Dense fine-mapping reveals FOXA2-bound sites as a genomic marker of type 2 diabetes risk. K. J. Gaulton1, T. M. Teslovich2, T. Ferreira1, M. Reschen3, A. Mahajan1, Y. Lee2, M. van de Bunt1, N. W. Rayner1, A. Raimondo4, C. O'Callaghan3, A. L. Gloyn4,6, A. P. Morris5, M. I. McCarthy1,4,6, DIAGRAM Consortium 1) Wellcome Trust Centre for Human Genetics, Oxford, UK; 2) Department of Biostatistics, University of Michigan, Ann Arbor MI, USA; 3) Oxford Centre for Cellular and Molecular Physiology, Oxford, UK; 4) Oxford Centre for Diabetes, Endocrinology and Metabolism, Oxford, UK; 5) Department of Biostatistics, University of Liverpool, Liverpool, UK; 6) Oxford NIHR, Churchill Hospital, Oxford, UK.
Genome-wide association studies (GWAS) for type 2 diabetes (T2D) have identified over 80 risk loci. These loci are broadly enriched for regulatory enhancers but the specific causal variants and factors driving a regulatory contribution to disease risk are mostly unknown. In this study we aimed to improve resolution of causal variant(s), and identify regulatory factors through which causal variants may influence T2D. We used Metabochip data from 27,206 T2D cases and 57,574 controls of European ancestry, supplemented by imputation up to 1000 Genomes Project data (March 2012 release) and performed fine-mapping of the 39 established T2D-loci captured on the chip. Conditional analyses revealed a total of 48 independent association signals at these loci: at each, we defined credible sets of SNPs that have >99% probability of including the causal variant. Credible sets included no more than twenty variants for 22 signals, with greatest refinement at MTNR1B (only rs10830963, 99.8% causal probability). Credible sets were then integrated with published ChIP-seq sites for 141 DNA-binding proteins (data from ENCODE). We compared the average probability of credible set variants in sites for each protein with the average probability of variants in site locations permuted within 1Mb. We found marked enrichment for FOXA2 binding sites (P=6x10-5), for which ChIP data was available from HepG2 cells and primary pancreatic islets. This enrichment was primarily driven by FOXA2 sites shared with at least one other factor (Pshared=8x10-5, Punique=.007), and sites identified in islets (Pislets=3x10-4, PHepG2=.008). To identify specific loci though which the FOXA2 signal might act, we considered FOXA2 binding site overlap at individual loci, finding nominally significant enrichment (P<.05) at 13 loci. We then identified 21 candidate regulatory variants in FOXA2 sites at these 13 loci predicted to disrupt a sequence motif (from JASPAR, ENCODE, and HOMER), including rs10830963 (MTNR1B). Electrophoretic mobility shift assays (EMSA) of rs10830963 confirmed allele-specific binding of NEUROD1 to the risk allele in rodent pancreatic beta-cell lines (MIN6, INS1). These results confirm the utility of fine-mapping experiments to resolve causal variants at established risk loci and provide potential molecular mechanisms. Further, FOXA2 binding appears to be a genomic marker of causal T2D variation, through which we identify a putative mechanism for the causal variant at MTNR1B.
You may contact the first author (during and after the meeting) at