A TRPM4-associated necrotic cell death signature for prognostic stratification in pancreatic cancer

Introduction

Pancreatic cancer is a highly lethal malignancy, projected to become the second leading cause of cancer-associated death in the United States by 2030 (1). The three main clinical challenges for pancreatic cancer patients include nonspecific symptoms, delayed diagnosis, and drug resistance (2). Pancreatic cancer tumorigenesis is driven not only by intrinsic genetic mutation but also by external factors such as the tumor microenvironment, providing opportunities for therapeutic intervention and disease progression control (3,4).

Cell death can be broadly categorized into two types: accidental cell death (ACD) and regulatory cell death (RCD). ACD is typically induced by irreversible physical or chemical damage and occurs in a non-programmatic manner. Conversely, RCD is regulated by intrinsic molecular signaling pathways and can be modulated by pharmacological or genetic interventions (5,6). Cell membranes, including plasma and organic membranes not only function as semi-permeable boundaries but also play a crucial role in regulating various extracellular and intracellular biological processes (BPs) (7). Ion channel proteins—the major membrane-localized proteins—maintain cellular ion homeostasis and modulate key metabolic activities (8,9). Dysfunction of ion channels, including those for calcium, iron, and copper, can trigger cell death (10-12). Sodium levels, in particular, are tightly regulated by the Na⁺/K⁺ pump and associated ion channels, and excessive intracellular sodium accumulation can induce cell death (13). TRPM4, a member of the transient receptor potential cation channel subfamily, is the first gene reported to be associated with necrosis by sodium overload (NECSO), a form of RCD that results in cell and tissue injury (14). Aberrant TRPM4 activation can cause necrotic cell death by driving a massive sodium influx and membrane depolarization. Currently, TRPM4 remains the only known gene linked to NECSO, highlighting the urgent need to identify additional genes or pathways involved in this process, which TRPM4-associated transcriptional patterns may reflect NECSO-related features.

Bioinformatics analysis serves as a powerful tool for elucidating the molecular network underlying various diseases, and may provide potential insights into their molecular functions (MFs). We retrieved comprehensive pancreatic cancer-related transcriptome and clinical data from The Cancer Genome Atlas Genomic Data Commons (TCGA GDC, https://portal.gdc.cancer.gov/) and Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/) databases. We aimed to investigate the potential association between TRPM4-associated necrotic cell death-related signature and tumor grade and prognosis in pancreatic cancer. Specifically, the cellular infiltration and molecular features of necrotic cell death-related transcription, as well as their association with the clinicopathological characteristics of pancreatic cancer, remain unexplored. Determining the clinical significance of TRPM4-related signature is valuable for elucidating their role in pancreatic cancer tumorigenesis and for exploring potential target drugs and therapeutic sensitivities. To our knowledge, this study represents the first integrative investigation of the role of TRPM4-associated signature in various pathophysiologic functions of pancreatic cancer. We present this article in accordance with the TRIPOD reporting checklist (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-1-968/rc).

Methods

Data acquisition

We obtained the gene expression profiles and clinical traits of pancreatic cancer patients from the Genomic Data Commons Data Portal (https://portal.gdc.cancer.gov/; accessed on September 2025). We also downloaded and processed clinical, mRNA, and tumor mutation burden (TMB) data from the TCGA-pancreatic adenocarcinoma (PAAD), following Zhou et al. (15). After data preprocessing, a total of 185 TCGA-PAAD samples with complete gene expression profiles and survival information were included in the co-expression and survival analyses. Samples with missing survival data or follow-up time equal to zero were excluded. The final sample size used for model construction and prognostic analyses was 179. We obtained 139 pancreatic cancer samples from GSE183795. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Because all data for this study were derived from public databases, ethical approval was not required.

Co-expression analysis and functional enrichment of genes involved in sodium over-load-induced necrosis

To derive a TRPM4-associated prognostic gene signature, we first explored genes co-expressed with TRPM4 in the TCGA-PAAD cohort. We used the TCGA-PAAD mRNA data to explore the relationship between TRPM4 and pancreatic cancer. Genes co-expressed with TRPM4 were identified using Pearson’s correlation analysis with |R|>0.3 and P<0.001. After screening out 645 genes, we visualized the top 50 genes using the R package “igraph” (v 2.1.4) (16). Subsequently, Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways were enriched in the co-expressed genes and analyzed using the R package “clusterprofile” (v 4.14.6) (17). The top five GO terms and the top 15 pathways of enrichment results were visualized using the “enrichplot” package (v 1.26.6), defined by a statistical significance threshold of P<0.05 (18). These TRPM4 co-expressed genes were subsequently subjected to survival analysis and model construction, as described below.

Established TRPM4-related gene signature

Based on the TRPM4 co-expressed genes identified above, univariate Cox regression was first performed to screen survival-related candidates, followed by least absolute shrinkage and selection operator (LASSO)-Cox regression using the R package “glmnet” (v 4.1.8) to establishe the final prognostic gene signature. The optimal penalty parameter (lambda) was selected via 10-fold cross-validation using the “cv.glmnet” function, and lambda.min (minimizing the cross-validated partial likelihood deviance) was chosen to derive the final gene signature (19-21). The TCGA-PAAD cohort was used as the training set, and the GSE183795 dataset was used as an external validation cohort to assess the robustness of the model (22). The risk score was calculated as follows: risk score = 0.0153 × (MSLN expression) + 0.0035 × (RAB27B expression) + 0.1352 × (MYEOV expression) + 0.0832 × (ECT2 expression) − 0.0118 × (KATNAL2 expression) − 0.2222 × (SAT2 expression) − 0.3456 × (SLC26A11 expression).

