In silico development and validation of a novel six-gene-derived signature in hepatocellular carcinoma

Introduction

Hepatocellular carcinoma (HCC), a primary malignancy arising from the liver parenchyma, ranks as one of the most prevalent and lethal forms of cancer worldwide (1). Its global burden is fueled by various risk factors, including chronic hepatitis B and C viral infections, excessive alcohol consumption, non-alcoholic fatty liver disease, and exposure to aflatoxin B1, among others (2). HCC typically presents at advanced stages due to its insidious onset and lack of specific early symptoms, which contributes to the poor prognosis associated with this disease (3). Although there have been significant advancements in surgical techniques, locoregional therapies, and systemic treatments, the overall survival (OS) for HCC patients continues to be alarmingly low, especially in cases of advanced or recurrent tumors (4). Hence, there is a critical need for novel approaches to improve early diagnosis, refine risk stratification, and tailor personalized treatment strategies for HCC patients, ultimately improving their clinical outcomes.

Pyruvate metabolism, which is a key process in cellular energy production and biosynthesis, plays a crucial role in the metabolic reprogramming of cancer cells (5). This metabolic shift, often referred to as the Warburg effect, involves increased glycolysis even in the presence of oxygen, leading to the production of lactate instead of complete oxidative phosphorylation (6). In HCC, aberrant pyruvate metabolism not only supplies the rapidly proliferating tumor cells with necessary precursors and adenosine triphosphate (ATP) but also fosters an acidic microenvironment that promotes invasion and metastasis (7). Moreover, it has been increasingly recognized that targeting pyruvate metabolism can unveil novel therapeutic vulnerabilities in cancer (8). Key enzymes and transporters involved in pyruvate metabolism, such as pyruvate dehydrogenase (PDH), lactate dehydrogenase (LDH), and monocarboxylate transporters (MCTs), have emerged as promising targets for pharmacological intervention (9,10). Recent research has also implicated dysregulation of pyruvate metabolism in modulating immune responses within the tumor microenvironment (11), further underscoring its multifaceted impact on cancer progression and treatment response.

Considering the essential role of pyruvate metabolism in HCC biology and its potential as a therapeutic target, this study aimed to exploit transcriptomic data from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases to identify molecular subtypes and prognostic signatures based on pyruvate metabolism-related genes (PMRGs). By uncovering the relationship between pyruvate metabolism dysregulation and HCC outcomes, we anticipate providing valuable insights into the disease heterogeneity, underlying pathogenic mechanisms, and novel therapeutic avenues for this challenging malignancy. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2024-2621/rc).

Methods

Data retrieval and acquisition of PMRG

Transcriptomic data and corresponding clinical information for TCGA-liver hepatocellular carcinoma (LIHC) samples were obtained from the University of California, Santa Cruz (UCSC) Xena platform (https://xenabrowser.net/). The GSE14520 cohort was sourced from the GEO database (https://www.ncbi.nlm.nih.gov/geo/). Fragments per kilobase million (FPKM) data were obtained and further log₂ transformed before using the data for model training and testing. The specific clinical and pathological characteristics of the two cohorts are shown in Table 1. A list of 40 PMRGs was extracted from the Kyoto Encyclopedia of Genes and Genomes (KEGG)_PYRUVATE_METABOLISM gene set within the Molecular Signatures Database (MSigDB, https://www.gsea-msigdb.org/gsea/msigdb/index.jsp); these genes are detailed in Table S1. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Identification and validation of a prognostic PMRG-related signature

Prognostic PMRGs were initially identified through univariate Cox regression analysis (P value <0.05) in the TCGA-LIHC training cohort. Subsequently, consensus clustering was conducted using the ConsensusClusterPlus R package (12), which employs a resampling-based algorithm to determine stable clusters of patients with distinct PMRG expression patterns. To optimize feature selection and mitigate overfitting, a least absolute shrinkage and selection operator (LASSO) Cox regression model was implemented using the glmnet R package with 10-fold cross-validation. The optimal penalty parameter (λ) was determined at the value yielding minimum partial likelihood deviance. The final prognostic signature was formulated as: riskscore = Σ(βi × Expr i), where βi represents the LASSO-derived coefficient for gene i, and log₂(FPKM +1) denotes the normalized expression level of each gene. This calculation method ensures feature stability by assigning non-zero coefficients only to genes with robust prognostic contributions. Risk scores were computed for all patients, with median values used to stratify cohorts into high- and low-risk groups in both training (TCGA-LIHC) and validation (GSE14520) datasets. The differences in OS between groups were evaluated using the Kaplan-Meier (KM) curves. The predictive accuracy of the signature was assessed suing the receiver operating characteristic (ROC) curves, and principal component analysis (PCA) were further conducted to visualize the distribution of clusters.

