A prognostic signature based on insulin-signaling-pathway genes for hepatocellular carcinoma

Abstract

Background

Hepatocellular carcinoma (HCC) is the most common subtype of liver cancer. HCC occurrence, metastasis and therapeutic effect are closely related to tumor metabolic microenvironment. However, the role of insulin-related glucose metabolism in HCC has also not been extensively studied.

Method

Transcriptional profiles and clinical data of HCC samples were retrieved from The Cancer Genome Atlas (TCGA). The insulin signaling pathway related genes were derived from GeneCards and only the protein-coding genes with the top 100 relevance score were retained. Univariate Cox analysis, least absolute shrinkage and selection operator (LASSO) regression analysis and multivariate Cox analysis were conducted to develop the prognosis model. The predictive performance of our prognosis model was assessed by using receiver operating characteristic (ROC) curve, calibration curve and nomogram. Further studies, such as enrichment analysis, drug sensitivity and immunotherapy response were performed to assess the tumor microenvironment and treatment response. The clinical significance of SLC2A1 in HCC was validated with an independent cohort.

Results

We constructed a prognostic signature based on 4 insulin pathway related genes: RHEB, PRKAA2, SLC2A1 and FOXO1. HCC patients divided into high-risk and low-risk group according to the median risk score. We evaluated the signature in training set, validation set and entire set. The Kaplan–Meier (K–M) survival curve revealed that patients in low-risk group had longer survival. Even in different clinicopathological subgroups, the prognostic signature had good prognostic performance. We also identified that commonly used chemotherapy agents such as 5-fluorouracil, gemcitabine, paclitaxel, sorafenib and sunitinib showed significant sensitivity in the high-risk group. The TIDE algorithm suggested that patients in high-risk group may be more sensitive to immunotherapy. SLC2A1 was selected as the core gene, and the Kaplan-Meier survival curve showed that SLC2A1 positive was significantly associated with prognosis in a HCC independent cohort. Univariate and multivariate Cox analysis demonstrated that SLC2A1 was an independent risk variable for poor prognosis.

Conclusions

In summary, we constructed a prognostic signature based on insulin signaling pathway genes. The excellent performance and applicability of our model underscores its advantages and reliability as a clinical tool. Moreover, we validated SLC2A1 was an independent prognostic factor in a HCC independent cohort.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-025-03763-x.

Introduction

Liver cancer is one of the most common malignancies of the digestive system, ranking sixth among new cases and third among causes of cancer death worldwide in 2022 [1]. Hepatocellular carcinoma (HCC) is a subtype of liver cancer that accounts for about 75–85%, most of which are diagnosed at an advanced stage, the 5-year survival rate of advanced-stage HCC is only 2% [2, 3]. Advanced HCC has primary chemotherapy and radiation resistance, limiting the available treatment options for these patients [4]. The remarkable development of tyrosine kinase inhibitor (TKI) and immune checkpoint inhibitor (ICI) for HCC treatment in recent years has led to improvements in the prognosis of HCC, yet further improvement is still necessary [5, 6]. Therefore, we need to explore new therapeutic targets and treatment strategies. The survival of HCC was primarily predicted by classical clinicopathological features, including stage, age, income, and other biomarkers, but the heterogeneity of HCC prevents these indicators from achieving satisfactory results [7]. Recent studies have also identified imaging-based parameters as valuable prognostic factors; for example, Li et al. demonstrated that portal venous and hepatic arterial coefficients can predict post-hepatectomy overall and recurrence-free survival in HCC patients, highlighting the importance of integrating multiple factors for accurate prognosis prediction [8]. Previous studies have proved that the occurrence, metastasis and therapeutic effect of tumors were closely related to the tumor metabolic microenvironment [9–12]. In energy metabolism, glucose is converted into metabolites for energy storage and signaling molecule synthesis.

