hours with a prescription for the non-selective beta-blocker propranolol HCl (Inderal LA), 80 mg once a day. She was told to discontinue the supplements that she was taking for the research trial. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Vital signs and pertinent events during emergency room admission ::: Blood Pressure ---------- ------ ---------------- ----- ---- ----- ------------------------------------------------------------- March 10 1301 226 117 87 20 1310 12 lead EKG performed Blood drawn for laboratory analyses 1334 203 115 75 18 1339 190 110 70 20 Taken to CT 1345 178 112 78 16 1349 176 100 78 18 1354 177 111 71 12 1357 177 118 72 10 1359 187 106 70 18 Medicated via IV: Labetolol HCl 20 mg Pt. stated ↓ headache 1404 162 96 67 11 1409 159 94 74 16 1414 169 111 72 15 1419 177 108 72 17 1424 181 104 73 27 1429 163 93 69 13 1434 169 97 69 15 1439 170 111 72 22 1444 162 97 84 14 1449 168 104 70 13 Pt. stated headache gone 1454 171 104 65 12 1459 166 92 69 14 1504 190 106 74 18 1509 181 107 69 34 1514 177 106 76 13 1519 163 104 71 20 1524 178 108 70 14 1529 176 106 64 20 1534 168 102 89 9 1539 162 100 70 16 1544 168 102 68 14 1549 165 102 70 12 Medicated PO: Inderal LA 80 mg 1618 179 117 69 N/A 1630 Discharged to home ::: The patient notified the trial investigator of the hypertensive event several days later. The hypertensive crisis was reported to the university\'s Institutional Review Board Research Compliance Office, and the research trial was allowed to continue with the stipulation that all participants submit to blood pressure monitoring weekly. Later that week, the participant\'s BP was measured by the primary investigator\'s staff and was still above normal limits. When the participant saw a cardiologist and her regular physician for further follow-up, no abnormalities in cardiac function, renal function or hormone levels were identified that could have led to the hypertensive crisis. The participant continued on antihypertensive medications for the next 12 months and was gradually able to decrease the dose of the medications over time. Discussion ========== One plausible explanation for the hypertensive crisis experienced by this participant is the inhibition of monoamine oxidase by the isoflavones (e.g., daidzin, daidzein) or their metabolites (e.g., equol). Rooke et al.\[[@B2]\] and Gao et al.\[[@B3]\] both reported that daidzin, the plant precursor of the mammalian metabolite daidzein, and some of its structural analogs can inhibit mitochondrial monoamine oxidase in vitro. Additionally, Dewar et al.\[[@B4]\] reported that equol, a mammalian metabolite of daidzein, was an effective inhibitor of rat liver monoamine oxidase in vitro. Since the soy isoflavone supplements used in the research trial consisted of 63% (178 mg aglycone units/g) genistein, 28% (79.1 mg aglycone units/g) daidzein and 9% (24.6 aglycone units/g) glycitein (percentages based on aglycone units), the daidzein in the supplement may have interacted with monoamine oxidase. Monoamine oxidase is responsible for the deamination of monoamines, including serotonin, epinephrine, norepinephrine, dopamine and tyramine. Its inhibition will cause an increase in the blood levels of these compounds. Since tyramine acts as a vasoconstrictor, an increased tyramine level will cause an increase in blood pressure \[[@B5],[@B6]\]. Review of the two-day food records recorded prior to the participant\'s entering the study in addition to dietary information obtained after the hypertensive event indicated the participant\'s normal diet typically contained multiple tyramine-containing foods. The participant confirmed that she had consumed several tyramine-containing foods during the study, including the day before and the day of her emergency room admission (Table [2](#T2){ref-type="table"}). Thus, the high dose of supplemental isoflavones \[397.5 mg isoflavones (aglycone units) containing approximately 111 mg daidzein (aglycone units)\], in conjunction with her typical moderate to high tyramine diet, may have contributed to a monoamine oxidase inhibitor-type reaction. Although the studies by Rooke et al.\[[@B2]\], Gao et al. \[[@B3]\] and Dewar et al.\[[@B4]\] suggest such a reaction might be possible, we believe this is the first report published of a possible monoamine oxidase inhibitor reaction and subsequent blood pressure spike occurring in vivodue to intake of a soy isoflavone supplement. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Participant\'s Dietary Intake ::: ------------------------------------------------------------------------------------------- Dinner on Day 9, 5:30 p.m. ------------------------------------------------------------------------------------------- \Yoplait fat free yogurt--12 ounces\ Peanuts, salted--1/4 cup\ Navel orange--1 medium\ \Banana, ripe--1 medium\ \Avocado, ripe--1 small\ Potato chips--1 handful\ Jelly beans--1/4 cup\ \3 Musketeers bar--1/3 of bar\ Vanilla ice cream--1/2 cup Breakfast on Day 10, 8:00 a.m. \Coffee--21 ounces\ \Bacon--3 slices\ Eggs, scrambled--2 whole\ Toast--1 slice with \~1 teaspoon margarine \* signifies tyramine-containing foods Foods from participant\'s Typical Diet containing tyramine or other pressor agents\[5,6\] Balsamic vinegar--1--2 teaspoons daily\ Cheddar cheese--2--4 ounces daily\ Mozzarella cheese--1 ounce daily\ Yogurt--16 ounces daily\ Dried beans or legumes--1/2 cup daily\ Coffee--17--21 ounces daily\ Bananas--1 every other day\ Avocado--3 times/week\ Tamari sauce--1 tablespoon 2 times/week\ Swiss cheese--2 ounces/week\ Cured meats--1 time/week\ Raisins--2--3 times/month\ Spinach--2--3 times/month\ Blue Cheese--2--3 times/month\ Chocolate--occasionally ------------------------------------------------------------------------------------------- ::: A second plausible explanation for the hypertensive crisis experienced by this participant is an imbalance in the renin-angiotensin system, an important regulator of blood pressure, due to the administration of the isoflavones. Isoflavones are known to bind to both the α and β estrogen receptors and exert weak estrogenic effects in vivo \[[@B7],[@B8]\]. Because angiotensinogen production by the liver is modulated by estrogens, the assumed increase in the serum isoflavone concentrations due to the high isoflavone intake may have stimulated an estrogenic response, thereby increasing hepatic angiotensinogen production and release into the plasma \[[@B9]-[@B11]\]. Once cleaved by renin, angiotensinogen becomes angiotensin I which is rapidly converted to angiotensin II by the angiotensin-converting enzyme \[[@B12]\]. Angiotensin II acts on the outer layer of the zona glomerulosa of the adrenal cortex, converting corticosterone to aldosterone, which subsequently increases renal sodium reabsorption as well as extracellular fluid and blood volume resulting in an increase in blood pressure \[[@B12]\]. Thus, the high dose of supplemental isoflavones consumed by this participant may have caused an imbalance in the renin-angiotensin system, the end result of which was the hypertensive crisis that the participant experienced. Conclusion ========== Due to the availability and increasing popularity of soy supplements, practitioners should be aware of the potential side effects associated with their use. This case study reports two plausible reactions, a monoamine oxidase inhibitor-type reaction or an imbalance in the renin-angiotensin system, which may have occurred with consumption of a high-dose isoflavone supplement resulting in the participant experiencing a hypertensive crisis. Although this reaction occurred within the context of a research study, it is possible that similar reactions might occur in general population if the dosage guidelines listed on the soy isoflavone supplements are exceeded. Practitioners counseling clients who are consuming soy isoflavone supplements should advise them that elevated blood pressure may be a potential side-effect to consider and monitor. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= AH participated in the design and coordination of the study and drafted the manuscript. IM participated in the design of the study, participated in conducting the laboratory analyses, and helped to draft the manuscript. CJ participated in the design of the study, performed the statistical analyses, and helped to draft the manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1472-6874/5/9/prepub> Acknowledgements ================ This study was funded by the Sustainable Technologies, Agribusiness and Resource Center, Arizona State University, Mesa, AZ 85212, USA. Written consent was obtained from the patient for publication of the study.	PubMed Central	2024-06-05T03:55:59.823861	2005-6-23	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181821/", "journal": "BMC Womens Health. 2005 Jun 23; 5:9", "authors": [ { "first": "Andrea M", "last": "Hutchins" }, { "first": "Imogene E", "last": "McIver" }, { "first": "Carol S", "last": "Johnston" } ] }
PMC1181822	Probably, most of the students who have taken (or will take) undergraduate courses based on the so-called exact sciences have faced (or will face at some point) the terror of being unable to solve some of the proposed problems or exercises, despite having a good theoretical knowledge of the topic. Basically, you may know the equations, you may understand the mathematical passages, you may have a grasp on the physical meaning of the equations and yet sometimes you are unable to get past the problem\'s enunciate! Unfortunately, this inability to solve problems often leads to frustration. And this feeling of helplessness starts a vicious circle, as the incapability of solving some problems results in diminished motivation to study and, as a consequence, more difficulties in solving future problems. Surely, problem solving is an important and complex aspect of our education; however it is frequently overlooked by teachers. For this reason, a book aimed at the development of problem solving skills should be welcomed by students and teachers. More than 50 years ago, G. Polya wrote a book on problem solving strategies which became very popular, \"How to Solve It\" \[[@B1]\]. Brigg\'s book is not intended to replace, but rather to complement Polya\'s classic book. For example, not only Polya\'s principles of analytical problem solving are cited and discussed in Brigg\'s book, but they are also adapted to computational problems. In fact, the interplay between the analytical and computational approaches to problem solving is very useful since computers became part of our daily lives. One of the main qualities of \"Ants, Bikes & Clocks: Problem Solving for Undergraduates\" is that it is a pleasant book to read. It is well written and the understanding of the text is facilitated by numerous tables, graphs, charts and figures. This readability is especially important considering the main target of the book, undergraduate students. But everyone would certainly enjoy browsing the chapters, reading the small notes on the sides of the pages, going through the various and inventive examples, etc. I did it! Another interesting aspect of Brigg\'s book is how the exercises are organized. At the end of the chapters, each exercise is followed by hints and respective answer. In addition, full solutions for all exercises can be found at the end of the book. Although the author cautions against checking the full solutions before trying to solve the problems, it is certainly good to know that the full solutions are within reach. In summary, this is an excellent and opportune book about problem solving, not only for undergraduates, but for everyone who is interested in problem solving strategies.	PubMed Central	2024-06-05T03:55:59.825420	2005-7-24	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181822/", "journal": "Biomed Eng Online. 2005 Jul 24; 4:45", "authors": [ { "first": "Eduardo", "last": "Abreu" } ] }
PMC1181823	Introduction ============ Thailand is one of the countries having most success fighting the epidemic of HIV/AIDS infection \[[@B1],[@B2]\]. Two randomised clinical trials conducted in Thailand have provided substantial impact on prevention of mother to child transmission of HIV/AIDS (PMTCT). The first demonstrated in 1999 that a short course of twice daily oral Zidovudine (AZT) was safe, well-tolerated and, in the absence of breast-feeding, lessened the risk for mother to child HIV-1 transmission from 18.9% to 9.4% \[[@B3]\]. This prompted the Thai Government to provide universal access to a short course of AZT in 2000 \[[@B4],[@B5]\]. The second trial recently reported that a combination of AZT and a single dose of Nevirapine (NVP), administered both to the mother during labour and to the newborn, resulted in only two percent of children being born with HIV \[[@B6]\]. On release of these results, the Thai National Perinatal HIV Prevention programme incorporated this new regimen into its policy \[[@B7]\]. By 2004 \"the current practice\" is 300 mg oral AZT twice a day started at 28 week of gestation, a single dose of 200 mg NVP at onset of labour plus intra-labour 300 mg AZT orally every three hours until delivery. Newborns receive a single dose of NVP 6 mg after birth and AZT 2 mg per kg every six hours for 7 days if the mother received 4 or more weeks of AZT. The children born to mothers arriving late in pregnancy or to mothers who received AZT for less than 4 weeks, are given a six-week AZT regimen \[[@B7]\]. The Thai PMTCT programme provides free services for two rounds of Voluntary Counselling and Testing (VCT) for all pregnant women, at first antenatal (ANC) visit and at 28 weeks. The reason for the second VCT is to detect the newly HIV infected during pregnancy. HIV infected pregnant women receive free antiretroviral drugs, breast milk substitutes for 12 months and counselling with their partner to test their newborn at 12 and 18 months. The Ministry of Public Health (MOPH) purchases drugs and artificial milk in bulk and distributes it via its regional networks \[[@B8]\]. While modelling in a Sub-Saharan African setting \[[@B9]\], revealed that a single NVP dose provided to both mother and baby could be highly cost-effective in high sero-prevalence settings, the cost-effectiveness of NVP therapy in an Asian setting where prevalence is lower is unknown. On the other hand, several studies have raised concern about the high rate of NVP resistance developing in mothers treated with a single regimen \[[@B10]-[@B13]\]. The rate was as high as 30--40% in South-Africa \[[@B14]-[@B16]\] and 17% in Thailand \[[@B17]\] and this would affect the choice of antiretroviral therapy (HAART) if the mother needs to be treated later on \[[@B18]\]. The purpose of this study is to appraise the cost-effectiveness of the regimen introduced in 2004, compared (regimen D in table [1](#T1){ref-type="table"}) with several alternatives: 1) the previous Thai practice -- a short course AZT regimen (regimen A); 2) the cheapest regimen consisting of a single NVP dose (regimen B); and 3) a mixed regimen of short course AZT for ANC arrivals at 34 weeks of gestation, and NVP for late arrivals beyond 34 weeks and for those who refuse the AZT regimen (regimen C). The last option is designed to maximise the effectiveness and minimise the problem of drug resistance. Furthermore, the analysis assesses the value of a second VCT round by comparing the cost-effectiveness of one and two maternal VCT for each of the four drugs options. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Protocol of four drug regimens for health economic evaluation in Thai HIV transmission cost-effectiveness model ::: Code Drug regimens Zidovudine (AZT) Nevirapine (NVP) ---------- ------------------------------------------------------ --------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------- -------------------------------------------------------- A A short course AZT (practice in Thailand to 2004) Starting from 32--34 week of gestation onward + intrapartum doses From birth for 7 days (6 weeks in the case of the mother receiving \<4 weeks AZT) Not provided Not provided B NVP alone (never adopted in the national policy) Not provide Not provide Intrapartum single dose Single dose after delivery C AZT or NVP (never adopted in the national policy) Start at 32 but not latter than 34 weeks of gestations + intrapartum doses but not with NVP From birth for 7 days (6 weeks in the case of the mother receiving \<4 weeks AZT) -- and given only the cases that mother received AZT If mother know HIV status after 34 weeks then give single dose but not with AZT Single dose after delivery if mother received NVP only D AZT and NVP (current practice, commencing from 2004) Starting from 28 week of gestation onward + intrapartum doses From birth for 7 days (6 weeks in the case of the mother receiving \<4 weeks AZT) In trapartum single dose Single dose after delivery ::: Design & Methods ================ Study model ----------- We use a hypothetical cohort of 100,000 pregnancies as the study model. The decision tree in Figure [1](#F1){ref-type="fig"} presents a flow of the programme options. Cost and outcome parameters are based on Thai settings. The combination of one and two VCT sessions and four antiretroviral therapy (ART) regimens leads to eight case options being considered within the model. For simplicity\'s sake we assign codes 1 and 2 for single and double VCT strategy, and A-D for four drug regimens. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### The decision tree used to model the prevention of HIV vertical transmission.(VCT = voluntary counseling and HIV testing, GA = gestational age). ::: ![](1478-7547-3-7-1) ::: Cost analysis is conducted from the perspective of the Ministry of Public Health (MOPH) as the Thai government pays all costs for VCT, ART and substitute feeding. We measure the programme outcomes as the net cost to the public-sector payer, total number of cases of paediatric HIV infection averted, and cost per paediatric HIV infection averted. The cost-effectiveness of each intervention is calculated as (IC+AC-HC)/IA, where IC are the programme intervention costs, AC additional healthcare cost due to NVP resistance, HC the lifetime health care cost of an HIV infected infant (or cost offset), and IA the number of HIV infections averted by the intervention. We converted all cost and effectiveness at the present value (2003) with discounting of 5%. Input parameters (table [2](#T2){ref-type="table"}) --------------------------------------------------- ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Estimated base-case values and their confidential intervals (CI) for input parameters in the Thai HIV cost-effectiveness model ::: Parameters Point estimate 95%CI for sensitivity analysis Parameter distribution Data sources ----------------------------------------------------------------------------------------------------------------------- -------------------- ------------------------------------ ---------------------------- --------------------------- Epidemiology Maternal HIV infection rate 1.5% Ref.19 ANC after gestational age 34 weeks 7.4% 6.7--8.0% Beta Ref.8 Rate of perinatal HIV transmission 18.9% 13.2--24.4% Beta Ref.3 Rate of transmission via breastfeeding 12.0% 7.0--17.0%\* Beta Ref.22 Percent of HIV infection detected by second VCT 4.7% 2.70--7.5% Beta Ref.20 Rate of HIV infected mothers who, treated with NVP, developed HIV resistance to NVP 17.4% 12.0--22.7% Beta Ref.17 Rate of HIV infected mothers who need to be treated AIDS within a year after delivery 33.9% 29.6--38.3% Beta Ref.28 Efficacy of Antiretrovial therapy Odds of transmitting the virus when mother received AZT \> = 4 weeks versus placebo 0.46 0.35--0.60 Normal Ref.23 Odds of transmitting the virus when mother received AZT \< 4 weeks versus receiving AZT \> = 4 weeks 1.40 0.82--2.38 Normal Ref.24 Risk of transmitting the virus with NVP regimen versus placebo 0.51 0.33--0.79 Normal Ref. 23 Risk of transmitting the virus with AZT+NVP regimen versus receiving AZT \> = 4 weeks 0.23 0.05--0.41 Normal Ref.6 Compliance to the Programme Infected pregnant women who know their HIV status before or at 36 week of gestation and accept AZT 75% 70--90% Beta Ref.8 Infected pregnant women who know their HIV status after 36 week of gestation and accept AZT 65% 55--90%\* Beta Assumption (see text) Infected pregnant women who know their HIV status before or at 36 week of gestation, do not accept AZT but accept NVP 50% 30--70%\* Beta Assumption (see text) Infected pregnant women who know their HIV status before or at 36 week of gestation and accept NVP 85% 70--90%\* Beta Assumption (see text) Infected pregnant women who know their HIV status after 36 week of gestation and accept NVP 75% 70--90%\* Beta Assumption (see text) Infected pregnant women who know their HIV status before or at 36 week of gestation and accept AZT+NVP 84% 80--90% Beta Ref.17 Infected pregnant women who know their HIV status after 36 week of gestation and accept AZT+NVP 75% 70--80%\* Beta Assumption (see text) Programme unit cost US\$ 2003 VCT for HIV negative pregnancy 2.69 1.57--7.79 Gramma Ref. 8 VCT for HIV positive pregnancy 7.10 3.82--14.54 Gramma Ref. 8 HIV testing for baby born by infected mother 5.61 3.18--11.65 Gramma Ref. 8 Cost of antepartum AZT (per weeks) 10.50 Thai Department of Health Cost of intrapartum AZT 2.30 Thai Department of Health Cost of infant AZT (per week) 17.20 Thai Department of Health Cost of NPV for mother and infant 3.10 Price survey by authors Breast milk substitutes (per 1 year) 175.90 Thai Department of Health Incremental cost of switching from NNRTI-base treatment regimen to PI-based regimen 497 147--847 Gramma Ref.29 Public sector health expenditure Life time pediatric HIV/AIDS treatment cost 1,680 1,340--2,015 Gramma Ref.30 Note that a range for sensitivity analysis derived from 95% CI of each parameter distribution except \* that based on assumption ::: For each parameter we determine base-case values. A maternal HIV infection rate of 1.5% was reported by national sentinel surveillance 2001 \[[@B19]\]. A prospective mother-to-child study from 1992 to 1994 in Bangkok \[[@B20]\] found that with a second VCT round during pregnancy an additional 4.7% (95% CI 2.7% and 7.5%) HIV positive mothers were detected who had become infected during pregnancy. In other word, we assume the second VCT picks up 4.7% of the incidence cases. In Thailand, all pregnant women have at least one ANC visit during pregnancy \[[@B21]\]. As it takes two weeks to know the maternal HIV status the 7.4% of pregnant women arriving late for ANC are likely to receive ART later than 36 weeks compromising the optimal period for effective AZT treatment of four weeks \[[@B8]\]. The risk of perinatal transmission without treatment is 18.9% \[[@B3]\] and with breast-feeding is assumed to be an additional 12% \[[@B22]\]. Lallemant et al \[[@B6]\] report that the odds ratio of transmitting the virus by regimen D versus control (no prevention) is 0.23 (95% CI 0.05 to 0.41). The effectiveness of other drug options is derived from a Cochrane systematic review \[[@B23]\] indicating an odds ratio compared to placebo of 0.46 (95% CI 0.35 to 0.60) for regimen A and 0.51 (95%CI 0.33 to 0.79) for regimen B. To estimate the risk of transmission among late arrivals for ANC who get AZT treatment less than 4 weeks in regimen A, we apply an odds ratio of 1.40 (95% CI 0.82 to 2.38) found in a Thai study \[[@B24]\] comparing the risk of transmission between a short (from 36 weeks onwards) and long (from 28 weeks) maternal course of AZT. In the absence of evidence of the efficacy of regimen D started after 28 weeks but before 36 weeks of gestation, we assume the same efficacy if treatment is started before 34 weeks, and the lower efficacy of regimen A for those starting after 34 weeks. In 13 provinces in the North of Thailand \[[@B8]\] 75% of infected pregnant women accepted AZT treatment after knowing their HIV status. A recent study found a higher proportion (84%) of pregnant women accepted regimen D \[[@B17]\]. We use the former figure as the base case scenario for infected women who knew their HIV status before 36 week of gestation and accepted AZT treatment in regimens A and C, and the latter for regimen D. As we know that women make their first antenatal visit late in pregnancy tend to report low education and socioeconomic status \[[@B25]\]. We, therefore, assume a lower proportion of 65% of infected pregnant women who know their HIV status after 36 weeks accept AZT in programmes A and C, and 75% of those accept AZT and NVP in programme D. For those who refuse AZT treatment for whatever reason, we assume 50% would accept the simpler regimen of NVP in programme C. For the regimen of single NVP, we propose 85% of infected pregnant women who know their serologic status before 36 weeks and 75% thereafter enroll in the programme B. This enrollment rate is slightly higher than with AZT as drug administration is simpler. The study presents all prices in US\$ (USD) at 2003 price units. Costs of providing VCT, maternal and infant antiretroviral treatment are derived from 160 MOPH hospitals \[[@B8]\]. Briefly, data on the units of each category of resources used and valued were gathered by mean of a questionnaire sent to each hospital participating in the study. Intervention unit costs include recurrent labour and non-labour expenditures but exclude capital depreciation. A study in Thailand found that 17.4% (95% CI 12.0% to 22.7%) of infected mothers who were treated with a single dose NVP developed a strain of HIV resistant to the drug \[[@B17]\]. Several studies report that the resistant virus spontaneously reverts to a wide-type genotype by one year \[[@B14],[@B26]\], meaning that the resistance would not have altered effectiveness of prevention in further pregnancies. Nevertheless, the concern is that those who were resistant and were then treated for AIDS following delivery would have a much more difficult time controlling the virus \[[@B13],[@B18]\]. AIDS experts therefore recommend a more expensive treatment regimen based on protease inhibitors (PI) as a first line treatment for mothers with the NVP resistant virus instead of the common and cheaper regimen, based on Non-Nucleoside Reverse Transcriptase Inhibitors (NNRTI) \[[@B27]\]. A report from MOPH \[[@B28]\] reveals 33.9% of infected mothers need treatment for AIDS by a year after delivery and the incremental cost of switching from the cheaper to the more expensive treatment regimen is 497 USD (95% 147 USD to 847 USD) assuming the incremental cost occurs only for the first three years of the treatment \[[@B29]\]. Assuming 80% of lifetime paediatric HIV/AIDS treatment cost is shouldered by the public sector, the lifetime medical care cost was estimated at 3,300 USD in 1997 (25 Baht per 1 USD) \[[@B30]\]. Converting the 1997 figures to 2003 values using the general consumer price index, the public sector health expenditure for a case of paediatric AIDS is 1,680 USD (40 Baht per 1 USD). A range of 1,340--2,015 USD is proposed for sensitivity analysis. A clinical trial of morbidity and mortality among breast fed and formula fed infants of HIV-1-infected women in Kenya found a similar overall mortality rate, incidence of diarrhea, pneumonia and other serious complications among the two groups \[[@B31]\]. Therefore, we do not add costs of treatments for excess diarrhoea and respiratory tract infections among formula-fed infants. Uncertainty analysis -------------------- To handle uncertainty in the model input parameters of interest are ascribed a distribution that reflects the uncertainty associated with their true value (table [2](#T2){ref-type="table"}) and entered in a probabilistic uncertainty analysis. For example, the beta-distribution was the choice of distribution for probability parameters which were bounded zero-one and the gamma distribution was modelled for unit cost parameters \[[@B32]\]. The results from 1,000 calculations are presented in a CE acceptability curve based on the concept of net-benefit approach suggested by Stinnett and Mullahy \[[@B33]\] and Briggs et al \[[@B34]\]. It shows the probability in the 1,000 iterations of the model that any of the eight interventions is the most cost-effective option given different willingness to pay thresholds. Results ======= The total programme cost, including costs of VCT, ART, artificial milk and dealing with drug resistance, of option 1A is the cheapest; option 2D is the most expensive, mainly due to the costs of second VCT and dealing with the problem of drug resistance (Table [3](#T3){ref-type="table"} and Figure [2](#F2){ref-type="fig"}). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Costs, effectiveness and cost-effectiveness of 6 intervention options for the Thai HIV cost-effectiveness model, US\$ 2003 ::: Programme model 1A 1B 1C 1D 2A 2B 2C 2D ------------------------------------------------------------------------- --------- --------- --------- --------- --------- --------- --------- ----------- Programme cost 560,000 500,000 580,000 600,000 840,000 770,000 880,000 880,000 Incremental cost of switching NNRTI-base treatment to PI-base treatment 160,000 30,000 150,000 160,000 30,000 160,000 Total programme cost 560,000 650,000 610,000 750,000 840,000 930,000 920,000 1,040,000 Life time treatment cost for pediatric HIV/AIDS 390,000 430,000 460,000 560,000 410,000 450,000 500,000 590,000 Net programme cost 170,000 220,000 160,000 190,000 430,000 480,000 410,000 450,000 Number of infections averted by the program 233 258 273 337 245 271 300 353 Cost-effectiveness ratio per averted infection 716 851 570 556 1,740 1,776 1,381 1,266 Note: numbers for programme costs are given to nearest 10,000 US\$, 2003 price levels ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Total programme cost component by voluntary counselling and HIV testing (VCT) and antiretroviral drug (ARV) options. ::: ![](1478-7547-3-7-2) ::: Options 1C and 2C have the lowest net programme cost in comparison to the other drug regimens, while 1B and 2B are the most expensive alternatives, on account of the higher cost of dealing with drug resistance and smaller offset-cost. The mixed regimen of AZT and NVP with 2 VCT sessions (programme 2D) is the most effective strategy, averting 353 infections. Compared to programme 1D, the second VCT session prevents a small number of 16 (353--337) additional cases at the additional cost of 260,000 USDThe least effective regimens are those based on AZT only, followed by the single NVP regimens (1B and 2B) which prevent 25 and 26 extra infections compared to the AZT only regimens (1A and 2A). Cost-effectiveness ------------------ Compared with no routine prophylactic ART, programme 1D is the most cost-effective regimen costing 556 USD to avert one paediatric HIV infection. Option 2B is the least cost-effective regimen at more than three times the cost per additional case prevented. With relative little benefit gained from the second VCT, all regimens with the single VCT strategy are more cost-effective than those with 2 VCT sessions. Uncertainty analysis -------------------- We plotted acceptability curves using net-monetary benefit approach for the choice of prevention strategy in figure [3](#F3){ref-type="fig"}. To explain this, we consider table [3](#T3){ref-type="table"} where the most effective drug option is programme D. The incremental cost of moving from 1D to 2D is 260,000 (450,000-190,000) USD for 16 (353-337) additional infections averted. In other words, the incremental cost-effectiveness ratio of giving up the 1D regimen and adopting 2D is 16,000 USD per additional infection averted. In Figure [3](#F3){ref-type="fig"} this incremental cost-effectiveness value is represented by showing the line of 1D crossed the line of 2D at the ceiling ratio of 15,000USD (they were not exactly the same value since the first is a deterministic and the latter is a probabilistic value). If the decision maker would prefer a confidence level greater than 95%, the threshold is 35,000 USD per infection averted. Further studies would be needed to improve the accuracy of the cost-effectiveness results. In particular, better information on the proportion of infected pregnancy detected by second VCT and the cost of VCT would improve the model because these parameters have the greatest bearing on uncertainty. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Acceptability curves using net-monetary benefit approach for the choice of prevention strategy.The proportion of simulations in which a strategy has the highest net-benefit across all strategies among 1,000 replications of the model (sum of all probability at each maximum willingness to pay for one HIV infection averted or \"ceiling ratio\" equal 1). ::: ![](1478-7547-3-7-3) ::: Having mentioned above that the rate of developing NVP resistant virus among Thais was considerably lower than of Sub-Saharan setting, we therefore tested the results assuming a rate of developing HIV resistant to NVP as high as in a Sub-Saharan setting (40%). Figure [4](#F4){ref-type="fig"} illustrates that the regimens that contains no NVP (A) or contains less NVP (C) are dominant. Programme 2D is still to be the preferable choice if the willingness to pay threshold greater than 40,000 USD per HIV infection averted with the statistical confidence level 95%. It is interesting to note that 1A is a significantly dominant regimen at the lowest threshold of zero willingness to pay, that is, where a decision has been made that no further resources would be attributed to healthcare. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Acceptability curves for the choice of prevention strategy at 40% of rate of HIV resistant to Nevirapine. ::: ![](1478-7547-3-7-4) ::: Discussion ========== This study has presented an economic evaluation of a comprehensive range of VCT and choices of drug regimen for PMTCT in Thailand. We excluded a long course of AZT, ACTG 076, in our assessment since it is relatively complicated (given intravenous form of AZT to mother during labour) and expensive, and may be difficult to adopt in developing countries \[[@B35],[@B36]\]. There are two major policy concerns in this study: whether to recommend single or double VCT, and which of four drug regimens is the most cost-effective investment per infection averted in a setting with moderately high HIV sero-prevalence in pregnancy. The assessment has proved by both point estimate and multivariate uncertainty analysis that the programme D has a lower cost per paediatric HIV infection averted than other alternatives. For the choice of VCT, though, we found that 1VCT is less costly than 2VCT. International experience with the accepted cost-utility ratios suggests that cost per Life Year Gained (LYG) or Quality Adjusted Life Year (QALY) threshold is 3 × per capita GDP \[[@B37]\]. This application would presently lead to a threshold value in Thailand of between 21,000 USD per LYG or QALY or 407,000 USD per paediatric HIV infection averted (assuming 19.4 LYG per HIV infection averted and the current practice, programme 2D, clearly represents value for money. Thus, our evidence supports the new policy of the Thai National Perinatal HIV Prevention programme having introduced programme 2D as a national regimen for PMTCT. This study is partly compatible with one conducted in Mexico, another low HIV prevalence setting, \[[@B38]\]. Both studies similarly identify that VCT has a major share of total programme cost but also is essential to the efficacy of the ART programme \[[@B39]\]. Minimisation of PMTCT cost in low-prevalence setting should therefore focus on VCT costs rather than drug cost. However, the result of this study provides additional information that providing artificial milk and dealing with the problem of drug resistance also add a considerable cost. Because the authors chose to explore costs and outcomes of the PMTCT in particular context of Thai setting and use only government perspective, applying these findings to somewhere else or using other viewpoints should be done with cautious. For example, the rate of accepting VCT in Thailand may be much higher than in other countries. Also, the percentage of detecting HIV infection by the second VCT used in this study is rather high. However, this study offers a useful and comprehensive framework for evaluation of the PMTCT, especially in developing countries where resources do not permit adequate development of basic health need but shoulder a major burden of HIV/AIDS. Acknowledgements ================ We appreciate programme grant support for Senior Research Scholar Program in Health Economic and Financing by Thailand Research Fund and Health Systems Research Institute. The first author is currently supported by the World Health Organization under the Fellowship Program award to study at the University of East Anglia. Without a close collaboration with Dr. Siriporn Kanshana and her team from Ministry of Public Health and all participating hospitals, the cost effectiveness study of PMTCT would not have been possible. We wish to thank Paul Coleman, S Bertozzi & S Bautists and an anonymous reviewer for their invaluable comments on the previous version of the manuscript.	PubMed Central	2024-06-05T03:55:59.826071	2005-7-18	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181823/", "journal": "Cost Eff Resour Alloc. 2005 Jul 18; 3:7", "authors": [ { "first": "Yot", "last": "Teerawattananon" }, { "first": "Theo", "last": "Vos" }, { "first": "Viroj", "last": "Tangcharoensathien" }, { "first": "Miranda", "last": "Mugford" } ] }
PMC1181824	Background ========== Public health recommendations for physical activity (PA) state that all individuals should minimally accumulate 30 minutes or more of daily moderate intensity activity, such as brisk walking \[[@B1]-[@B3]\]. Regular walking for exercise has been associated with numerous health benefits including reduced risk of coronary heart disease\[[@B4],[@B5]\] and diabetes\[[@B6]\], weight loss/weight maintenance\[[@B7]\], and lowered blood pressure\[[@B8]\]. There is also evidence that less structured walking, i.e., walking for transport, has similar health benefits \[[@B9]-[@B11]\]. The Compendium of Physical Activities\[[@B12]\] indicates that both types of walking meet minimal intensity requirements for health-related PA. Specifically, walking for exercise (Compendium code 17200) is a 3-MET activity and walking for transport (code 17270) is a 4-MET activity. A MET is a multiple of metabolic requirements at rest; ≥ 3 METs is considered at least moderate intensity\[[@B1]\]. In the USA, walking for exercise is consistently the most prevalent leisure-time PA\[[@B13],[@B14]\] and the most frequently reported activity among adults who meet public health PA recommendations\[[@B14]\]. Nevertheless, Rafferty et al\[[@B15]\] found that only 21% of self-defined walkers did so a minimum of 30 minutes five or more times per week. Similar findings have been documented for the Australian population. For example, in the Active Australia 1999 survey, 35% of 3,814 adults surveyed reported walking 5+ \'sessions\' in the immediate past week \[[@B16]\]. A secondary analysis of the 2001 Australian Sport Commission\'s Exercise, Recreation and Sport Survey indicated that walking was the most commonly reported activity undertaken by 13,659 individuals ≥ 15 years of age surveyed, yet only 16.8% engaged in a sufficient frequency of walking for health benefit (i.e., ≥ 5 days/wk) throughout the previous year\[[@B17]\]. Finally, based on a survey of 1,773 healthy workers and homemakers aged 18--59 years (living in the Perth greater metropolitan area of Western Australia), Giles-Corti and colleagues\[[@B18]\] found that although 72.1% reported any walking for transport and 68.5% reported any walking for exercise in the previous two weeks, only 17.2% of all walkers performed a sufficient amount likely to accrue health benefits. Walking for exercise is a purposeful or structured activity that can be captured relatively easily in surveys focused on leisure time activity. In contrast, walking for transport is an incidental activity that is likely to be missed using these same assessment approaches\[[@B19]\]. Although some surveillance systems are beginning to introduce questions related to walking for transport, their measurement properties and health benefits remain largely unknown at this time. Consequently, we are unfamiliar with patterns (e.g., daily bouts or their length) of walking for transport and are therefore limited in our ability to determine its relative importance to health. There is evidence to suggest that short-term recall and diaries are a more appropriate approach for capturing such incidental walking behaviours\[[@B20]\]. In the hierarchy of PA assessment approaches, diaries are considered to be direct methods of measurement, and therefore preferred, measures of PA\[[@B21]\]. The 1997 Australian Bureau of Statistics (ABS) Time Use Survey provides a unique opportunity to examine the patterns and relative contributions of these two types of walking to recommended levels of health-related PA. This nationally representative survey captures a detailed diary (recorded in 5-minute intervals) of all of a person\'s activities over the course of two consecutive days. In addition to providing an exhaustive record of daily time spent in sport/exercise, and more specifically, walking for exercise, it includes information on time spent in various modes of transport (including by walking). Therefore, the purpose of this population analysis was to utilize these existing data to describe the patterns of walking for transport and for exercise in relation to sports/exercise participation and meeting health-related PA recommendations. Methods ======= Australian Bureau of Statistics (ABS) Time Use Survey ----------------------------------------------------- The data analysed in this paper was collected by the ABS. Subject\'s informed consent, their anonymity and the confidentiality of the data they provide is guaranteed by the legislation which established the ABS. This is in compliance with the Helisinki Declaration. The 1997 Time Use Survey recruited household members ≥ 15 years of age from over 4,550 randomly selected private dwellings in Australia. Time use diaries were collected for two specifically designated consecutive days (representing each day of the week in equal proportions) for each person during all four seasons over the calendar year. Collection and analysis of two days of data is common in time use research; further, in large samples such as this we can expect atypical days to be minimized in the aggregate, assuring generalizability of findings\[[@B22]\]. The ABS diaries were formatted in five-minute time intervals, with space for respondents to record (in their own words) their primary activity, \'what else\' they were doing at the same time (i.e., secondary activity), the location of the activity, others present during the activity, and who they did this activity for\[[@B23]\]. A team of trained coders classified the respondents\' descriptions of their activities into a nesting 3-digit code. An exhaustive classification of the activities respondents described was de factostandardised in the 1960s\[[@B24]\]. In a remarkable piece of international collaboration under the directorship of Hungarian statistician Alexander Szalai, thirteen nations simultaneously conducted time use surveys using a commonly agreed upon activity classification. Most contemporary activity classifications are derived from this source including those used in Australian time use surveys \[[@B23]\]. The raw diary data were coded accordingly and summarized into bouts and minutes of primary and secondary activities representing nine major activity categories: 1) personal care; 2) employment; 3) education; 4) domestic; 5) child care; 6) purchasing; 7) voluntary work and care; 8) social and community interaction; and, 9) recreation and leisure. A coding scheme was also available for associated travel (including walking for transport) under each of these categories. Associated travel was always coded as a primary activity. The final sample included 7,247 persons (3,471 males and 3,776 females; representing 94% household response rate and 84% person response rate) providing 14,315 diary days of data. The age breakdown was as follows: 15--24 years = 18.5%, 25--39 years = 30.2%, 40--49 years = 32.1%, and 60+years+19.2%. Almost 74% of the sample was Australian-born, 62% were married or in a common-law relationship, and modal income (40%) was less than \$300AU per week. The most widely accepted benchmark for time use data quality is the average number of activity bouts recorded in each diary, with diaries containing over 20 bouts considered to have reached the threshold of acceptable quality\[[@B25],[@B26]\]. The average number of bouts recorded in the ABS 1997 Time Use Survey was 29.1 for Day 1 and 27.5 for Day 2, confirming the good quality of the data. Data treatment and statistical analysis --------------------------------------- Common to time use research in the social sciences, the unit of analysis was diary day. This is similar to incidence density expressed as person time of exposure in epidemiology. This approach permits us to determine population participation in target activities on any given day. In the development of its 1987 pilot survey, the Australian Bureau of Statistics compared the data quality from field tests of 24-hour, 48-hour and 7-day diaries and concluded that a 48-hour diary (i.e., representing two full days) produced data as good as a 24-hour one and that data quality fell off sharply beyond 48-hours of record keeping\[[@B27]\]. Regardless, before proceeding we confirmed that there were no differences in mean time spent in target activities (described below in detail) between the two days by verifying that all of the means for one day lay within the 95% CI of the second day, and vice versa (data not shown). This process validated the use of the diary day, rather than the individual as the unit of analysis. Bouts and minutes of walking for exercise were constructed from two separate recreation and leisure category variables: walking (including for exercise) and hiking/bushwalking. Similarly, bouts and minutes of sport/exercise were constructed from four recreation and leisure category variables: general sport and outdoor activities, organised sport (e.g., competitive sports), informal sport (e.g., non-competitive sports), and exercise (excluding walking). Minutes and bouts of walking for transport were constructed for each of the nine major activity categories. Finally, minutes and bouts of walking for exercise and for transport were combined to create a variable capturing total walking. Descriptive data are presented as mean ± SD and median (and 25^th^and 75^th^quartiles of distribution) for the entire sample of days and for the participant sub-sample. Although arguably not representative of the public health impact, study of the participant sub-sample allows us to examine behaviour patterns in the \'doers\'. This terminology (i.e., doers) is accepted in time use research referring to the participant sub-sample performing the said activity and was laid down during the foundational work of Szalai et al. \[[@B24]\]; it is appropriate to continue use of common terminology in this application of time use data to PA and public health concerns. Day of week patterns were examined for the proportion of days indicating any walking for exercise or transport as well as the number of bouts, and the accumulated time engaged in these activities. Between-day differences were modelled using logistic regression (i.e., using GEE in SAS) taking into account the clustered nature of the data. Analyses were not weighted since weighting did not produce different sample means (due to the very large number of days). Achievement of public health guidelines on any given day was evaluated using two different strategies: 1) by recorded participation in ≥ 30 accumulated minutes of either walking for exercise, walking for transport, all walking, other sports/exercise (excluding walking), and for all PA (sports/exercise/all walking); and, 2) recorded participation in ≥ 30 minutes of said activities where the shortest eligible bout was ≥ 10 minutes. The first strategy does not put a minimal requirement on bout length and values all time in moderate to vigorous intensity activity as indicative of increased energy expenditure. The second strategy is more rigorous and is based on empirical data that supports a minimal bout length to elicit cardiorespiratory and other health benefits \[[@B28]-[@B30]\]. Results ======= Accumulated daily minutes in walking for exercise, sports/exercise, and walking for transport for the total sample and for doer sub-samples are presented in Table [1](#T1){ref-type="table"}. Each PA variable was highly skewed (i.e., a large number of days with 0 minutes of the indicated activity); in all cases medians and quartile cut points for the total sample were zero. Of the three activities considered, the highest proportion of days indicated any walking for transport, followed by sports/exercise, and then walking for exercise. Based on doer sub-samples, sports/exercise (2100 doers representing 15% of diary days) provided the most accumulated minutes of PA over 1.4 ± 0.8 bouts/day (72 minutes/bout), followed by walking for exercise (1318 or 9%) over 1.2 ± 0.5 bouts/day (47 minutes/bout), and finally walking for transport (2879 or 20%) over 2.3 ± 1.4 bouts/day (12.5 minutes/bout). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Total sample (n = 14315 diary days) and doer sub-samples\* of accumulated daily minutes in walking for exercise, sports/exercise, and walking for transport ::: ---------------------------------------------------------------------------------------------------------------- PA variable Total sample Doer sub-sample\* ----------------------------------------------------- ------------------ ----------------------- --------------- Walking for exercise 5.4 ± 22.5 1318\ 58.6 ± 49.1\ (9%) 50 (30, 70) Sports/exercise (walking for exercise not included) 14.8 ± 50.3 2100\ 100.6 ± 92.8\ (15%) 70 (35, 130) Walking for transport 5.8 ± 16.6 2879\ 28.8 ± 26.5\ (20%) 20 (10, 40) ---------------------------------------------------------------------------------------------------------------- \* \'Doer\' is an accepted time use research term that represents a participant sub-sample of those reporting any of the indicated PA. \# of diary days varies within each PA variable category as indicated. \\ Median and 25^th^, 75^th^ percentiles are not presented for total sample since all values were zero. ::: Figures [1](#F1){ref-type="fig"}, [2](#F2){ref-type="fig"}, [3](#F3){ref-type="fig"} present the proportion of days, the number of bouts, and the accumulated minutes of walking for exercise and transport, respectively, by day of the week for those reporting any of these target activities. On any given day of the week, there was relatively greater participation in any walking for transport vs. walking for exercise (Figure [1](#F1){ref-type="fig"}; logistic regression p\< .05). Logistic regression also indicated that walking for transport participation was significantly more likely on either Wednesday or Thursday and that walking for exercise was more likely to occur on Sunday, Saturday or Wednesday (all p\< .05). With regards to Figure [2](#F2){ref-type="fig"}, walking for transport showed a significantly greater number of daily episodes compared to walking for exercise, although the number of episodes for both types of walking did differ by the day of the week, the most evident drop in walking for transport was between weekday and weekend day (all p\< .05). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Number of diary days of walking for exercise and transport by day of the week. ::: ![](1479-5868-2-5-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Number of bouts of walking for exercise and transport by day of the week in participant sub-sample (i.e., those reporting any of the target activity). ::: ![](1479-5868-2-5-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Accumulated mean minutes of walking for exercise and transport by day of the week in participant sub-sample (i.e., those reporting any of the target activity). ::: ![](1479-5868-2-5-3) ::: For those who report any walking, walking for exercise was approximately 30 minutes longer than walking for transport (Figure [3](#F3){ref-type="fig"}; p\< .05). Regardless, walking for transport provides almost 30 accumulated minutes of PA most days of the week for those who report any. The time spent walking for transport or walking for exercise did not vary significantly by the day of the week (p\> 0.05). Accumulated time spent walking as a mode of transport as related to each of the nine standard time use activity categories is presented in Table [2](#T2){ref-type="table"}. The greatest proportion of daily time spent walking for transport (i.e., proportionate to total time walking for transport) was related, in descending order, to purchasing (31.9%), employment (23.3%), social and community interaction (13.1%), and recreation and leisure (10.2%). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Total sample (n = 14315 diary days) and doer sub-samples\* of accumulated daily minutes walking for transport related to nine major activity categories ::: ---------------------------------------------------------------------------------------------- Activity Category Total sample Doer sub-sample\* ---------------------------------- ------------------ ----------------------- ---------------- Personal care 0.02 ± 0.59 18\ 13.9 ± 9.5\ (0.1%) 10 (5,20) Employment 1.36 ± 7.17 942\ 20.6 ± 19.6\ (6.6%) 15.0 (10,25) Education 0.52 ± 4.69 261\ 28.8 ± 19.9\ (1.8%) 25 (15,40) Domestic 0.05 ± 1.32 34\ 21.1 ± 17.4\ (0.2%) 17.5 (8.8, 30) Child care 0.49 ± 5.41 209\ 33.6 ± 30.0\ (1.5%) 25, (10,45) Purchasing 1.86 ± 8.61 1216\ 21.9 ± 20.8\ (8.5%) 20 (10,25) Voluntary work and care 0.14 ± 2.10 99\ 19.9 ± 15.8\ (0.7%) 15 (10,30) Social and community interaction 0.76 ± 5.58 472\ 23.1 ± 20.7\ (3.3%) 20 (10,30) Recreation and leisure 0.60 ± 5.51 323\ 26.4 ± 25.8\ (2.3%) 20 (10, 32.5) ---------------------------------------------------------------------------------------------- \* \'Doer\' is an accepted time use research term that represents a participant sub-sample of those reporting any of the indicated PA. \# of diary days varies within each PA variable category as indicated. \\ Median and 25^th^, 75^th^ percentiles are not presented for total sample since all values were zero. ::: The proportion of diary days on which public health guidelines are achieved are presented in Table [3](#T3){ref-type="table"}. Approximately 24% (95% CI = 23.3--24.7) of days met public health guidelines when all accumulated minutes were considered collectively for walking for exercise, walking for transport, and sports/exercise. Similar proportions of days met the guidelines solely through accumulated minutes walking for transport or walking for exercise. Implementing the more stringent criteria of counting only activity bouts of ≥ 10 minutes did not affect the proportion of days meeting the guidelines through walking for exercise (indicating that walking for exercise was undertaken consistently in minimal 10 minute bouts); the remaining categories were only reduced by less than 1%. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Proportion of diary days (and 95% CI) meeting public health guidelines by walking for exercise, walking for transport, all walking, sports/exercise (walking for exercise not included), and all PA ::: ----------------------------------------------------------------------------------------------------------------------------- PA variable Proportion of days\ meeting public health guidelines\ (≥ 30 minutes/day of at moderate+ activities) ----------------------------------------------------- --------------------------------------------------- ------------------- Walking for exercise (includes hiking/bushwalking) 7.3 (6.9--7.8) 7.3 (6.9--7.8) Walking for transport 7.3 (6.8--7.7) 6.6 (6.2--7.0) All walking (for exercise and for transport) 14.2 (13.6--14.7) 13.7 (13.1--14.3) Sports/exercise (walking for exercise not included) 11.9 (11.4--12.4) 11.6 (11.1--12.1) All PA (sports/exercise/all walking) 24.1 (23.3--24.7) 23.8 (23.1--24.5) ----------------------------------------------------------------------------------------------------------------------------- ::: Discussion ========== Although time use data have been used previously to study leisure activities ranging from sports/exercise to television watching\[[@B22]\], this exploration represents the first application of time use data and methods to PA and public health concerns. The detailed diary data collected by the ABS 1997 Time Use Survey represents a unique opportunity to study population walking patterns, relative to both exercise and transportation purposes. Time use data and methods have been extensively validated\[[@B22]\]. Although walking for exercise (typically undertaken in a singular lengthy bout) can be measured easily using questionnaire approaches to PA assessment, the incidental nature of walking for transportation (i.e., brief and episodic) makes it elusive to all but direct measures, as confirmed by the detailed records captured by time use methods. In addition, time use data do not suffer from social desirability bias associated with questionnaire approaches to PA measurement, since respondents are not a prioricharged with recording any specific activity. In PA epidemiology, such detailed daily records are considered direct measures of PA and are often used to validate other surveys and subjective indices of PA\[[@B31]\]. Another strength of these data is the very high response rate observed. Under Australian legislation respondents selected for official surveys can be required to participate. This nominal power, however, has rarely been enforced for a sample survey, and the ABS stringently imposes confidentiality and data security. Regardless, its existence is a likely explanation for the very high response rate to this survey. Walking prevalence is typically inferred from PA questionnaires that focus on the respondents who report engagement (to a specified level, including any) in that activity over a specific recall period (i.e., the past week, a typical week, the last month, etc.). In contrast, the detailed diaries collected as part of the time use methodology permit a unique examination beyond mere prevalence to the patterns of bouts and time spent walking for exercise and transport. The care taken in obtaining a nationally representative sample of diaries covering all days of the week and all seasons permit us to make confident conclusions about expected population behaviour on any given day. The fact that, for the total sample, all PA variable medians and quartile cut points were zero is a notable finding but one that is familiar to time use researchers. In fact, the conventional use and interpretation of time use statistics based on doer sub-samples is largely a result of earlier comment on the limitations of per capita average daily duration of activities \[[@B24]\]:\"if we learn that with a group of people a per capita average of eight minutes is spent daily on reading newspapers, we still do not know anything about the proportion of people who effectively read a newspaper on any given day, nor do we know how much time is devoted to newspaper reading by those people who do in fact indulge in this kind of activity. In short, aggregate average daily duration data as described above disclose nothing about typical duration of the individual at all -- neither about the typical frequency with which it is being performed during the day, nor about the proportion of its \'doers\' in the observed population on an average day.\" In terms of prevalence, however, we can conclude that on any given day, 24.1% of Australians achieve public health recommendations for PA (regardless of source). That is, they accumulate 30 minutes or more of at least moderate intensity PA (see Table [2](#T2){ref-type="table"}). Achieving public health guidelines also requires participation at a minimal intensity level (i.e., at least moderate intensity). Although the ABS diaries do not require the respondent to record intensity, PA records are typically scored using the Compendium of Physical Activities\[[@B12]\]. Accordingly, and as stated previously, both walking for exercise and for transport meet these minimal intensity requirements. As further support, walkers tend to naturally self-select a pace sufficient to meet these recommendations\[[@B32],[@B33]\]. Although the 10-minute minimal bout has empirical support for cardiorespiratory benefits \[[@B28]-[@B30]\], logic suggests, that in terms of energy balance at least, any PA that contributes to energy expenditure is important. Implementing the more stringent criteria for determining days achieving PA guidelines only reduced most proportions by less than 1% and had no effect at all on achieved PA guidelines by walking for exercise only. Expressed as a proportion of all related bouts (without limitation to only those days achieving PA guidelines), 35% of all walking for transport bouts were less than 10 minutes, compared to 5% for walking for exercise and 4% for sports/exercise (walking for exercise not included). Taken together, across a population, 35% of walking for transportation bouts might be missed with surveys that focus only on 10 minute bouts, but the impact on associated conclusions about achieving daily PA guidelines is minor. A remaining concern is that people are not engaging in activity frequently enough during the week (i.e., most if not all days of the week) to elicit important health benefits\[[@B15],[@B18],[@B34],[@B35]\]. Although public health guidelines are worded to encourage daily PA, five days out of the week is considered an acceptable minimal participation rate. Unfortunately, the ABS survey was limited to two days of data collection, so we are unable to make direct conclusions about the prevalence of meeting these behavioural criteria in the context of a week. As noted previously, time use researchers have documented that data quality falls off after 48 hours, a point after which a detailed diary becomes burdensome to respondents \[[@B27]\]. Although each individual only contributed 2 days, the study was designed to have days of the week well represented, throughout the year. This strategy, combined with the very large sample size, makes it very likely that the patterns observed are reflective of population behaviour. Hypothetically, during a week of days, the specific individuals who make up daily doer sub-samples will change repeatedly. As stated previously, Giles-Corti and colleagues\[[@B18]\] indicated that 72.1% of respondents to a 2-week recall self-reported performing any walking for transport and 68.5% had self-reported any walking for recreation (comparable to walking for exercise herein). The large discrepancy in doer sub-samples may be due in part to the longer time frame queried. Walking for exercise accounted for less than half of all walking compared to walking for transport (9% vs. 20%), yet either walking classification resulted in similar proportions meeting (≈7%) public health recommendations on any given day. This suggests that although the doer sub-sample approximates (≈28 mins/day) the public health recommendations, they do not exactly meet the cut point (i.e., 30 mins/day). This is confirmed in Table [3](#T3){ref-type="table"}. Giles-Corti et al\[[@B18]\] reported that only 13.6% of respondents achieved recommended PA levels by walking for transport in the previous 2 weeks compared to 31.7% of those who walked for recreation only. As stated above, the differences between this earlier study and the one herein are largely due to the different measurement approaches. The time use diary is a direct and unbiased record of two days that coded associated travel separately (and as a primary activity) from any other activity performed, therefore it is unlikely that either walking for transportation or walking for exercise were missed. In contrast survey methods ask for a general recall of any walking in the previous set time frame. This approach is considered to be indirect and is known to suffer from recall bias as well as social desirability bias\[[@B31]\]. Given these differences, the time use diary results are more likely to reflect the true state of affairs. Regardless, these ABS diary data indicate that, on any given day, walking for exercise appears to be less prevalent than walking for transport. Walking for exercise is typically undertaken in a single bout of approximately 47 minutes in duration compared to walking for transport which is performed in just over 2 bouts of 13 minutes each. Walking for exercise is a salient and purposeful activity, characteristic of those activities that can more easily be recalled in questionnaire approaches to PA assessment. Although not directly comparable, the median duration walking for exercise bout for participants was 20 minutes here, less than the median 30 minutes reported by American adults in the 1987, 1994, and 2000 Behavioral Risk Factor Surveillance System \[[@B36]\]. In contrast walking for transport is brief and can be incidental, characteristics that make it more elusive to capture by these traditional approaches but detectable using diaries. To emphasize, walking for transport that occurs with regularity as part of a lifestyle routine will be more easily encapsulated by recall compared to incidental, spontaneous, or haphazard bouts undertaken in the course of day-to-day living. Although walking is arguably a feature of other daily activities (e.g., household chores), we believe that we accounted for the preponderance of walking-related transport associated with the nine activity categories captured. Indeed, the level of detail derived from the time use diaries permitted identification of the specific activity categories most commonly associated with walking for transport: purchasing, employment, social and community interaction, and recreation and leisure. This demonstrates the advantages of using population data from non-health sector sources for a component of population health surveillance. Recent research has suggested that living within walking distance (defined as a 20-minute walk from home) of a department, discount or hardware store; or a park, biking or walking trail, was significantly related to higher pedometer-determined PA\[[@B32]\]. These findings have implications for improved design of PA questionnaires to capture these more elusive bouts of walking. For example, prompts should be regularly used to elicit maximal response about walking for transport related at least to purchasing, employment, social and community interaction, and recreation and leisure. As stated above however, it does not appear to be important to solicit bouts less than 10 minutes since their impact on achievement of PA guidelines is minimal. On a side note, it is interesting that approximately 10% of walking for transport was associated with recreation/leisure (walking to facilities, etc.); this type of walking for transport may be interpreted as part of the total recreation/leisure experience. Although participation in walking for exercise is highest on a Sunday, the number of bouts and time spent walking for exercise on Sunday is similar to other days of the week. In contrast, the percent participating, number of bouts, and duration of walking for transport trips was less on Sundays compared to other days of the week. The overall impact is that participation in all walking (in terms of total days) is lowest on Sundays. The negative effect of Sunday on overall walking behaviours has been shown before in American populations using pedometers that are sensitive to ambulatory activities \[[@B37]-[@B41]\]. Conclusion ========== In summary, we used a novel application of existing Australian time use data to examine the patterns of walking for transport and for exercise in relation to sports/exercise participation and health-related PA recommendations. These detailed records permitted evaluation of participation, bout length, and duration of walking behaviours on a daily basis in a nationally representative sample. Walking for transport is more common than walking for exercise on any given day. Further, although walking for transport is typically undertaken in multiple brief bouts, the accumulated time approximates public health guidelines for those who report any walking for transport. It is therefore likely an important source of healthy PA. In the future, time use data can be used to examine relationships between demographic variables, environment characteristics (for example, rural vs. urban residency) and behaviour. Although the ABS time use diaries do not collect simple health information such as height and weight at this time, it is possible that this improvement might yield more fruitful explorations of behaviour related to body mass index. It is also possible to ask sub-samples of time use study participants to wear motion sensors in order to capture another objective direct measure of physical activity behaviour. Obviously these time use data are robust and numerous questions can be investigated that are beyond the scope of this initial exploration of the utility of time use data for public health purposes, specifically with regards to PA behaviour patterns. Since similar time use data exist for numerous countries stretching back over several decades, this resource represents a promising avenue of exploration of PA time trends and international comparisons. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= CTL contributed to study design, analyses, and interpretation and led the writing of the manuscript. MB led the analysis and contributed to study design, interpretation, manuscript preparation. DM contributed to study design, analysis, interpretation and manuscript editing. AB contributed to study design, interpretation, and manuscript editing. All read and approved the final manuscript.	PubMed Central	2024-06-05T03:55:59.829144	2005-5-17	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181824/", "journal": "Int J Behav Nutr Phys Act. 2005 May 17; 2:5", "authors": [ { "first": "Catrine", "last": "Tudor-Locke" }, { "first": "Michael", "last": "Bittman" }, { "first": "Dafna", "last": "Merom" }, { "first": "Adrian", "last": "Bauman" } ] }
PMC1181825	Background ========== Globally, there are more than 1 billion overweight adults, at least 300 million of them obese. These alarming facts published by the World Health Organisation (WHO) \[[@B1]\] demonstrate that obesity has reached epidemic dimension in developed as well as in developing countries. Consequences on health range from several non-fatal but debilitating disorders that reduce quality of life to increased risk of premature death because of serious chronic diseases. Besides genetic factors and food consumption patterns exceeding the individual energy need, a sedentary lifestyle with lack of physical activity (PA) is one of the key causes \[[@B2]\]. The relationship between obesity, PA and chronic diseases is close and several epidemiological studies could show that regular PA can prevent from obesity and related chronic diseases, such as type-2 diabetes, cardiovascular disease, hypertension, stroke, cancers of different sites, osteoporosis, and contribute to maintain mental health \[[@B1],[@B3]\]. Thus, PA promotes health and well-being and has also enormous economic benefits considering the health care costs that could be attributed to obesity. However, the question of the adequate dose of exercise is still a matter of debate \[[@B4]-[@B6]\]. In order to provide a solid basis for obesity prevention strategies detailed knowledge of PA patterns in the target population is necessary. Therefore, we assessed short-term PA and sedentary behaviour of the Bavarian population by means of three unannounced 24-h recalls. Different activity domains contributing to total daily energy expenditure are described and their impact on obesity risk is quantified. Additionally, PA estimates in the Bavarian population are compared with current recommendations to prevent obesity and promote well-being and health. Methods ======= Study Design ------------ The Bavarian Nutrition Survey II (BVS II) is designed as a representative study of the Bavarian population to investigate dietary habits and PA. From September 2002 until June 2003, 1050 subjects aged 13--80 years were recruited by a three-stage random route sampling procedure from the German-speaking Bavarian population. This recruitment procedure included the selection of 42 communities as so-called sampling points (stratified by county and community characteristics), a random walk (every third household) with a given start address, and a random selection of one household member who meets the selection criteria. At baseline, subjects\' characteristics, lifestyle, socio-economic and health status were assessed by means of a computerized face-to-face interview. Within the following two weeks, participants were contacted by telephone on two workdays and one weekend day for recalling their dietary intake as well as PA on the day before. Within six weeks after recruitment, all adult study subjects (=18 years) were invited to their nearest health office for blood sampling and standardized anthropometric measurements. Participation rate in the whole study was 71 % (n = 1050). All adults that completed at least one 24-h dietary recall (n = 879) were invited to the health offices; from 65 % (n = 568) of those approached blood samples and anthropometric measurements could be obtained. For the present evaluation, 893 subjects who completed at least two 24-h activity-recalls were included. Within this group standardized anthropometric measurements were available from 552 subjects (61.8 %). All participants gave their written informed consent. The study was approved by the local ethical committee. Assessment of Physical Activity ------------------------------- According to a method described and validated by Matthews et al. \[[@B14]\], information on the short-term PA of each subject was collected by means of three unannounced computer-assisted telephone interviews. Trained interviewers asked the study participants to recall the exact type and time spent in activities of the following 5 categories during the last 24 hours: occupation, sports, other strenuous leisure time activities (LTPA~strenuous~), TV or PC use in leisure time and sleeping. In the categories sports and LTPA~strenuous~, the interviewers used a list of common activities laid on the screen in order to give examples to the participants and to fasten the interview process. Different types of walking (including walking for pleasure) were attributed to the category \'sports\' since this type of activity is very important in older age; the category LTPA~strenuous~included mainly leisure time PAs of moderate and vigorous intensity, such as different types of gardening, homemaking and household activities, or child caring. Although the wording of the question (\'strenuous\') may imply vigorous activities only, we actually assessed mainly activities of moderate intensity by means of this question (see results). Based on the results of their validation study, Matthews et al. \[[@B14]\] concluded that a series of three unannounced 24-h PA recalls provides an assessment of PA comparable to other short-term PA assessments that utilize activity monitors (Actillume monitoring) or the Baecke questionnaire. Deattenuated Pearson correlation coefficients between results from the 24-h recalls and the Baecke questionnaire ranged from 0.34--0.68 (p \< 0.01). A correlation coefficient of 0.64 (p \< 0.01) was reported for the association between 24-h recall results (total MET\h/d) and the Actillume measures (counts\min^-1^\d^-1^). They assessed four intensities of activity (light, moderate, vigorous, and very vigorous) in each of three activity domains (household, occupational, leisure-time) as well as sleeping time, and assigned 1.5 MET for light, 4.0 MET for moderate, and 6.0 MET for vigorous activities \[[@B14]\]. In our study, we more precisely assessed the time and type of activity spent in different PA categories and assigned individual MET values; however, except for TV/PC use, we did not actively assess the time spent with light activities during leisure time. As described in the compendium of physical activities by Ainsworth et al. \[[@B7],[@B8]\], multiples of the metabolic equivalent (METs) were used to estimate the relative intensity of each reported activity with one MET equal to the standard for resting energy expenditure (roughly 3.5 ml of oxygen consumed per kilogram of body weight/min) for the average adult. According to the assigned MET-values, all self-reported activities were classified as light (\< 3 METs), moderate (3--6 METs) or vigorous (\>6 METs) \[[@B4],[@B8]\]. The MET-values of occupational activities were determined by a combination of self-reported work-intensity (ranging from mainly sitting to laborious physical workload or actually not working) and respective job-title. When a description of activities was missing or the provided information unclear standardized mean MET-values were assigned. In particular, if job activities of students and retired persons were reported that could not be classified, a MET-value of 1.85 representing light work was assumed to be applicable. Type and intensity of the activity of homemakers was also difficult to evaluate; only for this group all reported strenuous activities belonging to the area of household activities were considered as being included in occupational household work and, therefore, not attributed to LTPA~strenuous~. To acknowledge homemakers\' activities as full occupation, we filled up the reported working time to at least 8 hours of work per weekday for all homemakers under 65 years. An intensity level of 2.5 METs representing \"multiple household tasks all at once, light effort\" \[[@B8]\] was assigned. Energy expenditure estimates (MET\h) independent from body weight were calculated by multiplying the reported duration of any activity (h) by respective intensity (MET) \[[@B7],[@B8]\]. By summing up all activities, participants\' daily MET\h were obtained for the different activity domains, e.g. sports-MET\h per day. In order to estimate a total daily PA score, it was necessary to introduce a new activity domain, called non-reported PA during leisure time (LTPA~non-reported~), according to a method described by Norman et al. \[[@B9],[@B10]\]. The difference between 24 hours per day and the total duration of self-reported activity/inactivity was considered as LTPA~non-reported~. These unknown activities were multiplied by an estimated MET-value of 1.75, which is between the suggested values of 1.5 MET \[[@B14]\] and 2.0 MET \[[@B9],[@B10]\]. The intensity factor corresponds to the mean of sitting (1.5 MET) and light home and self-care activities (2.0 MET) \[[@B7],[@B8]\]. Since our study participants mentioned also several light activities under the category LTPA~strenuous~-- which were multiplied with the most exact MET value given by Aintsworth et al. -- we tried not to overestimate the remaining non-reported time. The single recalls were weighted for weekday or weekend day to calculate a subject\'s total daily short-term PA and its components. We also estimated the participants\' short-term PA level (PAL~est.~) by dividing the individual total daily PA score (MET\h/d = kcal/(kg body weight\d) =\~ 1 kcal/(kg b.w.\min) \[[@B7],[@B8]\]) by the minimum score of 23.2 MET\h/d (assumption of 8 hours of sleep × 0.9 MET and 16 h being awake, but resting × 1.0 MET) \[[@B11]\]. Since 23.2 MET\h should reflect resting metabolic rate (RMR) expressed in units of MET\h, the resulting ratio gives the multiple of RMR \[[@B11]\], similar to the PAL value. However, it has to be emphasized that the calculated PAL~est.~values are of limited precision as compared to the PAL values mainly derived by means of the doubly labelled water method. Case definition --------------- To assess the prevalence of overweight and obesity, the subjects\' body mass index (BMI) was calculated as measured weight divided by the square of measured height (kg/m^2^). Self-reported figures were used for subjects who did not undergo anthropometric measurements. Following the WHO-guidelines \[[@B12]\] participants were classified into six categories as being underweight (\<18.5 kg/m^2^), normalweight (18.5-\<25 kg/m^2^), overweight (25-\<30 kg/m^2^), obese grade I (30-\<35 kg/m^2^), obese grade II (35-\<40 kg/m^2^) and obese grade III (≥ 40 kg/m^2^). All obese subjects (n = 144) with BMI ≥ 30 kg/m^2^were considered as cases and all other study participants served as controls in the logistic regression analyses. Statistical Analysis -------------------- The given descriptive results were weighted to correct for the deviation of the study group from the distribution of gender, age, and living area in the underlying Bavarian population. Since the PA data were not normally distributed, median and interquartile range are presented. Comparisons between gender and BMI groups were made by means of the Mann-Whitney U test. In order to examine the association between PA and obesity risk, logistic regression models were used. Risk calculations were conducted only for the subgroup with standardized measurement of weight and height. Additionally, subjects with an energy intake below 80% of the estimated basal metabolic rate (BMR, calculated by WHO-equations \[[@B13]\]) were excluded from risk estimations because of an increased likelihood of misreporting of PA. Thus, risk evaluation was conducted in a subgroup of 507 subjects. The activity estimates (MET\h/d) over each activity-domain as well as the total daily activity (MET\h/d and PAL~est~, respectively) were divided into four groups according to the distribution in the entire study population or by predefined cut points. Odds ratios (OR) and corresponding 95% confidence intervals (CI) are given for models adjusted for sex, age (\< 18 y, 18-\<30 y, 30-\<40 y, 40-\<50 y, 50-\<65 y, ≥ 65 y), energy intake (kcal/100/d), smoking (never, former, current) and socio-economic status (low, low-medium, medium, medium-high, high). Categorization of socio-economic status is based on the value of three characteristics on a point-scale including household net income, educational level of the one who is being interviewed and career position of the principal earner. Tests on trend were calculated using the quartile-based PA scores as a continuous variable as well as using the continuous variables (in MET\h/d). All statistical analyses were performed by means of the SPSS 11.0 software package (SPSS Inc., Chicago, USA). Results ======= Baseline characteristics and prevalence of obesity -------------------------------------------------- Baseline characteristics of the study participants are summarized in Table [1](#T1){ref-type="table"}. Significant gender differences existed for BMI groups, socioeconomic status, employment level, smoking habits and marital status; also anthropometric measures as well as basal metabolic rate (BMR) and energy intake differed by gender. The proportion of obese subjects in the whole sample (n = 893) was estimated to 17.1% in women and 16.1% in men. Excluding subjects with self-reported weight and height, the prevalence of obesity was even higher with 19.6% in women and 20.4% in men (overall 20.0%). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Baseline characteristics of the study participants^1^. ::: Total (n = 893)* Women (n = 528) Men (n = 365) p-value^2^ ------------------------------- --------------------- --------------------- ------------------- ---------------- ------------- ------ --------- Age (years) 0.713 \<18 48 7.4 20 6.4 28 8.6 18-\<30 99 13.5 65 14.0 34 12.9 30-\<40 196 21.0 125 20.2 71 21.8 40-\<50 182 19.0 119 19.7 63 18.2 50-\<65 228 23.8 129 23.5 99 24.2 ≥65 140 15.2 70 16.2 70 14.1 Body mass index (kg/m^2^) \<0.001 underweight (\<18.5) 35 4.2 25 5.9 10 2.4 normal (18.5-\<25) 402 44.9 265 49.3 137 40.0 overweight (25-\<30) 312 34.1 154 27.4 158 41.5 obese (≥30) 144 16.6 84 17.1 60 16.1 grade I (30-\<35) 99 11.2 54 10.5 45 12.0 grade II (35-\<40) 31 3.6 17 3.3 14 3.8 grade III (≥40) 14 1.9 13 3.5 1 0.2 Socioeconomic status 0.001 low 133 13.6 81 14.3 52 12.9 low-medium 230 25.5 129 25.7 101 25.4 medium 262 29.3 163 29.9 99 28.7 medium-high 178 21.2 118 24.0 60 18.2 high 90 10.3 37 6.2 53 14.8 Employment \<0.001 employed 429 48.1 241 41.2 188 55.6 homemaker 152 13.9 151 26.5 1 0.2 student/articled 78 12.3 35 10.3 43 14.6 unemployed/other 36 4.1 14 2.2 22 6.2 retired 198 21.5 87 19.7 111 23.4 Smoking status \<0.001 never 473 52.3 320 61.2 153 42.6 former 183 21.0 83 16.0 100 26.3 current 236 26.7 124 22.6 112 31.1 missing Data 1 0.1 1 0.2 0 0.0 Marital status \<0.001 single 176 21.6 82 17.1 94 26.6 married/cohabiting 578 67.0 337 65.8 241 68.4 divorced/widowed 138 11.3 108 17.1 30 5.0 missing data 1 0.0 1 0.0 0 0.0 mean ± SD Height (cm) 169.6 ± 9.1 164.0 ± 6.7 175.5 ± 7.3 \<0.001 Weight (kg) 74.3 ± 15.6 68.1 ± 13.9 81.0 ± 14.6 \<0.001 BMI (kg/m^2^) 25.8 ± 5.2 25.5 ± 5.8 26.2 ± 4.3 0.038 BMR (kcal) 1601 ± 253 1419 ± 132 1801 ± 197 \<0.001 Energy intake (kcal)^3^ 2001 ± 667 1704 ± 529 2326 ± 652 \<0.001 ^1^weighted for deviation from the underlying Bavarian population (sex, age, region) ^2^Chi-Square test or independent-samples t-test for gender differences ^3^3 × 24-hour dietary recall ::: Estimated Physical Activity --------------------------- Estimates of PA by activity domain (MET\h/d) and intensity are given in Table [2](#T2){ref-type="table"}, including also the corresponding duration of activities (h/d). Men as compared to women showed significantly higher values in total scores of sports activity, TV/PC use and total daily activity, while women reported a significantly longer sleeping time per day. This is also reflected in results by intensity sub-groups with men spending more time in PA with moderate or vigorous intensity. The most important intensity subgroup was occupational PA of light intensity showing the highest mean energy expenditure for both men and women. Non-reported time of PA in the 24-hour recalls was higher in women than in men. Total daily PA was estimated to 37.35 (5.58) MET\h/d (median, interquartile range) in women and 37.92 (8.80) in men, corresponding to PAL~est.~values of 1.61 (0.24) and 1.63 (0.38), respectively. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Estimated physical activity (h/d and MET\h/d) by sex, type, and intensity of activity^1^. ::: Type and intensity\ of activity/inactivity Total (n = 893) Women (n = 528) Men (n = 365) p-value^2^(MET\h/d)* ------------------------------------------------- --------------------- --------------------- ------------------- -------------------------- -------------- -------------- --------- Occupation total 2.86 (5.71) 5.36 (14.00) 2.86 (5.71) 5.14 (14.29) 2.95 (6.07) 6.83 (12.60) 0.799 light 82.6%^\#^ 68.1% 89.9% 82.4% 75.8% 55.2% moderate 13.2% 21.5% 8.8% 13.6% 17.7% 28.7% vigorous 4.0% 10.4% 1.7% 4.0% 6.5% 16.2% Sports total 0.12 (0.63) 0.48 (3.34) 0.08 (0.53) 0.38 (2.83) 0.14 (0.73) 0.59 (4.50) 0.028 moderate 57.5% 43.1% 66.7% 52.9% 51.1% 36.4% vigorous 40.0% 56.9% 33.3% 46.6% 48.9% 63.6% LTPA~strenuous~^3^ total 0.00 (0.54) 0.00 (1.71) 0.00 (0.57) 0.00 (2.10) 0.00 (0.36) 0.00 (1.55) 0.893 light 5.0% 3.2% 5.4% 4.4% 4.8% 2.2% moderate 92.5% 92.4% 91.95 90.4% 95.2% 94.0% vigorous 2.5% 4.5% 2.7% 5.2% 2.4% 3.8% TV/PC~leisure\ time~^4^ 1.64 (1.82) 1.64 (1.82) 1.38 (1.52) 1.38 (1.52) 2.00 (1.96) 2.00 (1.96) \<0.001 Sleeping^5^ 7.43 (1.39) 6.69 (1.25) 7.56 (1.37) 6.80 (1.23) 7.31 (1.36) 6.58 (1.22) 0.016 LTPA~non-reported~^6^ 10.93 (4.04) 19.12 (7.08) 11.26 (4.17) 19.71 (7.30) 10.61 (4.08) 18.56 (7.14) \<0.001 Total daily activity score MET\h 37.52 (7.18) 37.35 (5.58) 37.92 (8.80) \<0.001 PAL~est.~ 1.62 (0.31) 1.61 (0.24) 1.63 (0.38) \<0.001 ^1^Weighted for deviation from the underlying Bavarian population (sex, age, region) ^2^Mann-Whitney U-test ^3^LTPA~strenuous~= strenuous leisure time physical activity ^4^TV/PC~leisure\ time~(1.0 MET) ^5^Sleeping (0.9 MET) ^6^LTPA~non-reported~= non-reported leisure time physical activity (1.75 METs) \ Light (\<3 METs), moderate (3--6 METs), vigorous (\>6 METs) ^\#^Percentage of mean h/d and mean MET\h/d, respectively, of the corresponding activity domain ::: Table [3](#T3){ref-type="table"} shows the results for the estimated PA by type and intensity level in different BMI categories. Obese subjects reported less participation in occupational (women) and sports (men) activities but performed more LTPA~strenuous~than non-obese women and men. On the contrary, the time spent with TV/PC use during leisure time was highest in overweight and obese subjects. Sleeping time was shortest among obese women while underweight subjects slept most. Total daily activity scores were lowest in obese and underweight subjects, thus, the difference between obese and non-obese subjects did not reach statistical significance. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Estimated physical activity (MET\h/d) by weight class (BMI), type and, intensity of activity. ::: Underweight (\<18.5 kg/m^2^) Normalweight (18.5-\<25) Overweight (25-\<30) Obese (≥30) p-value ---------------------------- ---------------------------------- ------------------------------ -------------------------- ----------------- ------------- Median (Interquatile range) Occupation total 2.93 (9.76) 5.79 (14.29) 5.35 (14.29) 2.41 (11.68) 0.033 light 55.8%^\#^ 83.8% 81.6% 89.7% moderate 18.8% 12.2% 16.8% 10.3% vigorous 25.3% 3.9% 1.8% 0% Sports total 1.62 (3.24) 0.63 (3.12) 0.00 (2.30) 0.00 (1.73) 0.105 moderate 39.9% 47.0% 54.8% 80.4% vigorous 60.1% 53.0% 43.9% 19.6% LTPA~strenuous~^3^ total 0.00 (0.00) 0.00 (1.61) 0.00 (3.21) 0.00 (2.41) 0.026 light 0.0% 5.3% 2.2% 5.4% moderate 100.0% 87.6% 89.9% 94.6% vigorous 0 % 7.1% 7.3% 0% TV/PC~leisure\ time~^4^ 1.21 (1.11) 1.11 (1.33) 1.69 (2.04) 1.78 (1.61) 0.004 Sleeping^5^ 7.90 (2.07) 6.80 (1.24) 6.75 (1.03) 6.75 (1.25) 0.332 LTPA~non-reported~^6^ 20.77 (8.20) 19.26 (7.72) 19.49 (7.45) 19.89 (6.66) 0.325 Total daily activity score MET\h/d 36.56 (7.08) 37.78 (6.00) 37.20 (5.97) 36.70 (4.95) 0.083 PAL~est.~ 1.58 (0.31) 1.63 (0.26) 1.60 (0.26) 1.58 (0.21) 0.083 Men (n = 365)* n = 10 n = 137 n = 158 n = 60 Median (Interquatile range) Occupation total 3.47 (6.51) 7.48 (13.68) 5.09 (12.45) 3.01 (12.28) 0.098 light 100.0% 60.6% 46.0% 61.4% moderate 0% 18.2% 46.5% 10.7% vigorous 0% 21.3% 7.5% 28.1% Sports total 7.52 (15.67) 1.29 (7.86) 0.39 (3.21) 0.00 (2.05) 0.006 moderate 13.2% 33.0% 48.5% 28.6% vigorous 86.9% 67.0% 51.0% 71.4% LTPA~strenuous~^3^ total 0.00 (0.00) 0.00 (0.54) 0.00 (2.86) 0.29 (4.39) 0.007 light 0% 1.9% 3.1% 0.7% moderate 100.0% 87.5% 94.6% 98.7% vigorous 0% 11.5% 2.2% 0.7% TV/PC~leisure\ time~^4^ 1.46 (2.53) 1.77 (1.62) 2.07 (1.97) 2.46 (1.92) 0.001 Sleeping^5^ 8.21 (1.63) 6.62 (1.13) 6.56 (1.28) 6.48 (0.90) 0.040 LTPA~non-reported~^6^ 15.80 (6.43) 18.10 (7.45) 18.66 (6.95) 19.50 (7.84) 0.413 Total daily activity score MET\h/d 37.54 (11.56) 38.94 (8.81) 37.19 (8.42) 37.42 (8.92) 0.087 PAL~est.~ 1.62 (0.50) 1.68 (0.38) 1.60 (0.36) 1.61 (0.38) 0.087 ^1^Weighted for deviation from the underlying Bavarian population (sex, age, region) ^2^Mann-Whitney U-test ^3^LTPA~strenuous~= strenuous leisure time physical activity ^4^TV/PC~leisure\ time~(1.0 MET) ^5^Sleeping (0.9 MET) ^6^LTPA~non-reported~= non-reported leisure time physical activity (1.75 METs), \ Light (\<3 METs), moderate (3--6 METs), vigorous (\>6 METs) ^\#^Percentage of mean MET\h/d of the corresponding activity domain ::: Physical Activity and Risk of Obesity ------------------------------------- Risk estimations in the subgroup with measured weight and height and after exclusion of suspected miss-reporters revealed a significant inverse association between obesity and sports activity (Table [4](#T4){ref-type="table"}). After adjusting for sex, age, energy intake, socio-economic and smoking status the odds ratio (CI) for the subjects with more than 5 MET\h/d of sports activities was 0.37 (0.16--0.85; p = 0.037 for trend~cont.~) as compared to subjects with no sports activity. The use of TV/PC in leisure time was positively associated with obesity. As compared to subjects with less than 1 MET\h/d (1^st^quartile), the ORs (95% CI) in the 2^nd^, 3^rd^, and 4^th^quartiles, were 3.12 (1.42--6.87), 2.92 (1.29--6.58), and 2.51 (1.07--5.87), respectively (p = 0.059 for trend~cont.~). Obesity risk tends to decrease with increasing sleeping (p = 0.062 for trend~cont.~), except for the small group with \> 8 MET\h/d spent with sleeping. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Obesity risk by types of physical activity and total physical activity (n = 507^\#^)\. ::: Quartiles* ---------------------------------- --------------------------------- ------------------- ------------------- ------------------- ------------------- ------- ------- Occupation No. cases/controls 42 / 146 17 / 87 15 / 106 15 / 79 Limits of quartiles (MET\h/d) 0.00 0.25-\<8.00 8.00-\<14.50 ≥14.50 Median (MET\h/d) 0.00 4.29 11.23 17.68 odds ratio (95% Cl) 0.60 (0.28--1.30) 1 (ref.) 0.83 (0.38--1.83) 0.97 (0.44--2.18) Sports No. cases/controls 47 / 185 21 / 75 13 / 84 8 / 74 Limits of quartiles (MET\h/d) 0.00 0.10-\<2.00 2.00-\<5.00 ≥5.00 Median (MET\h/d) 0.00 0.94 2.86 8.57 odds ratio (95% Cl) 1 (ref.) 0.91 (0.49--1.69) 0.69 (0.34--1.39) 0.37 (0.16--0.85) 0.017 0.037 LTPA~strenuous~^3^ No. cases/controls 52 / 247 12 / 61 10 / 49 15 / 61 Limits of quartiles (MET\h/d) 0.00 0.10-\<2.00 2.00-\<4.50 ≥4.50 Median (MET\h/d) 0.00 1.07 3.21 6.93 odds ratio (95% Cl) 1 (ref.) 0.94 (0.45--1.94) 0.88 (0.40--1.93) 0.74 (0.37--1.48) 0.393 0.650 TV/PC~leisuretime~ No. cases/controls 10 / 129 29 / 116 26 / 93 24 / 80 Limits of quartiles (MET\h/d) \<1.00 1.00-\<2.00 2.00-\<3.00 ≥3.00 Median (MET\h/d) 0.5 1.43 2.34 3.65 odds ratio (95% Cl) 1 (ref.) 3.12 (1.42--6.87) 2.92 (1.29--6.59) 2.51 (1.07--5.89) 0.081 0.059 Sleeping No. cases/controls 27 / 97 36 / 166 21 / 135 5 / 20 Limits of quartiles (MET\h/d) \<6.00 6.00-\<7.00 7.00-\<8.00 ≥8.00 Median (MET\h/d) 5.46 6.57 7.35 8.23 odds ratio (95% Cl) 1 (ref.) 0.76 (0.42--1.37) 0.55 (0.28--1.07) 1.08 (0.33--3.51) 0.217 0.062 LTPA~non-reported~^4^ No. cases/controls 15 / 96 22 / 104 27 / 127 25 / 91 Limits of quartiles (MET\h/d) \<15.00 15.00-\<19.00 19.00-\<23.00 ≥23.00 Median (MET\h/d) 13.25 16.94 20.89 24.98 odds ratio (95% Cl) 1 (ref.) 1.25 (0.59--2.64) 0.75 (0.35--1.60) 0.94 (0.43--2.04) 0.587 0.275 Total daily PA score (PAL~est.~) No. cases/controls 31 / 92 39 / 189 10 / 87 9 / 50 Limits of quartiles (PAL~est.~) \<1.5 1.5-\<1.75 1.75-\<2.00 \>≥2.00 Median (PAL~est.~) 1.45 1.6 1.83 2.15 odds ratio (95% Cl) 1 (ref.) 0.59 (0.33--1.05) 0.35 (0.16--0.81) 0.56 (0.23--1.37) 0.038 0.728 ^\#^subgroup with measured weight and height and exclusion of suspected miss-reporters \adjusted for sex, age, energy intake (kcal/100/d), socioeconomic status (low, medium-low, medium, medium-high, high) and smoking status (never, former, current) ^1^tests on trend by using quartile-based PA scores as a continuous variable \[p~trend(cat.)~\] ^2^tests on trend by using uncategorized PA scores (MET\h) as a continuous variable \[p~trend(cat.)~\] ^3^LTPA~strenuous~= strenuous leisure time physical activity ^4^LTPA~non-reported~= non-reported leisure time physical activity ::: Obesity was inversely associated with total daily PA (PAL~est~values). The risk estimates declined over increasing PA quartiles (except for the 4th quartile) reaching statistical significance for the 3rd quartile with PAL values between 1.75 and 2.0. Combining all subjects with a PAL value of 1.75 or higher in one category (Q3 + Q4) the OR (95% CI) was 0.43 (0.21--0.85) indicating a strong inverse association with obesity. Meeting of Physical Activity Recommendations -------------------------------------------- When comparing the calculated PAL~est.~values in our population with the WHO recommendation of (measured) PAL =1.75, only 26.8% of women and 36.4% of men met this recommendation. The rates declined with increasing BMI and age (Table [5](#T5){ref-type="table"}), noting some exceptions (underweight subjects, age-groups \< 18 and 40-\<50). The public health recommendation of at least 30 minutes of moderate PA per day was met by 53.5% of women and 58.6% of men, including moderate to vigorous activities (≥ 3 METs) out of all relevant PA categories (occupation, sports, LTPA~strenuous~). Only the proportion of subjects with at least moderate (≥ 3 METs) sports activity for 30 min/d or longer was identified to decline with increasing BMI category; no such association can be seen when considering all leisure time PA or total PA (including also occupational activities). This indicates that a public health recommendation for obesity prevention in terms of an overall PA of at least 30 min/d of higher than light intensity may not work in this population. Such recommendations should be focused on sport activities only, a category that includes also walking. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Participants^1^meeting physical activity recommendations. ::: ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- a) by BMI (kg/m^2^) Total \ Underweight\ Normalweight \ Overweight \ Obese \ (n = 893) (BMI \<18.5)\ (BMI 18.5-\<25)\ (BMI 25-\<30)\ (BMI ≥ 30)\ (n = 35) (n = 402) (n = 312) (n = 144) ------------------------------------------------------------------------------------------------- ------------------------ ------------------- ----------------------- --------------------- ------------------------ --------------- --------------------- ----------------------- ------- ------- ------- ------- ------- ------- b) by age (years) Total\ \<18\ 18-\<30\ 30-\<40\ 40-\<50\ 50-\<65\ ≥65\ (n = 893) (n = 48) (n = 99) (n = 196) (n = 182) (n = 228) (n = 140) n % n % n % n % n % n % n % WHO-recommendation: PAL ≥ 1.75^§^\ 275 31.4 15 34.8 42 41.5 61 31.3 75 42.2 67 30.3 15 9.0 Public health recommendation (ACSM/CDC): ≥ 30 min/d of moderate-intense (≥ 3 METs) activity in all activity domains 498 55.9 33 67.7 54 52.5 96 49.2 105 57.8 134 59.1 76 55.6 in leisure time (sports, LTPA~strenuous~^2^) 452 50.2 33 67.7 52 50.0 78 37.7 93 49.7 122 53.8 74 54.1 in sports only 266 30.1 32 63.6 35 35.6 44 22.0 54 28.7 63 28.8 38 23.9 ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ^1^weighted for deviation from the underlying Bavarian population; ^2^LTPA~strenuous~= strenuous leisure time physical activity ^§^comparing the PAL~est.~values from the present study with limited accuracy with a recommendation based on precise PAL values given by the WHO ::: Discussion ========== The results of our investigation revealed that higher PA in the category sports and less use of TV/PC during leisure time were strongly and significantly associated with a decreased risk of obesity. Figure [1](#F1){ref-type="fig"} shows the mean BMI of subjects with respect to categories of sports activity and TV/PC use in leisure time. The mean BMI in the groups with higher sports activity and less time spent for TV/PC is distinctly lower than in subjects who were not active in sports and spent a long time watching TV or using a PC during leisure time. In general, sports are mostly of moderate or vigorous intensity and are often executed in one bout without long interruptions, especially endurance activities like walking, running or cycling. These sports activities demanding high energy costs were most popular among active subjects in the present study. Even people of older age (≥ 65 years) were still active in endurance sports by being engaged in walking although PA was declining with rising age. In comparison, obese subjects are more likely to be engaged in activities of moderate intensity, but hardly perform activities of high intensity, such as many sports \[[@B28]\]. This contrasts to TV/PC use which is associated with a very low energy expenditure. With increasing sedentary behaviour physical activities decreases \[[@B29]\]; moreover, especially television watching is associated with snacking, leading to high caloric intakes \[[@B30]\]. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Mean BMI (kg/m^2^) by sports activity and use of TV/PC~leisure\ time~(MET\h/d; n = 893) ::: ![](1479-5868-2-6-1) ::: Similar associations as reported here were found in other studies. An European study \[[@B31]\] investigating the PA pattern in samples from 15 EU member states found significant associations between BMI and leisure time PA \[OR of 0.52 (0.43--0.64)\] for subjects in the most active quintile compared to lowest and time spent sitting down \[OR of 1.61 (1.33--1.95)\] for subjects in the most inactive quintile compared to lowest, respectively. Cameron et al. \[[@B25]\] investigated the prevalence of obesity in Australian adults and examined its relationship with life-style factors. Strong associations between obesity and PA (OR of highest quintile: 0.70 men, 0.47 women) or TV watching (OR of highest quintile: 1.86 men, 1.82 women) were found. Similar associations between sedentary life-styles, mainly represented by TV watching, and PA have been shown by several previous studies \[[@B29],[@B32]-[@B37]\]. In the present study, we could not find distinct associations between PA in activity domains other than sports and TV/PC use and the risk of obesity. In contrast to the results reported by King et al. \[[@B38]\], occupational PA was unrelated to obesity risk. For unemployed subjects the lowest though not significant point estimate was found; this finding is possibly due to the fact that students and those who retired or were unemployed had more time left for sports or other recreational activities. In several \[[@B29],[@B40]\] but not all studies \[[@B39]\] an inverse association between occupational activity and leisure time PA was observed. The questioning for strenuous activities in leisure time mainly assessed moderate physical activities and contributed on average to only about 3 to 4 % of total daily energy expenditure. Risk estimates for obesity decreased with increasing activity in LTPA~strenuous~but did not reach statistical significance. This result may be affected by recall bias since obese subjects may have reported more activities in this category (Tab. 3) because of rating their activities more demanding. Two studies reported an inverse association between sleep duration and obesity \[[@B29],[@B41]\]. Except for the group with \>8 MET\h/d spent with sleeping, in our study risk estimates of obesity decreased with increasing time spent with sleeping; however, results were not statistically significant. In the present study, also non-reported activities were not associated with obesity risk. Our questionnaire did not assess light-intensity activities of common life (e.g. eating, car driving, self-care, etc.). Consequently, the high proportion of time attributed to this PA domain -- about half of estimated total daily energy expenditure -- was almost expected. On the other hand this result supports the view that only a small part of daily energy expenditure is spent in demanding activities which should be remembered best \[[@B42]\]. The estimated level of total physical activity in terms of MET\h/d in the present study population was very similar to that reported in the NHAPS Study \[[@B43]\]. This study is one of the few assessing 24-h PA with computer-assisted telephone-interviews; they found in 7.515 subjects (aged 18 years and over) mean values of 39.9 kcal/kg for men and 37.8 kcal/kg for women (MET\h corresponds to kcal/kg). For comparison, mean values in our study were 40.56 MET\h/d and 38.47 MET\h/d among men and women, respectively (for medians see table [2](#T2){ref-type="table"}). In a cohort of Swedish men aged 45--79 years, Norman et al. \[[@B9]\] reported a mean of 41.5 (SD: 4.9) MET\h/d for total daily activity assessed by questionnaire. In agreement with previous studies \[[@B10],[@B31],[@B32]\] total PA of the Bavarian subjects was found to be inversely associated with obesity. Subjects with a PAL value =1.75 (Q3 + Q4) had a 57 % reduced risk as compared to subjects with a PAL value \<1.5. These findings fit with the WHO-recommendation that a PAL of 1.75 or more is necessary to avoid excessive weight gain, a recommendation which is based on the review of 40 international studies \[[@B2]\]. Among normal-weight subjects, 35.1% met the recommendation, which is still low but clearly higher than the 22.8 % in the obese subjects (table [5](#T5){ref-type="table"}). Overall, this WHO-goal has only been reached by a total of 31.4 % of the study participants. The public health-recommendation from the Centers for Disease Control and Prevention (CDC) and the American College of Sports Medicine (ASCM) of at least 30 minutes of moderate PA per day \[[@B4]\] was met by a total of 55.9 %. An identical rate was even achieved by obese subjects, which might be astonishing at first sight, but if the recommendation was considered only in terms of sports activities, the percentage of sufficiently active obese subjects dropped to only 20.7 %. Taking into account that the recommendation of 30 minutes of moderate PA per day has minimum-character in the context of weight-management yet remembering the stricter guidelines of 60 minutes stated by the Institute of Medicine (IOM) \[[@B5]\], the data would turn out even worse. Nevertheless, considering diverging methods of assessment and PA recommendations these results are quite comparable with other studies. Brown and Baumann \[[@B18]\] found that the subjects\' percentage of meeting the current CDC/ACSM-recommendation in 2 Australian surveys ranged between 51.6 % and 60.2 %. Weyer et al. \[[@B44]\] observed that 61.5 % of 109 obese Germans did not meet any recommendation. This is less than the 87% of 7124 adults, who were not adequately active in the German General Health Survey in 1998 \[[@B45]\]. The obesity rate in this Bavarian sample is higher than in a recent survey published by the Federal Statistical Office of Germany \[[@B23]\] in 2004, but comparable to other German studies conducted since 1998. Bramlage et al. \[[@B24]\] reported on the prevalence of obesity comparing rates from the German \"Hypertension and Diabetes Risk Screening and Awareness\" (HYDRA)-study in 2001 (19.5 % in men, 20.3 % in women) with the German General Health Survey (GHS) 1998 data (18.8 % in men, 21.7 % in women). In comparison to the results of a former representative study in the Bavarian population in 1995 (BVS I), the prevalence of obesity increased in the last years as found also for other western countries \[[@B25]-[@B27]\]. The information about the participants\' short-term PA was collected by means of three 24-hour telephone recalls, a method validated by Matthews et al. \[[@B14]\] (see methods section). Other methods like behavioural observation, use of motion sensors, physiological markers (e.g. heart rate) and calorimetry are less subject to bias in the assessment of mainly long-term PA and energy expenditure. Especially the double-labeled water method is regarded as \'gold standard\'\[[@B15]\]. However, self-reported data obtained by means of diaries or recalls are most practical in large-scale population-based studies because of relatively low costs and low efforts for the participants \[[@B16]\]. In the present study, kind and duration of PA were assessed, but not the corresponding intensities (except for occupational PA). Instead, MET values were assigned to each specific activity. Consequently, some degree of error may have been introduced because of unclear description, misunderstanding or misidentification. In occupational PA, consideration of both self-reported job title and self-rated work-intensity at least reduced the great variability of subjects\' individual performances within the same job title \[[@B17]\]. However, using mean MET values to express the intensity of a PA assumes that there are no individual differences in performing the same types of activities, an assumption which in practice does not hold true \[[@B7],[@B8]\]. We further expressed PA in terms of MET\h/d and MET\*h/24 h but avoided to express PA in terms of \'kcal\' because the latter would have been strongly affected by body weight \[[@B7],[@B8]\] thus resulting in misclassification of individuals \[[@B18]\]. Potential bias must also be considered due to typical problems of self-report. First, the BMI variable might be affected by overestimation of height and underestimation of weight \[[@B19],[@B20]\] or in rare cases also by high muscle mass \[[@B21]\]. Using anthropometric measurements, valid BMI data could be obtained from a substantial part of the study subjects. Second, self-reported PA may be overestimated in order to create a more ideal picture of oneself \[[@B22]\]. And third, the quality of the survey is highly dependent on the respondents\' memory, a source of bias that should be minimized due to the short recalling period of 24 hours \[[@B14]\]; this should be one of the major strengths of the current study, besides its representativeness and its relatively large sample size. Conclusion ========== The overwhelming part of the Bavarian population did not reach current PA recommendations, and subjects meeting the recommendations showed a significantly lower risk of obesity. Our results strengthen the view of promoting sports activity in expense to TV/PC use in leisure time in order to counterbalance the rising prevalence of obesity in the Bavarian population. Other PA domains like occupation, LTPA~strenuous~, sleeping and LTPA~non-reported~showed weaker or no associations with obesity risk. However, due to the cross-sectional study design, no conclusion on causality can be drawn. Especially for the PA category sports activity, it remains unclear whether people are obese due to the low PA or the low PA is a consequence of their high body fat content. With respect to the weight development over time, probably both views are correct. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= NS carried out the coding of activities and the statistical analysis, drafted the manuscript. HS participated in collection and processing of data, participated in the statistical analyses. SH participated in collection and processing of data. GK, KG, GW participated in fund raising and the design of the study, JL senior author responsible for the design of the study, participated in collection and analyses of data, drafted the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ The study was supported by funds of the Bavarian Ministry of Environment, Health and Consumer Protection and the Kurt-Eberhard-Bode-Stiftung.	PubMed Central	2024-06-05T03:55:59.832458	2005-6-8	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181825/", "journal": "Int J Behav Nutr Phys Act. 2005 Jun 8; 2:6", "authors": [ { "first": "Nina", "last": "Schaller" }, { "first": "Henrike", "last": "Seiler" }, { "first": "Stephanie", "last": "Himmerich" }, { "first": "Georg", "last": "Karg" }, { "first": "Kurt", "last": "Gedrich" }, { "first": "Günther", "last": "Wolfram" }, { "first": "Jakob", "last": "Linseisen" } ] }
PMC1181826	Introduction ============ Cisplatin and its analog carboplatin have significant antitumor activity against a wide variety of solid tumors\[[@B1]\]. Furthermore, cisplatin also showed synergistic effect with other chemotherapeutic agents and therefore has been incorporated in many treatment regimens for solid tumors \[[@B1]\]. Like other chemotherapeutic agents, resistance to this drug is inevitable and often occurs after several cycles of treatment. Several laboratories have studied the mechanism(s) of cisplatin resistance in the past decades and many possible mechanism(s) have been identified \[[@B2]-[@B9]\]. These resistance mechanism(s) appears to fall into four major categories. The first category involves DNA damage/repair proteins. The second category involves drug retention (increased influx or decreased uptake). The third category involves increased drug inactivation or prevention of drug to reach the DNA target. The fourth category involves growth signaling via different pathways or increase in antiapoptotic protein(s). Nevertheless, it is generally accepted that cisplatin resistance most likely has multiple mechanisms and the mechanism of resistance may differ depending on the cell types. Therefore, overcoming cisplatin resistance may be difficult. We have developed two pairs of cisplatin resistant small cell lung cancer cell lines and three cell lines derived from patients who failed cisplatin, one from small cell lung cancer (SCLC) and two from non small cell lung cancer (NSCLC). Using microarray analysis in our cisplatin resistant cell lines as well as reviewing the available microarray data published in other laboratories, we have found that the there are many gene(s) which are overexpressed in these cisplatin resistant cell lines including those involved in DNA repair, signal transduction, invasion and metastasis, and antiapoptosis {for review see ref. \[[@B2]\]}. While different cisplatin resistant cell lines often overexpressed different genes involved in DNA repairs and/or signal transduction, antiapoptosis, a majority of cisplatin resistant cells overexpress elongation factor alpha and genes which are involved in ribosomal biogenesis. These findings have led us to hypothesize that after DNA damage by cisplatin, the surviving cells have to develop the ability to generate repair proteins and/or survival proteins to prepare for the next insult. Both ribosomal proteins and elongation factor are essential for translational process in protein synthesis. Thus, the common theme to survive cisplatin in these resistant cell lines is to increase these mRNAs. Consequently, if ribosomal proteins and/or elongation factor can be inhibited, one should be able to restore cisplatin sensitivity. It is well known that mTOR (mammalian target of rapamycin) also known as FRAP, RAFT, or RAPT is important in regulating translation of a set of mRNA which encode ribosomal proteins and elongation factor \[[@B10]-[@B12]\]. All of these mRNAs possess a sequence of pyrimidines at their extreme 5\'end (TOP mRNAs). Therefore, by inhibiting mTOR, one should be able to restore cisplatin sensitivity. In this report, we have investigated the possible role of a known mTOR inhibitor, rapamycin and its ester analog CCI-779 \[[@B13],[@B14]\] in restoring cisplatin sensitivity. It has been shown that the mTOR inhibitor, rapamycin, can also modulate other forms of drug resistance such as P-gp1 or MDR1 mediated drug resistance\[[@B15],[@B16]\]. MDR1 is a well characterized form of drug resistance which is primarily due to overexpression of an P-gp1 efflux pump \[[@B17],[@B18]\]. This efflux pump belong to ABC (ATP-binding-cassette) transporter superfamily, and is capable of effluxing many different chemotherapeutic agents, hence the term multidrug resistance. The resistance is due to decrease drug accumulation. Rapamycin has been shown to be able to reverse this form of drug resistance by blocking the efflux pump\[[@B15]\]. Another similar form of multidrug resistance which is due to decrease drug accumulation is the MRP1 mediated drug resistance. MRP1 also belongs to the ABC transporter superfamily \[[@B19]-[@B21]\], however, this efflux pump most likely transports the glutathione conjugated drug. In this report, we also investigated the role of mTOR inhibitor in these two forms of multidrug resistance. Methods ======= Cell lines ---------- SCLC1was established from the bone marrow of SCLC patient. SR-2 is the cisplatin resistant variant which was generated by intermittent exposure to cisplatin. The characteristics of these cell lines have been published previously \[[@B22]\]. All cell lines were maintained on RPMI media supplemented with 10% FBS. SCLC1Rwas generated by exposure of SCLC1 to rhodamine-123 \[[@B23]\]. This cell line overexpresses P-gp and exhibits 16 fold resistance to doxorubicin, 70 fold resistance to vinblastine and 5.3 fold resistance to VP-16, but not to cisplatin. SCLC1Aoverexpresses MRP1 and was selected by exposure of SCLC1 to doxorubicin \[[@B24]\]. This cell line exhibits 33 fold resistance to doxorubicin, 10 fold resistance to VP-16 and 42 fold resistance to vinblastine. SCLCBwas established from the supraclavicular lymph node of a SCLC patient. The characteristics of this cell line were previously published \[[@B25]\]. SCLCBC was generated by exposure SCLC to cisplatin similar to SR-2. SCLCBC exhibits 10 fold resistances to cisplatin, and 10 fold resistances to carboplatin but not to oxaliplatin. SCLCLwas established from the pleural effusion of a SCLC patient who relapsed on VP-16 and cisplatin. Unlike SCLC1 and SCLCB, this cell line grows as a floating aggregate. SCLCL does not overexpress P-gp 1 or MRP1 and is positive for chromogranin. NSCLCGwas established from a chest wall mass of a patient with adenocarcinoma of the lung who failed cisplatin and taxotere. This cell line is positive for keratin and does not overexpress P-gp1 and MRP1. NSCLCWwas established from the pleural effusion of a patient with poorly differentiated adenocarcinoma of the lung who failed gemcitabine + cisplatin, taxol + carboplatin and irinotecan + oxaliplatin. This cell line does not overexpress P-gp1 or MRP1. All cell lines were maintained on RPMI supplemented with 10% FBS. Compounds --------- CCI-779 and rapamycin were kindly provided by Wyerh-Ayrest. Cisplatin, carboplatin, oxaliplatin, VP-16, doxorubicin and vinblastine were obtained from the hospital pharmacy. Growth Inhibitory effect ------------------------ 1 × 106 cells were seeded onto 24-well plates and allowed 8 hr. for attachment. Various concentrations of rapamycin or its analogs CCI-779 or other chemotherapeutic drugs listed above were added to each well. Each concentration was performed in duplicate. After 72 hr. exposure, viable cells were counted in the presence of 0.2% trypan blue. The growth inhibitory effect (ID50) was determined by plotting the number of viable cells as a percentage of control against the compound concentration\[[@B26]\]. The growth inhibitory effect was performed after cells were kept in drug free media for 14 days. Assay for Apoptosis by Annexin V -------------------------------- Cells were treated with cisplatin with and without CCI-779 for 24 hrs. and then assay for apoptosis using Annexin V-FITC Apoptosis Detection kit (BD Bioscience). The kit includes annexin V conjugated to FITC and propidium iodide. For each sample, 10^5^or 10^6^cells were collected, washed in PBS and gently suspended in 100 μL of binding buffer (10 mM HEPES pH 7.4, 150 mM NaCl, 5 mM KCl, 1 mM MgCl~2~, 1.8 mM CaCl~2~) containing 0.25 μg annexin V-FTIC. Incubation lasted 15 min in the dark at room temperature. Finally, cells were washed and suspended in the binding buffer with 5 μg propidium iodide and analyzed with Coulter XL flow cytometer. At least 10,000 cells were counted per analysis. Westernblot Analysis of mTOR ---------------------------- Cells were lysed with RIPA buffer (10 mMTris pH7.4 100 mM NaCl, 1 mM EDTA, 20 mM Na~4~P~2~O~7~, 2 mM Na~3~VO~4~10%, 0.5% deoxycholate, 1 mM PMSF) and protease inhibitor cocktail from Sigma and passed several times through a 23G needle, and centrifuged. The total protein was separated on 8% SDS-PAGE, transferred onto membrane and immunoblot with rabbit polyclonal antibody to mTOR (purchased from Upstate) and detected by chemiluminescence. Westernblot Analysis of 4E-BP ----------------------------- The method described by Gingra et.al. was used \[[@B27]\]. Cells were seeded at 1 × 10^5^cells /ml onto 100 mm dishes, allowed overnight for attachment, then treated with various doses of cisplatin, with and without CCI-779. At 24 hrs, cells were harvested by scraping with RIPA buffer and protease inhibitor cocktail (purchased from Sigma). Lysis was completed by passing the cells through a 25 G needle. To bind elF-4E, 25 ul of 7methyl-GTP Sepharose was added to the lysates and incubated overnight at 4°C. The complexes were collected by centrifuge, washed with lysis buffer, then dissociated from sepharose by adding 50 ul of SDS-PAGE loading buffer, heated to 95°C then separated by SDS-PAGE, probed with 4E-BP antibody (purchased from Cell Signaling) and detected by chemiluminescence. Westernblot Analysis for p70-S6 Kinase -------------------------------------- This is similar to analyses of 4E-BP. Cells were seeded at 1 × 10^5^onto 100 mm dishes, exposed to cisplatin alone or in combination with CCI-779 overnight, lysed with RIPA buffer and then separated by SDS-PAGE. The membrane was probed with phospho-p70-S6 kinase antibody purchased from Cell Signaling (phospho-thre.389) and detected by chemiluminescence. This is a mouse monoclonal antibody which recognizes phospho-p70 S6 kinase. Westernblot Analysis for elongation factor α -------------------------------------------- Cisplatin resistant cells (SR-2 and SCLCBC) were treated with cisplatin or CCI-779 or combination of CCI-779 and cisplatin for 24 hours. Protein extracts were obtained and separated by SDS-PAGE and probed with anti-EF-1α monoclonal antibody (purchased from Upstate) and detected with chemiluminescence. This antibody detects a 53 kDa protein. Results ======= Growth Inhibitory Effect of rapamycin and CCI-779 ------------------------------------------------- Table [1](#T1){ref-type="table"} shows the result of growth inhibitory effect of rapamycin and CCI-779 in a panel of parental and resistant SCLC lines. Both CCI-779 and rapamycin are equally active in parental cell lines as well as their cisplatin resistant variants. Furthermore, CCI-779 is also active in cell lines which were established from patients who failed cisplatin containing regimens with the ID50 ranging from 0.1--0.5 ug/ml. In contrast, CCI-779 and rapamycin have no activity in P-gp1 or MRP1 overexpressing cell lines with the ID50 of 6.9 and 7.5 ug/ml, respectively. This is not surprising, since the structure of rapamycin suggest that it may be a P-gp substrate \[[@B28]\]. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Growth inhibitory effect of Rapamycin and CCI-779 (ID~50~) ::: SCLC1 SR-2 SCLCB SCLCBC SCLCL SCLCR SCLCA NSCLCW NSCLCG ----------------- ------------- ------------- --------------- ------------- ------------- ----------- ----------- ----------- ----------- Rapamycin ug/ml 0.66 ± 0.1 0.1 ± .0.05 0.05 ± 0.02 0.09 ± 0.03 0.2 ± 0.006 \>1 \>1 ND\* ND\* CCI-779 ug/ml 0.07 ± 0.02 0.06 ± 0.01 0.04 ± ..0.01 0.05 ± 0.02 0.3 ± 0.1 6.9 ± 2.4 7.5 ± 1.5 0.3 ± 0.1 0.5 ± 0.2 \ND = not done. The ID~50~represent the average of two experiments, each experiment done in duplicate. ::: Growth Inhibitory effect of CCI-779 in combination with cisplatin ----------------------------------------------------------------- Since CCI-779 is equally or slightly better than rapamycin in terms of growth inhibitory effects and is currently in clinical trial, we have further investigated whether the addition of CCI-779 can indeed increase cisplatin sensitivity, We have studied growth inhibitory effect of cisplatin in the presence of 10 ng/ml of CCI-779. At this dosage, the growth inhibitory effect is less than 10%. The results are shown in table [2](#T2){ref-type="table"}. CCI-779 is able to reduce the ID50 of cisplatin to 3--6 times in all cisplatin resistant cell lines. However, CCI-779 does not alter the ID50 in the parental line. Thus, it appears that CCI-779 can restore cisplatin sensitivity in otherwise cisplatin resistant cells. We further investigated whether CCI-779 can completely reverse cisplatin resistance. SR.2 and SCLCBC were treated with the cisplatin ID50 of sensitive cells (0.15 ug/ml in SCLC1 and 0.25 ug/ml in SCLCBC) and various concentrations of CCI-779. Our results indicate that CCI-779 at 0.03 ug/ml can completely reverse cisplatin resistance in both SR-2 and SCLCBC. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Growth inhibitory effect of Cisplatin (ID~50~) with and without CCI-779 ::: Cisplatin (ug/ml) Cisplatin (ug/ml)+ CCI-779 -------- ------------------- ---------------------------- SCLC1 0.15 ± 0.05 0.18 ± 0.02 SR-2 2.52 ± 0.51 0.45 ± 0.03 SCLCB 0.25 ± 0.12 0.14 ± 0.02 SCLCBC 2.57 ± 0.23 0.55 ± 0.04 SCLCL 0.95 ± 0.08 0.3 0 ± 0. 05 NSCLCW 1.2 ± 0.07 0.28 ± 0.04 NSCLCG 1.5 ± 0.5 0.4 ± 0.07 ::: We have further investigated whether combination of CCI-779 with cisplatin can also induce apoptosis in cisplatin resistant cells. The results are shown in fig [1](#F1){ref-type="fig"}. Cisplatin at 6 ug/ml cannot induce apoptosis in these resistant cells, while at a higher dose (10 ug/ml), the apoptotic effect is small but discernable. However, when CCI-779 (150 ng) was added to cisplatin, the effect is seen at 6 ug/ml of cisplatin and increases at 10 ug/ml. Thus, our data clearly shows that the addition of CCI-779 to cisplatin in these resistant cells not only augments the growth inhibitory effect, but also the apoptotic effect of cisplatin. Similar results were also obtained with SCLCBC cell lines. It is noteworthy that, CCI-779 alone at 300 ng/ml did not induce apoptosis (data not shown) in both cisplatin sensitive cells (SCLC1 and SCLCB). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Apoptosis assay using annexin-V by FACS analysis in cisplatin resistant cell line (SR-2). Cells were treated with cisplatin alone or in combination with 0.15 ug of CCI-779. Left panels are dual staining for propidium iodide uptake and annexin-V^FITC^, Right panels are corresponding distribution of annexin V^FITC^staining in populations of cells. 1A: control. 1B: SR-2 treated with 6 ug of cisplatin (No annexin V^FITC^was detected) 1C: SR-2 treated with 6 ug of cisplatin+CCI-779 (small peak of annexin V^FITC^was detected) 1D: SR-2 treated with 10 ug of cisplatin (a peak of annexin V^FITC^was detected) 1E: SR-2 treated with 10 ug of cisplatin+CCI-.779 (a large peak of annexin V^FITC^was detected) Note: CCI-779 alone at 0.15 and 0.3 ug/ml did not induce apoptosis (data not shown). ::: ![](1476-4598-4-25-1) ::: The effect of CCI-779 in reversing doxorubicin resistance in P-gp1 and MRP1 overexpressing cells ------------------------------------------------------------------------------------------------ It has been shown that rapamycin can block P-gp efflux pump \[[@B15],[@B16]\], but is less potent than cyclosporin. We have studied whether CCI-779 can block P-gp and MRP1 efflux pump and consequently increased doxorubicin sensitivity in these P-gp1 and MRP1 overexpressing cells. Our data demonstrated that at 10 ng/ml of CCI-779, there is no effect on the ID50 of doxorubicin in both P-gp1 and MRP1 overexpressing cells. However, at 1 ug/ml CCI-779 is able to decrease the ID50 from 0.04 to 0.009 ug/ml. The data are summarized in table [3](#T3){ref-type="table"}. Thus, our data suggest that CCI-779 is a weak P-gp blocker compared to those previously reported for rapamycin. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Growth inhibitory effect of Doxorubicin (ID~50~) with and without CCI-779 ::: DOX (ug/ml) DOX+0.01 ug CCI-779 DOX+1 ug CCI-779 ------- -------------- --------------------- ------------------ SCLCA 0.09 ± 0.02 0.07 ± 0.01 0.08 ± 0.02 SCLCR 0.042 ± 0.01 0.058 ± 0.02 0.009 ± 0.005 ::: MTOR expression by westernblot analysis --------------------------------------- It is known that mTOR is a target of CCI-779. It is possible that sensitivity to mTOR inhibitor may be related to mTOR levels. We have studied mTOR expression in the cisplatin sensitive and resistant cells. The result is shown in fig [2](#F2){ref-type="fig"}. There are no discernable differences in mTOR levels in these cell lines. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Westernblot analysis of mTOR in Lane1: SCLCB, Lane 2: SCLCBC. Lane 3: SR-2, Lane 4: SCLC1. Note: there are no differences in mTOR levels in these 4 cell lines. ::: ![](1476-4598-4-25-2) ::: 4E-BP phosphorylation --------------------- It is well known that mTOR phosphorylates 4E-BP and releases its binding from eIF-4E. This results in translation initiation. Inhibition of mTOR will prohibit its ability to phosphorylate 4E-BP and thereby inhibit the translation process. We have studied 4E-BP phosphorylation in SCLC1/SR2, SCLCB/SCLCBC. Our results are shown in fig [3](#F3){ref-type="fig"}. Cisplatin does not affect 4E-BP phosphorylation in cisplatin resistant cell lines. However, the addition of CCI-779 inhibits this process. We have tested two different doses of CCI-779 (150, and 300 ng/ml), our data indicate that 150 ug/ml is as efficient as 300 ug/ml in inhibiting 4E-BP phosphorylation. We have further reduced the dose to 100 ng and the effect on 4E-BP phosphorylation are similar to 150 ng/ml(data not shown). In contrast, in sensitive cells, the phosphorylation of 4E-BP decreased after the addition of cisplatin alone (fig [3](#F3){ref-type="fig"}, panel A&C). This effect is only slightly enhanced by CCI-779. These results correspond with the growth inhibitory effect illustrated in table [2](#T2){ref-type="table"} which showed that CCI-779 decreased the ID50 in cisplatin resistant cells but has only minimal effect on sensitive cells. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Westernblot analysis of 4E-BP phosphorylation in parental and cisplatin resistant small cell lung cancer cell lines. A: SCLC1 (parental line) was treated with 0.1 ug/ml of cisplatin alone, CCI-779 alone or in combination. Lane1: control (untreated cells) which showed high amount of phosphorylated 4E-BP. Lane 2: SCLC1 treated with 0.1 ug/ml of cisplatin which showed a decrease in the amount of phosphorylated 4E-BP, lane 3&4: SCLC1 treated with 150 ng/ml of CCI-779 alone and CCI-779 with cisplatin 0.1 ug/ml respectively. Both showed comparable amount of phosphorylated 4E-BP and are less than control cells and or cells treated with cisplatin alone. B: 4E-BP phosphorylation in SR-2 (cisplatin resistant cell line derived from SCLC1) after treatment with cisplatin with and without CCI-779. Lane 1: Control untreated cells which showed an abundant amount of phospho-4E-BP, lane 2, 3, and 4: SR-2 treated with 5,10 and 0.1 ug/ml of cisplatin. The amount of phospho-4E-BP are not affected by cisplatin at low dose, but slightly decreased at high dose. Lane 5,6 SR-2 treated with CCI-779 at 150 ng and 300 ng/ml respectively. Predominantly unphosphorylated 4E-BP is seen, Lane 7&8: SR-2 treated with 150 ng/ml of CCI-779 combined with cisplatin 5 ug and 10 ug/ml respectively. Small amounts of phosphorylated 4E-BP are seen and much less than those with cisplatin alone (lane 2,3&4). C:4E-BP phosphorylation in SCLCB(parental line). Lane 1: Control untreated cells which showed an abundant amount of phospho-4E-BP, lane 2: SCLCB treated with 0.1 ug/ml of cisplatin showed a decrease amount of phospho-4E-BP. lane 3 & 4 SCLCB treated with 150 ng/ml of CCI-779 alone and CCI-779 150 ng/ml with 0.1 ug/ml of cisplatin respectively. Both showed comparable amount of phospho-4E-BP and is slightly less than cisplatin alone. D: 4E-BP phosphorylation in SCLCBC (cisplatin resistant cells). Lane 1,2,3: SCLCBC treated with CCI-779 150 ng/ml combined with cisplatin at 1,5 and 10 ug/ml of cisplatin respectively. The amount of phospho-4E-BP is minimum and predominantly unphosphorylated form, lane 4&5 SCLCBC treated with CCI-779 at 150 ng and 300 ng/ml, only unphosphorylated 4E-BP predominate. Lane 6,7,8 SCLCBC treated with 10,5 and 1 ug/ml of cisplatin respectively. The amount of phospho4E-BP are similar to the control cells. Lane 9. Control untreated cells. Actin is used as control for protein loading for all cell lines ::: ![](1476-4598-4-25-3) ::: P-70S6 kinase phosphorylation ----------------------------- It is well known that CCI-779 decreases p-70 S6 kinase activity by inhibiting its phosphorylation.. We have determined the levels phospho-P-70 S6 kinase by westernblot analysis. Our results are shown in fig [4](#F4){ref-type="fig"}. Cisplatin at 0.1 ug/ml or 10 ug/ml did not affect the phosphorylation in two cisplatin resistant cell lines, however, the phosphorylation decreased after adding CCI-779. In contrast, in parental cells (SCLC1 and SCLCB), the levels of phospho-p-70 S6 kinase is decreased after adding cisplatin at 0.1 ug/ml and decreased slightly after the addition of CCI-779(data not shown). These findings are similar to the phosphorylation of 4E-BP. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Westernblot analysis of phospho-P70S6kinase in cisplatin resistant cells. A: SR-2 cells. Lane 1. SR-2 treated with cisplatin 10 ug/ml and CCI-779 150 ng/ml, lane 2: SR-2 treated with CCI-779 at 150 ng/ml. Both 1&2 showed only minimal amount of phospho-p70S6kinase. Lane 3&4 SR-2 treated with cisplatin at 0.1 and 10 ug/ml., cisplatin does not affect the amount of phospho-p70S6kinase and showed similar intensity to control cells. Lane 5: control untreated cells. B. SCLCBC cells. Lane 1: Control untreated cells, lane 2&3 SCLCBC treated with cisplatin at 0.1 and 10 ug/ml respectively. Similar amounts of phospho-p70S6kinase were seen. Lane 4 SCLCBC treated with cisplatin 10 ug/ml and CCI-779 150 ng/ml. The amount of phosphor-p70S6kinase is less. Lane 5 SCLCBC treated with CCI-779 at 150 ng/ml. No signal intensity for p-70S6kinase was seen. Actin was used as control for protein loading. ::: ![](1476-4598-4-25-4) ::: Assay for elongation factor alpha (eEF1-α) ------------------------------------------ It has been shown that rapamycin can selectively inhibit translation of mRNA encoding elongation factor alpha. In addition, our cisplatin resistant cell lines also overexpress elongation factor α (fig [5](#F5){ref-type="fig"}, panel A). We have examined whether CCI-779 has effect on elongation factor alpha. The data are shown in fig [5](#F5){ref-type="fig"} (panel B&C). Cisplatin at 10 ug/ml has no effect on eEF-1 α in these resistant cell lines. However, after adding CCI-779, the eEF-1 α is much less. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Westernblot analysis of eEF-1α. AeEF-1α in sensitive and resistant cells. Lane 1. SCLCB, lane 2: SCLCBC, lane 3: SR-2, lane 4: SCLC1. The signal intensity of the resistant cell line SCLCBC and SR-2 are much greater than their sensitive cells SCLCB and SCLC1 counterpart. BeEF-1α in SR-2 cell line. Lane 1: control, lane 2&3: SR-2 treated with 0.5 and 10 ug/ml of cisplatin respectively. Similar signal intensity are seen compared to control, lane 4&5: SR-2 treated with 10 ng and 50 ng/ml of CCI-779 respectively. There is no change in signal intensity of eEF-1α at 10 ng of CCI-779, but decreased after exposure to 50 ng/ml of CCI-779, lane 6: SR-2 treated with 10 ug/.ml of cisplatin with CCI-779 50 ng/ml, the signal intensity also less and similar to those treated with CCI-779 alone. C*eEF-1α in SCLCBC. Lane 1: control, lane 2&3 SCLCBC treated with 0.5 ug and 10 ug/ml of cisplatin respectively. Similar signal intensity of eEF-1α is seen compared to control, lane 4&5: SR-2 treated with 10 ng and 50 ng/ml of CCI-779 respectively, there is no changes in signal intensity at 10 ng/ml of CCI-779, but decreased at 50 ng/ml of CCI-779, lane 6 SCLCBC treated with 10 ug/ml of cisplatin with CCI-779 50 ng/ml. The signal intensity is less and similar to those treated with CCI-779 alone. ::: ![](1476-4598-4-25-5) ::: Discussion ========== The mechanisms of cisplatin resistance have been studied for the last several decades. To date, there is no effective pharmacological manipulation to overcome this form of complex resistance. In this report, we have found that mTOR inhibitor (CCI-779) is able to restore cisplatin sensitivity. This is most likely due to the fact that mTOR inhibitor can down regulate protein synthesis. mTOR regulates translation of proteins which are involved in cell cycle progression and cell growth \[[@B12],[@B29]-[@B32]\]. The two best characterized downstream targets of mTOR are two families of proteins that control translation, the ribosomal protein S6 kinase (S6K) and the eukaryotic initiation factor eIF4E binding protein (4E-BP)\[[@B12],[@B33]-[@B35]\]. S6K1 (p70S6 kinase) phosphorylates S6 which controls the translation of mRNA that possesses an unusual oligopyrimidine tract at its transcription starting site termed 5\'TOP. These 5\'TOP mRNAs encode major components of protein synthesis machinery (ribosomal protein and elongation factor). Another downstream target of mTOR is eLF-4E binding protein 4E-BP. eIF-4E is known to control translation of mRNA with secondary structure (often GC rich)\[[@B36]\]. These mRNAs are known to encode proteins related to proliferation. In this study, we have shown that in cisplatin resistant cells the phosphorylation of 4E-BP is not affected by cisplatin alone, but decreased after adding CCI-779 which indicates that CCI-779 can block the translational initiation. Similar results were obtained for the phosphorylation of p-70 S6 kinase. In contrast, in parental (cisplatin sensitive cells), cisplatin alone is able to decrease the phosphorylation of both 4E-BP and p-70S6 kinase, although this effect is slightly augmented with the addition of CCI-779. These results correspond with our growth inhibitory findings which show that the addition of CCI-779 only slightly augments the growth inhibitory effect of cisplatin in sensitive cells. The ability of cisplatin to inhibit 4E-BP and p-70 S6 kinase in cisplatin sensitive cells has been reported by others \[[@B37]\]. The exact mechanism is not known. Nevertheless, this mode of action appears to diminish in cisplatin resistant cell lines, and these cells are able to readily turn on the translational machinery. Overall, our findings support our notion that the reason why cisplatin resistant cell lines up regulated the mRNA encoding for ribosomal protein and elongation factors is due to the fact that these resistant cells have to turn on the protein synthesis machinery in order to survive. Consequently, by inhibiting the protein translational machinery using CCI-779, cisplatin resistance can be overcome. Furthermore, our preliminary results also show that CCI-779 can inhibit the synthesis of DNA repair proteins, cell cycle protein as well as telomerase which further support our hypothesis. Importantly, these findings also can be applied to de-novo cisplatin resistant cell lines derived from patients who failed treatment (SCLCL, NSCLCW and NSCLCG). The ID50 of cisplatin was decreased by 3--4 folds after CCI-779. The concentration we used to overcome cisplatin resistance is low (10 ng/ml) and is clinically achievable. \[[@B38],[@B39]\] These findings can have significant clinical implication for future use of mTOR inhibitors to overcome cisplatin resistance. It has been previously demonstrated that rapamycin can reverse MDR1 mediated drug resistance \[[@B15]\]. Furthermore, both P-gp1 and MRP1 mediated drug resistance have been reported in lung cancer \[[@B40]-[@B42]\]. Thus, we have also investigated the possible antitumor activity as well as reversal activity of CCI-779 in P-gp1 and MDR1 overexpressing cell lines. Our results demonstrated that both CCI-779 and rapamycin have no antitumor activity in P-gp1 or MRP1 overexpressing cell lines. It is possible that both rapamycin and CCI-779 are recognized by both P-gp1 and MRP1. This is not surprising since it has been reported that rapamycin can be transported by P-gp in renal proximal tubule \[[@B43]\]. We have also found that CCI-779 is a weak reversal agent when compared with those reported for rapamycin. CCI-779 at 1 ug/ml (1 uM) can only decrease the ID50 of doxorubicin by 4.5 fold whereas rapamycin at 1 uM can decrease the ID50 of daunorubicin by 10 fold \[[@B15]\]. This may be due to the fact that CCI-779 is more water soluble and thus is less effective as reversal agent \[[@B44]\] Conclusion ========== We conclude that CCI-779 is able to restore cisplatin sensitivity in small cell lung cancer cell lines which were selected for cisplatin resistance aswell as cell lines derived from patients who failed cisplatin. The most likely mechanism is due to inhibition of translation of proteins which are involved in cisplatin resistance. These findings should be further explored in the clinic. However, CCI-779 has no antitumor activity in P-gp1 and MRP1 overexpressing cell lines. Moreover, CCI-779 can only partially reverse P-gp1 at a higher dose (1 ug/ml). Therefore, CCI-779 is not a good candidate to use in MDR1 or MRP1 overexpressing cells. Acknowledgements ================ This work is supported By VA. Research fund to N. Savaraj and CA 79085 to MT. Kuo.	PubMed Central	2024-06-05T03:55:59.839598	2005-7-20	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181826/", "journal": "Mol Cancer. 2005 Jul 20; 4:25", "authors": [ { "first": "Chunjing", "last": "Wu" }, { "first": "Medhi", "last": "Wangpaichitr" }, { "first": "Lynn", "last": "Feun" }, { "first": "Marcus Tien", "last": "Kuo" }, { "first": "Carlos", "last": "Robles" }, { "first": "Theodore", "last": "Lampidis" }, { "first": "Niramol", "last": "Savaraj" } ] }
PMC1181827	Background ========== Long-term, community-based cohort studies designed to evaluate or compare both the risk of inherited and acquired risk factors for venous thromboembolism (VTE) in young women are lacking, only one Danish cohort study designed and conducted to examine cardiovascular risk factors and the relation with factor V Leiden (DNA was obtained later during follow-up) \[[@B1]\]. Information is rare on comparative differences in incidence rates and risk estimates across different clinical and inherited VTE risk factors generated in a community, i.e. making head-to-head comparisons possible within one study. A community-based thromboembolic risk factor study started in the mid-1990s in Bavaria, the [BA]{.underline}varian [T]{.underline}hrombo[E]{.underline}mbolic [R]{.underline}isk study (BATER), focused on women in the reproductive age \[[@B2]-[@B4]\]. Clinical and hereditary risk markers for VTE, the lifetime history of relevant conditions or medications, and the family history of cardiovascular diseases were documented from 1993 throughout the follow-up period until 2003, i.e., carefully reviewing complaints or findings possibly related to the occurrence of venous clots. The aim of this paper is to present risk estimates for VTE risk markers available in one study, and to provide incidence rates associated with these risk factors based on new VTE cases observed during follow-up. Methods ======= Material and methods of this long-term cohort study has been described in detail in earlier publications \[[@B2],[@B3]\] and particularly in a recent publication in this journal \[[@B4]\]. In brief, we examined a cohort of 4337 young women (18--55 years) living in Bavaria who had at least one follow-up. Data on demography, reproductive life, conditions/diseases, and particularly potential risk factors for VTE were collected through a self-administered questionnaire and subsequent telephone enquiries -- if necessary- to supplement, clarify and verify the information in the questionnaires to set up the year 1993 as common starting point for all cohort members. The source for the data on new (incident) VTE cases was the follow-up questionnaire (self-reported VTE or symptoms potentially compatible with VTE) completed by the study participants. This information was complemented by telephone interviews with the woman and with the treating physician. An external medical reviewer assigned all suspected VTE cases to one of five categories following an a priori defined decision scheme: DEFINITE(confirmation by imaging test), PROBABLE(unequivocal imaging test, other confirmatory tests positive, and anticoagulant therapy), POSSIBLE(unequivocal imaging test, suspicion in other tests, but no anticoagulant therapy), POTENTIAL(only clinical diagnosis without additional diagnostics, and no anticoagulant therapy), and NO VTE(alternative diagnosis). Details were given in a recent publication in this journal \[[@B4]\]. Possible and potential VTE cases were excluded from the analyses in this paper because of diagnostic uncertainty. Women with a history of cancer, with chronic liver diseases, or with known antiphospolipid syndrome were not present in the study. After having given informed consent the women included in this study gave a blood sample at entry into the study. An independent ethics committee approved all study related activities. Whole blood samples were obtained from resting subjects. Blood was put into tubes with trisodium citrate. Plasma was prepared soon after venipuncture by centrifugation for 15 minutes with 3000 to 4000 rpm at room temperature and stored at - 20°C. Protein C and antithrombin activities in plasma were measured by chromogenic substrate assays (Dade Behring, Marburg, Germany). For antithrombin (AT) the activity against factor IIa was determined, Protein C activity was measured after activation of the proenzyme by snake venom. Plasma activities are given as percentage (% of normal) of pooled human normal plasma. Protein C deficiency was defined as less than 77% of normal (5^th^percentile of the non-cases of the cohort). Antithrombin deficiency was assumed if less than 81% of normal (5^th^percentile) Genomic DNA was isolated by mean of QIAmp^®^DNA Blood Kit (Qiagen) according to the manufacturer\'s instructions. The genetic polymorphisms Factor V R506Q (G1691A), Prothrombin G2010A and 5-, 10-methylenetetrahydrofolate reductase (MTHFR) A223V (C677T) were determined using a multiplex PCR with allele-specific primers slightly modifying a previously described method \[[@B5]\]. All blood tests were performed in a blinded manner, i.e. the investigators had no clinical information, nor access to the clinical database. The database was structured to accommodate both concurrent as well as time-dependent variables. Concurrent variables are variables, which describe the woman\'s status at the time of questionnaire response, whereas the outcome variable is time-dependent. While concurrent variables were held in a fixed dataset, a periodic dataset containing information on the occurrence of VTE events along a time axis was created for each participant, using months as a unit of measurement. In this dataset, all exposures of interest in this paper, such as VTE risk factors including genetic markers, refer to the baseline point. Some of the variables of interest (age, BMI, Protein C, AT) were continuous. These variables were dichotomized in order to define a categorical exposure status (exposed -- non-exposed) for the analyses based on incidence or logistic regression. We arbitrarily separated the continuum in two roughly equal intervals such as age under/over 30 or BMI under/over 25 (kg/m2) in order to have sufficient cases for analyses with further stratification. For protein C and AT we used the lower 5th percentile (of the distribution in non-cases) as cut off point. This limit was considered as usual definition for \"deficiency\" and therefore clinically relevant \[[@B6]\]. Simple descriptive tables were prepared. All analyses concerning the occurrence of VTE events over time were performed by adding up individual observation time (1993 until the last contact) for different exposure cohorts and in total. Apart from the overall VTE incidence rate per 10,000 women-years of observation (WY), we calculated also incidence rate ratios to compare the incidence of different sub-groups, e.g., women with factor V Leiden (FVL) mutation compared with women without this genetic mutation. The calculation of the relative risk of occurrence of VTE is based on logistic regression analysis. Crude and adjusted odds ratios (OR) are reported with 95 % confidence intervals (95% CI). All analyses were performed with the statistical packages SAS 8.2 or STATA 8.2. Results ======= The overall cohort encompasses 4337 women with sufficient information in 1993 and one follow-up at minimum. The observational period for our current analysis was 32,656 WYs since 1993. Initially, 6082 eligible women were invited to join the study. Of these, 4372 (71.9%) agreed to participate in the follow-up and with the blood sampling. The main reason for non-participation was blood sampling. The follow-up was carried out until 2003 at most, or otherwise terminated at the time when the last contact was possible to get information about new conditions that may have had occurred. 2076 women could be followed up until 2002/3 (47.9 %), 595 (13.7%) women did not participate in follow-ups before 1999--2001, and the largest proportion of women dropped out during the first years before 1999 (38.4%). Thus, the follow-up period was censored some time before 2002/3 for approximately half of the cohort members, i.e. the last successful contact was defined as \"end of follow-up\". Thirty-four new cases of VTE occurred in the observational period. These cases were finally confirmed and categorized according to diagnostic certainty by an independent medical reviewer as definite (n = 31) or probable (n = 3). Cases with possible/potential VTE (n = 17) were excluded from further analyses because of low diagnostic certainty, i.e. it was not clear whether to classify them in the group cases or non-cases. Out of the 34 definite/probable VTE cases 18 cases (= 52.9%) were associated with \"clinical causes for VTE\" and 16 (= 47.1%) were so-called \"idiopathic\" VTEs. The following \"clinical causes\" were observed prior to occurrence of the new VTE: 4 with previous VTE, 3 pregnancy/puerperium, 4 after an accident, 2 after surgery, 3 immobilization, and 2 after long travel in sitting position. Table [1](#T1){ref-type="table"} depicts the profile of relevant data available at baseline (1993) to get an impression of the group under follow-up. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Distribution of clinically and genetically relevant variables in a cohort of women at baseline of the observational period 1993 -- 2003. The total number of women in this analysis is 4320, i.e. excluding 17 women with a final diagnosis of a possible/potential VTE. Deviations from this number are due to missing information ::: Variables ---------------------------------- ------ ------ -------------- Continuous variables n Mean (SD) Age (years) 4320 26.0 (8.6) Life births, number 1910 1.7 (0.8) BMI^§^ 4309 23.3 (4.1) Protein C (% of normal) 4315 102.4 (15.8) AT (% of normal) 4316 98.4 (11.3) Categorical parameters Percent (%) Own history of VTE No 4279 99.0 Yes 41 1.0 Family history No 3840 88.9 Yes 480 11.1 Age, alternative \<30 2843 65.8 ≥ 30 1477 34.2 Family history of varicous veins No 2395 55.4 Yes 1925 44.6 Family history of MI No 3830 88.7 Yes 490 11.3 BMI, alternative \<25 3218 74.7 ≥ 25 1091 25.3 Ever use of hormone replacement No 4031 93.7 Yes 270 6.3 Family history of stroke No 4013 92.9 Yes 307 7.1 Ever use of oral contraceptives No 346 8.0 Yes 3973 92.0 Education level: Abitur^&^ No 3119 73.2 Yes 1139 26.8 Ever smoker No 2022 46.8 Yes 2296 53.2 Laboratory & genetic markers Factor V Leiden mutation^1^ No 4035 93.7 Yes 271 6.3 Prothrombin mutation^1^ No 4088 96.6 Yes 142 3.4 MTHFR^1^ No 1798 42.5 Yes 2432 57.5 Protein C deficiency^\#^ No 4117 95.4 Yes 198 4.6 AT deficiency^\#^ No 4106 95.1 Yes 210 4.9 ^1^Homozygote & heterozygote together ^§^ Body mass index (kg/m^2^) ^&^maturity for university ^\#^definition see methods ::: The mean age was 26 ± 8.6 years, however, for the dichotomized age variable we used as cut-off point 30 years resulting in strata that contained VTE cases in both age groups. The frequency of other data, family history (first degree relatives) of potentially relevant diseases, conditions and genetic lab parameters is provided in the table [1](#T1){ref-type="table"}. Homo-and heterozygote carriers of mutation were analyzed together because of small numbers or homozygote carriers. Table [2](#T2){ref-type="table"} presents the characteristics of the 34 VTE cases and the remaining \"non-cases\" in the cohort. Varying total numbers in the table are due to missing information particularly in non-cases and genetic characteristics. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Description of the subgroup of VTE cases and non-cases regarding parameters considered in this study as potential VTE risk factors at baseline of the observation period 1993 -- 2003 Only definite and probable VTEs were considered as cases in this table, i.e. possible & potential VTE were excluded. Definitions of variables see text. ::: Non-cases VTE cases Total cohort ----------------------------------------- ----------- ------------- -------------- ----------- ------ ------------- N n (%) N n (%) N n (%) Demographic data Age\: 30+ years 4286 1456 (34.0) 34 21 (61.8) 4320 1477 (34.2) BMI\: 25+ 4275 1076 (25.2) 34 15 (44.1) 4309 1091 (25.3) OC use: yes, ever 4285 3941 (92.0) 34 32 (94.1) 4319 3973 (92.0) Other hormones: yes, ever 4267 267 (6.3) 34 3 (8.8) 4301 270 (6.3) Smoking: yes, ever 4284 2278 (53.2) 34 18 (52.9) 4318 2296 (53.2) Medical history Personal history of VTE\: yes 4286 37 (0.9) 34 4 (11.8) 4320 41 (1.0) Family history of VTE\: yes 4286 470 (11.0) 34 10 (29.4) 4320 480 (11.1) Family history of varicose veins\: yes 4286 1902 (44.4) 34 23 (67.7) 4320 1925 (44.6) Family history of MI\: yes 4286 482 (11.3) 34 8 (23.5) 4320 490 (11.3) Family history of stroke: yes 4286 302 (7.1) 34 5 (14.7) 4320 307 (7.1) Laboratory & genetic markers FVL mutation^§^: yes 4273 267 (6.3) 33 4 (12.1) 4306 271 (6.3) Prothrombin mutation^§^: yes 4199 140 (3.3) 31 2 (6.5) 4230 142 (3.4) Protein C deficiency ^\#^: yes 4281 4087 (95.5) 34 30 (88.2) 4315 4117 (95.4) AT deficiency ^\#^: yes 4282 4073 (95.1) 34 33 (97.1) 4316 4106 (95.1) MTHFR^§^: yes 4199 2415 (57.5) 31 17 (54.8) 4230 2432 (57.5) ^§^homo- and heterozygote together \* significant difference between VTE cases and non-cases (p \< 0.05) ^\#^definition see methods ::: There were some remarkable differences between cases and non-cases that affected differences in VTE risk estimates in further analyses such as incidence rates, incidence rate ratios as well as relative risk estimates (see below): cases were found to be older, to have a higher proportion of elevated BMI, of history of previous VTE, of family history of VTE (first degree relatives). Family history of varicose veins, myocardial infarction, and stroke were included not as VTE risk factors, but conditions with a potential for misclassification by respondents (past history was not validated). Although the frequency of known genetic risk factors for VTE (specifically FVL mutation and prothrombin mutation) seemed to be higher in cases than non-cases, this was not statistically significant (see comparisons below). Only 6 of the 34 women suffering from definite or probable VTE showed any established marker of thrombophilia in the laboratory screen. Two of 6 patients in this group had severe thrombophilia with the combination of Protein C deficiency (48 % activity) and heterozygous Factor V Leiden or a homozygous FVL mutation. The other 3 patients exhibited only one positive laboratory marker and were either heterozygous for FVL (n = 2) or prothrombin mutation (PTM, n = 1). The numbers however were too small for sub-analyses. 24 patients showed no detectable marker of thrombophilia, 2 out of these patients demonstrated a homozygous MTHFR mutation. We observed 34 new definite/probable VTE cases within the 32,508 WYs of observation, i.e., an incidence rate of 10.4 per 10,000 WYs. Table [3](#T3){ref-type="table"} shows incidence rates of VTE stratified by presence (= exposed) or absence (= non-exposed) of the variables considered as potential \"risk factors\" in this analysis. Marked differences of incidence rates were observed across the variables listed in table [3](#T3){ref-type="table"}, i.e., comparing the incidence between exposed and non-exposed in each of the variables. Several of the 15 compared parameters showed significantly elevated incidence rate ratios (relative risk): some demographic variables (advanced age, elevated BMI), data of the medical history (history of previous VTE, family history of VTE or family history of varicose veins). But to our surprise none of the established genetic VTE risk markers showed a significantly increased VTE risk in this analysis. However, taking the risk estimates at face-value, three of the five markers (positive FVL, prothrombin mutation, and protein C) depicted a 2-fold increase in risk, although statistically not significant. These however were only crude comparisons, i.e. do not account for the simultaneous influence of any of the other VTE risk factors. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Incidence rates for VTE (definite and probable) based on 4320 women and 32,508 WY of observation (1993 -- 2003). Tabulation by parameters considered as potential VTE risk factors. Descriptive tabulation of events per 10,000 WY in the exposed and non-exposed group. Incidence rate ratio (IR) and 95% confidence interval (95% CI). \"Exposed\" was defined as the group where the risk was assumed to be higher. ::: Exposed Non-exposed Overall Exp. vs. Non-exposed --------------------------------------- --------- --------------------------- --------- --------------------------- -------- --------------------------- ------------------ Exposed vs. non-exposed WY VTE incidence per 10^4^WY WY VTE incidence per 10^4^WY WY VTE incidence per 10^4^WY IR (95% CI) Demographic data Age: 30+ years vs. \<30 12,346 17.0 20,162 6.5 32,508 10.5 2.6 (1.3--5.7) BMI: 25+ vs. \<25 8,764 17.1 23,696 8.0 32,460 10.5 2.1 (1.01--4.4) OC use: ever vs never 30,388 10.5 2,116 9.5 32,504 10.5 1.1 (0.3--9.6) Other hormones: ever vs. never 2,484 12.1 29,896 10.4 32,380 10.5 1.2 (0.2--3.7) Smoking: yes, ever 16,878 10.7 15,615 10.2 32,493 10.5 1.04 (0.5--2.2) Medical history Personal history of VTE: yes 318 125.8 32,190 9.3 32,508 10.5 13.5 (3.5--38.3) Family history of VTE: yes 3,810 26.3 28,698 8.4 32,508 10,5 3.1 (1.3--6.8) Family history of varicous veins: yes 14,764 15.6 17,744 6,2 32,508 10.5 2.5 (1.2--5.7) Family history of MI: yes 3,919 20.4 28,589 9,1 32,508 10.5 2.2 (0.9--5.1) Family history of stroke: yes 2,480 20.2 30,028 9,6 32,508 10.5 2.1 (0.6--5.5) Laboratory & genetic markers FVL mutation^§^: yes 2,105 19.0 30,308 9.6 32,413 10.2 2.0 (0.5--5.7) Prothrombin mutation^§^: yes 1,010 19.8 30,817 9.4 31,827 9.7 2.1 (0.24--8.3) Protein C deficiency^\#^: yes 1510 26.5 30,958 9.7 32,468 10.5 2.7 (0.7--7.8) AT deficiency^\#^: yes 1648 6.1 30,828 10.7 32,476 10.5 0.6 (0.01--3.4) MTHFR^§^: yes 18,423 9.2 13,404 10.4 31,827 9.7 0.9 (0.4--1.9) ^§^homo- and heterozygote together ^\#^definition see methods ::: Similar to the evaluation of the relative risk estimates using the incidence rate ratio in the cohort approach, the evaluation with crude odds ratios showed an almost identical set of significant risk markers (table [4](#T4){ref-type="table"}): higher age, elevated BMI, personal history of previous VTE, family history of VTE & varicose veins, but in addition also family history of myocardial infarction. None of the five genetic markers was significantly associated with the VTE risk in the crude risk assessment. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Potential VTE risk factors and risk estimates for VTE (definite and probable) based on 4320 women. Comparative assessment with logistic regression analysis: Odds ratio (OR) and 95% confidence interval (95% CI) ::: Non-cases Cases Crude OR (95% CI) Adjusted\* OR (95% CI) ---------------------------------------------- ------------- --------- ------------------- ------------------------ ------------------ ------------------ Non-exposed Exposed Non-exposed Exposed Demographic data Age: 30+ years vs. \<30 2830 1456 13 21 3.1 (1.6--6.3) 2.4 (1.1--5.3) BMI: 25+ vs. \<25 3199 1076 19 15 2.3 (1.2--4.6) 2.0 (0.9--4.1) OC use: ever vs never 344 3941 2 32 1.4 (0.3--5.9) 1.3 (0.3--5.8) Other hormones: ever vs. never 4000 267 31 3 1.4 (0.4--4.8) 0.86 (0.2--2.9) Smoking: ever vs. never 2006 2278 16 18 1.0 (0.5--1.9) 0.8 (0.4--1.7) Medical history Personal history of VTE: yes vs. no 4249 37 30 4 15.3 (5.1--45.9) 6.6 (1.8--24.6) Family history of VTE: yes vs. no 3816 470 24 10 3.4 (1.6--7.1) 2.4 (1.0--5.4) Family history of varicous veins: yes vs. no 2384 1902 11 23 2.6 (1.3--5.4) 1.9 (0.8--4.1) Family history of MI: yes vs. no 3804 482 26 8 2.4 (1.1--5.4) 2.0 (0.8--4.6) Family history of stroke: yes vs. no 3984 302 29 5 2.3 (0.9--5.9) 1.2 (0.4--3.5) Laboratory & genetic markers FVL mutation^§^: yes 4006 267 29 4 2.1 (0.7--5.9) 2.0 (0.7--6.0) Prothrombin mutation^§^: yes 4059 140 29 2 2.0 (0.5--8.5) 2.3 (0.5--10.0) Protein C deficiency^\#^: yes 4087 194 30 4 2.8 (0.98--8.0) 3.0 (0.9--10.4) AT deficiency^\#^:yes 4073 209 33 1 0.6 (0.1--4.3) 0.5 (0.1--3.5) MTHFR^§^: yes 1784 2415 14 17 0.9 (0.4--1.8) 0.95 (0.5--1.97) ^§^homo- and heterozygote together ^\#^definition see methods \* adjusted for all other variables ::: When the risk assessment of all mentioned parameters underwent a fully adjusted analysis, i.e. controlling for all other respective variables, the overall findings approximately were the same, but only higher age, personal and family history of VTE increased significantly the risk of VTE within the 10-year period. None of the genetic markers had a statistically significant impact on VTE risk -- even not after adjustment for other potential risk factors. However, the two measured mutations and protein C remained at an apparently elevated VTE risk level -- although these results were not statistically significant. Virtually identical risk estimates were observed when the Cox regression was used instead of the logistic regression (data not shown), although based again on small numbers of exposed cases. Instable risk estimates due to small numbers and many adjustment variables cannot be excluded. No significant interaction terms were found in the analyses (data not shown). Discussion ========== To our knowledge, there are no long-term community-based cohort studies designed to evaluate or compare the risk of inherited or acquired risk factors for venous thromboembolism (VTE) in women under the age of 50 years, except for a Danish cohort study which, at least initially, targeted general at cardiovascular risk factors and not VTE specifically (DNA was obtained later during follow-up) \[[@B1]\]. Moreover, this study was specifically focused on factor V Leiden. It was our aim to evaluate or compare the absolute risk and risk ratio of established clinical or genetic risk markers for VTE. In the past, most risk factor studies for VTE were restricted to clinically available markers such as age, BMI, previous VTE, family history, or acute factors (immobilization, surgery, accidents, pregnancy/puerperium, and hormonal contraceptive use) and based on clinical or cross-sectional, observational studies or analyses in administrative databases. Many observational studies or cohort studies in young women did not consider inherited factors (overview about incidence and risk factor studies in \[[@B7],[@B8]\]). Cohort studies in the population rarely included or reported genetic markers for thrombophilia and acquired, lifestyle-related risk factors, except the Physicians Health Study for example -- the latter however only for males over 40 years of age \[[@B9]\], or the above mentioned Danish cohort study \[[@B1]\]. Other studies with focus on markers for hereditary thrombophilia were performed in patients (e.g. in anticoagulant clinics), in relatives of carriers of genetic mutations but not in the general population \[[@B6],[@B10]-[@B12]\]. Point estimates for thrombosis-free survival in carriers of major thrombophilic states are often restricted to the selected cohort of family members only (overview in Crowther \[[@B13]\]). Moreover, the evaluation of genetic markers often does not consider the impact of clinically available risk factors and the design was mainly restricted to clinical or case-control studies. Thirty-four VTE cases, classified as definite or probable, occurred within this period, which is equivalent with about 10 per 10,000 WYs. At the first glance, this incidence seems to be high. However, this might be the result of the specifics of our study: We put great effort on the detection of potential cases and -- even more important -- we included all definite and probable cases, whereas most of the reported incidence rates in young women refer only to \"confirmed\" and so-called \"idiopathic VTE\", i.e. excluded all cases that occurred in temporal relationship to other potential reasons such as pregnancy/puerperium, surgery, or immobilization, for example. A similar overall incidence rate of 12.3 events/10^4^person-years was observed in the Danish cohort study \[calculated from -- 1 -\], which however covers both gender and a higher mean age (45 years in the Danish study vs. 26 years in our study). Idiopathic VTE, however, reflects only a smaller part of all confirmed VTE cases \[[@B14]\]. In our cohort study we found roughly 53 % so-called \"idiopathic\" (primary) VTE cases, and the other roughly 47% of cases had a previous VTE in their history, or pregnancy/puerperium, surgery, or other reasons for immobilization/long bed-rest shortly prior to the VTE event. Thus, the incidence of \"idiopathic VTE\" observed in this study can be estimated to be about 5 per 10,000 WY. This is in agreement with other reported incidence rates in the general population \[[@B1],[@B7]\]. The incidence estimates for definite VTE ranges between 1 to 6 per 10^4^WY in non-users or oral contraceptives (OC) and 2 to 10 per 10^4^WYs in OC users. Older studies depicted almost-always higher incidence rates than more recently performed studies (see overview in \[[@B7]\]). A recent systematic review \[[@B8]\] came to a pooled incidence of definite VTE for the general population of 5 per 10,000 person years, similar in males and females, and found that around 40% of VTE cases were \"idiopathic\". We conclude that our data can be generalized for the female population of this region in the fertile age range. This conclusion is supported by results of a prospective, community-based cohort study \[[@B9]\] that found a VTE incidence rate of 2.7 \"primary VTE cases\" per 10^4^person-years (equal to idiopathic: no previous VTE history, no cancer, no surgery or trauma), however, in males aged 40--49 years. The absolute risks (incidence rates) varied markedly among those exposed or non-exposed by genetic and acquired VTE risk factors in our cohort study (see table [3](#T3){ref-type="table"}). The relatively small numbers of women exposed to genetic VTE markers (see also table [2](#T2){ref-type="table"}) should be taken into account before drawing conclusions. Thus, we have to be worry about statistically robust results in most of the sub-groups. The crude, not adjusted comparison between \"exposed\" (= risk factor present) and \"non-exposed\" (= risk factor absent) showed incidence ratios (OR) ranging from 0.6 (AT deficiency) to 13.5 (personal history of VTE). The highest VTE incidence rate was found for women with a history of previous VTE (125.8 VTE cases per 10^4^WY -- compared with 9.3 in women without VTE history), i.e., a 13.5fold increased incidence rate ratio (see table [3](#T3){ref-type="table"}). Other significant incidence rate ratios were observed for higher age, elevated BMI, family history of VTE (and for varicose veins). After fully controlling for all other risk factors (logistic regression) only age, personal and family history of VTE remained significant risk factors (see table [4](#T4){ref-type="table"}). A significant role of the family history of VTE has been reported in several studies \[[@B15]-[@B17]\]. It is surprising that in our study the aggregate variable \"family history\" was more important than the individual genetic or related markers. The impact of individual genetic markers on VTE risk (FVL, PTM, MTHFR) was not statistically significant in the logistic regression analysis (table [4](#T4){ref-type="table"}) -- possibly also due to the small number of incident cases. This is also true for Protein C and AT deficiency, i.e. when using the lower 5^th^percentile as usual definition for deficiency \[[@B18]\]. Even FVL mutation did not show significantly increased VTE risk, although the point estimate (OR = 2.0) -- at face-value -- is compatible with the estimate of another, but much larger cohort study \[[@B1]\], but lower than reported from several case-control studies. This difference between risk estimates for FVL in cohort and case-control studies is probably due to methodological reasons \[[@B1]\]. A similarly weak impact of genetic markers was also observed in follow-up studies of selected groups: Carriers of genetic polymorphisms have been followed prospectively and found a low absolute annual incidence of VTE \[[@B10],[@B12],[@B19]\]. Another prospective cohort observed a low incidence of VTE in otherwise healthy thrombophilic children \[[@B20]\] or asymptomatic family members who are carriers of factor V Leiden \[[@B11],[@B12]\] or other family studies \[[@B21]\]. Deficiency of AT and PC activity had also no significant impact on thrombotic risk. We cannot exclude however that a considerable part of such unexpected results for PC and AT activity may be related to pre-analytical conditions in our field study. From a laboratory perspective in a few women polymorphisms leading to decreased levels of analytical results but not to clinical manifest thrombophilia are quite possible. An explanation might be that silent mutations and polymorphisms cause reduced activities in the laboratory assays not reflecting a potential clinical problem \[[@B22],[@B23]\]. Without a positive personal or family history of VTE such results should at least be confirmed with additional laboratory tests and family examinations before informing the patients of a potential thrombophilic diathesis or before recommending respective medical prophylaxis. There is an increasing debate about the role of genetic factors in the prediction of future VTEs and thus sustaining a controversy about genetic screening. This issue is a current controversy in the literature \[[@B24],[@B25]\]. Clinical reports often suggest a high VTE recurrence rate in patients with previous VTE \[[@B26]\], but we found this phenomenon only in 4 of our 34 incident VTE cases. A recently published community-based cohort study of FVL carriers & non-carriers observed no significant difference on VTE incidence between both groups, except for women ≥ 60 years of age \[[@B27]\]. In general, our results support the notion that genetic parameters alone are relatively weak long-term risk factors; the occurrence of VTE requires interaction of both inherited and acquired risk factors or gene-gene-interactions \[[@B28]\]. The clinical VTE risk factors with significantly elevated incidence rate ratios such as advanced age and elevated BMI, but also history of previous VTE, family history of VTE or family history of varicose veins are not new. Most physicians are aware of these risk factors and consider them in practice. No incidence difference was also found for ever smoking which is rarely considered as risk factor for VTE. Hypertension was not analyzed. No significant prognostic impact was found for ever use of hormonal treatment/contraception at baseline. This is plausible because sex hormones effects are not general characteristics but short-term acting factors, which need another statistical approach, and was not the aim of this study. The influence of more acutely acting risk factors -- such as immobilization, surgery, long-haul flights, and use of drugs (e.g. OCs) was intentionally excluded from this analysis, although interactions between hereditary factors and acute, environmental pattern are known \[[@B21],[@B29],[@B30]\]. Only parameters that were available at baseline and likely to affect the long-term development were eligible for this analysis. Other influential risk factors or preventive measures have to be considered when discussing activities to reduce a predicted increased risk in the medical practice. It was not the aim of the study and data are neither available to test the effect of preventive measures nor the effect of additional risk factors in the immediate period before the event occurred. This would require another study design and a separate study with sufficient power for such questions. It is a limitation of this long-term cohort study, however, that the number of incident, confirmed (definitive and probable) VTE cases was still small in absolute numbers (n = 34) in this cohort observation period of 32,508 years of observation. The low incidence can be explained by the young average age (26 years at entry). Thus, incidence & and risk estimates have wide confidence intervals and conclusions are limited. Rare combinations of risk markers have not yet materialized in one single VTE case. This makes it difficult to further divide into cases that occurred in presumably exposed or unexposed sub-groups. This is particularly true if the potential risk factors (exposure) are infrequent, as it is for genetic markers. It is a problem of the study design that no attempt was made to confirm of deficiencies (second blood sample) and no family study was planned to assess inheritance of deficiencies such as AT or PC. In case of FVL mutation the adjusted risk is apparently increased about 2fold (non-significant) but only 4 cases had a positive test of FVL mutation (for prothrombin mutation only 2 cases). If one then adjusts for 14 other potential factors, there is obviously a statistical problem: Resulting risk estimates might be unstable, i.e. drifted in either direction. Even if one focuses on the crude OR, which is similar to the adjusted estimate in this case, careful interpretation is warranted. It cannot be excluded that the \"true risk\" of the mutation markers is 2fold increased, although the risk estimates do not favor of such a conclusion. Insofar, future analyses will benefit from an improved point of departure (longer observation, more cases). Conclusion ========== In conclusion, estimation of VTE risk cannot be based on genetic characteristics alone -- but only in combinations with available clinical information. Genetic markers play obviously a limited role in the long-term prediction of VTE. Genetic markers form together with the \"environmental circumstances\" the disposition, i.e. together with the family history of cardiovascular events, specifically venous events. If the disposition translates into an event, this is obviously more influenced by \"longstanding clinical VTE risk factors\", factors such as positive medical history, advanced age, and elevated BMI than specific genetic factors. However, there are obviously other important, more acutely affecting environmental factors such as immobilization, surgery, and treatment with drugs that influence coagulation. The latter factors can be used to reduce the basic risk determined by long-term acting acquired risk factors modulated by inherited factors. Acknowledgements ================ We thank Professor Dr. Liane Will-Shahab for the medical review of all suspected VTE cases. We also thank Dr. Sabine Möhner for executing the follow-up over the years, and Andrea Dick for their work with the blood samples and determination of lab data.	PubMed Central	2024-06-05T03:55:59.842490	2005-7-20	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181827/", "journal": "Thromb J. 2005 Jul 20; 3:8", "authors": [ { "first": "Michael", "last": "Spannagl" }, { "first": "Lothar AJ", "last": "Heinemann" }, { "first": "Thai", "last": "DoMinh" }, { "first": "Anita", "last": "Assmann" }, { "first": "Wolfgang", "last": "Schramm" }, { "first": "Rolf", "last": "Schürmann" } ] }
