The Ground State and Evolution of Promoter Region Directionality

Abstract

Eukaryotic promoter regions are frequently divergently transcribed in vivo, but it is unknown whether the resultant antisense RNAs are a mechanistic by-product of RNA polymerase II (Pol II) transcription or biologically meaningful. Here, we use a functional evolutionary approach that involves nascent transcript mapping in S. cerevisiae strains containing foreign yeast DNA. Promoter regions in foreign environments lose the directionality they have in their native species. Strikingly, fortuitous promoter regions arising in foreign DNA produce equal transcription in both directions, indicating that divergent transcription is a mechanistic feature that does not imply a function for these transcripts. Fortuitous promoter regions arising during evolution promote bidirectional transcription and over time are purged through mutation or retained to enable new functionality. Similarly, human transcription is more bidirectional at newly evolved enhancers and promoter regions. Thus, promoter regions are intrinsically bidirectional and are shaped by evolution to bias transcription toward coding versus non-coding RNAs.

Keywords: NET-seq; RNA polymerase; bidirectional; evolution; non-coding RNA; promoter; transcription; yeast.

Figure 1. Promoter region directionality across the Saccharomyces cerevisiae genome

(A) NET-seq analysis of the Saccharomyces cerevisiae genome. Upper panel shows the distribution of active RNAPII for the promoter region of the YBL068W gene. Transcription in the sense and antisense directions are plotted above and below the horizontal axis, respectively. Bottom panel shows the aggregate plot of NET-seq reads averaged over all gene promoter regions by aligning to their TSS. Only promoter regions between tandemly-oriented genes were included to ensure that the antisense transcript is non-coding. (B) Directionality score is defined as the log10 ratio of sense and antisense reads measured within 500 bp windows situated upstream and downstream of TSS (shown as boxes in (Figure 1a)). Genome-wide distribution of the directionality score is displayed in the middle panel. For promoter regions lacking sense (left panel) or antisense reads (right panel), distributions of the antisense or sense reads are displayed instead. Promoter regions were categorized as directional (yellow) if the sense to antisense ratio was ≥3, and bidirectional (pink) if the ratio was <3. See also Figure S1

Figure 2. Promoter region directionality is lost in a foreign environment

(A) Chromosome pieces extracted from K. lactis and D. hansenii were inserted into yeast artificial chromosomes (YACs) containing centromere and telomere sequences and selective markers on both arms. (B) NET-seq reads for two promoter regions from K. lactis (left) and D. hansenii (right) are shown in their native environments and in the foreign (S. cerevisiae) environment. (C) Genome-wide distributions of the directionality score native species and YAC S. cerevisiae strains are displayed in the middle panel. For promoter regions lacking sense (left panel) or antisense reads (right panel), distributions of the antisense or sense reads are displayed instead. The p-values of two sample Kolmogorov-Smirnov test (KS-test) for YAC and native distributions are 3.8 10⁻⁷ (K. lactis) and 2.0 10⁻⁹ (D. hansenii). See also Figures S1 and S2

Figure 3. Fortuitous promoter regions arise in foreign environments and produce bidirectional transcription

(A) Example of a fortuitous promoter region emerging within the coding sequence of the D. hansenii gene, DEHA2D15356g, when in a foreign (S. cerevisiae) environment. (B) Aggregate plot of the average NET-seq reads over the fortuitous promoter regions in native and YAC strains (upper two panels). Transcription in the sense and antisense directions are plotted above and below the horizontal axis, respectively. Aggregate plots are shown for TFIIB ChIP-seq in YAC strains (dark red) and MNase-seq in YAC (blue grey) and native (dark blue) strains over the fortuitous promoters are shown (bottom two panels). (C) Histogram of directionality scores for native D. hansenii (upper), corresponding YACs (upper), and the fortuitous promoter regions (bottom). Genome-wide distributions of the directionality score are displayed in the middle panels. For promoter regions lacking sense (left panels) or antisense reads (right panels), distributions of the antisense or sense reads are displayed instead. Two sample KS-test p-values are 7.1×10⁻¹² and 2.1×10⁻⁴ for when comparing fortuitous distribution to D. hansenii native and YAC distributions, respectively. (D) Transcription factors whose binding sites are significantly enriched at fortuitous promoter regions. P-values are determined through comparison of binding site density at fortuitous promoter regions compared to D. hansenii native promoter regions. Data summarized in Table S3. See also Figures S3, S5 and Table S1.

Figure 4. Evolutionary analysis of promoter region directionality

(A) Aggregate plots of genomic evolutionary rate profiling (GERP) scores, determined by multiple alignment of seven Saccharomyces genomes, for directional and bidirectional promoter regions as defined in Figure 1B. The p-value is 0.02 calculated by Kolmogorov-Smirnov test for the distributions of average GERP scores over directional and bidirectional promoters, i.e., 500bp upstream of TSS. (B) Evolutionary tree displaying the relationship between 24 yeast species. Saccharomyces sensu stricto species are boxed. (C) Directionality distributions for the promoter regions of the genes whose orthologs are present only in Saccharomyces sensu stricto (orange) and those whose orthologs are also present in other species (blue). The distributions are significantly different according to the KS-test (p-value<3.5 10⁻¹²). See also Figure S4 and Table S2.

Figure 5. Evolutionary trajectories of de novo promoter regions

(A) Schematic showing possible paths that a cell can take after an emergence of a fortuitous promoter region. First option is purging the TF binding site found in the coding sequence by mutation (left). Alternatively, the new transcripts produced by the fortuitous promoter regions could be retained (right). (B) Coding sequence binding site densities are calculated for Reb1 and Abf1 for the genomes of 24 yeast species, averaged across each clade and plotted against the branching point in the evolutionary tree relative to S. cerevisiae (see Figure 4B). Difference between the binding site densities for the genomes of the species at each branch point and densities for the S. cerevisiae genome was determined by a two-sample Poisson intensity test (Gu et al., 2008). **: p-value < 10⁻⁴, ***: p-value << 10⁻¹⁰. (C) Venn diagram shows the overlap between the transcription factors whose binding sites are enriched at fortuitous promoter regions and endogenous promoter regions of sensu stricto specific and other genes in S. cerevisiae. See also Table S4.

Figure 6. Human transcription is more bidirectional at newly evolved regulatory regions

(A) Directionality score histogram of human coding sequence promoter regions is shown. The directionality scores of the human promoter regions are calculated the same way as for yeast promoter regions, using NET-seq data from HeLa S3 cells (Mayer et al., 2015), with one alteration. The length of the upstream and downstream windows around the TSS is 1kb instead of 500bp, due to the ambiguity of human TSS annotation. Non-overlapping human CDS were curated as described in Mayer et al (Mayer et al., 2015). (B) Absolute values of directionality scores for enhancers (HeLa S3), human coding promoter regions (HeLa S3) and yeast coding promoter regions are plotted as cumulative distribution. Enhancer regions were identified as described in Mayer et al (Mayer et al., 2015). Fortuitous promoter regions and enhancers are not statistically significantly different (p-value=0.59 by KS test). (C) Directionality score distribution of recently evolved and older promoter regions are shown (p=0.03 by KS test). The list of recently evolved and older promoter regions is obtained from Villar et al. 2015.