CTCF-Mediated Enhancer-Promoter Interaction Is a Critical Regulator of Cell-to-Cell Variation of Gene Expression

Abstract

Recent studies indicate that even a homogeneous population of cells display heterogeneity in gene expression and response to environmental stimuli. Although promoter structure critically influences the cell-to-cell variation of gene expression in bacteria and lower eukaryotes, it remains unclear what controls the gene expression noise in mammals. Here we report that CTCF decreases cell-to-cell variation of expression by stabilizing enhancer-promoter interaction. We show that CTCF binding sites are interwoven with enhancers within topologically associated domains (TADs) and a positive correlation is found between CTCF binding and the activity of the associated enhancers. Deletion of CTCF sites compromises enhancer-promoter interactions. Using single-cell flow cytometry and single-molecule RNA-FISH assays, we demonstrate that knocking down of CTCF or deletion of a CTCF binding site results in increased cell-to-cell variation of gene expression, indicating that long-range promoter-enhancer interaction mediated by CTCF plays important roles in controlling the cell-to-cell variation of gene expression in mammalian cells.

Figure 1. CTCF binding sites interact with enhancers and promoters and positively correlate with gene activation

A. Interaction density among regulatory elements positively correlates with CTCF, Cohesin, GATA3, p300 and active histone modifications. The high-confidence interacting regions were sorted based on their interaction density (red: high; blue: low). The binding levels of chromatin proteins and histone modifications determined by ChIP-Seq were plotted as indicated at the bottom of the panel. B. Scatter plot shows a positive correlation between interaction density and H3K27ac level. C. Scatter plot shows a positive correlation between interaction density and CTCF binding level. D. Scatter plot shows a positive correlation between levels of CTCF binding and H3K27ac modification. E. More active genes are bound by CTCF than silent genes. The fraction of active or inactive promoters (+/− 2kb around TSS) bound by CTCF is plotted. F. Distribution of the closest p300 site relative to the CTCF sites in the genome. The location of CTCF sites is indicated by the arrow at bottom. The closest p300 sites are found and p300 binding levels are plotted (red: high; green: low). Displayed are the p300 sites located from 3 to 20kb away from the CTCF sites. G. Interaction density peaks at the CTCF and p300 binding sites. The interaction densities for the chromatin regions described in Panel F above are plotted. H. p300 sites interact with their neighboring CTCF sites. Plotted are the fractions of p300 sites that interact with the nearest CTCF sites. The background shows the interaction density with the chromatin regions without p300 binding at equal distance. I. Active promoters exhibit more looping with distal CTCF sites than silent promoters. Genes were separated to active genes (RPKM ≥3) and silent genes (RPKM ≤3). The number of CTCF binding sites (>5kb from TSS), which interact with the promoters, is plotted per promoter on Y-axis. J. Enhancers near or interacting with CTCF sites are more interactive. The p300-bound enhancers are separated to three categories: (1) with at least one CTCF site within a distance of 20 kb; (2) interacting with a CTCF site; and (3) others. The number of interacting promoters or enhancers is plotted for each category (Y-axis).

Figure 2. Knocking down of CTCF results in increased cell-to-cell variation

FACS analysis revealed increased cell-to-cell variation of expression of GATA3 (A), CD28 (B), CD90 (C), and CD5 (D). No significantly changed variation of expression was detected for Cohesin (E). Left panels show the distribution of gene expression with the x-axis indicating the expression level and y-axis indicating the cell density. Right panels are bar plots for the coefficient of variation that measures the expression variation of cells from individual replicate. APC signal were log10 transformed. Replicates for KD cells and control cells were paired based on experiment date. P-value was obtained by paired t-test.

Figure 3. CTCF binding sites at the Thy1 locus contribute to the functional interaction between Thy1 promoter and its enhancers and the expression noise control. CD90 protein is encoded by the Thy1 gene

A. The chromatin interactions surrounding the Thy1 gene locus are shown in the upper panel and H3K4me2, H3K4me3, H3K27ac and CTCF ChIP-Seq signals are shown in the lower panels. The red rectangle highlights the interactions between the CTCF binding site and Thy1 promoter and enhancers. The CTCF binding sites high-lighted by scissors were deleted separately in EL4 cells using CRISPR/CAS9. B. Deletion of the 1^st CTCF binding site decreased Thy1 mRNA levels. Total RNAs isolated from the wild type or CTCF site deletion EL4 clones were analyzed by quantitative reverse-transcription PCR and normalized to GAPDH. C. The 1^st CTCF site deletion abolished CTCF binding and compromised the H3K27ac modification at the Thy1 gene locus. The genome browser images show the ChIP-Seq data for CTCF binding, H3K27ac and chromatin input signals in wild type and CRISPR deletion EL4 cells. The high-lighted CTCF peak indicates the location of CRISPR deletion. D. The 1^st CTCF site deletion compromised the enhancer-promoter interaction in the Thy1 gene locus. Top panel shows the H3K4me2 and H3K27ac ChIP-Seq signals. The red and green horizontal lines below the ChIP-Seq tracks indicate the different TADs called by HMM. The bottom panel shows the relative chromatin interaction intensity of the Thy1 promoter with various enhancer regions indicated above the top panel (R1 to R10) by 3C analysis. The blue rectangle marked as R6 is the anchor site for the 3C analysis. The red rectangle marked R8 region is the deleted 1^st CTCF site. Data show average of two independent experiments and are represented as mean ± SEM. WT: control cells; KO: CRISPR/CAS9 deletion cells. E. Deletion of the 1^st CTCF site resulted in increased cell-to-cell variation of expression of CD90 protein encoded by the Thy1 gene as measured by FACS assay.

Figure 4. Single-molecule RNA-FISH shows increased cell-to-cell variation of Thy1 mRNA in the CTCF site-deleted cells

A. Typical images of RNA-FISH for detecting CTCF mRNA (red), Thy1 mRNA (green) and DNA (blue). B. Box plots showing the numbers of CTCF and Thy1 mRNA molecules per cell in wild type EL4 cells. The box plot is from one representative of 4 replicates. C. Box plots showing the numbers of CTCF and Thy1 mRNA molecules per cell in the 1^st CTCF site-deleted EL4 cells. The box plot is from one representative of 12 replicates. D. Deletion of the 1^st CTCF site results in increased coefficient of variation in the number of Thy1 mRNAs per cell. The bar plot shows the distribution of CVs of 4 replicates for WT and 12 replicates for KO. Data represented as mean ± SEM. P-value was obtained by t-test. E. The variation of Thy1 mRNA per cell caused by deletion of the 1^st CTCF site is related to the number of CTCF mRNA in the cell. On the X-axis, the cells are sorted to four groups according to the number of CTCF mRNAs per cell (0–14; 15–29; 30–44; and >44). Y-axis indicates the coefficient of variation of Thy1 mRNA. Grey bars indicate the wild type EL4 cells and red bars indicate the CTCF site deletion EL4 cells. Each bar represents one replicate. F. CTCF and Cohesin organize chromatin to large domains (left panel) and facilitate long-distance enhancer-promoter interaction to decrease fluctuation of expression and maintain robustness of expression (right panel).