Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2015 Jan 8.
Published in final edited form as: Clin Gastroenterol Hepatol. 2007 Jul 6;5(8):972–976. doi: 10.1016/j.cgh.2007.04.024

Association of variants of the interleukin-23 receptor (IL23R) gene with susceptibility to pediatric Crohn’s disease

Robert N Baldassano 1, Jonathan P Bradfield 2, Dimitri S Monos 3, Cecilia E Kim 2, Joseph T Glessner 2, Tracy Casalunovo 2, Edward C Frackelton 2, F George Otieno 2, Stathis Kanterakis 3, Julie L Shaner 2, Ryan M Smith 2, Andrew W Eckert 2, Luke J Robinson 2, Chioma C Onyiah 2, Debra J Abrams 1, Rosetta M Chiavacci 2, Robert Skraban 2, Marcella Devoto 4,5, Struan FA Grant 2,4,*, Hakon Hakonarson 2,4,*
PMCID: PMC4287202  NIHMSID: NIHMS29859  PMID: 17618837

Abstract

Background & Aims

Recently an association was demonstrated between the single nucleotide polymorphism (SNP), rs11209026, within the interleukin-23 receptor (IL23R) locus and Crohn’s disease (CD) as a consequence of a genome wide association study of this disease in adults. We examined the effects of this and other previously reported SNPs at this locus with respect to CD in children.

Methods

Utilizing data from our ongoing genome-wide association study in our cohort of 142 pediatric CD cases and 281 matched controls, we investigated the association of the previously reported SNPs at the IL23R locus with the childhood form of this disease.

Results

Using a Fisher’s exact test, the minor allele frequency (MAF) of rs1120902 in the cases was 1.75% while it was 6.61% in controls, yielding a protective odds ratio (OR) of 0.25 (95% CI 0.10 – 0.65; one-sided P = 9.2×10−4). Furthermore, of all the SNPs previously reported, rs11209026 was the most strongly associated. A subsequent family-based association test (which is more resistant to population stratification) with 65 sets of trios derived from our initial patient cohort yielded significant association with rs11209026 in a transmission disequilibrium test (one-sided P=0.0017). In contrast, no association was detected to the CARD15 gene for the IBD phenotype.

Conclusions

The OR of the IL23R variant in our pediatric study is highly comparable with that reported previously in a non-Jewish adult IBD case-control cohort (OR=0.26). As such, variants in IL23R gene confer a similar magnitude of risk of CD to children as for their adult counterparts.

Keywords: IL23R, gene, association, Crohn’s Disease

The inflammatory bowel diseases (IBDs), Crohn’s disease (CD) and ulcerative colitis (UC) are complex genetic disorders of the gastrointestinal tract that are recurring and chronic in nature. The exact etiology of the disease is unknown, but complex interplay between environmental risk factors and immunological components have been implicated in the triggering of the disease in those who are genetically susceptible to the disease. The hunt for genetic variants conferring risk to IBD has been ongoing for more than 10 years, with a number of susceptibility genes being implicated through linkage and candidate gene studies1–4, the most prominent and most consistently replicated being the caspase-recruitment domain 15 (CARD15) gene (formerly NOD2) 1, 2, 5 which encodes a protein involved in bacterial recognition by cells involved in innate immunity. However, CARD15 does not explain all cases of IBD, suggesting that other genetic factors may be involved.

Recently, a number of studies have reported remarkably strong, replicable associations with complex disease, including the complement factor H (CFH) gene in acute macular degeneration6–8 and the transcription factor 7-like-2 (TCF7L2) gene in type 2 diabetes9–11, with other such signals waiting for replication, such as insulin-induced gene 2 (INSIG2) in obesity12. In October this year, through the utilization of the Illumina HumanHap300 Genotyping BeadChip, Duerr et al 13 added to this repertoire by reporting a highly significant association between CD and the IL23R gene on chromosome 1p31. Specifically, an uncommon coding variant, rs11209026 (c.1142G>A, p.Arg381Gln), was shown to confer a strong protective effect against the disease and was then replicated in the same study in separate cohorts of patients with CD or ulcerative colitis.