To assess the prognostic ability of our model risk score, we conducted further analyses for the high- and low-risk groups, including survival, clinical traits, and Nomo-prediction. In the TCGA training cohort, patients were stratified into high- and low-risk groups using the median risk score as the cutoff. In the external GEO validation cohort, the optimal cutoff was determined using the maximally selected rank statistics method implemented in the “surv_cutpoint” function [“survminer” (v 0.5.1) package], and patients were classified accordingly (23). This approach was used to account for potential distribution differences of risk scores across cohorts. For overall and progression-free survival (PFS) analyses, we visualized Kaplan-Meier survival curves according to their high/low risk scores and used the log-rank test with P<0.05 indicating significance. The receiver operating characteristic (ROC) curve was used to verify prognostic performance. Time-dependent ROC curves were generated to evaluate the prognostic performance of the TRPM4-related signature at 1-, 3-, and 5-year overall survival (OS) using the R packages “survival” (v 3.8.3) (https://github.com/therneau/survival.git) and “timeROC” (v 0.4) (24). Finally, we developed the prognostic signature model using the risk score and other clinical traits from the TCGA-PAAD dataset.

Gene set enrichment analysis (GSEA)

To clarify the potential biological functions of different risk groups enriched in gene groups and pathways, enrichment terms were analyzed for the entire TCGA-PAAD dataset. This analysis revealed biological functions and pathways related toTRPM4-associated prognostic signature. For the GSEA, we selected the gene sets “c2.cp.kegg.Hs.symbols.gmt” and “c5.go.Hs.symbols.gmt”. Statistical significance was set at P<0.05, and the top five |Normalized Enrichment Score (NES)| values were visualized.

Relationship between TRPM4-related genes and immune infiltration

We obtained data on the infiltration of immune cells in PAAD using the cell-type identification by estimating relative subsets of RNA transcripts (CIBERSORT) algorithm. Subsequently, a violin plot was created to visualize the high- and low-risk groups, and a heatmap was constructed to illustrate the correlations among our model TRPM4-associated prognostic signature. The Mann-Whitney U test and Spearman correlation analysis were used to examine our results, for which P<0.05 was considered statistically significant.

Correlation among the potential model significance in TMB and staining in the protein atlas

Somatic mutation data of the PAAD cohort were obtained from the TCGA GDC (https://portal.gdc.cancer.gov/; accessed on November 2025). The top 15 most frequently mutated genes were identified in the high- and low-risk groups, and their somatic mutation profiles were visualized using the “maftools” package (v 3.20) (25). Revealing the potential relationship between TMB and model significance facilitated the evaluation of the clinical benefits of immunotherapy for patients with pancreatic cancer, as well as patient survival at different TMB levels. We also retrieved data on protein staining, patients, and antibodies from the Human Protein Atlas (https://www.proteinatlas.org/; accessed on November 2025). Immunohistochemistry (IHC) staining revealed the quantity and location of expressed genes.

Drug sensitivity analysis and molecular docking

Given that drug sensitivity is a significant factor in managing pancreatic cancer patients, the discovery of novel drugs is important. The Genomics of Drug Sensitivity in Cancer (GDSC) database (https://www.cancerrxgene.org/; accessed on November 2025) provides information on potential drug sensitivity related to both chemotherapy and immunotherapy. The three-dimensional (3D) structures of proteins were obtained from the Protein Data Bank (PDB) database (https://www.rcsb.org/; accessed on November 2025) and their molecular structures from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/; accessed on 20 June 2025). To strengthen the evidence supporting potential drugs, we performed molecular docking, which provided insights into ligand binding with receptors and Vina scores, using online docking tools (https://cadd.labshare.cn/cb-dock2/index.php; accessed on November 2025).

Statistical analysis

All statistical analyses were performed using R software (v 4.2.2, available on 31 October 2024; https://www.r-project.org/) and pearl (v 5.34.1). The Mann-Whitney U test was used to analyze the relationship between TRPM4-related signature and the clinical traits of pancreatic cancer patients. The Spearman correlation coefficient clarified the potential correlations between TRPM4-related signature and the immune microenvironment. Kaplan-Meier curves and the log-rank test were used to analyze survival. Results were considered statistically significant at P<0.05.

Results

TRPM4 was verified on the public database

Utilizing publicly available datasets [including TCGA-PAAD (Figure 1A) and GSE183795 (Figure 1B)], we found that TRPM4 expression exhibited significant differences according to the prognoses of pancreatic cancer patients. High TRPM4 expression was significantly associated with a poor prognosis. TRPM4 could not be detected in normal pancreatic tissue (Figure 1C) but could be detected in pancreatic tumor tissue at a medium intensity (Figure 1D), according to data obtained from the Human Protein Atlas (https://www.proteinatlas.org/search/TRPM4).

Figure 1 TRPM4 expression displayed differences. TRPM4 exhibits differences in overall survival in TCGA (A) and GSE183795 (B), and in protein expression in normal pancreatic tissue (https://www.proteinatlas.org/ENSG00000130529-TRPM4/tissue/pancreas) (C) and tumor tissue (https://www.proteinatlas.org/ENSG00000130529-TRPM4/cancer/pancreatic+cancer) (D). Survival differences were assessed using Kaplan-Meier analysis and the log-rank test. TCGA, The Cancer Genome Atlas.

