Construction of an eight-basement membrane related gene signature for predicting prognosis and immune response in hepatocellular carcinoma

Abstract

Background

The basement membrane plays a very vital role in impeding cancer progression and metastatic colonization. However, the relationship between basement membrane-related genes (BMRGs) and hepatocellular carcinoma (HCC) remains poorly understood.

Methods

The transcriptome and clinical data of HCC patients were gathered from the Cancer Genome Atlas (TCGA) and the International Cancer Genome Consortium (ICGC) database, and segmented into training and testing sets, respectively. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses enrichment subsequently analyzed the differentially expressed BMRGs. A novel prognostic BMRGs signature model was constructed by LASSO (least absolute shrinkage and selection operator) in training set and stratified patients into high-risk and low-risk groups, and which was verified in the testing set. Moreover, the accuracy of the new BMRGs signature was detected by receiver operating characteristic curve (ROC) analysis, followed by clinical correlation analysis, immune function analysis and multi-drug resistance analysis to verify the accuracy and clinical practical application value of the new signature.

Results

We created and verified a novel prognostic BMRGs signature for the prognosis of patients with HCC. The ROC analysis demonstrated that the 1-, 3-, and 5-year survival rates of HCC patients predicted by the prognostic BMRGs signature model based on the BMRGs signature were consistent with those of real patients. In clinical correlation analysis, univariate and multivariate Cox regression analyses validated that the model supports that the prognostic BMRGs signature can be independent risk factors for overall survival (OS) of HCC patients. In addition, the new signature can indeed differentiate the immunological analysis of HCC patients at different risk groups. In the drug sensitivity analysis, the expression levels of the BMRGs in the signatures were found to be related to the sensitivity of common chemotherapy drugs in HCC patients.

Conclusion

The newly prognostic BMRGs signature can be used as a prognostic indicator to predict the potential progression trajectory and therapeutic response in HCC, which may also provide sensible recommendations for immunotherapy and the selection of chemotherapeutic agents. Our findings provided a promising insight into BMRGs in HCC and a personalized prediction tool for prognosis and immune responses in patients.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-025-04050-5.

Introduction

Liver cancer is the second leading cause of cancer-related deaths worldwide, with about 850,000 new cases each year. Hepatocellular carcinoma (HCC) is one of the most common liver malignancies, accounting for approximately 90% of all cases of primary liver cancer [1, 2]. HCC leads to more than 700,000 deaths annually [3]. Currently, the treatment strategies for HCC are still very limited. Notably, understanding the molecular drivers of HCC is essential to develop novel biomarkers, improve the treatment and predict the prognosis of HCC [4, 5].

Basement membrane (BM) is layered cell-adherent extracellular matrices (ECMs), which exists in almost all tissues and is widely conserved across multicellular organisms [6]. As a special type of extracellular matrix, the BM is important for shaping tissues to normal size and maintaining functions by regulating cells to perceive biochemical signals [7]. BM is mainly composed of laminin, type IV collagen, nidogens and various proteoglycans. Among them, laminin is the main component of BM, and type IV collagen is the framework of BM [8]. Accordingly, BM is classified into epithelial BM and endothelial BM [9], in which endothelial BM wrap around cancer cells and prevent them from invading other normal tissues, while endothelial BM prevent the intravasation or extravasation of metastatic cancer cells into blood vessels and the lymphatic system [10]. Therefore, as the structural barrier of cancer cell invasion, intravasation and extravasation [9], BM plays a highly crucial role in the pathological mechanism of cancer, and are of great significance for the risk prediction and prognosis of cancer patients.

The invasion of cancer cells is mainly caused by the degradation of channel protease, which destroys the structure of BM. It is of great significance to maintain the integrity of BM structure in the treatment of cancer [11]. BM contains ~ 10 μm-sized nanoporous, which prevents most cells from passing through [12]. However, cancer tissue often shows the absence of type IV collagen and laminin [13], which makes the aperture of the BM larger and cancer cells easier to pass through the BM. BM is structurally degraded in tumor tissues, allowing cancer cells to enter the bloodstream and distant organs and colonize distant sites through the degraded BM [14]. It can be confirmed that the composition of BM in HCC has altered, which could be one of the causes of invasion and metastasis of HCC [15]. Therefore, maintaining the structural integrity of BM plays an essential role in preventing growth, invasion and metastasis of HCC.

In addition, it has become a hot topic to analyze the survival prediction and prognosis of cancer patients through the assessment of BM genes. In bladder cancer, the lncRNAs related to BM genes have been used to construct the risk signature, which could accurately predict the prognosis of bladder cancer patients. Knockdown of lncC00649 in bladder cancer 5637 cells, one gene of the risk signature, could cause the proliferation of cancer cells, indicating that the predicted risk BM genes is of great significance for cancer [16]. In addition, another experiment constructed the risk model by the BM genes in clear cell renal cell carcinoma (ccRCC), which can provide the potential progression trajectory and treatment response prediction of ccRCC [17]. However, whether BM related genes are correlated with HCC patient prognosis is still unknown.

