The TCGA database (https://portal.gdc.cancer.gov/) is a free website that provides a cancer genomics database, which contains the expression of genes, methylation, epigenetics, and proteome. The data on overall survival (OS), disease-specific survival (DSS), and progression-free interval (PFI) in pan-cancer were acquired via TCGA. The clinical data for PAAD were also obtained from TCGA.

The GEPIA database (https://gepia.cancer-pku.cn) is a free website that provides high-throughput RNA sequencing data for cancer. The expression of AQP5 in PAAD and normal pancreatic tissues was determined using GEPIA.

UALCAN (http://ualcan.path.uab.edu/) is a free website for the online analysis of gene or protein expression and clinical data in tumors. In PAAD, the correlation between the expression of AQP5 and pathological T stage and histologic grade was analyzed. Correlation analysis of AQP5 mRNA expression with the promoter methylation status was performed. The correlation between AQP5 promoter methylation status, tumor grade, and tumor stage was also analyzed and compared.

HPA is a free database that provides information on the expression of 24,000 proteins in tissues and organs. We used the HPA program to determine the distribution of AQP5 in PAAD and normal pancreatic tissues. We then acquired immunohistochemistry images of AQP5 proteins.

UCSC Xena (https://xenabrowser.net/) is a free website for cancer genomics analysis that contains other databases, including TCGA, IGGC, and CCL. Visitors can analyze and download a database. We used this website to acquire AQP5 RNA-seq data in PAAD and normal pancreatic tissues from TCGA and GTEx for ROC curve analysis. The mutation, CAN, and methylation of AQP5 in PAAD were also analyzed.

The TIMER database (https://cistrome.shinyapps.io/timer/) was used to analyze the correlation between gene expression and immune infiltration in tumor tissues. We used this website to analyze the correlation between AQP5 expression and six tumor-infiltrating immune cells in PAAD. We analyzed the abundance of tumor-infiltrating immune cells with different somatic copy number aberrations of AQP5 in PAAD. Cox regression analysis of AQP5 and the six tumor-infiltrating immune cells was performed using this website. The expression correlation between AQP5 and markers of immune cell genes was also determined.

The TISIDB database (http://cis.Hku.hk/TISIDB/) is a website that provides tumor immunity data. We used this website to analyze the relationship between AQP5 expression and 28 subtypes of tumor-infiltrating immune cells in different types of tumors. We also analyzed the correlation between AQP5 expression or methylation and tumor immune infiltration in PAAD.

The MethSurv database is a website for investigating the correlation between methylation biomarkers and the prognosis of patients with cancer. We used it to analyze DNA methylation at the CpG sites of AQP5 in PAAD and its prognostic value.

The Kaplan–Meier Plotter database (http://kmplot.com/analysis/) is a website that provides survival analysis of 54,675 genes in 21 tumor types. We performed a survival analysis of AQP5 expression levels in enriched immune cells in PAAD.

The LinkedOmics database (http://www.linkedomics.org/login.php) is a website that analyzes TCGA data. We searched for genes related to AQP5 expression in PAAD. The differentially expressed genes related to AQP5 were analyzed using Gene Ontology (GO) and classified as molecular function (MF), cellular component (CC), and biological process (BP). Gene set enrichment analysis and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analyses were performed using the link interpreter module.

The PAAD cell line (PANC1) and human pancreatic ductal epithelial cell line (HPDE6) were obtained from the ATCC. All the cell lines were cultured in DMEM containing 1% penicillin–streptomycin and 10% FBS. The normal pancreatic tissues of PAAD patients were obtained from Xijing Hospital. Western blotting, RT-PCR, immunofluorescence, and immunohistochemistry were performed as previously reported (21–23). The following antibodies were used: AQP5 (Abcam, Ab92320) and tubulin (Proteintech, 10094-1-AP). AQP5 primer sequence, Forward 5’-3’: GCCACCTTGTCGGAATCTACT; Reverse 5’-3’: GGCTCATACGT-GCCTTTGATG.

AQP5 knockdown and Transwell assays were performed as previously reported (23, 24). Small interfering RNA (si-RNA) against AQP5 and corresponding negative control (si-NC) were purchased from Tisingke Biotechnology Co., Ltd. (Beijing, China). si-AQP5 sequences and si-NC sequences were transfected into cells using Lipofectamine 3000 Reagent. si-AQP5-1 primer sequence, Forward 5’-3’: GGCACGUAUGAGCCUGACGTT; Reverse5’-3’: CGUCAGGCUCAUACGUGCCTT. si-AQP5-2 primer sequence, Forward 5’-3’: GCAUCUUCGCCUCCACUGATT; Reverse5’-3’: UCAGUGGAGG-CGAAGAUGCTT.

