MYC reshapes CTCF-mediated chromatin architecture in prostate cancer

Abstract

MYC is a well characterized oncogenic transcription factor in prostate cancer, and CTCF is the main architectural protein of three-dimensional genome organization. However, the functional link between the two master regulators has not been reported. In this study, we find that MYC rewires prostate cancer chromatin architecture by interacting with CTCF protein. Through combining the H3K27ac, AR and CTCF HiChIP profiles with CRISPR deletion of a CTCF site upstream of MYC gene, we show that MYC activation leads to profound changes of CTCF-mediated chromatin looping. Mechanistically, MYC colocalizes with CTCF at a subset of genomic sites, and enhances CTCF occupancy at these loci. Consequently, the CTCF-mediated chromatin looping is potentiated by MYC activation, resulting in the disruption of enhancer-promoter looping at neuroendocrine lineage plasticity genes. Collectively, our findings define the function of MYC as a CTCF co-factor in three-dimensional genome organization.

Fig. 1. Profiling of regulatory element interactome in PCa cell lines.

a Overview of HiChIP experiments and analyses in PCa cell lines. b The number of significant loops (FDR < 0.05) in VCaP HiChIP assays. From top to bottom, n = 74514, 61748, 46819, 9751, 24982, 24737, respectively. c Strength of H3K27ac and AR loops anchored at TSSs of all protein-coding genes. The frequency curves showed the normalized read number (loop strength) distribution at anchors distal to TSSs. The strength of each loop was normalized by the number of total ‘cis-far’ unique valid pairs. P-values were determined by paired t-test. For TSSs, n = 19962. d Strength of H3K27ac loops anchored at genes downregulated or upregulated by 2 h DHT treatment. The bar plot summarized the strength of loops-anchored gene TSSs. P-values were determined by paired t-test. For TSSs of upregulated and downregulated genes, n = 350 and 205, respectively. e Strength of AR loops anchored at genes downregulated or upregulated by 2 h DHT treatment. The bar plot summarized the strength of loops anchored at gene TSSs. P-values were determined by paired t-test. For TSSs of upregulated and downregulated genes, n = 350 and 205, respectively. f The expression of IL20RA was upregulated from the early time point (2 h) after DHT stimulation. n = 2. Data represent means ± SD. g For the IL20RA gene, AR loops were boosted as early as 2 h after DHT stimulation. h 3C–qPCR assay of the IL20RA genomic region. The data represents relative frequencies of interaction between the anchor region near the IL20RA TSS and selected PstI digestion sites (circles). n = 3. Data represent means ± SD. P-values were determined by Student’s t-test. *P < 0.05; **P < 0.01. Source data are provided as a Source Data file.

Fig. 2. Cell-type-specific CTCF looping regulates PCa-associated genes.

a Scatter plots showing the association of CTCF loop strength between two replicates in VCaP and 22Rv1, respectively. n = 167054 and 218739 from left to right, respectively. P-values were calculated using Pearson correlation. b Scatter plots showing the association of CTCF loop strength between VCaP replicate #1 and 22Rv1 replicate #1. n = 217761. P-value was calculated using Pearson correlation. c Scatter plots showing the association of CTCF loop strength between VCaP replicate #2 and 22Rv1 replicate #2. n = 210816. P-value was calculated using Pearson correlation. d Boxplots showing the strength and distances of CTCF loops classified based on cell-type specificity and H3K27ac status of two anchors from the same loop. Box plots indicating the mean (middle line), 25th and 75th percentile (box) and 10th and 90th percentile (whiskers), and n = 24661, 24708, 8804, 17267, 21223, 8967, 12496, 11277, 3348 for boxes from left to right, respectively. P-values were calculated by Wilcoxon signed-rank test. e The number of genes within CTCF loops. The gene numbers were normalized by CTCF loop distances. Box plots indicating the mean (middle line), 25th and 75th percentile (box) and 10th and 90th percentile (whiskers), and n = 24661, 24708, 8804, 17267, 21223, 8967, 12496, 11277, 3348 for boxes from left to right, respectively. P-values were calculated by Wilcoxon signed-rank test. f Average log2 expression fold changes (22Rv1 vs. VCaP) of genes within CTCF loops. Box plots indicating the mean (middle line), 25th and 75th percentile (box), and 10th and 90th percentile (whiskers), and n = 24661, 24708, 8804, 17267, 21223, 8967, 12496, 11277, 3348 for boxes from left to right, respectively. P-values were calculated by Wilcoxon signed-rank test. g KEGG pathway enrichment analysis of genes annotated to the anchors of CTCF loops. The proximity of genes to anchors is restricted to 3Kb. h Upper: A cell-type-specific CTCF binding insulates the interaction between a distal enhancer and TMC5 promoter by forming CTCF-CTCF loops. The 22Rv1-specific CTCF binding was highlighted within a red dashed rectangle. Bottom: TMC5 expression levels in VCaP and 22Rv1 cells based on RNA-Seq data. i Pearson correlation coefficients between DNA methylation ratio of CpGs and CTCF binding affinity at this CTCF site in 20 ENCODE cell lines. Each circle denotes a CpG within this CTCF site. j Scatter plot showing a positive correlation between average DNA methylation levels at this CTCF site and TMC5 expression levels in Changhai 2020 data set. The error band indicates SEM. P-value was calculated using Pearson correlation.