In this study, we first comprehensively analyzed the correlation between the differential expression of basal-related genes and the prognosis-related genes in the HCC patients. Then, a BM-related prognostic signature was constructed based on the Cancer Genome Atlas (TCGA) database, and further validated in the International Cancer Genome Consortium (ICGC) database. Furthermore, high-risk and low-risk groups were divided by the risk signature and immunocellular, immunofunctional and immunotherapeutic drug analyses were performed to explore the underlying mechanism.

Methods

Datasets

The RNA-seq (FPKM values) and corresponding clinical data of patients with Hepatocellular Carcinoma (LIHC) were downloaded from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/) via the GDC Data Portal. The data freeze (or release version) utilized was based on the download date of September 1, 2022. Patients with incomplete survival information (overall survival time < 30 days or missing) were excluded. A total of 374 HCC patients with complete transcriptome and clinical data were included as the training cohort. The validation dataset, comprising 240 HCC patients, was obtained from the International Cancer Genome Consortium (ICGC) database (https://dcc.icgc.org/) under the project code LIRI-JP. The data freeze corresponded to the download date of November 17, 2019. This dataset was used as the independent testing set. The ‘estimate’ R software was employed to compute ESTIMATE scores, which included stromal and immunological scores, and immune infiltration.

Selection of BM-related genes

Based on previous study [18], we screened out 223 BM-related genes (BMRGs) (shown in Supplementary Table S1). The scale method provided in the “limma” R package was used to identify the BM-related differentially expressed gene (DEGs) between tumor tissue (374 tumor samples) and adjacent nontumorous tissues (50 normal samples) in the TCGA dataset, with |log2 (fold change, FC)| >2 and false discovery rate (FDR) < 0.05 as filtering criteria. The “clusterProfiler” R package was utilized to conduct Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses based on the DEGs.

Evaluation of BMRGs in association with HCC prognosis

Univariate Cox regression was used to conduct a preliminary analysis of the prognostic value of BMRGs. The expression levels of the BMRGs and corresponding clinical data in HCC patients were included in the R software, and the “survival” R package was used for univariate regression analysis to screen prognostic BMRGs. P < 0.05 was the significant filtering condition.

Construction of a prognostic BMRGs signature

The entire TCGA dataset was assigned to the training set used to construct a prognostic BMRGs signature, and we utilized the training set to construct a BM model. Based on clinical univariate Cox analysis, the prognostic related BMRGs were screened. DEGs and prognostic related BMRGs were imported into R software, and the intersection genes were obtained by “Venn” package analysis. The LASSO penalized Cox regression were applied to generate a stable gene model by using ‘glmnet’ R package, and BMRGs closely associated with OS were screened. Multifactor Cox regression was further applied to analyze the selected BMRGs, and we finally constructed a prognostic BMRGs signature model from the screened BMRGs. Risk score was calculated by the following formula:

In which, β_BMRGs refers to the coefficient of BMRGs correlated with survival, and Expression_BMRGsn represents the expression of corresponding BMRG. Then, the patients were assigned into low- and high-risk groups, based on the median value of risk score of the training set.

Verification of the BMRGs signature

To validate the reliability of the BMRG signature, the entire ICGC cohort was assigned to a testing set. The risk score of the patients provided by the ICGC database was calculated according to the calculation formula obtained from the training set. Then, patients in testing set were assigned into low-risk group and high-risk group according to the median value of the risk score, and Kaplan-Meier survival curves were plotted between two risk groups.

Validation of the prognostic signature

Further analysis was performed between the low- and high-risk groups, including survival analysis, risk analysis, receiver operating characteristic curve (ROC), principal components analysis (PCA) and t-distributed stochastic neighbor embedding (t-SNE) analysis. To visualize the difference in survival between high- and low-risk groups, survival analysis was performed for each gene, survival curves were plotted by the “surviminer” R package. In order to visually display the survival risk of patients in the high- and low-risk groups, the “pheatmap” R package was used to construct the risk curve. To assess the predictive power of the gene signature, time-dependent 1-, 3-, and 5-year ROC curves were analyzed by “timeROC” R package. Based on the risk signature, PCA was carried out with the “prcomp” function of the “stats” R package and t-SNE was performed using the “Rtsne” R package to explore the distribution of different groups.

Clinical correlation analyses

Clinical correlation analysis of TCGC database data was performed using the the “ggpubr” R package. To generate nomograms of 1-, 3-, and 5-year OS of HCC, the “rms” R package was used, and the Hosmer Lemeshow test was applied to establish a calibration curve to indicate whether the predicted results were consistent with the actual results. In addition, univariate Cox and multivariate Cox analyses were utilized to verify whether the risk score represented an independent role. In order to show the relationship between the expression levels of the risk signature genes and clinical traits in the high- and low-risk groups, “pheatmap” R package was used to construct clinical correlation heatmap.

