3.1 Identification of immune related genes—CNV-driven genes VPS72

We investigated the relationship between immune-related genes (IRGs) and copy number variation (CNV)-driven genes in TCGA-LIHC data using the cBioPortal website. Based on previous studies, we analyzed two key gene sets: 212 IRGs identified through IRG scoring [33], and seven CNV-driven genes (VPS72, PSMD4, RFX5, PSMB4, UBAP2L, CKS1B, and PIGC) [28]. To examine the genomic alterations, we categorized the samples into two groups: an altered group (n = 215) containing samples with at least one Amplification or Deep Deletion in IRGs, and an unaltered group (n = 151) without any such changes. Our analysis revealed that CNV-driven genes were significantly enriched in the altered group, with VPS72 and CKS1B showing particularly strong enrichment patterns in both genomic alterations and mRNA expression levels (Fig. 1A, B). Furthermore, survival analysis demonstrated that patients with high VPS72 mRNA expression had significantly poorer prognosis compared to those with low expression (Fig. 1C). Given these findings—the significant enrichment of VPS72 in the altered group and its correlation with poor survival outcomes—we selected VPS72 for further investigation.

Fig. 1

Identification of immune related genes—CNV-driven genes VPS72 in this study. A, B the enrichment status of CNV-driven gene in the altered group of genomic alterations and mRNA. C The value prognosis of VPS72 and CKS1B were explored by GEPIA. p < 0.05 *; p < 0.01 **; p < 0.001***. p < 0.05 *; p < 0.01 **; p < 0.001***. The altered group consists of chromosomal gains and genomic amplifications

3.2 VPS72 is associated with immunity

We investigated the relationship between VPS72 mRNA expression and tumor immune infiltration using multiple HCC datasets (TCGA-LIHC, GSE63898, GSE25097, GSE14520, and GSE9843). Initial TIMER analysis revealed predominant enrichment of CD8 + T cells and myeloid dendritic cells in these datasets (Fig. 2A). Based on the technical limitations of the TIMER analysis platform, we are currently only able to analyze 6 major immune cell types (CD8 + T cells, B cells, CD4 + T cells, neutrophils, macrophages, and myeloid dendritic cells). Among these analyzable cells, CD8 + T cells and myeloid dendritic cells showed significant enrichment characteristics. Correlation validation: In validation analyses across multiple independent cohorts (see Table S1), we found a stable and significant correlation between VPS72 and CD8 + T cells, which provided reliable data support for focusing our research on CD8A, a marker gene for CD8 + T cells.

Fig. 2

VPS72 correlated with immune infiltration. A Relative proportion of immune cell infiltrations in HCC datas. B the relationship between VPS72 mRNA expression and T cell CD8 + _TIMER

To validate the association between VPS72 and CD8 + T cells, we conducted two levels of analysis. We analyzed multiple public datasets on hepatocellular carcinoma (HCC) to explore the relationship between VPS72 mRNA expression and CD8 + T cells. In the training set, we observed that patients with low VPS72 expression had higher CD8 + T cells expression compared to patients with high VPS72 expression. This finding was consistent across different datasets: TCGA (t = 3.819, p = 0.0002), GSE63898 (t = 5.620, p < 0.0001), GSE25097 (t = 2.718, p = 0.0070), GSE14520 (t = 3.943, p = 0.0001), and GSE9843 (t = 1.483, p = 0.1417) (Fig. 2B). Further investigation revealed significant negative correlations between VPS72 mRNA expression and CD8 + T cells across multiple datasets (Fig. 3A: TCGA: R = − 0.23; GSE63898: R = − 0.38;GSE25097:R = − 0.08; GSE14520; R = − 0.32; GSE9843: R = − 0.14). To explore the underlying relationship, we analyzed correlations between VPS72 and key immune checkpoint genes (SIGLEC15, TIGIT, CD274, HAVCR2, PDCD1, CTLA4, LAG3, and PDCD1LG2). Notably, VPS72 showed significant negative correlations with CD274, HAVCR2, PDCD1, and PDCD1LG2 across multiple datasets (Fig. 3B). While VPS72 showed a positive correlation with PDCD1 in TCGA, its consistent negative correlation with CD8 + T cells and PDCD1 in other databases suggests an overall negative relationship. These comprehensive analyses across multiple datasets strongly support a negative association between VPS72 expression and CD8 + T cell levels in HCC, suggesting VPS72's potential role in tumor immune infiltration.

