An externally validated lactylation-associated prognostic signature for overall survival prediction in lung adenocarcinoma identifies TUBA1C as a candidate gene

Background: Lung adenocarcinoma (LUAD) is characterized by marked prognostic heterogeneity. Although lactylation has been implicated in tumor progression and immune regulation, the clinical relevance of lactylation-related transcriptional programs in LUAD remains insufficiently defined. This study aimed to develop and externally validate a lactylation-related gene signature for overall survival prediction in LUAD and to prioritize candidate genes for further biological investigation.

Methods: We conducted a retrospective prediction model development and external validation study by integrating single-cell and bulk transcriptomic data. The Cancer Genome Atlas (TCGA)-LUAD was used as the model development cohort, whereas GSE31210 and GSE72094 were used as independent external validation cohorts, including 503, 226, and 398 patients, respectively. Overall survival was defined as the primary outcome. An optimized lactylation-related gene signature (LRGS) was constructed using machine learning strategies, and its predictive performance was evaluated using the concordance index, Kaplan-Meier survival analysis, and time-dependent receiver operating characteristic (ROC) analysis. In addition, pathway enrichment, tumor microenvironment, genomic alteration, and intercellular communication analyses were performed. Immunohistochemistry and reverse transcription quantitative polymerase chain reaction (RT-qPCR) were further used to validate TUBA1C expression in LUAD tissues and cell lines.

Results: The optimal model [StepCox (forward) + random survival forest (RSF)] achieved C-index values of 0.935, 0.668, and 0.637 in the TCGA-LUAD, GSE31210, and GSE72094 cohorts, respectively. The final LRGS consisted of 15 genes and stratified patients into high- and low-risk groups with significantly different overall survival across all cohorts (all P<0.001). Time-dependent ROC analysis demonstrated favorable predictive performance, with areas under the curve (AUCs) of 0.96, 0.98, and 0.99 at 1, 2, and 3 years in TCGA-LUAD. Multivariate Cox analysis confirmed that LRGS was an independent prognostic factor. High-risk tumors were associated with enhanced glycolysis, hypoxia, and PI3K-AKT-mTOR signaling, as well as a less immune-active tumor microenvironment. Single-cell ligand-receptor analysis further inferred relatively increased transforming growth factor-beta (TGF-β), vascular endothelial growth factor (VEGF), and C-X-C motif chemokine ligand (CXCL) communication patterns in the high-risk group. TUBA1C was prioritized as a candidate gene, and showed higher expression in LUAD tissues and NSCLC cell lines in preliminary validation assays.

Conclusions: LRGS may support prognostic stratification of LUAD based on lactylation-related transcriptional features. In addition, TUBA1C may be a potential biomarker worthy of further functional investigation.