Scales and mechanisms of somatic mutation rate variation across the human genome

Abstract

Cancer genome sequencing has revealed that somatic mutation rates vary substantially across the human genome and at scales from megabase-sized domains to individual nucleotides. Here we review recent work that has both revealed the major mutation biases that operate across the genome and the molecular mechanisms that cause them. The default mutation rate landscape in mammalian genomes results in active genes having low mutation rates because of a combination of factors that increase DNA repair: early DNA replication, transcription, active chromatin modifications and accessible chromatin. Therefore, either an increase in the global mutation rate or a redistribution of mutations from inactive to active DNA can increase the rate at which consequential mutations are acquired in active genes. Several environmental carcinogens and intrinsic mechanisms operating in tumor cells likely cause cancer by this second mechanism: by specifically increasing the mutation rate in active regions of the genome.

Introduction

The large-scale sequencing of tumors and healthy somatic cells presents a unique opportunity to learn about somatic mutation processes and how mutation rates vary across the human genome. The primary motivation for tumor genome sequencing was to identify the ‘driver’ mutations that cause cancer. Driver mutations are selected because they promote the expansion or survival of tumor clones. However, most somatic mutations in cancer genomes are inconsequential ‘passenger’ mutations that are under very weak or no selection and statistical analyses of these passenger mutations have provided many fundamental insights into the mutation processes that operate in human cells and how these processes vary across the genome, cell types and individuals.

Absolute mutation rates are difficult to determine for tumor cells, primarily because the number of cell divisions that a tumor cell has undergone is hard to establish. However, it has long been appreciated that many tumors have have an elevated mutation rate, for example because of inactivated DNA repair pathways [[1], [2], [3]]. In this review, we will not focus on the general acceleration in mutation rates in a cancer cell. Instead, we focus on relative mutation rates, which are more straightforward to quantify from regional densities of mutations in the genome. We provide an overview of the patterns of mutations that are observed across regions of the human genome and our current understanding about their mechanistic underpinnings when this is known (although often a detailed mechanistic understanding is still lacking). We place an emphasis on the insight and the novel hypotheses that cancer genomes have yielded about the organization of mutation processes. Our primary focus is on single nucleotide substitutions and short insertions and deletions. The reasons for this are pragmatic: structural variation is much more challenging to precisely infer using short-read sequencing and although progress is being made in both identifying and understanding structural variation, the influences on its regional rates in the soma are far less well understood [4].

Variability in mutation rates across the genome may result from two broad causes: differential accrual of DNA damage and also base mispairing during DNA replication (variation in mutation supply) or differential repair of damage and mispairs (variation in DNA repair). These influences are, of course, not mutually exclusive. Recent work has suggested, however, that the latter – differential DNA repair – appears to play a quantitatively more important role in shaping the mutation landscape in the human soma. This is consistent with the expectation that mutation rates are more sensitive to changes in repair rates than to changes in damage rates because the vast majority of instances of damage are repaired [5].