3DFAACTS-SNP: Using regulatory T cell-specific epigenomics data to uncover candidate mechanisms of Type-1 Diabetes (T1D)

Genome-wide association and fine-mapping studies have enabled the discovery of single nucleotide polymorphisms (SNPs) and other variants that are significantly associated with type 1 diabetes (T1D), and many other autoimmune diseases. However, many of the SNPs lie in non-coding regions that are not annotated with gene-coding information, limiting the identification of mechanisms that contribute to autoimmune disease progression. Autoimmunity results from a failure of immune tolerance, suggesting that regulatory T cells (Treg) are likely a significant point of impact for this genetic risk, as Treg are critical for immune tolerance. Focusing on T1D as a model of defective function of Treg in autoimmunity, we designed a SNPs filtering workflow called 3 Dimensional Functional Annotation of Accessible Cell Type Specific SNPs (3DFAACTS-SNP) that utilises overlapping profiles of Treg-specific epigenomic data (ATAC-seq, Hi-C and FOXP3-binding profiles) to identify regulatory elements potentially driving the effect of variants associated with T1D, and the gene(s) that they control. Using 3DFAACTS-SNP we identified 36 SNPs with plausible Treg-specific mechanisms of action contributing to T1D from 1,228 T1D fine-mapped variants, confirming previous known associations to 12 candidate gene regions and identifying 119 novel, 3D interacting regions in Tregs. In each case, candidate genes were surrounded by an extensive network of chromatin interactions, and were frequently associated with T cell super-enhancers. We further demonstrated the utility of the workflow by applying it on three other fine-mapped/meta-analysed SNP sets representing multiple sclerosis, 5 chronic inflammatory diseases and 4 autoimmune diseases, identifying 17 Treg-centric candidate variants and 35 interacting genes. Finally we demonstrate the broad utility of the 3DFAACTS-SNP workflow for functional annotation of any genetic variation datasets by applying the filtering approach to all common (>10% allele frequency) variants from the Genome Aggregation Database (gnomAD). We identified 7,900 candidate variants, generating a list of potential sites for future T1D or autoimmune research. We demonstrate that it is possible to further prioritise variants that contribute to T1D based on regulatory function, and illustrate the power of using cell type specific multi-omics datasets to determine disease mechanisms. The 3DFAACTS-SNP workflow can be customised to any cell type for which the individual datasets for functional annotation have been generated, giving broad applicability and utility.