Construction of molecular signatures based on the co-expression network of NECSO-related gene TRPM4 and its prognostic value in hepatocellular carcinoma

Introduction

Hepatocellular carcinoma (HCC) is a malignant tumor with high incidence and mortality rate globally (1,2). The pathogenesis of HCC is intricate, and patient prognosis exhibits significant heterogeneity (3). The currently employed tumor-node-metastasis (TNM) staging system is limited in its ability to accurately predict patient outcomes (4,5). Consequently, the identification of novel molecular biomarkers and the development of effective prognostic models are essential for enhancing the precision of HCC diagnosis and treatment.

Recently, a novel form of programmed cell death, termed necrosis by sodium overload (NECSO), has garnered attention due to its potential regulatory role in tumorigenesis. NECSO is induced by necrocide 1 (NC1) and is closely associated with imbalances in intracellular sodium homeostasis (6-9). The transient receptor potential cation channel subfamily M member 4 (TRPM4) is a pivotal calcium-activated non-selective cation channel protein that primarily facilitates sodium influx and is a key regulator of sodium balance and NECSO (10,11). TRPM4 is ubiquitously expressed across various organs and immune cells, and mounting evidence indicates its elevated expression in malignancies such as prostate and breast cancers, where it contributes to tumor invasion and metastasis (6,8).

Accumulating evidence has linked TRPM4-mediated sodium influx to the execution of NECSO, which modulates cancer cell survival, invasion, and immune microenvironment remodeling (12). However, the precise function of TRPM4 in HCC, particularly within the NECSO-associated downstream regulatory network, remains insufficiently investigated. In this study, we propose that genes co-expressed with TRPM4 in HCC form a biologically coherent network with significant prognostic and immunological implications. By integrating bulk and single-cell transcriptomics with genomic and immune profiling, our objectives were to: (I) identify core genes associated with TRPM4, (II) develop and validate a robust prognostic signature, (III) elucidate its relationship with the tumor immune microenvironment and mutational burden, and (IV) delineate cell-type-specific expression patterns that underlie its prognostic effects. We present this article in accordance with the TRIPOD reporting checklist (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0363/rc).

Methods

Data acquisition and processing

The comprehensive analytical workflow employed in this study is depicted in a graphical abstract (refer to Figure S1). In summary, RNA sequencing data in fragments per kilobase million (FPKM) format, along with corresponding clinical annotations and somatic mutation profiles for HCC, were sourced from The Cancer Genome Atlas-Liver Hepatocellular Carcinoma (TCGA-LIHC; comprising 371 tumor samples and 50 normal samples). External validation was conducted using the GSE76427 cohort from the Gene Expression Omnibus (GEO; n=115). All datasets were processed utilizing R software (version 4.2.0), with expression values transformed into transcripts per million (TPM) to ensure cross-sample comparability (13). This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Identification of TRPM4-associated core genes

Differentially expressed genes (DEGs) between tumor and normal tissue samples were identified using the limma package, with criteria set at |log₂FC|>1 and a false discovery rate (FDR) of <0.05. Genes exhibiting co-expression with TRPM4 were identified through Pearson correlation analysis, with thresholds of |R|>0.3 and P<0.05. The intersection of DEGs and TRPM4-coexpressed genes was used to define the core gene set.

Functional enrichment analysis

Functional enrichment analyses, including Gene Ontology (GO) categories—biological process (BP), cellular component (CC), and molecular function (MF)—as well as Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment, were conducted on the core genes utilizing the clusterProfiler package, with a significance threshold set at P<0.05 (6,9,14).

Prognostic signature development and validation

To identify signature genes and develop the risk score model, univariate Cox proportional hazards regression analysis (P<0.05) and least absolute shrinkage and selection operator (LASSO)-Cox regression were employed (9,15). Patients were categorized into high- and low-risk groups based on the median risk score of 1.682. The prognostic efficacy was assessed through Kaplan-Meier survival analysis using the log-rank test, time-dependent ROC curves, and the concordance index (C-index) (15). The risk score was further validated as an independent prognostic factor using multivariate Cox regression analysis, and a nomogram accompanied by calibration curves was developed for individualized prognostic prediction (16).

