Extended Data Fig. 4: Characteristics and evolution of bidirectional promoters.

a, Number of bidirectional and non-bidirectional promoters identified for each regulatory category. P values correspond to two-sided Fisher’s exact tests against ubiquitous promoters. b, Distribution of distance between bidirectional promoters in each species (amphioxus, 1,975; zebrafish, 549; and mouse, 876 pairs of promoters). The distance between amphioxus peaks closely corresponds to integral nucleosome spacing. c, Heat maps of TA, CG and nucleosome occupancy (derived from the NucleoATAC signal) around bidirectional promoter pairs in amphioxus (n = 1,975), mouse (n = 876) and zebrafish (n = 549), arranged by the distance between the two CAGE TSSs. In amphioxus, both TA and NucleoATAC signals indicate regions in which 0, 1 or 2 nucleosomes separate promoters. d, Enriched GO terms for genes associated with bidirectional promoters in amphioxus. Uncorrected P values correspond to two-sided Fisher’s exact tests as provided by topGO. e, Inferred evolutionary dynamics of 372 putatively ancestral bidirectional promoters among chordate groups. Red, number of inferred losses and disentanglements; black, number of detected bidirectional promoters by CAGE-seq (in brackets) or microsynteny (neighbouring genes in a 5′ to 5′ orientation) for each species. In parentheses, number of lost and disentangled (red) or retained (black) bidirectional promoters when considering only the cases supported by CAGE-seq. f, In vertebrates, disentanglement was not accompanied by a general increase in the fraction of bidirectional promoters with antisense non-coding transcription, as shown by the relative number of CAGE clusters identified as bidirectional promoters that are composed of two protein-coding genes (‘Prot-Prot’) or of one protein-coding and one non-coding or non-annotated locus (‘Prot-NC’). The total number of uniquely annotated, protein-coding-associated CAGE promoters was amphioxus, 11,789; mouse, 13,654; and zebrafish, 14,014.