Immune landscape assessment and mutation analysis

The CIBERSORT algorithm implemented in the IOBR package (13) was employed to quantify the infiltration levels of immune cells. Wilcox test was adopted to assess the differences between groups, and the correlation between PMRGs expression and immune infiltration was evaluated using Pearson correlation analysis. Somatic mutation analysis and visualization were conducted by the maftools package.

Drug sensitivity analysis

The pRRophetic package (14) was utilized to estimate the sensitivity of TCGA-LIHC samples to 45 drugs. Wilcoxon tests were carried out to compare drug sensitivity differences between groups, and the correlation between PMRG expression and drug sensitivity was investigated using Pearson’s correlation coefficient.

Gene set enrichment analysis (GSEA) and gene set variation analysis (GSVA)

Limma package was used to perform the differential expression analysis, with those satisfying adjusted P<0.05 and |log₂ (fold change)| >1 identified as differentially expressed genes. GSEA was conducted using the clusterProfiler package (15) to identify enriched Gene Ontology (GO) annotations and KEGG pathways. GSVA was applied to the KEGG_PYRUVATE_METABOLISM gene set using the GSVA package, with GSVA scores between groups was compared.

Construction and evaluation of a nomogram

The multivariate Cox regression analysis was performed to identify independent prognostic factors of HCC, and those variables with P<0.05 were incorporated into a nomogram to predict 1-, 3-, and 5-year OS probabilities. Calibration curves were plotted to examine the agreement between nomogram-predicted probabilities and observed survival rates. The predictive accuracy of the nomogram for 1-, 3-, and 5-year OS was evaluated using ROC curves. Additionally, the net benefit of the nomogram compared to other prognostic factors was assessed using decision curve analysis (DCA).

Statistical analysis

Statistical analyses and data visualization were performed using R statistical software (version 4.4.1). Group comparisons were analyzed using non-parametric Wilcoxon rank-sum tests, whereas linear associations between variables were examined through Pearson’s correlation coefficients. A two-tailed P value <0.05 considered statistically significant.

Results

Prognostic PMRGs identify two distinct HCC molecular subtypes

Of the 40 PMRGs, six met the criteria for differential expression, with three upregulated and three downregulated in HCC (Figure 1A). Univariate Cox regression analysis identified 20 PMRGs with prognostic relevance, among which ten were associated with favorable prognosis in HCC, while the remaining ten were associated with unfavorable prognosis (Figure 1B). The top three most frequently mutated PMRGs were ACACB, ACACA, and ACAT2 (Figure 1C). Based on the prognostic PMRGs, we classified HCC patients into two molecular subtypes, designated C1 and C2 (Figure 1D,1E). KM analysis demonstrated significantly better survival for subtype C2 compared to C1 (Figure 1F). PCA based on these genes effectively discriminated between C1 and C2 (Figure 1G).

Figure 1 PMRGs impact HCC prognosis. (A) Volcano diagram depicting expression of PMRGs in HCC. “NOT” means not dysregulation. (B) Distribution of hazard ratios for PMRGs associated with prognosis. (C) Somatic mutation landscape of PMRGs. (D,E) Consensus clustering of HCC based on prognostic PMRGs. (F) Kaplan-Meier survival curves comparing different molecular subtypes. (G) Principal component analysis of prognostic PMRGs. C1 and C2 represent the cluster 1 and cluster 2. CDF, cumulative distribution function; HCC, hepatocellular carcinoma; PC, principal component; PMRGs, pyruvate metabolism-related genes; TMB, tumor mutation burden.

A novel PMRG-related prognostic signature indicates HCC prognosis

LASSO was employed to refine the feature set, resulting in the selection of six PMRGs for constructing an HCC risk signature (Figure 2A,2B). The prognostic signature was generated: risk score = 0.096 × ACACA − 0.281 × ACAT1 + 0.318 × ACYP1 + 0.098 × DLAT + 0.304 × LDHA + 0.098 × ME1 (Figure 2C). TCGA-LIHC cohort were then divided into the high-risk (n=159) and low-risk (n=159) groups (Figure 2D). Patients in the high-risk group exhibited significantly shorter OS than their low-risk counterparts (Figure 2E, P<0.001). The results further demonstrated that the predictive accuracy for the 1-, 3-, and 5-year OS prediction using our signature were 0.786, 0.743, and 0.744, respectively (Figure 2F). PCA of the risk genes within the TCGA-LIHC cohort is depicted in Figure 2G. To further verify the predictive performance of the PMRG-related signature, risk scores were computed for the GSE14520 cohort and subjects were classified into high-risk (n=111) and low-risk (n=110) groups (Figure 2H). Survival analysis confirmed worse prognosis for high-risk HCC patients in the GSE14520 cohort compared to low-risk patients (Figure 2I, P=0.009). In GSE14520, the predictive accuracy for predicting 1-, 3-, and 5-year OS were observed to be 0.644, 0.636, and 0.665, respectively (Figure 2J). PCA of the risk genes within the GSE14520 cohort is presented in Figure 2K.