PMC1181828	Background ========== Colorectal cancer kills approximately 500 000 people worldwide each year \[[@B1]\], thus treatments which produce only modest improvements in survival may have an enormous health impact. Current practice is to base therapeutic decisions and prognostic advice on clinico-pathological data, however within conventional staging groups many genetic and molecular tumour subtypes exist. This heterogenicity of tumour genotypes accounts for much of the observed variation in recurrence rates and clinical responses to available therapies. Advances in our understanding of the molecular biology of colorectal cancer have fuelled the search for novel molecular prognostic markers with which to complement existing staging systems. Applied to clinical practice, these putative markers could be used to identify groups of patients with differing relative risks of recurrence and improve patient stratification for adjuvant treatment. Immunohistochemical studies in colorectal cancer have tended to investigate expression of individual proteins in relation to prognosis, and relatively few studies have focused on the analysis of multiple markers in combination. However, one such combination which may be of value in colorectal cancer is the combined p53 / Bcl-2 phenotype, as suggested by Manne et al, who described the p53/Bcl-2 phenotype of 134 patients with Dukes stage A-D tumours, finding that this combination gave independent prognostic information which was superior to that of either marker on its own \[[@B2]\]. The same group have subsequently shown in a larger cohort of 234 patients with tumours of the distal colo-rectum that combined p53/Bcl-2 analysis may provide stronger prognostic information than nodal status \[[@B3]\]. Other researchers have reported similar findings in differing groups of colorectal cancer patients \[[@B4],[@B5]\], with the p53(-)/Bcl-2(+) subset appearing to define a group of patients who appear to have a prolonged survival, although this has not always retained independence on multivariate analysis \[[@B6],[@B7]\]. Alternatively, it has been suggested that opposing p53(+)/Bcl-2(-) phenotype defines a particularly poor prognosis subset \[[@B8]\], or that Bcl-2 positivity in combination with either p53, p21 or mdm-2 confers a good prognosis \[[@B9]\]. Since its first description in 1998, tissue microarray (TMA) analysis \[[@B10]\] has been employed for the immunohistochemical analysis of target protein expression in a wide range of primary tumour types. Initial fears that the reduced amount of individual tumour tissue analysed using this technique might not be representative of the tumour as a whole appear largely unfounded \[[@B11]\], and the strengths of this approach lie in its ability to provide a rapid turnover of results from very large patient cohorts, whilst reducing variability in experimental conditions and reducing costs \[[@B12]\]. Two previous studies have utilised tissue microarray technology to investigate the combination of p53/Bcl-2 in rectal cancer. In the first, Hoos et al, studied 100 patients with lymph-node negative rectal cancer, and found an increased number of deaths in the patients with p53(+)/Bcl-2(-) tumors (7/33, 21.2%) compared with the p53(-)/Bcl-2(+) group (2/13, 15.4%) \[[@B13]\], and a recent study of 269 patients with Dukes stage A-D rectal cancer reported a lower rate of metastatic disease in p53(-)/Bcl-2(+) tumours, with p53(+)/Bcl-2(-) tumours showing a worse outcome \[[@B14]\]. The aim of the present study was to investigate the prognostic significance of the combined p53/Bcl-2 phenotype in a large scale, representative sample of colorectal cancers using tissue microarrays, in order to determine the potential utility of this combination of molecular markers in guiding treatment of patients in the UK. Patients and methods ==================== Patients and specimens ---------------------- The study population comprised 462 patients undergoing elective surgery for a single, non-metachronous histologically proven primary colorectal cancer at University Hospital Nottingham between 1^st^January 1994 and 31^st^December 2000. Data regarding tumour site, stage, histological type and tumour grade have been recorded prospectively for these patients. Only patients with lymph node positive disease were routinely treated with adjuvant chemotherapy, comprising 5-flurouracil and folinic acid. The original histopathological slide sets and pathological reports for all cases were obtained from the hospital archives, and reviewed to confirm the diagnosis and the accuracy of existing data. However, in a minority of cases only very limited slide sets were available and the full clinico-pathological dataset was incomplete. In addition, we did not attempt to ascertain the vascular invasion status of tumours where this information had not been previously recorded. Follow-up data regarding the date and cause of death for this cohort of patients has been provided prospectively by the UK Office for National Statistics. Follow-up was calculated from the date of resection of the primary tumour, and all surviving cases were censored for survival analysis at 31^st^December 2003. Disease specific survival was used as the primary end-point. The Local Research Ethics Committee granted approval for the study. Preparation of the tissue microarray ------------------------------------ Tissue microarrays were constructed as described previously \[[@B10]\]. 5 μm haematoxylin and eosin (H&E) stained slides were used to identify and mark out representative areas of viable tumour tissue. 0.6 mm needle core-biopsies from the relevant areas of corresponding paraffin-embedded blocks were then placed at defined coordinates in the recipient paraffin array blocks using a manual arrayer (Beecher Instruments, Sun Prarie, WI). Array blocks were constructed at a density of 80--150 cores per array. Analysis of each marker was performed on a single core from each tumour. This typically shows over 90% concordance with conventional whole section analysis of tumour markers and has been validated previously \[[@B11]\]. Immunohistochemical methods --------------------------- Immunohistochemistry was performed using a standard avidin-biotin peroxidase method with p53 mAb clone DO-7 (1:20 dilution, Dako Ltd, Ely, UK), and Bcl-2 mAb clone 124 (1:50, Dako Ltd, Ely, UK). Briefly, 5 μm array sections were deparaffinised with xylene, rehydrated through graded alcohol and immersed in methanol containing 0.3% hydrogen peroxide for 10 minutes to block endogenous peroxidase activity. Heat induced epitope retrieval consisting of 20 minutes microwave treatment in pH 6.0 citrate buffer was necessary with the DO-7 mAb. Endogenous avidin/biotin binding was blocked using an avidin/biotin blocking kit (Vector Labs, USA) and sections were treated with 100 μl of normal swine serum (NSS) for 10 min to block non-specific binding of the primary antibody. Test sections were incubated with 100 μl of primary antibody for 1 hr at room temperature. Positive control tissue comprised whole sections of human tonsil. Primary antibody was omitted from negative control sections, which were incubated in NSS. After washing with TBS sections were incubated with 100 μl of biotinylated goat anti-mouse/rabbit immunoglobulin (Dako Ltd, Ely, UK) diluted 1:100 in NSS for 30 min, followed by 100 μl of pre-formed streptavidin-biotin/horseradish peroxidase (HRP) complex (Dako Ltd, Ely, UK) for 60 min at room temperature. Staining was finally visualised using 3, 3\'-Diaminobenzidine tetrahydrochloride (DAB, Dako Ltd, Ely, UK). Evaluation of staining ---------------------- Evaluation of the staining was carried out by two observers (NFSW and ZM) blinded to the clinico-pathological data, with a consensus decision in all cases. For each antigen tumours were classified in two groups with the defined cut-off values of: p53 overexpression if \>10% tumour nuclei stained, irrespective of staining intensity, and Bcl-2 overexpression if cytoplasmic staining (any intensity) was identified in \>30% of the tumour cells. These values were determined from previous studies using the same reagents in colorectal cancer \[[@B13]\], although it has recently been shown in a meta-analysis that when using the DO-7 antibody, reducing the cut-off value to 1% of the nuclei stained does not alter the outcome of subsequent survival analysis \[[@B15]\]. In order to further investigate the significance of the combined p53 / Bcl-2 phenotypes, the individual staining patterns for these antigens were combined and then analysed as an additional group. Statistical methods ------------------- All statistical analyses were performed using the SPSS package (version 11 for Windows, SPSS Inc., Chicago, IL). Associations between categorical variables were examined using cross tabulation and the Pearson chi-square test for categorical variables. Kaplan-Meier curves were derived in order to assess disease-specific survival, and the significance of differences in disease-specific survival between groups was calculated using the log-rank test. Complete outcome data was available for all but one patient in the study. Patients whose deaths related to their colorectal cancer were considered in the disease-specific survival calculations. However, patients who died from postoperative complications (deaths within 1 month of the date of surgery), and patients with carcinoma in-situ (pT~is~tumors) were excluded from the survival analyses. Deaths resulting from non-colorectal cancer related causes were censored at the time of death. Multivariate analysis using the Cox proportional-hazards model was employed to determine hazard ratios and identify variables with independent prognostic significance in this cohort. In all cases Pvalues \<0.05 were considered statistically significant. Results ======= Clinico-pathological data ------------------------- The arrayed tumours were found to be broadly representative of the spectrum of colorectal cancer encountered in the UK (Table [1](#T1){ref-type="table"}). The median age at the time of surgery was 72 years, consistent with a median age at diagnosis of colorectal cancer of 70--74 years in the UK \[[@B16]\]. At the time of censoring for data analysis 49% of patients had died from their disease, 37% were still alive and 14% were deceased from non-colorectal cancer related causes. 38 deaths occurred within one month of surgery. The median length of follow-up available for surviving patients was 75 months (range 36--116). The overall median five-year disease-specific survival for the cohort was 58 months, which is comparable with the approximately 45% five-year survival seen for colorectal cancer in the UK \[[@B17]\]. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Clinico-pathological variables for the patient cohort (n = 462). ::: Age (years) Median 72 ---------------------------------- ------------------------------------- ----------- Gender Male 266 (58%) Female 196 (42%) Status Alive 169 (37%) Dead (colorectal cancer related) 228 (49%) Dead (non-colorectal cancer causes) 64 (14%) Unknown 1 Histological type Adenocarcinoma 392 (85%) Mucinous carcinoma 51 (11%) Columnar carcinoma 4 (1%) Signet ring carcinoma 7 (1%) Unknown 8 (2%) Histological grade Well differentiated 29 (6%) Moderately differentiated 353 (77%) Poorly differentiated 71 (15%) Unknown 9 (2%) Tumour site Colon 238 (52%) Rectal 181 (39%) Unknown 43 (9%) TNM stage 0 (T~is~) 3 (1%) 1 69 (15%) 2 174 (38%) 3 155 (33%) 4 54 (12%) Unknown 7 (2%) Extramural vascular invasion Negative 224 (48%) Positive 128 (28%) Unknown 110 (24%) ::: Antigen expression ------------------ The observed frequencies of expression for p53 and Bcl-2 are shown in table [2](#T2){ref-type="table"}. Although it has previously been stated that the number of cores uninterpretable due to tissue loss or damage in tissue microarray studies may exceed 20--30% of the total, this was not observed in the current study. The highest number of uninterpretable cases (30 cases / 6.5% of total) was seen when the results for p53 and Bcl-2 were combined. Representative examples of positive and negative staining for each antigen are shown in figure [1](#F1){ref-type="fig"}. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Frequencies of immunohistochemical expression of p53 and Bcl-2. ::: Antigen Number (%) positive (high) Number (%) negative (low) Number (%) missing --------------------------------- ---------------------------- --------------------------- -------------------- p53 221 (47.8) 224 (48.5) 17 (3.7) Bcl-2 199 (43.1) 238 (51.5) 25 (5.4) p53 (-) / Bcl-2 (+) combination 112 (24.2) 320 (69.3) 30 (6.5) ::: ::: {#F1 .fig} Figure 1 ::: {.caption} ###### A & B show cores from a tumour demonstrating positive staining for p53 (A), and negative staining for Bcl-2 (B). C & D show cores from a p53 negative (C) / Bcl-2 positive (D) tumour. All at x20 original magnification. ::: ![](1477-7819-3-47-1) ::: p53 expression -------------- p53 in its wild-type form is undetectable by conventional immunohistochemical methods, due to its short half-life. In contrast, most mutations in the p53 gene lead to expression of a stable protein product which accumulates in the nucleus and is detected by the DO-7 antibody. In this study overexpression of p53 was detected in 221/445 tumours (49.6%). On univariate analysis no associations were found between p53 expression and clinico-pathological variables including tumour grade, stage and the presence of extramural vascular invasion, however a significant relationship was noted between p53 expression and disease specific survival. On Kaplan-Meier analysis (figure [2](#F2){ref-type="fig"}), patients with p53(-) tumours demonstrated a significantly longer mean disease specific survival (DSS) of 76 (95% CI 69--83) months, as compared with a mean DSS of 64 (95% CI 57--71) months in patients with p53(+) tumours (p = 0.0240). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Kaplan-Meier plot for disease specific survival, p53 (+) vs p53 (-) tumours (n = 408). ::: ![](1477-7819-3-47-2) ::: Bcl-2 expression ---------------- Overexpression of Bcl-2 protein was detected in the cytoplasm of 199/437 evaluable tumours (45.5%). In contrast, no nuclear expression of Bcl-2 was observed. Although no strong associations between Bcl-2 expression and clinico-pathological variables were noted, there was a trend on univariate analysis towards increased cytoplasmic accumulation of Bcl-2 in moderate and poorly differentiated tumours compared with well differentiated tumours (p = 0.076). On Kaplan-Meier analysis, no association was found between Bcl-2 expression and survival (p = 0.0865, figure [3](#F3){ref-type="fig"}). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Kaplan-Meier plot for disease specific survival, Bcl-2 (+) vs Bcl-2 (-) tumours (n = 400). ::: ![](1477-7819-3-47-3) ::: p53(-) / Bcl-2(+) phenotype --------------------------- Comparing results for both Bcl-2 and p53, a significant reciprocal pattern of expression was seen (table [3](#T3){ref-type="table"}, p = 0.023). Combining the results for these two markers, a total of 112/432 tumours (25.9%) displayed the p53(-) / Bcl-2(+) phenotype. A trend was noted toward a higher prevalence of this phenotype in earlier stage tumours, with a prevalence of 33.9%, 26.3%, 24.1% and 14.3% in TNM stage I-IV tumours respectively (p = 0.078). Additional strong correlations were noted between the p53(-) / Bcl-2(+) phenotype and DSS on univariate analysis (p = 0.004), and on Kaplan-Meier analysis, with a mean DSS of 84 (95%CI 75--92) months in patients with p53(-) / Bcl-2(+) tumours compared with a mean survival of 65 (95% CI 59--71) months in the remaining patients (figure [4](#F4){ref-type="fig"}) (p = 0.0032). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Co-expression of p53 with Bcl-2. ::: Number (%) p53 (+) Number (%) p53 (-) ---------------------- -------------------- -------------------- Number (%) Bcl-2 (+) 129 (29.8) 106 (24.5) Number (%) Bcl-2 (-) 87 (20.1) 111 (25.6) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Kaplan-Meier plot for disease specific survival comparing p53 (-) / Bcl-2 (+) tumours with alternative phenotypes (n = 396). ::: ![](1477-7819-3-47-4) ::: Multivariate analysis --------------------- ### HRs for individual markers A multivariate analysis of factors influencing survival was performed using the Cox proportional hazards model. Analysis included all available cases. Of the conventional clinico-pathological variables analysed, significant independent prognostic value was demonstrated only for TNM stage, and for vascular invasion status. In contrast, neither p53 expression, nor Bcl-2 expression alone were found to be independent markers of prognosis. Considering the combined p53(-) / Bcl-2(+) phenotype, this was found to have independent prognostic value, conferring a significant survival advantage as compared with the other possible combinations of p53 / Bcl-2 expression (table [4](#T4){ref-type="table"}). ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Cox multivariate analysis. ::: Variable Category OR 95% CI pvalue -------------------------- ------------------------- -------- ---------------- ---------- TNM stage I 1 \<0.001 II 1.427 0.758--2.689 III 3.170 1.728--5.816 IV 22.070 11.309--43.068 Unknown 2.895 0.640--13.102 Vascular invasion status Negative 1 \<0.001 Positive 2.156 1.525--3.047 Unknown 1.221 0.798--1.869 p53/Bcl-2 status Non- p53(-) / Bcl-2 (+) 1 0.029 P53(-) / Bcl-2 (+) 0.659 0.452--0.959 ::: Discussion ========== Previous studies have suggested that combined p53 and Bcl-2 immunostaining may provide useful prognostic information in colorectal cancer; however these conclusions have been derived from varying study populations, with heterogeneous methodology and limited sample sizes. To our knowledge no previous study has investigated these findings in such a large, representative cohort of patients from the UK. Using widely validated antibodies and established techniques, our results indicate that patient stratification by combined p53/Bcl-2 phenotype provides tumour stage-independent prognostic information; specifically, that a subset of up to a quarter of colorectal cancer patients display a good prognosis p53(-)/Bcl-2(+) phenotype. In line with previous studies, we found no positive correlations between expression of the individual markers and our clinico-pathological data \[[@B18]\]. However, as has been the case previously \[[@B19]\], we noted significant reciprocity of p53 and Bcl-2 expression. Saleh et al, reported an association between Bcl-2 expression and lower p53 levels, a lower mean Ki-67 labelling index and favourable histopathological parameters \[[@B20]\]. Similarly, Popescu et al, demonstrated reciprocal expression of Bcl-2 and p53 mRNA in samples from 19 colorectal cancer metastases \[[@B21]\], however, other studies have failed to confirm this finding \[[@B22]\]. Although not statistically significant, we also noted a trend towards a higher prevalence of the good prognosis p53(-)/Bcl-2(+) subgroup in earlier stage tumours, in agreement with the findings of Gouissa et al, who demonstrated a trend for the p53(-)/Bcl-2(+) phenotype to be associated with lymph node negative tumours in a study of 108 colorectal adenocarcinomas \[[@B22]\]. Two recent papers have systematically reviewed the evidence for a prognostic role of p53 expression in colorectal cancer and illustrate some of the difficulties in interpreting the available literature. In the first, Anwar et al, described 35 studies (24 using immunohistochemical methods) in which p53 was found to be predictive of a worse outcome and 24 papers (16 employing IHC) found no association. In total only 4 of these studies included a comparable number of patients to ours, with 39 studies reporting on \<150 cases \[[@B23]\]. In the second review Munro et al, included 168 articles, reporting survival outcomes on a pool of over 18 000 patients. In this large review factors such as sources of heterogeneity between studies, the effect of publication bias and various assumptions concerning the techniques used for assessing p53 abnormalities led them to limit their conclusion to the fact that in patients with a better underlying prognosis, abnormal p53 has an adverse effect on outcome \[[@B15]\]. It appears logical that combined analysis of p53 with Bcl-2 should prove greater than the sum of its parts, as the p53 protein is known to regulate apoptosis via the Bcl-2/BAX pathway. Native p53 inhibits Bcl-2 gene expression by transcriptional activation of the pro-apoptotic gene BAX \[[@B24]\]. Bcl-2 itself is a proto-oncogene, whose expression varies in differing tissues. In the normal gastrointestinal tract Bcl-2 expression has been detected in the basal cells of crypts. Bcl-2 expression has also been described in the majority (up to 85%) of colorectal adenomas, with reduced expression on progression to invasive carcinoma \[[@B25],[@B26]\]. Despite the defined role of Bcl-2 protein in suppression of apoptosis, in colorectal cancer the evidence suggests that Bcl-2 expression is correlated with favourable parameters \[[@B27]\] and a better prognosis \[[@B28]\]. This may be explained by the finding that although Bcl-2 inhibits apoptosis it may also slow cell growth \[[@B29]\]. It has been reported that Bcl-2 contains an anti-proliferative domain, distinct from the domains required for its anti-apoptotic activity \[[@B30]\], and also that Bcl-2 can be converted to a BAX-like death effector by cleavage of a regulatory loop domain by caspases \[[@B31]\]. Whatever the underlying biological mechanism, in this study we have found that the combination of positive cytoplasmic Bcl-2 expression and negative nuclear p53 expression in colorectal cancer defines a population of patients with a good prognosis, indicating a clinically more indolent phenotype and a subset of patients for whom less aggressive adjuvant treatment is indicated. Conclusion ========== The use of molecular prognostic markers which complement traditional clinico-pathological staging information is not yet widespread in colorectal cancer, mainly due to the lack of high quality prospective studies and concerns regarding reproducibility and generalisation from existing data \[[@B32]\]. In particular, quantitative scoring of immunostaining for practical/diagnostic purposes is a very difficult affair and liable to much subjectivity. This is probably the major reason why so few quantitative immunomarkers have made their way into surgical pathology. These issues need to be addressed if we are to improve on existing prognostic criteria. However, our results suggest that analysis of p53 and Bcl-2 expression in colorectal cancer patients may provide useful prognostic information. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= LGD, IS, JHSand IOEconceived and designed the study. JHSwas responsible for maintaining the clinical database. NFSW, ZMand DSwere responsible for constructing the tissue arrays, performing the immunohistochemistry and analyzing the results. All authors were responsible for interpreting the results and drafting the article. All authors read and approved the final manuscript. Acknowledgements ================ The authors thank Claire Paish and John Ronan for their technical advice. This work was supported in part by a grant from the Special Trustees of Nottingham Hospitals (grant STR/03/M).	PubMed Central	2024-06-05T03:55:59.847486	2005-7-19	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181828/", "journal": "World J Surg Oncol. 2005 Jul 19; 3:47", "authors": [ { "first": "Nicholas FS", "last": "Watson" }, { "first": "Zahra", "last": "Madjd" }, { "first": "Duncan", "last": "Scrimegour" }, { "first": "Ian", "last": "Spendlove" }, { "first": "Ian O", "last": "Ellis" }, { "first": "John H", "last": "Scholefield" }, { "first": "Lindy G", "last": "Durrant" } ] }
PMC1181868	The stigma attached to mental illness is the greatest obstacle to the improvement of the lives of people with mental illness and their families \[[@pmed-0020136-b1]\]. Such stigma results in (1) a lower priority for mental health services, (2) difficulty getting staff of good quality to work in these services, (3) continuing problems in finding employment and housing for people who have had an episode of mental disorder, (4) the social isolation of people who suffer from mental illness and their families, and (5) poorer quality of care for physical illnesses occurring in people diagnosed as having had psychiatric illnesses \[[@pmed-0020136-b1]\]. These effects of stigma are true for all mental disorders, and in particular, for schizophrenia. The history of the stigmatisation of mental illness is long, but it is probable that intolerance to mental abnormality (and the rejection of people who had it) has become stronger in the past two centuries because of urbanisation and the growing demands for skills and qualifications in almost all sectors of employment. This, however, is only part of the story: mental illness is also linked to stigmatisation, discrimination, and intolerance in rural settings and in all countries, regardless of their level of industrialisation and sophistication of labour. Recent studies carried out in developing countries confirm that this stigma is universal \[[@pmed-0020136-b1]\]---indeed it is fair to say that stigma is attached to mental illness in different socio-cultural settings throughout the world, and that it is growing in strength and in its negative consequences. Programmes against Stigma {#s2} ========================= A number of programmes to diminish the stigma related to mental illness and its consequences have been started in recent years. Among those well known are campaigns undertaken in Australia, the United Kingdom, and Sweden---for example, "Changing Minds," an anti-stigma campaign, was launched in 1998 by the United Kingdom Royal College of Psychiatrists ([www.rcpsych.ac.uk/campaigns/cminds/index.htm](www.rcpsych.ac.uk/campaigns/cminds/index.htm)). A major international effort is The Global Programme against Stigma and Discrimination Because of Schizophrenia, launched by the World Psychiatric Association (WPA) in 1996 \[[@pmed-0020136-b2]\]. The WPA programme, known as "Open the Doors" (see [www.openthedoors.com](www.openthedoors.com) and [Figure 1](#pmed-0020136-g001){ref-type="fig"}), has five important characteristics that distinguish it from other previously developed programmes. First, it is an international and collaborative programme. Second, it is conceived as a long-term programme rather than as a campaign. Third, it involves family and patient organisations as well as governments, community agents, and health services at all stages of the programme, from its planning to its evaluation. Fourth, it emphasises the need for sharing experience and information obtained in the course of the programmes among all concerned, within and between countries. Finally, and perhaps most importantly, the programme\'s targets are selected on the basis of a process of consultation with people who have schizophrenia and their families rather than on the basis of theoretical constructs. This means that the targets of the programmes in different countries (and even in different regions of the same country) vary. It also means that the forces uniting the programme are shared convictions about the principal and overall goals of the programme rather than an imposed and artificial uniformity of specific short-term objectives. ::: {#pmed-0020136-g001 .fig} Figure 1 ::: {.caption} ###### The Logo of Open the Doors, in Nine Different Languages ::: ![](pmed.0020136.g001) ::: The WPA programme has already involved some 18 countries as follows: Austria, Brazil, Canada, Chile, Egypt, Germany, Greece, India, Italy, Japan, Morocco, Poland, Romania, Slovakia, Spain, Turkey, the UK, and the United States. It is likely that other countries, for example Zambia and the Czech Republic as well as others, will join it in the years to come. What Has the WPA Programme Achieved? {#s3} ==================================== To reach the main goal of the programme---the reduction or elimination of stigma and its consequences---the participating sites have undertaken studies aimed at a better understanding of the causes of stigma, its mechanism of action, and the factors that increase or lessen it \[[@pmed-0020136-b3],[@pmed-0020136-b4]\]. Participating sites have also implemented an array of measures ([Box 1](#box1){ref-type="boxed-text"}) that were selected taking into account the country and its specificities. Box 1. Measures Implemented by Participating Sites -------------------------------------------------- provision of information about schizophrenia, through the programme\'s websites, meetings, books, articles in health journals, newspapers, lectures, and congressesintroduction of specific legislation or rules for the behaviour of selected target groups (e.g., health staff in emergency services)organisation of artistic and cultural activitiessupport for the development of (physical) health services for people with mental disordersintroduction of anti-stigma activities into the training of different types of professionals (e.g., psychiatrists, police officers, teachers) Principles Guiding the Interventions {#s4} ==================================== Work with patients and relatives {#s4a} -------------------------------- A central priority must be to boost patients\' (and families\') self-esteem and self-respect. This facilitates patients\' socialisation, their active participation in the treatment and rehabilitation process, and their motivation for better personal care. Work with members of the health professions {#s4b} ------------------------------------------- Emphasis has been placed on the fact that health workers can do a great deal (as individuals) to prevent or diminish stigmatisation by: (1) helping their patients maintain self-esteem, (2) developing and implementing the plan of treatment together, (3) being constantly aware of the danger of labelling, which might harm their patients, (4) ensuring that they have respected their patients\' priorities rather then placing these priorities below those of the health care system, (5) working with families (learning from their experience and providing them with practical and useful information), and (6) in society, acting as advocates and models of tolerance and acceptance of people with mental illness \[[@pmed-0020136-b5]\]. Work with health authorities {#s4c} ---------------------------- Emphasis has been placed on the need to re-examine and improve legislation and procedures governing the health system to avoid its stigmatising potential. Work with journalists and other media professionals {#s4d} --------------------------------------------------- Journalists have been engaged in fighting stigma through better reporting about mental illness and about people with mental illness. In Ireland and the Philippines, for example, journalists have been led by the anti-stigma programme to the formulation of a voluntary code of non-stigmatising reporting. Work with the general public {#s4e} ---------------------------- The focus here has been on a change in behaviour rather than only a change of attitudes. The sites participating in the programme are learning from each other through contacts, visits, and the exchange of information. In addition, the programme has identified certain strategic directives that have been incorporated into the rules for new programmes. On the basis of experience, the key documents of the programme---a step-by-step guide on programme development and the manual attached to it---are constantly updated and improved. Many of these documents are available on the programme\'s website ([www.openthedoors.com](www.openthedoors.com)). Measuring Success, Overcoming Obstacles {#s5} ======================================= The success of the WPA programme is evaluated at country level and with direct reference to the targets identified by patients and their families as being particularly important for them. This is being done by focus group explorations of the experiences patients and families have had since the programme in their country began. Surveys of attitudes before the programme and during its course have been done in some countries (e.g., Canada) \[[@pmed-0020136-b6]\]. In addition, there are general indications of the success of the programme, including the increasingly wide use of programme materials, continuing collaboration among the sites, the number of publications and requests for presentation of the programme by both professional and nonprofessional organisations. A detailed description of these matters and references to publications from the participating sites are being published \[[@pmed-0020136-b3]\]. The main obstacle to success is the fact that changes in attitudes and behaviour take time. Continuous repetition of action and financial support have to be maintained over years---despite the fact that, in the beginning, anti-stigma programmes often produce only meagre results. Maintaining the motivation of all concerned over many years is very difficult. The programme also needs the lasting involvement of all structures of the health system (and of other social services), which must see the fight against stigma as one of their permanent and essential tasks. Conclusion {#s6} ========== In the descriptions of work done in the sites participating in the WPA Global Programme against Stigma and Discrimination Because of Schizophrenia, there are many examples of actions that have contributed to the lessening of stigma or to the prevention of its consequences. These examples underline the three basic principles that should be kept in mind when fighting stigma. First, the fight against stigma is a priority because stigmatisation is a major obstacle to any progress in the field of mental health. Second, programmes against stigma and discrimination should select their targets and evaluate their success with the active and concrete involvement of people with mental illness and their families. Finally, each of us, whether part of a major programme or alone, can do something to diminish or avoid stigmatisation by mental illness. It is just as important to ask what we can do ourselves to diminish stigmatisation as it is to urge others to do something about it. Citation: Kadri N, Sartorius N (2005) The global fight against the stigma of schizophrenia. PLoS Med 2(7): e136. WPA : World Psychiatric Association [^1]: Nadia Kadri is currently President of the Moroccan Society of Psychiatry and is the Moroccan local coordinator of the WPA Global Programme against Stigma and Discrimination Because of Schizophrenia. Norman Sartorius, previously Director of the Division of Mental Health of the World Health Organization and President of the WPA, is currently Scientific Director of the WPA Global Programme against Stigma and Discrimination Because of Schizophrenia. [^2]: Competing Interests: The authors declare that they have no competing interests.	PubMed Central	2024-06-05T03:55:59.850007	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181868/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e136", "authors": [ { "first": "Nadia", "last": "Kadri" }, { "first": "Norman", "last": "Sartorius" } ] }
PMC1181869	Although the 20th century saw a major expansion of the world economy, impressive military/security advances, and spectacular progress in science and technology, the grim reality in the first decade of the new millennium is that human life, health, and security remain under severe threat---but now from the adverse effects of inexorably widening disparities in wealth, health, and knowledge within and between nations. The gap between the income of the richest and poorest 20% of people in the world increased from a 9-fold difference at the beginning of the 20th century to 30-fold by 1960---and since then to over 80-fold by 2000 ([Figure 1](#pmed-0020143-g001){ref-type="fig"}). Although life expectancy has improved dramatically worldwide during this century, this trend has been reversed in the poorest countries in recent years \[[@pmed-0020143-b1]\]. The challenge of achieving improved health for a greater proportion of the world\'s population is one of the most pressing problems of our time and is starkly illustrated by the threat of infectious diseases. ::: {#pmed-0020143-g001 .fig} Figure 1 ::: {.caption} ###### Namibia\'s Worst Slum---Katutura in Windhoek The gap between the income of the richest and poorest 20% of people in the world increased from a 9-fold difference at the beginning of the 20th century to over 80-fold by 2000. (Photo: Jacob Holdt, [www.american-pictures.com](www.american-pictures.com)) ::: ![](pmed.0020143.g001) ::: Values {#s2} ====== The underlying basis for new threats to health, life, and security is our failure to adequately pursue the values that play an essential role in improving population health locally and globally. Such values include meaningful respect for human life, human rights, equity, freedom, democracy, environmental sustainability, and solidarity. Foremost among these is solidarity---without it, we ignore distant indignities, violations of human rights, inequities, deprivation of freedom, undemocratic regimes, and damage to the environment. We argue that this set of values---which combines genuine respect for the dignity of all people with a desire to promote the idea of human development beyond that conceived within the narrow, individualistic "economic" model of human flourishing---could serve to promote peaceful and beneficial use of new knowledge and power \[[@pmed-0020143-b2]\]. Good health and satisfying lives are determined both by "freedom from" want (basic subsistence and educational needs) and by "freedom to" undertake activities of one\'s choice to achieve personal goals \[[@pmed-0020143-b3]\]. Because freedom from want is dependent, at least to some extent, on the actions of others \[[@pmed-0020143-b4]\], achieving the goal of greater equity means that we must address the tension between individual freedom and solidarity. Liberty with responsibility exclusively to the self contradicts a view of social democracy that emphasizes that individuals arise from and are shaped by their societies, that their freedom to choose is embedded in social attachments, and that their social and economic rights must acknowledge solidarity as a balance between rights and responsibilities to themselves and others \[[@pmed-0020143-b5]\]. Solidarity is not discovered by reflection and reasoning, but rather by increasing our sensitivity (empathy) and adequate responses to the pain, suffering, and humiliation of others \[[@pmed-0020143-b6]\]. Extending the Bioethics Discourse in a Globalized World {#s3} ======================================================= Until the 1960s, discussions about ethics were largely confined to philosophical and theological studies. Advances in technology and medicine, together with increased concern for individual rights and freedoms, led to a new bioethics in which theologians, philosophers, lawyers, and other scholars engaged in a public discourse on applied ethics. Initially, this favored biomedical issues at the level of individual health---for example death and dying, reproductive medicine, and research ethics. However, since the birth of modern bioethics in the 1960s, the world has changed profoundly. Widening economic disparities, rapid population growth, the emergence of new infectious diseases, escalating ecological degradation, numerous local and regional wars, a stockpile of nuclear weapons, massive dislocations of people, advances in science and technology with profound implications for individuals and populations, and, most recently, new terrorist threats to life have demonstrated how interconnected we all are \[[@pmed-0020143-b7]\]. Growing global instability and threats from the widening gulf between the world\'s haves and have-nots call for new ways of thinking and acting. Distinctions between domestic and foreign policy have become blurred, and public health, even in the most privileged nations, is more closely linked than ever to health and disease in impoverished countries. The need for coherence between domestic and foreign policy was acknowledged by President Clinton when he declared HIV/AIDS a global emergency, and also by subsequent endeavors to foster a global response to this pandemic. Now, more than ever before, local action must be linked to a new global health ethics based on shared values to help make the world a more stable place. Such a new approach could facilitate transformation of current ideas about governance, the global political economy, and relations between countries. A framework that combines an understanding of global interdependence with enlightened long-term self-interest has the potential to produce a broad spectrum of beneficial outcomes, especially in the area of global health. An extended public debate, promoted by building capacity for this process through a multidisciplinary approach to ethics in education and daily life, could be the driving force for such change. These changes require that interest in health and ethics be extended beyond the microlevel of interpersonal relationships and individual health to include ethical considerations regarding public/population health at the level of institutions, nations, and international relations. Extending the discourse in this way could promote the new mindset needed to improve health and deal with threats to health on a global level. That mindset requires recognition that health, human rights, economic opportunities, good governance, peace, and development are all intimately linked within a complex, interdependent world. The challenge of the 21st century is to explore these links, to understand their implications, and to develop processes that can harness economic growth to human development, narrow global disparities in health, and promote peaceful coexistence. Five Transformational Approaches {#s4} ================================ A global agenda must thus extend beyond the rhetoric of universal human rights to include greater attention to duties, social justice, and interdependence. Health and ethics provide a framework within which such an agenda could be developed and promoted across borders and cultures. The relatively new interdisciplinary field of bioethics, when expanded in scope to embrace widely shared foundational values, could make a valuable contribution to the improvement of global health by providing the space for such a discussion to occur. Our vision, explicated in detail elsewhere, offers a way forward for global health reform through five transformational approaches \[[@pmed-0020143-b2]\]. Developing a global state of mind {#s4a} --------------------------------- First, developing a global state of mind about the world and our place in it is perhaps the most crucial element in the development of an ethic for global health. Achieving this will require an understanding of the world as an unstable complex system \[[@pmed-0020143-b8],[@pmed-0020143-b9]\], the balancing of individual goods and social goods, and the avoidance of harm to weak/poor nations through economic and other forms of exploitation that frustrate the achievement of human rights and well-being \[[@pmed-0020143-b10]\]. The emergence of a multi-faceted social movement, "globalization from below" (in which people at the grassroots around the world link up to impose their own needs and interests on the process of globalization), illustrates additional pathways to constructive change \[[@pmed-0020143-b11]\]. Promoting long-term self-interest {#s4b} --------------------------------- Second, in arguing that it is both desirable and necessary to develop a global mindset in health ethics, we suggest that this change need not be based merely on altruism, and that promoting long-term self-interest is also essential if we acknowledge that lives across the world are inextricably interlinked by forces that powerfully shape health and well-being. As an example, consider the long-term self-interest and mutual interdependence in the face of emerging new infectious diseases and microbial antibiotic resistance \[[@pmed-0020143-b12]\]. Striking a balance between optimism and pessimism {#s4c} ------------------------------------------------- Third, striking a balance between optimism and pessimism about globalization, solidarity, and progress will require a platform for dialogue among stakeholders, and a space where people can share different views about globalization. A broader conception of bioethics offers a basis for such a space. Developing capacity {#s4d} ------------------- Fourth, our vision for promoting an ethic for global health also features the development of capacity and a commitment to a broader discourse on ethics propagated through centers regionally and globally networked in growing and supportive North--South partnerships \[[@pmed-0020143-b13]\]. Achieving widespread access to public goods {#s4e} ------------------------------------------- Fifth, achieving widespread access to education, basic subsistence needs, and work requires collective action, including financing (to make sure they are produced), and good governance (to ensure their optimum distribution and use) \[[@pmed-0020143-b14]\]. Constructing new ways of achieving economic redistribution is the key to resolving many global problems. Conclusion {#s5} ========== While it would seem that nothing has changed since Lester Pearson noted over 30 years ago that "there can be no peace, no security, nothing but ultimate disaster, when a few rich countries with a small minority of the world\'s people alone have access to the brave, and frightening, new world of technology, science, and of high material living standards, while the large majority live in deprivation and want, cut off from opportunities of full economic development; but with expectations and aspirations aroused far beyond the hope of realizing them" \[[@pmed-0020143-b15]\], there is now perhaps a faint glimmer of hope that such progress is possible \[[@pmed-0020143-b16],[@pmed-0020143-b17]\]. SRB is supported by the University of Cape Town and the University of Toronto. ASD is supported by the McLaughlin Centre for Molecular Medicine. PAS is supported by a Distinguished Investigator award from the Canadian Institutes of Health Research (Ottawa, Canada). SRB and PAS are supported by grants from the Fogarty International Center of the United States National Institutes of Health (Bethesda, Maryland). ASD and PAS are supported by the Canadian Program on Genomics and Global Health (Toronto, Canada), whose funding sources are listed at [www.geneticsethics.net](www.geneticsethics.net). This article is abridged and modified from: Benatar SR, Daar AS, Singer PA (2003) Global health ethics: The rationale for mutual caring. Int Aff 79: 101--138. Citation: Benatar SR, Daar AS, Singer PA (2005) Global health challenges: The need for an expanded discourse on bioethics. PLoS Med 2(7): e143. [^1]: Solomon R. Benatar is Professor of Medicine and Director of the Bioethics Centre, University of Cape Town (Cape Town, South Africa) and Visiting Professor in Medicine and Public Health Sciences at the University of Toronto (Toronto, Canada). Abdallah S. Daar is Professor of Public Health Sciences and Surgery and Director of Ethics and Policy at the McLaughlin Centre for Molecular Medicine, University of Toronto. Peter A. Singer is Professor of Medicine at University Health Network and University of Toronto, and Sun Life Financial Chair and Director of the University of Toronto Joint Centre for Bioethics. [^2]: Competing Interests: The authors declare that they have no competing interests.	PubMed Central	2024-06-05T03:55:59.851354	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181869/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e143", "authors": [ { "first": "Solomon R", "last": "Benatar" }, { "first": "Abdallah S", "last": "Daar" }, { "first": "Peter A", "last": "Singer" } ] }
PMC1181870	Introduction {#s1} ============ Twins have long provided a unique opportunity to study how health is shaped from conception to death by biological and social factors \[[@pmed-0020162-b01]--[@pmed-0020162-b05]\]. At issue are the contributions, singly and combined, of genetic inheritance, in utero postzygotic events before and after twinning, and familial plus societal contexts, including the ways in which twins are treated by family members, each other, and society at large \[[@pmed-0020162-b01]--[@pmed-0020162-b07]\]. To explore these issues, one major trend in twin research has focused on comparing health status of twins raised separately since birth or early childhood \[[@pmed-0020162-b01]--[@pmed-0020162-b07]\]. Far fewer studies have investigated how twins raised together, but who differ in their postadolescent socioeconomic position, compare on adult health status \[[@pmed-0020162-b02],[@pmed-0020162-b04]\]. Yet such research could potentially inform current debates over the contribution of lifecourse socioeconomic conditions to adult health \[[@pmed-0020162-b08]--[@pmed-0020162-b11]\], given twins\' shared genetic inheritance and early life socioeconomic plus biological exposures. Twins afford an important opportunity to examine the additional impact of adult experiences on adult health in a population matched on early life experiences. In particular, one important unresolved issue in the burgeoning literature on lifecourse analysis of health concerns how well early life social circumstances are measured, since this these data are essential for distinguishing between the influence of early life and adult conditions on adult health status \[[@pmed-0020162-b08],[@pmed-0020162-b09]\]. At issue are the often limited data on childhood socioeconomic conditions \[[@pmed-0020162-b08],[@pmed-0020162-b09]\], plus the possibility of systematic error, by adult socioeconomic position, when adults recollect their early childhood circumstances \[[@pmed-0020162-b12]--[@pmed-0020162-b14]\]. Limited data, recall bias, and poor measurement together hinder obtaining accurate effect estimates, due to confounding by unmeasured factors. This concern is especially salient for studies investigating the social patterning of health, precisely because living and working conditions influence health through myriad discrete yet entangled pathways \[[@pmed-0020162-b08],[@pmed-0020162-b15]\]. Further complicating analysis of the impact of childhood and adult socioeconomic position on health are choices regarding the socioeconomic measure(s) employed \[[@pmed-0020162-b16]--[@pmed-0020162-b18]\]. As discussed in several comprehensive review articles \[[@pmed-0020162-b16]--[@pmed-0020162-b18]\], considerable evidence exists demonstrating that different socioeconomic measures---e.g., education, occupation, income, wealth, housing tenure, etc.---are not simply "exchangeable" with each other and instead often yield different estimates of the magnitude of the socioeconomic gradient and affect health by independent as well as correlated pathways. For example, while education has often been valued as a socioeconomic measure precisely because, once achieved, it is not subject to reverse causation (e.g., poorer health leading to lower income), it also has been shown to be insensitive to subsequent changes in adult socioeconomic position (e.g., income dynamics) that also can affect adult health status \[[@pmed-0020162-b16]--[@pmed-0020162-b22]\]. An important implication is that studies concerned with the joint impact of childhood and adult socioeconomic position on health must take into account how their choice of socioeconomic measures may influence their results. Of note, studies of adult twins, and especially monozygotic twins, can usefully address problems of capturing early life circumstances and assessing the contribution of childhood and adult conditions on adult health. This is because monozygotic twins raised together through childhood (a) are tightly matched on both genetic endowment and the socioeconomic circumstances characterizing their gestation and early life and childhood household resources \[[@pmed-0020162-b04]--[@pmed-0020162-b06]\], and (b) are the same biological sex, so they are likely to be accorded the same gender expectations and not have differential treatment or access to household resources because of their gender \[[@pmed-0020162-b06],[@pmed-0020162-b15]\]. Thus, even without any measurement of childhood conditions, a comparison of adult monozygotic twins who are concordant versus those who are discordant on adult socioeconomic position allows ascertainment of the extent to which socioeconomic position after childhood affects adult health, above and beyond the impact of childhood socioeconomic position. Twin analyses employing diverse socioeconomic measures capturing circumstances earlier versus later in adult life could also potentially yield insight into the impact of cumulative socioeconomic position on health, with interpretation of results for monozygotic twins, including patterns of within-pair variability, aided by comparison to results to same-sex dizygotic twins. Thus, our objective, framed by ecosocial theory and its concern with the lifelong embodiment of social conditions \[[@pmed-0020162-b15],[@pmed-0020162-b23]\], was to compare health status among a cohort of monozygotic and dizygotic women twins with shared upbringing (at least until age 14) and concordant versus discordant adult socioeconomic position. Outcomes included biological markers, anthropometric and health outcomes, and health behaviors. We employed data on both adult occupational class and educational level, hypothesizing that the former might capture relevant aspects of socioeconomic position occurring after completion of educational attainment. Methods {#s2} ======= Study Population {#s2a} ---------------- The study twin pairs were members of the Kaiser Permanente Women Twins Study Examination II, conducted in 1989--1990 in Oakland, California, United States \[[@pmed-0020162-b24]\]. The Examination I cohort included 434 twin pairs recruited in 1978--1979 from a twin registry established in 1974 at the Northern California Kaiser Permanente Medical Care Program. All participants resided in the San Francisco Bay Area at the time of Examination I and were born in or prior to 1960 (mean age, 41 y; range, 18--85 y). Zygosity for each pair was determined by analysis of 20 polymorphic loci, such that the probability of misclassification as monozygous was less than 0.001 \[[@pmed-0020162-b24]\]. For Examination II (1989--1990), original cohort members were sent a self-administered questionnaire on their health and sociodemographic characteristics plus an invitation to return for a physical exam \[[@pmed-0020162-b24]\]. Cohort retention was high: Only 72 women (8.3%) did not respond, of whom 36 were deceased. Among the 796 respondents, only 87 (10.9%) did not return for a physical exam. After additionally excluding five women whose twin was a nonrespondent, the Examination II cohort included 352 twin pairs (58% monozygotic and 42% dizygotic), representing 81.1% of the original cohort. Enrollment and study of the twins in both cohorts was approved by the Kaiser Permanente Medical Care Program, Northern California Region, Institutional Review board; analyses for this investigation were additionally approved by the Harvard School of Public Health Human Subjects Committee. Socioeconomic Data {#s2b} ------------------ Childhood and adult socioeconomic position were measured at Examination II using a self-administered questionnaire. We employed a modified version of Erik Olin Wright\'s occupational class schema \[[@pmed-0020162-b12],[@pmed-0020162-b16],[@pmed-0020162-b25]--[@pmed-0020162-b27]\], analogous to the United Kingdom\'s newly established National Statistics Socioeconomic Classification system (NS-SEC) \[[@pmed-0020162-b28]\]. Distinctions, in order of dominance, were between persons classified as "nonworking class" (NWC; own a business and employ others, self-employed, or supervisory employees), "working class" (WC; nonsupervisory employees), or not in the paid labor force \[[@pmed-0020162-b12],[@pmed-0020162-b16],[@pmed-0020162-b25]--[@pmed-0020162-b27]\]. We defined "childhood household social class" as the occupational class position of the person identified as the head-of-household when the respondent was age 14; we also ascertained the proportion of twins who lived together until at least age 14. We measured adult household social class using a validated, gender-appropriate approach, equal to the most dominant class position, taking into account the usual individual class position of the respondent and her partner or other head-of-household, if any \[[@pmed-0020162-b16],[@pmed-0020162-b25],[@pmed-0020162-b26]\]. Using a gender-neutral approach to measuring household class is increasingly recognized as a more valid means of assessing household socioeconomic position than one which automatically assigns it to either (a) the respondent, whether a woman or man, or (b) the occupation of the adult man in the household (if one is present), given the rise of dual wage-earner households \[[@pmed-0020162-b16]--[@pmed-0020162-b18],[@pmed-0020162-b27]--[@pmed-0020162-b32]\]. For example, the new UK NS-SEC measure explicitly rejects the prior conventional practice of "males taking precedence over females" when selecting the "household reference person" for assignment of household class, and instead chooses based upon "the person responsible for owning or renting or who is otherwise responsible for the accommodation," regardless of gender \[[@pmed-0020162-b29]\]. We also obtained data on the educational level attained by each twin and their father. No data were available, however, on childhood or adult household income, wealth, or debt, or on the educational level of the mother. Because the twin pairs were matched, by definition, on socioeconomic position in utero through age 14, we categorized twin pairs in relation to adult socioeconomic position. For adult household social class, three types of pairs were possible: two concordant (both WC:WC or NWC:NWC) and one discordant (WC:NWC). For education, the pairs were defined in relation to being concordant or discordant for fewer than 4 y versus 4 y or more of college. Health Outcome Data {#s2c} ------------------- The selected health outcomes were chosen because of their well-documented associations with socioeconomic position and because risk could plausibly be affected by both early life and adult circumstances \[[@pmed-0020162-b08],[@pmed-0020162-b09],[@pmed-0020162-b25]\]. Self-report data were analyzed for self-rated health (dichotomized as excellent/good versus fair/poor) and medication use. A validated interviewer-administered questionnaire was used to obtain data on physical activity (kcal per kg per y); this instrument assessed the typical amount of time spent in activities of varying intensity at home, at work, and during recreation \[[@pmed-0020162-b33]\]. Data on anthropometric and biological characteristics were obtained by physical examination and laboratory analysis \[[@pmed-0020162-b24]\]. Height was recorded to the nearest 0.5 cm, weight was measured to the nearest 0.1 kg, and these data were used to calculate body mass index (BMI, in kg/m^2^). Participants\' minimum waist girth was measured using a steel tape at the natural indentation or at a level midway between the iliac crests and the lower edge of the rib cage if no natural indentation was present; hip girth was measured at the level of the greatest protrusion of the buttocks. These measurements were recorded to the nearest 0.5 cm, and the averages of two measures (different by no more than 1 cm) were used to calculate the waist-to-hip ratio (WHR). After participants rested for 5 min, a mercury sphygmomanometer was used to take two measures each of systolic and diastolic blood pressure (seated, right arm); averages of these two measures were used for data analysis. High blood pressure was defined as systolic blood pressure 140 mm Hg or higher, or diastolic blood pressure 90 mm Hg or higher, or taking antihypertensive medication. Blood for lipid and lipoprotein measurement was obtained after participants had fasted overnight. It was collected into tubes containing ethylenediaminetetraacetic acid (EDTA). Total, high-density lipoprotein (HDL), and low-density lipoprotein (LDL) cholesterol were measured by standard methods \[[@pmed-0020162-b24]\]. Glucose level (mg/dl \[amount × 0.0555 = mmol/l\]; measured 2 h post-load) was determined using the glucose oxidase method \[[@pmed-0020162-b24]\]; analyses using data on glucose levels excluded the five twin pairs for whom one or both twins had values 300 mg/dl or higher. Data Analysis {#s2d} ------------- First, to establish the analytic cohort, we identified twin pairs for whom we could determine both that they had lived together until at least age 14 and their joint socioeconomic trajectory (n = 308 pairs). This analytic cohort excluded 44 twin pairs (21 missing data on duration lived together; one reporting separation prior to age 14; two where one twin said below age 14 and the other age 14 or above; plus 20 pairs for whom the joint data on adult household class was either missing, inconsistent, or not in the labor force). Second, we ascertained the retained twins\' sociodemographic and health characteristics. Third, for continuous outcomes, we calculated (a) the mean matched difference for twin pairs discordant on adult socioeconomic position, setting the twin with the most socioeconomic resources as the baseline, so as to determine both the magnitude and direction of differences in the outcomes among these pairs, and (b) the mean matched absolute difference for twin pairs in each socioeconomic stratum, to ascertain the variability of outcomes among the twin pairs both discordant and concordant on adult socioeconomic position. Additionally, for the categorical outcomes, we calculated the kappa statistic and associated 95% CI \[[@pmed-0020162-b34]\]. We do not report data on the 18 WC:WC twin pairs, because small numbers rendered the parameter estimates uninterpretable. All analyses were done in SAS \[[@pmed-0020162-b35]\]. Results {#s3} ======= As shown in [Tables 1](#pmed-0020162-t001){ref-type="table"} and [2](#pmed-0020162-t002){ref-type="table"}, the sociodemographic and health characteristics of the full cohort (n = 352 pairs) and the analytic cohort (n = 308 pairs) were quite similar, with about 40% having grown up in working class households and 80% in households in which the father had less than a 4-y college education. At Examination II, 32% of the twin pairs in the analytic cohort were discordant for adult household occupational class, and 20% were discordant for individual college attainment. ::: {#pmed-0020162-t001 .table-wrap} Table 1 ::: {.caption} ###### Sociodemographic Characteristics: Kaiser Permanente Women Twins Study, Oakland, California, United States, 1989-1990, Full Cohort (n = 352 Pairs) and Analytic Cohort (n = 308 Pairs) ::: ![](pmed.0020162.t001) ::: ::: {#pmed-0020162-t002 .table-wrap} Table 2 ::: {.caption} ###### Health Characteristics: Kaiser Permanente Women Twins Study, Oakland, California, United States, 1989-1990, Full Cohort (n = 352 Pairs) and Analytic Cohort (n = 308 Pairs) ::: ![](pmed.0020162.t002) ::: Results pertain to both the direction and magnitude of the difference in health status among the twin pairs, in relation to both different measures of adult socioeconomic position and zygosity. First, regarding the magnitude of health disparities among monozygotic twins discordant on adult socioeconomic position, for the analyses using data on adult occupational class ([Tables 3](#pmed-0020162-t003){ref-type="table"} and [4](#pmed-0020162-t004){ref-type="table"}), the WC twin had significantly higher systolic blood pressure (mean matched difference = 4.54 mm Hg; 95% CI, 0.10--8.97), diastolic blood pressure (mean matched difference = 3.80 mm Hg; 95% CI, 0.44--7.17), and LDL cholesterol than her NWC twin (mean matched difference = 7.82 mg/dl \[amount × 0.0259 = mmol/l\]; 95% CI, 1.07--14.57). Additionally, among the monozygotic twin pairs, a greater proportion of twin pairs discordant on occupational class were discordant for self-reported health compared to twin pairs concordant on occupational class (27.5% versus 6.9%; p = 0.0178). Poorer health was also more likely to be reported by the working class twin; among the 51 monozygotic pairs discordant on class, the proportion of pairs in which the WC twin reported fair or poor health while her NWC twin reported excellent or good health (17.6%) was almost twice that of the converse (9.3%, i.e., pairs in which the WC twin reported good or excellent health and the NWC twin reported fair or poor health). ::: {#pmed-0020162-t003 .table-wrap} Table 3 ::: {.caption} ###### Comparison of Health Outcomes for 290 Twin Pairs Concordant and Discordant on Adult Household Occupational Class (WC and NWC): Continuous Outcomes, by Zygosity: Kaiser Permanente Women Twins Study, Oakland, California, United States, 1989-1990 ::: ![](pmed.0020162.t003) ::: ::: {#pmed-0020162-t004 .table-wrap} Table 4 ::: {.caption} ###### Comparison of Health Outcomes for 290 Twin Pairs Concordant and Discordant on Adult Household Occupational Class (WC and NWC): Categorical Outcomes, by Zygosity: Kaiser Permanente Women Twins Study, Oakland, California, United States, 1989-1990 ::: ![](pmed.0020162.t004) ::: By contrast, corresponding analyses using data on educational level ([Tables 5](#pmed-0020162-t005){ref-type="table"} and [6](#pmed-0020162-t006){ref-type="table"}) revealed little difference in patterns of health among monozygotic twin pairs discordant on educational attainment. Dizgyotic twins discordant on adult socioeconomic position, whether categorized by occupational class or educational level, likewise did not notably differ on their adult health status ([Tables 3](#pmed-0020162-t003){ref-type="table"}--[6](#pmed-0020162-t006){ref-type="table"}). ::: {#pmed-0020162-t005 .table-wrap} Table 5 ::: {.caption} ###### Comparison of Health Outcomes for 290 Twin Pairs Concordant and Discordant on Educational Level (Less Than Versus at Least 4 y of College): Continuous Outcomes by Zygosity: Kaiser Permanente Women Twins Study, Oakland, California, United States, 1989-1990 ::: ![](pmed.0020162.t005) ::: ::: {#pmed-0020162-t006 .table-wrap} Table 6 ::: {.caption} ###### Comparison of Health Outcomes for 290 Twin Pairs Concordant and Discordant on Educational Level (Less Than Versus at Least 4 y of College): Categorical Outcomes by Zygosity: Kaiser Permanente Women Twins Study, Oakland, California, United States, 1989-1990 ::: ![](pmed.0020162.t006) ::: Second, regarding the variability in health outcomes among twin pairs in relation to their adult socioeconomic position, the mean matched absolute difference was similar among both monozygotic twins who were discordant and concordant on occupational class, and also was similar among dizygotic twins discordant and concordant on occupational class ([Tables 3](#pmed-0020162-t003){ref-type="table"} and [4](#pmed-0020162-t004){ref-type="table"}). Within occupational class strata, however, for all the continuous outcomes other than diastolic blood pressure, the magnitude of variability typically was greater for the dizygotic than the monozygotic twin pairs ([Tables 3](#pmed-0020162-t003){ref-type="table"} and [4](#pmed-0020162-t004){ref-type="table"}). Similar results were obtained for analyses based on educational level ([Tables 5](#pmed-0020162-t005){ref-type="table"} and [6](#pmed-0020162-t006){ref-type="table"}), with some important exceptions. Specifically, for several outcomes among the monozygotic twins, especially average systolic and diastolic blood pressure, post-load glucose, and physical exercise, variability was greatest among twin pairs in which both had fewer than 4 y of college, intermediate among discordant pairs, and least among those where both had 4 y of college or more. Discussion {#s4} ========== Our study provides novel evidence suggesting that correlations in health outcomes among adult women twin pairs who lived together through childhood vary by their subsequent socioeconomic position, with results sensitive to choice of socioeconomic measure. Although small numbers limit precision of estimates, cardiovascular factors differed more among twins who were discordant on adult occupation class than twin pairs concordant on being professionals, and, within twin pairs discordant on occupational class, the working-class twin typically fared worse than the professional twin. These patterns were much weaker or not evident for analyses using data on educational attainment. Together, these results, combined with our prior research showing that the twins who experienced cumulative deprivation had the worst health \[[@pmed-0020162-b26]\], lend additional support to the hypothesis that cumulative experiences across the lifecourse, including those after adolescence and after completion of educational attainment, and not just early life experiences, shape adult health \[[@pmed-0020162-b08],[@pmed-0020162-b09]\]. Additionally, the greater magnitude of variability in outcomes among dizygotic compared to monozygotic twins within the same socioeconomic strata is what would be expected, given the tighter matching on genetic endowment among the monozygotic twins \[[@pmed-0020162-b03]--[@pmed-0020162-b05]\]. However, the suggestive finding of greater magnitude of variability, within both the monozygotic and dizygotic twins, among pairs with the least education compared to the most education, especially for the cardiovascular-related results, has not to our knowledge previously been reported. Given that low educational attainment is highly correlated with low socioeconomic resources during childhood \[[@pmed-0020162-b16]--[@pmed-0020162-b18]\], our results lend tentative support to the hypothesis that increased variability of physiological traits such as blood pressure may be positively associated with greater early-life and cumulative exposure to economic deprivation \[[@pmed-0020162-b36]\]. A related body of research suggests that chronic exposure to social stressors associated with socioeconomic deprivation may result in repeated activation---and ultimately harmful dysregulation---of physiological systems that respond to stress, thereby increasing risk of elevated blood pressure, insulin resistance, and visceral fat deposition and thus risk of cardiovascular disease, obesity, and diabetes \[[@pmed-0020162-b37]--[@pmed-0020162-b39]\]. Study limitations include (a) the relatively small number of twin pairs (albeit similar to other twin studies \[[@pmed-0020162-b03],[@pmed-0020162-b11]\]); (b) lack of data on detailed occupational class position over time and on age at obtaining a college degree, plus prior or current data on income, poverty, wealth, and debt; (c) lack of data on gestational age, birth weight, birth order, and whether the twins had shared or separate chorions and amniotic sacs \[[@pmed-0020162-b02],[@pmed-0020162-b07],[@pmed-0020162-b11],[@pmed-0020162-b40]\]; (d) lack of data on differences in the twins\' childhood experiences and exposures (e.g., differential treatment accorded to first- versus second-born twins, and to monozygotic versus dizygotic twins \[[@pmed-0020162-b06]\]); and (e) lack of data on male twins; in addition, the small number of women twins who were concordant on adult working class position limits generalizability (but not internal validity) of results. Most studies assessing the impact of childhood socioeconomic position on health, however, have relied on occupational and sometimes educational data \[[@pmed-0020162-b08],[@pmed-0020162-b09],[@pmed-0020162-b26],[@pmed-0020162-b41]--[@pmed-0020162-b46]\], reflecting difficulties in obtaining income data across the lifecourse \[[@pmed-0020162-b16]--[@pmed-0020162-b18]\]. By contrast, strengths of our study include: (a) biological confirmation of zygosity; (b) identical gestational age; (c) identical biological sex, relevant to gender expectations and gendered exposures (more similar for same- versus opposite-sex twins \[[@pmed-0020162-b05],[@pmed-0020162-b06]\]); (d) data on age until which the twins lived together; (e) use of a validated and gender-appropriate household occupational class measure, plus data on education; and (f) measurement of anthropometric and physiologic characteristics, not just self-reported health. Moreover, by focusing on postadolescence divergence of socioeconomic position, the study avoided concerns affecting comparisons of twins raised separately versus together, e.g., difficulties in assessing similarities versus differences of the family of origin versus adoptive family \[[@pmed-0020162-b04]--[@pmed-0020162-b06]\]. A recent analysis of United Kingdom twins\' earnings in relation to educational level additionally underscores the utility of using twin analyses to gauge the impact of childhood and adult socioeconomic conditions, at the individual and the household level, on adult economic and health-related outcomes (e.g., smoking) \[[@pmed-0020162-b47]\]. Overall, results of this study are in accord with other research suggesting that cumulative exposures related to socioeconomic position, not only genetic inheritance and early life experiences, shape adult health \[8--11,26,41--46,48\]. As with our findings, these studies typically have documented the strongest joint impacts for outcomes pertaining to cardiovascular health \[8--11,26,41--46,48\]. Unlike prior research, however, the present study newly employed a same-gender twin design, affording comparatively tight matching on life circumstances through early adolescence, with monozygotic twins additionally matched on genetic inheritance, thereby circumventing important concerns raised about likely unmeasured confounders affecting results of prior studies dependent upon adult recall of---and limited data on---childhood socioeconomic position. Even so, generalizability of results to nontwins could be hampered if twins differ systematically from nontwins on factors influencing associations between socioeconomic position and adult health, as perhaps related to maternal and zygotic characteristics relevant to risk of monozygotic or dizygotic twinning or to exposures contingent upon being a twin in utero (e.g., down-regulation of growth) \[[@pmed-0020162-b02],[@pmed-0020162-b04]--[@pmed-0020162-b07],[@pmed-0020162-b49]--[@pmed-0020162-b52]\]. In summary, creative use of social and biological twin data concerning both social and biological aspects of twinship \[[@pmed-0020162-b01]--[@pmed-0020162-b03],[@pmed-0020162-b06]\] has the potential to inform current debates about the impact of lifecourse socioeconomic position on health. Suggesting such investigations could have public health import, prior research has estimated that a reduction of 2 mm Hg in the average diastolic blood pressure in the United States---i.e., about half the difference we observed in the comparison of working class to nonworking class monozygotic twins---would translate to a 17% decrease in hypertension, a 6% reduction in coronary heart disease, and a 15% reduction in risk of stroke and transient ischemic attacks \[[@pmed-0020162-b53]\]. Given the longstanding fascination with twins \[[@pmed-0020162-b01]--[@pmed-0020162-b03],[@pmed-0020162-b06]\], if additional and larger twin studies of economically diverse women and men twins confirmed the relevance of cumulative and intergenerational lifetime socioeconomic resources for health, and were also able to include a wider array of socioeconomic measures (e.g., income, wealth, debt, and mother\'s education) and data on gestational age and birth weight, the evidence would likely have high policy salience, plus importantly enhance understanding of how embodiment of societal conditions shapes population patterns of health, disease, and well-being \[[@pmed-0020162-b15],[@pmed-0020162-b23],[@pmed-0020162-b54]\]. Patient Summary {#sb1} --------------- ### Background {#sb1a} Important controversies exist about the extent to which people\'s health status as adults is shaped by their living conditions in early life compared to adulthood. These debates have important policy implications, with regard to directing resources for improving health: should they be focused on children, on adults, or both? One obstacle to determining the relative influence of early life compared to adulthood on health is a lack of sufficient high-quality data on childhood and adult socioeconomic position and adult health status. Twins research can be used to answer this question, because for twins raised together their social class early in life (here defined as before age 14) will be the same, and study of monozygotic (identical) twins further allows researchers to look at the impact of living conditions on people with the same genetic background. ### What Did the Researchers Do? {#sb1b} They looked at how much education each twin had and their social class in later life, and they analyzed these in relation to diverse health outcomes (blood pressure, cholesterol, body mass index) in 308 pairs of female twins recruited in California. ### What Did the Researchers Find? {#sb1c} They found that the monozygotic twins who differed later in life in their social class tended to have differences in health, with the working-class twin having higher blood pressure and higher cholesterol than her professional counterpart. By contrast, differences in education made no difference to these measures of health. ### What Do These Findings Mean? {#sb1d} It is already believed that social class in children may affect later health; these results suggest that even individuals who had the same social class in childhood may have different health because of adult social class, including their living conditions after completing their educations. The implication is that interventions to eliminate social inequalities in health must take into account adult as well as childhood living conditions. ### Where Can I Get More Information? {#sb1e} There are many twin sites on the Web. One site with many links, including to registries, is that of the International Society for Twin Studies. <http://www.ists.qimr.edu.au/links.html> The source of funding was the National Institutes of Health, National Heart, Lung, and Blood Institute (1 R29 HL51151--01). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Citation: Krieger N, Chen JT, Coull BA, Selby JV (2005) Lifetime socioeconomic position and twins\' health: An analysis of 308 pairs of United States women twins. PLoS Med 2(7): e162. BMI : body mass index CI : confidence interval NWC : nonworking class WC : working class WHR : waist-to-hip ratio [^1]: Competing Interests: The authors have declared that no competing interests exist. [^2]: Author Contributions: NK conceived and designed the study, interpreted the data, and prepared the manuscript. JTC helped devise the analytic plan, conducted the analyses, and assisted with interpreting the data and preparing the manuscript. BAC advised on the analytic plan, and assisted with analyzing and interpreting the data and with preparing the manuscript. JVS designed the Kaiser Permanente Women Twins Study II, and assisted with interpretation of the data and preparing the manuscript.	PubMed Central	2024-06-05T03:55:59.853093	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181870/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e162", "authors": [ { "first": "Nancy", "last": "Krieger" }, { "first": "Jarvis T", "last": "Chen" }, { "first": "Brent A", "last": "Coull" }, { "first": "Joe V", "last": "Selby" } ] }
PMC1181871	Introduction {#s1} ============ Hepatitis B virus (HBV) infection is a serious public health problem in many parts of Asia and Africa that results in complications such as cirrhosis and cancer of the liver \[[@pmed-0020163-b01]\]. Globally, despite effective vaccination against HBV and available antiviral treatments \[[@pmed-0020163-b02]--[@pmed-0020163-b07]\], there remain an estimated 350 million hepatitis B carriers with a lifetime risk for developing cirrhosis and hepatocellular carcinoma, as many individuals infected during infancy remain infectious carriers for decades. Although the association between HBV infection and the development of liver cancer has stirred profound interest in the development of therapeutics for both the prevention of HBV infection and the clearance of virus from chronic infection, HBV continues to be a pathogen of significant importance. Current antiviral therapies for HBV carriers include treatment such as alpha-interferon or lamivudine, but the long-term resolution of disease is disappointing due to low seroconversion rates and the development of drug-resistant viral mutants \[[@pmed-0020163-b07]\]. HBV belongs to a family of viruses known as Hepadnaviridae and encodes only four genes in a highly compact viral genome: the surface gene (S), the core gene (C), the X gene (X), and the polymerase gene (P). Viral replication has been shown to occur via an RNA intermediate in the cytoplasm, but, unlike retroviruses, integration of HBV DNA into the host genome is not required. Despite a wealth of information on the virus itself and the possible role that host factors may play in the viral infectious life cycle \[[@pmed-0020163-b08]--[@pmed-0020163-b11]\], the direct relationship between chronically infected patients and host--pathogen interactions is poorly understood. Host cellular factor heterogeneous nuclear ribonucleoproteins (hnRNPs) are known to be pre-mRNA binding proteins and shuttle intermediates between nucleus and cytoplasm. hnRNP K in particular has been implicated in diverse molecular and cellular functions, including nuclear--cytoplasmic shuttling \[[@pmed-0020163-b12]\], transcription, and translation \[[@pmed-0020163-b13]\]. Here, we aimed to study the importance of hnRNP K in the regulation of HBV viral replication. Both over-expression and knockdown assays were designed to enhance HBV viral load and to silence the endogenous hnRNP K, respectively, within the cell to verify the effects of hnRNP K on HBV viral replication. Methods {#s2} ======= Serological Assays {#s2a} ------------------ Serum samples were aliquoted and stored at −20 °C until they were defrosted for testing. We examined serum HBV DNA levels by the Hybrid Capture II HBV DNA Assay (Digene, Gaithersburg, Maryland, United States). PCR and Sequencing of the Enh II/Pre-Core Region {#s2b} ------------------------------------------------ DNA extracted from serum was amplified with primers 5′\--3′ and 5′- GCTTGGAGGCTTGAACAG-3′ (94 °C for 2 min, followed by 35 cycles of 94 °C for 15 s, 50 °C for 30 s, and 72 °C for 1 min, and lastly followed by 72 °C for 7 min) with Taq DNA polymerase (Roche, Basel, Switzerland). Nested primers 5′- GTCAACGACCGACCTTGAGG-3′ and 5′- ACCAATTTATGCCTACAGCCTC-3′ were used in a second round of PCR (94 °C for 2 min, followed by 35 cycles of 94 °C for 15 s, 55 °C for 30 s, and 72 °C for 1 min, and lastly followed by 72 °C for 7 min). The size of this nested PCR product was 116 base pairs (corresponding to nucleotides 1682--1798) and was resolved in 1.5--2% agarose gels. PCR products were cleaned up using the Qiaquick PCR Purification Kit (Qiagen, Valencia, California, United States) and sequenced directly to confirm the identity of the products using the ABI prism dRodamine Terminator Cycle Sequencing Ready Reaction Kit (ABI-Prism 310; Applied Biosystems, Foster City, California, United States). Results of the sequences were aligned and compared. Construction of Plasmids {#s2c} ------------------------ Plasmids pGL3-Control (a Luciferase plasmid with a simian virus 40 enhancer and promoter) and pGL3-Promoter (an enhancerless Luciferase plasmid with a simian virus 40 promoter upstream from the Luciferase gene) were obtained from Promega (Madison, Wisconsin, United States). Plasmid pGL3-Promo/A was constructed by amplifying the basic functional unit of Enh II by PCR using primers LucF (5′- GC CAACGACCGACCTTGAGG-3′) and LucR (5′- GC ACCAATTTATGCCTACAGCCTC-3′) comprising HBV nucleotide positions 1686--1801. The 131-base-pair PCR fragment was MluI/BglII-digested and ligated with MluI/BglII-digested pGL3-Promoter. The other mutant constructs were constructed using the Gene Editor Site-Directed in vitro Mutagenesis System (Promega) to introduce the HBV Enh II mutations at nucleotide position 1752. The first mutation was mutating nucleotide A to G (pGL3-Promo/G), the second was with nucleotide A to T (pGL3-Promo/T), and the third was with nucleotide A to C (pGL3-Promo/C). The sequences of the three mutant oligonucleotides were 5′- GGGGGAGGAG TTAGGTTAAA-3′, 5′- GGGGGAGGAG TTAGGTTAAA-3′, and 5′- GGGGGAGGAG TTAGGTTAAA-3′, respectively. Constructs were sequenced for verification. hnRNP K "variant 2" and "variant 3" clones were constructed by cloning a 1.4-kb RT-PCR fragment coding for the hnRNP K from total RNA extracted from HepG2 cells. The EcoRI- and XhoI-digested PCR fragments were cloned into EcoRI- and XhoI-digested pcDNA 3.1 separately. Cloning primers for "variant 2" were 5′- TAAAAGGAATTCAATATGCAAACTGAACAG-3′ and 5′- CTAGTCCTCGAGTTAGAAAAACTTTCCAGA-3′, and cloning primers for "variant 3" were 5′- TAAAAGGAATTCAATATGCAAACTGAACAG-3′ and 5′- CTTGCACTCGAGTTAGAATCCTTCAACATC-3′. The HBV 1752A full-length replicative clone (genotype A) was constructed by using a HBV genome-containing pBR325 plasmid ( ATCC, Manassas, Virginia, United States) as a template. Primers were designed to amplify two fragments: 1--1900 and 1600--3215. The region 1600--1900 contained the core promoter and the overlapping transcription termination region \[[@pmed-0020163-b14]\] ([Figure S1](#sg001){ref-type="supplementary-material"}). In-frame ligation of the two fragments using the internal EcoRI (1/3215) ensures continuous viral open reading frames were cloned into the NruI site in pcDNA 3.1, resulting in the replicative construct. Viral transcription was under its own promoter control as the viral insert was cloned at the NruI site, located upstream of the plasmid\'s CMV promoter \[[@pmed-0020163-b15]\]. The 1752ΔG, 1752ΔT, and 1752ΔC full-length replicative clones were constructed as described for the 1752A. The first fragment, 1600--3215, was generated as a PCR product from the HBV-pBR325 plasmid and cloned into pcDNA 3.1. The 1752G, 1752T, and 1752C mutations were each generated separately in the first fragment by the Quick-Change site-directed mutagenesis kit (Stratagene, La Jolla, California, United States). Sequencing was done for verification of the constructs. The second fragment, comprised of nucleotides 1--1900 of the HBV insert, was generated by PCR from HBV-pBR325 and cloned downstream of the first fragment in pcDNA 3.1. Cells {#s2d} ----- Four cell lines---HCCM, HepG2, PLC/PRF/5, and Sk-Hep-1---derived from human hepatocellular carcinomas were used in this study. HCCM and PLC/PRF/5 contained copies of the integrated HBV genome, while HepG2 and Sk-Hep-1 were obtained from patients with no history of HBV infection and HBV genome integration \[[@pmed-0020163-b16]--[@pmed-0020163-b18]\]. Cells were cultured and maintained in complete Dulbecco\'s modified Eagle\'s medium (Invitrogen, Carlsbad, California, United States) and supplemented with 10% fetal bovine serum (Hyclone, South Logan, Utah, United States) at 37 °C in humidified 5% carbon dioxide. Transient Transfection Assays {#s2e} ----------------------------- Cells were plated at an average seeding density of 1 × 10^6^ cells per well in 35-mm tissue culture dishes and transfected with Lipofectamine 2000 (Invitrogen) according to manufacturer\'s instructions. Briefly, 2 μg of plasmid DNA was used for each transfection mix and added dropwise onto the cells. After incubation for 48 h at 37 °C, the cells were subsequently harvested, followed by DNA isolation (Qiagen). HBV viral loads were measured by real-time PCR using the RealArt HBV LC PCR Kit (Artus GmbH, Hamburg, Germany) according to manufacturer\'s instructions in the LightCycler Instrument (Roche). Experiments were done in duplicate. Luciferase Assays {#s2f} ----------------- For the luciferase assays, 3 μg of plasmid DNA together with 1 μg of control/promoter luciferase plasmid-DNA was used for each transfection mix, and, after incubation for 48 h at 37 °C, cells were harvested with Cell Culture Lysis Reagent (CCLR; Promega). Next, 20 μl of cell lysates was mixed with 100 μl of Luciferase Assay Reagent (Promega), and luciferase activity was measured as relative light units determined with a Turner 20/20 luminometer (Promega). Relative luciferase activity was expressed as fold increase over vector without the enhancer element. Experiments were performed in triplicate. Preparation of Nuclear Protein Extracts and Gel-Shift Assays {#s2g} ------------------------------------------------------------ Cultures were trypsinized, rinsed twice with ice-cold 1× phosphate-buffered saline, and incubated on ice for 10 min with 5× original packed cell volume of Buffer A (10 mM N-2-hydroxyethylpiperazine-N′-ethane sulfonic acid \[HEPES\] buffer \[pH 7.9\], 1.5 mM magnesium chloride, 10 mM potassium chloride, and 1 mM dithiothreitol \[DTT\]). After centrifugation at 1,000 rpm for 3 min at 4 °C, cells were re-suspended in 2× original packed cell volume of Buffer A and homogenized in a Dounce homogenizer with an S pestle with 10 strokes. Nuclei fractions were sedimented by a 10-min centrifugation at 2,500 rpm, re-suspended in 1.5× Buffer B (20 mM HEPES \[pH 7.9\], 0.2 mM ethylene diamine tetra-acetic acid \[EDTA\], 1.5 mM magnesium chloride, 420 mM sodium chloride, 0.5 mM DTT, and 25% glycerol) and treated with another 10 strokes of Dounce homogenizer. Cell suspensions were then transferred to microcentrifuge tubes and incubated for 30 min at 4 °C with gentle rotation. Nuclear debris was removed by centrifugation at 13,000 rpm for 40 min at 4 °C. The supernatant was dialyzed for 4 h against two changes of 200 ml Buffer C (20 mM HEPES \[pH 7.9\], 0.2 mM EDTA, 20 mM magnesium chloride, 20 mM potassium chloride~,~ 420 mM sodium chloride, 25% glycerol, 0.5 mM DTT, and 0.5 mM phenylmethylsulfonyl fluoride) at 4 °C. After dialysis, nuclear extracts were clarified by centrifugation at 13,000 rpm for 20 min. Nuclear extracts were then aliquoted and stored at −80 °C. Protein concentration was quantitated with the Protein Assay kit (Bio-Rad Laboratories, Hercules, California, United States) using acetylated bovine serum albumin as standard. Binding reaction procedures were performed at 37 °C for 20 min in 20-μl reaction mixtures (10 mM Tris-HCl \[pH 7.5\], 50 mM sodium chloride, 1 mM EDTA, and 1 mM DTT) containing 10 μg of HepG2 nuclear extracts, 0.1--0.2 μg of non-specific competitor DNA poly (dI--dC) (Amersham Pharmacia Biotech, Piscataway, New Jersey, United States), and ^32^P-dATP end-labeled probe (1 × 10^4^ to 1 × 10^5^ cpm). Free DNA and DNA--protein complexes were resolved on 6% non-denaturing polyacrylamide gels. Gels were dried down under vacuum at 80 °C for 1 h before exposure to X-ray film (Biomax; Eastman Kodak, Rochester, New York, United States) at −80 °C. The sequences of the oligonucleotide probes (nucleotide changes are indicated) were: Probe 1, AGACTGTGTGTTTAATGAGTGGGAGGAG; Probe 2, AGTTGGGGGAGGAGATTAGGTTAAAGGT; Probe 3, AGACTGTGTGTTTAATGCGTGGGAGGAG; and Probe 4, AGTTGGGGGAGGAGGTTAGGTTAAAGGT. Affinity Capture of Host-Interacting Proteins, Two-Dimensional (2-D) Gel Electrophoresis, and Protein Identification {#s2h} -------------------------------------------------------------------------------------------------------------------- Nuclear protein extracts were obtained from HepG2 cells. HepG2 cells were harvested and rinsed twice with ice-cold Buffer A (0.15 M sodium chloride and 10 mM HEPES \[pH 7.4\]), and incubated on ice for 15 min with 5× original packed cell volume of Buffer B (0.33 M sucrose, 10 mM HEPES, 1 mM magnesium chloride, and 0.1% Triton X-100 \[pH 7.4\]). After centrifugation at 3,000 rpm for 5 min at 4 °C, the pellet was washed once with Buffer B and re-suspended gently on ice with 200 μl of Buffer C (0.45 M sodium chloride and 10 mM HEPES \[pH 7.4\]), with protease inhibitor cocktail \[Sigma P8340\]). The cell mixture was incubated for 15 min with gentle agitation followed by centrifugation at 13,000 rpm for 5 min. The supernatant was saved for DNA-binding proteins assay. Annealing of double-stranded oligonucleotides probes was done using 100 μl of deionized Milli Q water containing 1 nmol each of antisense probe and sense probe, which were labeled with biotin at the 3′ end, and 5′ end, respectively. Oligonucleotide mixture solutions were heated at 95 °C for 5 min and cooled slowly to room temperature. DNA-interacting proteins were captured as described. The oligonucleotide mixture was incubated with 5 mg Dynabeads M-280 streptavidin (Dynal Biotech, Oslo, Norway) at room temperature for 15 min in binding and washing buffer (5 mM Tris-HCl, 0.5 mM EDTA, and 1.0 M sodium chloride \[pH 7.5\]). The magnetic beads were then washed with binding and washing buffer and equilibrated with TGED buffer (20 mM Tris-HCl, 10% glycerol, 1 mM DTT, 0.01% Triton X-100, and 50 mM sodium chloride \[pH 8.0\]). 40 μg of extracted nuclear proteins was mixed 2:1 (w/w) with non-specific competitor DNA poly (dI--dC) (Amersham Biosciences, Little Chalfont, United Kingdom), and adjusted to 500 μl with TGED buffer. Nuclear proteins--poly (dI--dC) solution was added to the equilibrated magnetic beads--oligonucleotide probe at room temperature for 30 min. Unbound proteins were washed out with TGED buffer. Bound proteins were eluted with TGED buffer with 1 M sodium chloride. The same capturing and elution procedure was repeated another four times with new aliquots of nuclear proteins--poly (dI--dC) mixture. Eluted fractions were pooled and subjected to acetone precipitation. We performed 2-D gel electrophoresis according to the Amersham Biosciences protocol, with some modifications. Each sample containing acetone-precipitated proteins was made up to a volume of 350 μl with rehydration buffer (7 M urea, 2 M thiourea, 4% CHAPS, 0.5% IPG buffer \[pH 3--10\], and 1.0 mg of DTT). The mixture was mixed briefly by vortexing, and centrifuged at 13,000 rpm for 10 min. The supernatant was loaded to 18 cm (pH 3--10) nonlinear Immobiline DryStrips (Amersham Pharmacia Biotech), and rehydration was carried out actively at constant voltage (50 V) overnight. Isoelectric focusing was performed using IPGphor (Amersham Biosciences) at 20 °C in stepwise mode. Briefly, strips were focused at 500 V for 1 h, 2,000 V for 1 h, 5,000 V for 1 h, and 8,000 V for 12 h, with a total of 90 kVh accumulated. After isoelectric focusing, the IPG strips were incubated for 30 min in 15 ml of SDS equilibration buffer (50 mM Tris-HCl, 6 M urea, 30% glycerol, 2% SDS, 66 mM DTT, and a trace amount of bromophenol blue \[pH 8.8\]), followed by second incubation with the same buffer for 30 min with iodoacetamide (375 mg/15 ml) instead of DTT. 2-D vertical SDS-PAGE (Protein II XL, Bio-Rad Laboratories) was carried out using 10% gels at a constant voltage of 150 V for 6--8 h at 15 °C. Gels were stained with SilverQuest Silver Staining Kit (Invitrogen). Specific protein spots were cored out and de-stained according to manufacturer\'s instructions, after which the gel plug was dried and MS/MS analysis carried out by Proteomic Research Services (Ann Arbor, Michigan, United States). ### In-gel digestion {#s2h1} Samples were subjected to trypsin digestion on a ProGest (Genomic Solutions, Ann Arbor, Michigan, United States) workstation as follows: Gel plugs were soaked in ammonium bicarbonate solution and reduced with DTT. Alkylation was performed by using iodoacetamide. Samples were incubated at 37 °C overnight in the presence of trypsin. Formic acid was added to stop the reaction. Liquid chromatography/tandem mass spectrometry (LC/MS/MS) analyses: Peptides were cleaned up by using C18 ZipTips (Millipore, Billerica, Massachusetts, United States) and eluted with matrix (α-cyano-4-hydroxycinnamic acid) prepared in 60% acetonitrile and 0.2% TFA. 15 μl of eluent was processed on a 75-μm C18 column at a flow-rate of 20 nl/min. Eluent from the C18 column was fed into nano-LC/MS/MS on a Micromass Q-TOF2 mass spectrometer. MS/MS data were searched using a local copy of MASCOT search engine (Matrix Science, London, United Kingdom). RNA Interference {#s2i} ---------------- Small interfering RNA (siRNA) duplexes against hnRNP K were purchased from Dharmacon (SmartPool) (Lafayette, Colorado, United States), Qiagen (sequence: GCAGUAUUCU GGAAAGUUU), and Proligo (sequence: CUU GGGACUCU GCAAU AGATT) (Boulder, Colorado, United States). HepG2 cells were co-transfected in 24-well tissue culture plates with 1 μg of plasmid DNA (1752A replicative full-length clone) and the respective siRNA duplexes (2 μg) using 6 μl Lipofectamine 2000 (Invitrogen). After 48 h, the cells were collected and were RNA and DNA extracted. As controls, cells were also transfected with fluorescence-labeled non-silencing siRNA to monitor the transfection efficiencies. Transfections were performed in duplicate. Quantitation of hnRNP K Expression {#s2j} ---------------------------------- Total RNA was isolated with RNeasy kit (Qiagen) according to the manufacturer\'s instructions. The concentration and purity of extracted RNA were determined by measuring the A~260~ and A~280~, and adjusted to 250 ng/μl, which was then used as a template for real-time RT-PCR. 2 μl of RNA was used in a one-step real-time RT-PCR reaction with the LightCycler RNA Master SYBR Green kit (Roche) using primers 5′- AGACCGTTACGACGGCATGGT-3′ and 5′- GATCGAAGCTCCCGACTCATG-3′, and performed according to manufacturer\'s instructions. Absolute quantitation of RNA was obtained by using standard curves created with in vitro--transcribed RNA by the T7 RiboMax Express in vitro transcription system (Promega). The concentration of purified transcribed RNA was measured by RiboGreen RNA quantitation reagent (Invitrogen). Serial dilutions of in vitro--transcribed RNA were prepared in duplicate. Statistical Analysis {#s2k} -------------------- A multi-alignment sequence analysis of the HBV genomes with ranked HBV viral load was first performed, and we compared the frequencies of the mutations at nucleotide position 1752 using Fisher\'s exact probability test. Results {#s3} ======= Correlation between High and Low Viral Loads in Patients at a Precise Location in the HBV Genome {#s3a} ------------------------------------------------------------------------------------------------ To study the influence of HBV variants on the resultant dynamics of viremia in HBV carriers, we analyzed serum samples from 60 carriers for viral load using a commercially available test kit (Digene). The serum viral load among HBV carriers ([Figure 1](#pmed-0020163-g001){ref-type="fig"}) was found to vary by more than three orders of magnitude (0.1--376 pg/ml). About 60% (36/60) of the carriers had single-digit pg/ml (0.1--5 pg/ml) concentration of viral DNA, while 40% (24/60) had greater than 10 pg/ml (11--376 pg/ml) concentration ([Figure 1](#pmed-0020163-g001){ref-type="fig"}). To determine if viral variants had any link with viral load in the carriers, we amplified HBV DNA by PCR from serum, sequenced, and aligned the amplified fragments. Interestingly, within the entire viral genome, we observed a unique nucleotide position that showed significant correlation (p = 0.0001) with viral load as defined by the two broadly partitioned groups of serum viral load described above (using 10 pg/ml as an arbitrary cut-off). As there was no particular physiological phenotype linked to a threshold HBV viral load, we also applied cut-off values of 20 pg/ml and 50 pg/ml; in both instances, the results were similar to the 10-pg/ml value (data not shown). Although the HBV genotypes were predominantly genotype C, we observed a distinct tendency for carriers with high levels of serum HBV DNA to possess an A nucleotide at position 1752 of the virus sequence, while those with low levels of serum HBV DNA tend to have a G nucleotide at this position. This natural variation at nucleotide position 1752 was found to be located at the HBV genome\'s Enh II region, which has been shown to be a specific regulator of HBV replication \[[@pmed-0020163-b08],[@pmed-0020163-b09],[@pmed-0020163-b11],[@pmed-0020163-b19],[@pmed-0020163-b20]\]. ::: {#pmed-0020163-g001 .fig} Figure 1 ::: {.caption} ###### Distinct Segregation between High and Low Viremic HBV Individuals Is Correlated to Changes at Nucleotide Position 1752 Comparison of DNA sequences from nucleotide 1720--1769 with the HBV DNA concentration levels of the participants are illustrated. Sera was collected from 60 participants; DNA was isolated from sera and amplified with two rounds of PCR. Results of the sequences were aligned and compared. ::: ![](pmed.0020163.g001) ::: Natural Mutants of HBV with Different Viral Replication Efficiencies {#s3b} -------------------------------------------------------------------- As this variation resided in the viral enhancer element ([Figure 2](#pmed-0020163-g002){ref-type="fig"}A), we proceeded to study the effects of this base substitution on transcriptional efficiency by cloning a 131-base-pair Enh II fragment bearing 1752A upstream of a SV40 promoter--luciferase reporter gene vector. Site-directed mutagenesis was then carried out to generate three other constructs bearing 1752G, 1752T, and 1752C so that four constructs differing from each other at only this single base nucleotide could be tested ([Figure 2](#pmed-0020163-g002){ref-type="fig"}B). Transient transfections were carried out on the hepatic cell line HepG2 and subsequently assayed for luciferase activity. The results showed that the construct with 1752A strongly enhanced the SV40 promoter when linked to it in cis, resulting in higher levels of luciferase expression as compared to all other base substitutions ([Figure 2](#pmed-0020163-g002){ref-type="fig"}C). As a positive control, the vector containing both the SV40 promoter and enhancer sequences resulting in optimal luciferase expression was used ("+" in Column 1 in [Figure 2](#pmed-0020163-g002){ref-type="fig"}C). In addition, the cloning vector itself containing only the SV40 promoter--luciferase gene but without the enhancer element ("−" in Column 2 in [Figure 2](#pmed-0020163-g002){ref-type="fig"}C) served as the negative control. To determine if these results were specific to only the host HepG2 cell line, we repeated the transfections in three other hepatic cell lines: SKHep1, PLC/PRF/5, and HCCM. The results were consistent with the initial results observed with HepG2 ([Figure 2](#pmed-0020163-g002){ref-type="fig"}D, [2](#pmed-0020163-g002){ref-type="fig"}E, and 2F), further substantiating that the single base change at nucleotide position 1752 has a significant effect on the transcriptional efficiency of Enh II. These in vitro assays provide strong support for the effect of 1752A on higher HBV viral load, but it is worthwhile to note that this correlation is not absolute. As can be seen in [Figure 1](#pmed-0020163-g001){ref-type="fig"}, the nucleotide change in 1752A cannot alone account for the wide variation in HBV viral load, suggesting that there may be additional contributing factors. Subsequent experiments were carried out using the HBV negative HepG2 cell line, as it has better growth characteristics than SKHep1, while PLC/PRF/5 and HCCM have integrated HBV genomes. ::: {#pmed-0020163-g002 .fig} Figure 2 ::: {.caption} ###### Effect of Nucleotide 1752 Base Substitution on Enh II Activity \(A) The Enh II is located just upstream of the core promoter. The minimum sequence for enhancer activity has been previously defined at nucleotide 1687--1805 as shown. \(B) Site-directed mutagenesis on nucleotide 1752 in the Enh II (1752A, 1752G, 1752T, and 1752C) and amplified fragments were inserted upstream of the SV40 promoter in an enhancerless luciferase reporter vector. (C--F) Four cell lines---HepG2, PLC/PRF/5, SKHep1, and HCCM, all derived from human hepatocellular carcinomas---were transfected with the respective Enh II clones (1752A, 1752G, 1752T, and 1752C). For each cell type, the first column ("+") represents the luciferase activity of the internal positive control (promoter and enhancer). The second column ("−") represents the activity of the internal negative control (promoter only). The other columns of each cell line represent the vectors with the promoter and the Enh II (nucleotide 1752A, 1752G, 1752T, or 1752C). Results of the luciferase assay were normalized to the level of the internal positive control (arbitrarily set at 100%). ::: ![](pmed.0020163.g002) ::: Identification of Host Cellular Factor as hnRNP K {#s3c} ------------------------------------------------- In order to further establish and confirm the biological significance of this point mutation observed in the Enh II region, we wanted to demonstrate the presence of any direct physical HBV DNA--host interaction at the nucleotide position 1752 site. To this end, 28-mer oligonucleotide probes were designed to contain either the 1752A or 1752G nucleotide, with control probes taken from the immediate adjacent upstream sequence. Electrophoretic mobility shift assays were performed using HepG2 nuclear extracts with the respective probes, and the results showed a distinct DNA-binding protein was detected using the 1752A probe (Probe 2, lanes 5--8 in [Figure 3](#pmed-0020163-g003){ref-type="fig"}A) along with a weaker band of similar size using the 1752G probe (Probe 4, lanes 13--16 in [Figure 3](#pmed-0020163-g003){ref-type="fig"}A). Densitometric analysis of the bands indicated that the protein detected by Probe 2 was about 300% higher than that detected by Probe 4, suggesting that the 1752A probe has a higher binding affinity for the DNA-binding protein. These data lend support to the possibility that host DNA-binding proteins interact directly with the HBV viral sequence. ::: {#pmed-0020163-g003 .fig} Figure 3 ::: {.caption} ###### Evidence for the Involvement of a Host Cellular Protein in Enh II Activity and HBV Replication \(A) Electrophoretic mobility shift assays were performed using HepG2 nuclear extracts with four different probes. Probe 1, lanes 1--4; probe 2 (1752A), lanes 5--8; probe 3, lanes 9--12; probe 4 (1752G), lanes 13--16. Each set of probes contains increasing concentrations (0.0 μg, 0.05 μg, 0.10 μg, and 0.15 μg) of non-specific competitor DNA \[poly-(dI)-poly-(dC)\], respectively. \(B) 40 μg of nuclear protein extracts obtained from HepG2 cells was allowed to bind onto 5 mg Dynabeads M-280 streptavidin-biotin-oligonucleotides in the presence of 2:1 (w/w) ratio of non-specific competitor DNA poly (dI--dC). 1-D isoelectric focusing was followed by 2-D vertical separation on SDS-PAGE (10%). The estimated molecular weight of the specific protein spots detected by silver staining (arrow) is indicated. ::: ![](pmed.0020163.g003) ::: To further characterize this DNA-binding protein, nuclear extract of HepG2 was passed through an affinity column tagged with the oligonucleotide probe (bearing 1752A). Bound material was eluted and subjected to 2-D gel analysis. Silver staining of the gels demonstrated positive enrichment of a specific DNA-binding protein compared with the non-specific binding control oligonucleotide probe ([Figure 3](#pmed-0020163-g003){ref-type="fig"}B). The specific protein spots of approximately 56 kDa molecular weight were cored out and subjected to LC/MS/MS analysis ([Figure 4](#pmed-0020163-g004){ref-type="fig"}A). Peptide sequencing results showed that 21 fragments of the unknown DNA-binding protein turned out to match perfectly with that of a known protein, hnRNP K ([Figure 4](#pmed-0020163-g004){ref-type="fig"}B). The peptide sequence match and molecular weight of the predicted protein strongly suggest that the LC/MS/MS analysis has correctly identified the protein bound by the oligonucleotide-affinity column. The hnRNP K protein, which belongs to the family of hnRNPs, is about 56 kDa in size and has been shown to be involved in various cellular functions as described above. ::: {#pmed-0020163-g004 .fig} Figure 4 ::: {.caption} ###### Identification of Host Cellular Protein as hnRNP K \(A) Specific protein spots were cored out and de-stained, following which the gel plug was digested with trypsin. Sequence query of peptide fragments was carried out by Proteomic Research Services, using LC/MS/MS analysis. Results of the 21 sequenced peptides are illustrated. \(B) Results of peptide sequencing of the 56-kDa protein by LC/MS/MS showed high homology scores to hnRNP K in sequence alignments. ::: ![](pmed.0020163.g004) ::: hnRNP K Is an Essential Modulator of Viral Replication {#s3d} ------------------------------------------------------ To demonstrate the functional connection of hnRNP K to the regulation of HBV viral load, a 1.4-kb RT-PCR fragment coding for the full-length hnRNP K gene from total RNA extracted from HepG2 cells was cloned into the mammalian expression vector pcDNA 3.1 (Invitrogen). There are two known variants of hnRNP K, termed "variant 2" and "variant 3," which differ only at the last six amino acid residues of the C-terminal region. As there are no reported studies on the functional difference between the two variants, we decided to work with both variants. The two hnRNP K expression constructs were co-transfected separately into HepG2, together with a full-length 1752A replicative clone of HBV driven by the HBV core promoter rather than the CMV promoter \[[@pmed-0020163-b14],[@pmed-0020163-b15]\] (see [Figure S1](#sg001){ref-type="supplementary-material"}) to determine the effects of hnRNP K on the replicative efficiency of the HBV construct. As predicted, the HBV viral concentration increased in a dose-dependent manner with hnRNP K concentration, with no significant functional difference between variants 2 and 3 ([Figure 5](#pmed-0020163-g005){ref-type="fig"}). As a control, the empty expression vector pcDNA 3.1 did not have any effect on the HBV viral concentration. To further verify the impact of the single-base substitution at nucleotide position 1752, site-directed mutagenesis was carried out to generate three other full-length HBV replicative constructs bearing 1752ΔG, 1752ΔT, and 1752ΔC. Co-transfections with either hnRNP K variant 2 or variant 3 were performed as described above. All three constructs had 68--80% reduced HBV DNA when compared to the 1752A construct, indicating a lowered level of HBV viral load ([Figure 5](#pmed-0020163-g005){ref-type="fig"}), although a high dosage of hnRNP K was able to augment the replication efficiency of the three constructs. Taken together, this set of co-transfection experiments provide definitive evidence that we have precisely mapped an important virus element and associated host component required for regulating HBV replication. ::: {#pmed-0020163-g005 .fig} Figure 5 ::: {.caption} ###### hnRNP K Is Involved in Modulating Viral Replication HepG2 cells were co-transfected with full-length replicative HBV clones (indicated by "+") 1752A, 1752ΔG, 1752ΔT, and 1752ΔC (see [Methods](#s2){ref-type="sec"} and [Figure S1](#sg001){ref-type="supplementary-material"}) with increasing dosages (50, 250, and 1,250 ng/μl) of hnRNP K variant 2 (v2) or variant 3 (v3) as indicated. pcDNA 3.1 serves as a control. Transfections were performed in duplicate; standard deviations are shown. ::: ![](pmed.0020163.g005) ::: As a further critical proof of the regulatory role of hnRNP K in HBV viral load, the converse experiments to knock down hnRNP K should show a suppressive effect on viral load. To do this, siRNAs against hnRNP K were designed and obtained from three separate manufacturers. HepG2 cells were co-transfected with the replicative HBV full-length clone (1752A) together with the panel of hnRNP K siRNAs, and mRNA and DNA were analyzed 48 h after siRNA transfection. Non-silencing siRNA not matching to any genes and lamin A/C siRNAs were used as controls. The resultant hnRNP K mRNA levels as measured by quantitative real-time RT-PCR showed a 30% reduction relative to the non-transfected cells, non-silencing siRNA, and lamin A/C siRNA controls ([Figure 6](#pmed-0020163-g006){ref-type="fig"}A). HBV viral loads were correspondingly decreased by 50% using both siRNAs from source B and C, while the siRNAs from source A achieved a 15% reduction ([Figure 6](#pmed-0020163-g006){ref-type="fig"}B). The difference in effectiveness of siRNA from the three sources may be due to the different target regions that are selected to knock down hnRNP K, but the general suppressive trend was clear (see [Methods](#s2){ref-type="sec"}). The lamin A/C mRNA levels measured by real-time RT-PCR in HepG2 cells transfected with lamin A/C siRNA showed a 45% reduction relative to the non-transfected cells, while non-silencing siRNA and hnRNP K siRNAs had no effect on the lamin A/C mRNA levels ([Figure 6](#pmed-0020163-g006){ref-type="fig"}C). Taken together, these results confirm that hnRNP K plays a critical role in the process of HBV replication. The observation that changes in intracellular hnRNP K levels (both up and down) directly alter HBV replication efficiency points to a regulatory role for hnRNP K. ::: {#pmed-0020163-g006 .fig} Figure 6 ::: {.caption} ###### hnRNP K siRNAs Down-Regulate HBV Viral Replication \(A) HepG2 cells were co-transfected with 1752A full-length replicative HBV clone either with or without hnRNP K siRNA (2 μg). Non-silencing (Non-T) and lamin A/C (Lamin) siRNAs were used as controls. hnRNP K expression was measured by quantitative real-time RT-PCR. \(B) HBV viral load was quantitated by real-time PCR in cells transfected as described in (A). \(C) Lamin A/C expression was measured from real-time RT-PCR. Ratios were normalized to 100% for the non-transfected cells. The results represent two independent samples; standard deviations are shown. Black columns represent either non-transfected cells or cells transfected with non-silencing siRNA. White columns represent cells co-transfected with HBV and lamin A/C siRNA. Grey columns represent cells co-transfected with HBV and hnRNP K siRNAs (A, Dharmacon; B, Qiagen; C, Proligo). ::: ![](pmed.0020163.g006) ::: Discussion {#s4} ========== In this study, we were able to show that a host protein---hnRNP K---can be isolated by direct binding to a viral fragment derived from the HBV variant of the infected patient, and that hnRNP K binds to and modulates the replicative efficiency of HBV at a precise location at the Enh II regulatory region. Overexpression studies and RNA interference studies show a direct demonstration of the dependence of HBV on a host factor to modulate its replication efficiency, and hold promise as a new class of targets for the intervention of chronic hepatitis B infection. The interesting observation that chronic HBV carriers have different serum viral loads spanning three logs (0.1--376 pg/ml in this study) between patients prompted us to investigate the possible correlation with viral genotype. While no such particular relationship emerged, we detected a solitary yet distinct natural mutation at nucleotide position 1752 that had an "A" that segregated with viral loads greater than 10 pg/ml, while those with "G" segregated with the samples of lower than 10 pg/ml. Of the 60 HBV genomes sequenced, only one sample showed a "T" at position 1752, while "A" and "G" variants were present in almost equal proportions. No natural "C" variant was seen, but this could be due to our limited sample size or an effect related to geographical distribution patterns. To understand whether the natural variants at 1752 had any functional impact, we developed reporter constructs and tested all four possible 1752 variants for the ability to drive reporter gene transcription. It was evident that 1752A had higher activity than the three other non-A constructs. To investigate the underlying basis for this enhanced transcriptional activity, we proposed to search for evidence of possible physical interaction with DNA-binding proteins. Using an initial electrophoretic mobility shift assay followed by an affinity pull-down assay run on 2-D gel analysis, a 1752A oligonucleotide-affinity fragment was able to enrich a host binding factor sufficient for protein sequencing. The binding factor turned out to be hnRNP K, a known protein that has been shown to be involved in a number of cellular functions \[[@pmed-0020163-b12],[@pmed-0020163-b13]\]. It has multiple modular domains, such as the K homology domains \[[@pmed-0020163-b21]\] and RGG boxes that allow it to interact with both DNA and RNA \[[@pmed-0020163-b22]--[@pmed-0020163-b25]\]. Reports have indicated that interactions between hnRNP K and single-stranded DNA are mediated by three K homology domains \[[@pmed-0020163-b21]\], and that hnRNP K exhibits specific binding and transactivation within the c-myc promoter \[[@pmed-0020163-b26]--[@pmed-0020163-b30]\]. hnRNP K was also shown to interact physically with the proto-oncogenes c-src and vav \[[@pmed-0020163-b22],[@pmed-0020163-b24],[@pmed-0020163-b31],[@pmed-0020163-b32]\], thus allowing it to form multienzyme complexes and facilitate kinase cross-talk. This result suggests that hnRNP K is a versatile molecule that can act as a regulator of signal transduction and gene expression. An important next step was to show that hnRNP K acts on a full-length HBV genome and not just the 131-bp enhancer fragment that was tested earlier in the luciferase reporter assay. To this end, we constructed four full-length replicative HBV clones, which were identical except for a single base change at position 1752. Each replicative clone variant was co-transfected with two different hnRNP K expression constructs that represent the two known hnRNP K variants, differing only at the C-terminal end of six amino acids. The results once again confirmed that the 1752A was more efficient than the other three variants, while there was no significant difference between either variant of hnRNP K in enhancing viral load. With escalating doses of hnRNP K, the viral load of the 1752G, 1752T, and 1752C variants increased in a dose-dependent manner. This suggests that the increased cellular concentrations of hnRNP K compensated to some degree for the reduced affinity bought about by the nucleotide substitutions. In order to further demonstrate the role of hnRNP K in HBV replication, we wanted to test not only over-expression but also whether down-regulation of the cellular protein could have an effect. siRNAs designed to knock down endogenous hnRNP K were able to suppress both hnRNP K mRNA, and, along with it, the HBV viral load was greatly reduced. More importantly, the inclusion of control siRNA to lamin A/C suppressed lamin A/C mRNA but had no effect on HBV viral load, thus strengthening the link between hnRNP K and HBV. The mechanistic aspect of hnRNP K on HBV replication needs to be further explored, and efforts are currently under way. There are possibilities such as hnRNP K polymorphisms \[[@pmed-0020163-b33]\] (see [Figure S2](#sg002){ref-type="supplementary-material"}), phosphorylation states \[[@pmed-0020163-b21],[@pmed-0020163-b22]\], transcriptional regulation, and perhaps other cellular protein factors that may act in concert to support the replication activity of HBV in the host. That hnRNP K binding is mapped precisely to a single-base polymorphism in the HBV genome\'s regulatory region suggests that a dynamic interaction exists between specific natural mutants of the HBV Enh II region and the patient\'s genotype composition that together function to determine the overall outcome of viral replication efficiency. Indeed, a deeper understanding of how this works will provide insights into defining virulence and fitness of a virus. Our study shows that the overall viral replication efficiency is determined by a combination of both viral sequence and interaction with specific host proteins. While development of antivirals is an established path, targeting the host remains surprisingly unexplored. Interestingly, a recent study on anti-EGFR antibody treatment of breast cancer cells showed a decrease in the cell-replication rate with a corresponding reduction in hnRNP K expression levels \[[@pmed-0020163-b34]\]. This suggests that hnRNP K levels can be modulated by anti-EGFR treatment and holds possibilities for new indications for existing and approved pharmaceuticals as an approach to alter HBV viral load in chronic carriers. Overcoming entrenched, chronic viral infections is not a straightforward solution and will eventually involve a combination therapy of targeting the virus directly, blocking host support proteins, and simultaneously employing immuno-modulating agents to bring about long-term viral clearance. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### Construction of the Full-Length HBV Replicative Clone (53 KB PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### SNPs in the hnRNP K Gene from 18 Volunteers Compared with SNPs Extracted from Ensembl and Celera Databases (59 KB PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Patient Summary {#sb1} --------------- ### Background {#sb1a} Hepatitis B is a virus that can cause long-term health problems, including cirrhosis of the liver and liver cancer in people who are not able to clear the virus from their body. Although some drugs can suppress the multiplication of the virus, these drugs do not cure the patient of hepatitis B. Another way of tackling the virus might become clear if it were known exactly how the virus interacts with the patient\'s cells. ### What Did the Researchers Do? {#sb1b} They noticed that among patients who carried the hepatitis B virus (HBV), some had higher amounts of the virus in their blood compared to others. The patients with higher levels tended to have a virus that had one particular DNA type, and this particular virus type could bind more closely to a protein found in humans: hnRNP K. In the laboratory, when the investigators increased the amount of this protein in cells infected with HBV, the virus multiplied more; when they decreased it, the virus multiplied less. ### What Does This Mean? {#sb1c} This work suggests that one way to control the multiplication of HBV in infected people is indirectly, by altering levels of the protein hnRNP K. The next step will be to identify ways to do this safely and reliably. ### Where Can I Get More Information? {#sb1d} Medline Plus has a section on hepatitis B: <http://www.nlm.nih.gov/medlineplus/ency/article/000279.htm> The World Health Organization has a set of information on hepatitis, including hepatitis B: <http://www.who.int/topics/hepatitis/en/> We thank D. Simarmata, D. Kwek, L. V. Agathe, P. Tay (Genome Institute of Singapore), and H. Pan (Nanyang Technological University) for their excellent technical support, and E. Liu for helpful comments. We also thank the following contributors for patients\' serum used in this study: National University Hospital, Singapore General Hospital, and Tan Tock Seng Hospital. This work was supported by funding from the Agency of Science and Technology Research (A\STAR), Singapore. LFPN is supported by a Postdoctoral Fellowship from the Singapore Millennium Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Citation: Ng LFP, Chan M, Chan SH, Cheng PCP, Leung EHC, et al. (2005) Host heterogeneous ribonucleoprotein K (hnRNP K) as a potential target to suppress hepatitis B virus replication. PLoS Med 2(7): e163. DTT : dithiothreitol EDTA : ethylene diamine tetra-acetic acid HBV : hepatitis B virus HEPES : N-2-hydroxyethylpiperazine-N′-ethane sulfonic acid hnRNP : heterogeneous nuclear ribonucleoprotein LC/MS/MS : liquid chromatography/tandem mass spectrometry siRNA : small interfering RNA [^1]: Competing Interests:* The authors have declared that no competing interests exist. [^2]: Author Contributions: LFPN, MC, SHC, WNC, and ECR designed the study. LFPN, MC, PCPC, EHCL, and WNC performed the experiments. LFPN, MC, and ECR contributed to writing the paper.	PubMed Central	2024-06-05T03:55:59.855465	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181871/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e163", "authors": [ { "first": "Lisa F. P", "last": "Ng" }, { "first": "Marieta", "last": "Chan" }, { "first": "Soh-Ha", "last": "Chan" }, { "first": "Paul Chung-Pui", "last": "Cheng" }, { "first": "Eastwood Hon-Chiu", "last": "Leung" }, { "first": "Wei-Ning", "last": "Chen" }, { "first": "Ee-Chee", "last": "Ren" } ] }
