Th17-Related Genes and Celiac Disease Susceptibility

Th17 cells are known to be involved in several autoimmune or inflammatory diseases. In celiac disease (CD), recent studies suggest an implication of those cells in disease pathogenesis. We aimed at studying the role of genes relevant for the Th17 immune response in CD susceptibility. A total of 101 single nucleotide polymorphisms (SNPs), mainly selected to cover most of the variability present in 16 Th17-related genes (IL23R, RORC, IL6R, IL17A, IL17F, CCR6, IL6, JAK2, TNFSF15, IL23A, IL22, STAT3, TBX21, SOCS3, IL12RB1 and IL17RA), were genotyped in 735 CD patients and 549 ethnically matched healthy controls. Case-control comparisons for each SNP and for the haplotypes resulting from the SNPs studied in each gene were performed using chi-square tests. Gene-gene interactions were also evaluated following different methodological approaches. No significant results emerged after performing the appropriate statistical corrections. Our results seem to discard a relevant role of Th17 cells on CD risk.


Introduction
Celiac disease (CD) is an immune related disease mainly characterized by intestinal inflammation after gluten ingestion in genetically susceptible individuals. CD has been traditionally considered a Th1-mediated disease. However, accumulating evidence about the relevant role of the novel Th17 immune response in several autoimmune diseases [1] opened the possibility towards an involvement of this immunological pathway in CD pathogenesis. These cells seem to be involved in protective responses against extracellular pathogens but they can contribute to chronic inflammation and autoimmunity when dysregulated.
Studies based on murine models of several autoimmune diseases, as multiple sclerosis (autoimmune encephalomyelitis, EAE), rheumatoid arthritis (collagen-induced arthritis, CIA) and inflammatory bowel disease (experimental colitis), provided the first evidence about a role of Th17 cells in those conditions [3,4]. This idea was later supported by case-control studies, which associated genetic variants in IL23R with susceptibility to Crohn's disease, psoriasis and ankylosing spondylitis [5,6,7]. Nowadays, the Th17 immune response is considered as a relevant player in several autoimmune or inflammatory diseases. IL-17 mRNA or protein have been detected in biological fluids or the specific affected tissue in several autoimmune disorders [8] and genetic studies associated genes coding important Th17 related products with several diseases [9]. In addition, epistasis between IL23R and other Th17 related genes has been reported: with IL2/IL21 in UC [10] and with IL17A and IL17RA in Crohn's disease [11].
In 2008, a putative implication of the Th17 immune response in CD pathogenesis was suggested from two studies following different approaches. Our research group detected a significant association between a genetic polymorphism in the IL23R gene and CD [12] and Harris et al. found higher production of IL-23 after stimulation of human monocytes derived from CD patients with peptic fragments of wheat gliadin [13]. Subsequently, genetic linkage with the IL23R region was observed in Finnish families, although this result was not replicated in Hungarian pedigrees and no association with IL23R polymorphisms was observed in Finnish, Hungarian or Italian CD samples [14]. In addition, increased expression of several Th17-related cytokines or products was detected in patients with active CD [15,16] and very recently, gluten-specific IL-17A-producing cells have been found in the duodenum of CD patients [17], which supports a role of Th17 cells in CD pathogenesis.
Despite these results observed in CD, the role of the Th17 cells on this disease is still not well defined. We aimed at shedding more light upon this issue by performing an extensive genetic study including many genes coding distinctive cytokines, markers or transcription factors involved in the Th17 response. We will evaluate the individual influence of those genes on CD susceptibility and also the possible contribution of gene-gene interactions. Previous genome wide association studies (GWAS) did not find association with CD susceptibility of any Th17-related gene [18,19,20] (with exception of the IL2/IL21 locus, also involved in other processes), but we consider that a different scenario could emerge with this study: we cover most of the variability present in the studied genetic regions and we will evaluate the genetic interactions between the included polymorphisms, which has been proved as a valid approach to detect new susceptibility variants [21,22].

Ethics Statement
This study was approved by the ethical committee (CEIC) of the Hospital Clínico San Carlos. Samples were obtained after obtaining written informed consent.

Subjects
A total of 735 CD patients and 549 ethnically matched healthy controls were included in the initial study. A second sample set consisting of 294 CD patients and 475 controls was used for additional analysis. All these samples correspond to unrelated Spanish white individuals. CD patients were diagnosed following the European Society for Paediatric Gastroenterology and Nutrition (ESPGAN), 97% are positive for HLA-DQ2 and/or HLA-DQ8. Controls correspond mainly to blood donors and laboratory staff. CD samples were consecutively collected in two centres of the same region (Hospital La Paz and Hospital Clínico San Carlos, Madrid) and controls were collected at the Hospital Clínico San Carlos.

Markers and genotyping
We selected genes with a known functional role in the Th17 immune response. Accordingly, sixteen genes were studied: IL23R, RORC, IL6R, IL17A, IL17F, CCR6, IL6, JAK2, TNFSF15, IL23A, IL22, STAT3, TBX21, SOCS3, IL12RB1 and IL17RA. For all these genes except IL6R, JAK2 and STAT3, single nucleotide polymorphisms (SNPs) were selected following the ''aggressive tagging'' option present in the Haploview program with genetic data downloaded from the HapMap Project (http://hapmap.ncbi.nlm. nih.gov) (50 kb upstream and downstream of the transcription initiation site). To increase statistical power, only markers with a minor allele frequency (MAF).10% were included. In addition, SNPs located in those genes which code nonsynonymous changes or were previously associated with some autoimmune disease were also analysed independently of their MAF. In STAT3 and JAK2, only two SNPs previously associated with Crohn's disease, which share some susceptibility factors with CD, were included: rs744166 and rs10758669, respectively; and in IL6R we studied one functional polymorphism, rs8192284. SNPs located in IL6 and IL6R and two SNPs in IL23R, rs11209026 and rs7517847, were analysed in previous works [12,23], which included most of the samples initially analysed in this study, but their data were used to evaluate genetic interactions with other Th17 related genes.
A total of 101 SNPs were initially studied ( Table S1). All of them were genotyped by Veracode technology performed at the National Genotyping Center (http://www.cegen.org), except those that failed (rs10494269, rs9395767, rs608137, rs6927645, rs273506 and rs2241044) and those located in the IL6, IL6R and TBX21 genes, which were genotyped with specific TaqMan assays. Two SNPs (rs11209026 and rs7517847, both in the IL23R gene) were genotyped by those two technologies and identical results were obtained.
Additional analysis included the study by TaqMan technology of rs12070470, in the IL23R gene.