Immune infiltration and tumor mutational burden (TMB) analysis

The CIBERSORT algorithm was employed to estimate the infiltration proportions of 22 distinct immune cell subtypes (17). Spearman correlation analysis assessed the associations between signature genes/risk scores and immune infiltration levels. TMB was calculated as the total number of nonsynonymous mutations per sample (7). Differences in survival were assessed across TMB subgroups and combined TMB-risk subgroups using Kaplan-Meier analysis and receiver operating characteristic (ROC) curves. The correlations between signature genes and TMB were examined using the Spearman test.

scRNA-seq analysis

scRNA-seq data, generated using the 10x Genomics platform, were processed with the Seurat package (version 5.3.0). Low-quality cells, defined as those with a mitochondrial gene ratio exceeding 15% or fewer than 50 detected genes, were excluded from the analysis (16). Cell type annotation was conducted using SingleR (version 2.8.0) with integrated reference databases, including Blueprint, ENCODE, and the Human Primary Cell Atlas. This was followed by Uniform Manifold Approximation and Projection (UMAP) for dimensionality reduction, Leiden clustering, and validation through canonical marker genes (18). Single-cell risk scores were computed for each major cell cluster to quantify the contribution of each cell type to the bulk-tissue risk signature.

Statistical analysis

All analyses were conducted using R software (version 4.2.0). Intergroup comparisons were carried out using either the t-test or the Wilcoxon test, while survival analysis was performed using the Kaplan-Meier method with the log-rank test. Correlation analyses were conducted using either the Pearson or Spearman test. A two-tailed P value of less than 0.05 was considered statistically significant.

Results

Core gene screening, functional enrichment analysis, and construction of the LASSO-Cox prognostic model

To identify core genes that are co-regulated with TRPM4 and functionally relevant in HCC, we initially intersected 2,019 TRPM4-coexpressed genes with 4,212 DEGs from the TCGA-LIHC dataset using a Venn diagram (Figure 1A). This intersection yielded 294 overlapping genes, which constituted the core gene set. This focused set served as the foundation for subsequent functional and prognostic analyses.

Figure 1 Screening, functional annotation, and prognostic gene selection of TRPM4-associated core genes. (A) Venn diagram: intersection of TRPM4-coexpressed genes (n=2,019) and TCGA-LIHC DEGs (n=4,212) yielded 294 core genes. (B,C) GO enrichment (BP/CC/MF) and KEGG enrichment analysis. (D) Univariate Cox forest plot. (E,F) LASSO coefficient and LASSO deviance curve, respectively. BP, biological process; CC, cellular component; DEGs, differentially expressed genes; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; LASSO, least absolute shrinkage and selection operator; MF, molecular function; TCGA-LIHC, The Cancer Genome Atlas-Liver Hepatocellular Carcinoma.

GO enrichment analysis (Figure 1B) demonstrated a highly significant enrichment of these core genes (P<0.002, indicated by a color gradient) within BP such as bicarbonate transport and hypotonic response, which are linked to the ion channel function of TRPM4. In terms of CC, these genes were localized to the basal/apical plasma membrane, aligning with their roles in membrane-associated regulation. Furthermore, in the MF category, the genes were enriched in cadherin binding, suggesting their involvement in cell adhesion and intercellular communication.

KEGG pathway analysis (Figure 1C) revealed that the core genes were significantly enriched in pathways characterized by a high GeneRatio and strong statistical significance (P<0.002). These pathways include Human papillomavirus infection, recognized as an oncogenic pathway associated with HCC, the MAPK signaling pathway, which drives HCC proliferation, and ECM-receptor interaction, which is critical for HCC invasion. Collectively, these findings implicate the TRPM4 co-expression network in key malignant processes associated with HCC.

Univariate Cox regression analysis (Figure 1D) identified 87 core genes that were significantly associated with overall survival (OS) (P<0.05). Among these, 85 genes exhibited hazard ratios (HRs) greater than 1, indicating their role as prognostic risk factors, with many demonstrating P values less than 0.001. Conversely, two genes, GP6 and MYOM2, were associated with HRs less than 1, suggesting their function as protective factors. Subsequently, LASSO-Cox regression was employed to enhance model parsimony by utilizing coefficient paths (Figure 1E) and partial likelihood deviance curves (Figure 1F), resulting in the selection of 11 key genes: BRSK1, MMP1, GRIN2D, GP6, MYOM2, N4BP3, CCDC112, TSEN54, MAP3K9, SPP1, and B3GNT4. Each selected gene was retained based on statistical penalization in LASSO regression and biological relevance to HCC progression, stromal remodeling, or immune regulation, minimizing overfitting risks. The expression patterns of these 11 signature genes were further validated using TCGA-LIHC data, comparing normal and tumor tissues (Figure S2). All genes exhibited significantly elevated expression levels in tumor tissues compared with adjacent normal tissues (Figure S2A-S2K), corroborating their initial classification as DEGs within the cohort. The risk score formula was derived as follows: Risk Score = (0.005×BRSK1) + (0.080×MMP1) + (0.011×GRIN2D) – (0.092×GP6) – (0.022×MYOM2) + (0.020×N4BP3) + (0.037×CCDC112) + (0.031×TSEN54) + (0.049×MAP3K9) + (0.0434×SPP1) + (0.004×B3GNT4). Notably, the directionality of all coefficients was consistent with the results obtained from the univariate Cox analysis, thereby affirming the reliability of the prognostic effects associated with the selected genes.