Co-expression analysis of the TRPM4-related gene signature

It is well-established that TRPM4 is the only gene related to necrotic cell death identified in the published literature to date (14). Visualization of the top 50 co-expressed genes (|R|>0.5376, and P<0.001) revealed that most of the genes in the TRPM4-associated Pearson’s correlation-based gene co-expression network were positively correlated with TRPM4, with only one gene, CERS4, demonstrating a negative correlation (Figure 2A). Furthermore, GO and KEGG enrichment analyses were conducted to reveal the biological functions and metabolic pathways of these genes. GO terms, including BP, cellular component (CC), and MF, were enriched and are detailed in Figure S1. Among the BP terms, actin filament organization and myeloid cell differentiation had the highest enrichment counts (Figure 2B). The clustered GO bubble plot groups similar functions together (Figure 2C). KEGG pathways were enriched in signaling pathways, including the Ras, chemokine, and MAPK pathways (Figure 2D). Subsequently, the dendrogram displayed their inherent connections (Figure 2E), and the size of the dots reveals the scale among our samples.

Figure 2 TRPM4-related genes. The top 50 co-expressed genes are displayed (A). GO enrichment of biological processes (B) and clustered GO bubble plot (C) for the co-expressed genes. KEGG pathway enrichment of these related genes in bubble plot (D) and clustered KEGG bubble chart (E). BP, biological process; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; No., number of genes.

Establishment of the TRPM4-associated signature

The TCGA-PAAD dataset, comprising 185 samples, provides comprehensive clinical data on pancreatic cancer patients. Following data processing, we obtained 179 valid samples. Univariate Cox regression analysis screened 60 genes significantly associated with survival (Figure S2). The TCGA-PAAD dataset was combined with our datasets and used as the training group, and the GSE183795 dataset served as the testing group. Based on the LASSO coefficient profiles (Figure 3A) and a cross-validation plot (Figure 3B), the LASSO Cox algorithm identified seven optimal genes among the 60 candidate prognostic genes. Ultimately, we constructed TRPM4-related signatures based on these seven genes [risk score = 0.0153 × (Exp MSLN) + 0.0035 × (Exp RAB27B) + 0.1352 × (Exp MYEOV) + 0.0832 × (Exp ECT2) − 0.0118 × (Exp KATNAL2) − 0.2222 × (Exp SAT2) − 0.3456 × (Exp SLC26A11) (Exp: expressions)].

Figure 3 Established TRPM4-associated signature. LASSO coefficient profiles demonstrate the complexity of our model (A). Cross-validation plot showing the relationship between partial likelihood deviance and log (λ) (B). λ, penalty parameter selected by 10-fold cross-validation. LASSO, least absolute shrinkage and selection operator.

Verification of the TRPM4-associated signature of OS

The survival analysis provided putative evidence to support the relevance of the TRPM4-associated signature. Significant differences were observed within the different datasets (Figure 1A,1B), showing an association between higher TRPM4 expression and a poor prognosis. The Figure 4A was sourced from the TCGA-PAAD dataset as the training groups and show a statistically significant difference between the high- and low-risk groups (P<0.001). The Figure 4B was sourced from the GSE183795 dataset as the test group, and also demonstrated a significant difference between the two groups (P=0.006). In both datasets, the high-risk group was related to a worse prognosis for our TRPM4-associated signature. Additionally, there was a trend towards significance regarding the association between PFS and TRPM4 (P=0.050) in the TCGA-PAAD dataset (Figure S3A), while the association with risk score was statistically significant (P<0.001) (Figure S3B). The ROC curve indicated that the TRPM4-associated signature had higher sensitivity and superior specificity compared to TRPM4 alone (Figure S3C), with a greater area under the curve (AUC).

Figure 4 Validation of the TRPM4-associated prognostic signature in the TCGA-PAAD and GSE183795 cohorts. (A) TCGA-PAAD cohort: risk score distribution (patients ranked by increasing risk score), survival status and survival time distribution, heatmap of the seven-gene signature expression, and Kaplan-Meier overall survival curve for the high- and low-risk groups. (B) GSE183795 cohort: risk score distribution (patients ranked by increasing risk score), survival status and survival time distribution, heatmap of the seven-gene signature expression, and Kaplan-Meier overall survival curve for the high- and low-risk groups. High- and low-risk groups were defined based on the median risk score. In the TCGA-PAAD cohort, 89 patients were classified as high risk and 89 as low risk; in the GSE183795 cohort, 99 patients were classified as high risk and 27 as low risk. Overall survival differences were assessed using the log-rank test. PAAD, pancreatic adenocarcinoma; TCGA, The Cancer Genome Atlas.

