Clinicopathologic features of PC patients
A total of 85 PC patients were included in this study. The average age of these PC patients was 56.9 ± 6.5 years (ranging from 44 to 68 years), including 45 men and 40 women. There were 41 patients with tumor size less than 20mm, 44 patients with tumor size greater than or equal to 20mm, 71 patients with pancreatic head cancer, 14 patients with pancreatic body/tail carcinoma, 38 patients with serum CA19-9 ≤37U/mL, 47 patients with >37U/mL. Postoperative pathological examination confirmed that 29 patients with moderate/poor tumor differentiation, 56 patients with well tumor differentiation, 26 patients with non-metastatic lymph nodes, 59 patients with metastatic lymph nodes, 49 patients without PNI, 36 patients with PNI, 23 patients with stage I, and 62 patients with stage II. Thirty-five cases without early postsurgical recurrence and 50 cases with early postsurgical recurrence. A summary of clinicopathologic parameters is shown in Supplementary Table 1.
Expression of HSPA2 in PC tissues
Data from TCGA and GEPIA indicated that HSPA2 mRNA expression was evidently increased in PC specimens compared to nontumor samples (P <0.05; Fig. 1a-b). Immunostaining indicated that HSPA2 was basically distributed in the cytoplasmic space, as shown in Fig. 1c. Of the 85 PC specimens, overexpressed HSPA2 was observed in 70.6% (60/85). However, only 25.9% (22/85) of HSPA2 overexpression was found in the matched nontumor specimens. Statistical analysis revealed that HSPA2 expression was signally increased in PC specimens relative to surrounding paracarcinoma specimens (P =0.003; Fig. 1d and Supplementary Table 2). HSPA2 expression was observably higher in tumors with early postsurgical relapse compared with those without early postsurgical relapse (P <0.001; Fig. 1e and Supplementary Table 3). The correlation between HSPA2 expression and clinicopathological features was shown in Supplementary Table 4.
Logistic regression analysis of risk variables potentially associated with early postsurgical relapse of PC
To ascertain the variates that preestimate the early postsurgical recurrence of PC, we carried out univariate and multivariate analyses with logistic regression models. Univariate logistic regression analysis revealed that early neoplasm relapse was observably correlated with high serum CA19-9 levels before surgery (P =0.019), moderate/poor tumor differentiation (P =0.001), the presence of LNM (P <0.001), the presence of PNI (P =0.011), high tumor staging (P <0.001), and overexpressed HSPA2 (P <0.001) in PC patients after resection (Fig.2a and Supplementary Table 5).
In the multivariate logistic regression model, upregulated HSPA2 proved to be an independent hazard variate for early recidivation of PC following surgery [odds ratio (OR) =6.601; 95% confidence interval (CI) 1.635-26.652; P =0.008]. High tumor stage was also shown to be an independent variable for predicting early relapse of PC after operation (OR =8.049; 95% CI 1.783-36.334; P =0.007) (Fig.2b and Supplementary Table 5). The Hosmer-Lemeshow test showed a P value of 0.642 (greater than 0.05), indicating that this multivariate logistic regression model had a good goodness of fit.
Predictive nomogram for early postsurgical relapseof PC
A nomogram was established based on these risk variables associated with early postsurgical recurrence of PC. The results indicated that HSPA2 overexpression was the greatest risk factor, followed by high tumor stage, metastatic lymph nodes, presence of PNI, poor tumor differentiation and high serum CA19-9 levels (Fig. 2c). The C-index for internal validation of the nomogram model was 0.843 (95% CI 0.753-0.932). Internal calibration curve demonstated that nomogram prediction results were consistent with actual observation results (Fig. 2d).
Predictive performance of the nomogram
ROC curves were constructed to assess the reliability of these risk factors for predicting early relapse of PC following surgery. The area under the curve (AUC) was 0.760 (95% CI 0.649-0.871) for HSPA2, 0.630 (95% CI 0.509-0.751) for serum CA19-9, 0.671 (95% CI 0.552-0.791) for tumor differentiation, 0.750 (95% CI 0.638-0.862) for LNM, 0.641 (95% CI 0.522-0.760) for PNI, 0.756 (95% CI 0.643-0.868) for tumor stage. The combination of HSPA2 and the above-mentioned malignant clinicopathologic parameters improved the specificity and sensitivity over that of HSPA2 or these clinicopathologic parameters alone for predicting early postsurgical relapse of PC, with an AUC of 0.843 (95% CI 0.753-0.932) (Fig. 2e and Supplementary Table 6).
