What Is Pyroptosis Pyroptosis is a form of pro-inflammatory programmed cell death triggered by inflammasome activation. Named from Greek roots meaning 'fire falling,' it involves caspase-1 activation, gasdermin D cleavage, plasma membrane pore formation, cell swelling, and release of inflammatory cytokines like IL-18 and IL-1B. Unlike apoptosis, pyroptosis is explicitly inflammatory and generates a strong immune signal.
Pyroptosis in Melanoma The role of pyroptosis in cancer is paradoxical: inducing pyroptosis can kill tumor cells, but as a pro-inflammatory death, it can also create a microenvironment that promotes tumor growth. In skin cutaneous melanoma (SKCM), aberrant expression of pyroptosis-related genes (PRGs) has been linked to metastasis, invasion, and drug resistance. This study was the first to systematically evaluate PRG signatures for SKCM diagnosis and prognosis.
Study Design Researchers identified 20 PRGs from the GeneCards database and analyzed their expression across TCGA-SKCM and multiple GEO datasets. Nine machine learning algorithms were used to build diagnostic classifiers distinguishing SKCM from normal tissue. A LASSO Cox regression model then built a prognostic risk score from 9 PRGs, which was validated in independent datasets.
Expression Differences All 20 PRGs showed statistically significant differential expression between 471 SKCM tumor samples and 557 normal skin samples (q-value < 0.1). Eleven genes (including CASP1, PYCARD, IL18, GSDMA, NLRP1) were downregulated in tumors, while 9 genes (including NLRP3, AIM2, IL1B, GSDME, CASP3) were upregulated. This widespread dysregulation indicates pyroptosis is broadly altered in melanoma.
Survival Associations Thirteen of the 20 PRGs showed significant associations with overall survival. Eleven were protective factors (hazard ratio less than 1), suggesting their higher expression correlated with better outcomes, while 2 were risk factors. This mix makes single-gene prognosis unreliable, motivating a multi-gene signature approach.
Protein Interaction Network A protein-protein interaction (PPI) network analysis identified CASP1, CASP3, GSDMD, NLRP3, PYCARD, AIM2, and NLRC4 as central hubs in melanoma pyroptosis signaling. Immunohistochemistry from the Human Protein Atlas confirmed these proteins are indeed present at high levels in SKCM tissue, confirming that pyroptosis is actively occurring in a large proportion of melanoma tumors.
Classifier Performance Nine algorithms - KNN, logistic regression, SVM, ANN, decision tree, random forest, XGBoost, LightGBM, and CatBoost - were trained on PRG expression data from GSE98394 (primary melanoma vs. common nevi). All algorithms except the decision tree achieved AUC values above 0.900 in internal testing, and performed well in two external validation sets.
ANN as the Top Performer The artificial neural network (ANN) was identified as the most suitable algorithm, achieving excellent accuracy and robustness across all three datasets. Logistic regression, SVM, and random forest also performed consistently well. The decision tree underperformed in external validation, likely due to overfitting to the training set structure.
Clinical Relevance Since the classifiers were trained on a dataset containing benign nevi and melanoma, and validated against normal skin tissue and benign nevi separately, they have potential to distinguish malignant melanoma from both common nevi and normal skin - a clinically relevant diagnostic task. The classifiers using multi-gene PRG signatures substantially outperformed any single gene as a diagnostic marker.
Model Construction LASSO Cox regression on TCGA-SKCM training data selected 9 PRGs (GSDMD, GSDME, CASP4, GSDMC, NLRC4, APIP, AIM2, CASP3, IL18) and their coefficients to build a risk score formula. Patients were divided at the median risk score into low- and high-risk groups, which showed significantly different overall survival in Kaplan-Meier analysis.
Prognostic Performance Time-dependent ROC analysis demonstrated AUC of 0.640 for 3-year survival prediction and 0.711 for 5-year survival in the training set. Both the internal test set and external validation set (merged GSE54467 and GSE65904) showed consistent results, confirming the model generalizes beyond the training data.