Relationship between clinicopathological characteristics and prognostic risk score

Univariate Cox regression analysis demonstrated that age, grade, and risk score were associated with prognostic value (Figure 5A). Additionally, multivariate Cox regression analysis revealed that risk score may be the only independent prognostic factor for pancreatic ductal adenocarcinoma (PDAC) prognosis (Figure 5B), demonstrating a highly related TRPM4-associated signature. After combining the risk score with clinical characteristics, significant differences were observed in risk scores among the different age (Figure S4A), gender (Figure S4B), grade (Figure S4C), and stage (Figure S4D) groups. The ROC curve also illustrated that the risk score [AUC =0.748, 95% confidence interval (CI): 65.530–83.520] had higher sensitivity and specificity than other clinical traits (Figure S4E). Utilizing clinicopathological characteristics and the risk score, we developed an OS prognostic model (Figure 5C), which assigned weights to each factor. For example, if a patient with pancreatic cancer received a total score of 280, the analyzed 1-, 3-, and 5-year survival probabilities were 77.4%, 33.4%, and 24.3%, respectively. Calibration curves show that the OS rates analysis by our model are highly consistent with actual observations, especially for short-term survival probabilities (Figure 5D).

Figure 5 Clinical characteristics of the risk model. Univariate (A) and multivariate (B) Cox regression analyses demonstrate the clinical traits and risk score model for the independent prognostic index. Nomogram weighted clinical traits, predicted overall survival rate (C), and a calibrated plot display the similarity between predicted and observed overall survival (D). Univariate and multivariate Cox regression analyses were performed. Nomogram was constructed to predict overall survival. CI, confidence interval; OS, overall survival; HR, hazard ratio.

GSEA of the seven genes of the TRPM4-related signature

We performed GSEA to reveal functionally relevant gene sets, and then explored their biological relevance using GO and KEGG analyses in conjunction with the TCGA-PAAD dataset. In the high-risk group, GSEA revealed enrichment in processes related to cell keratinization and structural constituents of chromatin based on GO terms (Figure 6A). Additionally, cell cycle, spliceosome, and pyrimidine metabolism were enriched in the KEGG pathway (Figure 6B). In contrast, the low-risk group was enriched in immunoglobulin production, T cell activation, and plasma membrane receptor complex based on GO terms (Figure S5A). Similarly, the calcium signaling pathway, neuroactive receptor interaction, and primary immunodeficiency were enriched in the KEGG-based pathway (Figure S5B). The high-risk group exhibited proliferative and metabolic dysregulation, whereas the low-risk group displayed enhanced immune-related functions and preserved receptor signaling.

Figure 6 GSEA for the high-risk group. The enrichment of GO-based GSEA (A) and KEGG-based GSEA (B). GO- and KEGG-based gene sets were used. High risk samples in TCGA-PAAD (n=89) and GSE183795 (n=99), low risk group in TCGA-PAAD (n=89) and GSE183795 (n=27). BP, biological process; GO, Gene Ontology; GSEA, gene set enrichment analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; PAAD, pancreatic adenocarcinoma; TCGA, The Cancer Genome Atlas.

Immune cell infiltration, TMB, and IHC staining in relation to the TRPM4-associated signature

We employed CIBERSORT algorithm to further explore the relationship between the TRPM4-associated signature and immune cell infiltration. We observed a significant difference in immune cell infiltration, including CD8⁺ T cells (P=0.005) and eosinophils (P=0.046), between the high- and low-risk groups (Figure 7A). In addition, M0 macrophages and activated dendritic cells (DCs) were positively correlated with risk score, while CD8⁺ T cells were negatively correlated, according to Pearson’s correlation analysis (Figure 7B). Subsequently, we explored the relationship between the TRPM4-associated signature and TMB. The results revealed that the high-risk group had significantly higher mutation rates (Figure 7C-7E), with more frequent mutations in KRAS and TP53 (Figure 7D). Survival analysis based on TMB levels revealed a significant difference between the high- and low-TMB groups (Figure 7F), which was even more pronounced when combined with the risk score (Figure 7G). Patients with both high TMB and high-risk scores had the poorest survival probabilities. The Human Protein Atlas data showed the difference in the TRPM4-associated signatures between normal and tumor pancreas tissue. As shown in Figure S6, MSLN, MYEOV, and RAB27B expression was higher in tumor tissues than in normal tissues, with the opposite trend observed for KATNAL2 and SLC26A11. These results align with our findings and the modular risk score.

Figure 7 Association between TRPM4-associated signature and immune infiltration and tumor mutation state. Immune cells in high/low risk groups (A). Relationships among immune cells, TRPM4-associated signature, and risk score (B). Tumor mutation burden in the high risk (C) and low risk (D) groups. Violin plot displays the difference between high/low risk groups (E). Survival probability with high/low TMB (F). Survival probability is analyzed using the combination risk score and TMB (G). Survival analysis was performed using Kaplan-Meier curves and the log-rank test. High risk samples in TCGA-PAAD (n=89) and GSE183795 (n=99), low risk group in TCGA-PAAD (n=89) and GSE183795 (n=27). DCs, dendritic cells; H, high; L, low; NK, natural killer; PAAD, pancreatic adenocarcinoma; TCGA, The Cancer Genome Atlas; TMB, tumor mutation burden.