Glucose metabolic reprogramming is an important hallmark of cancer. Tumor cell metabolites affect their own proliferation and the activity of surrounding immune cells in various ways, promoting or inhibiting tumor growth [13]. Tumors exposed to high glucose levels had reduced the expression of glutamate-cysteine ligase catalytic subunit, a key component of glutathione biosynthesis, which in turn increased chemotherapy-induced oxidative anti-tumor damage [14]. Pancreatic ductal adenocarcinoma patients with normal blood glucose received chemotherapy experienced worse survival than those with hyperglycemia [15]. However, some studies have come to the opposite conclusions. Extracellular vesicles secreted by cancer cells inhibit insulin secretion, disrupt systemic glucose homeostasis, and promote tumor growth [16]. Type 2 diabetes (T2D) and insulin resistance are both associate with higher breast cancer incidence and poor survival [17–19]. Thus, systemic glucose homeostasis may play a variety of roles in different types of tumors. Growing evidence suggest that insulin related reprogramming of glucose metabolism contributes to HCC progression [20–22]. Consequently, interference with glucose metabolism may be a therapeutic target for HCC [23, 24]. Therefore, elucidating the role of the insulin signaling pathway in HCC is helpful to better understand the occurrence of tumor, guide treatment and improve the prognosis.

In recent years, interfering the glucose metabolism has attracted attention as a promising cancer treatment strategy, but has not been implemented in clinical practice. The role of insulin signaling pathway in HCC has also not been extensively studied. Therefore, we conducted an in-depth study on the significance of insulin signaling pathway genes in HCC. In this study, we constructed a prognostic signature based on 4 prognostic insulin signaling pathway genes in HCC. Subsequently, we evaluated the predictive ability of the prognostic signature for survival. A “hub gene”, SLC2A1 was selected for subsequent experiments. SLC2A1 is a glucose transporter that mediates basal glucose uptake and has potential as a tumor therapeutic target [25]. The clinical significance of SLC2A1 in HCC was evaluated in an independent cohort. The bioinformatics viewpoint from this study will be valuable for fully understand the role of insulin signaling pathway in HCC, and provide a theoretical basis for effective treatment.

Materials and methods

Data acquisition

Transcriptional profiles and clinical data of HCC samples were retrieved from TCGA (https://portal.gdc.cancer.gov/) and processed for subsequent analysis. The clinical data included sex, age, TNM stage, tumor grade, survival time and survival status. The insulin signaling pathway genes were derived from GeneCards (https://www.genecards.org/), and only the protein coding genes with top 100 relevance score were retained (Supplementary Table S1).

Signature construction

Totally 365 patients were randomly divided into training set (n = 183) and validation set (n = 182) according to 1:1 ratio, and 1000 bootstrap repetitions were conducted to assess the stability of the model. Univariate Cox analysis was performed on the training set using the “survival” R package to screen out the significant prognostic related insulin signaling pathway genes. The “pheatmap” package was used to visualize the expression of prognostic genes in normal and tumor tissues. Based upon the training set, the prognostic related insulin signaling pathway genes were further analyzed through least absolute shrinkage and selection operator (LASSO) regression analysis to avoid overfitting. Multivariate Cox analysis was used to construct the prognosis model. The risk score of each patient was calculated as follows: risk score = ∑(expression of signature genes * corresponding coefficient).

Prognostic signature validation

Patients were divided into high-risk group and low-risk group according to the median risk score. The R packages ‘‘survival” and ‘‘survminer” were used for survival analysis. Furthermore, we performed a stratified analysis to assess the prognostic value of prognostic signature in different clinicopathological features. Univariate and multivariate Cox analysis of risk scores and clinicopathological features were performed to identify independent prognostic factors. The nomogram model was constructed by R package ‘‘rms”. Receiver operating characteristic (ROC) curve and calibration curve were performed to evaluate the consistency and accuracy of prognostic signature.

Drug sensitivity and immunotherapy response

To compare drug sensitivity in high-risk and low-risk group, we calculated the IC50 of the most commonly used chemotherapy drugs in the GDSC database via R package “pRRophetic”. The sensitivity of patients to chemotherapy drugs was evaluated based on the predicted IC50 values, and the statistically differences were tested using the Wilcox test. To predict the immunotherapy response in high-risk and low-risk group, we calculated tumor immune dysfunction and exclusion (TIDE) scores and visualized through a violin plot.

Identify differentially expressed genes and enrichment analysis

Differentially expressed genes (DEGs) in high-risk and low-risk group were assessed by “limma” R package. The thresholds of DEGs were as follows: the |log2 FC | ≥ 2 and false discovery rate (FDR) < 0.05. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis were performed to explore the biological functions and signaling pathway of DEGs in high-risk and low-risk group using R package of “clusterProfiler”. In KEGG and GO, p-value < 0.05 and false discovery rate (FDR) < 0.05 were considered statistically significant.

