Pan-cancer analysis of bi-allelic alterations in homologous recombination DNA repair genes

Abstract

BRCA1 and BRCA2 are involved in homologous recombination (HR) DNA repair and are germ-line cancer pre-disposition genes that result in a syndrome of hereditary breast and ovarian cancer (HBOC). Whether germ-line or somatic alterations in these genes or other members of the HR pathway and if mono- or bi-allelic alterations of HR-related genes have a phenotypic impact on other cancers remains to be fully elucidated. Here, we perform a pan-cancer analysis of The Cancer Genome Atlas (TCGA) data set and observe that bi-allelic pathogenic alterations in homologous recombination (HR) DNA repair-related genes are prevalent across many malignancies. These bi-allelic alterations often associate with genomic features of HR deficiency. Further, in ovarian, breast and prostate cancers, bi-allelic alterations are mutually exclusive of each other. The combination of these two properties facilitates reclassification of variants of unknown significance affecting DNA repair genes, and may help personalize HR directed therapies in the clinic.Germline mutations in homologous recombination (HR) DNA repair genes are linked to breast and ovarian cancer. Here, the authors show that mutually exclusive bi-allelic inactivation of HR genes are present in other cancer types and associated with genomic features of HR deficiency, expanding the potential use of HR-directed therapies.

Fig. 1

Bi-allelic pathogenic alterations affecting HR-related genes affect multiple cancer types and are associated with genomics features of HR deficiency. a Incidence of bi-allelic pathogenic alterations in HR-related genes in TCGA samples stratified according germ-line or somatic origin. It is important to note that germ-line bi-allelic alterations most commonly consisted of a germ-line pathogenic mutation and a somatic loss of heterozygosity. b Incidence of bi-allelic alterations of HR-related genes according to cancer type. c Association of LST and signature 3 in HBOC cancers. Bi-allelic pathogenic alterations and bi-allelic VUSs in HR-related genes are associated with elevated LST and signature 3 (pathogenic: p < 2.2*10⁻¹⁶ and p < 2.2*10⁻¹⁶, respectively; VUS: p = 0.02 and p = 0.017, respectively). Mono-allelic pathogenic alterations were not associated with either LST or signature 3 (p = 0.26 and p = 0.14, respectively). d LST for each genotype; cases with dominant signature 3 colored red (see methods). The box plot center line represents the median, the box limits represent the 1st and 3rd quartiles, respectively, and the whiskers extend from box limits to the largest value up to 1.5 times the interquartile range. e Pan-cancer analysis shows bi-allelic pathogenic alterations associated with an elevated LST and signature 3 (p < 2.2*10⁻¹⁶ and p < 2.2*10⁻¹⁶, respectively). Mono-allelic pathogenic alterations were not associated with elevated LST or signature 3 (p = 0.98 and p = 0.76, respectively). All p-values from the Wilcoxon-rank sum test. Error bars represent s.e.m.

Fig. 2

Overall survival (OS) in TCGA ovarian cancer patients who received platinum-based therapy. a Bi-allelic alterations in HR-related genes are associated with significantly longer overall survival (OS, p = 0.005; log-rank test). b Forest plot showing results from a multivariate Cox proportional-hazards model adjusting for stage (continuous: I, II, III, IV) and age (continuous) demonstrates that bi-allelic pathogenic alterations are associated with significantly longer overall survival (hazard ratio (HR) = 0.46, 95% confidence interval (CI) = 0.28–0.75, p = 0.002, Genotype is a categorical variable, bi-allelic vs. other). Stage (HR = 1.67, 95% CI = 1.25–2.21, p < 0.001) and age (HR = 1.01, 95% CI 1.00–1.03, p = 0.02) are also associated with outcome on multivariate analysis

Fig. 3

Mutual exclusivity analysis of bi-allelic alterations in HR-related genes and reclassification of VUSs based on genomics information. a Bi-allelic pathogenic alterations are mutually exclusive using two different statistical tests (p = 0.022, WExT, saddle point approximation of COMET; p = 0.025, permutation, methods). Inclusion of VUSs also demonstrates mutual exclusivity(p = 0.04, WExT). b VUSs in CDK12 and BRCA1. Size of circle indicates LST scores, while the blue filling corresponds to the proportion of signature 3-related mutations. Interestingly, 3 of 4 BRCA1 VUSs occurred in regions enriched for loss of heterozygosity (LOH), and are associated with an increased likelihood of a functional deficit in HR. Variants discussed in text indicated in red