Figure 2 Construction and evaluation of an HCC risk signature derived from PMRGs. (A,B) LASSO Cox regression analysis. (C) Coefficients of genes contributing to the risk signature. (D-G) Prognostic assessment of the risk signature in the TCGA-LIHC cohort. (H-K) The performance of the risk signature in the GSE14520 cohort. AUC, area under the curve; HCC, hepatocellular carcinoma; LASSO, least absolute shrinkage and selection operator; LIHC, liver hepatocellular carcinoma; OS, overall survival; PC, principal component; PMRGs, pyruvate metabolism-related genes; TCGA, The Cancer Genome Atlas.

The PMRG-derived risk signature is associated with clinical pathological characteristics

We assessed differences in riskscores across various clinical pathological subgroups. Dead patients had higher riskscores than alive patients (Figure 3A), whereas no significant difference in riskscores was observed between elderly and younger patients (Figure 3B). Moreover, HCC patients with advanced pathological T stage had higher riskscores than those with early pathological T stage (Figure 3C), females displayed higher riskscores compared to males (Figure 3D), advanced- clinical stage patients had higher riskscores than early-clinical stage patients (Figure 3E). Comparison of different tumor grades revealed that patients with well-differentiated HCC had significantly higher risk scores compared to those with poorly differentiated tumors (Figure 3F).

Figure 3 Correlations between the PMRG-derived riskscore and clinical pathological features. Comparison of riskscores across different subgroups defined by (A) survival status, (B) age, (C) T stage, (D) gender, (E) stage, and (F) grade. ns, not significant; *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001. PMRGs, pyruvate metabolism-related genes.

Somatic mutation profiles between different risk groups

As shown is Figure 4A, the top three genes with higher frequent mutations in the high-risk HCC patients were TP53, CTNNB1, and TTN, whereas in the low-risk HCC patients, they were CTNNB1, TTN, and MUC16 (Figure 4B). No significant difference in tumor mutation burden (TMB) was observed between the high- and low-risk groups (Figure 4C). KM survival analysis revealed significant differences in prognosis among patients stratified by risk group and TMB (Figure 4D).

Figure 4 Somatic mutation characteristics associated with PMRG-derived signature. Somatic mutation profiles of (A) high- and (B) low-risk HCC patients. (C) Comparison of TMB between different risk groups. (D) Kaplan-Meier survival curves of patients with different risk scores and TMB. ns, not significant. HCC, hepatocellular carcinoma; PMRG, pyruvate metabolism-related gene; TMB, tumor mutation burden.

Differential gene expression pattern between different risk groups

GSVA demonstrated that the low-risk HCC patients exhibited a higher GSVA scores of the PMRG gene set as compared those with high-risk score (Figure 5A). Additionally, GSEA demonstrated significant activation of the PMRG gene set in the high-risk HCC patients (Figure 5B). In contrast, the low-risk group showed significant activation of acid catabolic processes and chromatid segregation (Figure 5C). KEGG pathway enrichment analysis demonstrated increased activation of drug metabolism pathways and complement and coagulation cascades, as well as suppression of IL-17 signaling pathway and cell cycle-related pathways in the low-risk HCC patients as compared to those with high-risk scores (Figure 5D).

Figure 5 PMRG-derived signature indicated gene expression patterns. (A) Comparison of PMRG GSVA scores between different risk groups. *, P<0.05. (B) Gene set enrichment analysis of PMRGs. (C) GO gene set enrichment analysis. (D) KEGG gene set enrichment analysis. GO, Gene Ontology; GSVA, gene set variation analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; PMRG, pyruvate metabolism-related gene.