Fig. 3

A The expression of VPS72 was related to CD8 + _TIMER (TCGA = − 0.23****; GSE63898 = − 0.38****; GSE25097 = − 0.08; GSE14520 = − 0.32****; GSE9843 = − 0.14). B the relationship between VPS72 mRNA expression and immune genes. p < 0.05 *; p < 0.01 **; p < 0.001***. p < 0.05 *; p < 0.01 **; p < 0.001***

3.3 Differentially VPS72 expressed analysis

To gain further understanding of the role played by VPS72 expression in hepatocellular carcinoma (HCC), we utilized normalized HCC data and categorized the population into high expression and low expression groups based on VPS72 expression levels. Subsequently, we conducted differential analysis to examine any significant differences between the two groups. Our analysis of the training sets (GSE9843, GSE63896, and GSE14520) and the validation set (GSE25097) revealed a total of 6771, 7375, 6232, and 7926 differentially expressed genes (DEGs)(|log2FC|≥ 1.0, p < 0.05), respectively. KEGG pathway analysis demonstrated the involvement of the differentially expressed genes in the T cell receptor signaling pathway (CD8A), PD-L1 expression and PD-1 checkpoint pathway in cancer (Figure S1A-C). Additional GO enrichment analysis revealed that the differentially expressed genes were associated with the activation of CD8 and T cells within biological processes (BP) (Figure S2A-B). These findings suggest a potential association between VPS72 and CD8 + T cells.

We conducted further investigations into the association between VPS72 and CD8 + T cells. We analyzed multiple public datasets on hepatocellular carcinoma (HCC) to explore the relationship between VPS72 mRNA expression and CD8A expression.We observed that patients with low VPS72 expression had higher CD8A expression compared to patients with high VPS72 expression. This finding was consistent across different datasets: GSE63898 (t = 3.5184, p = 0.005), GSE9843 (Mann–Whitney U, p = 0.0063), GSE14520 (Mann–Whitney U, p = 0.0713), and GSE25097 (t = 3.3078, p = 0.0011) (Fig. 4A).Furthermore, we examined the correlation between VPS72 expression and CD8A expression. We found a significant negative correlation between VPS72 mRNA expression and CD8A expression in multiple datasets: GSE9843 (r = − 0.2679, p = 0.0104), GSE63898 (r = − 0.3013, p = 0.0001), and GSE14520 (r = − 0.1539, p = 0.0209) (Fig. 4B).These findings suggest a strong association between VPS72 expression and CD8A expression in HCC.

Fig. 4

The expression of VPS72 was related of CD8 + T cells. A, B in the training set, the expression of VPS72 was related to CD8A of CD8 + T cells. C in the validation set, low VPS72 expression had higher CD8A expression than patients with high VPS72 expression. D VPS72 expression was associated with BCLC stages and prognostic value of VPS72 in our cohort p < 0.05 *; p < 0.01 **; p < 0.001***. p < 0.05 *; p < 0.01 **; p < 0.001***

3.4 Confirmation of VPS72 prognosis value

Our study characteristics are presented in Table S2. VPS72 expression was often significantly different in groups stratified by BCLC stage (BCLC-C > BCLC-A; BCLC-B > BCLC-A; Fig. 4C). VPS72 expression was significantly different in groups stratified by BCLC stage. Notably, the low VPS72 expression group exhibited a more favorable disease-free survival (DFS) outcome (P = 0.0331, with a follow-up time of 0–36 months, n = 80, Fig. 4D).

3.5 Validation of VPS72 is associated with CD8A

Subsequently, we assessed the expression of CD8A. According to the results summarized in Table 1, out of the 139 patients, 77 (55.4%) showed positive CD8A expression (IHC score > 0), among which 43 (30.9%) exhibited strong CD8A positivity (IHC score > 2), while only one patient (0.7%) demonstrated strong CD8A expression (IHC score ≥ 6).In our cohort, we performed a correlation analysis of CD8A with VPS72 expression. Firstly, patients were first compared between high and low groups according to VPS72 protein expression level. The findings revealed that Patients with VPS72-low expression had higher CD8A expression than patients with VPS72-high expression (p = 0.0043; n = 112; Fig. 5A). Two outliers (CD8A: 10; 4.6) were detected in Fig. 5A and were excluded from the subsequent correlation studies to ensure data reliability.Furthermore, the correlation analysis demonstrated a significant negative correlation between VPS72 protein expression and CD8A protein expression (r = − 0.2517, p = 0.008, n = 110; Fig. 5B).

Table 1 The distributions of PD1 and CD8A in HCCFig. 5

Validation of VPS72 is associated with CD8A immune infiltration. A, B VPS72 protein expression was correlated with the protein expression of CD8A. C, D the VPS72 knockout subcutaneous tumor model exhibited a significantly reduced tumor weight compared to the control group and VPS72 is associated with CD8A in vivo. E–F low VPS72 and CD8A coexpression subtype had a better DFS than the high VPS72 and low CD8A expression in our data and in the TCGA-LIHC cohort

Moreover, in our in vivo experiments, we observed that VPS72-knockdown subcutaneous tumors model revealed that reduced the weights of tumors than control tumors, as depicted in Fig. 5C. Additionally, the expression level of CD8A in the VPS72 knockout subcutaneous tumors was found to be higher than control tumors (Fig. 5D).

