Divergent mutational processes distinguish hypoxic and normoxic tumours

Abstract

Many primary tumours have low levels of molecular oxygen (hypoxia), and hypoxic tumours respond poorly to therapy. Pan-cancer molecular hallmarks of tumour hypoxia remain poorly understood, with limited comprehension of its associations with specific mutational processes, non-coding driver genes and evolutionary features. Here, as part of the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium, which aggregated whole genome sequencing data from 2658 cancers across 38 tumour types, we quantify hypoxia in 1188 tumours spanning 27 cancer types. Elevated hypoxia associates with increased mutational load across cancer types, irrespective of underlying mutational class. The proportion of mutations attributed to several mutational signatures of unknown aetiology directly associates with the level of hypoxia, suggesting underlying mutational processes for these signatures. At the gene level, driver mutations in TP53, MYC and PTEN are enriched in hypoxic tumours, and mutations in PTEN interact with hypoxia to direct tumour evolutionary trajectories. Overall, hypoxia plays a critical role in shaping the genomic and evolutionary landscapes of cancer.

Fig. 1. The pan-cancer landscape of tumour hypoxia.

We quantified tumour hypoxia in 1188 independent tumours spanning 27 different cancer types. a Hypoxia scores for 27 types of cancer, based on the Buffa mRNA abundance signature. Cancer types are sorted by the median hypoxia score (horizontal black line) for each cancer type. Each dot represents one tumour. Sample sizes for each cancer type are listed near the bottom along with the percent of tumours that have elevated hypoxia (hypoxia score > 0). The variability in hypoxia within cancer types was measured by the interquartile range (IQR), shown along the bottom. The IQR was particularly high in biliary adenocarcinoma (IQR = 43.0; Biliary-AdenoCA), lymphoid B-cell non-Hodgkin’s lymphomas (IQR = 36.0; Lymph-BNHL), lung adenocarcinoma (IQR = 34.0; Lung-AdenoCA) and breast adenocarcinoma (IQR = 32; Breast-AdenoCA). By contrast, chronic lymphocytic leukaemia (IQR = 2.0; Lymph-CLL) and thyroid adenocarcinoma (IQR = 11.0; Thy-AdenoCA) showed less variance in hypoxia score. b Analysis of hypoxia between 16 comparable cancer types in PCAWG and TCGA (Spearman’s ρ, AS89). Dots represent the mean of the scaled median hypoxia scores from three different mRNA-based hypoxia signatures. Error bars represent the standard deviation of the scaled median hypoxia scores. Overall, the pan-cancer quantification of hypoxia between the PCAWG and TCGA datasets shows strong agreement.

Fig. 2. The genomic correlates of tumour hypoxia.

We associated tumour hypoxia with mutational density and summary features, a, and driver mutations, b, across 27 cancer types using linear mixed-effect models. Hypoxia scores for all 1188 tumours are shown along the top. a Elevated tumour hypoxia was strongly associated with more deletions, elevated PGA, smaller CNAs, and a higher number of SNVs per megabase (n = 1188 independent tumours). Bonferroni-adjusted p-values are shown on the right. b We tested if driver mutations (e.g. any of SNV, CNA, SV or a compound event with more than one type of mutation) were associated with hypoxia in 1096 independent tumours with driver mutation data. Tumours with mutations in BCL2 showed lower levels of hypoxia while patients with mutations in TP53 showed remarkably elevated tumour hypoxia. Other driver mutations associated with elevated hypoxia include the oncogene MYC and the tumour suppressor PTEN. FDR-adjusted p-values are shown along the right. SV structural variant; PGA percentage of the genome with a copy-number aberration; CNA copy-number aberration; SNV single nucleotide variant; H-H head-to-head; T-T tail-to-tail. All associations were modelled using linear mixed-effect models while adjusting for cancer type, tumour purity, age and sex.

Fig. 3. Hypoxia-associated mutational signatures.

We associated hypoxia with the proportion of mutations attributed to specific mutational signatures using linear-mixed effect models. Hypoxia scores for 1188 independent tumours are shown across the top while FDR-adjusted p-values are shown on the right. a Hypoxia was associated with a series of single base substitution signatures with unknown aetiology including SBS5, SBS17a, SBS17b and SBS12. Some of these mutational signatures may reflect hypoxia-dependent mutational processes. Hypoxia was also associated with a lower proportion of attributed mutations to SBS1, which reflects deamination of 5-methylcytosine, and a higher proportion of attributed mutations to SBS3, which is related to deficiencies in DNA double-strand break repair and homologous recombination. b Several signatures of small insertions and deletions were also associated with hypoxia, including ID6 and ID2, which reflect defective homologous recombination and defective DNA mismatch repair, respectively. ID5, ID9 and ID4, all with unknown aetiology, were significantly associated with hypoxia score. All associations were modelled using linear-mixed effect models while adjusting for cancer type, tumour purity, age and sex.

Fig. 4. The subclonal hallmarks of tumour hypoxia.

We associated tumour hypoxia with features related to the subclonal architecture of 1188 independent tumours from 27 cancer types using linear mixed-effect models. a Hypoxia scores are shown along the top while Bonferroni-adjusted p-values are shown on the right. Hypoxia was not associated with the number of subclones in the tumour but elevated hypoxia was associated with a higher number of clonal mutations. b We also observed a significant interaction between hypoxia and altered PTEN where tumours with both of these features were particularly likely to be polyclonal. c The mRNA abundance of PTEN is modulated by both PTEN mutational status and tumour hypoxia. Tumours with altered PTEN and elevated hypoxia have the lowest abundance of PTEN mRNA. mRNA abundance is reported as FPKM with upper-quartile normalization. A Tukey box plot is shown. Box plots represent the median (centre line) and upper and lower quartiles (box limits), and whiskers extend to the minimum and maximum values within 1.5× the interquartile range. All associations were modelled using linear mixed-effect models while adjusting for cancer type, tumour purity, age and sex. d Altered PTEN and hypoxia may drive subclonal diversification. Many primary tumours have elevated hypoxia due to increased demand or decreased supply of oxygen. Tumours with elevated hypoxia tend to have altered PTEN. Elevated hypoxia and altered PTEN may drive subclonal diversification and poor outcomes.