PMRG-derived signature correlated with immune landscape

Immune infiltration analysis indicated that the HCC patients with high-risk scores had elevated levels of M0 macrophages, neutrophils, activated CD4⁺ T cells, and Treg cells, along with reduced infiltration of M2 macrophages, resting mast cells, resting CD4⁺ T cells, and CD8⁺ T cells compared to those with low-risk scores (Figure 6A). All five PMRGs contributing to the riskscore were significantly associated with M0 macrophage infiltration (Figure 6B). Furthermore, the high-risk patients displayed a higher StromalScore, while ImmuneScore and tumor purity remained similar between groups (Figure 6C-6F). Notably, the high-risk HCC showed an elevated tumor immune dysfunction and exclusion (TIDE) score than the low-risk HCC (Figure 6G), suggesting potential immune evasion mechanisms such as T-cell dysfunction or stromal exclusion. Paradoxically, true responders to immunotherapy exhibited significantly higher riskscores than false responders (Figure 6H), indicating that a higher riskscore is linked to enhanced immunotherapeutic response. This apparent contradiction implies that the PMRG-derived signature may reflect dual metabolic-immune interactions: while pyruvate metabolism dysregulation could promote immunosuppressive microenvironments, specific PMRG-driven processes might simultaneously enhance immunogenicity and therapeutic vulnerability.

Figure 6 PMRG-derived model indicated immune landscape and immunotherapy response. (A) Immune cell infiltration comparison between different risk groups. (B) Pearson correlation analysis between immune cell infiltration and PMRG expression. Comparison of (C) StromalScore, (D) ImmuneScore, (E) ESTIMATEScore, (F) tumor purity, and (G) TIDE score between different risk groups. (H) Riskscore comparison between true and false responders to immunotherapy. ns, not significant; *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001. PMRG, pyruvate metabolism-related gene; TIDE, tumor immune dysfunction and exclusion.

PMRG-derived signature correlated with drug sensitivity

As shown in Figure 7, significant differences in sensitivity to 34 drugs were observed between different risk groups, with the high-risk HCC displaying increased sensitivity to 23 drugs and decreased sensitivity to the remaining 11 drugs. Moreover, significant correlations were observed between the six PMRG genes comprising the risk signature and drug sensitivity.

Figure 7 Relationship between drug sensitivity, PMRGs, and risk signature. ns, not significant; *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001. PMRGs, pyruvate metabolism-related genes.

Nomogram of PMRG-derived signature

Univariate Cox regression analysis was performed to investigate whether the risk score and clinicopathological characteristics served as prognostic predictors. As a result, the risk score [hazard ratio (HR) =3.7, 95% confidence interval (CI): 2.7–5.2], T stage (HR =1.8, 95% CI: 1.4–2.1), and clinical stage (HR =1.8, 95% CI: 1.4–2.2) were found to be significantly associated with OS in HCC patients (Figure 8A). Next, PMRGs-derived risk score (HR =3.26, 95% CI: 2.24–4.73) was identified as the only independent risk factor for OS in HCC (Figure 8B). Consequently, we built a nomogram using PMRGs-derived signature to accurately predict the survival rates in HCC patients, with higher total scores indicating poorer survival (Figure 8C). The calibration curve revealed strong concordance between nomogram-predicted and observed survival rates across all timepoints, with minimal deviation from the ideal 45° line (Figure 8D). DCA demonstrated superior clinical net benefit of the nomogram compared to risk score or staging systems alone across threshold probabilities of 10–60%, justifying its clinical utility (Figure 8E). Finally, survival ROC analysis confirmed sustained predictive accuracy with aera under the curve (AUC) values of 0.761, 0.728, and 0.715 for 1-, 3-, and 5-year OS, respectively, outperforming single clinical variables (Figure 8F).

Figure 8 Generation and assessment of a novel nomogram using the PMRG-related risk signature. (A) Univariate and (B) multivariate Cox regression analyses evaluated the prognostic characteristics. (C) Nomogram for predicting OS in HCC patients. (D) Calibration chart assessing agreement between nomogram-predicted and actual OS rate. (E) Decision curve comparing the predictive performance of the nomogram with other clinicopathological factors. (F) ROC curves assessing the prognostic efficacy of the nomogram. AUC, aera under the curve; CI, confidence interval; HCC, hepatocellular carcinoma; HR, hazard ratio; OS, overall survival; PMRG, pyruvate metabolism-related gene; ROC, receiver operating characteristic.

Discussion

Pyruvate, as a pivotal intermediate in glycolysis, serves not only as an indicator of dysregulated energy metabolism in tumor cells but is also intimately associated with the malignancy and clinical outcomes of tumors. Moreover, the redirection of mitochondrial pyruvate metabolism has been shown to facilitate cancer progression (16,17). Alterations in pyruvate metabolism not only confer adaptive metabolic advantages to tumor cells, supporting their rapid proliferation and invasive capabilities, but may also serve as potential biomarkers for tumor grading and prognosis assessment. This study successfully employed PMRG analysis to identify clinically meaningful molecular subtypes in HCC, leading to the development of an innovative prognostic model with significant potential for personalized survival prediction and therapeutic response evaluation in HCC patients.

