Figure 1.
Pol II occupancy and promoter-proximal pausing at active genes correlates with cohesin levels in BG3 cells.
(A) PRO-seq density at the promoters of active genes divided into four groups with increasing levels of cohesin at the promoter as measured by Rad21 ChIP-chip. Active genes have at least 1 read per million in both the promoter and gene body regions. The four cohesin groups and number of genes in each are shown in Figure S1A. The box plots show the distributions of promoter PRO-seq density for each cohesin group. Asterisks indicate significant differences as determined by Kolmogorov–Smirnov statistical tests. (B–E) Rpb3, Ser2P Pol II, Cdk12, and CycT occupancy of active genes as measured by ChIP-chip. (F) Pause index distributions for each cohesin level group. Pause index was calculated using PRO-seq data (see text). (G) Example of PRO-seq and ChIP-chip data at the diminutive (myc) gene. The black bars above the log2 PRO-seq track indicate the regions defined as promoter (200 bp surrounding the transcription start site) and gene body (rest of annotated transcription unit). Bars beneath the ChIP-chip tracks indicate where occupancy is significant at p≤10−3.
Table 1.
Genome-wide correlations between Pol II, CycT, Cdk12, and Rad21 chromosome association by ChIP–chip.
Figure 2.
Cohesin binds nearly all predicted extragenic cis-regulatory modules (CRMs) in BG3 cells.
CRMs (enhancers or other regulatory sequences) in BG3 cells were predicted from modENCODE data [22]. For purposes of determining protein occupancy and changes, all CRMs were defined as 200 bp elements centered around the DNaseI hypersensitive site (DHS). 557 putative extragenic CRMs that are a minimum of 500 bp from transcription start sites were analyzed. (A) Percent of extragenic CRMs occupied by Pol II and Pol II kinases as determined by ChIP-chip at p≤10−3. Cohesin binding was determined by Smc1 and Nipped-B ChIP [9]. (B) Comparison of Rad21, Rpb3, Ser2P Pol II (ChIP-chip) and PRO-seq occupancy of active promoters (P) and CRMs. The PRO-seq data is separated into + and − strands for the CRMs. (C) Example of predicted CRMs (red vertical bars) upstream of the cut gene. They are between the wing margin enhancer that is sensitive to Nipped-B dosage in vivo [19], [25] and other tissue-specific enhancers [24]. The PRO-seq and ChIP-chip are as described in Figure 1. RNA-seq data is from modENCODE [51]. The Ser2P Pol II Δ track is the difference in ChIP MAT score between Rad21 RNAi-treated and mock control cells. Bars below the Δ track indicate where the decrease after Rad21 depletion is ≥2 σ for ≥105 bp. (D) Percent of extragenic CRMs showing decreases (DOWN) or increases (UP) in Pol II and Pol II kinases at determined by ChIP-chip. For each protein, only those CRMs binding the protein in the control cells were used for the calculation. Decreases and increases are defined as ≥2 σ from the median genome-wide difference over a region ≥105 bp.
Figure 3.
Cohesin and kollerin have very similar effects on the levels of transcriptionally engaged Pol II and promoter-proximal pausing.
(A) Plot of fold-change (log2) in PRO-seq reads in the 200 bp promoter regions of PRO-seq active transcription units after Rad21 (cohesin) depletion versus Nipped-B (kollerin) depletion. The correlation between the effect of Nipped-B and Rad21 depletion is in the upper left hand corner. Genes that bind cohesin at p≤10−3 as determined by cohesin ChIP are in red. Genes that don't bind cohesin are in blue, many of which are hidden underneath the cohesin-binding genes. (B) Same plot for the changes in the gene bodies. (C) Same plot for changes in the pause index. (D) Fold-changes in pause index after Nipped-B (left) or Rad21 (right) depletion versus low to high (blue to red) cohesin promoter occupancy.
Figure 4.
Cohesin depletion preferentially decreases transcriptional pausing and initiation at genes that lack cohesin, and increases Pol II kinase occupancy more frequently in the bodies of cohesin-binding genes.
(A) The box plots show the fold changes (log2) upon Rad21 depletion for all active cohesin-binding (red) or non-binding (blue) genes in total Pol II (Rpb3 ChIP) promoter occupancy, transcriptionally-engaged Pol II (PRO-seq) at the promoter, and pausing. Cohesin-binding was called at p≤10−3 to distinguish cohesin-binding from non-binding genes. (B) Percent PRO-seq active cohesin-binding (red) and non-binding (blue) transcription units with absolute decreases or increases of Rpb3 (total Pol II) and Ser2P Pol II in the promoter regions and gene bodies (differences in ChIP MAT scores ≥2 σ for ≥105 bp). The higher frequencies of Rpb3 and Ser2P decreases in the bodies of cohesin binding genes compared to non-binding genes are significant (Fisher's exact test p = 1×10−76 and 3×10−100). (C) Percent PRO-seq active transcription units with decreases or increases in CycT (P-TEFb) and Cdk12 in the promoter regions and gene bodies. Only genes binding CycT or Cdk12 at p≤10−3 in control cells were used for this analysis. The differences in CycT and Cdk12 increases in the bodies of cohesin-binding versus non-binding genes are significant (p = 2.6×10−18 and 9.6×10−68).
Figure 5.
Cohesin-binding genes produce more steady state mRNA per elongating Pol II complex in BG3 cells.
(A) The efficiency (log2 mRNA/Gene Body PRO-seq) of genes with different levels of cohesin at the promoter. (B) The fold-change in efficiency upon kollerin (Nipped-B). (C) The fold change in efficiency upon cohesin (Rad21) depletion.
Figure 6.
Key features of cohesin-binding genes and proposed roles for cohesin in genome-wide control of Pol II activity.
Cohesin-binding genes (upper left) have higher levels of Pol II and promoter-proximal Pol II pausing than other active genes (lower left). Cohesin-binding genes are enriched for GAGA factor (GAF) binding upstream of the promoter and TG repeats in the transcribed region, and unlike other active genes, lack the H3K36me3 histone modification [12]. Current findings indicate that cohesin facilitates looping and contact between enhancers and promoters (upper right), which primarily facilitates transition of paused Pol II to elongation. Cohesin at the promoter may also directly facilitate transition to elongation by increasing the efficiency of P-TEFb or Cdk12 phosphorylation of Pol II and the Spt5 and NELF pausing factors (lower right). To explain the broad effect on Pol II activity at genes that don't bind cohesin, we hypothesize that cohesin promotes expression of broadly acting transcription factors such as Myc that regulate many or most genes (lower left).