Topoisomerase IIb binding delineates localized mutational processes and driver mutations in cancer genomes

Abstract

Type-II topoisomerases resolve topological stress in DNA through double-strand breaks. While topoisomerases are chemotherapy targets linked to therapy-related genotoxicity, TOP2B is uniquely positioned to influence mutagenesis through its activity in non-dividing cells and sensitivity to topoisomerase poisons. To investigate this, we generated DNA-binding maps of TOP2B, CTCF, and RAD21 in human cancer samples and analyzed these for driver mutations and mutational processes across 6500 whole cancer genomes. TOP2B-CTCF-RAD21 and TOP2B-RAD21 sites are enriched in somatic mutations and structural variants, particularly at sites with evolutionary conservation, high transcription and long-range chromatin interactions. TOP2B binds driver genes such as TP53, MYC, FOXA1, and VHL, and many frequently mutated non-coding regions. We show that one non-coding TOP2B-bound element at the non-coding RNA gene RMRP drives tumor initiation and growth in vivo. Our study highlights TOP2B as a safeguard of genome integrity and a marker of mutational processes and hotspots in cancer, underscoring implications for cancer genomics research.

Fig. 1. Genome-wide binding landscape of TOP2B in human cancer cells.

a Overview of ChIP-seq analysis. TOP2B, CTCF, and RAD21 binding sites were identified in clinical samples of human hepatocellular carcinomas. Venn diagram shows sites bound by the three factors. b Evolutionary conservation, pan-tissue activity, and functional characteristics of binding sites. TOP2B-CTCF-RAD21 sites and CTCF-RAD21 sites are highly conserved based on DNA-binding profiles of CTCF and TOP2B in mammalian livers, occur frequently at constitutively active CTCF binding sites measured from 70 human cell lines, and are enriched in sites with double strand break (DSB) activity. P-values are from one-tailed hypergeometric tests. c Chromatin states of TOP2B binding sites in normal tissue types from the Roadmap Epigenomics project (X-axis). Chromatin states of TOP2B-CTCF-RAD21 sites and TOP2B-RAD21 sites were compared to sites lacking TOP2B binding (i.e., CTCF-RAD21 sites and RAD21-only sites, respectively). Colors in the heatmap show log odds ratios and asterisks show false discovery rate (FDR)-adjusted P-values (q-values) from two-tailed hypergeometric tests (FDR < 0.05). Tissues and chromatin states are ordered by total significance (TSS, transcription start site). d Transcriptional associations of TOP2B binding sites in cancer. Violin plots show mean gene expression levels of genes adjacent to binding sites. P-values are from two-tailed Wilcoxon tests. Pie charts below show the fractions of sites at near genes. P-values are from one-tailed hypergeometric tests. Pan-cancer transcriptome data from the Pan-cancer Analysis of Whole Genomes (PCAWG) study is shown. Boxplots span interquartile range (IQR; 25–75%), lines show median values, whiskers show values within 1.5x IQR, and site counts are shown below. e Long-range chromatin interactions involving TOP2B binding sites from 27 human cell types. P-values are from two-tailed Wilcoxon tests. Source data are provided as Source Data files.

Fig. 2. Mutational processes of small mutations at TOP2B binding sites in cancer genomes.

a Analysis overview. Mutation frequency in aligned binding sites was compared to flanking sequences using RM2, which quantifies mutational enrichments in genomic elements using trinucleotide context and megabase-scale mutation burden. b Mutational enrichments in TOP2B binding sites in primary and metastatic cancers (RM2, FDR < 0.05). Enrichment scores represent directional significance such that binding sites with enriched mutations are on the right and depleted sites on the left (OR, odds ratio). c Grassy hills plots show local mutation burden in pooled binding sites (colors; 600 bps) and flanking control sequences (grey; ±600 bps). FDR-adjusted two-tailed P-values (q-values) from RM2 are shown. d Mutations in TOP2B binding sites associate with conserved TOP2B binding in mice (left) and in vitro double-strand break (DSB) activity (right). e Mutational signatures of single base substitutions (SBS) in TOP2B binding sites. Indels were also included as a separate signature. In panels d and e, TOP2B-CTCF-RAD21 and TOP2B-RAD21 sites were compared to controls (CTCF-RAD21 and RAD21-only sites, respectively) using two-tailed hypergeometric tests (FDR < 0.05). Enriched signatures are displayed on the right and depleted signatures on the left. f Mutational processes of small mutations in TOP2B binding sites grouped by transcription or chromatin loops. Four bins of sites were analyzed (none, low, middle, high). Heatmap shows site types and cancer types having at least one bin with significant mutational enrichments (FDR < 0.05 from RM2). Positive associations are shown above (i.e., higher mutagenesis associated with more transcription or chromatin interactions) and negative associations are shown below. Color strips indicate site types, cancer types, and activity. g, h Examples of mutational processes at TOP2B-RAD21 binding sites. Grassy hills plots show mutation frequencies in sites binned by transcription or chromatin interactions. TOP2B-RAD21 sites (dark red) were compared to flanking sequences (grey) using RM2. A control analysis using RAD21-only sites is also shown (light blue). FDR-adjusted two-tailed P-values (q-values) from RM2 are shown. Loess smoothing in panels (c, g, h) is shown with 95% standard error bands. Source data are provided as Source Data files.

