Large-Scale Topological Changes Restrain Malignant Progression in Colorectal Cancer

Abstract

Widespread changes to DNA methylation and chromatin are well documented in cancer, but the fate of higher-order chromosomal structure remains obscure. Here we integrated topological maps for colon tumors and normal colons with epigenetic, transcriptional, and imaging data to characterize alterations to chromatin loops, topologically associated domains, and large-scale compartments. We found that spatial partitioning of the open and closed genome compartments is profoundly compromised in tumors. This reorganization is accompanied by compartment-specific hypomethylation and chromatin changes. Additionally, we identify a compartment at the interface between the canonical A and B compartments that is reorganized in tumors. Remarkably, similar shifts were evident in non-malignant cells that have accumulated excess divisions. Our analyses suggest that these topological changes repress stemness and invasion programs while inducing anti-tumor immunity genes and may therefore restrain malignant progression. Our findings call into question the conventional view that tumor-associated epigenomic alterations are primarily oncogenic.

Keywords: DNA methylation; chromatin; colon cancer; compartment; epigenetics; genome topology; nuclear architecture.

Figure 1.. Integrated Topological Maps Reveal Tumor-Specific Chromatin Loops and Stable TAD Structure

(A) Schematic of hierarchical genome organization with indication of genomic scale (left) and summary of genome-wide assays (center) and models (right). (B) Volcano plot presenting a differential analysis of loops between tumors and normal samples. Loops, represented as dots, with significantly stronger or weaker interactions in tumors compared with normal colon are highlighted in red and green, respectively. (C) Boxplots depicting expression fold change (log₂) between tumors and normal samples (y axis) for genes engaged in enhancer-promoter (E-P) loops. Genes are stratified by change in E-P loop strength between tumors and normal colon (x axis). (D) Genomic view of the EPHA2 locus (~130 kb), showing SMC1 HiChIP loops (arcs) and H3K27ac enrichment for normal colon (green) and colon tumor (purple). The width of the arcs corresponds to the average loop strength summarized for the set of 2 normal and 7 colon samples. An asterisk indicates the differential loop (STAR Methods). (E) Genomic view of the PDCD4 locus (~100 kb) as in (D). (F) Hi-C contact map showing pairwise contact frequencies (red heat) between genomic positions across chromosome 7 (rows, columns) in normal colon. Top: Hi-C eigenvector (PC1) based on long-range interactions demarcates compartments A (positive values, blue) and B (negative values, yellow). Right: inset with a magnified view of are presentative region reveals TAD structures (highlighted by black triangles). Rotation of this inset by 45° yields a horizontal display of TAD structures (see G). (G) Horizonal heatmaps showing local Hi-C contact patterns (red heat) across chromosome 14 for normal colon (green), colon tumors (purple), and cell lines (black). Exemplar TAD boundaries are indicated by black arrows.

Figure 2.. Compromised Partitioning and Positioning of Genome Compartments in Tumors

(A) Hi-C eigenvectors (PC1) based on long-range interactions demarcate compartments A (positive values, blue) and B (negative values, yellow) across a 45-Mb region of chromosome 6. Data show eigenvectors for normal colon (green), colon tumors (purple), and cell lines (black). (B) Heatmap showing pairwise correlations between the first Hi-C eigenvector (blue heat) in normal colon (green), colon tumors (purple), and cell lines (gray). Samples (rows, columns) are ordered according to complete linkage hierarchical clustering (top). (C) Heatmap showing fold change (log₂) in Hi-C contact frequencies between colon tumors and normal colon across chromosome 1. Data are based on an average of normal colons (n = 4) and tumors (n = 7). Interactions that increase in tumors (red) or decrease in tumors (green) are evident. Top left: the Hi-C eigenvector indicates compartment assignments in normal colon (A, blue; B, yellow). (D) Schematic of the maximum entropy modeling approach, in which structural models of genome organization are generated and iteratively refined to improve the correlation between Hi-C maps derived from these models in silico and the actual experimental Hi-C data. (E) Whole-nucleus maximum entropy models (1-Mb resolution) for are presentative normal colon sample (N1), showing compartment A in blue and compartment B in yellow. (F) Whole-nucleus maximum entropy models (1-Mb resolution) for a representative colon tumor sample (T1) as in (E). (G) Representative transmission electron microscopy (TEM) image of nuclei from normal colon epithelium, showing electron dense heterochromatin (HC) along the nuclear membrane and internally distributed euchromatin (EC). Scale bar, 1 um. (H) Representative EM image of colon tumor nucleus labeled as in panel (G). Scale bar, 1 um. (I) Top: Schematic of strategy for quantifying the internal HC in nuclei imaged by EM. Bottom: Histogram shows fraction of nuclei (y axis) that have indicated percentage of internal HC. Normal (green) reflects102 epithelial cell nuclei from 3 normal colon specimens. Tumor (purple) reflects 184 malignant cell nuclei from 3 colon tumors. Two-sided nested t test, p = 0.006. (J and K) Representative image of chromosome 12 DNA FISH in nuclei from normal colon epithelial cells (J) and malignant tumor cells (K), labeling compartment A and B regions in blue and yellow, respectively. (L) Histogram showing a fraction of the signal from compartment B DNA FISH probes in successive radial bins from the center to the periphery of the nuclei. Normal (green) reflects 47 epithelial cell nuclei from 2 normal colon specimens. Tumor (purple) reflects 82 malignant cell nuclei from two tumors for which chromosome 12 was copy number stable (T1 and T4). Innermost bins 1–10 were aggregated.