PMC1181872	Introduction {#s1} ============ Postherpetic neuralgia (PHN) is a chronic neuropathic pain syndrome that may complicate recovery from an acute attack of herpes zoster. Despite advances in antiviral therapy during acute herpes zoster and the more recent introduction of vaccination against varicella zoster, PHN continues to be a significant clinical problem, with up to 25% of patients developing persistent neuropathic pain after acute herpes zoster reactivation \[[@pmed-0020164-b01]\]. Acute herpes zoster, and, consequently, PHN, particularly afflicts the immunocompromised and elderly, a fact that has serious implications for health-care delivery in the context of ageing populations in the developed world and the worldwide spread of HIV disease. Left untreated, PHN can become a severe and debilitating condition affecting all aspects of a patient\'s life \[[@pmed-0020164-b02]\]. The nature of neuropathic pain in PHN is variable; it may be described as continuous or paroxysmal, evoked or spontaneous, burning or lancinating, and be associated with a range of other sensory abnormalities in the skin \[[@pmed-0020164-b03]\]. This variability in symptomatology could imply that a variety of different pain mechanisms may be operating in different patients with PHN or in the same patient at different points in time \[[@pmed-0020164-b03],[@pmed-0020164-b04]\]. This hypothesis has led to the suggestion that treatment plans could be optimised for individual patients on the basis of symptoms or even mechanisms \[[@pmed-0020164-b05]\]. However, the current situation is that the existing evidence base for therapies in PHN is constructed largely from clinical trials of analgesics that have examined PHN as a single disease entity. Furthermore, no substantial evidence base exists to relate specific sets of symptoms or signs to the efficacy of specific drugs, and no simple validated methods exist to determine which neuropathic pain mechanism(s) may be operating in a single patient. These factors dictate that current neuropathic pain treatment paradigms are forced to focus on PHN as a single disease entity. No single treatment has been shown to be completely effective for all sufferers of PHN, and in the practical clinical scenario combinations of analgesic drugs are usually required to achieve partial relief of pain. Although there are an increasingly large number of trials that compare various analgesics to placebo, very few directly compare single therapies for which an evidence base exists, or address the issue of combining treatments \[[@pmed-0020164-b06]\]. Although published systematic reviews have collated the evidence base for analgesic therapy in PHN \[[@pmed-0020164-b07]--[@pmed-0020164-b09]\] and neuropathic pain in general \[[@pmed-0020164-b10]\], the evidence base has fundamentally and substantially altered since the publication of these reviews, with the appearance of several major studies \[[@pmed-0020164-b11]--[@pmed-0020164-b21]\], partly because of the increasing recognition of the usefulness of PHN as a clinical model of neuropathic pain for trials. Therefore, we have conducted a systematic review and meta-analysis of analgesic therapy for PHN, which includes these more recent trials. Methods {#s2} ======= Inclusion Criteria {#s2a} ------------------ The time at which the pain of zoster-associated pain becomes PHN is debated in the literature. We have used the definition of pain persisting for longer than 3 mo after the crusting of skin lesions following an acute attack of herpes zoster. Trials were sought that examined adult patients with zoster-associated pain for greater than 3 mo, were blinded, randomised, and had at least one clinically relevant measure of pain outcome. Unpublished, letter, and abstract-only studies were excluded as were studies on prevention of PHN and anecdotes. Studies where data for PHN were not analysed separately from other neuropathic pain syndromes were also excluded. All randomised controlled trials (RCTs) identified up to October 2004 were included. Identification of Studies {#s2b} ------------------------- The following databases were searched without language restrictions: MEDLINE 1966--2004, EMBASE 1988--2004, CINAHL 1982--2002, and PubMed (29 October 2004). Search terms used were: "herpes zoster\," "postherpetic neuralgia\," "neuralgia\," "pain\," and "neuropathic\," in combination with "randomised," "random\," "random allocation," "double-blind," "controlled clinical trials," "trials," and "study." The Cochrane Controlled Trial Register and Cochrane Library (2004) database was searched using similar search terms. References of retrieved studies and review articles were also searched for further trials. Quality Assessment {#s2c} ------------------ Identified studies were independently assessed by at least two of four authors (KH, TJN, ASCR, and RWJ) in order to ascertain whether the inclusion criteria for PHN were met. The trials that met the PHN criteria were then quality scored by the same authors using the five point "Jadad" scoring system \[[@pmed-0020164-b22]\]. Studies were excluded if they achieved a score of less than three or if the enrolled study population was ten patients or less \[[@pmed-0020164-b23]\]. Disagreement between authors as to scores was adjudicated by ASCR to reach a consensus. Data Extraction {#s2d} --------------- The following data were extracted from each study: drug or treatment examined, number of patients enrolled and analysed, study design and duration, dosing regimen, outcome measures used, pain-relief outcomes, minor and major adverse events, and withdrawals. For crossover studies, patient episodes were calculated as one episode representing the result for one patient completing one limb of the crossover, i.e., one patient who completed both active and placebo arms of a trial was counted as two patient episodes. For crossover studies, information was extracted regarding the provision of washout periods and also as to whether verification of the adequacy of any washout period was performed (e.g., return of pain intensity to baseline before second treatment period). Data were extracted by two of the authors (KH, TJN, or ASCR) independently. For the measurement of treatment efficacy, an outcome was considered clinically relevant if an improvement of 50% or greater in pain relief was achieved. Where possible, dichotomous data were extracted; when dichotomous data were not presented in the published study, the corresponding authors were contacted and asked to supply such data if they were still available. Due to the length of time that had elapsed since many of the publications, these data were often not available. Data were extracted for the longest follow-up period reported in each trial. Because of the wide variety of outcome measures used across all of the reviewed papers, a hierarchy of outcome measures was implemented in line with the system used by several previous pain-treatment systematic reviews \[[@pmed-0020164-b08],[@pmed-0020164-b23],[@pmed-0020164-b24]\], as follows: 1) top two values on a five-point patient-reported global scale for pain relief or effectiveness or improvement; 2) top three values on a six-point patient-reported global scale for pain relief or effectiveness or improvement; 3) top value on a three-point patient-reported global scale for pain relief or effectiveness or improvement; 4) top two values on a four-point patient-reported categorical pain-relief scale; 5) 50% or greater reduction on a visual analogue or 11-point numerical rating scale for pain intensity. Where available, dichotomous data were collected on adverse events. A major adverse effect was defined as an event that precipitated the withdrawal of the patient from the study, while an adverse event was considered minor if the patient continued the treatment and completed the trial. Some studies reported withdrawals not considered to be related to the treatment separately. However, this approach was not consistently the case across the included studies, so we adopted a conservative approach and included all withdrawals in the figures for major harm. Comparator trials using active controls were excluded from the safety analysis, in keeping with the system used by Collins et al. \[[@pmed-0020164-b08]\]. Statistical Methods {#s2e} ------------------- A quantitative analysis was carried out on those trials where dichotomous data were available. Estimates of efficacy calculated were relative benefit (RB) and number needed to treat (NNT). The difference between active treatment and placebo was taken as statistically significant when the lower limit of the 95% confidence interval (CI 95%) for RB was greater than one. If tests of homogeneity were favourable, pooling of data was carried out for groups of similar treatments. A qualitative comment is made on those trials where no dichotomous data were obtainable. Safety analysis was carried out on placebo-controlled trials using relative risk (RR) and number needed to harm (NNH). A fixed-effects model was used for calculating RB and RR with 95% CIs. A statistically significant improvement was noted when the lower limit of the 95% CI for RB was greater than one. Where RB or RR was not statistically significant, the method proposed by Altman was used to describe the CI for NNT or NNH \[[@pmed-0020164-b25]\]. Number needed to treat to benefit (NNTB) and number needed to treat to harm (NNTH) are given, so that if NNT was calculated as ten and the 95% CI 40 to −20, then this would be expressed as NNT10 (NNTH 20 to : to NNTB 40). If there were no responders in either the active or placebo-treated groups, then the RB was undefined and this situation was dealt with by the method proposed by Fleiss \[[@pmed-0020164-b26]\]. When calculating odds ratios, 0.5 was added to each of the cell frequencies in the 2 × 2 table showing response by treatment allocation. Although the absolute risks in the arms may vary between trials and may be affected by factors such as varying lengths of follow-up, the assumption was made that the RRs would be similar across trials that addressed the same question. Where possible, combined figures for classes of drugs were calculated based on RR estimates from each of the trials. Heterogeneity tests were performed on these trials; a value of p \> 0.05 was interpreted to mean that it was appropriate to combine the RRs from different trials. Calculations were performed using the methods proposed by Armitage and Berry \[[@pmed-0020164-b27]\], programming being undertaken using Microsoft Excel 2000. Some trials addressing a particular question will show treatment benefits that are clearly greater or less than the typical benefit observed in all the other trials addressing that question; in this case, there is heterogeneity between the trial outcomes (i.e., trials with a RR that differs from the overall RR more than would be expected by chance). The method proposed by Galbraith was used to identify such atypical trials \[[@pmed-0020164-b28]\]. If the treatment benefit is greater, we have described such trials as "high"; if it is less, we have described them as "low." This review is reported in accordance with the QUOROM guidelines \[[@pmed-0020164-b29]\]. Results {#s3} ======= After the searches were completed, and obvious review articles, case and series studies, and anecdotes excluded, 62 articles were retrieved and independently reviewed by at least two of four of the authors (KH, TJN, RWJ, and ASCR). Twenty-seven studies were excluded at this stage ([Figure 1](#pmed-0020164-g001){ref-type="fig"}). Details of excluded papers and, therefore, interventions that must be regarded as not having been adequately tested, are shown in [Table 1](#pmed-0020164-t001){ref-type="table"}. ::: {#pmed-0020164-g001 .fig} Figure 1 ::: {.caption} ###### QUOROM Statement Flow Diagram ::: ![](pmed.0020164.g001) ::: ::: {#pmed-0020164-t001 .table-wrap} Table 1 ::: {.caption} ###### Studies Excluded from the Analysis and Therefore Intervention Not Adequately Tested ::: ![](pmed.0020164.t001) ::: Of 35 trials retained for further analysis, 18 were of a crossover design and 17 were of a parallel group design ([Table 2](#pmed-0020164-t201){ref-type="table"}). Thirty-one trials were placebo controlled (including "active" placebo). Four trials were comparator studies without a placebo group and therefore could not be included in the meta-analysis. These trials compared amitriptyline to nortriptyline \[[@pmed-0020164-b30]\], amitriptyline to maprotiline \[[@pmed-0020164-b31]\], and intrathecal steroid to epidural steroid \[[@pmed-0020164-b17]\] and two doses of levorphanol \[[@pmed-0020164-b21]\]. Of the remaining 31 trials, we were able to extract dichotomous outcome data for efficacy meta-analysis from 25 ([Table 3](#pmed-0020164-t301){ref-type="table"}). Qualitative comment has been made on the included studies from which dichotomous data could not be extracted. ::: {#pmed-0020164-t201 .table-wrap} Table 2 ::: {.caption} ###### Studies Included in the Efficacy Analysis ::: ![](pmed.0020164.t201) ::: ::: {#pmed-0020164-t202 .table-wrap} Table 2 ::: {.caption} ###### Continued ::: ![](pmed.0020164.t202) ::: ::: {#pmed-0020164-t301 .table-wrap} Table 3 ::: {.caption} ###### Summary of Data from Placebo-Controlled Trials for Which Dichotomous Data for Efficacy Could Be Extracted ::: ![](pmed.0020164.t301) ::: ::: {#pmed-0020164-t302 .table-wrap} Table 3 ::: {.caption} ###### Continued ::: ![](pmed.0020164.t302) ::: Data extracted were for the longest follow-up period reported in each trial. In the vast majority of trials, this was only until the end of the treatment period, with the exception of the intrathecal methylprednisolone studies \[[@pmed-0020164-b16],[@pmed-0020164-b17]\] that reported follow-up periods to 24 wk and 2 y, and one study that examined amitriptyline and followed ten "good responders" for 2 y \[[@pmed-0020164-b31]\]. In 14 studies, we could not find any reference to intent-to-treat analysis. In these studies, the percentage of non-completers varied between 1% and 24%. In seven studies, all recruited patients completed the study, and an additional 13 studies specifically indicated that an intent-to-treat analysis had been performed. Of the 31 included studies, 18 were a crossover design, notably those published longer ago (see [Table 2](#pmed-0020164-t201){ref-type="table"}). The design of 14 of these studies included a "washout" period, and ten of which included data that had the effect of verifying the adequacy of the washout period (e.g., return to baseline pain intensity before a treatment period). Efficacy of Antidepressants {#s3a} --------------------------- Seven RCTs (297 participants recruited) investigated a range of tricyclic antidepressants, and six of these were of a crossover design. Five of these trials had extractable dichotomous outcome data, four of which compared amitriptyline, nortriptyline, or desipramine to placebo ([Table 3](#pmed-0020164-t301){ref-type="table"}) \[[@pmed-0020164-b18],[@pmed-0020164-b32]--[@pmed-0020164-b34]\]. One trial used either nortriptyline or desipramine (the desipramine was used as a "back-up" medication when either the adverse effects of nortriptyline were intolerable or the dose could not be increased sufficiently). The reason for this design was that the primary aim of the study was to evaluate two classes of drug (opioid and tricyclic antidepressants) in general, rather than specific drugs \[[@pmed-0020164-b18]\]. These four trials accounted for 248 patient episodes and showed significant benefit associated with tricyclic antidepressant therapy, both individual trials and as pooled data (NNT 2.64 \[2.1--3.54\]) ([Table 3](#pmed-0020164-t301){ref-type="table"}). The fifth trial was a direct comparison of amitriptyline against maprotiline and therefore could not be included in the meta-analysis, but analysis of the data revealed that amitriptyline was associated with efficacy (50% pain reduction) in 47% (15/32) of patients and maprotiline in 38% (12/32); there was no significant difference between the treatments \[[@pmed-0020164-b31]\]. There were two trials where no dichotomous data were available. In one study, amitriptyline and nortriptyline were equally effective, with approximately 50% achieving a "good" response (top two values on five-point scale) \[[@pmed-0020164-b30]\]. A parallel design trial, with small numbers of patients in each arm, compared amitriptyline, fluphenazine, and a combination of the two to placebo \[[@pmed-0020164-b35]\]. Amitriptyline showed significant benefit; however, the addition of fluphenazine made no significant improvement. Fluphenazine alone was no better than placebo. Efficacy of Gabapentinoids {#s3b} -------------------------- Two parallel group trials that examined the efficacy of gabapentin in PHN were included in the meta-analysis (559 patient episodes) \[[@pmed-0020164-b11],[@pmed-0020164-b15]\]. In both trials, the dose of gabapentin was titrated over 1--2 wk. In one study, the target dose was 3,600 mg/d, a minimum of 1,200 mg/d was acceptable, and 65% achieved the target dose, while 83% received at least 2,400 mg/d \[[@pmed-0020164-b15]\]. The other trial compared the efficacy of two fixed doses of gabapentin, 1,800 mg/d and 2,400 mg/d \[[@pmed-0020164-b11]\]. Unlike the previous study, patients unable to attain the target doses were counted as withdrawals. Both doses showed efficacy with no significant difference between the two. The pooled results for gabapentin gave a NNT (50% pain reduction) of 4.39 (3.34--6.07) ([Table 3](#pmed-0020164-t301){ref-type="table"}). We located two parallel group placebo-controlled trials in which the analgesic efficacy of pregabalin was investigated (411 patient episodes) \[[@pmed-0020164-b13],[@pmed-0020164-b20]\], both trials revealing a superiority over placebo. In one trial, a pregabalin dose of either 300 or 600 mg/d, depending on creatinine clearance, was administered as an attempt to obtain a predicted pregabalin plasma concentration in all patients \[[@pmed-0020164-b13]\]. The pooled NTT for a 50% pain reduction with pregabalin was 4.93 (3.66--7.58) ([Table 3](#pmed-0020164-t301){ref-type="table"}). Efficacy of Opioids {#s3c} ------------------- We identified three RCTs that compared a conventional opioid to placebo, two crossover studies that investigated orally administered preparations \[[@pmed-0020164-b36]\], and one crossover trial with intravenous (i.v.) morphine \[[@pmed-0020164-b37]\]. A parallel group trial compared two doses of levorphanol \[[@pmed-0020164-b21]\]. Only the two crossover RCTs that examined orally administered medication had extractable dichotomous data (211 patient episodes), both of which demonstrated efficacy. Oxycodone, titrated up to a maximum of 60 mg/d \[[@pmed-0020164-b36]\], and controlled-release morphine (mean dose 91 mg/d) (or methadone \[mean dose 15 mg/d\] as a back-up medication if the morphine was not tolerated) \[[@pmed-0020164-b18]\] were associated with greater pain relief than placebo ([Table 3](#pmed-0020164-t301){ref-type="table"}). Pooled results for opioids yielded a NNT of 2.67 (2.07--3.77). The Raja et al. study also made a direct comparison of morphine/methadone to tricyclic antidepressants; the NNT for antidepressants was 3.73 (2.43--7.99) and for morphine/methadone 2.79 (2.01--4.6) \[[@pmed-0020164-b18]\]. In one non-placebo controlled parallel group study, two doses of levorphanol were compared \[[@pmed-0020164-b21]\]. Of the 26 PHN patients who entered the study, those treated with the 0.15-mg dose of levorphanol achieved a 14% reduction in baseline pain intensity, while the 0.75-mg dose was associated with a 33% reduction. When data from the 18 PHN patients who completed the study were analysed, this reduction was 10% and 42%, for the low- and high-dose groups, respectively. The RCT that investigated i.v. morphine (0.3 mg/kg over 1 h, average dose 19.2 mg) assessed pain relief for 120 min, and morphine treatment was associated with a significant improvement in pain relief, with a VAS score of 44.9 ± 35.6 for morphine compared to 22.2 ± 32.8 for placebo \[[@pmed-0020164-b37]\]. We identified a single placebo-controlled parallel group RCT (108 patient episodes) that demonstrated the efficacy of orally administered controlled release tramadol (average titrated dose 275.5 mg/d), which yielded an NNT of 4.76 (2.61--26.97) \[[@pmed-0020164-b14]\] ([Table 3](#pmed-0020164-t301){ref-type="table"}). The wide 95% CI indicates that replication of this study is required before efficacy can be firmly stated. Efficacy of Drugs Acting at N-Methyl-D-Aspartate (NMDA) Glutamate Receptors {#s3d} ----------------------------------------------------------------------------- We were able to extract dichotomous data from three RCTs (131 patient episodes) that compared NMDA receptor antagonists to placebo, none of which demonstrated a superior efficacy over placebo ([Table 3](#pmed-0020164-t301){ref-type="table"}). Two studies were of a crossover design: one compared dextromethorphan to placebo (using benztropine as an active placebo) \[[@pmed-0020164-b38]\], and the other had three treatment arms using dextromethorphan, memantine, and lorazepam (as an active placebo) \[[@pmed-0020164-b19]\]. The third study was of a parallel group design in which GV196771 (300 mg/d), an antagonist at the glycine binding site of NMDA receptor, was found to have no superiority over placebo when the data for the primary or secondary pain outcomes measures were analysed, although it did have a reducing effect on the area of static and dynamic mechanical allodynia \[[@pmed-0020164-b39]\]. In one further parallel group study, memantine (20 mg/d) was shown not to be superior to placebo, although we were unable to extract dichotomous outcome data from this study \[[@pmed-0020164-b40]\]. Efficacy of Topically Administered Treatments {#s3e} --------------------------------------------- Two parallel group studies (175 patient episodes) compared capsaicin cream (0.075% cream tds-qds) to placebo, analysis of which yielded a pooled NNT of 3.26 (2.26--5.85) ([Table 3](#pmed-0020164-t301){ref-type="table"}) \[[@pmed-0020164-b41],[@pmed-0020164-b42]\]. The problems of maintaining adequate blinding to treatment in capsaicin studies, because of the sensations that this treatment evokes on application, are well documented. However, the choice of a parallel group design does somewhat mitigate against this deficiency. Dichotomous data were extracted from two RCTs (178 patient episodes) that compared topical anti-inflammatory preparations to placebo. A crossover trial assessed single doses of aspirin (median dose 1,000 mg), indomethacin (median dose 75 mg), and diclofenac (median dose 100 mg) in diethyl-ether against placebo \[[@pmed-0020164-b43]\]. The aspirin and indomethacin preparations were associated with pain relief, while the diclofenac preparation was not ([Table 3](#pmed-0020164-t301){ref-type="table"}). However, significant heterogeneity was detected in these studies, and combined data could not be extracted. A crossover trial of benzydamine 3% cream found no significant differences between active and placebo treatments ([Table 3](#pmed-0020164-t301){ref-type="table"}) \[[@pmed-0020164-b44]\]. Three RCTs (123 participants recruited) of topical lidocaine were identified; however, only one (64 patient episodes) had extractable dichotomous data for efficacy estimates ([Table 3](#pmed-0020164-t301){ref-type="table"}): An enriched enrolment crossover trial that included only participants who had reported moderate or greater pain relief from lidocaine patches in an open label study, compared lidocaine patch (5% lidocaine, 700 mg/patch, up to 3 patches/d) to placebo \[[@pmed-0020164-b45]\]. The primary outcome measure was "time to exit" from treatment; however, a secondary measure (highest level of pain relief sustained over at least 5 d) allowed an NNT of 2.00 (1.43--3.31) to be calculated ([Table 3](#pmed-0020164-t301){ref-type="table"}). In another study, lidocaine 5% patches were compared to two placebos (vehicle and "no patch") in a single 12-h session; the authors reported that the lidocaine patch was associated with higher pain-relief scores compared to either placebo treatment \[[@pmed-0020164-b46]\]. In another RCT, 5% lidocaine gel was applied to painful areas of skin for single sessions of 24 h under an occlusive dressing or 8 h with no dressing \[[@pmed-0020164-b47]\]. The efficacy of lidocaine gel was compared to both vehicle placebo and application of lidocaine gel to a site remote from the area of PHN, while locally applied lidocaine gel was associated with analgesia efficacy compared to placebo. There was no significant difference between placebo and remote site treatments. Efficacy of i.v. Lidocaine {#s3f} -------------------------- Although not practical for long-term therapy of PHN, treatment with i.v. lidocaine has been tested in two RCTs, and we were able to extract dichotomous outcome data from one of these: Neither lidocaine 1 mg/kg (48 patient episodes) nor 5 mg/kg (48 patient episodes), infused over 2 h were associated with superior pain relief than saline infusion in a crossover study ([Table 3](#pmed-0020164-t301){ref-type="table"}) \[[@pmed-0020164-b48]\]. The second trial was also of a crossover design and compared placebo, morphine 0.3 mg/kg, and lidocaine 5 mg/kg infused for 1 h (average total dose of lidocaine 316 mg) and revealed no significant difference between lidocaine and placebo \[[@pmed-0020164-b37]\]. Efficacy of Intrathecally (i.t.) and Epidurally Administered Drugs {#s3g} ------------------------------------------------------------------ A parallel group study (total of 270 patient episodes) was identified that compared 3% lidocaine (3 ml, i.t.) (91 patient episodes), 3% lidocaine (3 ml, i.t.), plus 60 mg methylprednisolone (i.t.) (89 patient episodes) and no treatment (90 patient episodes) \[[@pmed-0020164-b16]\]. Intrathecal injections were carried out weekly for 4 wk and patients followed up for 2 y. Analysis of extracted dichotomous data revealed efficacy associated with the lidocaine/methylprednisolone treatment (NNT 1.13 \[1.05--1.22\]), but not with i.t. lidocaine alone ([Table 3](#pmed-0020164-t301){ref-type="table"}). Another parallel group study (25 patient episodes) that compared a mixture of lidocaine 2% (3 ml) plus 60 mg of methylprednisolone administered i.t. with lidocaine 2% (5 ml) plus 60 mg of methylprednisolone administered epidurally was identified \[[@pmed-0020164-b17]\]. The lack of a placebo group dictated that this study could not be included in the meta-analysis, but the data revealed that the intrathecal treatment was associated with efficacy (50% pain reduction) in 92% (12/13) of patients whilst efficacy was seen in 17% (2/12) of patients who received the epidural treatment. If the epidural treatment is assumed to be a placebo, an NNT for the intrathecal therapy could be calculated at 1.32 (0.99--2.00). Efficacy of Miscellaneous Treatments {#s3h} ------------------------------------ Two RCTs examined the benzodiazepine lorazepam: in one crossover study (65 patient episodes), lorazepam (0.5--6 mg/d) was not shown to be superior to placebo ([Table 3](#pmed-0020164-t301){ref-type="table"}) \[[@pmed-0020164-b33]\]. Lorazepam was also used as an active placebo in a crossover study (34 patient episodes) in which two NMDA receptor antagonists and lorazepam (mean dose 1.8 mg/d) were compared; no significant pain relief was reported between the groups ([Table 3](#pmed-0020164-t301){ref-type="table"}) \[[@pmed-0020164-b19]\]. We were able to extract dichotomous data from a parallel group study (21 patient episodes) that compared acyclovir (800 mg/4 h for 12 wk) to placebo, in which no benefit was attributable to acyclovir ([Table 3](#pmed-0020164-t301){ref-type="table"}) \[[@pmed-0020164-b49]\]. We were unable to extract dichotomous efficacy data from a further two crossover studies: one multi-arm study (160 patient episodes) compared single doses of codeine 120 mg, ibuprofen 800 mg, clonidine 0.2 mg to placebo, and there was no significant difference between treatments, but clonidine had some efficacy when compared to placebo \[[@pmed-0020164-b50]\]. In the second trial (33 patient episodes), the 5HT~1~ agonists buspirone and M-chloropiperazine were found to have no benefit when compared to placebo \[[@pmed-0020164-b51]\]. Transdermal iontophoresis of 0.01% (20 ml) vincristine showed no benefit for pain relief in a parallel group study (20 patient episodes) \[[@pmed-0020164-b52]\]. We were able to extract dichotomous data from one parallel group study (62 patient episodes) that compared auricular or body acupuncture to "mock TENS" as placebo; no analgesic benefit was associated with acupuncture ([Table 3](#pmed-0020164-t301){ref-type="table"}) \[[@pmed-0020164-b53]\]. We were unable to extract dichotomous outcome data for efficacy from a parallel group RCT (25 participants recruited, 22 completed) that compared subcutaneous injection of a mixture of bovine gangliosides (Cronassial) to placebo; no superiority over placebo was demonstrated \[[@pmed-0020164-b54]\]. Adverse Events {#s3i} -------------- Of the trials from which we were able to extract dichotomous data for efficacy, we were able to extract dichotomous data for minor and major harm from 18 and 20 reports, respectively ([Tables 4](#pmed-0020164-t401){ref-type="table"} and [5](#pmed-0020164-t501){ref-type="table"}). ::: {#pmed-0020164-t401 .table-wrap} Table 4 ::: {.caption} ###### Summary of Data from Placebo-Controlled Trials for Which Dichotomous Data for Minor Harm Could Be Extracted ::: ![](pmed.0020164.t401) ::: ::: {#pmed-0020164-t402 .table-wrap} Table 4 ::: {.caption} ###### Continued ::: ![](pmed.0020164.t402) ::: ::: {#pmed-0020164-t501 .table-wrap} Table 5 ::: {.caption} ###### Summary of Data from Placebo-Controlled Trials for Which Dichotomous Data for Major Harm Could Be Extracted ::: ![](pmed.0020164.t501) ::: ::: {#pmed-0020164-t502 .table-wrap} Table 5 ::: {.caption} ###### Continued ::: ![](pmed.0020164.t502) ::: For tricyclic antidepressants, the pooled analysis revealed that 84% of participants reported minor adverse events (NNH 5.67 \[3.34---18.58\]) ([Table 4](#pmed-0020164-t401){ref-type="table"}). In addition to dizziness and sedation, the most frequently reported were anticholinergic in nature (dry mouth, constipation). However, the incidence of these varied between studies (e.g., 13--74% incidence of dry mouth) \[[@pmed-0020164-b18],[@pmed-0020164-b34]\]. Three trials noted that adverse events were more frequent during the titration phase and then reduced during the maintenance phase \[[@pmed-0020164-b18],[@pmed-0020164-b34]\]. Major adverse events: withdrawals were relatively small in number and again related mostly to sedation and other anticholinergic effects, one patient treated with desipramine developed left bundle branch block, and one treated with amitriptyline erythema multiforme \[[@pmed-0020164-b32],[@pmed-0020164-b34]\]. The combined NNH for major adverse events was 16.9 (8.85--178) ([Table 5](#pmed-0020164-t501){ref-type="table"}). The pooled results for gabapentin gave a NNH of 4.07 (3.15--5.74) for minor harm and 12.25 (7.69--30.2) for major harm ([Tables 4](#pmed-0020164-t401){ref-type="table"} and [5](#pmed-0020164-t501){ref-type="table"}). The most frequently reported adverse events were dizziness and somnolence. One death was reported in each trial---both considered to be unrelated to study medication---one in a gabapentin-treated patient and one in a placebo-treated patient. Withdrawals in the first pregabalin trial were mostly related adverse events (28 out of 31 withdrawals in the active treatment group), giving a NNH of 4.27 (2.78--9.18) and 4.86 (2.86--9.21) for minor and major harm, respectively ([Tables 4](#pmed-0020164-t401){ref-type="table"} and [5](#pmed-0020164-t501){ref-type="table"}) \[[@pmed-0020164-b13]\]. Compared to the gabapentin trials, the titration period was more rapid, over 1 wk compared to 16 d and 4 wk. In the second pregabalin trial, an 83% incidence of minor adverse events was reported in the treatment group, but the incidence of withdrawal was highest in the placebo group \[[@pmed-0020164-b20]\]. Since there was significant heterogeneity between the two studies, the NNH data for pregabalin were not combined. The most frequent adverse events associated with pregabalin were dizziness and somnolence. For both oxycodone and controlled-release morphine, constipation, nausea, and sedation were reported more often in the study medication groups than with placebo \[[@pmed-0020164-b36]\]. It was noted that nausea and sedation became less troublesome with time, but not constipation. Compared to tricyclic antidepressants, morphine was associated with a higher incidence of adverse events, in 72% compared to 36% of treated patients. The incidence of withdrawals differed between trials, with no significant difference between oxycodone and placebo, but a significant difference between morphine or methadone and placebo. The combined NNH for major harm was 6.29 (4.16--12.8); due to the detection of significant heterogeneity, we were unable to calculate a combined NNH for minor harm, but from one study \[[@pmed-0020164-b36]\], this was 3.57 (2.16--10.23) ([Table 5](#pmed-0020164-t501){ref-type="table"}). For tramadol, the NNT for major harm was calculated at 10.81 ([Table 5](#pmed-0020164-t501){ref-type="table"}). For NMDA receptor antagonists, dichotomous data were available for minor adverse events from one trial 2.53 (1.38--14.6) \[[@pmed-0020164-b39]\] and for major adverse events from two trials \[[@pmed-0020164-b38],[@pmed-0020164-b39]\] combined NNH 14.27 (NNTH of 4.8 to approximately NNTB of 14.62). No systemic adverse events were reported in the capsaicin trials, but local irritation at the site of application of capsaicin was the most frequently reported local adverse event \[[@pmed-0020164-b41],[@pmed-0020164-b42]\]. It was noted in both studies that these symptoms decreased or disappeared during the course of the trials. From the pooled data, a NNH for minor harm of 3.94 (2.5--8.6) was calculated (see [Table 4](#pmed-0020164-t401){ref-type="table"}). The most frequently reported reason for withdrawal was burning pain at the site of application. Dichotomous major adverse-events data were only extractable from one RCT that yielded an NNH of 4.67 (3.13--9.19) ([Table 5](#pmed-0020164-t501){ref-type="table"}) \[[@pmed-0020164-b42]\]. Lidocaine patches \[[@pmed-0020164-b45]\] were reported as being well tolerated with no systemic adverse events and only minor local skin reactions noted. There was no significant difference between lidocaine patch and placebo for minor adverse events. There were only three withdrawals, two in the placebo group associated with a skin reaction and worsening pain and one in the lidocaine-treated group with severe depression. For lidocaine and placebo gels, 38% of participants reported local skin reactions and 19% reacted to the Tegaderm dressing; there was no statistical difference between lidocaine and placebo \[[@pmed-0020164-b47]\]. The topical anti-inflammatory treatments were well tolerated, with a low incidence of local skin irritation (see [Table 4](#pmed-0020164-t401){ref-type="table"}) \[[@pmed-0020164-b43],[@pmed-0020164-b44]\]. Whilst we were unable to extract dichotomous adverse-events data, no minor adverse effects were reported in either epidural or intrathecal lidocaine and methylprednisolone treated patients. Serial MRI scans were unchanged in all groups, and seven patients failed to complete the 2-y follow-up period, although no reasons were given for these withdrawals \[[@pmed-0020164-b16]\]. One patient who under went epidural injections had a cerebral haemorrhage 21 wk after injections, thought to be unrelated to treatment \[[@pmed-0020164-b17]\]. Discussion {#s4} ========== This systematic review of the literature suggests that there is evidence of analgesic efficacy (i.e., NNT \< 5.00) in established PHN for the following orally administered therapies: tricyclic antidepressants, opioids, gabapentin, tramadol, and pregabalin ([Table 3](#pmed-0020164-t301){ref-type="table"}). There is also evidence that some topically administered therapies are associated with analgesic efficacy in appropriate patients: lidocaine patch and capsaicin. Intrathecal administration of lidocaine and methyl prednisolone is associated with long-lasting analgesia. However, an important proviso must be made when interpreting data derived using a meta-analytic approach; the evidence supporting some therapies have been derived from several studies, compromising a relatively large number of patient episodes (e.g., tricyclic antidepressants, gabapentinoids, and opioids), whereas others have been derived from single studies (e.g., tramadol or intrathecal methylprednisolone) and/or only a relatively low number of patient episodes (e.g., topical lidocaine and capsaicin). The data extracted from small and/or single unreplicated studies needs to be viewed with a particular degree of caution (see [Tables 2](#pmed-0020164-t201){ref-type="table"} and [3](#pmed-0020164-t301){ref-type="table"}). Heterogeneity in clinical trial design and size is a frequent Achilles\' heel of a meta-analytic approach. The European Medicines Agency has recently published guidelines for neuropathic pain studies, which hopefully, at least for regulatory studies conducted in Europe, will dictate that future trial methods will be more homogenous and, therefore, the data more comparable \[[@pmed-0020164-b55]\]. Some evidence supports the use of anticonvulsant (anti-epileptic) drugs other than gabapentinoids in neuropathic pain conditions other than PHN \[[@pmed-0020164-b10]\], but we could find no suitable trials that examined their use specifically in PHN. Furthermore, we consider the classification of drugs as "anticonvulsants" somewhat meaningless in this context, since these drugs have disparate mechanisms of action. Compliance of patients with treatment is an important factor in the clinical effectiveness of therapies, and a major factor governing compliance is withdrawal due to side effects (major harm). However, the dichotomous data for adverse events that we have been able to extract must be viewed with some caution for a number of reasons. Firstly, we were unable to extract dichotomous adverse-events data for all studies from which we were able to extract efficacy data. Secondly, the methods for collecting adverse-event data varied widely between RCTs, with some trials actively seeking adverse-event details with specific scoring systems, while others relied only on patients volunteering such information. Thirdly, as with the efficacy data, many of the trials examined only short treatment periods---sometimes only single sessions---and also examined only a relatively small population of patients, a combination of factors that seriously reduces the ability of these RCTs to detect even relatively frequent adverse events. We elected to use rates for total adverse events, as many studies did not distinguish between events related or unrelated to therapy; however, the use of total adverse events as a measure could potentially yield a more global estimate of the pattern of adverse events and remove a subjective decision regarding such an association. Despite an appreciable frequency of minor adverse events, serious events were rare; the few deaths that were reported were considered to be unrelated to study medication ([Tables 4](#pmed-0020164-t401){ref-type="table"} and [5](#pmed-0020164-t501){ref-type="table"}). Some data suggest that the following therapies are not associated with efficacy in PHN, within the dose range examined: oral administration of certain NMDA receptor antagonists (memantine, GV196771, and dextromethorphan), codeine, ibuprofen, lorazepam, 5HT~1~ receptor agonists, and acyclovir. Furthermore, topical administration of benzydamine, diclofenac/diethyl ether, and vincristine (iontophoresis) are not efficacious. Intrathecal administration of lidocaine or epidural administration of lidocaine and methylprednisolone are not associated with analgesia, nor is i.v. therapy with lidocaine or subcutaneous injection of Cronassial. Acupuncture is not associated with analgesic efficacy in PHN. However, it should be pointed out that, in contrast to the "positive" trials that contain large numbers of patient episodes, many of the trials where an effect was not demonstrated represented comparatively low numbers of patient episodes or were single-dose studies, so it may be appropriate to regard such interventions as "not yet adequately tested" rather than demonstrating "no evidence of efficacy." Furthermore, the potential confounds of variations in pharmacokinetics and assessment of only a limited dose range/regimen dictate that a "lack of efficacy" statement cannot be made with complete confidence for these interventions. In 16 publications, cutaneous allodynia was measured either qualitatively or quantitatively. Only bedside methods were used for mechanical (15 papers) and thermal allodynia (two papers). Quantitative data were presented from seven studies; the rest were purely qualitative and usually accompanied by brief comments only. In general, improvement in allodynia, or lack of it, paralleled changes in global pain scores. Tricyclic antidepressants, oxycodone, i.v. and topical lidocaine, and i.t. methylprednisolone reduced the intensity and/or area of allodynia, whereas conventional NMDA receptor antagonists did not, although there was an effect on the area of dynamic and static mechanical allodynia associated with the glycine site antagonist GV196771 \[[@pmed-0020164-b39]\]. No useful data in this respect were available for gabapentin, pregabalin, or tramadol. The percentage of pain reduction or improvement that is clinically meaningful, and can therefore be employed as a dichotomous outcome measure, is controversial: the majority of systematic reviews published to date have reported dichotomous outcome data for efficacy in terms of 50% pain relief or reduction in baseline intensity. However, this measure, as far as we are aware, has been validated for acute pain only \[[@pmed-0020164-b56]\] and not for chronic neuropathic pain. Farrar et al. have validated the use of a 30% reduction in pain intensity as a clinically important dichotomous outcome measure for chronic pain \[[@pmed-0020164-b57]\]. We therefore attempted to obtain such 30% dichotomous outcome data for comparative purposes but were able to do so for only five studies; we have not reported such data since we consider that this would not reasonably reflect the breadth of included studies. However, we suggest that in future, RCTs in PHN should report responder rates for both 30% and 50% pain reduction, although of course it could be argued that a debate concerning the precise threshold for defining a responder is somewhat sterile, since it might be expected that the actual ratio of placebo to active treatment responders might be similar no matter which threshold was used, since the same threshold would be applied to both interventions. In many studies, evaluation of overall pain was measured using the McGill Pain Questionnaire, but rarely were any additional useful data presented. Triyclic antidepressants, oxycodone, and i.t. methylprednisolone appeared to reduce brief pain paroxysms as well. Mood was assessed using validated tools (BDI, POMS, HADS) in nine studies and quality of life (mainly SF36) in six. Simple linear or categorical scales were used in ten studies for sleep interference. In general, quality of life and sleep changed in parallel with improved pain scores, whereas changes in mood were inconsistent, the latter possibly due to methodological differences. Whilst the development of clinical management guidelines for PHN, which are based upon high-quality evidence, is a highly desirable goal, such guidelines can of necessity include therapies only where evidence satisfying current criteria exists. Justification for a number of therapies in current use by some practitioners is based upon clinical experience and anecdote combined with research standards of yesteryear, and therefore such therapies have not been evaluated here. Unfortunately, since many of these therapies are of little current commercial interest to the pharmaceutical industry and do not fall within areas of interest of the major research funding organisations, it is extremely unlikely that an evidence base supporting or refuting such therapies will ever be compiled. Other considerations also influence evidence gathering relating to novel treatments: where regulatory approval requires evidence of efficacy and safety compared with placebo, it may not be commercially advantageous to undertake head-to-head studies directly comparing efficacy and safety against a current standard treatment and thus such data are sparse. Furthermore, many PHN patients take more than one therapy, and there is only a very limited evidence base of the additive and synergistic effects of combing therapies. Nevertheless, from available evidence it is possible to produce guidelines for management, but only with the caveat that some therapies are excluded as they are not supported by high-quality evidence. Lack of evidence may be because adequate trials have shown no benefit or because no adequate trial has been undertaken. A further problem arises from the fact that there is no single pathophysiology that defines the generation and persistence of PHN \[[@pmed-0020164-b03]\]. Future studies should use quantitative sensory evaluation to clearly categorise subsets of patients contained within a study population and be powered to permit separate evaluation of the data obtained for such subsets. It is important to include economic considerations in development of a clinical management strategy, although it is recognised that this will vary between health-care systems. For example, where efficacy and adverse events of two drugs are similar, it would seem prudent to initiate treatment with the less expensive drug. For example, a treatment plan for oral therapy that proposes initial use of a tricyclic antidepressant, except where contra-indicated, might reserve the use of the more expensive gabapentinoids for patients, where the tricyclic antidepressant either fails to provide efficacy and/or is associated with unacceptable adverse events. Similarly, the efficacy of various opioids and tramadol supports their use, but perhaps as a second-line therapy in accordance with recommendations for the use of such drugs in chronic non-cancer pain \[[@pmed-0020164-b58]\]. With the above provisos, and using the evidence evaluation of this systematic review, the evidence base would support the first-line use of a tricyclic antidepressant for orally administered treatment of PHN, reserving the gabapentinoids for second-line use. However, a secondary proviso here might be a consideration of the wisdom of using, as first-line therapy, a group of drugs that have regulatory approval in many countries for the treatment of PHN (e.g., gabapentinoids) against a group that generally do not (e.g., tricyclic antidepressants). On the efficacy evidence, "strong" opioids could be considered as first- or second-line therapy, although the guidelines for use of opioids in non-malignant pain would suggest that opioids might be sensibly reserved for use following inadequate benefit from tricyclic antidepressants or gabapentinoids \[[@pmed-0020164-b58]\]. Logically, topical treatments might be considered to have a lower potential for generating systemic adverse effects than systemically administered therapies; therefore, the early use of topical lidocaine or capsaicin should be considered as a first line, especially where quantitative sensory evaluation (possibly supported by measurement of epidermal neuronal density in a skin biopsy \[[@pmed-0020164-b59]\]) has indicated that the patient falls into the "sensitised nociceptor" as opposed to "deafferentation" sub-group of PHN patients \[[@pmed-0020164-b03]\]. Intrathecal steroids appear to be associated with remarkable benefit in PHN patients \[[@pmed-0020164-b16]\], but this therapy may be potentially hazardous \[[@pmed-0020164-b60]--[@pmed-0020164-b62]\], and this trial has not yet been replicated. Therefore, we believe that a further high-quality RCT of this therapy is desirable before recommendations can be made regarding its use for PHN. This therapy is, of course, not suitable for PHN within the cranial nerves innervation territory. There is little evidence regarding possible synergistic effects of the various treatments to support or refute the concomitant use of combinations of, e.g., tricyclic antidepressants, opioids, and gabapentinoids. However, it certainly seems logical that concomitant use of drugs with different mechanisms of action may offer additional benefit to PHN patients. Although there is little direct evidence to support it, we would like to make a final recommendation that any treatment plan should recognise the importance of the biopsychosocial model of chronic pain and thus that any pharmacologically based management of PHN should be combined with advice on and management of psychological and social aspects \[[@pmed-0020164-b63]\]. Patient Summary {#sb1} --------------- ### Background {#sb1a} Postherpetic neuralgia (PHN) is the pain that people sometimes get after shingles. It can be severe. Although many treatments have been tried for it, doctors do not agree on how to best treat it. ### What Did the Researchers Do? {#sb1b} They looked systematically to find all the trials that have investigated treatments for PHN. They assessed each trial to see if it could provide useful results---e.g., if it was well designed, clear that they were treating patients with PHN, and that clear results could be taken from the trials. They also looked to see if the trials had assessed the possibility that the treatments could cause harm. They found that there were a wide range of treatments that had been tried. The most reliable oral treatments were tricyclic anti-depressants, morphine-like drugs (opioids), and gabapentin (and related drugs). Some topical treatments also worked: e.g., a local anaesthetic lidocaine and a cream made from the active ingredient of chilli peppers---capsaicin. However, for many treatments there was not enough evidence to assess if they worked. ### What Do These Results Mean? {#sb1c} This type of review is the most reliable form of evidence that doctors have available to them in deciding on treatment. Even so, the results are not conclusive. Future trials should be designed rationally to fill in the gaps of knowledge about the possible treatments for this disorder. In the meantime, however, there are some drugs that seem to work relatively well, and, outside of a clinical trial, these drugs should be used first. ### Where Can I Get More Information? {#sb1d} Medline Plus discusses shingles and neuralgia more widely: <http://www.nlm.nih.gov/medlineplus/ency/article/000858.htm> The Neuropathy Trust has a Web site with patient information: <http://www.neurocentre.com> The Varicella Zoster Research Foundation Web site also has useful information for patients: <http://www.vzvfoundation.org> ASCR has been paid for providing services (e.g., consultancy advice, lecture honoraria) and received travel expenses and entertainment from several pharmaceutical companies with interest in this area, e.g.: Pfizer, Boehringer-Ingelheim, GSK, Novartis, Vernalis, OHNO Pharma, Ionix, UCB Pharma, Organon and Pierre Fabry. ASCR is a member of the editorial board of PLoS Medicine. TJN has received consultancy fees, honoraria, and travel expenses from the following companies: GSK, Pfizer, Novartis, SchwarzPharma, AstraZeneca, Ortho McNeil, Napp, Endo, and Renovis. RWJ has received consultancy fees from Merck, Novartis, Menarini, and Yamanouchi and travel grants from Novartis. The authors recieved no funding for this study. Citation: Hempenstall K, Nurmikko TJ, Johnson RW, A\'Hern RP, Rice ASC (2005) Analgesic therapy in postherpetic neuralgia: A quantitative systematic review. PLoS Med 2(7): e164. CI : confidence interval i.t. : intrathecally i.v. : intravenous NMDA : N-methyl-D-aspartate NNH : number needed to harm NNT : number needed to treat NNTB : number needed to treat to benefit NNTH : number needed to treat to harm PHN : postherpetic neuralgia RB : relative benefit RCT : randomised controlled trial RR : relative risk [^1]: Competing Interests: Please see Acknowledgments. [^2]: Author Contributions: KH and ASCR designed the study. KH, TJN, RWJ, RPA, and ASCR analyzed the data. KH, TJN, RWJ, RPA, and ASCR contributed to writing the paper.	PubMed Central	2024-06-05T03:55:59.858845	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181872/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e164", "authors": [ { "first": "Kathleen", "last": "Hempenstall" }, { "first": "Turo J", "last": "Nurmikko" }, { "first": "Robert W", "last": "Johnson" }, { "first": "Roger P", "last": "A'Hern" }, { "first": "Andrew S.C", "last": "Rice" } ] }
PMC1181873	Introduction {#s1} ============ The past decade has seen a dramatic increase in the significance attached to infectious diseases from the public health perspective. This trend is due in part to the emergence of new and highly pathogenic infections such as Ebola \[[@pmed-0020174-b01]\], West Nile virus \[[@pmed-0020174-b02]\], and SARS \[[@pmed-0020174-b03]\]. There are also well-publicized concerns surrounding the deliberate introduction of pathogens as bioterrorism weapons \[[@pmed-0020174-b04],[@pmed-0020174-b05]\], and the continued persistence and resurgence of older infections, several of which now boast strains resistant to more than one drug \[[@pmed-0020174-b06]\]. In addition, there have been a number of high-profile and economically expensive disease outbreaks in domestic livestock \[[@pmed-0020174-b07]--[@pmed-0020174-b09]\] as well as wildlife populations \[[@pmed-0020174-b10]\]. The effective management and control of such infections is increasingly done with substantial input from mathematical models, which are used not only to provide information on the nature of the infection itself, through estimates of key parameters such as the basic reproductive ratio R ~0~ \[[@pmed-0020174-b11]\], but also to make predictions about the likely outcome of alternative courses of action \[[@pmed-0020174-b12]--[@pmed-0020174-b15]\]. During the 2001 outbreak of foot-and-mouth disease in the United Kingdom, for example, the former UK Ministry of Agriculture, Fisheries, and Food set up a committee that included two groups with expertise in mathematical modeling of disease dynamics. It is becoming increasingly important, therefore, that epidemiological models produce accurate quantitative predictions, and this in turn relies on accurate parameterization. Here, we examine the dynamical consequences of an unrealistic yet almost ubiquitous assumption embedded in such models concerning the distribution of the latent and infectious periods. In particular, we show that without greater care in model formulation, we may risk systematic biases when parameterizing models from data and may make overly optimistic policy recommendations. The most commonly used framework for epidemiological systems, is still the susceptible--infectious--recovered (SIR) class of models, in which the host population is categorized according to infection status as either susceptible, infectious, or recovered \[[@pmed-0020174-b16],[@pmed-0020174-b17]\]. Subsequent refinements of the model have incorporated an additional exposed (infected but not yet infectious) class (susceptible--exposed--infectious--recovered \[SEIR\] models) (see [Protocol S1](#sd001){ref-type="supplementary-material"} for mathematical equations). One of the fundamental mathematical assumptions in such models is that the rate of leaving the exposed or infectious class is constant, irrespective of the period already spent in that class. While mathematically very convenient, this assumption gives rise to exponentially distributed latent and infectious periods, which is epidemiologically unrealistic for most infections \[[@pmed-0020174-b18]--[@pmed-0020174-b20]\]. A more sensible formulation would be to specify the probability of leaving a class as a function of the time spent within the class, such that initially the chance of leaving the class is small, but the probability increases as the mean infectious/latent period is reached. This would give rise to a more realistic distribution of latent and infectious periods, with a stronger central tendency. A convenient way to describe such distributions is to write an expression for the infectious class (neglecting the latent class for this example) as follows: which translates mathematically into where β is the infection transmission rate and N is the total population size. The probability of remaining infectious through time is governed by the survivorship function, P(t), and as such the average infectious period, denoted by 1/γ, is given by \[[@pmed-0020174-b21]\]. The probability density function of the infectious period, p(t), is just the negative derivative of the survivorship function, −dP(t)/dt. Different functional forms for p(t) give rise to alternative assumptions concerning the distribution of the infectious period in the model. For example, setting p(t) equal to e^−γt^/γ recovers the classical exponentially distributed SIR model. More realistic distributions can be obtained by choosing p(t) to be a gamma probability density function \[[@pmed-0020174-b22]--[@pmed-0020174-b27]\], with parameters γ and n (see [Protocol S1](#sd001){ref-type="supplementary-material"}). An alternative (and computationally efficient) means of modeling infections with gamma distributions is to divide the infectious class into n subclasses with nγ being the rate of sequential progression through the subclasses. The advantage of this formalism is that when n = 1 we recover the exponentially distributed model, which has a large variance, while as n→∞ we obtain a fixed infectious period. The effects of n on the distribution of the infectious period are demonstrated in [Figure 1](#pmed-0020174-g001){ref-type="fig"}A; in [Table 1](#pmed-0020174-t001){ref-type="table"} we present some examples of latent and infectious period distributions estimated from data. ::: {#pmed-0020174-g001 .fig} Figure 1 ::: {.caption} ###### Gamma-Distributed Infectious Periods and Their Effects on the Epidemic Curve \(A) The change in the probability of remaining infectious as a function of time when the number of subdivisions within the infected class increases from n = 1 to n = 100. Irrespective of the value of n, the mean duration of the infectious period is 1 wk. When n = 1, the distribution of the infectious period is exponential, but as n increases the infectious period becomes closer to a constant length. \(B) The consequences of changes in n for the SIR-type epidemic without births or deaths. For the same basic reproductive ratio, R ~0~ = 5, and the same average infectious period, γ = 1, larger values of n lead to a steeper increase in prevalence and an epidemic of shorter duration. ::: ![](pmed.0020174.g001) ::: ::: {#pmed-0020174-t001 .table-wrap} Table 1 ::: {.caption} ###### Estimates of Latent and Infectious Period Durations Estimated from Data for a Number of Infections, Together with the Associated Gamma-Distribution Parameter (m and n) ::: ![](pmed.0020174.t001) ::: The dynamical consequences of these differences in the distribution of infectious and latent periods have received some attention over the past two decades. It has been shown, for example, that the precise distribution of the infectious period has no qualitative effects on the asymptotic values or properties of the system \[[@pmed-0020174-b21],[@pmed-0020174-b24]\], though perturbations to the endemic equilibrium take longer to die out as n increases \[[@pmed-0020174-b22],[@pmed-0020174-b26]\]. When contact rates vary seasonally, for example, to mimic the aggregation of children in schools \[[@pmed-0020174-b28],[@pmed-0020174-b29]\], changes in p(t) are known to have important consequences for the persistence likelihood of infections \[[@pmed-0020174-b25],[@pmed-0020174-b26],[@pmed-0020174-b30]\]. An issue that has received surprisingly little attention, despite its obvious applicability to emerging infections and possible "deliberate exposure," is the influence of latent and infectious period distributions on the invasion dynamics of an infection into a largely susceptible population. This is in contrast to the conceptually similar situation of within-host dynamics of viral disease, such as HIV, for which some models already adopt realistic distributions to describe stages in the cell life cycle \[[@pmed-0020174-b31],[@pmed-0020174-b32]\]. In particular, Lloyd \[[@pmed-0020174-b33]\] has shown how parameter estimates made from viral load data are affected by different assumptions about these distributions. Here we are interested in the application of this work to between-host transmission dynamics. As can be seen in [Figure 1](#pmed-0020174-g001){ref-type="fig"}B, changes in the gamma distribution parameter n have substantial quantitative consequences for the epidemic curve: in comparison to a gamma-distributed model, the epidemic given by the exponentially distributed model (i) takes off at a dramatically slower rate, (ii) predicts a significantly smaller (approximately 56%) peak number of cases, and (iii) lasts much longer (almost twice as long). Whether these marked differences between alternate model formulations may translate into potentially important public health concerns is a key question, which we address in two ways. First, we document systematic differences in the model parameters estimated from an epidemic using the exponential and gamma-distributed models. Second, we demonstrate that the use of exponential models produces overoptimistic predictions about the low levels of control required to subdue an epidemic. Methods {#s2} ======= The Relationship between R ~0~ and Initial Epidemic Growth Rate {#s2a} ----------------------------------------------------------------- During the early phase of an epidemic, the observed exponential growth rate, λ, is related to the basic reproductive ratio, R ~0~, of the infection. Mathematically, λ is just the dominant eigenvalue of the disease-free equilibrium, and one can show that λ must satisfy for the SEIR model with gamma-distributed latent and infectious periods (further details are given in [Protocol S1](#sd001){ref-type="supplementary-material"}). This equation translates into an expression for R ~0~ in terms of λ and the other parameters, as is presented in [equation 4](#pmed-0020174-e004){ref-type="disp-formula"}. Therefore, if we can estimate the growth rate λ from data, we can use [equation 4](#pmed-0020174-e004){ref-type="disp-formula"} to obtain an estimate of R ~0~. Contact Tracing and Isolation {#s2b} ----------------------------- To study the effects of contact tracing and isolation, we modify the assumptions of the SEIR epidemic model, while still incorporating gamma-distributed latent and infectious periods. In the new model, isolation of newly infectious cases occurs at a daily rate of d~I~ after a delay of τ~D~ days, which represents a period when infected individuals are infectious but asymptomatic or undetectable (I~A~). A fraction q of those who had contact with an infectious and symptomatic individual (I~S~) (but did not contract the infection) are removed to the quarantined susceptible class, S~Q~, where they spend exactly τ~Q~ days. An identical fraction of newly exposed individuals is also quarantined. Full details of the model equations are given in [Protocol S1](#sd001){ref-type="supplementary-material"}. Results {#s3} ======= In a typical management setting, such as the SARS outbreak of 2003, public health professionals are confronted with a novel (or perhaps a highly virulent variant of an existing) pathogen that is spreading rapidly through a predominantly susceptible population. One of the important tasks of any modeling exercise is to provide insights into some of the epidemiological characteristics of the invading infection, such as its transmissibility, virulence, and persistence dynamics. Of great interest is the estimation of the basic reproductive ratio of the infection (R ~0~), which measures the transmission potential of the infection, and determines the degree of control required. Some of these aspects can be explored by studying the range of model parameters that give (initial) outbreak dynamics consistent with the (short-term) epidemic data thus far gathered. One approach is to fit model parameters to data by "trajectory matching," where one looks for the combination of parameters that, in a statistically rigorous sense, give rise to dynamics most consistent with observed patterns \[[@pmed-0020174-b34],[@pmed-0020174-b35]\]. Alternatively, one may use the well-established result that during the initial stages of an epidemic, the growth rate is approximately exponential \[[@pmed-0020174-b17],[@pmed-0020174-b20]\], with the rate determined by R ~0~. First, we use this approach to examine, in general, how the distribution of the latent and infectious periods may affect the estimation of R ~0~ from initial epidemic data. To illustrate that our results are not specific to this methodology, we then take incidence data from an influenza outbreak and parameterize the epidemic models using trajectory matching. Estimating R ~0~ from Initial Epidemic Growth Rate {#s3a} ---------------------------------------------------- We may obtain an estimate for R ~0~ by calculating the initial growth rate (λ) of an infection from data and equating it to the growth rate of the equations, calculated from the dominant eigenvalue of the disease-free equilibrium ([see Methods](#s2){ref-type="sec"}). Such an exercise reveals that for any observed λ, the precise value of R ~0~ estimated is crucially dependent on the fundamental assumptions made concerning the distributions of latent and infectious periods. Specifically, we find that the following equation determines the relationship between R ~0~ and an empirically estimated epidemic growth rate, λ: where m and n represent the number of subclasses in the exposed and infectious categories, respectively. The mean latent and infectious periods are represented by 1/σ and 1/γ, respectively, and are assumed to be known or estimated from independent data. This relationship was first determined by Anderson and Watson \[[@pmed-0020174-b36]\] and has recently been applied in the context of viral life cycle dynamics by Lloyd \[[@pmed-0020174-b33]\]. The relationship between estimated R ~0~ and the distributions of the latent and infectious periods is demonstrated in [Figure 2](#pmed-0020174-g002){ref-type="fig"}A. It reveals a subtle yet very important interaction between model structure and estimated R ~0~. In general, as the infectious period becomes more tightly distributed (increasing n), lower values of R ~0~ are estimated for any given growth rate λ. On the other hand, as the variance in the latent period is reduced (increasing m), higher values of R ~0~ are estimated. Indeed, we may use the relationship given by [equation 4](#pmed-0020174-e004){ref-type="disp-formula"} to arrive at the following general principle: if we ignore the latent period, then models with an exponentially distributed infectious period will always overestimate the infection\'s basic reproductive ratio. When the latent period is included, however, this finding is reversed when the growth rate is large ([Figure 2](#pmed-0020174-g002){ref-type="fig"}B). In closely examining [equation 4](#pmed-0020174-e004){ref-type="disp-formula"}, we note that the basic reproductive ratio estimated from a model without an exposed class (1/σ = 0) is always lower than the estimate from the corresponding model when a latent period is included (1/σ \> 0) (see equation S1 in [Protocol S1](#sd001){ref-type="supplementary-material"}). Therefore, when faced with a rapidly spreading infection, either entirely ignoring the latent period or assuming exponential distributions will lead to an underestimate of R ~0~ and therefore will underestimate the level of global control measures (such as mass vaccination) that will be needed to control the epidemic. ::: {#pmed-0020174-g002 .fig} Figure 2 ::: {.caption} ###### Estimates of R ~0~ from Data on the Initial Growth Rate of an Epidemic \(A) The effects of changing the distributions of the latent and infectious periods on the estimated value of R ~0~, with λ assumed to be 100 per year and the average latent and infectious periods fixed at 1 wk. The gray grid surfaces show the asymptotic values of R ~0~ when the latent and infectious periods are both exponentially distributed (lower surface) or fixed (higher surface). We note that the shape of each surface is independent of the exact value of λ. \(B) At higher values of λ, R ~0~ may be substantially over- or underestimated using the classical exponentially distributed model (n = m = 1) compared to periods of fixed lengths (n = m→∞), depending on whether an exposed class is included (solid lines) or not (dashed lines). ::: ![](pmed.0020174.g002) ::: Estimating R ~0~ from Trajectory Matching {#s3b} ------------------------------------------- While the results described in the previous section are based on the rate of epidemic take-off, we reach the same qualitative conclusions about the effects of the distributions of latent and infectious periods when estimating R ~0~ by other data-fitting methods. For illustration, we use data from an influenza outbreak in an English boarding school \[[@pmed-0020174-b37]\] to estimate model parameters by trajectory matching. In the absence of independent data, this method can be used to provide estimates of the key infectious parameters. Of course, here we can also compare the parameter estimates to observed parameter ranges, since the influenza virus is known to have a latent period of between 1 and 4 d and infected individuals may transmit the virus up to 4 or 5 d after the onset of illness \[[@pmed-0020174-b38]\]. We determine the best fit of the model output to daily incidence data by minimizing the least squares errors for different values of the distribution parameters m and n. For comparison, we also determine the best-fit parameters in the absence of any latent period. The least squares errors and estimated R ~0~ of the best fit for a combination of m and n are presented in [Figure 3](#pmed-0020174-g003){ref-type="fig"}. These results clearly illustrate the points raised in the previous section: (i) entirely ignoring the latent period gives a significantly lower estimate of R ~0~, and (ii) the assumption of exponentially distributed latent and infectious periods results in consistently lower estimates of R ~0~ than their gamma-distributed counterparts. ::: {#pmed-0020174-g003 .fig} Figure 3 ::: {.caption} ###### Fitting Epidemic Models to Data from an Influenza Outbreak (A and B) The least squares error (LSE) (A) and R ~0~ (B) of the best-fit model under different assumptions about the distribution of the latent and infectious periods. (The label "w/o" denotes no latent class.) \(C) We plot the incidence data along with the SEIR best fit (m = 2, n = 2) and that obtained by ignoring any latent period (n = 1)---the SIR best fit. \(D) The best-fit estimate of R ~0~ changes for these two models as we increase the number of points used in the fitting procedure. When fitting the models, for each value of n (and m), we are estimating the average infectious period, 1/γ, and transmission parameter, β (and average latent period, 1/σ). The effective population size for the influenza outbreak was known to be N = 763. ::: ![](pmed.0020174.g003) ::: Despite visually similar solutions, the SIR best-fit and SEIR best-fit models ([Figure 3](#pmed-0020174-g003){ref-type="fig"}C) result in strikingly different estimates of R ~0~: 3.74 for the SIR model versus 35.9 for the SEIR model, which is partly a result of the small population size. However, the best-fit estimate of R ~0~ obtained from the gamma-distributed SEIR model (m = 2, n = 2) is much more sensitive to the number of points used to obtain the fit ([Figure 3](#pmed-0020174-g003){ref-type="fig"}D): the exponentially distributed SIR model gives the same estimate whether the first six (up to the peak in incidence) or more points are used. This difference is further emphasized if we use the first few points of the data to estimate the rate of epidemic take-off (λ = 1.0837 d^−1^), and then take the final estimates of the average latent and infectious periods to compute R ~0~ using [equation 4](#pmed-0020174-e004){ref-type="disp-formula"}. For the SIR model (n = 1, 1/γ = 2.2 d) we obtain an R ~0~ of 4.38 whereas for the SEIR model (m = 2, n = 2, 1/σ = 2.6 d, 1/γ = 2.1 d) we obtain an R ~0~ of 16.9. Thus the initial rate of increase in incidence does well in estimating R ~0~ for the exponentially distributed SIR model but significantly less well for the gamma-distributed SEIR model. Given that we are fitting an additional parameter, it is to be expected that a limited number of data points confounds the estimation of R ~0~ when we include a latent period in the model assumptions. However, this also highlights that even when incorporating a latent period, estimates of R ~0~ based on the initial epidemic growth rate may potentially underestimate the true value of R ~0~. Management Consequences {#s3c} ----------------------- The results outlined above highlight the pitfalls of making a priori assumptions concerning the distributions of latent and infectious periods when estimating parameters. Depending on the precise details, inappropriate model selection may give rise to either gross over- or underestimates for the basic reproductive ratio of an infection. However, even when parameter estimates are reliable, choice of model structure can also be very important when making recommendations concerning individual-level control strategies. Historically, it has been shown that contact tracing and the effective quarantine of infected individuals and those potentially exposed is an important means of infection management \[[@pmed-0020174-b13],[@pmed-0020174-b39],[@pmed-0020174-b40]\]. We introduce both these measures into the SEIR epidemic model, assuming that there is a small delay in detecting newly infectious individuals, which may represent an asymptomatic phase or uncertainty in diagnosing symptoms ([see Methods](#s2){ref-type="sec"} and [Protocol S1](#sd001){ref-type="supplementary-material"}). As we show in [Figure 4](#pmed-0020174-g004){ref-type="fig"}, the precise levels of isolation of infected individuals and of quarantining contacts required to control the outbreak and the predicted level of disease incidence are crucially affected by whether the classic exponentially distributed SEIR model or a more realistic framework is used. ::: {#pmed-0020174-g004 .fig} Figure 4 ::: {.caption} ###### The Predicted Effectiveness of Contact Tracing and Isolation of Infected Individuals in a Population of 1 Million Susceptible Individuals \(A) The proportion of the population contracting an introduced infection is depicted as a function of the infected isolation rate (d~I~) and the infectious period (1/γ). \(B) The consequences of contact tracing. In both, the surfaces represent predictions of the SEIR model with an exponential (colored surface) or gamma (black grid surface; m = n = 10) distribution of the latent and infectious periods, respectively. Model parameters: β = 0.5 per day, 1/σ = 5 d, τ~Q~ = 10 d, and τ~D~ = 2 d. In (B), 1/λ = 10 d. ::: ![](pmed.0020174.g004) ::: The process of isolating infected individuals results in a reduction in the mean infectious period (see [Protocol S1](#sd001){ref-type="supplementary-material"}). It is much more effective when the infectious period is exponentially distributed because it essentially truncates the tail of the distribution, so that the infectious period of a few individuals is dramatically reduced. This effect is not as pronounced in the gamma-distributed models because there is less variation in the infectious periods (see [Figure 1](#pmed-0020174-g001){ref-type="fig"}A). In the same way, a longer delay in detecting infected individuals has fewer consequences for the exponentially distributed model because during this time many individuals will have naturally left the infectious class. Under the assumption of a gamma-distributed infectious period most individuals are infectious for a minimum period of time so early detection is more important. While the predicted difference between the exponential and gamma-distributed models depends on the duration of the infectious period and the fraction of contacts traced, it is generally true that models with an exponentially distributed infectious period will give rise to overly optimistic predictions concerning the effectiveness of isolating infected individuals. To focus on the effects of the infectious period distribution on different courses of intervention we have assumed that all those who are quarantined and exposed are detected before the end of the quarantine period and are not released back into the general population. We have also formulated a model that takes into account the distribution of the latent period during quarantine and find similar qualitative results to those shown in [Figure 4](#pmed-0020174-g004){ref-type="fig"}. However, if the average latent period is increased relative to the fixed quarantine period and there is only a small amount of isolation of infected individuals, then the control measures are predicted to be more effective for the gamma-distributed model, because more exposed individuals in the exponentially distributed model will leave quarantine before they develop the infection. Discussion {#s4} ========== The use of models in epidemiology dates back almost a century, and while traditional models have often been highly successful in explaining observed dynamics \[[@pmed-0020174-b17],[@pmed-0020174-b20],[@pmed-0020174-b28],[@pmed-0020174-b29],[@pmed-0020174-b41]\], our results show that within a strict management setting, epidemiological details can make a crucial difference. Although a body of theoretical work \[[@pmed-0020174-b25],[@pmed-0020174-b26],[@pmed-0020174-b30]\] has demonstrated the importance of incorporating realistic distributions of latent and infectious periods into models of endemic disease, few studies have considered the effects associated with making predictions for an emerging disease \[[@pmed-0020174-b42]\]. The large discrepancies between estimates of R ~0~ from the exponentially distributed and gamma-distributed fits reiterate the importance of accurately determining the precise distributions of latent and infectious periods. Although the data required for such a task are often available from post hoc analyses of epidemics they are certainly lacking for a novel emerging infection. Instead, the uncertainty surrounding assumptions about the distributions should be incorporated into quantitative predictions made from epidemiological models, especially since this may well be greater than any uncertainty that arises from noise in the data. Of course, more sophisticated fitting methods than those used in this paper exist \[[@pmed-0020174-b43]--[@pmed-0020174-b46]\], but if the underlying structure of the model is inappropriate, the method of parameterization is largely irrelevant. The take home message from our work is that when developing models for public health use, we need to pay careful attention to the intrinsic assumptions embedded within classical frameworks. While some practitioners are already using the approach we advocate \[[@pmed-0020174-b03],[@pmed-0020174-b15],[@pmed-0020174-b34],[@pmed-0020174-b39],[@pmed-0020174-b47]\], the vast majority of applied epidemiological studies still use models that incorporate exponentially distributed latent and infectious periods. Perhaps this work points to the next steps in delivering quantitatively accurate epidemiological models. Supporting Information {#s5} ====================== Protocol S1 ::: {.caption} ###### Further Details and Analysis of the Mathematical Models (7 KB TEX). ::: ::: {.caption} ###### Click here for additional data file. ::: Patient Summary {#sb1} --------------- ### Background {#sb1a} When a new infectious disease emerges, such as SARS, it is important to try to predict how the disease will behave, e.g., how infectious it is and what its latent period is, so that the spread of the disease through the population can be estimated and appropriate public health measures such as quarantining can be decided on. ### What Did the Authors Do? {#sb1b} They assessed different currently used mathematical models of disease outbreaks, including models that took no account of latent periods, and another that assumed that the latent and infectious periods had a particular pattern---called exponential. They showed that both of these assumptions could potentially lead to underestimating the way the disease spreads. They then tested their predictions on a known outbreak of influenza that occurred in a school. ### What Do These Findings Mean? {#sb1c} Public health officials may need to rethink the way that they try to predict outbreaks of infectious disease. Minimally, they need to be sure that they put into any model the most accurate predictions of the behavior of the disease. ### Where Can I Get More Information? {#sb1d} The Health Protection Agency in the United Kingdom has a Web site that explains its work on assessing infectious disease outbreaks: <http://www.hpa.org.uk/infections/default.htm> The Centers for Disease Control and Prevention in the United States is a good place to start for information on any new infectious diseases: <http://www.cdc.gov/> We are grateful to Andy Dobson for comments that helped to improve the manuscript. HJW and PR are supported by the National Institutes of Health, and MJK is supported by the Royal Society. PR would also like to thank the Ellison Medical Foundation for funding. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Citation: Wearing HJ, Rohani P, Keeling MJ (2005) Appropriate models for the management of infectious diseases. PLoS Med 2(7): e174. SEIR : susceptible--exposed--infectious--recovered SIR : susceptible--infectious--recovered [^1]: Competing Interests: The authors have declared that no competing interests exist. [^2]: Author Contributions: PR conceived the idea. HJW, PR, and MJK designed the study. HJW formulated and analyzed the models. PR, HJW, and MJK contributed to writing the paper.	PubMed Central	2024-06-05T03:55:59.863126	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181873/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e174", "authors": [ { "first": "Helen J", "last": "Wearing" }, { "first": "Pejman", "last": "Rohani" }, { "first": "Matt J", "last": "Keeling" } ] }
PMC1181874	I appreciate Zlotkin and colleagues\' years of work on the Sprinkle product, and it sounds like the product is much improved from the pill form of micronutrient treatments \[[@pmed-0020188-b1]\]. I\'m not at all opposed to Sprinkle-type products or other nutrient pills for treatment (or in other special situations), but as ten years of international work experience on food and nutrition security issues has shown me, few programmes are supporting local solutions to problems. Once again, a message is being sent that nutrition comes from a pill or packet, made by a foreigner, and requires money. In the case of Sprinkles, the product could support local solutions by including a message on each sachet about the importance of eating a wide variety of local foods---or a picture of local fruits, vegetables, and legumes. Instead of just sprinkling a packet onto a bulky carbohydrate food, use the Sprinkles as treatment along with instruction about planting and eating less of that bulky carbohydrate in the first place. Even better would be to take all that research, time, energy, and money to teach people (or local manufacturers) how to make their own Sprinkles from local nuts, fruits, greens, oilseeds, insects, fish, and the like. The results could be just as immediate and dramatic, but with an impact that could last for generations to come. The organisations that support this type of permanent intervention could be mentioned during every teaching session along with big banners and flyers that announce them as the inventors and/or supporters. Just imagine a nice sprinkle powder that everyone can have on hand to improve their own nutrition without relying on a packet from an outside source that is manufactured with machines and jetted in with thousands of litres of petrol (or trucked across the country, if it is made in country). I\'m sure that pre-packaged, imported products have their place in wars, tsunamis, a few cities, and other disasters, but for the majority of the 750 million children in the developing world, their own indigenous foods would have just as much effect, with a longer-term impact on the society\'s nutritional health. I saw Zlotkin\'s presentation on Sprinkles at the International Congress of Dietetics conference in Chicago, Illinois, in 2004, and he did include a sentence about diversifying diets as part of the whole project, but it was strongly overshadowed by discussion of bringing in external resources and experts. When I asked him about using the same resources that went into developing, manufacturing, and transporting Sprinkles to create a local sprinkle product with an emphasis on local diversified diets, he immediately responded that it wouldn\'t work. How do we know, if no one really puts the effort into it at the level that products like Sprinkles get? I\'ve posted this message to several food and nutrition listservs and magazines, and I am now beginning to learn of some small projects working towards local sprinkle products. Zlotkin and team could assist these projects to research the work and scale it up to other countries with other local foods. Citation: Nordin S (2005) Sustainable super-sprinkle: Powdered local foods. PLoS Med 2(7): e188. [^1]: Competing Interests: The author has declared that no competing interests exist.	PubMed Central	2024-06-05T03:55:59.865518	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181874/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e188", "authors": [ { "first": "Stacia", "last": "Nordin" } ] }
PMC1181875	Shah Ebrahim says, in his answer to my statement in our debate \[[@pmed-0020189-b1]\], that plaque is so common that it makes sense to treat only those likely to have an acute event in the near future. Not everyone has a heart attack but everyone ages, nor can we be sure who will be lucky as they age and who will not. Anyone with a low-density lipoprotein level above 70 mg/dl is at risk \[[@pmed-0020189-b2]\]. Narrowing of arteries must certainly contribute to health decline in aging. He says that most plaque is stable and does not rupture. The half-million people who have a heart attack and the hundreds of thousands of individuals who have a stroke in the United States alone each year would disagree. Waiting decades more for further studies will not help all those now succumbing to the disease, as long as all the components have already been well vetted, as they have been. Ebrahim\'s statement that "most \[plaques\] are stable and unlikely to rupture" is based on his own study published in Stroke \[[@pmed-0020189-b3]\], but the paper has no data relating to how many patients found to have increased intima-media wall thickness (IMT) or plaque actually had heart attacks or strokes. It did find that intimal-wall changes were highly associated with ischemic heart disease. It is correct that risk factors can predict heart disease and stroke to some degree, though I have many patients for whom that approach fails while IMT detects the risk otherwise missed. However, risk factor analysis is not enough. Changes in smoking, diet, weight, blood pressure, and such are all targets for treatment, but how do doctors know whether they have controlled them adequately? If IMT increases, more control is required. If IMT stabilizes or reverses a little, then it means that control is at goal. Family history of premature coronary artery disease is a major but underused risk factor. The causative factors have not yet been elucidated by research, so we do not know what the goals are. However, IMT provides a highly satisfactory parameter by which to judge the effectiveness of treatment in familial coronary artery disease: IMT stabilization and reversal are good, whereas progression is bad and requires more intensive measures. Epidemiology and public-health planning correctly look at policies to apply to large populations. However, the practice of medicine is patient by patient, accomplished in the face-to-face doctor--patient relationship. Policy can be useful as a resource, but each patient should have the maximum individualized care and access that a doctor can provide. Patients should be able to make their own informed choices and not be dictated to by policies meant for masses. Shouldn\'t everyone with increased IMT at least be informed of the options available to limit it rather than waiting until acute events or advanced narrowing occur? It would be best to start a healthy lifestyle from birth, but fortunately by adulthood there is still time to make an enormous difference in practical terms if we take action at the stage when intimal widening is detectable with the highly sensitive ultrasound described in my original viewpoint \[[@pmed-0020189-b1]\]. Citation: Makover M (2005) A further response to Shah Ebrahim. PLoS Med 2(7): e189. [^1]: Competing Interests: The author has declared that no competing interests exist.	PubMed Central	2024-06-05T03:55:59.866052	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181875/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e189", "authors": [ { "first": "Michael", "last": "Makover" } ] }
PMC1181876	Introduction {#s1} ============ Malaria remains one of the most serious global health problems and a leading cause of childhood morbidity and mortality, especially in Africa \[[@pmed-0020190-b01]\]. Efforts to control malaria in Africa have been severely compromised by the emergence of resistance in Plasmodium falciparum to the inexpensive and widely used drugs, chloroquine (CQ) and sulfadoxine-pyrimethamine (SP) \[[@pmed-0020190-b02],[@pmed-0020190-b03]\]. Use of combination antimalarial therapy, particularly newer regimens containing artemisinin-based compounds, has been widely advocated \[[@pmed-0020190-b04]\]. However, concerns regarding the cost and availability of artemisinin-based combination therapy (ACT) remain, and limited data comparing ACT with other combination therapies in Africa are available \[[@pmed-0020190-b05]\] In 2000, Uganda replaced CQ with the combination of CQ + SP as the recommended first-line treatment for uncomplicated malaria, although efficacy data for this regimen were lacking. To support development of a rational antimalarial treatment policy, the Uganda Malaria Surveillance Project was formed as a collaborative effort between the Ugandan Malaria Control Program, the East African Network for Monitoring Antimalarial Therapy, and academic researchers to provide efficacy data on antimalarial therapies from multiple sites. In addition to providing efficacy data from diverse locales, this project offers the possibility of evaluating the impacts of varied malaria endemicities and drug resistance patterns on responses to antimalarial therapy. In a recent study from a relatively low-transmission site in Kampala, Uganda, combinations of amodiaquine (AQ) + SP and AQ + artesunate (AS) were found to be significantly more efficacious than CQ + SP. Directly comparing the two AQ-containing regimens over 28 d, AQ + SP was associated with a higher risk of recrudescence after therapy, and AQ + AS was associated with a higher risk of new infections, such that the overall risk of repeat therapy was similar \[[@pmed-0020190-b06]\]. Considering all recurrent infections, the ACT regimen did not offer an obvious advantage over the much less expensive AQ + SP combination. However, this conclusion led to concerns that results may have been unique to our study site and that the implications of recrudescent and new infections might be very different \[[@pmed-0020190-b07]\]. Here we report the results of randomized clinical trials comparing CQ + SP, AQ + SP, and AQ + AS at four additional sites in Uganda, conducted with the aim of providing data from areas of varying transmission intensity to help guide antimalarial treatment policy. Furthermore, we were interested in evaluating the interplay between transmission intensity and the efficacies of different combination regimens, considering effects both on recrudescences and new infections after therapy. Methods {#s2} ======= Study Design and Study Sites {#s2a} ---------------------------- We conducted single-blind, randomized clinical trials to compare the efficacy and safety of three combination antimalarial regimens at four district health centers, using a common study protocol. Study sites, selected for geographic diversity, were: Jinja (periurban, southern Uganda, medium-high endemicity), Arua (rural, northwest Uganda, very high endemicity), Apac (rural, central Uganda, very high endemicity), and Tororo (rural, southeast Uganda, very high endemicity) ([Figure 1](#pmed-0020190-g001){ref-type="fig"}). The level of transmission intensity was further characterized based on recent estimates of entomological inoculation rates (the number of infective bites per person per year): Jinja = 7, Arua = 393, Tororo = 591, Apac = 1,564 (P. Okello, Uganda Ministry of Health, personal communication). The study protocol was approved by the Uganda National Council of Science and Technology and the institutional review boards of the University of California, San Francisco and the University of California, Berkeley ([Protocols S1](#sd001){ref-type="supplementary-material"} and [S2](#sd002){ref-type="supplementary-material"}). ::: {#pmed-0020190-g001 .fig} Figure 1 ::: {.caption} ###### Map of Uganda Based on Malaria Endemicity Very low or no malaria: unstable malaria with varying parasite rates in children but generally below 5% (altitude above 1,700 m); low: parasite rates in children below 10% (altitude 1,500--1,700 m); medium to high: parasite rates in children above 10% and generally below 50%, except for seasonal peaks (altitude 1,300--1,500 m); very high: parasite rates in children above 50% (altitude below 1,300 m) (Malaria Control Program, Uganda Ministry of Health). EIR, entomological inoculation rate. ::: ![](pmed.0020190.g001) ::: Patients {#s2b} -------- Consecutive patients presenting to district health centers between November 2002 and May 2004 with symptoms suggestive of malaria and a positive-screening thick blood smear were referred to study physicians. Patients were enrolled if they fulfilled the following selection criteria: (1) age 6 mo or greater; (2) history of fever in the last 24 h or axillary temperature \> 37.5 °C; (3) no history of serious side effects to study medications; (4) no evidence of a concomitant febrile illness; (5) provision of informed consent; (6) no history of treatment with an antifolate or AQ during the previous week; (7) absence of pregnancy-based on history of last menstrual period; (8) no danger signs or evidence of severe malaria \[[@pmed-0020190-b08]\]; and (9) P. falciparum monoinfection with parasite density 2,000/μl--200,000/μl. Because laboratory results were generally not available until the following day, a patient could be excluded after randomization. Treatment, Randomization, and Blinding {#s2c} -------------------------------------- Patients were randomly assigned to receive one of three oral therapies: CQ (10 mg/kg on days 0 and 1, and 5 mg/kg on day 2) + SP (25 mg/kg sulfadoxine and 1.25 mg/kg pyrimethamine as a single dose on day 0); AQ (10 mg/kg on days 0 and 1, and 5 mg/kg on day 2) + SP; or AQ + AS (4 mg/kg on days 0, 1, and 2). Patients in the CQ + SP and AQ + SP treatment arms also received lactose placebo tablets on days 1 and 2. Randomization codes were computer-generated by an off-site investigator for two age groups (6--59 mo and 5 y or older) and provided to a study nurse responsible for treatment allocation. All other study personnel were blinded to the treatment assignments, and patients were not informed of their treatment regimen. Patients were directly observed for 30 min after treatment, and the dose was readministered if vomiting occurred. Patients who repeatedly vomited their first dose of study medication were excluded from the study. Follow-Up Procedures and Classification of Treatment Outcomes {#s2d} ------------------------------------------------------------- Following enrollment, patients were asked to return for follow-up visits on days 1, 2, 3, 7, 14, 21, 28, and any other day when they felt ill. Blood was obtained by finger prick for thick blood smears and storage on filter paper on all follow-up days except day 1. Hemoglobin measurements were repeated on day 28 or the day of treatment failure. Treatment outcomes were classified according to 2003 World Health Organization guidelines as early treatment failure (ETF; danger signs or complicated malaria or failure to adequately respond to therapy days 0--3), late clinical failure (LCF; danger signs or complicated malaria or fever and parasitemia on days 4--28 without previously meeting criteria for ETF), late parasitological failure (LPF; asymptomatic parasitemia on day 28 without previously meeting criteria for ETF or LCF), and adequate clinical and parasitological response (ACPR; absence of parasitemia on day 28 without previously meeting criteria for ETF, LCF, or LPF) \[[@pmed-0020190-b08]\]. Patients classified as treatment failures were treated with quinine (10 mg/kg three times daily for 7 d); however, their response to repeat therapy was not assessed. Patients were excluded after enrollment if any of the following occurred: (1) use of antimalarial drugs outside of the study protocol; (2) parasitemia in the presence of a concomitant febrile illness; (3) withdrawal of consent; (4) loss to follow-up; (5) protocol violation; and (6) death due to a nonmalaria illness. Laboratory Procedures {#s2e} --------------------- Blood smears were stained with 2% Giemsa for 30 min. Parasite densities were determined from thick blood smears by counting the number of asexual parasites per 200 WBCs (or per 500, if the count was less than 10 parasites/200 WBCs), assuming a WBC count of 8,000/μl. A smear was considered negative if no parasites were seen after review of 100 high-powered fields. Thin blood smears were used to detect nonfalciparum infections. Hemoglobin measurements were made using a portable spectrophotometer (HemoCue, Anglholm, Sweden). Molecular genotyping techniques were used to distinguish recrudescence from new infection for all patients failing therapy after day 3. Briefly, filter paper blood samples collected on the day of enrollment and the day of failure were analyzed for polymorphisms in merozoite surface protein-2 (MSP-2) using nested-PCR as previously described \[[@pmed-0020190-b09]\]. Genotyping patterns on the day of repeat therapy were compared with those at treatment initiation using GelCompar II software (Applied Maths, St-Martens-Latem, Belgium). An outcome was defined as recrudescence if all alleles present at the time of retreatment were present at the time of treatment initiation, and as a new infection otherwise. Statistical Analysis {#s2f} -------------------- The study was designed to test the hypothesis that treatment with AQ + SP or AQ + AS would change the risk of recrudescence compared to CQ + SP at each study site. We calculated that 176 patients were needed to be enrolled in each treatment arm at each site for a 0.05 two-sided type I error with 0.8 power, assuming a risk of recrudescence adjusted by genotyping of 15% in the CQ + SP group and 5% in the comparison groups. Data were double-entered and verified using EpiInfo 6.04 (Centers for Disease Control and Prevention, Atlanta, Georgia, United States), and analyzed using Stata version 8.0 (Stata Corporation, College Station, Texas, United States). Efficacy data were evaluated using a per-protocol analysis, which only included patients with treatment outcomes. Prior to completion of this study we decided that a per-protocol analysis would be more appropriate than an intention-to-treat (ITT) analysis. Our protocol was used to study CQ + SP versus AQ + SP at three additional sites prior to the completion of the studies presented here. During the analysis of these previous studies we decided that a per-protocol analysis provided better estimates of the true risk of treatment outcomes than an ITT analysis, as in an ITT analysis one must assign treatment outcomes to patients who did not complete the study. In addition, there were so few patients enrolled who did not complete the study at all our sites that the comparative results using a per-protocol analysis did not differ from that using an ITT analysis. Primary efficacy outcomes included the 28-d risks for all recurrent infection (ETF, LCF, or LPF), recrudescence (all ETF and any LCF or LPF categorized as recrudescence based on genotyping results), and new infections (any LCF or LPF categorized as new infection based on genotyping results). Risks of recrudescence and new infection were estimated using the Kaplan-Meier product limit formula with censoring for the competing risk (new infections censored when estimating risks of recrudescence and vice versa). Secondary outcomes included the risk of recurrent infection unadjusted by genotyping at day 14, presence of fever on days 1--3, parasitemia on days 2 and 3, change in hemoglobin level between the day of enrollment and the last day of follow-up, presence of gametocytes during any follow-up day, and the incidence of adverse events. Pairwise comparisons of treatment efficacy were made using risk differences with exact 95% confidence intervals (CIs). Other categorical variables were compared using chi-squared or Fisher\'s exact test, and continuous variables were compared using the independent samples t-test. All reported p-values were two-sided, without adjustment for multiple testing, and were considered statistically significant if less than 0.05. Results {#s3} ======= Enrollment {#s3a} ---------- Of 2,684 patients who underwent screening, 2,270 were randomized to treatment, and 2,160 were enrolled in the studies ([Figure 2](#pmed-0020190-g002){ref-type="fig"}). Primary efficacy outcomes, unadjusted and adjusted by genotyping, were available for 96% and 95% of patients enrolled, respectively. ::: {#pmed-0020190-g002 .fig} Figure 2 ::: {.caption} ###### Trial Profile Trial profile stratified by treatment group and site (Jinja, Arua, Tororo, Apac); screened: (693, 625, 718, 648); excluded during initial screening: (122, 74, 157, 61); randomized: CQ+SP (179, 184, 174, 196); AQ + SP (196, 185, 192, 197); AQ + AS (196, 182, 195, 194); excluded before enrollment: CQ + SP (4, 11, 8, 11); AQ + SP (5, 14, 11, 10); AQ + AS (8, 20, 1, 7); enrolled: CQ + SP (168, 180, 166, 185); AQ + SP (186, 180, 181, 183); AQ + AS (189, 174, 194, 174); excluded after enrollment: CQ + SP (8, 2, 4, 5); AQ + SP (13, 7, 9, 5); AQ + AS (8, 3, 13, 2); completed: CQ + SP (160, 178, 162, 180); AQ + SP (173, 173, 172, 178); AQ+AS (181, 171, 181, 172). ::: ![](pmed.0020190.g002) ::: Baseline Characteristics {#s3b} ------------------------ Baseline characteristics of the patients across the three treatment groups were similar at each site ([Table 1](#pmed-0020190-t001){ref-type="table"}). Between the four sites, patients differed in some baseline characteristics, as expected based on differences in transmission intensity ([Table 1](#pmed-0020190-t001){ref-type="table"}). Patients from Jinja were older, with 39% age 5 y or older, compared to less than 10% at the other sites (p \< 0.001), consistent with more common presentation of older individuals with malaria at a lower transmission site. Patients from Jinja also had higher baseline temperatures (p \< 0.001) and mean hemoglobin levels (p \< 0.001) compared to the other sites. Parasite densities varied significantly across the sites (p \< 0.001 for all pairwise comparisons) and decreased with increasing entomological inoculation rate estimates. The proportion of patients with gametocytes present in pretreatment samples varied considerably (p \< 0.001), ranging from 10% (Jinja and Tororo) to 51% (Apac). The proportion of patients providing a history of recent CQ use ranged from 5% (Arua) to 28% (Jinja), with 90% of patients reporting a partial treatment course (less than three doses) prior to presentation at the district health centers. ::: {#pmed-0020190-t001 .table-wrap} Table 1 ::: {.caption} ###### Baseline Characteristics of Patients Completing the Study ::: ![](pmed.0020190.t001) ::: Primary Outcome: Treatment Efficacy {#s3c} ----------------------------------- ETFs were uncommon, but significant differences between the three treatment groups were observed when data from all sites were combined (CQ + SP = 3.4%, AQ + SP = 1.0%, AQ + AS = 0.1%; p \< 0.04 for all pairwise comparisons). The risk of recurrent infection (unadjusted by genotyping) at day 14 was significantly higher for CQ + SP at all four sites (16%--57%) compared to AQ + SP (1%--11%) or AQ + AS (5%--13%) ([Table 2](#pmed-0020190-t002){ref-type="table"}). The risk of recurrent infection at day 28 was extremely high for CQ + SP at all four sites, ranging from 63% to 88% ([Table 2](#pmed-0020190-t002){ref-type="table"}). After adjustment by genotyping, the risk of recrudescence remained very high for CQ + SP at all four sites, ranging from 22%--46% ([Table 2](#pmed-0020190-t002){ref-type="table"}). Compared to CQ + SP, the risk of recurrent infection and recrudescence at day 28 was significantly lower for the AQ + SP (28%--59% and 7%--18%, respectively) and AQ + AS (19%--74% and 4%--12%, respectively) treatment groups (all ps \< 0.05, [Table 2](#pmed-0020190-t002){ref-type="table"}). ::: {#pmed-0020190-t002 .table-wrap} Table 2 ::: {.caption} ###### Primary Treatment Efficacy Outcomes ::: ![](pmed.0020190.t002) ::: The most interesting comparisons of treatment efficacy were seen with AQ + SP versus AQ + AS. AQ + SP was associated with a higher risk of recrudescence at three sites, although this reached statistical significance at only one site, Jinja (risk difference = 9%; 95% CI: 3%--15%; p = 0.009) ([Figure 3](#pmed-0020190-g003){ref-type="fig"}). In contrast, the risk of new infection associated with AQ + SP and AQ + AS was similar at two sites, and was significantly higher with AQ + AS at the two highest transmission intensity sites, Tororo (risk difference = 21%; 95% CI: 10%--31%; p \< 0.001) and Apac (risk difference = 15%; 95% CI: 5%--25%; p = 0.003) ([Figure 3](#pmed-0020190-g003){ref-type="fig"}). Overall, the risk of any recurrent infection (recrudescence or new infection) was similar at the two lower transmission intensity sites and significantly higher with AQ + AS at the two highest transmission intensity sites; Tororo (risk difference = 15%; 95% CI: 5%--25%; p = 0.005) and Apac (risk difference = 16%; 95% CI: 6%--26%; p = 0.004) ([Figure 3](#pmed-0020190-g003){ref-type="fig"}). ::: {#pmed-0020190-g003 .fig} Figure 3 ::: {.caption} ###### Comparison of AQ + SP versus AQ + AS Risk differences and 95% CIs for recrudescence (adjusted by genotyping), new infections (adjusted by genotyping), and any recurrent infection (unadjusted, recrudescence or new infection) at day 28. ::: ![](pmed.0020190.g003) ::: Although some patients classified as failures had asymptomatic parasitemia on day 28 (LPF), the majority (67%) were symptomatic on the day of failure (ETF or LCF). Comparisons of treatment efficacy (both unadjusted and adjusted by genotyping) considering only those patients who were symptomatic on the day of failure were similar to those above (data not shown). Comparisons of the characteristics of late clinical and parasitological failures due to recrudescence versus those due to new infections are presented in [Table 3](#pmed-0020190-t003){ref-type="table"}. Median duration to failure was shorter with recrudescences compared to new infections (26 d versus 27 d, p = 0.03), although the difference was marginal, and over 75% of both recrudescences and new infections occurred after 20 d of follow-up. No differences between recrudescences and new infections were found with respect to the proportion of patients who were symptomatic, the risk of complicated malaria, parasite density, or changes in hemoglobin. Gametocytes during follow-up were common with both outcomes, but more common with recrudescences (51% versus 43%, p = 0.02). ::: {#pmed-0020190-t003 .table-wrap} Table 3 ::: {.caption} ###### Comparison of Late Clinical and Parasitological Failures due to Recrudescence versus New Infection ::: ![](pmed.0020190.t003) ::: Secondary Outcomes {#s3d} ------------------ Treatment with AQ + AS was associated with a significantly lower risk of parasitemia on day 2 at all four sites (2%--9%) compared to CQ + SP (37%--74%) or AQ + SP (33%--58%); however, this was not consistently associated with clinical benefit as measured by presence of fever and hemoglobin level ([Table 4](#pmed-0020190-t004){ref-type="table"}). The proportion of patients with temperature \> 37.5 °C on day 2 was significantly lower in the AQ + AS group compared to CQ + SP at two sites, but there were no significant differences compared to the AQ + SP group ([Table 4](#pmed-0020190-t004){ref-type="table"}). Similar results were seen with temperature on days 1 and 3 and when considering subjective fever (data not shown). Increase in hemoglobin during follow-up was greatest in the AQ + SP group at all sites, reaching statistical significance at three sites when compared to CQ + SP and at one site when compared to AQ + AS ([Table 4](#pmed-0020190-t004){ref-type="table"}). The proportion of patients with gametocytes during follow-up was lowest in the AQ + AS group at all sites, reaching statistical significance at two sites when compared to CQ + SP and at one site when compared to AQ + SP ([Table 4](#pmed-0020190-t004){ref-type="table"}). When restricting the analysis to only those patients with no gametocytes on day 0, a similar trend was found, favoring AQ + AS, with statistical significance reached at all four sites when compared to CQ + SP and at two sites when compared to AQ + SP ([Table 4](#pmed-0020190-t004){ref-type="table"}). ::: {#pmed-0020190-t004 .table-wrap} Table 4 ::: {.caption} ###### Secondary Treatment Outcomes ::: ![](pmed.0020190.t004) ::: Safety and Tolerability {#s3e} ----------------------- Among 2,160 patients enrolled in the studies, 20 serious adverse events were reported in 16 patients (four CQ + SP, eight AQ + SP, four AQ + AS, p = 0.40). Serious adverse events included anemia (two AQ + SP, one AQ + AS), convulsion (one CQ + SP, two AQ + SP, one AQ + AS), dehydration (one AQ + AS), edema (one AQ + SP), malnutrition (one CQ + SP), mental status change (two AQ + SP), respiratory illness (one CQ + SP, two AQ + SP, one AQ + AS), vomiting (one CQ + SP), and weakness (one AQ + SP). One patient died of suspected severe malnutrition and one patient died of congestive heart failure due to a presumed congenital heart defect (both patients had received AQ + SP). All serious adverse events were deemed to be unlikely (ten events) or possibly (ten events) related to the study medications. Discussion {#s4} ========== Antimalarial drug resistance is one of the greatest threats to malaria control in Africa \[[@pmed-0020190-b10]\]. In response to widespread resistance to CQ and SP, use of combination antimalarial therapy, particularly ACT, has been strongly advocated \[[@pmed-0020190-b11]\]. This study compared the efficacy of three different combination therapies, CQ + SP, currently the recommended first-line regimen in Uganda; AQ + SP, an inexpensive regimen that has proven to be efficacious in recent studies; and AQ + AS, an ACT regimen. More important, evaluations were made across differing levels of transmission intensity, allowing us to assess the impact of endemicity on treatment responses. We found that CQ + SP was highly ineffective. Indeed, data from various parts of Uganda have shown that the pfcrt K76T mutation primarily responsible for CQ resistance is virtually ubiquitous \[[@pmed-0020190-b12],[@pmed-0020190-b13]\] and that the efficacy of CQ + SP is similar to that of SP alone \[[@pmed-0020190-b14]\]. Replacing CQ with AQ in combination with SP greatly reduced the risk of recrudescence and prevented new infections, a benefit which was most evident at the two highest transmission sites. The prevention of new infections by AQ + SP is likely due to the long elimination half-lives of both drugs \[[@pmed-0020190-b15],[@pmed-0020190-b16]\]. Compared to AQ + SP, treatment with AQ + AS was generally associated with a lower risk of recrudescence but a higher risk of new infection within the month after therapy. This is likely due to the fact that AS is rapidly eliminated, leaving only AQ to provide posttreatment prophylaxis. When comparing AQ + AS to AQ + SP, the overall risks of recurrent infection were similar at two study sites and significantly higher for AQ + AS at the two highest transmission sites. Differences in treatment efficacies between our study sites could not be compared formally, as patient populations differed between sites. Nonetheless, wide differences in efficacies between sites were seen. These differences may have been due both to variations in endemicity (and therefore antimalarial immunity) and to varied drug resistance patterns. Varied endemicity likely played an important role, as suggested by marked differences in baseline characteristics between the study populations. Considering varied drug resistance patterns, preliminary data analyzing parasites from the four sites suggest that the prevalences of mutations known to mediate resistance to CQ and SP were similar (unpublished data). Thus, the major factor mediating differences in outcomes between study sites appears to be differences in the antimalarial immunity of study populations. Antimalarial drug-efficacy studies that limit follow-up to 14 d or less may significantly underestimate the risk of recrudescence \[[@pmed-0020190-b17]\]. Our study shows that this underestimation is particularly problematic in regions with very high malarial endemicity. Overall, only 21% of recrudescences occurred within the first 2 wk of the 4-wk follow-up period, and this proportion decreased to 14% at the highest transmission site. Thus, in highly endemic areas recrudescence may be delayed, presumably due to the contribution of host immunity on initial parasite clearance, and this delay appears to increase with increasing transmission intensity \[[@pmed-0020190-b17]\]. Additionally, the risk of new infections can be substantial in highly endemic areas. Indeed, in our study 72% of all recurrent infections were due to new infections, and with our two most efficacious regimens (AQ + SP and AQ + AS) this proportion was 80%. Although comparisons of antimalarial drug efficacy generally do not consider the relative impact on the risk of new infections, this factor can play a large role in treatment outcomes, especially in high-transmission areas such as Africa. Our identification of frequent new infections emphasizes the importance of other malaria control measures, such as the use of bed nets, vector reduction, and possibly intermittent presumptive therapy, as ways of reducing the risk of new infections and maximizing the impact of therapy. Follow-up in this study was limited to 28 d. In a previous longitudinal study of SP alone or in combination with AQ or AS from Uganda, 72% of recrudescences occurred within 28 d and 81% within 42 d \[[@pmed-0020190-b18]\]. Thus, the overall risk of recrudescence in this study was likely underestimated. Similarly, we may have underestimated the impacts of the different therapies on risks of new infection. These limitations will be present in any study with distinct endpoints for efficacy assessment, but it is generally agreed that 28-d outcomes offer a reasonable comparative assessment of antimalarial therapies \[[@pmed-0020190-b08]\]. However, the best comparisons of the impacts of different therapies on the overall risk of repeat therapy can only be made using a longitudinal study design with extended follow-up covering multiple episodes of malaria, as we described previously \[[@pmed-0020190-b18]\]. We suggest that antimalarial therapies be judged not only by their impact on the risk of recrudescence, but also by their impact on the risk of new infections after therapy \[[@pmed-0020190-b06]\]. However, it has been argued that the consequences of recrudescence and new infection are not equal and that patients who have recrudescence face a greater risk of progression to complicated malaria and death \[[@pmed-0020190-b07]\]. Our data do not support this contention. In our study, the risk of ETF was remarkably low (\<2%), and only four of 2,081 patients progressed to severe malaria or danger signs requiring therapy with intravenous quinine during the first 3 d of follow-up. Among 1,133 failures identified 4--28 d after therapy, 760 (67%) patients were symptomatic (LCFs), and clinical differences between recrudescence and new infection were not apparent. The only episode of severe malaria/danger signs to occur after day 3 was caused by a new infection. Thus, we found no support for the argument that the clinical consequences of recrudescence after treatment are different from those of a new infection during our 28 d follow-up period. It has also been argued that recrudescences are more difficult to treat than new infections. We could not address this issue in this study as patients were not assessed following repeat therapy. However, treatment of recrudescent infections was compared to initial treatment in a previous longitudinal study. Although the treatment of recrudescent infections with SP was associated with a higher rate of treatment failure, interestingly, treatment with SP + AS was associated with a lower rate of treatment failure \[[@pmed-0020190-b19]\]. Additional studies with longer follow-up, covering repeated episodes of malaria, are needed to more fully evaluate the treatment implications of recrudescent infections. The classification of recrudescence and new infection depends on the genotyping methods used. In our study we used a highly specific definition of recrudescence that required all alleles present at the time of retreatment be present at the time of treatment initiation. As recurrent infections containing both new alleles and alleles present at the treatment initiation were classified as new infections, we may have underestimated the true risk of recrudescence. Defining recrudescence as having any allele at the time of retreatment present at the time of treatment initiation increased our point estimates of the risk of recrudescence. However, using this more sensitive (but less specific) definition of recrudescence did not have a significant impact on comparative efficacies. Indeed, we suggest that the most useful comparisons of the clinical consequences of different treatments should be based on treatment outcome results unadjusted by genotyping, including all retreatments for clinical illness. Artemisinin derivatives are highly attractive antimalarials because they act rapidly, are well tolerated, and are currently not limited by resistance \[[@pmed-0020190-b03]\]. However, artemisinin monotherapy is associated with a high risk of recrudescence, necessitating use of artemisinins in combination with other antimalarials to achieve maximum efficacy \[[@pmed-0020190-b20]\]. In Thailand, the combination of AS + mefloquine has been associated with sustained cure rates over 95% and a decreased incidence of malaria \[[@pmed-0020190-b21],[@pmed-0020190-b22]\]. The coformulated ACT, artemether-lumefantrine, has been shown to be highly efficacious in relatively low endemicity sites in Southeast Asia \[[@pmed-0020190-b23]--[@pmed-0020190-b25]\]. Published evidence for the effectiveness of ACT in highly endemic areas of Africa remains limited. The combination of CQ + AS was associated with a 47%--93% risk of treatment failure (unadjusted by genotyping) after 28 d in three West African countries \[[@pmed-0020190-b26]\]. SP + AS was associated with a higher risk of treatment failure (unadjusted and adjusted by genotyping) compared to AQ + SP in Uganda and Rwanda \[[@pmed-0020190-b18],[@pmed-0020190-b27]\]. AQ + AS was associated with a lower risk of recrudescence but a similar risk of overall treatment failure when compared to AQ + SP at a site with relatively low-transmission intensity in Uganda \[[@pmed-0020190-b06]\]. Although ACT is clearly better than standard monotherapies, the somewhat disappointing results of ACT in Africa probably relate to the high endemicity of malaria and the inclusion of inadequate partner drugs with the artemisinins. Considering endemicity, patients with symptomatic malaria from highly endemic areas such as Africa generally have higher pretreatment parasite densities compared to patients from lower endemic areas such as Southeast Asia\[[@pmed-0020190-b03]\], and higher pretreatment parasite densities have been associated with a higher risk of recrudescence after treatment with AS \[[@pmed-0020190-b28]\]. Additionally, the increased risk of new infections in highly endemic areas may contribute greatly to treatment outcome. These factors suggest that in Africa, more so than other areas, it is critical that artemisinins be combined with other highly effective drugs. However, whether partner drug should be short- or long-acting remains unclear. Long-acting partner drugs offer an extended prophylactic effect; however, they may encourage the selection of drug resistance. Considering the efficacy of ACTs, initial studies in Africa combining AS with available drugs (CQ, SP, and AQ), were relatively disappointing in areas where the efficacy of the partner drugs was likely limited by resistance, as was seen in this study \[[@pmed-0020190-b18],[@pmed-0020190-b26],[@pmed-0020190-b27]\]. ACT regimens with more effective partner drugs (e.g., artemether-lumefantrine and/or dihydroartemisinin-piperaquine) will likely show improved efficacy \[[@pmed-0020190-b04]\]. Despite the limitations noted, ACT and other antimalarial combination therapies offer great hope for Africa. However, the ideal combination regimens remain uncertain \[[@pmed-0020190-b04]\]. As African countries rapidly move toward replacement of ineffective monotherapies with combinations regimens, it is essential that controlled trials be conducted to compare efficacy, ideally considering the importance of varied transmission intensity. Studies with extended follow-up and longitudinal designs will provide the most useful comparisons, as different therapies may impact very differently on long-term outcomes. Based on the results of this study and others, Uganda has recently opted to replace CQ + SP with artemether-lumefantrine, joining many other African countries in choosing an ACT as first-line therapy. However, cost and availability of ACTs remain major concerns, and it appears that the sudden increase in demand for artemisinins may exacerbate these problems, at least in the short-term \[[@pmed-0020190-b05]\]. The inexpensive and widely available combination AQ + SP may still be appropriate for the treatment of uncomplicated malaria in areas where resistance to the drugs is relatively uncommon, in situations where ACT may not be feasible (e.g., such as the treatment of unconfirmed cases of malaria outside of the formal health sector), and, in the short term, while adequate supplies of ACT are not yet available in Africa. Supporting Information {#s5} ====================== Protocol S1 ::: {.caption} ###### CONSORT Checklist (58 KB DOC). ::: ::: {.caption} ###### Click here for additional data file. ::: Protocol S2 ::: {.caption} ###### Combination Therapies for Treatment of Complicated falciparum Malaria in Uganda: Evaluation of Efficacy, Safety, and Tolerability (1.0 MB DOC). ::: ::: {.caption} ###### Click here for additional data file. ::: Patient Summary {#sb1} --------------- ### Background {#sb1a} As resistance to current malarial regimens grows, it is increasingly important to find new combinations of drugs that will not only treat an ongoing infection, but will also prevent recurrence of the malaria (either a new infection or reappearance of the treated infection). One type of drug that is being assessed for preventing recurrence is based on artemisinin, which was originally derived from a plant, sweet wormwood. However, this drug is more expensive than older drugs ### What Did the Authors Do? {#sb1b} They did four randomized clinical trials in Africa at the same time; two in areas where malaria occurs very frequently and two where it is less frequent. They compared a combination of artemisinin with two other combinations of older drugs and looked to see how well the treatments worked on the present infection, on preventing recurrences, and on whether there were any serious adverse events. They found that the combination including artemisinin worked the best at treating current infections. However, patients given the artemisinin-based treatment were overall more likely to get new infections. Where malaria was very common, people treated with the artemisinin-based combination were more likely to get recurrent infections overall. ### What Do These Findings Mean? {#sb1c} Although the artemisinin-based treatment worked very well on present infections, for recurrent infections it did not perform better than, and was considerably more expensive than, an older combination of drugs. Artemisinin-based treatment should not automatically be assumed the best treatment for uncomplicated malaria in Africa. ### Where Can I Get More Information? {#sb1d} The World Health Organization has a Web site on Africa with links to malaria control: <http://www.afro.who.int/> The United Kingdom Department for International Development has information on malaria control in Africa: <http://www.lshtm.ac.uk/dfid/malaria/> We thank the clinical study team of Joy Bossa, Nelson Budaka, Oswald Byaruhanga, Dorothy Kirunda, Moses Musinguzi, Betty Nanzigu, Godfrey Buyinza, Isaac Kigozi, Felix Jurua, Joanita Nankabirwa, Fred Kizito, Sam Balikowa, and Grace Musimenta. We would also like to thank Moses Kiggundu, Dan Kyabayinze, and Regina Nakafeero for training the laboratory staff and providing laboratory quality control. We would also like to thank all the health workers at the health centers and their respective district administrators for allowing us to conduct these studies and for working alongside the study teams for lengthy periods. We are indebted to the administrative support of Sara Kibirango and Kenneth Mwebaze; drivers Nuhu Kibampawo, Joshua Sekitoleko, and Marx Dongo; and the data officer, John Patrick Mpindi. Financial support was provided by the Centers for Disease Control/Association of Schools of Public Health cooperative agreement, "Malaria Surveillance and Control in Uganda" (SA3569 and S1932--21/21), and the Department for International Development (DFID). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Citation: Yeka A, Banek K, Bakyaita N, Staedke SG, Kamya MR, et al. (2005) Artemisinin versus nonartemisinin combination therapy for uncomplicated malaria: Randomized clinical trials from four sites in Uganda. PLoS Med 2(7): e190. ACT : artemisinin-based combination therapy AQ : amodiaquine AS : artesunate CI : confidence interval CQ : chloroquine ETF : early treatment failure LCF : late clinical failure LPF : late parasitological failure SP : sulfadoxine-pyrimethamine [^1]: Competing Interests: The authors have declared that no competing interests exist. [^2]: Author Contributions: AY, KB, NB, SGS, MRK, AT, AK, AR, PJR, FWM, and GD designed the study. AY, KB, NB, SGS, MRK, AT, FK, SLN, MS, FWM, and GD performed the experiments. AY, KB, and GD analyzed the data. AY and KB enrolled patients. AY, KB, NB, SGS, MRK, AT, FK, SLN, AK, MS, AR, PJR, FWM, and GD contributed to writing the paper.	PubMed Central	2024-06-05T03:55:59.866810	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181876/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e190", "authors": [ { "first": "Adoke", "last": "Yeka" }, { "first": "Kristin", "last": "Banek" }, { "first": "Nathan", "last": "Bakyaita" }, { "first": "Sarah G", "last": "Staedke" }, { "first": "Moses R", "last": "Kamya" }, { "first": "Ambrose", "last": "Talisuna" }, { "first": "Fred", "last": "Kironde" }, { "first": "Samuel L", "last": "Nsobya" }, { "first": "Albert", "last": "Kilian" }, { "first": "Madeline", "last": "Slater" }, { "first": "Arthur", "last": "Reingold" }, { "first": "Philip J", "last": "Rosenthal" }, { "first": "Fred", "last": "Wabwire-Mangen" }, { "first": "Grant", "last": "Dorsey" } ] }
PMC1181877	PRESENTATION of CASE {#s1} ==================== A 25-y-old woman presented with pulmonary embolism. She had been taking, without apparent complication, norgestimate/ethynil estradiol (Ortho Tri-Cyclen; 0.180, 0.215, and 0.250 mg norgestimate cycles and 35 μg ethinyl estradiol) for 2 y, followed by the same drug combination but with a lower dose of ethinyl estradiol (Ortho Tri-Cyclen Lo; 25 μg ethinyl estradiol) for 14 mo. A nonsmoker, she lacked a relevant family history and was vigorously athletic. One month prior to presentation, she developed neck pain; disc protrusion at C5-C6 was detected by magnetic resonance imaging. The patient was prescribed valdecoxib, 20 mg twice a day (b.i.d.), for 2 wk. Her neck pain resolved. However, towards the end of this treatment period, she developed left-sided pleuritic chest and shoulder pain after a 6-h car ride. She was started on cyclobenzaprine, 10 mg b.i.d., and continued on valdecoxib. Her left-sided pain abated gradually. However, 18 d later, she developed right-sided chest and shoulder pain. A diagnosis of left iliac vein thrombosis and bilateral pulmonary emboli was based on a computed tomography scan. She was heparinized and continued on therapy with warfarin, 5 mg b.i.d., and enoxaparin, 60 mg b.i.d. Despite this, a ventilation-perfusion scan performed 13 d later showed multiple pulmonary emboli ([Figure 1](#pmed-0020197-g001){ref-type="fig"}). ::: {#pmed-0020197-g001 .fig} Figure 1 ::: {.caption} ###### Ventilation-Perfusion Scan \(A) After inhalation of 20.1 mCi of Xenon-133 gas, scintigraphic images were obtained in the posterior projection, showing uniform ventilation to lungs. \(B) After intravenous injection of 4.1 mCi of Technetium-99m-labeled macroaggregated albumin, scintigraphic images were obtained, shown here in the posterior projection. This and other views showed decreased activity in the following regions: apical segment of right upper lobe, anterior segment of right upper lobe, superior segment of right lower lobe, posterior basal segment of right lower lobe, anteromedial basal segment of left lower lobe, and lateral basal segment of left lower lobe. ::: ![](pmed.0020197.g001) ::: Currently, the patient is on warfarin, 2.5 mg daily, while completing a 6-mo warfarin regimen. Her warfarin is well tolerated and otherwise the patient is in good health. DISCUSSION {#s2} ========== Coxibs, selective inhibitors of cyclooxygenase-2 (COX-2), increase the risk of myocardial infarction and stroke \[[@pmed-0020197-b01]--[@pmed-0020197-b03]\], prompting concern for patients with established cardiovascular disease. Caution may also extend to individuals predisposed to thrombosis by genetic or environmental factors. Risk factors for spontaneous thrombosis include oral contraceptives, genetic predisposition to a hypercoaguable state, and prolonged periods of stasis \[[@pmed-0020197-b04]--[@pmed-0020197-b06]\]. At least two risk factors pertained to this patient. The patient had been taking oral contraceptives for 3 y prior to the index event, albeit without recognized thrombotic complication. It is possible that despite this extended period of apparent tolerance, the embolic events were solely related to use of the oral contraceptives \[[@pmed-0020197-b04]\]. The relatively small risk of venous thromboembolism attributable to oral contraceptive use may interact geometrically with the similarly small absolute risk of a procoagulant mutation, such as Factor V Leiden \[[@pmed-0020197-b05]\]. However, documented genetic risk factors, such as abnormalities in lupus anticoagulant, anti-thrombin III, proteins C and S, plasma homocysteine, anticardiolipin, β2 glycoprotein antibody, and prothrombotic mutations in Factor V and prothrombin were excluded. It remains possible that the patient was genetically predisposed to thrombosis by a mutation in an undetermined factor. Finally, prolonged stasis, such as that which occurred during the 6-h car trip, may account for the clinical event \[[@pmed-0020197-b06]\]. However, the absolute risk is small, and the patient had made this trip on multiple occasions devoid of apparent clinical complications during the prior 3 y. The risk of thrombosis from valdecoxib has now been established \[[@pmed-0020197-b03],[@pmed-0020197-b07]\], and this, together with its propensity rarely to cause Stevens-Johnson syndrome without a mitigating benefit over traditional nonsteroidal anti-inflammatory drugs, has led to its withdrawal from the market. The cardiovascular hazard of coxibs appears likely to be attributable to the suppression of COX-2-derived prostacyclin \[[@pmed-0020197-b01],[@pmed-0020197-b08]\]. Deletion of the prostacyclin receptor in mice does not cause spontaneous thrombosis, but rather enhances the response to thrombotic stimuli \[[@pmed-0020197-b09]\]. This is consistent with the fact that a cardiovascular signal from a coxib is most easily detected in patients with hemostatic activation, such as was observed under placebo-controlled conditions in two trials in patients undergoing cardiopulmonary bypass grafting \[[@pmed-0020197-b07]\] and anecdotally in patients with connective tissue disease \[[@pmed-0020197-b10]\]. Although this case does not establish a causative linkage with valdecoxib, the clinical event was not manifest until three potential risk factors were combined (oral contraceptives, prolonged stasis, and coxib treatment). Given the multiplicative interactions of risk factors for thromboembolic disease and the apparently untoward concurrence of two of them---the contraceptive pill and frequent road trips for the preceding 3 y---the rapid occurrence of the clinical event following initiation of the coxib suggests a causative link to the COX-2 inhibitor. However, it also remains formally possible that this temporal relationship was a coincidence. This case report has been reported to the regulatory authorities and to the manufacturer of valdecoxib. Just as the small absolute risks of thrombosis attributable to oral contraceptives and prothrombotic mutations may interact dramatically \[[@pmed-0020197-b05]\], selective inhibitors of COX-2 may also interact with genetic and environmental factors that predispose to the risk of thrombosis. Learning Points {#sb1} --------------- • The risk of thrombotic events on COX-2 inhibitors may extend to apparently healthy individuals. • Genetic and environmental risk factors for thrombosis may interact geometrically. • Patients should be screened for recognized genetic or environmental predisposition to thrombosis prior to deciding to initiate treatment with COX-2 inhibitors. Citation: Westgate EJ, FitzGerald GA (2005) Pulmonary embolism in a woman taking oral contraceptives and valdecoxib. PLoS Med 2(7): e197. b.i.d. : twice a day COX-2 : cyclooxygenase-2 [^1]: Competing Interests: The authors have declared that no competing interests exist. [^2]: Author Contributions: EJW drafted the paper, and GAF edited and expanded the manuscript.	PubMed Central	2024-06-05T03:55:59.870108	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181877/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e197", "authors": [ { "first": "Elizabeth J", "last": "Westgate" }, { "first": "Garret A", "last": "FitzGerald" } ] }
PMC1181878	We are writing in response to the letter by Stacia Nordin \[[@pmed-0020202-b1]\]. Independent of where a child is born in the world, the most appropriate feeding regimen is breast milk until six months of age, followed by a weaning or complementary food \[[@pmed-0020202-b2]\]. It is known that breast milk provides all the essential nutrients for a growing infant, except for vitamin D. It is also known that complementary foods should contribute to providing all of the essential nutrients when breast milk is no longer the sole source of nutrition after the first few months of life. However, as early as 1930, it was realized that typical complementary foods were generally poor sources of micronutrients (minerals and vitamins) and were often not sufficient to meet the micronutrient needs of growing children. For example, per 100 g, rice-based complementary foods contain about 1 mg of iron, and wheat-based complementary foods contain about 0.8 mg of iron \[[@pmed-0020202-b3]\]. Even a meat-based complementary food, such as commercial "toddler beef stew", contains only 1.2 mg of iron in each 170-g jar. Since the recommended dietary allowance for iron is 11 mg/day (for ages 7--12 months) \[[@pmed-0020202-b4]\], clearly a rice- or wheat-based complementary food, or even a dilute meat-based stew, would not provide an adequate amount of iron for a growing infant. Pablum, the first fortified baby food, was invented at the Hospital for Sick Children in Toronto, Canada, in the 1930s \[[@pmed-0020202-b5]\]. Subsequently, by the early 1960s in North America, all commercially manufactured infant cereals were fortified with iron. Today, the major source of iron in the diet of a North American child is fortified commercial infant cereals. And, indeed, the low rates of iron-deficiency anaemia in Canada and the United States are thought to be partly a result of the widespread use of commercially available iron-fortified cereals \[[@pmed-0020202-b6]\]. Another good example of a fortified food for young children in North America is fluid milk products, which are fortified with vitamin D in order to prevent the development of rickets. It is currently well accepted among nutritionists and pediatricians that most young children in North America depend on fortified foods to meet their micronutrient needs. In most developing countries, access to commercially processed baby foods (fortified with iron) is very limited mainly because of their high cost and limited availability \[[@pmed-0020202-b7]\]. It is noteworthy that recent research has demonstrated that even if dietary diversification and modification (such as soaking, fermentation, and germination) strategies are used at the household level, they may not be sufficient to overcome the deficits in iron and other micronutrients \[[@pmed-0020202-b3]\]. As a result, other options need to be considered for young children living in developing and poor countries to ensure that all of their nutrient requirements are met \[[@pmed-0020202-b8]\]. The use of Sprinkles is one such option \[[@pmed-0020202-b9],[@pmed-0020202-b10]\]. One of the greatest advantages of the Sprinkles concept is its emphasis on complementary food consumption because Sprinkles have to be mixed with food. When educating caregivers about anaemia and the use of Sprinkles, healthy weaning practices can be concurrently promoted to ensure the timely introduction of complementary foods at six months of age in addition to continued breast feeding (as recommended by the World Health Organization) \[[@pmed-0020202-b2]\]. This is an important benefit, as it is well known that in many developing countries poor weaning practices are common \[[@pmed-0020202-b3]\]. As a home fortificant, Sprinkles ensure that the food eaten contains adequate amounts of essential micronutrients. Indeed, Sprinkles are meant to improve the nutritional value of homemade baby foods, which are otherwise poor in micronutrient content. Sprinkles can enrich foods not only with iron but also with other essential micronutrients such as zinc, folic acid, and vitamins A and C. In addition, since Sprinkles can be easily mixed with any homemade semisolid foods, their use does not require any change in food practices; thus, they can be easily accepted in diverse cultural settings. With anaemia rates as high as 80% in young children in some developing countries, current food-based strategies alone are clearly not effective. All children should have the right to eat foods that meet their nutritional needs. The use of Sprinkles is one way to help these children meet their nutrient requirements. Unfortunately, a food-based strategy alone, using locally available unfortified foods, in most circumstances, is simply inadequate and may further predispose a growing child to various micronutrient deficiencies \[[@pmed-0020202-b8]\]. Citation: Zlotkin SH, Schauer C, Christofides A, Sharieff W, Tondeur MC, et al. (2005) Authors\' reply: Sprinkles as a home fortification strategy to improve the quality of complementary foods. PLoS Med 2(7): e202. [^1]: Competing Interests: SHZ is an occasional consultant to Bristol-Myers Squibb, and Mead Johnson. He owns the intellectual property rights to Sprinkles. The H. J. Heinz Company is supporting the technical development of Sprinkles on a cost-recovery basis. Any profit from royalty fees on the technology transfer of Sprinkles is currently donated to the Hospital for Sick Children Foundation.	PubMed Central	2024-06-05T03:55:59.870974	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181878/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e202", "authors": [ { "first": "Stanley H", "last": "Zlotkin" }, { "first": "Claudia", "last": "Schauer" }, { "first": "Anna", "last": "Christofides" }, { "first": "Waseem", "last": "Sharieff" }, { "first": "Mélody C", "last": "Tondeur" }, { "first": "S. M. Ziauddin", "last": "Hyder" } ] }
PMC1181879	Visceral leishmaniasis (VL), commonly known as kala-azar, from the Hindu vernacular, is a human systemic disease caused by parasitic protozoan species of the genus Leishmania. Transmitted by the bite of the tiny and seemingly innocuous female phlebotomine sandfly ([Figure 1](#pmed-0020211-g001){ref-type="fig"}), the parasite enters macrophages, where it multiplies and establishes the infection ([Figure 2](#pmed-0020211-g002){ref-type="fig"}). ::: {#pmed-0020211-g001 .fig} Figure 1 ::: {.caption} ###### Female Phlebotomus sp. Sandfly (Photo: WHO/CDC) ::: ![](pmed.0020211.g001) ::: ::: {#pmed-0020211-g002 .fig} Figure 2 ::: {.caption} ###### The Life Cycle of Leishmania spp., the Causal Agents of Leishmaniasis Leishmaniasis is transmitted by the bite of female phlebotomine sandflies. The sandflies inject the infective stage, promastigotes, during blood meals (1). Promastigotes that reach the puncture wound are phagocytized by macrophages (2) and transform into amastigotes (3). Amastigotes multiply in infected cells and affect different tissues, depending in part on the Leishmania species (4). This originates the clinical manifestations of leishmaniasis. Sandflies become infected during blood meals on an infected host when they ingest macrophages infected with amastigotes (5 and 6). In the sandfly\'s midgut, the parasites differentiate into promastigotes (7), which multiply and migrate to the proboscis (8). (Illustration: CDC/Alexander J da Silva/Melanie Moser) ::: ![](pmed.0020211.g002) ::: A multitude of clinical features of the disease ensue gradually, the most important being splenomegaly, recurring and irregular fever, anaemia, pancytopenia, weight loss, and weakness. Unlike malaria, there is no early dramatic fever to announce its arrival; the presentation is insidious, with symptoms appearing over a period of weeks or even months. Affected patients become progressively more anaemic, weak, cachectic, and susceptible to intercurrent infections. The disease is a silent killer, invariably killing almost all untreated patients \[[@pmed-0020211-b1]\]. VL affects not only the weakest in the community, such as children and those weakened by other diseases such as HIV and tuberculosis, but also healthy adults and economically productive social groups. > The disease is a silent killer, invariably killing almost all untreated patients An estimated 500,000 new cases of VL occur each year, and a tenth of these patients will die \[[@pmed-0020211-b2]\]. The actual death toll from the disease may be higher than this estimate, considering the existence of unidentified VL foci. Some 90% of those affected by the disease live in five countries: India (especially Bihar), Bangladesh, Nepal, (northeastern) Brazil, and Sudan \[[@pmed-0020211-b2]\]. VL often exists in areas that are either remote or not easily accessible, and where health facilities are barely available or inadequate. Those most likely to be infected are people who are poor, living in villages far from roads and health-care centres. Patients from such remote communities often die in the villages without seeking treatment. Some may attempt to report to distant health-care centres, but in many cases it is simply too late. Even if they can make the journey to a hospital, they would still succumb to the illness because of the absence of anti-leishmanial drugs. Thus, many decide to stay at home until they die. But in doing so, they act as a reservoir of infection, passing on the parasite to family and neighbours through the bite of sandflies. At present, approaches to the control of VL are varied. This variety is dictated first and foremost by the diverse epidemiological patterns of the disease, which range from domestic zoonosis (see Glossary) involving the dog (the Mediterranean littoral) or sylvatic zoonosis (South America, and possibly Africa) to anthroponosis (India and Africa). The major epidemics of VL that have occurred in India and Africa have primarily been a result of human-to-human transmission, be it in a primarily anthroponotic or zoonotic focus. Knowledge of the epidemiology, and ecological types, of VL is of paramount importance in designing a sound VL control strategy. Identifying Patients and Mapping the Distribution: The Need for Improved Tools {#s2} ============================================================================== Most VL infections occur in remote geographical areas where health facilities are not well established and where the infections often co-exist with malaria and other debilitating parasitic infections. Under these circumstances, the disease usually presents a diagnostic dilemma. To alleviate this difficulty, health workers need to be provided with up-to-date information on the geographical distribution of VL in endemic countries. The mapping of VL is a complex undertaking, as the distribution of the disease is multifocal in nature, with remarkable variation in its prevalence and incidence. Moreover, most clinical cases are neither treated nor reported. This difficulty is further compounded by the fact that most infections of Leishmania are subclinical. The main activities of mapping involve active surveillance by case-finding, leishmanin skin test surveys, and serological screening of populations. Effective mapping activities will also require additional information about the sandfly. As capacities for undertaking such surveys are limited, mapping activities may be based on the use of the Geographic Information System (GIS), by which risk areas of VL can be predicted in areas where surveillance data are not readily available \[[@pmed-0020211-b3],[@pmed-0020211-b4]\]. Using GIS, it is possible to produce predicted-risk maps of the disease, based on statistical associations between the spatially comprehensive environmental data available from satellites and previous knowledge about disease distribution. GIS has been used to produce risk maps of VL in some countries, and it is hoped that the maps will assist in planning future vector-control programmes. The leishmanin skin test has been used for geographical mapping and for epidemiological description of transmission patterns \[[@pmed-0020211-b5]\]. However, its large-scale use has been frustrated by the absence of a standardised test antigen and models for interpretation of field data. Periodic active case-finding surveys are often used in mapping activities. But the labour-intensive nature of the endeavour does not permit wide coverage. The development of simple and robust screening techniques, such as the fast agglutination screening test (FAST), can be a useful addition for mapping activities \[[@pmed-0020211-b8]\]. Using FAST, large numbers of blood samples can be tested in a short time and can provide an indicator of past or present disease or infection. Field experiences with FAST have indicated the need for further improvements of the test with respect to robustness as well as cost. Other test systems for use in epidemiological mapping are yet to be developed. Estimates of Incidence {#s3} ====================== Incidence of VL varies from place to place depending on the epidemiological characteristics. Areas of sporadic endemicity (low incidence) and endemo-epidemicity (high incidence) are known to exist. High incidence rates of VL are common in areas where human populations are despoiled by social instability, war, and migration. However, accurate data on the burden of VL do not exist in many VL foci, as large proportions of VL cases are not recorded. Researchers and clinicians working in the field estimate that in some countries less than 20% of patients are currently being reached, even though this may vary from country to country. The failure of governments to reach patients is one of the main reasons for the increasing death toll and the ever-increasing incidence of the disease. Control strategies for VL need to highlight the importance of treatment not only to reduce morbidity and mortality but also to prevent the accumulation of cases. These strategies require the availability of simple diagnostic tools and affordable, easy-to-administer drugs. Conventional Diagnostics: Invasive and Potentially Dangerous {#s4} ============================================================ Demonstration of parasites in stained smears of tissue aspirates from spleen, bone marrow, or lymph node remains the most accurate (specific) method available for diagnosis of VL. Spleen and bone marrow are both superior to lymph node but more invasive. Obtaining aspirates from the spleen can be dangerous in patients with haematological complications. Culture of the Leishmania parasite from tissue aspirates in Novy-MacNeal-Nicolle or Schneider\'s insect medium supplemented with 10% v/v foetal calf serum, if properly performed, is a more sensitive technique. Serological tests based on the detection of specific humoral antibodies are less invasive. Such tests include indirect fluorescent antibody test, direct agglutination test (DAT), enzyme-linked immunosorbent assay, and rK39 dipstick test \[[@pmed-0020211-b9],[@pmed-0020211-b10]\]. However, these tests, with the exception of the last, require trained personnel and considerable laboratory facilities. False positives can occur when these tests are used. Furthermore, serological tests may remain positive after successful treatment and give false negative results in patients with VL and HIV co-infection. Nonetheless, serological tests have been used in screening patients (to exclude other causes of febrile hepatosplenomegaly) and to support clinical diagnosis of VL. Control programmes for VL in some countries use diagnostic algorithms that also include DAT and the dipstick test systems. DAT has been used widely under field conditions. The existing dipstick test systems (such as rK39) are attractive options. Preliminary field trials of rK39 have shown promising but geographically variable results. Together with DAT, the rK39 dipstick test and urine antigen detection tests are currently under evaluation by the World Health Organization in different countries. The development of a sensitive, specific, simple, and affordable test for use in field settings will be a crucial step in the control of VL. Thus, the need to improve rK39, and the development of new dipstick systems, is vital. Current Treatments: Old, Toxic, and Difficult to Deliver {#s5} ======================================================== For many decades, the treatment of VL has been based on pentavalent antimonials, such as sodium stibogluconate (Pentostam) or meglumine antimoniate (Glucantime), given intramuscularly or intravenously for one month. Discovered 60 years ago, sodium stibogluconate remains the mainstay treatment of VL despite its cardiotoxicity in some patients. Treatment requires 30 days of intramuscular or intraveneous injections in a hospital setting. Although it is still effective in most endemic countries, with 95% cure rates, resistance is increasing in some regions, especially in northern Bihar, India, where it is up to 65% \[[@pmed-0020211-b11]\]. Other drugs, such as amphotericin B (Fungizone), liposomal amphotericin B (AmBisome), and miltefosine (Impavido), are available for the treatment of VL but are not optimal due to problems of toxicity, high price, or difficulty in administration. Amphotericin B {#s5a} -------------- This drug is highly efficacious but is associated with serious side effects and can only be given in a hospital setting. Liposomal amphotericin B {#s5b} ------------------------ This is considered to be the most effective of currently available anti-leishmanial drugs, but it is prohibitively expensive and has to be administered intravenously, making treatment more difficult under field conditions. However, recent clinical studies in India involving 203 patients showed that liposomal amphotericin B could be used as a single-dose treatment regime with a cure rate of 90% \[[@pmed-0020211-b12]\]. Such data are needed from African endemic areas, as it might be that response to liposomal amphotericin B can vary from species to species and in different populations. The varying doses of liposomal amphotericin B needed to achieve a cure in different endemic countries needs careful attention. Miltefosine {#s5c} ----------- This drug is effective against VL but is expensive and teratogenic \[[@pmed-0020211-b13]\], so it cannot be used to treat women of childbearing age. There is a theoretical risk of resistance developing quickly to it, if it is not used in combination with other drugs. Miltefosine is registered in India for first-line treatment of VL, and in Europe for treatment of VL in patients co-infected with HIV, especially in those patients unresponsive to other treatments \[[@pmed-0020211-b14]\]. Developing New Drugs {#s6} ==================== Given the problems associated with the handful of currently available drugs for VL, new and improved treatments to replace or complement existing therapy are needed urgently. Drug combinations for treating VL should provide advantages of protection from parasite resistance, as well as a reduction in treatment duration and overall toxicity. Paromomycin (also known as aminosidine), an antibiotic of the aminoglycoside family with proven anti-leishmanial activity, is a candidate drug for treatment of VL. Early clinical studies in Kenya and India \[[@pmed-0020211-b15]\] have shown that this drug is effective in the treatment of VL. The current treatment regimen for paromomycin is 21 days when used as a single agent, but could be reduced to 17 days when used in combination with sodium stibogluconate, as field experience of Médecins Sans Frontières has shown (unpublished data). The Drugs for Neglected Diseases Initiative is currently carrying out phase III clinical trials of paromomycin in east Africa with a view to registering the drug in Ethiopia, Sudan, and Kenya. The Institute for OneWorld Health, a nonprofit pharmaceutical company, has conducted phase III trials of paromomycin and is pursuing registration in India (see <http://www.oneworldhealth.org/diseases/leishmaniasis.php>). Vaccines: Progress and Frustrations {#s7} =================================== Extensive studies on the mechanisms of immuno-pathogenesis and protective immunity against leishmaniasis, especially in mice, have identified Leishmania species as good candidates for vaccine development. Leishmania species rarely undergo antigenic variation and show extensive cross-reactivity between different species \[[@pmed-0020211-b18],[@pmed-0020211-b19]\]. Furthermore, the observation that strong lifelong immunity follows after recovery from Leishmania infections in humans has provided a rationale for designing immuno-prophylactic strategies against leishmaniasis. It is now a well-established fact that protective immunity to leishmaniasis is a function of cell-mediated immunity mediated by Type 1 T helper cells. Vaccination against cutaneous leishmaniasis (CL) was a common traditional practice in the Middle East and the Soviet Union \[[@pmed-0020211-b20]\]. In this practice, scratched tissue from active lesions of patients with CL was applied to---or sandflies were allowed to bite---the skin of healthy individuals. Modern approaches of vaccination began by intradermal inoculation of live Leishmania in healthy individuals, in an attempt to produce mild, self-healing cutaneous lesions \[[@pmed-0020211-b21]\]. This process is often referred to as leishmanization. This term is also promiscuously used in the literature to describe the traditional practices described above. Self-healed cutaneous lesions induced by leishmanization usually confer protection against new infections. However, when leishmanization was applied to large populations, individuals developed complicated, severe, or persistent cutaneous lesions. Leishmanization has now been abandoned except in Uzbekistan \[[@pmed-0020211-b21]\]. Pioneered by Brazilian investigators, vaccination against Leishmania using killed preparations of the parasite stages has been attempted since the late 1930s \[[@pmed-0020211-b22]\]. Since then, killed-parasite preparations of various species and strains, with or without adjuvants, such as the autoclaved Leishmania major (ALM) + Bacille Calmette-Guérin (BCG) vaccines, have been extensively studied with variable success in Brazil, Venezuela, Colombia, Ecuador, Iran, and Sudan. Even though the first-generation vaccines were safe, efficacy data have not been convincing \[[@pmed-0020211-b23]\]. In Sudan, alum-precipitated ALM + BCG vaccine mixture was extensively studied and confirmed to be superior to ALM + BCG vaccine alone \[[@pmed-0020211-b28]\]. > There has been little progress toward development of vaccines. In general, first-generation vaccines, as attractive as they were, were met by disappointment from the scientific community, resulting in a shift of interest to novel approaches of vaccination using second-generation vaccines (recombinant molecules, and vaccines with live vectors encoding leishmanial antigens and sandfly salivary immunomodulators) \[[@pmed-0020211-b29]\]. Second-generation vaccines are still under development \[[@pmed-0020211-b30]\] with a number of ongoing safety and immunogenicity studies, but efficacy data are not expected before the next three to five years. In spite of the strong scientific conviction that leishmaniasis is prone to control by vaccines, and the extensive vaccine research carried out so far, especially in CL, no effective vaccine has yet emerged. In particular, there has been little progress towards the development of vaccines against VL. The Challenges Ahead {#s8} ==================== Progress towards the discovery of an effective vaccine against leishmaniasis has become a snail\'s race. Therefore, control of leishmaniasis by vaccines remains only a long-term goal \[[@pmed-0020211-b31]\]. Many leishmaniasis experts nowadays advocate vector control, especially for areas of anthroponotic transmission. History relates that in India VL was kept under control, inadvertently, by the large-scale spraying of DDT during anti-malaria campaigns \[[@pmed-0020211-b1]\]. Recent initiatives of the World Health Organization aim to eliminate VL from the Indian subcontinent by house-to-house spraying of DDT and to reduce epidemic CL in Kabul by a massive provision of insecticide-treated nets. Such nets have been used to reduce transmission of anthroponotic CL in Afghanistan \[[@pmed-0020211-b32]\]. Personal protection against the bites of Phlebotomus orientalis by insecticide-treated nets was considered a feasible VL control approach in Sudan \[[@pmed-0020211-b33]\]. In Latin America, and even more so in southern Europe, where VL is principally maintained by the domestic dog, opinions about control of VL are divided. In Southern Europe, the situation is further compounded by the increasing incidence of adult VL that is associated with HIV co-infection. In Africa, VL is transmitted mainly in rural areas either from a zoonotic source (in sporadic endemic areas) or human to human in secondarily anthroponotic foci. Owing to the complexity and diversity of transmission patterns, but also absence of health-care settings, control of VL in the African endemic countries will indeed be challenging. In Ethiopia, HIV co-infection in some endemic areas of VL ranges from 15%--40%, and is known to be much higher in hospitals in big cities. Significant co-infection rates are being documented in Sudan. In these countries, the surveillance of HIV co-infection in VL endemic areas has to be an integral component of national VL control programmes. The VL endemic countries provide a unique challenge to clinical research and development. Although the parasite also occurs in poor semi-urban environments, communities of affected patients are generally remote and far from health services. Government budgets are inadequate and health ministries are overstretched with many calls on their resources. In many areas hospital facilities are absent or underdeveloped. Tools for screening and identification of patients are inadequate. Current diagnostic techniques are invasive and complicated, and require trained staff. Treatments are toxic, expensive, and difficult to administer. These limitations have constrained the improvement of access to treatment. On the other hand, treatment possibilities by single-dose regimens of liposomal amphotericin B as well as the availability of miltefosine as an oral treatment of VL may provide opportunities for the development of simplified treatment regimes. Vector control can be a useful approach to reduce the incidence of VL. Nonetheless, this is easier said than done, given the huge amounts of funds required, as well as the absence of practical decision support systems in VL endemic areas. Aside from availability of up-to-date information on VL distribution, health policymakers and health workers should be able to carry out efficient and effective vector control programmes and to properly monitor impact \[[@pmed-0020211-b34]\]. Conclusion {#s9} ========== If we are to have an impact on the incidence of the disease and curtail the negative socio-economic consequences of VL epidemics that deter human development, the international community has to address the many issues that we have raised in this article. To achieve this complex and difficult task, substantial funding will be needed, and the task will require international cooperation for success. Glossary -------- Domestic zoonosis: Infection occurring via a domestic animal, e.g., dog. Sylvatic zoonosis: Infection occurring via a feral mammal, e.g., mongoose in Sudan. Anthroponosis: Direct human-to-human transmission by vector without an animal reservoir. Citation: Hailu A, Musa AM, Royce C, Wasunna M (2005) Visceral leishmaniasis: New health tools are needed. PLoS Med 2(7): e211. ALM : autoclaved Leishmania major BCG : Bacille Calmette-Guérin CL : cutaneous leishmaniasis DAT : direct agglutination test FAST : fast agglutination screening test GIS : Geographic Information System VL : visceral leishmaniasis [^1]: Asrat Hailu is at the Faculty of Medicine, Addis Ababa University, Addis Ababa, Ethiopia. Ahmed Mudawi Musa is in the Leishmaniasis Research Group, Institute of Endemic Diseases, University of Khartoum in Sudan. Catherine Royce is with the Drugs for Neglected Diseases Initiative in Geneva, Switzerland. Monique Wasunna is at the Kenya Medical Research Institute in Nairobi, Kenya. [^2]: Competing Interests: The authors are all members of the Leishmaniasis East Africa Platform and are currently engaged in the clinical development of paromomycin for treatment of visceral leishmaniasis. Paromomycin is manufactured as a low-cost generic drug by Grand Pharma, Hyderabad, India, and distributed by International Dispensary Association, Amsterdam, the Netherlands. The Drugs for Neglected Diseases Initiative is a not-for-profit foundation, as is the International Dispensary Association. None of the authors has any financial interest in Grand Pharma.	PubMed Central	2024-06-05T03:55:59.871618	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181879/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e211", "authors": [ { "first": "Asrat", "last": "Hailu" }, { "first": "Ahmed", "last": "Mudawi Musa" }, { "first": "Catherine", "last": "Royce" }, { "first": "Monique", "last": "Wasunna" } ] }
PMC1181880	Research into the etiology of schizophrenia has never been as interesting or as provocative as in the past three years. There has been progress on several fronts, but particularly regarding the molecular genetics of this complex disorder of mind and brain. At the same time, a number of critically important and unresolved issues remain that qualify the ultimate clinical and scientific validity of the results. However, the recent progress in this historically difficult area of inquiry does not seem to be widely appreciated. The purpose of this article is to provide a high-level review of progress, its limitations, and the implications for clinical research and clinical practice. The public health importance of schizophrenia is clear. The median lifetime prevalence of schizophrenia is 0.7--0.8% \[[@pmed-0020212-b1]\], with onset typically ranging from adolescence to early adulthood and a course of illness typified by exacerbations, remissions, and substantial residual symptoms and functional impairment \[[@pmed-0020212-b2]\]. Morbidity is substantial, and schizophrenia ranks ninth in global burden of illness \[[@pmed-0020212-b3]\]. In addition, schizophrenia is often comorbid with drug dependence (principally alcohol, nicotine, cannabis, and cocaine) and important medical conditions (obesity, Type 2 diabetes mellitus) \[[@pmed-0020212-b4]\]. Mortality due to natural and unnatural causes is considerable, and the projected lifespan for individuals with schizophrenia is some 15 years less than the general population \[[@pmed-0020212-b5]\]. The personal, familial, and societal costs of schizophrenia are enormous. Etiological Clues {#s2} ================= A substantial body of epidemiological research has established a set of risk factors for schizophrenia. A subset of this work is summarized in [Figure 1](#pmed-0020212-g001){ref-type="fig"}. Of a large set of pre- and antenatal risk factors \[[@pmed-0020212-b6]\], having a first-degree relative with schizophrenia is associated with an odds ratio of almost ten. The general impact of some of the risk factors in [Figure 1](#pmed-0020212-g001){ref-type="fig"} remains uncertain, and, additionally, migrant status, urban residence, cannabis use, and biological sex are supported as risk factors for schizophrenia. Although the attributable risk of some of these risk factors may be greater (e.g., place and season of birth) \[[@pmed-0020212-b7]\], the size of the odds ratio for family history suggests that searching for the familial determinants of schizophrenia is rational for etiological research. ::: {#pmed-0020212-g001 .fig} Figure 1 ::: {.caption} ###### Comparison of a Selected Set of Relatively Well-Established Risk Factors for Schizophrenia, Focusing Mainly on Pre- and Antenatal Factors \[[@pmed-0020212-b6]\] (abbreviations: CNS, central nervous system; depr, depression; Rh, Rhesus) ::: ![](pmed.0020212.g001) ::: Unpacking the Family History Risk Factor {#s3} ======================================== Studies of families, adoptees, and twins have been widely used to attempt to understand the relative contributions of genetic and environmental effects upon risk for schizophrenia. These "old genetics" approaches use phenotypic resemblance of relatives as an indirect means by which to infer the roles of genes and environment. There are many important assumptions and methodological issues with these studies \[[@pmed-0020212-b8]\]; however, genetic epidemiological studies of schizophrenia have yielded a remarkably consistent set of findings, as summarized in [Table 1](#pmed-0020212-t001){ref-type="table"} \[[@pmed-0020212-b9], [@pmed-0020212-b10]\]. ::: {#pmed-0020212-t001 .table-wrap} Table 1 ::: {.caption} ###### Summary of Studies of the Genetic Epidemiology of Schizophrenia ::: ![](pmed.0020212.t001) ::: To summarize this literature briefly, schizophrenia is familial, or "runs" in families. Both adoption and twin studies indicate that the familiality of schizophrenia is due mainly to genetic effects. Twin studies suggest the relevance of small but significant shared environmental influences that are likely prenatal in origin. Thus, schizophrenia is best viewed as a complex trait resulting from both genetic and environmental etiological influences. These results are only broadly informative, as they provide no information about the location of the genes or the identity of the environmental factors that predispose or protect against schizophrenia. Searching for genetic influences that mediate vulnerability to schizophrenia is rational, given the larger overall effect size and lesser error of measurement in comparison to typical assessments of environmental effects. Note that high heritability is no guarantee of success in efforts to identify candidate genes. Genomewide Linkage Studies of Schizophrenia {#s4} =========================================== Modern genotyping technologies and statistical analyses have enabled the discovery of genetic loci related to the etiology of many complex traits \[[@pmed-0020212-b11]\], such as Type 2 diabetes mellitus, obesity, and Alzheimer\'s disease. These "discovery science" approaches have been applied to schizophrenia, and are summarized in [Figure 2](#pmed-0020212-g002){ref-type="fig"}. The 27 samples shown here included from one to 294 multiplex pedigrees (see Glossary) (median 34) containing 32 to 669 (median 101) individuals affected with a narrow definition of schizophrenia. There were 310 to 950 (median 392) genetic markers in the first-stage genome scans. ::: {#pmed-0020212-g002 .fig} Figure 2 ::: {.caption} ###### Summary of Genomewide Linkage Studies of Schizophrenia The x-axis shows the location on the genome, from the telomere of the short arm of Chromosome 1 to the telomere of the long arm of Chromosome 22 (bottom row) along with 303 band chromosomal staining on the second-to-bottom row. The y-axis shows the 27 primary samples that reported first-stage genome scans for schizophrenia (i.e., excluding fine-mapping or partial reports) along with the results of a meta-analysis including most of the primary samples \[[@pmed-0020212-b12]\] (studies not included are shown with asterisks). Within each row, the height and color of the bars are proportional to the --log~10~(P-value), and the width of the bar shows the genomic location implicated by a particular sample. A selected set of candidate genes for schizophrenia are also shown. All genomic locations are per the hg16 build (<http://genome.ucsc.edu>). The physical positions of an inclusive set of the markers showing the best findings in the primary samples were plotted (assuming a confidence interval of ± 10 cM or, if mapping was uncertain, ± 10 megabases; seven markers from the primary samples did not map). ::: ![](pmed.0020212.g002) ::: "Hard" replication---implication of the same markers, alleles, and haplotypes in the majority of samples---is elusive. It is evident from [Figure 2](#pmed-0020212-g002){ref-type="fig"} that these studies are inconsistent, and no genomic region was implicated in more than four of the 27 samples. The Lewis et al. meta-analysis \[[@pmed-0020212-b12]\] included most of the studies in [Figure 2](#pmed-0020212-g002){ref-type="fig"} and found that one region on Chromosome 2 was stringently significant and several additional regions neared significance. Our focus on first-stage genome scans does not adequately capture the evidence supporting replication for certain regions (e.g., 6p) \[[@pmed-0020212-b13]\]. However, there appears to be "soft" replication across studies. It is unlikely that all of these linkage findings are true. The regions suggested by the Lewis et al. meta-analysis implicate more than 3,000 genes (18% of all known genes). For the 27 samples in [Figure 2](#pmed-0020212-g002){ref-type="fig"}, the percentages of all known genes implicated by 0, 1, 2, 3, and 4 linkage studies were 42%, 35%, 14%, 6%, and 3%, respectively. This crude summation suggests that linkage analysis is an imprecise tool---implausibly large numbers of genes are implicated and few genes are consistently identified in more than a small subset of studies. There are several potential reasons why clear-cut or "hard" replication was not found. With respect to the teams that conducted these enormously effortful studies, it is possible that no study possessed sufficient statistical power to detect the subtle genetic effects suspected for schizophrenia. For example, it would require 4,900 pedigrees to have 80% power to detect a locus accounting for 5% of variance in liability to schizophrenia at α = 0.001. These calculations make highly optimistic assumptions, and less favorable assumptions can lead to sample size requirements above 50,000 sibling pairs. For comparison, the total number of pedigrees in [Figure 2](#pmed-0020212-g002){ref-type="fig"} is less than 2,000. In addition, it is possible that etiological heterogeneity (different combinations of genetic and environmental causes between samples) and technical differences (different ascertainment, assessment, genotyping, and statistical analysis between samples) contributed; however, their impact is uncertain, whereas insufficient power is clear. If correct, the implication is that [Figure 2](#pmed-0020212-g002){ref-type="fig"} contains a mix of true and false positive findings. Association Studies of Schizophrenia {#s5} ==================================== Schizophrenia---like most other complex traits in biomedicine---has had a large number of genetic case-control association studies \[[@pmed-0020212-b19]\]. Although research practice is changing, interpretation of many studies is hindered by small sample sizes and a tendency to genotype a single genetic marker of the hundreds that might be available in a gene. For example, a widely studied functional genetic marker in COMT (rs4680) is probably not associated with schizophrenia \[[@pmed-0020212-b20]\], but nearby genetic markers assessed in a minority of studies may be \[[@pmed-0020212-b21]\]. However, as discussed in the next section, a number of methodologically adequate association studies of schizophrenia appear to support the role of several candidate genes in the etiology of schizophrenia. Similar to the linkage study data, "hard" replication remains elusive. Synthesis {#s6} ========= Despite the limitations of the accumulated linkage and association studies, there are good suggestions that these studies have identified plausible candidate genes for schizophrenia. [Table 2](#pmed-0020212-t002){ref-type="table"} summarizes the evidence in support of a set of possible candidate genes for schizophrenia. Reports supporting the role of many of these genes have appeared in top-tier international journals known for rigorous peer review. The evidence for several genes is encouraging but currently insufficient to declare any a clear-cut cause of schizophrenia. ::: {#pmed-0020212-t002 .table-wrap} Table 2 ::: {.caption} ###### Evidence Supporting 12 Potential Candidate Genes for Schizophrenia ::: ![](pmed.0020212.t002) ^1^Standard gene name (<http://www.gene.ucl.ac.uk/nomenclature>). ^2^Online Mendelian Inheritance in Man (<http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=OMIM>). ^3^Gene lies in a genomic region ("bin") implicated at a suggestive level in the Lewis et al. meta-analysis \[[@pmed-0020212-b12]\]. ^4^Evidence here includes studies not found in Figure 2 (e.g., fine-mapping studies or studies targeted to a particular region). ^5^+, positive study; −, negative study. ^6^From the Novartis Research Foundation (<http://symatlas.gnf.org>). +, expression above median over all tissues; ++, expression above 75th percentile. ::: The accumulated data provide particular support for DISC1, DTNBP1, NRG1, and RGS4. Each of these genes has received support from multiple lines of evidence with imperfect consistency: 1) The case for each of these as a candidate gene for schizophrenia is supported by linkage studies; 2) The preponderance of association study findings provides further support; 3) mRNA from each gene is expressed in the prefrontal cortex as well as in other areas of the brain; and 4) Additional neurobiological data link the functions of these genes to biological processes thought to be related to schizophrenia. For example, DISC1 modulates neurite outgrowth, there is an extensive literature on the involvement of NRG1 in the development of the CNS, and RGS4 may modulate intracellular signaling for many G-protein-coupled receptors. Moreover, DTNBP1 and RGS4 have been reported to be differentially expressed in postmortem brain samples of individuals with schizophrenia. This encouraging summation of work in progress masks a critical issue---the lack or consistent replication for the same markers and haplotypes across studies. The literature supports the contention that genetic variation in these genes is associated with schizophrenia, but it lacks impressive consistency in the precise genetic regions and alleles implicated. In contrast, association studies of other complex human genetic diseases have produced unambiguous, consistent, and clear-cut ("hard") replication. For example, 1) in Type 1 diabetes mellitus, the bulk of both the linkage and association data implicate the HLA region and INS \[[@pmed-0020212-b22]\]; 2) for Type 2 diabetes mellitus, there are a number of findings in the literature where the association evidence appears to be consistent and compelling (CAPN10, KCNJ11, and PPARG)---the data indicate that the same marker allele is significantly associated and has an effect size of similar direction and magnitude \[[@pmed-0020212-b22]\] (the linkage data are less congruent, probably due to power considerations); and 3) for age-related macular degeneration, at least five studies show highly significant association with the same CFH Y402H polymorphism \[[@pmed-0020212-b23]\] in a region strongly implicated by multiple linkage studies. For these findings, the data are highly compelling and consistent and provide a solid foundation for the next generation of studies to investigate the mechanisms of the gene--phenotype connection. Power/type 2 error appears to be a major factor---if the genetic effect is sufficiently large (HLA in Type 1 diabetes mellitus or CFH in age-related macular degeneration)---or, if the sample size is large, then there appears to be a greater chance of "hard" replication. At present, the data for schizophrenia are confusing, and there are two broad possibilities. The first possibility is that the current findings for some of the best current genes are true. This implies that the genetics of schizophrenia are different from other complex traits in the existence of very high degrees of etiological heterogeneity: schizophrenia is hyper-complex, and we need to invoke more complicated genetic models than other biomedical disorders. The alternative possibility is that the current findings are clouded by Type 1 and Type 2 error. Schizophrenia is similar to other complex traits: it is possible that there are kernels of wheat, but it is highly likely that there is a lot of chaff. At present, the second and more parsimonious possibility has not been rigorously excluded. The impact of Type 1/Type 2 error is likely, and it is not clear why schizophrenia should be inherently more complex. At present, we cannot resolve these possibilities. Public Health Implications {#s7} ========================== The public health importance of schizophrenia is clear, and the rationale for the search for genetic causes is strong. Schizophrenia research has never been easy: the current epoch of investigation into the genetics of schizophrenia provides a set of tantalizing clues, but definitive answers are not yet fully established. Findings from the accumulated literature appear to be more than chance yet sufficiently variable as to render "hard" replication elusive. The currently murky view of this literature may result from the competing filters of Type 1 and Type 2 error. The current literature could be a mix of true and false positive findings; however, it would be a momentous advance for the field if even one of the genes in [Table 2](#pmed-0020212-t002){ref-type="table"} were a true positive result. This body of work is not yet ready for wholesale translation into clinical practice. However, it is not premature to inform patients that this work is advancing and that it holds promise for new insights into etiology, pathophysiology, and treatment on the five- to ten-year horizon. On a larger scale, the treatment of the mentally ill mirrors the humanity of a society; in many societies, the return image is not flattering. If a specific genetic variation were proven to be causal to schizophrenia, this poor reflection might improve \[[@pmed-0020212-b28]\]. GLOSSARY -------- Multiplex pedigree: A family grouping of genetically related individuals with multiple affected individuals.First-stage genome scan: An initial survey of the genome to identify regions that may contain genetic variants that could cause the disease under study. Subsequent stages focus on a smaller genomic region.Type 1 error: The probability of rejecting a true null hypothesis (akin to a false positive result).Type 2 error: The probability of accepting a false null hypothesis (akin to a false negative result). Citation: Sullivan PF (2005) The genetics of schizophrenia. PLoS Med 2(7): e212 [^1]: Patrick F. Sullivan is at the Departments of Genetics, Psychiatry, and Epidemiology at the University of North Carolina at Chapel Hill, United States. E-mail: <pfsulliv@med.unc.edu> [^2]: Competing Interests: The author declares that he has no competing interests.	PubMed Central	2024-06-05T03:55:59.873658	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181880/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e212", "authors": [ { "first": "Patrick F", "last": "Sullivan" } ] }
