PW01-038 – Genomewide association study of Still’s disease

Methods We generated single nucleotide polymorphism (SNP) genotypes from the genomic DNA of 988 children with sJIA and 514 healthy control subjects. These data were combined with SNP genotypes in silico from 7370 additional healthy control subjects. After dividing the dataset into 9 strata by country of origin, we excluded samples and markers that did not meet our strict quality requirements. We performed haplotype phasing with ShapeIT, SNP imputation with IMPUTE2, and association testing with SNPTEST independently in each stratum. The results of association testing were subjected to fixedand randomeffects meta-analyses with GWAMA. A second round of more intensive “deep imputation” was performed in each region with pmeta<1E-7. Using the directly genotyped SNP data, we used imputation to deduce classical HLA types in each stratum. Significant associations were further evaluated with multivariate logistic regression using SNPTEST and SNP & Variation Suite 7. Results Using the above method, we ultimately tested a panel of over 1.6M SNPs for association with sJIA. Using metaanalysis of SNP association data from 9 strata, we identified 2 sJIA-associated regions that exceeded the stringent threshold for genome wide significance (pmeta<5E-8). The strongest association was located in the major histocompatibility complex locus, with one SNP nearest to HLADRB1 (rs112638393: pmeta=1.6E-10, OR 1.5 [1.3, 1.7]) and a second located nearest to BTNL2 (rs115945836: pmeta=2.4E-10, OR 2.8 [2.0, 3.9]). Conditioning on the effect of rs112638393 accounted for the majority of the effect around HLA-DRB1, while revealing a significant, independent association signal spanning BTNL2 and HLA-DRA. Additionally, meta-analysis of the imputed HLA type associations from 8 strata revealed a strong association between HLA-DRB1*1101 and sJIA (pmeta=1.2E-8, OR 2.1 [1.6, 2.7]). The second strongest regional association, which also exceeded genome wide significance, was located on Chr 1 nearest to LOC284661 (rs16838915: pmeta=5.4E-9, OR 2.0 [1.6, 2.5]). Logistic regression analysis demonstrated no residual association signal in this region after conditioning on rs16838915. In total, our study identified 11 loci that were suggestive of association with sJIA (p<5E-5).