Tissue microarray (TMA) of HCC

The HCC tissue microarrays with 93 cases of tumor tissues (cat. no: HLivH180Su10) were provided by Shanghai Outdo Biotech Company (Shanghai, China). The clinical data was collected from medical records and pathology reports.

Immunohistochemical (IHC) staining

The expression level of SLC2A1 protein was analyzed by IHC staining. The tissue section microarrays were dewaxed and hydrated in gradient ethanol solutions, and then microwaved in Tris-EDTA buffer for antigen recovery. After incubated with 3% hydrogen peroxide and blocked with 5% animal serum, the section was treated with primary antibody anti-SLC2A1 (CST, 73015 S, 1:500 dilution) at 4 ℃ overnight. The section was treated with secondary antibody for 1 h the next day. Finally, the section was stained with DAB solution and hematoxylin. All samples were divided into SLC2A1 positive and negative.

Statistical analysis

All bioinformatics analysis were performed in R language. The association between SLC2A1 expression and clinicopathological features was estimated using the Chi-square test or Fisher’s exact test. The Kaplan-Meier method and log rank-test were used evaluate the impact of SLC2A1 expression on OS. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001; ns, p > 0.05 were displayed as statistically significant or no significant (ns). Statistical analysis was performed using IBM SPSS (version 22.0).

Results

Construction of the prognostic signature

The top 100 insulin pathway related protein coding genes were included in this study. Totally 365 HCC patients were randomly divided into training set (n = 183) and validation set (n = 182) according to 1:1 ratio. A total of 10 significant prognostic genes were screened by univariate Cox analysis of the training set (Fig. 1A). The heatmap visualized the expression of prognostic related genes in tumor tissues and normal tissues (Fig. 1B). The prognostic genes were further analyzed through LASSO regression analysis and multivariate Cox analysis to construct the prognosis model (Fig. 1C, D). Finally, an insulin pathway related gene signature was constructed to predict prognosis. The risk score was calculated according to the gene expression level and regression coefficient (Fig. 1E), and the following formula was established: risk score = (RHEB * 0.545568) + (PRKAA2 * 0.281553) + (SLC2A1 * 0.349998) - (FOXO1 * 0.418667).

Fig. 1.

Construction of a prognostic signature based on insulin signaling pathway genes. A 10 significant prognostic related insulin signaling pathway genes were screened by univariate Cox analysis. B Heatmap of the 10 prognostic genes expression in HCC tissues and normal tissues. C LASSO coefficient profiles for the 10 prognostic genes. D LASSO regression with tenfold cross-validation found 5 prognostic genes using the minimum λ. E The coefficient value of the 4 prognostic genes selected by the multivariate analysis

Evaluation and validation of prognostic signature

Patients including training set, validation set and entire set were divided into high-risk and low-risk group according to the median risk score. There was no significant difference in clinicopathological features between training set and validation set (Table 1, all p > 0.05). The Kaplan-Meier survival curve showed that the OS of patients in high-risk group was shorter than that in low-risk group, regardless of training set (Fig. 2A, p < 0.001), validation set (Fig. 2B, p < 0.001) and clinicopathological features (Fig. 2C, p < 0.001). The distribution of risk scores and survival outcomes between high-risk group and low-risk group were displayed in Fig. 2D–I. As the risk score increased, the more patients died. The prognostic signature expression patterns were displayed in Fig. 2J–L, RHEB, PRKAA2 and SLC2A1 were highly expressed in high-risk group, and FOXO1 was highly expressed in low-risk group.

Table 1.

Clinicopathological features in the validation set vs. training set

Clinicopathological features	Validation set (%)	Training set (%)	p
Age (years)			0.6178
≤65	116 (63.74)	111 (60.66)
> 65	66 (36.26)	72 (39.34)
Gender			0.5264
Male	126 (69.23)	120 (65.57)
Female	56 (30.77)	63 (34.43)
Grade			0.6048
G1	29 (15.93)	26 (14.21)
G2	92 (50.55)	83 (45.36)
G3	53 (29.12)	65 (35.52)
G4	6 (3.3)	6 (3.28)
Unknow	2 (1.1)	3 (1.64)
Stage			0.2773
I	91 (50)	79 (43.17)
II	42 (23.08)	42 (22.95)
III	35 (19.23)	48 (26.23)
IV	1 (0.55)	3 (1.64)
Unknow	13 (7.14)	11 (6.01)

Fig. 2.