Figure 3.. An Intermediate Compartment I Is Also Reorganized in Tumors

(A) Heatmap depicting the Hi-C eigenvector used to define compartments, as in Figure 2A, except with the eigenvector values shown as heatmaps (blue, positive; yellow, negative). Data are shown for a ~2.3-Mb region (x axis) for normal colon samples (rows with green labels) and colon tumors (rows with purple labels). Hypomethylated blocks are indicated (black bars). (B) Heatmap depicting the first eigenvector for a 5-Mb region on chromosome 3 as in (A). (C) Hi-C contact map (top) and first eigenvector (PC1; center) for are presentative tumor sample. DNA methylation levels of low-density (open-sea) CpGs (bottom) are shown for three representative normal colon samples (green) and three representative tumors (purple). The represented region on chromosome 17 contains a 3.6-Mb TAD assigned to compartment B (black square) with a hypomethylated block (gray highlight). (D) Hi-C contact map, first eigenvector (PC1) and DNA methylation are shown for a 200 kb TAD assigned to compartment A with a hypomethylated block (data presented as in panel (C). (E) Hi-C contact map for chromosome 1 for a representative normal colon sample. Top left: colored bars indicate genomic regions assigned to compartment A (blue), compartment I (light blue), or compartment B (yellow). Magnified panels for representative regions highlight the intermediate long-range inter-compartmental interaction pattern typical of compartment I. (F) First and second eigenvectors of the chromosome 1 Hi-C matrix for normal colon. Each point represents one 100-kb bin, colored by compartment. (G) Whole-nucleus maximum entropy model (100-kb resolution) for normal colon, showing compartments A, I, and B. (H) Representative DNA FISH image (left) and high-magnification image (right) for HCT116 cell nuclei. Signal intensities are shown for compartment A (blue), I (light blue), and B (yellow) regions on chromosome 12. (I) Barplot indicating the percentage of cells for which the maximum DNA FISH signal intensity for compartment A, B, or I is located at the indicated radial position for 305 HCT116 cell nuclei. (J) Representative image of nuclei from normal colon epithelial cells. The image shows DNA FISH signal intensities of probes for compartments A, B, and I of chromosome 12. Two chromosome territories are magnified in the insets.

Figure 4.. Distinct Chromatin States Support a Three-Compartment Model

(A–C) Plots show 3-compartment model assignments for representative regions on chromosome 20 (A), chromosome 6 (B), and chromosome 4 (C). Dark blue, compartment A; light blue, compartment I; yellow, compartment B. DNA methylation levels of low-density (open-sea) CpGs for three normal samples (green) and three tumor samples (purple) are shown below, along with ChlP-seq profiles for H3K27ac, H3K27me3, and H3K9me3 for a representative normal colon sample (green) and tumor (purple). Hypomethylated blocks are shaded in gray. (D) Heatmap showing relative levels of H3K27ac, H3K36me3, H3K27me3, and H3K9me3 in normal colon (rows) for compartments A, I, and B (columns). DNA methylation differences between normal colon and tumors (for open-sea CpGs) and relative gene expression in normal colon (TCGA) are also shown. (E) Plots showing mean and standard deviation of fold change (log₂) in enrichment of the indicated modification between tumors and normal samples. The respective plots show data for different modifications and are stratified by compartment (x axes).