3.6 Development of Lo-immune and Hi-immune

To further characterize the population, we divided it into four subtypes based on the co-expression of VPS72 and CD8A: VPS72 high/CD8A high, VPS72 high/CD8A low, VPS72 low/CD8A high, and VPS72 low/CD8A low. We then examined the relationship between four subtypes expression in HCC tissue and the clinical data of our collected cohort (Table 2). Significant associations were found for tumor size among the VPS72 and CD8A coexpression subgroups (p < 0.05).Survival analysis revealed that the subtype characterized by low VPS72 and high CD8A expression exhibited a superior disease-free survival (DFS) compared to the subtype characterized by high VPS72 and low CD8A expression in the TCGA-LIHC cohort (P = 0.0036; Fig. 5F). This conclusion was further validated in the our cohort (P = 0.01, with a follow-up time ranging from 2 to 88 months, n = 103, low/high CD8A expression group is distinguished by the median; Fig. 5E).Moreover, these distinct survival outcomes prompted us to designate the low VPS72 and high CD8A group as "Lo-Immune," and the high VPS72 and low CD8A expression group as "Hi-Immune".

Table 2 Clinical characteristics of the patients with VPS72/CD8A subtypes in our IHC cohort3.7 Differentially Lo-immune and Hi-immune expressed analysis

To explore the potential mechanisms underlying the prognostic differences between the Lo-Immune and Hi-Immune groups, we conducted differential analysis using the training sets (GSE9843, GSE63896, and GSE14520) and the validation set (GSE25097). The analysis revealed a total of 6519, 7293, 5675, and 7926 differentially expressed genes (DEGs) (with a threshold of |log2FC|≥ 1.0 and p < 0.05). Out of these DEGs, 4296, 4173, 2689, and 3254 genes were found to be upregulated, while 2223, 3120, 2986, and 4672 genes were downregulated. The Venn diagram analysis of the training sets (GSE9843 vs. GSE63896 vs. GSE14520) revealed 1485 common differentially expressed genes (DEGs) among the three datasets. Additionally, KEGG pathway analysis unveiled the involvement of these DEGs in pathways related to PD-L1 expression and PD-1 checkpoint pathway in cancer (Figure S3A). In the validation set (GSE25097), KEGG analysis also demonstrated that these DEGs were associated with pathways related to PD-L1 expression and PD-1 checkpoint pathway in cancer (Figure S3B). Thus, we established a connection between the Lo-Immune and Hi-Immune groups and the PD-1 pathway.

3.8 Potential of Lo-immune as an indicator of immunotherapy response in patients with HCC

To further analyze the differences in PD-1 and other immune checkpoints gene (SIGLEC15, TIGIT, CD274, HAVCR2, CTLA4, LAG3, and PDCD1LG2) between the Lo-Immune and Hi-Immune groups, we utilized CBioPortal. The analysis revealed a significant enrichment of PD-1 in both the Lo-Immune and Hi-Immune groups (Fig. 6A). Specifically, in the training set, patients in the Lo-Immune group exhibited higher PD-1 expression compared to patients in the Hi-Immune group (p < 0.05) (Fig. 6B, C). Similarly, in the validation set, patients in the Lo-Immune group also demonstrated higher PD-1 expression (p < 0.05), and this finding was further validated in the TCGA-LIHC dataset (Fig. 6D, E). In our cohort of HCC patients, most cases exhibited positive PD-1 expression, with 62 cases (44.6%) showing positive PD-1 expression (IHC score greater than 0). Among them, 5 cases (3.6%) were classified as strongly positive, with an IHC score greater than 2 (Table 1). Additionally, our analysis indicated that the Lo-Immune group exhibited higher levels of PD-1 expression compared to the Hi-Immune group (p = 0.0425, n = 70, 112 were followed up completely during the study Table S3; Fig. 6F). Furthermore, we evaluated the prognostic value of the co-expression subtypes: Lo-Immune PD1 and Hi-Immune PD1. The results demonstrated that the co-expression subtype of Lo-Immune Hi-PD1 correlated with a better disease-free survival (DFS) outcome (p = 0.04, n = 63 follow-up time ranging from 2 to 88 months), and this finding was validated in the TCGA-LIHC dataset (Fig. 6G, H).

Fig. 6

Impact of PD1 expression and immune related genes (Lo-Immune and Hi-Immune) on clinical outcome. A The enrichment status of immune checkpoints gene in the altered group of genomic alterations. B–E Patients in the Lo-Immune group exhibited higher PD-1 expression compared to patients in the Hi-Immune group. F Comparison of the expression pattern of PD-1 between patients with differents Lo-Immune and Hi-Immune in our data. G–H Lo-Immune & Hi-PD-1 had a better DFS than the other subtypes in our data and in the TCGA-LIHC cohort