Although these IL23R findings are compelling, there are continuing concerns regarding the performance of association studies in complex traits; as such independent replication efforts are now considered mandatory14. With the many errors and biases that can blight any individual study, replication by others can ensure that the original findings are robust and can also provide a more accurate estimate of the likely effect size15, 16.

In this study we demonstrate that the variant, rs11209026, in the IL23R gene confers similar magnitude of protection of CD in children as reported in adults with IBD, further suggesting that the IL23 pathway may be causally linked to CD in children.

Materials and Methods

Study Subjects

The case-control study consisted of 142 pediatric patients with CD and 281 matched controls. All subjects were biologically unrelated ethnic Caucasian. Patients were consecutively recruited from the Greater Philadelphia area from 2000 to 2006 at the Children’s Hospital of Philadelphia. 65 cases had both parents also recruited for purposes of family based association analysis.

The diagnosis of CD was based on standard criteria17: 1) symptoms of abdominal pain, diarrhea, rectal bleeding, perirectal disease or growth failure; 2) two or more episodes of symptoms separated > 8 weeks apart or ongoing symptoms of >6 weeks duration, and; 3) signs of inflammation from endoscopic, histologic or radiologic evaluations. Involvement of the following intestinal regions was noted in each patient: esophagus, stomach, small bowel, terminal ileum, colon and their combinations based on objective evidence of mucosal ulcerations, cobblestoning, intestinal strictures or bowel wall thickening. All subjects had ileal or ileal-colonic disease and over 90 % of patients had involvement of the terminal ileum. Children with ulcerative colitis were excluded.

This study was approved by the Institutional Review Board of the Children’s Hospital of Philadelphia.

Genotyping

We performed high throughput genome-wide SNP genotyping using the Illumina Infinium™ II HumanHap500 BeadChip technology18, 19. The Infinium™ II Assay protocol enables effective multiplexing and genome-wide SNP access through a single-base extension (SBE) method with enzymatic SNP scoring. The minimal hands-on three-day Infinium™ II protocol began with 750ng of input DNA. In day one, genomic DNA was amplified 1000–1500-fold. Day two, amplified DNA was fragmented ~300–600bp, then precipitated and resuspended followed by hybridization onto a BeadChip. SBE utilizes a single probe sequence ~50bp long designed to hybridize immediately adjacent to the SNP query site. Following targeted hybridization to the bead array, the arrayed SNP locus-specific primers (attached to beads) were extended with a single hapten-labeled dideoxynucleotide in the SBE reaction. The haptens were subsequently detected by a multi-layer immunohistochemical sandwich assay, as recently described18, 19. Our technicians followed the standard operating procedures issued by Illumina from the step that DNA is added to the single-tube whole genome amplification up to the processing of the BeadChip genotyping files. The Illumina BeadArray Reader scanned each BeadChip at two wavelengths and created an image file. As BeadChip images were collected, intensity values were determined for all instances of each bead type, and data files were created that summarized intensity values for each bead type. These files consisted of intensity data that was loaded directly into Illumina’s genotype analysis software, BeadStudio. A bead pool manifest created from the LIMS database containing all the BeadChip data was loaded into BeadStudio along with the intensity data for the samples. BeadStudio used a normalization algorithm to minimize BeadChip to BeadChip variability. Once the normalization was complete, the clustering algorithm was run to evaluate cluster positions for each locus and assign individual genotypes. Each locus was given an overall score based on the quality of the clustering and each individual genotype call was given a GenCall score. GenCall scores provided a quality metric that ranges from 0 to 1 assigned to every genotype called. GenCall scores were then calculated using information from the clustering of the samples. The location of each genotype relative to its assigned cluster determined its GenCall score.

The resources available for this project included the Illumina technology platform itself plus nine Tecan pipetting robotic systems, eight scanners, a laboratory information management system (LIMS) and automated allele-calling software. The workflow was robotic-based for automatic sample processing and included algorithms for quality control of genotypes. The facility infrastructure had sufficient computational power and servers for data processing and storing, including a series of computers that were integrated (warehouse setting) to perform continuous datamining of all gathered and generated datasets.