Pyruvate metabolism represents an emerging and therapeutically targetable metabolic vulnerability in cancer biology, particularly holding substantial pathological relevance in HCC (18,19). PMRGs play crucial roles in governing core metabolic processes such as energy generation, biosynthesis, and redox homeostasis, while concurrently being intricately linked to tumorigenesis and progression. In liver cancer, upregulated ACACA expression is correlated with tumor advancement and poor prognosis (20). ACAT1, primarily responsible for cholesterol esterification, may indirectly influence tumor progression via modulation of lipid storage, signaling, and cellular membrane fluidity (21). Notably, ACAT1 inhibitors have demonstrated the ability to suppress cancer cell proliferation and migration, suggesting its potential as a therapeutic target (22). The CYP1 family, encompassing members like CYP1A1, CYP1A2, and CYP1B1, predominantly participates in the oxidative metabolism of drugs, environmental pollutants, and endogenous substances such as steroid hormones. Dysregulated expression of CYP1 family members in liver cancer tissues is implicated in disease progression, unfavorable prognosis, and chemotherapeutic drug responsiveness (23). DLAT, a constituent subunit of the PDH complex, is subject to suppression in tumor cells (e.g., through upregulation of PDH kinase), promoting the conversion of pyruvate to lactate and fueling the Warburg effect (24,25). Changes in DLAT expression or activity can directly impact this metabolic shift, thereby influencing the energy metabolic state and proliferation rate of tumors. LDHA catalyzes the reduction of pyruvate to lactate, functioning as a central enzyme in anaerobic glycolysis (the Warburg effect). It is frequently overexpressed in liver cancer, driving lactate production that supplies tumors with rapid energy, sustains an acidic tumor microenvironment conducive to invasion and metastasis, and impairs immune cell function (26). In liver cancer, elevated ME1 expression may facilitate energy supply and redox balance under glucose-limiting conditions, thereby promoting tumor growth (27).

Pyruvate metabolism is involved in shaping the tumor immune microenvironment, encompassing the regulation of chromatin remodeling during CD4⁺ T cell activation (28) and the infiltration of tumor-associated macrophage 2 (TAM2) populations (29). In the present study, PMRGs also exhibit associations with tumor immunology, as the immune microenvironment of patients with higher risk scores displays a more pronounced inclination towards immunosuppression and inflammation. ACACA encodes fatty acid synthase (FASN); elevated FASN expression may lead to secretion of inflammatory mediators such as prostaglandin E2, which promotes recruitment and function of immunosuppressive cells like myeloid-derived suppressor cells and Tregs, thereby dampening antitumor immunity (30). Members of the CYP1 family generate metabolites of aromatic hydrocarbons within tumor cells, such as those derived from 2,3,7,8-tetrachlorodibenzo-p-dioxin (31). These metabolites induce expression of various immunosuppressive molecules, including indoleamine 2,3-dioxygenase and programmed death-ligand 1, facilitating immune evasion (32). Moreover, CYP1 enzyme activity may influence intratumoral levels of hormones and lipid metabolites that modulate immune cell differentiation and function (33). High LDHA expression not only sustains the Warburg effect in tumor cells but may also affect tumor immunity by generating lactate (34).

Our findings reveal significant correlations between the PMRG-derived signature and HCC patients’ sensitivity to multiple chemotherapeutic agents, suggesting that the expression pattern of PMRGs may determine tumor responsiveness to specific chemotherapeutics and offer potential biomarkers for individualized chemotherapy regimens. While serum metabolomics provides direct insights into circulating metabolites (e.g., lactate, pyruvate), its utility in HCC biomarker discovery is constrained by systemic variability from non-hepatic tissues and transient dietary influences (35). In contrast, RNA-seq of liver biopsies captures tumor-specific transcriptional reprogramming of PMRGs, enabling precise identification of molecular subtypes and prognostic signatures rooted in persistent oncogenic adaptations. For example, LDHA upregulation in HCC tissues (detected via RNA-seq) directly reflects Warburg effect activation (36) whereas serum lactate levels may fluctuate due to extratumoral factors (e.g., muscle activity, renal clearance) (37). Nevertheless, integrating transcriptomic and metabolomic data could enhance early HCC detection by linking tumor-intrinsic PMRG dysregulation to systemic metabolic perturbations (38).