PMC1181881	Gary Schwitzer: The Agenda-Setting Role of Health Journalists {#s1} ============================================================= Some journalists say that their role and responsibility is no different in covering health information than it is in covering politics, business, or any other topic. These journalists say that their primary concern is accurate, clear reporting---they are less concerned about the consequences of their story once it is published \[[@pmed-0020215-b1]\]. But that approach may result in shoddy journalism and potential harm to the public \[[@pmed-0020215-b2]\]. I assert that it isn\'t sufficient to be accurate and clear when covering health news. Journalists have a responsibility to mirror a society\'s needs and issues, comprehensively and proportionally \[[@pmed-0020215-b3]\]. Often that doesn\'t happen in health news. Recently, I led an effort by the Association of Health Care Journalists to publish a statement of principles \[[@pmed-0020215-b4]\]. "Journalists have a special responsibility in covering health and medical news," the statement reads. "Association members know that readers and viewers may make important health care decisions based on the information provided in our stories." In our current era of entanglement, journalists must investigate and report the possible conflicts of interest among sources of health information and those who promote a new idea or therapy. Such conflicts may not be readily apparent, so journalists must look for them as a routine part of story research and interviews. They must investigate and report the possible links between researchers and private companies, researchers and public institutions, patient advocacy groups and their sponsors, celebrity spokespersons and their sponsors, and nonprofit health and professional organizations and their sponsors. To fail to do so may mean that journalists become unwitting mouthpieces for incomplete, biased, and imbalanced news and information. Journalists face unique challenges in covering health news. Some specialized skills, knowledge, and judgment are helpful. For example, some information based on poorly designed or poorly powered studies should not be reported unless the flaws are emphasized. Editors, reporters, and writers need to scrutinize the terminology used in health news. Vague, sensational terms (such as "cure," "miracle," and "breakthrough") may harm news consumers by misleading and misinforming \[[@pmed-0020215-b5]\]. At the core of journalism\'s values, such terms should not be used because they are meaningless. It is not the role of journalists to become advocates for causes. However, I believe that journalists have a responsibility to investigate and report on citizens\' needs as they struggle to understand and navigate the health-care system. People need help in understanding the ways in which scientists and policymakers reach conclusions. In that sense, there is an inherent educational role that journalists must assume. [](#pmed-0020215-g001){ref-type="fig"} ::: {#pmed-0020215-g001 .fig} ::: {.caption} ###### Journalists risk becoming unwitting mouthpieces for those with vested interests in their story (Illustration: Scott Mickelson) ::: ![](pmed.0020215.g001) ::: I have a special interest in how television journalists cover health and health policy news. Surveys consistently show that many Americans get most of their health news and information from television. One study documented troubling trends of brevity (an average of 45 seconds per story), absence of reporter specialization, sensational claims not supported by data, hyperbole, commercialism, disregard for the uncertainty of clinical trials, baseless predictions of treatments based on basic science studies, single-source stories, and a paucity of coverage of health policy \[[@pmed-0020215-b6]\]. Television viewers are likely to see many more one-sided political ads about health policy issues than balanced, comprehensive news stories about such issues. In my current research, I am analyzing health policy news coverage on three award-winning TV stations in three different parts of the United States in 2004. Despite the fact that American voters ranked health care as their third leading concern (after war and the economy) \[[@pmed-0020215-b7]\], the three stations I monitored devoted little time to health policy issues. My analysis shows that in ten months (326 hours of stations\' key late night newscasts) on these three stations, there was only one story on the uninsured. Presidential candidates\' health policy platforms drew a combined total of seven minutes of news---an average of 23 seconds per story, or about 15 seconds per station per month of the 2004 campaign. Whether it is preclinical news that is not ready for prime time, or clinical news that oozes optimism over unproven ideas, or a disdain for health policy news, television journalists seem to have abdicated their possible agenda-setting role. Journalists must weigh the balance between the amount of attention given news about medicine and the attention given news about health and the social determinants of health. There may be too much news about the delivery of medical services and not enough news about the cost of, quality of, and evidence for those services. The current imbalance may contribute to the nation\'s health-care cost crisis, driving up demand for expensive, unproven ideas. These are responsibilities journalists may not encounter in covering other topics. In health news, they are everyday issues. Ganapati Mudur: The Media May Be the Most Important Source of Health Information for the General Public {#s2} ======================================================================================================= Health reporting does involve "telling a story," but it also requires writers to take on additional responsibilities through the story cycle---finding the story, collecting information, and writing it. Standard news criteria such as timeliness and impact may be used to pick stories. But in health reporting, context is crucial. Research advances to be reported need to be placed in context. This may be achieved by citing earlier research on the topic and seeking out comments from independent experts who could put a new finding in perspective. Sometimes health research throws up contradictory findings. Is a gene linked to a disease? One study finds a link. Another does not. Such situations demand interpretative and analytical skills on the part of health writers. Otherwise, writers may mislead readers, or leave them confused. Health reporters need to find out who has funded the research and who might be likely to gain. And reporters must always double-check claims or else they may end up in embarrassing situations. Let me illustrate with an example. A top international science magazine last year reported that a novel stem cell therapy had cured patients with chronic aplastic anemia in Bombay, India \[[@pmed-0020215-b8]\]. The story was apparently based on claims made by the developers of the therapy, a private British company. A little more patience and investigation would have led the magazine to the real story: none of the patients had responded to the treatment, and the clinical collaborators in India had terminated the study \[[@pmed-0020215-b9]\]. When a public health situation is involved, health writers and the media can certainly play a role in quickly delivering important messages to the public. In a sense, then, they do serve as a component of the health provider community. And this makes it all the more important for health writers to ensure that they get it absolutely right. Given that most people do not interact with their doctors on a regular basis, the media is possibly the most significant source of health information for the general public. But health information in the media cannot substitute for personal medical advice. It is important that the public understands this. Regulatory mechanisms may be lax in some developing countries. India, for instance, has had a long history of unethical or illegal clinical trials. Drug regulatory authorities in India allow the sale of drugs---including pediatric formulations---that have never been approved in Western countries. This opens up opportunities for investigative health journalism, an opportunity for reporters to take up the traditional watchdog role of the press to find and report wrongdoing. David Henry and Amanda Wilson: Health Journalists Should Discuss Benefits and Harms of New Treatments and Use Independent Expert Sources {#s3} ======================================================================================================================================== Health reporting is a major growth area for the media, probably because it is in demand by the public and it is profitable. However, media coverage of medical news is generally of poor quality, particularly stories about new treatments \[[@pmed-0020215-b10]\]. Media Doctor is a Web site where the quality of stories in the Australian press is reviewed ([www.mediadoctor.org.au](www.mediadoctor.org.au)). We rate articles using ten evaluation criteria (see [Box 1](#box1){ref-type="boxed-text"}; [www.mediadoctor.org.au/content/ratinginformation.jsp](www.mediadoctor.org.au/content/ratinginformation.jsp)). In February/March 2005 articles that we rated achieved an average of only 52% on our "satisfactory" score ([www.mediadoctor.org.au/content/sourceinfo.jsp](www.mediadoctor.org.au/content/sourceinfo.jsp)). This was an improvement on the score from one year ago, but it is still inadequate. North American analyses of the quality of health reporting have had similar results \[[@pmed-0020215-b11]\]. The print media are clearly superior to the online news services \[[@pmed-0020215-b12]\]. The greatest differences between print and online services are in the use of independent information sources, and the quantification of the benefits and the coverage of potential harms of new treatments. Box 1. Criteria Used by Media Doctor to Evaluate News Stories ------------------------------------------------------------- Whether the treatment is genuinely newThe availability of the treatment in AustraliaWhether alternative treatment options are mentionedIf there is evidence of disease mongering in the storyIf there is objective evidence to support the treatmentHow the benefits of the treatment are framed (in relative or absolute terms)Whether harms of the treatment are mentioned in the storyWhether costs of the treatment are mentioned in the storyWhether sources of information and any known conflicts of interest of informants are disclosed in the articleWhether the journalist relied only on the press release for the story We recognise that there are different depths of journalism and that journalists face constraints, including commercial pressures and deadlines that give little time to reflect on stories, which are usually written on the same day as the press release arrives. Some journalists argue that the media are the messengers and not the message, and it is up to others to interpret their reporting. To a reporter who might otherwise exercise more caution, a well-written media release from a large public relations company describing a new pharmaceutical product must be attractive when a deadline is imminent. There is no danger that the company will allege plagiarism if it appears, almost intact, under the journalist\'s by-line. And even when they do have the time, journalists face two major challenges---understanding the clinical science and epidemiology, and dealing with powerful vested interests. Vested interests are not unique to medicine, but reporting on a new drug is different from, say, an MP3 player or a dishwasher. People will be intensely interested in a story about a new drug if it purports to treat a condition that they or their relatives have, and the story may become the basis of discussions with their physician and subsequent treatment decisions. We believe that in writing this type of story journalists have special responsibilities to ensure that they provide balanced information for their readers. In Australia, the Press Council believes the matter is of sufficient importance to provide advice to journalists \[[@pmed-0020215-b13]\]. In our view journalists will meet their responsibilities if they cover certain key issues when writing stories about new medical treatments. These include the accurate reporting of the comparative benefits, harms, and costs of the treatment and the extent to which their informants have ties with the manufacturer. It is helpful if journalists use independent expert sources to answer questions about the novelty of the treatment and the availability and efficacy of alternatives, although we acknowledge the practical difficulties in finding independent sources when time is limited. Journalists have indicated to us that they are concerned about these issues and are prepared to look critically at their own practices. It is unclear whether their editors and producers hold the same views and will provide the necessary resources, particularly time to do the job properly. But researchers and medical journal editors have responsibilities too. When reading medical news stories it is sometimes possible to tell whether the researchers and journals have done a good or bad job in communicating the essential facts to journalists. A number of medical journals issue press releases, and these have been found wanting \[[@pmed-0020215-b14]\]. Researchers should consider carefully what they wish to convey about the results of a new study and should ask to see and edit any press releases. We believe the criteria used by Media Doctor to evaluate news stories are a good starting point for researchers and editors. Merrill Goozner: Medical Reporters Must Get Beyond the Hype and Hope When Reporting on the Latest "Breakthrough" {#s4} ================================================================================================================ When I broke into the news business, the financial desk\'s primary source of breaking news was a Dow Jones wire clack-clack-clacking in the corner of the room. A bell rang whenever a major story broke. Sometimes two bells would go off, signaling a really big story. The day the stock market crashed in 1987, the newsroom sounded like St. Peters Square on Easter. I imagine something comparable occurs these days when the advance copies of leading medical journals cross science editors\' computer screens. Stories from the frontiers of medical research can make it onto page one---the most coveted real estate in daily journalism. News magazines have bolstered their sagging bottom lines with an endless stream of cover stories touting the latest breakthroughs in medicine. But is this news all that it is cracked up to be? Have the reporters properly weighed the importance of the studies they\'re touting? Have they asked the tough questions of the researchers and their sponsors to figure out the significance of the results? Have they presented the data in a fashion that is meaningful to health-care consumers? And in an age when most clinical trials are sponsored by private companies, have they fully informed their readers of the researchers\' conflicts of interest? Too often, the answer to these questions is no. Take recent reports from the American Society of Clinical Oncology, which met in mid-May in Orlando. One leading paper reported on a Veterans Administration review of the experience of over 40,000 women in the south central US. "The women taking statins were half as likely to have breast cancer as women who were not taking the drugs," the paper reported \[[@pmed-0020215-b15]\]. Put that way, it sounds like a dramatic reduction. But elsewhere in the story, it was reported that 12 percent of the women were taking the cholesterol-lowering medications and that only 1.4 percent of the total group contracted breast cancer. Only by massaging the numbers could one figure out that physicians would need to put 700 women on statins to eliminate one cancer case (in medical parlance, this is called number needed to treat). It sounds a lot less impressive that way. But the number needed to treat would be a lot more meaningful to women, especially those on tight budgets wondering if it is worth \$1,000 a year for a prescription. Reporting of surrogate endpoints instead of primary endpoints is another way that readers get misled. Reports on cancer drug trials often fall into this trap. A "lifesaving" drug that shrinks tumors by 50 percent sounds a lot better than a chemotherapy agent that prolongs life by two months. The same can be said for bone density and fractures, blood pressure and strokes, and cholesterol levels and heart attacks. While there may be a minor yet statistically significant reduction in the primary endpoint, the trial sponsors prefer to promote the more dramatic-sounding secondary endpoint. Too many reporters prominently feature the less meaningful number, while leaving out or delaying until late in the story the real bottom line \[[@pmed-0020215-b16]\]. Sadly, the media have only lately come around to taking seriously the issue of conflicts of interest in medical science. Last July, the National Heart, Lung, and Blood Institute\'s National Cholesterol Education Project updated its guidelines for cholesterol management. The update, touted in the front page of every major US paper \[[@pmed-0020215-b17]\], called for a dramatic reduction in the cholesterol levels now considered optimal for people who have never had heart disease but are considered moderately at risk. Prescribing physicians using these guidelines will likely put millions more Americans on these drugs in the next few years. Yet three days after the report came out, reporters at Newsday broke the story that eight of nine physicians on the National Cholesterol Education Project panel had financial ties to statin manufacturers, which had the most to gain from the new guidelines. Writing in the Washington Post, former New England Journal of Medicine editor Jerome Kassirer asked, "Why should we swallow what these studies say?" \[[@pmed-0020215-b18]\] The ensuing uproar contributed to a change in policy at the New York Times, which last fall circulated a memo to all reporters encouraging them to always report conflicts of interest of quoted sources in science stories, a policy that leading science and medical journals have had in place for many years \[[@pmed-0020215-b19]\]. In recent years the pharmaceutical and biotechnology industries have responded to complaints about the high cost of drugs by claiming they are needed to finance the medical miracles that are just around the corner. Meanwhile, the increase in life expectancy in the US has slowed and still remains far below other advanced industrial countries. The number of new drugs coming out of industry labs, despite a slight uptick last year, is actually down from a decade ago. In a health-care environment that is increasingly cost-constrained, it shouldn\'t be too much to ask that medical reporters get beyond the hype and hope when reporting on the latest "breakthrough." Maria Simbra: Whatever News Managers Want, Viewers Get---As Medical Reporters Are Pressed to Feed the Media Beast {#s5} ================================================================================================================= Reporters are surpassing doctors as a source of medical information. It\'s no secret health news sells. Producers and news directors take advantage of this to attract an audience for their newscasts. And viewers respond. In a survey by Rodale Press, 39% of the respondents said they turn to TV for health and medical information, and 37% said they would ask a health professional \[[@pmed-0020215-b20]\]. So as audience-appointed proxies for "health professionals," television medical reporters have a daunting task. They must be accurate, authoritative, and compassionate. They also need to understand the terminology, physiology, epidemiology, study design, and statistical analysis to keep health news in context for the viewer. But typically, this doesn\'t happen. The medical industry churns out volumes of information for medical reporters to quickly sift through every day. There\'s a lack of special training for medical journalists (the general assignment reporter can expect to get thrown into the medical beat from time to time). Usually local news reports are under 90 seconds. The pressure for ratings compounds the problem. Medical news is often simplified, or worse, sensationalized, because of industry pressures. Because health news sells, it can be and will be promoted---and in the process, distorted. What is a medical reporter to do? Well, alone, there\'s not much a reporter can do. Like medical errors, the problems with medical journalism are system wide. At the root is a clash of cultures. Medicine tends to be very methodical, slow, and subject to change. But the media want information that\'s definitive, they want it now, and, boy, it better be sensational. Also, people who go into journalism and ascend to management tend to be more inclined toward writing and creative interests. They may not understand (or may be openly hostile toward) the scientific process. [](#pmed-0020215-g002){ref-type="fig"} ::: {#pmed-0020215-g002 .fig} ::: {.caption} ###### TV reporters rarely cover medicine exclusively---one day it\'s finance, the next it\'s health (Illustration: Giovanni Maki) ::: ![](pmed.0020215.g002) ::: For TV reporters who are committed to the medical beat, educational opportunities are available through organizations such as the Association of Health Care Journalists ([www.ahcj.umn.edu](www.ahcj.umn.edu)) and the National Association of Medical Communicators ([www.ibiblio.org/namc](www.ibiblio.org/namc)). Journalists can learn how to interpret studies and present evidence-based balance in order to help viewers understand and make up their own minds about the latest developments in medicine, rather than just show the gee-whiz side of new technology. Unfortunately, it\'s rare to find a TV reporter who exclusively covers medicine. Stations view this as a luxury. It is more common for medical reporters to also be general-assignment reporters or anchors, and they have other priorities with their combination jobs. It\'s also rare to find management that\'s supportive of continuing education for their reporters. Even with the large profit margins at many TV stations, news directors generally do not provide financial support and ask that reporters interested in attending educational meetings use their own vacation time to do so. Furthermore, news directors, producers, and promotions staff don\'t seem to be themselves interested in learning about medical reporting. An article in JAMA says that viewers are acting on and making personal medical decisions based on health information in the mass media \[[@pmed-0020215-b21]\]. This trend has led some TV medical reporting experts, such as Gary Schwitzer (formerly of CNN, now at the University of Minnesota, and a contributor to this PLoS Medicine debate) and Dr. Timothy Johnson (ABC News), to call for credentialing of medical journalists. After all, some meteorologists are credentialed. Are personal health decisions less important than the weather? There\'s a disconnect between what station management values, what the reporters need, and what the viewers get. Right or wrong, the audience looks to TV medical reporters to educate and guide them on medical issues. It\'s an important responsibility that medical reporters and the mass media in general need to take seriously. Melissa Sweet: Remember the Commercial Imperative When Examining Media Coverage of Health {#s6} ========================================================================================= Many people make the mistake of using the terms "journalists" and "the media" interchangeably. They speak, in the same breath, of the terrible failings of journalists and the media in covering health or other issues. In so doing, they fail to make a distinction between the craft or profession of journalism and the competitive industry that is the media. They fail to understand that the goals and drivers of journalism are often in conflict with those of the media industry. The foremost goal of the media industry is, not surprisingly, to make profit. Many journalists are too idealistic to admit, even to themselves, that their job is to make money for their employers. Some believe they are there for the public interest, or even to interest the public. Some simply love to tell a yarn, to get the buzz that comes with uncovering a great story and breaking news. Some no doubt also come to enjoy the reflected glory of associating with the famous and the powerful. Indeed, many journalists have become celebrities themselves. Not coincidentally, this has benefits for their employers---nothing sells like celebrity. But only a brave, naïve, or foolhardy journalist would publicly admit these days to believing that one of their roles is to help provide a voice for those who otherwise have difficulty having their voices heard, such as the disadvantaged. It is not a career-enhancing move at a time when many media proprietors have decided that a key to improving profits lies with their so-called AB audiences. For those not up-to-date with marketing jargon, AB is shorthand for the affluent professionals so beloved by cashed-up advertisers. The theory goes that media outlets that attract audiences at the AB end of the socioeconomic scale are more likely to win advertisers or, even better, to get away with charging them premium rates. A senior manager at one of Australia\'s major newspaper groups recently explained why his company is focusing on boosting AB readership rather than total circulation \[[@pmed-0020215-b22]\]. "A good circulation result is one which attracts the readership we need and advertisers want," said Mark Scott, Editor-in-Chief of Fairfax\'s metropolitan newspapers, which include the respected broadsheet, The Sydney Morning Herald. "Sure, The Daily Telegraph \[a tabloid\] sells many more copies than The Sydney Morning Herald," said Scott, "but their ad rates are lower because the SMH \[Sydney Morning Herald\] has that AB audience." Scott said Fairfax\'s Sunday title The Sun-Herald is significantly more profitable at its present circulation level of about 513,000 than it was when it was selling 600,000 copies. "We have held that AB audience so our advertising revenue is up and our costs are lower." So what has this to do with how the media report health? Scott explained that his newspapers do extensive market research so they know what the AB market wants to read and how they want it presented. "We create our papers with those readers in mind and shape our marketing and promotions to reinforce their values and interests." In other words, the allocation of scarce resources in ever more stretched newsrooms is driven by what market researchers tell media managers about what AB audiences want to know about. This has implications for how the media cover all the areas that affect peoples\' health---politics, economics, and education, for example---as well as the coverage of health issues themselves. I haven\'t seen the market research, but it\'s not hard to guess what interests AB groups. They might want to know how to stay as healthy, smart, and good-looking as possible for as long as possible. They might want to know which biotech companies are good investments, and might be particularly interested in private health care. They are probably less interested in the needs of indigenous people, prisoners, the homeless, asylum seekers, or the poor, and it\'s probably a fair bet to say that they are also less interested in the ways in which disadvantaged groups have worse access to health care and prevention efforts. Some might think this is overly cynical. Perhaps AB people are not all self-centred; perhaps they care about broader issues than those that directly affect their own lives and personal well-being. Nor can the compliance of journalists be assumed. In the chaotic and anarchic world of journalism, there are many who try to do far more with their jobs than to make their bosses wealthy---even if they have to try and "sell" their stories to their news managers on the grounds that the stories will be of interest to the ABs. Many other factors also shape and influence news production. And a truly compelling story is likely to get a run, even if not of direct relevance to the wealthy. Nonetheless, it is important to remember the commercial imperative when examining media coverage of health. Many initiatives aimed at influencing health coverage target journalists, who are only one component of the media industry. Other powerful forces also shape how health is covered. An analogy can be drawn with efforts to improve the quality and safety of health care, another chaotic industry. Measures aimed at individual clinicians may be helpful in reducing medical errors, but it is also important to look at the broader system in which clinicians work. Katherine A. Baverstock: The Media Can Play a Special Role in Providing a Voice for People to Express Their Experiences of Illness {#s7} ================================================================================================================================== A registered pharmacist for the last 15 years, I was trained in the biomedical model of health, to measure and note signs and symptoms, make assessments, and advise about treatments on the basis of available scientific evidence. Becoming interested in the portrayal of medicines in the media whilst working in outback Australia (which is grossly underserved by health professionals), I began my doctoral project within this quantitative biomedical tradition. As I found during my literature review, research arising from this tradition assesses media writing about medicines for "quality." Such research focuses on certain categories of quantitative information about the medicine, such as the indication, associated risks and benefits, outcomes of treatment, contraindications, and cost, that would allow readers to analyse the evidence for themselves and decide whether they should use the medicine. The research in this area seems to be advocating a position for the health journalist as an educator. Australia\'s Quality Use of Medicines Strategy \[[@pmed-0020215-b23]\] has the objective of optimising the use of medicines within the Australian community. It lists the media as a partner in the strategy, together with consumers, health professionals, government, and the pharmaceutical industry. Similar to the other partners, the media have special responsibilities to ensure the quality use of medicines, as described in [Box 2](#box2){ref-type="boxed-text"}. Although many of these responsibilities sit comfortably within the codes of ethics observed by working journalists, some of these responsibilities made me uncomfortable as a health professional. Should journalists be viewed as de facto health educators with the same responsibilities as those of us in the registered health professions? Box 2. Australia\'s National Strategy for Quality Use of Medicines: Responsibilities of the Media ------------------------------------------------------------------------------------------------- The media are responsible for the following. Ethical and responsible reporting on health-care issuesReporting on medicines accurately and attempting to have errors corrected if they occurBeing aware of the variety of available information sources on medicines and the limitations of each sourceBeing aware of the impact of media reports on the use of medicines in the communityBeing aware of issues relevant to the broad context of medicine use, including risks of medicine use, non-drug alternatives, and the cost of medicine use to individuals and societyEncouraging dissemination of messages that enhance the quality of medication use Source: \[[@pmed-0020215-b23]\]. As I progressed in this quantitative framework, I began to feel more and more uncomfortable with the narrow examination of the newspaper stories I had collected for my research. As my analysis continued, it became apparent that the newspaper stories contained themes far richer and more interesting than quantitative information about how drugs work. The stories were an intriguing insight into how the community viewed issues surrounding medicines and the use of medicines. Even more interesting were the narratives about experiences with medicines. I decided to transfer my research to a communications faculty, and explored a far different perspective of medicines in the media. One of my first realisations was that the media are much more than the newspapers, television, and radio focussed on by so many biomedical researchers. They also include new media (the Internet), other print media, and small-scale media, such as leaflets and posters, and even the messages on the pens given out by pharmaceutical representatives. I would like to propose that rather than act as educators, the media can play a special role in providing a voice for people to express their experiences of illness and their interactions with the technologies of health. The advent of the Internet has democratised the media because this medium is accessible to everyone. The Internet can cross national boundaries and counteract isolation---not only geographic isolation, but also the isolation that may be caused by the experience of chronic illness and not knowing anyone who has lived your experience. People who were unable to have their stories heard within the traditional medical consultation now have a forum where they can be heard and have their stories validated. There is much research published within the sociological and anthropological literature examining the narrative surrounding health and illness within various types of media. Research now needs to examine how patients use information they find within the media, and whether it does make a difference to the medical encounter. Will an informed and questioning client leave us feeling threatened? Within the traditional health setting, lengthy communication between medical professionals and clients is often not possible. Many health professionals receive scant training in communication and counselling. The use of media technologies allows our clients to tell their story, a biography that may be ever-changing because of the experience of chronic illness. I would argue, that rather than being much maligned by health professionals, the media should be viewed as a tool that allows healing by facilitating the telling of stories. Citation: Schwitzer G, Mudur G, Henry D, Wilson A, Goozner M, et al. (2005) What are the roles and responsibilities of the media in disseminating health information? PLoS Med 2(7): e215. [^1]: Gary Schwitzer is an assistant professor and director of the Health Journalism Graduate Program at the University of Minnesota School of Journalism and Mass Communication, Minneapolis, Minnesota, United States of America. He worked in television news for 15 years, including time as the head of the Cable News Network (CNN) medical news unit. E-mail: <schwitz@umn.edu>. Ganapati Mudur is a special correspondent with The Telegraph, New Delhi, India. E-mail: <gsmudur@hotmail.com>. David Henry (E-mail: <david.henry@newcastle.edu.au>) is a professor of clinical pharmacology at the University of Newcastle, and Amanda Wilson is a medical writer and researcher; both are on the team of the Media Doctor project (<http://www.mediadoctor.org.au>), Newcastle Institute of Public Health, University of Newcastle, New South Wales, Australia. Merrill Goozner, a former journalist, is the director of the Integrity in Science Project at the Center for Science in the Public Interest (<http://www.cspinet.org>), Washington, D. C., United States of America. E-mail: <mgoozner@cspinet.org>. Maria Simbra is a neurologist and formally trained journalist. She is a medical reporter for KDKA-TV in Pittsburgh, Pennsylvania, and an adjunct professor of medical journalism at Point Park University, Pittsburgh, Pennsylvania, United States of Americal. E-mail: <msimbra@kdka.com>. Melissa Sweet is a freelance writer in Sydney, Queensland, Australia, who specializes in covering health and medical issues. Katherine A. Baverstock is a lecturer in pharmacy and pharmacology at Charles Sturt University in Wagga Wagga, New South Wales, Australia, and is completing a doctorate in social sciences at the University of South Australia studying the portrayal of prescription medicines in the Australian print media from a cultural studies perspective. E-mail: <kbaverstock@csu.edu.au>. [^2]: Competing Interests: The Integrity in Science Project, of which MG is the director, is funded in part by several foundations with a pro-environmentalist concern about industry influence over the scientific process and government advisory process. MS has previously worked at The Sydney Morning Herald. KB is on the Executive Committee of the Australasian Medical Writers Association.	PubMed Central	2024-06-05T03:55:59.875168	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181881/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e215", "authors": [ { "first": "Gary", "last": "Schwitzer" }, { "first": "Ganapati", "last": "Mudur" }, { "first": "David", "last": "Henry" }, { "first": "Amanda", "last": "Wilson" }, { "first": "Merrill", "last": "Goozner" }, { "first": "Maria", "last": "Simbra" }, { "first": "Melissa", "last": "Sweet" }, { "first": "Katherine A", "last": "Baverstock" } ] }
PMC1181882	Sadly, I fully agree with Richard Smith\'s opinion \[[@pmed-0020218-b1]\] that many medical journals have become marketing offices of pharmaceutical companies. Even worse is that few people seem to realize this, and there are many respectable academics who would wholeheartedly dispute these views. The "psychology of gift" operates in every social environment, being so pervasive because it is based on the profoundly human and universal norm of reciprocity \[[@pmed-0020218-b2],[@pmed-0020218-b3]\]. In academia and medical publishing it produces great returns to the pharmaceutical industry via clinical scientists who are not dishonest but in a state of denial about their motivations, as Jerome Kassirer, former editor of the New England Journal of Medicine, describes very clearly in his recent book \[[@pmed-0020218-b4]\]. What is happening is also extremely serious because the tainted trials we are offered can make evidence-based medicine a pointless enterprise. Again, something not widely appreciated. I think Smith\'s suggestions that there should be more public funding for clinical trials and that journals should critique rather than publish the results of the trials are very interesting. However, how are we going to get publicly, and adequately, funded trials given the current financial climate? A practical alternative to Smith\'s suggestion is to play one pharmaceutical company against the other in head-to-head trials, an approach that can help retain independence and that has been used to this purpose before. Another strategy would be for either the regulatory authorities or the academic review boards to demand of the pharmaceutical companies that whenever a new drug is tested in a phase III trial, it should always be done not only against placebo but also against the drug that the condition is generally treated with. This active comparator must be used at the appropriate dose, namely neither too low nor too high (in order to avoid the tested drug spuriously seeming more effective or safe). Moreover, any new licence should be accompanied by a legal requirement for a stringent system of post-marketing surveillance, to be run by the drug company but overviewed by the regulatory authority. These strategies could possibly help produce results that are more reliable from a scientific point of view, help reduce the number of expensive but not innovative "me too" drugs, and help protect patients\' safety more efficiently. Lastly, I hope Richard Smith will go straight into the lion\'s den and send his thoughts not only to the similarly minded editors and readers of PLoS Medicine but to some of the journals who are the most culpable of the policy he is exposing. Citation: Frighi V (2005) Medical journals, academia, and industry-sponsored clinical trials. PLoS Med 2(7): e218. [^1]: Competing Interests: The author has declared that no competing interests exist.	PubMed Central	2024-06-05T03:55:59.878994	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181882/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e218", "authors": [ { "first": "Valeria", "last": "Frighi" } ] }
PMC1181883	I agree with Richard Smith \[[@pmed-0020219-b1]\] that something needs to be done about the reporting of pharmaceutical industry trials. Like him, I believe that the solution should include compulsory publication on the Internet of trials \[[@pmed-0020219-b2]\]. However, I disagree that the problem has its origin with the pharmaceutical industry; it is inherent to medical publication. Of his eight ways of massaging data, the last five are dealt with by the International Conference on Harmonisation guidelines covering statistical principles for clinical trials (ICH E9)\[[@pmed-0020219-b3]\] that require prespecification of analyses. It is not possible to claim noninferiority on the basis of failure to prove a difference, and a paper describing appropriate approaches to equivalence trials that Richard Smith thought worth publishing in the BMJ \[[@pmed-0020219-b4]\] was doing no more than explaining what was common practice within the industry. The first three points are less easily policed, although choice of control group is taken extremely seriously, and, indeed, there is an appropriately entitled guideline \[[@pmed-0020219-b5]\] that covers this. The problems are inherent to publication not drug regulation. An instance: the New England Journal of Medicine published in January 2002 a paper claiming that voriconazole is a suitable alternative to amphotericin B preparations for empirical antifungal therapy in patients with neutropenia and persistent fever \[[@pmed-0020219-b6]\]. However, a letter to the editor in the same issue of that journal from scientists based at the United States Food and Drug Administration \[[@pmed-0020219-b7]\] pointed out that the analysis presented was not what was prespecified in the protocol and that not only had voriconazole failed to demonstrate noninferiority, but it was actually statistically significantly inferior to amphotericin B. Surely, responsibility for this discrepancy cannot be laid at the door of the Food and Drug Administration, nor can it be blamed on Pfizer. Rather, the authors and the New England Journal of Medicine owe readers some sort of explanation. Do the editors agree with Richard Smith (and me) that a published paper, whatever else it covers, should always identify the results of prespecified analysis, and if so, how do they check that this is so? Thus, I agree with Richard Smith that much is wrong with the publication of clinical trials sponsored by the pharmaceutical industry. I disagree that it is a particular problem for industry trials. It is the publication process that is in need of reform, and in particular we need to scrutinize carefully the motives of authors in publishing and the standards that editors apply in deciding what gets published. Citation: Senn S (2005) Bitter pills and puffed trials. PLoS Med 2(7): e219. [^1]: Competing Interests: SS has consulted for Actelion, Alcon, Amgen, Astra (now Astra-Zeneca), Astra-Zeneca, Aventis, Auxilium, Biosyn, Boehringer Ingelheim, Bracco, Bristol-Myer Squibb, Chiesi, Chiron, Ciba-Geigy (now Novartis), Covance, Dexcel, Elan, Eli-Lilly, Fournier, Glaxo-Wellcome (now GSK), GSK, INO Therapeutics, Janssen, Johnson&Johnson, Jouveinal (now Pfizer), Leiras, Merz, Novartis, Novartis Consumer Health, Novartis Ophthamology, Numico, Orion, Pleaid, Pfizer, Pharmacia (now Pfizer), Pharmapart, Phocus, Roche, Sandoz (now Novartis), Sanofi, Servier, Schein, Schering AG, Shire, Smith-Kline Beecham (now GSK), Statwood, Strakan, Wyeth, Zeneca (now Astra-Zeneca) and possibly other companies he has forgotten about. He used to work for Ciba-Geigy and in consequence owns some shares in Novartis. This note has been prepared without consulting any of the above companies, without their knowledge, and without their express permission, and none of his views should be attributed to any of the above. SS is an academic whose career is furthered by publishing. His views should also not be ascribed to his current employer, Glasgow University.	PubMed Central	2024-06-05T03:55:59.879610	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181883/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e219", "authors": [ { "first": "Stephen", "last": "Senn" } ] }
PMC1181884	Smithereens are better than no Smith at all. It was grand to see Richard Smith in full flight again \[[@pmed-0020220-b1]\], a raptor this time, relegating the randomized controlled trials he previously championed in the BMJ to the ether, to be replaced by printed "commentaries." In doing so, he laid three problematic eggs. First, he shoved systematic reviews and meta-analyses, surely the least biased summaries of efficacy, out of the nest before he took off. Second, the canaries who write commentaries often live in gilded cages provided by the drug industry and printing their pronouncements would make matters worse. Finally, the fledglings who conduct nondrug health-care trials, especially in low- and middle-income countries, shouldn\'t have their careers stunted by not being able to publish their work in print journals. Citation: Sackett D (2005) Might banning trial publication do more harm than good? PLoS Med 2(7): e220. [^1]: Competing Interests: DS has been wined, dined, supported, transported, and paid to speak by countless pharmaceutical firms for over 40 years, beginning with two research fellowships and interest-free loans that allowed him to finish medical school. Dozens of his randomized trials have been supported in part (but never in whole) by pharmaceutical firms, who never received or analysed primary data and never had veto power over any reports, presentations, or publications of the results. He has twice worked as a paid consultant to advise pharmaceutical firms on whether their products caused lethal side effects; on both occasions he told them yes. He has testified as an unpaid expert witness for a patient with stroke who successfully sued a manufacturer of oral contraceptives, and as a paid expert in preparing a class-action suit against a manufacturer of prosthetic heart valves.	PubMed Central	2024-06-05T03:55:59.880160	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181884/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e220", "authors": [ { "first": "David", "last": "Sackett" } ] }
PMC1181885	One solution for fair-minded doctors not mentioned by Smith \[[@pmed-0020221-b1]\] might be to keep away from major high-impact journals and subscribe instead to those with a lower profile but that serve their specialty. I analysed all original papers published in the last 12 issues of Archives of Disease in Childhood. Of 198 such papers, there were seven (3.5%) manufacturer-funded studies dealing with drugs, vaccines, or infant foods. Another ten papers (5%) dealt with drugs or vaccines, including three reports of adverse events, but were not funded by industry. The funding of one was obscure. This pristine record was somewhat spoiled by a sponsored supplement, but clearly labelled as such, about a particular medication. It provoked an angry correspondence on the subscribers\' message board of one of the co-publishers. It seems that at least paediatrics, a far-away specialty of which Smith may know little, treads a careful path. Citation: Marcovitch H (2005) Little fish are less likely to take the bait. PLoS Med 2(7): e221. [^1]: Competing Interests: HM was Editor of Archives of Disease in Childhood for nine years, and is now Associate Editor for the BMJ. He has no pharmaceutical company sponsorship.	PubMed Central	2024-06-05T03:55:59.880687	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181885/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e221", "authors": [ { "first": "Harvey", "last": "Marcovitch" } ] }
PMC1181886	Richard Smith has correctly highlighted the potential distortion of the evidence base caused by the publication of commercially sponsored trials \[[@pmed-0020222-b1]\]. However, his proposed solution could do with more thought. First, let us be clear that the problem is possibly much wider than drug-company trials. There is the risk of systematic bias in reports of any research funded by a body that has an interest in the results. This "sponsor-induced bias" has been well documented in the area of tobacco-company-funded research on the effects of direct and indirect smoking \[[@pmed-0020222-b2]\]. In addition, governments, charities with an interest in a disease, and other bodies may also help to ensure that the results of research they sponsor (including trials) or the reporting of research favour one particular outcome. Lastly, individuals who carry out research, even if not funded by an interest group, may also bring prejudices to the table that influence the results and the published report. In other words, the tendency to bias is omnipresent. The issue of commercially funded trials is simply one of the degree and influence that trials have on clinical practice and health-care spending. Following Smith\'s thinking to its logical conclusion, we would not publish any research but simply critiques. Smith proposes that instead of publishing trials, journals should concentrate on critically describing them. If he is not confident that the current system of peer review is sufficiently robust to identify weaknesses, why should he be any more confident in the critiquing process (which is a form of peer review)? Journal peer review is often ad hoc (especially when my work is rejected) and is in desperate need of professionalizing, but I suspect along with Smith that this is not sufficient protection. Surely the way to deal with the systematic risk of bias is a reform not in the publication but in the production of evidence, which in turn reflects the way it is funded, conducted, analysed, and reported. My alternative solution in the case of trials is as follows. Companies (or indeed any body with a particular interest) should not be allowed to directly fund a clinical trial and no journal should publish a company-sponsored trial. Instead industry should pay a public or independent trials body, staffed by the best methodologists around and possibly established on an international scale. This international infrastructure should be publicly funded so that its staff do not feel dependent on industry business for security. The body, in conjunction with clinical experts from around the world, should conduct the study, ensuring that the questions are in the public\'s interest and fair (consumers would have an important role to play here). This infrastructure would ensure that the research was of the highest standard and reported accurately. Once the funding had been agreed on, there would be a compulsion to register the trial and to publish no matter what the results. This body would also have the ability to carry out or commission economic modelling (which is even more susceptible than trials to sponsor-induced bias) \[[@pmed-0020222-b3]\]. The resulting data would be held in a publicly accessible data archive. We should have an international agreement that no phase III trials would be permitted other than through this route. While still a rather bureaucratic response, it would ensure that the evidence base was less contaminated. Drug companies might even find that such a social solution results in trials being cheaper and easier to run. Citation: Sheldon T (2005) Focus on the funding and production of evidence rather than its publication. PLoS Med 2(7): e222. [^1]: Competing Interests: The author has declared that no competing interests exist.	PubMed Central	2024-06-05T03:55:59.881061	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181886/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e222", "authors": [ { "first": "Trevor", "last": "Sheldon" } ] }
PMC1181887	We congratulate and acknowledge PLoS Medicine for initiating a students\' forum \[[@pmed-0020223-b1]\] and for its efforts to encourage students\' participation across the globe. There is really a need for medical students, especially from the developing world, to be actively conducting research, reading journals, publishing papers, and staying in touch with current developments in the field of medicine. Many developing countries lack a national-level student medical journal for students to exchange their views and ideas, which thereby pacifies their thinking and makes them hypnotic to issues such as the influence of drug companies and the neglected health problems of poorer countries. It will be really motivating for students from developing countries to actively take part in debate through the Student Forum of PLoS Medicine, which is composed of articles selected by student representatives across the world. The unique integration of student associations with PLoS Medicine, and the journal\'s policy of not publishing advertisements for drugs or medical devices, will also enlighten students about the influence of drug companies in medical practice and enable students to realize their priorities in poorer countries for the future. Thereby, students may focus their attention on becoming professionals in developing new strategies to combat killer infectious diseases like malaria and tuberculosis, and malnutrition---such as vitamin deficiencies among children and iron deficiency among pregnant women---that are dreaded and very common in poorer countries. Citation: Kumar CJ, Deoker A (2005) Applause to PLoS Medicine for initiating students\' forum. PLoS Med 2(7): e223. [^1]: Competing Interests: The authors have declared that no competing interests exist.	PubMed Central	2024-06-05T03:55:59.881567	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181887/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e223", "authors": [ { "first": "C. Jairaj", "last": "Kumar" }, { "first": "Abhizith", "last": "Deoker" } ] }
PMC1181888	In response to Michael Makover\'s comments \[[@pmed-0020224-b1]\], the important point at issue here is evidence---and not whether there is room for both population approaches and high-risk approaches. We can certainly identify plaques using carotid ultrasound. We can use risk factor scoring schemes to identify those at high risk of suffering a cardiovascular event. We can give patients a range of drugs that have been shown in trials to reduce risk of these events. However, the relevant evidence from randomised controlled trials of risk factor screening (using either scores or carotid ultrasound) and intervention is simply not available. So what do we do---ignore the lack of evidence? Or do we get on with organising the trials? Citation: Ebrahim S (2005) Author\'s reply. PLoS Med 2(7): e224. [^1]: Competing Interests: The author has declared that no competing interests exist.	PubMed Central	2024-06-05T03:55:59.881893	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181888/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e224", "authors": [ { "first": "Shah", "last": "Ebrahim" } ] }
PMC1181889	Steven Senn \[[@pmed-0020225-b1]\] raises several very important points about the publication of trials. In an attempt to improve transparency of reporting, we require authors to submit their protocol along with the trial so that it is available for reviewers and editors to compare with the journal article, and we encourage the protocol to be published with the trial so that readers can check these results for themselves. One problem in trial reporting is the relatively unstructured nature of trial reports in medical journals compared with, for example, Trial Bank (<http://rctbank.ucsf.edu/>). We are currently considering how we report trials at PLoS; a much more structured and, hence, more transparent report may make it much harder to hide results (or the lack of them). Citation: Barbour V, Cohen B, Yamey G (2005) Editors\' reply. PLoS Med 2(7): e225. [^1]: Competing Interests: The authors are editors for PLoS Medicine.	PubMed Central	2024-06-05T03:55:59.882160	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181889/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e225", "authors": [ { "first": "Virginia", "last": "Barbour" }, { "first": "Barbara", "last": "Cohen" }, { "first": "Gavin", "last": "Yamey" } ] }
PMC1181890	Richard Smith\'s key suggestion \[[@pmed-0020226-b1]\] is that medical journals "should stop publishing trials" and concentrate on "critically evaluating them." This bold and radical suggestion deserves wide debate. It\'s obvious that many medical journals are losing relevance as vehicles for scientific information, but it\'s unclear what will save them. Even as journals strive to better enforce their conflicts-of-interest disclosure rules, drug companies will strive to find or create other publication outlets that can communicate to physicians precisely what advertisers wish to communicate. In sum, an unanticipated effect of purging clinical trial reports from medical journals might be an even larger proliferation of frank advertising outlets and messages that might more effectively catch doctors\' attentions. Citation: Cohen D (2005) Bold suggestion by Smith. PLoS Med 2(7): e226. [^1]: Competing Interests: DC is a former editor of Ethical Human Sciences and Services, which published several articles critical of the drug industry, and has authored articles critical of drug industry sponsorship and influence on clinical psychopharmacology trials.	PubMed Central	2024-06-05T03:55:59.882419	2005-7-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181890/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e226", "authors": [ { "first": "David", "last": "Cohen" } ] }
PMC1181891	Twins\' research is a favorite tool of the human geneticist, but it has a controversial history. Nazi doctor Josef Mengele, infamous for his work at Auschwitz, was fascinated by twins. He sought them out at the extermination camp and used them in violent experiments. Later, British psychologist Cyril Burt worked on the heredity of intelligence, producing findings that some suspected were "too good"---and which later were shown to be based on fraudulent data involving invented twins. In this month\'s PLoS Medicine, Nancy Krieger and colleagues examine the health of female twins from a very different perspective. To understand the impact of lifetime socioeconomic position on health, they studied twins who were raised together but who had different socioeconomic position after adolescence. Many studies have compared the health of twins raised separately since birth or early childhood, but few have investigated the adult health of twins raised together but who had different post-adolescent socioeconomic position. Such a study could clarify the impact of adult experiences on adult health in a population matched on early life experiences, the authors say. Krieger\'s team employed data from a cohort of 434 twins who lived in the San Francisco Bay area in 1978--1979; the average age at recruitment was 41 years old. The cohort was given a health and sociodemographic questionnaire. Data on anthropometric and biological characteristics were obtained by physical examination and laboratory analyses; the process was repeated at a second examination around 1989--1990. At the second phase of the study, 72 women (8.3%) did not return, of whom 36 had died, which left 352 twin pairs (58% monozygotic, 42% dizygotic), representing 81.1% of the original cohort. The sociodemographic and health characteristics of the full cohort (352 pairs) and the cohort analyzed (those pairs where it was known both that they had lived together until at least age 14 and their joint socioeconomic was available---i.e., 308 pairs) were quite similar: 40% grew up in working-class households and 80% in households where the father had fewer than four years of college education. At the second examination, 32% of the twin pairs in the analytic cohort had a difference in their adult household occupational class, and 20% had different levels of college education. The team found that health outcomes among monozygotic adult female twins who lived together through childhood varied by their subsequent socioeconomic position. Twins with differing occupational class differed in health status compared with twins with similar occupational class. The working-class twin had significantly higher systolic blood pressure, diastolic blood pressure, and LDL cholesterol than her professional, non-working-class twin. Twins discordant on educational level, however, had similar health status, likely reflecting the fact that current occupational class is a better measure for investigating the impact of lifetime socioeconomic position than is education (which is completed usually in early adulthood). Dizgyotic twins with differing adult socioeconomic position, either occupational class or educational level, did not notably differ in their adult health status. These novel findings lend support to the hypothesis that health is shaped not only by early life experiences but also by cumulative experiences across the lifecourse, the authors say. As with many twin studies, the numbers of individuals studied is relatively limited. The lack of some biological and personal data such as detailed occupational class position over time, lack of data on income, poverty, wealth, debt, gestational age, birthweight, and birth order also limits the conclusions that can be drawn. These limitations are countered, however, by the tight matching of the twins on early life and childhood exposures, as well as by the matching for genetic endowment among the monozygotic twins. Together, the findings have important implications for health policy, since they suggest that adult socioeconomic position does have an impact on adult health above and beyond early life exposures, and they also add to our understanding of how societal conditions shape population patterns of health, disease, and well-being.[](#pmed-0020235-g001){ref-type="fig"} ::: {#pmed-0020235-g001 .fig} ::: {.caption} ###### Twins in the womb (from the first US textbook on midwifery, An Abridgement of the Practice of Midwifery, by W. Smellie, 1786) ::: ![](pmed.0020235.g001) :::	PubMed Central	2024-06-05T03:55:59.882817	2005-7-26	{ "license": "Creative Commons Zero - Public Domain - https://creativecommons.org/publicdomain/zero/1.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181891/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e235", "authors": [] }
PMC1181892	There are at least 300 million acute cases of malaria each year globally, resulting in more than a million deaths. Ninety percent of deaths due to malaria occur in Africa, south of the Sahara, mostly in young children. The number of deaths is increasing, and one key factor linked to this has been widespread drug resistance of Plasmodium falciparum to conventional antimalarials, such as sulfadoxine-pyrimethamine (SP); such resistance is widespread in southeast Asia, South America, and Africa. The inappropriate use of antimalarials during the past century has contributed to this increase in resistance. For example, there has been overreliance on quinolines (such as chloroquine) and antifolates (such as pyrimethamine) resulting in cross-resistance among these drug classes. However, in the past decade, a new group of antimalarials---the artemisinin compounds, such as artesunate, artemether, and dihydroartemisinin---have been deployed on an increasingly large scale. These compounds produce a very rapid therapeutic response, are active against parasites resistant to multiple drugs, are well tolerated, and reduce gametocyte carriage. To date, no parasite resistance to these compounds has been detected.[](#pmed-0020236-g001){ref-type="fig"} ::: {#pmed-0020236-g001 .fig} ::: {.caption} ###### A trial site in Uganda ::: ![](pmed.0020236.g001) ::: If used alone, the artemisinins will cure falciparum malaria in seven days, but studies in southeast Asia have shown that combinations of artemisinin compounds with certain synthetic drugs produce high cure rates after just three days of treatment. There is also some evidence that combinations of therapies could greatly retard development of resistance to the partner drug. Although combinations including artemisinins have been widely advocated, they are expensive and relatively untested in highly endemic areas. In this month\'s PLoS Medicine, Adoke Yeka and colleagues compared artemisinin-based compounds and other combination therapies in four districts with varying transmission intensity in Uganda in 2,160 patients aged six months or greater with uncomplicated falciparum malaria. The team tested the combination of chloroquine and SP, currently the first-line therapy in Uganda, the combination of amodiaquine and SP, a cheap regimen proven to be efficacious in previous trials, and the combination of amodiaquine and artesunate. During the 28-day study they collected data on the efficacy of the different regimens and examined the effect on recrudescence and new infections after therapy. Combined amodiaquine and artesunate was the most efficacious regimen for preventing recrudescence, but this benefit was outweighed by an increased risk of new infection. This result was probably due to artesunate being rapidly eliminated, leaving only amodiaquine to provide post-treatment prophylaxis. Considering all recurrent infections, the combination of amodiaquine and SP was at least as efficacious as the other combinations at all sites and superior at the highest transmission sites. In all, 72% of all recurrent infections were due to new infections, and with the two most efficacious regimens (amodiaquine and SP, and amodiaquine and artesunate) this proportion was 80%. The identification of new infections stressed the need for other malaria control measures, such as bed nets, said the authors. They also suggested that antimalarials should be judged not just on their impact on recrudescence but also on their impact on the risk of new infections after therapy. Previous studies have suggested that patients who suffer recrudescence have a higher risk of complicated malaria and death. Artemisinins are highly attractive antimalarials, but when used as monotherapy, they have a high risk of recrudescence and hence must be combined with other antimalarials to achieve maximum efficacy. But whether the partner drug should be long or short acting remains unclear, said the authors. Altogether, artemisinin combinations offer great hope for Africa, the authors say, although the ideal combination regimen remains uncertain and cost is a problem. To compare the efficacy of the different therapies, bigger and longer controlled trials are needed in conditions of varied transmission intensity. Nevertheless, based on the results of this study and others, Uganda has chosen a combination of artemether and lumefantrine as its first-line therapy against malaria.	PubMed Central	2024-06-05T03:55:59.883459	2005-7-26	{ "license": "Creative Commons Zero - Public Domain - https://creativecommons.org/publicdomain/zero/1.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181892/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e236", "authors": [] }
PMC1181893	Hepatitis B is a serious global public health problem but is preventable with safe and effective vaccines that have been available since 1982. Despite these vaccines, about 2 billion people have been infected with hepatitis B virus (HBV), and more than 350 million have lifelong infections. These chronically infected people are at high risk of death from cirrhosis of the liver and liver cancer, which both kill about 1 million people each year. Suppression of viral replication in chronic carriers of HBV is an effective approach to controlling disease progression. Current antiviral therapies include lamivudine and alpha-interferon, but long-term resolution of the disease is disappointing because of low seroconversion rates and the development of drug-resistant viral mutants. In this month\'s PLoS Medicine, Lisa F. P. Ng and colleagues describe the identification of a host factor that has a significant effect on viral replication efficiency. The team began by examining the serum viral load of a group of carriers of hepatitis B in relation to the HBV genome carried. They found a significant association between high serum viral load and a natural sequence variant within the HBV enhancer II regulatory region at position 1752. Upon testing all four possible 1752 variants, the 1752A variant had the highest transcriptional activity. Further investigation of this enhanced transcriptional activity revealed evidence of possible interaction with host DNA binding proteins. The team found that a protein present in the human host---hnRNPK---could be isolated by direct binding to a viral fragment derived from the HBV variant of these infected patients. hnRNPK has previously been shown to be involved in several cellular functions---for example, as a regulator of signal transduction and of gene expression. On further examination of the role of hnRNPK in HBV replication, they established that hnRNPK is capable of acting on the full length of HBV, rather than just a partial fragment. They compared four full-length replicative HBV clones, identical except for a single base change at position 1752, that were transfected with two different hnRNPK expression constructs and showed that 1752A was more efficient at promoting replication than the other three variants. To further show the role of hnRNPK in HBV replication, the team tested the effect of over-expression and down-regulation of the cellular protein. Using siRNA, designed to reduce endogenous hnRNPK, they showed suppression of both hnRNPK mRNA and HBV viral load, whereas a control siRNA had no effect on HBV viral load. Despite these findings, the mechanism behind hnRNPK on HBV replication needs further exploration, the authors say, concluding that viral replication efficiency was determined by a combination of viral sequence and interaction with specific host proteins. However, they suggest that these results indicate that although drug development of antivirals is an established research avenue, targeting the host is an untapped opportunity. They describe parallels with anti-EGFR antibody treatment of breast cancer cells, which produced a decrease in cell replication rate and corresponding reduction in hnRNPK expression levels; this result suggested that hnRNPK levels could be modulated by anti-EGFR treatment, thus highlighting new treatment options for altering the HBV viral load in chronic carriers. The authors conclude that the future of long-term viral clearance will require combination therapy of targeting the virus directly, blocking host support proteins, and using immuno-modulating agents.	PubMed Central	2024-06-05T03:55:59.884012	2005-7-26	{ "license": "Creative Commons Zero - Public Domain - https://creativecommons.org/publicdomain/zero/1.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181893/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e237", "authors": [] }
PMC1181894	Postherpetic neuralgia (PHN) is a chronically painful condition that is a complication of shingles (acute herpes zoster), a recurrence of the varicella-zoster virus, which initially causes chickenpox. Although shingles usually resolves within a month, some people continue to feel the pain of PHN long after the rash and blisters heal, because of nerve damage (neuropathic pain) caused by the shingles. Not everyone who has had shingles develops PHN, although it is a common complication of shingles in older adults. Despite advances in antiviral therapy during acute herpes zoster and the more recent introduction of vaccination against varicella zoster, PHN continues to be a significant clinical problem, with 10--20% of patients developing persistent neuropathic pain after acute herpes zoster reactivation. The nature of PHN pain is variable, which implies that a variety of mechanisms might be operating. This variability has led to the hypothesis that treatment plans could be optimised for individual patients on the basis of the individual pattern of their symptoms or the underlying mechanism of the pain. However, the current evidence base for therapies in PHN is based on clinical trials of analgesics, which have examined PHN as a single disease entity. Furthermore, there is little evidence for the efficacy of drugs for specific sets of symptoms and no simple way to determine which pain mechanisms might be operating in an individual patient. In this month\'s PLoS Medicine, Andrew Rice and colleagues reassessed the evidence base by doing a systematic review and meta-analysis of analgesic therapy for PHN, which has fundamentally changed in the wake of several major new trials. The authors searched the literature for trials of PHN and retrieved 62 articles, of which 35 were kept for final analysis. Their analysis confirmed several previous research findings, although they cautioned that the meta-analytic study design of collecting data from a range of trials had several inherent pitfalls, and it is difficult to directly compare treatments across different trials. However, they found evidence for analgesic efficacy in established PHN for orally administered therapies, such as tricyclic antidepressants, some opioids, gabapentin, tramadol, and pregabalin. Some topically administered therapies, such as lidocaine and capsaicin, were associated with analgesic efficacy in selected patients. However, it appeared that therapies such as oral administration of certain NMDA receptor antagonists, codeine, ibuprofen, lorazepam, 5HT1 receptor agonists, and acyclovir were not efficacious in PHN. Altogether, the authors conclude that the evidence base supports the first-line use of a tricyclic antidepressant for orally administered treatment of PHN, reserving the gabapentinoids for second-line use. Topical treatments, such as lidocaine or capsaicin, should be considered as first-line treatment if a patient falls into the "sensitised nociceptor" as opposed to "deafferentation" sub-group of PHN patients. The role of intrathecal steroid is still not clear: one trial indicated that intrathecal steroids were associated with benefits in patients with PHN, but this therapy might be hazardous, and the authors and other researchers have concluded that further high-quality trials of this therapy are needed. The authors found little evidence regarding possible synergistic effects of the various treatments to support or refute the concomitant use of combinations of drugs. They stressed that any treatment plan must recognise the importance of the biopsychosocial model of chronic pain and that any pharmacologically based management of PHN should be combined with advice on and management of psychological and social aspects. Finally, as there is no single pathophysiology that underlies PHN, they propose that future studies should use quantitative sensory evaluation to clearly categorise subsets of participants for better interpretation of treatment effects.	PubMed Central	2024-06-05T03:55:59.884759	2005-7-26	{ "license": "Creative Commons Zero - Public Domain - https://creativecommons.org/publicdomain/zero/1.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181894/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e238", "authors": [] }
PMC1181895	When the SARS epidemic showed the first signs of waning, the World Health Organization proclaimed that the turnaround was a testament to the efficient response of health systems worldwide and justified its decisive action in issuing a global alert. That swift response was partly due to infectious disease experts being able to use models of disease spread, even though SARS was a newly emerging disease, to help plan their next move. In fact, epidemiologists have used mathematical models to predict and understand the dynamics of infectious diseases for more than 200 years. The emergence of diseases such as Ebola, SARS, and West Nile virus, and multi-drug-resistant malaria---as well as the potential for diseases to be introduced by bioterrorism---has attached even greater importance to this management tool. Models are used to provide information on such infections and predict the effect of alternative courses of action. In this month\'s PLoS Medicine, Helen Wearing and colleagues suggest, however, that many off-the-shelf models are inappropriate for making quantitative predictions because substantial biases have been introduced by two important, yet largely ignored, assumptions. The authors warn that if such biases are not corrected, health authorities risk making overly optimistic health policy decisions. They begin with the "SEIR" class of models, in which the host population is classified according to infectious status, i.e., individuals are susceptible, exposed, infectious, or recovered. This model assumes that the rate of leaving the exposed or infectious class is constant, irrespective of the time already spent in that class. Although mathematically convenient, this assumption gives rise to exponentially distributed latent (incubation) and infectious periods, which is epidemiologically unrealistic for most infections, say the authors, who suggest instead that it would more sensible to specify the probability of leaving a class as a function of the time spent within the class. Hence, initially, the chance of leaving the class is small but then increases as the mean infectious/incubation period is reached. This assumption would give a more realistic distribution of incubation and infectious periods.[](#pmed-0020239-g001){ref-type="fig"} ::: {#pmed-0020239-g001 .fig} ::: {.caption} ###### Estimates of variation in the basic reproductive ratio ::: ![](pmed.0020239.g001) ::: The authors also note another issue that has received surprisingly little attention in infectious disease models, namely, the influence of incubation and infectious period distributions on the invasion dynamics of an infection into a largely susceptible population---despite its obvious application to emerging infections and possible "deliberate exposure." The impact of these differences on models could translate into potentially important public health concerns, say the authors. They tested their theory by using analytical methods to show that, first, ignoring the incubation period or, second, assuming exponentially distributed incubation and infectious periods (when including the incubation period) always resulted in underestimating the basic reproductive ratio of an infection from outbreak data. They then illustrated these points by fitting epidemic models to data from an influenza outbreak. Their results suggested that within a strict management setting, epidemiological details could make a crucial difference. Although previous studies have shown the importance of using realistic distributions of incubation and infectious periods in endemic disease models, few studies have considered the effects associated with making predictions for an emerging disease. Discrepancies between estimates of reproductive ratio from exponentially distributed and gamma-distributed fits confirm the need to have precise distributions of incubation and infectious periods. Although such data are available from post hoc analyses of epidemics, they are lacking for novel emerging infections. The key point is that uncertainty about these distributions should be incorporated into models when making quantitative predictions. The take home message is that when developing models for public health use, policy makers need to pay attention to the intrinsic assumptions within classical models. The authors note that while some practitioners are using their approach, most applied epidemiological studies still use models that incorporate exponentially distributed incubation and infectious periods; the authors hope their work will point to the next steps in delivering quantitatively accurate epidemiological models.	PubMed Central	2024-06-05T03:55:59.885351	2005-7-26	{ "license": "Creative Commons Zero - Public Domain - https://creativecommons.org/publicdomain/zero/1.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1181895/", "journal": "PLoS Med. 2005 Jul 26; 2(7):e239", "authors": [] }
PMC1182131	The legal risks associated with health research involving human subjects have been highlighted recently by a number of lawsuits launched against those involved in conducting and evaluating the research \[[@B1]\]. Some of these cases have been fully addressed by the legal system, resulting in judgments that provide some guidance. The vast majority of cases have either settled before going to trial, or have not yet been addressed by the courts, leaving us to wonder what might have been and what guidance future cases may bring. What is striking about the lawsuits that have been commenced is the broad range of individuals/institutions that are named as defendants and the broad range of allegations that are made. Plaintiffs cast a wide net: the range of defendants named ========================================================= A review of recent Canadian and American cases demonstrates that in commencing lawsuits over alleged research misconduct, plaintiffs cast a wide net, naming as defendants anyone who had anything to do with the research in question. Named defendants have included the researchers, the research ethics committee/board/institutional review board (REC) that approved the research and its individual members, as well as bioethicists who consulted on the research project. For example, in Gelsingerv. Trustees of the University of Pennsylvania\[[@B2]\], an eighteen year-old who had volunteered to participate in a corrective gene study died during the course of the study. In that case, the trustees of the university and two hospitals affiliated with the research, the investigators, the company that sponsored the research, the former medical school dean and a bioethicist, were all originally named as defendants on the bases (among others) of wrongful death, assault and battery linked to a lack of informed consent, and common law fraud/misrepresentation linked to the informed consent process. The case settled for an undisclosed amount \[[@B3]\]. In Robertsonv. McGee\[[@B4]\], the REC had approved a protocol for a Phase I study of a cancer vaccine. Many of the patients who enrolled in the study had advanced disease, were unresponsive to standard therapies and had very poor prognoses. According to news reports, 94 subjects received the vaccine and 26 died during the study, although the deaths were not attributed to the vaccine itself \[[@B5]\]. On January 29, 2001, a number of subjects and subject representatives filed a lawsuit seeking actual and punitive damages. The Robertsonplaintiffs sued the hospital, the principal investigator, the pharmaceutical sponsor, a top university official, the individual members of the REC, and the university bioethicist who consulted with the REC. The issues in this case were never decided by the court because the court held that it did not have jurisdiction over the allegations made in the complaint and dismissed the case. In the case of Weissv. Solomon\[[@B6]\], a research subject suffered a cardiac arrest and died after undergoing a fluorescein angiogram as part of a research study. The deceased\'s family sued the principal investigator, the hospital and a physician who referred one of his patients to the study. The Court hearing the case ultimately found liability against the primary investigator and hospital only. Interestingly, the liability against the hospital was based in part upon the fact that the hospital\'s REC had approved the research protocol and the consent form which was determined by the Court to be deficient. All of these cases demonstrate that plaintiffs cast a wide net when deciding whom to name as defendants in cases where research goes wrong. It is possible for almost anyone involved in a research project, no matter how remotely connected, to be named in legal proceedings if something goes wrong. Unfortunately, there are not enough decided cases to predict with any degree of certainty how far the courts in Canada and the United States (U.S.) will be prepared to go in attributing fault in research negligence cases to those that have less than a direct connection to the injuries that have been sustained. Broad range of allegations made =============================== A review of the lawsuits that have been commenced recently demonstrates that there are a broad range of allegations that can potentially be made against those involved in research. The allegations are such that it is open to a plaintiff to sue not only the primary investigator and others directly involved in the research but also those with a much less tangible relationship to the research subject. What follows is a brief overview of some of the more common allegations that have been made and what, if anything, the courts have said about these allegations. Deficiencies in the consent process ----------------------------------- It is perhaps obvious for most health care professionals that liability will arise where research is conducted without informed consent from the subjects (or their substitute decision makers) being obtained. What may be less known is that it appears from existing case law as though the standard for obtaining informed consent in the research context is higher than in the therapeutic context. The other interesting development has been that it appears that courts may be willing to hold a hospital responsible where the informed consent form approved by the REC of the Hospital is determined to be less than adequate. In the Weisscase for example, the Quebec Superior Court found that the duty to inform in matters relating to purely scientific experimentation is the most exacting possible and includes the disclosure of all known risks including those which are rare or remote, especially if they may entail grave consequences \[[@B7]\]. This would suggest a standard that is higher than the standard for consent to treatment which requires that only material risks be disclosed. In addition, the court in Weiss found that the hospital was liable and attributed some of that liability to the fact that the hospital\'s REC failed to ensure that the consent form used for the research was appropriate. Failure to follow laws, regulations, policies, procedures and guidelines ------------------------------------------------------------------------ Some of the recent lawsuits have alleged that the defendants failed to follow applicable laws, regulations, policies, procedures and/or guidelines. If these allegations can be proven, it is likely that liability will follow, as a failure to comply with applicable laws, regulations, policies, procedures and/or guidelines will likely be interpreted by the courts as a clear sign that the defendants failed to meet the standard of care. Government initiated legal proceedings -------------------------------------- In addition to being sued by research subjects who allege being injured by their participation in a study, there is also a possibility that the researchers and institutions involved in health research involving human subjects will face legal battles with governmental bodies/agencies with jurisdiction over research. In the U.S., such bodies have imposed drastic sanctions where there is evidence of research misconduct. The defendants in the Gelsingercase not only settled with the plaintiffs, they also settled with the U.S. government in a separate civil action commenced by the government on the basis of breach of the False Claims Act\[[@B8]\]. The settlement resulted in a total of over one million dollars in payments to the government by the institutions involved in the research as well as restrictive controls being placed on three investigators involved in the research with respect to their future clinical research activities \[[@B9]\]. Conflict of Interest ==================== In the case of Gelsinger, the complaint alleged, among other things, that the university was to receive an ownership stake in Genovo (the sponsor) in lieu of funding of the gene transfer research program and that the University and various physicians associated with the research program had substantial financial and equity interests with respect to the vectors employed in the research \[[@B10]\]. The extent of these financial interests were not disclosed to Jesse Gelsinger before he made his decision to participate in the research as a subject \[[@B11]\]. If this case had not settled and instead proceeded to a trial, one of the key issues would have been the circumstances under which an undisclosed conflict of interest will result in liability for the individual(s) and/or entities who are in the conflict position and whether other parties who know of the conflict or who ought reasonably to know about the conflict and are in a position of authority have a legal obligation to intervene to prevent the research from proceeding on the basis of the conflict. The existence of an undisclosed conflict of interest may also put a plaintiff in the position of being able to argue that the consent given was not truly informed. Lawsuits based on principles of international human rights ========================================================== In the case of Abdullahiv. Pfizer, Inc. \[[@B12]\], Pfizer was the defendant in a lawsuit that alleged that it improperly administered an experimental antibiotic to children in Nigeria during an outbreak of bacterial meningitis, measles and cholera in Kano, Nigeria. The guardians of certain of those children instituted an action in the Southern District of New York, alleging violations of the Nuremberg Code, the Declaration of Helsinki, Article 7 of the International Covenant on Civil and Political Rights, U.S. Food and Drug Administration\'s regulations and other norms of international law. They also asserted that the court had jurisdiction over the matter under the Alien Torts Claim Act. This case was sent back to the District Court to determine whether the U.S. or Nigeria is the appropriate forum to hear the case \[[@B13]\]. In the Robertsoncase, several of the causes of action were derived from international human rights law rather than standard medical malpractice or tort law, including allegations of breaching the right to be treated with dignity, citing the Nuremberg Code and Declaration of Helsinki concerning biomedical research \[[@B14]\]. Conclusion ========== Admittedly, the number of judgments rendered in research misconduct lawsuits in North America to date is small. That being said, experience to date demonstrates that: \(a) plaintiffs will likely cast a wide net, naming anyone or any institution that had even the slightest involvement or connection to the research; \(b) plaintiffs have found a number of different ways to frame their claims; \(c) the Government may be a potential plaintiff in these lawsuits; and \(d) the standards applicable in the research context may, in fact, be higher than those applicable in the therapeutic treatment context. The research community should take this early experience as a warning and should reflect carefully on practices where research involving human subjects is concerned. Competing interests =================== The author(s) declare that they have no competing interests. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1472-6939/6/4/prepub>	PubMed Central	2024-06-05T03:55:59.886007	2005-6-13	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182131/", "journal": "BMC Med Ethics. 2005 Jun 13; 6:4", "authors": [ { "first": "Randi", "last": "Zlotnik Shaul" }, { "first": "Shelley", "last": "Birenbaum" }, { "first": "Megan", "last": "Evans" } ] }
PMC1182327	Published research findings are sometimes refuted by subsequent evidence, with ensuing confusion and disappointment. Refutation and controversy is seen across the range of research designs, from clinical trials and traditional epidemiological studies \[[@pmed-0020124-b1]\] to the most modern molecular research \[[@pmed-0020124-b4],[@pmed-0020124-b5]\]. There is increasing concern that in modern research, false findings may be the majority or even the vast majority of published research claims \[[@pmed-0020124-b6]\]. However, this should not be surprising. It can be proven that most claimed research findings are false. Here I will examine the key factors that influence this problem and some corollaries thereof. Modeling the Framework for False Positive Findings {#s2} ================================================== Several methodologists have pointed out \[[@pmed-0020124-b9]\] that the high rate of nonreplication (lack of confirmation) of research discoveries is a consequence of the convenient, yet ill-founded strategy of claiming conclusive research findings solely on the basis of a single study assessed by formal statistical significance, typically for a p-value less than 0.05. Research is not most appropriately represented and summarized by p-values, but, unfortunately, there is a widespread notion that medical research articles should be interpreted based only on p-values. Research findings are defined here as any relationship reaching formal statistical significance, e.g., effective interventions, informative predictors, risk factors, or associations. "Negative" research is also very useful. "Negative" is actually a misnomer, and the misinterpretation is widespread. However, here we will target relationships that investigators claim exist, rather than null findings. "It can be proven that most claimed research findings are false" As has been shown previously, the probability that a research finding is indeed true depends on the prior probability of it being true (before doing the study), the statistical power of the study, and the level of statistical significance \[[@pmed-0020124-b10],[@pmed-0020124-b11]\]. Consider a 2 × 2 table in which research findings are compared against the gold standard of true relationships in a scientific field. In a research field both true and false hypotheses can be made about the presence of relationships. Let R be the ratio of the number of "true relationships" to "no relationships" among those tested in the field. R is characteristic of the field and can vary a lot depending on whether the field targets highly likely relationships or searches for only one or a few true relationships among thousands and millions of hypotheses that may be postulated. Let us also consider, for computational simplicity, circumscribed fields where either there is only one true relationship (among many that can be hypothesized) or the power is similar to find any of the several existing true relationships. The pre-study probability of a relationship being true is R/(R + 1). The probability of a study finding a true relationship reflects the power 1 - β (one minus the Type II error rate). The probability of claiming a relationship when none truly exists reflects the Type I error rate, α. Assuming that c relationships are being probed in the field, the expected values of the 2 × 2 table are given in [Table 1](#pmed-0020124-t001){ref-type="table"}. After a research finding has been claimed based on achieving formal statistical significance, the post-study probability that it is true is the positive predictive value, PPV. The PPV is also the complementary probability of what Wacholder et al. have called the false positive report probability \[[@pmed-0020124-b10]\]. According to the 2 × 2 table, one gets PPV = (1 - β)R/(R - βR + α). A research finding is thus more likely true than false if (1 - β)R \> α. Since usually the vast majority of investigators depend on a = 0.05, this means that a research finding is more likely true than false if (1 - β)R \> 0.05. ::: {#pmed-0020124-t001 .table-wrap} Table 1 ::: {.caption} ###### Research Findings and True Relationships ::: ![](pmed.0020124.t001) ::: What is less well appreciated is that bias and the extent of repeated independent testing by different teams of investigators around the globe may further distort this picture and may lead to even smaller probabilities of the research findings being indeed true. We will try to model these two factors in the context of similar 2 × 2 tables. Bias {#s3} ==== First, let us define bias as the combination of various design, data, analysis, and presentation factors that tend to produce research findings when they should not be produced. Let u be the proportion of probed analyses that would not have been "research findings," but nevertheless end up presented and reported as such, because of bias. Bias should not be confused with chance variability that causes some findings to be false by chance even though the study design, data, analysis, and presentation are perfect. Bias can entail manipulation in the analysis or reporting of findings. Selective or distorted reporting is a typical form of such bias. We may assume that u does not depend on whether a true relationship exists or not. This is not an unreasonable assumption, since typically it is impossible to know which relationships are indeed true. In the presence of bias ([Table 2](#pmed-0020124-t002){ref-type="table"}), one gets PPV = (\[1 - β\]R + uβR)/(R + α − βR + u − uα + uβR), and PPV decreases with increasing u, unless 1 − β ≤ α, i.e., 1 − β ≤ 0.05 for most situations. Thus, with increasing bias, the chances that a research finding is true diminish considerably. This is shown for different levels of power and for different pre-study odds in [Figure 1](#pmed-0020124-g001){ref-type="fig"}. Conversely, true research findings may occasionally be annulled because of reverse bias. For example, with large measurement errors relationships are lost in noise \[[@pmed-0020124-b12]\], or investigators use data inefficiently or fail to notice statistically significant relationships, or there may be conflicts of interest that tend to "bury" significant findings \[[@pmed-0020124-b13]\]. There is no good large-scale empirical evidence on how frequently such reverse bias may occur across diverse research fields. However, it is probably fair to say that reverse bias is not as common. Moreover measurement errors and inefficient use of data are probably becoming less frequent problems, since measurement error has decreased with technological advances in the molecular era and investigators are becoming increasingly sophisticated about their data. Regardless, reverse bias may be modeled in the same way as bias above. Also reverse bias should not be confused with chance variability that may lead to missing a true relationship because of chance. ::: {#pmed-0020124-g001 .fig} Figure 1 ::: {.caption} ###### PPV (Probability That a Research Finding Is True) as a Function of the Pre-Study Odds for Various Levels of Bias, u Panels correspond to power of 0.20, 0.50, and 0.80. ::: ![](pmed.0020124.g001) ::: ::: {#pmed-0020124-t002 .table-wrap} Table 2 ::: {.caption} ###### Research Findings and True Relationships in the Presence of Bias ::: ![](pmed.0020124.t002) ::: Testing by Several Independent Teams {#s4} ==================================== Several independent teams may be addressing the same sets of research questions. As research efforts are globalized, it is practically the rule that several research teams, often dozens of them, may probe the same or similar questions. Unfortunately, in some areas, the prevailing mentality until now has been to focus on isolated discoveries by single teams and interpret research experiments in isolation. An increasing number of questions have at least one study claiming a research finding, and this receives unilateral attention. The probability that at least one study, among several done on the same question, claims a statistically significant research finding is easy to estimate. For n independent studies of equal power, the 2 × 2 table is shown in [Table 3](#pmed-0020124-t003){ref-type="table"}: PPV = R(1 − β^n^)/(R + 1 − \[1 − α\]^n^ − Rβ^n^) (not considering bias). With increasing number of independent studies, PPV tends to decrease, unless 1 - β \< a, i.e., typically 1 − β \< 0.05. This is shown for different levels of power and for different pre-study odds in [Figure 2](#pmed-0020124-g002){ref-type="fig"}. For n studies of different power, the term β^n^ is replaced by the product of the terms β~i~ for i = 1 to n, but inferences are similar. ::: {#pmed-0020124-g002 .fig} Figure 2 ::: {.caption} ###### PPV (Probability That a Research Finding Is True) as a Function of the Pre-Study Odds for Various Numbers of Conducted Studies, n Panels correspond to power of 0.20, 0.50, and 0.80. ::: ![](pmed.0020124.g002) ::: ::: {#pmed-0020124-t003 .table-wrap} Table 3 ::: {.caption} ###### Research Findings and True Relationships in the Presence of Multiple Studies ::: ![](pmed.0020124.t003) ::: Corollaries {#s5} =========== A practical example is shown in [Box 1](#box1){ref-type="boxed-text"}. Based on the above considerations, one may deduce several interesting corollaries about the probability that a research finding is indeed true. Box 1. An Example: Science at Low Pre-Study Odds ------------------------------------------------ Let us assume that a team of investigators performs a whole genome association study to test whether any of 100,000 gene polymorphisms are associated with susceptibility to schizophrenia. Based on what we know about the extent of heritability of the disease, it is reasonable to expect that probably around ten gene polymorphisms among those tested would be truly associated with schizophrenia, with relatively similar odds ratios around 1.3 for the ten or so polymorphisms and with a fairly similar power to identify any of them. Then R = 10/100,000 = 10^−4^, and the pre-study probability for any polymorphism to be associated with schizophrenia is also R/(R + 1) = 10^−4^. Let us also suppose that the study has 60% power to find an association with an odds ratio of 1.3 at α = 0.05. Then it can be estimated that if a statistically significant association is found with the p-value barely crossing the 0.05 threshold, the post-study probability that this is true increases about 12-fold compared with the pre-study probability, but it is still only 12 × 10^−4^. Now let us suppose that the investigators manipulate their design, analyses, and reporting so as to make more relationships cross the p = 0.05 threshold even though this would not have been crossed with a perfectly adhered to design and analysis and with perfect comprehensive reporting of the results, strictly according to the original study plan. Such manipulation could be done, for example, with serendipitous inclusion or exclusion of certain patients or controls, post hoc subgroup analyses, investigation of genetic contrasts that were not originally specified, changes in the disease or control definitions, and various combinations of selective or distorted reporting of the results. Commercially available "data mining" packages actually are proud of their ability to yield statistically significant results through data dredging. In the presence of bias with u = 0.10, the post-study probability that a research finding is true is only 4.4 × 10^−4^. Furthermore, even in the absence of any bias, when ten independent research teams perform similar experiments around the world, if one of them finds a formally statistically significant association, the probability that the research finding is true is only 1.5 × 10^−4^, hardly any higher than the probability we had before any of this extensive research was undertaken! Corollary 1: The smaller the studies conducted in a scientific field, the less likely the research findings are to be true. Small sample size means smaller power and, for all functions above, the PPV for a true research finding decreases as power decreases towards 1 − β = 0.05. Thus, other factors being equal, research findings are more likely true in scientific fields that undertake large studies, such as randomized controlled trials in cardiology (several thousand subjects randomized) \[[@pmed-0020124-b14]\] than in scientific fields with small studies, such as most research of molecular predictors (sample sizes 100-fold smaller) \[[@pmed-0020124-b15]\]. Corollary 2: The smaller the effect sizes in a scientific field, the less likely the research findings are to be true. Power is also related to the effect size. Thus research findings are more likely true in scientific fields with large effects, such as the impact of smoking on cancer or cardiovascular disease (relative risks 3--20), than in scientific fields where postulated effects are small, such as genetic risk factors for multigenetic diseases (relative risks 1.1--1.5) \[[@pmed-0020124-b7]\]. Modern epidemiology is increasingly obliged to target smaller effect sizes \[[@pmed-0020124-b16]\]. Consequently, the proportion of true research findings is expected to decrease. In the same line of thinking, if the true effect sizes are very small in a scientific field, this field is likely to be plagued by almost ubiquitous false positive claims. For example, if the majority of true genetic or nutritional determinants of complex diseases confer relative risks less than 1.05, genetic or nutritional epidemiology would be largely utopian endeavors. Corollary 3: The greater the number and the lesser the selection of tested relationships in a scientific field, the less likely the research findings are to be true. As shown above, the post-study probability that a finding is true (PPV) depends a lot on the pre-study odds (R). Thus, research findings are more likely true in confirmatory designs, such as large phase III randomized controlled trials, or meta-analyses thereof, than in hypothesis-generating experiments. Fields considered highly informative and creative given the wealth of the assembled and tested information, such as microarrays and other high-throughput discovery-oriented research \[[@pmed-0020124-b4],[@pmed-0020124-b8],[@pmed-0020124-b17]\], should have extremely low PPV. Corollary 4: The greater the flexibility in designs, definitions, outcomes, and analytical modes in a scientific field, the less likely the research findings are to be true. Flexibility increases the potential for transforming what would be "negative" results into "positive" results, i.e., bias, u. For several research designs, e.g., randomized controlled trials \[[@pmed-0020124-b18]\] or meta-analyses \[[@pmed-0020124-b21],[@pmed-0020124-b22]\], there have been efforts to standardize their conduct and reporting. Adherence to common standards is likely to increase the proportion of true findings. The same applies to outcomes. True findings may be more common when outcomes are unequivocal and universally agreed (e.g., death) rather than when multifarious outcomes are devised (e.g., scales for schizophrenia outcomes) \[[@pmed-0020124-b23]\]. Similarly, fields that use commonly agreed, stereotyped analytical methods (e.g., Kaplan-Meier plots and the log-rank test) \[[@pmed-0020124-b24]\] may yield a larger proportion of true findings than fields where analytical methods are still under experimentation (e.g., artificial intelligence methods) and only "best" results are reported. Regardless, even in the most stringent research designs, bias seems to be a major problem. For example, there is strong evidence that selective outcome reporting, with manipulation of the outcomes and analyses reported, is a common problem even for randomized trails \[[@pmed-0020124-b25]\]. Simply abolishing selective publication would not make this problem go away. Corollary 5: The greater the financial and other interests and prejudices in a scientific field, the less likely the research findings are to be true. Conflicts of interest and prejudice may increase bias, u. Conflicts of interest are very common in biomedical research \[[@pmed-0020124-b26]\], and typically they are inadequately and sparsely reported \[[@pmed-0020124-b26],[@pmed-0020124-b27]\]. Prejudice may not necessarily have financial roots. Scientists in a given field may be prejudiced purely because of their belief in a scientific theory or commitment to their own findings. Many otherwise seemingly independent, university-based studies may be conducted for no other reason than to give physicians and researchers qualifications for promotion or tenure. Such nonfinancial conflicts may also lead to distorted reported results and interpretations. Prestigious investigators may suppress via the peer review process the appearance and dissemination of findings that refute their findings, thus condemning their field to perpetuate false dogma. Empirical evidence on expert opinion shows that it is extremely unreliable \[[@pmed-0020124-b28]\]. Corollary 6: The hotter a scientific field (with more scientific teams involved), the less likely the research findings are to be true. This seemingly paradoxical corollary follows because, as stated above, the PPV of isolated findings decreases when many teams of investigators are involved in the same field. This may explain why we occasionally see major excitement followed rapidly by severe disappointments in fields that draw wide attention. With many teams working on the same field and with massive experimental data being produced, timing is of the essence in beating competition. Thus, each team may prioritize on pursuing and disseminating its most impressive "positive" results. "Negative" results may become attractive for dissemination only if some other team has found a "positive" association on the same question. In that case, it may be attractive to refute a claim made in some prestigious journal. The term Proteus phenomenon has been coined to describe this phenomenon of rapidly alternating extreme research claims and extremely opposite refutations \[[@pmed-0020124-b29]\]. Empirical evidence suggests that this sequence of extreme opposites is very common in molecular genetics \[[@pmed-0020124-b29]\]. These corollaries consider each factor separately, but these factors often influence each other. For example, investigators working in fields where true effect sizes are perceived to be small may be more likely to perform large studies than investigators working in fields where true effect sizes are perceived to be large. Or prejudice may prevail in a hot scientific field, further undermining the predictive value of its research findings. Highly prejudiced stakeholders may even create a barrier that aborts efforts at obtaining and disseminating opposing results. Conversely, the fact that a field is hot or has strong invested interests may sometimes promote larger studies and improved standards of research, enhancing the predictive value of its research findings. Or massive discovery-oriented testing may result in such a large yield of significant relationships that investigators have enough to report and search further and thus refrain from data dredging and manipulation. Most Research Findings Are False for Most Research Designs and for Most Fields {#s6} ============================================================================== In the described framework, a PPV exceeding 50% is quite difficult to get. [Table 4](#pmed-0020124-t004){ref-type="table"} provides the results of simulations using the formulas developed for the influence of power, ratio of true to non-true relationships, and bias, for various types of situations that may be characteristic of specific study designs and settings. A finding from a well-conducted, adequately powered randomized controlled trial starting with a 50% pre-study chance that the intervention is effective is eventually true about 85% of the time. A fairly similar performance is expected of a confirmatory meta-analysis of good-quality randomized trials: potential bias probably increases, but power and pre-test chances are higher compared to a single randomized trial. Conversely, a meta-analytic finding from inconclusive studies where pooling is used to "correct" the low power of single studies, is probably false if R ≤ 1:3. Research findings from underpowered, early-phase clinical trials would be true about one in four times, or even less frequently if bias is present. Epidemiological studies of an exploratory nature perform even worse, especially when underpowered, but even well-powered epidemiological studies may have only a one in five chance being true, if R = 1:10. Finally, in discovery-oriented research with massive testing, where tested relationships exceed true ones 1,000-fold (e.g., 30,000 genes tested, of which 30 may be the true culprits) \[[@pmed-0020124-b30],[@pmed-0020124-b31]\], PPV for each claimed relationship is extremely low, even with considerable standardization of laboratory and statistical methods, outcomes, and reporting thereof to minimize bias. ::: {#pmed-0020124-t004 .table-wrap} Table 4 ::: {.caption} ###### PPV of Research Findings for Various Combinations of Power (1 - ß), Ratio of True to Not-True Relationships (R), and Bias (u) ::: ![](pmed.0020124.t004) The estimated PPVs (positive predictive values) are derived assuming a = 0.05 for a single study. RCT, randomized controlled trial. ::: Claimed Research Findings May Often Be Simply Accurate Measures of the Prevailing Bias {#s7} ====================================================================================== As shown, the majority of modern biomedical research is operating in areas with very low pre- and post-study probability for true findings. Let us suppose that in a research field there are no true findings at all to be discovered. History of science teaches us that scientific endeavor has often in the past wasted effort in fields with absolutely no yield of true scientific information, at least based on our current understanding. In such a "null field," one would ideally expect all observed effect sizes to vary by chance around the null in the absence of bias. The extent that observed findings deviate from what is expected by chance alone would be simply a pure measure of the prevailing bias. For example, let us suppose that no nutrients or dietary patterns are actually important determinants for the risk of developing a specific tumor. Let us also suppose that the scientific literature has examined 60 nutrients and claims all of them to be related to the risk of developing this tumor with relative risks in the range of 1.2 to 1.4 for the comparison of the upper to lower intake tertiles. Then the claimed effect sizes are simply measuring nothing else but the net bias that has been involved in the generation of this scientific literature. Claimed effect sizes are in fact the most accurate estimates of the net bias. It even follows that between "null fields," the fields that claim stronger effects (often with accompanying claims of medical or public health importance) are simply those that have sustained the worst biases. For fields with very low PPV, the few true relationships would not distort this overall picture much. Even if a few relationships are true, the shape of the distribution of the observed effects would still yield a clear measure of the biases involved in the field. This concept totally reverses the way we view scientific results. Traditionally, investigators have viewed large and highly significant effects with excitement, as signs of important discoveries. Too large and too highly significant effects may actually be more likely to be signs of large bias in most fields of modern research. They should lead investigators to careful critical thinking about what might have gone wrong with their data, analyses, and results. Of course, investigators working in any field are likely to resist accepting that the whole field in which they have spent their careers is a "null field." However, other lines of evidence, or advances in technology and experimentation, may lead eventually to the dismantling of a scientific field. Obtaining measures of the net bias in one field may also be useful for obtaining insight into what might be the range of bias operating in other fields where similar analytical methods, technologies, and conflicts may be operating. How Can We Improve the Situation? {#s8} ================================= Is it unavoidable that most research findings are false, or can we improve the situation? A major problem is that it is impossible to know with 100% certainty what the truth is in any research question. In this regard, the pure "gold" standard is unattainable. However, there are several approaches to improve the post-study probability. Better powered evidence, e.g., large studies or low-bias meta-analyses, may help, as it comes closer to the unknown "gold" standard. However, large studies may still have biases and these should be acknowledged and avoided. Moreover, large-scale evidence is impossible to obtain for all of the millions and trillions of research questions posed in current research. Large-scale evidence should be targeted for research questions where the pre-study probability is already considerably high, so that a significant research finding will lead to a post-test probability that would be considered quite definitive. Large-scale evidence is also particularly indicated when it can test major concepts rather than narrow, specific questions. A negative finding can then refute not only a specific proposed claim, but a whole field or considerable portion thereof. Selecting the performance of large-scale studies based on narrow-minded criteria, such as the marketing promotion of a specific drug, is largely wasted research. Moreover, one should be cautious that extremely large studies may be more likely to find a formally statistical significant difference for a trivial effect that is not really meaningfully different from the null \[[@pmed-0020124-b32]\]. Second, most research questions are addressed by many teams, and it is misleading to emphasize the statistically significant findings of any single team. What matters is the totality of the evidence. Diminishing bias through enhanced research standards and curtailing of prejudices may also help. However, this may require a change in scientific mentality that might be difficult to achieve. In some research designs, efforts may also be more successful with upfront registration of studies, e.g., randomized trials \[[@pmed-0020124-b35]\]. Registration would pose a challenge for hypothesis-generating research. Some kind of registration or networking of data collections or investigators within fields may be more feasible than registration of each and every hypothesis-generating experiment. Regardless, even if we do not see a great deal of progress with registration of studies in other fields, the principles of developing and adhering to a protocol could be more widely borrowed from randomized controlled trials. Finally, instead of chasing statistical significance, we should improve our understanding of the range of R values---the pre-study odds---where research efforts operate \[[@pmed-0020124-b10]\]. Before running an experiment, investigators should consider what they believe the chances are that they are testing a true rather than a non-true relationship. Speculated high R values may sometimes then be ascertained. As described above, whenever ethically acceptable, large studies with minimal bias should be performed on research findings that are considered relatively established, to see how often they are indeed confirmed. I suspect several established "classics" will fail the test \[[@pmed-0020124-b36]\]. Nevertheless, most new discoveries will continue to stem from hypothesis-generating research with low or very low pre-study odds. We should then acknowledge that statistical significance testing in the report of a single study gives only a partial picture, without knowing how much testing has been done outside the report and in the relevant field at large. Despite a large statistical literature for multiple testing corrections \[[@pmed-0020124-b37]\], usually it is impossible to decipher how much data dredging by the reporting authors or other research teams has preceded a reported research finding. Even if determining this were feasible, this would not inform us about the pre-study odds. Thus, it is unavoidable that one should make approximate assumptions on how many relationships are expected to be true among those probed across the relevant research fields and research designs. The wider field may yield some guidance for estimating this probability for the isolated research project. Experiences from biases detected in other neighboring fields would also be useful to draw upon. Even though these assumptions would be considerably subjective, they would still be very useful in interpreting research claims and putting them in context. Citation: Ioannidis JPA (2005) Why most published research findings are false. PLoS Med 2(8): e124. PPV : positive predictive value [^1]: John P. A. Ioannidis is in the Department of Hygiene and Epidemiology, University of Ioannina School of Medicine, Ioannina, Greece, and Institute for Clinical Research and Health Policy Studies, Department of Medicine, Tufts-New England Medical Center, Tufts University School of Medicine, Boston, Massachusetts, United States of America. E-mail: <jioannid@cc.uoi.gr> [^2]: Competing Interests: The author has declared that no competing interests exist.	PubMed Central	2024-06-05T03:55:59.887425	2005-8-30	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182327/", "journal": "PLoS Med. 2005 Aug 30; 2(8):e124", "authors": [ { "first": "John P. A.", "last": "Ioannidis" } ] }
PMC1182328	As a neurologist subspecializing in epilepsy at a respected academic institution, I (DH) assumed that I knew everything I needed to know about epilepsy and patients with epilepsy. I was wrong. In September of 1994, John Lester, my colleague in the Department of Neurology at Massachusetts General Hospital, showed me an online bulletin board for neurology patients that he had created \[[@pmed-0020206-b1]\]. In reading through the online messages, I observed hundreds of patients with neurological diseases sharing their experiences and discussing their problems with one another. I knew that many patients with chronic diseases had been making use of online medical information \[[@pmed-0020206-b2]\]. Nonetheless, I was shocked, fascinated, and more than a bit confused by what I saw. I\'d been trained in the old medical school style: my instructors had insisted that patients could not be trusted to understand or manage complex medical matters. Thinking back through my years of training and practice, I realized that there had always been an unspoken prohibition against groups of patients getting together. I had the uncomfortable sense that by promoting interactions between patients and de-emphasizing the central role of the physician, I might be violating some deep taboo. Remarkably Complex Stories {#s2} ========================== My initial doubts notwithstanding, I found dozens of well-informed, medically competent patients sharing information on a variety of topics. I was especially struck by the many stories recounting the development of a particular patient\'s illness, the patient\'s efforts to manage it, and the resulting interactions with health professionals. By telling their stories in such elaborate detail, experienced group members could offer a great deal of useful advice and guidance to those newly diagnosed, based on what they had learned in their own online research, what they had been told by their clinicians, and what they had deduced from personal experiences with the disease. These "patient stories" often included a number of empowering elements that set them apart from the advice patients typically receive from their clinicians: role modeling by an active, critical, well-informed "expert patient" (\[[@pmed-0020206-b1]\]; <http://patientweb.net>), comparative reviews and recommendations of clinicians and treatment facilities \[[@pmed-0020206-b2]\], and advice about how to handle the practical details of living with a chronic illness \[[@pmed-0020206-b6]\] (such as how to organize a home medical record, manage treatment side effects, find the best drug prices, and deal with less-than-perfect health professionals and health-care provider systems, and a wide variety of other topics relating to effective medical self-management). These extended patient narratives---no two alike---thus gave rise to an accumulated body of what my colleagues and I began to think of as an expert patient knowledge base. We concluded that these patient narratives could be invaluable resources for clinicians and researchers, interested in taking an in-depth look at the changing roles of patients and clinicians in the Internet age. The constant outpouring of sympathy and support that we observed in interactions among community members surpassed anything a patient might conceivably expect to receive at a doctor\'s office. As Richard Rockefeller, President of the Health Commons Institute, has suggested, disease-specific online patient networks provide their members with an invaluable type of around-the-clock support that he has called the "chicken soup of the Internet" \[[@pmed-0020206-b7]\]. Working with several colleagues, I initiated an observational study to analyze the ways in which E-patients were using this new medium. Since I am an epilepsy specialist, we decided to focus on an epilepsy support group at the site Lester had created, BrainTalk Communities (<http://www.braintalk.org>) ([Figure 1](#pmed-0020206-g001){ref-type="fig"}) \[[@pmed-0020206-b8]\]. The BrainTalk Communities currently host more than 300 free online groups for neurological conditions (such as Alzheimer disease, multiple sclerosis, Parkinson disease, chronic pain, epilepsy, and Huntington disease) for patients across the globe. More than 200,000 individuals visit the BrainTalk Communities\' Web site on a regular basis. This site is now owned and operated by an independent nonprofit group, BrainTalk Communities, and is no longer formally associated with Massachusetts General Hospital. ::: {#pmed-0020206-g001 .fig} Figure 1 ::: {.caption} ###### Logo of the BrainTalk Communities---Online Patient Support Groups for Neurology ::: ![](pmed.0020206.g001) ::: What we found surprised us. We assumed that most interactions would be support related, with some members describing their medical experiences and others offering active listening, sympathy, and understanding. But while such interactions were an important part of the group process, they were observed in only about 30% of the postings. In the remaining 70% of the postings, group members provided each other with what amounted to a crash course in their shared disease, discussing topics such as the anatomy, physiology, and natural history of the disorder; treatment options and management guidelines for each form of treatment; and treatment side effects, medical self-management, the day-to-day practicalities of living with the disease, and the effects of their condition on family and friends ([Table 1](#pmed-0020206-t001){ref-type="table"}). ::: {#pmed-0020206-t001 .table-wrap} Table 1 ::: {.caption} ###### BrainTalk Communities Online Epilepsy Support Group: Types of Questions Asked by Users ::: ![](pmed.0020206.t001) Source: \[[@pmed-0020206-b8]\]. ::: A Source of Information {#s3} ======================= Much of the information that the group provided to members was similar to what I routinely provided to my own clinic patients. So I was surprised to learn that many of the clinicians caring for group members provided considerably less information, guidance, and support. And some, apparently, provided none at all. Statements such as "My provider is too busy," "My provider doesn\'t care," or "My provider doesn\'t seem to know about such-and-such" were alarmingly common. About 10% of the members\' posts spontaneously mentioned that they had been unable to get the medical information that they needed from their own clinicians. When we surveyed members directly, more than 30% said that they had been unable to obtain all the medical information they would have liked from their physicians ([Table 2](#pmed-0020206-t002){ref-type="table"}). This was a primary reason for many members\' participation in the group. ::: {#pmed-0020206-t002 .table-wrap} Table 2 ::: {.caption} ###### BrainTalk Communities Epilepsy Support Group: Responses to a Survey of Users ::: ![](pmed.0020206.t002) Source: \[[@pmed-0020206-b11]\]. ::: Some other types of information, especially practical tips for living with epilepsy and the social aspects of the disease, went far beyond what I had been providing for my own patients. I am a board-certified epilepsy specialist at one of the most highly respected medical centers in the United States, yet I learned a great deal about these topics from the support group. I now share many of the things I learned from group members with my clinic patients. The BrainTalk Communities epilepsy support group that we observed was facilitated by volunteer patient moderators, with little or no professional input. About 6% of the postings contained information that some of our medical reviewers considered at least partly mistaken, misinterpreted, outdated, or incomplete. We observed that other group members frequently corrected such misinformation. And group participants appeared to understand that they should not take uncorroborated statements as hard facts. They seemed well aware that some postings were erroneous, and in fact seemed to substantially overestimate the incidence of questionable materials. We observed no serious problems as a result of these questionable postings, and saw many reports by patients who had obtained better care, prevented medical mistakes, or averted serious injury because of the information and advice they received from fellow group members. We concluded that, as Ferguson and Frydman have suggested, many professionals have seriously overestimated the risks and underestimated the benefits of online support groups and other online health resources for patients, probably because they do not operate within our familiar professionally centered constructs \[[@pmed-0020206-b9]\]. What I\'ve Learned {#s4} ================== In retrospect, the most important thing I (DH) have learned from our online group was that patients want to know about, and in most cases are perfectly capable of understanding and dealing with, everything their physician knows about their disease and its treatments. After observing the group, I realized that I had been providing my patients with a very limited subset of what I knew about their condition. Today, there is nothing that I know about epilepsy that I would hesitate to share with a patient. For example, I now offer my patients an open and frank discussion of the very rare sudden unexpected death in epilepsy syndrome. I had previously not mentioned this rare but alarming complication, fearing that some patients might become overly concerned with it. But once I discovered that BrainTalk Communities group members discussed this topic quite openly and freely online, reviewing the scientific data in a sophisticated way, I began to share my knowledge on this topic with my clinic patients. My newfound frankness has been much appreciated. And none of my patients have become unduly troubled by these discussions. I have also learned that an online group like the BrainTalk Communities epilepsy group is not only much smarter than any single patient, but is also smarter, or at least more comprehensive, than many physicians---even many medical specialists. While some postings do contain erroneous material, online groups of patients who share an illness engage in a continuous process of self-correction, challenging questionable statements and addressing misperceptions as they occur. And while no single resource, including physicians, should be considered the last word in medical knowledge, the consensus opinion arrived at by patient groups is usually quite excellent. And if more expert clinicians offered to consult informally with the online support groups devoted to their medical specialties---as I now do---we could help group members make information and opinion shared in these groups even better. I had been taught to believe that patients could only be "empowered" by their clinicians. And while I do believe that clinicians can help in this regard by sharing their knowledge openly and by encouraging patient self-reliance, it now seems quite clear that growing numbers of patients are perfectly capable of empowering themselves, with or without their clinician\'s blessing. Physicians and other health professionals should do all they can to support them in this worthy effort. As a result of what we\'ve learned from these online patient networks, our research group has developed a password-protected Web site, PatientWeb (<https://fisher.mgh.harvard.edu/>), for the patients that we see in the clinic---all those patients with epilepsy who receive medical care at the Massachusetts General Hospital and Brigham and Women\'s Hospital. Thanks to what we have learned from these online groups, we plan to pilot new ways for private, local online groups made up of patients with the same disease and receiving care from the same clinicians to collaborate with each other, and with their clinicians, more effectively. Conclusions {#s5} =========== Clinicians have overestimated the downsides, while seriously underestimating the benefits, of condition-specific online patient support communities. These free online resources now provide invaluable services 24 hours a day, seven days a week, for patients across the country and around the world. It would be unfortunate indeed if medical professionals let their uneasiness at this emerging trend toward patient empowerment and autonomy cloud their ability to assess the impressive benefits these groups provide. Many patients are now ready, willing, and able to take a more active role in their own care, and the care of others with related diseases. By encouraging patients to do more for themselves and for each other, clinicians can help mitigate many of the negative effects of contemporary time-pressured medical practice. Thus, even though there may now be less time for the counseling, storytelling, support, information sharing, and empowerment-based training that was once a routine part of the typical office visit, we can now help our patients obtain such services by referring them to online patient networks. The distributed expertise of online support groups is by no means limited to the emotional aspects of the illness and to the practical logistics of living with the disorder. It can also include current reviews of the literature, reports from the latest medical meetings, accounts of behind-the-scenes activities at the best treatment centers, sophisticated guidance on dealing with medical professionals, and excellent advice on dealing with complex aspects of medical management. Finally, I have concluded that few, if any, physicians could have created a system like BrainTalk Communities. As a tech-savvy non-physician intimately familiar with both the inner workings of medical care and the power of information technology systems to create effective online communities, John Lester was less proprietary than most physicians are about medicine\'s proper professional "turf." He was also less inhibited by professional biases regarding the potential value of the medical contributions that "unqualified" individuals might make. This is not an isolated occurrence. We suspect that the intensely professionally centered enculturation most physicians receive in their training and practice environments may render them, in the words of John Seely Brown and Paul Draguld, "blinkered if not blind" to the emergence of many promising new technocultural changes, which currently present new opportunities for health-care innovation \[[@pmed-0020206-b10]\]. Thus, physicians who seek to innovate in these areas might benefit greatly---as I have---from joining forces with Web developers, Net-savvy social scientists, experienced E-patients, and other colleagues unencumbered by the limiting belief systems that may result from our traditional medical training. In light of their empowering social dynamics and volunteer economics, we suspect that patient-led online groups may prove to be a considerably more promising and sustainable health-care resource than professionally moderated therapy groups. And we are convinced that networked work teams linking patients, caregivers, and medical professionals will be an important model for future health-care innovation. This article was written collaboratively but presents DH\'s point of view and reflects his experience. The authors wish to thank and acknowledge John Lester, Stephanie Prady, and Joshua Fogel for reviewing earlier drafts of this article and offering helpful suggestions. Citation: Hoch D, Ferguson T (2005) What I\'ve learned from E-patients. PLoS Med 2(8): e206. [^1]: Dan Hoch is an assistant professor of Neurology at Harvard Medical School, Boston, Massachusetts, United States of America. Tom Ferguson is a senior research fellow at the Pew Internet and American Life Project, Austin, Texas, United States of America. [^2]: Competing Interests: DH is a secretary for and TF is a director of the nonprofit organization BrainTalk Communities.	PubMed Central	2024-06-05T03:55:59.890514	2005-8-9	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182328/", "journal": "PLoS Med. 2005 Aug 9; 2(8):e206", "authors": [ { "first": "Dan", "last": "Hoch" }, { "first": "Tom", "last": "Ferguson" } ] }
PMC1182345	We previously demonstrated the presence of delay-period activity in midbrain dopamine neurons, and provided evidence that this activity corresponds to uncertainty about reward. An alternative interpretation of our observations was recently put forth in which it was suggested that the delay-period activity corresponds not to uncertainty but to backpropagating TD prediction errors. Here we present evidence that supports our original proposal but appears inconsistent with the alternative interpretation involving backpropagating errors. Because the activity of dopamine neurons appears to code reward prediction error, it has been suggested that dopamine neurons may provide a teaching signal in analogy to the prediction error found in temporal difference (TD) models of reinforcement learning. Taking the analogy a step further, it has also been proposed that particular TD models may describe the activity of dopamine neurons \[[@B1],[@B2]\]. More recently, we have reported that dopamine neurons show a gradual increase in activity that occurs between onset of a conditioned stimulus (CS) and reward when the CS is associated with uncertainty about the reward outcome \[[@B3]\]. Niv et al \[[@B4]\] have now suggested how a conventional TD model might account for this observation without reference to uncertainty. Their explanation relies on the fact that, in certain TD models, prediction errors \"backpropagate\" in time over consecutive CS presentations. In our experiments, on a particular trial a prediction error occurs immediately after reward onset, which occurs 2 seconds after CS onset. According to the backpropagation model favored by Niv et al, on the next trial in which that same CS is presented, an internally timed \"prediction error\" would occur at a shorter delay, perhaps at 1.9 seconds after CS onset. On each subsequent trial, the error would occur at a shorter delay until finally it immediately follows the onset of the CS. This model would require that neurons show sudden increases or decreases in activity at long but precisely timed delays after stimulus onset. Although the implementation of such a scheme by real neurons is questionable, it nonetheless might account for the observed delay period activation if one makes the additional assumption that neuronal firing rate has a particular nonlinear relationship to prediction error. For example, Niv et al argue that the difference between 1 and 2 spikes per second has a much greater functional impact in terms of prediction error than the difference between 9 and 10 spikes per second. Thus, adding activity across trials, as we did to generate histograms, would result in the appearance of neuronal activation despite the fact that the average activity at all times (except immediately after CS onset) would correspond to a prediction error of zero. Below we present some of the reasons that we are skeptical of the interpretation of Niv et al. First, the nonlinear relationship suggested by Niv et al between the firing rate of dopamine neurons and the functional prediction error is opposite to the experimentally observed nonlinear relationship between firing rate and dopamine concentration in mesolimbic target regions. Chergui et al \[[@B5]\] found that there is more extracellular dopamine per impulse at higher firing rates than at lower firing rates. Second, inspection of the published data appears inconsistent with the model of Niv et al. They suggest that the delay-period activity is an artifact of averaging over trials to generate histograms, and that the sustained increase in activity does not occur in single trials. Contrary to their proposal, there does appears to be strong and sustained activation within single trials, as shown in figure 2 of our original report \[[@B3]\] and in data from another neuron shown here in figure [1A](#F1){ref-type="fig"}. It is difficult to be certain whether or not activity increases on single trials, in part because Niv et al have not specified precisely what a single-trial increase in delay-period activity should look like, and in part because of the general problem within neuroscience of how to interpret spike trains. Indeed, it would seem that any spike could conceivably represent a backpropagating positive error, and any inter-spike interval could correspond to a negative error. However, if we take a more constrained, conventional approach based on firing rates over tens of milliseconds, then firing rate appears to increase during the delay period on single trials. Similarly, gradual changes in neural activity related to reward expectation are observed in many other types of neuron \[\[[@B6],[@B7]\], for example\], and are widely believed to represent meaningful increases in activity on single trials rather than artifacts of averaging over trials. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### The delay-period activity of dopamine neurons does not appear to depend on the reward outcome of the last presentation (trial) of the same conditioned stimulus (CS). Additional analysis was performed on the data for p = 0.5 from figure 3 of Fiorillo et al \[3\]. (and reproduced by Niv et al \[4\] in their figure 1A). Rasters and histograms were generated after segregating trials according to whether the last presentation of the same CS was or was not rewarded. Both rewarded and unrewarded trials are shown. According to the hypothesis of Niv et al, one might expect to see less delay-period activity if the last presentation of the CS was followed by no reward. However, no difference was observed (see statistics in main text). A.Activity in a single dopamine neuron. B.Average activity in a population of 28 neurons selected for the presence of delay-period activity. ::: ![](1744-9081-1-7-1) ::: Third, additional analysis of the data (averaged over trials) challenges the interpretation of Niv et al. If the activity during the delay period is due to backpropagating \"error\" signals that originated in previous trials, then the activity in the last part of the delay period should reflect the reward outcome that followed the last exposure to that same CS. Thus there should be more activity at the end of the delay period if the last trial was rewarded, and less if it was unrewarded. We have analyzed trials in which the CS predicted reward at p = 0.5, and found no dependence of neural activity on the outcome of the preceding trial of the same CS (Fig. [1A,B](#F1){ref-type="fig"}) (comparing either the last 100 or 500 ms before reward: p \> 0.05 in 51 of 54 neurons, Mann-Whitney test; p \> 0.4 for the population of 54 neurons, Wilcoxon test). Thus the delay-period activity does not appear to depend on the outcome of the last trial, as suggested by Niv et al. Fourth, our more recently published results \[[@B8]\] are inconsistent with the model of Niv et al. Each of three conditioned stimuli predicted two potential reward outcomes of equal probability. The discrepancy in liquid volume between the two potential reward outcomes varied according to the CS. The greater the discrepancy, the more pronounced was the sustained, ramp-like increase in neural activity (Fig [2A](#F2){ref-type="fig"}) \[[@B3]\]. However, the phasic response following reward (or omission of reward) was identical across the three conditions, revealing an adaptation of the prediction error response to the expected discrepancy in reward magnitude (Fig. [2B](#F2){ref-type="fig"}) \[[@B8]\]. If one were to incorporate these recently published results \[[@B8]\] into the backpropagation TD model of Niv et al, then one would find that since the reward prediction error response at the end of each trial in these experiments is the same, the delay-period activity representing the backpropagating errors would also be the same. However, the data are inconsistent with the model, since the delay period activity increases with the discrepancy between potential reward magnitudes (Fig. [2A](#F2){ref-type="fig"}) \[[@B3]\]. Our results \[[@B8]\] show that although the phasic activity of dopamine neurons corresponds well to a general definition of reward prediction error, it is inconsistent with the explanation of the delay period activity proposed by Niv et al. Fifth, it should be noted that the backpropagating prediction error in the model of Niv et al does not reflect an inherent necessity of TD models, but is rather a consequence of the specific temporal stimulus representation chosen. The implementation of different temporal stimulus representations can lead to quite different results. The original TD model \[[@B9]\] and recent versions \[[@B10]\] have used temporal stimulus representations in which the transfer of the neuronal response to the CS is accomplished in a manner that appears more biologically plausible than backpropagation. In TD models utilizing backpropagation, neural signals during the delay period are precisely timed but are without functional consequence, since the sequence of positive and negative errors are self-generated (occurring in the absence of any external events) but are presumed to cancel each other out. This strikes us an odd notion that is neither efficient, nor elegant, nor necessary to the principles of TD learning. When discrepancies between TD models and responses of dopamine neurons have been noted in the past, such as the absence of depression at the usual time of reward on trials in which reward is delivered earlier than usual \[[@B11]\], TD models have been modified accordingly to better describe the neural activity \[[@B10],[@B12],[@B13]\]. Although TD models have proven very useful, one would not necessarily expect to find the formal structure of any current TD model implemented in the brain. Present TD models exhibit a number of characteristics that appear to be motivated by the need for simplification rather than by any empirical or theoretical constraint. For example, the prediction in TD models is typically equated with expected reward value and ignores the uncertainty in prediction. To illustrate this in terms familiar to people (and perhaps also of relevance to dopamine neurons), a 10% chance of gaining \$100 is clearly not equivalent to a 100% chance of gaining \$10, yet TD models do not discriminate amongst these two scenarios. TD models have evolved over the years to become more useful and realistic. We believe this process will continue and hope that the study of neurons might be helpful in this regard. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### When visual stimuli predict reward at p = 0.5, the delay-period activity of dopamine neurons increases with the discrepancy between potential reward magnitudes, but the phasic response to reward delivery is independent of reward magnitude. A.The onset of each of three distinct visual stimuli was followed by either of two potential liquid volumes with equal probability. Histograms represent the average activity of 35 dopamine neurons; these neurons were not selected for the presence of any task-related modulation. Delay-period activity increased with the discrepancy between potential liquid volumes. See figure 4 of Fiorillo et al \[3\] for a full summary of this data. B.Population histograms showing the average response of dopamine neurons to the delivery of reward in the same experiment as illustrated in A (n = 57). Data are from figure 4 of Tobler et al \[8\], and include data from 22 neurons tested with trace conditioning (as described in \[3\]) that were not included above in panel A. No differences were observed in the phasic activation to reward in trace versus delay conditioning. The critical observation here is that the delay-period activity varies (from top to bottom) in panel A, but the phasic \'prediction error\' response to reward does not (as shown in B). This data is inconsistent with the proposal of Niv et al., according to which the delay-period activity shown in each panel in A should scale with the corresponding phasic prediction error response shown in B. Since the two responses do not scale together, it appears that the delay-period activity cannot be accounted for by the backpropagation of prediction errors. ::: ![](1744-9081-1-7-2) ::: Acknowledgements ================ This work was supported by the Wellcome Trust (W.S. and P.N.T.) and the Howard Hughes Medical Institute (C.D.F.).	PubMed Central	2024-06-05T03:55:59.892059	2005-6-15	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182345/", "journal": "Behav Brain Funct. 2005 Jun 15; 1:7", "authors": [ { "first": "Christopher D", "last": "Fiorillo" }, { "first": "Philippe N", "last": "Tobler" }, { "first": "Wolfram", "last": "Schultz" } ] }
