We have established an international consortium for the purpose of identifying and obtaining DNA specimens from at least 1000 well-characterized cases of sJIA and a collection of ethnically-matched controls. This consortium includes investigators from several of the major pediatric rheumatology centers in North America, including the Cincinnati Children's Hospital Medical Center, the University of Utah, Stanford University, and the Hospital for Sick Children in Toronto;from several of the major pediatric rheumatology centers in Europe, including Great Ormand Street Hospital in London, University Children's Hospital in Muenster, and the Gaslini Institute in Genova;and other centers and study groups from the United States, United Kingdom, Germany, Turkey, Spain, Argentina and Brazil. In May of 2010, we hosted a meeting of international collaborators at the NIH, in which there were discussions regarding recruitment strategies, appropriate informed consent for genome-wide association studies (GWAS), sample shipment, genotyping, and data analysis. Because of the difficulties for some centers to recruit appropriate ethnically matched controls, where necessary we planned to obtain in silico controls from existing online databases, using bioinformatic tools to ensure appropriate matching by ancestry. All genotyping has been performed in the Inflammatory Disease Section of the NHGRI, using Illumina Human Omni 1M-Quad, v1.0 bead chips and an Illumina iScan Beadarray scanner. Altogether, we have generated single nucleotide polymorphism (SNP) genotypes on 988 children with sJIA and 431 healthy control subjects. These data were combined with SNP genotypes, in silico, from 7579 additional healthy control subjects. After dividing the dataset into 9 strata by country of origin, we excluded samples and markers that failed to meet quality control standards. Haplotype phasing, SNP imputation, and association testing were performed independently in each stratum, and we then subjected the association results from greater than 1.6 million SNPs to meta-analyses. A second round of more intensive imputation employing a more densely genotyped set of reference haplotypes was performed in each region with a minimal P value less than 10 to the negative seventh. Regions with association signals exceeding genomewide significance (P less than 5 times 10 to the negative eighth) were further evaluated with logistic regression and conditional analysis. Genome wide meta-analysis of sJIA identified disease associations with two regions, the major histocompatibility complex (MHC) locus and an intergenic region on chromosome 1, each of which were subjected to a second round of SNP imputation. In both regions, meta-analyses of the second round of imputation data identified significant associations. Meta-analyses of the MHC locus identified three strong association signals, the strongest of which is located between BTNL2 and HLA-DRA (P = 7.1 times ten to the negative fifteenth, odds ratio = 2.2), a second located nearest to HLA-DQA1 (P = 4.5 times ten to the negative eleventh, odds ratio = 1.8), and a third centered around HLA-DRB1 (p = 1.6 times ten to the negative tenth, odds ratio = 1.5). Univariate regression controlling for the strongest of these signals revealed that variants near HLA-DQA1 influenced sJIA risk, independent of the primary regional association signal. On chromosome 1, we identified an sJIA-associated cluster of 9 SNPs (P = 5.4 times 10 to the negative ninth, odds ratio = 2.0) that was nearest to LOC284661, encoding a long intergenic noncoding RNA. By cross referencing data from the ENCODE project with the 9 sJIA-associated SNPs, we found evidence of important regulatory sites in close proximity to the sJIA-associated SNP haplotype. Furthermore, several of the sJIA-associated SNPs were located within known histone marks or transcription factor binding sites. To better understand the contribution of the MHC locus to sJIA, we employed HLA imputation, a process where SNP data from individuals is used to predict with high accuracy the classical HLA types of those individuals. Using this methodology, we determined the classical HLA types of 6 of the study populations. Association meta-analysis of these data revealed a strong association between sJIA and the MHC class II (MHC-II) allele, HLA-DRB1*11, which influences sJIA risk in all 6 study populations. Moreover, haplotype analysis of the U.K. stratum, the largest subpopulation of the study, identified a risk haplotype that includes the MHC-II alleles HLA-DRB1*11, HLA-DQA1*05, and HLA*DQB1*03. Overall, these data implicate the MHC-II loci, HLA-DR and HLA-DQ, in the pathogenesis of sJIA. While the intergenic location of the most strongly associated variants raises the possibility that an alteration of HLA-DR regulation or expression may underlie its role in sJIA, the strong association of HLA-DRB1*11 with sJIA may also indicate a role for antigen presentation to helper T-cells. The data implicating a long intergenic noncoding RNA is particularly intriguing, since these molecules may play an important role in coordinate regulation of multiple functionally related genes. A manuscript describing these findings is nearly complete. To further investigate these findings we will embark on a targeted deep resequencing study of 100 candidate genes implicated by this GWAS study, by previous genetic studies of sJIA, or by their known roles in other phenotypically similar inflammatory diseases. In addition to investigating sJIA, we have also undertaken investigations of a second genetically-complex autoinflamatory disease, Behcet's disease (BD). In collaboration with members of Dr. Daniel Kastner's group at NHGRI (and others), we have performed a series of studies of an extremely well-characterized and meticulously assembled BD case-control population from Turkey that have resulted in several high-profile publications in previous years. During the current reporting period, we have completed a follow-up study designed to better understand the mechanism through which the predominant BD risk factor, the MHC class I (MHC-I) molecule, HLA-B*51, influences disease risk. To do this, we employed SNP data from our previous GWAS to perform imputation of classical MHC types and their subordinate amino acid positions. Using logistic regression analyses, we identified multiple independent BD-associated MHC-I alleles, including both risk alleles (HLA-B*51, HLA-B*15, HLA-B*57, HLA-B*27) and protective alleles (HLA-A*03, HJLA-B*49). Moreover, we identified a group of seven amino acid positions within MHC-I molecules that independently influence BD risk. Six of these residues are located in and around the antigen binding groove of the MHC-I molecule, and one is located in the MHC-I signal peptide. The shape and size of the antigen binding groove determines which peptides can be bound by a given MHC-I molecule, and in turn, presented to cytotoxic T-cells. In addition to interacting with the T-cell receptors (TCR) of cytotoxic T-cells, MHC-I molecules also engage killer immunoglobulin-like receptors (KIR) on both T-cells and natural killer (NK) cells, acting to regulate these cytotoxic cell types. The BD-associated residues of the antigen binding groove are known to play roles in defining the interactions between peptide/MHC-I complexes and both TCRs and KIRs. The MHC-I signal peptide is also known to regulate cytotoxic cells through another family of receptors, the C-type lectin like receptors, which act independently from the antigen binding groove. This is the first evidence linking the MHC-I signal peptide to the pathogenesis of BD. Taken together, these data implicate the regulation of cytotoxic cells in the pathogenesis of BD. These results were recently published in Proceedings of the National Academy of Sciences USA.
Showing the most recent 10 out of 16 publications