Our analysis revealed distinct drug sensitivity patterns between risk groups. Sorafenib, a first-line tyrosine kinase inhibitor for advanced HCC, showed no differential sensitivity between high- and low-risk groups, likely because its mechanism of action (VEGFR/PDGFR inhibition) is independent of pyruvate metabolism pathways governed by our PMRG signature. In contrast, metformin exhibited high sensitivity across both groups, consistent with its known AMPK/mTOR-mediated metabolic effects that broadly suppress tumor growth regardless of pyruvate metabolic status (39). Notably, gemcitabine demonstrated significantly greater sensitivity in the low-risk group, potentially attributable to enhanced drug uptake in tumors with lower lactate export (40) or reduced nucleotide synthesis capacity (41), both features associated with favorable PMRG profiles. Additionally, in the identified signature, ACAT1 has been identified as a facilitator of enzalutamide resistance (42) and is implicated in the development of resistance to gemcitabine (43) and doxorubicin (44). CYP1 family members participate in drug metabolism, particularly in the oxidative biotransformation of many chemotherapeutic agents. Inhibition of CYP1 enzymes may play a crucial role in preventing cancer and overcoming chemoresistance to anticancer drugs (45). Additionally, DLAT has been implicated in mediating paclitaxel chemosensitivity in prostate cancer (46), and LDHA-dependent glycolytic reprogramming drives CDKN3-mediated cisplatin resistance in bladder cancer (47). Furthermore, high-risk groups exhibit higher TIDE scores, and genuine immune therapy responders display significantly higher risk scores than pseudo-responders, implying a link between the PMRG-derived risk score and enhanced responsiveness to immunotherapy.

Several limitations in our study should be acknowledged. First, the prognostic model derived from the analysis of public datasets necessitates validation in prospective, multicenter cohorts to assess its generalizability. Second, the functional roles of risk signature genes in HCC initiation and progression warrant further exploration. Lastly, experimental investigations into the fundamental mechanisms by which PMRGs regulate pyruvate metabolism are needed.

Conclusions

This study presents a novel risk signature for HCC rooted in genes associated with PMRGs, which demonstrates robust prognostic value. The signature integrates expression levels of six pivotal genes—ACACA, ACAT1, CYP1, DLAT, LDHA, and ME1—yielding a six-gene panel that significantly stratifies patient survival in both TCGA and GSE14520 cohorts into high- and low-risk groups, with marked differences in OS. Notably, this risk profile exhibits strong correlations with patients’ immune microenvironment, drug responsiveness, and immune therapy outcomes, thereby underscoring its potential to inform personalized chemotherapy and immunotherapy decision-making. In summary, this distinct and efficacious PMRG-based risk signature not only furnishes a powerful tool for HCC prognosis but also offers valuable insights to tailor treatment strategies for individual patients.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2024-2621/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2024-2621/prf

