Comparative genomic profiling identifies targetable brain metastasis drivers in non-small cell lung cancer

Brain metastases (BM) severely impact the prognosis and quality of life of patients with non-small cell lung cancer (NSCLC). To identify targetable drivers of NSCLC-BM, we profiled somatic copy number alterations (SCNAs) in 51 matched pairs of primary NSCLC and BM samples from 33 patients with lung adenocarcinoma (LUAD) and 18 patients with lung squamous cell carcinoma (LUSC). BM consistently had a higher burden of SCNAs compared to the matched primary tumors, and SCNAs were typically homogeneously distributed within BM, as revealed by multi-region SCNA profiling in 15 BM samples, suggesting that BM do not undergo extensive evolution once formed. By comparing focal SCNAs in matched NSCLC-BM pairs, we identified BM driving alterations affecting multiple cancer genes, including several targetable genes such as CDK12, DDR2, ERBB2, and NTRK1, which we validated in an independent cohort of 84 BM samples. We explored the metastatic potential of CDK12 and DDR2 in vitro and in vivo and found that overexpression of either gene alone in murine lung cancer cells causes the induction of key genes involved in epithelial-mesenchymal transition. Finally, we performed whole-exome sequencing of 40 NSCLC-BM pairs and identified BM driver mutations in multiple cancer genes, including several genes involved in epigenome editing and 3D genome organization, such as EP300, CTCF, and STAG2, which we validated by targeted sequencing of an independent cohort of 115 BM samples. Our study represents the most comprehensive genomic characterization of LUAD and LUSC BM available to date, uncovering potentially targetable NSCLC-BM drivers and inspiring the design of novel precision treatment strategies for NSCLC patients.