Validation of the prognostic signature. A–C Kaplan-Meier analysis of high-risk and low-risk patients in training set, validation set and entire set. D–F The scatter plot of patients’ OS status in training set, validation set and entire set. G–I The ranked dot plot of risk score distribution in training set, validation set and entire set. J–L The heatmap of prognostic signature expression in training set, validation set and entire set

To further validate the value of this prognostic signature in subgroups with different clinicopathological features, we performed a stratified analysis. The patients were divided into different subgroups according to age (Fig. 3A, B), sex (Fig. 3C, D), grade (Fig. 3E, F) and stage (Fig. 3G, H). The results showed that the low-risk group had better OS in all subgroups, consistent with previous observations.

Fig. 3.

Stratified analysis of the prognostic signature. Kaplan-Meier analysis under different clinical subgroups: A, B age, C, D sex, E, F grade, G, H stage

Establishment of nomogram

Additionally, we performed univariate and multivariate Cox analysis for clinical features and risk score. In the univariate Cox analysis, we observed that age (HR: 1.010, 95% confidence interval (CI) 0.996–1.025, p = 0.174), gender (HR: 0.776, 95%CI 0.531–1.132, p = 0.188) and grade (HR: 1.133, 95%CI 0.881–1.457, p = 0.330) were not associated with survival, however, stage (HR: 1.680, 95%CI 1.369–2.062, p < 0.001) and risk score (HR: 1.340, 95%CI 1.209–1.484, p < 0.001) were significantly related to survival (Fig. 4A). In the subsequent multivariate Cox analysis, we observed that stage (HR: 1.558, 95%CI 1.255–1.933, p < 0.001) and risk score (HR: 1.273, 95%CI 1.136–1.425, p < 0.001) were independent prognostic factors (Fig. 4B). Based on the clinicopathological features and risk score, a nomogram was conducted to predict the 1-, 3-, and 5-year OS probabilities (Fig. 4C). The calibration curve showed a high degree of agreement between the predicted OS and actual OS (Fig. 4D). The AUC values of 1-year, 3-year and 5-year OS were 0.777, 0.698 and 0.672, respectively (Fig. 4E). Moreover, compared with age, gender, grade, and stage, risk score showed greater predictive power, with the largest area under the ROC curve (Fig. 4F).

Fig. 4.

Construction and validation of the risk score related nomogram. A, B Univariate Cox analysis and multivariate Cox analysis of risk score and clinicopathological features. C Construction of a nomogram predicting the 1-, 3-, and 5-year OS of HCC. D Calibration curve for the nomogram predicting the 1-year, 3-year, and 5-year OS of HCC. E The time-dependent ROC curves of the prognostic signature, AUC was evaluated at 1 year, 3 years and 5 years, respectively. F The time-dependent ROC curves of the nomogram, AUC was evaluated at 1 year

Enrichment analysis and immune infiltration analysis

To clarify the underlying molecular function of DEGs in high-risk and low-risk group, we performed a functional enrichment analysis. KEGG analysis showed that DEGs were mainly enriched in epidermis development, extracellular matrix organization, extracellular structure organization, external encapsulating structure organization, apical part of cell and other pathways (Fig. 5A, B). GO analysis showed that DEGs were mainly enriched in epidermis development, extracellular matrix organization, extracellular structure organization, external encapsulating structure organization and other pathways (Fig. 5C, D).

Fig. 5.

A, B GO functional enrichment analysis. C, D KEGG pathway enrichment analysis

Next, we used CIBERSORT algorithm to calculated the immune cell infiltration. We found that the low-risk group had significantly higher infiltration of B cell, CD8⁺T cell, neutrophil, NK cell, TIL cell, however, high-risk group had significantly higher infiltration of aDC, macrophages (Fig. 6A). Subsequently, in the comparison of immune function, we found that cytolytic activity, type I IFN response, type II IFN response were higher in the low-risk group, while MHC class I was significantly down-regulated (Fig. 6B). Finally, we compared the expression of common immune checkpoints between the two groups. The results showed that there was no difference in CD274, PDCD1 and PDCD1LG2 expression between the two groups, while the expression of CTLA4 (p < 0.001) in the high-risk group were significantly higher (Fig. 6C–F).

