Prognostic implications of Aquaporin 9 expression in clear cell renal cell carcinoma

All of the study designs and test procedures were performed in accordance with the Helsinki Declaration II. Study protocols were obtained by Fudan University Shanghai Cancer Center (FUSCC) (Shanghai, China) included in this work.

A total of 533 ccRCC patients with available RNA-sequence data from the Cancer Genome Atlas (TCGA) database were consecutively recruited in analyses [17]. The gene expression profile was measured experimentally using the Illumina HiSeq 2000 RNA Sequencing platform by the University of North Carolina TCGA genome characterization center. Level 3 data was downloaded from TCGA data coordination center. X-tile software was utilized to take the cut-off value of mRNA expression of AQP9, in concordance of which overall participants were divided to two groups, respectively. Student’s t tests were used to compare differential transcriptional expressions levels of AQP9 between paired AJCC stages or ISUP grades, marked in asterisk. The overall statistical expression difference of AJCC stages or ISUP grades was measured using One-way ANOVA test.

We next enrolled a total of 380 ccRCC patients from the Department of Urology, Fudan University Shanghai Cancer Center (FUSCC; Shanghai, China) from Aug 2009 to May 2018 in analyses. Tissue samples, including ccRCC and normal tissues, were collected during surgery and available from FUSCC tissue bank.

In this study, transcriptional expression profiles of AQP9 in ccRCC patients were obtained from Oncomine database using Oncomine online database (http://​www.​oncomine.​com) [18]. Difference of transcriptional expression was compared by Students’t-test. Cut-off of p value and fold change were as following: p-value = 0.01, fold change = 1.5, gene rank = 10%, Data type: mRNA.

The Human Pathology Atlas project (https://​www.​proteinatlas.​org) contains immunohistochemistry (IHC) data using a tissue microarray-based analysis on 44 different normal tissue types, and proteome analysis of 17 major cancer types [19]. Staining intensity, quantity, location and patients’ information in patients with the respective cancer types were available online. In this study, representative proteins expressions of IHC images of AQP9 were detected in ccRCC and normal tissues in Human Protein Atlas.

Total RNA sequence was extracted using TRIzol^® reagent (Invitrogen Life Technologies, USA) from 380 paired tumor and para-carcinoma normal samples. Primers were diluted in ddH₂O with SYBR Green PCR Master Mix (Applied Biosystems, Japan). Transcriptional expression was determined as the fold change of AQP9 relative to β-Actin. PCR primers sequence for AQP9 are as follows: forward are 5′-TTGCCCAAGCTATTCTCAGTCGA-3′ and reverse are 5′-CAGAGACACCGCCAGCCACAT-3′. The AQP9 mRNA expression was represented as ΔCt = Ct_(AQP9) − ΔCt_(β-actin). Relative expression in ccRCC was represented using the ratio of AQP9 expression in Tumor/Normal tissues (T/N). “Low AQP9 expression” and “High AQP9 expression” denote the T/N ratio of AQP9 mRNA expression with median cutoff in FUSCC cohort.

Immunostaining of AQP9 was performed using a mouse monoclonal anti- AQP9 antibody (1:100, Cat. ab84828, Abcam, USA). Positive or negative staining of a certain protein in one FFPE slide was independently assessed by two experienced pathologists, and determined as follows. The overall IHC score grading from 0 to 12 was evaluated according to the multiply of the staining intensity and extent score, as previously described [20].

Phenotype and expression profiles of hub genes in 533 ccRCC patients from TCGA were analyzed and displayed. Survival comparison between distinct mRNA expression groups of AQP9 was analyzed in ccRCC patients. The primary end point for patients was progression-free survival (PFS), and overall survival (OS) was the secondary end point, which was evaluated from the date of first therapy to the date of death or last follow-up. The follow-up duration was estimated using the Kaplan–Meier method with 95% confidence intervals (95% CI) and log-rank test in separate curves. Univariate and multivariate analysis were performed with Cox logistic regression models to find independent variables, including age at diagnosis, age (ref. < 60 years), gender (ref. Male), pT stage (ref. T1–T2), pN stage (ref. N0), pM stage (ref. M0), AJCC stage (ref. I–II), ISUP grade (ref. 1–2) and AQP9 expression (ref. Low). X-tile software was utilized to take the cut-off value [21]. All hypothetical tests were two-sided and p-values less than 0.05 were considered significant in all tests. Integrated score was identified as sum of the weight of AQP9 and significant clinicopathological prognostic indicators.