The expression of AQP5 in pan-cancer and normal tissues was acquired via the UCSC Nena website; the R package was used to analyze the difference. The ROC curve was plotted using R software. The survival data and clinical data were obtained via TCGA, the R package was used to analyze correlation, and the p-value was calculated by a log-rank test. Logistic regression analysis was used to analyze the relationship between AQP5 levels and clinical characteristics. Univariate and multivariate Cox regression analyses were performed to confirm independent risk factors. The measurement data are presented as the mean ± SD. Spearman’s correlation analysis was used for nonparametric tests. Statistical significance was set at p < 0.05.

We used the UCSC Xena database to analyze AQP5 mRNA expression in pan-cancer and normal tissues. The results indicated that AQP5 mRNA expression was very high in most tumors, such as adrenocortical carcinoma, pancreatic carcinoma, cholangiocarcinoma, and colon adenocarcinoma, compared to the corresponding normal tissues (Figure 2A). In TCGA, GSE16515, and GES71729 datasets, we found that AQP5 was expressed at higher levels in PAAD tissues than in normal tissues (Figure 2B). We also found that AQP5 expression in PAAD was higher than that in normal pancreatic tissues in the TIMER database (Figure 2C). The AQP5 expression was higher in histologic G2, G3, and G4 patients than that in G1 patients (Figure 2D). AQP5 expression was higher in patients with pathological T3 and T4 stage disease than in those with T1 and T2 stage disease (Figure 2E). In the HPA database, we found that AQP5 protein expression levels were significantly increased in PAAD tissues compared to normal pancreatic tissues (Figure 2F). The protein levels of AQP5 were examined in pancreatic cell lines and tissues. Figure 2G shows that AQP5 protein levels were overexpressed compared to immunohistochemistry in clinical PAAD tissue. Similarly, we confirmed this in the PAAD cell lines. AQP5 mRNA and protein levels were overexpressed compared with real-time PCR, Western blotting, and immunofluorescence (Figures 2H–J).

Kaplan–Meier analysis was used to compare the different prognoses between the AQP5 high- and low-expression groups in different types of tumors. A forest plot showed AQP5 as a risk factor for OS, PFI, and DSS prognostic values in PAAD (Figures 3A–C). We plotted survival curves for OS (HR = 1.9, p = 0.015), DSS (HR = 2.32, p = 0.033), and PFI (HR = 1.7, p = 0.028) (Figures 3D–F). The resulting ROC curve is shown in Figure 3G. These results showed that AQP5 may predict a worse prognosis in patients with PAAD.

We analyzed the association between AQP5 expression and the PAAD clinical index using logistic regression. The results indicated that high expression AQP5 was significantly associated with pathological T stage classification (OR = 0.133, p = 0.036 for T2 vs. T1; OR = 3.455, p = 0.013 for T3 vs. T2) (Table 1).

We investigated the prognostic risk factors for PAAD using the Cox regression analysis. Univariate Cox regression analysis indicated that pathologic stage II (HR = 2.333, p = 0.033), histologic G3 (HR = 2.625, p = 0.007), and AQP5 (HR = 1.142, p = 0.009) were factors that affected the survival of patients with PAAD. Multivariate Cox regression analysis showed that histologic G3 (HR = 2.191, p = 0.007) and AQP5 (HR = 1.134, p = 0.009) were independent risk factors influencing the survival of patients with PAAD (Table 2).

We used the TIMER database to analyze the relationship between AQP5 expression and immune cell infiltration in various cancers. The results indicated that the AQP5 expression was significantly correlated with gene markers of macrophage (r = 0.068, p = 4.32e-0.2) (Figure 4A). Using the TISIDB database, we analyzed the correlation between AQP5 expression and 28 subtypes of tumor-infiltrating immune cells in different types of tumors (Figure 4B). The results showed that CD56 (bright) cells (r = 0.345, p = 2.7e-6), CD56 (dim) cells (r = 0.287, p < 0.01), neutrophils (r = 0.15, p = 0.045), CD4⁺ central memory T cells (r = 0.275, p < 0.01), CD8⁺ central memory T cells (r = 0.207, p = 0.006), CD4⁺ effective memory T cells (r = −0.178, p = 0.018), and T helper cell 17 (r = 0.37, p = 4.28e-7) were significantly correlated with AQP5 expression in PAAD (Figure 4C). We used the Kaplan–Meier plotter database to investigate the survival of AQP5 expression, followed by subgroups of immune cells in PAAD. The results showed that low AQP5 expression levels in PAAD-enriched B cells, natural killer cells, CD8⁺ T cells, and Th2 cell cohorts were associated with a favorable prognosis (Figure 4D). We also found that the abundance of tumor-infiltrating immune cells was affected by different somatic copy number aberrations of AQP5 in PAAD (Figure 4E). In addition, we used the TIMER database to investigate the relationship between AQP5 expression and the gene markers of tumor-infiltrating immune cells in PAAD (Table 3). We found that AQP5 expression correlated with the gene markers of B cells (CD79A), TAM (CD68), Th1 (STAT1), and neutrophils (CEACAM8) in PAAD.