Fig. 6.

Association between the risk score and immune-related characteristics. A, B The scores of immune infiltrating cells and immune-related functions in high and low risk groups. C–F The expression level of immune checkpoint markers (CTLA4, PD-1, PD-L1, and PD-L2) in high and low risk groups

Comparisons of drug sensitivity

To further investigate the clinical application of risk model in the precision treatment of HCC, we evaluated the therapeutic efficacy of commonly used chemotherapy agents in different risk populations. According to the findings, patients in high-risk group were more sensitive to 5-fluorouracil (Fig. 7A), gemcitabine (Fig. 7B), paclitaxel (Fig. 7C), sorafenib (Fig. 7D) and sunitinib (Fig. 7E). In addition, we calculated the potential immunotherapy sensitivity of each patient based on the TIDE algorithm and found that patients in high-risk group may be more sensitive to immunotherapy (Fig. 7F).

Fig. 7.

The potential drug sensitivity and immunotherapy response in high-risk and low-risk groups. A 5-fluorouracil. B Gemcitabine. C Paclitaxel. D Sorafenib. E Sunitinib. F Immunotherapy response of TIDE algorithm. G Representative IHC staining of SLC2A1 in HCC tumor tissue. H The Kaplan-Meier analysis of SLC2A1 in a HCC independent cohort

Clinical significance of SLC2A1 in HCC

Among these four hub genes, we selected SLC2A1 to further investigate its clinical significance in HCC. To evaluate the expression level of SLC2A1 in HCC patients, we performed IHC staining of SLC2A1 on HCC tissue microarrays with 93 cases of tumor tissues (Fig. 7G). We subsequently analyzed the relationship between SLC2A1 expression and clinicopathological features (Table 2). SLC2A1 expression was positively related with tumor grade (p = 0.003), but not with age, gender, stage (all p > 0.05). Kaplan-Meier curve showed SLC2A1 positive was significantly associated with prognosis in patients with HCC (Fig. 7H, p = 0.0011). Furthermore, univariate and multivariate Cox analysis revealed that SLC2A1 positive was an independent risk variable for poor prognosis (Table 3).

Table 2.

Correlation of SLC2A1 expression with clinicopathological features in patients

Clinicopathological Features	SLC2A1 negative (%)	SLC2A1 positive (%)	p
Age (years)			0.447
≤ 55	32 (61.5)	20 (38.5)
> 55	25 (62.5)	15 (37.5)
Unknow	0 (0)	1 (100)
Gender			0.929
Male	51 (61.4)	32 (38.6)
Female	6 (60)	4 (40)
T stage			0.139
T1–2	31 (72.1)	12 (27.9)
T3–4	22 (52.4)	20 (47.6)
Unknow	4 (50)	4 (50)
N stage			0.685
N0	51 (61.4)	32 (38.6)
N1	1 (100)	0 (0)
Unknow	5 (55.6)	4 (44.4)
M stage			0.345
M0	53 (63.1)	31 (36.9)
M1	0 (0)	1 (100)
Unknow	4 (50)	4 (50)
Stage			0.139
Stage I–II	31 (72.1)	12 (27.9)
Stage III–IV	22 (52.4)	20 (47.6)
Unknow	4 (50)	4 (50)
Grade			0.003
G1–2	44 (72.1)	17 (27.9)
G3	13 (40.6)	19 (59.4)

Table 3.

Univariate Cox and multivariate Cox analysis of prognostic factors influencing survival

Clinicopathological features	Univariate HR (95% CI)	p	Multivariate HR (95% CI)	p
Age (years)
≤ 55	Reference
> 55	0.904 (0.539–1.516)	0.701
Unknow	–	–
Gender
Male	Reference
Female	1.453 (0.581–3.635)	0.424
Grade
G1–2	Reference
G3	1.258 (0.742–2.134)	0.394
Stage
Stage I–II	Reference		Reference
Stage III–IV	2.804 (1.579–4.976)	0.000	2.484 (1.388–4.447)	0.002
Unknow	3.234 (1.350–7.746)	0.008	2.914 (1.209–7.027)	0.017
SLC2A1 expression
Negative	Reference		Reference
Positive	2.284 (1.362–3.828)	0.002	1.932 (1.143–3.267)	0.014