Search Tool for the Retrieval of Interacting Genes (STRING; http://​string-db.​org) (version 10.0) online database was used to predict PPI network of co-regulated hub genes and analyzing the functional interactions between proteins [22]. An interaction with a combined score > 0.4 was considered statistically significant.

Subsequently, the gene ontology (GO): BP (biological process), GO: CC (cellular component), GO: MF (molecular function) and KEGG pathways analyses for hub genes in this module were performed using Database for Annotation, Visualization and Integrated Discovery (DAVID; http://​david.​ncifcrf.​gov; version 6.8) online database [23], and then visualized in bubble chart. p-value < 0.05 was considered statistically significant. Cytoscape (version 3.5), an open source bioinformatics software platform, was used to visualize molecular interaction networks [24]. ClueGO is a Cytoscape plug-in that visualizes the non-redundant biological terms for large clusters of genes in a functionally grouped network [25]. The biological process from GO and KEGG pathway analysis of hub genes was performed and visualized using ClueGO (version 2.5.3) and CluePedia (version 1.5.3), a Cytoscape plug-in that visualizes the non-redundant biological terms for large clusters of genes in a functionally grouped network [26]. Gene set enrichment analysis (GSEA) was used to predict potential hallmarks using transcriptional sequences in TCGA database. A permutation test with 1000 times was used to identify the significantly changed pathways [27]. Adj. p less than 0.01 and FDR less than 0.25 were identified as significant related genes. Statistical analysis and graphical plotting were conducted using R software (version 3.3.2).

Tumor Immune Estimation Resource (TIMER, https://​cistrome.​shinyapps.​io/​timer/​) was used to perform comprehensive correlation analysis between tumor-infiltrating immune cells signatures and selected hub genes. An integrated repository portal for tumor-immune system interactions (TISIDB, http://​cis.​hku.​hk/​TISIDB/​index.​php) [28] was utilized to examine tumor and immune system interactions in 28 types of TILs across human cancers. The relative abundance of TILs were inferred by using gene set variation analysis based on AQP9 expression profile. Spearman’s test was used to measure correlations between AQP9 and TILs. All hypothetical tests were two-sided and p-values less than 0.05 were considered significant in all tests. All of these statistical analyses were performed in R or corresponding R packages survival and survminer.

To first assess the association of AQP family member expressions with prognosis, we obtained follow-up and transcriptional expression data from TCGA and evaluated the impact of AQP0–11 expressions on the prognosis of ccRCC patients (Additional file 1: Figure S1). It suggested that high expression of AQP9 significantly predicted poor OS (p < 0.001) and PFS (p < 0.001).

We next examined the correlations of AQP9 expression and clinicopathological characteristics in TCGA and FUSCC cohorts. As shown in Table 1, increased AQP9 mRNA expression in ccRCC patients significantly correlated with advanced pT (p < 0.001), pN (p < 0.001), and pM stage (p < 0.001), AJCC stage (p < 0.001) and ISUP grade (p = 0.004) in the FUSCC cohort. In the TCGA cohort, increased AQP9 mRNA expression significantly correlated with advanced pT (p < 0.001), pN (p = 0.004), and pM stage (p < 0.001), AJCC stage (p < 0.001) and ISUP grade (p < 0.001) (Additional file 3: Table S1).

