Common deleterious germline variants shape the urothelial cancer genome

Abstract

The prevalence and biological consequences of deleterious germline variants (DGVs) in urothelial cancer (UC) are unknown. We performed whole-exome sequencing (WES) of 158 tumors and corresponding germline DNA from 80 UC patients at Weill-Cornell Medicine (WCM). We developed a novel computational framework (DGVar) to detect DGVs from germline WES data and predict their biological functions. We used strict criteria to identify truncating variants in 1604 tumor suppressor genes (TSGs) from germline WES data. We assessed germline-somatic interactions over the lifetime of each tumor. We confirmed our findings by applying DGVar to germline WES from 398 patients from the Cancer Genome Atlas (TCGA). We performed extensive validation of our results in other UC cohorts and whole-exome sequencing data from more than 13,000 non-cancer subjects. DGVs were identified in 45/80 patients (56%) of the WCM UC cohort and 315 DGVs in 48% (190/398) of patients in the TCGA UC cohort. DGVs were significantly more common in UC patients of WCM and TCGA compared to 2,504 subjects from the 1000 genome project (1KGP) population (p<0.0001). Similarly, our DGVs were more common in WCM and TCGA UC compared to 11,209 adult parents of European ancestry from the SPARK study of autism. KLK6, POLQ, and ITGA7 variants were the most common DGVs in WCM (15%) and TCGA UC (5%) cohorts. DGVs in the DNA repair pathway genes, including BRCA2, FANCA, POLQ, and XPA were identified in 7.5% of patients. Select DGVs were validated by Sanger sequencing of germline DNA including a novel XPA Leu200* variant, which eliminated the DNA-binding domain. DGVar identified 60 DGVs which are not detectable using currently used germline clinical targeted sequencing panels. Functional modeling showed that 96% of DGVs clustered with somatic variants within 15 Angstroms in genes with known crystal structures. Consistent with the pathogenic role of identified DGVs, we discovered frequent bi-allelic inactivation events arising from somatic losses of wild-type alleles. Furthermore, we observed deepening loss of heterozygosity of DGV genes in serial tumor samples as UC progressed from primary to metastatic sites. Finally, we identified extensive intra-patient heterogeneity in germline-somatic variant interactions and characterized their biological consequences. We show that close half of UC patients harbor DGVs, which potentially play a critical role in UC initiation and progression. Our results redefine the landscape of germline variants in urothelial cancer and provide a pivotal new understanding of their unique role in the biology of the disease.