Fine-mapping and functional studies highlight potential causal variants for rheumatoid arthritis and type 1 diabetes

Abstract

To define potentially causal variants for autoimmune disease, we fine-mapped^1,2 76 rheumatoid arthritis (11,475 cases, 15,870 controls)³ and type 1 diabetes loci (9,334 cases, 11,111 controls)⁴. After sequencing 799 1-kilobase regulatory (H3K4me3) regions within these loci in 568 individuals, we observed accurate imputation for 89% of common variants. We defined credible sets of ≤5 causal variants at 5 rheumatoid arthritis and 10 type 1 diabetes loci. We identified potentially causal missense variants at DNASE1L3, PTPN22, SH2B3, and TYK2, and noncoding variants at MEG3, CD28-CTLA4, and IL2RA. We also identified potential candidate causal variants at SIRPG and TNFAIP3. Using functional assays, we confirmed allele-specific protein binding and differential enhancer activity for three variants: the CD28-CTLA4 rs117701653 SNP, MEG3 rs34552516 indel, and TNFAIP3 rs35926684 indel.

Fig. 1 ∣. Imputation accuracy and quality of datasets.

Datasets were imputed with different reference panels: the European subpopulation of 1000 Genomes (EUR), full 1000 Genomes (COSMO), full 1000 Genomes imputed with PBWT (COSMO(PBWT)), and HRC. a, We sequenced 799 1-kilobase regions in 568 individuals with ImmunoChip genotypes and called 1,854 common (MAF>1%) variants. We calculated the imputation accuracy (genomic R²) by correlating imputed genotypes using each reference panel with genotypes called from the sequencing experiment. b, INFO scores for each reference panel in the rheumatoid arthritis dataset.

Fig. 2 ∣. Variants in the 95% credible sets of significant loci determined by the Bayesian factor.

a, The inner ring of dots indicates whether the locus has ≤10 variants in the credible set and a significant association signal (filled circles). Comb., combined; RA, rheumatoid arthritis. The middle ring shows variants in each credible set. Highlighted segments indicate loci with a candidate causal variant. The color intensity indicates the posterior probability (PP), gray representing a lack of significance. The outer ring shows indel, promoter, and missense coding annotation for each variant in the credible set. b, Number of variants in the 95% credible sets within significant loci. We narrowed down the list of probable causal variants to ≤5 in 5 out of 20 significant RA loci, and 10 out of 34 significant T1D loci.

Fig. 3 ∣. Analysis of the CD28-CTLA4 locus.

a, A regional association plot for the combined analysis (20,787 rheumatoid arthritis or T1D cases, and 18,616 unique controls; Methods) shows a single variant (rs3087243) near CTLA4 in the credible set. Conditioning on rs30872043 identifies rs117701653 as an independent association. Logistic regression −log₁₀[P] values are from a two-sided χ² test (n = 39,403). b, An exhaustive pairwise analysis for rheumatoid arthritis shows that the rs3087243 + rs117701653 pair has the strongest association. Logistic regression −log₁₀[P] values are from a two-sided χ² test (n = 27,345). c, Haplotype analysis using rs30872043 and rs117701653, with the AG haplotype as a reference. The C allele of rs117701653 shows the largest decrease in risk for rheumatoid arthritis (RA), and the A allele of rs30872043 shows the largest decrease in risk for T1D. Logistic regression odds ratios (dots) and 95% confidence intervals (bars) are from a two-sided χ² test (combined n = 39,403; T1D n = 20,445; RA n = 27,345). d, EMSA with Jurkat nuclear extract using probes containing rs117701653 and rs3087243. A representative blot of three independent experiments is shown. e, Luciferase assay in Jurkat T cells transfected with pGL3 plasmids containing rs117701653. RLUs are normalized to cells transfected with the empty plasmid (pGL3). Dots represent independent experiments, each of which was performed in duplicate. Means±s.d. and two-tailed P values of the grouped comparisons (unpaired t-test) are shown.

Fig. 4 ∣. Analysis of the MEG3 locus.

a, Regional plot for the MEG3 locus in T1D (9,334 cases, 11,111 controls). We observe two variants in the credible set (rs56994090 and the rs34552516 indel). We did not observe secondary signals when conditioning on rs56994090. Logistic regression −log₁₀[P] values are from a two-sided χ² test (n = 20,445). b, EMSA with Jurkat nuclear extract using probes containing rs354552516 and rs56994090. A representative blot of three independent experiments is shown. c, Luciferase assay in Jurkat T cells transfected with pGL3 plasmids containing rs34552516. RLUs are normalized to cells transfected with the empty plasmid (pGL3). Dots represent independent experiments, each of which was performed in duplicate. Means±s.d. and two-tailed P values of the grouped comparisons (unpaired t-test) are shown.

Fig. 5 ∣. Analysis of the TNFAIP3 locus.

a, Regional plot for the TNFAIP3 locus in rheumatoid arthritis (11,475 cases, 15,870 controls). The variant with the strongest posterior probability in this locus is rs35926684, a G/GA indel. Conditional on rs35926684, we observe a significant secondary association with rs58721818. Logistic regression −log₁₀[P] values are from a two-sided χ² test (n = 27,345). b, Exhaustive pairwise association analysis in rheumatoid arthritis indicates that there are 11 pairs with a lower P value than rs35926684 + rs58721818, although the top-associated pair (rs69220220 + rs58721818) has an equivalent Pvalue −log₁₀[P] =13.95 versus 14.21. Logistic regression −log₁₀[P] values are from a two-sided χ² test (n = 27,345). c, Haplotype analysis with rs35926684 + rs58721818 and previously reported variants rs6920220 and rs5029937 shows that rs35926684 and the previously reported top variant rs6920220 are often located on the same haplotype (GAGC), although a rare haplotype exists with only the alternative allele of rs35926684, which causes a similar increase in risk, but with a larger confidence interval. Logistic regression odds ratios (dots) and 95% confidence intervals (bars) are from a two-sided χ² test (n = 27,345). RA, rheumatoid arthritis d, EMSA with Jurkat nuclear extract using probes containing the G or GA allele of rs35926684. A representative blot of three independent experiments is shown. e, Luciferase assay in Jurkat T cells transfected with pGL3 plasmids containing rs35926684. RLUs are normalized to cells transfected with the empty plasmid (pGL3). Dots represent independent experiments, each of which was performed in duplicate. Means±s.d. and two-tailed P values of comparisons (unpaired t-test) are shown.