Independent Prognostic Factor Multivariate Cox regression including tumor stage, T-stage, N-stage, M-stage, and the risk score showed the risk score remained an independent prognostic factor (HR = 2.078, p < 0.001). High-risk patients had more stage-IV tumors and more T1 classification extremes, and the risk score provided information beyond what staging alone could capture.
GSEA Findings Gene set enrichment analysis (GSEA) identified 53 immune-related pathways significantly enriched in low-risk patients, including the chemokine signaling pathway, Toll-like receptor signaling, T cell receptor signaling, leukocyte transendothelial migration, cytokine-cytokine receptor interaction, and NK cell-mediated cytotoxicity. No pathways were significantly enriched in the high-risk group.
Macrophage Polarization CIBERSORT immune deconvolution revealed that low-risk patients had higher fractions of activated CD4+ memory T cells, gamma-delta T cells, and M1 macrophages, while high-risk patients had higher M2 macrophage fractions. Survival analysis confirmed that high M1 or low M2 macrophage content was associated with better outcomes, pointing to macrophage polarization as a key mediator of PRG-related prognosis.
The M1/M2 Connection M1 macrophages are pro-inflammatory and anti-tumor, presenting antigens via MHC class I and II molecules. M2 macrophages are anti-inflammatory and pro-tumor. The finding that high-risk (low pyroptosis activity) patients have more M2 tumor-associated macrophages (TAMs) suggests that pyroptosis may normally help maintain an anti-tumor immune microenvironment by promoting inflammatory M1 polarization.
Risk-Correlated Gene Analysis Pearson correlation analysis between the risk score and individual gene expression identified genes with strong associations (|correlation| greater than or equal to 0.6, p < 0.05). Six genes showed particularly strong correlations, and all were associated with better survival when more highly expressed. NLRC4 was the most strongly correlated gene with the overall risk score.
NLRC4 Biology NLRC4 is an inflammasome component that activates caspase-1. Animal studies have shown that Nlrc4-knockout mice develop more melanoma when injected with B16F10 cells, suggesting NLRC4 normally suppresses tumor growth. Its strong correlation with the risk score suggests that NLRC4 inflammasome activity may serve as a key gatekeeper of anti-tumor pyroptosis signaling in SKCM.
GSDME and the Caspase-3 Pathway GSDME, which is cleaved by caspase-3 to trigger pyroptosis, was found to be expressed in the majority of SKCM tumors. Its expression level was not significantly linked to immune cell infiltration, suggesting GSDME-mediated pyroptosis in melanoma cells is not simply a byproduct of immune activity. The Casp3/GSDME pathway appears to be a distinct route by which melanoma cells undergo pyroptosis and release inflammatory cytokines.
Transcriptomic Focus This study relied exclusively on mRNA expression data. Protein levels, post-translational modifications, and spatial patterns of pyroptosis within tumors were not examined. Future studies should validate key PRGs at the protein level and use spatial transcriptomics or multiplexed immunohistochemistry to understand the cell-type-specific distribution of pyroptosis activity within tumors.
Therapeutic Implications The authors hypothesize that PRG-based risk scores could monitor responses to BRAF and MEK inhibitors, which are known to affect the immune microenvironment through pyroptosis pathways. Similarly, patients on immune checkpoint therapy (like ipilimumab) may have altered PRG expression through autoimmune mechanisms. Testing whether PRG scores predict immunotherapy response is an important next step.
Multi-Omics Extension The study used only transcriptomic data. Extending the analysis to include proteomics, metabolomics, and single-cell RNA sequencing would provide a more complete picture of pyroptosis biology in melanoma. Single-cell data could resolve which specific cell populations - melanoma cells, stromal cells, immune cells - are undergoing pyroptosis and how this affects tumor evolution and drug resistance.