Prognostic performance of the TRPM4-associated risk model (Figure 2)

Figure 2 Prognostic performance of the TRPM4-associated risk model and nomogram validation. (A) Forest plot of univariate Cox regression: associations of risk score and clinical stage with OS. (B) Forest plot of multivariate Cox regression: independent prognostic value of the risk score after adjusting for clinical stage (HR =5.419, P<0.001). (C) Risk score distribution: patient stratification by median risk score (1.682, dashed line). (D) Heatmap: expression patterns of signature genes in high- and low-risk subgroups (red = upregulated; green = downregulated). (E) Kaplan-Meier survival curve: OS differences between high- and low-risk subgroups (log-rank P<0.001). (F) Kaplan-Meier survival analysis in this external dataset (GEO cohort). (G) Time-dependent ROC curves: 1-, 3-, and 5-year predictive performance of the risk model (AUC =0.779, 0.693, 0.701). (H) Comparative ROC analysis: AUC comparison of the risk model with clinical variables (age, gender, grade) and clinical stage. (I) Nomogram: integrated model of risk score and clinical stage for predicting 1-, 3-, and 5-year OS probabilities. (J) C-index analysis: concordance index comparison between the risk model and clinical variables. (K) Calibration curves: agreement between predicted and observed 1-, 3-, and 5-year OS rates (dashed line = ideal calibration). AUC, area under the curve; GEO, Gene Expression Omnibus; HR, hazard ratio; OS, overall survival; ROC, receiver operating characteristic.

Univariate Cox regression analysis (Figure 2A) identified the risk score and clinical stage as significant predictors of OS, with the risk score exhibiting a HR of 6.509 (P<0.001) and clinical stage an HR of 1.680 (P<0.001). Multivariate Cox analysis (Figure 2B) further corroborated the risk score as a robust independent prognostic factor (HR =5.419, P<0.001), even after adjusting for clinical stage. This finding underscores the prognostic value of the risk score, independent of the TNM staging system, which is the clinical gold standard.

Upon establishing the independent prognostic significance of the risk score, we stratified patients from the TCGA-LIHC cohort into high- and low-risk subgroups based on the median risk score of 1.682 (Figure 2C). This threshold effectively differentiated patients according to their risk score distribution, with high-risk individuals clustering above the dashed line and low-risk individuals below it. The accompanying heatmap (Figure 2D) illustrates the expression pattern of the signature: genes associated with high risk, such as BRSK1 and MMP1, were predominantly upregulated, indicated by intense red coloration in the high-risk group. Conversely, protective genes like GP6 and MYOM2 were downregulated, shown in deep green, thereby creating a distinct expression profile that was aligned with the risk status.

In alignment with the risk stratification, Kaplan-Meier survival analysis (Figure 2E) revealed significant OS differences between two subgroups. The high-risk subgroup had significantly shorter median OS, with a steeply declining survival curve, while the low-risk subgroup exhibited a plateaued survival curve (log-rank P<0.001), highlighting the pronounced divergence in outcomes between the two groups. To confirm the model’s generalizability beyond the TCGA cohort, we applied the same 11-gene risk score formula to an independent cohort from the GEO database. Kaplan-Meier survival analysis in this external dataset (Figure 2F) similarly demonstrated that high-risk patients had significantly shorter OS than low-risk patients (log-rank P=0.02), thereby replicating the prognostic stratification pattern observed in the TCGA-LIHC cohort and supporting the model’s robust and reproducible performance across independent patient populations.

In addition to survival stratification, we assessed the model’s predictive accuracy. Time-dependent ROC curves (Figure 2G) were employed to assess the model’s predictive performance, yielding area under the curve (AUC) values of 0.779, 0.693, and 0.701 for 1-, 3-, and 5-year predictions, respectively. Comparative ROC analysis (Figure 2H) showed the risk score achieved a higher AUC than age (0.531/0.512/0.520), gender (0.509/0.508/0.511), and tumor grade (0.499/0.524/0.516) at 1-, 3-, and 5-year time points, and was comparable to clinical stage (0.671/0.670/0.668). C-index analysis (Figure 2J) further indicated that the risk score consistently surpassed non-stage clinical variables in C-index throughout the follow-up period. To validate the model’s robustness across various clinical subgroups, patients were stratified based on key clinical characteristics, including gender, age, tumor grade, and clinical stage, followed by subgroup survival analysis (Figure S2). In all subgroups—female/male (Figure S3A), age ≤60/>60 years (Figure S3B), tumor grade G1–2/G3–4 (Figure S3C), and clinical stage I–II/III–IV (Figure S3D)—patients classified as high-risk consistently exhibited significantly shorter OS than low-risk patients (all log-rank P<0.001). These results confirm the risk model’s ability to stratify prognosis regardless of baseline clinical traits, reinforcing its broad clinical utility.