Immunofunction analyses

To assess the differences of different immune cells between the low- and high-risk groups, an immune cell correlation analysis heatmap was constructed using the immune cell infiltration file collated from the TCGA website (http://timer.comp-genemics.org) and R software installed with the “limma” and “pheatmap” packages. The patient data of high and low risk group obtained from TCGA cohort were scored by the single-sample gene set enrichment analysis (ssGSEA) to obtain an immune function score for each patient. It was calculated by the “GSVA” R package and included the infiltrating score of 16 immune cells and the activity of 13 immune-related pathways. To determine whether there were differences in risk scores among immune subtypes, a differential analysis of immune subtypes, including wound healing, IFN-gamma dominant, inflammatory, and lymphocyte depleted (C1-C4), was performed using “ggpubr”.

Pathway enrichment analyses

GSEA software (v.4.1.0) was used to conduct enrichment of KEGG pathway and the GSVA enrichment heat map of risk genes in low- and high-risk groups to obtain risk signature data. Then, the top 5 KEGG pathways in low- and high-risk groups were selected.

Drug sensitivity

The RNA-seq data and compound activity data were obtained from CellMiner database (http://discover.nci.nih.gov/cellminer/home.do). Drug sensitivity analysis was performed using the “ggpbur”, “ggplot2”, “limma” and “impute” R packages.

Statistical analyses

All statistical analyses were performed with R software (version 3.5.3) and its resource packages were employed. Student’s t-test was utilized with p < 0.05 as the threshold for statistical significance to determine whether differences between different risk groups were significant.

Results

Identification of differentially expressed BMRGs and functional analyses

In order to show the investigation procedure clearer, the flow chart of this study is illustrated in Fig. 1. A total of 374 HCC samples contained mRNA expression profiles and corresponding clinical data were downloaded from the TCGA website and were used as the training set. Moreover, a total of 240 HCC samples were obtained from the ICGC website and enrolled as the testing set. By comparing the expression of 223 BMRGs in 374 tumor and 50 normal samples from HCC patients in the TCGA cohort, there were 51 BM-related DEGs in tumor samples with |log₂ FC| >2 and FDR < 0.05 (Fig. 2A). Results showed that ADAMTS13 and ECM1 genes were of low expression, whereas the other 49 genes were highly expressed in the tumor samples (S2 Table). These BM-related DEGs could be chosen in the following analysis, including GO enrichment and KEGG pathway analyses (Fig. 2B, C). As expected, in the GO rich concentration, these DEGs are mainly concentrated in the extracellular matrix organization, extracellular structure organization and external encapsulating structure organization in the biological process and they are closely related to extracellular matrix structural constituent, integrin binding and metalloendopeptidase activity in molecular function (Fig. 2B). In KEGG enrichment, these DEGs are mainly involved in the ECM-receptor interaction, focal adhesion, human papillomavirus infection and PI3K-AKTsignaling pathways (Fig. 2C).

Fig. 1.

Flow chart for the identification of prognostic BMRGs signature in HCC.

Fig. 2.

Exhibition of differentially expressed BMRGs of HCC patients in the TCGA cohort. A Volcano plot of DEGs between tumor and normal samples. The Y-axis represents the adjusted FDR, while X-axis represents the value of the log2 fold change (log2FC). Representative results of GO (B) and KEGG analyses (C). The most significant GO enrichment and KEGG pathways of the DEGs from the TCGA cohort. D Venn diagram to identify DEGs correlated with overall survival (OS)

Identification of prognostic BMRGs in the TCGA cohort

We conducted univariate regression analysis between BMRGs and OS status, and identified a total of 56 BMRGs significantly associated with OS as the prognostic genes (p < 0.05, S3 Table). Next, we selected the intersection genes of BM-related DEGs and prognostic genes. Finally, 14 BM-related DEGs that were strongly related to patient survival were identified as prognostic BMRGs (p < 0.05, Fig. 2D).

