Significant correlation between HSPA4 and prognosis and immune regulation in hepatocellular carcinoma

Immunohistochemical staining and ethics statement

Five pathologically diagnosed patients with HCC from the Second Hospital of Dalian Medical University were randomly selected for immunohistochemical staining to validate HSPA4 expression in HCC tissues (Anti-HSPA4 antibody (EPR14166); Abcam,Cambridge, England). Both cancerous and adjacent noncancerous tissues were collected simultaneously.

The ethics committee of the Second Hospital of Dalian Medical University approved this study in strict adherence to the Declaration of Helsinki protocol (number 2021059). Written informed consent was obtained from each patient.

Comparison of the HSPA4 expression level

Data from the TCGA database were used to analyze the expression of HSPA4 in 26 types of human cancer, in 374 HCC tissues with 50 normal liver tissues, and in 50 HCC tissues with their paired adjacent normal liver issues. In addition, gene expression profiles GSE62232 and GSE101685 were extracted from the GEO database to analyze the expression of HSPA4.

Correlation analysis of HSPA4 and cancer stage and prognosis

Correlation analysis of HSPA4 and cancer stage, including T stage, pathologic stage, histological grade, and AFP level was performed with R using the ggplot2 package. Kaplan–Meier plots were created and log-rank test was performed using survival package. The ROC curve of diagnosis, time-dependent curve of diagnosis, and nomogram model analysis were created using R packages, including the pROC, timeROC, and survival packages. The clinical data used in this section were retrieved from the TCGA database.

Genetic alteration in patients with HCC

The genomic profiles of HSPA4 in cBioPortal (www.cbioportal.org) were analyzed using three datasets, namely, hepatocellular carcinomas from TCGA, Firehose Legacy; AMC Hepatology 2014; and INSERM, Nat genet 2015. Kaplan–Meier plots were created and log-rank test was performed to identify the significance of the difference between the survival curves, and differences with p < 0.05 were considered statically significant.

DNA methylation information of HSPA4

The MethSurv database (https://biit.cs.ut.ee/methsurv/) was applied to analyze the DNA methylation sites of HSPA4 in the TCGA database. Moreover, the prognostic values of CpG methylation in HSPA4 was evaluated, and the survival outcome was the overall survival (OS).

Correlation analysis of HSPA4 with immune cell infiltration and immune checkpoints

TIMER (https://cistrome.shinyapps.io/timer/) was used to analyze the correlation of HSPA4 expression with immune cell infiltration and immune cell biomarkers in HCC. In addition, the correlation of HSPA4 expression with immune checkpoints in HCC was evaluated using TIMER and R with the ggplot2 package based on TCGA. Results with a p < 0.05 were considered statistically significant.

Gene ontology (GO) analysis and kyoto encyclopedia of genes and genomes (KEGG) analyses

The STRING database was used to obtain PPI network information of HSPA4. When the protein interaction score was above 0.9, the interaction was considered statistically significant. Ten functional partner genes were obtained for the GO term enrichment and KEGG pathway analyses to investigate the functions of HSPA4.

Statistical methods

R (v.3.6.3) was used to perform statistical analysis. Differences between groups were compared using the Wilcoxon rank-sum test or Student’s t-test, as appropriate. Correlations were determined using Pearson or Spearman correlation tests, as appropriate. Kaplan-Meier plots were created and log-rank tests were performed to identify the significance of the difference between the survival curves, and differences with a p < 0.05 were considered statistically significant.

Higher HSPA4 expression levels in cancerous than in noncancerous tissues

To explore the possible roles of HSPA4, we first analyzed its expression in 26 types of human cancer. As shown in Fig. 1, HSPA4 was significantly unregulated in 25 cancer types as compared with the corresponding normal tissues, including bladder urothelial carcinoma (BLCA), breast invasive carcinoma (BRCA), breast invasive carcinoma (CHOL), colon adenocarcinoma (COAD), glioblastoma multiforme (GBM), kidney chromophobe (KICH), liver hepatocellular carcinoma (LICC), lung squamous cell carcinoma (LUSC), rectum adenocarcinoma (READ), thyroid carcinoma (THCA), kidney renal papillary cell carcinoma (KIRP), lung adenocarcinoma (LUAD), prostate adenocarcinoma (PRAD), and uterine corpus endometrial carcinoma (UCEC). However, no significant difference was observed in kidney renal clear cell carcinoma (KIRC).

A higher expression of HSPA4 was observed in the GSE62232 and GSE101685 datasets (p < 0.001) (Figs. 2A, 2B). Overexpression of HSPA4 was observed in both paired and unmatched comparative studies based on the TCGA database (Figs. 2C, 2D). Immunohistochemical staining of clinical HCC samples also confirmed that the level of HSPA4 in tumor tissues was higher than that in adjacent normal liver tissues (Figs. 2E, 2F).

HSPA4 expression correlated with cancer stage and prognosis

The expression of HSPA4 remarkably correlated with T stage (Fig. 3A), pathologic stage (Fig. 3B), and histologic grade (Fig. 3C). Patients who were in a more advanced cancer stage tended to express a higher level of HSPA4. Higher HSPA4 expression was also associated with the AFP level (Fig. 3D).

According to the Kaplan–Meier survival curves, HCC cases with higher HSPA4 expression showed a lower OS (p < 0.001), a poorer disease specific survival (DSS) (p = 0.005), and a progression-free interval (PFI) (p = 0.039) (Fig. 4). Overexpression of HSPA4 was associated with a poorer prognosis.