DCA was employed to quantify the probability of net benefits of risk factors to further assess their predictive performance and potential clinical application value for early recurrence of PC after surgery. The results demonstated that HSPA2 expression had a larger net benefit than other risk variables. The combination of all risk variables presented the highest net benefit relative to individual risk variables, indicating that HSPA2 expression combined with other malignant clinicopathological characteristics had a great advantage in predicting the possibility of early postoperative relapse (Fig. 2f).
Coexpression genes associated with HSPA2 in PC
To understand the biological function of HSPA2 in PC, we analyzed the coexpressed genes of HSPA2 using the LinkFinder module of LinkedOmics and performed GO and KEGG pathway analyses of coexpressed genes using the DAVID tool. The results of coexpression gene analysis suggested that there were 1,240 and 342 genes with distinct positive (red dots) and negative (blue dots) association with HSPA2, respectively (FDR <0.05; Fig. 3a and Additional File 1). The top 50 genes with positive and negative correlations with HSPA2 were displayed by heat maps (Fig. 3b).
GO and KEGG pathway analyses of HSPA2‑related coexpressed genes
The GO functional annotation results showed that HSPA2‑related coexpressed genes were principally involved in extracellular matrix organization, proteinaceous extracellular matrix, extracellular matrix, cell adhesion, focal adhesion, endoplasmic reticulum lumen, extracellular matrix binding, collagen trimer, collagen catabolic process, extracellular matrix structural constituent, integrin binding, calcium ion binding, extracellular space, Z disc, extracellular region, etc. (FDR <0.001; Fig. 3c). The KEGG pathway analysis indicated that HSPA2‑related coexpressed genes were enriched in multiple tumor-related signaling pathways, including ECM-receptor interaction, focal adhesion, PI3K-Akt signaling pathway, proteoglycans in cancer, pathways in cancer, cGMP-PKG signaling pathway, regulation of actin cytoskeleton, MAPK signaling pathway, TGF-β signaling pathway, etc. (FDR <0.05; Fig. 3d).
Identifification of HSPA2-associated gene sets by single-gene GSEA
To explore the potential mechanism of HSPA2 promoting the progression of PC (including biological processes and signaling pathway), we identified the gene sets enriched in the HSPA2 high expression dataset through single-gene GSEA based on the TCGA data. In the reference Hallmark gene set, NOM p value less than 0.05 indicated a significantly enriched gene set. The results revealed that there were 14 tumor-associated gene sets significantly enriched in the dataset with high HSPA2 expression, including epithelial mesenchymal transition, TGF-β signaling pathway, angiogenesis, apoptosis, IL6/JAK/STAT3 signaling, hypoxia, mitotic spindle, hedgehog signaling, peroxisome, Wnt β-catenin signaling, TNFA-α signaling via NF-κB, Kras signaling up, heme metabolism and P53 pathway (Fig 3e and Table 1).
In the reference KEGG gene set, FDR q value less than 0.05 suggested significant enrichment. The results showed that there were 26 tumor-related gene sets involved in focal adhesion, extracellular matrix (ECM) receptor interaction, TGF-β signaling pathway, regulation of actin cytoskeleton, pathways in cancer, basal cell carcinoma, small cell lung cancer, adherens junction, renal cell carcinoma, prostate cancer, axon guidance, NOD-like receptor signaling pathway, neurotrophin signaling pathway, Wnt signaling pathway, melanoma, colorectal cancer, gap junction, Fcγr mediated phagocytosis, MAPK signaling pathway, pancreatic cancer, glioma, endometrial cancer, chemokine signaling pathway, tight junction, cell adhesion molecules (CAMs) and hedgehog signaling pathway were significantly enriched in the phenotype with high expression of HSPA2 (Fig 3f and Table 2).
Expression of HSPA2 protein in human PC cell lines
The above functional and pathway enrichment analyses have revealed that HSPA2 expression was strongly associated with PC progression. Therefore, we next performed several in vitro cellular experiments to further validate the effect of HSPA2 expression on the biological behavior of human PC cells.
We obtained several human PC cell lines (BxPC-3, PANC-1 and SW1990) and further identified the protein expression pattern of HSPA2. The results of Western blotting determined that HSPA2 protein could be detected in BxPC-3, PANC-1 and SW1990 cells, among which the protein expression intensity in BxPC-3 cells was highest, and that in PANC-1 cells was lowest (Fig. 4a). Therefore, BxPC-3 and PANC-1 cells were selected for the downregulation and upregulation of HSPA2 in subsequent experiments, respectively.
