The Challenge of LUAD Prognosis. Lung adenocarcinoma (LUAD) accounts for nearly 55% of lung cancer deaths and presents major challenges due to high heterogeneity in patient outcomes. Even patients at the same disease stage often experience dramatically different responses to therapy because of variations in the tumor microenvironment and immune status. Conventional staging systems cannot capture this heterogeneity, creating a need for robust molecular biomarkers.
What Is Lactylation. Lactylation is a post-translational protein modification driven by lactate -- a metabolic byproduct of glycolysis. In lung cancer, lactate stimulates cancer-associated fibroblasts and promotes immunosuppressive macrophage polarization (M2/tumor-associated macrophages). The oncogenic transcription factor ETV4 activates mTORC1 signaling through glycolysis and lactate synthesis. Lactylation also promotes drug resistance by driving cell cycle progression and activating multidrug resistance pathways.
What Is PANoptosis. PANoptosis is a newly characterized form of programmed cell death that simultaneously integrates apoptosis, necroptosis, and pyroptosis through a unified inflammatory response. In cancer, PANoptosis significantly suppresses tumor proliferation across multiple cancer types. It is orchestrated by PANoptosome multiprotein complexes that coordinate these distinct cell death programs. TNF-alpha and IFN-gamma can jointly trigger PANoptosis in lung cancer cell lines by activating caspases, GSDMD, GSDME, and MLKL.
Why Study Both Together. Lactylation-mediated reprogramming in macrophages regulates tumor immunity, while PANoptosis drives macrophage-induced tumor cell death. Evidence from multiple contexts (including sepsis-induced lung injury and epilepsy) shows these two pathways interact through inflammatory-metabolic crosstalk. Despite their individual prognostic significance, the combined lactylation-PANoptosis axis had not been characterized in lung adenocarcinoma.
Multi-Dataset Approach. The researchers analyzed both bulk RNA sequencing and single-cell RNA sequencing (scRNA-seq) datasets from multiple LUAD cohorts. The training set combined TCGA-LUAD (507 tumor samples, 51 normal samples) with GSE50081 (127 samples). Four independent external validation sets were used: GSE31210 (226 samples), GSE30219 (85 samples), GSE29016 (67 samples), and GSE42127 (131 samples). This approach enhanced sample diversity and reduced overfitting risk.
Identifying Lactylation and PANoptosis Genes. Starting from 5,566 bulk differentially expressed genes in TCGA-LUAD, the researchers identified 90 lactylation-related DEGs (64 upregulated, 26 downregulated) and 24 PANoptosis-related DEGs (10 upregulated, 14 downregulated). Survival analysis showed that upregulated lactylation-related genes correlated with worse prognosis, while downregulated lactylation and PANoptosis genes showed protective effects. The downregulated genes in both categories showed a strong positive correlation (R = 0.84), suggesting shared regulatory mechanisms.
WGCNA and Single-Cell Analysis. Weighted Gene Co-expression Network Analysis (WGCNA) identified nine co-expression modules in the TCGA-LUAD dataset. The green module, containing 1,027 genes, showed the strongest correlation with both downregulated lactylation and PANoptosis genes (R = 0.88 for both). Parallel scRNA-seq analysis of 22 LUAD samples identified 30 cell clusters including mast cells, T cells, B cells, epithelial cells, endothelial cells, fibroblasts, and NK cells. The intersection of single-cell DEGs with the WGCNA module yielded 507 genes related to both lactylation and PANoptosis.
Machine Learning Model Construction. From the 507 candidate genes, univariate Cox regression identified 85 prognostic markers. These were then used to train 117 different algorithm combinations using 10-fold cross-validation. Models were ranked by averaged concordance index (C-index) across training and all validation sets. The optimal model combined StepCox (forward stepwise selection) with Ridge regression regularization, which was selected for its superior and consistent performance across independent datasets.
Building the LAPRS Model. The StepCox plus Ridge model incorporating all 85 signature genes was designated the Lactylation and PANoptosis-Related Signature (LAPRS). The model achieved AUC values of approximately 0.7 in external validation sets for 1-, 3-, and 5-year overall survival prediction, with consistent performance across training and validation cohorts. Patients were stratified into high- and low-risk groups using the median LAPRS risk score, with the low-risk group showing significantly better overall survival in nearly all datasets.
Outperforming 55 Existing Models. The LAPRS model was systematically compared to 55 previously published LUAD prognostic models. Hazard ratio analysis demonstrated that LAPRS consistently achieved a higher C-index across all training and validation datasets, ranking among the top-performing models in AUC comparisons. The superior performance is attributed to the synergistic effect of incorporating both lactylation and PANoptosis pathways, which likely capture complementary regulatory mechanisms not addressed by previous models.
Clinical Integration Through Nomogram. A predictive nomogram was developed incorporating LAPRS risk score alongside clinical variables including age, T stage, N stage, and overall pathological stage. Calibration plots for 1-, 3-, and 5-year overall survival showed strong concordance between predicted and actual survival rates. Decision curve analysis demonstrated superior net clinical benefit compared to individual clinical factors alone, confirming the nomogram's utility for personalized prognostic assessment.
Independence as a Prognostic Factor. Univariate and multivariate Cox regression analyses across all datasets confirmed LAPRS as an independent prognostic factor for overall survival (p less than 0.001 in all analyses). This independence from standard clinical variables strengthens the case for LAPRS as a clinically applicable biomarker that adds information beyond conventional tumor staging.
