CDK12 alterations and ARID1A mutations are predictors of poor prognosis and therapeutic targets in high-grade salivary gland carcinoma: analysis of the National Genomic Profiling Database.

Abstract

Due to the diversity of histopathologic types in salivary gland carcinoma, genomic analysis of large cohorts with next-generation sequencing by histologic type has not been adequately performed. We analysed data from 93 patients with salivary duct carcinoma and 243 patients with adenoid cystic carcinoma who underwent comprehensive genomic profiling testing in the Center for Cancer Genomics and Advanced Therapeutics database, a Japanese national genome profiling database. We visualised gene mutation profiles using the OncoPrinter platform. Fisher's exact test, Kaplan-Meier analysis, log-rank test and Cox regression models were used for statistical analysis. In salivary duct carcinoma, a population with CDK12 and ERBB2 co-amplification was detected in 20 of 37 (54.1%) patients with ERBB2 amplification. We identified five loss-of-function variants in genes related to homologous recombination deficiency, such as BRCA2 and CDK12. Cox survival analysis showed that CDK12 and ERBB2 co-amplification is associated with overall survival (hazard ratio, 3.597; P=0.045). In salivary duct carcinoma, NOTCH1 mutations were the most common, followed by mutations in chromatin modification genes such as KMT2D, BCOR, KDM6A, ARID1A, EP300 and CREBBP. In the multivariate Cox analysis, activating NOTCH1 mutations (hazard ratio, 3.569; P=0.009) and ARID1A mutations (hazard ratio, 4.029; P=0.034) were significantly associated with overall survival. CDK12 and ERBB2 co-amplification is associated with a poor prognosis in salivary duct carcinoma. Chromatin remodelling genes are deeply involved in tumour progression in adenoid cystic carcinoma. One such gene, ARID1A, was an independent prognostic factor. In salivary duct carcinoma and adenoid cystic carcinoma, there might be minor populations with mutations that could be targeted for treatment with the synthetic lethality approach.