PMC1182346	Background ========== Complementary and alternative medicine (CAM) has a reputation for good value among health conscious consumers \[[@B1]\]. In the United States consumers spend over \$34 billion per year on CAM therapies \[[@B2]\], dollars spent outside the conventional health care financing system. Such evidence on out-of-pocket expenditures is a testament to the widely held belief that CAM therapies have benefits that outweigh their costs. Regardless of public opinion, there is often little more than anecdotal evidence on the health and economic implications of CAM therapies. The paucity of outcomes research in CAM has likely depressed access to CAM therapies by impeding their integration into financial mechanisms commonly found in conventional health care. Most US consumers who have health insurance coverage, either through public or private institutions, bear the entire cost of CAM therapies out-of-pocket \[[@B3]\]. Theoretically, CAM therapies seem effective and a good candidate for cost savings because they avoid high technology, offer inexpensive remedies, and harness the power of vis medicatrix naturae(the body\'s natural ability to heal itself). As such, a thorough and external review of economic and health outcomes of CAM is necessary for evidence-based consideration of CAM therapies as a covered expense. That being said, it is also known that affirmative evidence on economic and health outcomes is a necessary, but not sufficient step toward CAM coverage, and not the decision itself. Other factors such as historical demand, political expediency, consumer demand, and practitioner enthusiasm may also be considered in the decision to incorporate CAM into a health insurance policy \[[@B1],[@B4],[@B5]\]. The need for economic evaluations is also growing in conventional healthcare. An increasing number of health plans and hospitals have moved from a simple budgetary focus in formulary decisions to requiring detailed evidence on the economic value of considered therapies relative to alternatives \[[@B6],[@B7]\]. Beyond their use in decisions concerning health insurance coverage, economic outcomes of both CAM and conventional therapies also influence health policy, justify licensure of practitioners, inform industry investment decisions, provide general evidence to consumers about potential economic benefits, and can guide future research efforts through identifying decision-critical parameters for additional research \[[@B8],[@B9]\]. In their systematic review of CAM economic evaluations, White and Ernst \[[@B4]\] identified 34 economic evaluations of CAM conducted between 1987 and 1999; only eleven of which were full economic evaluations (ie, compared both economic and health outcomes between two or more alternatives) \[[@B10]\]. Quality was evaluated by noting whether cost data were collected prospectively and whether comparison groups were comparable -- ie, assigned randomly. Unfortunately, their search strategy included the term \"alternative medicine\" but not \"complementary medicine.\" Therefore, all single therapy studies in their review are of CAM therapies that are usually used as substitutes (alternatives) to conventional care (eg, acupuncture, homeopathy, and spinal manipulation). No studies of complementary therapies (those used in conjunction with conventional care) were included, despite the use of the term \"complementary\" in their conclusion that spinal manipulative therapy may have benefits for back pain, but \"there was a paucity of rigorous studies that could provide conclusive evidence of differences in costs and outcomes between other complementary therapies and orthodox medicine \[[@B4]\].\" The objectives of this paper are: 1) to introduce concepts commonly applied in economic evaluations of health technologies (often called technology assessment) so that practitioners and CAM users can translate and benefit from published evidence; and 2) present a systematic review of the current scope and quality of economic evaluations of CAM. We begin with an overview of economic evaluation, including didactic examples from the CAM economic literature to help clarify the concepts presented. Readers familiar with this type of analysis can skip this section and proceed directly to the methods section. In our systematic review we expand upon and update the initial review by White and Ernst. We evaluate study quality in more detail, using both additional study design criteria and quality of reporting criteria, and present a summary of the results from exemplary studies. While their review was the first of its kind, economic evaluations in the CAM literature have improved greatly in the last five years. We end the paper with a description of the attributes of CAM that make economic evaluation challenging and how these issues may be addressed. We hope that practitioners\' interest in economic evaluation will continue to grow, leading to greater incorporation of this research into CAM trials. What is an economic evaluation? ------------------------------- An economic evaluation is a comparison of outcomes among alternative ways of achieving common objectives. These analyses are conducted according to explicit, systematic, and consistent criteria, and take into account both the positive and negative consequences of each alternative. Consequences may include economic, clinical, and humanistic outcomes, known as the ECHO model \[[@B11]\]. Economic outcomes represent the consumption and production of resources and their monetary value from the perspective of a decision maker. Clinical outcomes are medical events that are professionally meaningful. Humanistic outcomes are a broad category of intangible personal attributes, typically collected through self-report. Humanistic outcomes include quality of life characteristics such as sense of safety, physical comfort, enjoyment, meaningful activity, relationships, functional competence, dignity, privacy, individuality, autonomy, and spiritual well-being. Conventionally, clinical and humanistic outcomes are considered health outcomes, and we follow this convention for the remainder of the article. There are several forms of economic evaluations that can be performed (cost-effectiveness analysis being only one of these) and each differs based on the selection and measurement of health outcomes. The perspective (or point of view) taken for the analysis also influences the selection and measurement of consequences, because not all outcomes are important to all decision makers. Generally, there are three perspectives for economic analysis: individual (eg, patient), institutional (eg, health maintenance organization), or societal. The societal perspective accumulates all outcomes, while individual and institutional analyses are more selective. Regardless of perspective, the objective of an economic evaluation is to provide information on consequences relating to alternatives faced by a decision maker. The most basic form of economic evaluation is a table that lists the individual economic and health outcomes of alternative interventions. This table is known as a cost-consequence study. Cost-identification studies and cost-minimization analyses only address economic outcomes and are discussed below in that section. The remaining forms of economic evaluations summarize economic and health outcomes into a single result (Table [1](#T1){ref-type="table"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Three forms of full economic evaluations ::: ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Cost-benefit Analysis (CBA) Cost-effectiveness Analysis (CEA) Cost-utility Analysis (CUA) --------------------------- ---------------------------------------------------- -------------------------------------------------------- ---------------------------------------------------------------------------------- Number of Health Outcomes Multiple outcomes One outcome Multiple outcomes Unit of Health Outcomes Summary measure in monetary units (eg, US dollars) Natural units (eg, reduction in number of hot flashes) Summary measure in quality of life units (eg, quality-adjusted life-years, QALY) Results Net benefits\ Cost-effectiveness ratio\\ Cost-utility ratio\\ (B~1~+ B~2~- C~1~- C~2~) (C~1~- C~2~) /(E~1~- E~2~) (C~1~- C~2~) /(QALY~1~- QALY~2~) ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- \* Results are calculated when both the costs and the effects (health outcomes) of one therapy are higher than those of another. When the costs are lower and the effects are higher for one therapy, it is said to dominate the alternative (and the alternative is said to be dominated) and no ratio is presented. C~1~= total costs of alternative 1; C~2~= total costs of alternative 2; B~1~= monetary value of health outcomes of alternative 1; B~2~= monetary value of health outcomes of alternative 2; E~1~= health effects of alternative 1; E~2~= health effects of alternative 2; QALY~1~= quality-adjusted life-years of alternative 1; QALY~2~= quality-adjusted life-years of alternative 2. ::: The advantages of performing cost-benefit and cost-utility analyses are that multiple outcomes are summarized into a single unit, either monetary units such as dollars (CBA) or QALYs (CUA) and that therapies with different sets of health outcomes can be compared based on the differences in the summary measures. Cost-benefit analysis has the additional benefit of directly indicating whether the therapy pays for itself. The disadvantages of CBA and CUA come from the techniques required to produce a summary measure. Cost-benefit analysis requires putting a monetary value on all health outcomes (and ultimately on life), and cost-utility analysis assigns value to health outcomes based on their contribution to quality of life under the presumption of population-based preferences. An extensive literature addresses the methodological and theoretical issues involved in the construction of these summary measures. The process usually occurs in two steps. In the first step, health outcomes of the intervention are measured, and in the second the outcomes are valued in summary units and aggregated. Cost-benefit analyses often assess the monetary value of health outcomes based on willingness-to-pay using a technique called conjoint analysis \[[@B12]-[@B14]\]. Willingness-to-pay inherently places a lower values of life on individuals with low income, because they can not pay what they do not have. Cost-utility analyses have multiple methods to place quality of life values on health outcomes, also known as social tariffs. Summary measures of quality of life may not be sensitive enough to pick up short-term changes such as for acute conditions and will not pick up specific clinical outcomes like blood pressure control \[[@B15]\]. Examples of instruments used to capture these general health states include the EuroQoL (EQ-5D) \[[@B16]\] and the Health Utilities Index \[[@B17]\]. Cost-effectiveness analysis (CEA) is the current standard in the literature, and has the most straight forward interpretation. Under CEA, therapies useful for a specific disease or condition can be directly compared using a metric of effectiveness relevant to that condition, such as blood pressure control. Although these types of analyses do not allow a summary of multiple outcomes they tend to respond well to the most urgent questions, such as how much would it cost to reduce the number of gestational diabetes cases by 10%? Clearly, a reduction in gestational diabetes cases has measurable implications in quality of life and economic units, but the creation of a summary measure is not necessary to address the decision maker\'s question. No matter the approach taken, it is recommended that the estimated outcomes (economic, clinical and humanistic) of health care alternatives used in economic evaluation are best estimated in pragmatic clinical trials that directly and realistically compare the therapies of interest \[[@B10]\]. Rarely are the results of placebo-controlled trials appropriate \[[@B1],[@B4],[@B18]-[@B20]\]. Also, since many CAM therapies target chronic disease, it is important that the study period be long enough to capture the full benefits and costs of each therapy, and that future costs and benefits be discounted to the present for comparison. Finally, all economic evaluations should include some type of sensitivity analysis to test the robustness of results to the various assumptions made \[[@B1],[@B20],[@B21]\]. What are economic outcomes? --------------------------- Economic outcomes are the net bundle of resources forgone due to an intervention valued at the opportunity cost of those resources (the value of their next best use or \"opportunity\"). Since the cost of a therapy differs depending on whether you are a patient, a health plan, or a health care provider, the economic outcomes (ie, costs) of each therapy depend on the perspective of the study. Studies that only measure the economic outcomes of interventions are known as cost-identification studies. A study that describes the economic and health outcomes of a single therapy can also be called a cost-identification study. These studies inform full economic evaluations. That is, they provide the data needed to better design future studies that consider both the economic and health outcomes of two or more alternative therapies. A cost-minimization analysis (CMA) explicitly assumes equivalence in health outcome among alternative therapies, and examines only economic outcomes. In practice, it appears the same as a cost-identification study, but under the assumption of equivalence, a CMA is a full economic evaluation. Table [2](#T2){ref-type="table"} has been summarized from other references \[[@B1],[@B20],[@B22]\] and gives a list of the types of economic outcomes and the perspective of analysis where each is considered. Note that these types of economic outcomes should be inclusive of both the full costs of the therapy and of any treatment for adverse effects, which can be expensive. In economic evaluations, the safety of a therapy is addressed through accounting for the cost of treating these adverse events as well as through their impact on clinical and quality of life outcomes. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Economic outcomes to include in economic evaluation ::: ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Type of Cost Examples Perspectives in Which This Cost is Included --------------------------- -------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------- Direct costs: Medical Intervention costs:\ Portion paid by health plan included in institutional perspective\ Practitioner fees\ Portion paid by patient included in individual perspective\ Diagnostic costs\ All included in societal perspective Therapy costs\ Service costs:\ Facilities and equipment, including hospitalization or clinic/office costs Ancillary staff Direct costs: Non-medical Transportation costs\ Usually all paid by the patient, so often included in individual perspective\ Time off work for appointments/hospitalization All included in societal perspective Indirect costs Lost work productivity during recuperation\ Usually all paid by the patient, so often included in individual perspective\ Lost leisure time\ All included in societal perspective Child care costs\ Costs to care givers Intangible costs Pain\ Not usually included as costs; instead, may be included in humanistic outcomes in cost-utility analysis Suffering\ Grief ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Summarized from similar tables in other references \[1, 20, 22\]. ::: It is recommended that economic outcome data are best collected prospectively as part of a pragmatic clinical trial \[[@B1],[@B4],[@B19],[@B20]\]. Inclusion and exclusion criteria for cost data should be established in the protocol, as for clinical outcome measurements, but provision must be made to add extra categories of costs which only become apparent after the trial has commenced \[[@B1],[@B20]\]. Many studies try to collect cost data retrospectively, often after a therapy has shown clinical effectiveness. However, retrospective data collection is seldom fertile, adapted, or exhaustive, and it is subject to bias \[[@B18],[@B20]\]. Examples of the different forms of economic evaluations of CAM -------------------------------------------------------------- Our systematic review of the CAM economic evaluation literature (presented below) revealed no cost-consequence studies and no cost-benefit analyses. However, we did find examples of a cost-identification study, cost-minimization analysis, cost-effectiveness analysis, and cost-utility analysis. These examples are presented below. ### Cost-identification study Frenkel and Hermoni, 2002 \[[@B23]\], performed a retrospective comparison of medication consumption costs from computerized medication charts three months before and three months after a homeopathic intervention for atopic and allergic disorders. The review was performed on 48 consecutive self-referred patients in one clinic over one year with a diagnosis of an atopic condition who agreed to a classical homeopathic treatment in addition to usual conventional care. Of the 31 medication users (prescription and non-prescription allergy-related medications) before the intervention, 27 reduced their use, two increased their use, and two had their medication level unchanged after the intervention. Of the 17 who had not used medication before the intervention, 4 began medication after the intervention. There was an average drop in 3-month medication costs after homeopathy of \$14 (1998 US\$) or 54% per person. ### Cost-minimization analysis Herron and Hillis, 2000 \[[@B24]\], retrospectively compared government payments to physicians for 1418 Quebec health insurance enrollees who practiced the Transcendental Meditation (TM) to payments for 1418 randomly selected and matched enrollees who did not. Long term health outcomes were assumed to be equal for both groups. Before starting meditation, the groups were similar in the yearly rate of increase in payments. After starting TM, annual physician payments for the meditation group declined 1 to 2% per year, while those for the non-TM group increased annually over the six year period. The difference in the annual change in payments was statistically significant at a rate between 5 and 13% per year. ### Cost-effectiveness analysis Franzosi et al, 2001 \[[@B25]\], prospectively gathered health and economic outcomes during the 3.5 year follow-up period of a large randomized open-label study (n = 5664) of omega-3 polyunsaturated fatty acids (n-3 PUFA) as secondary prevention for patients with recent myocardial infarction. The perspective was that of a third-party payer; accordingly only direct health care costs (hospital admissions, laboratory and diagnostic tests, and medications) were considered. The incremental number of life-years saved by n-3 PUFA treatment over the 3.5 years (discounted at 5%) was 0.0332 per patient. The incremental cost discounted over the same period was 817€ per patient. Therefore, the incremental cost-effectiveness ratio is 24,603€ (approximately \$25, 415 in 1999 US\$ \[[@B26]\]) per life-year saved. ### Cost-utility analysis Korthals-de Bos et al, 2003 \[[@B27]\], performed an economic evaluation alongside a randomized controlled trial to compare manual therapy, physiotherapy, and care by a general practitioner for neck pain. The study used the societal perspective and collected direct and indirect costs (including hours of help from family and friends, and hours of absenteeism from work or other activities) through the use of cost diaries kept by patients over one year. Data on each patient\'s overall health state were gathered at baseline and at one year using a survey instrument called the EuroQoL \[[@B16]\]. The utility of these health states were then calculated by using \"society\'s\" preferences for each of those health states. Society\'s preferences were estimated from a sample of the general population by the developers of the EuroQoL instrument. Using the comparison of manual therapy to general practitioner care, manual therapy had a lower one-year cost (\$402, US\$) than general practitioner care (\$1241). The QALYs were 0.82 for manual therapy and 0.77 for general practitioner care. Since the costs were lower and the QALYs higher for manual therapy as compared to usual care, manual therapy is said to dominate general practitioner care and no cost-utility ratio is calculated. Methods ======= The National Center for Complementary and Alternative Medicine (NCCAM) defines CAM as \"a group of diverse medical and health care systems, practices, and products that are not presently considered to be part of conventional medicine \[[@B28]\].\" We further defined CAM as including only those therapies that could be prescribed (or recommended) and/or performed by a CAM practitioner who does not also have a conventional medical license (eg, doctor of medicine -- MD, or doctor of osteopathy -- DO). Therefore, we did not include therapies such as chemotherapy regimens nor therapies requiring surgical implantation (such as neuroreflexotherapy \[[@B29]\]) as CAM therapies even though these therapies do appear in searches using the keywords complementary and/or alternative medicine. We also did not include well-accepted vitamin and mineral supplementation therapies such as calcium and vitamin D for osteoporosis, niacin for dyslipidemia, and vitamin B12 and folic acid for homocysteine reduction. Search strategy --------------- We searched the following electronic databases from January 1999 to October 2004: Medline, AMED, Alt-HealthWatch, and the Complementary and Alternative Medicine Citation Index via NCCAM and the National Library of Medicine (NLM). Searching was restricted to English language journals and human studies with the keywords: complementary medicine or alternative medicine, and costs or cost analysis or cost-benefit or cost-effective or economic analysis or economic evaluation. We removed duplicates from the search results and selected papers that reported original data on specific CAM therapies from any form of standard economic analysis, analysis of costs, or economic modeling. Studies were then excluded if they were cited in the White and Ernst review \[[@B4]\], or if they were case studies or case series of five or fewer subjects. Data analysis ------------- The following data were extracted from each of the included studies: full citation information (author(s), date, title, journal, etc), form of economic evaluation (stated or inferred), the therapies being compared and whether the CAM therapies were being used in addition to usual care (complementary) or instead of usual care (alternative), the perspective of analysis (stated or inferred), the study design, the sample size, and summary results. The studies were categorized as either full economic evaluations (defined as a comparison between two or more alternatives and considering both costs and consequences \[[@B10]\]) or partial economic evaluations (those studies that did not contain a comparison, or only addressed costs). Studies that estimated resource utilization were included as full economic evaluations even if resources were not valued. We captured data on quality of the full economic evaluations using two approaches. The first approach was to gather from each study the data needed to assess quality according to a 35-item checklist developed by the BMJEconomic Evaluation Working Party \[[@B30]\]. This checklist was developed to improve the quality of published economic evaluations, and was chosen because it is thorough, and entails an objective assessment of whether essential components of an economic evaluation are reported in the article. Therefore, the checklist is mainly a measure of reporting quality and not necessarily of study quality. We also report available results from several other general reviews of economic evaluations of conventional therapies that use this checklist for comparison. As the purpose of economic evaluations is to inform clinical practice and health policy decisions, the best evaluations are timely and use the best data available at the time \[[@B10]\]. On the other hand, an evaluation is only as good as the data upon which it is based. It has been suggested that the ideal situation for data collection is to collect economic data along side health outcomes in a randomized pragmatic trial \[[@B10]\]. Pragmatic trials offer a compromise between the goals of internal and external validity. To assess study quality, we went beyond White and Ernst\'s \[[@B4]\] criteria of randomization (to reduce bias by creating comparable groups) and prospective collection of economic outcome data (to ensure all costs are captured) to include two additional indicators of whether a pragmatic (effectiveness or \"real world\") rather than efficacy trial was conducted. The first is that the comparison group was usual care, and the second was that the study was not blinded and not mandatory -- ie, that physicians and patients could react realistically to the therapy \[[@B10]\]. These criteria relate to the external validity or generalizability of the study. Other indicators of a study\'s generalizability, such as the determination of whether study participants could be assumed to represent a normal case load, were not used as they required detailed knowledge as to the appropriateness of the inclusion and exclusion criteria for each condition studied -- a level of expertise not held by the study\'s authors. Based on the study quality criteria, we report summarized results of the exemplary studies -- ie, those meeting all four study quality criteria. If the health outcomes for one therapy are better than that of its alternative and the economic outcomes are better or equal (lower or equal costs), that therapy is said to dominate (be clearly better than) its alternative. This is also the case if both therapies have equal health outcomes and one has lower costs. In all other cases, the decision maker must elect whether the increase (loss) in health benefits is worth the increase (savings) in cost. Results ======= The database search rendered 1765 potential studies to screen. Application of inclusion and exclusion criteria reduced the list to 56 economic evaluations \[[@B23]-[@B25],[@B27],[@B31]-[@B82]\]. The therapies compared, study design employed, sample size used, and a summary of study results are provided for each study in an appendix \[see [Additional file 1](#S1){ref-type="supplementary-material"}\]. The list contains 39 full evaluations and 17 partial evaluations. The evaluations cover a range of CAM therapies applied to a variety of conditions (Table [3](#T3){ref-type="table"}). Some therapies, such as acupuncture, homeopathy, and manual therapy, were studied mainly as alternative therapies (ie, as substitutes or alternatives for conventional care). Other therapies, such as guided imagery, were studied as complementary therapies (ie, used in addition to conventional care). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Types of complementary and alternative medicine (CAM) therapies studied for various conditions (full/partial economic evaluations) ::: Acupuncture Homeopathy Manual therapy Spa therapy Mind-body therapy Hypnosis Botanical medicine Nutritional supplements Diet Biofeedback Hyperbaric oxygen therapy Miscellaneous\* TOTALS† ------------------------------------ ------------- ------------ ---------------- ------------- ------------------- ---------- -------------------- ------------------------- ------ ------------- --------------------------- ----------------- --------- Populations with mixed conditions‡ 3/2 2/1 0/1 0/1 10 Back, neck, and/or leg pain 1/0 5/0 1/0 1/0 8 Surgery 2/1 2/0 5 Cardiac patients 2/0 1/0 1/0 4 Rheumatic disorders 0/1 1/0 1/0 3 Epilepsy 0/3 3 General costs 0/1 0/2 3 Allergy 0/1 1/0 2 Cancer chemotherapy 2/0 2 Diabetic ulcers 2/0 2 Dyspepsia 1/0 1/0 2 EENT in children 1/1 2 Headache/migraine 2/0 2 Midwifery/obstetrics 1/0 0/1 2 Miscellaneous§ 1/0 2/0 1/0 1/0 1/2 2/0 10 TOTALS† 5 11 7 3 8 3 2 7 6 2 2 4 60 EENT = Eye, ear, nose, and throat conditions \* Miscellaneous CAM therapies include: multivitamins, shoe orthoses, electrodermal screening, and aromatherapy. † Some studies compared more than one CAM therapy. Therefore, totals exceed the number of studies found. ‡ Populations with mixed conditions include: patients with chronic disease, patients at one general practice (4 studies), long-term care workers, persons in Quebec health system, inner city children, and older adults (2 studies). §Miscellaneous conditions include: anxiety, Parkinson\'s, psoriasis, uterine fibroids, urinary tract infection, macular degeneration, severe burn, AIDS, obesity, and hypertension. ::: Reporting quality checklist --------------------------- Table [4](#T4){ref-type="table"} shows the results of the application of the BMJ35-item quality checklist \[[@B30]\] to the 39 full economic evaluations. For comparison, Table [4](#T4){ref-type="table"} also contains comparable results from systematic reviews in conventional medicine \[[@B6],[@B83],[@B84]\]. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Reporting quality of complementary and alternative medicine (CAM) economic evaluations and comparable results of similar reviews in conventional medicine ::: -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Items from the BMJChecklist \[30\] (Indented items apply only to a subset of studies) Review of CAM Studies N (%) Reviews of Conventional Medicine Studies N (%) ------------------------------------------------------------------------------------------- ----------------------------- ------------------------------------------------ Study design \(1) The research question is stated 39 (74) 43 (16)\* \(2) The economic importance of the research question is stated 39 (51) \(3) The perspective of the analysis is stated 39 (33) 228 (52)† \(4) The rationale for choosing the alternatives is stated 39 (69) \(5) The alternatives being compared are clearly described 39 (74) 228 (83)† \(6) The form of economic evaluation used is stated 39 (49) \(7) The choice of form of economic evaluation is justified 39 (3) 43 (7)\* Data collection \(8) The source(s) of effectiveness estimates are stated 38 (100) (9) Details of the effectiveness study are given 36 (94) or (10) Details of the review or meta-analysis are given 2 (50) \(11) Primary outcome measures are clearly stated 39 (95) (12) Methods to value health states are stated 4 (100) 228 (75)†\ 43 (79)‡ (13) Details of the subjects from which values were obtained are given 4 (25) 228 (76)†\ 43 (46)‡ (14) Productivity changes are reported separately 8 (88) (15) The relevance of productivity changes is discussed 8 (25) \(16) Quantities of resources are reported separately from unit costs 39 (67) 43 (19)‡ \(17) Methods for the estimation of quantities and unit costs are described 39 (67) \(18) Currency and year are recorded 39 (41) 228 (68)† \(19) Details of adjustments for inflation or currency conversion are given 39 (21) 43 (21)\* (20) Details of any model used are given 3 (100) (21) The choice of the model and its key parameters are justified 3 (100) Analysis and interpretation of results \(22) Time horizon of costs and benefits is stated 39 (100) (23) The discount rate is stated 4 (50) 228 (65)† (24) The choice of discount rate is justified 4 (25) 43 (16)\\ 34 (21)‡ (25) An explanation is given if costs and benefits not discounted 4 (50) 8 (12)‡ (26) Details of statistical tests and confidence intervals are given for stochastic data 38 (87) (27) The approach to sensitivity analysis is given 5 (100) 43 (2)\ (28) The choice of variables for sensitivity analysis is justified 5 (40) 39 (79)‡ (29) The ranges over which variables are varied are stated 5 (100) 228 (57)†\ 38 (66)‡ \(30) Relevant alternatives are compared 39 (36) 228 (57)† (31) Incremental analysis is reported 13 (54) 228 (46)† \(32) Major outcomes are presented disaggregated and aggregated 39 (85) \(33) The answer to the study question is given 39 (69) \(34) Conclusions follow from the data reported 39 (100) \(35) Conclusions are accompanied by the appropriate caveats 39 (67) 228 (84)† -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- \* Comparable estimates available from Jefferson et al, 1998 \[83\]. † Comparable estimates available from Neumann, 2004 \[6\], a systematic review of cost-utility analyses. ‡ Comparable estimates available from Gerard et al, 2000 \[84\], a systematic review of cost-utility analyses. ::: ### Study design These checklist items indicate whether essential components of the study design were reported. About half the studies stated the form of the economic evaluation, however, several were stated incorrectly and only one justified the form chosen. The bulk of the studies presented cost-effectiveness analyses (36 or 92%), five presented cost-utility analyses, and one was a cost-minimization study \[[@B24]\]. Only one-third of studies stated the perspective of the analysis, however, it could be determined from the costs included for all studies. Ten used a societal perspective, and the majority (33 or 85%) used some sort of institutional perspective (eg, health insurance company or hospital). Note that the totals by form and perspective add to more than 39. This is because individual studies can include analyses using more than one form of economic evaluation and can report costs from more than one perspective. ### Data collection These checklist items relate to the presence of information essential to the generalizability of study results. All studies that included health outcomes (ie, all except the one cost minimization study \[[@B24]\]) reported the source of their effectiveness estimates. In 36 of the 38 cases the source was a single study, often the economic evaluation itself. The two other studies were modeling studies \[[@B49],[@B78]\] where reviews were used as the source of effectiveness estimates. Items 12 and 13 are appropriate for cost-utility analyses (where health states are valued in terms of utility) and there were four such studies \[[@B27],[@B35],[@B51],[@B78]\], only one of which gave details on the subjects from whom the valuations were obtained \[[@B35]\]. Productivity changes (items 14 and 15) are appropriate for studies using the societal perspective. Eight studies included the costs of changes in productivity from improvement in back or leg pain \[[@B47],[@B82]\], neck pain \[[@B27]\], migraine \[[@B32]\], anxiety \[[@B44]\], ankylosing spondilitis \[[@B51]\], psoriasis \[[@B49]\], and children\'s rhinopharyngitis \[[@B41]\]. All but one \[[@B49]\] reported these amounts separate from total costs. However, few discussed the relevance to the study of productivity changes. About two-thirds of studies reported resource use quantities separate from unit costs, or described the methods used to estimate both quantities and unit costs. Whereas, almost all reported the currency used, only a minority (16 or 41%) reported the currency year. A smaller number reported the details of adjustments for inflation or currency conversion, but this was not often required in studies collecting and reporting data in the same year and currency. Models (one decision tree model \[[@B78]\] and two multiplicative-type or impact \[[@B49],[@B57]\] models) were used in three studies and in all cases the details of the model were given and justified. ### Analysis and interpretation of results All studies stated the time horizon for costs and benefits and most (35 or 90%) reported a time horizon of one year or less. Items 23 through 25 apply only to the four remaining studies with time horizons longer than one year. The discount rate is reported in two of these studies (one with a time horizon of 42 months \[[@B25]\] and the other that included a 12-year projection \[[@B78]\]), but only one justified the choice of discount rate \[[@B78]\]. Two studies gave an explanation for why they did not discount costs and benefits, however, neither needed to -- one had a one-year time horizon \[[@B35]\] and the other stated its time horizon as one course of chemotherapy \[[@B55]\]. Five studies performed sensitivity analyses \[[@B25],[@B27],[@B35],[@B51],[@B78]\]. In all cases the approach and the range of variables tested were stated, but the choice of variables to test was only justified in two cases \[[@B35],[@B51]\]. In about one-third of studies there was some comparison of study results to that of other studies. In most cases this was done as a simple statement noting that the results were either similar, or that they were dissimilar and that this might be because of differences in study design. Incremental cost-effectiveness or cost-utility ratios are usually only required when one therapy offers clearly better health outcomes than the other, but at a higher cost. In the 13 studies where this was the case over one-half reported incremental analyses. In most cases the major outcomes of the studies were shown disaggregated, and the study question was answered. We did require that a proper research question be stated (see the answers to item 1) for it to be answered. In all cases, we felt that the conclusions followed the data, but in about one-third of cases the conclusions were not presented with the appropriate caveats. For example, if a study did not explicitly discuss its limitations, it was not included as meeting the last item. Measures of study quality ------------------------- Twenty-seven studies (69%) gathered cost data prospectively and 21 (54%) used randomly assigned comparison groups. In 32 studies (82%) the physicians and patients were not blinded to the treatment received and participation was not mandatory (a worksite intervention \[[@B57]\]), and therapies were compared to usual care in 34 (87%) of studies. Fourteen studies \[[@B25],[@B27],[@B32],[@B34],[@B35],[@B50],[@B51],[@B53]-[@B55],[@B68],[@B74],[@B76],[@B82]\] met all study quality criteria, and a summary of their results is shown in Table [5](#T5){ref-type="table"}. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Summary of the results of complementary and alternative medicine (CAM) economic evaluations with exemplary study quality ::: ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- CAM Therapy Compared to Usual Care\* Patient Population Form of Economic Evaluation Health Effects of CAM Compared to Usual Care† Cost of CAM Compared to Usual Care† --------------------------------------------- -------------------------------------------------------------------------------- ------------------------------------------------------------------------- ----------------------------- ----------------------------------------------- ------------------------------------- Liguori et al, 2000 \[32\] Acupuncture Patients with migraine CEA Better Lower‡ Wonderling et al, 2004 \[35\] Acupuncture Patients with chronic headache CUA Better Higher‡ Paterson et al, 2003 \[34\] Acupuncture Patients with dyspepsia CEA Similar Similar Homeopathy CEA Similar Similar Korthals-de Bos et al, 2003 \[27\] Manual therapy Patients with neck pain CEA CUA Better\ Lower¶ Similar Brefel-Courbon et al, 2003 \[50\] Spa therapy Patients with Parkinson\'s disease CEA Similar Lower Van Tubergen et al, 2002 \[51\] Combined spa-exercise therapy Patients with ankylosing spondylitis CEA CUA Better\ Higher¶ Better Tusek et al, 1999 \[53\] Complementary guided imagery Cardiac surgery patients CEA Better Lower van Dixhoorn and Duivenvoorden, 1999 \[54\] Complementary relaxation therapy Patients with previous myocardial infarction CEA Better Lower Jacobsen et al, 2002 \[55\] Complementary professionally-administered stress management training Cancer patients undergoing chemotherapy CEA Similar Higher‡ Complementary self-administered stress management training CEA Better Lower‡ Franzosi et al, 2001 \[25\] Complementary omega-3 polyunsaturated fatty acids Patients with recent myocardial infarction CEA Better Higher Smedley et al, 2004 \[68\] Complementary preoperative and post operative oral nutritional supplementation Patients undergoing lower gastrointestinal tract surgery CEA Better Similar Norris et al, 2004 \[56\] Potassium-rich diet Postoperative cardiac patients CEA Similar Lower Ryan and Gevirtz, 2004 \[76\] Biofeedback-based psychophysiological treatment Patients with \"functional\" disorders (e.g., irritable bowel syndrome) CEA Better Lower Larsen et al, 2002 \[82\] Complementary custom-made biomechanical shoe orthoses Recent military conscripts CEA Better Higher ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Boldentries indicate that the CAM therapy was shown to be clearly superior to (dominate) usual care. CEA = cost-effectiveness analysis; CUA = cost-utility analysis \* The use of the term \"complementary\" in this column indicates CAM therapies used in addition to usual care. † If tests of statistical significance were performed, costs must be significantly higher or lower (and health effects significantly better or worse), or they were considered \"similar.\" ‡ This study used both a societal and an institutional perspective, and the results were in the same direction. ¶This study used a societal perspective only. All other studies used an institutional perspective only. ::: Discussion ========== The number of economic evaluations of CAM has increased in recent years, even if we only count full evaluations of alternative therapies. Study quality has also increased, and although reporting quality can use improvement, it is on the whole similar to that seen in economic evaluations of conventional medicine. Nevertheless, there are still too few good quality evaluations to draw many conclusions about the cost-effectiveness of specific CAM therapies for particular conditions. Potential reasons for paucity ----------------------------- A possible explanation for the paucity of studies is that there may be less of an incentive to perform economic evaluations of CAM. Consumers are already spending a large amount of their disposable income on CAM without formal proof of effectiveness or cost effectiveness. Economic evaluations are typically required for the incorporation of therapies under traditional financing mechanisms and for adjustment of coverage under these mechanisms. Therefore, the market for economic evaluation in CAM may be small due to reduced involvement of third-party payers in CAM financing. Some CAM practitioners do not see the need for economic evaluations. An interesting study by Kelner et al \[[@B85]\] asked chiropractors, homeopaths, and Reiki practitioners about the need to demonstrate the effectiveness, safety and cost effectiveness of their therapies. The chiropractors agreed that high quality economic evaluations are essential to their practice, but Reiki practitioners could see no reason for this research, and the homeopaths were divided on these issues. There may be good reason why some practitioners resist economic evaluation. If studies are performed that show economic benefit of CAM therapies, third party reimbursement may follow which could reduce practitioner autonomy. Coverage may also be restricted to the standardized forms of botanical medicines, nutritional supplements, or protocols used in the studies \[[@B86]\]. This could dramatically change how CAM is practiced by decreasing the use of multidimensional multicomponent interventions, by institutionalizing care into conventional health care systems, and by limiting the individualization of care. Relative quality of evaluations ------------------------------- The reporting quality was poor for certain items, but was comparable to the quality found by systematic reviews of economic evaluations in conventional medicine \[[@B6],[@B83],[@B84]\]. Although the BMJchecklist was mostly objective (ie, required the least amount of judgment compared to the other checklists available), a fair amount of interpretation was still required for many items. For example, in our review we interpreted Item 1 as whether the study stated either a specific research question or study objectives in terms of economic and health outcomes. Three-quarters of the full economic evaluations of CAM met this criterion. However, in Jefferson et al, 1998 \[[@B83]\], only 16% of the 43 economic evaluations of conventional medicine reviewed where identified as fulfilling Item 1. It is likely that Jefferson et al took a more restrictive interpretation of this quality criterion. Several studies have shown that at least some aspects of quality in economic evaluations improve over time \[[@B6],[@B87]\]. Our findings suggest a trend of quality improvement in these studies in CAM. We found that 69% (27 of 39) of the cohort of full economic evaluations collected cost data prospectively as compared to 45% (5 of 11) in White and Ernst\'s review. Similarly, we found that 54% (21 of 39) of our studies used randomization to create the comparison groups as compared to 45% (5 of 11) in White and Ernst\'s review. We found that 14 (36%) of full economic evaluations met all four study quality criteria and were identified as exemplars. However, the evidence from these criteria must be interpreted cautiously; meeting all study quality criteria does not guarantee an adequate study design. Some aspects of what makes a good pragmatic trial could not be judged by what was reported. For example, pragmatic trials enroll patients typical of normal caseload in typical settings with average physicians following them under routine conditions \[[@B10]\]. Judgments as to whether these criteria were met were not possible because of vague reporting. It is also not generally agreed across all health economists that a pragmatic trial, even a well-designed one, can fully represent the real world of health care. These economists advocate for the collection of cost data using an observational study design. A study may also be of \"poor\" quality because it applied the CAM therapy inappropriately. This can happen when a study is designed by researchers not familiar with a therapy. In response to this problem researchers and practitioners of several CAM therapies have begun development of standards for research and reporting. Reporting standards do not guarantee that the therapy was used appropriately, but they at least allow determination of what was done. One such set of reporting standards are the STRICTA recommendations for acupuncture \[[@B88]\]. Of the four full evaluations of acupuncture, two (one of which was included in Table [5](#T5){ref-type="table"}\[[@B32]\]) met STRICTA reporting standards. As these types of guidelines are not yet available for all CAM therapies, we did not assess whether CAM therapies were applied appropriately in the studies reviewed. Cost-effectiveness of CAM ------------------------- The exemplary studies summarized in Table [5](#T5){ref-type="table"} indicate that a number of CAM therapies may be considered cost-effective compared to usual care for a number of conditions: acupuncture for migraine, manual therapy for neck pain, spa therapy for Parkinson\'s, complementary guided imagery for cardiac surgery patients, complementary relaxation therapy for patients with previous myocardial infarction, complementary self-administered stress management for cancer patients undergoing chemotherapy, complementary pre- and post-operative oral nutritional supplementation for lower gastrointestinal tract surgery, potassium-rich diet (rather than potassium supplements) for postoperative cardiac patients, and biofeedback for patients with \"functional\" disorders such as irritable bowel syndrome. Acupuncture and homeopathy were both found to be equivalent in terms of effects and costs to usual care for dyspepsia. The attractiveness of the other CAM therapies shown in Table [5](#T5){ref-type="table"} depends on whether the increased health benefits are worth the additional cost, or whether other aspects of the therapy make them attractive, such as patient preference. Only one of the studies summarized in Table [5](#T5){ref-type="table"} reported results of a CAM therapy being dominated by (clearly inferior to) usual care. The use of professionally-administered stress management for cancer patients undergoing chemotherapy was shown to have higher costs, but no additional health benefits over usual care. It is important for CAM that this contradictory evidence is also known for best clinical practice and the efficient use of CAM resources. On the surface one might expect that therapies that substitute for usual care (alternative medicine) would be much more likely to be cost effective. In this sample of exemplar studies, of the nine study comparisons where CAM therapies were shown to be superior to usual care (better effects and lower costs, similar effects and lower costs, or better effects and similar costs), four were studies of complementary therapies. Therefore, there is evidence that even though complementary therapies are given in addition to usual care, they can improve clinical outcomes without increasing costs. Issues specific to the economic evaluation of CAM ------------------------------------------------- In many ways the economic evaluation of CAM therapies is similar to that of conventional medicine. However, there are a number of issues specific to CAM that must be considered. These issues can roughly be divided into three groups: those involved with the impact of economic evaluation on CAM in general, those involving the estimation of health outcomes (ie, issues involved with estimating the efficacy or effectiveness of CAM), and those specific to CAM\'s economic and humanistic outcomes. The first group of issues has already been addressed above under the potential reasons for paucity. The methodological challenges involved in determining the clinical effectiveness of CAM have been discussed at length in a number of papers. These include the appropriateness of population-based studies when individualized treatments are used and individualized outcomes are expected \[[@B89]-[@B91]\], reductionist focus on one therapy for one outcome when that therapy comes from a holistic healing system \[[@B92]-[@B94]\], the difficulties with blinding when no appropriate placebo is available \[[@B94],[@B95]\], and the requirement for randomization when most CAM users have strong preferences for their therapy of choice and will often either refuse to be randomized, or will bypass the randomization if it is not to their liking \[[@B94]\]. These challenges are relevant to economic evaluations since they are dependent on effectiveness studies for health outcomes. Also since humanistic and economic outcomes are ideally measured alongside health outcomes in the same trials \[[@B1],[@B4],[@B19],[@B20]\], the challenges above are also relevant to their measurement. However, there are several additional issues specific to CAM humanistic and economic outcome measurement which must be considered. First, although CAM therapies can be used to treat acute conditions, they are more commonly used to treat chronic disease, to prevent future disease (risk reduction), and to optimize health and well-being. Using CAM for those indications requires that long term studies be performed \[[@B96]\]. However, there are a number of challenges inherent in long term studies in addition to the increase in cost (eg, increased loss to follow-up through patient attrition) \[[@B97]\]. In our systematic review we found only two clinical trials that followed patients prospectively longer than one year: a five-year study of relaxation therapy for patients with a previous myocardial infarction \[[@B54]\], and a 3.5-year study of n-3 PUFA as secondary prevention for patients with previous myocardial infarction \[[@B25]\]. Economic evaluations in CAM must recognize that the process of healthcare itself can be effective for patients. Attributes of the process of using CAM that may have value include patient empowerment, the operationalization of patient preference for a particular type of intervention, the length and process of the consultation, and still having treatment options open when other medical approaches have failed \[[@B4],[@B98]\]. Therefore, economic evaluation of CAM needs to measure and include this value where appropriate. Optimizing health, maximizing wellness, and enhancing well-being are patient-centered outcomes -- ones that by definition require subjective measurement \[[@B99]\]. Economic evaluation of CAM must include appropriate measurement of these humanistic outcomes to account for the full value of CAM therapies. Our systematic review found five studies where humanistic outcomes were captured. The more well-known instruments used to measure health status in these studies included the SF-6D \[[@B35]\] and the EuroQoL (EQ-5D), and health status was translated into quality of life units using population-based preferences \[[@B27],[@B51]\]. Sensitivity of these instruments to the changes in quality of life is an important concern for the evaluation of CAM therapies. Although the use of the EuroQoL for manual therapy for neck pain \[[@B27]\] resulted in a statistically insignificant change in quality of life, two other studies demonstrated small, but statistically significant differences in quality of life using the SF-6D for acupuncture for chronic headache \[[@B35]\], and using the EuroQoL for spa therapy for ankylosing spondylitis \[[@B51]\]. Therefore, it is possible to measure a change in humanistic outcomes for CAM therapies with these instruments. The collection of economic outcome data is complicated by that fact that in the United States and other countries many CAM therapies are available over the counter and/or are often paid for out-of-pocket. The lack of administrative claims data on CAM therapies in countries where these costs are not covered or reimbursed means that cost studies require primary data collection (eg, patient self-report instruments) \[[@B100]\]. In their study on manual therapy for neck pain, Korthals-de Bos and colleagues used weekly cost diaries to obtain economic outcomes \[[@B27]\]. The second, related challenge is that many over-the-counter products, such as certain botanical medicines and nutritional supplements, are not standardized and of inconsistent quality. Standardization and quality will affect both the costs of the therapy and its outcomes. Finally, since there is often no provider \"gatekeeper\" controlling access to CAM therapies, monitoring of patient use can be complicated and labor intensive. Recommendations for future research ----------------------------------- Despite the challenges described for economic evaluations of CAM therapies, these studies ought to be done. Every planned trial of CAM therapies should at least consider the feasibility of including an evaluation of economic impacts. Observational studies should also include these data, and as information accumulates regarding economic impacts, these costs and cost savings can be estimated more accurately. Although in the ideal every cost category shown in Table [2](#T2){ref-type="table"} should be measured and outcomes should include a measure of quality-adjusted life-years, the estimation of direct medical costs and savings associated with the therapy (eg, practitioner fees, lab fees, and the cost of herbs or other supplements prescribed) will be fairly straightforward for most studies, and the planned primary outcome of the study can serve as the measure of effects to determine cost effectiveness. Even if the clinical outcomes of a CAM therapy are similar or slightly less beneficial than those of usual care, a lower cost of care can still make these therapies attractive to decision makers. However, if no cost data are available, even highly effective therapies can be easily overlooked. Limitations ----------- The limitations of this study are similar to those of the other reviews. First, the reader was not blinded to journals and article authors, which may have influenced results. Second, our measures of study quality depend on the information reported in an article, and no attempt was made to judge the merits of clinical or modeling assumptions made in the analyses. Third, only one reader read all the papers and extracted all the data. This may have lead to inaccurate reporting of results, and/or a biased interpretation of study quality. To maximize accuracy, data extraction was performed at least twice for each paper with several months break between extractions. Also, the approach and assumptions used to determine study quality were discussed at length with the other authors. These discussions led to a homogeneous approach being taken to both the application of the reporting quality criteria and the definition as to what constitutes an economic evaluation. Conclusion ========== As health care costs continue to rise, decision makers must allocate their increasingly scarce resources toward therapies which offer the most benefit per unit of cost. Economic evaluations inform evidence-based clinical practice and health policy. To be considered by these decision makers, CAM therapies and their outcomes must be known and compared to conventional approaches. However, CAM practitioners must themselves decide whether the cost of performing these studies is worth the potential impacts to their profession of being considered in managed care. Nevertheless, these evaluations will be done and they will be better done with practitioner involvement. Whereas the number and quality of these studies has increased in recent years and more CAM therapies have been shown to be good value, there are still not enough studies to measure the cost effectiveness of the majority of CAM. If CAM providers wish to increase the provision of therapies to improve population health, they must report the potential outcomes of CAM therapies widely and well. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= PH had the main responsibility for the manuscript and for bringing together the concepts of CAM and economics. PH also read and evaluated the quality of all papers included in the review. BC ensured that the health economic concepts were presented appropriately. PH and OC conceived the idea for the paper. OC contributed clinical and methodological insights. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1472-6882/5/11/prepub> Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Descriptions of included studies ordered by complementary and alternative medicine (CAM) modality, form of economic evaluation, and publication date. The appendix contains a table summarizing each of the 56 economic evaluations found in the systematic review. For each evaluation the following are reported: therapies compared, study population, study design and sample size, whether it was a full or partial economic evaluation, form of the evaluation, perspective, and summary results. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ This project was supported by grant \#T32 AT01287-03 from the National Center for Complementary and Alternative Medicine.	PubMed Central	2024-06-05T03:55:59.893567	2005-6-2	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182346/", "journal": "BMC Complement Altern Med. 2005 Jun 2; 5:11", "authors": [ { "first": "Patricia M", "last": "Herman" }, { "first": "Benjamin M", "last": "Craig" }, { "first": "Opher", "last": "Caspi" } ] }
PMC1182347	Background ========== Although massage is one of the oldest healthcare practices in the world, with references to it found in ancient Chinese medical texts as well as in the writings of Hippocrates, medical doctors in the US have not practiced therapeutic massage for nearly 100 years \[[@B1]\]. In the 1930\'s and 1940\'s, massage fell out of favor with nurses and physical therapists as well. However, since the 1970\'s, interest in massage therapy has burgeoned and it is now one of the most popular complementary and alternative medical (CAM) modalities. In the US, Eisenberg, et al. \[[@B2]\] found 11% of randomly surveyed Americans had used massage for treating common medical conditions in 1997, with 62% of these receiving massage from a trained massage therapist. They found that the number of visits to massage therapists exceeded that to all other CAM providers except chiropractors, with trained massage therapists providing an estimated 114 million massage sessions to Americans in 1997. Eighty percent of randomly surveyed US adults with health insurance claimed they would be \"likely\" to use massage, making it the most popular of the 11 therapies included in the survey \[[@B3]\]. Palinkas \[[@B4]\] reported that massage was the third most commonly used type of CAM among primary care patients, with 17.2% of CAM users reporting use of massage within the last year for the same reason they were seeking primary care. Despite this growth in the popularity of massage, little is known about the practices of licensed massage therapists. We included massage therapists in our study of random samples of licensed CAM practitioners and their practices \[[@B5],[@B6]\]. In this report, we present new information about the demographic and training characteristics of licensed massage therapists, the reasons patients seek their care, the assessment process massage therapists use during visits, and the treatments and self-care recommendations they provide. We have included information about massage efficacy and safety and communication between massage therapists and physicians in the Discussion section to assist biomedical healthcare providers in placing our findings in the broader context of patient care. Methods ======= Original study -------------- The data presented in this paper were collected as part of a larger study of four licensed CAM professions, including massage therapy. The methods are described in detail elsewhere \[[@B5],[@B6]\] and summarized here. Our goal was to obtain data on 20 consecutive visits to 50 randomly selected massage therapists in one Northeastern state (Connecticut) and one Western state (Washington) who gave at least 10 massage treatments per week. Massage therapists were randomly sampled from state licensure listings in Washington (1998) and Connecticut (1999). In both states, licensing requirements for massage therapists including having 500 hours of education and a passing score on the national examination. We excluded providers without identifiable telephone numbers and those not currently practicing. The proportion of ineligible practitioners was 47% in Connecticut and 33% in Washington. About 84% of ineligible Connecticut massage therapists lacked identifiable phone numbers, while in Washington ineligible therapists were about equally divided between those who were not practicing and those who lacked identifiable phone numbers. All participating massage therapists were interviewed about their demographic, training, and practice characteristics. Those with at least 10 visits in a typical week were then invited to participate in visit-based data collection. A sample of those seeing 5 to 9 visits per week were also invited to collect data on patient visits. Massage therapists with less than 5 visits per week were not asked to collect visit data and provided about 2% of all massage visits \[[@B6]\]. We obtained approval from the Group Health Cooperative, University of Washington, and Beth Israel Deaconess Medical Center Institutional Review Boards. Visit data were collected between May and September in 1998 in Washington and between June 1999 and March 2000 in Connecticut. Massage therapists were given visit forms marked with unique identification codes and were asked to record data on 20 consecutive visits (even if the same patient was seen more than once). Practitioners were randomly assigned weekdays to begin data collection. Visit form ---------- The one-page visit form was modeled after those used in the National Ambulatory Medical Care Survey (NAMCS) \[[@B7]\] and a copy of the visit form is found in [Additional File 1](#S1){ref-type="supplementary-material"}. Whenever possible, questions were worded identically to those in the NAMCS (e.g., demographic characteristics, smoking status, reason for visit, referral source, source of payment, visit duration, visit disposition). New questions asked if the patient was receiving care from a conventional medical provider for the primary problem and if the massage therapist had communicated about this problem with a conventional provider who also provided care for the patient\'s main problem. We also designed special questions to capture information about massage treatments, including information on use of specific assessment techniques, massage techniques, and lifestyle recommendations. We asked practitioners to record up to five \"complaints, symptoms, or other reasons for this visit\" using the patients own words, listing the most important complaint or reason first. These data were classified using the NAMCS Reason for Visit Classification System, which distinguishes among symptoms, diseases, diagnostic/screening/preventive interventions, treatments, and injuries \[[@B7]\]. Individual reasons for visit were then clustered into larger categories that correspond to International Classification of Diseases, Ninth Edition(ICD9) chapters. No information was collected on adverse experiences as part of this study. Analysis -------- In the massage therapist analyses, Chi-square and Fisher Exact tests were used to compare proportions, and Kruskal Wallis tests were used to compare medians. Even though standard errors are not presented, they are always within 5 percentage points of the estimate. Analyses were performed using SAS version 8 (SAS Institute, Cary, NC). In the visit analyses, each visit in the sample was weighted by the inverse of its sampling probability, which reflected both the chance that the particular provider participated and the estimated proportion of that provider\'s annual visits included in the study. Consequently, our results represent estimates of all visits made to massage therapists in each state, except for the 2% of visits made to providers with fewer than 5 visits per week or visits to therapists who were not licensed. Because of the two-stage sampling design, we used SUDAAN software (Research Triangle Institute, Research Triangle, NC) to calculate standard errors and confidence intervals using Taylor series linearization. Because of the large sample sizes (965 and 1040 visits) the weighted percentages presented in the tables have small standard errors, generally between 0.5 and 2.5 percentage points and rarely exceeding 3 percentage points. As a result, moderate to large differences between the states are also statistically significant. Therefore, the standard errors are not included in the tables. Results ======= Participation rates ------------------- Participation rates for the massage therapist interview were 86% (114 of 133) in Connecticut and 84% in Washington (112 of 134). Of the massage therapists who saw enough clients per week to be eligible to collect visit data, 66% in Connecticut (61 of 93) and 70% in Washington (65 of 93) complied. Data were collected on 965 visits in Connecticut and 1040 visits in Washington. Characteristics of the massage therapists ----------------------------------------- In both states, massage therapists were typically white, female and had a median age of 42 years (Table [1](#T1){ref-type="table"}). Virtually all of them received their basic training in the US, with most having trained in the state where they were currently practicing. A small fraction had no formal training. In both states, massage therapists reported training a median of about 600 hours. Massage therapists reported a median of 4 to 5 years in practice, with only 18% in Connecticut and 13% in Washington reporting more than 10 years. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Demographic and training characteristics of massage therapists ::: State ------------------------------------------------- ----------- ----------- -------- Demographic Characteristics Women 85% 85% White 95% 95% Hispanic 4% 4% Median Age 41.5 yrs. 41.5 yrs. Basic Training Formal Schooling 93% 94% US -- other states 12% 8% US -- same state 81% 85% Foreign 1% 1% No Formal Schooling 6% 6% Median Years in Practice 5 yrs. 4 yrs. Post-graduate Training Any 79% 52% \\\* Craniosacral 14% 12% Neuromuscular 10% 10% Reflexology 10% 6% Reiki 13% 6% Polarity 5% 5% Lymph Drainage 3% 5% Meridian -- based (Shiatsu, Tuina, acupressure) 22% 10% \* Myofascial Release 14% 3% \\ Pregnancy Massage 6% 1% \* p \< 0.05; \\ p \< 0.01; \\\* p \< 0.001 ::: Most massage therapists (82% in Connecticut and 89% in Washington) reported additional hours of training after graduation, receiving a median of 60 hours. Nearly 80% of the massage therapists in Connecticut and about half in Washington reported \"specialty or advanced training\" (i.e., continuing education), with 43% and 31%, respectively, reporting multiple types of such training. Continuing education was extremely heterogeneous, with practitioners noting 56 different types of training in Connecticut and 37 types in Washington. However, only 4 types of training were received by more than 10% of practitioners in Connecticut (meridian -based therapies, craniosacral, myofascial release and Reiki) and only one type of training was received by more than 10% of practitioners in Washington (craniosacral therapy) (Table [1](#T1){ref-type="table"}). Ten percent of massage therapists in Connecticut and 8% of those in Washington held other healthcare profession licenses. All but one of those (acupuncture) were in biomedical areas, most commonly nursing. Connecticut massage therapists reported a median of 10 patient visits per week and 12 hours of direct patient care per week, compared with 15 patient visits per week and 17 hours of direct patient care, for massage therapists in Washington (p \< 0.02 for hours of direct patient care). Reasons for visits to massage therapists ---------------------------------------- Visits to massage therapists were for a limited number of conditions. About 60% of visits were for musculoskeletal symptoms, particularly back, neck, and shoulder symptoms (Table [2](#T2){ref-type="table"}). Visits for \"wellness\" (i.e., relaxation) accounted for another 20% of visits and mental health concerns, largely anxiety and depression, for another 6 to 9% of visits. Virtually all other visits were for general body symptoms (mostly generalized pain) or \"nervous system\" symptoms (most commonly headache). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Most common reasons for visits to massage therapists licensed in Connecticut (1999) and Washington (1998) by broad and specific categorization ::: Connecticut Washington --------------------------------------------- ------------------------ ------------ --------------------------------------------- ------------------------ ------------ (N = 965 visits) (N = 1040 visits) Broad Categories\* \% with Primary Reason Broad Categories\* \% with Primary Reason 1\. Musculoskeletal Symptoms 59.2 1\. Musculoskeletal Symptoms 63.0 2\. Wellness\\ 19.5 2\. Wellness\\ 18.7 3\. Psychological and Mental Health Symtoms 8.8 3\. Psychological and Mental Health Symtoms 5.7 4\. General Symptoms 4.5 4\. Nervous System Symptoms 4.9 5\. Nervous System Symptoms 3.7 5\. General Symptoms 3.7 Top 5 Categories 95.7 Top 5 Categories 96.0 \% with \% with Specific Reasons Primary Reason Any Reason Specific Reasons Primary Reason Any Reason 1\. Back Symptoms 20.4 34.4 1\. Back Symptoms 20.2 39.8 2\. Massage Wellness 19.5 25.8 2\. Neck Symptoms 20.0 38.5 3\. Neck Symptoms 13.0 24.1 3\. Massage Wellness 18.7 26.5 4\. Shoulder Symptoms 8.4 23.1 4\. Shoulder Symptoms 7.4 26.6 5\. Anxiety or Depression 8.8 17.4 5\. Anxiety or Depression 5.2 12.3 6\. Leg Symptoms 5.0 10.0 6\. Headache 3.7 8.4 7\. Unspecified Muscle Symptoms 4.0 6.3 7\. Leg Symptoms 2.6 6.3 8\. Generalized Pain 3.1 4.5 8\. Generalized Pain 2.1 3.5 9\. Headache 1.6 5.2 9\. Hip Symptoms 1.9 6.7 10\. Unspecified Joint Symptoms 1.4 2.2 10\. Arm Symptoms 1.8 5.6 Top 10 reasons 85.2 Top 10 reasons 83.6 \* Broad Categories of Primary Reason for Visit Codes correspond to ICD chapters \\ Wellness was not originally part of the NAMCS Reason for Visit Classification. Most of these visits are for relaxation. ::: Most visits were for chronic problems, either problems that were ongoing (41% in Connecticut and 32% in Washington) or for flare-ups of chronic problems (12% in Connecticut and 15% in Washington). About a quarter to a third of all visits were for non-illness care (32% in Connecticut and 27% in Washington) and the remainder of visits were for acute problems (15% in Connecticut and 17% in Washington). Interaction with other healthcare providers and insurance --------------------------------------------------------- Most massage visits resulted from self-referrals (64% or 75%) but 4% in Connecticut and 11% in Washington resulted from referrals by medical or osteopathic physicians (virtually all for musculoskeletal symptoms). Although massage therapists discussed the care of the patient with another provider in 22% of visits in Connecticut and 30% in Washington, that provider was a medical or osteopathic physician less than one-third of the time. The most frequent consultations were with chiropractors. Massage therapists indicated that medical or osteopathic physicians were treating their patients for the same condition for 24% (Connecticut) or 32% (Washington) of visits. Massage therapists noted that they had discussed their patients\' care with the physicians of 29% (Connecticut) or 49% (Washington) of their physician-referred patients compared with only 12 -- 14% of their other physician-managed patients. Two percent of visits in both states ended with a referral to a medical or osteopathic physician. Insurance covered only 8% of visits in Connecticut and 26% of visits in Washington, and almost all the remainder were paid for by the patient. Care during visits to massage therapists ---------------------------------------- Massage therapists performed assessments in about two-thirds to three-quarters of the visits (Table [3](#T3){ref-type="table"}). The most common methods were tissue assessment via palpation, range of motion, and postural assessment. Multiple assessments were used in 38% (Connecticut) or 48% (Washington) of visits. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Diagnostic assessments performed by massage therapists licensed in Connecticut (1999) and Washington (1998) ::: Connecticut Washington -------------------------------------------- ----------------- ------------ Diagnostic Assessment Percent Using At least one 67.2 74.0 Applied Kinesiology 2.0 5.8 Postural Assessment 19.8 30.7 Range of Motion 34.9 46.0 Tissue Assessment 56.3 60.8 Other (e.g., acupressure point assessment) 7.1 2.7 ::: Virtually all visits included a massage that emphasized at least two techniques (Table [4](#T4){ref-type="table"}). The most commonly emphasized techniques were Swedish massage, deep tissue, and trigger point/pressure point techniques. Massage therapists in both Connecticut and Washington emphasized five other techniques in between 14% and 25% of visits: energy work, hot/cold therapy, movement re-education, craniosacral, and reflexology. Massage therapists in Connecticut were more likely to emphasize Oriental bodywork (i.e., meridian based techniques such as shiatsu) while those in Washington were more likely to emphasize neuromuscular therapy. Definitions of some of the most commonly emphasized techniques are provided in [Additional File 2](#S2){ref-type="supplementary-material"}. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Massage techniques emphasized during visits to massage therapists licensed in Connecticut (1999) and Washington (1998) ::: Connecticut Washington ------------------------------ ----------------- ------------ Techniques Emphasized Percent Using Any 99.4 99.9 Craniosacral 15.3 15.1 Deep Tissue 62.8 65.3 Emotional Bodywork 5.7 4.0 Energy Work 24.9 17.2 Guided Imagery 5.3 4.7 Hot/Cold Therapy 19.9 24.2 Manual Lymph Drainage 3.8 6.3 Movement Re-education 19.2 24.2 Neuromuscular Therapy 5.8 20.5 Oriental Bodywork 16.6 8.6 Pregnancy Massage 1.4 0.7 Reflexology 15.0 15.4 Somatherapy 1.2 5.0 Swedish Techniques 80.6 76.8 Trager 6.7 14.1 Trigger Point/Pressure Point 51.5 45.6 Other (e.g., Esalen, Thai) 7.1 4.2 Two or more techniques 86.7 92.5 ::: More than 80% of visits included self-care recommendations (Table [5](#T5){ref-type="table"}), with 50% (Connecticut) or 64% (Washington) of visits including multiple recommendations. Increasing water intake, movement (especially active movement), body awareness, and breathwork were the most common recommendations. Visits lasted a median of 60 minutes. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Self-care recommendations given by massage therapists licensed in Connecticut (1999) and Washington (1998) ::: Connecticut Washington --------------------------------------- ----------------- ------------ Self-Care Recommendations Percent Using Any 81.1 84.6 Body Awareness 37.2 37.7 Breathwork 28.4 25.2 Hot/Cold Therapy 29.0 33.2 Movement -- any 39.2 44.6 Movement -- active 26.6 35.1 Movement -- passive 17.3 13.5 Movement -- resisted 7.2 7.8 Visualization 8.3 8.7 Water Intake, Increase 48.4 56.1 Other (e.g., self-massage, relaxation 5.6 3.4 ::: Discussion ========== To our knowledge, this is the first study that describes the demographic and training characteristics of US massage therapists and uses systematically collected visit data to describe their treatment patterns. Strengths of the study are the collection of data from licensed massage therapists practicing in geographically separated parts of the country where CAM use is relatively common, random sampling of providers from state licensing lists, relatively high response rates, and large sample sizes. The main limitation is that we collected data from only two states, which may not be representative of massage practice in other states. However, licensure requirements in Connecticut and Washington are similar to those in most other states with licensure requirements. As of December, 2004, 33 states and the District of Columbia had passed legislation regulating massage practice. Of those, 21 require exactly 500 hours of training for licensure and 12 require between 570 and 1000 hours \[[@B8]\]. Licensure in both Connecticut and Washington requires 500 hours of training plus a passing score on the national certification exam administered by the National Certification Board for Therapeutic Massage and Bodywork (NCBTMB). The latter is required for licensure in 24 states and is an option for licensure in another 5 states. In some states, including Massachusetts and California, massage regulations vary within the state (i.e., between townships, cities or counties). By contrast, the two provinces in Canada with regulatory requirements mandate that massage therapists receive 2500 hours (Ontario) or 3300 hours (British Columbia) of training. Characteristics of the massage therapists ----------------------------------------- Our study describes an eclectic group of health professionals. Most massage therapists have taken continuing education training that includes both Western-oriented treatment techniques (e.g., neuromuscular therapy, myofascial release), and non-Western oriented treatment techniques (e.g., Reiki, meridian-based massage). Our finding that most massage therapists are white females with a median age around 40 is consistent with the findings of the only other published study of the characteristics of massage therapists, which surveyed 82 massage practices in the Boston area \[[@B9]\]. However, that study reported that the median length of practice was 7 years (compared to our 4 to 5 years), that providers received a median of 1000 hours of clinical training (compared to our 600 hours), and that practitioners saw a median of 20 patients per week (compared to our 10 to 15 visits per week). The other study used the telephone book in a single urban area to recruit massage therapists whereas we used state -- wide licensing lists. Their restriction to an urban area, their recruitment methods and their lower response rate may have biased their sample toward busier practitioners. Why patients visit massage therapists and evidence for efficacy --------------------------------------------------------------- The majority of visits to massage therapists focused on musculoskeletal conditions, possibly reflecting the extensive use of massage by physical therapists for rehabilitation during the first half of the 20^th^century \[[@B10]\]. These are conditions for which Western medical care is often of limited value, which may explain why back and neck pain are the most common reasons why patients seek CAM care in general \[[@B2]\]. While massage as a relaxation technique has received abundant attention in the popular culture, we found that less than one-third of all visits to licensed massage therapists focused on non-illness care. CAM is also commonly used for self-defined anxiety and depression \[[@B2],[@B11]\]. Among such a group of respondents to a national survey, 5% and 2% of respondents said that they used massage therapy to treat these conditions, respectively \[[@B11]\]. Since massage therapists do not make diagnoses, no information is available on whether patients\' visiting for anxiety and depression in our study actually had these disorders diagnosed by physicians. We could find no other published studies presenting data on patients\' reasons for visits to massage therapists from a large population-based sample of visits, so we do not know how comparable these results are. A survey of a representative sample of US adults reported that massage therapy was one of the most common CAM therapies used for back problems, neck problems and fatigue \[[@B2]\]. While fatigue was not a commonly listed reason for visiting massage therapists in our study, some patients who received wellness care or care for anxiety or depression could conceivably have had fatigue as a symptom. The use of massage for treating medical conditions has grown substantially since 1990 \[[@B2]\]. Although massage is one of the most popular forms of CAM care and has been found to have intriguing physiological effects (reviewed by Field \[[@B12]\]), few studies with moderate to large sample sizes have been conducted to evaluate its clinical effectiveness, even for most musculoskeletal conditions, conditions for which massage is frequently sought and for which conventional medicine has few good treatments. Three recent studies, including two that were well designed and had reasonable sample sizes, evaluated therapeutic massage as a treatment for subacute or chronic back pain and all three found positive results \[[@B13]\]. In addition, several studies of acupressure for back pain have also found positive results \[[@B14],[@B15]\]. A recent Cochrane review of massage for back pain \[[@B16]\] concluded that \"massage might be beneficial for patients with subacute and chronic non-specific back pain, especially when combined with exercises and education. More studies are needed to confirm these conclusions\". While even fewer studies of massage have been conducted for other musculoskeletal pain conditions, there are small studies suggesting that massage may have benefits for patients with fibromyalgia \[[@B17]\], shoulder pain \[[@B18]\] and diffuse chronic pain \[[@B19]\], while Irnich \[[@B20]\] did not find massage effective for neck pain. Most of those studies lacked follow-up after the treatments had stopped, but Hasson found that the benefits of massage did not persist three months after the last treatment. A recent meta-analysis of randomized trials of massage for various conditions found that massage had its greatest short-term benefits in reducing trait anxiety and depression, but no studies have evaluated these effects after the end of the treatment period \[[@B21]\]. A systematic review of massage for symptom relief in cancer patients found preliminary evidence that massage had short term benefits on psychological well-being and possibly anxiety \[[@B22]\], but called for additional studies to confirm and extend these findings. The modest evidence base for massage therapy\'s clinically important effects provides physicians with little information for advising patients about its effectiveness for conditions other than subacute or chronic back pain. However, given the safety profile and preliminary evidence of effectiveness for back pain, physicians should feel comfortable recommending massage for selected patients with musculoskeletal conditions and, possibly, for mild stress-related anxiety. Care during visits to massage therapists ---------------------------------------- Massage therapists in Washington were more likely than those in Connecticut to use postural assessment and range of motion as assessments tools. Such differences likely reflect differences in training. In general, these differences in assessment were not associated with differences in the massage techniques emphasized by practitioners. Swedish, deep tissue, and trigger (pressure) point were by far the most popular techniques in both states. In their survey of massage therapists in Boston, Lee and Kemper \[[@B9]\] found similar results: 90% of practitioners reported using Swedish techniques and more than half reported using trigger point massage, sports massage, myofascial release, and aromatherapy. A substantial minority of visits included techniques with a non-Western origin, such as some forms of energy work (e.g., Reiki) and meridian-based massage. In addition, this study as well as a previous study \[[@B23]\], found that massage therapists often emphasize self-care (e.g., drinking more water, movement, body awareness). Recommendations often include increasing the patients\' awareness of how they are using their bodies coupled with exercises designed to enhance movement and posture, based on the assumption that many musculoskeletal conditions result from poor use of the body. While these recommendations have not been scientifically validated, they are likely to be safe and may enhance the patient\'s sense of well-being. Safety of massage ----------------- In a review of the safety of massage therapy, Ernst \[[@B24]\] found 16 case reports and 4 case series in the biomedical literature over a 6 year period describing adverse effects associated with various forms of massage. However, only 3 reports (including 7 cases) described adverse effects that were probably attributable to treatments by massage therapists practicing Western forms of massage. These included the displacement of a ureteral stent, a hepatic hematoma after deep tissue massage \[[@B25]\] and the deterioration in hearing among patients who received neck massage. Ernst found three additional reports of adverse events associated with shiatsu, the most serious of which was retinal artery embolism with partial loss of vision after application of shiatsu to the upper neck. Although the rate of adverse effects over this period of time is unknown, in the US alone an estimated 113 million visits were made to massage therapists in 1997 \[[@B2]\], suggesting that serious adverse experiences due to massage are extremely rare. Despite these scattered reports of adverse experiences, common forms of massage (e.g., Swedish, deep tissue, and neuromuscular) are considered very low risk, especially when massage is tailored appropriately to the individual (e.g., possible pressure or anatomic site restrictions), as massage therapists are commonly trained to do \[[@B10]\]. While it is still generally assumed that patients with deep vein thrombosis should not receive massage to the lower extremities, many previous contraindications, such as proscribing massage to patients with metastatic cancer, are no longer considered warranted. Massage therapists are trained not to massage anatomic sites containing localized conditions such as skin injuries or burns. Communication between massage therapists and physicians ------------------------------------------------------- Massage therapy is an increasingly popular form of care used by patients who are often also being treated by a physician for the same condition. Nevertheless, we found that massage therapists and physicians rarely communicated with each other. Possible barriers to communication include our observation that most patients who see both a physician and a massage therapist for a particular condition were not referred to massage by the physician. Furthermore many massage therapists are not trained in charting language familiar to physicians, nor are they permitted to make \"diagnoses\". In addition, referring patients to massage therapists has not been part of the training of physicians. Finally, we suspect that most massage therapists, who are typically part-time solo practitioners, lack office staff and record systems to assist with administrative tasks, including routine (and written) communication with other care providers. We believe that patients may benefit from increased communication between their physicians and massage therapists. Physicians can foster improved communication by asking patients about the care they are receiving from a massage therapist and learning about the treatment plan. Some patients will want to try massage therapy only after consultation with their physician. In these circumstances, physicians can use the framework recommended by Eisenberg \[[@B26]\] to guide patients through the process of selecting a well-trained, therapeutically-oriented massage therapist, jointly negotiating the treatment plan, and monitoring the effects of the treatment over time. Conclusion ========== While substantial barriers to the full integration of massage therapy into the healthcare system remain (e.g., variability between states in licensure and practice regulations, lack of widespread insurance reimbursement, lack of solid studies on efficacy for many frequently-treated conditions, ambivalence on the part of massage therapists as to the advisability of mainstreaming)\[[@B27]\], the information provided in this report should be informative to physicians and other healthcare providers interested in advising their patients about massage therapy. Competing Interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= KJS participated in the design of the overall project and the data analyses and drafted this manuscript. DCC was the PI on one of the grants funding the study, designed and directed the data collection and analysis of the overall project. JK helped design the data collection instruments. JE participated in the design of the overall project, directed the data collection, quality control, and participated in the analyses for this paper. AH directed the data collection for Connecticut. RD participated in the design of the overall project and data collection procedures and helped to obtain funding. DME was the PI on one of the grants funding the study and participated in the design of the overall project and data collection procedures. All authors read and approved the manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1472-6882/5/13/prepub> Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Massage Care Survey. The visit form used for each of the massage therapy visits ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Glossary of Selected Massage Techniques. Definitions of selected massage techniques ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ This project was supported by grants from the Group Health Foundation, Grants \#HS09565 and \#HS08194 from the Agency for Health Care Policy and Research and Grant \#AR43441-04S1 from the National Institutes of Health. In-kind support was provided by the Centers for Disease Control and Prevention. The contents of this manuscript are solely the responsibility of the authors and do not necessarily represent the official views of the National Center for Complementary and Alternative Medicine, National Institutes of Health. We thank the 33 members of the original massage therapy study team for data collection and Kristin Delaney for help with data analysis.	PubMed Central	2024-06-05T03:55:59.901049	2005-6-14	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182347/", "journal": "BMC Complement Altern Med. 2005 Jun 14; 5:13", "authors": [ { "first": "Karen J", "last": "Sherman" }, { "first": "Daniel C", "last": "Cherkin" }, { "first": "Janet", "last": "Kahn" }, { "first": "Janet", "last": "Erro" }, { "first": "Andrea", "last": "Hrbek" }, { "first": "Richard A", "last": "Deyo" }, { "first": "David M", "last": "Eisenberg" } ] }
PMC1182348	Background ========== The use of complementary medicines in Australia has become commonplace. In 2002 it was estimated that 52% of the Australian population had used at least one non-physician-prescribed complementary medicine in the previous year\[[@B1]\]. These medicines are widely available from many sources including health food stores, supermarkets, direct marketing, natural therapy clinics and pharmacies. Although self-prescription of oral complementary medicines is common, research has also shown an increase in visits to alternative practitioners, in particular naturopaths and herbal therapists\[[@B1]\]. Both of these types of practitioners commonly prescribe oral complementary medicines. The Expert Committee on Complementary Medicines in the Health System in Australia was convened in 2003 and asked to consider the regulatory, health system and industry structures necessary to ensure that the objectives of the National Medicines Policy (NMP) and in particular that of the National Strategy for Quality Use of Medicines (QUM), were met in relation to complementary medicines. The report from the Expert Committee raised a number of concerns surrounding the information available to healthcare providers and consumers regarding complementary medicines, and one of the recommendations of the committee was the commissioning of a study to determine the needs of healthcare professionals and consumers on complementary medicines and the options available for conveying this information\[[@B2]\]. Previous research has surveyed the naturopathic and Western herbalists workforce\[[@B3]\], providing an overview of the types of therapies and diagnostic procedures used by these therapists, the form and labelling of complementary medicine preparations used in the practice, cost of treatment to the client, practitioner income, and adverse events experienced within the practice. Whilst this survey found that nutritional medicines were generally provided as over-the-counter commercial formulations, the methods for preparing and dispensing herbal and homoeopathic medicines varied, with one half of practitioners mixing or combining 90% or more of their herbal medicines in their own clinics. The practitioners in this survey also reported a substantial number of adverse events associated with these medicines. However this research did not address the practice behaviours of these practitioners in regard to the provision of information to clients on the complementary medicines which were prescribed at the consultation or information on complementary medicines accessed by the practitioner themselves. Research in the USA has also examined practice patterns of naturopathic physicians\[[@B4]\]. The demographics of these practitioners were found to be remarkably similar to those of naturopaths in Australia. American naturopathic physicians commonly prescribed oral complementary medicines, with the most common of these being botanical medicines, vitamins, minerals and homoeopathic preparations. Again, this study did not address questions relating to the provision of or access to information on oral complementary medicines by naturopathic practitioners. In Australia the profession of naturopathy is essentially unregulated. A level of self-regulation is exerted by the professional associations in the form of minimum qualification levels and monitoring of annual requirements such as Continuing Professional Education, however there are a large number of these associations and entry requirements vary between the groups. Several Australian States are currently exploring options for the regulation of naturopaths and other complementary medicine practitioners \[[@B5]\]. The current agreed minimum qualification level enforced by the main professional associations is Advanced Diploma in Naturopathy. This qualification is contained within the Health Training Package HLT02 and represents a consistent base level of training within the industry. The Advanced Diploma courses are taught at privately-owned Vocational Education and Training Colleges which are registered through the various State Education Authorities. Some of these colleges are also registered to deliver degree courses, and several universities also deliver courses in naturopathy. The purpose of this study was to examine the practice behaviours of naturopaths in relation to the provision of and access to information on oral complementary medicines. In particular we were interested to know what are their counseling and advice-giving behaviors and are these behaviors adequate to ensure safe and judicious use of oral complementary medicines. The study had two aims: 1\. to explore the information provided by naturopaths to clients on these medicines, circumstances under which information would not be provided to a client, and the types of questions clients ask in respect to oral complementary medicines, and 2\. to examine the skill base of naturopaths with accessing information, how practitioners find out about adverse events and changes in regulation to complementary medicines and their confidence in respect to answering client questions about these medicines. Methods ======= We designed a short self completion questionnaire with the aim of collecting data from naturopaths describing their practice behaviors, education and training, information gathered during a consultation with a client, experience with accessing information and knowledge about complementary and alternative therapies and socio demographic characteristics. The questionnaire adopted and adapted questionnaires that had been used to assess practice behaviors and practitioner knowledge around CAM and the use of other over the counter medicines \[[@B6]-[@B9]\]. A total of 36 questions were included with the inclusion of some open ended questions to examine access to information on CAM, any barriers to the provision of information on complementary medicines and suggestions for reducing any perceived barriers. The questionnaire was piloted with a small group of naturopaths in South Australia and minor revisions made. The questionnaire took approximately 20 minutes to complete. The study was approved by the Human Research Ethics Committee at the University of South Australia. The sampling frame was based on a population of 3,000 naturopaths from three States in Australia. The sample size was estimated at 300, based on an estimate that 90% of naturopaths would actively prescribe herbs, with a 95% confidence interval, this allows for a 5% error and a 50% response rate. A stratified sample was dawn from three States in Australia. Data were collected from a national survey of practicing naturopaths in Australia. A representative sample of 300 naturopaths was undertaken from a listing of naturopaths held by the Medical Benefits Fund (MBF), a large national health insurance fund in Australia. It was decided to sample from a database held by a health fund because of the absence of a central database of naturopaths held by any professional association. MBF did not release the list of practitioners but undertook the sampling procedure under direction of the research team and provided the administrative support to mail out the questionnaires. The Fund generated a list of active naturopaths defined as those who have had a claim paid in the previous 12 months. Subjects were assigned a random number selected by a researcher independent from the study team. Naturopaths were sent a questionnaire with a covering letter explaining the purpose of the study and a reply paid envelope. Two reminders were sent out to facilitate the return of questionnaires. Statistical analysis was performed using SPSS version 11.5 (Chicago, IL, USA)\[[@B10]\]. Frequencies and percentages and other descriptive statistics were used to describe the data. Results ======= In total 300 questionnaires were sent out, and 110 were returned. However, five of these were non practicing practitioners giving a response rate of 35%. The majority identified themselves as Caucasian and female and were in the 45--54 age category. A diploma was the standard qualification (Table [1](#T1){ref-type="table"}). Naturopaths received training in the modalities of Western herbal medicine (96%), nutritional supplementation (94%), diet therapy (90%), remedial therapies (78%) and homoeopathy (74%). Over a third of practitioners worked full time in practice, working for 29 (+/- 14.5 SD) hours in clinic per week and over half of practitioners (51%) had been in practice for less than 10 years. Where demographic data was available age, gender and education were compared with data from the naturopaths and Western herbalists workforce survey\[[@B3]\]. The demographics appear consistent with 76% of practitioners being female with a mean age of 44 years. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Socio-demographic characteristics of Naturopaths ::: Socio demographic characteristics of naturopaths Demographic data from naturopathic workforce survey\\\* -------------------------------------------------- ----- ------ ----------------------------------------------------------- ------ Currently practicing as a naturopath 105 95.5 Gender Male 31 30.0 191 24.0 Female 73 70.0 604 76.0 Age 44.0 10.4 18--34 years 20 19.0 35--44 29 27.6 45--54 34 32.4 55+ 21 20.0 Educational level\\ 3.1 years Diploma 49 46.7 Advanced diploma 47 44.8 Degree 36 34.3 Higher degree 9 8.6 Ethnic background Caucasian 100 96.2 Asian 3 2.9 Other 1 1.0 \Mean and SD, \\* Not mutually exclusive percentages do not add up to 100% because a practitioner may hold more than one qualification. \\ \Data from Bensoussan A, Myers SP, Wu SM, O\'Connor K (2004) ::: Practice behaviour ------------------ Most practitioners (103, 98%) reported having a dispensary in their clinic, with 101 (97%) naturopaths performing the dispensing themselves, and a small number (7%) using unqualified staff. Seventy-eight percent of naturopaths always advised their patients to purchase their products from their clinic, with the remainder advising their patients to purchase their products from the pharmacy or health food shop. Twenty eight (27% of the sample) never dispensed a repeat medication before undertaking a follow up consultation, while 74 (71%) practitioners often or sometimes did so. Generally, there was a reluctance to dispense medications without a consultation. Practitioners reported never dispensing a product without full consultation: (1) if the patient was presenting with a different condition (45% of practitioners), (2) if the patient had consulted with another naturopath (49%), and (3) if the patient requested it (58%). Counselling, screening and advice giving behaviour -------------------------------------------------- The majority of naturopaths (96%) gave verbal information to their patients on complementary medicines. Most also provided written information which included practitioner handwritten notes (83%), printed information (75%) and less often printed information from journal articles (26%), text books (25%) or manufacturer information (35%). The content of information presented to patients varied between practitioners (Table [2](#T2){ref-type="table"}). Information on dose was given by almost all practitioners. Over 75% of practitioners always gave information on the full name, reason for prescribing the product, possible interactions, expected response, contra-indications, and the action of the product. Seventy five percent of naturopaths did not \'always\' present information relating to safe use, for example information on possible adverse effects, interaction with other substances, names and ingredients and information on effectiveness. Lack of time was given as the main factor why individual ingredients were not discussed with the patient. All naturopaths asked new patients about their use of orthodox Western medicine, and over 90% of naturopaths always asked about the use of Western medicines when they recommended or dispensed CAM or were concerned about a potential interaction. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Information given to patients about prescribed products ::: Always Sometimes Rarely Never Missing -------------------------------------- -------- ----------- -------- ------- --------- ------ --- ----- ---- ----- Recommended dose 101 96.2 1 1.0 3 2.9 Full name 91 86.7 11 10.5 3 2.9 Reason for prescribing 89 84.8 6 5.7 1 1.0 9 8.6 Possible interaction with medication 85 81.0 14 13.3 4 3.8 2 1.9 Medical condition requiring caution 84 80.0 17 16.2 3 2.9 1 1.0 Expected response 84 80.0 121 10.5 4 3.8 2 1.9 4 3.8 Product actions 82 78.1 17 16.2 3 2.9 1 1.0 2 1.9 Possible adverse reactions 78 74.3 20 19.0 2 1.9 1 1.0 4 3.8 Interaction with other substances 62 59.0 30 28.6 11 10.5 2 1.9 Names of ingredients 54 51.4 40 38.1 8 7.6 2 1.9 21 1.0 Evidence of effectiveness 22 21.0 52 49.5 23 21.9 8 7.6 ::: Confidence in their own knowledge --------------------------------- The approach in counselling on appropriate use of naturopathic products will be influenced by the practitioner\'s training and their skills in life long learning. Ninety one (83%) naturopaths reported their formal training met their needs when performing the day to day practice of naturopathy. There are a number of information resources available to practitioners and we asked naturopaths to indicate the frequency with which they used a particular resource. The resources used most oftenby practitioners to obtain information on CAM included articles in professional newsletters (91%), reference textbooks (72%), continuing professional education (CPE) seminars run by manufacturers (70%), patient feedback (66%), personal observation of patient response (58%), and CPE activity run by other industry bodies (55%). Use of the scientific literature or discussion with other health professionals was reported less often. For example the resources used sometimesincluded discussion with naturopathic colleagues (49%), journal articles describing a case study (43%), journal articles describing randomised clinical trials (41%), information from course notes (43%), popular health magazines (44%) and reference web sites (37%). Information sources used rarelyby practitioners included interaction with pharmacists (46% of naturopaths), medical doctors (44% of naturopaths) or other health care practitioner (40% of naturopaths). Naturopaths\' views on the adequacy of information resources they can draw on to help answer questions relating to CAM are reported in Table [3](#T3){ref-type="table"}. The majority of naturopaths (76%) reported the available resources describing the benefits of CAM were very adequate. Over 50% of naturopaths identified five areas of practice where information resources were not very adequate. Consequently, in response to another question, at least 40% of practitioners responded not feeling very confident in answering questions relating to these five areas namely 1- use of CAM during pregnancy and when breastfeeding, 2- adverse effects, 3- safety in relation to medical conditions, 4- questions about interactions between CAM and other medicines and 5- questions on the regulatory status of CAM. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Views on the adequacy of information resources to answer questions about CAM ::: Very adequate Somewhat adequate Not adequate Not applicable ---------------------------------------------- --------------- ------------------- -------------- ---------------- ---- ------ --- ----- Benefits of CAM 78 75.7 22 21.4 2 1.9 1 1.0 Safety of CAM re children 59 57.3 36 35.0 8 7.8 Safety of CAM re pregnancy and breastfeeding 54 48.5 45 43.7 8 7.8 Adverse effects of CAM 51 49.0 42 40.4 9 8.7 1 1.0 Safety of CAM re medical conditions 45 43.3 43 41.3 16 15.4 1 0.9 Interactions of CAM and other medicines 33 31.4 53 50.5 17 16.2 2 2.0 ::: A number of factors were expressed as being very important in influencing practitioner confidence with providing information to their patients. This included: knowledge of these products (95%), access to scientific or clinical information (85%), belief in the effectiveness of these products (84%), belief in the quality of the products (84%) and belief in the safety of the products (81%). Information awareness regarding safety of CAM --------------------------------------------- A safety concern relating to the use of CAM are adverse events associated with herbal medicines. Over 96% of naturopaths reported they were notified about adverse events through manufacturer or distributor newsletters. Other sources of notification were from professional associations (88%) and professional association journals, website or email discussion lists (86%), CPE (69%) and informal discussion with colleagues (60%). Less than 30% of the naturopaths used the Therapeutic Goods Administration (TGA) (the Australian regulatory authority for therapeutic goods) website as a source of notification of adverse events. Practitioners were asked how their knowledge of adverse events could be improved. Fifty percent of practitioners identified formal education as a need and 58% identified improved access to CPE activities and other relevant information sources were needed. Formal training or improved skills to assess quality scientific or clinical evidence was viewed positively by 30% of naturopaths. Practitioners were asked open ended questions regarding whether the information they were able to access about CAMs was adequate, what they considered the barriers to the provision of information about CAMs to patients were, and what suggestions they had for ways of overcoming these barriers. The majority of practitioners were very positive about the information they could access although there were common themes in these answers regarding the time taken to access quality information. A number of practitioners mentioned the need for unbiased information and expressed concerns regarding the objectivity of the information that was supplied by the manufacturers of those products. In general, practitioners expressed strong opinions about what they perceived to be the barriers to their own access to information. The perceived division between orthodox and complementary medicine practitioners was the strongest theme presented. Themes identified less frequently were perceived Government bias against complementary medicines, the cost of accessing information and the lack of research. Other comments related to the need for professional registration of naturopaths and a need for orthodox medical practitioners to have a greater awareness of and training in complementary medicines. With regard to the information provision to patients, the barriers were seen as misleading or incorrect information in the media, time constraints within the consultation, information overload on the part of the client, information supplied with the product in language too complex for the client, and self-prescribing of CAMs through over-the-counter sales of these products. When suggesting ways to overcome these barriers, practitioners sought greater recognition and status of the naturopathic profession and increased integration and communication with the orthodox health community. Many practitioners also felt that restricting CAM products from over-the-counter sales and allowing them to be supplied by complementary practitioners only would assist with information provision to the client. There were also calls for manufacturers to include more information on their products and for greater balance in media reporting of CAM-related issues. Discussion ========== Self reported data from this study reported on the practice behaviour of naturopaths in relation to providing information on complementary medicines. Our data suggest most naturopaths are concerned about safety in relation to possible interaction between orthodox Western medicine and CAM and this is an area naturopaths explore with their patients during the consultation. There is scope to improve the practice behaviour of some naturopaths to ensure greater safety by giving consideration to reducing the number of practitioners who would dispense a product without consultation (on solely the patient\'s request), and not using unqualified staff to dispense a product. We found both consistencies and inconsistencies between practitioners where information on the product was presented. Most practitioners reported providing information on the dose, the full name, reason for prescribing the product, possible interactions, expected response, contra-indications, expected response and the action of the product. However, information was not provided by 40% or more of practitioners on possible adverse effects, interaction with other substances, and the names and ingredients of the herbal product. A bias resulting from the self report of behaviour can not be excluded. With an increasing profile of safe use of complementary medicines, improved labelling by the naturopath has the potential to reduce public health concerns and increase the judicious use of herbal medicines. Although the majority of prescribing took place in the clinic, 22% of practitioners advised their patients to purchase product from the pharmacy or health food store. Patients who are sent to these outlets may or may not be given a written directive with the item to be purchased. Patients may be sold product which is different to that recommended by the naturopath, be given conflicting advice regarding dosage or may simply forget the instructions given to them, particularly when those instructions differ from that printed on the label. Initial naturopathic training met the needs of naturopaths with performing the day to day practice of naturopathy. We have identified a number of deficiencies that will need to be addressed to ensure naturopaths can remain up to date with increasing quality information on CAM and continue to provide high standards of counselling. Naturopaths identified that information resources relating to the safety of CAM use in children, use during pregnancy and breastfeeding, the adverse effects of CAM, the safety of using CAM as related to certain medical conditions and interactions of CAM and other medicines as not very adequate. Given the insufficient research base describing any adverse effects in some of these patient groups, providing evidence based practice in these areas will remain inadequate and naturopaths will need to continue practising with caution. Naturopaths use a wide variety of information sources to obtain information on CAM, although conventional health care practitioners and practitioners from another CAM discipline were not widely used as a resource to access information. This is similar to the findings from the workforce survey where practitioners reported that they felt well prepared for practice within their clinical training with the exception of the area of inter-professional communications\[[@B3]\]. Our findings also suggest that practitioners do not rely heavily on information sources such as scientific literature that are most likely to provide quality evidence on the use or non use of complementary medicines. A notable proportion of practitioners identified a need for continuing professional education in this area and further development of skills with assessing quality scientific or clinical evidence may lead to greater reliance on evidence based information sources. The naturopathic and Western herbal medicine workforce survey reported that adverse events were relatively common, and calculated that naturopaths would on an average encounter 1.2 adverse events in their patients per year of full time practice\[[@B3]\], or one event each 11 months of practice\[[@B2]\]. This indicates that the access to quality information related to adverse events associated with oral CAMs is critical to the safe use of these medicines. The Governments\' response to Complementary Medicines in the Australian Health System*(2003) identified a need to improve access to information about adverse reports\[[@B11]\]. Our study also highlighted the need for a greater awareness of the Australian regulatory TGA website within the practitioner population. There is also a need for the positive promotion of the TGA as an information source, as the \'anti-CAM\' perception of the Government held by some of the practitioners may contribute to these practitioners not accessing Government based information sources. Practitioners also identified that their knowledge of adverse events could be improved. A limitation of this study is the low response rate which may limit the generalisations we can make relating to the national population of naturopaths. The demographic characteristics of our study population allow limited comparison with published data from national surveys from Australia and the United States\[[@B2],[@B3]\]. Data from the two surveys in Australia are comparable with respect to gender and age ranges of naturopaths. The questions asked on education can not be compared directly, however three years training in naturopathy in Australia would equate to the award of a diploma. Data from the USA survey indicates the majority of practitioners are Caucasian and this is comparable to data in our survey. The similarities between the two studies suggest the results may be generalisable to a wider population of naturopaths in Australia. The self reported questions may give rise to a bias and this highlights a need for observational research of practice behaviour to confirm our findings. Conclusion ========== The findings from this study are timely in relation to the Australian Government\'s recognition of the need to identify the information and skills needed by health care professionals to assess the quality of evidence concerning the use of complementary medicines. This study provides baseline data for describing the practice behaviors in naturopaths in relation to counseling and advice-giving behaviors and their skills in accessing information on CAM adverse events. It will facilitate the evaluation of practice behaviour over time. The majority of naturopaths in Australia report behaviours that suggest they provide appropriate and judicious use of oral CAMs for their patients but the practitioners identify the need for further training to ensure safe and judicious practice continues in the future. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= CS conceptualised the research, analysed the data and took the lead with writing the paper. KM participated in conceptualising the research, study design, analysed the data and contributed to the preparation of the manuscript. EH participated in the study design and reviewed the manuscript. SS participated in the study design and reviewed the manuscript. GB participated in the study design and reviewed the manuscript. DR participated in the study design and reviewed the manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1472-6882/5/15/prepub> Acknowledgements ================ This research was supported by a grant from the University of South Australia\'s Collaborative Research Grant Scheme. We would like to thank the Medical Benefits Fund (MBF) of Australia for their administrative support.	PubMed Central	2024-06-05T03:55:59.905772	2005-7-11	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182348/", "journal": "BMC Complement Altern Med. 2005 Jul 11; 5:15", "authors": [ { "first": "Caroline", "last": "Smith" }, { "first": "Karen", "last": "Martin" }, { "first": "Elizabeth", "last": "Hotham" }, { "first": "Susan", "last": "Semple" }, { "first": "Geraldine", "last": "Bloustien" }, { "first": "Deepa", "last": "Rao" } ] }
PMC1182349	Background ========== Automatic functional annotation is essential for high-throughput sequencing projects. Typically, large datasets undergo annotation by means of \"annotation jamborees\", where groups of experts are assigned to manually annotate a designated portion of an organism\'s genome. More recently, various tools have become available to streamline this process \[[@B1]-[@B9]\]. However, limitations encountered with these tools are that many require web-submission of data \[[@B2]\], need substantial manual intervention \[[@B1],[@B4]\], supply only a single output format, are part of a large sequence analysis package \[[@B3]\] and most importantly, do not combine a broad range of information resources. To address these shortcomings, we developed a new annotation pipeline, which we term \"AutoFACT\". Unique to AutoFACT, is its hierarchal filtering system for determining the most informative functional annotation. This paper describes AutoFACT\'s functional assignment capabilities, outlining the procedure for annotating unknown nucleotide or protein sequence data. We assess the validity of AutoFACT by comparing annotations to four previously annotated and phylogenetically diverse organisms, including human, yeast and both eukaryotic and bacterial pathogens. AutoFACT has been applied to the EST sequencing project of Acanthamoeba castellanii, a free-living soil amoeba and opportunistic human pathogen. This example highlights AutoFACT\'s performance, which yields a \~50% increase in functional annotations over a top-BLAST-hit approach against NCBI\'s non-redundant database or against UniProt\'s expert-annotated UniRef90 database. Implementation ============== AutoFACT is a command-line-driven program written in PERL for LINUX/UNIX operating systems. It uses BioPerl \[[@B10]\] modules to parse and analyze BLAST \[[@B11]\] reports. Average annotation time is 2.5 hours for 5000 sequences of approximately 500 bp in length on a desktop workstation (BLAST time not included). A web version of AutoFACT is available where users can submit up to 10 sequences at a time for annotation. For large sequencing projects, it is recommended that the user download and install the local version of AutoFACT. Results ======= Methodology ----------- AutoFACT takes a single FASTA-formatted sequence file as input, automatically recognizes the sequence type as nucleotide or protein and proceeds to ask the user for preferences regarding which databases to use, the order of database importance and bit score cutoff. The bit score is a measure of sequence similarity independent of the size of the database used (unlike E-values). It is derived from the raw alignment score in which the statistical properties of the scoring system used have been taken into account. Bit scores are normalized with respect to the scoring system and hence can be used to compare alignment scores from different searches \[[@B12]\]. Each sequence in the FASTA-formatted file is then assigned to one of six annotation classes: (1) Ribosomal RNA (rRNA), (2) \[Functionally annotated\] protein, (3) Unassigned protein, (4) \[Domain name\]-containing protein, (5) Unknown EST (when using EST data) or (6) Unclassified (Table [1](#T1){ref-type="table"}, Figure [1](#F1){ref-type="fig"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### AutoFACT annotation classes ::: Annotation Class Hit to LSU or SSU rRNA database Hit to UniRef, nr, KEGG and/or COG Hit is inform-ative Hits share common inform-ative terms Hit to Pfam or Smart Hit to est\_others --------------------------------------------- ------------------------------------- ---------------------------------------- ------------------------- ------------------------------------------ -------------------------- ------------------------ \"Ribosomal RNA\" YES N/A N/A N/A N/A N/A \" \[Functionally Annotated\] protein\" NO YES YES YES N/A N/A \"Unassigned protein\" NO YES YES/NO NO NO N/A \" \[Domain name\]-containing protein\" NO YES/NO NO NO YES N/A \"Unknown EST\" NO NO N/A N/A NO YES \"Unclassified\" NO NO N/A N/A NO NO ::: ::: {#F1 .fig} Figure 1 ::: {.caption} ###### AutoFACT methodology. Sequences are classified into one of six annotation categories (purple boxes). The user decides which bit score cutoff to use (default 40) before a BLAST hit is considered significant. For database references, see text. ::: ![](1471-2105-6-151-1) ::: AutoFACT assigns classification information, based on a hierarchal system, from a collection of specialized resources, currently nine databases (Table [2](#T2){ref-type="table"}), using BLAST comparison \[[@B13]\]. Since not all descriptions from top BLAST hits are genuinely informative, AutoFACT adopts the \"uninformative rule\" \[[@B5]\], by which the highest scoring BLAST hit with a biologically informative description is considered informative. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Databases searched and classification information assigned by AutoFACT ::: --------------------------------------------------------------------------------------------------------------------------- Database Classification Information Reference ------------------------------------------------ ---------------------------------------------------------- --------------- European Ribosomal Database Large subunit (LSU) ribosomal RNAs\ \[25\] Small subunit (SSU) ribosomal RNAs Uniprot\'s UniRef 90 GeneOntology terms Enzyme Commission numbers Locus names \[16,26\] Uniprot\'s UniRef100 Clusters of Orthologous Groups (COG) Functional categories \[27,28\] Kyoto Encyclopedia of Genes and Genomes (KEGG) Metabolic pathways Enzyme Commission numbers Locus names \[29,30\] Protein Familes Database (Pfam) Protein domains \[31\] Smart Signaling domains\ \[32\] Domain architectures NCBI\'s non-redundant database (nr) N/A \[33\] NCBI\'s est\_others database --------------------------------------------------------------------------------------------------------------------------- ::: Figure [1](#F1){ref-type="fig"} outlines the AutoFACT methodology. When analyzing nucleotide data, AutoFACT begins by using BLAST to search the nucleotide sequences in the input file against the set of user-specified databases. If a match to the rRNA dataset is found with a minimum match length and percent sequence identity (default: 50 bp and 84% identity), the sequence is classified as a \"ribosomal RNA\". If no match is found the sequence is then searched against the remaining set of user-specified databases. In step 2 (or step 1 for protein data), description lines of significant hits, based on a user-specified bit score cutoff (default \<40), are examined for the presence of functionally uninformative terms such as \'hypothetical\', \'unknown\', \'chromosome\', etc. When a hit contains an uninformative term, the next best hit is scrutinized and so forth, until a description line without uninformative terms is found, e.g. \'proton-transporting ATP synthase\'. The user specifies the number of top BLAST hits the program should filter. In step 3, a search for common terms among the informative hits from each database is performed. For annotation transfer, the user specifies a database order of importance so that informative terms from the first database are searched against informative terms from the remaining databases in a given order. For example, if the user specifies the database order as UniRef90, nr, KEGG and COG, informative terms in the informative hit from UniRef90 are first searched for matches to the informative hits from the other databases. If a match is found between at least one informative term from the UniRef90 hit and at least one other informative database hit (e.g., \'proton-transporting [ATP synthase]{.underline}\' matches \'H^+^-pumping [ATP synthase]{.underline}\'), the description line of the UniRef90 hit is assigned to the input sequence. If there are no matches to UniRef90 terms, the informative terms from the informative hit of the next database (nr, in this example) are then queried in the same way as above, until a functionally informative description line has been assigned to the sequence. We prefer to use UniRef90 as the first database in the order of importance for two reasons. First, as a member of UniProt it is one of the better annotated and curated of the available databases. Second, because UniProt entries with 90% sequence similarity are combined into a single record, the description lines are species-independent and tend to be more general in their descriptions. On the other hand, description lines from NCBI\'s nr database are often several lines long and contain repetitive information. Testing showed that using various database combinations does not significantly change the annotation results. A user\'s choice of db order is therefore dependent on the format of the description line one would prefer to assign to the sequence in question (Table [3](#T3){ref-type="table"}). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Database description line formats from ACL00000101 BLAST hits ::: Database Description Line -------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- UniRef90 ATP synthase beta chain related cluster UniRef100 ATP synthase subunit beta \[Salmonella typhimurium\] NCBI\'s nr ATP synthase beta chain \[Erwinia carotovora subsp. atroseptica SCRI1043\] emb\\|CAG77407.1\\| ATP synthase beta chain \[Erwinia carotovora subsp. atroseptica SCRI1043\] KEGG atpD; membrane-bound ATP synthase, F1 sector, beta-subunit \[EC:3.6.3.14\] \[KO:K02112\] COG \[C\] COG0055 F0F1-type ATP synthase, beta subunit ::: AutoFACT proceeds to step 4 when there are no common informative terms between any of the databases, or when only uninformative hits are found. In this step, a sequence with significant similarity to one or more sequences in the Pfam or SMART databases is classified as a \' \[domain name\]-containing\' protein or a \'multi-domain-containing protein\'. A sequence containing no domains is simply classified as an \'unassigned protein\'. A sequence is also classified as a \' \[domain name\]-containing protein\' when the only significant hit is to a domain database. It is considered \'unclassified\' when no hits are found to any of the specified databases. When EST sequences are being annotated, the last step in the annotation pipeline is to check the sequence against NCBI\'s est\_others database. If a significant match is found, the sequence is classified as an \'unknown EST\'; otherwise it remains \'unclassified\'. In step 5, functionally annotated sequences are then classified according to KEGG pathways, COG functional groups, Enzyme Commission (EC) numbers, GeneOntology (GO) terms and locus names. Putative KEGG pathways are assigned if an informative term from the automatically assigned description line matches a term in the informative KEGG hit. The same reasoning is used to assign putative COG functional categories. EC numbers \[[@B14]\] are assigned in one of two ways, either from parsing the KEGG description line or by mapping the accession number of the informative UniRef hit to an enzyme viaExPASy\'s enzyme.dat file \[[@B15]\]. GO terms are assigned by mapping the UniRef accession number of the informative hit viathe gene\_association.goa\_uniprot file \[[@B16]\]. Three different output formats are generated by AutoFACT: HTML web pages (Figure [2](#F2){ref-type="fig"}) for easy viewing and browsing, a General Feature Format (GFF) file \[[@B17]\] to facilitate data transfer to the user\'s private database and a simple tab-delimited text file for easy data extraction and manipulation. A log file is also generated to document all decision-making steps in the annotation process. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Distribution of informative versus uninformative annotations. A. castellaniiESTs (5,130 clusters) were annotated in three ways: (A) by top BLAST hit to NCBI\'s nr database; (B) by top BLAST hit to UniProt\'s UniRef90 database; and (C) by AutoFACT. The \"uninformative rule\" (Andrade et al., 1999) was used to query description lines assigned by all methods. AutoFACT yields an \~50% increase in informative annotations compared to top BLAST hits against NCBI\'s nr and the UniRef90 databases. AutoFACT\'s annotation source is shown in parentheses (). ::: ![](1471-2105-6-151-4) ::: Validation ---------- To assess the validity of AutoFACT annotations, we compared results for 200 randomly chosen cDNA sequences across four previously annotated and phylogenetically diverse organisms: i) Homo sapiens, annotated by the Ensembl Annotation Pipeline \[[@B8]\]; ii) Saccharomyces cerevisiae, annotated by MIPS/PEDANT \[[@B18],[@B19]\]; iii) Plasmodium falciparum, annotated by The Institute For Genomic Research (TIGR) \[[@B20]\]; and iv) Rickettsia prowazekii, previously annotated by GeneQuiz \[[@B5]\]. We used AutoFACT\'s default values and considered hits to genes from the same species as uninformative. Figure [3](#F3){ref-type="fig"} compares the annotation results of 200 randomly chosen sequences for each species/pipeline. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Sample HTML output for AutoFACT annotation of Acanthamoeba castellaniiEST cluster ACL00000152. Automatic annotation results are displayed at the top of the page and all data used to infer the annotation are represented in the bottom part of the table. Percent sequence identity is the extent to which two (nucleotide or amino acid) sequences, in a High Scoring Segment Pair (HSP), are invariant. In the case of the est\_others data, the reported % sequence identity refers to a \"translated nucleotide -- translated nucleotide\" comparison. The \"Informative Hit\" value specifies whether the first, second, etc., BLAST hit in the corresponding database was informative. The \"Color Key for Alignment Scores\" displayed at the top of the diagram is from NCBI\'s BLAST Results page. The scores for the annotation and for the source of the annotation, 627 in this example, are highlighted according to the color key. The page also contains links to relevant database entries. ::: ![](1471-2105-6-151-2) ::: Homo sapiens\[Ensembl\] ------------------------- Comparison of Human Ensembl annotations to AutoFACT revealed no significant differences in annotation assignments. There were 2/200 (1%) sequences that AutoFACT annotated as \'unassigned protein\', either because the only BLAST hits were to other human sequences or because the informative terms could not be matched across database sources. Had we been less strict in our annotation criteria and considered hits to the same species as informative, AutoFACT would then have assigned the same annotations as Ensembl to these two sequences. The high similarity between annotation results is primarily due to the fact that the source of most of the Ensembl annotations is UniProt/SWISSPROT, which AutoFACT also uses viaUniRef90, the database of highest importance in the AutoFACT database order. Saccharomyces cerevisiae\[MIPS/PEDANT\] ----------------------------------------- AutoFACT and PEDANT annotations for a set of 200 cDNAs differed by 5% (10/200). We examined the original annotations for these 10 sequences in the expertly curated Saccharomyces Genome Database (SGD). Because AutoFACT considered hits to Saccharomyces cerevisiaeas \'uninformative\', 6/10 sequences were classified as \' \[domain name\]-containing proteins\'. We do not consider these annotations to be false positives, merely less specific annotations. In 1/10 of the assignments, AutoFACT was better than PEDANT (Table [4](#T4){ref-type="table"}). The remaining 3/10 annotations are considered to be false positives, suggesting an overall error rate of 1.5% (3/200). ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Differences found between AutoFACT and PEDANT annotations for Saccharomyces cerevisiae ::: ID PEDANT Annotation AutoFACT Annotation AutoFACT Score AutoFACT E-value AutoFACT % Identity --------- ----------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------- ---------------- ------------------ --------------------- yal048c vacuolar aspartic protease GON1; possible rho-like GTPase involved in secretory vesicle transport 1724 0.0 50% (360/718) yhr064c SSZ1 -- regulator protein involved in pleiotropic drug resistance multi-domain protein 651 3.00E-68 28% (154/539) yhr046c INM1 -- inositol-1(or 4)-monophosphatase \Protein qutG related cluster 378 4.00E-35 31% (99/310) yhr143w DSE2 -- glucan 1,3-beta-glucosidase activity* multi-domain protein 229 2.00E-19 25% (70/278) yhl043w ECM34 -- involved in cell wall biogenesis and architecture DUP domain-containing protein 205 5.00E-17 36% (26/72) yal047c SPC72 -- Stu2p Interactant \Repeat organellar protein related cluster 160 2.00E-09 20% (124/620) yhr167w THP2 -- subunit of the THO complex, which appears to functionally connect transcription elongation with mitotic recombination* \Myosin heavy chain related cluster 129 2.00E-06 24% (51/210) yhr154w RTT107 -- Establishes Silent Chromatin* BRCT domain-containing protein 118 4.00E-06 28% (24/83) yhl020c OPI1 -- negative regulator of phospholipid biosynthesis pathway multi-domain protein 114 5.00E-06 24% (30/123) yhr196w UTP9 -- U3 snoRNP protein Borrelia\_orfA domain-containing protein 104 1.00E-04 19% (75/376) Annotations in bold are the same as the original annotations found in the Saccharomyces Genome Database. AutoFACT annotations marked with an asterisk (\) are considered false positives. ::: Plasmodium falciparum\[TIGR\] ------------------------------- We compared TIGR\'s preliminary annotations for a set of 200 Plasmodium falciparumcDNAs to annotations generated by AutoFACT. TIGR\'s preliminary annotations are automatically assigned by searching nucleotide and protein databases for \"good\" matches. At this preliminary stage, none of the annotations are examined or verified by human annotators. We found that between the two fully automatic pipelines, 4% (8/200) of the annotations differed, half of which were annotated by AutoFACT as \' \[domain name\]-containing proteins\' (Table [5](#T5){ref-type="table"}). Because TIGR\'s preliminary annotations have not been examined by human annotators, we cannot estimate the % false positives in this instance. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Differences found between AutoFACT and TIGR preliminary annotations for Plasmodium falciparum* ::: ID TIGR Preliminary Annotation AutoFACT Annotation AutoFACT Score AutoFACT E-value AutoFACT % Identity ------------- --------------------------------------------------------------------------------------------- ------------------------------------------------------------- ---------------- ------------------ --------------------- 1396.m03572 PF14\_0675 reticulocyte binding protein 2 homolog B, putative Reticulocyte Binding protein; multi-domain protein 157 1E-10 18% (60/320) 1396.m03591 PF14\_0655 RNA helicase-1, putative Eukaryotic translation initiation factor 4A related cluster 1591 1E-177 79% (310/388) 1396.m03721 PF14\_0530 ferlin, putative heat shock protein DNAJ pfj4 534 6E-53 40% (103/252) 1396.m04144 PF14\_0112 POM1, putative Twinkle related cluster 152 6E-08 38% (34/89) 1396.m04178 PF14\_0078 HAP protein Asp domain-containing protein 535 8E-55 26% (100/371) 1396.m04220 PF14\_0036 acid phosphatase, putative Metallophos domain-containing protein 134 2E-08 20% (45/220) 1396.m04244 PF14\_0015 aminopeptidase, putative hydrolase, alpha/beta fold family 179 5E-12 22% (66/288) 1396.m04296 PF14\_0382 metalloendopeptidase, putative multi-domain protein 118 0.000006 16% (50/297) ::: Rickettsia prowazekii\[GeneQuiz\] ----------------------------------- AutoFACT annotations for Rickettsia prowazekii\[[@B21]\] were compared to annotations previously assigned by GeneQuiz (\[[@B5],[@B22]\]). AutoFACT differed from GeneQuiz annotations at 4.5% (9/200) of the sequences, yet differed only by 1% (2/200) from the more accurate original annotations \[[@B21]\], which are based on human inspection and include phylogenetic information. GeneQuiz estimates an overall error rate of 2.5--5%, which is confirmed in our comparison here (Table [6](#T6){ref-type="table"}). Based on these automatic annotation results, AutoFACT is the more accurate of the two pipelines, with an error rate of only 1%. ::: {#T6 .table-wrap} Table 6 ::: {.caption} ###### Differences found between AutoFACT and GeneQuiz annotations for Rickettsia prowazekii ::: ID Gene Quiz Annotation AutoFACT Annotation AutoFACT Score AutoFACT E-value AutoFACT % Identity ------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------- ---------------- ------------------ --------------------- RP103 PKM101 CONJUGATION PROTEINS (TRAL), (TRAM), (TRAA), (TRAB), (TRAC), (TRAB), (TRAC), (TRAD), (TRAN), (TRAE), (TRAO), (TRAF), (TRAG), ENTRY EXCLUSION PROTEIN (EEX), (KIKA), (KORB), (KORA) AND ENDONUCLEASE (NUC) GENES, COMPLETE CDS (TRAM) (TRAB) (TRAB) (TRA VIRB4 PROTEIN related cluster 4159 0.0 100% (805/805) RP151 NEMPA PROTEIN PRECURSOR. Aspartyl/glutamyl-tRNA(Asn/Gln) amidotransferase subunit B related cluster 2004 0.0 82% (398/483) RP259 D-STEREOSPECIFIC PEPTIDE HYDROLASE PRECURSOR. *Penicillin binding protein 4\ related cluster 2048 0.0 96% (398/414) RP268 NADH-UBIQUINONE OXIDOREDUCTASE CHAIN 2 (EC 1.6.5.3). Heme exporter protein B related cluster 794 3E-84 74% (160/215) RP282 NADH DEHYDROGENASE SUBUNIT 2.** \HyfB domain-containing protein related cluster 1821 0.0 74% (380/512) RP287 CAVEOLIN-2. VIRB8 PROTEIN related cluster* 1047 1E-114 85% (212/247) RP291 CONJUGAL TRANSFER PROTEIN TRBI. VIRB10 PROTEIN related cluster 2016 0.0 85% (413/483) RP293 CONJUGAL TRANSFER PROTEIN TRAG. VIRD4 PROTEIN related cluster 3002 0.0 97% (577/591) RP414 LPS BIOSYNTHESIS RFBU RELATED PROTEIN. \Glycosyltransferase related cluster 1614 1E-180 92% (314/338) Annotations in bold are the same as the original annotations by Andersson et al. (1998). AutoFACT annotations marked with an asterisk (\) are considered false positives. ::: Case Study: Acanthamoeba castellanii -------------------------------------- AutoFACT is currently used by the Protist EST Program (PEP) \[[@B23]\], a pan-Canadian genomics initiative involving investigators at six Canadian universities. The objective of PEP is to survey, through EST sequencing, the expressed portions of the genomes of a phylogenetically comprehensive selection of protists (30--40 of these mostly unicellular eukaryotes). Under the PEP initiative, 12,937 individual EST reads yielding 5,130 clusters (consensus sequences) have been obtained to date for A. castellanii. We compared AutoFACT annotations for these clusters to annotations taken from top BLASTx hits against NCBI\'s nr database and from top BLASTx hits against UniProt\'s well-annotated UniRef90 database. AutoFACT compared the A. castellaniisequences against a total of seven databases. UniRef90, KEGG, COG and NCBI\'s nr were searched using BLASTx ; Pfam and SMART were searched using RPS-BLAST; and NCBI\'s est\_others database was searched using tBLASTx. In each instance, a bit score cutoff of 40 was used and the top 10 BLAST hits were filtered for uninformative terms. The database order of importance was UniRef90, KEGG, COG, NCBI\'s nr. Figure [4](#F4){ref-type="fig"} shows an \~50% increase in functionally informative annotations with AutoFACT (58% informative hits) compared to the quick and easy top-BLAST-hit approach (\~ 32%). Scanning the top 10 hits for informative terms in AutoFACT\'s UniRef90 source alone results in a 10% increase in informative annotations over the top-BLAST-hit approach against both nr and UniRef90. This result demonstrates the power of the \"uninformative rule\" alone. As such there is a significant decrease (from 19% to 6%) in \'uninformative\' hits when using AutoFACT. By searching against the domain databases Pfam and SMART, AutoFACT reduces the number of \'no hits found\' by approximately 10% in comparison to the datasets annotated by the top-BLAST-hit approach. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Comparison of AutoFACT annotations across four phylogenetically diverse organisms previously annotated by well-established automatic pipelines. Two hundred previously annotated cDNAs from Homo sapiens\[Ensembl Annotation Pipeline\], Saccharomyces cerevisiae\[MIPS/PEDANT\], Plasmodium falciparum\[TIGR\] and Rickettsia prowazekii\[GeneQuiz\] were re-annotated with AutoFACT using a bit score cutoff of 40 and a database order of importance as follows: UniRef90, KEGG, COG, NCBI\'s nr, Pfam and SMART. The top 10 BLAST hits to each database were filtered for functionally uninformative terms. BLAST hits to the species itself were considered uninformative. The portion of the bar representing different results from AutoFACT (dark purple) should not be construed as false positives. For example in the case of GeneQuiz (4.5% differences), it is the AutoFACT annotation that is the better of the two in almost all instances (see Results section). Numbers printed directly on columns represent the number of cDNA sequences (out of 200) in each category. ::: ![](1471-2105-6-151-3) ::: AutoFACT annotations for each organism mentioned above can be viewed at <http://megasun.bch.umontreal.ca/Software/AutoFACT.htm> Conclusion ========== To efficiently and fully exploit the wealth of sequence data currently available, thorough and informative functional annotations are paramount. Considering the ever-growing number of EST sequencing projects, it becomes increasingly important to fully automate the annotation process and to make optimal use of the various available annotation resources and databases. Because no two annotation systems are exactly alike, choice of system is very much dependent on the user\'s end goal. AutoFACT uses a hierarchal filtering system for determining the most informative functional annotation. It provides a means of classification by identifying EC numbers, KEGG pathways, COG functional classes and GeneOntology terms. AutoFACT supplies three different output formats and a log file, which are versatile and adaptable to user requirements. Importantly, it allows users to maintain data locally, whereas many other systems require sequence submission elsewhere for annotation. By combining multiple resources, AutoFACT associates sequences with a broad range of biological classifications and has proven to be very powerful for annotating both EST and protein sequence data. The A. castellaniicase study shows that in comparison to the \'quick and easy\' top-BLAST-hit approach against either NCBI\'s nr or UniProt\'s UniRef databases, AutoFACT substantially improves functional annotations of sequence data. Comparisons to other well-established annotation pipelines show that AutoFACT performs equally well and in some cases better than the alternative. We have also demonstrated that AutoFACT exhibits an equivalent level of performance (1--2% error rate) when it is used to annotate sequences across different domains of life. Finally, we caution that over-prediction is common when using sequence similarity to infer protein function. Examples of similar sequences that do not share the same or even related functions have been documented \[[@B24]\]. Automatic annotations therefore may require further validation in certain cases. Availability and requirements ============================= Project name: AutoFACT Project homepage: <http://megasun.bch.umontreal.ca/Software/AutoFACT.htm> Operating system(s): LINUX/UNIX Programming language: PERL Other requirements: BioPerl and BLAST License: GNU General Public License (GPL) Any restrictions to use by non-academics: None Authors\' contributions ======================= LBK designed, developed and implemented AutoFACT. MWG provided the Acanthamoeba castellaniidata used to test and validate AutoFACT. GB and BFL supervised the study, making significant design contributions. All authors read and approved the final manuscript. Acknowledgements ================ This work has been conducted in the context of the Protist EST Program (PEP) and is supported by Genome Canada/Atlantic/Quebec. We thank Eric Wang and Pierre Rioux for their suggestions and script testing. Thank you to Amy Hauth for her critique of the manuscript and ever-helpful discussions, to Emmet O\'Brien and Beatrice Roure for their useful feedback and to BioneQ for access to their high-performance computer cluster. Computer resources financed by a grant from the Canadian Institutes of Health Research (CIHR, grant \# MOP15331) have also been used in this work.	PubMed Central	2024-06-05T03:55:59.909873	2005-6-16	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182349/", "journal": "BMC Bioinformatics. 2005 Jun 16; 6:151", "authors": [ { "first": "Liisa B", "last": "Koski" }, { "first": "Michael W", "last": "Gray" }, { "first": "B Franz", "last": "Lang" }, { "first": "Gertraud", "last": "Burger" } ] }