To transform the risk model into a quantifiable clinical instrument, we developed a nomogram (Figure 2I) that integrates the TRPM4-associated risk score and clinical stage, both of which were previously identified as independent prognostic factors. The nomogram allocates weighted point values to each factor, with higher risk scores corresponding to greater point allocations. By summing these points, clinicians can directly estimate a patient’s 1-, 3-, and 5-year OS probabilities. Calibration curves (Figure 2K) were employed to validate the predictive accuracy of the nomogram. The 1-year OS predictions demonstrated a close alignment with the observed survival rates, as evidenced by the calibration curve nearly overlapping the 45° ideal line. Meanwhile, the 3- and 5-year predicted probabilities deviated by less than 5% from the actual outcomes. These findings underscore the nomogram’s capability for precise, individualized prognostic stratification, highlighting its potential utility in clinical decision-making for patients with HCC.

Immune microenvironment and TMB analysis

Subsequently, we investigated the relationship between the TRPM4-associated risk model and the immune microenvironment of HCC, as well as TMB. Utilizing CIBERSORT to quantify 22 immune cell subtypes, we established correlations between signature genes/risk scores and immune cell infiltration, which are illustrated in a heatmap (Figure 3A). Risk-associated genes, such as MMP1 and BRSK1, exhibited positive correlations with immunosuppressive subsets, including regulatory T cells and M0 macrophages. Conversely, protective genes, such as GP6 and MYOM2, were associated with anti-tumor effector cells, such as memory B cells. The risk score demonstrated a strong positive correlation with M0 macrophages, a key immunosuppressive cell type, and negative correlations with naive B cells and memory B cells.

Figure 3 Characteristics of the immune microenvironment and synergistic prognostic stratification based on TMB and risk. (A) Correlation heatmap: displays the associations between signature genes/risk score and immune cell infiltration, with color indicating correlation strength and statistical significance at P<0.001. (B) Immune cell composition: illustrates the differences in immune cell proportions between high- and low-risk subgroups. (C) TMB distribution: compares TMB between high- and low-risk subgroups, with a P value of 0.17. (D) TMB survival curve: depicts overall survival differences between high- and low-TMB subgroups, with a log-rank P value of 0.015. (E) Combined TMB-risk survival curve: demonstrates the synergistic prognostic stratification of TMB and risk score, with a log-rank P value of less than 0.001. TMB, tumor mutational burden.

Further analysis of the involvement of the 11 signature genes in the regulation of the tumor microenvironment was conducted by evaluating their associations with key tumor microenvironment scores, such as the immune score and stromal score (Figure S4). Each gene was correlated with distinct microenvironmental characteristics; for example, high expression of MMP1 was associated with lower immune scores, whereas elevated expression of SPP1 corresponded to higher stromal scores. These results confirm the signature genes modulate multiple functional dimensions of the HCC tumor microenvironment.

In accordance with these correlations, the composition of immune cells exhibited significant differences between the high- and low-risk subgroups (Figure 3B). Specifically, the high-risk group showed a significantly higher proportion of M0 macrophages (P<0.001), neutrophils, and an elevated M2/M1 macrophage ratio (P<0.001), indicating impaired anti-tumor macrophage polarization and enhanced immune suppression (Figure 3B). The low-risk group was characterized by abundant naive B cells, memory B cells, and resting CD4⁺ memory T cells. These observations confirm that the risk model reflects an immunosuppressive microenvironment in high-risk patients, and potentially contributing to their adverse prognosis. To further elucidate the molecular connections between the signature genes and immune regulation, we examined their correlations with immune-related molecules, such as checkpoints, receptors, and mediators (Figure S5). Each gene appears to drive distinct yet synergistic immunosuppressive processes that collectively influence the microenvironment of high-risk HCC. Specifically, MMP1, linked to TNFSF members, and MAP3K9, associated with TRAF6/NF-κB, enhance chronic inflammatory signals that induce immune tolerance. Meanwhile, GP6, negatively correlated with CD co-stimulatory receptors, and SPP1, linked to PDCD1/CD163, impair T cell activity and promote the polarization of immunosuppressive macrophages; Additionally, BRSK1 (associated with CXCR4) and CCDC112 (linked to ITGB1) recruit suppressive immune cells and fortify stromal barriers to obstruct effector cell infiltration. Furthermore, MYOM2 (associated with TGFBR1) and GRIN2D (linked to NLGN1) exacerbate immune exclusion through stromal activation and neuro-immune crosstalk. N4BP3, TSEN54, and B3GNT4 further compromise innate immunity, immune cell proliferation, and antigen presentation, respectively. These gene-specific immune molecular networks converge to establish an immunosuppressive-exclusionary microenvironment phenotype that is associated with poor prognosis in high-risk patients.