Analysis

Both the genetic matching of the cases and controls and the statistical tests for association were carried out using the software package plink (http://pngu.mgh.harvard.edu/~purcell/plink/index.shtml). Controls were genetically matched to the Caucasian CD cases by clustering of the pair-wise identity-by-state distances. Complete linkage agglomerative clustering was used with restrictions based on a pair-wise population concordance test so that one case was matched with two controls. Controls were matched with cases based on their pair-wise identity-by-state as long as a pair-wise population concordance test had a p-value > 0.001. In some limited cases, only one genetically matched control could be found based on the restrictions. A one-sided Fisher’s exact test was used to calculate p-values on allele count differences between 142 cases and 281 controls. Odds ratios and the corresponding 95% confidence intervals were calculated for the association analysis. A binomial exact transmission disequilibrium test was used to calculate one-sided p-values on differences between transmitted and untransmitted allele counts in 65 trios derived from the study cohort.

Results

Association between rs11209026 and pediatric CD risk

In this replication attempt, we genotyped 142 Caucasian pediatric CD cases (71 male, 71 female) and 281 matched controls (140 male, 141 female) recruited from the greater Philadelphia area with the Illumina HumanHap550 Genotyping BeadChip as part of our ongoing genome-wide association study of the disease. The diagnosis of CD was based on standard criteria (see Study Subjects section).

The probe for rs11209026 is present on our 550K BeadChip, so as a first pass we queried the data with a single test for this SNP to investigate if this marker is associated with CD in our pediatric cohort as reported by Duerr et al13. Using a Fisher’s exact test, we observed a significant protective effect of the minor A allele on the risk of CD. The minor allele frequency (MAF) in the cases was 1.75% while it was 6.61% in controls, yielding an odds ratio (OR) of 0.25 (95% CI 0.10 – 0.65; one-sided P = 9.2×10−4) – see Table 1. This OR is very much in line with that reported previously in the non-Jewish adult case-control cohort (OR=0.26) and is also clearly lower than that reported in the adult Jewish case-control cohort (OR=0.45)13.

Table 1.

Crohn’s disease (CD) case-control association study results for IL23R region markers with P-values < 0.0001 in Duerr et al13. Minor allele frequencies (MAF), allelic test one-sided P-values, and odds ratios (OR) with 95% confidence intervals (CI) are shown for each SNP. The ORs shown are for the minor allele. r2 values, derived from HapMap20, are shown with respect to each SNP’s linkage disequilibrium with rs11209026.

Marker Location MAF
(Cases n= 142)		 MAF
(Controls n= 281)		 One-sided P-value OR [95% CI] r2 to rs11209026
(HapMap)
rs1004819 intron 0.385 0.296 0.0060 1.49 [1.10,2.00] 0.01
rs7517847 intron 0.339 0.441 0.0026 0.65 [0.48,0.87] 0.013
rs10489629 intron 0.399 0.465 0.039 0.76 [0.57,1.02] 0.067
rs2201841 intron 0.378 0.326 0.077 1.26 [0.93,1.69] 0.027
rs11465804 intron 0.021 0.067 0.0024 0.30 [0.13,0.72] 0.881
rs11209026 Arg381Gln 0.017 0.066 9.2×10−4 0.25 [0.09,0.68] 1
rs1343151 intron 0.268 0.353 0.0073 0.67 [0.49,0.92] 0.111
rs10889677 exon-3′UTR 0.378 0.326 0.079 1.25 [0.93,1.69] 0.027
rs11209032 inter-genic 0.402 0.323 0.014 1.41 [1.05,1.90] 0.032
rs1495965 inter-genic 0.539 0.454 0.012 1.40 [1.05,1.87] 0.046
Open in a new tab

Duerr et al 13 also summarized all other SNPs in the locus that yielded a P<0.0001. As we had definitively replicated the association with rs11209026, we subsequently addressed these same SNPs to investigate their association with CD. The results in Table 1 indicate that the results are comparable to the original publication and that rs11209026 is indeed the most strongly associated SNP among those previously reported. In line with the original study13, many of these SNPs represent independent association signals, with six of the seven additional nominally significant SNPs having an r2 <0.2 with rs11209026 (based on HapMap data20), thus they cannot be fully explained by the most significant association with rs11209026 (Table 1).