PMC1182350	Background ========== One of the biggest challenges facing biologists today is the structural and functional classification and characterization of protein sequences. For example, in humans, the number of proteins for which the structures and functions are unknown makes up more than 40% of the total number of proteins. As a result, over the past couple of decades, extensive research has been done on trying to identify the structures and functions of proteins. It is well known that the subcellular localization of proteins plays a crucial role in their functions \[[@B1]\]. A number of computational approaches have been developed over the years to predict the localization of proteins, including recent works like \[[@B2]-[@B12]\]. Initial efforts relied on amino acid compositions \[[@B13],[@B14]\], the prediction of signal peptides \[[@B15]-[@B19]\] or a combination of both \[[@B20],[@B21]\]. Later efforts were targeted at incorporating sequence order information (in the form of dipeptide compositions etc.) in the prediction algorithms \[[@B22]-[@B27]\]. There are drawbacks associated with all these methods. For example, prediction algorithms based on amino acid compositions suffer from the drawback that there is a loss of contextual information. As a result, sequences which are completely different in function and localization but that have a very similar amino acid composition would both be predicted as belonging to the same region of the cell. On the other hand, approaches that rely on predicting signal peptides can lead to inaccurate predictions when the signals are missing or only partially included \[[@B13]\]. Recent efforts have also focused on the use of physicochemical properties to predict subcellular localization of proteins \[[@B28],[@B29]\]. Bhasin et al. \[[@B30]\] created an algorithm which was a hybrid of four different predictive methods. In addition to using amino acid compositions and dipeptide composition information, they also included 33 different physicochemical properties of amino acids, averaged over the entire protein. However such a globally averaged value again leads to a loss of contextual information. Bickmore et al. \[[@B31]\] studied the characteristics of the primary sequences of different proteins and concluded that motifs and domains are often shared amongst proteins co-localized within the same sub-nuclear compartment. Since the structure and hence the function of proteins is dictated by the different interacting physicochemical properties of the amino acids making up the protein, it would stand to reason that co-localized proteins must share some conservation in the different properties. In this paper, we present a new algorithm called pSLIP: Prediction of Subcellular Localization in Proteins. We use multiple physicochemical properties of amino acids to obtain protein extracellular and subcellular localization predictions. A series of SVM based binary classifiers along with a new voting scheme enables us to obtain high prediction accuracies for six different localizations. Results and discussion ====================== We implemented our algorithm on Park and Kanehisa\'s dataset \[Table [1](#T1){ref-type="table"}\]. We divided the dataset into clusters based on sequence length and ran N-fold cross validation (NF-CV)tests for each of the protein clusters. The accuracies for each of these clusters were recorded and finally, these cluster accuracies were combined to produce overall accuracies. Table [2](#T2){ref-type="table"} lists them along with results from Park and Kanehisa\'s work \[[@B21]\] and Chou and Cai\'s work \[[@B8]\]. The protein subcellular localization method used in \[[@B21]\] is based on amino acid compositions. Chou and Cai\'s method \[[@B8]\] uses a hybrid algorithm called GO-FunD-PseAA \[[@B5]\] that combines gene ontology \[[@B32]\], functional domain decomposition \[[@B33]\] and pseudo-amino acid composition \[[@B26]\] for localization prediction. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### The number of proteins in the dataset. \* These classes were not considered as they have too few proteins to achieve reliable training. ::: Subcellular Location Number of entries ------------------------- ------------------- Chloroplast 671 Cytoplasmic 1241 Cytoskeleton\* 40 Endoplasmic reticulum\* 114 Extracellular 861 Golgi apparatus\* 47 Lysosomal\* 93 Mitochondrial 727 Nuclear 1932 Peroxisomal\* 125 Plasma membrane 1674 Vacuolar\* 54 Total 7579 ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Sensitivity (sens) and Specificity (spec) (in %) on Park and Kanehisa\'s dataset \[21\]. First two columns show results from Park and Kanehisa\'s algorithm \[21\] obtained by 5-fold crossvalidation. The next column shows results from Chou and Cai\'s work \[8\] obtained using leave one out test. The last set of results are from our algorithm obtained using 5-fold and 10-fold crossvalidation. ::: Subcellular P & K P & K Chou Cai pSLIP ----------------- ------- ------- ---------- ------- ------ ------ ------ sens sens sens spec sens spec sens Chloroplast 72.3 70.3 93.9 89.9 84.8 93.5 92.4 Cytoplasmic 72.2 73.9 91.5 86.2 84.1 91.6 87.7 Cytoskeleton 58.5 59.8 80.0 \- \- \- \- ER 46.5 39.0 90.3 \- \- \- \- Extracellular 78.0 77.1 90.0 96.3 92.0 98.4 93.7 Golgi Apparatus 14.6 \- 76.6 \- \- \- \- Lysosomal 61.8 62.4 92.5 \- \- \- \- Mitochondrial 57.4 53.5 83.6 75.9 86.8 85.7 93.5 Nuclear 89.6 89.0 95.3 89.4 90.4 93.1 93.1 Peroxisomal 25.6 \- 82.4 \- \- \- \- Plasma Membrane 92.2 91.9 95.0 95.1 94.1 95.0 97.0 Vacuolar 25.0 \- 66.7 \- \- \- \- TA 78.2 79.1 92.4 89.5 93.1 LA 57.9 68.5 \- 88.7 92.9 ::: The results reported by Park and Kanehisa are obtained after 5-fold cross validation testing. To ensure fairness in comparing results, we ran a 5-fold test on our algorithm. As is apparent from Table [2](#T2){ref-type="table"}, our method provides good overall accuracy of 89.5% which is significantly higher than 78.2% and 79.1% obtained for the two different cases from Park and Kanehisa\'s paper. Even more interesting is the fact that the accuracies obtained by Park and Kanehisa are skewed towards those locations that have the most number of proteins in the dataset, viz., nuclear and plasma membrane. Total accuracies can sometimes present a misleading picture about the efficacy of a classification technique. Local accuracies, on the other hand can provide a more realistic view of classification efficiencies. We obtained a local accuracy of 88.7% which is only slightly less than the overall accuracy (89.5%) of the technique. On the other hand, the local accuracies obtained by Park and Kanehisa are significantly lower than the corresponding total accuracies (57.9% and 68.5% when compared with total accuracies of 78.2% and 79.1% respectively.) Chou and Cai have used the leave one out cross validation (LOO-CV)test to assess the performance of their GO-FunD-PseAA predictor. Due to reasons described later, we\'ve used only NF-CV tests. In order to make a reasonable comparison with their results, we did a 10-fold test which provides a good trade-off between bias and variance in test results. As results in Table [2](#T2){ref-type="table"} show, our algorithm performs as well as the GO-FunD-PseAA predictor and the obtained accuracy of 93.1% compares favorably with the 92.4% accuracy obtained by Chou and Cai. Although Chou and Cai\'s work tackles the harder problem of classifying over more subcellular locations than we do, the results do show the promise in the approach of using physicochemical properties for localization prediction. Table [3](#T3){ref-type="table"} shows the classification performance for each of the individual clusters. We found that the prediction accuracies for each of the classes is largely uniform across the different clusters. However, for the cluster with base length of 450, the classification performance for all classes are uniformly lower than the accuracies obtained for the other clusters. This is probably due to the presence of sequences of lengths far greater than the base length of that cluster. The was because of an insufficient number of sequences of lengths greater than 1350 (in order to form a separate cluster of their own.) However, it is clear that if not for this aberration, the overall accuracies for this method would be higher. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Cluster-wise Specificity (spec) and Sensitivity (sens) (in %) for pSLIP using 10-fold cross validation. ::: Base length 50 150 300 450 Overall --------------- ------ ------ ------ ------ --------- ------ ------ ------ ------ ------ spec sens spec sens spec sens spec sens spec sens Chloroplast 99.3 97.9 95.9 93.0 97.3 89.1 83.0 90.2 93.5 92.4 Cytoplasmic 95.6 90.3 94.6 93.4 89.4 92.9 89.9 78.5 91.6 87.7 Extracellular 99.1 100 99.0 98.1 97.8 89.3 96.3 80.4 98.4 93.7 Mitochondrial 95.6 96.6 92.1 95.0 88.5 91.0 73.3 92.0 85.7 93.5 Nuclear 93.6 94.6 95.3 95.8 93.6 93.6 91.4 90.8 93.1 93.1 Plasma Memb 95.9 98.9 96.9 98.8 97.4 99.0 92.7 94.8 95.0 97.0 TA 96.4 95.9 93.8 89.3 93.1 LA 96.4 95.7 92.5 87.8 92.9 ::: Cross validation experiments are frequently prone to an optimistic bias \[[@B34]\]. This occurs because the experimental setup can be such that the choice of the learning machine\'s parameters becomes dependent on the test data. We\'ve tried to minimize the possible effect of this by using only a small subset (ninety sequences of each type) of the available sequences for parameter search, as described later in this paper. As a further experiment, we\'ve also carried out an independent dataset (ID)test using the the eukaryotic sequences dataset developed by Reinhardt and Hubbard \[[@B14]\]. This dataset has also been widely used for subcellular localization studies. Instead of doing cross-validation testing on this dataset, we use the SVM classifiers generated by our method using Park and Kanehisa\'s dataset and predict the subcellular localization of all sequences in the Reinhardt and Hubbard dataset. The results of this test, along with the results obtained by others using this dataset, are shown in Table [4](#T4){ref-type="table"}. The first set of results are from Reinhardt and Hubbard\'s work \[[@B14]\] in which a neural network based classifier built upon amino acid composition as input is used. The next set of results are from Hua and Sun \[[@B13]\] who used an amino acid composition based SVM method. Esub8 \[[@B27]\] uses some sequence information along with amino acid composition for classification using SVM. ESLpred \[[@B30]\] is a hybrid model combining amino acid composition, dipeptide composition and average physicochemical properties as features for sequences that are then classified by SVM. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Classification performance (sensitivity) (in %) on Reinhardt and Hubbard\'s dataset \[14\]; NF-CV: Results are given by N-fold cross validation. LOO-CV: Results are given by leave one out cross validation test. ID: Results are given by directly testing entire dataset, without any training on this dataset. ::: Subcellular R&H SubLoc Esub8 ESLpred pSLIP --------------- ------ -------- ------- --------- ------- Cytoplasmic 55 76.9 80 85.2 75.9 Extracellular 75 80 86.5 88.9 76.3 Mitochondrial 61 56.7 67.6 68.2 85.3 Nuclear 72 87.4 91.2 95.3 84.2 TA 66.0 79.4 84.14 88.0 81.0 LA 65.8 75.3 81.3 84.4 80.4 ::: The results in Table [4](#T4){ref-type="table"} illustrate the importance of incorporating sequence order information in the classification method. The first two methods ignore order information entirely and we believe that their prediction accuracy suffers as a result of this. Furthermore, although the prediction accuracies of Esub8 and ESLpred are better than those of our method, it must be borne in mind that these results are from training and testing on the same dataset while our results are from training the classifiers on a different dataset. It must be noted however that the prediction accuracies for mitochondrial proteins, which are notoriously difficult to predict, are significantly higher using our method than any of the other methods (85.3% as compared to accuracies between 56% and 68.2% for the other methods). The GO-FunD-PseAA predictor, whose classification performance on the Park and Kanehisa dataset is shown in Table [2](#T2){ref-type="table"}, has also been tested on the Reinhardt and Hubbard dataset. The predictor performs well on this dataset too and yields the highest total accuracy of 92.9% \[[@B5]\] using the rigorous leave one out cross validation test. However, we could not include these results in Table [4](#T4){ref-type="table"} since the results in \[[@B5]\] do not provide a subcellular location-wise breakdown of prediction performance. We have implemented our algorithm for predicting subcellular localizations as a web server which can be accessed at <http://pslip.bii.a-star.edu.sg/>. Conclusion ========== Protein subcellular localization has been an active area of research due to the important role it plays in indicating, if not determining, protein function. A number of efforts have previously used amino acid compositions as well as limited sequence order information in order to predict protein localization. In this work, we have developed a novel approach based on using multiple physicochemical properties. In order to use sequence order information, we divide the set of proteins into four different clusters based on their lengths. Within each cluster, proteins are mapped onto the lowest length in that cluster (50, 150, 300 and 450 for the four clusters). We then developed multiple binary classifiers for each cluster. For each protein, the output from each binary classifier was encoded as a binary bit sequence to form a meta-dataset. To predict the localization of a query protein, a similar binary sequence was generated based on the outputs of the different binary classifiers and the nearest neighbor to this protein was sought in the meta-dataset. We obtained significantly higher classification accuracies (93.1% overall and 92.1% local) for the Park and Kanehisa dataset. The prediction accuracies obtained for mitochondrial and extracellular proteins in particular are among the highest that have been achieved so far. The clustered approach was chosen to not only be able to include sequence order information beyond that of di-, tri-and tetra-peptide information but also to mitigate the effects of over-averaging. One of the problems we encountered was the small number of proteins of length greater than 1350. As a result, these were averaged down to a base length of 450 leading to a drop in accuracies for the 450 cluster. Obviously larger datasets, with more representative samples in the length range greater than 1350 might yield greater accuracies. Methods ======= Dataset ------- We used the protein sequences dataset^1^created by Park and Kanehisa \[[@B21]\]. The dataset consists of 7579 eukaryotic proteins drawn from the SWISS-PROT database and classified into twelve subcellular locations. The protein sequences were classified based on the keywords found in the CC (comments or notes) and OC (organism classification) fields of SWISS-PROT. Proteins annotated with multiple subcellular locations were not included in the dataset. Further, proteins containing B, X or Z in the amino acid sequence were excluded from the dataset. Finally, proteins with high sequence similarity (greater than 80%) were not chosen for inclusion. Table [1](#T1){ref-type="table"} summarizes the number of sequences in each of the twelve subcellular locations. For some of the locations such as cytoskeleton, there were too few sample sequences to achieve reliable training accuracies using SVM, the machine learning algorithm used in this work. Hence, we considered only sequences of type: chloroplast, cytoplasmic, extracellular, mitochondrial, nuclear and plasma membrane resulting in a dataset with 7106 eukaryotic protein sequences. Support vector machine ---------------------- The concept of Support Vector Machines (SVM) was first introduced by Vapnik \[[@B35],[@B36]\] and in recent times, the SVM approach has been used extensively in the areas of classification and regression. SVM is a learning algorithm which, upon training with a set of positively and negatively labeled samples, produces a classifier that can then be used to identify the correct label for unlabeled samples. SVM builds a classifier by constructing an optimal hyperplane that divides the positively and the negative labeled samples with the maximum margin of separation. Each sample is described by a feature vector. Typically, training samples are not linearly separable. Hence, the feature vectors of all training samples are first mapped to a higher dimensional space Hand an optimal dividing hyperplane is sought in this space. The SVM algorithm requires the solving of a quadratic optimization problem. To simplify the problem, SVM does not explicitly map the feature vectors of all the samples to the space H. Instead, mapping is done implicitly by defining a kernel function ![](1471-2105-6-152-i1.gif) between two samples with feature vectors ![](1471-2105-6-152-i2.gif) and ![](1471-2105-6-152-i3.gif) as: ![](1471-2105-6-152-i4.gif) where Φ is the mapping to the space H. For a detailed description of the mathematics behind SVM, we refer the reader to an article by Burges \[[@B37]\]. For the present study, we used the SVMlightpackage (version 6.01) created by Joachims \[[@B38]\]. The package is available online^2^and is free for scientific use. Multi class SVM --------------- A multi-class classification problem, such as the subcellular localization problem, is typically solved by reducing the multi-class problem into a series of binary classification problems. In the method employed here, called the 1-vs-1 method, a binary classifier is constructed for each pair of classes. Thus, a c-class problem is transformed into several two-class problems: one for each pair of classes (i, j), 1 ≤ i, j≤ c, i≠ j. We use the notation ![](1471-2105-6-152-i15.gif) to refer to both the binary classification problem of separating samples of classes iand jas well as the SVM classifier which is used to solve this problem. The classifier for the two-class problem ![](1471-2105-6-152-i15.gif) is trained with samples of classes iand j, ignoring samples from all other classes. Since SVM is a symmetric learning algorithm, the classifier ![](1471-2105-6-152-i15.gif) is the same as the classifier ![](1471-2105-6-152-i24.gif). Thus, for the purpose of classification, it is sufficient if we consider only those classifiers ![](1471-2105-6-152-i15.gif) for which i\<j Hence, we only construct classifiers for each pair of classes (i, j), 1 ≤ i, j≤ c, i\<jand there are c(c- 1) /2 such classifiers. To illustrate, if a classification problem involved three classes a, band c, the 1-vs-1 method would require construction of the binary classifiers ![](1471-2105-6-152-i16.gif) and ![](1471-2105-6-152-i17.gif). We term those classifiers which are trained to differentiate the true class of the test sample from other classes as relevantclassifiers while the remaining classifiers are termed irrelevantclassifiers. For example, using the three class example cited above, if the (as yet unknown) true class of a test sample is a, then the relevant classifiers would be ![](1471-2105-6-152-i18.gif) and ![](1471-2105-6-152-i19.gif) while the classifier ![](1471-2105-6-152-i17.gif) would be irrelevant. An unlabeled test sample is tested against all the c(c- 1) /2 binary classifiers and the predictions from each of these classifiers are combined by some algorithm to assign a class label to the sample. The design of the combining algorithm should be such that the predictions from the relevant classifiers gain precedence over those from the irrelevant classifiers. A simple voting schemeis one such algorithm that has been used earlier \[[@B27]\]. In this scheme, a vote is assigned to a class every time a classifier predicts that the test sample belongs to that class. The class with the maximum number of votes is deemed to be the true class of the sample. The prediction performance of the voting scheme approach relies on the assumption that the relevant classifiers for the unlabeled sample perform very well and the number of votes they cast in favor of the true class outnumber the number of votes obtained by any other class from the irrelevant classifiers. In practice, we found this not to be the case. Some of the relevant classifiers performed poorly and a wrong class frequently got the highest number of votes by virtue of many irrelevant classifiers voting for it. To solve this problem with combining classifier predictions, we replaced the voting scheme with a new classifier, called the meta-classifier. Described in \[[@B39]\], the meta-classifier works as follows: The set of all the two-class classifiers described above is first built using the available training samples. After this, each of the training samples is tested against all the binary classifiers (Figure [1(a)](#F1){ref-type="fig"}) and the class predictions are encoded in a binary (0 or 1) bit sequence (Figure [1(b)](#F1){ref-type="fig"}). We choose a bit sequence of length c^2^with one bit for each possible combination ![](1471-2105-6-152-i25.gif). All the bits in the sequence are initialized to zero. If a binary classifier ![](1471-2105-6-152-i20.gif) predicts the training sample to belong to class p, then the bit corresponding to the position (p,q) is set to 1; if the prediction is for class q, then the bit corresponding to the position (q,p) is set to 1. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Meta-data set construction.This figure shows, using two samples, how the meta-data set is constructed for a three class problem. Figure 1(a) lists the predictions obtained by testing the two samples against all the binary classifiers. Figure 1(b) lists the bit sequences corresponding to the obtained predictions. ::: ![](1471-2105-6-152-1) ::: Thus, after testing a sample with all the classifiers, we get an encoded representation of the classifier predictions in the form of a bit sequence. Since this sample is a training sample, it\'s true class (or label) is known. The true class and the bit sequence together constitute a meta-data instance derived from the training sample. The collection of all such meta-data instances derived from all the training samples is termed the meta-data set. Figure [1](#F1){ref-type="fig"} provides an illustration of this process. When an unlabeled test sample is presented for classification, it is first tested against all the binary classifiers (Figure [2(a)](#F2){ref-type="fig"}) and the class predictions are encoded in a bit sequence (Figure [2(b)](#F2){ref-type="fig"}) as explained above. Then, we seek an instance in the meta-data set whose bit sequence most closely matches this test sample\'s bit sequence. To search for such a matching sequence, we use a nearest neighbor approach with the distance between two sequences being the number of bits in which they differ. The Exclusive OR(XOR) operator can be used to count the number of differing bits. Table [5](#T5){ref-type="table"} shows the truth table for the XOR operator. When we carry out a XOR operation between the test bit sequence and a sequence from the meta-data set, we get another bit sequence. The number of 1-bits in this resulting sequence gives a measure of distance with a higher number implying greater distance between the sequences. Figure [2](#F2){ref-type="fig"} describes this procedure with an example. This distance calculation method is similar to the hamming decodingmethod described in \[[@B40]\]. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Meta-classification.The unlabeled sample is first tested against all the binary classifiers and Figure 2(a) shows the predictions obtained after such a test. Next, its bit sequence representation is constructed \[Figure 2(b)\]. The XOR operation is performed between this bit sequence and each of the sequences from the meta-data set \[Figures 2(c) & 2(d)\]. The distance between samples is the number of 1-bits in the XOR operation result. Since the distance of the unlabeled sample from sample 1 is less than that from sample 2, the unlabeled sample is assigned the label of sample 1, i.e. it is assigned to the class a. ::: ![](1471-2105-6-152-2) ::: ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Truth table for the Exclusive OR (XOR) operator. Thus, for example, 101 XOR 011 would be 110. ::: XOR 0 1 ----- --- --- 0 0 1 1 1 0 ::: After the sequence from the meta-data set that is closest to the test sequence is identified, we assign its known class label to the test sample. In case multiple (equidistant) sequences are found after the nearest neighbor search, resulting in a tie between two or more classes, we pick that particular class which would have got the maximum votes of the tied classes; the votes being counted according to the voting scheme described earlier. By implementing this meta-classification approach, we found an improvement in accuracy of 10% -- 15% over the voting method. Feature vectors --------------- Many previous efforts have used amino acid composition as the feature to determine protein subcellular localization. In these efforts, the feature vector corresponding to an amino acid sequence is typically a 20-dimensional vector with each element of the vector representing the frequency of occurrence of an amino acid in that particular sequence. As highlighted earlier, this approach leads to a complete loss of sequence order information. On the other hand, the averaging of physicochemical properties over the entire length of the protein sequence also results in a loss of sequence order information. We believe that sequences that are co-localized must share some similarity across certain physicochemical properties, regardless of their length. To overcome the shortcomings of earlier efforts, we employ a novel method of building feature vectors which is based on an idea first proposed in \[[@B41]\]. Consider an amino acid sequence of length L. Suppose we wish to use Mdifferent physicochemical properties in the feature representation of the amino acid sequence. Corresponding to each amino acid i, we build property vectors![](1471-2105-6-152-i21.gif) (1,\..., 20) where ![](1471-2105-6-152-i22.gif) (1,\..., M) is the vector of normalized values of the Mphysicochemical properties for the amino acid i. Then, for the sequence of length L, we concatenate the property vectors of each of the amino acids in the sequence in succession to get a vector of dimension L× M. ![](1471-2105-6-152-i6.gif) where ![](1471-2105-6-152-i7.gif) is one of ![](1471-2105-6-152-i5.gif) depending upon which amino acid iis present at location kin the sequence. While this method allows us to build feature vectors using any number of desired physicochemical properties, it results in vectors whose dimension is a function of the length of the amino acid sequence. One of the problems with using physicochemical properties averaged over the entire sequence length, in localization prediction efforts so far, has been the difference in lengths of the different protein sequences. Since SVM requires equal length feature vectors, this has always been a deterrent to utilizing sequence order information. Hence, we apply a local averaging process to scale all generated feature vectors to a standard dimension. Suppose we take our standard dimension to be Kand wish to construct a feature vector of this dimension for a protein sequence of length L. For now, let\'s assume that we are building amino acid property vectors using just one physicochemical property. So the feature vector ![](1471-2105-6-152-i8.gif), obtained by concatenating property vectors as described above, will also be of dimension L. We then sequentially group the feature vector\'s elements such that we end up with Kgroups or partitions of the vector. Withing each part, we take the average values of the constituent elements and then build a vector out of these averaged values. This operation is like constructing a mapping ![](1471-2105-6-152-i23.gif)and can be thought of conceptually as reducing the protein sequence of length Lto a standard length land here l= K. If we used Mphysicochemical properties instead of one, a conceptual scaling of a protein sequence of length Lto a standard length lis equivalent to a scaling of ![](1471-2105-6-152-i8.gif) (of length L× M) to a standard dimension of K= l × M. To do this operation, we partition the amino acid sequence of length Linto lnearly equal parts and then take the average value of the property vectors within these partitions. Figure [3](#F3){ref-type="fig"} shows an example where the feature vector for an amino acid sequence of length L= 6 is constructed. The example uses M= 2 physicochemical properties and the target standard dimension is of length K= 10. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Feature vector construction.In this example, the physicochemical properties chosen are hydropathyand hydrophobicity. Their property vectors ![](1471-2105-6-152-i5.gif) are concatenated in the order of occurrence of the residues in the sample protein sequence. The ![](1471-2105-6-152-i8.gif) vector thus obtained is scaled by averaging to the required target dimension. ::: ![](1471-2105-6-152-3) ::: There is a loss of sequence order information due to this averaging process and this loss is significant when scaling down proteins of very long lengths to a much shorter length. To minimize this information loss, we divided our dataset into clusters defining a base length for each cluster. This base length is equivalent to the conceptual protein sequence standard length l. Within each cluster, no sequence has a length less than land the length of the longest sequence is no greater than three times l. Initially, we built clusters using the base lengths as (10, 30, 90, 270, 810, 2430) but this resulted in an uneven population distribution of sequences across clusters that caused problems in the SVM training stage. We adjusted the cluster sizes and finally chose the base lengths as (50,150, 300, 450). For sequences of length less than fifty, we extended their length to fifty by suitably repeating the residues. For example, if the sequence AMKMSFof length six needs to be scaled to a length of ten, we would repeat residues to get AAMMKKMMSF. Parameter selection ------------------- Once the set of sequence clusters is built, we treat each of these clusters as a separate multi-class SVM problem. For each of the clusters, we build a set of binary classifiers as explained earlier. For the SVM algorithm itself, the kernel ![](1471-2105-6-152-i1.gif) is to be defined first. We chose the widely used Radial Basis Function (RBF)kernel which is defined as: ![](1471-2105-6-152-i9.gif) Further, the SVM optimization model employs a regularization parameter Cwhich controls the trade-off between the margin of separation (between positive and negatively labeled samples) and the error in classification. Thus, the process of parameter selection for the classification problem repeats over the set of binary classifiers per cluster and then for each of the clusters. There are three parameters that need to be ascertained: 1\. The set of Mphysicochemical properties to represent the amino acid sequences 2\. The value of γ for the kernel function 3\. The value of the regularization parameter C For the set of physicochemical properties, we used the Amino Acid index database \[[@B42]\] available at <http://www.genome.jp/dbget/aaindex.html>. An amino acid index is a set of 20 numerical values representing any of the different physicochemical properties of amino acids. This database currently contains 484 such indices. The process of selecting the parameters is carried out using the approach shown in Figure [4](#F4){ref-type="fig"} which essentially involves doing a parameter space search for the different possible combinations of sequence clusters, classifiers and parameters. We first determine the prediction performances that can be obtained by the different classifiers in the different clusters by building feature vectors with each of the 484 indices taken one at a time. During this search over the indices, we let SVMlightassign default values to Cand γ. Once this search is done, we pick the top five best performing indices to be the representative physicochemical properties (for that particular combination of cluster and classifier.) ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Parameter search.This shows the approach taken to find the best parameters for the different possible combinations of sequence clusters and classifiers. ::: ![](1471-2105-6-152-4) ::: We then build feature vectors using these top five indices and do a search over the Cand γ space looking for the best performing combination of these parameters. At the end of this parameter search, we obtain the best combination of amino acid indices, Cand γ for each classifier in each sequence cluster. We then look at which sequence cluster achieved the best overall prediction performance and pick the set of best parameters for classifiers in that cluster as the best set for all the clusters. The set of top five amino acid indices for each of the classifiers as found using this parameter search for the Park and Kanehisa dataset have been provided \[see [Additional file 1](#S1){ref-type="supplementary-material"}\]. Validation ---------- To assess the prediction performance of the proposed algorithm, a cross-validation test must be performed. The three methods most often used for cross-validation are the independent dataset (ID)test, the leave one out cross validation (LOO-CV)test and the N-fold cross validation (NF-CV)test \[[@B43]\]. Of the three, the LOO-CV test is considered to be the most rigorous and objective \[[@B44]\]. Although bias-free, this test is very computationally demanding and is often impractical for large datasets. Further, it suffers from possibly high variance in results depending on the composition of the dataset and the characteristics of the classifier. The NF-CV test provides a bias-free estimate of the accuracy \[[@B45]\] at a much reduced computational cost and is considered an acceptable test for evaluating prediction performance of an algorithm \[[@B46]\]. In NF-CV tests, the dataset is divided into Nparts with approximately equal number of samples in each part. The learning machine is trained with samples from N- 1 parts while the Nthpart is used as testing set to calculate classification accuracies. The learning-testing process is repeated Ntimes until each part has been used as a testing set once. Since the number of protein sequence samples in each class are all different, it is obvious that during training phase of a binary classifier, the number of training samples in the two classes will not be equal. If the SVM is trained on these unequally sized sets, the resulting classifier will be inherently biased toward the more populous class; it is more likely to predict a test sample to belong to that class. The greater the disparity in populations between the two classes, the more pronounced the bias is. It is difficult to prevent this bias in the training stage without adjusting more parameters on a per classifier level. To prevent this problem, we reduce the training set to an equisized set by randomly selecting msamples from the larger set; mbeing the size of the smaller set. To quantify the performance of our proposed algorithm, we use the widely used measures of Specificityand Sensitivity. Let Nbe the total number of proteins in the testing dataset and let kbe the number of subcellular locations (classes). Let ![](1471-2105-6-152-i10.gif) be the number of proteins of class iclassified by the algorithm as belonging to class j. The specificity, also called precision, for class imeasures how many of the proteins classified as belonging to class itruly belong to that class. ![](1471-2105-6-152-i11.gif) The sensitivity, also called recall, for class imeasures how many of the proteins truly belonging to class iwere correctly classified as belonging to that class. ![](1471-2105-6-152-i12.gif) We further define Total Accuracyto measure how many proteins overall were correctly classified. ![](1471-2105-6-152-i13.gif) It is expected that the most populous classes will dominate the total accuracy measure and a classifier biased towards those classes will perform well according to this measure, even if the prediction performance for the smaller sized classes is not good. Hence, we consider another measure termed Location Accuracyand defined as: ![](1471-2105-6-152-i14.gif) Location accuracy reveals any poor performance by an individual classifier by providing a measure of how well the classification works for each class of proteins. The definitions of Total Accuracy and Local Accuracy used here are equivalent to those used by Park and Kanehisa \[[@B21]\]. Authors\' contributions ======================= DS and GC designed and implemented the pSLIP algorithm along with the voting scheme. KL and AK conceived the study, participated in its design and supervised the process. AK and DS drafted the manuscript. All authors read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 This file lists the top five amino acid indices found by parameter search for each of the binary classifiers. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ The authors would like to acknowledge Tariq Riaz and Jiren Wang for their inputs during several helpful discussions.	PubMed Central	2024-06-05T03:55:59.914513	2005-6-17	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182350/", "journal": "BMC Bioinformatics. 2005 Jun 17; 6:152", "authors": [ { "first": "Deepak", "last": "Sarda" }, { "first": "Gek Huey", "last": "Chua" }, { "first": "Kuo-Bin", "last": "Li" }, { "first": "Arun", "last": "Krishnan" } ] }
PMC1182351	Background ========== Multiple sequence alignments are a crucial prerequisite for a diverse set of methods ranging from the reconstruction of phylogenies and the quantification of adaptive evolution, to the detection of conserved RNA secondary structures and protein motifs. In this contribution we present a novel alignment tool that has been developed primarily with the aim of improving multiple alignments of viral genomes. The genomes of RNA viruses often carry conserved RNA structures that perform vital functions during the life cycle of the virus. In many cases only a small part of the viral genome is functionally relevant at the level of RNA. Algorithms for the systematic search of conserved secondary structure patterns in large RNA, such as QRNA \[[@B1]\], alidot \[[@B2]-[@B4]\], RNAz \[[@B5]\], and RNAdecoder \[[@B6]\] are based on the observation that a small number of point mutations is very likely to cause large changes in the secondary structures \[[@B7]\]. Secondary structure elements that are consistently present in a group of sequences with less than, say, 95% average pairwise identity are therefore most likely the result of stabilizing selection, not a consequence of the high degree of sequence conservation. A comprehensive analysis of the genomic secondary structure features using alidot is available for Picornaviridae \[[@B8]\], Flaviviridae \[[@B9]\], and Hepadnaviridae \[[@B10],[@B11]\]. A preliminary survey across a large subset of the available sequence data was presented very recently \[[@B12]\]. The comparative approach to detect conserved RNA structures is obviously dependent upon high-quality multiple alignments: even a relative small number of alignment errors, which act like randomly placed mutations, will obscure the signals from consistent and compensatory point mutations and, hence, decrease the sensitivity of the RNA detection algorithms. While we eventually need an alignment of the genomic nucleic acid sequence, we observe that an overwhelming part of a viral genome codes for one or more proteins in one or more (overlapping) frames. In contrast to the protein sequences, which are often easily alignable, the sequence similarities are drastically reduced on the nucleic acid level due to the redundancies of the genetic code, see Fig. [1](#F1){ref-type="fig"}. It is desirable, therefore, to utilize the amino acid sequence when aligning codingnucleic acid sequences with higher sequence divergence. This is sometimes done by aligning protein sequences and subsequently back-translating to nucleic acids. In many cases, however, only a part of the sequence of interest is translated in vivo. In addition, there may be alternative proteins encoded in overlapping reading frames within the same nucleic acid sequences. Such overlapping reading frames are best known from viruses, including Hepatitis B \[[@B13],[@B14]\], Influenza \[[@B15]\], and Umbraviruses \[[@B16]\]. Recently, however, examples have been found in prokaryotic \[[@B17],[@B18]\] and even in eukaryotic genomes \[[@B19],[@B20]\]. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Example for the higher sequence heterogeneity on the level of nucleic acids. A hypothetical amino acid alignment on top represents a high degree of similarity. See the same sequences below on the level of nucleic acids with very low sequence similarity. The pairwise identity is only 33%, just slightly above the 25% identity expected for two random nucleic acid sequences. ::: ![](1471-2105-6-160-1) ::: In this contribution we describe a progressive alignment tool that implements an extended scoring scheme to incorporate simultaneously information on translation products in one or more (\[partly\] overlapping) reading frames which allows the user to combine all information from both the nucleic acid and amino acid sequences (if any). It makes explicit use of information about overlapping open reading frames, as they occur in many functional sequences, and allows arbitrary weighting for almost all scoring parameters, in order to gain more reliable scoring at certain regions of the nucleic acid sequences where additional amino acid scoring of underlying protein sequence can be utilized. Implementation ============== The codaln program implements Gotoh\'s algorithm for pairwise sequence alignments with affine gap cost functions \[[@B21]\]. The only change compared to this standard recursive algorithm for nucleic acid sequence alignment concerns the (mis)match score σ(x~i~, y~j~) of position ifrom sequence xwith position jfrom sequence y. Instead of taking into account only the nucleic acid letters, each position is considered as a vector containing the nucleic acid letter andthe amino acid that would arise from translation in each of the three possible reading frames providedthe frame in question is actually translated. Thus, we have σ(x~i~, y~j~) = β~0~σ~n~(x~i~, y~j~) + β~1~σ~p~(π\[x~i~x~i+1~x~i+2~\], π\[y~j~y~j+1~y~j+2~\]) + β~2~σ~p~(π\[x~i-1~x~i~x~i+1~\], π\[y~j-1~y~j~y~j+1~\]) + (1) β~3~σ~p~(π\[x~i-2~x~i-1~x~i~\], π\[y~j-2~y~j-1~y~j~\]) where π\[uvw\] denotes the amino acid corresponding to the codon uvw. Here β~k~= 1, k= 1, 2, 3, if both xand yare translated in the k-th reading frame, while β~k~= 0 if the k-th reading frame is not actually translated in either xor y(or if one chooses to ignore a particular reading frame). β~0~is the relative weight of the the nucleic acid match score, usually 1. In non-coding (untranslated) regions we therefore retain only the nucleic acid score. Fig. [2](#F2){ref-type="fig"} gives an example. Further, it is possible that one gives different weights for alternative reading frames, maybe dependent upon parameters such as preferred codon usage. Default is no preference. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Application of the scoring model to a hypothetical alignment. Note that there are no amino acid contributions in the right hand part of the example because of the single indel that causes a frameshift. For illustration we show BLOSUM62 scores and simple scores for nucleic acids and gaps rather than the rescaled default values (His/Gln has score 0). ::: ![](1471-2105-6-160-2) ::: The score model is much simpler than the one proposed by Hein \[[@B22],[@B23]\] and implemented in combat \[[@B24]\] and CAT \[[@B25]\]. In contrast to these approaches, which enforce gap lengths that are multiples of three in order to maintain the reading frame, codaln does not use special gap penalties at all. Instead, it relies on the match scores from the coding regions to guide the alignment back into the correct reading frame after a frameshift insertion or deletion. This results in an algorithm that is both faster and able to handle overlapping reading frames. In its current implementation codaln can deal with 18 different codon tables, including the standard genetic code (default), various non-canonical tables for certain groups of organisms, and 11 distinct codon tables for mitochondrial genomes. The codon tables link the nucleic acid triplets with their encoded amino acids. They are used both for an automatic search for start and stop codons and for translation in the scoring function; see Tab. [1](#T1){ref-type="table"}. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### 18 codon tables can be utilized by the program for linking the nucleic acid triplets with their corresponding amino acids. ::: option organism featuring this codon table --------- ------------------------------------------------------------ univ universal genetic code (default) acet Acetabularia ccyl Candida cylindrica tepa Tetrahymena, Paramecium, Oxytrichia, Stylonychia, Glaucoma eupl Euplotes mlut Micrococcus luteus mysp Mycoplasma, Spiroplasma mitocan canonical mitochondrial code mitovrt Vertebrates -- mitochondrial code mitoart Arthropods -- mitochondrial code mitoech Echinoderms -- mitochondrial code mitomol Molluscs -- mitochondrial code mitoasc Ascidians -- mitochondrial code mitonem Nematodes -- mitochondrial code mitopla Plathelminths -- mitochondrial code mitoyea Yeasts -- mitochondrial code mitoeua Euascomycetes -- mitochondrial code mitopro Protozoans -- mitochondrial code ::: The program furthermore provides a flexible scheme for modifying the scoring model. Both amino acid and nucleic acid scores can be either taken from built-in defaults or read in from parameter files. A number of scaling factors can be specified in order to determine the relative weights of nucleic acids and/or amino acids in all the different reading frames. Tab. [2](#T2){ref-type="table"} summarizes the most important defaults. The program reads sequences in Pearson\'s (FASTA) format, GenBank file format, ViennaRNA format as well as completely unformatted sequences in any combination. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Default scoring parameters (can be arbitrarily weighted or changed by user defined settings). ::: parameter default value ----------------------- ------------------------- protein scores BLOSUM62 ×50 nucleic acid scores identity 1000, else 300 gap open penalty -1500 gap extension penalty -300 Remark.The effect of the positive default score for nucleic acid mismatches is to reduce the influence of nucleic acid mismatches relative to the amino acid and gap scores. ::: The program uses the information about translated regions, if contained in the input file. Alternatively, codaln attempts to detect all theoretically possible open reading frames which have a user-defined minimal length. Exons and fragmented coding regions are joined, translated, and the resulting amino acid sequences are then used for the scoring function in addition to the nucleic acid sequences. The program can optionally regard a sequence as circular so that a coding region can wrap around the ends of the sequence and is still scored correctly. An intermediate output reports the structure of annotated and inferred exons and open reading frames both in a text and in PostScript format, Fig. [3](#F3){ref-type="fig"}. At this stage, the user can stop the process, edit the annotation file, and restart the alignment procedure with the modified annotations. The coding regions that are used for scoring can be automatically defined, user defined, modified, or eliminated. Before restarting the alignment process, codaln again provides a text and PostScript output summarizing the modified annotation. If necessary, this process can be repeated. Multiple alignments are built from the pairwise alignments using the same progressive scheme that is used e.g. by ClustalW \[[@B26]\]: A guide tree is inferred from the pairwise distances and determines the order in which profiles are constructed from alignments of two sequences, a sequence and a profile, or two profiles. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Reports on the annotated and inferred structure of the input sequences are automatically generated by codaln, respecting all user intervention. ::: ![](1471-2105-6-160-3) ::: The profile alignments respect the full model of both nucleic acid and amino acid (mis)match scores. In the present implementation, the sequences within a profile are unweighted; it would be straightforward, however, to include a weighting scheme that takes the relative distances of the sequences into account to reduce the weight of very similar sequences, as implemented e.g. in ClustalW. Results ======= More plausible alignments ------------------------- Not surprisingly, we observe that codaln multiple alignments of coding DNA sequences have a much larger fraction of gaps with a length divisible by three than ClustalW multiple alignments. This is the desired effect of including amino acid-based scoring contributions since it reduces biologically implausible frameshifts. In itself, this is of course not a direct evidence for real improvements of multiple nucleic acid sequence alignments, but it is a good indication that, in coding regions, codaln preferentially makes insertions and deletions at the protein level. Unfortunately, good hand-curated multiple alignments of partially coding sequences do not seem to be available, so that a systematic quantitative analysis (using, e.g., the BAliBASE tools \[[@B27]\]) could not be performed. Pairwise alignments of coding DNA sequences are barely distinguishable from those obtained with combat \[[@B24]\] provided the amino acid contributions dominate codaln\'s scoring function. We therefore concentrate on a qualitative assessment of codaln alignments in particular application contexts. Hox genes and their genomic neighborhood ---------------------------------------- Hoxgenes were first described in the fruitfly Drosophila melanogaster. They code for homeodomain containing transcription factors \[[@B28]\] and are characterized by a 60 amino-acid helix-turn-helix DNA binding homeodomain. This domain is highly conserved on the level of protein, but can be quite variable at the DNA level. Vertebrates, in contrast to all invertebrates examined, have multiple Hoxgene clusters that have arisen from a single ancestral cluster in the most recent common ancestor of chordates \[[@B29],[@B30]\]. This ancient cluster in turn evolved through tandem gene duplications from a more ancient hypothetical protohox cluster \[[@B31]\]. We applied both ClustalW and codaln to the genomic sequences at the Hox4locus. Exon 2of Hox4, which contains the homeobox, is highly conserved also on the level of nucleic acid, while exon 1has a well-conserved amino acid sequence but exhibits high variability at the nucleic acid level. The non-coding sequence in the intron and the flanking sequences are highly variable. Thus, this example is a hard test case for our approach. Fig. [4](#F4){ref-type="fig"} summarizes the gap lengths in the Hox4alignments. A comparison of the number of gaps with a length divisible by 3 with the other gaps of other lengths is a useful indicator whether coding regions are reasonably aligned: Base triplets preferentially should not be disrupted as amino acids within a protein sequence cannot be disrupted. In this example, codaln produces 436 gaps with a length divisible by 3 (ClustalW: 330) and 797 others (ClustalW: 1113). While codaln produces a significantly higher fraction of gaps that are a multiple of 3 and correctly aligns the coding sequences in both exons, ClustalW only treats exon 2correctly, which is highly conserved on the level of nucleic acids. The nucleic acid alignment for the more variable exon 1, in contrast, is much more divergent. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Relative distribution of gaps in an alignment of genomic Hox4sequences. The alignment is essentially gap-less in exon 2. ClustalW (above) returns a very poor alignment of exon 1in which gaps occur with a broad distribution. In contrast, codaln respects the coding region so that almost all gap lengths in this area are divisible by 3. ::: ![](1471-2105-6-160-4) ::: Conserved RNA secondary structures in Levivirus genomes ------------------------------------------------------- Virus genomes serve as an ideal test case for a procedure that makes explicit usage of information about (overlapping) coding regions to improve the resulting alignments. Improved alignments, as we shall see, increase the sensitivity of the detection of secondary structure elements. The members of the genus Levivirus infect eubacteria (Procarya). All members of the family Leviviridae (Levivirus and Allolevivirus) are ssRNA positive-strand viruses. The replication cycle includes no DNA stage. The virions are neither enveloped nor tailed with nucleocapsids that are isometric, 24--26 nm in diameter. The total genome length is 3466 up to 4276 nucleotides depending on type of strain. Most Levivirus species have four (partly) overlapping genes, while some exceptions exist which contain only three open reading frames \[[@B32],[@B33]\]. We have investigated 8 complete genomic sequences of the Levivirus genus: The Enterobacteria phages MS2, KU1, GA, and fr. Alignments of the genomic sequences were prepared using codaln and scanned for conserved RNA secondary structures using the alidot method \[[@B3]\]. The results are compared to earlier studies based on ClustalW alignments \[[@B10],[@B12]\]. The two different alignment processes produce results that seem to be similar at a first glance: The number of gaps and a visual interpretation of the quality of the alignment only does not already announce the significantly different results when further processing the alignments by alidot. Interestingly, the combination of codaln and alidot produces a weak signal at the basis of the Hogeweg mountain plot (see Fig. [5](#F5){ref-type="fig"}). ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Hogeweg mountain plots of conserved RNA structures in Levivirus genomes. Above: ClustalW, below: codaln. Colors indicate the number of consistent mutations: red 1, ochre 2, green 3, turquoise 4, blue 5; Saturated colors indicate that there are only sequences that are compatible to the structure prediction. Decreasing saturation of the colors indicates 1 or 2 non-compatible sequences. The thickness of the slabs is proportional to the average frequency of the base pair in the thermodynamic equilibrium. For further details see \[3\]. ::: ![](1471-2105-6-160-5) ::: Long range interactions, so called panhandlestructures, are known to play a role e.g. in the replication of Bunyaviridae \[[@B34]\] and Flaviviridae \[[@B35]\]. It will be interesting to see if the long-range interactions can be experimentally verified in Leviviridae as well. At the 5\'-terminal end of the Levivirus sequences we furthermore detect a short GC-rich hairpin(tetraloop) adjacent to an unpaired GGG element, see Fig. [6](#F6){ref-type="fig"}. This feature is probably the analogon to the recognition signal site for the RNA replicase in Alloleviviruses. This stem-loop-structure is well known and defined in Qβ(Allolevivirus). ::: {#F6 .fig} Figure 6 ::: {.caption} ###### The 5\'-terminal hairpin in Levivirus (left) is probably the analogon to the recognition signal site for the RNA replicase in Alloleviviruses which is well analyzed in Qβ(right). In Qβthe replicase amplifies RNA templates autocatalytically with high efficiency. This recognition element in Levivirus likely has a similar function. ::: ![](1471-2105-6-160-6) ::: The Qβreplicase amplifies RNA templates autocatalytically with high efficiency, and the recognition element, consisting of a hairpin and a short unpaired region at the 5\'-terminus, is essential for recognition \[[@B36],[@B37]\]. Discussion ========== Algorithms for the the automatic detection of biologically functional secondary structure elements, such as the ones used here, are dependent upon accurate alignments. Clearly, the quality of alignments can be enhanced by including additional biological information. In the case of codaln, we make use of the information on the coding properties of a nucleic acid sequence into the alignment process. We demonstrate this in the case of alignments of the Hox4genomic region which consists of non-coding regions and two coding exons, one containing the highly conserved homeodomain, while the other exon is poorly conserved on nucleic acid level. As expected, the quality of the alignment in the coding region can be increased significantly. Virus genomes can serve as an ideal test case for a procedure that makes explicit usage of information about various (overlapping) coding regions. Above we have seen that additional conserved secondary structure elements become detectable with the improved alignment. Leviviruses are, despite their short genome, a quite complex example. The sequences are at least in part highly divergent at the nucleic acid level, so that the information on the coding sequences in codaln significantly improves the quality of the alignment. Using codaln instead of a purely nucleotide-based alignment program, we found a putative recognition signal site, analog to the one for the RNA replicase in Alloleviviruses. Conclusion ========== The codaln program was specifically developed for applications to genomic sequences of RNA viruses with their partially overlapping reading frames and untranslated regions. The Hoxgene example shows, however, that codaln is also applicable to other partially coding sequence data. The recent discovery of ORFs that overlap with different reading directions \[[@B38]-[@B40]\] suggest to extend codaln to such cases as well. Our framework makes such an extension straightforward. Availability and requirements ============================= C source code and documentation may be downloaded from <http://www.bioinf.uni-leipzig.de/Software/> or <http://www.tbi.univie.ac.at/~roman/Codaln/>. Hox4 data sources ----------------- The Hox4sequences are taken from GenBank for Homo sapiens(HsA join(AC004080.2rc+AC010990 \[201-6508\]rc+AC004079 \[75001-end\]rc) \[125253 126761\], HsB NT\_010783 \[5306154 5309021\]rc, HsC NT\_009563 \[586220 584941\]rc, HsD NT\_037537 \[4224691 4225996\]), Mus musculus(MmA NT\_039343 \[3862302 3864022\]rc, MmB AC011194 \[114551 116043\], MmC NT\_028016 \[137212 139414\], MmD AC015584 \[164151 165456\]), and Morone saxatilis(MsA AF089743 \[29109 30386\]). For Danio reriothe sequences are retrieved both from the web server of the Danio rerio Sequencing Project\[[@B41]\] and GenBank (DrAa AC107365rc \[61628 62827\], DrBa AL645782.2 \[33537 35809\], DrCa ctg23.10700001-10870000 \[75679 77005\]rc, DrD ctg13407.19000-191000 \[61789 63580\]rc). rc = reverse complement; sequence intervals are listed in brackets. Authors\' contributions ======================= RS implemented the algorithm, RS and CF performed quantitative comparisons, ILH and PFS conceived this study. All four authors closely collaborated in writing the manuscript. Acknowledgements ================ This work is supported by the Austrian Fonds zur Förderung der Wissenschaftlichen Forschung, Project Nos. P-13545-MAT and P-15893, and the German DFGBioinformatics Initiative.	PubMed Central	2024-06-05T03:55:59.919534	2005-6-28	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182351/", "journal": "BMC Bioinformatics. 2005 Jun 28; 6:160", "authors": [ { "first": "Roman R", "last": "Stocsits" }, { "first": "Ivo L", "last": "Hofacker" }, { "first": "Claudia", "last": "Fried" }, { "first": "Peter F", "last": "Stadler" } ] }
PMC1182352	Background ========== Here we describe PROMI, a system to discover group-specific conservation characteristics in the amino acid distribution of profiles. For this we understand the sequences forming a general profile to be associated with a user-defined biological classification label, where the number of labels should be much smaller than the number of rows in the profile. In detail relations between profile columns and the applied group affiliation of the sequences forming the profile shall be investigated. The relations will be apparent by constituting significant amino-acid conservations, leading either to distinct amino acid consensus patterns in the analyzed groups or to knowledge about affinity between the groups \[[@B1]\]. To tackle this aim the mutual information (MI) is used as an interdependence measure of random variables X~i~and Y\[[@B2]-[@B5]\]. The interdependence between X~i~(in our case column of a profile X) and Y(here group affiliation) is understood as the knowledge one gains about Yif X~i~is known and vice versa \[[@B6],[@B7]\]. Small values imply small gain of knowledge between the variables, whereas high values point out a higher gain. The calculated MI-profile of the whole alignment consisting of all kgroups as well as all ![](1471-2105-6-164-i1.gif) pairwise profiles together with computed sequence logos finally allow conclusions regarding group-specific amino acid-positions where the distribution differ significantly and thus a group-discrimination on the basis of one profile-position is possible. Moreover the mean value of each pairwise MI-profile leads to formation of an elementary distance matrix D, where low MI-profile-mean-values state that the molecular similarity between groups of sequences is high opposed to higher MI-profile-mean-values with a higher molecular distance in the underlying groups. Further, by applying hierarchical clustering to D, a phylogenetic tree reflecting the distance between its constituents can be constructed. In the following we use \"class\" and \"classification\" synonymously with \"group\" and \"group affiliation\". Implementation ============== PROMI is implemented in Perl as a web based service running on an apache web server and available for free use. Depicted in Figure [5](#F5){ref-type="fig"} the selecting of matches relating to consensus sequences in PROSITE format \[[@B8]\] or given as a regular expression is performed by using the EXPASY ScanProsite tool \[[@B9]\], a Perl reference implementation for dealing with PROSITE motifs. The chosen instances of the motif were aligned using ScanProsite as well and the organism-specific origin was assigned by splitting up the NCBI none redundant protein database file \[[@B10]\] into species-specific \"proteome\" flatfiles. By upload of user-prepared sequences in FASTA format any other user-defined classification, alternative to the classification by organism identifier, can be applied. All calculations are implemented in the R environment \[[@B11]\]. To fulfil this, the RSPerl \[[@B12]\] and RSvgDevice \[[@B13]\] packages were used to embed R inside of Perl and to offer high order output in svg-format rather than the default png-format (svg output requires a plug-in for the web browser as provided by Adobe \[[@B14]\]). The computation of the sequence logos is done on the server-side by local utilisation of the Berkeley weblogo software. The Bioperl \[[@B15]\] module Bio::SeqIO is used to handle files of protein sequences. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Workflow of the web service PROMI. In step one the user specifies the motif and selects (may be user submitted) protein files. For sane results step two can be used to refine the selection derived by step one (by disabling false positive matches) and to limit the matches for a balanced ratio within the classes. Step three is the presentation of the complete and pairwise calculations in plots (compare Figure 2) as well as tables breaking down the multiple alignments into ratios of amino acids per position. ::: ![](1471-2105-6-164-5) ::: Results and discussion ====================== Sliding a window from column 1to nof the profile, as can be seen in Figure [1](#F1){ref-type="fig"}, leads to a MI-profile for the motif where low MI-values correspond to positions with a high degree of conservation among their constituent groups, whereas high MI-values correspond to positions with good discriminatory power in terms of the applied group affiliation. For evaluation of significance a label permutation test giving a p-value for each mean and deviation is performed. Therefore the MI-profile resulting from the given (unshuffled) multiple alignment is compared against histograms of 100 profiles derived by shuffling each column of the underling multiple alignment. By observing separation of shuffled and unshuffeld values in both histograms (synonymous to a p-value near zero) a positive supporting factor for significance is given. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Schematic multiple alignment with mfragments of length nbelonging to kuser defined groups. During the MI-profile-calculation used vectors X~i~and Yare marked in red, whereas X~i~behaves like a sliding window from left to right (i= 1\...n). X= (X~1~,\..., X~i~,\..., X~n~)and Yare derived by applying the enlarged Prosite motif PDOC00028 as a search-pattern to user-defined groups of protein sequences. By selecting fragments corresponding to the motiv a quasi multiple alignment was formed. In multitask the vector Yis formed from the user-given group labels. Here, to get a general idea, the alignment is colored and arranged per group-label of the vector Y. For demonstration conserved positions within the alignment (corresponding to the motif) are darkend. In case of pairwise alignments, only 2 groups would be stated. ::: ![](1471-2105-6-164-1) ::: Visualisation of mutual information profiles -------------------------------------------- Figure [2](#F2){ref-type="fig"} shows a computed MI profile at the top and the corresponding sequence logo aligned below. The profile is derived by applying the PROSITE motif PDOC00028 with the consensus pattern C-x(2,4)-C-x(3)-\[LIVMFYWC\]-x(8)-H-x(3,5)-H for the Zinc finger C~2~H~2~-type domain on several organism-specific \"proteomes\" using the tool PROMI. To take the structurally important linker region in zinc finger proteins of animals into consideration this pattern was enlarged at the N-and C-terminus by 7 amino acids. Each bar within the MI-profile corresponds to a specific column of the multiple alignment showing the obtained MI-value. As can be seen the MI-values differ over the alignment and the cysteines and histidines at positions 8, 13, 26, and 32 show the lowest MI-values (0.0) since these amino acid positions have to be conserved among all sequences. The blue line indicates the mean level, and two red lines correspond to the standard deviation. The row under the bar plot indicates the gaps\' frequency according to the alignment (white/grey ratio equals no gaps/gaps ratio). The histogram to the top right explicitly shows the MI standard deviation of 100 shuffled sample sets joined by the unshuffled standard deviation value marked in red whereas the histogram below (bottom right) shows the same for the mean (in blue). Thus, the histograms provide significance levels by showing the separation of shuffled and unshuffled MI-values. All MI-values larger or smaller than the marked red lines are therefore obvious significant positions with organism-specific differences within the amino acid distributions. At the bottom of Figure [2](#F2){ref-type="fig"} the sequence logo derived by the Berkeley weblogo tool \[[@B16]\] is depicted. Sequence logos are computed using entropy, and thus reveal the predominant amino acids available at specific positions within multiple alignments. Strong conservation between species is illustrated particularly lucidly at positions 8, 13, 26, and 32, which show very low MI-values. In contrast high MI-values point out positions with a high discrimination between the species, as visualized for positions 21, 24 and 25. The increased MI-values in the region of positions 33--38 coincides with the well described animal-specific TGEKP motif \[[@B17]\]. However, each position within the motif is valued differently -- especially position 38 -- suggesting a more differentiated view on this motif in a species-specific comparison. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### MI-profile for a multiple alignment of sequences derived from 6 organisms (top) and the corresponding Sequence logo (bottom). Each bar corresponds to one column in Figure 1, reflecting the value for the mutual information between the position X~i~and the a priori applied user-classification Yof the underlying multiple alignment. ::: ![](1471-2105-6-164-2) ::: Phylogenetic tree ----------------- In case of kgiven groups (see Figure [1](#F1){ref-type="fig"}) ![](1471-2105-6-164-i1.gif) pairwise MI-profiles can be constructed. The resulting MI-average of each pair provides a distance matrix interpretable as a species-specific distance measure (schematically given in Figure [3](#F3){ref-type="fig"}), which is sensitive enough to reflect relations in the underlying profiles. As can be seen in Figure [4](#F4){ref-type="fig"}, the resulting phylogenetic tree for the C~2~H~2~-domain of comprised species Arabidopsis thaliana, Homo sapiens, Mus musculus, Oryza sativa, Rattus norvegicusand Saccharomyces cerevisiaereveals the known phylogenetic relationship of the zinc finger domains of the observed organisms on the basis of group-and motif-specific differences. Of particular interest is the fact that S. cerevisiaeis clustered with animals rather than with plants. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Construction of distance matrix D(r,s)(b) in our case of k= 6 groups by calculating each mean of all ![](1471-2105-6-164-i2.gif) pairwise MI-profiles (a). rand sare placeholder of the user-applied classification labels, where D(r,s)is the distance of a group of sequences labeled rto a group of sequences labeled sand visa verse. By applying hierarchical clustering with the complete linkage method to Da phylogenetic tree reflecting the distances can be constructed (c). ::: ![](1471-2105-6-164-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Phylogenetic tree derived by using the mean of the pairwise computed MI-profiles as distances. The scale provides the absolute distance values of the build clusters. ::: ![](1471-2105-6-164-4) ::: Seemingly a very naive and rough approach disregarding protein steric structures as well as amino acid proximities the method is coherent. MI as a distance measure satisfies the four axioms of a metric: non-negativity, identity of indiscernibles, symmetry and triangle inequality \[[@B18]\]. Conclusion ========== As aforementioned the method we have outlined may not only be applied in a species-specific context, but may also be understood in terms of a phylogenetic identifier, gene expression values or any other desired classification by the user. Influence of potential negative thresholds, as for example false positive matches using PROSITE motifs for sequence search, or a possible bias because of highly differing amounts of matching sequence fragments per group, could be decreased with little extra effort. For this the inclusion of a user-prepared multiple alignment, derived by BLAST and CLUSTAL is conveniently possible. The number of selected instances per group can be balanced by limiting the fragments to a suitable number. Regardless of the abovementioned issues, the proposed method is a fast and convenient approach for the motif-specific analysis of hundreds of sequences derived by homology of ortholog or paralog gene and protein domain families. Methods ======= Multiple alignment and mutual information ----------------------------------------- Given a motif, describing sequence fragments of length n, the potential mmatching sequence fragments with length ncan be arranged as a multiple alignment as seen in Figure [1](#F1){ref-type="fig"}. Thereby each of the ncolumns corresponds to one position of the motif. Furthermore the mfragments belong to kgroups reflecting the user-applied classification (m\>\> k). By setting up this structure n m-dimensional vectors X~i~(with i= 1\...n) are constituted, each consisting of characters from the amino acid one letter code plus one additional letter for gaps. All X~i~combine as the matrix X= (X~1~, X~2~,\..., X~i~,\..., X~n~) reflecting the profile. Additionally an m-dimensional vector (Y) is formed from the classification consisting of a discrete set of group labels, where each label corresponds per row to the group affiliation of a sequence fragment in matrix Xat the same row. A convenient measure of correlation between two discrete random variables such as in our case X~i~and Yis the mutual information I(X~i~, Y) using the entropy and joint entropy respectively ![](1471-2105-6-164-i3.gif) and has been applied recently \[[@B1]-[@B5]\]. The entropy H(X~i~) is calculated using the number of occurrences of each character xwithin the vector X~i~of length m. The same is used for H(Y). The joint entropy H(X~i~, Y) is derived by concatenating the characters xand the group labels yand again calculating the likelihood of all possible combinations. Furthermore each MI value is multiplied by a factor giving the likelihood of gaps in each column. This heuristic approach was adapted from the C4.5 machine learning algorithm \[[@B19]\]. Iterating each column i(i= 1\...n) of the profile and calculating I(X~i~, Y) a so-called mutual information profile (MI-profile) can be established, incorporating all MI-values. Distance matrix and phylogenetic tree ------------------------------------- Given sequences affiliated to kgroups, ![](1471-2105-6-164-i1.gif) pairwise orderless combinations of groups can be formed. By calculating MI-profiles for all pairwise combinations so called pairwise MI-profiles can be created. As one potential measure the (k\*k)-dimensional distance matrix Dis formed up by calculating the mean over all columns of each pairwise MI-profile, schematically shown in Figure [3](#F3){ref-type="fig"}. This mean describes the distance of amino acid conservation between the groups of sequences. In the case of our study these groups are species-derived, but can also be of any other classification order. Given rand sas one possible combination of group labels ![](1471-2105-6-164-i4.gif) can be calculated. (X,Y)~r,\ s~are all these sequence fragments belonging either to group ror to group s. Eventually, the phylogenetic tree reflecting the distances can be constructed by applying hierarchical clustering (complete linkage clustering) to D. Availability and requirements ============================= • Project name: PROMI • Project home page: <http://promi.mpimp-golm.mpg.de/> • Operating system: server side Linux; client side platform independent • Programming language: Perl, R • Other requirements: EXPASY ScanProsite tool, R, Bioperl, RSPerl, RSvgDevice, Berkeley weblogo software • License: GNU GPL • Any restrictions to use by non-academics: no The source code of the web service script and the R script are available upon request. Other used software components are available at the according sources. Authors\' contributions ======================= JH assembled and developed the software components, conceived the idea of interpreting the MI as distance measure and participated in writing and drafting the manuscript. NK provided protein sequences, computed the biological results and drafted the manuscript. WW conceived the biological application of analysing the zinc finger C~2~H~2~-type protein domain. Further he interpreted the results and participated in the design of the study and drafted the manuscript. JS contributed to the conception of this study, participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.	PubMed Central	2024-06-05T03:55:59.921947	2005-6-29	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182352/", "journal": "BMC Bioinformatics. 2005 Jun 29; 6:164", "authors": [ { "first": "Jan", "last": "Hummel" }, { "first": "Nima", "last": "Keshvari" }, { "first": "Wolfram", "last": "Weckwerth" }, { "first": "Joachim", "last": "Selbig" } ] }
PMC1182353	Background ========== The Tat (Twin arginine translocation) pathway, which operates in parallel to the well characterized Sec export pathway, has recently been discovered in bacteria \[[@B1]-[@B3]\]. Substrates of the Tat pathway are often redox cofactor binding proteins which acquire their cofactors, and therefore fold in the cytoplasm. Thus, in contrast to protein export via the Sec pathway, Tat substrates are folded prior to export \[[@B1],[@B4]\]. Indeed, there is good evidence to suggest that the Tat pathway has the ability to recognize the folded state of a substrate protein and to reject unfolded proteins \[[@B5],[@B6]\]. Proteins entering the Tat pathway have signal peptides with a tripartite structure that is much like classical Sec signal peptides and indeed are probably also cleaved by leader peptidase \[[@B7]\]. However, in contrast to Sec signal peptides, a striking twin-arginine motif is found at the border between the n- and h-regions of the Tat signal peptide. A consensus sequence for this twin-arginine motif has previously been defined as (S/T)RRxFLK \[[@B1]\], where the two consecutive arginines are invariant (see Figure [1](#F1){ref-type="fig"}). With an average length of 37 amino acids, Tat signal peptides are significantly longer than classical Sec signal peptides. (The extended length is usually found in the n-region). In addition, the h-region of Tat signal peptides has a lower average hydrophobicity than classical Sec signal peptides \[[@B8]\]. Tat signal peptides are not limited to the bacterial domain, and they have also been found in Archaea \[[@B9]\] and in plants where they mediate Tat-dependent translocation across the chloroplast thylakoid membrane \[[@B10]\]. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Tat signal peptides.Sequence logo of the positive training set aligned at the two consecutive arginines. None of the variant forms of Tat signal peptides were included. Sequence conservation is shown in bits and constructed as previously described \[29\]. ::: ![](1471-2105-6-167-1) ::: Although the twin-arginine sequence motif has previously been defined, some variations of this may still be accepted by the Tat machinery. For example, the tetrathionate reductase from Salmonella entericawas found to lack one of the (previously thought) invariant arginines of the twin-arginine motif \[[@B11]\]. Moreover, analysis of the E. colipre-propenicillin amidase Tat signal peptide showed that it targets the Tat pathway. This signal peptide deviates from the consensus by having an asparagine between the two arginines in the twin-arginine motif \[[@B12]\]. On several occasions, mutational analysis has shown that the two consecutive arginines are not absolutely required for export via the Tat translocon. Either arginine of the E. coliSufI protein could be substituted by lysine without blocking Tat-dependent transport, although the rate of transport was reduced \[[@B13]\]. Substituting both arginines for lysine, or the first arginine for alanine, blocked transport completely. A similar result was found by Ize et al.\[[@B14]\], working with the E. coliTorA Tat signal peptide fused to green fluorescent protein (GFP) \[[@B14]\]. Buchanan et al.\[[@B15]\] found that the first arginine of the E. coliTorA Tat signal peptide could be substituted for lysine without blocking Tat-dependent transport \[[@B15]\]. Also using a TorA-GFP construct, DeLisa et al.\[[@B16]\] found that the second arginine could be substituted not only by lysine but also by asparagine or glutamine without blocking export. Numerous studies have used regular expressions in combination with SignalP for identification of putative Tat signal peptides. The SignalP method for prediction of Sec signal peptides \[[@B17],[@B18]\], can to some extend correctly predict Tat signal peptides; although the h-regions of Tat signal peptides tend to be less hydrophobic than classical Sec signal peptides and SRP signal sequences. However, accuracy by the SignalP approach for prediction of Tat signal peptides is low, and the position of the predicted cleavage site is often wrongly assigned. This is probably due to the extended length of the Tat signal peptide and the fact that frequently the c-regions of Tat signal peptides contain basic amino acids \[[@B8]\]. Two different targeting pathways converge at the Sec translocon. Proteins carrying SRP (signal recognition particle) signal sequences are targeted via SRP to the Sec machinery, mostly for integration into the membrane. Proteins carrying Sec signal peptides are translocated across the bacterial membrane and the signal peptide is cleaved off in order to release the mature protein. Discrimination of Sec and SRP signals is apparently based on the hydrophobicity of the signal sequence itself. Increasing the hydrophobicity of a Sec signal peptide reroutes it to the SRP targeting pathway \[[@B19],[@B20]\]. Recently, a Perl program for prediction of Tat signal peptides, TATFIND, based on regular expressions was developed \[[@B9],[@B21]\]. In addition to regular expressions, TATFIND uses a hydrophobicity measure in a rule based classification scheme. Because of a limited number of variant Tat signal peptides (not containing two consecutive arginines), the program was designed to ignore these. Therefore it is not capable of identifying the variant Tat signal peptides unless the user possesses some programming experience for changing the TATFIND program code. In this report we describe a new method, termed TatP, for the identification of Tat signal peptides. TatP differs from the previously described TATFIND because it integrates pattern matching and machine learning. Comparison of TatP with TATFIND on various independent test sets indicates that TatP has slightly more false negative predictions than TATFIND. Nevertheless, TatP generates far less false positive predictions than TATFIND. Implementation ============== Data set extraction ------------------- All sequence data were extracted from Swiss-Prot release 42.0 \[[@B22]\]. We extracted putative Tat signal peptide sequences with the database cross-reference identifier (DR: TAT\_signal\_seq) from the TIGR database of protein families (Ac: TIGR01409). The TIGR family of Tat signal peptides is based on predictions using a hidden Markov model. Of the 117 sequences extracted matching the above criteria, twelve were found by comments in the \'CC -!- PTM\' line to have experimental evidence for utilizing the Tat export pathway and were removed from the positive training set into an independent test set (see below). Due to limited amount of data, the positive training set was homology partitioned, thereby bringing the most homologous sequences together into 5 different subsets <http://www.cbs.dtu.dk/services/TatP>. 1178 cytoplasmic sequences all carrying two consecutive arginines (RR) within the first 30 amino acids were extracted from Swiss-Prot. Following redundancy reduction, a resulting set of 462 cytoplasmic sequences together with 62 Gram-positive and 70 Gram-negative Sec signal peptides randomly chosen from the SignalP 3.0 dataset without the \'Tat\_signal\_seq\' identifier, were used as a negative training set. The negative training set was homology reduced using the same scheme as for SignalP \[[@B17],[@B23]\]. In the negative training set we do not distinguish between Sec signal peptides or SRP signal sequences. Twelve Tat substrates (DMSA\_ECOLI, MBHS\_ALCEU, MBHS\_ALCHY, MBHS\_AZOVI, MBHS\_ECOLI, MBHS\_OLICA, MBHS\_WOLSU, MBHT\_ECOLI, OPGD\_ECOLI, PHNS\_DESVM, PHSS\_DESBA, TORA\_ECOLI) were removed from the positive training set into an independent test set, together with two additional sequences from a mutation study of TorA \[[@B8]\]. This test set of known Tat substrates is biased, as most of these proteins function as the small subunit of \[Ni-Fe\] hydrogenases. Therefore, we also included additional 23 Tat substrates from E. coliin the independent test set, having a total of 35 known Tat substrates. A curated set of E. coliTat signal peptides can be found at <http://www.jic.bbsrc.ac.uk/staff/tracy-palmer/signals.htm>. Furthermore, a test set of 15 sequences all with putative Tat signal peptides from B. subtilis, was found by literature studies \[[@B24],[@B25]\]. Additionally, we have used the remaining 713 bacterial cytoplasmic sequences (removed by homology reduction) all carrying two consecutive arginines (RR) within the first 30 amino acids of each sequence as an evaluation set, as well as a test set of 632 transmembrane proteins also carrying two consecutive arginines within the initial 30 amino acids. Regular expressions ------------------- Perl syntax regular expressions are allowed as filtering of potential Tat signal peptides. Furthermore, this assists the user in identifying the position of a potential Tat signal peptide motif. A regular expression of \'RR. \[FGAVML\] \[LITMVF\]\', where \'.\' means any amino acid, is used by default and this regular expression covers 97% of the sequences found in the training set. This pattern was extracted from an ungapped multiple sequence alignment of the positive training set where the position of the two arginines was fixed. Neural network architecture --------------------------- We used two different neural networks for coping with the Tat signal peptide prediction problem. One network for recognition of the cleavage site, and one network for determining whether a given amino acid belongs to the Tat signal peptide or not (discrimination). A more thorough description of the neural network has been presented in previous papers \[[@B17],[@B23]\]. Performance evaluation of the prediction method was carried out on different data sets. During training of the method, we used the same measure of performance as used in SignalP i.e.cross-validation on the training set. This means that the data set was split into five equally sized parts followed by training on four and tested on one for a total of five times until all sequences independently had been used for training and testing, respectively. Furthermore, we evaluated the method on various independent test sets as described above. The cleavage site network was based on an asymmetric window of 19 positions to the left and 3 positions to the right of the cleavage site. The cleavage site network had 2 units in the hidden layer. The discrimination network performs best using a symmetric or nearly symmetric window. The window size used in this work was symmetric covering 31 positions. The discrimination network had 3 units in the hidden layer. Composition and positioning information was included in the neural network as previously described \[[@B17]\]. Description of scores --------------------- Five different scores are presented in the output of TatP. Two scores, the S-score and the C-score, are generated by the discrimination and cleavage site network and are calculated for each position in the amino acid sequence. The positioning of the predicted cleavage site is indicated by the Y-score. The Y-score is defined as, ![](1471-2105-6-167-i1.gif) where Δ~d~S~i~is the difference between the average S-score of d positions before and d positions after position i: ![](1471-2105-6-167-i2.gif) The mean S-score is calculated as the average of the S-score in the predicted signal peptide region separated by the Y-score for maximal cleavage site prediction. As for SignalP version 3.0, the best discrimination of signal peptides versus non-signal peptides, is calculated as an average of the mean S-score and the maximal Y-score. We call it the D-score. If the D-score is above the defined cutoff value and a Tat signal peptide motif is found the sequence is nominated as a Tat substrate. Results and Discussion ====================== Performance evaluation of TatP ------------------------------ The TatP prediction server is a regular expression and neural network based tool that can be used to predict the presence of bacterial Tat signal peptides. The user is allowed to independently specify a preferred consensus pattern by Perl regular expressions as a filtering of input sequences. As default, we have implemented a complex consensus pattern which covers 97% of putative Tat signal peptides in the training set. The used regular expression is generated from the training data and can be identified in 103 of the 105 positive training sequences. The reason for excluding two sequences was to generate a consensus pattern that was as strict as possible. We allow the user to change the consensus pattern before running the entire prediction, which allows for large flexibility of the method. In eleven of the 105 positive training sequences, TATFIND did not identify a Tat signal peptide motif. The strength of allowing the user to choose a preferred consensus pattern is obvious when it comes to prediction of variant Tat signal peptides, which under default conditions are discarded. Our method uses a regular expression for identifying putative Tat substrates and a neural network for coping with the less hydrophobic h-region of the Tat signal peptide. The h-region is generally less hydrophobic in Tat signal peptides than classical Sec or SRP targeting sequences. A numeric and graphic output is generated to assist the user to interpret borderline cases and localize the predicted Tat signal peptide cleavage site (See Figure [2](#F2){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Server output.Graphical output from the prediction server. Here is shown the prediction results from SUFI\_ECOLI, a known Tat substrate. The S-score indicates whether an amino acid belongs to a Tat signal peptide or not. C- and Y-scores indicate the positioning of a potential cleavage site. ::: ![](1471-2105-6-167-2) ::: For evaluation of TatP, we have tested the performance of our method in several different ways. We have evaluated the method on twelve experimentally verified Tat signal peptides extracted from Swiss-Prot together with a curated set of E. coliTat signal peptides. Furthermore, we have analyzed sequences of TorA (E. coliTMAO reductase) where the wildtype signal peptide targets a passenger protein to the Tat pathway, and where mutational analysis shifted the targeting of the protein to the Sec pathway. We also examined whether the method is able to discriminate the Sec signal peptides emerging from the generation of the training set for SignalP version 3.0 from Tat signal peptides \[[@B17]\]. A proteomic study on proteins secreted in Bacillus subtiliswas also used for validation of the TatP method \[[@B24],[@B25]\]. Moreover, we also comment on a few known cases of the variant Tat signal peptides, as well as prediction on transmembrane helices. Discrimination between Sec and Tat signal peptides -------------------------------------------------- Numerous studies have used SignalP for prediction of Tat signal peptides. To some extent, SignalP is able to correctly classify Tat signal peptides, although the accuracy is lower than the TatP method presented here. Of the 35 known Tat substrates used in this study, only 83% receive a positive prediction by SignalP, even though they all have the consensus pattern. This shows that the SignalP can be used for Tat signal peptide prediction, although a more accurate method is preferable. The method presented here -- TatP -- has a higher accuracy than the approach mentioned above. The TatP method predicts 91% of the 35 known Tat substrates as having a Tat signal peptide. This is comparable to TATFIND which positively predicted 97% of these. Unfortunately, three known Tat substrates were not correctly identified by TatP, PHSS\_DESBA, YAGT\_ECOLI and FHUD\_ECOLI. TATFIND was not able to correctly classify YCDO\_ECOLI where it found no consensus motif for a Tat signal peptide. In addition to the classification prediction, TatP also reports a potential cleavage site for the Tat signal peptide (see Figure [2](#F2){ref-type="fig"}), a feature which is not reported in the output by TATFIND. Unfortunately, not all of the 35 known Tat signal peptides have experimentally verified cleavage sites. Merging information from Swiss-Prot and our curated E. colidataset we expect to have a correct cleavage site annotation in all known Tat substrates. When using the merged annotations, the TatP method has a correct cleavage site prediction in 84% of true-positive predicted Tat substrates. Assessment of whether TatP and TATFIND were able to discriminate Tat signal peptides from Sec signal peptides was carried out on the training set for SignalP version 3.0. Additionally, we evaluated the performance of TatP and TATFIND on 713 cytoplasmic proteins carrying two consecutive arginines located within the initial 30 amino acids. By searching these datasets with the regular expression pattern alone, we find that the false positive rates are 4% and 7% for Gram-positive and Gram-negative proteins, respectively. For cytoplasmic proteins with two consecutive arginines within the initial 30 amino acids, the false positive rate of the regular expression alone is 15%. TATFIND greatly reduces the amount of false positive predictions by including a rule-based classification scheme in addition to regular expressions, resulting in a false positive rate of 1.4%. In TatP we apply a neural network to identify the less hydrophobic stretches of the h-region in Tat signal peptides and the false positive rate is lowered to only 0.4% (see Table [1](#T1){ref-type="table"}). Indeed, this shows the power of the neural network, which is capable of discriminating between the hydrophobicity of Sec/SRP and Tat targeting signals. In the Gram-positive test set, only slightly more false positives were found by TatP than by TATFIND. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### False positive predictions. The table shows the rate of false positive predictions on Gram-positive, and Gram-negative proteins which are secreted via the Sec-dependent pathway. Furthermore is shows cytoplasmic proteins and transmembrane helix proteins, both carrying two consecutive arginines within the initial 30 amino acids. Used methods for prediction are regular expression (Regex), TATFIND and TatP (regular expressions in combination with a neural network (NN)). ::: Method Gram+ Gram- Cyt. (RR) Transmem (RR) ----------------- ----------- ----------- --------------- ------------------- Regex 4% 7% 15% 23% TATFIND 1% 6% 1.4% 7.0% TatP (Regex+NN) 3% 5% 0.4% 3.5% ::: It should be mentioned that some of these false positive predictions by TatP and TATFIND might represent true Tat signal peptides in the database. The used Gram-negative and Gram-positive data sets were extracted from Swiss-Prot before Tat signal peptides were annotated as such. Moreover, many of these sequence entries were entered in the Swiss-Prot database even before the Tat export pathway was discovered, thus a few Tat signal peptides might be erroneously annotated. Positive amino acid residues have a topological effect on transmembrane helices \[[@B26]\]. It is interesting to see whether TatP and TATFIND are able to discriminate transmembrane helices from Tat signal peptides. 632 proteins carrying transmembrane domains and also carrying two consecutive arginines within the initial 30 amino acids were analyzed. We found 22 proteins to be predicted as having Tat signal peptides, whereas TATFIND identified 46. Many of these were indeed metalloproteins but had a transmembrane domain annotation instead of a Tat signal peptide annotation. The reasons for this may be the same as described above. Rieske proteins remain anchored to the membrane by an uncleaved Tat signal sequence in thylakoids \[[@B27]\], which is also believed to be the case in prokaryotes since the Fe-S cluster of the protein must reach the trans-side of the membrane. The exact amount of true Tat signal peptides in this transmembrane data set is unknown. Nevertheless, it shows the power of TatP to identify putative Tat-substrates which are erroneously annotated as being transmembrane helices. TATFIND seems to over-predict the transmembrane proteins having two consecutive arginines, as some of the 46 predicted proteins are chaperones, which are unlikely to carry a Tat signal peptide. Prompted by the lack of experimental evidence for a large number of Tat substrates, we looked at 22 bacterial genomes, where genes encoding components of the Tat translocation system failed to be identified on the basis of homology \[[@B21]\]. Ideally, no Tat substrates should be found in these genomes by any Tat signal peptide prediction method. In these 22 bacterial genomes, TATFIND identified 20 Tat substrates whereas TatP identified only 4. Cristóbal et al.\[[@B8]\] showed that Tat signal peptides are generally less hydrophobic than both Sec signal peptides and SRP signal sequences. They demonstrated that the Tat signal peptide of the E. coliTMAO reductase (TorA) could be converted into a Sec signal peptide without changing the RR motif, but instead replacing by the hydrophobic h-region with a more hydrophobic region taken from a Sec signal peptide. Both TorA signal peptide mutants and the wildtype Tat signal peptide carry the consensus pattern. According to the hydrophobicity discrimination from the neural network, the wildtype TorA is correctly predicted to be secreted via the Tat pathway. The TorA \[10:10\] signal peptide mutant, which is more hydrophobic than the wildtype but having the same length, is falsely predicted positive, while the TorA \[19:10\] signal peptide mutant, which is both more hydrophobic and shorter, is correctly predicted negative. Similar results are found for TATFIND. Tat signal peptides in B. subtilis ---------------------------------- Jongbloed et al. \[[@B24]\] compiled a list of 69 B. subtilisproteins with potential Tat signal peptides. In the light of the above-mentioned mutation studies which showed that the first arginine in some cases could be substituted by a lysine without loss of Tat-dependence, they considered both RR- and KR-motifs. All sequences with an RRXΦΦ or KRXΦΦ (where Φ denotes a hydrophobic residue) followed by a hydrophobic region were selected from the genome sequence. Subsequently, the authors looked for the selected proteins in the B. subtilissecretome. Strikingly, only one of the 69 proteins was actually found to be Tat-dependent, the phosphodiesterase PhoD. It is positively predicted by TatP. 13 other proteins were found to be secreted, but were also secreted by a tatdeletion mutant, indicating Sec-dependence. All of these 13 proteins carry a sequence matching the consensus motif, but after applying the neural network for analyzing the hydrophobicity of these signal peptides all 13 receive a negative and correct prediction by TatP. Three of the 13 proteins, LipA, WapA and YolA, were furthermore shown to be dependent on SecA \[[@B24]\]. TATFIND made one false positive prediction, but was also capable of correctly predicting twelve Sec-dependent proteins not to carry a Tat signal peptide. The above mentioned sequences were all predicted to have a Sec signal peptide using SignalP version 3.0 (without truncation), except for PhoD (See Table [2](#T2){ref-type="table"}). Thus, in the case of B. subtilis, putative Tat signal peptides were all correctly classified as being Sec signal peptides by SignalP and none of them received a positive prediction by TatP, showing the power of regular expression based searches also incorporating the neural network for hydrophobicity discrimination. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Putative Tat substrates in B. subtilis. The table shows the predictions by SignalP version 3.0, regular expressions, TATFIND and TatP on B. subtilisproteins which have a putative Tat signal peptide motif, but nevertheless are secreted via the Sec-dependent pathway. Only PhoD and YwbN are secreted via the Tat secretion pathway. ::: Protein Regex SignalP TATFIND TatP ------------- ----------- ------------- ------------- ---------- LipA Yes Yes No No AbnA Yes Yes No No BglC Yes Yes No No BglS Yes Yes No No LytD Yes Yes No No OppA Yes Yes No No PbpX Yes Yes No No WapA Yes Yes No No WprA Yes Yes No No YdhF Yes Yes No No YfkN Yes Yes Yes No YhcR Yes Yes No No YolA Yes Yes No No PhoD Yes No Yes Yes YwbN Yes No Yes Yes ::: Very recently, a second Tat-dependent substrate was identified in B. subtilis. YwbN was found to be secreted in a strictly Tat-dependent manner \[[@B25]\]. YwbN is positively and correctly predicted by TatP. In the B. subtilisgenome, TATFIND predicts seven Tat signal peptides \[[@B21]\], while TatP with the default regular expression predicts five. If we follow Jongbloed et al.\'s idea and allow KR- as well as RR-motifs (but keep the rest of the regular expression unchanged), the number of positive predictions is ten. The exact number of Tat substrates in this organism, however, must await experimental verification. Variant Tat signal peptides --------------------------- TatP without the regular expression (or with a variant regular expression) could potentially predict variant Tat signal peptides without the canonical twin arginines. The TtrB subunit of Salmonella entericatetrathionate reductase carrying a lysine instead of an arginine in the first position of the Tat signal peptide motif (Genbank AC: NP\_805056) \[[@B11]\], was correctly and positively predicted by TatP if the regular expression was expanded with a lysine in the first position. TATFIND does not directly allow variations in the used regular expression (consensus motif), although it can be changed if one has basic programming skills. Thus, the TtrB subunit receives a negative prediction by TATFIND. Moreover, the mutant Tat signal peptides generated by Stanley et al.\[[@B13]\], Buchanan et al.\[[@B15]\], Ize et al.\[[@B14]\], and DeLisa et al.\[[@B16]\] are easily predicted by TatP if the user modifies the regular expression to allow for the relevant substitutions in the RR positions. However, the naturally occurring variant Tat signal peptide reported by Ignatova et al.\[[@B12]\] was unfortunately not predicted positively by TatP. We did not specifically investigate Tat signal peptides fused to heterologous passenger proteins \[[@B5],[@B28]\]. However, we expect the majority of these fusion proteins to obtain a correct prediction (with the caveat that the passenger protein is competent for export by the Tat pathway) as the prediction method mainly uses the N-terminal region of the protein sequence. Database annotations -------------------- As described above, errors can enter a database due to lack of knowledge. Many Tat signal peptides are annotated as Sec signal peptides, simply because the corresponding sequences entered the database before any knowledge of the Tat secretion pathway. From release 42.0 to 44.0 of Swiss-Prot many new Tat signal peptides have been annotated using the TIGR HMM (see Methods and Materials). Even though Swiss-Prot is manually curated, errors have been found regarding annotation of Tat signal peptides. It is doubtful whether the following Swiss-Prot entries carry Tat signal peptides, as the Tat secretion pathway so far has not been identified in eukaryotic ER membrane translocation. We have found the following six eukaryotic sequences carrying the TIGR Tat signal peptide identifier (TAT\_signal\_seq), which did not have associations with chloroplast thylakoid membrane trafficking, CLC7\_HUMAN, CLC7\_MOUSE, CLC7\_RAT, CLCC\_ARATH, CO1B\_HUMAN and PSP1\_ARATH. The above mentioned cases seem to be false positive predictions made by the TIGR HMM. Conclusion ========== The TatP method presented here integrates regular expressions and a neural network based method for prediction of twin-arginine (Tat) signal peptides. Our study shows that regular expressions alone are not capable of correctly identifying Tat signal peptides. The further addition of a neural network aids correct classification of Tat signal peptides and Sec signal peptides. Although the TatP method only slightly outperforms TATFIND, the method presented here is far superior to a combination of simple pattern matching and classical signal peptide prediction methods. Thus, this method is highly useful as it is publicly available as an easy to use online prediction server. In the area of investigating variant Tat signal peptides, no programming skills are needed. We believe that the TatP prediction method will serve as a nice complement to the SignalP method for prediction of both Tat and classical Sec signal peptides. Availability and requirements ============================= The TatP prediction server is available at <http://www.cbs.dtu.dk/services/TatP/> and the SignalP server for prediction of classical Sec signal peptides is available at <http://www.cbs.dtu.dk/services/SignalP/>. The TatP prediction method is available for the IRIX, SunOS and Linux platforms and is dependent on a range of standard Unix tools. A license for both academic and non-academic users can be obtained by following the instructions found on the above web page. Authors\' contributions ======================= JDB carried out sequence retrieval, neural network training and optimization and drafted the manuscript. HN provided input on the manuscript and method. DW tested the robustness of the method. TP provided a curated E. colidataset and wrote parts of the biological background. Finally, SB provided general inputs and improvements to the manuscript. Acknowledgements ================ This work was supported by grants from the Danish National Research Foundation, the Danish Natural Science Research Council, the Danish Center for Scientific Computing. TP is funded by the BBSRC and MRC. We also thank Mechthild Pohlschröder for kindly providing the TATFIND program and Gunnar von Heijne for valuable comments on the manuscript.	PubMed Central	2024-06-05T03:55:59.923500	2005-7-2	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182353/", "journal": "BMC Bioinformatics. 2005 Jul 2; 6:167", "authors": [ { "first": "Jannick Dyrløv", "last": "Bendtsen" }, { "first": "Henrik", "last": "Nielsen" }, { "first": "David", "last": "Widdick" }, { "first": "Tracy", "last": "Palmer" }, { "first": "Søren", "last": "Brunak" } ] }
PMC1182354	Background ========== Plants can recognize microbial pathogens by a specific interaction system, which was historically named the gene-for-gene interaction, because particular matching genes must be present in the pathogen as well as in the plant. A successful recognition triggers a hypersensitive reaction of individual plant cells, which is a form of programmed cell death in plants. Though a dead cell on its own might already stop the growth of biotrophic pathogens more importantly the cell death program by itself generates unknown signals for neighboring cells. Thereby the plant immune system is activated locally in some cell layer around the original infection to prepare the plant cells for the next microbial attack. Often this signal from the first infection spreads throughout the whole plant and turns on a long lasting broad pathogen resistance called the systemic acquired resistance. Despite the enormous efforts to dissect the machinery for the hypersensitive reaction many details are still unknown except for the early recognition of the microbial molecules. Often the programmed cell death in plants requires the signaling compound salicylic acid downstream of the recognition process to proceed beyond restrictions points in the cell death program \[[@B1]\]. A conclusive role for salicylic acid has not been worked out but it is likely to function in signal amplification \[[@B2],[@B3]\] and transcriptional activation of genes are very likely \[[@B4],[@B5]\]. We have isolated a gene from soybean which is strongly induced during the hypersensitive reaction and serves as a marker for programmed cell death in this system \[[@B6]\]. The gene is not directly responsive to salicylic acid but transcription can be amplified in the presence of this signal molecule. The gene encodes a protein consisting of two domains. The N-terminal domain is extremely rich in the amino acid asparagine (\~25%) and was therefore called N-rich protein (NRP) \[[@B6]\]. The exact biological function of the NRP-gene remains to be elucidated. Here we describe the analysis of a protein domain found in the soybean NRP-protein and other plant proteins associated with development. The biological processes associated with these proteins lead us to name this novel domain DCD for their role in [d]{.underline}evelopment and [c]{.underline}ell [d]{.underline}eath. Results and discussion ====================== Sequence analysis revealed a significantly conserved region, hence novel domain DCD. The DCD domain is an approximately 130 amino acid long stretch that contains several mostly invariable motifs (Fig. [1](#F1){ref-type="fig"}). These include a FGLP and a LFL motif at the N-terminus and a PAQV and a PLxE motif towards the C-terminus of the domain. Several amino acids are positionally conserved in all members with a DCD domain indicating a critical role of these residues in structure and function (Fig. [1](#F1){ref-type="fig"}). In particular three cysteines are almost generally (red asterisks in Fig [1](#F1){ref-type="fig"}) or subfamily specifically (green asterisks in Fig. [1](#F1){ref-type="fig"}) conserved, which putatively possess a metal binding feature. The predicted secondary structure is mostly composed of beta strands and confined by an alpha-helix at the N- and at the C-terminus. Using the metaserver 3D-Jury \[[@B7]\] no similarities to any other known structural folds could be assigned. The modular nature of the DCD domain is supported by the presence in several protein families with different domain architecture (Fig. [2](#F2){ref-type="fig"}). The DCD domain is only found in plant proteins but absent from bacteria, fungi and animals. The two fully sequenced plant genomes from rice and Arabidopsiscontain 11 and 7 members with a DCD domain, respectively. At least four subgroups of proteins can be identified by phylogenetic comparison of the DCD domain each having members in the rice and in the Arabidopsisgenome (Fig. [2](#F2){ref-type="fig"}). A similar picture emerges from the analysis of plant EST-sequences, which also cluster to the different subgroups (data not shown). The four subgroups differ in the architecture where the DCD domain is located within the protein (Fig. [2](#F2){ref-type="fig"}). Whereas in subgroup I the DCD domain is found in the C-terminus of the protein, it is found more towards the middle of the protein in subgroup II. The third (III) subgroup is more variable; the proteins are mostly characterized by a DCD domain at the N-terminus and in one case it is found subsequent to a ParB domain. The fourth (IV) subgroup shares a DCD domain at the N-terminus but contains several KELCH repeats at the C-terminal part of the protein. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Multiple sequence alignment of DCD-domains.The alignment was built using T-coffee \[21\] and refined manually. First column: database accession numbers (Genbank, if available); second column: species names (at: Arabidopsis thaliana; cp: Citrus X paradisi; cr: Ceratopteris richardii; gm: Glycine max; mt: Medicago truncatula; os: Oryza sativa; tp: Thalassiosira pseudonana); third column: start of the domain in the respective sequences. The aligment is coloured by chroma \[22\]. (conserved prolines: white on grey; conserved glycines and alanines: green on grey; conserved leucines, isoleucines, phenylalanines, cysteines, valines and tyrosines: yellow on grey; conserved asparagines and glutamines: dark red on grey; conserved glutamic acids: light red on grey; conserved threonines and serines: light blue on grey; conserved aliphatic residues: grey on yellow; conserved hydrophobic residues: black on yellow; conserved small residues: dark green on white; conserved positively charged residues: blue on white; conserved polar residues: dark blue on white; conserved charged residues: pink on white; conserved aromatic residues: blue on yellow; conserved big residues: blue on light yellow; conserved negatively charged residues: red on white) The consensus sequence (conserved in 80% of the sequences) shown below; h, p, s, l, b, c, a, + and -- indicate hydrophobic, polar, small, aliphatic, big, charged, aromatic, positively charged and negatively charged residues, respectively. The predicted secondary structure taken from the consensus of the alignment (H, helix or E, beta sheet predicted with expected average accuracy \> 82%; h, helix or e, beta sheet predicted with expected average accuracy \< 82%) using PhD \[23\]. Independent predictions have been performed for Psipred 17 using representatives of distinct groups (accession number: gi\\|2369766, gi\\|50932255, gi\\|51535545). Asterisks on the top of the alignment indicate conserved Cystein residues (red: present in almost all DCD domains, green: present subfamily-specific) ::: ![](1471-2105-6-169-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Phylogenetic tree of DCD domains in plants and related domain architecture.The tree topology was calculated using the Neighbor-Joining algorithm. Hypothetical proteins putatively involved in cell death with c-terminal DCD domains (I. Genbank accession numbers in red), hypothetical proteins with central DCD domains (II. Genbank accession numbers in blue), further hypothetical proteins with N-terminal DCD-domains, ParB domain preceding a DCD-domain and of unclear architecture (III. Genbank accession numbers in black), Kelch-repeat containing proteins (IV. Genbank accession numbers in yellow). The not clearly grouped sequence of Thalassiosirais colored in green. The accession numbers are followed by species names (at: Arabidopsis thaliana; cp: Citrus X paradisi; cr: Ceratopteris richardii; dc: Daucus carota; gm: Glycine max; mt: Medicago truncatula; os: Oryza sativa; pp: Physcomitrella patens; ps: Pisum sativum; tp: Thalassiosira pseudonana) Numbers in nodes indicate bootstrap values (only essentials are shown). The domain names are according to the Simple Modular Architecture Research Tool \[19, 20\] <http://smart.embl-heidelberg.de> \* this sequence contains may be an incorrectly sequenced C-terminal part ::: ![](1471-2105-6-169-2) ::: Whereas the majority of DCD domains (families II, III. IV) contain a second conserved cysteine, directly following the N-terminal one, family I possess a putative functional substitute in the \"central loop\" of the domain (Fig. [1](#F1){ref-type="fig"}, green asterisks on the top of the alignment). We could only identify the DCD domain in a variety of plants, using PSI-BLAST. The domain seems to be present in ESTs from dicots (e.g. Arabidopsis), monocots (e.g. rice), gymnosperm trees (e.g. pine), ferns, and mosses (e.g. Physcomitrella). The available sequences from algae are very limited, but the recently sequenced diatom Thalassiosira pseudonana\[[@B8]\] contains a distant member of this domain in a hypothetical protein (Fig. [2](#F2){ref-type="fig"}). At least the DCD domain is present from early in plant evolution before the separation of diatoms and green algae, leading to higher plants, occurred about 1 billion years ago. For three of the proteins with a DCD domain, all clustering into group I, some biological data have been published. These proteins include the B2-protein from carrot, which was found to be strongly and early induced during the developmental shift from undifferentiated cell cultures to somatic embryogenesis \[[@B9]\]. Though the exact function of the protein still has to be elucidated a role in developmental processes is supported by the finding from Arabidopsistranscript profiling with microarrays. Here the DCD containing protein At2g32910 is only weakly expressed throughout the whole life cycle of Arabidopsisexcept during embryogenic development. A similar pattern is observed for the gene At5g01660, which has several KELCH repeats next to the DCD domain. This gene is most abundantly expressed in embryos but also in the meristem of the shoot apex. A second protein with a DCD domain was identified by \[[@B10]\] in pea. Here the so called Gda1gene is strongly expressed in peas during the vegetative phase but rapidly disappears after shifting the plants into the reproductive phase. The transition is mediated by a change of the light period from short to long days. Interestingly the GDA1gene can be rapidly induced by the phytohormone gibberellic acid, a key player in the developmental change from the vegetative to the reproductive phase in plants. The GDA1transcript accumulated only 15 min after application of gibberellic acid, indicating that the GDA1gene is a primary response gene to this phytohormone. A third protein with a DCD domain was isolated by \[[@B6]\]. This protein was named N-rich protein (NRP) because of the extreme high content of asparagine (\~25%) in the N-terminal half in front of the DCD domain. The NRP-gene is rapidly induced during programmed cell death in soybean, caused by inoculation with avirulent bacteria. Isogenic bacteria, lacking a single Avr-gene are not recognized by soybean cells and neither trigger programmed cell death nor the induction of the NRPgene. The gene is induced early in the cell death program well before the cells lose control of their membrane integrity. Using Phytophthoraas a fungal pathogen to inoculate soybean plants, the same response was found as with bacteria, indicating that the NRP-gene is responding to the cell death program rather than to specific molecules from a particular pathogen. The putative Arabidopsisortholog (At5g42050) is induced by several stress conditions including ozone, osmotic and cold stress as indicated by publicly available transcript profiling data (Genevestigator: <https://www.genevestigator.ethz.ch/>). Ozone treatment leads to small lesions with cell death similar to a hypersensitive reaction caused by avirulent pathogens. A similar set of genes is activated by both inducers of programmed cell death. The DCD domain is quite well conserved on the amino acid level throughout the plant kingdom. The domain is present in proteins with different architectures. Some of these proteins contain additional recognizable motifs, like the KELCH repeats or the ParB domain. The latter domain has been attributed to the partitioning of plasmids and chromosomes in bacteria and has a nuclease activity \[[@B11]\]. KELCH motifs are typically composed of \~50 amino acid long stretches which form a beta sheet \[[@B12]\]. They occur as 5 to 7 repeats that form a beta propeller tertiary structure. KELCH motifs are widespread and have been identified in viruses, plants, fungi and mammals. Most of the characterized KELCH motifs are interfaces for protein protein interaction, often by interaction with proteins from the cytoskeleton \[[@B13]\]. Conclusion ========== The occurrence of the conserved DCD domain in plant proteins of variable length and different architecture, but present throughout the plant kingdom, suggests a role in protein-protein interaction. Transcription profiling reveals that the genes encoding a DCD domain are upregulated during plant development and programmed cell death. It is tempting to speculate, that the DCD domain mediates the signaling in these processes and could thus be used to identify interacting proteins to gain further molecular insights into these processes. Methods ======= Using the protein sequence of Glycine max(gi\\|57898928) as query for a PSI-Blast search \[[@B14]\] after one iteration we retrieve homologs in several plant families with high significance (E-value \> = 8\e^-30^). A conserved region of \~130 amino acids could be identified and the borders of the shared region were defined according to the PSI-BLAST pairwise alignments. Further PSI-BLAST searches with this region converge within the first iteration. A multiple sequence alignment was build using T-coffee and refined manually; additional HMM searches \[[@B15]\] with profiles based on this alignment of non redundant representatives support the findings. Phylogenetically distant sequences of diatom (Thalassiosira pseudonana), fern (Ceratopteris richardii) and moss (Physcomitrella patens*) derive from searches against genome database and EST database, respectively. Two different methods, Phd/PROF\_sec \[[@B16]\] and Psipred \[[@B17]\], were used to predict the secondary structure. A phylogenetic tree was reconstructed using the non redundant alignment of 29 sequences (including fragments and one translation (O49932) that likely contains a frameshift at the C-terminus) in MEGA \[[@B18]\], calculated with the neighbor-joining algorithm. Similar topologies were obtained using other methods e.g minimum evolution (data not shown) and bootstrap values were calculated to test significance. The domain architecture is predicted and displayed by the Simple Modular Architecture Research Tool <http://smart.embl-heidelberg.de>\[[@B19],[@B20]\]. Authors\' contributions ======================= RT carried out the molecular genetic studies and initially identified the described domain. TD performed the sequence based and phylogenetic analysis and PB coordinates the project. All authors contributed to the writing of the manuscript.	PubMed Central	2024-06-05T03:55:59.927007	2005-7-11	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182354/", "journal": "BMC Bioinformatics. 2005 Jul 11; 6:169", "authors": [ { "first": "Raimund", "last": "Tenhaken" }, { "first": "Tobias", "last": "Doerks" }, { "first": "Peer", "last": "Bork" } ] }
PMC1182355	Background ========== Specific protein-protein and protein-nucleic acid interaction are in the focus of many biochemical studies. The exact nature of these interactions is not known. Some scientists argue that the macromolecular interactions are determined by long sequence domains that are involving many residues (amino acids and nucleotides), while others found that there is some degree of specificity already on a single residue level, i. e. some residue pairs are preferentially co-located on interacting interfaces. The existence of preferred residue pairs within, as well as between, macro-molecular structures are supported by numerous statistical analyses of protein-RNA \[[@B1]\] regulatory protein-DNA \[[@B2]\], restrictions enzyme-DNA cut site \[[@B3]\], protein-protein \[[@B4]-[@B8]\] structures and interfaces. Although many studies are performed for statistical analyses of residue co-location, it was not possible for us to find a publicly available tool for this purpose. We found only a reference for the existence of a commercially available tool, the QUANTA modeling software \[[@B9],[@B10]\]. Implementation ============== Any structure files (.pdb) may be selected for analyses from the main window. (Figure [1](#F1){ref-type="fig"}). The tool automatically provide the title of the selected PDB file, a list of sequences present in the file and a list of every common atom in the residues of the respective sequences. These possible backbone atoms are N, CA: C~alpha~, C and O in proteins; and P, O1P: O~1~P, O2P: O~2~P, O5\: O~5~\', C5\: C~5~\', C4\: C~4~\', O4\: O~4~\'. C3\: C~3~\', O3\: O~3~\', C2\: C~2~\', C1\: C~1~\', C5: C~5~, C6: C~6~, N1: N~1~, C2: C~2~,, N3: N~3~and C4: C~4~in nucleic acids. It is possible to exclude one or more sequence from analyses by selecting the \"no-one\" option in the Common Atoms list. The user is asked to define a spherical space around the selected core atoms by choosing a minimum and maximum detection radius around these atoms (between 0 to 15 Ångströms). It is usually not interesting to detect residue co-locations related to neighbor residues in the same sequences. Therefore it is possible to exclude up- and downstream neighbors in the same sequence. (Ex +/-: 0--10). The program ignores terminal residues if they are annotated as HETATM i.e. non-standard residues. The SeqX program makes a list of atoms (and the corresponding residues) which are located within the defined radius around the pre-selected common atoms and are not excluded as neighbor residues. This list is accessible as a Residue Table that contains the Residue Contact Map elements. The atomic distances are calculated by the Pythagoras theses. The results of these analyses are visualized in a Residue Contact Map and summarized in a statistical table. The Residue Contact Map is a dot-plot like graph where every residue in every sequence in the PDB structure is compared to each other, and residue co-locations are indicated by a square. The color of the squares indicate the type of molecular contacts (blue: nucleic acid -- nucleic acid, red: protein -- nucleic acid, black: protein -- protein). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Main features of the SeqX analytical tool. A: Main Window, B: a dot-plot like Residue Contact Map, C: Statistical Analyses of residue co-location in a matrix format and D: the details of the Residue Contact Map elements. ::: ![](1471-2105-6-170-1) ::: The main features of the protein secondary structure are indicated by background colors (blue/green: beta sheet, yellow: alpha helix, gray: turn), if they are annotated (not always) in the pdb source files. It is possible to zoom in the center of the Map and move it into optional directions (using the mouse). Primary structure (the sequence) is available along the coordinates. If the sequence is too long, it is necessary to zoom in the Map to make the sequence readable.. Protein sequence is indicated with the 20 one-letter codes (capital letters), while the nucleic acid sequence with the a, t/u, g, c letters. Clicking on any co-locations highlights the corresponding 2 letters in the sequences (green letter coloring). A simple statistical analysis is performed and the number of every possible residue combinations is listed in a Residue Contact Table. It is possible to save the results of the analysis in JPG and XLS (or similar) files. It is also possible to save even the Residue Contact Map in binary form and XLS format for future statistical processing (\"Save binary\" saves the map as 0 a 1 numbers). Results ======= The Residue Contact Map provides a 2D dot-plot like graph of residue co-locations in protein, nucleic acid or nucleoprotein complexes. (Figure [2](#F2){ref-type="fig"}). This plot is simple and as easy to understand as any other dot-plot. The main right diagonal line corresponds to residue co-locations in the same polymeric chain (neighbors) and it is possible to eliminate by neighbor exclusion. (Figure [3](#F3){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### SeqX: Residue co-locations in a nucleo-protein complex (1A6Y). Residue Contact Map of a nucleo-protein complex (1A6Y) was generated by SeqX. Residue co-locations were detected within 6--9 Å radiuses from C~alpha~amino acid and C~1~\' nucleotide atoms and it was not necessary to use the closest neighbor exclusion option, Ex (+/-): 0. The different kind of molecular co-locations and the main features of the protein secondary structure is color coded. It is possible to zoom in the central portion of the map (middle) and navigate it in optional directions using the mouse (left and right part of the figure). The primary structure (sequence) is readable along the x and y axis in the magnified maps and it is possible to highlight individual co-locations (green letters). ::: ![](1471-2105-6-170-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Effect of neighbor residue exclusion on the residue contact map of a nucleoprotein complex. Residue Contact Maps of 1EQZ nucleoprotein structure using constant 5--9 Å radius around C~alpha~and C~1~\' atoms but varying the Ex +/- value from 0 to 10 Å Note the successive disappearance of the right diagonal line which corresponds to co-locations of neighbor residues in the same strand. ::: ![](1471-2105-6-170-3) ::: The Residue Contact Table contains all possible residue combinations (20 × 20 amino acid to amino acid, 4 × 4 nucleic acid to nucleic acid and 20 × 4 amino acid to nucleic acid combinations) and lists the frequency of theses co-locations in the observed structure. Some of the listed co-locations are specific (true) while other is aspecific (false) co-locations. It is possible to estimate the specificity of the results only in the case of nucleic acids where the Watson-Crick base pairs are known to be specific co-locations. The Residue Contact Table provides data for the 16 (4 × 4) different type of nucleic acid base co-locations, however it is known that only adenine-thymine (a-t, t-a) and guanine-cytosine (g-c, c-g) co-locations indicate true (T) and specific base-pairs, while the 8 other pairs are false (F). The estimated specificity of the SeqX tool on dsDNA is up to 60% (T/F \~ 1.4), (Figure [4](#F4){ref-type="fig"}). The specificity is greatly improved by proper distance selection and exclusion of residue neighbors. (Figure [5](#F5){ref-type="fig"}). It is easy to explain the reason for these observations. (Figure [6](#F6){ref-type="fig"}, [7](#F7){ref-type="fig"}, [8](#F8){ref-type="fig"}). ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Effect of detection radius on specificity of SeqX (B-DNA). The number of nucleotide co-locations was monitored by SeqX using a Palindromic 146 Base-pairs long B-DNA Fragment (in 1AOI). The minimal detection radius varied from 0--6 Å (from the C~1~\' atom), (inserted legend) while the maximal radius was 1--10 Å (X-axis). The ratio of the detected true and false Watson & Crick base-pairs was regarded to be the specificity of the measurements. EX.NEXT indicates when the closest base neighbors on the same strand were automatically excluded from the detection by the tool. ::: ![](1471-2105-6-170-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Residue co-location vs. neighbor exclusion value. Residue co-locations were detected by SeqX in a helical protein (1FXK). The detection radius was kept constant (5--9 Å) around the C~α~atoms but the neighbor exclusion value (EX +/-) was varying. ::: ![](1471-2105-6-170-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Atomic distances in dsDNA (upper view). The distance between C~1~\' atoms and a cytosine (C) -- guanine (G) base pair is indicated. Circles are drawn around C1\' atoms (radius: 3-5-7-9 Å). ::: ![](1471-2105-6-170-6) ::: ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Atomic distances in dsDNA (side view). Three base pairs are indicated in a dsDNA. Circles are drawn around C~1~\' atom of a cytosine (C) using 3-5-7-9 Å radius. G: Guanine. ::: ![](1471-2105-6-170-7) ::: ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Effect of monitoring radius on detection of residue co-locations in a dsDNA. Residue Co-locations are indicated in a dsDNA (D1-D2). Circles show distances (r1, r2, r3) from the C~1~\' atom of a cytosine (C3). Residue co-locations ware detected by SeqX tool and summarized in the table. The number of detected thymine (T), guanine (G), adenine (A) depended on the radius. Shaded circle indicate area between r3 and r2 radius. G8, the expected complementary base to C3 was most selectively found using r2 radius and Ex. +/- 1 option which excluded closest neighbors to C3 in D1 from the detection. ::: ![](1471-2105-6-170-8) ::: It is more difficult to find optimal SeqX parameters for studying residue co-locations in- and between proteins. In contrast to the DNA it is not known which (if any) amino acid pairs represent specific residue co-locations. Furthermore some protein structures are very compact and, for example, in the case of alpha helical proteins many amino acid neighbors might interfere with the specificity of the detection (Figure [9](#F9){ref-type="fig"}) and the exclusion of more than one neighbor is necessary to improve the specificity of the detection. ::: {#F9 .fig} Figure 9 ::: {.caption} ###### Atomic distances in a protein structure of two parallel alpha helices. Distances between C~-α~atoms are indicated around ILE88. The distance to the -4^th^ARG84is 6.2 A and to the +4^th^GLU92 is 6.0 A. This indicates that Ex. +/- 4 might be necessary to exclude the aspecific neighbor effects. ::: ![](1471-2105-6-170-9) ::: We found that detection radius between 5--9 Å and exclusion of +/-8 neighbors gives the best results for analyzing alpha helical protein structures. A real specificity estimation is not possible to do on protein sequences (not even in receptor-ligand structures), because the amino acids are not known to be complementary to each other. Therefore the frequency of amino acid co-locations found by SeqX (preferentially in alpha helical proteins) is compared to the frequency of residue co-locations data from literature \[[@B5],[@B6],[@B8]\]. Our results showed highly significant correlation to data from the literature (p \< 0.0001, n = 210). (Figure [10](#F10){ref-type="fig"}). ::: {#F10 .fig} Figure 10 ::: {.caption} ###### Frequency of residue co-locations\_1-3. A subset of amino acid co-locations (34630) were identified by SeqX in proteins which contained preferentially alpha helices. The frequency of these co-locations were compared to the frequency of residue co-locations with data from literature such as (1) Helical Interface, HI\_M+S \[8\]; (2) Residue-Residue Contacts, Cij \[6\] and (3) chaperon, CH3 data \[5\]. ::: ![](1471-2105-6-170-10) ::: Some cautious and preliminary estimation is still possible even for the specificity of detected residue co-locations in protein structures. Namely, it is known from physico-chemical studies, that some amino acids are attractive while others are repulsive to each other. The known physico-chemical laws suggest that pair-formation (co-location) is probably preferred between amino acids having similar hydrophobicity or different charge, while pair-building between amino acids with different hydrophobicity or similar charge are strongly prohibited. To test this assumption we generated a pool of artificial random protein sequences by translating randomized nucleic acid sequences. The nucleic acids contained equal amount of each nucleotide bases (4 × 25%) and, by that way, the average frequency of amino acids in the translated artificial proteins became very similar to the amino acid frequency of the entire human proteome. The residue co-locations within and between these sequences are determined by statistical lows if we assume that the spatial mobility of the residues in these proteins is free and independent of each other. The calculated probability of any residue co-locations (P~ab~) will be P~ab~= n~a~n~b~/T^2^, T = n~a~+n~b~\...+n~20~wher n is the number of a given amino acid. The calculated relative frequency of a given co-locating pair (C~ab~) is proportional to P~ab~and might be calculated by the C~ab~= P~ab~/ (P~ab~+\...P~xy~) 100 formula, where x and y indicate any of the 20 possible amino acids and the number of xy pairs is 400. The relative frequency of physico-chemically favored co-locations is significantly higher (and the relative frequency of un-favored co-locations is significantly lower) in real protein structures (determined by SeqX) than it is calculated for random interactions (p \< 0.001, n = 80 and n = 10, respectively). (Figure [11](#F11){ref-type="fig"}.) ::: {#F11 .fig} Figure 11 ::: {.caption} ###### Real vs. calculated residue co-locations. The relative frequency of real residue co-locations were determined by SeqX in 80 different protein structures and compared to the relative frequency of calculated co-locations in artificial, random protein sequences (C). The 200 possible residue pairs provided by the 20 amino acids were grouped into 4 subgroups regarding their physico-chemical compatibility to each other i.e. favored (+) and un-favored (-) regarding hydrophobicity and charge. (HP+: hydrophobe -- hydrophobe and lipophobe -- lipophobe; HP-: hydrophobe -- lipophobe; CH+: positive -- negative and hydrophobe -- charged; CH-: positive-positive and negative -negative and lipophobe -- charged interactions). The bars represent the mean +/- S.E.M. (n = 80 for real structures and n = 10 for artificial sequences). Student\'s t-test was applied to evaluate the results. ::: ![](1471-2105-6-170-11) ::: This example indicates that the number of false positive (un-favored) co-locations is about 20% and the specificity of SeqX methods for proteins might be as much as \~80% However this is a very crude estimate, because the number of true co-locations is not surely known. Discussion ========== To understand the nature of specificity of macromolecular interactions is a major challenge in bioinformatics. We were successful in providing evidence to support the view that some degree of specificity already exists on residue level \[[@B3]\]. Therefore we decided to continue our studies of frequency analyses of residue co-locations in nucleoprotein structures. The SeqX tool is specifically designed for this purpose. The 2D Residue Contact Map is a simple and easy to understand display of nucleic acid and protein structures. There are some very sophisticated analytical tools which also even incorporate this feature, like MOLTALK \[[@B11]\], STING Millennium \[[@B12]\], STRIDE \[[@B13]\] MolSurfer \[[@B14]\] MOLPROBITY \[[@B15]\]. The major advantage of this approach is its simplicity. The effective usage of 3D tools and learning the \"3D thinking\" usually requires lengthy training which often is not affordable for general bioinformaticians. We have further developed the concept of Residue Contact Map and added many new features that are not present in existing tools. Such features are 1., The option to choose different backbone atoms (in addition to the conventional C~alpha~and C~1~\' atoms; 2., The option to exclude neighbor atoms and to improve the specificity of the method; 3\. The direct connection to a Residue Contact Table which automatically provides a basic statistical analyzes of the residue co-locations. It is expected, that statistical analyses of residue co-locations in protein and nucleic acid sequences will provide further insight and understanding the rules of macromolecular interactions. The ultimate goal of these types of studies is to find short \"complementary\" or \"compatible\" sequences/motifs even for specific nucleic acid -- protein and protein -- protein interactions, something similar to the well known Watson-Crick rules of specific nucleic acid -- nucleic acid contacts. It is well known that in studies of protein interactions, protein engineering and drug design the most important are the interactions between side chains. However, the recent SeqX program is a general purpose tool (for nucleic acids as well as for proteins) for statistical analyzes and visualization of entire-residue co-locations and it does not pay particular attention to side chains and the pattern of the side chain interactions. It does not limit the usefulness of this tool for its original purpose: any significant residue co-locations (i.e. that which are different from random) are necessarily caused by the side chains (\'R\' in amino acids, \'bases\' in nucleic acids) because they are the variable elements of the structures. However a future implementation might focus on analyzes of side chain to side chain co-locations and examine whether that will improve the specificity of this tool. Conclusion ========== The SeqX is a simple, easy to use specialized tool for visualization and statistical analyses of protein and nucleic acid residue co-locations. It is mainly and specifically developed to study known and novel specific residue interactions. Authors\' contributions ======================= JCB designed and tested the tool, and wrote this article. GF implemented the software. GF is the winner of the first prize of the First Hungarian George Gamow Competition and Fellowship in 2004 with his contribution. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 SeqX\_1.041\_05601.jar. see this article ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ The general support of Z. Benyo and B. Benyo is greatly appreciated. Grants were provided by the Homulus Foundation (Stockholm, Sweden).	PubMed Central	2024-06-05T03:55:59.928568	2005-7-12	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182355/", "journal": "BMC Bioinformatics. 2005 Jul 12; 6:170", "authors": [ { "first": "Jan C", "last": "Biro" }, { "first": "Gergely", "last": "Fördös" } ] }
PMC1182356	Background ========== Human artificial chromosomes are currently being developed as tools for functional annotation of the genome and as potential vectors for gene therapy and other biotechnological applications (reviewed in \[[@B1],[@B2]\]). Strategies for the creation of artificial or engineered chromosomes can be broadly divided into two classes: top down, based on the truncation of an existing chromosome into a much smaller mini-chromosome suitable for further manipulation, and bottom up, whereby defined, cloned chromosomal elements are assembled in vitrointo a prefabricated unit that is capable of nucleating formation of a HAC de novoupon introduction into human cells \[[@B1]-[@B4]\]. These cloned chromosomal elements may also be assembled in cultured cells through a combination of non-homologous recombination and end-joining mechanisms \[[@B5]\]. Thus far, both approaches have resulted in the creation of a de novoHAC composed of large concatamers of the input DNA species (reviewed in \[[@B2]\]). These de novoHACs are mitotically stable in the absence of selection, associate with key centromere and kinetochore proteins and are functionally comparable to the native chromosomes of the host cell. Furthermore, HACs containing two genomic loci, for HPRT and GCH1, have demonstrated evidence of functionality in certain cell culture models, establishing the potential application of HACs as vectors for gene transfer \[[@B6]-[@B8]\]. Creation of artificial chromosomes de novominimally requires a cloned centromeric element of either natural \[[@B9]\] or synthetic \[[@B5]\] origin. Only higher-order alpha-satellite DNA, found at the centromeres of all normal human chromosomes \[[@B10]\], has been shown to be capable of nucleating centromere formation de novo. Alpha-satellite DNA consists of a hierarchical structure of tandem repetitive monomers of \~170 bp, which may be further organized into higher-order repeat units over many hundreds of kilobases \[[@B11]\]. Higher-order alpha-satellite DNA is capable of establishing the assembly of a protein/DNA complex, the kinetochore, which mediates the interactions between the chromosome and the spindle apparatus during cell division \[[@B12],[@B13]\]. These and other lines of evidence suggest that alpha-satellite DNA of this type represents the functional centromere in normal human chromosomes \[[@B10],[@B14]\]. In addition to a functioning centromere, linear artificial chromosomes require synthetic telomeres, which are capable of seeding large telomeric arrays in vivo\[[@B15]\]. However, telomeric DNA is not required for the creation of circular HACs de novo, and its presence or absence appears to have no significant impact on the stability of such HACs \[[@B16]\]. Finally, HAC vectors require origins of DNA replication that are functionally analogous to Autonomously Replicating Sequence (ARS) elements in yeast \[[@B17]\]. However, mammalian origins of replication remain poorly defined \[[@B18]\], although at least some mammalian origin elements (\"replicators\") have been documented that continue to behave as such when translocated to an ectopic chromosomal location \[[@B19],[@B20]\]. Notwithstanding uncertainty about the genomic features that constitute an origin, mammalian origins of replication have been shown to occur on average once every \~100 kb \[[@B21],[@B22]\]. De novoHAC formation at frequencies of at least 10% has been documented from BAC vectors containing only cloned alpha-satellite DNA \[[@B2]-[@B4],[@B16]\], implying that replication origin function must be supplied by elements within alpha-satellite or the BAC vector backbone. Notwithstanding this result, we reasoned that we could augment de novoHAC formation by providing any genomic fragment of at least 100 kb in cisto the centromeric element of the HAC vector, thereby providing additional origin function and potentially additional unidentified functional elements, or simply by providing improved stability as a consequence of an increase in the size of the HAC vector. To circumvent technical difficulties in the manipulation of high-molecular weight DNA by traditional cloning techniques \[[@B23]\], we have developed a novel transposition-based approach to rapidly retrofit genomic BAC clones with telomeres and other key functional elements. Ligation of the linearized derivatives of these retrofitted BAC vectors (referred to as \"BAC-GEN\" vectors) with a complementary linearized BAC vector containing a synthetic D17Z1 alpha-satellite array \[[@B5],[@B24]\] and a telomere (referred to as \"BAC-CEN\") results in the assembly of a linear prefabricated HAC vector containing the defined genomic fragment of interest. Here, we apply this method to the construction and validation of HAC vectors containing different large fragments from the human genome, representing a diverse group of functionally validated de novoHACs containing human genes. In the course of this work, we observed that certain genomic loci appear to greatly facilitate the formation of de novoHACs, suggesting the existence of at least one additional parameter to be optimized during the development of future iterations of HAC vectors. Such loci may contain origins of replication \[[@B20],[@B25]\], scaffold or matrix attachment regions (S/MARs) \[[@B26]\] or other functionally significant chromosomal elements that might contribute to HAC formation and/or stability. We have applied this approach to the construction of a prefabricated HAC vector incorporating the entire 200 kb ß-globin genomic locus, which contains a well-defined mammalian origin of replication \[[@B20]\]. We demonstrate efficient rates of formation of these ß-globin HACs and provide evidence of persistent gene expression. Taken together, the ability to rapidly create multiple, functionally validated BAC-based HAC vectors incorporating any defined genomic locus represents a promising advance in the development of HAC vector technology. Results ======= Assembly of linear, prefabricated HAC vectors --------------------------------------------- The bimolecular BAC-based HAC vector system is comprised of a centromere-containing \"CEN arm\" containing an 86 kb D17Z1-derived synthetic alpha-satellite array \[[@B5]\] and a \"GEN arm\", incorporating a defined, large (\>100 kb) genomic fragment. Both BAC-CEN and BAC-GEN additionally contain \~800 bp synthetic telomeres \[[@B15]\] and selectable markers as indicated in Figure [1](#F1){ref-type="fig"}. A linearized CEN arm is generated by digestion of BAC-CEN with the ultra-rare homing endonucleases I-CeuI and PI-SceI, which creates a unique, non-self complementary overhang. Any genomic BAC vector may be retrofitted to form a BAC-GEN vector by transposition with a custom-built Tn5 based transposon \[[@B27]\] incorporating a telomere, selectable markers and appropriately oriented recognition sites for I-CeuI and PI-SceI \[[@B3]\]. Transposition of the telomere cassette is non-site specific, and insertions into either the BAC vector backbone or genomic insert can be isolated. We were able to generate vector backbone transpositions for all VJ104-based genomic BACs (generated by \"shotgun\" subcloning, see Methods) and genomic transpositions for BACs containing the HGH, PKD1 and ß-globin loci. For the latter, the integration site of the transposon was established by direct sequencing, and the transposon was confirmed to not interrupt either the target gene or its established regulatory elements (data not shown). The BAC-GEN arm is linearized in a similar manner to BAC-CEN, generating an overhang that is complementary only to the residual PI-SceI overhang created from the CEN arm. A ligation between the linearized CEN and GEN arms generates a prefabricated, linear HAC vector (Figs. [1](#F1){ref-type="fig"},[2](#F2){ref-type="fig"}), which may be gel-purified and introduced into mammalian cells by transfection or direct nuclear microinjection; alternatively, the entire ligation reaction may be transfected directly. We assembled a collection of thirteen BAC-GEN vectors representing different genomic loci (100--200 kb) that were shotgun subcloned or identified through the public databases as containing genes of potential therapeutic interest (see Methods). A summary of the sizes and chromosomal origins of each of these genomic fragments is indicated in Table [1](#T1){ref-type="table"}. Prefabricated HAC vectors containing each of these genomic loci were generated by the methodology described in Figures [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"} and transfected into HT1080 cells, resulting in the formation of large, cytogenetically visible de novoHACs presumably composed of concatamers of the initial DNA species. Assembly of the prefabricated HAC was monitored in all cases by PFGE or FIGE (Fig. [2](#F2){ref-type="fig"}) (see Methods for additional details). Impact of different genomic loci on de novoHAC formation ---------------------------------------------------------- The efficiency of de novoHAC formation from prefabricated HAC vectors containing different genomic loci is summarized in Table [1](#T1){ref-type="table"}. All HACs were validated structurally by FISH analysis with probes against D17Z1 alpha-satellite, BAC vector, genomic insert and telomeric DNA, as shown in Figure [3A--C](#F3){ref-type="fig"}. De novocentromere formation was demonstrated by the localization of CENP-C to the HAC (Figure [3D](#F3){ref-type="fig"}); CENP-C is an established marker of functional centromeres \[[@B28],[@B29]\]. Although the numbers of clones are modest, it is clear from the data in Table [1](#T1){ref-type="table"} that not all the genomic loci examined form HACs at similar efficiency. For example, some genomic loci (e.g. G2, G17 and G19) form HACs only rarely (\<10%). In contrast, fragments G6 (75%) and G16 (71%) appear to facilitate de novoHAC formation much more efficiently. Fragment G6 is a 101,704 bp NotI fragment represented on the genome scaffold by BAC accession numbers AC004854.2 and AC004847.3. Fragment G16 is a 84,886 bp NotI fragment represented on the genome scaffold by BAC accession numbers AC104698 and AC016907. Other loci, including the HGH and PKD1 genomic loci, form HACs at intermediate frequencies (19% and 21%, respectively). Interestingly, the HGH locus in a prefabricated vector formed HACs at approximately the same frequency as reported earlier for the same locus in a different vector system \[[@B3]\], suggesting that this intermediate frequency is indeed a property of the genomic locus. Functionality of ß-globin HAC vectors ------------------------------------- It is important to determine whether genes introduced as part of HAC vectors are functional in the recipient host cell. As a proof of principle, we have generated evidence of sustained gene expression from cytogenetically validated HACs containing the entire 200 kb ß-globin genomic locus, establishing the potential application of future iterations of these HACs for gene transfer. As shown in Figure [4](#F4){ref-type="fig"}, expression from the third exon of ß-globin is continuously detectable by RT-PCR from clones containing ß-globin HACs after 30 days of culture in the absence of selective pressure in the cell line HT1080, a fibrosarcoma line that does not express ß-globin. Expression of ß-globin continues to be detectable in the absence of selection for \>90 days of continuous culture (data not shown). Discussion ========== Design and validation of bimolecular BAC-based HAC vectors ---------------------------------------------------------- HACs are believed to function by reproducing the three known critical elements of naturally occurring chromosomes: centromeres, telomeres and origins of replication \[[@B1],[@B2]\]. Optimization of HAC formation may theoretically be achieved by the systematic identification and manipulation of factors that affect the efficiency of formation and subsequent stability of each of these key functional elements. For example, in previous studies, we and others have used de novocentromere formation as an assay to design and evaluate synthetic D17Z1-based alpha-satellite arrays with modifications in the density and distribution of the consensus CENP-B box, a protein binding site known to impact the effectiveness of de novocentromere formation \[[@B3],[@B30]\]. We have shown that D17Z1-based arrays containing an increased number of CENP-B boxes relative to native D17Z1 show a corresponding increase in the efficiency of de novoHAC assembly \[[@B3]\]. In this report, we employ de novoHAC formation as an assay to identify genomic loci that are highly efficient in HAC formation and are thus candidates for containing origins of replication, S/MARs or other cis-acting functional elements that may impact the formation and/or maintenance of HACs. We assembled a collection of genomic DNAs in the 100--200 kb size range (a size range providing a reasonable expectation of containing at least one of these functional units \[[@B21]\]) and assayed their ability to support the formation of de novoHACs. Construction of multiple BAC-based HACs demands the development of novel BAC modification methodologies, owing to the difficulties inherent in the manipulation of high-molecular weight DNAs by traditional subcloning techniques \[[@B23]\]. A first step towards achieving eventual defined composition of matter for HAC vectors \[[@B2]\] requires moving away from uncontrolled in vivomechanisms for HAC vector assembly \[[@B5],[@B6],[@B8]\] towards construction of pre-fabricated HAC vectors containing clearly defined centromeric and other genomic elements. The controlled and systematic generation of large synthetic or naturally derived alpha-satellite arrays is itself difficult, but manageable \[[@B3],[@B5],[@B30],[@B31]\]. Once derived however, these alpha-satellite arrays must be efficiently and predictably brought together with the genomic fragment of interest to create a prefabricated HAC vector. A number of in vivosite-specific recombination approaches to join large alpha satellite arrays with large genomic fragments have been reported \[[@B32],[@B33]\], involving multi-step recombinogenic methodologies whose limitations are evidenced by the fact that few genomic loci have to date been reported to have been successfully incorporated into a HAC vector. In contrast, our transposon-based strategy for the construction of bimolecular BAC-based HAC vectors has facilitated the high-throughput creation of de novoHACs from multiple large genomic loci. Additionally, our laboratory has recently reported a complementary transposon-based methodology for the rapid retrofitting of genomic BACs into unimolecular BAC-HAC vectors, by the mobilization of a single transposable element containing alpha-satellite, telomeric DNA and mammalian selectable markers \[[@B3]\]. This latter approach may be even more efficient overall, avoiding the requirement for an in vitroligation between two distinct DNA species and providing the flexibility to create either circular or linear derivatives of the same BAC-HAC vector as needed \[[@B3]\]. On the other hand, the strategy detailed in the current report involves modification of a target genomic BAC with a much smaller transposable element, enabling the retrofitting of the target BAC with the desired functional cassettes to be achieved in a more straightforward and efficient manner. However, we cannot rule out the possibility that trace amounts of recircularized CEN or GEN arms (Figure [2](#F2){ref-type="fig"}) are co-purified with the prefabricated species and contribute to de novoHAC formation, or, if the entire ligation reaction is used, determine to what extent individual, linearized CEN or GEN arms contribute to de novoHAC assembly by end-joining or non-homologous recombination mechanisms \[[@B5]\]. Given that the resultant de novoHACs are ultimately produced by the uncontrolled concatamerization of the starting DNA species, this point, while noteworthy, is not significant in our view. It is important to view the current strategy attempting to create prefabricated HAC vectors in the context of iterative progress towards the eventual achievement of defined composition of matter for de novoHACs, while understanding that this remains a goal yet to be accomplished \[[@B2]\]. Only when this objective has been reached will dissection of the relative contributions of \"contaminating\" alternative forms of the starting vectors be truly meaningful. Nevertheless, the overall effectiveness of the current methodology has facilitated the construction and functional validation of multiple de novoHACs derived from a significant collection of genomic loci, thereby establishing HAC technology as a general one suitable for analysis of in principle any locus in the genome. cis-acting genomic loci affect de novoHAC formation ------------------------------------------------------- Although the current study was initiated to provide a functional platform for the identification of functional genomic elements in a manner analogous to that first used to identify Autonomously Replicating Sequences (ARS elements) in yeast \[[@B17]\], we stress that we currently have no independent biochemical or other confirmation that the observed effects on HAC formation frequencies are actually related to the presence or absence of replication origins, S/MARs or any other specific functional elements. Nevertheless, it is not unreasonable to propose variation in origin function as one hypothesis to explain the observed differences in de novoHAC assembly, and further experiments will be required to explore this possibility. As seen in Table [1](#T1){ref-type="table"}, the majority of genomic fragments surveyed support de novoHAC formation at frequencies consistent with previous reports using vectors that lack genomic fragments or contain a limited number of other genomic fragments \[[@B3],[@B4],[@B16],[@B34]\]. Indeed, our own previous results using BAC vectors containing only the synthetic, 86 kb D17Z1-based alpha-satellite array used in the current report generates a baseline for de novoHAC formation of 10.5% from 38 analyzed clones \[[@B3]\]. The majority of genomic fragments surveyed (G2, G8, G10, G14, G17, G19) appear to support de novoHAC formation at frequencies comparable to alpha-satellite alone \[[@B3]\]. We note that the unimolecular BAC-HAC vector reported in \[[@B3]\] incorporating the same HGH genomic region used in the current report forms de novoHACs at similar intermediate frequencies comparable to that observed here (15% in \[[@B3]\], 19% in the current report). Other genomic loci (G4, G6, G11, G16 and the ß-globin locus) do appear to facilitate de novoHAC formation at efficiencies above the baseline (see Table [1](#T1){ref-type="table"}). Most notably, genomic loci G6 and G16 support de novoHAC formation at frequencies of over 70%, substantially higher than other genomic fragments of similar size. These effects clearly cannot be explained as the result of a simple increase in the size of the HAC vector, as this would result in a general increase in de novoHAC formation regardless of which specific genomic fragment was utilized. Further dissection of these 100 kb fragments is currently underway to isolate smaller subfragments that may be incorporated as a functional cassette into the design of future iterations of HAC vectors. Cell line specific effects on de novoHAC formation and gene expression ------------------------------------------------------------------------ The current report is based on the analysis of de novoHAC formation in the HT1080 fibrosarcoma cell line, consistent with all previously reported studies on the assembly of de novoHACs (reviewed in \[[@B1],[@B2]\]). Although no systematic examination of the role of the host cell environment on rates of de novoHAC formation has yet been reported, it remains formally possible that the cis-acting effects of adjacent genomic loci on de novoHAC formation are contingent on certain cellular environments. Although the choice of the HT1080 cell line for use in this and other related studies \[[@B1],[@B2]\] is largely historical, we have observed comparable rates of de novoHAC formation in the 293 and other closely related cell lines using BAC vectors containing only cloned alpha-satellite DNA (our unpublished observations). The observation that de novoHACs incorporating a 216 kb ß-globin genomic locus do in fact express ß-globin in the non-erythroid HT1080 cell line is noteworthy (Figure [4](#F4){ref-type="fig"}). Although these HACs contain the entire 5\' and 3\' Locus Control Regions established as being critical for the regulation of globin gene expression in a physiologically appropriate manner \[[@B37]\], it appears that the cloned ß-globin genomic DNA, upon introduction into the cell nucleus in the context of a HAC vector, does not adopt the repressive chromatin configuration found at the endogenous, host cell ß-globin locus. This observation may potentially be highly significant if found to be consistent with the behavior of additional genes upon introduction into the nucleus as HAC vectors, as it impacts on the ability to reproducibly and reliably obtain cell- and tissue specific-patterns of gene expression for applications in biotechnology. Finally, although we have not rigorously quantified ß-globin gene expression from clones containing de novoß-globin HAC vectors over time, we do note that ß-globin gene expression is stably observed by the RT-PCR assay of Figure [4](#F4){ref-type="fig"} over the 90-day time period used in the current report (data not shown). Conclusion ========== HAC vectors provide a novel approach to human genome annotation and gene transfer that may ultimately circumvent many of the technical difficulties currently associated with standard, retroviral-based gene therapy vectors \[[@B2]\]. These include position effects on gene expression and transgene silencing (reviewed in \[[@B35]\]) and a strict, upper packaging limit permitting delivery of only about 8 kb of foreign DNA \[[@B36]\], precluding the incorporation of critical genomic elements that may be required for physiologically meaningful expression of a therapeutic transgene \[[@B37]\]. Additionally, integration of viral vectors into the host genome has been shown to result in oncogene activation leading to cancer \[[@B38]\]. Finally, viral vectors used in recent clinical trials have resulted in severe immunological reactions and death \[[@B39]\]. In contrast to the capacity limits of conventional gene transfer vectors, we have been able to design and construct HAC vectors that contain the entire 200 kb ß-globin genomic region, including the 5\' Locus Control Region required for ß-globin gene regulation and expression \[[@B37]\], thereby bypassing the requirement to identify, dissect and repackage critical regulatory elements into mini-genes capable of being delivered by potentially immunogenic viral vectors. We have shown sustained gene expression from these ß-globin HACs in the absence of selective pressure during at least 90 days of continuous culture. Similarly, we have designed and fabricated HAC vectors incorporating the entire HGH (159 kb) and PKD1 (208 kb) genomic loci; the PKD1 cDNA itself is over 14 kb, well outside the range of retroviral gene therapy vectors \[[@B40]\]. In summary, we anticipate that the functional identification and optimization of individual chromosomal components using the HAC vector systems described here and elsewhere \[[@B2],[@B3]\] will eventually permit the design and construction of prefabricated, custom built HAC vectors incorporating any therapeutic gene in the context of its full complement of endogenous, genomic regulatory elements. HAC vectors may therefore not only fulfill their potential in biotechnology, but will additionally lead to significant advances in the functional annotation of the genome. Methods ======= Construction of BAC-CEN and BAC-GEN ----------------------------------- BAC-CEN is a derivative of pBAC108L, modified to include \~800 bp of synthetic telomeric sequence (made as described in \[[@B5]\]), a puromycin resistance cassette and an adapter containing the recognition sites for the homing endonucleases I-CeuI and PI-SceI (New England Biolabs). 86 kb of synthetic, D17Z1-based alpha-satellite DNA (representing 32 tandem copies of the 2.7 kb higher-order repeat) was subcloned as a BamHI-BglII fragment into a unique BamHI site on BAC-CEN, to create BAC-CEN17a32. Individual BAC-GEN vectors were generated by transposon-mediated retrofitting of defined, genomic BACs \[[@B27]\]. To create the transposon targeting vector, the EZ:TN transposon (Epicentre Technologies, Madison WI) was modified to include \~800 bp of synthetic telomeric DNA, a neomycin/kanamycin resistance marker and an adapter containing the recognition sites for the homing endonucleases I-CeuI and PI-SceI, as described above. Transposition reactions were carried out as recommended by the manufacturer. Target genomic BACs were identified and procured through the genome project databases (PKD1: CTD2517G10-208 kb, ß-globin: CTD264317-216 kb, HGH: CTD2202F23-159 kb, all obtained from Research Genetics), or were created by \"shotgun\" subcloning of size-selected, NotI-digested whole genomic DNA into the BAC vector VJ104, a pBAC108L derivative \[[@B5]\]. Transposition of the telomeric unit into the vector backbone of a genomic BAC was identified as an upward shift in the electrophoretic mobility of the corresponding vector band upon digestion with NotI (data not shown). Preparation of the prefabricated HAC vector ------------------------------------------- 10 µg (equimolar amounts) of each of BAC-CEN and the selected BAC-GEN DNAs were mixed together into a single 1.5 ml eppendorf tube and digested with PI-SceI and I-ceuI in a total volume of 200 µl for 3 hours. The homing endonucleases were heat inactivated, and ATP (Epicentre) was added to a final concentration of 1 mM. The linearized CEN and GEN arms were ligated together overnight at room temperature by addition of T4 DNA Ligase (New England Biolabs). In all cases, the assembly of the prefabricated HAC vector was monitored by resolution of the individual species within the ligation reaction using Pulsed Field Gel Electrophoresis (PFGE) (Bio-rad, DR-III), or Field Inversion Gel Electrophoresis (FIGE) (Bio-rad) and confirmed to have proceeded efficiently as shown in Figure [2](#F2){ref-type="fig"}. Only ligation reactions showing efficient assembly of the prefabricated HAC vector were used for transfections. The ligation product representing the prefabricated vector species could then be gel purified by electroelution of the target band out of the gel slice into 0.5X TBE. The electroeluted DNA species was then dialyzed into ddH~2~O and concentrated into the smallest possible volume using a Microcon YM-100 spin column (Amicon) according to the manufacturer\'s instructions. The concentrated HAC vector was used directly for transfection as described below. In some cases, the ligation reaction was transfected directly without additional gel-purification of the prefabricated species. Cell transfection ----------------- Human fibrosarcoma HT1080 cells were transfected using the Fugene-6 (Roche) reagent according to the manufacturer\'s instructions, and stable clones identified through resistance to puromycin at 3 µg/ml and neomycin at 600 µg/ml. Clones appeared after 7--10 days and were subsequently expanded to generate clonal lines for further analysis. Multiple independent transfections were performed for all 13 HAC species, and the data pooled to generate Table [1](#T1){ref-type="table"}. Cytogenetic analysis -------------------- FISH analysis of clonal lines was carried out according to established procedures. Briefly, a D17Z1-specific alpha-satellite probe was used to initially identify putative HACs. Further structural confirmation of all putative HACs was done by FISH with probes specific to the BAC vector backbone, the genomic insert and telomeric DNA. In all cases, probes were labeled by nick translation in the prescence of Spectrum Green or Spectrum Orange conjugated dUTP (Vysis). For immunofluorescence analysis, slides were immunoreacted with a rabbit anti-CENP-C antibody \[[@B5]\] at a concentration of 1/2000 in PBS and detected with FITC conjugated goat anti-rabbit IgG (Molecular Probes). Images were acquired through a Zeiss fluorescence microscope and CCD camera. Clones containing cytogenetically confirmed HACs were further validated by additional cytogenetic examination following continuous culture in the absence of selection for at least 90 days (data not shown). Gene expression --------------- RT-PCR reactions to assay ß-globin gene expression were carried out using the ExpressDirect system (Pierce) according to the manufacturer\'s instructions. RNA was prepared from HT1080 cells and HAC-containing clones, using standard techniques. List of abbreviations ===================== BAC- Bacterial artificial chromosome HAC- Human artificial chromosome ARS- Autonomously replicating sequence CENP- Centromere protein HPRT- Hypoxanthine guanine phosphoribosyltransferase S/MARs- Scaffold/matrix attachment regions GCH1- Guanosine triphosphate cyclohydrolase 1 PFGE- Pulsed Field Gel Electrophoresis FIGE- Field Inversion Gel Electrophoresis FISH -- Fluorescence in situ hybridization Authors\' contributions ======================= JB designed and constructed the vectors, planned and coordinated the experiments and wrote the manuscript. GS and GC contributed to vector construction and executed the experiments. GVB and HFW provided critical intellectual feedback and assisted in writing the manuscript. Acknowledgements ================ This work was funded by Athersys, Inc. Publication charges were paid for by Duke University Institute for Genome Sciences & Policy. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Strategy for construction of a bimolecular, prefabricated, linear HAC vector.Digestion of BAC-CEN and BAC-GEN vectors with the ultra-rare homing endonucleases I-CeuI and PI-SceI permits directional ligation of both \"arms\" to form a linear HAC vector. ::: ![](1472-6750-5-21-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Pulsed Field Gel Electrophoresis showing creation of HAC vector.Ligation reactions were set up using linearized BAC-CEN alone (Lane 1), linearized BAC-GEN alone (Lane 2), or linearized BAC-CEN and BAC-GEN together (Lane 3). A clear ligation product (arrow), the linear HAC vector, is only visible in Lane 3. Note that trace amounts of re-circularized CEN and GEN arms are detectable in Lanes 1 and 2, but these do not significantly affect assembly or purification of the HAC. ::: ![](1472-6750-5-21-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### *Cytogenetic validation of de novoHAC vectors containing the ß-globin genomic locus.A) D17Z1 alpha-satellite (green), BAC vector (red). (B) D17Z1 alpha-satellite (green), ß-globin genomic locus (red). (C) D17Z1 alpha-satellite (green), telomere DNA (red). (D) D17Z1 alpha-satellite (green), CENP-C (red). In all cases, DNA is in blue (DAPI). Arrows point to de novoHACs. ::: ![](1472-6750-5-21-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Analysis of ß-globin mRNA expression from cell lines containing de novoß-globin HACs after 30 days in the absence of selection.Poly A+ RNA from individual cell clones was subjected to first strand synthesis and PCR with primers specific to exon-3 of the ß-globin gene. Arrow indicates the ß-globin PCR product. 1) Untransfected HT1080. (2) Untransfected HT1080, no reverse transcriptase (RT). (3) ß-globin HAC clone \#1, +RT. (4) ß-globin HAC clone \#1, -RT. (5) ß-globin HAC clone \#2, +RT. (6) ß-globin HAC clone \#1, -RT. (7) ß-globin genomic DNA control ::: ![](1472-6750-5-21-4) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Influence of different genomic loci on de novoHAC formation ::: Genomic Locus* Size (kb) Location No. Transfections Freq. HAC formation^1^ ------------------- --------------- -------------- ----------------------- ---------------------------- G2 100 9q33 12 1/21 (5%) G4 98 8q24.3 10 8/19 (42%) G6 90 7p13 6 9/12 (75%) G8 125 15q23 5 1/8 (12%) G10 125 3p16 9 3/16 (19%) G11 100 22q11.4 11 9/33 (27%) G14 125 1q32 2 3/14 (21%) G16 97 2p23 5 5/7 (71%) G17 97 4q22 8 2/19 (10%) G19 95 11p13 6 1/26 (4%) ß-globin 216 11p15.5 6 7/27 (26%) HGH 159 17q24.2 8 3/16 (19%) PKD1 208 16p13.3 10 4/19 (21%) 1:HAC formation frequency is listed as (\# clones containing HAC functionally validated by cytogenetic analysis and mitotic stability for at least 90 days)/(Total \# clones screened from all transfections with the genomic locus of interest) :::	PubMed Central	2024-06-05T03:55:59.930364	2005-7-5	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182356/", "journal": "BMC Biotechnol. 2005 Jul 5; 5:21", "authors": [ { "first": "Joydeep", "last": "Basu" }, { "first": "George", "last": "Compitello" }, { "first": "Gregory", "last": "Stromberg" }, { "first": "Huntington F", "last": "Willard" }, { "first": "Gil", "last": "Van Bokkelen" } ] }
