RNA-seq data from matching primary cells and tissues showed that ~95% of RNAs originating from enhancers were unspliced and typically short (median 346 nt) - a striking difference to mRNAs (19% unspliced, median 56 nt) (Fig. 2a, and Supplementary Fig. 11a-c). Unlike TSSs of mRNAs, which are enriched for predicted 5’ splice sites but depleted of downstream polyadenylation (pA) signals11,12, enhancers showed no evidence of associated downstream RNA processing motifs, and thus resemble antisense PROMoter uPstream Transcripts (PROMPTs)11 (Fig. 2b, and Supplementary Fig. 11d). Most CAGE-defined enhancers gave rise to nuclear (>80%) and non-polyadenylated (~90%) RNAs13 (Supplementary Fig. 11e). Based on RNA-seq, few enhancer RNAs overlap exons of known protein-coding genes or lincRNAs (9 and 1 out of 4208 enhancers detected, respectively), suggesting that they are not a substantial source of alternative promoters for known genes (as in ref. 14).
