Pan-cancer analysis of mutations in open chromatin regions and their possible association with cancer pathogenesis

Abstract

Background: Open chromatin is associated with gene transcription. Previous studies have shown that the density of mutations in open chromatin regions is lower than that in flanking regions because of the higher accessibility of DNA repair machinery. However, in several cancer types, open chromatin regions show an increased local density of mutations in activated regulatory regions. Although the mutation distribution within open chromatin regions in cancer cells has been investigated, only few studies have focused on their functional implications in cancer. To reveal the impact of highly mutated open chromatin regions on cancer, we investigated the association between mutations in open chromatin regions and their possible functions.

Methods: Whole-genome sequencing data of 18 cancer types were downloaded from the PanCancer Analysis of Whole Genomes and Catalog of Somatic Mutations in Cancer. We quantified the mutations located in open chromatin regions defined by The Cancer Genome Atlas and classified open chromatin regions into three categories based on the number of mutations. Then, we investigated the chromatin state, amplification, and possible target genes of the open chromatin regions with a high number of mutations. We also analyzed the association between the number of mutations in open chromatin regions and patient prognosis.

Results: In some cancer types, the proportion of promoter or enhancer chromatin state in open chromatin regions with a high number of mutations was significantly higher than that in the regions with a low number of mutations. The possible target genes of open chromatin regions with a high number of mutations were more strongly associated with cancer than those of other open chromatin regions. Moreover, a high number of mutations in open chromatin regions was significantly associated with a poor prognosis in some cancer types.

Conclusions: These results suggest that highly mutated open chromatin regions play an important role in cancer pathogenesis and can be effectively used to predict patient prognosis.

Keywords: TCGA; bioinformatics; cancer genetics; genomics; medical genetics.

FIGURE 1

Mutation distribution and characteristics of open chromatin regions. (A) Distribution of the total number of mutations (top), total length of open chromatin regions (upper), total number of mutations per open chromatin region (lower), and total exon lengths (bottom) per 100 kb on chromosome 11. “OC” on the horizontal axis represents mutations within open chromatin regions. (B) Open chromatin categories used in this study. Open chromatin regions were divided into three categories defined as N‐OC, L‐OC, and H‐OC based on the number of mutations. (C) Proportion of the frequency of 15 chromatin states in open chromatin regions for 15 cancer types. The horizontal axis represents the three categories of open chromatin regions in each cancer type and the vertical axis represents the proportion of 15 chromatin states. Ctrl represents the average proportion of the 15 chromatin states in the whole genome. Asterisks represent the cancer types with a significantly higher proportion of the chromatin state “1_TssA” or “7_Enh” in H‐OC regions. The upward arrows indicate that the proportion is significantly higher for open chromatins with more mutations, whereas the downward arrows indicate that the proportion is significantly lower for open chromatins with more mutations. (D) TF footprint length per open chromatin in the three categories for 12 cancer types. The horizontal axis represents the three categories of open chromatin for 12 cancer types and the vertical axis represents the TF footprint length per open chromatin (bp)

FIGURE 2

Proportion of genome amplification in open chromatin regions. (A) Outline of the analysis for genome amplification in open chromatin regions. The average copy number (ave CN) was determined for each open chromatin region and the proportion of amplified open chromatin regions was calculated. (B) Proportion of open chromatin regions with an average copy number of ≥4 in 18 cancer types. The horizontal axis represents regions outside open chromatin regions (C: Control) and the three categories of open chromatin regions (N, L, and H) for each cancer type. The black asterisks represent the cancer types with a significantly higher proportion of genome amplification in open chromatin regions than that outside the open chromatin regions. The red asterisk represents the cancer types with a significantly higher proportion of genome amplification in H‐OC regions than those in other open chromatin categories. (C) Genomic region around EGFR with H‐OC regions accumulation in GBM. The black vertical lines represent the positions of H‐OC region. (D) Genomic region around ERBB2 with the accumulation of H‐OC regions in BRCA. (E) Ratio between the observed sample numbers with the co‐occurrence of mutation and genome amplification in each H‐OC region and the expected sample numbers. Red line: ratio = 1. (F) Ratio between the observed sample numbers with the co‐occurrence of mutation and genome amplification in each L‐OC region and the expected sample numbers. Red line: ratio = 1

FIGURE 3

Characteristics of possible target genes of H‐OC regions. (A) Number of possible target genes for each open chromatin region category. The horizontal axis represents the three categories of open chromatin regions (N, L, and H) in 18 cancer types. The black asterisks represent the cancer types with a significantly higher number of possible target genes per open chromatin in H‐OC regions than those in other open chromatin categories. (B) Proportion of genes present in the Cancer Gene Census (CGC) and tumor suppressor genes (TSGs) registered in TSGene 2.0 for each of the three categories. The horizontal axis represents the three categories of open chromatin regions (N, L, and H), and the vertical axis represents the proportion of cancer‐related genes among the selected 1007 genes. (C) GO analysis using the possible target genes of H‐OC regions. The horizontal axis represents the −log10(p‐value), and the vertical axis represents the GO terms arranged in the order of descending p‐values. (D) The relationship between the possible target genes of H‐OC regions and human diseases was obtained using the DisGeNET database. The horizontal axis represents the −log10(p‐value), and the vertical axis represents the human diseases arranged in the order of descending p‐values

FIGURE 4

Survival analysis according to the number of mutations in H‐OC regions. (A) Kaplan–Meier curves of samples with <29 (red line) or ≥29 (green line) mutations in H‐OC regions. The horizontal axis represents the time (in days), and the vertical axis represents the survival probability. (B) Comparison of the C‐indexes obtained with the Cox regression model using the TMB (red box) and those obtained using the number of mutations in H‐OC regions (green box) in 14 cancer types. The black asterisk represents the cancer types with a significantly higher C‐index obtained using the TMB model. The red asterisk represents the cancer types with a significantly higher C‐index obtained using the H‐OC model. The C‐index was calculated by 100 times 5‐fold cross‐validation. The red line represents a C‐index of 0.5, and the blue line represents a C‐index of 0.7