Family-based association between individual IL23R polymorphisms and pediatric CD risk

We then looked at the data using a family-based association test; this test is more resistant to population stratification and presents us with the opportunity to confirm that the observations from our case-control analysis are accurate. Utilizing 65 sets of trios, derived from our initial patient cohort, we also observed significant association with rs11209026 in a transmission disequilibrium test (one-sided P=0.0017). In addition, investigating the other SNPs outlined in Table 1 with this family-based approach again revealed no other association stronger than that of this key SNP – see Table 2.

Table 2.

Family-based association results using 65 trios of SNPs previously investigated by Duerr et al13. Family-based association one-sided P-values were computed using a binomial exact transmission disequilibrium test. r2 values, derived from HapMap20, are shown with respect to each SNP’s linkage disequilibrium with rs11209026.

Marker Location One-sided P-value r2 to rs11209026
(HapMap)
rs1004819 intron 0.13 0.01
rs7517847 intron 0.11 0.013
rs10489629 intron 0.5 0.067
rs2201841 intron 0.39 0.027
rs11465804 intron 0.011 0.881
rs11209026 Arg381Gln 0.0032 1
rs1343151 intron 0.097 0.111
rs10889677 exon-3′UTR 0.39 0.027
rs11209032 inter-genic 0.21 0.032
rs1495965 inter-genic 0.22 0.046
Open in a new tab

Assessment of association to the CARD15 gene

Duerr et al 13 pointed out that the CARD15 variants on the BeadChip, rs2066843 and rs2076756, in fact produced the best association in their analysis. In our cohort, rs2066843 had a MAF of 30.4% in cases and 28.2% in controls OR=1.11, [95% CI (0.82 – 1.52; one-sided P = 0.27)] and rs2076756 had a MAF of 30.3% in cases and 25.9% in controls [OR=1.24 (95% CI 0.91 – 1.70; one-sided P = 0.10)].

Discussion

From an interim analysis of our ongoing genome-wide association study of pediatric CD, we have investigated variation in the IL23R locus previously reported to be associated with adult IBD13. Consequently, we have replicated association of this gene with CD by demonstrating its effect in the childhood form of the disorder. More specifically, the uncommon coding variant, rs11209026 (c.1142G>A, p.Arg381Gln), was shown to confer a strong protective effect against the disease with a highly comparable odds ratio (0.25) to that previously observed in an adult non-Jewish cohort (0.26). In addition, the association seen within the family-based association analysis discounts the possibility that the original or our findings had arisen as a result of population substructure.

Although the size of the cohort in the original genome wide association study was larger, consisting of 567 non-Jewish, European ancestry patients and 571 non-Jewish controls13, the amount of testing in our cohort was very restricted to a focused effort of specifically investigating if SNPs at a single locus also yield association in the same direction as previously reported. Our cohort is sufficiently powered to ask this straight-forward validation question, as the original finding impressively suggested that those individuals carrying the A allele of this SNP were approximately four-fold more protected from IBD that those carrying the G allele.

As the association we observe is indeed of a very similar magnitude to that of the original report, this independent replication confirms IL23R as a genuine IBD susceptibility gene. As such, the “winner’s curse” that is often seen for other complex trait susceptibility genes21 is not observed in this instance. What is of particular note is that the association observed in children is almost identical to that of adults. With the gene-environment interaction22 models in mind, we have been motivated to look at the genetics of childhood disease in order to more readily distill the genetic component in these phenotypes due to the fact that environmental exposure and impact has been for a relatively short period of their lifetime. However, with the magnitude of the association being so comparable between children and adults, this particular research outcome suggests that the environmental interaction with this variant over time is negligible and in fact this variant may be primarily associated with early onset IBD. This may be further supported by the fact that the CARD15 variants failed to show association in our pediatric cohort (and also debated by others23) as opposed to significant association in the original genome-wide study in adults13. While the lack of association with the CARD15 variants may be related to power issues, it is clear that the IL23R association is readily detected in our pediatric cohort.