PMC1182357	Background ========== The search for molecular markers to predict response to treatment and prognosis in patients with cancer has always occupied a major role in cancer research. To date, glycoproteins, polypeptides, or other macromolecules such as CEA, CAI5-3, AFP, βHGC, etc. -- that are determined by biochemical methods -- have a definitive, albeit modest, role for patient management in some tumor types \[[@B1],[@B2]\]. Knowledge of tumor biology and availability of molecular biology techniques have allowed testing of a vast array of potential molecular markers. Among these, those focused on apoptosis have gained considerable attention because of the major role that cell death plays in tumorigenesis and response to therapy \[[@B3],[@B4]\]. It is well-known that free or extracellular DNA circulates in serum/plasma of patients with cancer and in healthy individuals \[[@B5]\]. Such an observation has been raised of late and has focused increased attention particularly on the field of tumor markers. In patients with cancer, there is often a correlation between tumor load and amount of free DNA in circulation \[[@B6]\]. Because part of this DNA exists in the form of oligo- and mononucleosomes, the cell death detection plus -ELISA- kit (Boehringer Mannheim) that is based on a quantitative sandwich-enzyme-immunoassay-principle using mouse monoclonal antibodies directed against DNA and histones, respectively, were developed as an in vitrotest; nonetheless, this subsequently has been applied to quantification of oligo and mononucleosomes in tumors \[[@B7]\] and in the plasma/serum \[[@B8]\] of cancer patients. We recently demonstrated in mice and humans that at least part of nucleosomes in circulation originates from tumor cells, that the early rise of nucleosomes observed in patients undergoing chemotherapy is the consequence of the tumor apoptosis induced by chemotherapy, and that due to an efficient mechanism of depuration a lower level of nucleosomes post-chemotherapy indicates a favorable tumor response in patients with cervical cancer \[[@B9]\]. To date, the majority of studies that have evaluated the nucleosome level pre- and post-therapy with the enzyme-linked immunosorbent assay (ELISA) method have been performed on a small and rather heterogeneous group of patients with regard to tumor type and chemotherapy schedule \[[@B10],[@B11]\]. Consequently, the potential value of serial determination of serum nucleosomes to predict response to therapy and prognosis remains to be investigated. In this study, we have addressed this issue by performing pre- and post-chemotherapy analyses of serum nucleosomes in a homogeneous group of patients with locally advanced cervical cancer treated with paclitaxel and carboplatin as neoadjuvant therapy followed by surgery. Methods ======= Forty one untreated patients with histologic diagnosis of cervical carcinoma and FIGO staged as IB2-IIIB were studied. Patients needed to meet the following inclusion criteria: 1) aged between 18 and 70 years; 2) performance status 0--2 according to Word Health Organization (WHO) criteria; 3) normal hematological, renal, and hepatic function according to standard parameters; 6) a normal chest x-ray, and 7) signature on written informed consent. Exclusion criteria included 1) severe systemic or uncontrolled disease (infection, central nervous system, metabolic, etc.) that precluded use of chemotherapy, 2) pre-existing neuropathy due to any cause, 3) pregnancy and lactation, 4) mental illness, and 5) previous or concomitant malignancies (exception, melanoma skin cancer). The protocol was approved by the regulatory boards of the Instituto Nacional de Cancerología (INCan) in Mexico City. Treatment. Treatment consisted of carboplatin at a dose of 6 area under the curve (AUC) diluted in 500 mL of 5% glucose administered over 1 h followed by paclitaxel at 175 mg/m^2^administered over 3 h, both drugs on day l; a total of three 21-day courses were administered. Complete blood cell counts and biochemical profiles were performed at days 14 and 21 of each cycle. After neoadjuvant chemotherapy, patients were submitted to radical hysterectomy. Adjuvant post-operative chemoradiation with cisplatin was performed in all cases with positive surgical margins, one or more positive pelvic lymph node, and in patients with disease in parametria (all high-risk factors for recurrence) \[[@B12]\], as well as in cases whose residual disease contained vascular or lymphatic permeation and/or deep invasion into the middle or internal thirds of the cervical stroma (both intermediate-risk factors for recurrence) \[[@B13]\]. Blood sampling for nucleosome measurements. The day prior to beginning the first course of chemotherapy, a sample of 6 mL of blood was drawn from the patient\'s arm and centrifuged within the next h at 1,500 gfor 20 min to obtain serum. Subsequently, we added 10 rnM EDTA (pH8) to the serum and samples were stored at -20°C until analysis. The same procedure was performed for taking and processing samples at day 7 of the first course of chemotherapy and at day 7 of the third course of chemotherapy (hereafter referred as the first and third courses). Analysis of nucleosomes by ELISA. This assay was carried out with the cell death detection kit plus -ELISA- (Roche) as follows: 20 μL of serum samples (diluted 1:4 with incubation buffer) were placed into a streptavidin-coated microtiterplate and incubated with a mixture of anti-histone-biotin and anti-DNA-POD for 2 h at 15--25°C. Antibodies bind to the histones and DNA of nucleosomes and fix the immunocomplexes to the microtiterplates by the streptavidin-biotin interaction. After the incubation period, unbound antibodies are removed by washing with incubation buffer; then, the ABTS (2,2-Azino di-3-ethylbenzthiazolin-sulfonate) solution is added for 30 min. Incubation of the retained POD-linked complexes with ABTS as substrate permits the spectrophotometrical quantification of the nucleosomes at 405 nm. As positive control, we used the complex DNA-histone provided in the kit and as negative control, the substrate solution only. We constructed a standard curve with serial dilutions of apoptotic material ranging from 3,500 arbitrary units (AU) to 200. All determinations were performed by duplicate, and then the values of the double absorbance measurements of the samples were averaged. Afterward, background value of the immunoassay from each of these averages was subtracted and final values were expressed as AU; interassay variability was 10%. Tumor response criteria and efficacy assessment. Response to neoadjuvant chemotherapy was evaluated pathologically in the surgical specimens and classified as complete or partial. Complete response was considered when there was no evidence of tumor cells or when there was residual microscopic disease in absence of positive pelvic or para-aortic lymph nodes, disease in parametria, positive surgical margins, vascular and/or lymphatic permeation, or deep invasion into the middle or internal thirds of the cervical stroma. Partial pathologic response was considered with any macroscopic residual, or any microscopic residual plus at least one high-risk and/or intermediate-risk factor for recurrence. Time to progression was considered from the date that chemotherapy began until the date of confirmed progressive and/or recurrent disease, whereas overall survival was considered from date of diagnosis until the date of patient\'s death. Statistical analysis. Baseline level of nucleosomes was compared with those obtained at first and third courses of chemotherapy using the paired ttest. Patients were divided into two groups depending on the differential values of nucleosomes above and below the median to correlate with pathologic response using the chi square test. Survival analysis for time to progression and overall survival was performed using the Kaplan-Meier method \[[@B14]\]. Differences in survival between groups were analyzed by the log-rank method \[[@B15]\]. Cox proportional hazards regression analysis was used to assess prognostic significance of a number of covariates in a multivariate setting \[[@B16]\]. Two-tailed probability values of \"conventional\" 0.05 or less were considered significant. Analysis was performed with SPSS for Windows version 10 (SPSS, Inc., Chicago, IL, USA, 1999). Results ======= Patient characteristics ----------------------- A total of 41 patients were studied. Characteristics of these patients are shown in Table [1](#T1){ref-type="table"}. Median patient age was 45 years (range, 24--67 years); the majority of patients had squamous tumors (8 %) and were staged IB2-IIB(84%). Serum nucleosomes ----------------- We evaluated nucleosome levels in all 41 patients before (baseline) and post-treatment at first and third chemotherapy courses. As shown in Figure [1](#F1){ref-type="fig"}, mean (CI 95%) baseline level was 1,085 AU (median, 825) with a range from 419--2,980 AU. Measurement at the first chemotherapy course showed a mean of 1,120 AU (CI 95%), (median, 891) ranging from 499--3,200 AU, and at the third course there was a mean of 978 AU (CI 95%) (median, 897) with a range of 471--2,930 AU. There were no statistically significant differences as compared with baseline values or between these. To further evaluate nucleosome behavior during treatment, differential values of nucleosomes were calculated between baseline and first and third courses. In the first course, levels diminished in 20 patients to a mean of -411 (standard deviation \[SD\] 692) and a median of -170 AU, whereas values increased in 21 patients with a mean of 561 (SD, 492) and median of 417 AU. On the other hand, during the third chemotherapy course the mean value for the 20 patients in whom values decreased was -611 (SD, 628) and the median of -327 AU, while in the remaining patient values mean increase was 430 (SD, 462) with a median of 217 AU. The paired ttest was statistically significant (p\<0.0001). Pathologic response and clinical outcome ---------------------------------------- Induction chemotherapy was well tolerated. A total of 123 cycles were administered, of which only two were delayed for 1 week. Most common toxicities were nausea/vomiting and neuropathy, which were mainly grades 1--2. There were no cases of leukopenia grade 3 but neutropenia grades 3 and 4 were present in 12 and 3% of the courses, respectively. Other toxicities were mild and uncommon. All patients completed the three chemotherapy courses and underwent radical hysterectomy. Surgical specimen analysis showed complete pathologic response in 16 cases (39%), while partial pathologic response was found in 25 of 41 patients (61%). At a mean follow-up of 23 months (range, 7--26 months), projected progression time and overall survival were 80.3 and 80.4%, respectively. Correlation of nucleosome levels with pathologic response --------------------------------------------------------- For establishing associations between pathologic response and prognosis, patients were divided into two groups depending on differential values of nucleosomes above and below the median. Nucleosome levels at first chemotherapy course showed no statistically significant correlation with response, (χ^2^test, p= 0.160) (Table [2](#T2){ref-type="table"}). On the other hand, there was a statistically significant -- albeit marginal -- correlation between response and nucleosome levels measured at the third course, and there were 11 complete responses among the 20 patients who had lowered nucleosomes and 16 partial responses in patients with elevated nucleosomes (χ^2^test, p= 0.041) (Table [3](#T3){ref-type="table"}). Correlation of nucleosome levels with clinical outcome ------------------------------------------------------ Nucleosome levels at first chemotherapy course failed to correlate with outcome. Contrariwise, survival analysis conducted with values obtained at the third course showed a statistically significant better time to progression and overall survival in patients who demonstrated lower levels (p= 0.0243) and p= 0.0260, respectively) (Figure [3](#F3){ref-type="fig"}). Clinical stage and tumor size were also analyzed. Earlier staged (IB2-IIB) patients had longer time to progression and survival as compared with patients staged at IIIB (p= 0.0164 and p= 0.0105, respectively), whereas patients with smaller tumors had longer time to progression and survival (p= 0.0096, and p= 0.0089, respectively). In addition, the status of complete pathologic response showed a trend for better time to progression and survival times (p= 0.0998 and p= 0.0953, respectively) (data not shown). We assessed the prognostic significance of a number of covariates in a multivariate setting. Covariates considered for inclusion were nucleosome values at third course, stage, tumor size, and pathologic response. A forward selection procedure was used to find the combination of pairs of covariates that were significant or that showed a trend (p\<0.1) in survival. Cox regression analysis of all patients showed that nucleosome levels and tumor size were significant prognostic factors for survival. Response rate (RR) showed that nucleosome increase in the third course increased risk of death to 6.86 (CI 95%, 0.84--56.04), whereas a tumor size \>30 cm^2^increased risk of death to 9.08 (CI 95%, 1.11--74.20) (Table [4](#T4){ref-type="table"}). Discussion ========== The majority of anticancer agents in use at present were developed using empirical screens designed to identify agents that selectively kill tumor cells. However, until recently pathologists noticed that radiation and chemotherapy can induce cell death with the morphologic features of apoptosis \[[@B17]\], although the significance of these observations was not widely appreciated. Currently, it is well-established that anticancer agents induce apoptosis and that disruption of apoptotic programs can reduce treatment sensitivity \[[@B18]\]. This connection between response to treatment and apoptosis has led to evaluation of apoptosis levels by diverse methods the levels before or during treatment with variable results. While single or baseline levels of apoptosis, or key pro- or anti-apoptotic gene products, yield mixed results in its prognostic significance \[[@B7],[@B19]-[@B25]\], evaluation of the treatment-induced apoptotic effect has shown to be more useful \[[@B26],[@B27]\]; however, this is associated with the risk and discomfort of repeated biopsies. In this scenario, measurement of serum nucleosomes by ELISA, a simple and non-invasive test, has facilitated the testing of its prognostic significance in cancer treatment. Previous studies of serum nucleosomes in patients with cancer have shown that high heterogeneity in baseline levels of nucleosomes among the different types of cancer and even among patients with the same tumor type. Our present data from a homogeneous population of patients with cervical cancer confirm these observations: values ranged from 419--2,980 AU. These values are nevertheless very similar to those found by our group in a previous study in patients with cervical cancer similar clinical stages, which ranged from 627--2,313 AU \[[@B9]\]. On the other hand, we found no correlation of baseline levels with clinical-pathologic factors. It appears that this correlation could be tumor type-specific, as reported by Holdenrieder et al., who showed a highly significant correlation between stage and serum nucleosomes in colon but not in breast carcinoma \[[@B28]\]. To evaluate the potential prognostic significance of serum nucleosomes, it is important to choose adequate evaluation time points that would reflect the proapoptotic effect of chemotherapy; to date, optimal time points for measurement have not been established. Holdenrieder et al. analyzed a total of 42 patients (colorectal, lung, and lymphoma) receiving primary adjuvant of second-line chemotherapy with a variety of agents and schedules (administration in 1--5 days). Despite such heterogeneity in the study, a pattern of nucleosome increase within the first 3 days to a subsequent nucleosome decrease in the treatment-free period emerged, as well as correlation of post-treatment levels of nucleosomes with response \[[@B10]\]. Similar data was reported by Kuroi et al., in 18 patients with recurrent breast carcinoma undergoing treatment with docetaxel every 3--4 weeks; these authors found a statistically significant correlation between decrease in nucleosomes and response. In addition, they evaluated the kinetics of nucleosomes before and 72 h thereafter in four patients who received single-agent docetaxel; in this instance, the authors found a peak of nucleosomes at 24 h and a decrease at 72 h \[[@B11]\]. To the contrary, we previously reported that in patients with cervical cancer receiving oxaliplatin, gemcitabine at day 1 the rise occurred within the first hours to decrease at 24 h in the responding patients \[[@B9]\]. Based on these data, we decided to evaluate here nucleosomes at day 7 of the first and third courses of treatment with carboplatin and paclitaxel. At such time points, it could be expected that no significant drug levels would be in circulation and that therefore, nucleosomes would reflect tumor mass or activity as the maximum level of cytotoxicity could be attained in the days prior to measurement, as demonstrated in a murine model of tumors treated with platinum \[[@B29]\] or paclitaxel \[[@B30]\]. Interestingly, as a whole there were no nucleosome reductions either in the first course (mean baseline, 1,085 vs. 1,120) or at the third treatment course (baseline, 1,085 vs. 978). However, differential levels showed a higher decrease in the third than in the first course as compared with the baseline (-611 vs. -411, respectively). These results would suggest that as a whole, tumor reduction induced by one chemotherapy course was insufficient to produce significant reduction in nucleosome levels in this patient population receiving the carboplatin plus paclitaxel combination. On the other hand, Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"} show that nucleosomes at the third -- but not at the first -- course correlated with pathologic response. This is illustrated in Figure [2](#F2){ref-type="fig"}, which shows a trend for lower levels of nucleosomes in both first and third courses in patients who achieved complete response with regard to those who did not, suggesting that these tumors were sensitive to chemotherapy and that they failed to develop resistance during subsequent therapeutic courses. Our results to date demonstrate that decrease in serum nucleosomes during chemotherapy treatment -- although marginally -- correlates with response under these experimental conditions. To investigate whether the kinetics in nucleosome levels possesses prognostic significance, we analyzed time to progression-free survival and overall survival. Results showed that while nucleosome levels at the first cycle failed to correlate with time to progression and survival, third-cycle levels did correlate with both parameters; on the other hand, earlier stages (IB2-IIB) and smaller tumors also showed longer time to progression and survival. Interestingly, neither time to progression nor survival were correlated with pathologic response. To investigate which variables maintained their prognostic significance for survival, a Cox proportional hazard model was undertaken using pairs of covariates to ascertain which of these pairs better predicted outcome. Nucleosomes determined at the third course and tumor size emerged as statistically significant; however, these results should be considered with caution, as confidence intervals were very wide, perhaps as a result of small sample size and small number of events. The biological meaning of circulating nucleosomes and the factors that govern their production and clearance from the body are yet to be understood. A body of experimental evidence demonstrates that in mice, apoptosis develops rapidly, within hours, after cytotoxic treatments and that it is dose-dependent \[[@B31]\]; moreover, the extent of apoptosis is higher with the first treatment course than after subsequent doses \[[@B30]\]. In this study, we found that nucleosomes after the first course lowered to a lesser degree than after the third cycle. This finding may reflect either that a single measurement after the initial cycle is not sufficient and would require repeated measurements during the first course of chemotherapy, as recently reported by Holdenrider et al. These investigators found that the area under the curve of nucleosomes during days 1--8 of the first chemotherapy cycle carried the strongest prognostic impact in patients with lung cancer \[[@B32]\]. Our finding on the correlation between nucleosome levels at the third course with prognosis may suggest that the kinetics of circulating nucleosomes may not only reflect the immediate, chemotherapy-induced debulking effect, but that other factors such as amount and/or activity of serum nucleases \[[@B33]\], efficiency of nucleosome clearing from circulation \[[@B34],[@B35]\], and perhaps the macrophage number and function \[[@B36]\] may also influence the behavior of nucleosomes, which in turn may have possess on their own an effect on tumor-host interactions that influences the patient\'s prognosis. This is suggested by data demonstrating that chromatin fragments impair the natural killer-mediated cytotoxicity of tumor cells \[[@B37]\]. Conclusion ========== Serum nucleosome level as determined in the third course of neoadjuvant chemotherapy in patients with cervical cancer was marginally related with tumor response and survival. Our findings, along with those reported by other authors on diverse tumor types, suggest that circulating nucleosome measurements could be a useful predictive and/or prognostic factor in the management of patients with cancer. Further studies are needed to better define not only the predictive/prognostic role of circulating nucleosomes in cancer, but also their potential participation in their own right in response to therapy. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= C T-B, L T-C, and A V-C performed sampling and analysis of nucleosomes, while LF O-O carried out the statistical analysis, LC cared for patients, and A D-G conceived the study and wrote the manuscript. All authors read and approved the final manuscript Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2407/5/65/prepub> Acknowledgements ================ This work was supported by CONACyT, México grant 34649-M and PAPIIT grant IN214902-2, both awarded to AD-G. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Nucleosome values. Expressed as mean values and CI 95%, had a mild increase in the first course to decrease thereafter, however, these differences were not statistically significant. ::: ![](1471-2407-5-65-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Nucleosome value distribution according to response (expressed as mean and CI 95%). A clear trend for decrease in both measurements is shown in responding patients. ::: ![](1471-2407-5-65-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Kaplan-Meier estimates of survival at a mean follow-up of 23 months (range, 7--26 months) according to nucleosome values determined in the third chemotherapy course (p= 0.0260). Time to progression-free survival interval was also statistically significant longer in patients who experienced decreased nucleosomes. ::: ![](1471-2407-5-65-3) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Patient characteristics ::: -------------------- ------------- Number 41 Age (years) 45 (24--67) Clinical stage IB2-IIB 34 (84) IIIB 7 (16) Tumor size (cm^2^) \<30 22 (53) \>30 19 (47) Histology Squamous 33 (81) Adeno/adesq 8 (19) -------------------- ------------- ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Correlations of nucleosome levels in the first course and response to chemotherapy ::: Complete response Partial response p value --------------------- ------------------- ------------------ ------------ Nucleosome decrease 10 10 Nucleosome increase 6 15 p= 0.160 ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Correlations of nucleosome levels in the third course and response to chemotherapy ::: Complete response Partial response p value --------------------- ------------------- ------------------ ------------ Nucleosome decrease 11 9 Nucleosome increase 5 16 p= 0.041 ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Cox model for survival ::: Variable SE RR CI 95% p value ----------------------------------------------- ------ ------ ---------- --------- Nucleosome in third course Decrease vs Stable 1.07 6.86 0.84--56 0.072 Tumor size \<30 cm^2^vs \>30 cm^2^ 1.07 9.08 1.10--74 0.040 SE, standard error; RR, response rate. CI 95%, 95% confidence interval. :::	PubMed Central	2024-06-05T03:55:59.933231	2005-6-27	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182357/", "journal": "BMC Cancer. 2005 Jun 27; 5:65", "authors": [ { "first": "Catalina", "last": "Trejo-Becerril" }, { "first": "Luis F", "last": "Oñate-Ocaña" }, { "first": "Lucía", "last": "Taja-Chayeb" }, { "first": "América", "last": "Vanoye-Carlo" }, { "first": "Lucely", "last": "Cetina" }, { "first": "Alfonso", "last": "Duenas-Gonzalez" } ] }
PMC1182358	Background ========== Most human cancers arise in epithelial cells, underlining the importance of understanding the molecular biology of cancer and the complex balance of proliferation and differentiation in this cell type. Increased knowledge of these processes may provide unique targets for the future development of pharmacotherapy aiming at halting or reversing metastasis and cancer growth. N-acetyl-L-cysteine (NAC) is a membrane permeable aminothiol that functions as a nucleophilic ROS scavenger and antioxidant as well as a precursor of intracellular cysteine and glutathione (GSH). The reduced cysteine represents the active form, as opposed to the inactive oxidized cystine dimer. To date, NAC is used as a mucolytic and as acute treatment of fulminant hepatic failure following paracetamol poisoning. However, cancer preventing and therapeutic effects have also been suggested. In particular, NAC has been demonstrated to induce anti-proliferative and differentiating effects in normal human epidermal keratinocytes (NHEK), as well as in the epithelial colon cancer cell line Caco-2 \[[@B1]\]. Primary normal human epidermal keratinocytes (NHEK) undergo spontaneous terminal differentiation over 30 days in culture. However, if supplemented with 2 mM NAC 24 hrs after seeding, an accelerated differentiation process can be observed. Three days post NAC exposure, differentiation of NHEK is demonstrated by an increased number of intercellular junctions, basal localization of cytokeratin and apical localization of actin determined by scanning electron micrographs of cells and sub-structures and high resolution confocal fluorescence immuno micrographs of for example β-catenin, E-cadherin, actin and cytokeratins. Furthermore ceased proliferation can be demonstrated by ^3^H thymidine incorporation without accompanying apoptosis experimentally verified by properdium iodide labelling and flow cytometry. Interestingly, an epithelial colon cancer cell line responded in analogy with the normal epithelial cells. Caco-2 differentiate spontaneously over a period of around 25--30 days in culture \[[@B2]\]. However, when a single supplement of 10 mM NAC was given to Caco-2 cells 24 hrs after seeding, the proliferation decreased and the cells progressed to a differentiated state in three days without any sign of apoptosis \[[@B1]\]. Here the differences in gene expression was studied overtime for both NHEK and Caco-2 cells using microarray technology with subsequent confirmation of a selected set of genes. The results are discussed from the perspective of accelerated differentiation and growth arrest. Methods ======= Cell cultures ------------- Normal human epidermal keratinocytes, NHEK (Cambrex, San Diego, CA), plated at a density of 8 × 10^3^cells/cm^2^, were grown in KGM^®^medium plus KGM^®^SingleQuots^®^(Cambrex). The Caco-2 human colon carcinoma cells were seeded at a density of 9 × 10^3^cells/cm^2^, grown in Dulbecco\'s modified Eagle minimum essential medium (DMEM, GIBCO Labs, Grand Island, N.Y.), supplemented with 10% (v/v) heat-inactivated fetal calf serum (GIBCO Labs), L-glutamine (2 mM), penicillin (50 IU/ml) and streptomycin (50 mg/ml). The N-acetyl-L-cysteine (NAC, Sigma Chem. Co., St Louis, MO) stock solution (20 mM in KGM and 100 mM in DMEM) was stored at 4°C in the dark and used within 1 week from preparation. Twenty-four hrs after seeding, filter sterilized NAC solution was added to the cell cultures to a final concentration of 2 mM and 10 mM in NHEK and Caco-2 respectively. The concentrations were selected after performing a dose-dependent inhibition analysis for each of the two cell types \[[@B1]\]. Scanning electron microscopy ---------------------------- Cells grown on coverslips were fixed with 2.5% glutharaldehyde in 0.1 M Millonig\'s phosphate buffer (MPB) at 4°C for 1 hr. After washing in MPB, cells were post-fixed with 1% OsO~4~in the same buffer for 1 hr at 4°C and dehydrated in increasing acetone concentrations. The specimens were critical-point dried using liquid CO~2~and sputter-coated with gold before examination on a Stereoscan 240 scanning electron microscope (Cambridge Instr., Cambridge, UK). Experimental set up ------------------- From the normal (NHEK) cells, RNA was obtained from cultures grown for 1, 12 and 24 hrs after NAC treatment as well as from untreated cells at the same time points. Replicates at the level of individually performed cDNA synthesis were used for hybridisation in duplicate for the 1, 12 and 24 h time points except for one fall out of a 24 hrs post NAC treated sample. RNA was extracted from both treated and non-treated cancer (Caco-2) cell cultures at 1, 12 and 24 hrs. Biological replicates were used for hybridisation in a duplicate fashion for both treated and non-treated samples. RNA extraction, purification and quality control ------------------------------------------------ The total RNA was extracted from cell cultures using Trizol (Gibco BRL, NY, USA) according to the manufacturers instructions. Thereafter mRNA was extracted by oligo dT Dynabeads (Dynal, Oslo, Norway) and the quality of mRNA was validated using the Bioanalyzer 2100 (Agilent technologies, Waldbrunn, Germany). Target preparation and hybridisation ------------------------------------ A total of 8 micrograms of mRNA from each sample was used to perform cDNA synthesis. Following in vitro transcription, the biotin labelled cRNA was fragmented and a total of 15 micrograms were subsequently hybridised and analysed on the Affymetrix GeneChip™, all according to the manufacturers instructions (Affymetrix, Santa Clara, USA) with scanning performed on the GeneArray 2500A Scanner (Affymetrix, Santa Clara, USA). Data analysis ------------- The MAS 5.0 software package (Affymetrix Inc.) was used to compute cell intensity files (.cel) for each chip. For the purpose of inspection of parameters important to quality control, chip files (.chp) were also generated for each chip using the statistical expression algorithm. Next, individual probeset intensity values were computed based on the cell intensity files using the RMA algorithm within the Bioconductor package. All data was analysed in three separate batches (2 separate sample preparation batches for the Caco-2 cell line and one for NHEK cell line. Each batch also included RNA samples obtained from 30 min, 3 hrs and 48 hrs treated and untreated cell cultures which are not presented in this manuscript). The RMA background correction, quantile normalization, only PM probe correction and a median polishing were applied. Next, data probeset intensities from each sample were converted to a biological fold ratio between the treated and control sample. To generate the ratio, signal intensities of each treated sample were divided by the control sample intensities for each corresponding time point, separately within each RMA batch. Because of a missing 24 hrs control for NHEK cell line due to fallout, time point 12 hrs control was used instead (detailed investigation showed that control samples at each time point were extremely similar to each other, rationalizing this choice). The average of replicates for each time point in two cell lines was used in the downstream analysis. Due to the quality control issues encountered with time point 24 hrs NHEK cell line, only one replicate was used. The standard deviation statistics was calculated using the Global Error Model based on deviation from one (GeneSpring, Silicon Genetics). A final total of 6 conditions were generated (1, 12, 24 hrs time points for NHEK and Caco-2). The MIAME compatible dataset is made available at the ArrayExpress expression data repository at EMBL. In order to identify genes that show significant changes during treatment we searched for genes that show significant up- or down-regulation in treatment relative to the control, in at least one of the time point for one of the cell lines (filtering was done using the t-test p-value \< 0.01, indicating significant deviation from the control value or from the ratio of 1). Two-way hierarchical clustering of genes and conditions was performed using the set of pre-filtered genes (2054) and Standard correlation. In order to identify genes with significant NAC treatment response at each time point for each cell line we applied a two-step filtering based on the t-test p-value of less than 0.1 (indicating statistically significant changes from the control value), and a biological fold cut-off of 1.5 fold (up or down). A total of 6 genelists (3 for each cell line) was produced and visualized graphically for common distinct patterns. Data filtering and visualization was performed using GeneSpring (Silicon Genetics). To assess the overrepresentation of functional groups, according to Gene Ontology, the publicly available tool EASE (v2.0) was used \[[@B3]\]. Real-time kinetic PCR --------------------- Validation of microarray results was performed by quantifying relative mRNA expression levels of several genes of interest. Two RT-PCR formats were used for the investigated genes. The first relied on primers and TaqMan probes obtained from Applied Biosystems and their Assay-on-demand system and the second approach was based on in-house designed primers and SYBR green detection. The assays used either the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) or the transferrin receptor (TFR) as the endogenous internal reference gene. For the first method, cDNA synthesis was performed using SuperScript™ III (Invitrogen) in a 20 μl reaction containing: 1x first strand buffer, 5 mM DDT, 40 U RNasin (Promega), 5 μg total RNA, 200 U SuperScript™ III, 250 mM oligo(dT)20 primer, 0,5 mM dNTP each dATP, dGTP, dCTP, dTTP. Primer, total RNA and nucleotides were heated to 65°C for 5 min and subsequently cooled on ice for 1 min. After addition of the other reagents the samples were incubated at 50°C for approximately 60 min. Finally, the reaction was inactivated by heating to 70°C for 15 min. Assays-on-Demand (Applied Biosystems), containing gene specific primers and probes labelled with 6-carboxyfluorescein (FAM) at the 5\' end and with 6-carboxytetramethylrhodamine (TAMRA) at the 3\' end for the following genes: GAPDH(Genbank nb. BC029618), MMP9(J05070), AKR1C3(D17793), AQP3(N74607), PLAT(M15518), HBP1(AF019214), PTGS2(U04636), ERBB3(M34309) and PNRC1(U03105) (see Table [3](#T3){ref-type="table"}), were used together with TaqMan Universal PCR Master Mix (Applied Biosystems) in the TaqMan real time PCR reaction as described by the manufacturer. 25 μl reactions were done in 96-well plates. Amplification and detection was carried out using the iCycler iQ Multicolor Real-Time PCR Detection System (Bio-Rad Laboratories, Hercules, CA, USA). All samples were run in triplicate. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Relative gene expression levels measured by real time kinetic PCR in Caco-2 and NHEK cells after NAC treatment as compared to untreated cells. ::: Caco-2 ----------------------------------------------- ------------ ------------- ------------ -------------------- ------------ ---------------- ------------ ---------------- ------------ ---------------- ------------ ---------------- RT-PCR Affymetrix RT-PCR Affymetrix RT-PCR Affymetrix RT-PCR Affymetrix matrix metalloproteinase 9\* MMP9 J05070 na nd no change up na up na na aldo-keto reductase\* AKR1C3 D17793 na nd no change up up (2.8) nd na na aquaporin 3\* AQP3 N74607 no change down up (6.8) up up (5.0) up na na plasminogen activator, tissue\* PLAT M15518 na nd no change nd no change nd na na HMG-box transcription factor 1\* HBP1 AF019214 no change nd up (2.6) up up (2.1) nd na na Cox-2\* PTGS2 U04636 no change nd no change nd down (2.2) down na na v-erb-b2\* ERBB3 M34309 na nd up (3.7) up no change nd na na proline-rich nuclear receptor coactivator 1\* PNRC1 U03105 na nd no change up up (2.4) nd na na inhibitor of differentiation\* ID1 AA457158 down (3.8) down na nd na nd na nd E-cadherin\\ CDH1 Z13009 na nd no change nd up nd no change na hsp27\\ HSPB1 U90906 na nd up nd up nd no change na p53\\ TP53 U94788 na nd up nd no change nd no change na N-myc downstream regulated gene 1\\ NDRG1 D87953 na nd up nd up up up na NHEK 30 min 3h 12h 24h 48h Gene Common Genbank RT-PCR Affymetrix 1 h RT-PCR Affymetrix RT-PCR Affymetrix RT-PCR Affymetrix RT-PCR Affymetrix matrix metalloproteinase 9\* MMP9 J05070 no change nd no change nd na up na up up (11.1) nd aldo-keto reductase\* AKR1C3 D17793 na nd down (4.0) nd na nd na nd down (5.3) nd aquaporin 3\* AQP3 N74607 no change nd no change nd na nd na up up (2.8) nd plasminogen activator, tissue\* PLAT M15518 no change nd no change nd na nd na down no change nd HMG-box transcription factor 1\* HBP1 AF019214 down (6.3) nd down (4.0) nd na up na nd no change nd Cox-2\* PTGS2 U04636 down (7.7) nd down (6.3) nd na nd na down down (8.3) nd v-erb-b2\* ERBB3 M34309 na nd down (2.4) nd na up na nd down (4.5) nd proline-rich nuclear receptor coactivator 1\* PNRC1 U03105 na nd down (5.3) nd na up na nd no change nd inhibitor of differentiation\* ID1 AA457158 down (8.3) down na nd na nd na nd na nd E-cadherin\\ CDH1 Z13009 na nd na nd no change nd up nd up nd hsp27\\ HSPB1 U90906 na nd na nd up nd up nd up nd p53\\ TP53 U94788 na nd na nd no change nd up nd up nd N-myc downstream regulated gene 1\\ NDRG1 D87953 na nd na nd no change nd no change nd no change nd Significant fold change: \<0.5 for down regulated and \>2.0 for up regulated genes, respectively. na = not analysed; nd = no data. \* Taqman RT-PCR was done on these genes using Assays-on-demand (Applied Biosystems). GAPDH was used as a reference gene. \\SYBR green RT-PCR was done on these genes using in house designed primers. Transferrin receptor was used as a reference gene. ::: For the second method, in-house designed gene-specific primers for six human transcripts (Transferrin receptor (TFR): NM\_003234 (5\'-TCCCTAGGAGGCCGTTTCC-3\'and 5\'-GCCTACCCATTCGTGGTGAT-3), COX-2: NM\_000963 (5\'-GCATTGGAAACATCGACAGTGT-3\'and 5\'-TGACGTCTTTTTACTTGAATTTCAACTTATAT-3); NDRG1: NM\_006096 (5\'-CCAGTGCGGCTGCCAG-3\'and 5\'-TTCCTATGAGAAAATCCACGGTG-3); TP53: NM\_000546 (5\'-CCTTGAGGGTGCCTGTTCC-3\'and 5\'-CCCTCTACCTAACCAGCTGCC-3\');CDH1: NM\_004360 (\'5\'-TGAAGACCTTTAATGGCTTCCC-3\'and 5\'-CACACTTACTCAGAACAAGTCACTGG-3\'); HSPB1: NM\_001540) (5\'-AAAATCCGATGAGACTGCCG-3\' and 5\'-GCACAGCCAGTGGCGG-3\') were designed using the Primer Express software (Applied Biosystems, Foster City, CA, USA). Real time RT-PCR analysis was performed in triplicate using ds cDNA synthesized from mRNA obtained from the investigated cells. All PCRs were performed at 60°C annealing temperature and the TFRgene was used as internal standard. A PCR mastermixture was prepared using the SYBRGreen PCR Core Reagents (Applied Biosystems, Foster City, USA) and aliquoted into microplate wells together with 1 μl template and 5 pmol of each primer for a final volume of 25 μl per reaction. The iCycler (Bio-Rad) was used for PCR and detection of fluorescent signals. Standard curves (C~T~versus log concentration) were generated for each primer pair using duplicate cDNA samples in a series of consecutive 5-fold dilutions. Efficiency calculations (E=(10\^(-1/amplificationslope)-1) were performed to validate compatibility of investigated genes with the internal control. The compatibility of all pair-wise compared amplification efficiencies were confirmed with a maximal deviation of 10% (data not shown). The specificity of all individual amplification reactions was confirmed by meltcurve analysis (data not shown). The relative mRNA expression levels of the target genes in each sample were calculated using the comparative C~T~method. The C~T~value was defined as the number of PCR cycles required for the fluorescence signal to exceed the threshold value. C~T~values were defined as the absolute value of the difference between the C~T~of the target RNA and the Ct of the housekeeping gene RNA for each sample. The level of significance was set to a 2-fold relative difference between samples, i.e. significant fold change 0.5 \< 2^-ΔΔC^~T~\> 2 for down- and up-regulated genes respectively. Results ======= Experimental overview --------------------- The aim of this work was to study the molecular effects of the antioxidant N-actetyl- cysteine (NAC) on proliferating cells, by gene expression analysis using the Affymetrix GeneChip platform. Previous studies on cell culture systems, demonstrated mainly by morphological and biochemical data, have indicated that NAC addition stimulates proliferating cells to go into differentiation phase \[[@B1]\]. Figure [1](#F1){ref-type="fig"} displays a scanning electro micrograph image of the cell types tested and demonstrates the morphological shift from proliferating to differentiating cells after addition of NAC. We have performed a microarray-based gene expression analysis of the human colon carcinoma cell line (Caco-2) and normal human epidermal keratinocytes (NHEK) over time (1, 12, and 24 hrs) after addition of NAC, compared to untreated, control samples at the same time points. Data obtained from GeneChip analysis were processed using the RMA analysis approach and samples were normalized to their corresponding controls. Filtering was done based on two criteria (p-value \< 0.1 and \>1.5 fold up-or down-regulation) for each time point. The labelling and hybridisation was done in duplicate, except at 24 hrs NAC treatment of NHEK, where a single measurement was made. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Scanning electron micrographs showing the effect of NAC treatment on the morphology of NHEK (A) and Caco-2 (B) cells.A1) NHEK untreated controls show a heterogeneous polygonal morphology, with a villous surface and relatively broad intercellular space; A2) After 72 h with 2 mM NAC cells grow flat in a thin monolayer with a regular polygonal morphology and a smooth surface with loss of the fine microvillous structure, also showing dramatically reduced intercellular space. B1) Proliferating Caco-2 cells display an irregular morphology with scarce microvillous structures and large intercellular space; B2) After 72 h with 2 mM NAC Caco-2 cells had the morphology of end-stage differentiated cells, with a regularly polygonal and about three times thicker than controls, with a relevant number of brush border microvilli at the cell surface (black arrows) as well as a dramatically reduced intercellular space (white arrows). Bars: 20 μm ::: ![](1471-2407-5-75-1) ::: Analysis -------- Gene lists of up- and down-regulated genes at each time point for NHEK and Caco-2 (1, 12, 24 hrs) were generated. In addition, time points within each cell type were analysed to identify similarities/differences among early and late time-point responses. Any results that produced zero genes were skipped. A summary of the top ten differentially expressed genes in Caco-2 and NHEK, at all time points, is depicted in additional files [1](#S1){ref-type="supplementary-material"} and [2](#S2){ref-type="supplementary-material"}. The complete lists of differentially expressed genes are accessible as additional datafiles [3](#S3){ref-type="supplementary-material"} and [4](#S4){ref-type="supplementary-material"} or at <http://biobase.biotech.kth.se/NACmicroarray>. There were no general differences in the number of transcripts detected in treated as compared to untreated samples in any of the cell lines. Following NAC induced differentiation in Caco-2 and NHEK, 253 and 414 targets, respectively, were significantly differentially expressed across the time series, according to above statistical criteria (Table [1](#T1){ref-type="table"}). Generally, for both cell types, the amount of significantly up/down-regulated genes is limited to \< 200 genes after consideration of multiple appearances. In Caco-2, a tendency towards more down-regulated genes was identified, while in NHEK, the numbers of genes induced was approximately equal to those repressed (Table [1](#T1){ref-type="table"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Number of genes induced/repressed at each time point. ::: Time points Caco-2 NHEK -------------------- ------------ ---------- 1 hr (repressed) 8 16 12 hrs (repressed) 54 10 24 hrs (repressed) 86 171 1 hrs (induced) 1 26 12 hrs (induced) 69 28 24 hrs (induced) 35 163 ::: Both Caco-2 and NHEK exhibit a relatively limited early response at 1 hr, followed by an increasingly stronger response at 12 and 24 hrs. The initial response could be characterized as transient, since most of the genes induced/repressed initially either changed in the opposite direction later on, or simply went back to normal expression levels. The induced/repressed genes in the two cell types are generally quite different despite their morphological similarities after NAC treatment. This is for example demonstrated in that most of the genes induced/repressed in Caco-2 show no change in expression levels in NHEK and vice versa (see below). A direct comparison of the corresponding differentially expressed genes is also provided. Caco-2 analysis --------------- The general trends for Caco-2 cells are depicted in Figure [2](#F2){ref-type="fig"} based on the collection of genes that passed the set criteria. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Matrix view of significantly affected genes.Genes induced/repressed in Caco-2 relative to control and their corresponding expression profiles in NHEK. The color is specific for genelist of affected genes at each time point (unless these are the same genes as from one of the earlier time points). Clearly, the early response at 1 hr is very limited (1,8 genes in Caco-2). There are also almost no genes induced/repressed at 1 hr that continue the trend at later time points -- indicating an early transient response. This is markedly different for 12 hr and 24 hr time points that show many common genes. It is also clear that many genes affected by NAC treatment in one cell line don\'t show the same response in the other cell line. ::: ![](1471-2407-5-75-2) ::: The graph shows the behaviour of genes induced/repressed in Caco-2 across the time courses in both cell types (signal represents the fold difference between each treatment and the average of corresponding controls). As indicated above, there is a small transient change at 1 hr, and a more significant change, taking place at later time points. In particular, it is observed that the majority of genes repressed at 12 hrs continue their trend at 24 hrs (blue lines correspond to 12 hrs genes, dark green line corresponds to 24 hrs genes only). Similarly, many genes up-regulated at 12 hrs continue the trend at 24 hrs (red line 12 hrs genes, orange line 24 hrs genes only). (Gene lists showing multiple appearances in the same cell type, Caco-2, at more than one time point can be found in [additional file 4](#S4){ref-type="supplementary-material"} or at <http://biobase.biotech.kth.se/NACmicroarray>). The expression pattern of the corresponding genes in the other cell system NHEK shows weak or no correlation. Table [1](#T1){ref-type="table"} represents the number of genes induced/repressed at each time point for Caco-2 cells and NHEK cells, including those with changes across multiple time points. Again, it is evident that early activity (1 hr) is limited and quite different from the activity at later time points and that few genes induced/repressed early on (1 hr) continue their trend at later time points (Figure [2](#F2){ref-type="fig"}). This has also been observed after 24 and 48 hrs using another microarray platform based on spotted cDNA arrays (data not shown). NHEK analysis ------------- The NHEK cells were analysed in a similar manner (see [additional file 3](#S3){ref-type="supplementary-material"} or <http://biobase.biotech.kth.se/NACmicroarray>). The t-test p-value for NHEK 24 hrs is not included, since there were no treatment replicates; an error model was used instead. However, some of the values in the error model do not correspond, in this case, to other t-test p-values and are therefore not included. A similar trend to that described for Caco-2 takes place in these cells, with early transient response at 1 hr, followed by a more extensive response at later time points (12, 24 hrs) as demonstrated in Figure [3](#F3){ref-type="fig"}. Again, the genes active at 1 hr do not appear to be active at later time points (except for a few genes in each case). On the other hand, genes active at 12 hrs also appear active in the same direction at 24 hrs. Table [1](#T1){ref-type="table"} includes the exact numbers of induced/repressed genes. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Matrix view of significantly affected genes.Genes induced/repressed in NHEK relative to control. The color is specific for genelist of affected genes at each time point (unless these are the same genes as from one of the earlier time points). Also in these cells, quite a limited early response is observed at 1 hr (16 and 26 genes in NHEK). There are almost no genes induced/repressed at 1 hr that continue the trend at later time points -indicating an early transient response. This is markedly different for 12 hrs and 24 hrs treatments that show many common genes. It is also clear that many genes affected by NAC treatment in one cell line don\'t show the same response in the other cell line. ::: ![](1471-2407-5-75-3) ::: NHEK vs. Caco-2 analysis ------------------------ Comparisons of NAC treated Caco-2 and NHEK show that the responses are very different. Only a few genes were regulated similarly in both cells (Table [2](#T2){ref-type="table"}). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Genes common in both Caco-2 and NHEK at the different time points. ::: Time point Regulation Probe ID Gene ------------ ------------ ------------------------------ ---------------------------------------------------------------------------------------------------------------------------------------------- 1 hr repressed 36618\_g\_at Inhibitor of DNA binding 1, dominant negative helix-loop-helix protein 12 hrs repressed none 24 hrs repressed 33720\_at 1069\_at Putatative 28 kDa protein, Human cyclooxygenase-2 (hCox-2) gene, complete cds 1 hr induced none 12 hrs induced 39809\_at 36980\_at 1585\_at HMG-box transcription factor 1 proline-rich nuclear receptor coactivator 1 v-erb-b2 erythroblastic leukemia viral oncogene homolog 3 (avian) 24hrs induced 39248\_at Aquaporin 3 ::: Even if we consider that statistical considerations (filtering etc.) inaccurately prevented a number of genes from overlapping in analysis of two cells, the clustering in Figure [4](#F4){ref-type="fig"} show that the responses are indeed very different. The hierarchical clustering is based on filtered genes (significant change in at least one time point) for all time points and each cell type. At the time point 1 h after treatment, there is no cell-specific clustering. Actually, both cell types are in the same cluster for 1 hr. It is in the later time points, with much stronger and distinct response, that the cell-specific clusters form (blue = NHEK, green = Caco-2). ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Hierarchical clustering diagram at 1, 12 and 24 hrs.A set of 2054 genes which show significant changes in expression in at least one of the time points in either of the two cell line clustered using the standard correlation. Cluster of the corresponding time points for each cell line indicates very close similarity of NHEK and Caco-2 at time point 1 hr, at which the NAC effect is not yet pronounced (both cell lines form a tight separate cluster at 1 hr). However, during the later time points the changes are much more significant and the differences between Caco-2 and NHEK become very pronounced as well (each cell line forms a separate cluster that includes both 12 and 24 hrs time points). ::: ![](1471-2407-5-75-4) ::: A set of genes indicated to be differentially expressed were chosen to validate the analysis (Table [3](#T3){ref-type="table"}). Additional time points were also used in this analysis. The genes represented both markers of proliferation and motility as well as new candidates in these and other related processes and in many of these cases the differential gene expression could be confirmed. Gene ontology analysis of Caco-2 and NHEK ----------------------------------------- In order to perform Gene Ontology analysis of significantly regulated genes in Caco-2 and NHEK we created combined lists of genes up- and down-regulated at any of the three time points. In order to statistically assess the overrepresentation of each category we used the publicly available tool EASE. For each cell line, the down-regulated genes were assessed for gene ontology enrichment relative to the up-regulated genes in the same cell line, within the universe of all genes being the Affymetrix U95aVer2 probe set. The final table of gene ontology groups was selected based on the number of overlapping genes, EASE score, illustrating common biological mechanisms or pathways (e.g. motility, cell division). The results of analysis in both cell lines indicate a strong tendency towards differentiation by halting cell proliferation and related processes, Table [4](#T4){ref-type="table"} (NHEK) and Table [5](#T5){ref-type="table"} (Caco-2). However, the mechanisms inhibiting cell proliferation in NHEK and Caco-2 cell lines appear to differ significantly. As seen from the two tables, genes responsible for proliferation are shut down in both cell lines, with NHEK showing a stronger direct proliferation effect (more genes are shut down). This effect, however, can be explained by a considerably higher number of significantly regulated genes identified in NHEK vs. Caco-2. Furthermore, the mechanism of inhibiting proliferation in NHEK appears to be directly related to the shut down of processes such as nuclear division, mitosis, cytokinesis etc. -- all required for cells to divide and proliferate. In Caco-2, on the other hand, the mechanism appears to involve cell structure inhibition, through the shut down of genes involved in adhesion and cytoskeleton changes. Finally, the motility processes critical to Caco-2 inhibition of proliferation are affected in NHEK, while cell division seems to be affected to a lesser degree in Caco-2. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Gene ontology for differentially expressed genes found in NHEK cells ::: Gene Category Down (genes) Up (genes) Combined list Affy probes EASE score -------------------------------------------------------------------- ------------------ ---------------- ------------------------------- ---------------- mitotic cell cycle 41 4 49 1.96E-14 cell cycle 52 12 76 2.03E-12 M phase 22 1 24 7.73E-09 DNA replication and chromosome cycle 22 1 24 7.73E-09 cell proliferation 56 22 100 0.000000024 nuclear division 20 1 22 7.23E-08 M phase of mitotic cell cycle 19 1 21 0.000000217 DNA replication 19 1 21 0.000000217 S phase of mitotic cell cycle 19 1 21 0.000000217 mitosis 19 1 21 0.000000217 DNA metabolism 29 6 41 0.000000403 DNA dependent DNA replication 13 0 13 0.00000641 cytokinesis 12 0 12 0.0000199 metabolism 100 79 256 0.000126 regulation of cell cycle 26 11 48 0.00137 cell growth and/or maintenance 79 62 202 0.00241 nucleobase\\, nucleoside\\, nucleotide and nucleic acid metabolism 53 36 124 0.00262 biosynthesis 15 5 24 0.00475 protein metabolism 43 29 101 0.00992 physiological process 132 130 389 0.0103 obsolete biological process 17 7 31 0.0125 macromolecule biosynthesis 11 4 18 0.0267 protein modification 22 14 50 0.0639 cytoplasm organization and biogenesis 12 6 23 0.0693 cell organization and biogenesis 17 13 42 0.211 phosphorus metabolism 16 12 40 0.244 phosphate metabolism 16 12 40 0.244 protein amino acid phosphorylation 12 9 30 0.326 phosphorylation 12 10 32 0.424 macromolecule catabolism 10 8 26 0.435 protein catabolism 10 8 26 0.435 proteolysis and peptidolysis 10 8 26 0.435 cell adhesion 4 16 36 1 cell motility 2 6 14 0.995 Bold:processes significantly affected in Caco-2 pointing to differentiation Italicized:processes significantly downregulated in NHEK, pointing to differentiation Cutoff (EASE score \< 5.00E-01) Cutoff (\#overlapping genes \> 9, larger than in Caco-2 due to a large initial list size). ::: ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Gene ontology for differentially expressed genes found in Caco-2 cells ::: Gene Category Down (genes) Up (genes) Combined list Affy probes EASE score -------------------------------------------------------------------- ------------------ ---------------- ------------------------------- ---------------- cell motility 8 1 10 0.0459 cytoplasm organization and biogenesis 8 1 10 0.0459 muscle development 7 1 9 0.0872 morphogenesis 18 7 33 0.0879 development 22 10 43 0.107 cell adhesion 11 3 18 0.112 cell surface receptor linked signal transduction 11 3 18 0.112 organelle organization and biogenesis 6 1 8 0.158 cytoskeleton organization and biogenesis 6 1 8 0.158 cell organization and biogenesis 11 4 19 0.159 organogenesis 16 7 31 0.179 response to stress 8 3 14 0.282 macromolecule biosynthesis 9 4 17 0.331 cell-cell signaling 6 2 10 0.346 signal transduction 19 5 30 0.0118 immune response 9 1 11 0.0233 protein metabolism 15 4 23 0.0235 cellular process 58 34 126 0.0316 defense response 10 2 14 0.0467 transcription\\, DNA-dependent 17 10 37 0.358 transcription 17 10 37 0.358 response to biotic stimulus 10 5 20 0.374 nucleobase\\, nucleoside\\, nucleotide and nucleic acid metabolism 24 16 56 0.434 regulation of transcription\\, DNA-dependent 16 10 36 0.442 regulation of transcription 16 10 36 0.442 cell cycle 9 5 19 0.486 nucleotide metabolism 3 2 7 0.835 nuclear division 3 3 9 0.927 mitosis 3 3 9 0.927 mitotic cell cycle 5 3 11 0.691 cytokinesis 2 2 6 0.952 Bold:processes significantly downregulated in NHEK pointing to differentiation Italicized:specific processes significantly downregulated in Caco-2, pointing to differentiation Cutoff (EASE score) \< 5.00E-01 Cutoff (Number of overlapping genes \> 5 ::: Discussion ========== Deregulation of proliferation is a characteristic of tumorigenesis and therapeutic approaches for cancer treatment targets apoptosis, cell cycle arrest and differentiation. NAC has been shown to induce a multitude of molecular changes related to tumorigenesis \[[@B4]\]. Recently, NAC has been demonstrated to inhibit apoptosis \[[@B5],[@B6]\], possess anti-inflammatory activities \[[@B7]\] and inhibit proliferation \[[@B8]\]. Here, we have monitored the reflection in global gene expression profiles of the transition from proliferation to a differentiated state in normal and cancer cells in vitro, as induced by NAC. Two out of three previous studies of the global gene expression that accompanies the spontaneous differentiation of Caco-2 report a general down-regulation of gene expression in differentiated cells as compared to the proliferating counterpart \[[@B9]-[@B11]\]. A similar, but not as pronounced, trend is reflected in the number of genes differentially expressed following NAC induced differentiation in Caco-2. The expression level of 253 targets in Caco-2 and 414 in NHEK were statistically differentially expressed at different time points. Multiple appearances of differentially regulated transcripts were common, resulting in detection of totally less than 200 unique genes, respectively. This is fewer than expected in comparison to previous reports on differential regulation during spontaneous Caco-2 differentiation and probably due to the difference in stringency of the algorithms used for data analyses (MAS 4.0 vs RMA), rather than related to functional biological discrepancies. Initial response ---------------- In both data sets, the early responses were relatively limited and appeared to be transient, indicating that a large part of initial immediate early events occur at the level of translation and post-translational modifications. The nature of initial regulatory events is also expected to be transient due to feedback inhibition as well as to a restricted number of NAC induced mechanisms. Interestingly, significant transcriptional down-regulation of the inhibitor of differentiation 1 (ID-1)was found at 1 hr in both Caco-2 and NHEK, suggesting a common mechanism of NAC induced differentiation and inhibition of proliferation in NHEK and Caco-2 epithelial cells. Analysis of ID-1 expression levels by real-time quantitative PCR confirmed the suppression of the transcript in both cell types (Table [3](#T3){ref-type="table"}). ID-1 has been demonstrated to bind helix-loop helix transcription factors, preventing them from binding DNA \[[@B12]\]. In particular, ID-1 has been shown to be required for G1 progression, and its constitutive expression inhibited the lineage commitment and differentiation in B-cells \[[@B13],[@B14]\]. Previous reports have also shown that ID-1 is a negative transcriptional regulator of CDKN2A(p16/p14/p19), which induces G1 arrest through the inhibition of Rb phosphorylation by cdk -4 and -6 \[[@B15]\]. Overexpression of ID-1was also reported in psoriatic involved skin \[[@B16]\]. Inhibitors of histone deacetylase activity are emerging as a potentially important new class of anticancer agents. The cell cycle blockade and differentiation caused by such a drug, trichostatin A, caused decreased levels of ID-1 consistent with cell cycle senescence and differentiation of A2780 ovarian cancer cells \[[@B17]\]. Vitamin D is also known to promote differentiation and was shown by others to down-regulate ID-1through a suppressive vitamin D response sequence in the 5\'of the gene \[[@B18]\]. The ID-1expression is regulated by a protein complex containing the immediate-early response gene EGR1\[[@B19]\]. The growth regulatory properties of EGR1 have been found to involve coordinated regulation of TGF-β~1~and fibronectin (FN1). The resulting proteins are secreted and lead to increased expression of plasminogen activator inhibitor-1 (PAI1). Both the secreted FN1 and PAI1 functions to enhance cell attachment and normal cell growth \[[@B20]\]. We detect the induction of both fibronectin and PAI1 in NHEK cells at both 12 and 24 hrs, suggesting a role of EGR1 pathways in the NAC mediated mechanism at least in this cell type. Other interesting down-regulated genes at 1 h after NAC treatment in NHEK included for example regulated in development and DNA damage response 1(REDD1), squamous cell carcinoma antigen 1(SCCA1), highly expressed in cancer(HEC), s100a9and kallikrein 7(also termed stratum corneum chymotryptic enzyme-- SCCE). REDD1has previously been shown to be down-regulated in differentiating primary human keratinocytes and ectopic expression inhibits in vitro differentiation \[[@B21]\]. The suppression of SCCA1has been demonstrated to inhibit tumour growth \[[@B22]\], HECexpression is increased in tumours \[[@B23]\] and S100A9has been found to be up-regulated in psoriasis patients displaying keratinocyte hyperproliferation and altered differentiation \[[@B24]\]. SCCEhas been suggested to play a role in desquamation and its up-regulation is associated with poor prognosis of ovarian and breast cancer \[[@B25],[@B26]\]. A number of interesting up-regulated genes, such as activin Aand WEE1, were also identified in NHEK 1 hr after NAC treatment. Over expression of WEE1 inhibits cell cycle progression by inactivation of the CDC2/cylin B complex \[[@B27]\], while activin Ais a member of the TGF-β family of cytokines which is known to promote growth arrest and differentiation in several tissues including intestinal epithelia \[[@B28],[@B29]\]. In fact it was first identified as a protein that exhibits a potent differentiation-inducing activity \[[@B30]\]. In Caco-2, the proto-oncogene fosand the transcription factor HNF3A, which have been observed to be amplified in human malignancies \[[@B31]\], was two of the identified genes being repressed at 1 hr after NAC treatment. A single gene, integrin alpha 2, was found up-regulated at this time-point. It is clear that these data collectively describe molecular changes associated with the mediation of a differentiated epithelial phenotype. A large part of the differentially expressed genes have clear implications in withdrawal of mitogenic signals and in promotion of growth arrest. Multiple signalling pathways are suggested to be involved. Late response (multiple occurrences) ------------------------------------ Progressively more genes were affected in both cell types, and many showed similar trends in direction, over later time points. Cluster analysis revealed tightly linked genes between the 12 and 24 hrs time points in the same cell type and genes with such multiple appearances are potentially more strongly implicated in the differentiation process. In NHEK cells, for example the down-regulated mitogens neuregulin 1(heregulin) and melanoma growth stimulatory activity(MGSA) \[[@B32]\], belong to this group. MGSAbelongs to a super family of chemochines, including IL-8, which is involved in inflammatory processes. Heregulin is known to activate the oncogenic ERBB2 receptor \[[@B33]\]. Cdc-6, which are regulated in response to mitogenic signals, binds PCNA and is required for initiation of DNA replication \[[@B34]\], was also repressed at both 12 and 24 hrs after NAC treatment, implying programs involving withdrawal of mitogenic factors as important mechanisms for NAC mediated inhibition of proliferation and increased differentiation in NHEK cells. The expression of Topoisomerase II(TOP2) was also repressed, confirming results obtained in NAC treated CHO cells \[[@B35]\]. Topoisomerases control and alter the topologic states of DNA, and the relaxation activity of TOP2 is essential for productive RNA synthesis on nucleosomal DNA \[[@B36]\]. The list of correspondingly important up-regulated genes in NHEK was extensive and included activin A, transglutaminase 2(TGM2), ErbB3(HER3), matrix metalloproteinase 9(MMP-9), fibronectin (FN!), PAI1and TGFβ among others. Notably, activin Awas the only gene found to be up-regulated at all investigated time points, demonstrating a sustained growth inhibitory and differentiation promoting signal. TGM2 catalyses cross-linking of proteins, demonstrates G-protein function in receptor signalling \[[@B37]\] and was recently reported to phosphorylate IGFB3 \[[@B38]\]. IGFB3 in turn has a major role in regulation of proliferation as a growth inhibitor through IGF2 binding and alternative IGF2 independent pathways \[[@B39]\]. Thus implicating potential regulatory functions of TGM2in proliferation and differentiation. Surprisingly, ErbB3, which promotes proliferation through the Wnt signalling pathway, was also up-regulated. RT-PCR could confirm the induction (Caco-2, 12 hrs) and a study investigating spontaneous Caco-2 differentiation is also in agreement with our data on up-regulation \[[@B11]\]. In contrast, a recent publication reported its up-regulation in breast cancer \[[@B40]\], suggesting a dual role of ERBB3 in cell cycle regulation. The induction of MMP9, confirmed by RT-PCR (48 h NHEK) is in additional contrast to our observations of NAC-induced cell differentiation and proliferation. MMP9 has been associated with angiogenesis, tumour progression and metastasis as mediated through degradation of the extracellular matrix (ECM) \[[@B41]\] and stimulation of hyperproliferation \[[@B42]\]. However, tumours with low levels of MMP9 were found to be less differentiated. Thus, although MMP9 stimulates proliferation, it is also implied in positive regulation of differentiation \[[@B42]\]. NAC has been proposed to inhibit activation of latent MMP9 protein in ECM reservoirs by removal of its propeptide and by competing for the zink ion which is necessary for enzymatic function \[[@B43]\]. Hence, our data may imply that post-transcriptional regulation of MMP9 prevails over the transcriptional changes as the major control mechanism. A large number of multiple occurring differentially expressed transcripts were demonstrated in Caco-2, including intestinal trefoilfactor 3(TFF3) and Aquaporin 3(AQP3) among others. TFF3 has been shown to have a central role in the maintenance and repair of intestinal mucosa \[[@B44]\] and upregulation is expected during differentiation. Aquaporins (AQPs) are water channel proteins, important for the transport of water and other small proteins across the cell membrane \[[@B45]\]. AQP1has previously been shown to be involved in cell cycle control \[[@B46]\], suggesting that AQP3may also have a role in the progression of cancer. AQP3has been reported to be highly expressed in several types of stratified epithelial cells in rat, including the epidermis \[[@B47]\] and the differentiated cells of the gastrointestinal tract \[[@B48]\]. The expression of AQP3was reported to be up-regulated in differentiating Caco-2 cells \[[@B11]\], while expression was shown to be down-regulated in differentiating primary keratinocytes \[[@B49]\]. In this study AQP appears to have a transient behaviour in Caco-2 cells with a repressed behaviour at initial time point and induced pattern at later time points (confirmed by RT-PCR), indicating a remodelling of cell membrane constituents. The genes repressed in Caco-2 included for example Cyclin D1, Inhibin beta B, BMP-2and FHL-2. The D1 cyclin is involved in β-catenin-TCF signalling and its down-regulation induce G1 arrest \[[@B50]\]. FHL-2 has been demonstrated to be a coactivator of β-catenin from cyclin D and IL-8 promoters in a colon cell line \[[@B51]\], suggesting that repression of FHL-2 may also repress growth. Inhibin betaand BMP-2are members of the TGFfamily of genes. Inhibin beta is an antagonist of activin A activity and consequently represses differentiation and promotes growth \[[@B52]\]. BMP-2on the other hand, has been demonstrated both to induce apoptosis \[[@B53]\] and growth inhibition/differentiation \[[@B54]\]. In contrast, another recent study demonstrated the ability of BMP-2to enhance the growth of tumours \[[@B55]\]. ### Late response (single occurrence) A vast number of additional interesting genes with potentially important roles in mediation and manifestation of the differentiated epithelial phenotype was identified as significantly induced or repressed in a single time point. As an example, up-regulation in NHEK after 24 hrs was seen for transcripts encoding CDKN2B(p15), which is believed to be an effector of TGF-β induced G1 arrest and inhibition of proliferation \[[@B56]\], and for CDKN1C, which is a p21homologue and negative regulator of cell proliferation \[[@B57]\]. Other up-regulated genes at this time point were the transcription factor Junand BTG1. Junwas recently demonstrated to be a regulator of erythroid differentiation \[[@B58]\] and Jun Bknock out mice have been shown to develop a proliferative disease resembeling human chronic myeloid leukemia \[[@B59]\]. BTG1has been proposed to belong to a family of antiproliferative genes \[[@B60]\]. Up-regulation was also found for cadherin 13, a gene with growth inhibitory functions that is expressed in normal cells but not in the majority of human tumour cells of epithelial origin \[[@B61]\]. While not identified by global transcript analysis, RT-PCR investigation revealed increased levels of E-cadherinin both Caco-2 (24 hrs) and NHEK (24 and 48 hrs), this in agreement with the previously reported immunohistochemical data \[[@B1]\] that showed increased staining of E-cadherin in NAC treated cells. In Caco-2 cells, the differentiation-related gene NDRG1, which is expressed during differentiation and down-regulated in colorectal neoplasms \[[@B62]\], was up-regulated at 12 hrs. This increase was also demonstrated by RT-PCR at 12, 24 and 48 hrs after NAC treatment. The induction of Cdx2may also be a functional change, since reduced expression of Cdx2has been shown to be important in colon tumorigenesis \[[@B63]\]. Interestingly, in correlation with the controversial results from NHEK, an up-regulation of ErbB3was identified in Caco-2 at 12 hrs. In addition, the oncogene mycwas also up-regulated in contrast to the expected decrease. The down-regulation of Cox-2and BMP-2was also in concordance with NHEK data. In addition, the TGF-β family member BMP-4was repressed. This correlates well with our results on repressed ID-1expression, since both BMP-2 and -4 up-regulate ID-1\[[@B64],[@B65]\]. Hence, although induced by the same mechanism (NAC) and yielding the same end-stage of growth inhibition and differentiation, the processes in NHEK and Caco-2 are on the whole quite different. This is demonstrated by gene specific differences that result in lack of correlation between cell types at the same time point after treatment as identified by cluster analysis. This is furthermore supported by the Gene Ontology analysis, tables [4](#T4){ref-type="table"} and [5](#T5){ref-type="table"}, indicating that the two cell lines achieve their differentiated states using two distinct mechanisms, this in concordance with previously observed effects of NAC treatment on cell morphology and growth arrest \[[@B1]\]. It should be noted that apoptosis does not appear to act as a regulating mechanism, since only a very small proportion of apoptotic genes are affected in either direction (2 to 12 genes out of 317, in either direction). This is also supported by the previous study \[[@B1]\], which analysed apoptosis by propidium iodide labelling and flow cytometry. As additional testimony to the lineage specific differentiation programs, only very few genes were identified as being similarly regulated in both cell-types. These included ID-1, AQP3and ErbB3(as described above). Cox-2was also similarly down-regulated in both cell types after NAC treatment. When investigating Cox-2expression by RT-PCR we could confirm the down-regulation at 24 hrs in Caco-2 and identify an additional repression in NHEK at 1, 3 and 48 hrs after NAC treatment. Overexpression of Cox-2has been shown to promote cell migration and invasion in Caco-2 cells \[[@B66]\] and to regulate colon carcinoma induced angiogenesis by production of angiogenic factors \[[@B67]\]. In addition, epidermal differentiation is also affected by Cox-2over expression. Cox-2seems to prevent entrance into the postmitotic state, which is coupled to the switching on expression of differentiation-associated proteins, allowing keratinocytes to proliferate \[[@B68]\]. In correlation with our data, a NAC mediated inhibition of Cox-2expression have previously been demonstrated in colorectal cancer \[[@B69]\]. Hence, it is likely that Cox-2repression is a NAC specific event endorsing differentiation/growth arrest in both NHEK and Caco-2. HBP1was also up-regulated in both Caco-2 and NHEK after 12 h of NAC treatment and RT-PCR could confirm an increase in Caco-2 at 12 and 24 hrs. This corresponds to previous findings, where HBP1has been seen to have a negative effect on tumours. It has previously been established that HBP1is a target of the retinoblastoma pathways \[[@B70],[@B71]\] and that HBP1negatively regulates Wnt/β-catenin, thus inhibiting proliferation and suggesting that HBP1may have a tumour suppressor function \[[@B72]\]. Two additional proteins, putative 28 kDa proteinand proline rich nuclear receptor coactivator-1(PNRC1), were also identified as differentially expressed in both cell types. RT-PCR analysis was able to confirm an up-regulation of PNRC1in Caco-2 24 hrs after treatment. However, these genes are not previously described and potential functions remain unresolved. A proportion of the differentially expressed transcripts were not possible to predict as being part of the differentiation context. For example, ErbB3, fos, TGFβIand mycwere found to be expressed at higher levels in differentiated cells, in contrast to their roles in promotion of proliferation. Interestingly, a similar contrasting increase in fosand ErbB3levels was found in normal colonic cells as compared to colorectal cancers in a SAGE study \[[@B73]\]. It is important to note that spontaneous morphological and functional differentiation in Caco-2 have been demonstrated not to be coupled, with independent mosaic patterns of proliferating and differentiated cells present adjacent in the cell culture \[[@B74]\], which may in part explain some of the contrasting results. However, many of the detected differentially expressed genes in this study have previously been described as altered in differentiated epithelial cells. In NHEK cells we could for example confirm the expected increased expression of SPRR1Band SPRR2A. In Caco-2 the up-regulation of AQP3, NDRG1and TFF3were among the genes that validated the results from the global transcription profile analysis. Interestingly, several genes with major relevance in psoriasis have been found differentially expressed as a consequence of NAC treatment in the particular epithelial cells included in this study, for example S100A9, ID-1 and Cox-2. These findings could give a mechanistic background to the ongoing clinical studies being carried out based on empirical NAC treatment of patients having psoriasis. Alternative NAC signalling mechanisms at the level of proteins and metabolites may also be important. For instance the phospholipid modulator, platelet activating factor, has been demonstrated to induce differentiation and inhibit proliferation in colon cells \[[@B75]\], and inhibit proliferation in cultured human keratinocytes \[[@B76]\]. Accordingly, we are performing supplementary proteome and metabolome studies. In addition, analyses of additional cell lines for finding a common pathway of molecular changes that result from NAC induced differentiation are being considered. Conclusion ========== Our data demonstrate that NAC stimulated differentiation induces a limited and transient early transcriptional change, followed by a more constitutive and extensively different expression at later time points in both NHEK and Caco-2 cells. The genes affected are to a large extent related to inhibition of proliferation and stimulation of differentiation, but the responses are almost completely lineage specific. This and further analysis of NAC mediated expression changes provide a description of the complex molecular mechanisms of sulphydryl reductant treatment and potential targets for the development of new drugs for treatment of proliferation related epithelial disorders. List of abbreviations ===================== Inhibitor of differentiation 1 (ID-1) N-acetyl-L-cysteine (NAC) Normal human epidermal keratinocytes (NHEK) Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) Transferrin receptor (TFR) Fibronectin (FN1) Plasminogen activator inhibitor-1 (PAI1) Regulated in development and DNA damage response 1 (REDD1), Squamous cell carcinoma antigen 1 (SCCA1) Highly expressed in cancer (HEC) Stratum corneum chymotryptic enzyme (SCCE) Melanoma growth stimulatory activity (MGSA) Topoisomerase II (TOP2) Transglutaminase 2 (TGM2) Matrix metalloproteinase 9 (MMP-9) Intestinal trefoilfactor 3 (TFF3) Aquaporin 3 (AQP3) Proline rich nuclear receptor coactivator-1 (PNRC1) Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= ACG participated in the design of the study, drafted the manuscript, coordinated and carried out real-time kinetic PCR and Affymetrix experiments as well as performed initial data processing and data analysis. IK performed data processing, statistical analysis and data analysis as well as assisted with the manuscript. EER performed real-time kinetic PCRs and assisted in drafting the manuscript. LBL, GG and EKK cultured cells and isolated totRNA. JL, TP, TL and MCR directed the teams that carried out this study. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2407/5/75/prepub> Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Summary of the top 10 differentially expressed genes in Caco-2, at all time points studied. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Summary of the top 10 differentially expressed genes in NHEK, at all time points studied. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 3 All differentially expressed transcripts in NHEK at all time points. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 4 All differentially expressed transcripts in Caco-2 at all time points. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We kindly acknowledge The Knut and Alice Wallenberg Foundation, the Swedish Research Council, the Wallenberg Consortium North (WCN) for financial support, and the Nactilus AB, Malmö, Sweden (to LB-L, GG, EKK, TP).	PubMed Central	2024-06-05T03:55:59.936063	2005-7-7	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182358/", "journal": "BMC Cancer. 2005 Jul 7; 5:75", "authors": [ { "first": "Anna C", "last": "Gustafsson" }, { "first": "Ilya", "last": "Kupershmidt" }, { "first": "Esther", "last": "Edlundh-Rose" }, { "first": "Giulia", "last": "Greco" }, { "first": "Annalucia", "last": "Serafino" }, { "first": "Eva K", "last": "Krasnowska" }, { "first": "Thomas", "last": "Lundeberg" }, { "first": "Luisa", "last": "Bracci-Laudiero" }, { "first": "Maria-Concetta", "last": "Romano" }, { "first": "Tiziana", "last": "Parasassi" }, { "first": "Joakim", "last": "Lundeberg" } ] }
PMC1182359	Background ========== Gliomas are the most common subgroup of primary brain tumors. The majority of gliomas are classified on the basis of the histological appearance of the neoplastic cells: astrocytoma (WHO grades I to IV), oligodendroglioma (ODG; WHO grades II to III), and oligoastrocytoma (OAC; WHO grades II to III). ODG consist of cells morphologically resembling oligodendrocytes, and OAC consist of a spectrum of neoplastic cells, some with oligodendroglial features and others with astrocytic features. ODG and OAC often recur at a higher pathological grade, either as anaplastic ODG or OAC (grade III), or astrocytoma grade IV (glioblastoma, GBM) \[[@B1]\]. The exact genetic basis of the progression of lower-grade oligodendroglial tumors to higher-grade tumors is unclear. Both ODG and OAC have significantly better prognosis than GBM, with survival times ranging from 3 to 6 years, compared to 1 year for GBM \[[@B1]-[@B3]\]. Several genetic alterations in ODG and OAC, particularly the allelic loss of chromosome arms 1p and 19q, are used as prognostic indicators to predict longer recurrence-free survival after radiation therapy and/or chemotherapy \[[@B4]-[@B6]\]. In a previous study, we used gene expression profiling of a panel of 29 GBM specimens to identify a group of 50 named genes, designated as survival-associated genes, whose increased expression was inversely correlated with patient survival \[[@B7]\]. The tumor specimens were derived from 25 GBM tumors; in 4 of these specimens, 2 distinct regions from each tumor were analyzed. A two-step algorithm was used to identify the survival-associated genes \[[@B7]\]. First, by analyzing patient survival in relation to gene expression, a cluster of approximately 500 genes was identified whose increased expression in aggregate best inversely correlates with patient survival. This cluster of genes may represent a group of transcriptionally co-regulated genes, so it is plausible that they are involved in related biological processes. In the second step, we further selected 50 genes within this cluster that showed significantly greater variation in expression between different tumors than between the paired samples from the 4 specimens in which two distinct regions were analyzed. This subset of 50 genes is more likely to encode intrinsic properties related to the behavior and prognosis of individual GBM tumors. The prognostic value of one gene, brain-type fatty acid-binding protein (FABP7), was validated using two independent sets of GBM specimens \[[@B7]\]. The same microarray analyses also demonstrated that there was increased expression of the 50 survival-associated genes in ODG and OAC tumors compared to normal brain tissue \[[@B7]\]. This observation suggests that overexpression of the survival-associated genes may be common to different glioma subtypes. Expression of other genes in the group, such as Ptprz1\[[@B8]\], Bcan\[[@B9]\], and Crmp5\[[@B10]\], is also detected in oligodendrocytes in developing or injured brains, but their expression has not been reported in ODG and OAC. Two genes in the group, Olig1and Olig2, were of particular interest since OLIG1 and OLIG2 are present in neural progenitor cells and oligodendrocytes but not in astrocytes and neurons. Olig1and Olig2encode basic helix-loop-helix (bHLH) transcription factors that play a central role in oligodendrocyte development in the central nervous system \[[@B11],[@B12]\]. Olig1/Olig2-null mice show complete failure of oligodendrocyte development, and the precursor cells destined for an oligodendrocyte fate differentiate instead into astrocytes \[[@B12]\]. An initial report suggested that OLIG1 and OLIG2 might be markers for identification of oligodendroglial tumors \[[@B13]\]. More recent work, however, has demonstrated that OLIG1 and OLIG2 are expressed in all glioma subtypes including astrocytoma, ODG, and OAC \[[@B14],[@B15]\]. In the group of 50 survival-associated genes, we did not identify any gene that has been previously reported to be a target of OLIG1 or OLIG2, or encode any protein known to interact with OLIG1 and OLIG2. However, in the larger cluster of 500 genes, Id4, a member of the helix-loop-helix (HLH) transcription factor family, was present. Id proteins dimerize with bHLH factors to prevent transcriptional induction of downstream target genes, and regulate a variety of functions in normal and neoplastic cells \[[@B16]\]. Several lines of evidence suggest the importance of Id4 in oligodendrocyte development. Id4 is expressed in oligodendrocyte precursor cells and is thought to control the timing of oligodendrocyte differentiation \[[@B17]\]. Enforced expression of Id4 stimulates proliferation and blocks differentiation of oligodendrocyte precursor cells \[[@B17]\]. Id4 was recently found to directly interact with OLIG1 and OLIG2 in neural progenitor cells and to mediate the inhibitory effects of bone morphogenetic protein-4 (BMP-4) on oligodendroglial differentiation \[[@B18]\]. OLIG1/2 and Id4 are co-expressed in neural progenitor cells, in which OLIG1 and OLIG2 are localized in the nucleus whereas Id4 is localized in the cytoplasm; when Id4 expression is induced after BMP-4 treatment, OLIG1/2 localized predominantly to the cytoplasm and co-localizes with Id4 \[[@B18]\]. Despite this evidence in vitro, the expression of Id4 as it relates to OLIG1 and OLIG2 expression in ODG and OAC has not been reported. In this study, we analyzed in silicothe levels of expression for each survival-associated gene in the UniGene libraries derived from GBM and ODG specimens. We then examined the expression of Id4 in tissue samples from ODG and OAC using immunohistochemistry and found that, in contrast to the patterns of expression for OLIG1 and OLIG2, Id4 is preferentially expressed in cells of astrocytic lineage in ODG and OAC tumors, which is similar to the expression patterns of FABP7 (unpublished data). The differential expression of Id4 and FABP7 in cells of astrocytic phenotype in GBM and oligodendroglial tumors suggests that specific transcription factors functioning in cell-fate determination during development might also determine the development of histopathologically distinct glioma subtypes. Methods ======= In silicoanalysis ------------------- UniGene libraries for GBM and ODG were used to examine the normalized expression of the 50 survival-associated genes. Library derived from OAC was not available at the time of analysis. UniGene assigns all human Expressed Sequence Tag (EST) sequences that meet minimal standards of quality to distinct \"clusters\". Each cluster contains sequences that represent a unique human expressed gene, and related information such as the tissue types from which the libraries generating the sequences were prepared \[[@B19]\]. The normalized expression for each survival-associated genes as well as for Id4 in tissues was recorded from the SOURCE database <http://genome-www5.stanford.edu/cgi-bin/source/sourceSearch>. The normalized expression of a gene represents the relative expression level (in percentile) in different tissues and is normalized for the number of EST clones from all libraries reported by UniGene. Tissue specimens ---------------- Paraffin-embedded and frozen specimens were obtained from the Brain Tumor Research Center Tissue Bank at the University of California, San Francisco after approval from the Committee on Human Research. Microarray analysis ------------------- Sample preparation and microarray methods followed those of our previous study \[[@B7]\]. Briefly, total RNA was extracted from frozen tissue specimens using Trizol (Invitrogen; Carlsbad, CA) followed by mRNA purification using FastTrack (Invitrogen). Messenger RNA was reverse transcribed to cDNA and directly labeled with Cy dyes (Amersham Biosciences; Piscataway, NJ) before hybridization. Other detailed protocols can be found in web supplement <http://microarray-pubs.stanford.edu/gbm/>. Antibodies ---------- FABP7 antiserum was a gift from Dr. Nathaniel Heintz (Rockefeller University, New York, NY). A dilution of 1 to 400 was used for both immunostaining and migration assays, and a dilution of 1 to 500 was used for immunoblotting. Dilutions of antibodies against glial fibrillary acidic protein (GFAP) (ICN; Costa Mesa, CA) and Id4 (Santa Cruz Biotechnology, Santa Cruz, CA) for immunohistochemistry were 1:1000 and 1:100, respectively. A dilution of 1 to 100 of the Id4 antibody was used for immunoblotting. Peroxidase-conjugated and biotinylated secondary antibodies were obtained from Vector Laboratories (Burlingame, CA). Western blot analysis --------------------- The protein fraction of frozen tissue specimens was purified by isopropanol precipitation after removing total RNA and genomic DNA using Trizol, washed several times in 0.3 M guanidine hydrochloride in 95% ethanol, and resuspended in 1% SDS. The protein concentration of each sample was quantitated by using a D~c~Protein Assay Kit (Bio-Rad; Hercules, CA), and equal amounts of protein for each sample were separated by SDS-PAGE and transferred to nitrocellulose membranes (Bio-Rad), blocked with 10% skim milk, incubated with specific antibodies, and visualized using a Super Signal West Pico Chemiluminescent kit (Pierce; Rockford, IL). Immunohistochemistry -------------------- All frozen tissue sections used for immunohistochemistry were fixed in 4% formaldehyde, treated with H~2~O~2~, blocked with normal serum, incubated with primary antibodies at 4°C overnight or at room temperature (RT) for 2 hours, and incubated with biotinylated secondary antibody and peroxidase-labeled streptavidin at RT for 30 minutes. The DAB Reagent kit was used to visualize the immunoreactivity (KPL; Gaithersburg, MA). Staining of paraffin-embedded sections followed the same protocol, except for prior de-waxing and antigen retrieval using microwave heating. Statistical analyses -------------------- All statistical analyses used SPSS for Windows (Release 11.5.0). Comparison of Id4 mRNA levels in gliomas was analyzed using Student\'s t test. Correlation of Id4 and FABP7 expression with patient survival was analyzed using the Cox proportional hazards regression. A p value ≤0.05 was considered statistically significant for all tests. Results ======= Transcripts of most survival-associated genes identified in glioblastomas are also detected in oligodendrogliomas ----------------------------------------------------------------------------------------------------------------- In our previous microarray analyses, we found elevated expression of the GBM survival-associated genes in ODG and OAC as compared to normal brain specimens \[[@B7]\]. However, expression of these 50 survival-associated genes in ODG and OAC tumors has not been previously studied except Olig1and Olig2\[[@B14]\]. To corroborate our previous findings, we examined in silicothe normalized expression for each survival-associated gene as well as for Id4in UniGene libraries derived from GBM and ODG tumors. We found that, among the 50 genes, EST clones from 36 of the genes (72%) were present in GBM libraries, and clones from 30 of them (83%) were detected in libraries derived from ODG specimens (Table [1](#T1){ref-type="table"}). As expected, significant numbers of EST clones representing Olig1, Olig2, and Id4appeared in libraries from ODG. This result indicates that, at least at the mRNA level, most of the survival-associated genes identified in GBM can also be detected in ODG, which correlates with the results from our previous microarray analyses. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Normalized expression of the survival-associated genes in GBM and oligodendrogliomas ::: Clone ID Gene Name GBM (%)^1^ Oligodendroglioma (%) --------------- ----------- ------------ ----------------------- IMAGE:222457 ARC 56.16 3.53 IMAGE:383898 ARHN 5.1 28.36 IMAGE:1416420 ASCL1 5.79 7.39 IMAGE:32687 BCAN 12.5 38.22 IMAGE:290213 CACNG4 7.3 24.99 IMAGE:249688 CCND2 1.85 0.8 IMAGE:295116 CHRFAM7A ND^2^ ND IMAGE:220096 CNGB1 ND ND IMAGE:26295 CNR1 16.02 ND IMAGE:248485 CRB1 ND 9.04 IMAGE:26505 CRMP5 14.59 5.42 IMAGE:41720 CRMP5 14.59 5.42 IMAGE:43771 CROC4 29.82 14.65 IMAGE:301104 CSPG5 8.02 16.41 IMAGE:33854 DBCCR1L ND 0.01 IMAGE:951022 DCX 19.01 3.78 IMAGE:361688 DPP6 2.18 14.1 IMAGE:81320 ETV1 1.76 4.17 IMAGE:279195 FABP7 ND 1.44 IMAGE:345626 FABP7 ND 1.44 IMAGE:773976 FABP7 ND 1.44 IMAGE:75759 FLJ10748 18.43 1.17 IMAGE:669379 GLCCI1 ND 0.44 IMAGE:49987 GRIA2 1.4 3.13 IMAGE:730016 KCNJ10 23.1 2.88 IMAGE:50939 KIAA0523 2.05 4.62 IMAGE:192543 KIAA0773 2.48 15.04 IMAGE:927112 KIAA0773 2.48 15.04 IMAGE:397495 LRRTM2 ND ND IMAGE:787860 MGC45428 0.34 0.52 IMAGE:838478 NCALD 1.22 1.92 IMAGE:838668 NSE1 0.46 0.3 IMAGE:743187 NUDT10 ND ND IMAGE:41214 OLIG1 2.92 33.23 IMAGE:26884 OLIG2 13.58 34.21 IMAGE:287468 PCDH17 ND 0.95 IMAGE:166934 PCDHGC3 9.46 8.46 IMAGE:753071 PHLDA1 4.27 3.71 IMAGE:866702 PTPN13 ND ND IMAGE:30175 PTPRO ND 1.12 IMAGE:785148 PTPRZ1 5.34 8.2 IMAGE:161436 PTPRZ1 5.34 8.2 IMAGE:283976 PURG 13.16 ND INAGE:1636226 SALL1 ND 20.31 IMAGE:207087 SCN 5.43 1.96 IMAGE:179534 SCNQ2 3.55 20.51 IMAGE:361659 SGEF 2.27 ND IMAGE:357220 SHC3 ND ND IMAGE:45877 SLC29A4 ND 1.44 IMAGE:38015 SLITRK2 ND ND IMAGE:758266 THBS4 ND 1.19 IMAGE:34204 TRIM9 6.54 4.31 IMAGE:951705 ZCCHC6 0.76 ND IMAGE:359684 ZDHHC22 32.39 17.7 Id4 2.1 8.29 ^1^Normalized expression level of each gene is defined as the number of EST clones detected in tumor libraries (GBM or oligodendrogliomas) divided by the number of clones detected in all libraries reported by UniGene. ^2^ND, not detected. ::: Id4 protein is readily detected in glioblastomas and oligoastrocytomas but not in oligodendrogliomas ---------------------------------------------------------------------------------------------------- Because the in silicoanalysis of tumor specimens cannot distinguish the cellular source of expression, we used immunohistochemistry to localize the expression of Id4 in GBM and oligodendroglial tumors. The specificity of Id4 antibodies was demonstrated by detection of a 15 kD band on immunoblot using Trizol extracts of GBM and normal brain specimens (Figure [1A](#F1){ref-type="fig"}). Id4 immunoreactivity was found in 10% to 100% of tumor cells in 12 of the 13 GBM specimens we examined (92%), and Id4 in both nucleus and cytoplasm could be detected in immunoreactive cells (Figure [1B](#F1){ref-type="fig"}). Among the 21 ODG tumors examined, expression of Id4 was seen in some reactive astrocytes but was essentially negative in neoplastic oligodendrocytes (Figure [1C](#F1){ref-type="fig"}), while unequivocal Id4 staining was observed in some neurons associated with tumor cells (data not shown). We also examined 10 OAC specimens, all of which had GFAP-positive neoplastic cells and 5 of them (50%) had detectable Id4 immunoreactivity in GFAP-positive areas (Figure [1D](#F1){ref-type="fig"}). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Id4 is preferentially expressed in cells of astrocytic lineage. The specificity of anti-Id4 antibodies was confirmed with immunoblotting by using Trizol extracts of a GBM specimen and white matter (WM) from a normal subject (A). Cytoplasmic Id4 immunoreactivity was prominent in most GBM examined (B). In oligodendrogliomas, neoplastic oligodendrocytes were essentially negative for Id4, whereas some reactive astrocytes could be found Id4-positive (C). Id4 immunoreactivity could also be seen in the astrocytic component of some oligoastrocytomas (D). Bar in B, 50 μm. ::: ![](1472-6890-5-6-1) ::: Our in silicoand immunohistochemistry data showed that Id4 mRNA had comparable levels in UniGene libraries derived from GBM and ODG specimens (Table [1](#T1){ref-type="table"}), while Id4 protein was detected only in neoplastic astrocytes but not in neoplastic oligodendrocytes (Figure [1](#F1){ref-type="fig"}). To confirm the discrepancy of the Id4 expression between mRNA and protein levels in ODG tumors, we examined expression of Id4 mRNA in selected GBM (N = 7), ODG (N = 3), and OAC specimens (N = 3) from a previous microarray study \[[@B7]\]; all these tumor specimens had been examined for Id4 expression with immunohistochemistry and showed Id4 immunoreactivity in neoplastic astrocytes but not in neoplastic oligodendrocytes. Unlike protein expression, we found that Id4 mRNA had similar levels in specimens derived from all three tumor types examined in this analysis (p = 0.6, Figure [2](#F2){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Microarray analysis of Id4 mRNA in glioma tissues that have been examined with immunohistochemistry. The expression levels of Id4 were represented by the log 2-transformed ratio of the fluorescence signal in tumor to the signal in hybridization reference. Two IMAGE clones (788234 and 789369) corresponding to different regions of the Id4gene with interpretable data were selected for analysis. All three tumor types examined in this analysis demonstrated similar levels of Id4 mRNA (p values between 0.33 and 0.75). ::: ![](1472-6890-5-6-2) ::: Id4 and FABP7 are co-expressed in microgemistocytes of oligodendrogliomas and in neoplastic astrocytes of oligoastrocytomas --------------------------------------------------------------------------------------------------------------------------- Although Id4 has an expression pattern similar to OLIG1 and OLIG2 in GBM, Id4 protein was preferentially detected in reactive and neoplastic astrocytes but not in neoplastic oligodendrocytes. We then examined whether this difference occurred with other genes in the survival-associated group. We chose to study expression of FABP7 in ODG and OAC and found that, similar to Id4, FABP7 was expressed in OAC but not in the neoplastic oligodendrocytes (unpublished data). We therefore examined whether both genes are co-expressed in ODG and OAC tumors. Microgemistocytes are a small subset of neoplastic cells that may be seen in ODG tumors, but the exact cellular origin of this type of cells is still controversial. Six ODG specimens had an easily discernible group of GFAP-positive microgemistocytes (Figure [3A](#F3){ref-type="fig"} and [3B](#F3){ref-type="fig"}), and in 4 of these specimens both FABP7 and Id4 immunoreactivity was present in the cytoplasm (Figure [3C](#F3){ref-type="fig"} and [3D](#F3){ref-type="fig"}). No Id4 immunoreactivity was detected in FABP7-negative microgemistocytes (Figure [3E](#F3){ref-type="fig"} to [3H](#F3){ref-type="fig"}). Among the 10 OAC specimens examined, 8 were positive for FABP7, and 5 of these 8 specimens had detectable Id4 immunoreactivity. In this subgroup of 5 OAC specimens (Figure [4A](#F4){ref-type="fig"}), GFAP (Figure [4B](#F4){ref-type="fig"}), FABP7 (Figure [4C](#F4){ref-type="fig"}), and Id4 (Figure [4D](#F4){ref-type="fig"}) were observed in the same population of neoplastic astrocytes. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### FABP7 and Id4 are co-expressed in microgemistocytes in oligodendrogliomas. H & E staining showed typical morphology of microgemistocytes with eosinophilic cytoplasm (A) with strong GFAP immunoreactivity (B). In consecutive sections, cytoplasmic FABP7 (C) and Id4 (D) staining was detected in the same group of cells. However, microgemistocytes are negative for both FABP7 and Id4 in some oligodendrogliomas. H & E (E) and GFAP (F) staining showed the presence of microgemistocytes that were essentially negative for FABP7 (G) and Id4 (H). Some FABP7-positive cells were reactive astrocytes (G). Bar in A, 50 μm. ::: ![](1472-6890-5-6-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### FABP7 and Id4 are co-expressed in astrocytic component of oligoastrocytomas. H & E staining demonstrated the presence of neoplastic astrocytes with eosinophilic cytoplasm and elongated nuclei (A) that were also GFAP-positive (B). The same population of cells was positive for both FABP7 (C) and Id4 (D) in consecutive sections of the frozen specimen. The scale of the photomicrographs is the same as in Figure 1B. ::: ![](1472-6890-5-6-4) ::: Expression of Id4 and FABP7 in relation to the clinical outcome of patients with oligodendrogliomas and oligoastrocytomas ------------------------------------------------------------------------------------------------------------------------- Given the findings that nuclear FABP7 immunoreactivity predicts the outcome of patients with GBM, and that Id4 and FABP7 are co-expressed in OAC and in microgemistocytes of ODG, we tested whether expression of Id4 and FABP7 might have prognostic value for patients with ODG or OAC. Among the tumor specimens on which we performed immunohistochemistry, we analyzed 15 ODG (grade III) and 10 OAC (1 grade II and 9 grade III) for which we could obtain clinical information about the patients (Table [2](#T2){ref-type="table"}). The median age for this group of patients with ODG and OAC was 51 and 45 years old, respectively. The follow-up period from the time of initial diagnosis ranged from 3 to 9 years. Although they were not treated under the same protocol, all had surgical resections and received both radiotherapy and alkylating chemotherapy. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Summary of clinical data of patients with ODG and OAC, and immunohistochemistry of their tumors. ::: Tumor ID Diagnosis Alive/Dead^4^ Survival Time (year)^5^ GFAP FABP7 Id4 ---------- ------------ --------------- ------------------------- ------------- ------------------- --------------- 2650 ODG III^1^ A 9.12 \- (MG+)^6^ \- (MG+) \- (MG+) 2963 ODG III A 7.28 \- (MG+) \- (MG+) \- (MG+) 3081 ODG III A 6.97 \- (MG+) \- (MG+) \- (MG+) 3474\* ODG III A 5.67 \+ (?)^7^ \- \- 3492 ODG III D 5.29 \- \- \- 3551\* ODG III A 5.42 NI^8^ \- \- 3607 ODG III D 3.91 \- \- \- 3651\* ODG III A 5.09 \- \- \- 3767 ODG III A 4.73 \- \- \- 3768 ODG III A 4.73 \- \- \- 3815 ODG III A 4.59 \- (MG+) -/+^9^(MG-) \- (MG-) 4115 ODG III A 3.64 \- (MG+) \- (MG+) \- (MG+) 4142 ODG III A 3.55 \- (MG+) \- (MG-) \- (MG-) 4167 ODG III A 3.48 -/+ -/+ \- 4253 ODG III A 3.26 \- \- \- 2450 OAC III^2^ A 9.9 \+ \+ (cyto/nuc)^10^ \+ (cyto) 2459 OAC III A 9.9 \+ \+ (cyto/nuc) \+ (cyto) 2533\* OAC III A 6.55 \+ \+ (cyto) \- 2609\* OAC III A 9.27 \+ \+ (cyto/nuc) \- 2622\* OAC III D 5.75 \+ \- \- 2668 OAC III D 8.97 \+ \+ (cyto/nuc) \+ (cyto) 3304 OAC III A 6.27 \+ \+ (cyto) \+ (cyto) 3378 OAC III A 5.97 \+ \+ (cyto/nuc) \+ (cyto/nuc) 3729 OAC II^3^ A 4.85 \+ \+ (cyto) \- 4141 OAC III A 3.55 \+ \- \- ^1^ODG III, oligodendroglioma, grade III. ^2^OAC III, oligoastrocytoma, grade III. ^3^OAC II, oligoastrocytoma, grade II. ^4^Alive or dead of the patients was determined by the date of the most recent follow-up (9/30/04). ^5^The date of diagnosis to 9/30/04 was defined as the survival time for patients who were alive. ^6^Positive in microgemistocytes (MG). ^7^This was a frozen section with 50% of tumor cells positive for GFAP, but H & E did not show microgemistocytes; the result was therefore questionable. ^8^This was a frozen section with lots of freezing artifect; the result was not interpretable. ^9^A small group (\<5%) of tumor nuclei were positive. ^10^Both cytoplasm and nuclei were positive. \* Specimens that have been analyzed by cDNA microarrays. ::: Two of the 15 patients with ODG have died, and their tumors were negative for Id4/FABP7 immunoreactivity (Table [2](#T2){ref-type="table"}). Two of the 10 patients with OAC have died; one tumor had no detectable Id4 and FABP7 immunoreactivity, while the other one showed a similar degree of cytoplasmic/nuclear FABP7 and cytoplasmic Id4 staining to the other 4 specimens derived from patients who were still alive (Table [2](#T2){ref-type="table"}). There were no characteristics that distinguished the living subjects from those diseased. We used Cox regression to analyze the correlation of Id4 and FABP7 immunoreactivity with patient survival by the presence of immunoreactive microgemistocytes (for stratifying ODG patients), or by the immunoreactivity in nucleus alone or in both nucleus and cytoplasm (for stratifying OAC patients). Our limited data showed that expression of Id4 and FABP7 did not correlate with the survival of patients with ODG or OAC tumors (Table [3](#T3){ref-type="table"}). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Cox regression analyses of FABP7 and Id4 immunoreactivity in relation to the survival of ODG and OAC patients ::: Tumor Markers Immunoreactivity Status no. of cases p ------- ------------- ------------------------- -------------- ------- ODG FABP7 & Id4 MG+ vs. MG- 4 vs. 11 0.524 FABP7 cyto/nuc+ or cyt+ vs. - 8 vs. 2 0.859 OAC nuc+ vs. nuc- 5 vs. 5 0.581 Id4 cyto/nuc+ or cyt+ vs. - 5 vs. 5 0.573 ::: Discussion ========== Oligodendroglial tumors are the third most common type of primary brain tumor. They have a diffusely infiltrative growth pattern similar to astrocytic tumors but tend to have a better prognosis. The exact genetic basis for the development of gliomas with astrocytic, oligodendroglial, or mixed (oligoastrocytic) phenotypes is unknown. However, it has been recognized that specific genetic backgrounds do affect the clinical outcome of patients with oligodendroglial tumors. A subgroup of oligodendroglial tumors with allelic loss of chromosome arms 1p and 19q respond better to treatment \[[@B20]\]. Several genetic aberrations frequently detected in GBM, such as mutations in the p53gene, Egfroverexpression or amplification, and deletion of the p16/Cdkn2Agene, are also associated with poor prognosis in both ODG and OAC. Defining the genetic features of these glial neoplasms has both diagnostic and potentially therapeutic benefits. Previously, we identified a group of genes that stratifies patients with histologically identical GBM into two groups with markedly different survival \[[@B7]\]. The expression of FABP7, in particular, demonstrated a strong inverse correlation with patient survival, and may be involved in the pathogenesis of GBM (unpublished data). Our data suggested that this group of genes that includes Olig1and Olig2might be also expressed in ODG and OAC \[[@B7]\]. Therefore, we hypothesized that our group of survival-associated genes originally identified in GBM might also be associated with the pathobiology of oligodendroglial neoplasms. We examined in silicothe expression of the transcripts for each survival-associated gene in UniGene libraries derived from GBM and ODG tumors, and most could be identified in both tumor types. The transcripts of the Fabp7gene were not detected in the GBM libraries from UniGene (Table [1](#T1){ref-type="table"}). This might be because FABP7 immunoreactivity was observed in only half to two-thirds of the GBM specimens as shown in our previous studies \[[@B7]\], and the number of libraries reported by UniGene may not be sufficient to \"capture\" all potentially positive tumors. Another limitation of this type of analysis is that the cellular source for positive gene expression cannot be determined. For this reason, we validated our analysis by examining specific genes at the tissue level using immunohistochemistry. Our data demonstrate that Id4 and FABP7 proteins are primarily localized in cells with astrocytic features but not in neoplastic oligodendrocytes. Therefore, the expression of survival-associated genes in ODG and OAC can be divided into at least two subgroups; genes such as Olig1and Olig2that are expressed in all glial tumors, and genes such as Fabp7that are restricted to astrocytes. This division is supported by our immunohistochemical analysis of another survival-associated gene, Ptprz1, which has an expression pattern in ODG and OAC similar to Id4 and FABP7 (unpublished data). Examining the biological roles of OLIG1, OLIG2, and Id4 in developing central nervous system provides a plausible explanation for these varying patterns of gene expression. OLIG1 and OLIG2 play a central role in oligodendrocyte development, as illustrated by their expression in fully differentiated oligodendrocytes \[[@B21]\], the failure of oligodendrocyte differentiation in Olig1/Olig2-knockout mice \[[@B12]\], and the expansion of oligodendroglial cells in vivofollowing overexpression of OLIG1 and OLIG2 \[[@B22],[@B23]\]. Oligodendrocyte development exemplifies the principle of \'switching\' of cell fate by alteration of transcription-factor expression levels. For example, overexpression of BMP-4 in mice results in a significant decrease in the number of oligodendrocytes, accompanied by an increase in astrocytes \[[@B24]\], coincident with the gradual decline of OLIG1 and OLIG2 \[[@B18]\]. Ectopic expression of Id4 in neural progenitor cells recapitulates the effect of BMP-4 by blocking oligodendrocyte maturation, stimulating cell-cycle progression of oligodendrocyte precursor cells, and promoting astrocyte formation \[[@B17],[@B18]\]. Id4 expression decreases when the precursor cells are induced to differentiate into oligodendrocytes \[[@B17]\]. OLIG1, OLIG2, and Id4 are co-expressed in neural progenitor cells but in different subcellular compartments (OLIG1 and OLIG2 are localized in the nucleus and Id4 is localized in the cytoplasm). BMP-4 treatment induces Id4 expression along with translocation of OLIG1 and OLIG2 to the cytoplasm; therefore, it was proposed that cytoplasmic Id4 must reach a critical level to effectively sequester OLIG1 and OLIG2 to prevent their nuclear translocation, and subsequent specification of the cell fate \[[@B18]\]. Both astrocytomas and oligodendroglial tumors in humans demonstrate diffuse expression of OLIG1 and OLIG2, although expression of OLIG2 is lower and increasingly variable and heterogeneous in GBM as compared to ODG tumors \[[@B14]\]. Conversely, Id4 is negative in neoplastic oligodendrocytes, positive in neoplastic astrocytes in half of the OAC specimens examined, and positive in almost all GBM specimens. This pattern of Id4 expression parallels its expression in neural progenitor cells during their commitment to either an oligodendrocytic or astrocytic lineage. Based on these observations, one possibility is that astrocytomas and oligodendroglial tumors may arise from a common cell type with features similar to the OLIG1/2-positive neural progenitor cells. Increased expression of Id4 by growth-factor stimulation or genetic mutation may switch transformed cells towards an astrocytic phenotype, whereas failure of Id4 expression would give rise to pure ODG. The astrocytic component of OAC may represent a focal area of Id4 expression in tumor subpopulations occurring at later stages after tumor initiation, leading to the co-expression of Id4 with OLIG1 and OLIG2 in neoplastic astrocytes and lack of Id4 expression in neoplastic oligodendrocytes. This potential mechanism for the presence of mixed cell populations in OAC is consistent with genetic analyses demonstrating that astrocytic and oligodendroglial components within an oligoastrocytoma might be of monoclonal origin \[[@B25]\]. Furthermore, the magnitude and the timing of Id4 induction could explain the heterogeneity of OAC, in that one group is genetically related to ODG, and the other one related to astrocytomas \[[@B26],[@B27]\]. However, the astrocytic component in OAC may have a different transitional nature from the GFAP/Id4/FABP7-positive cell types in ODG, since a recent study found that no OLIG1 expression is seen in GFAP-positive microgemistocytes in ODG tumors \[[@B28]\]. In this study, we did not find a correlation between Id4 and FABP7 expression and the clinical outcome of patients with OAC, but it may be of interest to repeat the analyses in a larger cohort of astrocytoma-like OAC as defined in the study by Maintz and others \[[@B26],[@B27]\]. In conclusion, we show that Id4 expression is present primarily in neoplastic astrocytes in both oligodendroglial tumors and GBM, which is in contrast to the expression patterns of Olig1and Olig2, two survival-associated genes originally identified in GBM tumors. Two other survival-associated genes, Fabp7and Ptprz1, show expression patterns similar to Id4. Based on the function of Id4, OLIG1, and OLIG2 during normal development, our data suggest that Id4 and potentially other genes such as Fabp7and Ptprz1might account for the formation of distinct glioma subtypes during oncogenic progression. Conclusion ========== Id4 expression is present primarily in neoplastic astrocytes in both oligodendroglial tumors and GBM, which is in contrast to the expression patterns of OLIG1 and OLIG2, two survival-associated genes originally identified in GBM tumors. FABP7 show expression patterns similar to Id4, indicating that the survival-associated genes can be divided into at least two groups in ODG and OAC. The functions of Id4, OLIG1, and OLIG2 during normal development, and our data suggest that Id4 and potentially other proteins such as FABP7 and PTPRZ1 might account for the formation of distinct glioma subtypes during oncogenic progression. List of abbreviations ===================== FABP7, brain-type fatty acid-binding protein; HLH, helix-loop-helix; GBM, glioblastoma; ODG, oligodendroglioma; OAC, oligoastrocytoma; WHO, World Health Organization. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= YL designed the study, performed the experiments, and wrote the manuscript. AWB participated in evaluating the immunohistochemistry. MKN participated in collecting clinical data of patients. NG participated in editing the manuscript and provided laboratory resources for experiments. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1472-6890/5/6/prepub> Acknowledgements ================ We thank the Brain Tumor Research Center Tissue Bank at UCSF for contributing tissue specimens for this study. We would also like to thank Dr. Nathaniel Heintz for providing FABP7 antiserum. This work was supported by funding from the Department of Neurological Surgery at UCSF. UCSF is an NCI-designated Specialized Program of Research Excellence for Brain Tumors.	PubMed Central	2024-06-05T03:55:59.942841	2005-7-15	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182359/", "journal": "BMC Clin Pathol. 2005 Jul 15; 5:6", "authors": [ { "first": "Yu", "last": "Liang" }, { "first": "Andrew W", "last": "Bollen" }, { "first": "M Kelly", "last": "Nicholas" }, { "first": "Nalin", "last": "Gupta" } ] }
PMC1182360	Background ========== Pax6is expressed in the developing eye and brain, where it affects both progenitor cell production and neuronal differentiation \[[@B1]-[@B5]\]. The Pax6 protein contains two DNA binding domains, a paired domain (PD) and a paired-type homeodomain (HD) (Fig. [1A](#F1){ref-type="fig"}). The PD consists of two separate helix-turn-helix motifs, termed PAI and RED, which act on different target sequences \[[@B6]\]. The best characterised Pax6alternative splicing event, involving the insertion of a 42 bp exon (exon 5a) into the PAI subdomain of the PD, results in two major Pax6 isoforms \[Pax6 and Pax6(5a)\] with different DNA-binding properties. In vitrostudies have shown that the PAI subdomain of Pax6 binds preferentially to a consensus sequence (P6CON) \[[@B7]\] but that PAI subdomain disruption in Pax6(5a) allows the RED subdomain to bind an alternative sequence (5aCON) \[[@B6]\]. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Detection of Pax6and Pax6(5a). (A) Riboprobe design. The ssRNA antisense riboprobe used to detect both Pax6and Pax6(5a)spans the Pax6mRNA from exon 3 to exon 5a. Two protected fragments are produced, Pax6(218nt) and Pax6(5a)(247nt). The PD- and HD-coding regions are represented above the Pax6mRNA sequence, which is shown to scale. Grey and black bars are exons. Broken line shows part of the riboprobe derived from the vector. (B) Example of an RNase protection assay showing GAPDHand Pax6signal detected in total RNA from E12.5 eye (E), diencephalon (D) and telencephalon (T). 1 μg total RNA was used in each sample. Central lane is a DNA ladder, which allows approximate sizing of bands (fragment sizes are indicated above bands). (C) Quantification of RNase protection assays. Pax6(lower band) and Pax6(5a)(upper band) in the telencephalon, diencephalon and eye at E12.5 and E18.5. Gel images and densitometric traces of bands are shown. Level of background estimated using the rolling disk method (Quantity One software, Biorad) is indicated on each trace. ::: ![](1471-213X-5-13-1) ::: Pax6 affects early progenitor cell proliferation and later neuronal differentiation in the developing brain \[[@B4],[@B5]\]. This changing role might be due at least in part to a shift in the relative concentrations of Pax6 and Pax6(5a) during development but there has been no previous report of such alterations during neurogenesis in the brain in vivo. We anticipated that even relatively small changes in the Pax6 : Pax6(5a) ratio might be important since stronger effects on gene activity viaP6CON and 5aCON are observed if Pax6and Pax6(5a)are introduced into cultured cell lines at ratios of 1:1 or 8:1 than at ratios of 2:1, 4:1 or 16:1 \[[@B8]\]. We selected a direct method for quantification of the ratio (RNase protection assays) and carried out multiple assays so as to obtain statistically analysable data on key Pax6-expressing brain tissues at a range of ages throughout neurogenesis. We found changes in the ratio similar in magnitude to those shown previously to alter target gene activity in vitro. Results and Discussion ====================== RNase protection assays were carried out on tissues dissected from wild type mouse embryos aged E12.5-E18.5 and quantified by densitometry (Fig. [1](#F1){ref-type="fig"}). The Pax6: Pax6(5a)ratio varied throughout neurogenesis in the telencephalon, diencephalon and hindbrain, following a similar pattern in each (Table [1](#T1){ref-type="table"}). In the telencephalon, the ratio was about 6:1 early in neurogenesis, at E12.5, but was significantly lower (about 2:1) at each subsequent age (p \< 0.05--p \< 0.01; Table [1](#T1){ref-type="table"}). In the diencephalon and hindbrain, ratios were about 8:1 and 10:1 at E12.5 but were lower (about 2:1 to 4:1) from E14.5-E18.5 (Table [1](#T1){ref-type="table"}); these decreases were statistically significant when data from all three age groups were combined (p \< 0.05 for both tissues). Combining results from the telencephalon, diencephalon and hindbrain showed a significant fall in the Pax6: Pax6(5a)ratio between E12.5 and E14.5, from approximately 8:1 to 3:1 (p \< 0.05), with a further slight decrease at E16.5 and E18.5 (p \< 0.01; Table [1](#T1){ref-type="table"}). Data on the Pax6: Pax6(5a)ratio in the eye at E12.5-E18.5 did not show the same trend (Table [1](#T1){ref-type="table"}). Ratios varied from about 10:1 at E14.5 to about 4:1 at E18.5 but none of the differences were statistically significant. Although we did not detect significant changes in the eye, it is possible that non-synchronised changes in the ratio do occur within its component Pax6-expressing neural and non-neural tissues (cornea, lens and retina). Neither Pax6nor Pax6(5a)was detected in samples from the feet. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Mean Pax6: Pax6(5a)ratios ± SEMs. \"Combined\" data are from telencephalon, diencephalon and hindbrain considered together at each age. For each tissue and for combined brain tissues, significant differences between values at E12.5 and values at subsequent ages are shown (unpaired Student\'s t-tests; \* = p \< 0.05, \\ = p \< 0.01). Numbers of individual assays are indicated in brackets. ::: E12.5 E14.5 E16.5 E18.5 ------------------- -------------------- ----------------------- ----------------------- ----------------------- Telencephalon 6.25 +/- 0.68 (5) 2.12 +/- 0.77\\ (2) 2.63 +/- 1.52\* (3) 2.53 +/- 0.37\\ (4) Diencephalon 9.72 +/- 3.37 (3) 3.63 +/- 0.63 (3) 1.41 (1) 3.00 +/- 0.63 (2) Hindbrain 7.66 +/- 3.68 (4) 3.87 +/- 1.62 (2) 1.66 +/- 0.35 (2) 2.23 +/- 0.12 (3) Combined 7.59 +/- 1.41 (12) 3.27 +/- 0.55\* (7) 2.10 +/- 0.72\\ (6) 2.53 +/- 0.21\\ (9) Eye 6.81 +/- 2.08 (4) 9.52 +/- 3.64 (3) 5.95 +/- 1.90 (3) 3.55 +/- 1.30 (3) ::: There is evidence that Pax6 and Pax6(5a) have different functions in vivoin both vertebrates \[[@B9]-[@B11]\] and invertebrates \[[@B12],[@B13]\]. Studies of Pax6- and Pax6(5a)-related genes in Drosophila melanogaster, ey/ toyand eyg/ toe, have shown that they promote, respectively, differentiation and proliferation of eye precursor cells \[[@B12],[@B13]\]. Overexpression of Pax6 and Pax6(5a) can alter the expression of different sets of genes in mammals \[[@B10],[@B11]\]. Mammalian brain cells reduce their proliferation in response to overexpression of Pax6 or Pax6(5a) and increase their neurogenesis in response to overexpression of Pax6 \[[@B9]\]. It is possible that a reduction in the Pax6: Pax6(5a)mRNA ratio from E14.5 is involved in programming progenitor cells to initiate processes that occur later in embryogenesis. Such processes are potentially numerous; they might, for example, include the transition from predominantly neurogenesis to the major phase of gliogenesis or the development of specific sets of later-generated neurones such as the superficial layers of the cerebral cortex. Conclusion ========== We conclude that the Pax6: Pax6(5a)ratio falls in the telencephalon, diencephalon and hindbrain during neurogenesis and, moreover, the magnitude of the change is in the range that alters target gene expression in vitro\[[@B8]\]. This finding allows the possibility that changes in the relative expression levels of isoforms of a single Pax6gene might result in changes in the functions of this gene in mammalian brain development. Methods ======= For each RNase protection assay, wild-type mice (CD-1) were time-mated \[the day of conception was designated embryonic day 0.5 (E0.5)\] and killed on E12.5, E14.5, E16.5 or E18.5. The telencephalon, diencephalon, hindbrain, eyes and feet (used as a Pax6 non-expressing control tissue) were dissected from each embryo. Tissues from multiple embryos were combined so as to obtain sufficient material for each assay. Total RNA was isolated from tissue snap-frozen in liquid nitrogen, its integrity was checked by agarose gel electrophoresis and its concentration was measured using a fluorimeter. The riboprobe was designed to span the Pax6mRNA from exon 3 to exon 5a so as to protect Pax6and Pax6(5a)mRNAs (Fig. [1A](#F1){ref-type="fig"}). Oligonucleotides 5\'-AAG TGG ACG TAT ATC CCA GTT CTC-3\' and 5\'-AGC ACC TGG ACT TTT GCA TC-3\' were used to amplify sequence from mouse Pax6cDNA. Pax6cDNA was cloned into pCR-BluntII-TOPO and sequenced across the Pax6insert in both directions to identify clones with Pax6sequence in the appropriate orientation to generate antisense probe. The plasmid pTRI-GAPDH (Ambion) was used for synthesis of a murine glyceraldehyde-3-phosphate dehydrogenase (GAPDH) riboprobe. Riboprobe synthesis was carried out using the Maxiscript SP6 in vitrotranscription kit (Ambion) in the presence of 40 μCi α-^32^P-UTP 800 Ci/mmol (Amersham). Riboprobes were purified using a MicroSpin G25 column (Amersham Biosciences). RNase protection assays were performed on 1 μg total RNA from E12.5 and E14.5 embryonic tissues, or 4 μg total RNA from E16.5 and E18.5 embryonic tissues, using the Hyb Speed kit (Ambion). The quality the undigested full-length riboprobe was examined and that of all total RNA samples was assessed using a GAPDHriboprobe (e.g. Fig. [1B](#F1){ref-type="fig"}). Products were resolved on 6% polyacrylamide, 8 M urea gels that were then fixed (15% methanol, 5% acetic acid for 30 minutes), dried and exposed to film with an intensifying screen. Where GAPDHsignal was weak, this was a sign that protein or DNA contamination had led to inaccurate RNA quantification, and these samples were excluded from subsequent analyses. Gel bands were quantitated using a GS-710 densitometer and the Quantity One software package (BioRad: background subtraction was performed using the rolling disk method \[[@B14]\]; Fig. [1C](#F1){ref-type="fig"}). Each assay was repeated several times (n values for the number of assays for each tissue at each age are in Table [1](#T1){ref-type="table"}). Authors\' contributions ======================= JP carried out all the RNase protections assays, analysed the results and wrote a first draft of the manuscript. JP, JM, IS and DP all helped conceive and design the study and all worked on and approved the final manuscript. Acknowledgements ================ Many thanks to Karen Chapman for advice on RNase protection, and to Tom Pratt and Jane Quinn for help with dissections. This work was funded by the Wellcome Trust.	PubMed Central	2024-06-05T03:55:59.946528	2005-7-19	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182360/", "journal": "BMC Dev Biol. 2005 Jul 19; 5:13", "authors": [ { "first": "Jeni", "last": "Pinson" }, { "first": "John O", "last": "Mason" }, { "first": "T Ian", "last": "Simpson" }, { "first": "David J", "last": "Price" } ] }
PMC1182361	Background ========== Ever since the isolation of human embryonic stem cells (hESCs) in 1998 \[[@B1]\], the implications for their use in a number of disease therapies have been highly regarded. Additionally, these cells also find value as a model to study basic human development. However, in all aspects of ESC research, hESCs must first be appropriately defined or characterized. One way to characterize hESCs is to utilize the large number of glycoproteins and carbohydrates existing on the cell surface as a way to delineate pluripotent or differentiated cell types. The most common hESC surface pluripotency markers are the stage specific embryonic antigens -3 and -4 (SSEA-3, -4) and tumor rejection antigens-1-60 and -1-81 (TRA-1-60, -1-81). SSEA-3 and -4 are globoseries cell surface glycoproteins that were first used to delineate embryological changes in the developing mouse embryo \[[@B2],[@B3]\]. Both of these antigens were found to recognize sequential regions of a mouse ganglioside epitope, with SSEA-4 (MC813-70 antibody) recognizing the terminal portion of the sequence and SSEA-3 (MC613) recognizing the internal region of he sequence. Thus, two antibodies were used to define this unique embryonic antigen. In mouse embryonic stem cells (mESCs), SSEA-3 and -4 are expressed on the 2--8 cell and morula stages of preimplantation embryos and are also found on unfertilized oocytes; however, there is a loss of expression in the inner cell mass (ICM) of mESCs \[[@B2],[@B3]\]. Yet in hESCs, there is no expression of SSEA-3 or -4 at the 2--8 cell or morula stage; however, these are expressed on the ICM of human blastocysts and on isolated hESCs \[[@B4]\]. It has been well documented that these cell surface carbohydrates change both with development and differentiation in vitro \[[@B5],[@B6]\], but there may be other undiscovered developmentally regulated cell surface carbohydrates. The hematopoetic field has numerous cell surface antigens which have been identified that define various bone marrow and blood stem cells, such as CD4, CD8, CD34, CD38, CD44, CD45, c-kit, Mac-1, Muc -18, Lin, Sca-1, and Thy-1 \[[@B7]-[@B11]\] and there has been some progress in identifying other hESC surface markers that can be used both for defining the pluripotent state and for cell sorting and cell identification. Cell surface markers such as CD9 and CD24 are presented on pluripotent hESC surfaces, while gene expression analysis indicates such genetic markers as REX1 \[[@B12]\], Cripto/TDGF1 \[[@B12]\], OCT-4 \[[@B13]\], DNMT3B \[[@B13]\], LIN28 \[[@B13]\], Nanog \[[@B14]\], and others that can denote pluripotency in hESCs \[[@B13]\]. Since carbohydrate antigens on hESC surfaces could provide further potential markers defining the pluripotent state, we targeted glycosylation patterns on the cell surface of SSEA-4 enriched hESCs using lectins to bind to these glycans. Lectins, carbohydrate binding proteins that recognize diverse sugar structures, have been extensively used to identify and characterize cell surface glycosylation patterns. For example, lectins have been used to investigate metastatic processes in many types of cancer \[[@B15]-[@B21]\] and to identify cell types based on presentation of specific cell surface carbohydrates \[[@B22]-[@B27]\]. Carbohydrate analysis using lectins has also led to the delineation of embryologic developmental stages in some species. For example, many developmentally regulated glycans identified as lectin receptors on mESCs are displayed on cell surfaces at the preimplantation and implantation stages of development. These include ConcanavalinA (Con A) defined by the hapten methylmannoside \[[@B28]\], Peanut agglutinin (PNA) from Arachis hypogaeadefined by the hapten galactose \[[@B29]\], Wheat Germ Agglutinin (WGA) from Triticum vulgarisdefined by hapten N-acetylglucosamine \[[@B30]\], Dolichos biflorusagglutinin (DBA) defined by the hapten N-acetylgalactosamine \[[@B31]\], and Ricinus communisagglutinin (RCA) defined by the hapten galactose or lactose \[[@B32]\]. These results suggest that glycans may be involved in cell-cell interactions serving a specific developmental function and also indicate that they can be used as markers to define these stages of mouse embryogenesis. In this study, we analyzed the pluripotent state of two NIH approved hESC lines, BG01 and BG02. These cell lines have previously been characterized for positive staining of the pluripotency markers TRA-1-60 and -1-81, OCT-4, alkaline phosphatase, and SSEA-3 and -4 \[[@B33]\]. Lectin binding percentages on SSEA-4 enriched hESC surfaces were determined by flow cytometry using a panel of 14 lectins and SSEA-4, which enabled us to probe the surface of pluripotent hESCs for a variety of carbohydrates and carbohydrate linkages. Rosler et al showed that in long-term culture, hESCs maintained pluripotency as determined by measurement of a variety of characteristic markers, including SSEA-4, TRA-1-60 and TRA-1-81 \[[@B34]\]. Moreover, there was a high correlation between expression of SSEA-4 and TRA-1-60 and -1-81, and the presence of these markers correlated with undifferentiated morphology. Thus, we defined pluripotent hESCs as those populations with 98--99% expression of SSEA-4, a developmentally regulated cell surface pluripotency marker that is frequently used to delineate pluripotent hESCs \[[@B1],[@B4],[@B5],[@B35]\]. We were able to obtain this high level of SSEA-4 expression by using magnetic bead sorting to select SSEA-4 expressing cells. The SSEA-4 antibody (MC 813-70) specifically recognizes a carbohydrate chain (Figure [1](#F1){ref-type="fig"}); thus, we analyzed a panel of lectins that are likely to be presented on hESC surfaces in order to identify other potential markers for pluripotent hESCs. Immunocytochemistry using SSEA-4 and each of the 14 lectins was analyzed to determine localization of lectin binding within adherent hESC colonies and to validate binding specificity of each lectin using appropriate competitive sugars. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Structure of Stage Specific Embryonic Antigen 4 (SSEA4).SSEA-4, one of the most commonly used hESC surface pluripotency markers, is a globoseries glycolipid that is characteristically downregulated upon hESC differentiation. ::: ![](1471-213X-5-15-1) ::: We found that a variety of lectins binding unique carbohydrate moieties had distinct binding patterns. Our findings demonstrate that there are many surface carbohydrate antigens that could be exploited to further characterize the pluripotent state of hESCs, and these lectins may provide a source of unique markers with which to characterize subpopulations that exist in colonies of adherent hESCs. Results ======= Carbohydrate analysis using flow cytometry ------------------------------------------ HESCs were trypsin passaged and selected for SSEA-4. It is important to note that manual passaging would not be feasible for flow cytometry studies, since atleast 250,000 single cells are needed per treatment for flow cytometry analysis. After SSEA-4 enrichment using magnetic bead sorting, 99.3 ± 0.31% of hESCs expressed SSEA-4 (Figure [2](#F2){ref-type="fig"}). To determine the presence and the percent binding of the 14 chosen lectins on pluripotent hESC surfaces, both karyotypically normal hESC lines (BG01 and BG02) were analyzed using flow cytometry with the panel of lectins and the pluripotency marker, SSEA-4. A broad range of binding percentages was observed (Figure [3](#F3){ref-type="fig"}). The highest binding percentages were detected using TL, RCA, ConA, WFA, and SNA. Binding percentages for TL, RCA, and ConA were similar to that of SSEA-4, ranging from 98--99% of enriched hESCs. Figure [4B](#F4){ref-type="fig"} shows a shifted histogram plot of TL as a representative of these high percentage binding lectins. WFA, PNA and SNA were also found to bind to over 60% of hESCs. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### SSEA-4 expression in enriched hESCs determined by flow cytometry.The histogram plots of unstained, control hESCs (2A) and cells stained with SSEA-4 antibody after enrichment using magnetic bead sorting at day 0 (2B). ::: ![](1471-213X-5-15-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Quantitation of lectin binding on pluripotent hESC surfaces with 14 different lectins.The percent of cells with specific carbohydrate expression as determined by flow cytometry using 14 different lectins. The data are means +/- SD of 3 independent assays of BG01 and BG02 hESC lines. hESCs from each line were stained with one of 14 lectins and SSEA-4 immediately following enrichment for SSEA-4 expression. abcd: Means with different letters are significantly different, p \< 0.05. ::: ![](1471-213X-5-15-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Flow cytometry histograms of lectin binding in unstained and stained HESCs.A-D shows histograms of SSEA-4 binding and representative lectins that were used in this study. To validate that double staining can be performed without signal interference, we determined that SSEA-4 expression was not found in the FL1 channel. Figure 4A shows a histogram plot with the overlay image of SSEA4 (black tracing) matching the histogram plot of unstained cells (grey fill). Figure 4B shows a positive peak shift in the histogram overlay with Tomato lectin (TL) in black tracing and unstained cells (grey fill). Figure 4C shows lack of Lotus tetragonolobuslectin (LTL) (black tracing) binding in the histogram overlay with unstained cells (grey fill). Figure 4D shows a histogram overlay with MAA (black tracing) binding and shows two peaks -one representing a large population that overlays unstained cells and a smaller population denoted by arrow that shows a smaller population of MAA+/SSEA4+ cells. Figure 4E shows plots of unstained hESCs, and Figure 4F shows subpopulations of cells characterized as SSEA-4+/MAA- or as SSEA-4+/MAA+ cells (arrow). ::: ![](1471-213X-5-15-4) ::: Two lectins, DBA and LTL, did not bind to hESCs. Analysis of DBA and LTL histograms showed no peak shifts using these lectins (Figure [4C](#F4){ref-type="fig"}), and analysis of flow cytometry plots also did not reveal any peak shifts (supplementary data). Some lectins were found to partially bind to hESC colonies. These lectins revealed two distinct subpopulations of cells that included a SSEA4+/lectin- population and a SSEA4+/lectin+ population. Figure [4D](#F4){ref-type="fig"} shows a representative histogram plot of MAA which shows a large population of cells that overlap the unstained control cells; however, the arrow indicates another smaller population of cells that show a peak shift representing positive MAA binding. Figure [4E](#F4){ref-type="fig"} shows the comparison of unstained control cells to the shifted plot of MAA (Figure [4F](#F4){ref-type="fig"}) in which there are two distinct populations of cells present: a SSEA-4+/lectin- population containing the majority of cells, and a smaller but still distinct SSEA-4+/lectin+ population indicated by the arrow. Lectins that bound in this way included PHA-E, VVA, UEA, PHA-L, and MAA, and these exhibited a large binding range varying from 58% to 13% binding of enriched SSEA-4 positive cells (Figure [3](#F3){ref-type="fig"}). Carbohydrate analysis using immunocytochemistry ----------------------------------------------- Both hESC lines, BG01 and BG02, were analyzed by immunocytochemistry to determine localization of carbohydrates and to determine whether particular staining patterns were present in adherent colonies maintained in culture. Cells were either passaged manually or by trypsinization and we did not observe obvious differences in hESC colony staining patterns between these passage methods. Immunocytochemistry results of all 14 lectins supported our flow cytometry analysis. Five lectins (TL, RCA, ConA, WFA, SNA, and HHL) bound throughout the colonies without any localized patterns of binding. TL, RCA, ConA, and WFA appeared to bind to cells that also expressed SSEA-4. Figure [5A--C](#F5){ref-type="fig"} shows a hESC colony that represents uniform lectin binding. RCA binding (Figure [5A](#F5){ref-type="fig"}) is shown throughout a colony that is SSEA-4 positive (Figure [5B](#F5){ref-type="fig"}) and is also stained with DAPI nuclear stain (Figure [5C](#F5){ref-type="fig"}). ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Carbohydrate expression as determined by lectin binding using immunocytochemistry.Figure 5A--C shows a hESC colony that represents uniform lectin binding. Ricinus Communisagglutinin (RCA) binding in red (5A) is shown throughout this SSEA-4 positive colony in green (5B). The DAPI nuclear stain image (blue) is also shown (5C). Other lectins showed partial binding patterns, such as Vicia Villosaagglutinin (VVA) binding (red), which is shown in a hESC colony (5D) that has uniform SSEA-4 antibody binding (green) (5E). Arrows denote distinct SSEA-4 positive regions lacking VVA binding. DAPI nuclear staining (blue) is also shown (5F). PHA-E binding is shown in two separate images in Figure 5G--H. In 5G there are two adjacent colonies, one that expresses strong binding of SSEA-4 antibody (green) and weak to no binding of PHA-E (red), and an adjacent colony showing binding of PHA-E without SSEA-4 antibody binding. (DAPI nuclear staining in blue). 5H, shows another colony with a streak of stacked cells (as determined by high DAPI expression, see arrow) in the middle of the colony that are beginning to lose SSEA-4 expression (green), but have strong PHA-E binding (red). However, the rest of the colony adjacent to this streak of cells is uniformly positive for SSEA-4 but is lacking PHA-E binding. 5I shows lack of DBA binding and presence of SSEA-4 and DAPI staining. Images and scale bars: 5A-G) 20× magnification, 100 μm. H-I) 10× magnification, 100 μm. ::: ![](1471-213X-5-15-5) ::: We also observed interesting patterns of carbohydrate expression for the lectins that showed two subpopulations of cells when analyzed using flow cytomtery. VVA bound to distinct regions of a SSEA-4 positive colony (Figure [5D--F](#F5){ref-type="fig"}), in contrast to the more uniform binding described above. In Figure [5D](#F5){ref-type="fig"} the arrow shows a region where there is no VVA binding, but this region is SSEA-4 positive as shown in Figure [5E](#F5){ref-type="fig"}. The DAPI nuclear stain of this same colony is shown in Figure [5F](#F5){ref-type="fig"}. Figures [5G](#F5){ref-type="fig"} & H show PHA-E binding in different colonies. In Figure [5G](#F5){ref-type="fig"} there are two adjacent colonies with one colony showing strong binding of SSEA-4 antibody (green) and little PHA-E binding (red). The adjacent colony shows binding of PHA-E but a lack of SSEA-4 antibody binding. In Figure [5H](#F5){ref-type="fig"} there is one colony that shows a streak of stacked cells as determined by the high expression of DAPI staining in the middle of the colony (depicted by arrow). These cells are beginning to lose SSEA-4 antibody binding (green) but have strong PHA-E binding (red). The rest of the colony adjacent to this streak of cells is uniformly positive for SSEA-4 but lacks PHA-E binding. Using immunocytochemistry we also demonstrated that DBA and LTL did not bind to hESC colonies. Figure [5I](#F5){ref-type="fig"} shows a hESC colony which is SSEA-4 positive (green) and is stained with DAPI nuclear stain (blue) but shows only debris of the DBA lectin (red) with no regions of DBA binding. Validation of lectin binding ---------------------------- We used complementary competitive sugars that are known to block lectin binding (listed in Table [1](#T1){ref-type="table"}) to validate lectin binding specificity. The concentrations listed in Table [1](#T1){ref-type="table"} were all found to inhibit each lectin\'s binding as detected by immunocytochemistry. As a representative of competitive sugar inhibition, Figure [6](#F6){ref-type="fig"} shows blocking of RCA binding by addition of 200 mM galactose (Figure [6A](#F6){ref-type="fig"}). In the absence of galactose, RCA binding is evident (red, Figure [6B](#F6){ref-type="fig"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Comparison of the specificity for monosaccharides and oligosaccharides of a panel of 14 biotinylated lectins used in immunocytochemistry and flow cytometry. ::: Lectin Origin Monosaccharide Specificity Inhibitor ------------------------------------------------ -------------------------------- ------------------------------------------------ Concanavalin A(Con A) Man or Glc 200 mM α-methylmannoside or α-methyl glucoside Phaseolus vulgariserythro-agglutinin (PHA-E) Gal galactose Phaseolus vulgarisleuco-agglutinin (PHA-L) Gal galactose Sambucus nigraagglutinin (SNA) Sialic Acid 500 mM lactose in acetic acid Arachis hypogeapeanut (PNA) Gal 200 mM galactose Vicia villosaagglutinin (VVA) GalNAc 200 mM N-acetylgalactosamine Maackia amurensis(MAA) Gal 200 mM lactose Ricinus communisagglutinin (RCA) Gal 200 mM galactose or lactose Wisteria floribundaagglutinin (WFA) GalNAc 200 mM N-acetylgalactosamine Ulex europaeusagglutinin (UEA) Fuc 50--100 mM L-fucose Lotus tetragonolobuslectin (LTL) Fuc 50--100 mM L-fucose Dolichos biflorusagglutinin (DBA) GalNAc 200 mM N-acetylgalactosamine Hippeastrum hybridlectin (HHL) Man 100 mM mannose Lycopersicon esculetumtomato (TL) GlcNAc Chitin Hydrolysate ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Addition of competitive sugar inhibitor blocks lectin expression as determined by immunocytochemistry.Addition of 200 mM galactose can block RCA binding in a hESC colony as shown by lack of RCA binding (6A), but does not affect DAPI stain (blue). However, in the absence of galactose, uniform binding of RCA (red) was detected 6(B). 6A-B) 20× magnification, 100 μm ::: ![](1471-213X-5-15-6) ::: Discussion ========== This study is the first of which we are aware to use lectins to probe the surface of hESCs in order to analyze carbohydrate presentation. To quantitate lectin binding percentages on pluripotent hESC surfaces using flow cytometry, we first enriched our colonies for SSEA-4 expression. Using enriched pluripotent hESCs, we found a large binding percentage range for this chosen panel of 14 lectins. Some lectins bound throughout the colonies of pluripotent hESCs, and the carbohydrate moieties represented by this group include alpha-linked mannose (Con A), galactosyl(β1--3) N-acetylgalactosamine (PNA), terminal galactose and N-acetylgalactosamine (RCA), GalNAc β4-Gal (WFA), N-acetylglucosamine (TL) and sialic acid alpha 2,6 GalNAc (SNA) and alpha-linked mannose residues only (HHL). These results indicate that there is high expression of these carbohydrates with these particular linkages on pluripotent hESC surfaces. As ConA can bind both mannose and glucose moieties, we see greater lectin binding of ConA as opposed to HHL, which specifically binds only mannose residues. These lectins may be additional markers to denote pluripotent hESCs. Similarly, WFA, PNA and SNA were found to bind to over 60% of hESCs, suggesting that these lectins could be indicative of a pluripotent phenotype as well; however, they do not have binding percentages that correlate with SSEA-4 expression. We also found that some lectins showed interesting partial binding patterns. Two distinct populations of cells were observed as determined by flow cytometry plot analysis and confirmed by immunocytochemistry using these lectins and SSEA-4 antibody. These hESC populations were identified as either SSEA-4+/lectin+ or SSEA-4+/lectin- regions. Lectins showing this binding pattern included PHA-L, VVA (see Figure [5D--F](#F5){ref-type="fig"}), UEA, PHA-E (Figure [5G--H](#F5){ref-type="fig"}), and MAA. Table [1](#T1){ref-type="table"} shows their carbohydrate binding specificities. As shown by immunocytochemistry of the representative lectin, PHA-E, it appeared that some areas that were beginning to lose expression of SSEA-4, show binding of the lectin (Figure [5G--H](#F5){ref-type="fig"}). A decrease of SSEA-4 expression can occur even under the most stringent culturing conditions and in this research, is important because it underscores the need for enriching for desired populations of cells prior to experimentation and also supports the hypothesis that cell-cell interactions are important in carbohydrate presentation. As clusters of cells lose SSEA-4 expression, and therefore differentiate, new carbohydrates are presented on the cell surface. Furthermore, these differentiated cells could be indicative of progenitor subpopulations existing in the colonies. For example, another lectin in our study that was found to bind partially to colonies is UEA, which binds fucose moieties and has been determined to be a marker for endothelial cells \[[@B16]\]. The role these partially presented carbohydrates play was not investigated in this study, but others have suggested that certain carbohydrates may serve a developmental function. For example, Panin and colleagues have identified O-fucose on epidermal growth factor-like repeats of Notch, and elongation of O-fucose has been implicated in the modulation of Notch signaling by Fringe \[[@B36],[@B37]\]. Notch receptors and associated proteins are important in a number of signaling pathways that direct cell fate decisions, proliferation and apoptosis. We also observed that two lectins, DBA and LTL, did not bind hESCs. DBA binds primarily to α-linked N-acetylgalactosamine and LTL binds fucose moieties that are α-linked to GlcNAc. These non-binding lectins suggest that these particular carbohydrates and their associated linkages are either not present on pluripotent hESC surfaces or if they are present, they have been modified and therefore are not susceptible to lectin binding. Interestingly, however, PNA and WFA, which bind to a high percentage of hESCs, bind to β-linked N-acetylgalactosamine. Thus, it appears that the linkage of this carbohydrate is important. Similarly, UEA binds to about 20% of our cells and recognizes fucose moities like LTL. However, the linkages that these lectins recognize are different and cause different binding percent outcomes, further suggesting that the linkage of these carbohydrates are important. Developmentally regulated cell surface antigens have been identified by monoclonal antibodies in a variety of species and stem cell types. Many of these antigens have proven to be side chains of membrane glycolipids and glycoproteins. For example, the SB10 antigen found on human mesenchymal stem cells was shown to be the activated leukocyte-cell adhesion molecule (ALCAM) that possesses N-linked oligosaccharide side chains \[[@B38]\]. Muramatsu and colleagues discovered a protein called embigin that has a developmentally regulated carbohydrate chain that is lost upon mouse embryogenesis \[[@B39]\]. Furthermore, other researchers showed that the monoclonal antibody GCTM-2, which binds to human embryonic carcinoma cells and primate ES cells, recognizes an epitope on a keratan sulfate proteoglycan \[[@B40],[@B41]\]. These studies have shown not only that carbohydrates on cell surfaces change with embryological development but also that they can be used to identify different types of stem cells. In addition, there are a number of carbohydrate epitopes that have been found on the mammalian preimplantation embryo. Ofcourse, the tumor rejection antigens and stage specific antigens have been well characterized on human embryos \[[@B4]\], but there are many others such as the TEC antigens (TEC1- TEC4), LeX, LeY, CD46, CD55, and CD59. The TEC antigens have been shown to recognize either the carbohydrate moiety of embryoglycan or a developmentally regulated protein epitope on both mouse and bovine preimplantation embryos \[[@B42]\]. LeX \[Galbeta1-4(Fucalpha1-3)GlcNAc\] is first detected on the blastomeres of the 8-cell stage of the mouse embryo and LeY \[Fucalpha1-2Galbeta1-4(Fucalpha1-3)GlcNAc\] is highly expressed on the surface of the mouse blastocyst \[[@B43],[@B44]\]. Also, studies of human preimplantation embryos show that the glycosyl phosphatidylinositol anchored proteins CD46, CD55 and CD59 are expressed on oocytes and plasma membrane \[[@B45]\]. In this study, we analyzed glycans on the cell surface of hESCs using a panel of 14 lectins and SSEA-4 antibody and found that pluripotent hESC surfaces have a large variety of surface carbohydrates that can be targeted for marker sources. Conclusion ========== This work represents a fundamental base to systematically classify pluripotent hESCs, and in future studies these lectins may be used to distinguish differentiated hESC types based on carbohydrate presentation that accompanies differentiation. It will be important to determine the potential of these lectin positive cell types and to identify the cell surface glycoproteins or glycolipids to which these lectins bind. Because there are many different enzymes that add or remove carbohydrate structures on cell surfaces, it will also be interesting to investigate the function of some of these glycosyltransferases in both pluripotent and differentiated hESCs. Methods ======= Cell culture and passaging -------------------------- NIH approved hESC lines BG01 and BG02 with normal karyotypes were obtained from BresaGen, Inc. \[[@B33]\]. Cells were grown in DMEM/F12 supplemented with 15% FCS (HyClone), 5% knockout serum replacer, 1× non-essential amino acids, 20 mM L- Glutamine, 0.5 U/ml penicillin, 0.5 U/ml streptomycin, 4 ng/ml FGF-2 (Sigma), (all from Gibco Invitrogen unless otherwise labeled). Human ESCs were maintained on mitotically inactivated primary mouse embryonic fibroblasts (MEF) feeder layers for routine maintenance. Cells were grown in 100 cm tissue culture treated dishes (Falcon). Cells were passaged every 3 days using either a pretreatment with 10 mg/ml collagenase for 2 minutes followed by 0.05% trypsin for 1 minute or manual dissection with a fire pulled Pasteur pipette \[[@B33]\]. Immunocytochemistry was performed on routinely maintained adherent hESC colonies, and flow cytometry was performed using routinely maintained hESC colonies that were first enriched for SSEA-4. Enrichment of SSEA-4 positive cells ----------------------------------- Cultures of hESCs were grown in 100 cm dishes and trypsin passaged into single cell suspensions as described above. Cells were incubated on ice for 15 minutes in 1:10 dilution of SSEA-4 (MC 813-70, Developmental Studies Hybridoma Bank \[DSHA\]; Iowa City; in stain buffer (SB). SB consisted of 0.5 ml 0.5 U/ml penicillin, 0.5 U/ml streptomycin, 0.5 ml of 100 mM EDTA, 46.5 ml phosphate buffered saline (PBS) and 2.5 ml fetal bovine serum (FBS). After incubation 10 ml SB was added and cells were resuspended and centrifuged at 3000 × g for 5 minutes. Supernatant was removed, leaving the cell pellet intact, and cells were washed again in SB and resuspended. After centrifugation, a 1:4 dilution of secondary anti-mouse IgG in stain buffer was added to the cell pellet, and the resuspended pellet was incubated on ice for 25 minutes. After incubation 10 ml SB was added to wash the cell pellet, followed by centrifugation (3000 × g) for 5 minutes. This process was repeated two more times in 5 ml SB. Cells were finally resuspended in 500 μl of SB before being applied to a pre-washed magnetic bead column. The flow through from the column was collected and saved for counting, and the retained eluate was collected separately. Both flow through and eluate were brought up to 5 ml in SB after collection and counted \[[@B46]\]. Cells were then subjected to flow cytometry. Flow cytometry results presented here are from hESCs that were used immediately after the enrichment procedure. Flow cytometry -------------- Cell surface antigen and carbohydrate expression of SSEA-4+ enriched hESCs was assessed by indirect immunofluorescence detected by flow cytometry to provide a quantitative binding percentage of both SSEA-4 and lectin. Human ESCs were harvested into single cell suspensions using trypsinization as described above. Cells were enriched for SSEA-4 as described above. Then, cells were fixed in 2% paraformaldehyde in 1× PBS. After blocking in 1% bovine serum albumin (BSA) for 30 minutes, cells were placed in sterile conical tubes in aliquots of 500,000 cells each and double stained with one of the 14 lectins at 5 μg/ml and SSEA-4 in a 1:100 dilution. Cells were washed 3 times with PBS and then stained with secondary antibodies that included streptavidin -allophycocyanin (1:250, BD Biosciences; Franklin Lakes, NJ;) for recognition of biotinylated lectins and antigoat Mouse IgG conjugated Alexa 488; (1:2000) for recognition of SSEA-4. These secondary antibodies were chosen so that there would not be overlap in the emission/excitation wavelengths and so that double staining could be performed. Unstained, enriched hESCs and enriched hESCs stained with secondary antibodies alone were used as controls. Table [1](#T1){ref-type="table"} shows the chosen tested lectins, their commonly abbreviated name, and the specificity of these lectins for their respective monosaccharides to act as a quick reference guide and the concentration of hapten, or competitive sugar, that was used to verify specificity of each lectin for its carbohydrate. Specificity of each lectin has previously been described in detail by Cummings and colleagues \[[@B47]\]. Cytometry was performed using a Beckman Coulter Cytomics FC 500 Flow Cytometer. Data analysis was performed using the RXP Analysis Software by Beckman Coulter and Windows Multi Document Interface for Flow Cytometry (WinMDI 2.8). Three independent assays were carried out using both BG01 and BG02 hESC lines. Immunocytochemistry ------------------- Immunocytochemistry was used to analyze the localization of cell surface carbohydrate expression and SSEA-4 on routinely maintained adherent cultures of hESCs. Human ESCs were harvested from 100 cm^2^dishes by either trypsin/collagenase dissociation or manual dissociation and fixed in 4% paraformaldehyde (Fisher Scientific) in 1× PBS (Gibco) for 30 minutes. After blocking in 1% BSA solution for 30 minutes, cells were equally divided into 4-well chamber slides which had previously been plated with mitotically inactivated MEFs. Cells were then double stained with one of 14 biotinylated lectins (all lectins obtained from Vector Laboratories; Burlingame, CA; 10 μg/ml) and SSEA 4 (MC 813-70, Developmental Studies Hybridoma Bank \[DSHA\]; Iowa City;1:100) for 20 minutes at 37°C, followed by 3 washes in PBS. Secondary antibodies included streptavidin conjugated Alexafluor 594 (Molecular Probes; Eugene, OR; 1:250 dilution) and antigoat Mouse IgG conjugated Alexa 488 (Molecular Probes; 1:2000 dilution). During staining procedure, cells were kept at 37°C. Cells were washed three times after secondary antibody incubation for 5 minutes in PBS. Cells were then post-stained with 5 ng/ml DAPI (Sigma) to detect cell nuclei and washed overnight in PBS. Controls of unstained cells were obtained by incubation with secondary antibodies alone. All slides were mounted and visualized using a Nikon TS100 inverted microscope. Individual color channels were captured separately and merged in Adobe Photoshop. Abbreviations used ------------------ human embryonic stem cells (hESCs) stage specific embryonic antigen-4 (SSEA-4) inner cell mass (ICM) N-acetylgalactosamine (GalNAc) Fibroblast growth factor-2 (FGF-2) mouse embryonic fibroblasts (MEF) stain buffer (SB) phosphate buffered saline (PBS) fetal bovine serum (FBS) bovine serum albumin (BSA) Concanavalin A(Con A) Phaseolus vulgariserythro-agglutinin (PHA-E) Phaseolus vulgarisleuco-agglutinin (PHA-L) Sambucus nigraagglutinin (SNA) Arachis hypogeapeanut (PNA) Vicia villosaagglutinin (VVA) Maackia amurensis(MAA) Ricinus communisagglutinin (RCA) Wisteria floribundaagglutinin (WFA) Ulex europaeusagglutinin (UEA) Lotus tetragonolobuslectin (LTL) Dolichos biflorusagglutinin (DBA) Hippeastrum hybridlectin (HHL) Lycopersicon esculetumtomato (TL) Authors\' contributions ======================= AV carried out the maintenance and culturing of the hESCs, flow cytometry and immunocyotochemical studies and drafted the manuscript. MM helped in maintenance of hESCs, immunocytochemistry, and aided in manuscript review. KJ participated in flow cytometry set up. SS, IL, and MP participated in the design of the study. SS participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript. Links ===== Developmental Studies Hybridoma Bank <http://www.uiowa.edu/~dshbwww> BD Biosciences <http://www.bdbiosciences.com> Vector Laboratories <http://vectorlabs.com> Molecular Probes <http://www.probes.com> Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Histogram plots of other lectins tested.Figure S1 shows peak shifts of RCA (A), PNA (B), ConA (C), WFA (D), SNA (E), and PHA-L (F) binding in black tracing, and unstained cells in grey fill. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Flow cytometry plots of other lectins tested.Figure S2 shows plots of some lectins that have 2 populations of cells such as PHA-L (A), UEA (B), and VVA (C). The plot of CON A (D) indicates only one population of cells showing a positive shift for lectin and SSEA-4 antibody binding, while the plot of DBA (E) and LTL (F) also show only one population of cells with no shift in lectin binding, but a positive shift in the SSEA-4 detection channel. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We would like to acknowledge Scott Noggle for magnetic bead sorting protocol; Deb Weiler for culture and maintenance of mouse embryonic fibroblast feeder layers; MaryAnne Della-Fera and Raj Rao for manuscript review and Bresagen, Inc. for supplying the hESC cells.	PubMed Central	2024-06-05T03:55:59.947997	2005-7-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182361/", "journal": "BMC Dev Biol. 2005 Jul 21; 5:15", "authors": [ { "first": "Alison", "last": "Venable" }, { "first": "Maisam", "last": "Mitalipova" }, { "first": "Ian", "last": "Lyons" }, { "first": "Karen", "last": "Jones" }, { "first": "Soojung", "last": "Shin" }, { "first": "Michael", "last": "Pierce" }, { "first": "Steven", "last": "Stice" } ] }