Statistical analysis
Deviations from Hardy-Weinberg proportions were assessed in all the SNPs studied.
A case-control analysis using chi-square tests was performed for each individual SNP and for the haplotypes resulting from the SNPs studied in the same genetic region.
Interactions between genes were evaluated following four different approaches: logistic regression, random forests (RF), classification and regression trees (CART) and multifactor dimensionality reduction (MDR).
The comparison of genotypic frequencies between cases and controls for all the SNPs analysed achieved a nominal significant value in twelve polymorphisms located in eight different genetic regions (Table 1). Although none of them withstand Bonferroni correction, we tried to replicate associations involving SOCS3 and IL23R using a second sample set. These two genes show the lowest case-control p-values in the present analysis and additionally some SNP in those genes showed a nominal significance in previous CD GWAS [20].
The initial IL23R data analysis also evidenced one haplotype significantly associated with CD susceptibility (rs4655683-rs10889667-rs1569922-rs790632-rs7517847-rs10489629-rs7528924-rs2201841-rs4655530-rs11209026-rs6682033-rs6693831, G-C-C-C-T-A-G-T-A-G-A-C): 9.2% in CD patients vs. 6.3% in controls (p = 0.0067). For replication purposes, the SNP rs12070470, highly correlated with that haplotype (r 2 = 1 according to http://hapmap. ncbi.nlm.nih.gov/) was studied in the second sample set instead of the 12 SNPs initially considered. No significant associations involving IL23R were observed in the replication set. Regarding the SNP rs4969170, in the SOCS3 gene, a significant association was observed pooling the original and the replication sets: p = 0.0012 OR = 0.64 95% CI 0.49-0.84 (Table 2). Statistical power limitations probably precluded us to obtain a significant result in the replication set.
No consensus exists as to the best methodology to evaluate epistasis. Therefore, we used four different statistical methods to evaluate genetic interactions between all the studied polymorphisms located in different genes. We did not find statistically significant results with any methodological approach.

Discussion
With the development of genome wide association studies (GWAS), the number of discovered genes involved in CD susceptibility has highly increased. However, the percentage of disease heritability explained has not experienced such an increase. Genetic variants not included in GWAS and genetic interactions could be underlying some missing heritability. We bear this in mind when studying the relevance of the Th17 immune response on CD susceptibility. We performed an extensive case-control study including sixteen genes which code relevant factors involved in that immune response. Tag SNPs were selected to cover most of the variability present in each gene, with exception of IL6R, STAT3 and JAK2. SNPs coding nonsynonymous changes or those previously associated with other autoimmune diseases were also included in order to increase the a priori probability of obtaining a significant result. Additionally, we evaluated the possibility that interactions between the studied genes were involved in disease susceptibility. Our results seem to discard a relevant role of Th17 cells on CD risk, since no significantly associated SNP or gene-gene interaction was consistently observed, with the only exception of rs4969170, located in SOCS3, which deserves further research. However, although SOCS3 is later confirmed, its functional role must be elucidated, since it is involved in different functional pathways and it would be expected that more than one Th17 gene was associated with CD susceptibility, as it has been observed with other Th17mediated diseases.
The discovery of the IL-23 cytokine prompted the reexamination of the dominant Th response in many autoimmune diseases, primarily in those considered as skewed towards a Th1 phenotype. Studies based on murine models of multiple sclerosis, rheumatoid arthritis and inflammatory bowel disease related these conditions with a Th17 response. However, a similar conclusion is not drawn from GWAS results [24,25,26]. Although several Th17related genes have been associated with IBD and RA, the list of MS susceptibility genes does not suggest a Th17 related etiology. This intriguing issue is probably far away from being answered. Nowadays it seems clear that Th immune responses are not independent and plasticity exists between Th cell subsets. A shift between Th1 and Th17 can occur during the inflammatory process and it is possible to speculate that the relative contribution and the timing of each subset will determine which genes would be involved in disease risk. Moreover, the cytokine microenvironment can determine the shift towards a specific immune response. From this point of view, genetics could not be so relevant if other compensatory mechanisms exist. This evidence, as previously suggested, that overlap between autoimmune diseases must be observed with caution. Th17 cells seem to mediate several autoimmune diseases but their impact in disease etiology seems to be different.
In summary, gene expression studies link CD pathogenesis to Th17 cells, but we evidenced that polymorphisms in Th17-related genes do not seem to be crucial for disease development. This is concordant with observations on MS. Although, in general, genetic data provide clues that ratified by functional studies unravel disease pathogenesis, this time it makes necessary to do somehow the other way around, with the special difficulty of explaining the divergent genetic results observed in different immune mediated diseases. Therefore, much more work is expected in this field.

Supporting Information
Table S1 Genes and SNPs studied ordered by chromosome and position. (DOC)