HR hazard ratio, CI confidence interval

Discussion

Despite significant advancement in the diagnosis and treatment of HCC over the past decade, the survival of HCC remains poor. Therefore, there is an urgent need to find new strategies to guide the treatment of HCC. As an energy source for cell metabolism, glucose has attracted more and more attention in recent years. Preclinical studies suggest that interfering with glucose metabolism may be a strategy to improve the efficacy of HCC treatment [26, 27]. Inhibition of the glucose transporter by canagliflozin (a drug used for type-II diabetes) effectively synergized with sorafenib to treat HCC [27]. Until now, the specific role of insulin pathway genes in HCC remains unclear. In this context, the aim of this study was to construct a novel insulin pathway gene related prognostic signature for HCC patients.

In this study, we constructed a novel prognostic signature through LASSO regression analysis and multivariate Cox analysis, ultimately identifying four hub genes: RHEB, PRKAA2, SLC2A1, and FOXO1. Among them, RHEB, PRKAA2 and SLC2A1 were highly expressed in HCC, which were correlated with poor survival of HCC, however, FOXO1 was highly expressed in normal tissue, which was associated with good survival of HCC. RHEB is crucial activator of insulin-mediated cell signal transduction, and previous studies have shown that insulin elevates its GTP-binding state [28, 29]. RHEB promotes tumor proliferation and inhibits autophagy [30]. It is highly expressed in a variety of tumors, including colorectal cancer and HCC [31, 32]. PRKAA2 encodes the α2 catalyic subunit of AMPK, which is involved in the glucose metabolism [33]. In gastric cancer, PRKAA2 expression contributing to the activation of autophagy and chemoresistance [34]. In another study, PRKAA2 facilitate cell growth and glucose metabolic reprogramming [35]. FOXO1 is a critical regulator of hepatic and systemic insulin sensitivity, controling glucose homeostasis [36, 37]. In addition, it also regulates tumor cell cycle arrest, cell apoptosis, DNA damage repair and cell growth [38]. FOXO1 is down-regulated in non-small cell lung cancer and FOXO1 silencing correlates with cancer progression, while the opposite biological effect may exist in breast cancer [39, 40].

To construct a prognostic signature based on a risk score, we divided patients into high-risk and low-risk group according to the median risk score. We found that patients in high-risk group had shorter OS than those in low-risk group, regardless of training set, validation set and clinicopathological features. The ROC curve revealed that the prognostic signature has high accuracy in predicting the prognosis of HCC patients, with the AUC values of 1-year, 3-year and 5-year OS were 0.777, 0.698 and 0.672, respectively. More importantly, this signature still has significant prognostic value in patients with different clinical features. In the univariate Cox analysis and multivariate Cox analysis, we observed that stage and risk score were independent prognostic factors. Finally, we constructed a risk score related nomogram to predict HCC prognosis, the calibration curve showed a high degree of agreement between the predicted OS and actual OS. These results indicated that our our risk model has an excellent prediction performance. Specifically, the insulin pathway-based prognostic model can be integrated with routine clinical indicators (e.g., TNM stage, AFP level, and vascular invasion) to stratify HCC patients into high and low-risk groups in clinical practice. For high-risk patients identified by the model, more aggressive adjuvant therapies (e.g., postoperative targeted therapy combined with immunotherapy) are recommended to reduce recurrence risk; in contrast, low-risk patients may benefit from de-escalated treatment intensity to avoid unnecessary adverse events, balancing therapeutic efficacy and quality of life.

