In this study, we describe the most comprehensive characterization to date of the epigenomic profile of unstimulated human pancreatic islets. Using DNase- and ChIP-seq techniques, we profiled open chromatin, CTCF binding sites, H3K4me3, H3K4me1, and H3K79me2 across the entire genome in human islets. Integrated analysis of these large-scale datasets identified ~18,000 putative TSSs, ~30% of which were previously unannotated by RefSeq. Further computational genomic analyses revealed that at least several hundred of these are islet-active TSSs, including those for major islet miRNAs previously implicated in the control of glucose homeostasis (Lynn, 2009
). Interestingly, active chromatin marks (H3K4me3, DHS, H3K79me2) were absent from a subset of highly islet-expressed genes, including those encoding islet-specific hormones (INS, GCG, SST, IAPP, PPY,
). This observation suggests that some genes critical for islet function have an unconventional promoter chromatin signature indicative of a unique transcriptional control mechanism. Mutskov and Felsenfeld (2009)
have proposed such a model based on detailed analysis of the INS
locus in human islets.
We also identified ~34,000 candidate distal regulatory elements in human islets. A substantial number of these putative elements were clustered (<1000 bp from each other). Comparisons with other cell types indicated that these clustered elements are significantly enriched for islet-unique sites and thus may represent islet-specific regulatory modules worthy of more extensive future investigation. Based on CTCF binding profiles, ~22% of the ~34,000 candidate distal regulatory elements are predicted insulator sites. Previous studies have reported that the H3K4me1 signal is enriched in distal regulatory elements (Heintzman and Ren, 2007, 2009
). Though our analyses confirm this finding in aggregate, we show that H3K4me1 enrichment may not be a reliable predictor of regulatory activity for individual elements.
Fifty SNPs associated with islet-related diseases and traits map to within 500 bp of a candidate non-promoter regulatory element. Focusing on T2D, 4 of 12 elements that function as enhancers in vitro (FTO, KCNQ1, TCF7L2, and WFS1 loci) harbor T2D-associated SNPs, including 2 (TCF7L2 and WFS1 loci) that exhibit significant allele-specific differences in activity. These results suggest that altered enhancer activity plays a role in the molecular mechanism underlying at least a subset of T2D genetic association signals.
These datasets should provide functional context for non-coding variants identified through additional association, targeted resequencing, or whole genome sequencing studies. Further analysis of the repertoire of regulatory elements in the human islet will enhance the understanding of gene regulation in the islet and should offer additional insight into the molecular mechanisms that underlie diabetes susceptibility.