Construction of prognostic BMRGs signature model in the TCGA cohort

The 14 prognostic BMRGs were used to establish a prognostic BMRGs model by using Lasso penalized Cox regression analysis on the TCGA training set. The λ value, the best penalization parameter, was selected in Lasso regression analysis to determine which BMRGs could be used for subsequent risk signature analysis (Fig. 3A, B). Finally, an 8-BMRGs signature was identified based on the optimal value of λ, which included matrix metallopeptidase 1 (MMP1), matrix metallopeptidase 14 (MMP14), integrin subunit alpha2 (ITGA2), laminin subunit gamma 1 (LAMC1), meprin A subunit alpha (MEP1A), roundabout guidance receptor 1 (ROBO1), SPARC like 1 (SPARCL1), tubulointerstitial nephritis antigen (TINAG). All the 8 genes in the signature were high expressed in the HCC (Fig. 3C). Interestingly, the forest map shows that SPARCL1 is a low-risk gene and the other 7 genes are high risk genes. It means that SPARCL1 is low expressed in the high-risk group in the risk signature (Fig. 3D). The risk score was calculated as follows: Risk score = β_MMP1 × Expression_MMP1 + β_MMP14 × Expression_MMP14 + β_ITGA × Expression_ITGA + β_LAMC1 × Expression_LAMC1 + β_MEP1A × Expression_MEP1A + β_ROBO1 × Expression_ROBO1 + β_SPARCL1 × Expression_SPARCL1 + β_TINAG × Expression_TINAG.

Fig. 3.

Prognostic BMRGs signature of HCC patients in the TCGA cohort. A Cross validation of adjustment parameter (λ) selection in LASSO. B LASSO coefficient spectra of 14 core genes. C The heatmap of the gene expressions of 8-BMRGs signature between tumor samples and normal samples in HCC patients of the TCGA cohort. D Forest plots showing the results of the multivariate Cox regression analysis, and 8 genes were enrolled in prognostic BMRGs signature, including MMP1, MMP14, ITGA2, LAMC1, MEP1A, ROBO1, SPARCL1, TINAG. A P-value < 0.05 would be considered statistically significant. Hazard ratio (HR) > 1 means the gene is high risk gene and HR < 1 means it is low risk gene

Evaluation of the prognostic BMRGs signature model in TCGA cohort

In the training set, the established BMRG signature gave each HCC patient a risk score by the prognostic model. A total of 365 patients were screened from the TCGA cohort by removing patients with a survival time of 0. Patients are ranked by risk score from lowest to highest, and the median number was the optimal cut-curve risk score (Fig. 4A). The patients were divided into high-risk group (n = 182) and low-risk group (n = 183) according to the optimal cut-off value of risk score. The distribution of risk scores and the correlation between risk scores and OS were presented in the form of scatter plots (Fig. 4B). Results showed that, in terms of survival time, patients in high-risk group had a significantly lower survival time than those in the low-risk group (Fig. 4B). In addition, the Kaplan-Meier survival curve plotted in Fig. 4C showed that patients in high-risk group had a lower survival probability (p = 5.311e^− 4). The ROC curve showed the predictive effect of risk score on OS. The area under curve (AUC) values of the 1-, 3- and 5- year survival patients were 0.705, 0.679 and 0.589, respectively (Fig. 4D), which provided further evidence that the verified prognostic BMRGs signature possessed reliable predictive ability in the training set. PCA and t-SNE analysis indicated the patients in different risk groups were distributed in two directions (Fig. 4E-F).

Fig. 4.

Prognostic analysis of the BMRGs signature model in the TCGA cohort. A The distribution of risk scores obtained from the prognostic BMRGs signature model and median value of the risk scores in the TCGA cohort. The patients were divided into low- and high-risk groups according to the median value of the risk score. B The distributions of OS status, OS and risk score in the TCGA cohort. C Kaplan-Meier survival curves of patients in the low- and high- risk groups classified by the BMRGs signature in the TCGA cohort. Patients in high-risk group had a poorer outcome than those in low-risk group (P < 0.05). D Time-dependent ROC curves verified the prognostic performance of the BMRG signature model in the TCGA cohort. When 0.5 < AUC < 1 means the signature model that we have built has significant predictive value. E PCA analysis of 8 BMRGs in TCGA cohort. F t-SNE analysis of 8 BMRGs in the TCGA cohort. The blue and red points represent the patients in low- and high-risk groups, respectively

Validation of the prognostic BMRGs signature model in ICGC cohort

To test the robustness of the prognostic model constructed from the TCGA cohort, the risk score of each patient from the ICGC cohort was calculated with the established prognostic BMRGs signature. And the patients were categorized into high- or low-risk groups based on the median risk score value of the testing set (Fig. 5A). As shown in Fig. 5B and C, HCC patients in high-risk group showed poor survival status compared with those in low-risk group in ICGC cohort. In addition, the ROC curve was drawn based on the ICGC cohort to evaluate the reliability of the prognostic BMRG signature. In the ICGC cohort, the AUC of ROC curve was 0.688, 0.706 and 0.684 at 1-, 3- and 5 years (Fig. 5D), respectively. Similarly, patients in the two different risk groups were also distributed in discrete direction by PCA and t-SNE analysis (Fig. 5E-F).

Fig. 5.