Fig. 3. Structural variant (SV) breakpoints (SVBPs) are enriched at TOP2B binding sites in cancer genomes.

a Comparison of SVBP frequency in TOP2B binding sites in pooled pan-cancer data from primary cancers and metastases. TOP2B-bound sites and non-bound sites were compared (TOP2B-CTCF-RAD21 vs CTCF-RAD21, TOP2B-RAD21 vs RAD21-only). As genome-wide controls, tiled genomic windows of matching size were used. b Frequency of SVBPs in TOP2B binding sites relative to functional activity at the sites. TOP2B binding sites were grouped into four bins by pan-cancer gene expression or chromatin interaction frequency. c Frequency of SVBPs in TOP2B binding sites with functional annotations: TOP2B conservation in mice, constitutive CTCF binding in human tissues, or DSB activity. P-values from one-tailed Wilcoxon tests are shown in panels (a–c). d Analysis of SVBPs in TOP2B-CTCF-RAD21 and TOP2B-RAD21 sites relative to small mutations. Sites were compared to control sites lacking TOP2B binding using two-tailed hypergeometric tests and filtered results are shown (FDR < 0.05). Enrichment scores represent directional significance such that sites with enriched mutations are on the right and depleted sites on the left (OR, odds ratio). e Relative distances of binding site midpoints and the closest SV breakpoints. Pooled SV breakpoints from HMF and PCAWG datasets within 1 kbps from site midpoints were included. The dotted line corresponds to the 600 bps window defining binding sites in this study (i.e., ±300 bps). Boxplots span interquartile range (IQR; 25–75%), lines show median values, whiskers show values within 1.5x IQR, and site counts are shown below. P-values are from one-tailed Wilcoxon tests. Source data are provided as Source Data files.

Fig. 4. Frequently mutated regulatory elements (FMREs) at TOP2B binding sites.

a FMREs identified at the binding sites through significant enrichment of small mutations (SNVs, indels) or structural variant breakpoints (SVBPs) in individual cancer types in primary cancers (PCAWG) and metastases (HMF) (one-tailed tests from ActiveDriverWGS, FDR < 0.05). Known cancer genes are labelled. Color strips below the bar chart show site types, primary or metastatic cancers, and the type of mutation involved (small mutation or structural variant). b Circos plot of translocations at FMREs combined from primary cancers and metastases. Known cancer genes at FMREs are labelled. c Functional characteristics of FMREs relative to other binding sites. FMREs in the highest category (bin4) correspond to sites with highest third of gene expression or chromatin interactions from 27 human tissue types, or sites with constitutive CTCF binding in >90% of human tissues. P-values from one-tailed Wilcoxon tests (top) or one-tailed hypergeometric tests (bottom) are shown. d FMREs ranked by the number of promoter-enhancer chromatin interactions. Top genes with FMREs are labelled. e Genes with differential expression associations with alterations in adjacent FMREs (one-tailed F-test, FDR < 0.05). Bars show fold-change values of genes with respect to FMRE mutations. Symbols indicate associations identified from SNVs and indels (circles) or SVBPs (triangles). f Enrichment map of biological processes and molecular pathways enriched in non-coding FMREs (ActivePathways, FDR < 0.05). The network shows enriched pathways as nodes in which similar pathways that share many genes are connected by edges. Pathways were prioritized by mutational enrichments across primary and metastatic cancers. Colors indicate cancer types in which pathways were identified, and white corresponds to pathways only detected across multiple cancer types. Source data are provided as Source Data files.