Funding: This work was supported by the Zhejiang Province Traditional Chinese Medicine Science and Technology Project (No. 2024ZL645).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2024-2621/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. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. Yang JD, Hainaut P, Gores GJ, et al. A global view of hepatocellular carcinoma: trends, risk, prevention and management. Nat Rev Gastroenterol Hepatol 2019;16:589-604. [Crossref] [PubMed]
3. Piñero F, Dirchwolf M, Pessôa MG. Biomarkers in Hepatocellular Carcinoma: Diagnosis, Prognosis and Treatment Response Assessment. Cells 2020;9:1370. [Crossref] [PubMed]
4. Harding-Theobald E, Louissaint J, Maraj B, et al. Systematic review: radiomics for the diagnosis and prognosis of hepatocellular carcinoma. Aliment Pharmacol Ther 2021;54:890-901. [Crossref] [PubMed]
5. Miao J, Chen B, Zhang L, et al. Metabolic expression profiling analysis reveals pyruvate-mediated EPHB2 upregulation promotes lymphatic metastasis in head and neck squamous cell carcinomas. J Transl Med 2025;23:316. [Crossref] [PubMed]
6. Vaupel P, Schmidberger H, Mayer A. The Warburg effect: essential part of metabolic reprogramming and central contributor to cancer progression. Int J Radiat Biol 2019;95:912-9. [Crossref] [PubMed]
7. Du D, Liu C, Qin M, et al. Metabolic dysregulation and emerging therapeutical targets for hepatocellular carcinoma. Acta Pharm Sin B 2022;12:558-80. [Crossref] [PubMed]
8. Rattigan KM, Brabcova Z, Sarnello D, et al. Pyruvate anaplerosis is a targetable vulnerability in persistent leukaemic stem cells. Nat Commun 2023;14:4634. [Crossref] [PubMed]
9. Feng Y, Xiong Y, Qiao T, et al. Lactate dehydrogenase A: A key player in carcinogenesis and potential target in cancer therapy. Cancer Med 2018;7:6124-36. [Crossref] [PubMed]
10. Payen VL, Mina E, Van Hée VF, et al. Monocarboxylate transporters in cancer. Mol Metab 2020;33:48-66. [Crossref] [PubMed]
11. Elia I, Rowe JH, Johnson S, et al. Tumor cells dictate anti-tumor immune responses by altering pyruvate utilization and succinate signaling in CD8(+) T cells. Cell Metab 2022;34:1137-1150.e6. [Crossref] [PubMed]
12. Wilkerson MD, Hayes DN. ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking. Bioinformatics 2010;26:1572-3. [Crossref] [PubMed]
13. Zeng D, Ye Z, Shen R, et al. IOBR: Multi-Omics Immuno-Oncology Biological Research to Decode Tumor Microenvironment and Signatures. Front Immunol 2021;12:687975. [Crossref] [PubMed]
14. Geeleher P, Cox N, Huang RS. pRRophetic: an R package for prediction of clinical chemotherapeutic response from tumor gene expression levels. PLoS One 2014;9:e107468. [Crossref] [PubMed]
15. Wu T, Hu E, Xu S, et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation (Camb) 2021;2:100141. [Crossref] [PubMed]
16. Cui J, Quan M, Xie D, et al. A novel KDM5A/MPC-1 signaling pathway promotes pancreatic cancer progression via redirecting mitochondrial pyruvate metabolism. Oncogene 2020;39:1140-51. [Crossref] [PubMed]
17. Ursprung S, Woitek R, McLean MA, et al. Hyperpolarized (13)C-Pyruvate Metabolism as a Surrogate for Tumor Grade and Poor Outcome in Renal Cell Carcinoma-A Proof of Principle Study. Cancers (Basel) 2022;14:335. [Crossref] [PubMed]
18. Alshamleh I, Kurrle N, Makowka P, et al. PDP1 is a key metabolic gatekeeper and modulator of drug resistance in FLT3-ITD-positive acute myeloid leukemia. Leukemia 2023;37:2367-82. [Crossref] [PubMed]
19. Yang C, Lee D, Zhang MS, et al. Genome-Wide CRISPR/Cas9 Library Screening Revealed Dietary Restriction of Glutamine in Combination with Inhibition of Pyruvate Metabolism as Effective Liver Cancer Treatment. Adv Sci (Weinh) 2022;9:e2202104. [Crossref] [PubMed]
20. Shen Y, Wang X, Ni Z, et al. Identification of acetyl-CoA carboxylase alpha as a prognostic and targeted candidate for hepatocellular carcinoma. Clin Transl Oncol 2023;25:2499-513. [Crossref] [PubMed]
21. Gu L, Zhu Y, Lin X, et al. Stabilization of FASN by ACAT1-mediated GNPAT acetylation promotes lipid metabolism and hepatocarcinogenesis. Oncogene 2020;39:2437-49. [Crossref] [PubMed]
22. Sun T, Xiao X. Targeting ACAT1 in cancer: from threat to treatment. Front Oncol 2024;14:1395192. [Crossref] [PubMed]
23. Krkoška M, Svobodová J, Kabátková M, et al. Deregulation of signaling pathways controlling cell survival and proliferation in cancer cells alters induction of cytochrome P450 family 1 enzymes. Toxicology 2021;461:152897. [Crossref] [PubMed]
