Metabolic signature of phosphoethanolamine and arginine pathways predicts prognosis and immunotherapy resistance in lung adenocarcinoma

Abstract

Background Metabolic reprogramming is a hallmark of lung adenocarcinoma (LUAD). However, the prognostic significance of phosphoethanolamine (pEtn) and arginine metabolism-related genes remains poorly defined. This study aimed to develop and externally validate a pEtn- and arginine metabolism-related gene signature for predicting prognosis in LUAD and to characterize its associations with tumor biology, immune response, and therapeutic sensitivity. Methods We analyzed transcriptomic data from The Cancer Genome Atlas (TCGA) LUAD (TCGA-LUAD, n=514) and independent Gene Expression Omnibus (GEO) cohorts. Candidate genes were identified from Molecular Signatures Database (MSigDB) gene sets, followed by differential expression, Cox regression, and protein-protein interaction (PPI) analyses. An 18-gene metabolic signature was constructed using least absolute shrinkage and selection operator (LASSO) regression and validated across three external datasets. Functional enrichment, immune infiltration profiling, Tumor Immune Dysfunction and Exclusion (TIDE) analysis, and drug sensitivity prediction, Cancer Therapeutics Response Portal (CTRP), and Genomics of Drug Sensitivity in Cancer (GDSC) were performed to explore biological and therapeutic implications. Results We identified 114 metabolism-related genes, of which 27 were significantly associated with overall survival (OS). An 18-gene prognostic signature stratified patients into high- and low-risk groups, with high-risk patients showing significantly poorer survival [TCGA: hazard ratio (HR) =2.81, P<0.001; external validation: HR 2.60–3.24, all P<0.01]. High-risk tumors were enriched in cell cycle, glycolysis, and epithelial-mesenchymal transition (EMT) pathways, and exhibited elevated but dysfunctional CD8+ T cell infiltration, consistent with immune exclusion and reduced immunotherapy response. Drug sensitivity analysis suggested that high-risk patients may benefit from dasatinib, lovastatin, irinotecan, and talazoparib. Conclusions An 18-gene metabolic signature demonstrates robust prognostic value in LUAD, linking metabolic reprogramming with tumor aggressiveness, immunotherapy resistance, and drug vulnerability. This model provides a potential tool for risk stratification and personalized treatment.