Fig. 3. Deletion of a CTCF site near the MYC promoter leads to re-organization of CTCF looping.

a The scatter plot showing the fold changes of CTCF binding affinities at CTCF sites looping to gene promoters and the expression fold changes of corresponding genes. The interactions between CTCF sites and gene promoters were determined by CTCF HiChIP loops. VCaP and 22Rv1 CTCF ChIP-Seq data were obtained from ENCODE and gene expression data were retrieved from GSE25183. The CTCF-gene pairs with consistent binding/expression fold changes and both absolute fold changes > 1.5 are labelled red, and opposite fold changes are labelled blue. The dots and triangles indicate CTCF sites with and without H3K27ac overlapping, respectively. Genes in representative CTCF-gene pairs were annotated. n = 2088. P-value was calculated by Chi-square test. b Upper: The highlighted CTCF site was connected to MYC promoter by a CTCF loop. The CTCF binding affinities at this site (−10 Kb from MYC promoter) were negatively correlated with MYC expression in prostate cancer cell lines. The CTCF site −10 Kb from MYC promoter was deleted by CRISPR/Cas9-mediated knock-out in 22RV1 cells. The control (sgCtrl) and CTCF deletion (sgDele-10Kb) cells were then used for H3K27ac and CTCF HiChIP experiments. Bottom: Significantly changed H3K27ac loops at MYC region by the “−10 Kb CTCF site” deletion. For each group of H3K27ac HiChIP, two biological replicates were performed. c Number of significantly changed CTCF loops at each chromosome before and after “−10Kb CTCF site” deletion. d Enrichment analysis of MYC binding at the anchors of dysregulated CTCF or H3K27ac loops. Orange points represent the actual ratio of dysregulated loop anchors with MYC binding. Each box represents 500-time random sampling from all CTCF or H3K27ac loop anchors. Same number of loop anchors as in the sgCtrl-specific or sgDele-10Kb-specific anchor set was used for random sampling, respectively. O/E (observed vs. expected) was calculated by comparing the overlap percentage of actual dysregulated loop anchors with that of the average of randomly sampled anchors. Box plots indicating the mean (middle line), 25th and 75th percentile (box), and 10th and 90th percentile (whiskers). Points were highlighted by red if P < 0.05. P-values were determined by Student’s t-test. n = 215, 297, 2896, 2322 from left to right, respectively. e Motifs enriched in the CTCF peaks at CTCF anchors of indicated loops. f Motif distribution in CTCF peaks at CTCF anchors of indicated loops. For each CTCF peak, the location of the best-ranked CTCF motif was used as the center for the motif density plot. g The aggregated CTCF ChIP-Seq signal in 22Rv1 cells at indicated peaks.

