Uncovering Hidden Enhancers Through Unbiased In Vivo Testing

Transcriptional enhancers are a predominant class of noncoding regulatory elements that activate cell type-specific gene expression. Tissue-specific enhancer-associated chromatin signatures have proven useful to identify candidate enhancer elements at a genome-wide scale, but their sensitivity for the comprehensive detection of all enhancers active in a given tissue in vivo remains unclear. Here we show that a substantial proportion of in vivo enhancers are hidden from discovery by conventional chromatin profiling methods. In an initial comparison of over 1,200 in vivo validated tissue-specific enhancers with tissue-matched mouse developmental epigenome data, 14% (n=286) of active enhancers did not show canonical enhancer-associated chromatin signatures in the tissue in which they are active. To assess the prevalence of enhancers not detectable by conventional chromatin profiling approaches in more detail, we used a high throughput transgenic enhancer reporter assay to systematically screen over 1.3 Mb of mouse genomic sequence at two critical developmental loci, assessing a total of 281 consecutive 5kb regions for in vivo enhancer activity in mouse embryos. We observed reproducible enhancer-reporter activity in 88 tissue-specific elements, 26% of which did not show canonical enhancer-associated chromatin signatures in the corresponding tissues. Overall, we find these hidden enhancers are indistinguishable from marked enhancers based on levels of evolutionary conservation, enrichment of transcription factor families, and genomic positioning relative to putative target genes. In combination, our retrospective and prospective studies assessed only 0.1% of the mouse genome and identified 309 tissue-specific enhancers that are hidden from current chromatin-based enhancer identification approaches. Our findings suggest the existence of tens of thousands of active enhancers throughout the genome that remain undetected by current chromatin profiling approaches and are an unappreciated source of additional genome function of import in interpreting growing whole human genome sequencing data.