Fig. 5. TOP2B binding sites at established cancer driver mutations.

a Known cancer driver genes at frequently mutated regulatory elements (FMREs) identified from enriched small mutations or SV breakpoints (SVBPs) (one-tailed tests from ActiveDriverWGS, FDR < 0.05). Stacked barplots show cancer samples affected at each FMRE, types of mutations, and binding sites. Colors on the left indicate cancer types. b Genes with FMREs at mutational hotspots and loss-of-function mutations in known driver genes. Mutations ( ≥ 5) in affected genes are grouped by amino acid substitutions or impact (frameshift (FS), non-frameshift (NFS), or nonsense (STOP)). Colors indicate cancer types. Mutation counts are shown across primary and metastatic cancers. c–f The prominent FMRE at TP53 promoter is a TOP2B-CTCF-RAD21 binding site that is enriched in SVBPs. c Genomic locus of the FMRE at TP53 with SVBPs in the FMRE (black) and adjacent flanking sequence (grey) (top), corresponding ChIP-seq signals from representative experiments (black) and MACS2 peaks (pink) (middle). Counts of alterations and cancer samples colored by alteration types and cancer types are shown below. FDR-adjusted P-values from one-tailed tests in ActiveDriverWGS are shown. d Circos plot of translocations at the TP53 promoter FMRE. The FMRE locus is shown as a triangle and translocations involving the FMRE are displayed as arcs (brown). Putative target genes of translocations are listed. e TP53 expression in metastatic prostate cancers in samples having SVBPs at the FMRE (n = 10) and other samples (n = 156). FDR-adjusted P-value from one-tailed a F-test is shown. Boxplots span interquartile range (IQR; 25–75%), lines show median values, and whiskers show values within 1.5x IQR. f Long-range chromatin interactions at the TP53 FMRE from 27 human tissue types. Source data are provided as Source Data files.

Fig. 6. Experimental validation of the FMRE at the RMRP/CCDC107 locus.

a Genomic locus of the frequently mutated regulatory element (FMRE) at the non-coding RNA gene RMRP. The plot shows small mutations (SNVs, indels) in primary breast cancers (PCAWG) and metastases (HMF) (top), TOP2B-CTCF-RAD21 binding site (middle), corresponding ChIP-seq signals from representative experiments (black) and MACS2 peaks (pink) (bottom). b FMRE Mutations grouped by cancer types. c Locus overview with pan-cancer mutation frequency (top), FMRE and binding sites (middle), and adjacent genes (bottom). Triangles indicate sgRNAs used in genome editing. d Experimental validation. Two sgRNAs targeting the orthologous mouse FMRE were delivered by lentivirus using ultrasound-guided in utero injection into mouse embryos. In parallel, two sgRNAs targeting the human FMRE were delivered into MCF10A human mammary epithelial cell lines (not shown). e Genome editing of mouse FMRE causes earlier tumor onset in vivo. Kaplan-Meier plot compares tumor-free survival of mice having FMRE-targeted sgRNAs (red or purple lines) and mice having control sgRNAs (grey lines). Hazard ratio (HR) and P-value were derived from Cox proportional-hazards regression with Wald test. HR is shown with 95% confidence intervals. f Three-dimensional (3D) Matrigel growth assay of MCF10A cells compares control (Scr)-treated cells (left) and FMRE-mutated cells (right). One representative experiment for sgRNA sg6-1835p is imaged on day 28. g Differential expression analysis of FMRE-mutant MCF10A cells. Scatterplot shows differential gene expression from RNA-seq data of 2D (X-axis) and 3D assays (Y-axis) of edited MCF10A cells relative to Scr-treated controls. Genes were prioritized based on either joint up-regulation or down-regulation in 2D and 3D assays and significant genes are colored (merged FDR < 0.05 from DPM). Directional significance scores display up-regulated genes on the top or right and down-regulated genes on the bottom or left. Cancer genes from Cancer Gene Census are labelled. h Significantly enriched pathways visualized as an enrichment map. Significant pathways and processes are shown as nodes that that are connected into subnetworks if the pathways share many genes (family-wise error rate (FWER) < 0.05 from ActivePathways). Node color corresponds to transcriptomics evidence from the two types of growth assays. Source data are provided as Source Data files.