We compared the mRNA expression of AQP9 between ccRCC samples and adjacent normal tissues based on RNA-sequence data from TCGA and independent cohorts in silico. AQP9 mRNA was highly expressed in 533 ccRCC tissues compared with 72 healthy tissues (p < 0.0001), as shown in Fig. 1a. AQP9 expression was also significantly higher in ccRCC primary tumors in comparison with adjacent normal tissues in GSE11151 (Yusenko Renal dataset; ***p < 0.001) [29], GSE14994 (Beroukhim Renal dataset; *p < 0.05) [30] and GSE6344 (Gumz Renal dataset; *p < 0.05) [31] (Fig. 1b–d).

IHC staining indicated that AQP9 staining was not detected in normal kidney tissues, while medium levels of expression (as defined in Methods) were observed in ccRCC tumor tissues (Fig. 1e, f). Taken together, these results suggested that AQP9 was highly expressed at transcriptional and proteomic levels in ccRCC tissues compared with normal tissues.

After integrating clinicopathological and survival data from TCGA, we found significantly elevated AQP9 mRNA expression in ccRCC samples compared with normal samples. As shown in Fig. 2a, AQP9 mRNA expression in ccRCC samples was significantly correlated with advanced clinical stage (p < 0.001), and the highest AQP9 mRNA expression was found in stage 4 cases. Figure 2b shows the relationship between AQP9 mRNA expression and different pathological grades, and the results suggested that AQP9 mRNA expressions were significantly correlated with pathological grade (p < 0.001). Similarly, the highest AQP9 mRNA expressions were found in grade 4 cases. Survival analysis using the Kaplan–Meier method showed that elevated AQP9 expression was significantly correlated with shorter PFS (p = 0.009) and OS (p < 0.001) in TCGA cohorts (Fig. 2c, d). Overall, elevated AQP9 mRNA expression was significantly associated with advanced clinicopathological parameters and poor prognosis in ccRCC patients from TCGA cohort.

To validate AQP9 mRNA expression in ccRCC tissues, we performed RT-qPCR in 380 paired tumor and normal samples with available clinical follow-up data from FUSCC cohort. We found dramatically increased AQP9 mRNA expression in ccRCC samples: 97.9% of ccRCC patients had higher levels of AQP9 expression in tumor tissues than normal tissues (Fig. 3a, b). To assess the level of AQP9 protein expression in FUSCC tumor samples, we performed IHC staining and found significant elevated AQP9 expression in terms of density and intensity in ccRCC tissues compared with adjacent normal kidney tissues in FUSCC cohort (p < 0.001, Fig. 3c, d).

In univariate Cox regression analysis models, traditional prognostic factors such as pTNM stage, AJCC stage, and ISUP grade were significantly relevant to PFS (p < 0.05; Additional file 3: Table S2) and OS (p < 0.001,; Additional file 3: Table S3) in ccRCC patients in both the TCGA and FUSCC cohorts. Importantly, AQP9 amplification markedly correlated with poor PFS (TCGA: hazard ratio [HR] = 8.141, p < 0.001; FUSCC: HR = 2.593, p < 0.001) and poor OS (TCGA: HR = 2.262, p < 0.001, FUSCC: HR = 2.774, p < 0.001).

In multivariate Cox regression analysis, traditional prognostic factors, specifically pM stage, were still relevant to PFS (TCGA: HR = 2.690, p = 0.043; FUSCC: HR = 2.593, p = 0.018; Table 2) and OS (TCGA: HR = 1.763, p < 0.001; FUSCC: HR = 1.895, p = 0.014; Table 3) in ccRCC patients. In addition, pT stage, pN stage, AJCC stage and ISUP grade were significant both in PFS (pT stage: p = 0.023, pN stage: p = 0.003, AJCC stage: p = 0.006, ISUP grade: p < 0.001) and OS (pT stage: p = 0.045, pN stage: p = 0.008, AJCC stage: p < 0.001, ISUP grade: p = 0.004) in the FUSCC cohort. Importantly, elevated AQP9 expression was significantly associated with poor PFS (TCGA: HR = 3.443, p = 0.040; FUSCC: HR = 1.714, p = 0.001) and poor OS (TCGA: HR = 1.714, p = 0.026; FUSCC: HR = 1.514, p = 0.026) in both cohorts of ccRCC patients.