We subsequently assessed TMB in relation to risk status, finding that the high-risk group exhibited a marginally elevated TMB compared to (median TMB: high-risk vs. low-risk =5.2 vs. 4.1 mutations/Mb, P=0.17; Figure 3C). Stratification of patients based on TMB revealed that individuals in the high-TMB category experienced significantly reduced OS compared to those in the low-TMB category (log-rank P=0.015, Figure 3D). Further analysis of the correlation between each of the 11 signature genes and TMB (Figure S6) identified significant negative correlations for GP6 (R=−0.22, P=2.2e−05), GRIN2D (R=−0.15, P=0.0034), MYOM2 (R=−0.3, P=5.3e−09), and N4BP3 (R=−0.15, P=0.0051), whereas the remaining genes did not exhibit statistically significant associations with TMB. Notably, the integration of TMB with risk stratification resulted in improved prognostic stratification (Figure 3E): patients in the high-TMB/high-risk subgroup had the worst OS (median OS =15.3 months), while those in the low-TMB/low-risk subgroup had the most favorable outcomes (median OS =68.7 months); the intermediate subgroups (high-TMB/low-risk, low-TMB/high-risk) showed distinct survival gradients. This synergistic effect (log-rank P<0.001) confirms TMB and the risk model together refine prognostic stratification of HCC patients.

Cell subtype-specific expression patterns of signature genes revealed by scRNA-seq

We conducted scRNA-seq on both control and treatment cohorts to elucidate the cellular specificity of the TRPM4-associated 11-gene signature. Utilizing UMAP clustering (Figure 4, top-left) we identified distinct cell subtypes, including natural killer (NK) cells, CD4⁺ T cells, macrophages, and B cells, which exhibited divergent distribution patterns between the two groups. Differential expression of the 11 signature genes was observed across specific cell subtypes (Figure 4). For BRSK1, expression in the control group was predominantly localized to CD4⁺ memory T cell clusters, as indicated by localized red signals in the control UMAP plot. In contrast, the treatment group exhibited an expanded expression of BRSK1 to activated CD4⁺ T cell subsets, accompanied by a moderate increase in intensity. MMP1 expression was negligible across all control cell clusters; however, it was robustly induced, as indicated by intense red signals, in a distinct cancer-associated fibroblasts (CAFs) cluster within the treatment group. GRIN2D expression in the control group was faint and restricted to a small neuronal-like cell subset, whereas in the treatment group, it was upregulated in CAFs clusters, with signals extending across a broader cell population. GP6: in the control group, expression was predominantly localized to vascular endothelial cell clusters, as indicated by bright red signals. In contrast, the treatment group exhibited a near-complete silencing of expression, with only sparse and faint signals observed. MYOM2: expression in the control group was confined to resting CD4⁺ T cell subsets. However, in the treatment group, expression expanded to activated B cell clusters and maintained a sustained intensity. N4BP3: Expression in the control group was restricted to resting NK cell clusters, whereas in the treatment group, it shifted to activated NK cell populations, characterized by expanded and intensified signals. CCDC112: the control group displayed faint and scattered expression across small cell subsets. In the treatment group, expression was significantly upregulated and concentrated in macrophage clusters, as evidenced by dense red signals. TSEN54: in the control group, expression was limited to vascular endothelial clusters, whereas in the treatment group, it was dispersed across T cells and macrophages, exhibiting moderate intensity. MAP3K9: expression in the control group was concentrated within monocyte subsets, while in the treatment group, it extended across multiple immune cell clusters, displaying heterogeneous levels. SPP1: in the control group, expression was robust and localized to macrophage clusters (intense red); however, in the treatment group, expression was markedly suppressed, with only faint and scattered signals. B3GNT4: expression in the control group was low and confined to small B cell subsets, but in the treatment group, it remained low yet expanded to include additional B cell populations. These alterations in cell subtype-specific expression across all 11 signature genes suggest that the TRPM4-associated signature may modulate the functional status of distinct cell populations (e.g., immune cell activation, stromal cell remodeling) and thereby mediate HCC progression and therapeutic response. Single-cell risk score analysis confirmed that CAFs and macrophages exhibited the highest risk scores among all cell types, identifying these two compartments as the main cellular sources driving the bulk-tissue prognostic signature. These cell-type-specific expression changes directly shape the tumor microenvironment and determine the overall risk score, establishing a mechanistic connection between single-cell observations and the prognostic model. Owing to the lack of matched spatial transcriptome data, the spatial distribution of CAFs and macrophages in high-risk tumor regions was not validated in this study, this will be explored in future prospective research.