Validation of the prognostic BMRGs signature model in the ICGC cohort. A The distribution of risk scores obtained from the prognostic BMRGs signature model and median value of the risk scores in the ICGC cohort. The patients were divided into low- and high-risk groups according to the median value of the risk score. B The distributions of OS status, OS and risk score in the ICGC cohort. C Kaplan-Meier survival curves of patients in the low- and high- risk groups classified by the BMRG signature in the ICGC cohort. Patients in high-risk group had a poorer outcome than those in low-risk group (P < 0.05). D Time-dependent ROC curves verified the prognostic performance of the BMRGs signature model in the ICGC cohort. E PCA analysis of 8 BMRGs of HCC in ICGC cohort. F t-SNE analysis of 8 BMRGs of HCC in the ICGC cohort. The blue and red points represent the patients in low- and high-risk groups, respectively

Construction of the nomogram and evaluation of the prognostic BMRGs signature

To determine the signature’s potential use in the clinical setting, we created 1-, 3-, and 5-year calibration plots and a nomogram based on the patient’s clinical information from TCGA cohort, and found that some clinical factors were accorded well with the OS prediction (Fig. 6A, B). The univariate Cox regression showed the correlation between some single factor and prognosis, such as age (HR = 1.010, p = 0.174), gender (HR = 0.188, p = 0.776), grade (HR = 1.133, p = 0.330), stage (HR = 1.680, p < 0.001) and risk score (HR = 3.450, p < 0.001). Among them, the stage and the risk score were the most significantly related to OS (Fig. 6C). The multivariate Cox regression analysis revealed that stage (HR = 1.585, p < 0.001) and risk score (HR = 3.086, p < 0.001) were still independent predictors for OS (Fig. 6D). And there were significant differences in risk score at different stages (Fig. 6E). Clinical correlational heat map analysis showed the change trend of the gene expressions of BMRGs signature in different risk groups and different related parameters (Fig. 6F).

Fig. 6.

Clinical correlation analyses. A The nomogram associated with clinical parameters to predict the 1-, 3-, 5-year OS in patients with HCC from TCGA cohort. B The calibration curves of the nomogram for the prediction of 1-, 3-, and 5-year OS. The line in grey represents an ideal nomogram, and lines in other color shows the current nomograms. The ideal and current lines approach together means the nomogram performed well. Vertical lines represent 95% confidence interval (CI). C Univariate Cox regression analyses for the risk score model as an independent prognostic factor regarding OS in TCGA cohort. P value < 0.05 means the close relationship between the factor and the prognosis of HCC patients. D Multivariate Cox regression analyses. E Correlation between clinical stage and risk score. F The heatmap of the clinical correlation between low- and high-risk groups in TCGA cohort

Analysis of immune infiltration, function, and immune checkpoints

Recent studies have shown that cancer can be effectively suppressed through immunotherapy, and immunocheckpoint inhibitors (ICIs) have been shown to have positive therapeutic effects on patients with advanced HCC [19]. Although both BM and immune cells are important components of the tumor microenvironment, the link between them remains unclear, which prompted our follow-up analysis. Found in existing studies,

In view of this, we subsequently compared the immune infiltration, functions and immune checkpoints between the two risk groups. The relationship between risk score of prognostic BMRGs signature model and specific immune cell infiltration in human cancers was subsequently performed by GSVA. The results showed that the levels of immune infiltration of B cells, CD4⁺ T cells and CD8⁺ T cells were significantly different between high- and low-risk groups (Fig. 7A). It suggests that the immune cell levels of high-risk patients were higher than those of low-risk patients. The expression distributions of immune checkpoint gene in the two groups are shown in Fig. 7B, and the expression levels of CD274, CD276, TNFSF4, CD200 and CD27 were higher in the high-risk group than in low-risk group (Fig. 7B). In the difference analysis of immune function, it was found that APC costimulatory point, immune checkpoint, T-cell co-inhibition and stimulation were significantly different between high- and low low-risk groups (Fig. 7C). The risk scores of different immune subtypes were significantly different (Fig. 7D), which provide a reference for the research and development of immunotherapy drugs for HCC. These results indicated that the BMRG signature is of great significance in the evaluation of immune function.

Fig. 7.

Investigation of tumor immune factors and immunotherapy between different risk groups in TCGA cohort. A Immune correlation heatmap. Different methods for immuno-oncology are used to evaluate the infiltration of immune cells, such as TIMER, CIBERSORT, CIBERSORT-ABS, QUANTISEQ, MCPCOUNTER, XCELL, EPIC. B Comparison of immune checkpoints between different risk groups. C Comparison of the ssGSEA scores of 13 immune-related functions between different risk groups. D Relationship between immune typing and risk score. C1-C4 are wound healing, IFN-gamma dominant, inflammatory, and lymphocyte depleted, respectively. p < 0.05 means that patients with different immune subtypes had different risk scores. *p < 0.05, **p < 0.01, ***p < 0.001

