In addition to mutations in genes, aberrant enhancer element activity at non-coding regions of the genome is a key driver of tumorigenesis. Here, we perform epigenomic enhancer profiling of a cohort of more than forty genetically diverse human colorectal cancer (CRC) specimens. Using normal colonic crypt epithelium as a comparator, we identify enhancers with recurrently gained or lost activity across CRC specimens. Of the enhancers highly recurrently activated in CRC, most are constituents of super enhancers, are occupied by AP-1 and cohesin complex members, and originate from primed chromatin. Many activate known oncogenes, and CRC growth can be mitigated through pharmacologic inhibition or genome editing of these loci. Nearly half of all GWAS CRC risk loci co-localize to recurrently activated enhancers. These findings indicate that the CRC epigenome is defined by highly recurrent epigenetic alterations at enhancers which activate a common, aberrant transcriptional programme critical for CRC growth and survival.

The human genome can be segmented into topologically associating domains (TADs), which have been proposed to spatially sequester genes and regulatory elements through chromatin looping. Interactions between TADs have also been suggested, presumably due to variable boundary positions across individual cells. However, the nature, extent, and consequence of these dynamic boundaries remain unclear. Here, we combine high-resolution imaging with Oligopaint technology to quantify the interaction frequencies across both weak and strong boundaries. We find that chromatin intermingling across population-defined boundaries is widespread but that the extent of permissibility is locus-specific. Cohesin depletion, which abolishes domain formation at the population level, does not induce ectopic interactions but instead reduces interactions across all boundaries tested. In contrast, WAPL or CTCF depletion increases inter-domain contacts in a cohesin-dependent manner. Reduced chromatin intermingling due to cohesin loss affects the topology and transcriptional bursting frequencies of genes near boundaries. We propose that cohesin occasionally bypasses boundaries to promote incorporation of boundary-proximal genes into neighboring domains.

No abstract

SNPs associated with disease susceptibility often reside in clusters of gene enhancers, or super enhancers. Constituents of these enhancer clusters cooperate to regulate target genes, and often extend beyond the linkage disequilibrium blocks containing GWAS risk SNPs. We identified “outside variants”, defined as SNPs in weak LD with GWAS risk SNPs that physically interact with risk SNPs as part of the target gene’s regulatory circuitry. These outside variants explain additional target gene expression variation beyond that of GWAS associated SNPs. Additionally, the clinical risk associated with the GWAS SNPs is considerably modified by the genotype of the outside variant. Collectively, these findings suggest a potential model whereby outside variants and GWAS SNPs that physically interact in 3D chromatin collude to influence target transcript levels as well as clinical risk. This model offers an additional hypothesis for the source of missing heritability of complex traits.