Figure 4 The cell subtype-specific expression of the TRPM4-associated 11-gene signature using single-cell RNA sequencing. The top-left panel presents a UMAP plot depicting cell clusters, which are color-coded by subtype, including NK cells, CD4⁺ T cells, macrophages, among others, across both control and treatment groups. The subsequent panels display UMAP plots that represent the expression levels of each signature gene, with a color gradient indicating expression intensity (red denotes high expression), in the control group (left) and treatment group (right). NK, natural killer; UMAP, Uniform Manifold Approximation and Projection.

Discussion

HCC is characterized by significant heterogeneity, resulting in limited prognostic accuracy and suboptimal responses to immunotherapy (19). Traditional clinicopathological staging systems are inadequate in capturing the molecular, immune, and cellular determinants that influence patient outcomes (20). The NECSO pathway, a novel programmed cell death mechanism associated with ion homeostasis and tumor progression (6-9), remains poorly understood in terms of its regulatory network and clinical significance in HCC. TRPM4 acts as a key mediator of sodium influx and NECSO execution, regulating cancer cell death, invasion, and immune microenvironment remodeling (10-12). However, its downstream network and clinical role in HCC remain unclear (6,8). In this study, we developed and validated an 11-gene prognostic signature derived from the TRPM4 co-expression network, linking a functional cell death program to HCC prognosis, immune status, and single-cell expression patterns. Our findings not only provide a robust independent prognostic classifier but also uncover a previously unrecognized TRPM4-mediated axis involving immune, stromal, and mutational components that influences immune exclusion, microenvironmental remodeling, and clinical outcomes in HCC.

This study introduces a significant advancement through the development of a biologically anchored and rigorously validated prognostic signature, which surpasses traditional clinical variables in predictive accuracy. Unlike agnostic transcriptomic signatures, our model is based on TRPM4-dependent NECSO, thereby linking statistical prediction to a functionally pertinent cell death program (10). This signature is a robust independent prognostic predictor with superior or comparable performance to traditional clinical variables. The 1-, 3-, and 5-year AUC values were higher than age, gender, and grade, and comparable to clinical stage. The model was constructed using LASSO-Cox regression to reduce overfitting and validated in an independent GEO cohort, ensuring stability and generalizability. Each of the 11 genes was selected based on statistical robustness and biological relevance to HCC, including stromal activation, immune regulation, and oncogenic signaling, justifying their inclusion in the model. Importantly, the stratification capability of this signature remained consistent across various clinical subgroups, suggesting that it captures molecular risk information that is distinct from traditional clinicopathological factors. When incorporated with clinical staging into a prognostic nomogram, this model facilitated precise and individualized survival predictions with high calibration accuracy, underscoring its direct translational potential for clinical risk stratification.

Our analysis of the immune microenvironment indicates that this signature serves as a molecular proxy for an immunosuppressive and pro-metastatic TME. High-risk scores were significantly correlated with increased infiltration of M0 macrophages, a characteristic indicative of innate immune dysfunction and tumor-promoting inflammation in HCC (21), and an elevated M2/M1 macrophage ratio, both of which indicate an immunosuppressive tumor microenvironment. M0 macrophage enrichment reflects impaired polarization toward anti-tumor M1 subsets and a predisposition toward tumor-promoting M2 macrophages, which together mediate immune suppression and contribute to poor prognosis. Conversely, low-risk tumors exhibited abundant infiltration of naive B cells, memory B cells, and resting CD4⁺ memory T cells, cellular subsets closely associated with effective anti-tumor immune activity (22). At the molecular level, the signature genes orchestrate distinct yet synergistic immunosuppressive pathways: MMP1 and SPP1 facilitate extracellular matrix remodeling, stromal activation, and T-cell exclusion (23,24). BRSK1 and CCDC112 are involved in the recruitment of suppressive immune subsets and the reinforcement of stromal barriers. GP6 and MYOM, identified as protective factors, are associated with enhanced anti-tumor immunity and a reduced mutational burden. Furthermore, MAP3K9, GRIN2D, TSEN54, N4BP3, and B3GNT4 play roles in modulating inflammatory signaling, antigen presentation, and innate immune function. This coordinated immune-stromal interaction suggests that the signature not only reflects tumor proliferation but also quantifies the functional immune state of the TME, which is a crucial factor in therapeutic resistance and metastatic progression.