High-Risk Patients Show More Genomic Instability. Comparative genomic analysis revealed significant differences in mutational landscapes between LAPRS risk groups. High-risk patients exhibited substantially elevated overall mutation frequencies. TP53 mutation frequency was 62% in high-risk patients versus 36% in low-risk patients. TTN mutations were 27% more frequent in high-risk patients. Co-occurring mutations were more common in high-risk patients, while low-risk patients showed preferential mutual exclusivity among mutations. Tumor mutation burden (TMB) was significantly higher in the high-risk group.
Low-Risk Patients Have Better Immune Infiltration. Seven independent immune infiltration algorithms consistently showed a negative correlation between LAPRS risk score and immune cell presence. Single-sample GSEA analysis confirmed that B cells, dendritic cells, and NK cells were significantly more abundant in the low-risk group. The low-risk group also showed higher scores for APC co-stimulation, checkpoint activity, MHC expression, T cell co-stimulation, and type II IFN response -- all markers of active anti-tumor immunity.
High-Risk Patients Are Less Likely to Benefit from Immunotherapy. High LAPRS score correlated with elevated TIDE scores, indicating greater immune evasion potential and reduced immunotherapy responsiveness. Conversely, low-risk patients showed higher expression of immune checkpoint molecules including CD80, CD86, CTLA4, ICOS, and TIGIT, suggesting greater potential benefit from immune checkpoint blockade therapy. The ESTIMATE algorithm confirmed lower tumor purity and higher immune/stromal scores in the low-risk group.
Differential Drug Sensitivity Profiles. The two risk groups showed distinct drug sensitivity patterns. The low-risk group was more sensitive to Dactolisib, AZD8055, Sabutoclax, and Topotecan. The high-risk group showed lower sensitivity to Paclitaxel, Docetaxel, AZD7762, and BI-2536. These differential sensitivity profiles may help guide treatment selection based on LAPRS risk stratification, offering a pathway toward more personalized chemotherapy choices.
Identifying APOL1 as the Central Hub. To find the most critical gene in the LAPRS, eight computational algorithms identified the most frequently recurring candidates among the 85-gene signature. APOL1 (apolipoprotein L1) emerged as the most consistent poor prognostic factor. High APOL1 expression was associated with shorter survival across multiple cohorts, while APOL1 was significantly upregulated in the high-risk LAPRS group. APOL1 also showed a positive correlation with LAPRS scores across all datasets.
How APOL1 Connects Lactylation and PANoptosis. Gene set enrichment analysis showed that APOL1 suppresses aberrant lactate accumulation while enhancing aerobic glycolysis. Enrichment analysis linked APOL1 to two caspase cascade-associated pathways, one of which overlapped with both downregulated lactylation and PANoptosis genes. Protein interaction analysis demonstrated that APOL1 connects to the lactylation-related gene VIM (vimentin) through TNF -- a PANoptosis-related gene -- forming a mechanistic bridge between the two pathways.
Cell-Type Specific Expression Patterns. Single-cell RNA sequencing analysis revealed high APOL1 expression in endothelial cells and fibroblasts. APOL1 and TNF were co-expressed in NK cells and T cells, correlating with immune activation. APOL1, TNF, and VIM were co-expressed in myeloid cells, suggesting that PANoptosis-lactylation interplay in myeloid cells may counteract APOL1's tumor-promoting effects through caspase cascade activation.
Experimental Validation in Cell Lines. Laboratory experiments in A549 and H1299 lung cancer cell lines confirmed that APOL1 overexpression significantly increased cell proliferation and viability compared to controls. Despite APOL1 showing overall low expression in LUAD tumor tissue (both at the transcriptomic and proteomic levels), its overexpression consistently promoted tumor growth, confirming its oncogenic role. This apparent paradox suggests APOL1 may have concentration- and context-dependent dual functions, similar to other cancer genes like MAP3K4 and STAT3.
A New Framework for LUAD Stratification. The LAPRS model offers an innovative approach to personalized prognosis and treatment guidance for LUAD patients. By capturing the combined regulatory role of lactylation and PANoptosis -- two complementary molecular mechanisms -- the model provides superior stratification compared to existing approaches. The accompanying nomogram integrates LAPRS with standard clinical variables to generate individualized survival predictions directly applicable to clinical practice.
Therapeutic Targeting Opportunities. The distinct drug sensitivity profiles identified for high- and low-risk LAPRS groups provide actionable guidance for treatment selection. The identification of immune checkpoint expression patterns in the low-risk group specifically supports the use of ICB therapy in this patient subset. The APOL1-TNF-VIM axis represents a potential therapeutic target, with the lactylation-PANoptosis interplay offering the possibility of counteracting APOL1's oncogenic effects.
Limitations and Future Work. The study relies primarily on publicly available RNA sequencing datasets, and APOL1's mechanistic role modulated by the lactylation-PANoptosis axis requires further experimental validation. The paradoxical dual roles of APOL1 -- promoting tumor proliferation at low basal expression while potentially participating in immune activation -- need additional investigation. Future studies should explore whether targeting the APOL1-TNF-VIM pathway can improve outcomes in high-risk LAPRS patients.
Broader Significance. This research demonstrates the value of combining multi-omics data integration with machine learning for prognostic model development. The finding that lactylation and PANoptosis share common protective genes in LUAD opens new avenues for understanding tumor metabolism and programmed cell death interactions. The LAPRS framework may serve as a template for developing similar integrative prognostic signatures in other cancer types.