A mutational signature for brain metastasis tropism in lung adenocarcinoma

Lung adenocarcinoma brain metastasis (LUAD-BM) is a leading cause of cancer-related mortality, yet the genomic determinants driving brain tropism and reliable predictive biomarkers remain poorly defined.

We analyzed the mutational profiling of 2602 primary LUAD and LUAD-BM samples. Differentially mutated genes were used to construct a mutational signature-based classification for BM tropism, which was further characterized by corresponding multi-omic data. Independent validation was conducted in TCGA (N = 503) and Oncosg (N = 302) cohorts.

Mutational profiling of 172 LUAD-BM revealed frequent alterations in TP53, EGFR, and KRAS, with enrichment of apoptosis-, E2F-, and cell cycle-related pathways. Twenty-five genes were identified as significantly more frequently mutated in LUAD-BM, most of which also displayed higher mutation allele frequencies and preferential enrichment in LUAD patients with isolated brain metastasis. Using this 25-gene mutational signature, primary LUAD samples were classified into three distinct clusters with significantly different survival outcomes. Cluster 3 was strongly associated with brain metastasis, accounting for 62% of LUAD-BM cases and 59% of LUAD patients with isolated brain metastasis, and showed the highest incidence of BM development (25.3%, P < 0.001). Meanwhile, Cluster 3 was characterized by markedly elevated TMB, extensive copy number variations across multiple chromosomes, and an almost universal TP53 mutation rate (99.9%), representing a highly unstable genomic subtype resembling LUAD-BM. The prognostic value of this classification was independently validated in TCGA-LUAD and Oncosg cohorts, where the mutation-defined clusters retained distinct survival patterns and genomic features. Transcriptomic analyses further demonstrated that Cluster 3 was enriched for cell cycle, mitotic, and chemokine signaling pathways, whereas Cluster 1 showed activation of lung function-associated pathways, highlighting the biological and clinical relevance of the mutation-based stratification.

We identified a high-risk LUAD subtype defined by a distinct mutational and copy number landscape, providing a practical framework for brain metastasis risk stratification and potential guidance for targeted surveillance strategies.