The integration of TMB further enhances clinical utility by revealing a synergistic prognostic interaction between mutational burden and immune risk. Although patients classified as high-risk exhibited slightly elevated TMB, the combined stratification approach delineated four distinct survival categories with markedly divergent clinical outcomes, thereby demonstrating that this dual stratification system surpasses the prognostic capability of either marker used independently. These findings address the well-documented limitation of TMB as an inconsistent standalone prognostic indicator in HCC (24). Notably, the genes GP6, GRIN2D, MYOM2, and N4BP3 exhibited significant negative correlations with TMB, identifying them as novel molecular links between the mutational landscape and immune regulation. Critically, this dual stratification approach may enhance the selection of patients for immunotherapy: patients classified as high-risk with high TMB may require combination strategies to overcome immune exclusion, whereas low-risk patients might benefit from checkpoint inhibitor monotherapy.

The scRNA-seq component overcomes a key limitation of bulk transcriptomics by clarifying cell-type-specific expression and dynamic remodeling of the signature across control and treatment conditions (16). Our data demonstrate that the 11 signature genes exhibit distinct, context-dependent expression patterns across immune, stromal, and endothelial compartments, with prominent regulation observed in CAFs, macrophages, and T-cell subsets. Single-cell analysis further revealed that MMP1 is specifically induced in CAFs while SPP1 is downregulated in macrophages, and single-cell risk score quantification confirmed that CAFs and macrophages represent the primary cellular drivers of the bulk-tissue prognostic signature. This cell-type-specific programming indicates that the TRPM4 network acts heterogeneously across cellular compartments to orchestrate context-dependent remodeling of the TME. Together, these findings elevate the signature from a purely statistical predictor to a biologically interpretable framework for dissecting HCC progression. However, this study retains the inherent limitations of retrospective bioinformatic analyses: all findings are based on public datasets and require prospective clinical validation, the causal roles of TRPM4 and its co-expressed signature genes in NECSO execution, HCC progression, and immune regulation remain to be verified by functional experiments, and the protective mechanisms of GP6 and MYOM2—especially their involvement in ion homeostasis, mutagenesis suppression, and anti-tumor immunity—merit further mechanistic exploration. Future scRNA-seq studies using paired pre- and post-treatment samples will be critical to delineate dynamic cellular remodeling during therapeutic intervention.

Conclusions

In conclusion, this study identifies the TRPM4-NECSO axis as a pivotal regulator of HCC prognosis and immune microenvironment remodeling. We establish and validate a TRPM4-centered 11-gene signature that is biologically anchored in a novel programmed cell death pathway (10), exhibits robust prognostic performance independent of clinicopathological variables, acts as an informative marker for immune suppression and exclusion, reflects cell-type-specific functions at the single-cell level, and provides biologically interpretable insights for risk stratification. By linking NECSO to immune regulation, stromal remodeling, and TMB, our work redefines the molecular landscape of HCC and supports the future development of NECSO-targeted therapeutic strategies (25). Since this study is based on public transcriptomic and genomic datasets without in vitro or in vivo functional validation, future investigations will employ TRPM4 overexpression/knockdown, sodium flux measurement, and NECSO detection to verify the causal roles of TRPM4 and the signature genes in HCC progression. Importantly, this 11-gene signature serves as a potential tool for HCC prognostic stratification, prospective clinical validation is mandatory prior to real-world clinical application. Ultimately, this signature may advance precision oncology by improving risk evaluation and guiding individualized immunotherapeutic strategies for patients with HCC.

Acknowledgments

The authors would like to thank the researchers who provided open access to the raw data in the TCGA database and GEO database.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0363/rc