Pathway enrichment analyses

GSEA analysis was used to analyze the transcriptional information of HCC patients in two risk groups. According to KEGG pathway enrichment analysis, drug metabolism CYP450, fatty acid metabolism, glycine, serine and threonine metabolism, tryptophan metabolism, and valine, leucine and isoleucine degradation were mainly enriched in the high-risk group (Fig. 8A). Pathways in the low-risk group were mainly enriched in cell circulation, endocytosis, pyrimidine metabolism, RNA degradation, and spliceosome pathways (Fig. 8B). GSVA enrichment was then performed to further explore potential signaling pathways enriched by the risk signature. It is easy to find that adipocytokine signaling pathway and PPAR signaling pathway were negatively correlated with risk score. Pathway enrichment of the previously screened 8 prognostic BMRGs revealed that these genes were significantly enriched and positively correlated with most cancer-related signaling pathways, including WNT, mTOR and P53 signaling pathways. In certain pathways, these genes showed consistent negative correlation, such as PPAR and adipocytokine signaling (Fig. 8C). In addition, T cell receptor signaling pathway, B cell receptor signaling pathway and other immune-related signaling pathways were also enriched, which further confirmed the association between signature and HCC immunity.

Fig. 8.

Pathway enrichment analyses. A, B GSEA analyses of the top five enriched pathways in the high- and low-risk group in the TCGA cohort, respectively. C KEGG enrichment heat map of the 8 BMRGs in the signature

Drug sensitivity

When exploring the relationship between risk score and the efficacy of common chemotherapy drugs in the treatment of HCC, it was found that several drugs had lower IC₅₀ in high-risk patients who were identified by our prognostic model, including Oxaliplatin, Dexrazoxane, Crizotinib and Raloxifene (Fig. 9). Moreover, Irofulven and Vemurafenib were more sensitive to patients with low-risk score (Fig. 9). These results provide an important basis for the research and development of clinical anti-HCC drugs in the future and provide new strategies for further research to find potential therapeutic targets.

Fig. 9.

Multi-drug resistance analyses. The analyses show the drug susceptibility correlation between the gene in prognostic BMRGs signature and chemotherapy drug commonly used in HCC. The X-axis represents the amount of gene expression and the Y-axis represents the strength value of drug sensitivity. p < 0.05 is considered to be significantly different. Cor value is the correlation coefficient. Cor > 0 means that the gene expression was positively correlated with drug sensitivity, while Cor < 0 represents negative correlation. The blue line is the correlation curve

Discussion

Globally, the prevalence of HCC risk factors has varied significantly over the years [20]. It is of great value to find meaningful risk signatures to predict the prognostic effect of HCC patients [21]. BM plays an important role in the invasion and metastasis of malignant tumors [22]. Thus, BMRGs have been considered as potential targets to inhibit cancer development, and the change of its expression provides an important basis for the prognosis of HCC. Therefore, it is of considerable research value to seek the risk signature of BMRGs for HCC.

In this study, we systematically investigated the association between the 223 BMRGs and DEGs in a large-scale HCC samples. ECMs play a key role in cancer as drivers of tumor metastasis, and increased expression of genes encoding to mediate ECM constitution is associated with increased mortality in patients with breast, lung and gastric cancers [23]. Our results indicated that DEGs were mainly enriched in ECM organization and structure in BP, and enriched in interign binding, metallopeptidase and metallopeptidase activities in MF. In KEGG analysis, the DEGs were mainly enrich in ECM-receptor interaction pathway and PI3K-AKT pathway. Therefore, the screened DEGs play an important role in the occurrence and development of HCC. A novel prognostic model integrating 8 BMRGs was constructed in TCGA cohort and was validated in ICGC cohort, demonstrating its sensitivity and specificity in our study. The established signature separated HCC patients into high-risk group and low-risk group with significantly different survival time in the TCGA and ICGC cohorts, indicating its satisfied prognostic performance. The independence of the prognostic BMRGs signature in predicting the OS of HCC patients was not affected by clinical factors, including age, gender and grade. Notably, prognostic BMRGs signature is of great significance in the evaluation of immune function by assessing immune infiltration, functions, checkpoints and subtype, and may provide basis for screening more effective chemotherapeutic agents for HCC.

Our eight-BMRGs signature consists of MMP1, MMP14, ITGA2, LAMC1, MEP1A, ROBO1, SPARCL1 and TINAG. Among them, all genes are high-risk genes except SPARCL1 which is a low-risk gene in HCC samples. We found that all of them were closely associated with HCC patients’ OS. Matrix metalloproteinases (MMPs) play a key role in degrading ECMs and BM components [24], and previous studies have indicated that MMPs can promote HCC progression. MMP1 is highly expressed in cancer tissues and is associated with the migration and invasion of cancer cells [25]. Knockdown of MMP1 can block the HCC cell proliferation and invasion [26]. MMP1 has been shown to be a biomarker of poor HCC prognostic [27], which was also confirmed in our study. The expression of MMP1 was increased in HCC patients, and it exists as a high-risk gene in our constructed prognostic BMRGs signature. MMP14 is also a member of MMPS. The difference between MMP1 and MMP14 is that MMP14 is a membrane-type (MT) MMPs while MMP1 is a collagenase [24]. MMP14 has been reported to promote potential metastasis of tumor cells [28], and significantly increased expression of MMP14 in a variety of tumors, which can be involved in cellular migration, inflammation, and angiogenesis, has also been demonstrated [29]. In our study, the expression of MMP14 was increased in HCC patients and was identified as a high-risk gene.

