Causes and consequences of large-scale copy number alterations in cancer evolution

Abstract

Cancer genomes are characterised by genomic instability and frequent large-scale copy number alterations across the genome. The aim of this thesis was to investigate the consequences of these alterations from an evolutionary perspective, with a focus on the development of extreme ploidy and the impact of dosage sensitivity in cancer. In chapter 2, I investigated genetic and environmental factors involved in ploidy alterations. While several factors are known to cause polyploidy in vitro, their relevance within patient cohorts is not clear. By analysing polyploid and hypodiploid patients across The Cancer Genome Atlas, I identified replication rate, TP53 mutations, hypoxia and HBV infection as significant predictors of polyploidy in patient tumours. Results from a random forest classifier based on these predictors support the idea that the rate of polyploidy in primary tumours is largely determined by the rate of occurrence of stochastic mitotic errors in cells with a permissive G1 checkpoint. TP53 mutations, replication rate and hypoxia were also significant predictors of hypodiploidy, indicating shared underlying factors between these two opposite ploidy states. In chapter 3, I analysed patterns of chromosome loss in the establishment of hypodiploidy and in the evolution of post-WGD tumours. Multiple lines of evidence suggest two groupings of cancer types with frequent hypodiploidy: one where hypodiploids form a distinct subgroup with stereotyped chromosome loss patterns, and one in which chromosome losses are largely random and hypodiploid cases merely represent one end of a spectrum of chromosomal instability. Loss patterns are generally not explained by chromosome features, and in particular there is no evidence that hypodiploid tumours evolve to avoid monosomy of germline dosage-sensitive genes. I then explored the commonalities between extreme ploidy states further, finding evidence for a general chromosomal instability phenotype that contributes to both increases and decreases in ploidy and to both large-scale and small-scale copy number alterations. Additionally, I formalised and tested a clinical heuristic to identify a diagnostically-challenging subset of acute lymphoblastic leukaemias based on karyotype alone, which allowed me to calculate its prevalence and to rigorously estimate the rate of genome doubling in hypodiploid leukaemia. In chapter 4, I performed a functional screen to identify haploinsufficient tumour suppressor genes based on observable changes in cell behaviour upon monoallelic TSG inactivation. This screen allowed me to identify novel haploinsufficient TSGs and to quantify TSG haploinsufficiency across tissues and functions. Our results support a model of continuous dosage-responsiveness, wherein TSGs are not cleanly divided into `two-hit' vs `haploinsufficient' classes but instead influence tumorigenesis depending on both their degree of inactivation and their importance in regulating cell behaviour. The results of this thesis emphasise the weakness of negative selection and the unifying role of chromosomal instability in cancer copy number evolution.