RNA Interactions Are Essential for CTCF-Mediated Genome Organization

Abstract

The function of the CCCTC-binding factor (CTCF) in the organization of the genome has become an important area of investigation, but the mechanisms by which CTCF dynamically contributes to genome organization are not clear. We previously discovered that CTCF binds to large numbers of endogenous RNAs, promoting its self-association. In this regard, we now report two independent features that disrupt CTCF association with chromatin: inhibition of transcription and disruption of CTCF-RNA interactions through mutations of 2 of its 11 zinc fingers that are not required for CTCF binding to its cognate DNA site: zinc finger 1 (ZF1) or zinc finger 10 (ZF10). These mutations alter gene expression profiles as CTCF mutants lose their ability to form chromatin loops and thus the ability to insulate chromatin domains and to mediate CTCF long-range genomic interactions. Our results point to the importance of CTCF-mediated RNA interactions as a structural component of genome organization.

Keywords: CTCF; RNA binding; RNA deficient-mutants; TADs; chromatin domains; chromatin loops; chromatin organization; gene expression; transcriptional inhibition.

Figure 1.. Transcriptional Inhibition Disrupts CTCF Binding Predominantly in TSS

Transcription was inhibited in mESCs for 4 h with co-incubation of DRB and triptolide. Cells incubated with DMSO served as control. (A) Shows CTCF ChIP-seq heatmaps centered and rank-ordered on CTCF-binding sites. Corresponding average density profiles are plotted at the top of the heatmaps to illustrate differences between DMSO and 4 h of TI. B) A subset of peaks exhibit dramatically reduced CTCF enrichment after 4 h of TI. Overlapping peaks for TSS are highlighted in blue. (C) Average density profiles for the same ChIP-seq as (A) but centered on TSS. (D) Boxplot showing the motif affinity scores for CTCF-binding sites lost after transcriptional inhibition versus a random set of CTCF-binding sites in the control (Mann-Whitney test, p < 0.0001). (E) Representative example of a CTCF peak with decreased binding to the TSS of the Slain2 gene. ChIP-seq tracks for DMSO (gray) and 4 h of TI (red) are overlapped for comparison. (F) 5C heatmap depicting the interaction frequency between restriction fragments across a 4 Mb region surrounding the HoxA cluster (data were binned in 15 kb windows; step size 5 kb; the median is shown). Comparative 5C heatmap shows increased (red) and decreased (blue) interactions after TI. Overlapped ChIP-seq tracks above illustrate decreased binding of CTCF. Darker colors represent increasing interaction frequency. (G) Zoom into a chromatin domain delimited by CTCF sites (top). Overlapped ChIP-seq tracks for DMSO (gray) and 4 h of TI (red) illustrate no change in CTCF binding for the loop enclosed in a rectangle (bottom).

Figure 2.. Deletions in ZF1 and ZF10 Independently Abolish CTCF Binding to RNA

(A) Schematic representation of known domains of WT CTCF with its 11 zinc fingers being numbered (top); smoothed residue-level RBR-ID score (He et al., 2016), plotted along the primary sequence (bottom). (B) FACS analysis highlighting percentage of GFP+ or mCherry+ CTCF-AID-GFP mESCs with or without rescue of CTCF: WT, ZF1Δ, or ZF10Δ. (C) Immunoprecipitation of all rescue cell lines indicated and immunoblots for CTCF and Rad21. (D) Representative image of GFP-CTCF incubated with each rescue, immunoprecipitated with a GFP antibody and blotted against CTCF (left); bar graph quantification of each rescue protein relative to the GFP-CTCF (n = 5) (right). (E) PAR-CLIP of stably expressed WT and mutant CTCF in mESCs. Autoradiography for ³²P-labeled RNA (top) and control western blot (middle and bottom). (F) Schematic representation of ZF1 and ZF10 of CTCF; mutations found in breast and endometrial cancer that alter zinc binding are shown in black; mutations that do not alter zinc binding are in blue, and RBR-ID deletions are in brackets.

