Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2012 Nov;22(11):2163-75.
doi: 10.1101/gr.136507.111. Epub 2012 Jul 10.

Cohesin Regulates Tissue-Specific Expression by Stabilizing Highly Occupied Cis-Regulatory Modules

Affiliations
Free PMC article

Cohesin Regulates Tissue-Specific Expression by Stabilizing Highly Occupied Cis-Regulatory Modules

Andre J Faure et al. Genome Res. .
Free PMC article

Abstract

The cohesin protein complex contributes to transcriptional regulation in a CTCF-independent manner by colocalizing with master regulators at tissue-specific loci. The regulation of transcription involves the concerted action of multiple transcription factors (TFs) and cohesin's role in this context of combinatorial TF binding remains unexplored. To investigate cohesin-non-CTCF (CNC) binding events in vivo we mapped cohesin and CTCF, as well as a collection of tissue-specific and ubiquitous transcriptional regulators using ChIP-seq in primary mouse liver. We observe a positive correlation between the number of distinct TFs bound and the presence of CNC sites. In contrast to regions of the genome where cohesin and CTCF colocalize, CNC sites coincide with the binding of master regulators and enhancer-markers and are significantly associated with liver-specific expressed genes. We also show that cohesin presence partially explains the commonly observed discrepancy between TF motif score and ChIP signal. Evidence from these statistical analyses in wild-type cells, and comparisons to maps of TF binding in Rad21-cohesin haploinsufficient mouse liver, suggests that cohesin helps to stabilize large protein-DNA complexes. Finally, we observe that the presence of mirrored CTCF binding events at promoters and their nearby cohesin-bound enhancers is associated with elevated expression levels.