Drug sensitivity screening

As an exploratory analysis, we screened the sensitivity of 94 drugs using the TCGA and GEO datasets (Table S1). According to the latest National Comprehensive Cancer Network (NCCN) and European Society for Medical Oncology (ESMO) clinical guidelines, the first-line treatments recommended for pancreatic cancer are based on gemcitabine and 5-fluorouracil (5-FU), including GN, NALIRIFOX, and FOLFIRINOX (26,27). Three clinically recommended drugs were screened, including irinotecan (PubChem ID: 60838), olaparib (PubChem ID: 23725625), and oxaliplatin (PubChem ID: 9887053). Protein structures were obtained from the PDB database, including MSLN (PDB ID: 7U9J) (28), RAB27B (PDB ID: 2F7S) (29), and ECT2 (PDB ID: 3L46) (30). The plots showed the half maximal inhibitory concentration (IC50) of sensitive drugs for the TCGA-PAAD (Figure 8A) and GSE183795 (Figure 8B) datasets. Molecular docking simulations were conducted to assess binding affinities between proteins and drugs, which included the lowest Vina scores for binding, such as MSLN binding with irinotecan (binding energy: −11.1 kJ/mol) (Figure 8C), MSLN with olaparib (binding energy: −9.9 kJ/mol) (Figure 8D), and RAB27B with oxaliplatin (binding energy: −4.6 kJ/mol) (Figure 8E). The summary table presents the docking results of the three main model proteins with the selected drugs, including information about the drugs and proteins and the Vina scores (Table 1).

Figure 8 Drugs sensitivity. The half maximal inhibitory concentration based on TCGA-PAAD (A) and GSE183795 (B) datasets. Three-dimensional molecular docking and Vina scores shown for MSLN with irinotecan (C), MSLN with olaparib (D), and RAB27B with oxaliplatin (E). Molecular docking was performed as an exploratory analysis, and Vina scores represent predicted binding affinity. MSLN, mesothelin; PAAD, pancreatic adenocarcinoma; TCGA, The Cancer Genome Atlas.

Discussion

NECSO is a recently discovered regulated mode of cell death with only one associated gene having been reported in the literature—TRPM4 (14). Identifying TRPM4-associated genes is important for explaining the underlying molecular related mechanism and may provide insights into various diseases. Pancreatic cancer is a highly aggressive and lethal malignancy, with a median survival of approximately 4 months and a 5-year survival rate of 13%. Investigating the role of necrotic cell death in cancer can be helpful for our understanding of the potential related MFs underlying pancreatic cancer progression and its treatments (2).

In this study, we applied integrative analysis to construct prognostic signature and investigated the relationship between the TRPM4-associated signature and the tumor immune microenvironment (TIME) in pancreatic cancer. To identify the TRPM4-associated prognostic signature, we employed two significant public databases: the TCGA and the GEO. To our knowledge, this study represents the first attempt to develop a TRPM4-associated prognostic risk score in pancreatic cancer, which may aid prognostic stratification and provide a hypothesis-generating framework for future studies. The results of this integrative transcriptomic analysis showed that high risk scores were associated with poor prognosis, as well as differences in inferred immune cell infiltration patterns and somatic mutation profiles. In addition, univariate Cox analyses supported that the risk score model acted as a superior independent prognostic factor in pancreatic cancer patients compared to TRPM4 alone. Our analyses are based on retrospective bulk transcriptomic datasets. Therefore, the findings should be interpreted as association-level and hypothesis-generating rather than mechanistic or treatment-predictive.

Seven TRPM4-associated signatures were established in the TCGA-PAAD dataset, including MSLN, RAB27B, MYEOV, ECT2, KATNAL2, SAT2, and SLC26A11, for which Pearson’s correlation analysis revealed the same correlation as for our model risk score. Utilizing the TCGA-PAAD dataset for our training group and GSE183795 as the test group, we found that a high-risk score was related to shorter survival times and poor prognosis. Univariate Cox regression analyses revealed that risk score could serve as an independent factor to analysis prognosis, and the nomogram plot showed that the risk score weighted highly in analyzing OS, with the results aligning closely with observed OS. It makes sense to use our approach to analysis the prognosis of pancreatic cancer patients using their signature genes’ expression. It is well-established that acquired genetic mutations drive most forms of tumors and are involved in six cancerous hallmarks, including sustained proliferative signaling, evasion of growth suppressors, cell death resistance, replicative immortality, and activation of invasion and metastasis (31). Several mutations in key genes—TP53, ATM, BRCA1/2, SMAD4 and CDKN2A—are significantly involved in disease progression and the prognosis of pancreatic cancer patients (32-34). All these results highlight the significant value of key genes and their molecular features in the prognostic assessment of pancreatic cancer. Significantly, our prognostic model constructed based on TRPM4-related gene features outperformed TRPM4 alone in computational simulations. Prospectively, such molecular-level analysis models may provide a framework for future hypothesis-driven studies and contribute to a better understanding of prognostic heterogeneity in pancreatic cancer.

Furthermore, enrichment in GO and KEGG terms through the GSEA correlated with the epithelial-to-mesenchymal transition (EMT) and genetically BPs, especially in the high-risk score group. The EMT drives embryonic development and carcinoma progression and metastasis and plays a role in detaching the basement membrane and invading nearby biological structures (35,36). Further, immune infiltration analysis based on bulk transcriptomic deconvolution indicated that CD8⁺ T cell abundance was negatively correlated with the risk score. These findings should be interpreted as descriptive and hypothesis-generating, rather than direct evidence of immunotherapy benefit. Less immune cell infiltration may increase immune-associated biological functions, which is absent in immune-associated treatment in pancreatic cancer (37,38). However, it should be noted that PDAC is widely recognized as an immunologically cold tumor (39). Therefore, immune infiltration and TMB analyses based on bulk RNA sequencing data should be interpreted with more caution. The immune cell proportions inferred in this study reflect computational deconvolution rather than direct measurements. Subsequently, the high-risk score group uncovered a higher number of KRAS and TP53 mutations, and the OS of the combined risk score and TMB was significant. These observations suggest a potential association between higher risk scores and distinct somatic mutation patterns, which may reflect differences in tumor biology. However, no direct conclusions regarding immune escape or immunotherapy response can be drawn from the present data.