Downregulation and upregulation of HSPA2 expression in PC cells
To realize the function of HSPA2 in the biologic behaviors of PC cells, such as propagation, apoptosis, migration and invasion, the expression of HSPA2 in PC BxPC-3 cells and PANC-1 cells was downregulated and upregulated, respectively. The siRNAs (siRNA-HSPA2-1, siRNA-HSPA2-2 and siRNA-HSPA2-3) targeting three different regions of HSPA2 were introduced into PC BxPC-3 cells to downregulate the expression of HSPA2. The siRNA without a specific target sequence served as a control (siRNA-NC). The recombinant plasmid with the human HSPA2 gene was introduced into PC PANC-1 cells to upregulate the expression of HSPA2. Untransfected cells and cells transfered with the mock vehicle were utilized as controls. The protein expression quantities of HSPA2 were measured using immunoblotting to assess the knockdown and overexpressing efficiency. Of the three siRNAs used, two (siRNA-HSPA2-2 and siRNA-HSPA2-3) worked well, and the other (siRNA-HSPA2-1) did not (Fig. 4b). The two siRNAs with higher silencing efficiency were selected for the downregulation of HSPA2 in subsequent experiments. The expression of HSPA2 in PANC-1 cells transferred with the recombinant plasmid (HSPA2 vector) was enhanced relative to the non-transfection group (blank control) and the empty vector group (Fig. 4c).
Effects of HSPA2 on proliferation of PC cells
To estimate the influences of HSPA2 silencing and overexpressing on the multiplication of BxPC-3 and PANC-1 cells, respectively, we implemented a CCK-8 assay. The data manifested that the proliferative ability of BxPC-3 cells in the two HSPA2 downregulation groups decreased relative to siRNA-NC, but there were no conspicuous differences (P >0.05; Fig. 5a). Consistently, the proliferation ability of PANC-1 cells in the HSPA2 upregulation group was increased compared with the non-transfection group (blank control) and the empty vector group, but there were no obvious differences (P >0.05; Fig. 5b).
Effects of HSPA2 on apoptosis of PC cells
To appraise the impacts of HSPA2 silencing and overexpressing on the apoptosis of BxPC-3 and PANC-1 cells, respectively, we implemented FCM analysis. The data revealed that HSPA2 silencing had no obvious impact on apopmsis of BxPC-3 cells versus the controlled siRNA (siRNA-NC) (P >0.05; Fig. 5c). Likewise, HSPA2 overexpression also had no significant effect on the apopmsis of PANC-1 cells in comparison to the non-transfected group (blank control) and the empty plasmid group (P >0.05; Fig. 5d).
Effects of HSPA2 on migration and invasion of PC cells
To identify the impacts of HSPA2 silencing and overexpressing on the migratory and invasive capabilities of BxPC-3 and PANC-1 cells, respectively, we carried out cell migration and invasion experiments with transwell cell culture chambers. The results showed that HSPA2 knockdown remarkably reduced the migratory number of BxPC-3 cells (P <0.01 for siRNA-HSPA2-2 vs. siRNA-NC, P <0.05 for siRNA-HSPA2-3 vs. siRNA-NC; Fig. 5e), while HSPA2 overexpression significantly increased the migratory number of PANC-1 cells (P <0.01 for blank control vs. HSPA2 vector, P <0.05 for empty vector vs. HSPA2 vector; Fig. 5f). Similarly, HSPA2 downregulation significantly decreased the invasive number of BxPC-3 cells (P <0.05 for siRNA-HSPA2-2 or siRNA-HSPA2-3 vs. siRNA-NC;Fig. 5e), and HSPA2 overexpression significantly elevated the invasive number of BxPC-3 cells (P <0.01 for blank control vs. HSPA2 vector, P <0.01 for empty vector vs. HSPA2 vector; Fig. 5f).
Effect of HSPA2 on VEGF expression in PC cells
To preliminarily understand the influence of HSPA2 on the angiogenesis of PC, VEGF concentrations in the supernatant of cell culture medium were detected by the ELISA method after HSPA2 was downregulated in BxPC-3 cells and upregulated in PANC-1 cells, respectively. The results indicated that the VEGF value in the supernatant of BxPC-3 cells did not change significantly after HSPA2 silencing compared to siRNA-NC (P >0.05;Fig. 5g). Similarly, the VEGF value in the supernatant of PANC-1 cells did not change significantly after HSPA2 overexpressing relative to the untransfected group and the mock carrier group (P >0.05;Fig. 5h). These results revealed that HSPA2 had no significant influence on the expression of VEGF in PC cells.