Our results lend further support for the protective role of the IL23R gene in IBD, suggesting that interventions at the IL23 pathway level may be of value in patients who suffer from this disease. The IL23R gene encodes a subunit of the receptor for the pro-inflammatory cytokine, interleukin-23. There are at least six alternatively spliced forms of the mRNA that in turn lead to diverse isoforms of the receptor protein24. The variants that we observe association to may directly dictate splicing or some other regulatory mechanism but more likely are in linkage disequilibrium with the causative variant(s).

Once our genome-wide association study is complete, we will have the opportunity to look for other variants in the genome that are associated with IBD, both CD and ulcerative colitis, as a consequence of our use of a higher resolution BeadChip. In addition, we will explore the IL23R gene further to elucidate other potential variants that may confer genetic susceptibility to this debilitating disorder in our cohort.

Acknowledgments

We would like to thank Alejandrina Estevez, Kenya Fain, Pamela Kline, Erin Santa, Alexandria Thomas and LaShea Williams for their expert assistance with the data collection and management. We would also like to thank Smari Kristinsson, Larus Arni Hermannsson and Asbjörn Krisbjörnsson of Raförninn ehf for their extensive software design and contribution. This research was financially supported by the Children’s Hospital of Philadelphia, the Edmunds Fund, the Heineman Foundation, the IBD Family Research Council and NIH grant for the General Clinical Research Center (GCRC) 5-MO1-RR-000240.