The TRPM4-associated signature, comprising four positively correlated genes (MSLN, RAB27B, MYEOV, and ECT2) and three negative genes (KATNAL2, SAT2, and SLC26A11), represents a prognostic model for risk stratification in pancreatic cancer. The positive components of this signature are all functionally linked to tumor progression. For instance, MSLN enhances cell adhesion, tumor invasion, progression, and metastasis, and is an ideal candidate for targeted therapy (13). It is currently being studied in a clinical trial for pancreatic cancer with high MSLN, indicating its potential dual role as a biomarker and therapeutic target (40,41). Its overexpression may enhance TRPM4-mediated calcium signaling, facilitating cellular motility and invasion. Similarly, RAB27B belongs to the RAS oncogene family and is involved in the multivesicular body sorting pathway and the positive regulation of exocytosis (https://www.ncbi.nlm.nih.gov/gene/5874). Bioinformatics analysis indicated that it can serve as an independent marker for pancreatic cancer (42). RAB27B-mediated treatment could be a promising therapeutic strategy for metastatic pancreatic cancer (43), potentially in tandem with TRPM4-releated membrane regulation. MYEOV (44) and ECT2 (45) promote cell cycle progression and DNA synthesis, and their upregulation may synergize with TRPM4 in enhancing the proliferative capacity of PDAC cells. In contrast, the three negatively correlated genes—KATNAL2 (46,47), SAT2 (48), and SLC26A11 (49)—are generally associated with cell cycle regulation, epigenetic stability, and ion transport. Their downregulation may reflect a suppression of tumor-suppressive pathways. Notably, KATNAL2, involved in microtubule severing, may regulate cell division fidelity (47); its loss could allow TRPM4-associated aberrant cell division to proceed unchecked. Collectively, the signatures of these seven genes not only effectively risk-stratify pancreatic cancer patients, but also reveal a potential related biological function by which tumor cells are tilted towards pro-oncogenic signaling pathways. From an integrative perspective, TRPM4 expression may be associated with a broader transcriptional state reflected by the joint expression of the signature genes. However, the present study does not provide direct mechanistic evidence, and these associations should be interpreted cautiously.

We conducted molecular docking with these four positive genes, illustrating TRPM4-associated signatures according to the Vina score of several clinical drugs. MSLN, RAB27B and ECT2 were found to be positively associated with our risk score model. The final drugs selected for molecular docking were chosen based on their clinical relevance in pancreatic cancer treatment according to current NCCN and ESMO guidelines, and their predicted differential sensitivity between high- and low-risk groups in the GDSC-based analysis. This strategy was intended to prioritize clinically relevant compounds for exploratory in silico evaluation. In addition, molecular docking analyses were performed as exploratory investigations to assess potential ligand-protein interactions. These results provide supportive, hypothesis-generating information rather than direct biological or clinical evidence, which warrant further experimental validation.

To summarize, this is the first study to construct a risk model based on TRPM4-associated genes in pancreatic cancer. The model showed that a high-risk score was associated with a poor prognosis, suggesting that it may be a related prognosis biomarker. Further analysis of the risk model involved the TIME, finding that T cell CD8⁺ was correlated with low-risk patient scores. High risk scores were associated with somatic mutation in KRAS and TP53; these significant mutations prevail in enhanced oncogenes, resulting in tumorigenesis (50,51).

Several limitations of this study should be acknowledged. First, all analyses were based on bulk RNA sequencing data, which cannot resolve cell-type-specific expression patterns or establish causal relationships. Second, although LASSO regularization and an external validation cohort were applied, the risk of model overfitting cannot be fully excluded. The validation cohort was relatively small and methodologically similar to the training dataset, and no resampling or gene stability analyses were performed, which may limit the generalizability of the proposed TRPM4-associated signature. Further validation in larger, independent, and ideally prospective cohorts across different platforms will be required to confirm the robustness and reproducibility of the model. Third, this study relied on publicly available retrospective datasets, which may introduce selection bias. In addition, immune infiltration analyses were inferred computationally and should be interpreted with caution. Finally, although the proposed TRPM4-related nomogram demonstrated prognostic relevance, its clinical applicability remains exploratory. Prospective studies and experimental validation are required before any potential clinical implementation.

Conclusions

In conclusion, we identified a TRPM4-associated gene signature that enables prognostic stratification of patients with pancreatic cancer. While exploratory analyses suggested associations with immune-related features and drug sensitivity patterns, the present study does not support predictive or therapeutic claims. Further studies incorporating treatment response data and prospective validation are required to assess any predictive utility.

Acknowledgments

The authors would like to thank the Translational Research Center (TRC), Erlangen Hospital, for their support and assistance. This work was supported by the China Scholarship Council (CSC).

