Multi-cohort validation of a semaphorin prognostic signature to identify an immune- excluded subtype and therapeutic sensitivity in lung adenocarcinoma.

Abstract

e20016 Background: Semaphorins, classically implicated in axonal guidance, have increasingly recognized roles in tumor progression and immune regulation. However, their integrated prognostic significance and biological relevance in lung adenocarcinoma (LUAD) remain incompletely characterized. Methods: A retrospective multi-cohort study used TCGA-LUAD RNA-sequencing data (N = 503) for training and 12 independent GEO LUAD cohorts (total N = 1,820) for external validation. Differential expression analysis of 20 semaphorin genes was performed between short- and long-progression-free survival (PFS) subgroups (median cutoff: 438 days). Six prognostic candidates (SEMA3A, SEMA4A, SEMA6B, SEMA4B, SEMA3F, SEMA7A; P < 0.05) were incorporated into a multivariable Cox model with ridge regularization (λ = 0.1) to mitigate multicollinearity. A risk score was derived from penalized regression coefficients, with patients stratified by median score. External validation used multivariable Cox models adjusted for age, sex, and stage, with random-effects meta-analysis (DerSimonian–Laird). Functional enrichment (GSEA; GO and KEGG), tumor microenvironment profiling (ESTIMATE, xCell, ssGSEA), immune checkpoint analysis, and drug sensitivity prediction (oncoPredict/CTRP2) were performed. Results: The 6-gene signature demonstrated robust prognostic stratification. In TCGA, high-risk patients exhibited inferior overall survival (OS; HR = 1.43, 95% CI 1.17–1.74, P < 0.001) and PFS (HR = 1.27, 95% CI 1.10–1.48, P = 0.001). Meta-analysis across external cohorts confirmed consistent performance for OS (12 cohorts, N = 1,820; pooled HR = 1.41, 95% CI 1.26–1.58, P < 0.001) and recurrence (6 cohorts, N = 995; HR = 1.35, 95% CI 1.09–1.67, P = 0.006). High-risk tumors displayed a pro-proliferative, immune-excluded phenotype, with enrichment of mitotic spindle checkpoint regulation, collagen fibril organization, and p53 signaling. Immunologically, these tumors showed marked depletion of CD8⁺ T cells, B cells, and myeloid dendritic cells, with significantly lower Immune and Stromal Scores (all P < 0.0001). Despite this immune-excluded state, high-risk tumors selectively overexpressed CD276 (B7-H3; P < 0.0001) and CD274 (PD-L1; P < 0.01). Drug sensitivity modeling predicted increased susceptibility to taxane-based chemotherapy (docetaxel and paclitaxel; P < 0.0001) and reduced sensitivity to carboplatin (P < 0.01). Conclusions: We identified a semaphorin-based signature consistently associated with survival, immune contexture, and therapeutic response patterns in LUAD, supported by large-scale external validation. These findings highlight the potential role of semaphorin signaling as a clinically informative axis for risk stratification and treatment optimization.