Meta-analysis of genome-wide association studies in celiac disease and rheumatoid arthritis identifies fourteen non-HLA shared loci

Abstract

Epidemiology and candidate gene studies indicate a shared genetic basis for celiac disease (CD) and rheumatoid arthritis (RA), but the extent of this sharing has not been systematically explored. Previous studies demonstrate that 6 of the established non-HLA CD and RA risk loci (out of 26 loci for each disease) are shared between both diseases. We hypothesized that there are additional shared risk alleles and that combining genome-wide association study (GWAS) data from each disease would increase power to identify these shared risk alleles. We performed a meta-analysis of two published GWAS on CD (4,533 cases and 10,750 controls) and RA (5,539 cases and 17,231 controls). After genotyping the top associated SNPs in 2,169 CD cases and 2,255 controls, and 2,845 RA cases and 4,944 controls, 8 additional SNPs demonstrated P<5 × 10(-8) in a combined analysis of all 50,266 samples, including four SNPs that have not been previously confirmed in either disease: rs10892279 near the DDX6 gene (P(combined) = 1.2 × 10(-12)), rs864537 near CD247 (P(combined) = 2.2 × 10(-11)), rs2298428 near UBE2L3 (P(combined) = 2.5 × 10(-10)), and rs11203203 near UBASH3A (P(combined) = 1.1 × 10(-8)). We also confirmed that 4 gene loci previously established in either CD or RA are associated with the other autoimmune disease at combined P<5 × 10(-8) (SH2B3, 8q24, STAT4, and TRAF1-C5). From the 14 shared gene loci, 7 SNPs showed a genome-wide significant effect on expression of one or more transcripts in the linkage disequilibrium (LD) block around the SNP. These associations implicate antigen presentation and T-cell activation as a shared mechanism of disease pathogenesis and underscore the utility of cross-disease meta-analysis for identification of genetic risk factors with pleiotropic effects between two clinically distinct diseases.

Figure 1. Established CD and RA SNPs and their association across diseases.

(A) Known CD SNPs in RA. The figure represents OR and CI for the established CD SNPs (p<5×10⁻⁸, one SNP per locus) in RA meta-analysis (5,539 auto-anbitody positive cases and 17,231 controls). (B) Known RA SNPs in CD. The figure represents OR and CI for the established RA SNPs (p<5×10⁻⁸, one SNP per locus) in CD meta-analysis (4,533 cases and 10,750 controls). For the six shared loci established in both diseases, figure 1A includes the top CD SNP and figure 1B the top RA SNP. From six shared loci, three (TNFRSF14, IL2/IL21 and TNFAIP3) are associated with same SNP or a good proxy (r²>0.9) in both diseases; in other three loci – CTLA4, REL and TAGAP – the most associated SNPs in CD and RA are not in strong LD with each other (r²<0.3), which is reflected in moderate association (CTLA4) or no association (REL) of these SNPs in the second disease. The TAGAP SNPs show association to opposite alleles in CD and RA.

Figure 2. QQ plot of CD associated SNPs in RA and RA associated SNPs in CD.

QQ plot of CD associated SNPs (p<0.001) in RA (green) and RA associated SNPs (p<0.001) in CD (black). The most strongly associated SNPs (after removing known risk loci) in one disease were further filtered for P<0.001, and the resulting LD-pruned SNP sets were then tested for their distribution of association in the other disease. The QQ-plots indicate excess sharing of moderately associated SNPs across CD and RA.

Figure 3. QQ plot of CD-RA meta-analysis by directional method and opposite allelic effect.

CD-RA inverse variance weighted meta-analysis assuming allelic effects in the same direction in the two diseases (panel A) and opposite allelic effects (panel B). Black – all loci except the MHC region (chr. 6: 20–40 Mb). Green – all loci except MHC and established CD and RA regions (1 MB around previously validated SNPs excluded).