SLC2A1, known as glucose transporter 1, plays a crucial role in cell glycometabolism both in cancer cells and normal cells [41]. SLC2A1 is highly expressed in a variety of cancers, and SLC2A1 overexpression promotes the tumor growth and metastasis, such as lung cancer, breast cancer and gastric cancer [42–44]. In our study, we found that SLC2A1 was highly expressed in HCC and correlated with poor survival. Subsequently, we validated the clinical significance of SLC2A1 in a HCC independent cohort. Survival analysis showed that SLC2A1 positive was significantly associated with poor prognosis in patients with HCC. Univariate and multivariate Cox analysis revealed that SLC2A1 positive was an independent risk variable for prognosis. This makes SLC2A1 a potential target for HCC treatment. There are several ways to inhibit SLC2A1, such as using glucose transporter inhibitors, small molecules, anti-glucose transporter antibodies and siRNA. Inhibitors of SLC2A1 were used, compounds such as STF-31, phloretin, hydroxylated phenyl esters and others [25]. Unfortunately, most of these studies were only in preclinical studies and have not yet been tested in clinical trials. With the development of science, targeting SLC2A1 on the surface of tumor cells may become a strategy for anti-tumor therapy. Notably, SLC2A1, as a core independent risk factor, provides a potential target for personalized treatment. For patients with high SLC2A1 expression, metabolic intervention strategies (e.g., combination therapy with metformin or SLC2A1 inhibitors) could be considered to suppress tumor growth by regulating glucose metabolism. Additionally, the model can guide stratified follow-up protocols: high-risk patients may require more frequent monitoring to enable early detection of recurrence, while low-risk patients can be followed up every 6 months, reducing medical costs and patient burden. These applications directly assist healthcare professionals in optimizing treatment plans and improving patient management.

As advanced HCC has primary chemotherapy and radiation resistance [4], understanding the mechanism of drug resistance is of great significance to improve survival. Function analysis showed that the DEGs were mainly enriched in epidermis development, extracellular matrix organization, extracellular structure organization, external encapsulating structure organization and other pathways. The extracellular matrix ECM is an important component of the tumor microenvironment, which can alter the therapy response of tumor cells [45]. The drug distribution of most compounds in tumor tissues is known to occur primarily through diffusion [46]. In this regard, the concentrated extracellular matrix in tumors represents a barrier that limits intratumor drug transport and therapeutic efficacy [47]. Here, we further analyzed the relationship between the risk score and drug sensitivity, and identified commonly used antineoplastic agents as candidates for HCC patients. Specifically, fluorouracil, paclitaxel, gemcitabine, sorafenib, and Sunitinib showed better therapeutic effects in the high-risk group.

To summarize, our study described a novel prognostic signature based on insulin signaling pathway genes in HCC. However, some limitations are inevitable in our research. Firstly, this study was based on TCGA database, and more multicenter prospective clinical studies are needed to verify the accuracy of our prognostic signature. Second, some clinical features were missing, such as AFP level, which may affect the predictive power of our model. Thirdly, we validated the clinical significance of SLC2A1 with an independent cohort, but without in vitro and in vivo experiments.

Conclusion

In summary, we constructed a prognostic signature based on insulin signaling pathway genes using bioinformatic analysis. The excellent performance and applicability of our prognostic signature underscores its advantages and reliability as a clinical tool. Once clinically validated, it may help identify high-risk patients to enhance decision-making and identify low-risk patients to avoid overtreatment. Moreover, we validated SLC2A1 was an independent prognostic factor in HCC with an independent cohort. Overall, this study provides new insights into the prognosis and treatment of HCC.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

YY. Z. Conceptualization, methodology, writing-original draft. HQ. S. methodology, formal analysis, writing-original draft. WS. C. writing-review & editing, software, visualization, data curation. KW. P. writing-review & editing, validation, project administration, funding acquisition. All authors reviewed the manuscript.

Funding

This work was supported by grants from Guangzhou Science and Technology Foundation (2023A04J0591) and Plan on enhancing scientific research in GMU (2024).

Data availability

The datasets analyzed in this study were obtained from the publicly available TCGA database (https://portal.gdc.cancer.gov/) and GeneCards (https://www.genecards.org/). For additional details or access to the complete datasets, interested readers are encouraged to contact the corresponding author.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the ethical standards of the Institutional Review Board and Ethics Committee of The Second Affiliated Hospital, Guangzhou Medical University. The use of the hepatocellular carcinoma (HCC) tissue microarray was approved by the Ethics Committee of Shanghai Outdo Biotech Company (Ethical Number: SHYJS-CP-1804007). All participants provided informed consent prior to participation in the research. All authors consent to participate.

Consent for publication

All patients signed informed consent forms and agreed to provide clinical data for research purposes. Signed formal consents are available to ensure that the publisher has the authors’ permission to publish the research findings.

Competing interests

The authors declare no competing interests.