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-1-968/rc

Peer Review File: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-1-968/prf

Funding: This work was supported by the China Scholarship Council (CSC) (No. 202408080222).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-1-968/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Rahib L, Wehner MR, Matrisian LM, et al. Estimated Projection of US Cancer Incidence and Death to 2040. JAMA Netw Open 2021;4:e214708. [Crossref] [PubMed]
2. Stoop TF, Javed AA, Oba A, et al. Pancreatic cancer. Lancet 2025;405:1182-202. [Crossref] [PubMed]
3. Graham S, Dmitrieva M, Vendramini-Costa DB, et al. From precursor to cancer: decoding the intrinsic and extrinsic pathways of pancreatic intraepithelial neoplasia progression. Carcinogenesis 2024;45:801-16. [Crossref] [PubMed]
4. Jeiroshi A, Deng J, Xu Z, et al. Navigating the paradox of senescence and chemoresistance in pancreatic cancer. Semin Cancer Biol 2025;114:60-72. [Crossref] [PubMed]
5. Galluzzi L. Metabolic switches in cell death regulation. Cell Metab 2025;37:1252-4. [Crossref] [PubMed]
6. Chen F, Kang R, Tang D, et al. Ferroptosis: principles and significance in health and disease. J Hematol Oncol 2024;17:41. [Crossref] [PubMed]
7. Leonard TA, Loose M, Martens S. The membrane surface as a platform that organizes cellular and biochemical processes. Dev Cell 2023;58:1315-32. [Crossref] [PubMed]
8. Levin M. Bioelectric signaling: Reprogrammable circuits underlying embryogenesis, regeneration, and cancer. Cell 2021;184:1971-89. [Crossref] [PubMed]
9. Waxman SG, Zamponi GW. Regulating excitability of peripheral afferents: emerging ion channel targets. Nat Neurosci 2014;17:153-63. [Crossref] [PubMed]
10. An J, Zhang L, Duan Y, et al. Sodium’s role and therapeutic targeting in cancer. Trends Pharmacol Sci 2026;47:53-65. [Crossref] [PubMed]
11. Wang F, Görgülü K, Algül H, et al. The Role of Alcohol in Pancreatic Diseases: A Comprehensive Perspective. Gastroenterology 2026;170:268-86. [Crossref] [PubMed]
12. Tsvetkov P, Coy S, Petrova B, et al. Copper induces cell death by targeting lipoylated TCA cycle proteins. Science 2022;375:1254-61. [Crossref] [PubMed]
13. Yi S, Noh K, Kim H, et al. Advancing pancreatic cancer therapy by mesothelin-specific nanobody conjugation. Mol Cancer 2025;24:124. [Crossref] [PubMed]
14. Fu W, Wang J, Li T, et al. Persistent activation of TRPM4 triggers necrotic cell death characterized by sodium overload. Nat Chem Biol 2025;21:1238-49. [Crossref] [PubMed]
15. Zhou Y, Lu Y, Czubayko F, et al. Identification of Cancer Associated Fibroblasts Related Genes Signature to Facilitate Improved Prediction of Prognosis and Responses to Therapy in Patients with Pancreatic Cancer. Int J Mol Sci 2025;26:4876. [Crossref] [PubMed]
16. CsárdiGNepuszTMüllerKigraph for R: R interface of the igraph library for graph theory and network analysis.Zenodo 2025. doi: .10.5281/zenodo.14347716
17. Yu G. Thirteen years of clusterProfiler. Innovation (Camb) 2024;5:100722. [Crossref] [PubMed]
18. Yu G. enrichplot: Visualization of Functional Enrichment Result. 2025. Available online: https://bioconductor.org/packages/enrichplot/
19. Engebretsen S, Bohlin J. Statistical predictions with glmnet. Clin Epigenetics 2019;11:123. [Crossref] [PubMed]
20. Simon N, Friedman J, Hastie T, et al. Regularization Paths for Cox’s Proportional Hazards Model via Coordinate Descent. J Stat Softw 2011;39:1-13. [Crossref] [PubMed]
21. Friedman J, Hastie T, Tibshirani R. Regularization Paths for Generalized Linear Models via Coordinate Descent. J Stat Softw 2010;33:1-22. [Crossref] [PubMed]
22. Yang S, Tang W, Azizian A, et al. Dysregulation of HNF1B/Clusterin axis enhances disease progression in a highly aggressive subset of pancreatic cancer patients. Carcinogenesis 2022;43:1198-210. [Crossref] [PubMed]
23. Kassambara A, Kosinski M, Biecek P. survminer: Drawing Survival Curves using ‘ggplot2’. 2025. Available online: https://cran.r-project.org/package=survminer
24. Blanche P, Dartigues JF, Jacqmin-Gadda H. Estimating and comparing time-dependent areas under receiver operating characteristic curves for censored event times with competing risks. Stat Med 2013;32:5381-97. [Crossref] [PubMed]
25. Mayakonda A, Lin DC, Assenov Y, et al. Maftools: efficient and comprehensive analysis of somatic variants in cancer. Genome Res 2018;28:1747-56. [Crossref] [PubMed]
26. Network NCC. NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines®) Pancreatic Adenocarcinoma. 2025. Available online: https://www.nccn.org/professionals/physician_gls/pdf/pancreatic.pdf