ITGA2 encodes the transmembrane receptor alpha subunit of collagens and related proteins and is overexpressed in several types of tumors [30]. In the recent years, there is growing evidence that ITGA2 may play an essential role in modulating tumor cell migration, invasion, and metastasis [31]. Some genes related to cancer progression can induce proliferation, metastasis and migration of HCC by binding to the promoter region of ITGA2 gene, and up regulating its expression to activate ERK1/2 signaling pathway [32]. MEP1A is meprin A subunit alpha relevant to cancer invasion [33], and has been found to be overexpressed and induce cell migration and invasion in HCC [34]. In recent research, MEP1A levels were significantly elevated in HCC tumor tissues compared with matched adjacent nonneoplastic tissues and nonmalignant liver disease tissues. Immunohistochemical analysis showed that positive expression of MEP1A in tumor cells was an independent and significant risk factor for survival after curable resection [35]. In our study, ITGA2 and MEP1A were highly expressed in HCC patients provided by TCGA cohort, which were also is the high-risk genes in our signature. They are very important for adjusting the HCC prognosis and the survival rate of HCC patients. LAMC1, laminin subunit gamma 1, is a chain of laminin [36]. Laminins are involved in key cellular events involved in tumor angiogenesis, cell invasion, and metastasis development [37]. In recent research, LAMC1 was highly expressed in HCC tissues and cells and HCC can be improved by inhibiting LAMC1 expression [38], and the expression of LAMC1 is up regulated in HCC patients in our study. ROBO1 is round about guidance receptor 1, which can be activated by SLIT-family proteins. Macrophage ROBOs has been considered as a potentially novel regulator of the glioblastoma microenvironment and immunotherapeutic target for tumors [39]. TINAG is tubulointerstitial nephritis antigen encoding a glycoprotein that is restricted within the kidney to the basement membranes underlying the epithelium of Bowman’s capsule and proximal and distal tubules [40]. In our study, it was found that the increased expression of these genes as high-risk gene, can aggravate the prognostic risk of HCC patients. Notably, TINAG has not been reported to be related to HCC, which deserves further investigating.

Interestingly, SPARCL1 was identified as low-risk gene in our study, which possesses the ability to inhibit cancer cell proliferation, invasion, and glycolysis, and induce cell apoptosis [41]. SPARCL1 plays an important role in tumor pathogenesis and is significantly correlated with clinicopathological features and tumor microenvironment. It is significantly downregulated in some cancer tissues [42], and SPARCL1 was the only low-risk gene in our signature. This implies that high expression of SPARCL1 in HCC patients may improve the prognosis of patients.

Through the strengthening of the patient’s immune system, immunotherapy has been proved to be highly succeeded in making cancer a treatable disease for various malignancies [43]. Fairly extensive preclinical literatures emphasize that immune-base therapeutic strategies provide survival benefits for HCC. In our study, we divided high and low risk groups by risk signature, and analyzed immune infiltration, immune function, immune checkpoints and immune subtypes by high and low risk groups. With the increase of risk score, infiltration of B cells, T cells and NK cells increased significantly, and this was more obvious in the high-risk group than in low-risk group. The expression levels of immune-related genes, such as CD274, CD276, CD27, TNFSF4 and the other immune checkpoints, were also significantly increased in the high-risk group. Immune functions APC co-stimulation, CCR, T cell co-stimulation, and T cell co-stimulation were significantly enhanced in high-risk group, with different risk scores for different immune subtypes. The above results imply that the prognostic BMRGs signature constructed in our study can indeed differentiate the immune function, immune checkpoint, and immune infiltration of HCC patients at different risk groups, which can serve as an important reference for future clinical immunotherapy of HCC.

Moreover, BM genes in the signature, such as LAMC1 and MMP14, are related to the sensitivity of chemotherapy drugs. We found that patients showed decreased sensitivity to Oxaliolatin, Dexrazoxane, Crizotinib, and Raloxifene with increasing LAMC1 expression. In addition, the sensitivity of patients to Irofulven and Vemurafenib increased with the increasing expression levels of LAMC1 and MMP14, respectively. The signature screened common chemotherapy agents that are more sensitive to HCC patients in different risk groups, which provided rational recommendations for clinical application of more effective chemotherapeutic agents in HCC in the future.

