A Nutrient-sensing–Inflammatory Metabolic Transcriptional Phenotype Defines Immune-dysregulated Lung Adenocarcinoma with Adverse Prognostic Direction: An Integrated TCGA and Multi-cohort Transcriptomic Study
Description
Background: Lung adenocarcinoma (LUAD) is marked by molecular and immune-microenvironmental heterogeneity. Existing transcriptional subtypes, including TRU, PI and PP, are informative but do not fully capture the continuous coupling of inflammation, nutrient sensing, metabolic stress, stromal remodeling and poor prognosis. We proposed a pathway-defined nutritional–inflammatory metabolic phenotype (NIMP), where “nutritional” denotes tumor-level nutrient-sensing and metabolic-stress transcriptional programs rather than dietary intake or clinical nutritional status. Methods: TCGA-LUAD was used as the discovery cohort. A NIMP-core score was calculated by GSVA from ten prespecified Hallmark pathways, with continuous within-cohort standardized NIMP_core_z as the primary exposure. Cox models reported HRs per 1-SD increase. External validation was performed independently in GSE31210, GSE68465 and GSE72094, followed by random-effects REML meta-analysis and Bayesian evidence calibration. TCGA analyses further assessed clinicopathologic features, mutation context, immune programs, Hallmark pathway positioning and TRU/PI/PP subtype assignment. Results: The TCGA discovery cohort included 504 primary tumors and 182 OS events. Higher NIMP_core_z was associated with adverse OS in univariable Cox analysis (HR=1.2736, 95% CI 1.1007–1.4738, P=0.001161) and remained adverse after adjustment for age, sex and AJCC stage in 486 complete cases (HR=1.1781, 95% CI 1.0172–1.3644, P=0.02867). Adding NIMP_core_z to the clinical-only model modestly improved fit (ΔAIC −2.81; ΔC-index +0.0076; likelihood-ratio P=0.0283). In external validation, all three OS cohorts showed HRs >1: GSE31210 HR=1.90, GSE68465 HR=1.12 and GSE72094 HR=1.18; GSE31210 RFS also showed an adverse association (HR=1.58, P=0.000659). Random-effects meta-analysis of external OS yielded a pooled HR of 1.31 (95% CI 0.981–1.75, P=0.0668; I²=82.7%). Bayesian calibration showed posterior HR median 1.28 (95% CrI 0.829–2.05) and Pr(HR>1)=91.7%, indicating moderate directional evidence rather than definitive pooled validation. Biologically, higher NIMP was associated with myeloid inflammation, SPP1-associated macrophage programs, TGF-β/EMT, CAF/stromal activation, immune-checkpoint activity, IFN-γ response and T-cell dysfunction signatures. NIMP-high was not explained by higher nonsynonymous mutation count or canonical KRAS/STK11/KEAP1 immune-cold contexts. Hallmark positioning linked NIMP-high tumors to inflammatory–cytokine, stromal-remodeling, angiogenic/coagulative, PI3K–AKT–mTOR and IFN-related programs. TRU/PI/PP analysis showed PI enrichment without subtype redundancy. Conclusions: NIMP represents a clinically adverse, subtype-linked but non-redundant immune–inflammatory–metabolic transcriptional axis in LUAD. Current evidence supports NIMP as a pathway-level risk and biology-stratification axis, not as a clinically translated independent prognostic biomarker or immunotherapy-predictive tool.