Binding domain mutations provide insight into CTCF's relationship with chromatin and its contribution to gene regulation

Abstract

Here we used a series of CTCF mutations to explore CTCF's relationship with chromatin and its contribution to gene regulation. CTCF's impact depends on the genomic context of bound sites and the unique binding properties of WT and mutant CTCF proteins. Specifically, CTCF's signal strength is linked to changes in accessibility, and the ability to block cohesin is linked to its binding stability. Multivariate modeling reveals that both CTCF and accessibility contribute independently to cohesin binding and insulation, but CTCF signal strength has a stronger effect. CTCF and chromatin have a bidirectional relationship such that at CTCF sites, accessibility is reduced in a cohesin-dependent, mutant-specific fashion. In addition, each mutant alters TF binding and accessibility in an indirect manner, changes which impart the most influence on rewiring transcriptional networks and the cell's ability to differentiate. Collectively, the mutant perturbations provide a rich resource for determining CTCF's site-specific effects.

Keywords: CTCF; CTCF mutations; cancer; chromatin accessibility; chromatin organization; cohesin; gene regulation; neurological disorder; residence time.

Graphical abstract

Figure 1

CTCF complementation system (A) Scheme of the CTCF doxycycline-inducible degron system. (B) Experimental strategy for the expression of WT and mutant transgenic CTCF. (C) Flow cytometry showing the level of GFP (endogenous CTCF) and mRuby (transgenic WT CTCF). (D) Left, scheme showing the locations of the different types of CTCF mutations within a ZF. Amino acids making contacts with the DNA are shown in shades of pink, residues that coordinate the zinc ion in red, boundary residues in purple, and residues that contact the sugar phosphate backbone of DNA in blue. Right, representation of CTCF showing the locations of each mutation under investigation. (E) Bar graph showing the expression levels (mean fluorescence intensity) of endogenous CTCF (GFP) and transgenic WT or mutant CTCF (mRuby2) after ID treatment. The error bars represent the standard deviation between the two replicates. (F) Western blot of endogenous CTCF (CTCF antibody) and transgenic CTCF (FLAG antibody) in untreated (WT-U), IAA, ID conditions. See also Figures S1–S5.

Figure 2

CTCF mutations have unique chromatin binding profiles (A) Scheme showing the locations of the CTCF mutations within the ZF. The consensus CTCF motif highlights the triplet to which each mutant ZF binds (top). The most common motif for the de novo binding sites is shown below. Each bar graph shows the percentage of WT-only, common, and mutant-only CTCF binding sites. The heatmaps show the profile of CTCF and ATAC-seq signal at those sites. The UN condition corresponds to the FLAG control. (B) Profiles of ATAC-seq in WT and mutants. Wilcoxon p values were coded as follow: NS, not significant, ∗5 × 10⁻²–5 × 10⁻³, ∗∗5 × 10⁻³–5 × 10⁻⁴, ∗∗∗5 × 10⁻⁴–5 × 10⁻⁵, ∗∗∗∗<5 × 10⁻⁵. Data were generated on 2 replicates. See also Figures S6–S9.

Figure 3

Each mutation uniquely impacts CTCF’s chromatin bound fraction, residence time, and interaction with DNA (A) Plots of FRAP dynamics for WT and mutant CTCF. The bold lines show the fitted model of the average recovery, and the outlines give the 95% confidence intervals (95% CIs). (B) Violin plots of specific bound fractions. (C) Violin plots of specific residence times (min). p values were determined by bootstrapping (n = 2,500). (D) Heatmaps show the proportion of CTCF-cohesin versus CTCF-only binding sites. UN corresponds to the FLAG control in untreated cells. (E) Correlation between residence time and the percentage of CTCF-cohesin overlap. (F) Correlation between the FRAP-specific bound fraction relative to WT and the fraction of common CTCF sites relative to all potential binding sites. (G) Correlation between the FRAP-specific bound fraction relative to WT and the effect of CTCF binding on ATAC-seq signal at CTCF-SMC3 sites (Figure 2B). For (E)–(G), data were generated in 2 replicates, the p values were calculated using linear regression, and the shaded area corresponds to the 95% CI. See also Figures S10–S14.