Figure 5.. Compartmental Reorganization Is Closely Associated with DNA Hypomethylation and Accumulated Divisions

(A) Association between Hi-C eigenvector (PC1) and hypomethylation of 100-kb windows in tumor samples. Data are stratified by compartment and extent of hypomethylation. Points and horizontal bars represent point estimates and 95% confidence intervals from a linear regression model. (B) Data visualized as in (A), showing the association between Hi-C eigenvector and hypomethylation for HCT116 cells treated with 5′-aza versus DMSO. (C) Plot showing DNA methylation forWI-38 cells at passages 16 (black), 30 (dark gray), and 40 (light gray). Data are stratified by compartment. Bars represent the average of two replicates (dots). (D) Plot showing DNA methylation (for open-sea CpGs) for a representative region on chromosome 10. Traces represent methylation values for WI-38 fibroblasts at passages 16 (black, n = 2), 30 (dark gray, n = 2), and 40 (light gray, n = 2). Compartment assignments for WI-38 are shown at the top (A, blue; I, light blue; B, yellow). (E) Data visualized as in (A), showing the association between Hi-C eigenvector and hypomethylation for late (passage 40 [P40]) versus early (P16) WI-38 cells. (F) Plot showing change over time in Hi-C eigenvector for 100-kb windows that show more than 20% hypomethylation in late-passage (P40) versus early-passage (P16) WI-38 fibroblasts.

Figure 6.. Compartmental Reorganization Linked to Tumor-Suppressive Transcriptional Programs

(A) Volcano plot depicting the association (x axis) between expression and block hypomethylation for genes in compartment B, computed across tumors in the TCGA colorectal cohort with a purity of more than 40%. The y axis indicates the significance of the association. Genes (points) plotted toward the top right are upregulated in association with block hypomethylation. They are highly enriched for cancer germline antigens (CGAs; red) and ERV elements (green). Genes plotted toward the top left are downregulated in association with hypomethylation. A magnified panel (below) highlights high-confidence downregulated genes after excluding genes expressed in non-epithelial cell types (black; STAR Methods). Genes related to EMT, Wnt signaling, invasion, and metastasis are labeled. (B) Data presented as in (A) for compartment I genes. (C) Boxplot showing association between expression change and DNA block hypomethylation for genes in compartments A, B, and I. Negative values indicate repression in association with DNA hypomethylation, and positive values indicate upregulation in association with hypomethylation. (D) Functional gene set annotations enriched (false discovery rate [FDR] < 20%) among 146 high-confidence downregulated B/I genes (from the insets in A and B). *, enriched annotations included multiple overlapping sets related to embryonic development (Table S6). (E) Plot showing average block methylation levels for normal colon biopsies (y axis) as a function of donor age (x axis). Each point represents one sample from a low-risk (green) or high-risk (magenta) donor (Wang et al., 2020). Linear regression fit (lines) and 95% confidence intervals (shades) are indicated. (F) Plots showing average expression of the high-confidence downregulated B/I genes (from the insets in A and B) in clinical specimens. Each point corresponds to a different sample from a cohort of colorectal tumors and normal colons (Cancer Genome Atlas Network, 2012). The y axis represents log₂-normalized counts. (G) Kaplan-Meier curve depicting survival outcomes of patients stratified by their average tumor expression of the high-confidence downregulated compartment B/I genes.

Figure 7.. Compartment Shifts in Excessively Replicated Cells Restrain Malignant Progression

The schematic depicts compartment shifts and proposed functional consequences. (A) In normal nuclei, compartments A, B, and I are robustly partitioned and spatially segregated. In tumor or aging cells that have accumulated excess divisions, compartmental organization is compromised, and compartment-specific epigenetic states are altered. (B) Exemplar loci are shown for each compartment in normal (top) or pathologic (bottom) states. Hypomethylation of compartment B induces ERVs and CGAs, which promote anti-tumor immunity. Repressive chromatin in compartments B and I downregulates genes associated with EMT, invasion, and stemness.