Patterns and functional implications of rare germline variants across 12 cancer types

open access -

Published 22 December 2015

 

...This study of over 4,000 cancer cases is the largest integrated analysis of germline and somatic variants to date. Our systematic analysis indicated that an estimated 18% of cancer cases from the TCGA cohort had ≥1 rare truncations in 624 genes associated with cancer. Further, there was significant enrichment of rare truncation variants in 13 genes and suggestive evidence of increases in 21 more, comprising 8.3% (333 out of 4,034) of TCGA cancer cases.

We observed several significant associations in specific cancer types: RAD51C in AML, ATM in PRAD and PALB2 in STAD. Across cancer types, a higher percentage of breast and ovarian cancer cases were identified as having rare truncation variants in cancer genes versus other cancer types, due predominantly to high frequencies in BRCA1/2. The percentage of breast and ovarian cancer cases carrying BRCA1/2 germline truncation variants in the TCGA cohort was 4.4% and 11.6%, respectively, consistent with previous reports^{39, 40, 41, 42}. Interestingly, stomach cancer has the second highest percentage of rare truncations
in 114 genes previously reported¹, largely due to the contributions from ATM, BRIP1, PALB2, XRCC2 and others. In contrast, for KIRC and GBM, truncation variants in the 34 associated germline genes were uncommon, identified in only less than 6% of cases (Fig. 2d). These results contribute to our understanding of the genetic architecture of cancers, complementing the known effect of common and tagged variants from array-based studies⁴³, as well as the estimate of overall heritability from twin studies in multiple cancer types⁴⁴.

Our results indicated that germline truncation and missense variants in several genes were under selection in the tumour, with ATM, BRCA1, BRCA2 and RAD51C determined as significant from both truncation and missense analyses and BAP1, BRIP1, FANCM, PALB2 and RAD51D from truncation analysis alone. As a proof-of-concept, we performed functional validation for 68 BRCA1 missense variant sites using HDR assay; our experimental efforts identified 9 variants from 14 patients with complete or partial defective HDR function and validated our LOH analysis for effective enrichment of variants under functional selection (an estimated eightfold enrichment in BRCA1).

More importantly, our integrated germline and somatic study identified BRCA1, BRCA2, RAD51C, RAD51D, FANCM, EME1 and MSH6 germline truncations significantly associated with increased somatic mutation frequencies in specific cancer types, suggesting that germline defects in DNA repair expand to the somatic level. Further, our search for co-occurring or mutually exclusive germline truncation/somatic mutations across 12 cancer types revealed a number of important insights in terms of genes and pathways involved including: (1) the association between germline BRCA1/2 germline truncations and frequent TP53 and infrequent PIK3CA somatic mutations confirm breast cancer clinical subtype classification; and (2) ATM as a bona fide (third frequently truncated) susceptibility gene demonstrated by both burden and LOH analyses is the only common gene highly mutated at both germline and somatic levels.

Although our study has been revealing at a genetic level, we are mindful of the limitations of the TCGA data set, including the lack of detailed family history information that would further inform the potential pathogenicity of germline variants. Despite the large sample size overall, our inferences are limited for specific cancer types because of small case numbers. In addition, the vast majority of TCGA cases in our sample set were of European background, emphasizing the need for the development of a reference source of genomic data on germline cancer predisposition variants from ancestrally diverse population groups. Nonetheless, this study is the largest to date that has integrated somatic and germline alterations to identify important genes across 12 major types contributing to cancer susceptibility and our results provide a promising list of candidate genes for definitive association and functional analyses. The combination of high throughput discovery and experimental validation should identify the most functionally and clinically relevant variants for cancer risk assessment......