Figure 4

Effect of CTCF binding and accessibility on SMC3 overlap (A) Bar graph showing the independent effect of CTCF (top) and ATAC-seq (bottom) signal on SMC3 enrichment in WT. The error bars correspond to the 95% CIs. The p values were calculated using a multivariate logistic model. (B) Percentage of SMC3 overlap in WT and CTCF mutants, stratified by ATAC-seq and CTCF signals. (C) SMC3 profiles in WT and mutant CTCF. p values were calculated using Wilcoxon tests. See also Figures S15, S16, and Table S1.

Figure 5

CTCF mutations alter gene expression, cellular reprogramming, and TF binding (A) Heatmap showing supervised clustering of the cell lines based on the expression levels of DEGs identified across the comparisons to WT. (B) Gene set enrichment analysis of DEGs in IAA condition and in H455R. The volcano plots below highlight the DEGs belonging to these enriched pathways. (C) Radar plots showing the averaged expression of developmental germ layer genes in WT and mutant mESCs cultured in LIF and no LIF conditions. (D) Heatmap showing predicted differentially bound TFs in WT, mutants and IAA. CTCF is highlighted with an asterisk (∗). (E) Volcano plots highlighting the differentially expressed target genes of CTCF and MBD2 in IAA and CTCF mutants. The metrics for enrichment of the target genes among the DEGs are reported on top on the volcanos (ORs and logistic p values). (F) Examples of altered CTCF binding and footprinting at the Rerg promoter (left) and altered MYC footprinting at the Brdt promoter (right). All data in this figure were generated on 2 replicates. See also Figures S17–S20 and Table S2. Differentially expressed genes (DEGs) identified using DE-seq. DEGs were defined as adjusted genes with p-value<0.05 and absolute log2 fold change > log2(1.5) compared to WT, Table S3. CTCF mutations altered cell differentiation, Table S4. CTCF mutations altered TF binding.

Figure 6

CTCF mutations alter chromatin interactivity (A) The top panels show the aggregated differential TAD analysis between mutants and WT. In the first panel, (a) indicates the intra-TAD interaction, (b) and (c) the inter-TAD interactions. The lower panels show the aggregated differential peak analysis. The data were generated by Hi-C on 2 replicates. (B) Correlation between the interaction counts and the FRAP residence times. (C) Correlation between the insulation score at CTCF peaks and the FRAP residence times. (D) Correlation between the loop extrusion length and the FRAP bound fractions. (E) Example of differential interactions between mutants and WT (right). The left matrix shows interactions in WT within a 10 Mb region (40 kb resolution) with the insulation score on the side. (F) Profiles show the averaged insulation score at WT-only, common, and mutant-only binding sites. (G) Profiles show the insulation score at CTCF binding sites stratified by CTCF signals. The bar graph shows the independent effect of CTCF signal on insulation score. (H) Profiles show the insulation score at CTCF binding sites stratified by ATAC signals. The bar graph shows the independent effect of ATAC signal on insulation score. For (F)–(H), p values reported in the profiles were calculated using Kruskal-Wallis tests. The fitted estimates for the bar graphs in (G) and (H) were obtained using a mixed multivariate model. The error bars correspond to the 95% CI. See also Figures S21–S24.

Figure 7

Changes in gene expression are linked to changes in chromatin interactivity but to a lesser extent than changes in TF binding at gene promoters (A) Bar graphs showing the enrichment of over-expressed (top) or under-expressed (bottom) genes in gained (top) or lost (bottom) loops in IAA and mutants compared to WT. ORs and p values were calculated using logistic models. (B) Example of 2 loci (blue and red rectangles) with a direct effect of gain in CTCF binding and chromatin interactivity. The interaction matrices (left) show gain of both intra- and inter-TAD interactions in some mutants compared to WT. The left panels show the zoom-in tracks of these loci. Only significant differential chromatin loops are shown. Overexpressed genes are highlighted in red. (C) Bar graph showing the percentage of DEGs resulting from direct or indirect effect of CTCF binding distinguishing loop dependent and independent effect.