Value of HSPA4 overexpression in diagnosis and predicting prognosis

From the ROC curve of diagnosis, HSPA4 expression showed an ability to accurately identify tumor from normal tissue (AUC = 0.957) (Fig. 5A). Time-dependent survival ROC curve of HSPA4 was created to predict 1-, 3-, and 5-year survival rates. All these AUC values were above 0.6, which is considered suitable for prediction (Fig. 5B). By integrating clinicopathologic factors (including age, T stage, and prothrombin time) and the HSPA4 level, a nomogram model was built, which can be used to predict the survival probabilities at 1-, 3-, and 5-years for patients in clinical practice (Fig. 5C).

HSPA4 genetic alteration in patients with HCC

A total of 916 patients with HCC from three datasets were analyzed—including TCGA, Firehose Legac; AMC Hepatology 2014; and INSERM, Nat genet 2015. The percentage of HSPA4 genetic alterations in HCC was 1.3% (Fig. 6A), and the alteration rate ranged from 0.82 (2/243) to 1.59% (6/377) (Fig. 6B). Kaplan–Meier plots and log-rank tests indicated no significant differences in OS (p = 0.464) (Fig. 6C) and disease-free survival (p = 0.180) (Fig. 6D) between patients with alterations and those without alterations.

HSPA4 methylation in patients with HCC

DNA methylation levels of HSPA4 with the prognostic value of each single CpG were investigated using the MethSurv tool. The results of MethSurv suggested 14 methylation CpG sites, among which cg05996250 and cg07474441 had the highest DNA methylation (Fig. 7). The methylation level of seven CpG sites correlated with prognosis, namely, cg02067788, cg24159723, cg26853048, cg04161464, cg13778073, cg05996250, and cg07474441 (p < 0.05) (Table 1). Patients with high HSPA4 methylation of these CpG sites had a worse overall survival than patients with low HSPA4 methylation.

Relationship between HSPA4 expression and immune cell infiltration

The relationship between HSPA4 expression and immune cell infiltration adjusted by purity (B cells, CD4+ T cells, CD8+ T cells, DCs, neutrophils, and macrophages) was investigated using TIMER. The results demonstrated that the expression level of HSPA4 positively correlated with infiltrating levels of B cells (r = 0.288, p = 5.36e−08), CD8+ T cells (r = 0.132, p = 1.46e−02), CD4+ T cells (r = 0.196, p = 2.48e−04), macrophages (r = 0.301, p = 1.50e−08), neutrophils (r = 0.33, p = 3.13e−10), and DCs (r = 0.205, p = 1.43e−04) in HCC but negatively correlated with tumor purity (r = −0.112, p = 3.8e−02) (Fig. 8).

Correlation between HSPA4 expression and biomarkers of immune cells in HCC

To further explore the role of HSPA4 in the tumor immune micro-environment, we identified the correlation between HSPA4 expression and biomarkers of immune cells in HCC using the TIMER database (Table 2). As listed in Table 2, HSPA4 significantly positively correlated with B cell biomarkers (CD19, CD20, and CD38), CD8+ T cell biomarkers (CD8A), other T cell subsets (Tfh, Th1, Th2, Th9, Th17, Th22, and Treg), M1 macrophage biomarkers (IRF5 and PTGS2), M2 macrophage biomarkers (CD115), TAM biomarkers (PDCD1LG2, CD80, CD40, and TLR7), natural killer cell biomarkers (CD7 and XCL1), neutrophil biomarkers (ITGAM and FUT4), and dendritic cell biomarkers (CD1C and ITGAX) in HCC. These findings support that HSPA4 is positively linked to immune cell infiltration.

Relationship between HSPA4 expression and immune checkpoints in HCC

PD-1 (PDCD-1) and CTLA-4 are important immune checkpoints that are responsible for tumor immune escape. Considering the potential oncogenic role of HSPA4 in HCC, the relationship of HSPA4 with PD-1 and CTLA-4 was assessed in both the TIMER and TCGA databases. There was a significant positive correlation of HSPA4 with PD-1 and CTLA-4 in HCC (Figs. 9A–9D).

GO and KEGG pathway analyses

Based on the STRING database, a PPI network of HSPA4 was constructed. The top ten functional partner genes were selected with a high degree of connectivity (Fig. 10A). These genes were BAG1, BAG3, DNAJB1, DNAGB4, HSP90AA1, HSPA8, HSPA9, HSPH1, STIP1, and STUB1 (Fig. 10B). The KEGG pathway analysis revealed that HSPA4, HSPA8, and HSP90AA1 were enriched in antigen processing and presentation (Fig. 10C; Table 3). The HSPA4 expression level positively correlated with the levels of HSPA8 (Fig. 10D) and HSP90AA1 (Fig. 10F) in patients with HCC (r = 0.570, p < 0.001; r = 0.550, p < 0.001). Overexpression of HSPA8 and HSP90AA1 was also associated with a poorer prognosis in HCC (p < 0.001) (Figs. 10E, 10G). The GO enrichment analysis included three main functions, namely, biological process, cellular components, and molecular functions (p < 0.05) (Table 3). The biological process mainly included “response to temperature stimulus”, “response to topologically incorrect protein”, “autophagy”, and “regulation of protein ubiquitination”. Molecular functions mainly included “heat shock protein binding”, “ubiquitin protein ligase binding”, “misfolded protein binding”, “cadherin binding”, “cell adhesion molecule binding”, and “MHC class II protein complex binding”. Cellular components mainly included “chaperone complex”, “endocytic vesicle lumen”, “cytoplasmic vesicle lumen”, “inclusion body”, “lysosomal lumen”, and “fificolin-1-rich granule”.