The translation of our eight-BMRGs signature into clinical practice holds significant promise. Technically, the expression levels of these eight genes could be assessed using quantitative RT-PCR or Nanostring-based assays on tumor biopsy specimens, making it a feasible companion diagnostic tool. In the clinical scenario, this signature could serve multiple purposes. Firstly, it could stratify patients at diagnosis into distinct prognostic groups, identifying high-risk patients who might benefit from more aggressive adjuvant therapy or closer surveillance. Secondly, given its strong association with immune cell infiltration and checkpoint expression, the signature could help identify patients who are more likely to respond to immunotherapy, such as ICIs targeting PD-1/PD-L1 (e.g., antibodies against CD274). Compared to single biomarkers like AFP (Alpha-fetoprotein), which has limited sensitivity and specificity, our multi-gene signature offers a more robust and mechanistically informed prognosis. It also complements existing pathological staging by adding a molecular dimension that reflects the tumor’s invasive and immunomodulatory potential.

Nevertheless, several limitations should be mentioned in this study. This is a preliminary study of the prognostic value of BMRGs, with the purpose of providing some theoretical assistance for follow-up studies. However, we constructed a prognostic BMRGs signature model and validated it using retrospective data from public databases. More prospective real-world data is needed to confirm its clinical effectiveness. Next, we plan to conduct further studies to complement our findings. Not only limited to these studies, but also include validation of the expression of the prognostic genes and further investigation in the role of the prognostic genes. Furthermore, the accuracy of the prognostic BMRGs signature for the prognosis and immunomodulation in HCC patients remains a clinically important issue. The guideline for the clinical use of this prognostic BMRGs signature model needs to be defined further [4].

Several prognostic models for HCC have been established based on different biological perspectives, such as ferroptosis-related genes [4] and vascular invasion patterns [44]. The study highlights the significance of hemodynamic features from the portal venous and arterial systems in predicting postoperative outcomes, offering a radiological approach to prognosis [44]. In contrast, our eight-BMRGs signature provides a molecular-based tool rooted in the biology of the tumor microenvironment, specifically the basement membrane’s role in cancer invasion and immunity. The strengths of our model lie in its robust validation across two independent international cohorts (TCGA and ICGC), its integration with comprehensive immune landscape and drug sensitivity analyses, and its focus on a biologically coherent set of genes. This makes it particularly suitable for predicting responses to immunotherapy and chemotherapy. A potential limitation, shared by many transcriptome-based models, is the requirement for gene expression profiling, which may not be as readily available in all clinical settings as imaging or serum biomarkers. Future studies could integrate our BMRGs signature with radiological features [44]. And other molecular markers to build a more comprehensive and multi-modal prognostic framework for HCC, ultimately advancing towards personalized medicine.

Conclusions

In summary, our study defined a novel prognostic BMRG signature containing 8 basement membrane-related genes. This model proved to be independently associated with OS in both the TCGA and ICGC cohorts, providing insight into the prediction of HCC prognosis. The BMRG signature is not only closely linked to the prediction of HCC prognosis but can also provide sensible recommendations for immunotherapy and the selection of chemotherapeutic agents. The prognostic BMRG signature provides patients access to personalized risk assessments, which may be utilized to direct future care.

Abbreviations

BM

Basement membrane

BMRG

Basement membrane-related gene

ccRCC

Clear cell renal cell carcinoma

DEG

Differentially expressed gene

ECM

Extracellular matrice

FDR

False discovery rate

GO

Gene Ontology

GSVA

Gene set variation analysis

HCC

Hepatocellular carcinoma

ICGC

International cancer Genome Consortium

ICI

Immunocheckpoint inhibitors

ITGA2

Integrin subunit alpha2

KEGG

Kyoto Encyclopedia of Genes and Genomes

LASSO

Least absolute shrinkage and selection operator

LAMC1

Laminin subunit gamma 1

MEP1A

Meprin A subunit alpha

MMP1

Matrix metallopeptidase 1

MMP14

Matrix metallopeptidase 14

OS

Overall survival

PCA

Principal components analysis

ROC

Receiver operating characteristic curve

ROBO1

Roundabout guidance receptor 1

SPARCL1

SPARC like 1

ssGSEA

Single-sample gene set enrichment analysis

TCGA

The Cancer Genome Atlas

TINAG

Tubulointerstitial nephritis antigen

t-SNE

t-distributed stochastic neighbor embedding

Author contributions

Conceived and designed: Lina Xu; Writing-original draft: Guanlin Wu; Data analysis: Jiayao Li; Data collection: Xiaohan Zhang; Debug software: Lianhong Yin; Project administration: Xu Han, Meng Gao, Xuerong Zhao.

Funding information

Dalian Science and Technology Innovation Foundation. Grant/Award Numbers: 2022JJ13SN066.

Data availability

Data included in this study was available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.