AQP5 expression is significantly elevated in PAAD. We assessed the cause of the elevated PAAD levels. Gene mutations, copy number variation (CNV), and DNA methylation are critically involved in genetic and epigenetic regulation and are highly associated with cancer progression. We verified DNA methylation, gene mutation, and CNV levels of the AQP5 in PAAD using the UCSC Xena database. The heatmap indicated that the expression of AQP5 mRNA was correlated with DNA methylation but not with CNV and somatic mutations in PAAD (Figure 5). Therefore, we suggest that DNA methylation may contribute to elevated levels of AQP5 in PAAD.

The methylation status of AQP5 was low in high-stage and high-grade tumors (Figure 6A). Furthermore, correlation analysis indicated that AQP5 mRNA expression was significantly negatively correlated with its methylation status (Figure 6B). We then used the TISIDB database to further explore the relationship between AQP5 levels and 28 tumor-infiltrating immune cell subtypes (Figure 6C). These results showed that AQP5 is associated with some immune cell subtypes in PAAD (Figure 6D). The results revealed that Th2 cells (rho = −0.177, p = 0.0181), Th17 cells (rho = −0.199, p = 0.00772), Tem-CD4 cells (rho = 0.178, p = 0.0175), Tcm-CD4 cells (rho = −0.199, p = 0.000426), Tcm-CD8 cells (rho = −0.261, p = 0.0143), iDC (rho = 0.164, p = 0.0279), eosinophils (rho = 0.317, p = 1.69e-05), CD56 cells (rho = −0.3, p = 4.91e-05), Act-DCs (rho = −0.184, p = 0.0138), Act-B cells (rho = 0.191, p = 0.0104), and AQP5 were moderately correlated. The heatmap shows the methylation of CpG islands in AQP5 (Figure 6E). Consistently, CpG sites of AQP5, including cg04450003 and cg16403326, indicated that PAAD methylation at these CpG sites was correlated with favorable prognosis in patients with PAAD. However, cg08266366 indicated that PAAD methylation at these CpG sites correlated with poor prognosis (Figure 6F and Table 4). These results reveal that the methylation levels of AQP5 act as an effective prognostic biomarker for patients with PAAD, demonstrating that AQP5 may play a pivotal role in tumor progression.

We used LinkedOmics to perform enrichment analysis of co-expression genes associated with AQP5 in PAAD. As shown in the volcano plot (Figure 7A), 1,424 positively associated genes (red) and 1,592 negatively associated genes (green) were identified in PAAD (p < 0.05, FDR < 0.01). A heatmap was used to display the top 50 positively and negatively correlated genes in PAAD (Figures 7B, C). GO-BP analysis showed that co-expressed genes were involved in transcription, DNA-templated viral process protein ubiquitination, cell–cell adhesion, intracellular protein transport, and the type I interferon signaling pathway. GO-CC analysis showed that these genes were mainly enriched in the nucleoplasm, cytosol, cytoplasm, nucleus, and perinuclear regions of the cytoplasm and membrane. GO-MF analysis showed that these genes were predominantly enriched for protein binding, ubiquitin-protein transferase activity, ligase activity, transcription coactivator activity, transcription factor activity, sequence-specific DNA binding, and zinc ion binding (Figures 7D–F). GO term annotation showed that co-expressed genes of AQP5 were mainly involved in Staphylococcus aureus infection, cell adhesion molecules, hematopoietic cell lineage, ECM–receptor interaction, viral myocarditis, JAK-STAT signaling pathway, osteoclast differentiation, phagosome, Th17 cell differentiation, and inflammatory bowel disease (Figure 7G). KEGG pathway analysis showed that the co-expressed genes were enriched in renal cell carcinoma, ErbB signaling pathway, Shigellosis protein processing in the endoplasmic reticulum, neurotrophin signaling pathway, and ubiquitin-mediated proteolysis pathways in cancer in PAAD (Figure 7H). Figure 8 shows the detailed GSEA enrichment plots.

To investigate whether AQP5 affects the migration and invasion of PAAD cells, we performed Transwell assays. As shown in Figures 9A, B, AQP5 knockdown significantly impaired the migration and invasion capabilities of PANC-1 cells. These results suggest that AQP5 promotes migration and invasion of PAAD cells.