Fine-mapping, trans-ancestral and genomic analyses identify causal variants, cells, genes and drug targets for type 1 diabetes

Abstract

We report the largest and most diverse genetic study of type 1 diabetes (T1D) to date (61,427 participants), yielding 78 genome-wide-significant (P < 5 × 10^-8) regions, including 36 that are new. We define credible sets of T1D-associated variants and show that they are enriched in immune-cell accessible chromatin, particularly CD4⁺ effector T cells. Using chromatin-accessibility profiling of CD4⁺ T cells from 115 individuals, we map chromatin-accessibility quantitative trait loci and identify five regions where T1D risk variants co-localize with chromatin-accessibility quantitative trait loci. We highlight rs72928038 in BACH2 as a candidate causal T1D variant leading to decreased enhancer accessibility and BACH2 expression in T cells. Finally, we prioritize potential drug targets by integrating genetic evidence, functional genomic maps and immune protein-protein interactions, identifying 12 genes implicated in T1D that have been targeted in clinical trials for autoimmune diseases. These findings provide an expanded genomic landscape for T1D.

Extended Data Fig. 1. Fine mapping of the chromosome 6q22.32 region

European (EUR, top panel) and African (AFR, middle panel) ancestry group association z-score statistics and posterior probabilities (bottom panel) from multi-ethnic fine mapping of EUR and AFR using PAINTOR. z-scores are colored by linkage disequilibrium (LD) to the lead PAINTOR-prioritized variant.

Extended Data Fig. 2. Fine mapping of the chromosome 18q22.2 region

European (EUR, top panel) and African (AFR, middle panel) ancestry group association z-score statistics and posterior probabilities (bottom panel) from multi-ethnic fine mapping of EUR and AFR using PAINTOR. z-scores are colored by linkage disequilibrium (LD) to the lead PAINTOR-prioritized variant.

Figure 1 |. Fine-mapping T1D regions using a Bayesian stochastic search algorithm.

a, Number of variants in GUESSFM-prioritized groups with group posterior probability > 0.5. Candidate gene names and lead variants for each group are shown on the y-axis. b, Manhattan plot of the UBASH3A region from the EUR case-control analysis, highlighting the lead variant from the univariable analysis, rs11203203:G>A (grey), and the three variants prioritized using GUESSFM, rs9984852:T>C (blue), rs13048049:G>A (red) and rs7276555:T>C (green). c, Comparison of model AIC in the UBASH3A region for models fit using EUR cases and controls only, comparing combinations of alleles prioritized either in univariable (grey) or GUESSFM analyses (red, green and blue). d, Analysis of haplotypes associated with T1D in the UBASH3A region. The most common haplotype (H1: T-G-G-T for rs7276555-rs13048049-rs11203203-rs9984852) is presented on the far left; alternative haplotypes (H2-H6) are shown with white squares highlighting the differentiating alleles (C, A, A, or C, respectively). The frequency and effect estimates for association with T1D relative to the baseline haplotype (H1) are shown above the grid (the point and error bars represent the log odds ratio and 95% confidence interval of the log odds ratio, respectively); for example, the log odds ratio for T1D risk for haplotype H3 (T-G-A-T) relative to the baseline haplotype (H1) is close to zero and the 95% confidence interval crosses zero. Haplotype analyses were performed based on n = 33,601 unrelated EUR individuals (13,458 T1D cases and 20,143 controls).

Figure 2 |. Fine-mapping of the chromosome 4p15.2 region.

a, European (EUR, top panel) and African (AFR, middle panel) ancestry group association z-score statistics; posterior probabilities (bottom panel) from multi-ethnic fine-mapping of EUR and AFR using PAINTOR; z-scores are colored by linkage disequilibrium (LD) to the lead PAINTOR-prioritized variant. b, Overlay of T1D-credible variants with open chromatin ATAC-seq peaks in immune cells, with variants prioritized by PAINTOR (posterior probability > 0.1) highlighted with blue dashed lines. Normalized ATAC-seq read count shown for effector CD4+ T cells, B cells, and CD8+ T cells, under stimulated and non-stimulated conditions.

Figure 3 |. Functional annotation of T1D-associated variants in the BACH2 region.

a–c, Position of T1D credible variants (rs72928038:G>A and rs6908626:G>T) relative to introns and exons of BACH2 (a), chromHMM tracks across diverse immune cell types from the BLUEPRINT consortium (red, active promoter; orange, distal active promoter; dark green, transcription; light green, genic enhancer; yellow, enhancer; white, quiescent; light grey, Polycomb repressed; dark grey, repressed; blue, heterochromatin) (b), and interactions with the BACH2 promoter in published PCHi-C data from naïve CD4⁺ T cells (grey squares indicate boundaries of target (left) and bait (right)) (c). Chromatin coordinates and scale are identical and aligned in figures a–c. d, Accessibility of regions overlapping rs72928038:G>A and rs6908626:G>T by genotype; peak accessibility is quantified as normalized transposase cut frequency (Online Methods); center line, median; box limits, upper and lower quartiles; whiskers, 1.5x interquartile range (n = 115 individuals). e, Allele-specific accessibility of chromatin within heterozygous individuals at rs72928038:G>A (n = 14 heterozygous individuals) and rs6908626:G>T (n = 15 heterozygous individuals). f, Chromatin accessibility profiles in the region overlapping rs72928038:G>A across resting and activated CD4⁺ and CD8⁺ T cells (published data). Height of tracks represent transposase cut frequency; all tracks are plotted using the same vertical scale. g, LocusCompare plots showing colocalization between T1D association, the caQTL for chr6:90266766–90267715 (left), and the eQTL for BACH2 (right).