Figures

Figure 1.
Figure 1.
(A) Genome-wide occupancy of cohesin, CTCF, tissue-specific and ubiquitous TFs in primary mouse liver as measured by ChIP-seq and shown near the Pck1 gene. Cohesin colocalizes with CTCF as well as with clusters of transcription factors in the absence of CTCF, one of which can be seen overlapping the TSS of the Pck1 gene. (B) Venn diagram showing CTCF and cohesin (RAD21, STAG1, STAG2) occurrence within CRMs. The pie charts indicate genomic locations of all CRMs (background), as well as those containing CTCF and CNC. The latter occur within promoter regions at a higher relative frequency compared with the other two classes.
Figure 2.
Figure 2.
Within-CRM binding correlations reveal distinct modes of cohesin binding in diverse cell types. The number of ChIP fragments (mapped reads extended to the estimated fragment length) overlapping a given CRM was used as a measure of binding strength for each data set. Factors were clustered along both axes based on the similarity in their colocalization profiles. (A) Heatmap visualization of all pairwise correlations between all ChIP-seq data sets in mouse liver cells illustrates cohesin subunits (RAD21, STAG1, STAG2) clustered with CTCF. Cohesin also correlates with key tissue-specific TFs (FOXA1, HNF4A, and HNF1A) independently of CTCF as well as with histone modifications associated with transcriptional activity (H3K4me1, H3K4me3, H2AK5ac) and coactivators (EP300 and CREBBP). (B) All pairwise correlations between previously published ChIP-seq data sets in mouse embryonic stem cells. Cohesin binding strength (SMC1A, SMC3) correlates with CTCF while also forming a distinct cluster with key regulators of stem cell identity (POU5F1, SOX2, NANOG, MYC), components of the mediator complex, as well as RNAP2. Similar results were obtained by performing the correlation analysis separately on CRMs with CNC and CTCF (see Supplemental Fig. S3).
Figure 3.
Figure 3.
Cohesin-non-CTCF (CNC) binding occurs preferentially at multiply bound CRMs. (A) Results from K-means clustering (K = 10) of the binary presence/absence of ChIP-seq peaks corresponding to the 11 sequence-specific factors within 210,067 CRMs containing at least one of these factors. Factors were clustered based on the similarity in their binary occupancy profiles. The clusters were indexed and sorted by the proportion of CRMs with CNC in each cluster (increasing from left to right). (B) The binary presence/absence of ChIP-seq peaks for various chromatin features (non-sequence-specific factors and histone modifications) visualized according to the K-means results in A. Genomic location with respect to promoters (≤2.5 kb from an annotated TSS), exons, introns, and gene distal regions, is also indicated. The proportion of CRMs with CNC sites in each cluster is indicated at the bottom (increasing from left to right). (C) Barplot indicating the mean number of distinct TFs within each CRM cluster. Bar widths correspond to the number of CRMs within each cluster.
Figure 4.
Figure 4.
CNC sites are associated with liver-specific gene expression. (A) Ratio of CNC-containing CRMs versus those with CTCF (log-fold change) for CRM classes with 0–10 TFs. Each class of CRMs was also tested for association with 107 genes signficantly up-regulated in mouse liver cells (see Methods). The significance of the association (negative-log-transformed Fisher's exact test P-values) are indicated. (*) P < 0.01; (**) P < 0.001. The enrichment of CNC-containing CRMs reaches threefold when seven TFs are present, and coincides with highly significant enrichment for an association with liver-specific gene expression for the same class. (B) CRMs with high numbers of colocalizing TFs are associated with increased promoter proximity (≤2.5 kb from an annotated TSS) and characteristics of transcriptional activity (RNAP2 and H3K4me3 ChIP-seq peaks). Likewise, the associated absolute gene expression value increases significantly with the number of bound TFs. Error bars indicate the 95% confidence interval of the median.
Figure 5.
Figure 5.
Cohesin ChIP signal is significantly associated with TF motif score. (A) Cartoon heatmap representation of correlations between each sequence-specific factor's motif score and the ChIP signal of all available ChIP-seq data sets. Correlations with CRM occupancy (number of distinct TFs present) and promoter proximity (distance to the nearest canonical TSS) are also shown. For each factor, the motif score correlation was calculated on the set of CRMs that contained a ChIP-seq peak for the same factor. Correlations with cohesin and coactivator ChIP signal were averaged over subunits (RAD21, STAG1, STAG2) and family members (EP300, CREBBP), respectively. Heatmap rows were ordered by increasing correlation with cohesin ChIP signal (from top to bottom). As a visual summary, only the top- and bottom-ranking correlations involving TFs are shown (see Supplemental Figs. S6, S7 for all correlations). (B) Increased cohesin ChIP signal at TF binding events without motifs. For each sequence-specific factor, the number of cohesin ChIP fragments within CRMs without high-scoring motifs was compared with that of CRMs with motifs. The 95% confidence intervals shown are based on a normal approximation of the Hodges–Lehmann estimate (median of all possible differences).
Figure 6.
Figure 6.
ONECUT1 ChIP-seq in heterozygous Rad21+/− mouse liver cells shows preferential loss of TF binding events where no motif is present. (A) Sample region near the BC031353 gene showing overall reduction in RAD21 ChIP signal in heterozygous Rad21+/− cells (responsive RAD21) and associated significant reduction in ONECUT1 ChIP signal within two CRMs (responsive ONECUT1). The ONECUT1 binding event overlapping the TSS contains no ONECUT1 motif. (B) WT ONECUT1 CRMs without motifs show a preferential decrease in ChIP signal (FDR < 0.1) in heterozygous Rad21+/− mouse liver cells (Fisher's exact test P = 10−4). Regions of interest (ROI) are those CRMs where RAD21 binding was ablated in heterozygous Rad21+/− mouse liver cells (responsive RAD21).
Figure 7.
Figure 7.
Simultaneous CTCF binding within promoters and nearby enhancers is associated with elevated expression levels. (A) Violin plots showing gene expression distributions. Genes with CTCF binding events both within their promoters and nearby their associated enhancers show significantly elevated expression levels over those of the other three indicated classes (Mann-Whitney U-test P < 10−3). (B) Sample region near the liver-expressed Agxt gene, where CTCF binds within the core promoter, as well as near putative upstream cohesin-bound enhancers. Note that while CTCF is absent from the enhancers (CNC), it co-binds with HNF4A and EP300 within the Agxt promoter.

Similar articles

See all similar articles

Cited by 82 articles

See all "Cited by" articles

Publication types

MeSH terms

Feedback