Abbreviations used in this paper

CD

Crohn’s Disease

IBD

Inflammatory Bowel Disease

SNP

single nucleotide polymorphism

IL23R

interleukin-23 receptor

OR

odds ratio

CI

confidence interval

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  • 1.Hugot JP, Chamaillard M, Zouali H, et al. Association of NOD2 leucine-rich repeat variants with susceptibility to Crohn’s disease. Nature. 2001;411(6837):599–603. doi: 10.1038/35079107. [DOI] [PubMed] [Google Scholar]
  • 2.Ogura Y, Bonen DK, Inohara N, et al. A frameshift mutation in NOD2 associated with susceptibility to Crohn’s disease. Nature. 2001;411(6837):603–6. doi: 10.1038/35079114. [DOI] [PubMed] [Google Scholar]
  • 3.Peltekova VD, Wintle RF, Rubin LA, et al. Functional variants of OCTN cation transporter genes are associated with Crohn disease. Nature Genetics. 2004;36(5):471–5. doi: 10.1038/ng1339. [DOI] [PubMed] [Google Scholar]
  • 4.Stoll M, Corneliussen B, Costello CM, et al. Genetic variation in DLG5 is associated with inflammatory bowel disease. Nature genetics. 2004;36(5):476–80. doi: 10.1038/ng1345. [DOI] [PubMed] [Google Scholar]
  • 5.Radford-Smith G, Pandeya N. Associations between NOD2/CARD15 genotype and phenotype in Crohn’s disease-Are we there yet? World J Gastroenterol. 2006;12(44):7097–103. doi: 10.3748/wjg.v12.i44.7097. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Klein RJ, Zeiss C, Chew EY, et al. Complement factor H polymorphism in age-related macular degeneration. Science. 2005;308(5720):385–9. doi: 10.1126/science.1109557. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Edwards AO, Ritter R, 3rd, Abel KJ, Manning A, Panhuysen C, Farrer LA. Complement factor H polymorphism and age-related macular degeneration. Science. 2005;308(5720):421–4. doi: 10.1126/science.1110189. [DOI] [PubMed] [Google Scholar]
  • 8.Haines JL, Hauser MA, Schmidt S, et al. Complement factor H variant increases the risk of age-related macular degeneration. Science. 2005;308(5720):419–21. doi: 10.1126/science.1110359. [DOI] [PubMed] [Google Scholar]
  • 9.Grant SF, Thorleifsson G, Reynisdottir I, et al. Variant of transcription factor 7-like 2 (TCF7L2) gene confers risk of type 2 diabetes. Nature Genetics. 2006;38(3):320–3. doi: 10.1038/ng1732. [DOI] [PubMed] [Google Scholar]
  • 10.Florez JC, Jablonski KA, Bayley N, et al. TCF7L2 polymorphisms and progression to diabetes in the Diabetes Prevention Program. The New England Journal of Medicine. 2006;355(3):241–50. doi: 10.1056/NEJMoa062418. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Zeggini E, McCarthy MI. TCF7L2: the biggest story in diabetes genetics since HLA? Diabetologia. 2006 doi: 10.1007/s00125-006-0507-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Herbert A, Gerry NP, McQueen MB, et al. A common genetic variant is associated with adult and childhood obesity. Science. 2006;312(5771):279–83. doi: 10.1126/science.1124779. [DOI] [PubMed] [Google Scholar]
  • 13.Duerr RH, Taylor KD, Brant SR, et al. A genome-wide association study identifies IL23R as an inflammatory bowel disease gene. Science. 2006;314(5804):1461–3. doi: 10.1126/science.1135245. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Patterson M, Cardon L. Replication publication. PLoS Biology. 2005;3(9):e327. doi: 10.1371/journal.pbio.0030327. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Page GP, George V, Go RC, Page PZ, Allison DB. Are we there yet? Deciding when one has demonstrated specific genetic causation in complex diseases and quantitative traits. American Journal of Human Genetics. 2003;73(4):711–9. doi: 10.1086/378900. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Hirschhorn JN, Lohmueller K, Byrne E, Hirschhorn K. A comprehensive review of genetic association studies. Genet Med. 2002;4(2):45–61. doi: 10.1097/00125817-200203000-00002. [DOI] [PubMed] [Google Scholar]
  • 17.Walker-Smith JA, Sanderson IR. Diagnostic Criteria for Chronic Inflammatory Bowel Disease in Childhood. In: Bistrian BR, Walker-Smith JA, editors. Inflammatory Bowel Diseases. Vol. 2. Nestlé Nutrition Workshop Series Clinical & Performance Programme, Nestec Ltd; Vevey/S Karger AG, Basel: 1999. pp. 107–20. [DOI] [PubMed] [Google Scholar]
  • 18.Gunderson KL, Steemers FJ, Lee G, Mendoza LG, Chee MS. A genome-wide scalable SNP genotyping assay using microarray technology. Nature genetics. 2005;37(5):549–54. doi: 10.1038/ng1547. [DOI] [PubMed] [Google Scholar]
  • 19.Steemers FJ, Chang W, Lee G, Barker DL, Shen R, Gunderson KL. Whole-genome genotyping with the single-base extension assay. Nat Methods. 2006;3(1):31–3. doi: 10.1038/nmeth842. [DOI] [PubMed] [Google Scholar]
  • 20.A haplotype map of the human genome. Nature. 2005;437(7063):1299–320. doi: 10.1038/nature04226. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Lohmueller KE, Pearce CL, Pike M, Lander ES, Hirschhorn JN. Meta-analysis of genetic association studies supports a contribution of common variants to susceptibility to common disease. Nature genetics. 2003;33(2):177–82. doi: 10.1038/ng1071. [DOI] [PubMed] [Google Scholar]
  • 22.Hemminki K, Lorenzo Bermejo J, Forsti A. The balance between heritable and environmental aetiology of human disease. Nature Reviews. 2006;7(12):958–65. doi: 10.1038/nrg2009. [DOI] [PubMed] [Google Scholar]
  • 23.Ferraris A, Knafelz D, Torres B, Fortina P, Castro M, Dallapiccola B. Analysis of CARD15 gene variants in Italian pediatric patients with inflammatory bowel diseases. The Journal of Pediatrics. 2005;147(2):272–3. doi: 10.1016/j.jpeds.2005.03.039. [DOI] [PubMed] [Google Scholar]
  • 24.Zhang XY, Zhang HJ, Zhang Y, et al. Identification and expression analysis of alternatively spliced isoforms of human interleukin-23 receptor gene in normal lymphoid cells and selected tumor cells. Immunogenetics. 2006;57(12):934–43. doi: 10.1007/s00251-005-0067-0. [DOI] [PubMed] [Google Scholar]

RESOURCES