In TCGA cohorts, survival analysis showed that elevated AQP9 expression was significantly correlated with shorter PFS (p = 0.009) and OS (p < 0.001). In FUSCC cohort, survival curves suggested that elevated AQP9 mRNA levels in patients significantly correlated with poorer PFS and OS (p < 0.001; Fig. 3e, f). For high AQP9 expression patients, the median PFS was 39.5 months and the median OS was 59.5 months. For low AQP9 expression patients, the median PFS was 66 months and the median OS was 72 months. ROC curves were generated to identify the ability of the gene model to predict prognosis events.

After integrating all the significant clinicopathological parameters and gene expression profiles in the Cox regression models (Table 2), we generated the formula: 1.782 × pT stage (ref. T1–T2) + 1.937 × pN stage (ref. N0) + 1.763 × pM stage (ref. M0) + 2.425 × AJCC stage (ref. I–II) + 1.812 × ISUP grade (ref. 1–2) + 1.714 × AQP9 expression (ref. Low) for PFS; and another formula: 1.692 × pT stage (ref. T1–T2) + 1.837 × pN stage (ref. N0) + 1.895 × pM stage (ref. M0) + 3.553 × AJCC stage (ref. I–II) + 1.751 × ISUP grade (ref. 1–2) + 1.514 × AQP9 expression (ref. Low) for OS. The AUC indices for the FUSCC-PFS and FUSCC-OS were 0.823 and 0.828, respectively (p < 0.001; Fig. 4a, b).

A network of AQP9 and its co-expression genes is shown in Fig. 5a. As illustrated in Fig. 5b, functional enrichment analyses of 11 involved genes were performed and the results are visualized in a bubble chart. Significant genes were significantly involved in polyol transport, defense response, and immune response, markedly participated in plasma membrane and were integral to membrane and intrinsic to membrane and the plasma membrane part. As shown in Fig. 5c, functional annotation using ClueGO indicated that changes in the biological processes of the AQP9 were significantly associated with the transport, integral component of membrane and the endomembrane system. Detailed functional annotations information and the percentage of each term were illustrated in Additional file 2: Figure S2.

A total of 100 significant genes were obtained by GSEA with positive and negative correlations. GSEA was used to perform hallmark analysis for AQP9. The results suggested that the most of the involved significant pathways included complement, coagulation, IL6/JAK–STAT3 signaling, inflammatory response, hypoxia, IL2–STAT5 signaling, allograft rejection and TNF-A signaling via NF-κB. The details are shown in Fig. 6a–h. In addition, transcriptional expression profiles of the 100 significant genes are shown by heat map in Fig. 6i.

After determining the prognostic value of AQP9, we performed correlation analysis between AQP9 and immune infiltration level for ccRCC. Elevated AQP9 was significantly associated with B cell, T cell, monocyte, macrophage, tumor-associated macrophage, and neutrophil cell infiltration (p < 0.05), leading to a general increase in immune infiltration. The Spearman’s correlation showed estimated statistical significance between AQP9 expression and immune cell signature infiltration in Additional file 3: Table S4. Partial correlation and correlation adjusted by tumor purity are also provided. Important signatures of a variety of immune cells including CD8 + T cells, T cells (general), B cells, monocytes, tumor-associated macrophages, M1 macrophages, M2 macrophages, neutrophils, natural killer cells, dendritic cells, Th1, Th2, Tfh, Th17, regulatory T cells (Treg), T cell exhaustion are illustrated in Additional file 3: Table S4. Additionally, we also found significant correlations of AQP9 with 28 types of TILs across human heterogeneous cancers (Fig. 7a). AQP9 significantly correlated with abundance of central memory CD8 T cells (Tcm_CD8 T cells; rho = 0.536, p < 0.001), macrophage (rho = 0.528, p < 0.001), natural killer T cells (NK T cells; rho = 0.485, p < 0.001), myeloid derived suppressor cells (MDSC; rho = 0.485, p < 0.001), gamma delta T cells (Tgd cells; rho = 0.479, p < 0.001) and Treg (rho = 0.459, p < 0.001) in Fig. 7b–g.