PMC1182362	Background ========== Molecular phylogenetics has expanded understanding of relationships among all major angiosperm groups and has thereby strongly impacted their classification \[[@B1]\]. More recently, such advances have also included some nonphotosynthetic holoparasites whose phylogenetic positions had previously been uncertain, such as Hydnoraceae \[[@B2]\] and Rafflesiales \[[@B3]\]. The latter study documented that Rafflesiales are actually a polyphyletic assemblage of three or four independent evolutionary lineages. The losses and reductions in features that are pervasive in parasitic plants have resulted in remarkable morphological convergences, thereby explaining their erroneous placement by traditional methods. Previous molecular phylogenetic work with such holoparasites highlighted the need to employ gene sequences from different subcellular compartments and analytical methods that accommodate rate heterogeneity, thus avoiding long-branch attraction artifacts \[[@B3]\]. These steps are justified because congruence among different gene trees provides evidence that the organismal tree is being recovered, while incongruence suggests the presence of nonstandard processes such as introgression, lineage sorting, and horizontal gene transfer (HGT) \[[@B3]-[@B6]\]. In addition to the abovementioned holoparasites, another group with uncertain placement is Balanophorales. For over 150 years, there has been taxonomic debate as to whether the genus Cynomoriumis part of Balanophoraceae \[[@B7],[@B8]\] or whether it should be classified as a separate family, Cynomoriaceae \[[@B9]\]. These plants are fleshy, monoecious or dioecious holoparasitic herbs that often produce swollen tuberous haustorial root connections to their host plants. Their stems bear scalelike leaves and their unisexual flowers represent the ultimate in reduction among angiosperms \[[@B10]\] (Figure [1](#F1){ref-type="fig"}). Balanophoraceae (in the strict sense) contains 17 genera and 44 species in the neo- and paleotropics whereas Cynomoriumhas two species in eastern Asia (C. songaricum) and the Mediterranean (C. coccineum). The presence of bisexual flowers, an unusual bimodal karyotype \[[@B11]\] and features of the stamens, ovules, and embryo sac \[[@B12]\] in Cynomoriumall support its segregation from Balanophoraceae. Whether considered one or two families, all past and present classifications accept a relationships between Cynomoriumand Balanophoraceae. Previous molecular phylogenetic work using only nuclear small-subunit ribosomal DNA \[[@B13]\] suggested placement of Cynomoriumwith Saxifragales and Balanophoraceae with the sandalwood order (Santalales). The validity of these results could not be assessed because high substitution rates in these plants could potentially confound phylogeny estimation \[[@B14]\] and because confirmatory data from other genes was lacking. Here we utilize both nuclear and mitochondrial gene sequence data, analyzed with maximum parsimony (MP) and Bayesian inference (BI) methods, to determine whether Balanophoraceae and Cynomoriaceae are related to one another and where these clades fit in the global angiosperm phylogeny. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### *1 Morphology of Cynomorium coccineum(a-e) and Balanophora fungosa(f). a, Habit showing inflorescences emerging from the ground. b, Scanning electron micrograph (SEM) of staminate flower. c, SEM of carpellate flower. d, SEM of bisexual flower. e, SEM of pollen. f, SEM of carpellate flowers of Balanophorashowing remarkable convergence with those of Cynomorium. ::: ![](1471-2148-5-38-1) ::: Results ======= Results of analyses of of the global data set using both MP (Figure [2](#F2){ref-type="fig"}) and BI ([Additional File 3](#S3){ref-type="supplementary-material"}) strongly indicate that Cynomoriumis not closely related to Balanophoraceae but is a component of Saxifragales. Small-subunit (SSU) nuclear ribosomal DNA alone analyzed with BI ([Additional File 4](#S4){ref-type="supplementary-material"}) and ML ([Additional File 5](#S5){ref-type="supplementary-material"}) gave the same result as obtained with the global data set. Similarly, mitochondrial matRanalyzed alone with BI ([Additional File 6](#S6){ref-type="supplementary-material"}) showed Cynomoriumwell separated from Balanophoraceae. Moreover, strong support is obtained from multigene and separate gene analyses for a relationship between Balanophoraceae and Santalales. Given these results, attempts were made to precisely determine the sister taxon of Cynomoriumwithin Saxifragales, a clade that has been subjected to extensive molecular phylogenetic work \[[@B15],[@B16]\]. The shortest MP trees contained a clade composed of Cynomoriumand Crassulaceae (Figure [3](#F3){ref-type="fig"}), however, this relationship did not receive high bootstrap (BS) support. A similar lack of resolution for this clade was also seen in the BI tree ([Additional File 7](#S7){ref-type="supplementary-material"}). On both the MP and BI trees, Cynomoriumis part of a clade with moderate support that contains Crassulaceae, Paeoniaceae, Aphanopetalaceae, Tetracarpaeaceae, Haloragaceae, Pterostemonaceae, Iteaceae, Grossulariaceae, and Saxifragaceae. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Maximum parsimony strict consensus tree from the globaldata set. Data set of combined nuclear SSU rDNA, chloroplast rbcLand atpBand mitochondrial matR. Strict consensus of 4 trees. Bootstrap support values are shown above the lines, Bayesian posterior probabilities below. The minimum length tree was 9405 steps (consistency index minus uninformative sites 0.3126, retention index 0.4301, rescaled consistency index 0.1709). Cynomoriumis not part of Balanophoraceae but a taxon within Saxifragales. ::: ![](1471-2148-5-38-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Maximum parsimony strict consensus tree from the Saxifragalesdata set. Data set of combined nuclear SSU and LSU rDNA, chloroplast rbcL, atpB, and matK. Strict consensus of 9 trees. Bootstrap support values are shown above the lines. The minimum length tree was 8282 steps (consistency index minusuninformative sites 0.3975, retention index 0.5609, rescaled consistency index 0.2750). Although the sister relationship between Cynomoriumand Crassulaceae is seen in the shortest trees, bootstrap support for this clade is low. ::: ![](1471-2148-5-38-3) ::: Discussion ========== This work documents that Cynomoriumand Balanophoraceae are not closely related, a result in conflict with past and present classifications. Assuming our molecular gene trees reflect the organismal tree, these data indicate that morphological features, such as inflorescences bearing numerous highly reduced flowers, are convergent and were thus attained independently by these two holoparasite lineages. Cynomoriumin Saxifragales --------------------------- From an historical perspective, Hooker \[[@B7]\] allied Cynomoriumand Balanophoraceae with Haloragaceae (the latter is a family in Saxifragales, see Figure [3](#F3){ref-type="fig"}) because both groups have epigynous flowers with stamens opposite the valvate perianth lobes. He also noted the striking similarities between the female flowers of Gunneraand Lophophytum(Balanophoraceae). Our analyses that included Gunnera(data not shown) indicate it is not closely related to either Cynomoriaceae or Balanophoraceae, a result in agreement with other work that placed Gunneraas sister to all core eudicots \[[@B17],[@B18]\]. Other 19^th^century workers such as Hoffmeister proposed a relationships between Cynomoriumand Hippuris(Plantaginaceae, an asterid) because both have a single, epigynous stamen attached to the top of a unicarpellate ovary. This relationship was not confirmed following a detailed floral morphological study \[[@B19]\] nor was Cynomoriumshown to be related to asterids following molecular analyses reported here. The exact position of Saxifragales within the core eudicot clade remains uncertain because various molecular phylogenetic analyses are in conflict (c. f. \[[@B17],[@B18],[@B20],[@B21]\]). These studies have proposed relationships with rosids, asterids, caryophyllids, and Santalales, i.e. nearly all members of the core tricolpate clade (see review by Judd and Olmstead \[[@B22]\]). The results reported here also demonstrate uncertainty with regard to the topology of the above clades (Figure [2](#F2){ref-type="fig"}). Given the variety of ways these clades could resolve, the possibility exists that Balanophoraceae and Cynomoriumare more closely related than graphically depicted in Figure [2](#F2){ref-type="fig"}. Although future molecular phylogenetic studies may support this, it is unlikely that these two holoparasite groups will become sister given that the Balanophoraceae + Santalales clade has BS support of 94% and Cynomoriumis sister to Peridiscuswith 98% BS support. Moreover, the latter two genera are sister to Hamameliswith a 98% BS value. Long-branch attraction has long been suggested as a confounding factor when conducting phylogenetic analyses of organisms with marked among-lineage rate heterogeneity \[[@B23]\]. Indeed, our group has shown that Rafflesiales, another holoparasitic angiosperm group, is susceptible to long-branch attraction, particularly when analyzed with MP \[[@B3]\]. The results reported here differ from that study in several respects. First, MP analysis of the global data set does not indicate that Balanophoraceae and Cynomoriumexhibit particularly long branches with respect to other angiosperms, nor are these holoparasite clades \"attracted\" to each other ([Additional File 8](#S8){ref-type="supplementary-material"}). This result differs from that obtained when the Rafflesiales molecular data are analyzed with MP where the four distinct lineages (Apodanthaceae, Cytinaceae, Mitrastemonaceae and Rafflesiaceae) appear monophyletic \[[@B3]\]. When the global data set is analyzed with BI, some degree of branch length heterogeneity is seen ([Additional File 9](#S9){ref-type="supplementary-material"}), however, Balanophoraceae and Cynomoriumstill remain on distinct clades. It should be pointed out that for the global data set, branch length comparisons between holoparasites and photosynthetic plants are somewhat biased given that the chloroplast genes were coded as missing for the holoparasites. Taken together, we feel our results are not being influence by long-branch attraction and that they provide strong evidence for the independent evolution of these two holoparasite lineages. Attempts to find morphological features that link Cynomoriumwith Saxifragales are complicated, first because the latter order is morphologically heterogeneous and second because of morphological reductions in the holoparasite (see below). The heterogeneity of Saxifragales is evidenced by the fact that pre-molecular classifications placed many of its component families in distinct subclasses \[[@B8]\]. Although a close relationship between Cynomoriumand Crassulaceae was not seen on the BI tree, the shortest MP tree showed these genera as sister taxa, thus this possible affinity will here be considered. Tricolporate pollen (Figure [1e](#F1){ref-type="fig"}), stem succulence, and proanthocyanidins (red pigments, see \[[@B24]\]) are possible characters that may reflect a relationship between Cynomoriumand Crassulaceae, but these are all general features found in many groups. Other than such facies, there are few floral characters that could be considered synapomorphies between the extremely reduced flowers of Cynomoriumand those of Crassulaceae. The staminate flower of Cynomoriumis composed of four to six distinct perianth parts forming a whorl below the single stamen and a wedge-shaped structure interpreted as a pistillode \[[@B25]\]. The apical portion of the pistillode is concave and accepts the base of an anther theca prior to filament elongation (Figure [1b](#F1){ref-type="fig"}) and its internal lateral surface has a groove that accepts the filament. Stamen morphology is not unlike many tricolpate angiosperms: the anthers are dorsifixed, tetrasporangiate, dithecal and open by longitudinal slits. The carpellate flower (Figure [1c](#F1){ref-type="fig"}) is composed of a single carpel with an elongated, grooved style. The perianth (or perigonial scales) are much smaller than in the staminate flower, reduced to small papillae at the summit of the ovary or along its sides. The bisexual flowers are similar to carpellate flowers except for the addition of a stamen at the base of the ovary (Figure [1d](#F1){ref-type="fig"}). As with many holoparasitic flowering plants, the reduction or loss of morphological features interferes with the comparative morphological approach when searching for relatives among less modified photosynthetic plants. Balanophoraceae in Santalales ----------------------------- Although precedents exists for placing Balanophoraceae with Santalales \[[@B8],[@B26]-[@B29]\], recovering this relationship with both nuclear and mitochondrial genes is surprising because any morphological similarities (e.g., reduction in the gynoecium) have usually been assumed to be cases of convergence \[[@B30]\]. Moreover, BI of the Saxifragales data set indicates a derivation of Balanophoraceae from withinSantalales, not as its sister (Figure [3](#F3){ref-type="fig"} and [Additional File 8](#S8){ref-type="supplementary-material"}). Although a multigene analysis of Santalales with robust taxon sampling exists \[[@B31]\], it was generated with nuclear ribosomal DNA and chloroplast genes. To conduct analyses with Balanophoraceae and Santalales, mitochondrial genes from the latter will be needed to test relationships. Cronquist \[[@B8]\] viewed the sandalwoods as the best candidate ancestral group for Balanophoraceae, however, he admitted any similarities might reflect convergent adaptations to parasitism. Of the 17 genera in the family, only Dactylanthus, Hachetteaand Mystropetalonwere used in this study because they are sister (and basal) to the remaining genera and have lower substitution rates in SSU rDNA, thus avoiding potential long-branch artifacts. The floral morphology of Mystropetalonis less reduced than other genera in the family. Given that its staminate flowers have pistillodes (vestigial carpels) and carpellate flowers have staminodes (vestigial stamens), it is assumed this plesiomorphic condition reflects evolution from an ancestor with bisexual flowers. Floral features in common with Santalales include a valvate perianth and stamens with \"typical\" morphology (i.e. anther sacs and filaments as opposed to synangia as seen in Balanophora). When examining the haustorial structure of Mystropetalon, Weber \[[@B32]\] noted the presence of a collapsed zone, graniferous tracheary elements, and \"runners,\" which he associated with Santalales. Similarly, reduction in ovules is seen in several santalalean families, particularly the mistletoes in Viscaceae where no true ovule exists. Either an embryo sac is embedded within central placenta called a \"mamelon\" or the archesporium develops hypodermally in the ovary, functioning directly as megaspore mother cells \[[@B33]\]. Similarly, ovules have been lost in some Santalaceae (Exocarpos) and Loranthaceae. During the course of floral reduction, the loss of ovules most influenced Fagerlind \[[@B27]\] who drew a direct ancestor-descendant relationship between Santalaceae and Balanophoraceae. As mentioned above, such apparent similarities between Balanophoraceae and Santalales have also been interpreted as convergent. Kuijt \[[@B10]\] suggests that the reduction of santalalean ovules may permit retention of meristematic activity, a requirement for the formation of complex fruits in epiparasites. For root parasites such as Balanophoraceae, he suggests ovular reduction is related to selection for tiny seeds that require host stimulants for germination, as is seen in Orobanchaceae. In addition to ovules, the androecia of both Viscaceae and Balanophoraceae are diverse with both showing trends toward the formation of synandria and porose dehiscence. Whether these trends are convergent or follow from homologous suites of genes may yield to direct testing using floral homeotic mutants. Although herbaceous root parasites have evolved in Santalaceae (e.g. Thesium, Quinchamalium), most members of Santalales are woody. Completely holoparasitic species do not occur in this order, with the possible exception of Daenikera, an endemic monotypic genus of New Caledonia whose photosynthetic capabilities remain to be determined. As adults, dwarf mistletoes (Arceuthobium, Viscaceae) fix only about 30% of the carbon needed for growth \[[@B34],[@B35]\] but prior to host attachment seedlings actively photosynthesize. For this reason these mistletoes must be considered advanced obligate hemiparasites, not holoparasites. Thus, to include with Santalales the entirely holoparasitic group Balanophoraceae requires further investigation to identify the exact sister group. Ethnobotany and conservation biology of Cynomorium* ---------------------------------------------------- Cynomorium coccineum, known to the Muslim world as \"tarthuth,\" has been harvested from the deserts of north Africa and the Middle East for thousands of years. Arabs and Bedouins eat the interior portions of fresh young stems, prepare infusions of older stems to treat colic or stomach ulcers, or dry and pulverize the plant for use as a spice or condiment with meat dishes \[[@B36]\]. Medicinal uses of tarthuth can be traced to Al-Kindi, Al-Razi (Rhazes), Ibn Masawayh, Ibn Wahshiya, and Maimonides but the plant became known to Europeans only in the 16^th^century. A group called the Knights Hospitaller of St. John operated a hospital in Jerusalem and learned of the medicinal qualities of tarthuth from local Muslim physicians. When the Crusaders lost Jerusalem to the Muslims, they moved to the island of Malta where Cynomoriumwas also native. The site where the \"Maltese Mushroom\" grew (Fungus Rock) was thereafter vigorously guarded and thieves were imprisoned or made galley slaves. The \"treasure of drugs,\" as the Arabs called it, was used for a variety of purposes, including treating apoplexy, venereal disease, high blood pressure, vomiting, irregular menstrual periods and as a contraceptive and toothpaste. Modern biomedical and phytochemical research on Cynomorium coccineumhas demonstrated a variety of activities from plant extracts. Effects of Cynomoriumextracts on mammalian reproductive cells modulation of pituitary gonadotrophins \[[@B37]\] as well as changes in testicular development \[[@B38]\] and epididymal sperm patterns \[[@B39]\] in rats. Cynomorium songaricum, known in Chinese medicine as \"suo yang,\" has been shown to contain triterpenes with HIV protease inhibitory activity \[[@B40],[@B41]\]. Interest in herbal medicines is growing at a rapid pace, and attention has been focused upon Cynomoriumas evidenced by its availability via hundreds of distributors advertising (via the internet) herbal remedies for kidney and intestinal ailments as well as for impotence. Although it is not at present clear how many original distributors of the plant exist, and whether plant material being marketed as Cynomoriumis actually this plant, what is clear is that it is not being cultivated, thus authentic herbal preparations must be obtained from wild populations. Very little information exists on the cultivation of Cynomorium\[[@B25]\] and certainly commercial suppliers are not practicing sustainable harvest by cultivating this obligate holoparasite. There is evidence that overexploitation of this plant has resulted in localized extinction \[[@B42]\]. For these reasons, we here raise concerns for the conservation of both species of Cynomoriumand strongly voice the need to develop cultivation methodologies. Given the potential and actual biomedical applications of extracts from this plant, and conservation concerns given extensive harvesting from wild populations, information on its phylogenetic position within angiosperms is timely. Further molecular work will likely illuminate its closest relatives within Saxifragales. These taxa should then become the subject of phytochemical analyses to determine whether they also contain compounds of biomedical interest. Cultivation of photosynthetic plants would be more straightforward than the holoparasite, thus possibly relieving some of the pressure to harvest this more sensitive species from the wild. Conclusion ========== All previous classifications have allied Cynomoriumand Balanophoraceae, likely owing to the holoparasitic habit and the presence of inflorescences with numerous tiny flowers. Molecular phylogenetic analyses using nuclear and mitochondrial gene sequences both indicate that these taxa are not closely related and that perceived similarities are a result of convergent evolution. Cynomoriumis strongly supported as being a member of Saxifragales, however, its exact position within this order remains unresolved. Surprisingly, both nuclear and mitochondrial genes place Balanophoraceae with the sandalwood order, a relationship previously proposed but explained by some as a case of convergent evolution. Both species of wild Cynomoriumare being harvested from wild populations for use in herbal medicines, and evidence exists for overexploitation. Given that methods to cultivate these holoparasites are not being employed, we here raise conservation concerns. If further molecular phylogenetic work identifies the nearest photosynthetic relatives of Cynomorium, these plants should be examined for phytochemical activity. Methods ======= Taxon sampling and data collection ---------------------------------- Voucher information for all newly sequenced taxa is given in [Additional Files 1](#S1){ref-type="supplementary-material"} and [2](#S2){ref-type="supplementary-material"}. DNA was extracted, PCR amplified, cloned, and sequenced following reported methods \[[@B3],[@B43]\]. The nuclear and mitochondrial sequences were amplified using primers reported elsewhere \[[@B14],[@B44],[@B45]\]. Sequencing was conducted using an ABI Prism^®^377 automated DNA sequencer (Applied Biosystems) according to the manufacturer\'s protocols. Two multiple sequence alignments were constructed. The purpose of the *globaldata set was to test the position of Cynomoriumand the three Balanophoraceae genera among angiosperm orders. This matrix included nuclear small-subunit (SSU) rDNA, chloroplast rbcLand atpB, and mitochondrial matRfor 67 taxa with Laurusand Cinnamomumas outgroups. Balanophoraceae and Cynomorium lack the chloroplast genes rbcLand atpB, but they were included in the global data set to stabilize the overall tree topology as previously demonstrated \[[@B2],[@B3]\]. Given the strong support for a relationship between Cynomoriumand Saxifragales obtained from this global analysis, a Saxifragalesdata set was constructed to further resolve the position of the parasites within this clade. Here the genes used were nuclear SSU and LSU rDNA, chloroplast rbcL, atpB, and matKfor 62 taxa with Leaand Vitisas outgroups. In both alignments, chloroplast genes were coded as missing for the holoparasites. Accession numbers for newly sequenced genes are: AY957440 -- AY957454. Phylogenetic analyses --------------------- The globaland Saxifragalesdata sets were analyzed using maximum parsimony (MP) in PAUP\ 4.0b10 \[[@B46]\] and Bayesian inference (BI) methods in MrBayes 3.0b4 \[[@B47]\]. MP searches were performed using 100 random addition sequence replicates with tree-bisection-reconnection (TBR) branch-swapping, holding ten trees at each addition step, with all sites equally weighted. Fully partitioned Bayesian analyses were performed by partitioning each data set by gene and, for the protein-coding genes, by codon position. This resulted in a total of 10 partitions for the full data set and 11 partitions for the Saxifragales data set (Table [1](#T1){ref-type="table"}). MP trees were constructed for each gene (following the protocol described above), and these trees were used in PAUP\* to evaluate 24 nucleotide substitution models for each data partition. For example, models were evaluated for the \"rbcLPos1\" partition (the first codon position of the rbcLdata set) on the MP tree for the rbcLdata set. MrModelTest 2.0 \[[@B48]\] was used to select an appropriate model from the PAUP\* output via a second-order version of the Akaike Information Criterion (AICc) that takes sample size into account, as recommended by Posada and Buckley \[[@B49]\]. AICcvalues were computed using both the total number of characters and the number of variable characters per partition. For nearly all partitions, the Akaike weight for the chosen model was much higher than the Akaike weight of the next best model, so model-averaged analyses were not performed. The best-fitting models and their Akaike weights for each data partition are listed in Supplementary Data. Partitioned Bayesian analyses were performed with all model parameters unlinked (i.e., model parameters for each data partition were estimated separately from each partition), and with topology and branch lengths linked. Two separate analyses (with different MCMC seeds, random starting trees and default uniform priors for all parameters) were run for each data set for 15 million generations, with trees sampled every 500 generations. Trees recovered during the first 2.5 million generations (the first 5000 trees) in both runs were discarded as burn-in, leaving a total of 25,000 trees which were used to construct majority-rule consensus trees. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Summary of output from MrModelTest for each data partition in the globaland Saxifragalesdata sets § ::: Data Set Data Partition Chosen Model (AICc) Chosen Model (hLRT) Characters (total/variable) Akaike Weight of Chosen Model --------------------------------- -------------------- ----------------------------------- ------------------------- --------------------------------- ----------------------------------- Both globaland Saxifragales atpBPos3 GTR+I+Γ GTR+I+Γ/GTR+Γ 490/416 0.9997 (all) rbcLPos1 GTR+I+Γ GTR+I+Γ 467/130 1.0000 (all) rbcLPos2 SYM+I+Γ SYM+I+Γ/JC+I+Γ 467/84 0.7397 (all) rbcLPos3 GTR+I+Γ GTR+I+Γ 467/391 0.9911 (all) nu SSU GTR+I+Γ SYM+I+Γ/GTR+I+Γ 1750/428 0.8264 (all) Just global atpBPos1\* GTR+I+Γ GTR+I+Γ 490/139 1.0000 (all) atpBPos2\* GTR+I+Γ GTR+I+Γ 490/68 0.9872 (all) matRPos1 GTR+Γ GTR+Γ 743/298 0.6598 (all) matRPos2 GTR+Γ GTR+Γ/HKY+Γ 743/285 0.5628 (all) matRPos3 GTR+Γ SYM+Γ/GTR+Γ 742/376 0.6739 (all) Just Saxifragales atpBPos1\* GTR+G GTR+G/GTR+I 480/85 0.4127 (all) atpBPos2\* GTR+I+Γ (all)/ HKY+I+Γ (variable) HKY+I/HKY+I+Γ/ GTR+I+Γ 480/45 0.3988 (variable) matKPos1 GTR+I+Γ GTR+Γ 528/297 0.5389 (all) matKPos2 GTR+I+Γ GTR+I+Γ/GTR+Γ 528/250 0.9208 (all) 1. matKPos3 GTR+Γ GTR+Γ 528/369 0.7387 (all) nu LSU GTR+I+Γ GTR+I+Γ 3391/808 1.0000 (all) §Models selected by the hLRT can vary depending on the model parameter addition hierarchy used; models used for analysis were those chosen by the AICc and at least one version of the hLRT \[for matKpos 1, AICc was used\]. The listed numbers of variable characters in the \"both globaland Saxifragales\" portion of the table refer to the globaldata set. Akaike weights of the chosen model were computer for the total number of characters (all) and the number of variable characters (variable). \* Both the globaland Saxifragalesdata sets included atpBcodon positions 1 and 2, but the models chosen for these data partitions differed between the two data sets (simpler models were chosen for these partitions with the Saxifragalesdata set, which comprises more closely related taxa than does the globaldata set). ::: List of abbreviations ===================== AICc-- Akaike Information criterion atpB-- ATP synthase beta subunit BI -- Bayesian inference BS -- bootstrap GTR -- general time reversible model (a model of DNA sequence evolution) HIV -- human immunodeficiency virus I + Γ -- invariant sites plus gamma distribution matK-- chloroplast maturase K matR-- mitochondrial maturase R MCMC -- Markov Chain Monte Carlo (a simulation method used to approximate the posterior probabilities of trees) ML -- maximum likelihood MP -- maximum parsimony rbcL-- ribulose bisphosphate carboxylase/oxygenase, large subunit nu SSU -- nuclear small-subunit ribosomal DNA nu LSU -- nuclear large-subunit ribosomal DNA TBR -- tree bisection-reconnection branch swapping Authors\' contributions ======================= DLN coordinated all aspects of the study, obtained directly or through colleagues all genomic DNAs used herein, conducted sequence aligments, and drafted the manuscript. JPD obtained all of the newly generated sequences. FEA performed all phylogenetic analyses. All of the authors read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Taxon sampling -- global data set.MS Excel file giving taxon sampling by gene and GenBank accession numbers for the *globaldata set. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Taxon sampling -- Saxifragales data set.MS Excel file giving taxon sampling by gene and GenBank accession numbers for the Saxifragalesdata set. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 3 BI tree from global data set.Bayesian inference majority rule consensus tree of 50,000 trees derived from the globaldata set composed of nuclear SSU rDNA, chloroplast rbcL, atpB, and mitochondrial matR. Trees were generated in two separate BI analyses, each run for 15 million generations with trees from the first 2.5 million generations removed as burn-in. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 4 BI tree from the nuclear SSU rDNA partition.Bayesian inference majority rule consensus tree of 50,000 trees derived from the nuclear SSU rDNA partition. Trees were generated in two separate BI analyses, each run for 15 million generations with trees from the first 2.5 million generations removed as burn-in. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 5 ML tree from the nuclear SSU rDNA partition.Phylogram from maximum likelihood (ML) analysis of the nuclear SSU rDNA partition (GTR+I+Γ model. The tree was generated using a successive approximations approach in which a MP tree was generated and used as a starting tree for additional branch swapping under ML. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 6 BI tree from the mitochondrial matR partition.Bayesian inference majority rule consensus tree of 50,000 trees derived from the mitochondrial matRpartition. Trees were generated in two separate BI analyses, each run for 15 million generations with trees from the first 2.5 million generations removed as burn-in. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 7 BI tree from the Saxifragales data set.Bayesian inference majority rule consensus tree of 50,000 trees derived from the Saxifragalesdata set. Trees were generated in two separate BI analyses, each run for 15 million generations with trees from the first 2.5 million generations removed as burn-in. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 8 MP phylogram from the global data set.Data set of combined nuclear SSU rDNA, chloroplast rbcLand atpBand mitochondrial matR. Tree one of four is shown, with branch lengths drawn proportional to the number of changes. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 9 BI majority rule phylogram from the global data set.Bayesian inference majority rule consensus phylogram from the globaldata set (nuclear SSU rDNA, chloroplast rbcLand atpBand mitochondrial matR). Branch lengths are means of the branch length posterior probability distribution across all post burn-in trees from one of the two MrBayes runs (branch lengths resulting from the other run are nearly identical). ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ The authors wish to thank all who assisted this project by collecting plant samples or assisting in field work: A. Aparicio and I. Sanchez García (Cynomorium coccineum), C. Ecroyd (Dactylanthus taylori), W. Forstreuter (Benthamina alyxifolia), J.-M. Groult (Hachettea austrocaledonica), R. Narayana (Santalum album), K. Steiner and A. Wolfe (Mystropetalon thomii*). Data analysis was greatly facilitated by David Swofford and Peter Foster who generously allowed access to computer clusters at Florida State University and The Natural History Museum (London), respectively. J. Bozzola and S. Schmitt at the SIUC IMAGE facility assisted in SEM photography. Financial support was provided by the National Science Foundation to DLN. The helpful comments of S. Stefanovic improved the final manuscript.	PubMed Central	2024-06-05T03:55:59.950829	2005-6-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182362/", "journal": "BMC Evol Biol. 2005 Jun 21; 5:38", "authors": [ { "first": "Daniel L", "last": "Nickrent" }, { "first": "Joshua P", "last": "Der" }, { "first": "Frank E", "last": "Anderson" } ] }
PMC1182363	Background ========== The vestibular system is the sensory mechanism of the inner ear (labyrinth) that helps the body maintain its postural equilibrium. There are two distinct sets of end organs in the labyrinth: the utricle and saccule within the vestibule, which respond to linear accelerations and changes in the position of the head with respect to gravity; and the semicircular canals, which respond to rotational movements (angular acceleration). The information that these organs deliver is proprioceptive in nature. The left and right utricular sensory epithelia (maculae) are in the same, approximately horizontal plane and because of this position they appear to be the dominant partner and are more useful than the saccular maculae in providing information about the position of the head and its side-to-side tilts when a person is in an upright position. The maculae are stimulated by shearing forces between the otolithic membrane and the cilia of the hair cells beneath it. However, the measurement or quantification of this \'otolith function\' in patients is very difficult. Many methods have been proposed to try and evaluate otolith function: ocular counter-rolling (OCR) induced by lateroflexion, whole body roll, eccentric rotation and translational acceleration have all been explored and promoted as indicators of vestibular otolith function \[[@B1]\]. However, these methods showed poor sensitivity and specificity, thereby preventing a sound clinical application \[[@B1]\]. Lateroflexion or body roll is a simple physiological test that changes the orientation of the head and the otolith system in space and measures the responses, such as eye movements in response to a counter roll. The subject tilts their head to one side and, as a consequence the eyes counter roll to a certain extent. Unfortunately, this simple test is associated with very low sensitivity and specificity and there is a large overlap between patients and healthy subjects. For example, an extensive study in healthy subjects can reveal a wide range of ocular counter rolling, from 3 to 11 degrees, induced by this simple lateroflexion test \[[@B2]\] It is also possible to measure responses to a linear acceleration (translation), which is also one of the specific stimuli for the otolith system. A linear sled device can change a subject\'s position in space, is motor driven and can move very fast (up to 1.2 G). However, if the sled is moving very fast and thus causing substantial motion, it is important to reduce the movements of the head using a mask specifically designed for each subject. In addition, the sled involves complex and advanced technology, and so is very expensive, and again there is limited sensitivity and a large overlap between patients and healthy subjects. Responses can also be elicited by eccentric rotation in a \'human centrifuge\' that can rotate up to 7 cycles/second and induce up to 6G \[[@B3]\]. Ocular counter rolling can be measured in this way, at constant rotation velocities with the amplitude of the response depending on the centrifugal force acting upon both labyrinths. When combining a centrifuge with a motor driven linear sled, it is also possible to test each of the labyrinths separately by rotation around one labyrinth in order to centrifuge the other labyrinth alone. Although this centrifuge technology has been used for many years, it is also associated with many problems and low sensitivities. The equipment is expensive, the eye movement responses are very small and correct position of the labyrinths difficult and responsible for false positive outcomes. The aim of this study was therefore to investigate whether the thresholds for the perception of linear acceleration might allow for a better measure for the clinical evaluation of the otolith function than measurement of eye movements. When auditory and visual cues are excluded and body movement is minimised, the detection of dynamic motion stimuli of small intensity appears to be primarily dependent on the otolith and the somatosensory systems response to pressure changes on the body surface \[[@B4]\]. Previous studies have shown that when oscillatory stimuli of 0.3 -- 0.4 Hz are employed, the thresholds for detection of linear movements in the horizontal plane range from 1.8 -- 6.3 cm/s^2^for anterior-posterior (AP) accelerations and 1.9 -- 5.7 cm/s^2^for lateral accelerations \[[@B5]\]. However, the literature shows that the acceleration thresholds vary with the stimulus profile used to determine the thresholds (sinus, parabolic, linear, steps), but that thresholds expressed in terms of velocity are more constant and less variable with the stimulus profile \[[@B6]-[@B9]\]. For example, Gianna et al observed mean normal thresholds of 4.84 cm/s2 using acceleration steps, 12.1 cm/s2 for linear ramps and 16.7 cm/s2 for parabolic stimuli \[[@B8],[@B9]\]. Expressed in terms of velocity all thresholds were close to 20 cm/s. A practical problem of these stimulus profiles is that they require a long sled and that after each stimulus a deceleration period and adaptation period is required, which makes the test procedure long lasting. We therefore investigated the threshold for perception of the direction of linear horizontal motion using a raised cosine bell profile. Methods ======= Whole body motion stimulus was generated by a motor driven linear sled running on a horizontal track of 4.2 metres (maximum velocity 3.7 m/s; maximum acceleration 1.2 m/s^2^adjustable in steps of 1 cm/s^2^). The seat could be changed into one of two positions in which the AP or the traverse axis of the head was parallel to the direction of motion. Twenty-eight healthy individuals with no previous complaints of dizziness and no history of audio-vestibular disease volunteered to participate in the study; 15 males and 13 females (22--60 years; seven subjects/decade). The subjects were seated upright with their feet on a footrest; head fixed against a headrest and the body restrained with safety belts. To eliminate external visual cues subjects were tested with eyes closed and in complete darkness; to eliminate auditory cues, subjects wore headphones; to mask proprioceptive cues, the sled was vibrated continuously by adding a sinusoidal signal to sled motor control profile (70 Hz sinus, 0.1 cm/s^2^peak amplitude). Upon request, all subjects indicated that this continuous vibration prevented them from using the motor vibrations as a cue for detection of sled motion per se. The stimulus was a simple raised cosinus bell cycle of 0.1 Hz (5 periods maximum per test), with peak accelerations of 0 -- 20 cm/s^2^. The impact of wind on movement perception is very difficult to prevent. However, speeds and accelerations were low, and the continuous cosinusoidal movement so smooth, that in a previous pilot study none of the subjects upon request indicated that they could make use of wind as a movement direction cue. Preliminary data from patients with complete congenital bilateral vestibular areflexia (no responses to caloric irrigations and rotations, no Vestibular Evoked Myogenic Potentials, no galvanic induced body sway) showed velocity thresholds that exceeded the normal values observed in this study (\>40 cm/s). This was considered to be a strong indication that proprioceptive cues played a minor role in perception of the sled movement in the experimental setup (results to be published elsewhere). Thresholds for the perception of motion were obtained under two conditions: the seat into a position in which the AP axis of the head was parallel to the direction of motion and with the seat into a position in which the traverse axis of the head was parallel to the direction of motion. Subjects were required to correctly indicate whether they were moving forward or backward in an AP motion, left or right in lateral motion, or were stationary for all five outward and return cycles. Sled movement was initially set with a maximum acceleration of 5 cm/s^2^and the subjects were asked to indicate whether they could detect motion. Patient were not pre-informed about the possible sled profile. If they were unable to detect the direction of the sled correctly, the maximum acceleration was increased by 1 cm/s^2^until perception was correct. If detection was correct, acceleration was decreased by 1 cm/s^2^until detection failed. This procedure was repeated until six thresholds were detected (three upper levels and three lower levels) (Figure [1](#F1){ref-type="fig"}). In Figure [1](#F1){ref-type="fig"}, the large filled squares indicate when motion was detected, after which time acceleration was decreased by 1 cm/s^2^until the subject could not accurately determine motion or the direction of motion. To rule out the possibility that differences in thresholds in the two directions might be attributable to training or to loss of concentration, back and forth movement was randomly performed as the first or last experiment. The subjects participated in the study after signing their informed consents approved by the institutional Ethics Committee. All studies were conducted in accordance with the Declaration of Helsinki. Results ======= The Shapiro-Wilk test shows that both the thresholds in the AP motion and those in lateral motion are not normally distributed. Table [1](#T1){ref-type="table"} shows an overview of the acceleration and velocity threshold data. The AP and lateral thresholds were not significantly different (signed rank test, p \< 0.192) and were correlated (Pearson; r = 0.561; p \< 0.01) and the reproducibility was good (within 17%). No significant correlation was found between thresholds and sex. Figures [2](#F2){ref-type="fig"} and [3](#F3){ref-type="fig"} show the thresholds as a function of age. The AP thresholds increased with age (median velocity threshold = -2.8 + 0.42 × age (p \< 0.001, r = 0.775), but the threshold for lateral movement was not correlated with age (p = 0.215, r = 0.242). Discussion ========== The thresholds for perception of the direction of linear acceleration using an oscillation (cosine) observed in this study (median 6.5 -- 8.5 cm/s^2^) were higher than those previously reported by Gundry \[[@B4]\]: 1.8 -- 6.3 cm/s^2^for AP accelerations and 1.9 -- 5.7 cm/s^2^for lateral accelerations. In addition, Gundry\'s results were obtained using oscillatory stimuli, with different techniques and equipment, and at frequencies of 0.3 -- 0.4 Hz compared to 0.1 Hz used in our study \[[@B5]\]. These findings are also in agreement with those of Benson et al., who suggested that the reason thresholds for the detection of discrete movements in the AP direction decreased with stimulus frequency might be due to an increased contribution of propriocepsis to detection of movement at high frequencies \[[@B7]\]. Furthermore, in the studies performed by Gundry, only the threshold for perception of movement was evaluated and not that of movement direction \[[@B5]\]. In the current study, we detected a more specific threshold, in this case the threshold for detection of motion direction. A major factor, however, is that acceleration thresholds reported in the literature vary widely with the stimulus profile used. It was suggested by Gianna et al. that velocity thresholds depend less upon the profile applied \[[@B8],[@B9]\]. In the present study, median velocity thresholds ranging from 3 to 36.6 cm/s were measured. Mean thresholds were 13.9 cm/s for AP and 10.4 cm/s for lateral movements. In the literature, the mean velocity thresholds for lateralmovements reported are all close to 20--22 cm/s (range 5 -- 50 cm/s), which is substantially larger than the mean values observed in our study (AP: 13.5 cm/s and lateral: 12.2 cm/s; Table [1](#T1){ref-type="table"}). The procedure (scoring criteria) followed by Gianna et al. \[[@B8],[@B9]\] and Benson et al. \[[@B7]\] to determine velocity thresholds were different from the method used here and could account for differences. The oscillatory stimulus by its repetitive and predictive character makes it relatively easy to lower the threshold by learning already within one test \[[@B10]\]. There is also an age dependency that to our knowledge has not been reported previously in the literature: especially in elder people the velocity threshold for AP movements increases. The correlation with age is in agreement with the findings of Igarashi et al., who found that statoconia volume in elderly people was significantly less than that in young children \[[@B11]\]. It is not evident why this age dependency holds for AP and not for lateral movements. A fundamental aspect of importance might be that subjects in general are more habituated to low frequency long-lasting AP linear movements (trains, cars) than to lateral movements. Nevertheless, the thresholds for the two movement directions are not significant different and we do not see how habituation would affect age dependency. It is well known that the vestibular system detects accelerations but controls eye velocity. When bode plots are made of the VOR it is obvious that within the normal sensitivity frequency range of the canals (roughly 0.1 -- 10 Hz) the vestibular system detects head velocity and controls eye velocity (zero phase shift and constant gain). The observation that velocity threshold reflects vestibular sensitivity more robustly than acceleration thresholds complies with the assumption that the otolith system also detects primarily head velocity (via integration of the acceleration) and that motion perception is related to velocity, providing that sufficient acceleration occurs to stimulate the accelerometer (hair cells) in the labyrinth. Conclusion ========== The perception thresholds for linear acceleration might reflect otolith function better than the rudimentary, non-functional ocular counter rolling reflex but, similar to the OCR, a large range of normal values was observed. Our recent experiments indicate that this variability is primarily associated with the fact that many subjects are unfamiliar with the perception of minor linear accelerations. After a training session, variability seems to decrease; perception thresholds might then reflect more the sensitivity of the otolith system per se. Competing interests =================== The author(s) declare that they have no competing interests. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1472-6815/5/5/prepub> Acknowledgements ================ The author wants to thank Solvay Pharmaceuticals who financed publication of this research in this journal by their institutional membership of BioMed Central and especially Dr. Claudio Sandner who assisted the author to establish publication of this research in an open access journal. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Schedule for threshold detection. ::: ![](1472-6815-5-5-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Velocity threshold of anterior-posterior linear movement as a function of age. Thresholds increase with age (median velocity threshold = -2.8 + 0.42 × age (p \< 0.001, r = 0.775). ::: ![](1472-6815-5-5-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Velocity threshold of lateral linear movement as a function of age. The threshold for lateral movement is not correlated with age (p = 0.215, r = 0.242). ::: ![](1472-6815-5-5-3) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Threshold perception for AP and lateral accelerations ::: Median Threshold Mean Range SD --------------------------------- ---------------------- ---------- ------------- -------- AP accelerations (cm/s^2^) 8.5 8.8 3 -- 16 4.1 AP velocities (cm/s) 13.9 13.5 4.8 -- 25.5 6.5 Lateral accelerations (cm/s^2^) 6.5 7.7 3 -- 23 4.5 Lateral velocities (cm/s) 10.4 12.2 4.8 -- 36.6 7.1 :::	PubMed Central	2024-06-05T03:55:59.953998	2005-6-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182363/", "journal": "BMC Ear Nose Throat Disord. 2005 Jun 22; 5:5", "authors": [ { "first": "H", "last": "Kingma" } ] }
PMC1182364	Background ========== Benzodiazepines (BZDs) are approved for treating clinically significant anxiety and insomnia. They have a better safety profile in comparison to barbiturates -- their predecessors, and many physicians have overestimated their safety. It is now widely accepted that BZD prescribing has many risks, including tolerance, dependence and misuse, as well as BZD-induced depression, cognitive impairment, disinhibition and psychomotor impairment \[[@B1]\]. BZD has become more restricted recently. The UK data sheets for diazepam and temazepam recommend that BZD should be used for short-term (2 -- 4 weeks) management. They are unlikely to be efficacious in the treatment of anxiety after 4 months \[[@B2]\]. BZD addiction can occur when doses within the clinical range are taken regularly over about 6 months \[[@B3]\]. In addition, a recent review suggests that, even in general anxiety disorder -- a major indication for BZDs -- this treatment may do more harm than good \[[@B4]\]. BZD use has been studied in high income countries, and it appears that prescribing levels have come down as the risks of addiction and adverse effects have become more widely publicised \[[@B5]\]. However, little research has been carried out in developing countries. Limited evidence suggests that BZD use is common in these countries. A recent community survey in Lebanon found that 9.6% (N = 1000) of the population had taken BZD in the previous month \[[@B6]\]. This study also found that approximately 30% of BZD users took these medications for more than 12 months. Although many developing countries allow BZDs to be sold over-the-counter in pharmacies, a literature review of BZD use in Brazil found that the vast majority of BZD consumption was due to medical prescription \[[@B7]\]. While psychiatrists are specialised in caring for people with mental health and substance use problems, their numbers are very limited, especially in developing countries. Therefore, GPs play an important role in prescribing BZDs. For example, a survey in Chile found that 69% of those taking BZDs received the medications from community clinics \[[@B8]\]. A cross-sectional study of prescriptions in 3,368 patients visiting a primary health unit in Brazil also showed that 20.6% of prescriptions included BZDs \[[@B9]\]. A survey of community hospitals in a rural area of Thailand demonstrated that 15% of all adult outpatients received benzodiazepines \[[@B10]\]. These results seem to agree with a previous questionnaire survey among Thai GPs, which found that approximately 50% of them prescribed BZDs for more than 25% of their patients \[[@B11]\]. As there is a trend to increase the provision of mental health care at the primary care level \[[@B12]\], GPs may prescribe more BZDs in the future. A survey of BZD prescribing practice and attitudes would increase understanding of BZD prescribing problems, important in developing a strategy to reduce BZD prescribing. We therefore proposed to evaluate the prescribing behaviour and attitudes of GPs practicing in Northern Thailand. Two major foci of the survey were the indications for and appropriate duration of BZD treatment. Because most GPs in developing countries are overloaded by a large number of patients in their everyday practice, patient medical records, e.g. diagnosis, are usually not completed or inaccurate. We therefore decided to conduct a survey using mailed questionnaires. This survey concerned only the GPs\' practice at public settings, where pharmacists take a role in dispensing the prescribed drugs, and GPs practice as non-dispensing physicians. This decision was made because most Thai physicians have their own private practice and are allowed to dispense almost all prescribed drugs, including BZDs, in their private settings. The results of a previous study have shown that dispensing and non-dispensing physicians have different patterns of prescribing \[[@B13]\]. Methods ======= Study population ---------------- Questionnaire survey to all 100 general practitioners working in community hospitals located in Chiang Mai and Lampoon, Northern Thailand. Questionnaire ------------- The questionnaire for this study included four parts: i\) GP\'s were asked to say whether they would prescribe BZDs and/or brief (3--5 minutes) supportive psychotherapy/advice for five case vignettes of anxiety/insomnia following a stressful life event, panic disorder, depression, essential hypertension and uncomplicated low back pain (see Annex). Brief supportive psychotherapy/advice could be given by the GPs or other medical professionals in the community hospitals. ii\) GPs were asked whether they agreed with the use of BZDs for clinically significant insomnia, anxiety and depression, as well as non-psychiatric illnesses, by using a 10 cm line of visual analogue scale (VAS) ranging from totally disagree to totally agree. A VAS score of 0.00--1.9, 2.0--3.9, 4.0--6.0, 6.1--8.0, and 8.1--10.0 were classified as totally disagree, disagree, neutral, agree and totally agree, respectively. iii\) the same VAS and its scoring system were used to ask the GPs whether they agreed with the regular use of BZDs for less than 1 week, 1 week to 1 month, more than 1 month to 4 months, more than 4 months to 6 months and more than 6 months. iv\) finally, GPs were asked if they thought they over-prescribed BZDs and if so why (more than one reason could be chosen). Procedures ---------- The GPs were informed that the survey did not intend to assess their knowledge but wished to understand their practice. The answers should be based on their everyday practice in public settings only (mainly community hospitals). The GPs were also asked to return the questionnaires with no answer if: i) they did not wish to respond or ii) they saw less than five adult out-patients per week. To maximize responses, 6 weeks after the first mailing, we sent the questionnaire to all GPs who had not replied. Six weeks after the second mailing, we sent the questionnaire for the third (and last) time to GPs who had still not responded. The survey was carried out between July and November 2003. Statistical analysis -------------------- Descriptive statistics (percentages, means and standard deviations) were calculated. Results ======= Fifty-eight of 100 GPs (58%) returned the questionnaires to us. Three were excluded; two GPs refused to answer the questionnaire, and another one saw less than five adult patients per week. Questionnaires from 55 GPs (32 males and 23 females) were subsequently analyzed. The GPs mean (SD) age in years was 31.6 (7.1), with a mean of 6.7 (5.8) years in practice and working for an average of 38.7 (18.2) hours per week. The mean number of patients seen per week (SD) was 392.8 (243.9). Of 13 GPs who had special training, four were paediatricians, three were internists, two each were surgeons and orthopaedists, and there was one family practitioner and one obstetrician/gynaecologist. Table [1](#T1){ref-type="table"} shows the number of GPs giving BZDs and brief supportive psychotherapy/advice for the 5 case vignettes. The majority of GPs would give BZDs for anxiety/insomnia following a stressful life event and panic disorder. The numbers of GPs giving brief supportive psychotherapy/advice for anxiety/insomnia and panic disorder were lower (22% and 15%, respectively). While 47% of the GPs would give BZDs for depression, 27% and 18% would give BZDs for essential hypertension and uncomplicated lower back pain, respectively. Of 26 GPs (47% of the GPs) who gave BZDs for depression, 16 of them would administer antidepressants concurrently. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Number of GPs giving BZDs and brief supportive psychotherapy or medical advices for the 5 case vignettes (N = 55) ::: Clinical conditions No. of GPs (%) giving BZDs No. of GPs (%) giving brief supportive psychotherapy or advice --------------------------------------------------- ---------------------------- ---------------------------------------------------------------- 1\. Stressful life event and anxiety and insomnia 51 (92.7) 39 (70.9) 2\. Panic disorder 43 (78.2) 35 (63.6) 3\. Depression 26 (47.2) 29 (52.7) 4\. Essential hypertension 15 (27.3) 46 (83.6) 5\. Uncomplicated low back pain 10 (18.2) 38 (69.1) ::: Agreement on the indications and durations of BZD use are shown in Table [2](#T2){ref-type="table"}. In respect to indications, 75%, 62% and 29% of the GPs agreed or totally agreed with the use of BZD for insomnia, anxiety, and depression, respectively. Twenty-eight GPs specified the use of BZDs for nonpsychiatric illnesses, especially the use of BZDs as muscle relaxants in 17 GPs. Eighty percent, 47%, 16%, 5% and 2% of the GPs agreed or totally agreed with the regular treatment of BZD for less than 1 week, 1 week -- 1 month, more than 1 month -- 4 month, more than 4 months -- 6 months and more than 6 months, respectively. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Agreement on the use of BZD assessed by using a 10 cm line of visual analogue scale from totally disagree to totally agree\* ::: Issues Totally disagree, n (%) Disagree, n (%) Neutral, n (%) Agree, n (%) Totally agree, n (%) ------------------------------------------------------- ------------------------- ----------------- ---------------- -------------- ---------------------- 1\. For clinically significant insomnia 2 (3.6) 1 (1.8) 11 (20.0) 14 (25.5) 27 (49.1) 2\. For clinically significant anxiety 7 (12.7) 4 (7.3) 10 (18.2) 15 (27.3) 19 (34.5) 3\. For clinically significant depression 16 (29.1) 11 (20.0) 12 (21.8) 5 (9.1) 11 (20.0) 4\. For regular use of less than 1 week 2 (3.6) 1 (1.8) 8 (14.5) 13 (23.6) 31 (56.4) 5\. For regular use of 1 week -- 1 month 15 (27.3) 2 (3.6) 12 (21.8) 15 (27.3) 11 (20.0) 6\. For regular use of more than 1 month -- 4 months 28 (50.9) 10 (18.2) 8 (14.5) 5 (9.1) 4 (7.3) 7\. For regular use of more than 4 months -- 6 months 39 (70.9) 9 (16.4) 4 (7.3) 1 (1.8) 2 (3.6) 8\. For regular use of more than 6 months 47 (85.5) 4 (7.3) 3 (5.5) 0 (0.0) 1 (1.8) \*A VAS score of 0.00--1.9, 2.0--3.9, 4.0--6.0, 6.1--8.0, and 8.1--10.0 were classified as totally disagree, disagree, neutral, agree, and totally agree ::: Twenty-five GPs (45.5%) accepted that they excessively used BZDs in the past year. The reasons for the over-prescribing were lack of time (17 responses), lack of knowledge and skills (14 responses), intention to keep doctor-patient relationship (i.e., patient demand -13 responses), lack of alternative treatment to BZDs (12 responses) and saving costs (10 responses). Discussion ========== This survey was conducted in young Thai GPs practicing in community hospitals. One of the surprising findings was these doctors reported that they would use BZDs for essential hypertension and uncomplicated low back pain, as well as the use as muscle relaxants. Almost half of the GPs agree that they over-prescribe BZDs. BZD prescribing for essential hypertension and low back pain is relatively common in developing countries \[[@B10]\]. Although there is some evidence supporting the benefits of BZDs for these conditions \[[@B14]-[@B16]\], the administration of these drugs may be detrimental \[[@B17]\]. This practice should be obsolete as a number of inexpensive drugs with preferred risk/benefit profiles are widely available, e.g., propanolol, orphenadrine citrate. The results of this and previous studies \[[@B18]\] demonstrate that BZD prescribing is a dilemma for GPs. Many of them realize the harmful effects of BZDs, but cannot control their prescribing. Improvement of knowledge and skills alone may not solve the problem. GPs in this survey have to see approximately 10 patients per hour, and this time pressure should be taken into account in developing any strategy to solve the problem. The findings of this survey are helpful in developing a strategy to reduce BZD use, especially among GPs practicing in developing countries. Because almost half of the GPs have already identified BZD prescribing as a problem, a simple and practical strategy to reduce prescribing would be welcomed. As lack of knowledge and skills contributes to the problem, an educational programme should be a part of the strategy. Firm evidence showing that fluoxetine can be used to treat patients with anxiety and/or depression safely and cost-effectively in primary care settings of low-income countries \[[@B19]\] should be presented to the GPs. GPs perceived causes of BZD over-prescribing, e.g., lack of knowledge and skills, lack of alternative treatment to BZDs and saving cost of treatment, could be solved. Improving GPs\' communication and training other health professionals, e.g., nurses, to provide brief supportive psychotherapy/advice may be also helpful, especially, in maintaining health professional-patient relationship. In addition, this can be used as an alternative or an adjunct to BZD treatment. Because sufficient consultation time is a key for quality patient care, it should be kept in mind that the impact of any strategy may be reduced if this problem cannot be mitigated. Some limitations should be considered in interpreting the study findings. First, a questionnaire survey may not reflect the \'real world\' practice. As most physicians have realized the detrimental effects of BZDs, the answers of many respondents may be based on their knowledge, or what they perceive to be \'best practice\', but not their actual practice. Due to this limitation, we focus our interpretation and discussion only on outstanding findings, e.g., the use of BZDs for physical illnesses, the acceptance of BZD over-prescribing. Second, the moderate response rate (58%) of this survey may have some impact on the validity of this study. Third, the results of present study may not be widely applicable because the GPs in Thailand may have different backgrounds from those in other parts of the world, e.g., culture, education, health care systems. However, the findings may provide insight for further studies elsewhere particularly developing countries. Last, most (16 of 26) GPs who gave BZDs for treating depression would also administer antidepressants concurrently. BZDs and anti-depressants should not be co-prescribed for longer than four weeks \[[@B20]\]. As the survey did not assess duration of BZD co-prescribing in depressed patients, we cannot determine the appropriateness of these co-prescriptions. Conclusion ========== This survey found that a considerable proportion of GPs in Chiang Mai and Lampoon, Thailand inappropriately use BZDs for physical illnesses, especially essential hypertension and uncomplicated low back pain. However, many GPs are aware that they over-prescribe BZDs. The problems of lack of time, knowledge and skills should be taken into account in improving the prescribing behaviour and attitude. List of abbreviations ===================== BZD = benzodiazepine GP = general practitioner VAS = visual analog scale Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= MS conceived and initiated the study, conducted the survey and analyzed the data. PG and JC conceived and initiated the study. NW conducted the survey. All authors participated in the writing of successive drafts of the manuscript and all have read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2296/6/27/prepub> Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Five case vignettes used in the survey. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ This study was supported by a grant from Effective Health Care Programme Alliance, International Health Research Group, Liverpool School of Tropical Medicine, Liverpool, U.K. The authors would like to thank the 58 general practitioners who responded to our mailed questionnaires.	PubMed Central	2024-06-05T03:55:59.955781	2005-6-23	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182364/", "journal": "BMC Fam Pract. 2005 Jun 23; 6:27", "authors": [ { "first": "Manit", "last": "Srisurapanont" }, { "first": "Paul", "last": "Garner" }, { "first": "Julia", "last": "Critchley" }, { "first": "Nahathai", "last": "Wongpakaran" } ] }
PMC1182365	Background ========== The α-proteobacteria comprise a large and extremely diverse group of Gram-negative bacteria which form a part of the largest known phyla within prokaryotes, namely the proteobacteria \[[@B1]\]. The vast diversity of the α-subdivision is clearly evident through the lifestyle differences among its members making them important in agricultural, medical and industrial fields. Such examples include the animal and human intracellular pathogens (Rickettsia, Bartonella, and Brucella) \[[@B1]-[@B3]\], the plant pathogens and symbiotic soil bacteria (Agrobacterium, Sinorhizobium, Mesorhizobium, and Bradyrhizobium) \[[@B1],[@B4]-[@B6]\], the Drosophilaendosymbiont (Wolbachia) \[[@B1]\] and a number of other free-living bacteria occupying a wide variety of ecological niches \[[@B1]\]. Furthermore, this group exhibits a wide spectrum of characteristics in terms of morphology (spiral, rod, stalked), metabolism (phototrophs, heterotrophs, and chemolithotrophs), physiology and cell division mechanisms \[[@B1],[@B7],[@B8]\]. In addition to their great diversity in these regards, this group of species is also of central importance due to compelling evidence indicating that a large proportion of the genes in eukaryotic cells, especially those related to mitochondria, have an α-proteobacterial ancestry \[[@B9]-[@B16]\]. In the current view, the α-subdivision are thought to form a more recently branching monophyletic taxon emerging after the epsilon and delta but before the beta and gamma subdivisions or Classes of proteobacteria \[[@B1],[@B13],[@B17]\]. Although this group is distinguished from other major bacterial groups based on 16S rRNA and other gene phylogenies \[[@B7],[@B13],[@B17]-[@B19]\], no set of criteria exists to clearly define and circumscribe the α-proteobacteria in clear and unambiguous molecular terms \[[@B1]\]. Thus, the following question remains: what defining molecular characteristics distinguish an α-proteobacterium and its subgroups from all other bacteria? The task of identifying such markers is aided by the availability of 18 completely sequenced α-proteobacterial genomes along with 10 partially sequenced genomes \[[@B11],[@B20]-[@B33]\], belonging to the following orders: Rhizobiales, Rickettsiales, Caulobacterales, Rhodobacterales, Sphingomonadalesand Rhodospirillales\[[@B34]\]. The comparative analyses of genomes provides a valuable resource and a very powerful means for identifying characteristics that are unique to a particular group of species \[[@B6],[@B16],[@B27],[@B28],[@B32],[@B35],[@B36]\]. We have used these data to identify a large number of conserved inserts and deletions (indels) in protein sequences that are distinctive characteristics of different groups of bacteria and provide molecular means for their identification and characterization \[[@B13],[@B37]-[@B40]\]. Recently, we have also identified many conserved indels in protein sequences that are useful for defining the α-proteobacteria group, and its various subgroups, in molecular terms \[[@B17]\]. The distribution pattern of these signatures in different α-proteobacteria has been used to deduce a working model to describe the interrelationships as well as the branching order among the α-proteobacteria species \[[@B17]\]. In the present study, a new type of taxonomic marker is described which provides an additional means to define the α-proteobacteria group as well as the relationship within this group. These new markers consist of whole proteins that are specific to certain groups or subgroups of bacteria and are not found in any other phyla \[[@B35]\]. In this work we have identified a large number of proteins which are specific to either the α-proteobacteria group as a whole or its various subgroups. These signature proteins were identified in BLASTP searches \[[@B41]\] of individual proteins from the genomes of three α-proteobacterial species (viz. Rickettsia prowazekii, Caulobacter crescentusand Bartonella quintana) \[[@B11],[@B24],[@B32]\], which show important differences in lifestyles and physiology. Results of this study presented here will prove useful in developing a clearer picture of α-proteobacterial phylogeny as well as aid in the identification of bacterial strains belonging to this group and its subgroups. Functional studies on these α-proteobacteria specific proteins should prove instrumental in the discovery of novel physiological characteristics that are uniquely shared by members of this large and diverse group of bacteria. Results ======= These studies were undertaken with the aim of identifying proteins that are uniquely found in α-proteobacteria and which could provide novel molecular means for defining and identifying bacteria belonging to this group and its subgroups. To identify proteins which are specific to α-proteobacteria or its subgroups, BLAST searches were carried out individually on every single annotated protein present in the genomes of three different α-proteobacteria, C. crescentus, R. prowazekiiand B. quintana. These genomes were chosen because of their different sizes (R. prowazekii, 1.11 Mb with 835 open reading frames (ORFs); B. qunitana, 1.58 Mb, 1142 ORFs; C. crescentus, 4.02 Mb with 3737 ORFs) and because these species display important differences in life-style and other characteristics \[[@B11],[@B24],[@B32]\]. Results of the BLAST searches were inspected in order to identify proteins which are only found in α-proteobacteria, as well as proteins where the only acceptable BLAST scores as indicated by their expected values (E values) were from α-proteobacteria \[[@B41]\]. These studies have resulted in the identification of 61 signature proteins, which appear distinctive of α-proteobacteria and are generally not found in any other Bacteria. For all of these proteins, the lengths of the query proteins as well as the E values obtained from BLAST searches for different hits are shown (Tables [1](#T1){ref-type="table"}, [2](#T2){ref-type="table"}, [3](#T3){ref-type="table"}, [4](#T4){ref-type="table"}, [5](#T5){ref-type="table"}, [6](#T6){ref-type="table"}, [7](#T7){ref-type="table"}, [8](#T8){ref-type="table"}, [9](#T9){ref-type="table"}). The former values are important in determining the significance of the observed BLAST scores (See Methods section). Additionally, for all of the α-proteobacteria specific proteins, the length of the hit protein over the query sequence is shown in brackets to show that the homologues in different species are of similar length. Most of the α-proteobacterial signature proteins that we have identified are of hypothetical function as annotated in the NCBI database <http://www.ncbi.nlm.nih.gov/genomes/MICROBES/Complete.html>. For the sake of presentation and discussion, we have arbitrarily divided these proteins into ten groups based on their distribution patterns among α-proteobacteria. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Signature Proteins Specific for the Alpha-proteobacteria.^a^ ::: Protein CC2102\[16126341\] CC3292\[16127522\] CC3319^b^\[16127549\] CC1887^c^\[16126130\] CC1725\[16125969\] CC1365\[16125614\] ------------------- ------------------------ --------------------------- --------------------------- --------------------------- ------------------------ --------------------------- Length 162 224 89 105 100 161 Mag. mag. 2e-19 (1.04) 9e-49 (1.29) 2e-15 (0.99) 3e-20 (1.10) 1e-08 (0.88) 2e-21 (0.87) Rhod. rubr. 3e-17 (1.09) 1e-42 (0.92) 1e-12 (1.04) 4e-21 (1.12) 1e-08 (1.09) 3e-17 (0.96) Nov. aro. 1e-19 (1.02) 3e-45 (1.04) 4e-07 (0.88) 2e-25 (1.30) \- 1e-11 (0.98) Z. mobilis \* 4e-24 (1.01) 6e-40 (1.00) \- \- \- 5e-11 (1.02) C. cres.\* 1e-64 (1.00) e-106 (1.00) 1e-34 (1.00) 3e-61 (1.00) 4e-49 (1.00) 7e-79 (1.00) Sil. pom.\* 3e-16 (0.93) 2e-44 (1.15) 8e-11 (0.97) 3e-22 (1.21) 8e-06 (0.99) 4e-12 (0.96) Sil. sp. 1e-13 (0.98) 1e-43 (1.14) 8e-09 (0.99) 5e-24 (1.22) 1e-05 (0.88) 9e-12 (0.89) Rh. spha. 1e-16 (0.94) 3e-43 (1.11) 4e-08 (0.94) 8e-24 (1.18) 2e-08 (1.00) 1e-11 (0.91) Bra. jap.\* 5e-21 (1.06) 8e-47 (1.11) 2e-12 (1.01) 2e-29 (1.29) 2e-09 (1.05) 2e-11 (1.15) Rho. pal.\* 8e-22 (1.10) 4e-45 (1.06) 7e-13 (1.45) 2e-25 (1.30) 2e-08 (1.05) 2e-12 (1.00) Agr. tum.\* 4e-21 (1.04) 7e-49 (1.02) 4e-10 (0.96) 3e-26 (1.26) 5e-06 (1.14) 7e-13 (1.10) Sino. meli. \* 4e-23 (1.06) 2e-49 (1.04) 6e-12 (1.01) 2e-22 (1.25) 5e-05 (0.93) 1e-14 (1.04) Bru. mel.\* 5e-19 (1.17) 4e-48 (1.02) 1e-12 (1.03) 1e-23 (1.30) 1e-07 (1.09) 4e-15 (1.04) Bru. suis\* 5e-19 (1.17) 4e-48 (1.02) 2e-12 (0.97) 1e-23 (1.30) 2e-08 (1.09) 4e-15 (1.07) Meso. loti\* 5e-18 (1.10) 9e-49 (1.00) 9e-18 (0.97) 9e-23 (1.26) 4e-09 (1.20) 3e-14 (1.40) Meso. sp. 2e-17 (1.09) 2e-47 (1.03) 5e-16 (0.98) 6e-27 (1.26) 8e-06 (1.05) 7e-15 (1.33) B. henselae\* 8e-14 (1.12) 5e-41 (1.04) 4e-12 (0.96) 2e-23 (1.26) 5e-08 (1.09) 4e-10 (0.99) B. Quintana\* 5e-14 (1.12) 8e-42 (1.02) 4e-12 (0.96) 5e-21 (1.26) 1e-08 (1.09) 3e-11 (0.99) R. conorii\* 5e-10 (0.97) 3e-41 (0.83) 1e-07 (0.88) 1e-05 (0.94) 5e-11 (1.01) 7e-07 (0.98) R. prowazekii\* 4e-08 (0.98) 4e-40 (0.83) 3e-07 (0.88) 5e-07 (0.95) 2e-12 (1.07) 8e-06 (0.95) R. typhi\* 1e-08 (0.98) 2e-40 (0.83) 3e-07 (0.88) 4e-07 (0.95) 5e-11 (1.07) 3e-06 (0.98) R. akari 6e-09 (0.97) 9e-41 (0.90) 1e-07 (0.88) 2e-05 (0.94) 7e-12 (1.07) 3e-07 (0.98) R. rickettsii 1e-10 (0.97) 9e-41 (0.83) 1e-07 (0.88) 1e-05 (0.94) 5e-11 (1.01) 7e-07 (0.98) R. sibirica 2e-10 (0.97) 5e-41 (0.83) 1e-07 (0.88) 1e-05 (0.94) 5e-11 (1.01) 3e-07 (0.98) Wolbachia\* 5e-11 (0.90) 3e-20 (0.89) 2e-07 (0.99) 1e-08 (0.99) 1e-05 (1.04) 3e-04 (0.70) Ana. mar.\* 2e-09 (0.98) 2e-40 (0.92) \- \- 9e-07 (1.19) \- Ehr. canis 2e-07 (1.01) 6e-39 (0.95) 2e-06 (1.04) 5e-10 (0.94) \- 0.023 (0.90) Ehr. rum.\* 1e-08 (0.94) 2e-39 (0.99) 2e-05 (1.07) 8e-11 (0.89) \- 4e-05 (0.84) Non-Alpha \- Strep. glau. 1.6 (1.89) Fuso. nucl. 2.9 (3.87) Burk. fung. (3.69) Azo. sp. 0.44 (0.91) Myc. pneum. 0.98 (2.63) ^a^Alpha-specific proteins were identified by BLAST searches on individual protein sequences on three α-proteobacterial genomes as described in the Methods section. The expected (E) values for various alpha-proteobacteria species as well as the first non-alpha species in the BLAST results are shown here. The values in brackets after the E values represent the ratios of the length of the hit protein divided by the query protein and a value close to 1.0 indicates that the homologues are of similar lengths. The CC numbers indicate the protein identification number in the C. crescentusgenome. GenBank accession numbers for the query sequence are shown in square brackets. An asterisk (\) identifies bacterial genomes which are completely sequenced, whereas other sequences are from partially or incompletely sequenced genomes. Proteins not found in a given species are indicated with a dash (-). Abbreviations: Agr. tum., Agrobacterium tumefaciens; Ana. mar., Anaplasma marginale; Azo. sp., Azoarcus sp. EbN1; B. henselae, Bartonella henselae; B. quintana, B. quintana; Bra. jap., Bradyrhizobium japonicum; Bru. mel., Brucella melitensis; Burk. fung., Burkholderia fungorum; C. cres., Caulobacter crescentus; Ehr. canis, Ehrlichia canis; Ehr. rum., Ehr ruminantium; Fuso. nucl., Fusobacterium nucleatum; Mag. mag., Magnetospirillum magnetotacticum; Meso., Mesorhizobium; Myc. pneum., Mycoplasma pneumoniae; Nov. aro., Novosphingobium aromaticivorans; Rhod. rubr., Rhodospirillum rubrum; Rh. spha., Rhodobacter sphaeroides; Rho. pal., Rhodopseudomonas palustris; Sil. pom., Silicibacter pomeroyi; Sil. sp., Silicibacter sp. TM1040; Sino. meli., Sinorhizobium meliloti; Strep. glau., Streptomyces glaucescens; Wolbachia, Wolbachia endosymbiont of Drosophila melanogaster. ^b^Magnetococcus sp. MC-1(unclassfied) found in BLAST search with E value of 3e-09 \[48832993\]. ^c^Protein also found in Eukaryotes. The E values for a few representative eukaryotic species are as follows: Homo sapiens; 8e-10 \[30583279\], Chlamydomonas reinhardtii; 3e-09 \[34334022\], Caenorhabditis elegans; 5e-06 \[7332202\]. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Signature Proteins specific for Alpha-proteobacteria, except Rickettsiales.^a^ ::: Protein* CC1211\[16125461\] CC1886\[16126129\] CC2245\[16126484\] CC3470\[16127700\] CC0520^b^\[16124775\] --------------- ---------------------------- ------------------------------ ----------------------------- ------------------------------ -------------------------------- Length 167 223 190 253 284 Mag. mag. 3e-19 (1.09) \- 3e-16 (0.96) 5e-23 (0.97) 2e-38 (0.89) Rhod. rubr. 3e-17 (0.73) 1e-07 (0.73) 3e-22 (0.64) 5e-04 (0.82) 2e-36 (0.93) Nov. aro. 9e-13 (1.41) 3e-05 (0.91) 7e-13 (1.24) 6e-20 (0.84) \- Z. mobilis 8e-12 (1.38) \- 5e-12 (1.21) 1e-15 (0.82) \- C. cres. 5e-93 (1.00) 3e-58 (1.00) 1e-75 (1.00) e-115 (1.00) e-146 (1.00) Sil. pom. 9e-21 (1.19) 2e-04 (0.53 7e-16 (1.04) 2e-10 (0.78) 5e-27 (0.88) Sil. sp. 8e-20 (1.17) 5e-07 (0.65) 1e-15 (1.03) 3e-07 (0.78) 1e-27 (0.87) Rh. spha. 1e-19 (1.24) 2e-06 (0.55) 2e-16 (1.09) 8e-07 (0.78) 2e-30 (0.92) Bra. jap. 4e-20 (1.01) 1e-10 (1.61) 1e-25 (0.89) 5e-17 (0.94) 2e-44 (0.87) Rho. pal. 1e-18 (1.28) 5e-10 (1.48) 3e-25 (1.12) 2e-17 (0.86) 5e-45 (0.91) Agr. tum. \- 2e-10 (0.65) \- 1e-15 (0.85) 8e-46 (1.01) Sino. meli. 1e-19 (1.04) 3e-12 (0.67) 4e-25 (0.91) 7e-16 (0.87) 1e-45 (0.89) Bru. mel. 4e-22 (1.11) 2e-10 (0.68) 1e-26 (0.97) 1e-18 (0.82) 2e-41 (1.08) Bru. suis 4e-22 (1.08) 2e-10 (0.70) 5e-26 (0.95) 8e-18 (0.83) 1e-41 (0.91) Meso. loti 4e-24 (1.02) 5e-09 (0.89) 1e-27 (0.90) 8e-21 (0.83) 1e-44 (0.92) Meso. sp. 1e-21 (1.07) 2e-10 (0.55) 5e-25 (0.94) 1e-18 (0.83) 1e-42 (0.91) B. henselae 6e-20 (1.06) 9e-14 (0.63) 9e-22 (0.93) 5e-15 (0.84) 1e-28 (0.86) B. quintana 2e-19 (1.06) 3e-12 (0.63) 3e-22 (0.93) 2e-11 (0.84) 2e-27 (0.89) Non Alpha Kin. radio. 0.097 (3.89) Vib. para. 7.7 (0.82) Pse. aeruginosa4.2 (1.88) Vib. cholerae0.035 (1.96) Polaromonas sp. 0.003 (0.95) Protein CC0365\[16124620\] CC0366^c^\[16124621\] CC1977^d^\[16126220\] CC3010^e^\[16127240\] CC0100\[16124355\] Length 169 177 241 216 576 Mag. mag. 2e-06 (1.05) 2e-06 (0.93) 1e-32 (0.98) 4e-24 (0.90) 7e-34 (0.92) Rhod. rubr. 2e-08 (1.08) 4e-06 (0.91) 2e-36 (1.07) 8e-18 (0.94) \- Nov. aro. 3e-07 (1.09) 3e-05 (0.93) 5e-29 (0.95) 3e-13 (0.91) 5e-26 (1.09) Z. mobilis \- \- 1e-35 (0.97) 6e-12 (0.96) 2e-35 (1.02) C. cres. 1e-44 (1.00) 1e-44 (1.00) e-111 (1.00) 6e-93 (1.00) 0.0 (1.00) Sil. pom. 4e-09 (1.12) 1e-07 (1.02) 2e-40 (0.99) 1e-19 (0.95) 2e-34 (0.99) Sil. sp. 4e-08 (1.09) 1e-11 (1.02) 6e-38 (0.97) 2e-18 (0.96) 1e-33 (1.00) Rh. spha. 1e-06 (1.09) 1e-08 (0.90) 6e-36 (0.98) 2e-20 (0.98) 7e-30 (0.98) Bra. jap. 8e-07 (0.95) 2e-12 (1.06) 1e-28 (1.10) 8e-20 (0.94) 3e-35 (1.02) Rho. pal. 9e-06 (0.96) 9e-10 (1.05) 2e-33 (1.08) 2e-19 (0.99) 5e-35 (1.09) Agr. tum. 1e-06 (0.95) 7e-10 (1.21) 2e-35 (1.10) 3e-20 (0.98) 8e-35 (0.98) Sino. meli. 2e-08 (0.95) 4e-10 (1.15) 3e-38 (1.08) 2e-20 (0.84) 2e-31 (0.94) Bru. mel. 0.002 (0.84) 1e-10 (1.02) 1e-32 (1.08) 4e-20 (0.97) 4e-36 (0.99) Bru. suis 2e-07 (0.94) 1e-10 (1.18) 8e-33 (1.08) 4e-20 (0.97) 5e-36 (0.94) Meso. loti 1e-08 (0.96) 8e-08 (1.09) \- 5e-22 (0.95) 2e-34 (0.92) Meso. sp. 7e-11 (0.94) 5e-11 (1.09) 1e-27 (1.09) 6e-23 (0.95) 1e-34 (0.99) B. henselae 1e-06 (0.97) 3e-08 (1.06) 1e-27 (1.08) 2e-20 (0.95) 2e-27 (0.99) B. quintana 5e-07 (0.97) 1e-07 (1.06) 2e-29 (1.08) 2e-20 (0.94) 4e-30 (0.99) Non-Alpha Myc. galli. 1.1 (1.17) Rhodo. baltica0.005 (1.47) M. thermo. 0.28 (1.03) Syn. elongatus0.002 (1.33) Coryn. efficiens0.11 (0.52) ^a^Abbreviations and other details regarding BLAST results can be found in Table 1. E values of 0.0 indicate an extremely high degree of similarity between protein sequences. Additional abbreviations: Coryn., Corynebacterium; Kin. radio., Kineococcus radiotolerans; M. thermo., Moorella thermoacetica; Myc. galli., Mycoplasma gallisepticum; Pse., Pseudomonas; Rhodo., Rhodopirellula; Syn., Synechococcus; Vib. para., Vibrio parahaemolyticus. ^b,c^Magnetococcus sp. MC-1(unclassified) contains homologues of both proteins with E values of 1e-18 \[48832519\] and 8e-06 \[48833234\] respectively. ^d^Protein also found in Eukaryoteswith examples from representative species as follows: Homo sapiens; 1e-22 \[21735485\], Oryza sativa; 2e-18 \[50939575\] and Cryptococcus neoformans; 3e-17 \[57225838\]. ^e^One BLAST hit is Pseudomonas sp. with E value of 8e-20 \[94976\]. ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Alpha-proteobacteria specific proteins which are absent in the Rickettsialesas well as (A) the Bartonellaceaefamily, or (B)the Rhodobacterales^a^ ::: Protein CC2345\[16126584\] CC3115\[16127345\] CC3401\[16127631\] CC3467\[16127697\] CC1021\[16125273\] --------------- -------------------------------- ----------------------------- ---------------------------- ----------------------------- ------------------------- Length 159 136 120 152 130 Mag. mag. 2e-38 (0.99) 2e-29 (0.90) 3e-14 (1.03) 7e-16 0.86 3e-14 (1.16) Rhod. rubr. 1e-39 (1.01) \- \- \- \- Nov. aro. 4e-35 (1.03) 8e-32 (1.08) 1e-09 (1.23) 1e-19 (1.10) 5e-06 (1.13) C. cres. 2e-86 (1.00) 9e-79 (1.00) 8e-56 (1.00) 3e-82 (1.00) 1e-57 (1.00) Sil. pom. 2e-38 (1.01) 1e-09 (0.96) 3e-09 (1.08) 5e-23 (0.99) 3e-14 (1.00) Sil. sp. 3e-40 (1.00) 5e-10 (1.05) 5e-10 (1.19) 5e-24 (1.34) 8e-16 (1.10) Rh. spha. 6e-40 (0.99) 1e-07 (1.00) 9e-07 (1.13) 1e-25 (1.02) 4e-10 (1.05) Bra. jap. 2e-44 (1.04) 4e-27 (0.98) 6e-11 (1.22) 5e-28 (1.07) 6e-17 (1.12) Rho. pal. 2e-45 (1.03) 8e-24 (0.96) 5e-13 (1.09) 5e-27 (1.05) 9e-15 (1.09) Agr. tum. 7e-44 (0.99) 6e-30 (0.94) 2e-15 (1.07) 3e-28 (1.06) 1e-10 (0.98) Sino. meli. 3e-44 (0.99) 1e-25 (0.83) 4e-14 (1.06) 7e-30 (1.14) 7e-13 (1.16) Bru. mel. 2e-43 (1.01) 1e-28 (1.12) 3e-15 (1.27) 9e-29 (1.08) 1e-10 (1.17) Bru. suis 2e-43 (1.01) 4e-29 (0.93) 2e-15 (1.06) 9e-29 (1.08) 1e-11 (1.18) Meso. loti 1e-43 (1.01) 1e-28 (0.98) 3e-14 (1.11) 2e-27 (1.12) 5e-14 (1.15) Meso. sp. 3e-43 (1.01) 1e-22 (0.79) 4e-10 (1.20) 5e-30 (1.09) 5e-13 (1.11) Non-Alpha Vib. para. 2.1 (1.04) Symbio. therm. 1.3 (0.89) Burk. cepacia0.96 (2.53) Noc. farcinica0.17 (1.76) Pse. syringae1.7 2.35 Protein CC1652\[16125898\] CC2247\[16126486\] CC3295\[16127525\] CC1035\[16125287\] Length 250 46 169 224 Mag. mag. 2e-07 (0.91) 2e-06 (1.61) 2e-09 (0.52) \- Rhod. rubr. 7e-08 (1.01) \- \- \- Nov. aro. \- 7e-05 (1.41) 5e-06 (0.99) 3e-33 (1.19) Z. mobilis \- 1e-04 (1.65) \- \- C. cres. 7e-99 (1.00) 5e-21 (1.00) 3e-92 (1.00) e-103 (1.00) Bra. jap. 5e-12 (0.87) 7e-07 (1.63) 6e-29 (1.00) 7e-43 (0.90) Rho. pal. 5e-12 (0.92) 5e-06 (2.41) 2e-27 (1.00) 3e-42 (0.97) Agr. tum. 4e-05 (1.00) 1e-04 (1.67) 3e-26 (1.01) 1e-39 (0.88) Sino. meli. 2e-09 (0.89) 7e-04 (1.67) 3e-24 (1.16) 2e-35 (0.89) Bru. mel. 2e-11 (0.89) 0.008 (1.70) 3e-28 (1.04) 4e-39 (0.88) Bru. suis 1e-11 (0.89) 0.008 (1.70) 2e-28 (0.99) 4e-39 (0.88) Meso. loti 2e-06 (0.88) 5e-05 (1.67) 9e-24 (1.05) 1e-38 (0.93) Meso. sp. 4e-11 (0.89) 2e-04 (1.63) 2e-25 (1.04) 2e-38 (0.80) B. henselae 1e-06 (0.89) 0.030 (1.70) 3e-18 (1.02) 1e-34 (0.88) B. quintana 2e-06 (0.89) 0.002 (1.70) 8e-18 (1.02) 2e-35 (0.88) Non-Alpha Meth. flagellatus0.17 (1.53) \- Burk. cepacia0.13 (2.89) Bdell. bacter. 0.70 3.14 ^a^The manner in which these alpha specific proteins were identified is as described in Table 1. Additional abbreviations: Bdell. bacter., Bdellovibrio bacteriovorus; Burk., Burkholderia; Meth., Methylobacillus; Noc., Nocardia; Pse., Pseudomonas; Symbio. therm., Symbiobacterium thermophilum; Vib. para., Vibrio parahaemolyticus. ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Signature Proteins Specific for the Rickettsialesor the Rickettsiaceaefamily^a^ ::: Protein RP104^b^\[15603981\] RP105\[15603982\] RP106^c^\[15603983\] RP766\[15604600\] RP192\[15604066\] RP030\[15603909\] RP187^d^\[15604061\] --------------------- -------------------------- ------------------------- --------------------------- ------------------------- ------------------------- --------------------------- -------------------------- length 1124 672 971 92 128 219 194 R. prowazekii 0.0 (1.00) 0.0 (1.00) 0.0 (1.00) 8e-37 (1.00) 2e-41 (1.00) e-118 (1.00) e-108 (1.00) R. conorii 0.0 (0.88) 0.0 (0.98) 0.0 (0.99) 3e-33 (1.18) 3e-30 (0.93) e-107 (1.01) e-100 (2.56) R. typhi 0.0 (1.01) 0.0 (1.00) 0.0 (1.00) 1e-27 (0.85) 3e-40 (1.00) e-116 (1.00) e-102 (2.56) R. akari 0.0 (0.90) 0.0 (1.00) 0.0 (1.01) 2e-25 (0.85) 2e-35 (1.02) e-108 (1.01) 3e-99 (2.56) R. rickettsii 0.0 (0.88) 0.0 (0.98) 0.0 (0.99) 4e-27 (0.85) 6e-32 (0.93) e-109 (1.00) e-99 (2.56) R. sibirica 0.0 (0.88) 0.0 (0.98) 0.0 (0.99) 4e-27 (0.85) 2e-31 (0.93) e-106 (1.01) e-101 (2.56) Ehr. canis 7e-22 (0.75) 4e-36 (1.25) 2e-26 (1.49) \- \- \- \- Ehr. rum. 2e-22 (0.73) 2e-37 (1.22) 3e-23 (1.57) \- \- \- \- Ehr. chaf. 5e-23 (0.73) 8e-36 (1.23) \- \- \- \- \- Wolbachia 1e-22 (0.78) 7e-27 (1.27) 3e-20 (0.82) \- \- \- \- Ana. mar. 2e-17 (0.78) 4e-25 (1.31) 2e-24 (1.05) \- \- \- \- Non-Rickettsiales Meso. loti0.004 (0.32) Sil. sp. 0.002 (0.53) Xyl. fast. 2e-07 (0.36) Leg. pneu. 1.3 (2.60) M. thermo. 8.1 (2.47) Myc. pulm. 0.001 (4.72) Camp. lari0.69 (4.09) ^a^Rickettsialeand Rickettsiaspecific proteins were identified by whole-genome BLAST searches using protein sequences as probes from the fully sequenced R. prowazekiigenome. The RP numbers refer to the protein identification number in the R. prowazekiigenome. Other details are as in Table 1 and in Methods section. Additional abbreviations: Camp., Campylobacter; Ehr. chaf., Ehrlichia chaffeensis; Leg. pneu., Legionella pneumophila; M. thermo., Moorella thermoacetica; Myc. pulm., Mycoplasma pulmonis; Xyl. fast., Xylella fastidiosa. ^b,c^BLAST hits for the family Anaplasmataceaedo not show homology over the entire range of the protein and may represent a conserved protein domain. ^d^BLAST hits for other Rickettsiastrains are longer (497 aa) but contain a region that is almost completely homologous to the query sequence. ::: ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Signature Proteins Specific for the Rizobialesorder^a^ ::: Protein BQ00140\[49473701\] BQ00720 \[49473755\] BQ03880\[49474026\] BQ12030\[49474691\] BQ07670\[49474353\] BQ11900\[49474679\] ------------------ ------------------------------ -------------------------- ------------------------------- --------------------------- ----------------------------- ---------------------------- Length 222 83 198 91 336 172 Bra. jap. 5e-18 (1.10) 2e-09 (1.08) 2e-20 (0.97) 2e-07 (1.05) 5e-46 (1.06) 1e-17 (0.98) Rho. pal. 3e-14 (1.11) 3e-09 (1.06) 2e-18 (0.97) 1e-07 (1.37) \- \- Agr. tum. 5e-13 (1.06) 6e-11 (1.02) 1e-26 (0.98) 2e-12 (1.03) 7e-66 (0.97) 2e-24 (1.37) Sino. meli. 3e-20 (1.00) 2e-13 (1.02) 7e-23 (0.98) 2e-14 (0.98) 5e-62 (1.02) \- Bru. mel. 9e-39 (0.98) 4e-19 (1.04) 1e-38 (0.97) 3e-13 (0.54) 8e-70 (0.96) 2e-26 (1.02) Bru. suis 4e-39 (0.98) 4e-19 (0.96) 1e-38 (1.03) 6e-20 (0.99) 8e-70 (0.98) 6e-27 (0.98) Meso. loti 3e-25 (1.07) 2e-13 (1.18) 1e-25 (0.98) 4e-16 (0.99) 2e-67 (0.95) 1e-26 (1.03) Meso. sp. 7e-18 (1.06) 2e-13 (1.02) 2e-25 (0.98) 2e-14 (1.02) 3e-61 (0.93) 7e-31 (1.02) B. henselae 1e-92 (0.97) 6e-43 (1.00) 5e-92 (1.00) 2e-39 (1.00) e-158 (1.01) 3e-81 (1.00) B. Quintana e-127 (1.00) 4e-43 (1.00) e-107 (1.00) 1e-43 (1.00) 0.0 (1.00) 1e-91 (1.00) Non -- Rizobiale Bdell. bacter. 0.25 (1.77) Sil. sp. 0.46 (0.82) Vibrio fischeri0.005 (2.49) St. pyogenes0.12 (0.86) St. agalactiae0.38 (2.65) Croc. watsonii6.1 (2.85) ^a^Signature proteins that are distinctive of the order Rhizobialeswere identified by carrying out BLAST searches of all proteins found in the genome of B. quintana. The BQ numbers refer to the protein identification number in the B. quintanagenome. Abbreviations and further details regarding BLAST results are as in Table 1 and the Methods section.. Additional abbreviations: Bdell. bacter., Bdellovibrio bacteriovorus; Croc., Crocosphaera; St., Streptococcus. ::: ::: {#T6 .table-wrap} Table 6 ::: {.caption} ###### Signature Proteins specific for the Rizobialesexcept the Bradyrhizobiaceaefamily^a^ ::: Protein BQ01660\[49473833\] BQ02450\[49473907\] BQ03770\[49474017\] BQ13470\[49474819\] ------------------ ------------------------------------- --------------------------------- ------------------------------------ ------------------------- Length 119 199 280 179 Bra. jap. \- \- \- \- Rho. pal. \- \- \- \- Agr. tum. 6e-15 (1.04) 3e-06 (1.07) 2e-07 (1.09) 4e-11 (1.01) Sino. meli. 1e-15 (1.03) 1e-05 (1.03) 1e-11 (1.06) 1e-10 (1.00) Bru. mel. 2e-23 (1.06) 8e-11 (1.01) 1e-17 (1.07) 4e-26 (0.99) Bru. suis 2e-23 (1.06) 1e-11 (1.12) 1e-17 (1.21) 4e-26 (0.99) Meso. loti 3e-12 (1.39) 1e-05 (1.25) 3e-13 (0.96) 3e-24 (1.00) Meso. sp. 2e-12 (1.08) 2e-09 (1.02) 6e-17 (0.95) 8e-09 (0.99) B. henselae 2e-55 (1.00) 1e-64 (0.99) 2e-91 (1.01) 2e-67 (0.99) B. Quintana 1e-64 (1.00) e-102 (1.00) e-131 (1.00) 2e-99 (1.00) Non -- Rizobiale Bacillus licheniformis0.77 (1.78) Treponema denticola2.9 (2.32) Therm. tengcongensis0.001 (2.79) Mag. mag. 0.60 (1.04) See table 1 legend for abbreviations and additional information pertaining to BLAST results. Additional abbreviations: Therm., Thermoanaerobacter. ::: ::: {#T7 .table-wrap} Table 7 ::: {.caption} ###### Signature Proteins specific to the Bartonellaceaefamily.^a^ ::: Protein BQ12190\[49474706\] BQ11460\[49474647\] BQ11450\[49474646\] BQ11430\[49474645\] BQ11380\[49474640\] BQ11160\[49474626\] BQ11120\[49474623\] BQ11100\[49474621\] BQ11030\[49474614\] ------------------ --------------------------- ----------------------------- ------------------------------ --------------------------- --------------------------- ------------------------------ ----------------------------- ---------------------------- ---------------------------- Length 94 103 129 65 76 104 264 231 148 B. henselae 2e-27 (1.00) 2e-48 (1.02) 2e-52 (1.00) 3e-22 (1.00) 1e-22 (0.83) 4e-41 (1.05) e-103 (1.00) 2e-94 (1.00) 3e-64 (0.99) B. quintana 1e-40 (1.00) 3e-52 (1.00) 3e-61 (1.00) 2e-31 (1.00) 4e-38 (1.00) 6e-54 (1.00) e-145 (1.00) e-131 (1.00) 4e-83 (1.00) Other Rizobiales \- \- \- \- \- \- \- \- \- Non-Rizobiale Prov. rettgeri 1.0 (2.39) Symbio. therm. 0.024 (2.16) Bacillus subtilis 3.9 (3.59) Lacto. gasseri 2.3 (13.8) Staph. aureus 0.12 (5.84) Oceano. iheyensis 1.7 (6.16) Dehalo. etheno. 0.57 (3.08) Bacillus cereus 1.7 (1.82) Citro. freundii 1.8 (5.72) ^a^Abbreviations and further details regarding BLAST results can be found in table 1. Additional abbreviations: Citro., Citrobacter; Dehalo. etheno., Dehalococcoides ethenogenes; Lacto., Lactobacillus; Oceano., Oceanobacillus; Prov., Providencia; Staph., Staphylococcus; Symbio. therm., Symbiobacterium thermophilum. ::: ::: {#T8 .table-wrap} Table 8 ::: {.caption} ###### Other Alpha-proteobacterial Specific Proteins^a^ ::: ------------------------------------------------------------------------------------------------------------- Protein CC0189\[16124444\] CC0569\[16124823\] CC0331\[16124586\] -------------- ------------------------------ -------------------------------- ------------------------------ Length 88 288 186 C. cres. 5e-38 (1.00) e-126 (1.00) e-102 (1.00) Other Alphas Mag. mag.; 8e-08 (0.93)\ Nov. aro.; 1e-40 (1.12)\ Bra. jap.;2e-19 (0.71)\ Rhod. rubr.; 8e-08 (0.76)\ Meso. loti; 3e-50 (1.00) Agr. tum.; 2e-23(0.87)\ Nov. aro.; 1e-10 (0.75)\ Bru. mel.; 1e-17 (0.76)\ Sil. pom.; 2e-09 (0.67)\ Bru. suis; 2e-23 (0.92)\ Sil. sp.; 2-09 (0.68)\ Meso. loti; 6e-20 (0.89) Rh. spha.; 2e-10 (0.67) Non-Alpha Nitro. euro.; 0.089 (0.84) Desulf. haf.; 7e-05 (0.88) Micro. deg.; 0.007 (0.96) Protein CC0349\[16124604\] CC2323\[16126562\] CC2637\[16126872\] Length 265 377 374 C. cres. e-138 (1.00) 0.0 (1.00) 0.0 Other Alphas Mag. mag.; 2e-29 (0.98)\ Mag. mag.; 2e-70 (0.98)\ Mag. mag.; 3e-19 (0.99)\ Rhod. rubr.; 8e-32 (0.67)\ Rhod. rubr.; 2e-68 (1.00)\ Rhod. rubr.; 1e-23 (1.09)\ Nov. aro.; 3e-44 (0.92)\ Bra. jap.; 3e-40 (1.04)\ Bra. jap.; 7e-22 (0.99)\ Sil. sp.; 5e-27 (0.68) Rho. pal.; 4e-36 (1.03)\ Rho. pal.; 8e-19 (0.97)\ Agr. tum.; 7e-23 (1.02)\ Agr. tum.; 2e-04 (0.91)\ Sino. meli.; 3e-27 (1.00)\ Sino. meli.; 8e-13 (1.01)\ Meso. sp.; 7e-23 (1.01) Meso. loti; 1e-07 (0.92)\ Meso. sp.; 1e-06 (1.02) Non-Alpha Staph. epi.; 0.006 (1.34) Trep. pallidum; 0.085 (1.14) Pse. fluor.; 0.32 (1.74) ------------------------------------------------------------------------------------------------------------- ^a^Abbreviations and other details regarding BLAST results can be found in Table 1. Additional abbreviations: Desulf. haf., Desulfitobacterium hafniense; Micro. deg., Microbulbifer degradans; Nitro. euro., Nitrosomonas europaea; Pse. fluor., Pseudomonas fluorescens; Staph. epi., Staphylococcus epidermidis; Trep., Treponema. ::: ::: {#T9 .table-wrap} Table 9 ::: {.caption} ###### Alpha-proteobacterial specific proteins with lateral gene transfers. ^a^ ::: ----------------------------------------------------------------------------------------- Protein CC2585\[16126823\] CC0226\[16124481\] ---------------------- --------------------------------- -------------------------------- Length 209 132 Alpha-Proteobacteria Mag. mag.; 2e-16 (0.79)\ Rhod. rubr.; 8e-14 (0.59)\ Rhod. rubr.; 2e-16 (1.03)\ C. cres.; 6e-72 (1.00)\ C. cres.; 2e-96 (1.00)\ Agr. tum.; 2e-17 (0.64)\ Rho. pal.; 7e-21 (1.05)\ Sini. mel.; 2e-15 (0.64)\ Agr. tum.; 7e-19 (1.02)\ Meso. loti.; 1e-18 (0.75)\ Sini. mel.; 2e-19 (1.14)\ Meso. sp.; 5e-21 (0.64) Bru. mel.; 5e-19 (1.04)\ Bru. suis.; 5e-19 (1.04)\ Meso. loti.; 6e-35 (1.32) Other Bacteria Gamma-proteobacteria:\ Gamma-proteobacteria:\ Az. vine.; 2e-07 (1.05)\ Pse. aer.; 6e-21 (0.64)\ Pse. fluor.; 3e-07 (1.04)\ Sal. ent.; 1e-20 (0.64) Pse. aer.; 1e-06 (1.05)\ Pse. syr.; 1e-06 (1.04) Non-Alpha Des. vulgaris; 0.073 (1.40) Pro. marinus; 0.26 (2.54) Protein CC2790\[16127022\] RP382\[15604247\] Length 567 510 Alpha-Proteobacteria Mag. mag.; 9e-49 (0.41)\ R. prowazekii; 0.0 (1.00)\ Nov. aro.; 2e-74 (0.78)\ R. conorii; e-155 (0.99)\ C. cres.; 0.0 (1.00)\ R. akari; e-141 (1.01)\ Sil. pom.; 2e-85 (0.72)\ R. rickettsii; e-155 (0.99)\ Sil. sp.; 2e-79 (0.76)\ R. sibirica; e-152 (0.99)\ Rh. spha.; 1e-79 (0.74)\ Ehr. canis; 9e-27 (0.65)\ Rho. pal.; 3e-77 (0.68)\ Ehr. rum.; 2e-32 (0.85)\ Agr. tum. 6e-82 (0.72)\ Wolbachia; 2e-32 (0.82)\ Bru. mel.; 1e-86 (0.76)\ Ana. mar.; 6e-20 (0.85) Bru. suis;; 6e-83 (0.68)\ Meso. sp.; 2e-28 (1.40) Other Bacteria Beta-proteobacteria:\ Aquificales:\ Burk. cepacia; 4e-18 (0.81) Aqu. aeolicus; 7e-11 (0.76) Non-Alpha Strep. coelicolor; 2.7 (0.65) Bac. frag.; 0.097 (1.00) ----------------------------------------------------------------------------------------- ^a^Abbreviations and other details regarding BLAST results can be found in Table 1. Additional abbreviations: Az. vine., Azotobacter vinelandii; Aqu., Aquifex; Bac. frag., Bacteroides fragilis; Burk. Burkholderia; Des., Desulfovibrio; Lep. int., Leptospira interrogans; Pro., Prochlorococcus; Pse. aer., Pseudomonas aeruginosa; Pse. fluor., Pse. fluorescens; Pse. syr., Pse. syringae; Sal. ent., Salmonella enterica. ::: The first grouping of α-proteobacterial markers consists of 6 proteins that are specific to nearly all sequenced α-proteobacterial species and are not found in any other Bacteria(Table [1](#T1){ref-type="table"}). These proteins clearly distinguish the α-proteobacteria as a distinct group from all other Bacteria. Even though some genes have been lost from certain species, these proteins remain largely distinctive of the α-subdivision. Interestingly, no homologues were detected in Zymomonas mobilisfor three of these signature proteins (CC3319, CC1887, CC1725). Z. mobilisis also lacking a number of other signature proteins described in this study and this may be attributed to the genetic loss of a variety genes resulting in its small genome size (2.06 Mb) \[[@B33]\]. A number of genes for the tricarboxcylic acid cycle as well as other functions have previously been documented as missing in this genome \[[@B33]\]. One of these signature proteins (CC1725) is also not found in Novosphingobium aromaticivoransindicating it was lost from members of the Sphingomonadalesfamily. A homologue of the protein CC3319 was detected in the currently unclassified Magnetococcus sp. MC-1genome suggesting that this species may be distantly related to the α-proteobacteria \[[@B42]\]. A number of α-proteobacteria-specific indels (i.e., inserts or deletions) are also present in Magnetococcus\[[@B17]\], supporting the above inference. Finally, the protein CC1887 is also found in the α-proteobacteria as well as a variety of Eukaryotessupporting the derivation of mitochondrion from an α-proteobacterial lineage \[[@B9]-[@B13]\]. Another group of 10 signature proteins showing a high affinity for sequenced alphas are those distinguishing all other α-proteobacteria from the order Rickettsiales(Table [2](#T2){ref-type="table"}). In this case, the Rickettsialesshow no detectable homologues of otherwise α-specific proteins. These results suggests that the genes for these proteins have either been lost from the Rickettsialesor it forms one of the earliest branching lineage within α-proteobacteria \[[@B2],[@B43]\]. These proteins are present in almost all other sequenced α-proteobacteria with few exceptions. The proteins CC0520 and CC0366 have homologues in Magnetococcus sp. MC-1again lending support to the inference that this unclassified species is distantly related to the alpha-group. The protein CC1977 is also found in Eukaryotesand the E values for a few representative eukaryotic species are given in the Table [2](#T2){ref-type="table"} legend. One protein (CC3010), showing a very high affinity for this grouping as noted by low E values, is also found in a single gamma proteobacterium (Pseudomonas sp.). This finding is most likely due to a non-specific event such as a lateral gene transfer (LGT) of which additional examples will be presented later. The next grouping of signature proteins are those which are found in almost all sequenced α-proteobacteria excluding the intracellular pathogens belonging to the Bartonellaceaefamily and the order Rickettsiales(Table [3A](#T3){ref-type="table"}). This grouping outlines a case in which proteins have probably been lost independently from two unrelated groups within the α-proteobacteria most likely due to their intracellular lifestyles \[[@B2],[@B3],[@B44]\]. Five proteins of this type were identified with minimal loses seen in other α-proteobacteria. CC2345 provides a good example of this type of protein since it is highly conserved in all available α-proteobacterial genomes as indicated by low E values. The other four proteins also show a high affinity for this category with losses occurring only in Z. mobilisand Rhodospirillum rubrum. A variation on the above theme is a collection of 4 α-specific proteins that are absent in the orders Rickettsialesand Rhodobacterales(Table [3B](#T3){ref-type="table"}). However, a key feature distinguishing these proteins from those presented in Table [3A](#T3){ref-type="table"} is the free-living lifestyle of the Rhodobacteralesas opposed to the intracellular Bartonellas. Since Rickettsialesand the Rhodobacteralesare not known to share any unique characteristic, it is possible that the loss of these proteins from these two orders has occurred due to unrelated reasons. Also, some additional losses are seen in this grouping. For example the protein CC1652 is absent in the Sphingomonadaleswhile the protein CC1035 is absent in the Rhodospirillales. Note that the protein CC2247 exhibits high E values for BLAST hits representing Brucellaceaeand Bartonellaceaebut this high E value is acceptable due to the very short length of this protein (46 amino acids) and the fact that besides α-proteobacteria no other BLAST hits were observed (Table [3B](#T3){ref-type="table"}). The Blast searches on proteins found in the R. prowazekiigenome have led to identification of a number of signature proteins which are specific to species belonging to the order Rickettsiales. This order is made up of two families: the Anaplasmataceae(Anaplasma, Ehrlichia and Wolbachia) and Rickettsiaceae(Rickettsias) \[[@B2],[@B43]\]. The first group of such proteins (RP104, RP105, and RP106) are present in all species belonging to the order Rickettsiales, but are not found in any other α-proteobacteria (Table [4](#T4){ref-type="table"}). It should be noted that the proteins RP104 and RP106 do not show homology over the entire length of the homologous proteins in members of the Anaplasmataceaefamily. Thus, additional domains that are specific for the Rickettsiaceaefamily may be present in these proteins. These signature proteins are highly conserved within this order, as indicated by their very low E values (Table [4](#T4){ref-type="table"}) and represent interesting examples of genes that were likely introduced in a common ancestor of the Rickettsiales. Note that the first non-RickettsialeBLAST hit for the protein RP106 appears at 2e-07 (Xyella fastidiosa). RP106 is still included as a Rickettsiales-specific protein because the Xyellaprotein is only 348 amino acids in length and thus it is likely a different protein. Another group of 4 proteins are specific to the Rickettsiaspecies and are not found in other members of the Rickettsiales(Table [4](#T4){ref-type="table"}). These proteins (RP766, RP192, RP030 and RP187) are highly conserved and represent cases in which genes were introduced into a common ancestor of the Rickettsiaceae. Homologues of the protein RP187 are much longer in other Rickettsiastrains (194 vs 497 aa) but the region representing the query sequence is highly conserved. It is possible that other Rickettsiaspecies have acquired an additional protein domain during the course of evolution. In addition to the Rickettsiales, the Rhizobialesform a major order within the α-proteobacteria \[[@B1],[@B17],[@B42]\]. To identify proteins which are distinctive of the Rhizobiales, BLAST searches were carried out on all ORFs in the genome of B. quintana. Six proteins have been identified that are conserved amongst all sequenced Rhizobialeswith minimal evidence of gene loss occurring (Table [5](#T5){ref-type="table"}). The protein BQ07670 is absent in Rhodopseudomonas palustriswhile the protein BQ11900 is absent in this strain as well as in Sinorhizobium meliloti. The presence of these proteins solely in the Rhizobialesindicates they were likely introduced in a common ancestor of this order. Other signature proteins that are useful in defining the Rhizobialesare those which are present in all sequences members of this order, except the Bradyrhizobiaceaefamily (Table [6](#T6){ref-type="table"}). Four proteins of this type have been identified with no losses occurring in any species. These proteins indicate that the Bradyrhizobiaceaefamily is more distantly related to other members of the Rhizobiales. The deeper branching and distinctness of Bradyrhizobiaceaeand Methylobacteriaceaefrom other Rhizobialesis also strongly supported by phylogenetic analyses based on different gene sequences and conserved indels in many proteins \[[@B1],[@B17],[@B45]\]. A number of proteins have also been identified which are unique to the Bartonellaspecies. Nine examples of such proteins are shown in Table [7](#T7){ref-type="table"}. These proteins are highly conserved amongst both sequenced Bartonellaspecies with no gene losses occurring. The presence of these proteins solely in this family of α-proteobacteria indicates that they should provide useful markers for the Bartonellaceaefamily. Six other α-specific signature proteins were identified that do not show any distinct pattern but are sporadically present in α-proteobacterial species (Table [8](#T8){ref-type="table"}). These proteins are more randomly distributed among a limited number of sequenced α-proteobacteria and it is likely that gene losses for these proteins have occurred independently in various species or groups. Nevertheless, these proteins are still unique to the α-proteobacteria. The protein CC0189 is represented in the Rhodospirillales, Novosphingomonadales, Caulobacteralesand Rhodobacteralesbut is not found in any Rhizobiales. One protein (CC0331) is represented in various families within the Rhizobialeswhile two others (CC2323 and CC2637) show a similar trend and are also present in Rhodospirillales. A final grouping of 4 signature proteins consists of those where limited lateral gene transfers (LGTs) have apparently occurred (Table [9](#T9){ref-type="table"}). Three of these proteins (CC2585, CC0226 and CC2790) were isolated from the Caulobactergenome and represent cases in which genes were also present in a limited numbers of gamma or beta-proteobacteria. Specifically, a homologue of the protein CC2585 was detected in a number of gamma-proteobacteria belonging to the Pseudomonadaceaefamily while CC0226 was only detected in Pse. aeruginosaand the enteric bacterium Salmonella enterica. The protein CC2790 shows some similarity to a Superfamily I DNA and RNA helicase found in Burkholderia cepacia(beta-proteobacteria). However, this BLAST hit only shows conservation over 142 amino acids of the 567 amino acids C. crescentusprotein. Furthermore, all alpha BLAST hits are annotated as hypothetical proteins indicating this non-alpha BLAST hit probably represents a different protein with a shared protein domain that was transferred. Interestingly, one of the proteins, RP382, which is otherwise highly specific for the order Rickettsiales, is also found in Aquifex aeolicus. In each of these cases, the direction of gene transfer remains unclear. Discussion ========== The α-proteobacteria forms an extremely diverse group showing vast differences in such characteristics as morphology, metabolism, and physiology \[[@B1]\]. In the current view, this group is distinguished from all other Bacteriabased on 16S rRNA phylogenetic trees \[[@B1],[@B8],[@B19],[@B46]\]. Few molecular or physiological characteristics were known which clearly distinguish this group from all other Bacteria\[[@B1],[@B7]\]. However, our recent work has identified a large number of conserved inserts and deletions in protein sequences which are distinctive characteristics of α-proteobacteria and its subgroups and not found in any other groups of Bacteria\[[@B17]\] (see also <http://www.bacterialphylogeny.com>). These signatures provide useful tools for identifying α-proteobacteria within Bacteriaas well as for understanding the interrelationships and branching order within this group. Here, we describe 61 signature proteins that are largely specific for the α-proteobacteria. Almost all of these proteins are of hypothetical functions, and in view of their α-proteobacterial specificity, it is likely that they are involved in functions that are limited to only this group of bacteria. Because such genes are likely involved in specialized functions, the loss of some of these genes from certain α-proteobacterial species is not surprising. Based on signature proteins described here, along with various α-proteobacteria-specific conserved inserts and deletions \[[@B17]\], a clearer picture of α-proteobacteria phylogeny and taxonomic classification can be derived. Figure [1](#F1){ref-type="fig"} presents a model for α-proteobacterial evolution which indicates the evolutionary stages where these proteins are suggested to have evolved or been introduced. The model based on these signature proteins is identical to that deduced independently based upon a large number of conserved indels in different proteins \[[@B17]\], indicating its reliability. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Summary diagram showing the distribution pattern of various α-proteobacteria signature proteins. The arrows indicate the evolutionary stages where these signature proteins were likely introduced. Some proteins, which are sporadically present in α-proteobacteria are not shown here. The branching position of α-proteobacteria relative to other bacterial groups was deduced as described in earlier work \[13,17,40\]. ::: ![](1471-2164-6-94-1) ::: Several signature proteins are specific to nearly all α-proteobacteria. These proteins provide additional support to various alpha-distinguishing indels, which are found only in the α-proteobacteria and not in any other groups of bacteria. Examples of such indels include the following: an 8 amino acid insert in the α subunit of ATP synthase complex, 3 amino acid insert in prolipoprotein-phosphatidylglycerol transferase, and a 1 amino acid deletion in the FtsK protein \[[@B17]\]. The simplest and most parsimonious explanation for the presence of these α-specific signatures (both proteins and indels) is that they were introduced once in a common ancestor of all α-proteobacteria and their presence in various α-proteobacterial species is due to vertical transmission \[[@B47],[@B48]\]. It is difficult to explain the presence of these genes in various α-proteobacteria by other non-specific means such as lateral gene transfers \[[@B49]\]. The finding of these unique genes and conserved indels in various α-subdivision members strongly indicates that all such bacteria carry out certain physiological functions that are unique to the members of this group. Therefore, studies aimed at determining the functional roles of these proteins and indels are of much interest. The largest group of signature proteins discovered are those found in all α-proteobacteria excluding the order Rickettsiales. These proteins indicate that the Rickettsialesconstitute a distinct clade within the α-subdivision, which is in accordance with phylogenetic analyses based on different gene sequences \[[@B2],[@B17],[@B43],[@B50]\]. Phylogenetic studies based on 16S rRNA and many other genes \[[@B2],[@B43],[@B45],[@B50]\], as well as our studies based on conserved indels in several proteins that are present in various α-proteobacteria but absent in Rickettsialesas well as other groups of bacteria \[[@B17]\], provide evidence that the Rickettsialescomprise the deepest branching group within α-proteobacteria. In view of this, the most logical explanation for these signatures is that they were introduced in a common ancestor of other α-proteobacteria after the divergence of the Rickettsiales(Figure [1](#F1){ref-type="fig"}). An interesting group of α-specific signature proteins are those which are absent in the intracellular pathogens belonging to the order Rickettsialesand the family Bartonellaceae. The latter group of species form a family within the Rhizobialesorder \[[@B1],[@B17]\]. Because these two groups are phylogenetically unrelated, it is likely that the genes for these proteins were selectively lost in these two groups independently due to their intracellular lifestyles. It is logical to assume that the cellular functions of these proteins are either not required in the intracellular environment, or they are provided for by the host cells leading to the loss of these genes from these organisms. These proteins could have been introduced in either a common ancestor of all α-proteobacteria and subsequently lost in the Rickettsialesand Bartonellaceae, or introduced after the divergence of the Rickettsialesand lost in the Bartonellaceae. It is interesting that the Brucellas(also intracellular pathogens) have retained all of these proteins indicating that this group differ in its physiological requirements from other α-proteobacterial intracellular pathogens \[[@B1],[@B3],[@B51]\]. Several α-specific signature proteins that are absent in both the Rickettsialesas well as Rhodobacteraleswere also identified. Since there is no evidence to suggest any sort of relationship between these two groups \[[@B1],[@B17]\], the simplest explanation is that these genes were introduced after the divergence of the Rickettsialesand lost preferentially by the Rhodobacterales. Other signature proteins were isolated pointing to a variety of relationships. For instance, the protein CC0189 which is only present in Caulobacterales, Rhodobacteriales, Rhodospirillalesand Novosphingomonadalesindicates a close relationship between these deep branching orders within α-proteobacteria. This relationship is also seen from the protein CC0349 but to a lesser extent since losses have occurred in some species. These findings are supported by indels in a variety of proteins that indicate these orders show a closer relationship and have branched prior to the Rhizobiales\[[@B17]\]. Other signature proteins are found in a selection of these above orders and are also found in some but not all families within the Rhizobiales(CC0331, CC2323 and CC2637). A close relationship between Caulobacterand Rhodobacteralesis generally indicated by phylogenetic trees and is also supported by a conserved 11 amino acid insert in the protein aspargine-glutamine amido transferase \[[@B1],[@B17]\]. Thus, it is somewhat surprising that in our analysis of the Caulobactergenome, we did not identify any signature protein that was uniquely shared by these two α-proteobacterial orders. However, a 12 amino acid insert in the protein DNA ligase indicates that Rhodobacteralesmay be more closely related to Rhizobialesin comparison to Caulobacterales\[[@B17]\]. In view of these results, and the fact that C. crescentusrepresents the only fully sequenced bacterium within its order \[[@B24]\], additional sequence information is required to further clarify the evolutionary relationships amongst Rhizobiales, Rhodobacteralesand Caulobacterales. Several signature proteins were found to be specific for either the order Rickettsialesor the family Rickettsiaceae. These proteins provide molecular markers for these groups and they likely originated in common ancestors of these groups. The distinctness of these groups is also supported by a number of conserved indels in different proteins which are uniquely present in the species from these groups, but not found in any other bacteria \[[@B17]\]. It should be noted that McLeod et al. \[[@B28]\] based upon their comparative analysis of the Rickettsiasgenomes have identified a number of hypothetical proteins that are only found in particular Rickettsias. These proteins were grouped into the following classes: R. typhiORFs not found in R. conoriior R. prowazekii; R. typhiORFs found in R. conoriibut not in R. prowazekii; and R. typhiORFs found in R. prowazekiibut not in R. conorii. However, no proteins that were specific for all Rickettsiasor Rickettsialeswere described in the McLeod et al. study \[[@B28]\]. A number of signature proteins identified here are useful in defining and characterizing the Rhizobialesorder. Of the six Rhizobiales-specific proteins described here, four (viz. BQ00140, BQ00720, BQ03880 and BQ12030) are completely conserved amongst all sequenced Rhizobialesand should provide good molecular markers for this order. Two other proteins (BQ07670 and BQ11900) also show a high affinity for this grouping with a few gene losses in some species. We have previously described a conserved indel in tryptophanyl-tRNA synthetase that is present in all sequenced Rhizobialesbut is absent in all other bacteria \[[@B17]\]. These signatures were likely introduced in a common ancestor of the Rhizobialesorder (Figure [1](#F1){ref-type="fig"}). Four additional proteins that were identified here are completely specific to all sequenced Rhizobiales, except for the Bradyrhizobiaceaefamily. Phylogenetic analysis based on a number of gene sequences as well as conserved indels in a number of proteins (viz. Trp-tRNA synthetase, LytB metalloproteinase) provide evidence that that the Bradyrhizobiaceaefamily is distantly related to other Rhizobiales(Rhizobiaceae, Brucellaceae, Phyllobacteriaceae), and it has branched prior to the latter groups of species \[[@B1],[@B17],[@B45]\]. Thus, it is likely that these signature proteins evolved in a common ancestor of various other Rhizobialesafter the divergence of the Bradyrhizobiaceaefamily (Figure [1](#F1){ref-type="fig"}). A number of signature proteins that are unique for the Bartonellaspecies were later introduced in that particular branch of the tree (Figure [1](#F1){ref-type="fig"}). Although most of the signature proteins identified here are specific for only the α-proteobacteria, we have also come across a few examples where lateral gene transfer seems to have occurred between α-proteobacteria and a few species from other groups of bacteria. The rarity of such proteins in comparison to those which exhibit strict group-specificity indicates that most newly acquired alpha-specific genes have been predominantly transmitted via vertical descent and LGT and other non-specific mechanisms play relatively minor role in their transmission. It should be mentioned that although our analyses of proteins in R. prowazekii, C. crescentusand B. quintanagenomes have identified a large number of signature proteins, based on these studies signature proteins for certain other groups within α-proteobacteria (e.g. Rhizobiaceae, Rhodobacterales, Sphignomonadales, etc.) will not be detected. It should be possible to identify signatures for these groups by carrying out similar analysis using protein sequences from these genomes. Daubin and Ochmann \[[@B52]\] and Lerat et al. \[[@B36],[@B47]\] have previously examined the gene repertoire of γ-proteobacteria and have indicated the presence of many ORFans genes (i.e. ORFs that have no known homologs) that are limited to either certain bacterial strains or particular subgroups of γ-proteobacteria. The ORFan genes were found to be present in their studies in different monophyletic clades at different phylogenetic depths, which is similar to what we have reported here for the signature proteins in the α-proteobacteria taxon. The other characteristics of ORFans genes noted by these authors were that they are generally short (between 400--500 bp), A+T rich, and evolve faster than other genes which are more broadly distributed \[[@B47],[@B52]\]. Many of the signature proteins identified in the present work are of similar lengths as the ORFans genes. These earlier studies also indicate that ORFans genes generally encode for functional proteins, and once acquired they are vertically transmitted, and based on them it possible to make robust phylogenetic inference as we have been able to do in the present study for α-proteobacteria. Although the source of ORFans genes in different genomes remains to be determined, it has been suggested that many of them are derived from bacteriophages \[[@B47],[@B52]\]. The concept that mitochondria have originated from an α-proteobacterial ancestor is supported by a large body of evidence including phylogenetic analysis and shared presence of many common indels \[[@B9]-[@B14]\]. The homologues of two of the α-proteobacterial signature proteins (CC1887 and CC1977) are also present in Eukaryotesproviding further support for this inference. For the remainder of the proteins no eukaryotic homologues were detected which supports the observation of Boussau et al. \[[@B44]\] that for a large fraction of genes in α-proteobacterial genome no homologs are found in the eukaryotes. Currently, it is thought that within α-proteobacteria the species belonging to the order Rickettsialesare the closest relatives of mitochondria \[[@B10]-[@B12],[@B53]-[@B55]\]. However, of the two proteins which are commonly found in eukaryotes, only one of them (CC1887) is present in the Rickettsiales. A specific relationship of mitochondria to the Rickettsialesis also supported by only some conserved indels, but not all \[[@B17]\]. In a recent study, where the relationship of alpha proteobacteria to mitochondria was examined based on a large number of individual and concatenated protein sequences \[[@B56]\], the closest relationship of mitochondria was seen for Rhodospirillum rubrumrather than the Rickettsiales. In earlier work, we have described two conserved signatures (a 37 aa insert in valyl-tRNA synthetase and 1 aa indel in LonA protein), which were commonly shared by all eukaryotic homologs and certain other groups of bacteria but which were not found in any α-proteobacteria \[[@B13]\]. An update of these signatures indicates that they still constitute exceptions to the α-proteobacterial derivation of the mitochondrial/eukaryotic homologs (R.S. Gupta, unpublished results). These observations in conjunction with the recent conflicting observations regarding the possible origins of NADH dehydrogenase subunits from Trichomonas vaginalis\[[@B57],[@B58]\] indicate that additional work is necessary to clarify the sources of different mitochondrial and nuclear cytosolic genes of eukaryotic proteins. Conclusion ========== Whole-genome analyses of B. quintana, Ri prowazekiiand C. crescentusproteins have led to the discovery of 61 signature proteins which are distinctive characteristics of the α-proteobacteria and its subgroups. These signature proteins provide additional support to our recent work based on a large number of conserved inserts and deletions in protein sequences that are either specific for the α-proteobacteria or provide information regarding the interrelationships and branching order within this group \[[@B17]\]. Sequence information from additional α-proteobacterial species will be useful in testing the predicted presence or absence of various identified molecular signatures (indels and proteins) in different groups, thus validating the suggested relationships. Studies aimed at understanding the cellular functions of these α-specific signature proteins should be of much interest since they will likely provide novel insights into unique physiological characteristics shared by this important group of bacteria and its various subgroups. Studies on proteins which are specific to the intracellular pathogens, such as Rickettsialesand Bartonella, could also provide new drug targets for their associated diseases. Methods ======= Identification of α-Proteobacteria Specific Proteins ---------------------------------------------------- To identify signature proteins which are specific to the α-proteobacteria or its various subgroups, all proteins in the genomes of C. crescentus, R. prowazekii, and B. quintanawere analyzed. BLAST searches were carried out \[[@B41]\] on each individual protein in these genomes to identify all other bacteria containing proteins with similar sequences. These results were visually inspected for homologues showing specificity to α-proteobacteria with no other similar homologues present in any other Bacteria. Expect values (E values) were analyzed for putative α-specific proteins. The E values, which are calculated by the BLAST software, indicate the probability that the observed similarity between the query protein and any other protein detected by the BLAST search arose by chance \[[@B41]\]. In BLAST searches, the E values are lowest (closer to 0) for BLAST hits with a high degree of homology to the query sequence and they increase as BLAST hits are detected with lower similarity. The results of BLAST searches were inspected for sudden increases in E values from the last α-proteobacteria in the search to the first non-alpha bacteria. This increase in E values was important when the next non-alpha BLAST hit was in a range where the observed similarity could occur by chance (\> 10^-05^). However, higher E values were sometimes allowed and could be significant for smaller proteins since they contain fewer characters resulting in higher E values (for statistical reasons) for their true homologs. For all α-specific signature proteins described here, E values were recorded for each blast hit as well as for the first non-alpha bacterium in a given search. Although E values take into account the length of the sequence over which the similarity is observed between the query sequence and a BLAST hit, low E values can sometime result if high degree of homology is observed between two different proteins over a short sequence region. Therefore, we have also inspected BLAST results for homology over the entire protein length and for similarity in protein length. The length ratios of the hit proteins over the query protein are shown in brackets beside the E values and these values are expected to be close to 1.00 if the identified proteins are of similar lengths as the query protein. It should be mentioned that BLAST searches can sometime indicate misleading similarity, particularly when no close relatives of the query species are in the database \[[@B59]\]. However, in the present study where most of the BLAST hits correspond to α-proteobacteria, such a possibility is highly unlikely. All proteins indicated in the Tables [1](#T1){ref-type="table"}, [2](#T2){ref-type="table"}, [3](#T3){ref-type="table"}, [4](#T4){ref-type="table"}, [5](#T5){ref-type="table"}, [6](#T6){ref-type="table"}, [7](#T7){ref-type="table"}, [8](#T8){ref-type="table"}, [9](#T9){ref-type="table"} are specific for the α-proteobacteria based on these criteria unless otherwise mentioned. Authors\' contributions ======================= PK carried out BLAST searches on different proteins and was responsible for the initial evaluation of the results. RSG conceived and directed this study and was responsible for the final evaluation of the results. PK prepared a rough draft of the manuscript under RSG\'s directions, which was revised and modified by RSG. All authors read and approved the final manuscript. Acknowledgements ================ The work was supported by a research grant from the National Science and Engineering Research Council of Canada.	PubMed Central	2024-06-05T03:55:59.957586	2005-6-16	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182365/", "journal": "BMC Genomics. 2005 Jun 16; 6:94", "authors": [ { "first": "Pinay", "last": "Kainth" }, { "first": "Radhey S", "last": "Gupta" } ] }
PMC1182366	Background ========== The hepatitis B virus (HBV) and hepatitis C virus (HCV) often cause persistent infection, leading to chronic liver diseases, cirrhosis and hepatocellular carcinoma \[[@B1],[@B2]\]. Given the burden of these diseases and the current potential for cure, there is a compelling need for diagnosis of active HBV and HCV infection. A variety of HBV and HCV markers have been used to detect HBV and HCV infection. Gene amplification tests, such as PCR-based \[[@B3]-[@B7]\] assays are used to diagnose and monitor the efficacy of treatment. However, these methods require cumbersome procedures and expensive equipment, thus requiring considerable skills and high costs. Immunoassays are generally easy and inexpensive. So far, some immunological methods such as enzyme-linked immunosorbent assays (ELISA) and rapid diagnostic paper have been used in clinical practice. While the value and significance of these methods are beyond argument, they suffer from several disadvantages, mainly their inability to produce results simultaneously. Ruo-Pan Huang \[[@B8]\] has detected multiple cytokines and antibodies simultaneously on nitrocellulose membrane, utilizing horseradish peroxidase (HRP)-conjugated antibodies as detecting reagents and visualizing the signals with an enhanced chemiluminescence (ECL) system. However, this method is time-consuming and requires expensive set-up, limiting its large-scale application. Mezzasoma et al. \[[@B9]\] have detected serum antibodies against the TORCH antigens on amino-silane-activated glass slides with fluorescently labelled secondary antibodies. Unfortunately, this method is also limited in clinical applications due to the cost of the assay. In the past few years, protein chip and microarray technology has shown its great potential in the functional analysis of the proteome, clinical diagnostics and drug discovery. It allows fast, easy and parallel detection of thousands of addressable elements in a single assay. For instance, the potential of this technology to diagnose human diseases, such as leukemia, breast cancer and, potentially, heart failure, has stimulated much interest. In our previous studies, we established a platform on which gene chips with a high sensitive visual detection based on two-probe sandwich hybridization/nanoparticle amplification have been employed, and HBV and HCV gene fragments were detected on a glass slide by visual inspection \[[@B10],[@B11]\]. In this paper, we developed a protein chip technology based on NIASS method. A protein chip was devised to detect antibodies of HBV and HCV easily and simultaneously. In this assay, the enhancing solution was the physical developer that contained both silver ions and a reducing agent, buffered to an acidic pH. During silver enhancement, the colloidal nano-gold served as a nucleation site for the deposition of metallic silver and the particles grew in size, giving an intensely dark signal which could be visualized with the naked eyes. Colloidal nano-gold labelled SPA was used as a detecting reagent which could bind specifically to the Fc portion of immunoglobulin from many mammals. The clinical performance of this assay was validated with a collection of serum samples previously characterized with commercial ELISA for their reactivity against the selected antigens. The data displayed that no distinct difference (P \> 0.05) existed between the results determined by our assay and ELISA respectively. In a preliminary test, our assay detected up to 3 ng/ml anti-IgG, which was close to that in the fluorescent detection method. Methods ======= Preparation of nano-gold particles ---------------------------------- Colloidal nano-gold solutions were prepared by the citrate reduction of HAuCl~4~according to the literature \[[@B12]\], filtered through a 0.45 μm nylon filter, and stored at 4°C. Prior to use, all glassware was immersed in cleaning solution (200 g potassium dichromate and 500 ml of concentrated sulfuric acid dissolved in distilled water to the volume of 2000 ml) for 24 hours, rinsed 3 times with ddH~2~O and then dried in oven. A transmission electron microscope (TEM, HITACHI H-8100, Japan) was used to determine the size of the colloidal nano-gold particles and UV-Vis spectrum was used to analyze the uniformity of the nano-gold particles. Determination of the optimum ratio of colloidal nano-gold solution to SPA concentration --------------------------------------------------------------------------------------- Prior to the preparation of colloidal nano-gold labelled SPA, the optimum ratio of colloidal nano-gold solution to SPA concentration should be determined. Colloidal nano-gold solution was centrifuged (500 g) for 15 minutes, and the precipitate was eliminated. Serial volume SPA solution (0.05 mol/L) was added to 1000 μl of the colloidal nano-gold solution which had been adjusted to pH 5.5 with 0.2 M potassium carbonate, followed by incubation for 15 minutes at room temperature. After 100 μl of 10% (w/v) NaCl solution was added to the mixture, the absorbance at 520 nm was measured, and the SPA solution volume showing maximum absorbance was regarded as the optimum. The volume of SPA solution in practical operation generally increases 10%--20%. Preparation of nano-gold-SPA probes ----------------------------------- Thirteen microlitres of a SPA solution was incubated with 1000 μl of the colloidal nano-gold solution (pH 5.5) for 15 minutes at room temperature, followed by addition of 5% (w/v) bovine serum albumin (BSA) solution to a final concentration of 1%. The mixture was allowed to stand for over 15 minutes, and was centrifuged (50000 g, 4°C) for 45 minutes twice. After each cycle, the supernatant was eliminated and the red precipitate was washed with 20 mmol/L Tris buffer (pH6.5, filtrated through a 0.22 μm membrane) including 1% (w/v) BSA and 0.02% (w/v) NaN~3~. Following the final cycle, the precipitate was suspended in the same buffer and stored at 4°C. Preparation of the protein chip ------------------------------- Super-flat optic glass slides (BAIO, China) were prepared according to the literature \[[@B13]\]. Briefly, the slides were soaked in 10% (w/v) NaOH and then 0.1 M HCl for 2 hours, respectively. After a thorough rinsing with ddH2O and boiling for 1 hour, the glass slides were set in 1% (v/v) 3-glycidoxypropyltrimethoxysilane (GOPTS) solution (GOPTS dissolved in 95% ethanol) at 37°C for 6 hours, followed by drying and incubating at 135°C for 1 hour, and finally stored at 4°C. Nought point one microlitres of the antigen solution (100 μg/mL) was manually printed onto the chemically derivated slide. Arrays included four antigens (HBsAg, HBeAg, HBcAg and HCVAg which was a mixture of NS3, NS5 and core antigens) printed in four replicates. SPA and HIVEnv36 were used as positive and negative control respectively. All the antigens were diluted with phosphate-buffer saline (PBS, 137 mmol/L NaCl, 2.7 mmol/L KCl, 10 mmol/L Na~2~HPO~4~, 2 mmol/L KH~2~PO~4~, adjusted to pH7.4 with HCl) with 40% glycerol to the final concentration. In the assay of evaluating the detection limit of the protein chip technology, SPA, human IgG and HIVEnv36 (100 μg/mL) were spotted onto the GOPTS-activated slides in four replicates to prepare a set of model arrays. The slides were placed at 37°C with the humidity maintaining at 90% for 1 hour, followed by rinsing three times with PBS to remove the unbound proteins. Blocking of the protein chip ---------------------------- The chip was immersed in 1% (w/v) BSA (dissolved in PBS) solutions for 30 minutes with gentle shaking, rinsed three times with PBST solution (Tween-20 dissolved in PBS to a final concentration of 0.05%) and finally rinsed twice with PBS. Detection of antibodies in patients\' sera ------------------------------------------ Forty microlitres of the colloidal nano-gold-SPA solution was mixed with 10 μl of serum sample. The mixture was diluted with PBS buffer, bringing the total volume to 1000 μl, and incubated for 30 minutes at 37°C. The mixed solution was placed on the surface of the chip for 30 minutes at 37°C with the humidity maintaining at 90%. The chip was rinsed three times with PBS, dried and covered with silver enhancer solution, which was composed of 700 μl of citric acid buffer (2.55 g citric acid and 2.35 g sodium citrate dissolved in 1000 ml of ddH~2~O, pH3.5), 300 μl of hydroquinone solution (1.7 g hydroquinone dissolved in 30 ml of ddH~2~O) and 20 μl of silver nitrate solution (2.5 mg silver nitrate dissolved in 100 μl of ddH~2~O). After treatment with the silver enhancer solution, the chip was rinsed with water, and then air-dried. Data collection and analysis ---------------------------- The slides were analyzed with a scanner (FOUNDER, China) and the density of each spot was determined by the software taken by the scanner. To compare the protein chip assay with ELISA, a total of 305 serum samples showing different activities in the ELISA (low, medium and high) collected from Xiehe Hospital (Wuhan, China) were detected. The use of the serum samples was approved by the patients. The sera were divided into 4 groups (Table [1](#T1){ref-type="table"}) and each sample was assayed individually. The sera were stored at -4°C and previously characterized with commercial ELISA (KHB, China), which were performed according to the manufacturer\'s instructions. The results were collected and analyzed. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### The division of serum samples for detection ::: HBsAb HBeAb HBcAb HCVAb ----------------- ------- ------- ------- ------- Group-1(n = 90) \+ \- \- \- Group-2(n = 85) \- \+ \+ \- Group-3(n = 80) \- \- \- \+ Group-4(n = 50) \- \- \- \- ::: Results ======= Size and configuration of colloidal nano-gold and nano-gold-SPA --------------------------------------------------------------- A TEM was used to determine the size distribution of the colloidal nano-gold particles. The resulting particles remained stable and formed uniform spheres. At least 200 particles were measured with TEM, of which the average diameter was 15 ± 2 nm. The so prepared nano-gold-SPA particles were stained by phosphate tungstate and observed by TEM, shown in Figure [1](#F1){ref-type="fig"}. Colloidal nano-gold particles surrounded by SPA were still well-distributed, maintaining uniform and stable. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Transmission electron micrograph of nano-gold-SPA probes. The arrowheads pointed to nano-gold-SPA probes. The structure of halo disturbance surrounding the nano-gold was SPA proteins. In keeping the objective of reacting with antibodies, SPA was chosen, for that, SPA could be site-specifically attached to Fc region of antibody, preserving its antigen-combining activity. And nano-gold clustered for silver enhancement. ::: ![](1471-2334-5-53-1) ::: Comparison of UV-Vis spectra between colloidal nano-gold and nano-gold-SPA -------------------------------------------------------------------------- A UV-Vis spectrophotometer (UNICO™, China) was employed to scan the absorbance of colloidal nano-gold in the range of 500-550 nm with a scanning precision of 1 nm. Only one peak was displayed in the scanning curve (figure not shown), which indicated that so prepared colloidal nano-gold was uniform. This method was also used to scan the absorbance curve of colloidal nano-gold and colloidal nano-gold-SPA in the range of 400-700 nm with a scanning precision of 6 nm. The results were shown in Figure [2](#F2){ref-type="fig"}, in which the peak wavelength was 520 nm, whereas that of nano-gold-SPA was 526 nm. Still only one peak was displayed in the scanning curve. The shift of λ~max~was not due to the surface modification of colloidal nano-gold, but to the change in gold particle distribution during centrifugation of colloidal nano-gold-SPA. It meant that λ~max~of colloidal nano-gold didn\'t shift along with the mutual reaction between nano-gold and SPA. The optimum ratio of colloidal nano-gold solution to SPA concentration was determined in the wavelength of 520 nm. According to the Figure [3](#F3){ref-type="fig"}, the optimum volume of SPA solution was about 12-14 μl (0.05 mol/L). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Comparison of UV-Vis spectra between colloidal nano-gold and nano-gold-SPA probes. The dash dot curve denoted UV-Vis spectrum of colloidal nano-gold within the range of 400-700 nm. The solid curve denoted UV-Vis spectrum of colloidal gold-SPA probes within the range of 400-700 nm. A(gold)/520 was 1.636920, whereas A(gold-SPA)/520 was 1.185927, and the loss of colloidal nano-gold was about 27.55%, which was due to centrifugation during preparation of nano-gold-SPA probes. ::: ![](1471-2334-5-53-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Determination of ratio between colloidal nano-gold and SPA. The value of abscissa denoted the volume of SPA solution added into each tube of colloidal nano-gold solution, while the value of ordinate denoted O.D value of each tube at the wavelength of 520 nm. The O.D value increased with the adding of SPA solution, and reached the maximum when colloidal nano-gold was saturated by SPA. ::: ![](1471-2334-5-53-3) ::: The effects of various silver enhancement time on signal intensity (relative density) ------------------------------------------------------------------------------------- Immunogold silver staining enhancement technique was applied to amplify the detection signals, which greatly enhanced the sensitivity of the assay. Therefore, the results could be visualized by the naked eyes because the small nano-gold particles could easily be grown to a useful size. The data in Figure [4](#F4){ref-type="fig"} indicated that the detected results were related to the silver enhancement time. The optimum time ranged from 8 to 12 minutes. When the time exceeded 15 minutes, the background darkened in a decrease in signal intensity which was determined by both spot intensity and background intensity. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### The effects of various silver enhancement time on signal intensity. The concentration of antigens spotted on the slides was 100 μg/ml. The silver enhancement time varied from 5 to 15 minutes. The result indicated the optimum time for silver development ranged from 8 to 12 minutes. Both too short (less than 5 minutes) and too long (more than 15 minutes) development might bring on darkened background. ::: ![](1471-2334-5-53-4) ::: Detection limit of the protein chip assay ----------------------------------------- The detection limit of the protein chip assay was demonstrated by incubation of model arrays with different concentrations of anti-IgG. The experiment was for guidance only. As shown in Figure [5](#F5){ref-type="fig"}, the anti-IgG as low as 3 ng/ml could be detected, such a sensitivity was close to that in the fluorescent detection method (1 ng/ml) \[[@B14]\]. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Detection limit of the protein chip assay. a: The concentration of anti-IgG was 3 μg/ml. b: The concentration of anti-IgG was 300 ng/ml. c: The concentration of anti-IgG was 30 ng/ml. d: The concentration of anti-IgG was 3 ng/ml. The first, second and third row denoted the negative control, detection and positive control spots respectively. The concentration of human IgG spotted on the slides was 100 μg/ml. The silver enhancement time was 10 minutes. ::: ![](1471-2334-5-53-5) ::: Comparison of protein chip assay and ELISA ------------------------------------------ The results for comparison of the protein chip assay with ELISA were shown in Table [2](#T2){ref-type="table"}, [3](#T3){ref-type="table"} and [4](#T4){ref-type="table"}. Paired chi-square test indicated that no distinct difference (P \> 0.05) existed between the results determined by our assay and ELISA respectively. While with ELISA formats, each analyte was detected in a separate assay, thereby increasing time and costs. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Comparison of the protein chip assay with ELISA test protocols (Group-1 and Group-4 serum samples) ::: Protein chip assay ELISA Total samples ---------------------------------------- ------- --------------- ---- HBsAb(+), HBeAb(-), HBcAb(-), HCVAb(-) 86 0 86 HBsAb(-), HBeAb(-), HBcAb(-), HCVAb(-) 4 50 54 P \> 0.05 ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Comparison of the protein chip assay with ELISA test protocols (Group-2 and Group-4 serum samples) ::: Protein chip assay ELISA Total samples ---------------------------------------- ------- --------------- ---- HBsAb(-), HBeAb(+), HBcAb(+), HCVAb(-) 80 0 80 HBsAb(-), HBeAb(-), HBcAb(-), HCVAb(-) 5 50 55 P \> 0.05 ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Comparison of the protein chip assay with ELISA test protocols (Group-3 and Group-4 serum samples) ::: Protein chip assay ELISA Total samples ---------------------------------------- ------- --------------- ---- HBsAb(-), HBeAb(-), HBcAb(-), HCVAb(+) 75 0 75 HBsAb(-), HBeAb(-), HBcAb(-), HCVAb(-) 5 50 55 P \> 0.05 ::: Furthermore, no reactivity was detected against HBeAg, HBcAg, HCVAg and negative control in Group-1, against HBsAg, HCVAg and negative control in Group-2, against HBsAg, HBeAg, HBcAg and negative control in Group-3, and against HBsAg, HBeAg, HBcAg HCVAg and negative control in Group-4, thus indicating the absence of cross reaction in the assays (Figure [6](#F6){ref-type="fig"}). ::: {#F6 .fig} Figure 6 ::: {.caption} ###### The demonstration of protein chips for detection of serum samples. a: The chip for detection of serum sample from Group-1 b: The chip for detection of serum sample from Group-2 c: The chip for detection of serum sample from Group-3 d: The chip for detection of serum sample from Group-4 Line 1: positive control spots Line 2: HBsAg spots Line 3: HBeAg spots Line 4: HBcAg spots Line 5: HCVAg spots Line 6: negative control spots The concentration of each protein spotted on the slides was 100 μg/ml. The silver enhancement time was 10 minutes. ::: ![](1471-2334-5-53-6) ::: Discussion ========== The preparation of colloidal nano-gold-SPA was a key process of the entire experiment. Many different routes to produce colloidal nano-gold particles have been reported \[[@B15]-[@B19]\]. The vast majority of processes involved the reduction of gold compounds in both aqueous and nonaqueous media. In our assay, we applied a versatile precipitation method capable to generate mono-dispersed, spherical nano-gold particles using aqueous solution of gold chloride and sodium citrate as the reducing agent. During the manipulation of preparing the nano-gold-SPA, pH of the gold solution was carefully controlled. Excessively acidic or basic conditions could easily cause the nano-gold-SPA to deposite. Colloidal nano-gold particles carried positive charges at acidic condition, allowing attachment of SPA through steric interaction. Therefore, adjusting pH to or above the isoelectric point of SPA could increase negative charges on SPA surface. In our assay, Tris buffer was undertaken to store the nano-gold-SPA. And the stability of nano-gold-SPA solution could be destroyed at room temperature or frozen condition. Four centigrade was becoming to store the nano-gold-SPA solution. Actually silver enhancement reaction processes rapidly. Silver ions in solution nucleate around nano-gold particles and precipitate as silver metal. Particle grows in size with time of development. It was of importance to control the reaction time because both too short and too long development might bring on darkened background. Besides, the distribution of nano-gold on the chip and the uniformity of nano-gold played a part in the enhancement result. Protein A is a cell wall component of Staphylococcus aureusthat binds specifically to the Fc portion of immunoglobulin from many mammals \[[@B20],[@B21]\]. And the binding does not interfere with antigen-antibody reaction. Also SPA can bind to colloidal nano-gold through static interaction, forming the labelling probes. In our assay, the well-prepared colloidal nano-gold-SPA could be preserved for over 1 year, keeping stable and active. Compared with horseradish peroxidase-conjugated antibodies and fluorescently labelled secondary antibodies \[[@B8],[@B9]\], the colloidal nano-gold-SPA was more economical and convenient, thus reducing the cost of the assay and simplifying the manipulation of the experiment. The common supports used for immobilization are either glass microscope slides or membranes; however, glass is preferred because it is more amenable to automation and exhibits a lower background signal \[[@B22]\]. To attach proteins to a solid substrate, the surface of the substrate has to be modified to achieve the maximum binding capacity. Coating the glass surface with poly-L-lysine (PLL) is a convenient method which is also applied in DNA microarrays \[[@B23]\]. The attached proteins are passively adsorbed to the surface in random orientation through non-specific interactions and can be washed off under stringent washing conditions \[[@B24]\]. There are several methods available for the surface activation of glass slides, many of which have been described \[[@B25]-[@B33]\]. Most methods introduced a chemical group onto the surface of the glass and then react with active groups of protein molecules. In our assay, the glass slides were activated by GOPTS. The preparation of the slides was convenient while the immobilization efficiency was high. The use of the chemically derivated slides coupled with printing solutions whose concentrations were optimized in our assay allowed us to generate visible results suitable for a study aimed at assessing the clinical performance of the protein chip assay. We compared the detection results tested by the protein chip assay and ELISA, respectively. In Group-1, 4 samples were detected as HBsAb negative. This might be due to the very low concentration of HBsAb in serum samples which could not be detected by the protein chip assay. The same applied to Group-2 and Group-3. Paired chi-square test indicated that no distinct difference (P \> 0.05) existed between the results determined by the two methods. While the protein chip assay could detect in parallel antibodies on one chip without non-specific reactions, thus it is more convenient and costs less time than ELISA. Moreover, it required small quantity of both samples and reagents, which made it economical and useful for those conditions when large quantities of samples were not easy to get. Furthermore, it is cheaper and safer than the fluorescent and isotope ones. The well-prepared protein chip could be preserved for over 1 year without losing of activity. Conclusion ========== The protein chip assay described is sensitive and specific, easy to perform, and could provide results in less than 40 minutes. In addition, several kinds of antibodies could be detected simultaneously without cross reaction. The results of this study suggest that there is potential for the application of our method in clinical diagnosis of various infectious diseases. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= Lianlian Duan and Yefu Wang were responsible for conception and design of the study. Lianlian Duan collected all data and wrote first draft of the manuscript. Zhixiang Wan and Jianxin Zhai were responsible for statistical analysis. Yefu Wang supervised the study and together with Shawn Shun-cheng Li participated in design and critical review of the manuscript. All authors have read and approved the manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2334/5/53/prepub> Acknowledgements ================ This work was supported by the Hubei Provincial R&D Program (Wuhan). Many thanks are also acknowledged to Xiehe Hospital (Wuhan) for supplying the serum samples used in this study.	PubMed Central	2024-06-05T03:55:59.965158	2005-7-6	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182366/", "journal": "BMC Infect Dis. 2005 Jul 6; 5:53", "authors": [ { "first": "Lianlian", "last": "Duan" }, { "first": "Yefu", "last": "Wang" }, { "first": "Shawn Shun-cheng", "last": "Li" }, { "first": "Zhixiang", "last": "Wan" }, { "first": "Jianxin", "last": "Zhai" } ] }

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.