24. Han A, Chua V, Baqai U, et al. Pyruvate dehydrogenase inactivation causes glycolytic phenotype in BAP1 mutant uveal melanoma. Oncogene 2022;41:1129-39. [Crossref] [PubMed]
25. Yang Q, Zeng S, Liu W. Roles of cuproptosis-related gene DLAT in various cancers: a bioinformatic analysis and preliminary verification on pro-survival autophagy. PeerJ 2023;11:e15019. [Crossref] [PubMed]
26. Wang X, Zhang P, Deng K. MYC Promotes LDHA Expression through MicroRNA-122-5p to Potentiate Glycolysis in Hepatocellular Carcinoma. Anal Cell Pathol (Amst) 2022;2022:1435173. [Crossref] [PubMed]
27. Fujiwara-Tani R, Nakashima C, Ohmori H, et al. Significance of Malic Enzyme 1 in Cancer: A Review. Curr Issues Mol Biol 2025;47:83. [Crossref] [PubMed]
28. Mocholi E, Russo L, Gopal K, et al. Pyruvate metabolism controls chromatin remodeling during CD4(+) T cell activation. Cell Rep 2023;42:112583. [Crossref] [PubMed]
29. Li X, Du Y, Jiang W, et al. Integrated transcriptomics, proteomics and metabolomics-based analysis uncover TAM2-associated glycolysis and pyruvate metabolic remodeling in pancreatic cancer. Front Immunol 2023;14:1170223. [Crossref] [PubMed]
30. Schroeder B, Vander Steen T, Espinoza I, et al. Fatty acid synthase (FASN) regulates the mitochondrial priming of cancer cells. Cell Death Dis 2021;12:977. [Crossref] [PubMed]
31. Kwon YJ, Shin S, Chun YJ. Biological roles of cytochrome P450 1A1, 1A2, and 1B1 enzymes. Arch Pharm Res 2021;44:63-83. [Crossref] [PubMed]
32. Kado SY, Bein K, Castaneda AR, et al. Regulation of IDO2 by the Aryl Hydrocarbon Receptor (AhR) in Breast Cancer. Cells 2023;12:1433. [Crossref] [PubMed]
33. Zhao M, Ma J, Li M, et al. Cytochrome P450 Enzymes and Drug Metabolism in Humans. Int J Mol Sci 2021;22:12808. [Crossref] [PubMed]
34. Osaka N, Sasaki AT. Beyond Warburg: LDHA activates RAC for tumour growth. Nat Metab 2022;4:1623-5. [Crossref] [PubMed]
35. Dalamaga M. Clinical metabolomics: Useful insights, perspectives and challenges. Metabol Open 2024;22:100290. [Crossref] [PubMed]
36. Huang J, Xie H, Li J, et al. Histone lactylation drives liver cancer metastasis by facilitating NSF1-mediated ferroptosis resistance after microwave ablation. Redox Biol 2025;81:103553. [Crossref] [PubMed]
37. Deulkar P, Singam A, Mudiganti VNKS, et al. Lactate Monitoring in Intensive Care: A Comprehensive Review of Its Utility and Interpretation. Cureus 2024;16:e66356. [Crossref] [PubMed]
38. Yang T, Wang SY. Integrating transcriptome and metabolomics analyses of hepatocellular carcinoma to discover novel biomarkers and drug targets. Clin Res Hepatol Gastroenterol 2025;49:102546. [Crossref] [PubMed]
39. Zamanian MY, Golmohammadi M, Yumashev A, et al. Effects of metformin on cancers in experimental and clinical studies: Focusing on autophagy and AMPK/mTOR signaling pathways. Cell Biochem Funct 2024;42:e4071. [Crossref] [PubMed]
40. Amrutkar M, Vethe NT, Verbeke CS, et al. Differential Gemcitabine Sensitivity in Primary Human Pancreatic Cancer Cells and Paired Stellate Cells Is Driven by Heterogenous Drug Uptake and Processing. Cancers (Basel) 2020;12:3628. [Crossref] [PubMed]
41. Mullen NJ, Singh PK. Nucleotide metabolism: a pan-cancer metabolic dependency. Nat Rev Cancer 2023;23:275-94. [Crossref] [PubMed]
42. Kohrt SE, Awadallah WN, Phillips RA 3rd, et al. Identification of Genes Required for Enzalutamide Resistance in Castration-Resistant Prostate Cancer Cells In Vitro. Mol Cancer Ther 2021;20:398-409. [Crossref] [PubMed]
43. Ueno G, Iwagami Y, Kobayashi S, et al. ACAT-1-Regulated Cholesteryl Ester Accumulation Modulates Gemcitabine Resistance in Biliary Tract Cancer. Ann Surg Oncol 2022;29:2899-909. [Crossref] [PubMed]
44. He W, Li Y, Liu SB, et al. From mitochondria to tumor suppression: ACAT1's crucial role in gastric cancer. Front Immunol 2024;15:1449525. [Crossref] [PubMed]
45. Su CY, Wu A, Dong Z, et al. Tumor stromal topography promotes chemoresistance in migrating breast cancer cell clusters. Biomaterials 2023;298:122128. [Crossref] [PubMed]
46. Wen H, Qu C, Wang Z, et al. Cuproptosis enhances docetaxel chemosensitivity by inhibiting autophagy via the DLAT/mTOR pathway in prostate cancer. FASEB J 2023;37:e23145. [Crossref] [PubMed]
47. Li M, Che N, Jin Y, et al. CDKN3 Overcomes Bladder Cancer Cisplatin Resistance via LDHA-Dependent Glycolysis Reprogramming. Onco Targets Ther 2022;15:299-311. [Crossref] [PubMed]