Genomic landscape of 1007 pan-cancer brain metastases

Background: The pan-cancer genomic landscape of brain metastases (BM) has not been well-characterized. Herein, we evaluate genomically profiled BM tumor samples and additional sequenced tumor samples from other sites to further understand disease evolution. Methods: We analyzed BM specimens from 1007 patients who underwent craniotomy between 2014 and 2024. Targeted sequencing was performed with MSK-IMPACT, a next-generation sequencing assay which detects genomic alterations in up to 505 genes. The FACETS algorithm was used to estimate tumor purity, fraction of genome altered (FGA) and whole-genome duplication (WGD) status. We analyzed matched sample pairs from patients who had additional sequenced tumor samples resulting in 227 primary-BM (P-BM) pairs, 189 extracranial metastasis-BM (ECM-BM) pairs and 60 BM-BM pairs. One pair per patient was selected for each category based on maximum purity. For paired comparisons, the Wilcoxon signed-rank test was used to compare continuous features and McNemar's test was used to compare WGD, with Benjamini-Hochberg p-value adjustment. The Jaccard index was computed to assess mutational concordance between pairs. Private mutation analysis was limited to genes with driver mutations in >5 pairs in each P-BM/ECM-BM group. Results: Median intracranial progression-free survival (iPFS) and overall survival (OS) from craniotomy were 11.3 and 25 months, respectively. The most frequent histologies in the cohort were non-small cell lung cancer (NSCLC; n = 360), breast (n = 181), and melanoma (n = 128). The genes with the highest proportion of oncogenic alterations were TP53 (62.0%), CDKN2A (25.0%), TERT (23.4%), KRAS (19.7%), and ERBB2 (12.3%). At least one structural variant or mutation was shared by 88.1% of P-BM pairs, 90.0% of ECM-BM pairs and 98.3% of BM-BM pairs. The mean number of shared driver mutations was 2.48 for P-BM pairs, 2.43 for ECM-BM pairs and 4.02 for BM-BM pairs. In a pan-cancer paired analysis, FGA, WGD and tumor purity were higher in BM in P-BM pairs (q < 0.01 for all) and in ECM-BM pairs (q < 0.01 for all). FGA and purity were higher in BM for P-BM pairs in upper gastrointestinal (GI) and NSCLC (q < 0.01), and in ECM-BM pairs for NSCLC (q < 0.01). By histology, in P-BM pairs, lower GI (n = 22) had the highest mean Jaccard index (J = 0.71) and prostate cancer (n = 10) had the lowest (J = 0.33), while in ECM-BM pairs melanoma (n = 23) had the highest (J = 0.77) and sarcoma (n = 9) had the lowest (J = 0.36). TP53 was the most commonly mutated gene in both P-BM (n pairs = 147, shared proportion = 0.816) and ECM-BM pairs (n pairs = 111, shared proportion = 0.869). Mutations in NFE2L2 and KMT2B were most commonly private to the BM in P-BM pairs (4/6 and 5/8 pairs with driver mutations private to BM, respectively), while NF1 mutations were most often BM-private in ECM-BM pairs (4/7). Conclusion: There is a high degree of concordance in alterations between P-BM and ECM-BM pairs. Alterations more commonly private to BM warrant further investigation.