Assessment of RAS Dependency for BRAF Alterations Using Cancer Genomic Databases

Importance: Understanding RAS dependency and mechanisms of RAS activation in non-V600 BRAF variant cancers has important clinical implications. This is the first study to date to systematically assess RAS dependency of BRAF alterations with real-world cancer genomic databases. Objective: To evaluate RAS dependency of individual BRAF alterations through alteration coexistence analysis using cancer genomic databases. Design and Setting: A cross-sectional data analysis of 119538 nonredundant cancer samples using cancer genomics databases including GENIE (Genomics Evidence Neoplasia Information Exchange) and databases in cBioPortal including TCGA (The Cancer Genome Atlas) (accessed March 24, 2020), in addition to 2745 cancer samples from Mayo Clinic Genomics Laboratory (January 1, 2015, to July 1, 2020). Frequencies and odds ratios of coexisting alterations of RAS (KRAS, NRAS and HRAS) and RAS regulatory genes (NF1, PTPN11 and CBL) were calculated for individual BRAF alterations, and compared according to the current BRAF alteration classification; cancer type specificity of coexisting alterations of RAS or RAS regulatory genes was also evaluated. Main Outcomes and Measures: Primary outcome measurement is enrichment of RAS (KRAS, NRAS and HRAS) alterations in BRAF variant cancers. Secondary outcome measurement is enrichment of RAS regulatory gene (NF1, PTPN11, and CBL) in BRAF variant cancers. Results: A total of 2745 cancer samples from 2708 patients (female/male ratio: 1.0) tested by Mayo Clinic Genomics Laboratory and 119538 patients (female/male ratio: 1.1) from GENIE and cBioPortal database were included in the study. In 119538 nonredundant cancer samples, class 1 BRAF alterations and BRAF fusions were found to be mutually exclusive to alterations of RAS or RAS regulatory genes (odds ratio range 0.03-0.13 and 0.03-0.73 respectively), confirming their RAS independency. Both class 2 and class 3 BRAF alterations show variable and overlapping levels of enriched RAS alterations (odds ratio range: 0.03-5.9 and 0.63-2.52 respectively), suggesting heterogeneity in RAS dependency and a need to revisit BRAF alteration classification. For RAS-dependent BRAF alterations, the coexisting alterations also involve RAS regulatory genes by enrichment analysis (for example, S467L shows an odds ratio of 8.26 for NF1, 9.87 for PTPN11, and 15.23 for CBL) and occur in a variety of cancer types with some coalterations showing cancer type specificity (for example, HRAS variations account for 46.7% of all coexisting RAS alterations in BRAF variant bladder cancers, but 0% in non-small cell lung cancers). Variant-level assessment shows that BRAF alterations involving the same codon may differ in RAS dependency. In addition, RAS dependency of previously unclassified BRAF alterations could be assessed. Conclusions and Relevance: Current BRAF alteration classification based on in vitro assays does not accurately predict RAS dependency in vivo for non-V600 BRAF alterations. RAS-dependent BRAF variant cancers with different mechanisms of RAS activation suggest the need for different treatment strategies.