Figure 3.. Gene Expression Defects Are Partially Preserved between RBR Deletions

(A) Principal-component analysis (PCA)-based representation of single-cell RNA-seq data for rescue cell lines from WT (gray), ZF1Δ (black), and ZF10Δ (red). Each dot represents a single cell, and dots are arranged on the basis of PCA. The final number of cells sequenced per condition is noted in parentheses. (B) Heatmaps depicting differentially expressed genes from scRNA-seq. (C) Venn diagrams showing the overlap between differentially expressed genes for the different conditions and levels of significance. (D) Bar graph illustrating the percentage of genes that have at least one CTCF-binding site for CTCF in the promoter region or gene body. (E) Boxplot showing the motif affinity scores for CTCF-binding sites within DEG represented in (B) compared with a random sample of genes (Mann-Whitney test, p < 0.0001).

Figure 4.. Deletion of RBRs in CTCF Disturb Its Chromatin Binding

(A) CTCF ChIP-seq for WT (gray), ZF1Δ (black), and ZF10Δ (red) rescue cell lines. Heatmaps were generated by centering and rank-ordering on CTCF-binding sites. Those lost for ZF1Δ (top) or ZF10Δ (bottom) are shown. (B) De novo motif discovery was called for binding sites in (A), and a black box encloses the eighth position in which A to G was specifically preferred by ZF1Δ. (C) Boxplot showing the motif affinity scores for CTCF-binding sites in (A) compared with unchanged sites (Mann-Whitney test, p < 0.0001). (D) Bar graph representing the top three genomic regions for CTCF sites in (A). (E and F) Mean expression levels for differentially expressed gene Cdkn2a (E) and corresponding ChIP-seq tracks (F) under each condition.

Figure 5.. RBR Mutants Disturb Chromatin Loops

(A) Aggregate peak analysis (APA) was used to measure the aggregate strength of chromatin loops annotated by HICCUPS on the WT rescue. Loop strength is indicated by the extent of focal enrichment at the center of the plot. APA scores are shown on the bottom left. (B) Bar graph representing APA scores between rescue conditions. (C) Bar graph representing the number of chromatin loops annotated by HICCUPS for each individual condition. (D) Representative contact matrix (at 5 kb resolution) shows that the chromatin loop in the WT rescue (left) disappears in the ZF1Δ (middle) or loses strength in ZF10Δ (right), while CTCF binding is lost at one of the anchors. (E) Same as (D), but in this example CTCF remains bound under all conditions.

Figure 6.. RBR Mutants Disturb Chromatin Loops

(A) Intra-domain interaction changes in WT versus ZF1Δ and WT versus ZF10Δ for common domains. CTCF mutant rescues are associated with gain (red) and loss (blue) of intra-domain interactions. (B) Boxplots representing the correlation between DEG and chromatin domains whose interactions are increased. Only downregulated genes for ZF1Δ are significantly correlated with increased intra-domain interactions, while all others are not significantly correlated. (C) Representative contact matrix (at 5 kb resolution) show that the chromatin loop anchor on the left overlaps with the promoter of the Syne2 gene. In the WT rescue (left), the loop disappears in the ZF1Δ (middle) and is stable in ZF10Δ (right), while CTCF binding is lost only at the promoter and anchor for ZF1Δ. Syne2 is upregulated only under the ZF1Δ condition. (D) Graphical representation of a chromatin loop formed by two CTCF proteins (green) and cohesin rings (blue), stabilized by an RNA in the WT condition (left diagram). Two outcomes are observed for the RNA binding-deficient mutants (ZF1Δ and ZF10Δ): (1) the loop is lost and a CTCF protein loses its binding to chromatin (top, right diagram), or (2) both CTCF proteins remain bound to chromatin yet the chromatin loop is still lost (bottom, right diagram).