Bidirectional transcription initiating at enhancers has been proposed to represent the signature of enhancer activity. Here we show that bidirectional transcription is a pervasive feature of all forms of accessible chromatin, including enhancers, promoters, CTCF-bound sites and other DNase hypersensitive regions. Transcription is less predictive for enhancer activity than epigenetic modifications such as H3K4me1 or the accessibility of DNA when measured in both enhancer assays and at endogenous loci. Bidirectional transcription initiation from accessible chromatin is therefore not sufficient for, nor specific to, enhancer activity. The stability of enhancer initiated transcripts does not influence measures of enhancer activity and we cannot detect any evidence of purifying selection on the resulting enhancer RNAs within the human population. Our results suggest that transcription initiating at enhancers is frequently a by-product of promiscuous RNA polymerase activity at accessible chromatin, and may not generally play a functional role in enhancer activity.
