Interplay between whole-genome doubling and the accumulation of deleterious alterations in cancer evolution

Abstract

Whole-genome doubling (WGD) is a prevalent event in cancer, involving a doubling of the entire chromosome complement. However, despite its prevalence and prognostic relevance, the evolutionary selection pressures for WGD in cancer have not been investigated. Here, we combine evolutionary simulations with an analysis of cancer sequencing data to explore WGD during cancer evolution. Simulations suggest that WGD can be selected to mitigate the irreversible, ratchet-like, accumulation of deleterious somatic alterations, provided that they occur at a sufficiently high rate. Consistent with this, we observe an enrichment for WGD in tumor types with extensive loss of heterozygosity, including lung squamous cell carcinoma and triple-negative breast cancers, and we find evidence for negative selection against homozygous loss of essential genes before, but not after, WGD. Finally, we demonstrate that loss of heterozygosity and temporal dissection of mutations can be exploited to identify novel tumor suppressor genes and to obtain a deeper characterization of known cancer genes.

Figure 1. Prevalence of whole genome duplication (WGD) and loss of heterozygosity (LOH) in NSCLC.

a-b) Proportion of WGD, subclonal WGD and non-WGD genomes (left), and proportion of the genome subject to LOH and haploid LOH in WGD vs nWGD (right) in LUSC (a) and LUAD (b). For all boxplots, the upper whisker indicates the largest value (no further than 1.5* inter-quartile range (IQR) of the box-edge), and the lower whisker corresponds to the smallest value at most 1.5* IQR of the box-edge; the median is indicated by the thick horizontal line; and the first and third quartiles are indicated by box edges; data beyond whiskers are ‘outliers’ and plotted individually. c) Differences in the proportion of clonal vs subclonal LOH in TRACERx data. Significant differences between groups was assessed with a t-test. d) Proportion of WGD tumors vs haploid LOH in nWGD tumors across 34 cancer types in the TCGA cohort. LUSC, lung squamous carcinoma; LUAD, lung adenocarcinoma

Figure 2. Whole genome doubling (WGD) buffers the deleterious effect of passenger alterations.

a) The principle of Muller’s ratchet in asexual and sexual organisms. The red dots represent the mutations acquired over time on chromosome segments. Asexual organisms with no recombination accumulate mutations in an irreversible manner leading towards cell death or extinction, while sexual populations with recombination are viable for longer periods of time. b) WGD buffers the effects of late (post-WGD) deleterious mutations in regions of loss of heterozygosity (LOH) by providing additional mutation-free segments. c) Proportion of WGD cells in the tumor at the end of simulations with varying values of WGD-associated cost, sWGD, and passenger fitness costs, s_p. The cost relates to a proportional reduction in birth rate, such that 0.5 cost represents a 50% increase in the waiting time to the next birth. d) Relationship between the proportion of cells subject to WGD and deleterious alteration rate, u_p, for two different values of s_p.

Figure 3. Timing mutations relative to whole genome doubling (WGD).

a) Those mutations occurring before genome duplication (red dots) will be present at multiple copies, whereas those occurring after the duplication event (blue dots) will only be present at only one copy. b) Validation of the mutation timing approach using an isogenic genome doubling system involving genome-doubled HTC-116 clones deriving from a non-WGD common ancestor. Barplots show the proportion of mutations correctly vs incorrectly classified as either pre-WGD or post-WGD by our timing approach for common mutations in the tetraploids (left) and tetraploid private mutations (i.e., not present in the diploid genomes) (right). c) Number of mutations in regions of LOH across LUSC and LUAD datasets, grouped by WGD status and timing. LUSC, lung squamous carcinoma; LUAD, lung adenocarcinoma

Figure 4. Purifying selection before but not after whole genome doubling (WGD).

a) dNdS values for truncating mutations in WGD tumors calculated for all genes, essential genes and lung-specific cancer genes, grouped by loss of heterozygosity (LOH) status and timing of the mutations in lung squamous carcinoma (LUSC) and lung adenocarcinoma (LUAD) from the TRACERx dataset (n=93), LUSC from TCGA (n=325) and, LUAD from TCGA (n=398). A dNdS ratio of 1 (red line) is consistent with neutrality. Values significantly higher than 1 (consistent with positive selection) are shown in dark blue. Values significantly lower than 1 (indicating purifying selection) are shown in red. Each point represents a dNdS estimate, with 95% CI shown. b) In WGD tumors difference in the number of chromosome arms in haploid LOH found through simulations vs the observed data in the TCGA NSCLC cohorts (n=970) and the ploidy of the tumors (below). Tumors with very high ploidy (>4.5) rarely exhibit any haploid LOH in simulations or in observed data. No multiple test correction was performed. c) Pearson Correlation between the essential genes score and the frequency of haploid LOH per chromosome arm (n=39) for TCGA NSCLC (n=970). d) Correlation between the essential genes score and the frequency post-WGD losses for TCGA NSCLC. Shaded region indicates 95% confidence interval, and Pearson’s correlation coefficient is indicated.

Figure 5. Exploiting loss of heterozygosity (LOH) to identify cancer genes.

a) dNdS selection coefficients for truncating mutations in early mutations in LOH (i.e. mutations present on all copies of remaining allele that is not subject to LOH) (x-axis) vs truncating mutations in genomic regions without evidence of LOH (y-axis). The background color indicates whether the gene was identified as significant using mutations in early LOH (dark blue) using all mutations (grey), or identified as significant in both cases (light blue) (q-value<0.05). The border color represents whether the gene is currently included in the COSMIC/NCG databases as a cancer gene. Data from TCGA and TRACERx LUSC (n=356) and TCGA and TRACERx LUAD (n=460) is shown b) dNdS selection coefficients for truncating mutations in early mutations in LOH (“LOH”) vs truncating mutations in regions without evidence of LOH (“noLOH”) across cancer types. Only genes that are significant in at least one cancer type in the LOH category are shown. Barplots show the total number of cancer genes that are significantly (q-value<0.05) identified using the “all approach” (grey) and the number of cancer genes that are only identified (q-value<0.05) using the “LOH” approach (dark blue). The latter is also represented with a number above the bars, and the specific genes are marked with dots in the heatmap. Only those cancers where we identify additional cancer genes using the “LOH” approach (this is, only looking at early mutations in LOH) are shown. (BRCA TNBC= triple negative breast cancer, HNSC=head and neck squamous cell carcinoma, BRCA ER+=ER positive breast cancer, COAD=colon adenocarcinoma, LUSC=lung squamous cell carcinoma, CHOL=cholangiocarcinoma, KIRP=kidney renal papillary cell carcinoma, LIHC=liver hepatocellular carcinoma, OV=ovarian serous cystadenocarcinoma, BLCA=bladder urothelial carcinoma, CESC=cervical squamous cell carcinoma and endocervical adenocarcinoma, KIRC=kidney renal clear cell carcinoma, READ=rectum adenocarcinoma, UCS=uterine carcinosarcoma.)