27. Conroy T, Ducreux MESMO Guidelines Committee. Electronic address: clinicalguidelines@esmo. ESMO Clinical Practice Guideline Express Update on the management of metastatic pancreatic cancer. ESMO Open 2025;10:104528. [Crossref] [PubMed]
28. Zhan J, Lin D, Watson N, et al. Structures of Cancer Antigen Mesothelin and Its Complexes with Therapeutic Antibodies. Cancer Res Commun 2023;3:175-91. [Crossref] [PubMed]
29. Kukimoto-Niino M, Sakamoto A, Kanno E, et al. Structural basis for the exclusive specificity of Slac2-a/melanophilin for the Rab27 GTPases. Structure 2008;16:1478-90. [Crossref] [PubMed]
30. Zou Y, Shao Z, Peng J, et al. Crystal structure of triple-BRCT-domain of ECT2 and insights into the binding characteristics to CYK-4. FEBS Lett 2014;588:2911-20. [Crossref] [PubMed]
31. Gavish A, Tyler M, Greenwald AC, et al. Hallmarks of transcriptional intratumour heterogeneity across a thousand tumours. Nature 2023;618:598-606. [Crossref] [PubMed]
32. Hsu FC, Roberts NJ, Childs E, et al. Risk of Pancreatic Cancer Among Individuals With Pathogenic Variants in the ATM Gene. JAMA Oncol 2021;7:1664-8. [Crossref] [PubMed]
33. Hayashi A, Hong J, Iacobuzio-Donahue CA. The pancreatic cancer genome revisited. Nat Rev Gastroenterol Hepatol 2021;18:469-81. [Crossref] [PubMed]
34. Qian Y, Gong Y, Fan Z, et al. Molecular alterations and targeted therapy in pancreatic ductal adenocarcinoma. J Hematol Oncol 2020;13:130. [Crossref] [PubMed]
35. Kiri S, Ryba T. Cancer, metastasis, and the epigenome. Mol Cancer 2024;23:154. [Crossref] [PubMed]
36. Jinesh GG, Brohl AS. Classical epithelial-mesenchymal transition (EMT) and alternative cell death process-driven blebbishield metastatic-witch (BMW) pathways to cancer metastasis. Signal Transduct Target Ther 2022;7:296. [Crossref] [PubMed]
37. Meier P, Legrand AJ, Adam D, et al. Immunogenic cell death in cancer: targeting necroptosis to induce antitumour immunity. Nat Rev Cancer 2024;24:299-315. [Crossref] [PubMed]
38. Wagner J, Gößl D, Ustyanovska N, et al. Mesoporous Silica Nanoparticles as pH-Responsive Carrier for the Immune-Activating Drug Resiquimod Enhance the Local Immune Response in Mice. ACS Nano 2021;15:4450-66. [Crossref] [PubMed]
39. Kung HC, Zheng KW, Zimmerman JW, et al. The tumour microenvironment in pancreatic cancer - new clinical challenges, but more opportunities. Nat Rev Clin Oncol 2025;22:969-95. [Crossref] [PubMed]
40. Pang N, Shi J, Qin L, et al. IL-7 and CCL19-secreting CAR-T cell therapy for tumors with positive glypican-3 or mesothelin. J Hematol Oncol 2021;14:118. [Crossref] [PubMed]
41. Baldo P, Cecco S. Amatuximab and novel agents targeting mesothelin for solid tumors. Onco Targets Ther 2017;10:5337-53. [Crossref] [PubMed]
42. Mei Y, Liang D, Ai B, et al. Genome-wide identification of A-to-I RNA editing events provides the functional implications in PDAC. Front Oncol 2023;13:1092046. [Crossref] [PubMed]
43. Yang J, Zhang Z, Zhang Y, et al. ZIP4 Promotes Muscle Wasting and Cachexia in Mice With Orthotopic Pancreatic Tumors by Stimulating RAB27B-Regulated Release of Extracellular Vesicles From Cancer Cells. Gastroenterology 2019;156:722-734.e6. [Crossref] [PubMed]
44. Shen H, Ye F, Xu D, et al. The MYEOV-MYC association promotes oncogenic miR-17/93-5p expression in pancreatic ductal adenocarcinoma. Cell Death Dis 2021;13:15. [Crossref] [PubMed]
45. Xu D, Wang Y, Wu J, et al. ECT2 overexpression promotes the polarization of tumor-associated macrophages in hepatocellular carcinoma via the ECT2/PLK1/PTEN pathway. Cell Death Dis 2021;12:162. [Crossref] [PubMed]
46. DeSpenza T Jr, Singh A, Allington G, et al. Pathogenic variants in autism gene KATNAL2 cause hydrocephalus and disrupt neuronal connectivity by impairing ciliary microtubule dynamics. Proc Natl Acad Sci U S A 2024;121:e2314702121. [Crossref] [PubMed]
47. Stathatos GG, Dunleavy JEM, Zenker J, et al. Delta and epsilon tubulin in mammalian development. Trends Cell Biol 2021;31:774-87. [Crossref] [PubMed]
48. Wei W, Hu Q, Li W, et al. The Role of Ferroptosis Signature in Overall Survival and Chemotherapy of Pancreatic Adenocarcinoma. DNA Cell Biol 2022;41:116-27. [Crossref] [PubMed]
49. Zhao LP, Wang HJ, Hu D, et al. β-Elemene induced ferroptosis via TFEB-mediated GPX4 degradation in EGFR wide-type non-small cell lung cancer. J Adv Res 2024;62:257-72. [Crossref] [PubMed]
50. Larrayoz M, Garcia-Barchino MJ, Celay J, et al. Preclinical models for prediction of immunotherapy outcomes and immune evasion mechanisms in genetically heterogeneous multiple myeloma. Nat Med 2023;29:632-45. [Crossref] [PubMed]
51. Zhang Y, Zou J, Chen R. An M0 macrophage-related prognostic model for hepatocellular carcinoma. BMC Cancer 2022;22:791. [Crossref] [PubMed]