Fig. 4. MYC facilitates CTCF chromatin binding.

a Multi-omics assays to evaluate the function of MYC on CTCF binding and looping in 22Rv1 cells. b Western blot assay showing the efficiency of MYC overexpression. c The overlap of MYC, CTCF and H3K27ac peaks between control (Ctrl) and MYC overexpression (MYC-OE) cells. d Heatmaps showing the ChIP-Seq signal of MYC, CTCF and H3K27ac at MYC and CTCF peaks. From top to bottom, MYC and CTCF peaks were separated into shared, MYC-only, and CTCF-only groups. e The overlap between MYC and CTCF peaks in MYC-OE cells. f The aggregated CTCF ChIP-Seq signal in Ctrl and MYC-OE cells. CTCF peaks were divided into four groups based on MYC and H3K27ac status. g The aggregated H3K27ac ChIP-Seq signal in Ctrl and MYC-OE cells. H3K27ac peaks were divided into four groups based on MYC and CTCF status. h Co-immunoprecipitation to detect the protein-protein interaction between CTCF and MYC in Ctrl and MYC-OE cells. i Western blot and corresponding Coomassie blue staining of GST pull-down assay. 22Rv1 cell lysates were subjected to pulldowns with immobilized GST only or recombinant GST-CTCF protein. Bound protein was probed with anti-CTCF and anti-MAX antibodies by Western blot. The red arrow indicates the GST-MYC band. j 22Rv1 cells expressing GFP-MYC and Flag-CTCF were used in co-IP assay. The input and IPed proteins were analyzed by Western blot with anti-GFP and anti-Flag antibodies. k Proximity Ligation Assay to detect the in-situ interaction between MYC and CTCF proteins. Nuclei were stained with DAPI. Scale bar, 10 μm. For b, h, i, j, and k, these experiments were repeated independently three times with similar results. Source data are provided as a Source Data file.

Fig. 5. MYC represses neuroendocrine genes by promoting CTCF looping.

a Volcano plot showing dysregulated genes by MYC overexpression. n = 2. b GO BP enrichment analysis of 479 MYC-repressed genes. The GO terms with high similarity were clustered together and summarized. n = 2. c GSEA plot showing the pan-NET genes were overall suppressed by MYC overexpression. d Boxplots showing H3K27ac loop strength fold changes at the promoters of upregulated or downregulated genes after MYC overexpression. Box plots indicating the mean (middle line), 25th and 75th percentile (box) and 10th and 90th percentile (whiskers). n = 854, 5024 from left to right, respectively. P-value was determined by Student’s t-test. e The length distribution of CTCF loops dysregulated by MYC. f Normalized CTCF HiChIP contact matrices of Ctrl and MYC-OE cells at a genomic region of chromosome 2. The black rectangles indicate TAD structures. Black arrows highlight the intra-TAD CTCF looping enhanced by MYC. Red arrows highlight the corresponding CTCF binding increased by MYC. g The analysis workflow to identify genes with H3K27ac loops disrupted by increased CTCF looping after MYC overexpression. h Ranked dot plot showing the decreased H3K27ac loop strength (MYC-OE - Ctrl) at promoters of 4557 genes from the analysis in (g). Red and blue circles indicate genes up- and downregulated by MYC, respectively. The annotated circles were downregulated pan-NET genes. Bar plot showing the overlapping between MYC-dysregulated genes and the 4557 genes. The P-value was determined by Chi-square test. i Dysregulated CTCF and H3K27ac loops at the genomic region spanning the CDK5R2 gene in Ctrl and MYC-OE cells. MYC-induced CTCF binding and looping were highlighted. j Gene set z-scores of MYC targets and pan-NET genes were negatively correlated in two PCa RNA-Seq data sets. Spearman rho and P-value were shown. Box plots indicating the mean (middle line), 25th and 75th percentile (box) and 10th and 90th percentile (whiskers). n = 83, 167, 83, 9, 16, 9 from left to right, respectively. k Top: Three CRISPRi sgRNAs were designed to target the three CTCF sites upstream of CDK5R2 gene, respectively. Bottom: CTCF ChIP-qPCR and RT-qPCR with or without CRISPRi in MYC overexpressed 22Rv1 cells. n = 3. Data represent means ± SD. P values were two-sided Student’s t test. *P < 0.05; **P < 0.01. Source data are provided as a Source Data file.