1068 - Mutational Gene Signature Defines a High-Risk Subtype of Lung Adenocarcinoma with Genomic Instability and High Propensity for Brain Metastasis

Purpose/Objective(s):

Brain metastasis (BM) is a major cause of mortality in lung adenocarcinoma (LUAD), yet reliable genomic predictors of BM risk and prognosis remain limited. We aimed to define a mutation-based classification of primary LUAD that captures brain metastatic tropism and clinical outcome.

Materials/Methods:

Gene mutation, copy number variation (CNV), and clinical data of LUAD and LUAD-BM samples were obtained from the MSK MetTropism cohort and used as the discovery dataset. Independent validation was performed using TCGA-LUAD and Oncosg cohorts. Non-synonymous somatic mutations were analyzed using maftools. Differentially mutated genes between LUAD and LUAD-BM were identified by Fisher’s exact test with false discovery rate correction. A 25-gene mutational signature enriched in LUAD-BM was used to cluster LUAD samples via UMAP and hierarchical clustering. Cluster labels were transferred to validation cohorts using a k-nearest neighbor approach. CNV landscapes were analyzed at both segment and gene levels. Transcriptomic features were explored using Genomap and differential expression analyses. Survival associations were evaluated using Kaplan–Meier and Cox proportional hazards models.

Results:

LUAD-BM samples exhibited significantly worse overall survival, higher tumor mutation burden, higher fraction of genome altered, and distinct trinucleotide mutation signatures compared with primary LUAD. Twenty-five genes showed significantly higher mutation frequencies in LUAD-BM and were enriched in apoptosis, cell cycle, and hypoxia-related pathways. Based on this signature, primary LUAD samples were classified into three mutation-defined clusters with distinct clinical outcomes. Cluster 3 showed the poorest prognosis, the highest risk of brain metastasis, near-universal TP53 mutation, and markedly elevated genomic instability resembling LUAD-BM. This classification was independently validated in TCGA and Oncosg cohorts, where Cluster 1 consistently showed favorable survival and low mutational burden. CNV analyses revealed extensive amplifications and deletions in Cluster 3 across multiple chromosomes. Transcriptomic profiling demonstrated that Cluster 1 was enriched for lung function–related genes, Cluster 2 for metabolic pathways, and Cluster 3 for cell cycle and chemokine signaling pathways. Mutation-defined clusters showed weak concordance with RNA-based molecular subtypes.

Conclusion:

We identify a robust mutation-based classification that stratifies LUAD patients by brain metastatic potential and prognosis. This genomic framework provides complementary information to transcriptomic subtyping and may facilitate early risk stratification and precision management of LUAD patients at high risk for brain metastasis.