Peer Review File: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0363/prf

Funding: This work was supported by the Guangxi Natural Science Foundation (No. 2025GXNSFBA069041), the Basic Research Capabilities of Middle-aged and Young Teachers in Guangxi’s Colleges and Universities (No. 2025KY0119), and the Youth Science Foundation of Guangxi Medical University (No. GXMUYSF202539).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0363/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. This 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. Siegel RL, Kratzer TB, Giaquinto AN, et al. Cancer statistics, 2025. CA Cancer J Clin 2025;75:10-45. [Crossref] [PubMed]
2. Singh SP, Madke T, Chand P. Global Epidemiology of Hepatocellular Carcinoma. J Clin Exp Hepatol 2025;15:102446. [Crossref] [PubMed]
3. Moris D, Martinino A, Schiltz S, et al. Advances in the treatment of hepatocellular carcinoma: An overview of the current and evolving therapeutic landscape for clinicians. CA Cancer J Clin 2025;75:498-527. [Crossref] [PubMed]
4. Ji K, Zhu H, Zhang C, et al. Nomogram-based prognostic stratification for patients with large hepatocellular carcinoma: a population study of SEER database and a Chinese cohort. J Gastrointest Oncol 2024;15:2201-15. [Crossref] [PubMed]
5. Faria SC, Szklaruk J, Kaseb AO, et al. TNM/Okuda/Barcelona/UNOS/CLIP International Multidisciplinary Classification of Hepatocellular Carcinoma: concepts, perspectives, and radiologic implications. Abdom Imaging 2014;39:1070-87. [Crossref] [PubMed]
6. Zhu Y, Guo Y, Su Y, et al. Necrosis by sodium overload-associated genes TRPM4 and SLC9A1: biological roles and clinical implications in breast cancer progression. Front Immunol 2025;16:1623511. [Crossref] [PubMed]
7. Shi Y, Qu X, Liu D, et al. Integration of single-cell and bulk transcriptomics to explore prognostic and immunotherapeutic characteristics of sodium overload-induced cell death in lung adenocarcinoma. Transl Cancer Res 2025;14:6330-47. [Crossref] [PubMed]
8. Dai Y, Wang H, Wang W, et al. Multidimensional pan cancer analysis of the sodium induced cell death gene TRPM4. Sci Rep 2025;15:29768. [Crossref] [PubMed]
9. Qu X, Zhang Y, Shi Y, et al. Construction of a prognostic model for hepatocellular carcinoma based on necrosis by sodium overload-related genes and identification of ANKRD13B as a new prognostic marker. Funct Integr Genomics 2025;25:192. [Crossref] [PubMed]
10. 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]
11. Guo J, She J, Zeng W, et al. Structures of the calcium-activated, non-selective cation channel TRPM4. Nature 2017;552:205-9. [Crossref] [PubMed]
12. Qiao Y, Wang J, Wang B, et al. Sodium disrupts mitochondrial energy metabolism to execute NECSO. Nat Commun 2025;17:493. [Crossref] [PubMed]
13. Hutchison WJ, Keyes TJ, et al. The tidyomics ecosystem: enhancing omic data analyses. Nat Methods 2024;21:1166-70. [Crossref] [PubMed]
14. Kanehisa M, Furumichi M, Sato Y, et al. KEGG: biological systems database as a model of the real world. Nucleic Acids Res 2025;53:D672-7. [Crossref] [PubMed]
15. Liu H, Zhang Y, Dong Y, et al. Construction of a prognostic model for disulfidptosis related ferroptosis genes lung adenocarcinoma and the role of DECR1 in lung adenocarcinoma. Front Immunol 2025;16:1658897. [Crossref] [PubMed]
16. Shen KY, Chen BY, Gao WC. Single-Cell RNA Sequencing Reveals the Role of Epithelial Cell Marker Genes in Predicting the Prognosis of Colorectal Cancer Patients. Dis Markers 2022;2022:8347125. [Crossref] [PubMed]
17. Newman AM, Steen CB, Liu CL, et al. Determining cell type abundance and expression from bulk tissues with digital cytometry. Nat Biotechnol 2019;37:773-82. [Crossref] [PubMed]
18. Yan P, Sun H, Wang S, et al. Molecular Subtypes and Risk Prediction Model Based on Malignant Cell Differentiation Trajectories in Breast Cancer. J Cell Mol Med 2025;29:e70680. [Crossref] [PubMed]
19. Fu L, Li S, Mei J, et al. BIRC2 blockade facilitates immunotherapy of hepatocellular carcinoma. Mol Cancer 2025;24:113. [Crossref] [PubMed]
20. Zhang Z, Li Y, Quan Z, et al. Constructing immune and prognostic features associated with ADCP in hepatocellular carcinoma and pan-cancer based on scRNA-seq and bulk RNA-seq. Front Immunol 2024;15:1397541. [Crossref] [PubMed]
21. Jabbarzadeh Kaboli P, Roozitalab G, Farghadani R, et al. c-MET and the immunological landscape of cancer: novel therapeutic strategies for enhanced anti-tumor immunity. Front Immunol 2024;15:1498391. [Crossref] [PubMed]
22. Pandey A, Linxweiler M, Kuo F, et al. Patterns of immune equilibrium and escape in indolent and progressing tumors. Cancer Cell 2023;41:1389-91. [Crossref] [PubMed]
23. Sugita S, Matsuo Y, Ishida M, et al. Kaempferol Inhibits MMP-1-Mediated Migration and Invasion in Gemcitabine-Resistant Pancreatic Cancer Cells. Nutrients 2026;18:380. [Crossref] [PubMed]
24. Wei S, Tan J, Huang X, et al. Metastasis and basement membrane-related signature enhances hepatocellular carcinoma prognosis and diagnosis by integrating single-cell RNA sequencing analysis and immune microenvironment assessment. J Transl Med 2024;22:711. [Crossref] [PubMed]
25. Yin Z, Li G, Zhong Q, et al. Cinobufagin overcomes bortezomib resistance in multiple myeloma strains by targeting SEC62/TRPM4-mediated NECSO. Phytomedicine 2025;147:157171. [Crossref] [PubMed]