Biased chromatin signatures around polyadenylation sites and exons

Abstract

Core RNA-processing reactions in eukaryotic cells occur cotranscriptionally in a chromatin context, but the relationship between chromatin structure and pre-mRNA processing is poorly understood. We observed strong nucleosome depletion around human polyadenylation sites (PAS) and nucleosome enrichment just downstream of PAS. In genes with multiple alternative PAS, higher downstream nucleosome affinity was associated with higher PAS usage, independently of known PAS motifs that function at the RNA level. Conversely, exons were associated with distinct peaks in nucleosome density. Exons flanked by long introns or weak splice sites exhibited stronger nucleosome enrichment, and incorporation of nucleosome density data improved splicing simulation accuracy. Certain histone modifications, including H3K36me3 and H3K27me2, were specifically enriched on exons, suggesting active marking of exon locations at the chromatin level. Together, these findings provide evidence for extensive functional connections between chromatin structure and RNA processing.

Figure 1. Nucleosome enrichment and depletion in the vicinity of core sites of RNA processing and controls

Nucleosome read signal, centered on (A) 3′ss, (B) exon centers and (C) 5′ss, with approximate exon sizes indicated by black box below. Nucleosome signal relative to (D) sequence-matched decoy 3′ss, (E) regions of exonic nucleotide composition and (F) decoy 5′ss. For reference, we have plotted nucleosome signal on the same y-axis for (G) transcription start sites of expressed genes and (H) PAS.

Figure 2. Exon-biased distribution of specific histone H3 methylation marks

(A) ChIP enrichment for exons, relative to flanking intronic regions (see methods), compared to 1.0 (CTCF and Pol II) or histone overall average of 1.3 (purple dashed line). Error bars are 95% confidence intervals (resampling). ** indicates p<0.01 after correction for multiple testing (resample test, Bonferroni-corrected). (B) Histone marks are similarly enriched in highly and lowly expressed genes. Profiles centered on exons for (C) mono-methyl histone marks, (D) di-methyl histone marks and (E) tri-methyl histone marks and Pol II, H2AZ and the negative control CTCF (F). (C)–(F) are normalized to average library ChIP signal across the displayed region.

Figure 3. Increased exonic bias of specific histone H3 methylation marks in exons with long flanking introns or weaker 3

′ss motifs. (A) Exon enrichment, relative to flanking introns for isolated exons (flanking introns > 5 kb, top bar of each pair) and clustered exons (flanking introns between 0.5 kb and 1.0 kb). Error bars are 95% confidence intervals. * indicates p<0.05 and ** indicates p<0.01 after Bonferroni correction for multiple testing (resample test). (B) Nucleosome signal profile for exons with short and with long flanking introns. (C) Nucleosome enrichment on exons is inversely correlated with 3′ splice site strength.

Figure 4. Nucleosome depletion and downstream nucleosome enrichment at high usage PAS

(A) Mean nucleosome density around human PAS of low (blue) or high (red) usage, normalized to average ChIP signal. (B) Mean NAS for positions around human PAS of low or high usage. Wilcoxon rank sum test p-values shown for 150 bp windows centered on indicated positions.