CA9 transcriptional expression determines prognosis and tumour grade in tongue squamous cell carcinoma patients

Abstract CA9 is a member of the carbonic anhydrases’ family, that is often expressed in cancer cells under hypoxic condition. However, the role of CA9 in the molecular mechanisms of tongue squamous cell carcinoma (TSCC) pathogenesis remains unclear. CA9 expression was analysed using the TCGA database, and its influence on survival was performed using Kaplan‐Meier, LASSO and COX regression analyses. The correlation between CA9 and immune infiltration was investigated by CIBERSORT and ESTIMATE. Moreover, the relationship between CA9 expression and downstream molecular regulation pathways was analysed by GSEA, GO and WGCNA. CA9 expression correlated with clinical prognosis and tumour grade in TSCC. Moreover, CA9 expression potentially contributes to the regulation of cancer cell differentiation and mediates tumour‐associated genes and signalling pathways, including apoptosis, hypoxia, G2M checkpoint, PI3K/AKR/mTOR signalling and TGF‐beta signalling pathways. However, the follicular helper T cells, regulatory T cells, immune and stromal scores showed no significance between high and low CA9 expression groups. These findings suggested that CA9 plays a critical role of TSCC prognosis and tumour grade. CA9 expression significantly correlated with the regulation of cell differentiation, various oncogenes and cancer‐associated pathways.

studies indicated that the hypoxic condition of the cancer microenvironment regulates the expression CA9. Moreover, CA9 targeting, with immune-checkpoint blockade, improves the response and survival of patients with hypoxic solid malignancies. 6 In response to hypoxia, pancreatic cancer cells, that express activated KRAS, increase the expression of CA9 by stabilizing the transcription factor HIF1A and HIF2A, and mediate pH and glycolysis in the tumour microenvironment. 7 Eckert et al 8 demonstrated that CA9 levels in oral squamous cell carcinoma (OSCC) samples were higher than those in precancerous tissues. Additionally, immunohistochemical analysis of CA9 expression, in oral lichen and leukoplakia, suggested that its expression was also observed in oral precancerous lesions. 9,10 CA9 plasma levels may be used as a non-invasive method for monitoring OSCC progression. 11 LCN2 was able to target the CA9 transcriptional and post-transcriptional regulation in OSCC cells, which suppressed metastasis. 12 However, the molecular mechanism and regulation of CA9 gene expression in TSCC still require further investigation. The main goal of our study was focused on investigating the mechanism by which CA9 expression influences TSCC progression. Our results indicated that CA9 high expression was related to TSCC patients' poor outcome and that was mediated by oncogenes and tumour-associated pathways. In addition to the role of CA9 in hypoxia, we found that CA9 had a significant correlation with TSCC tumour grade and cell differentiation.

| Data source
Expression data and corresponding clinical information of TSCC pa-
The heatmaps were generated with the 'pheatmap' package in R (R version 3.5.3).

| Prognostic analysis of carbonic anhydrases
The prognostic analysis of TSCC patients was evaluated using the Kaplan-Meier method, the univariate Cox proportional hazards regression and the multivariate Cox regression analysis. The samples were stratified as high or low expression around the median of each carbonic anhydrases' expression (high expression: n = 73, low expression: n = 73).

| Identification of CA9 differentially expressed genes (DEGs)
DEGs were identified in TSCC tissues by comparison with CA9 high and low expression groups and using the 'edgeR' package in R (R version 3.5.3). The heatmap and volcano plot, with clustering for the significantly differentially expressed mRNAs in TSCC between CA9 high and low expression groups, were generated with the 'ggplot2' package. GO analysis was constructed by DAVID 6.8. 13,14

| Gene set enrichment analysis (GSEA)
The signalling pathway, underlying the association between CA9 expression levels, was explored with GSEA (version 3.0.). The number of permutations was set at 1000, and the P-value < .05 and an FDR < 0.25 were considered statistically significant. Multiple GSEA plots were produced by 'plyr', 'ggplot2' and 'grid' packages.

| The clinical features correlated with of CA9 expression level
The samples with insufficient clinical data were filtered out. Filtered clinical information from TSCC patients, including age, gender, tumour grade and stage, was obtained from TCGA. Patients were divided into two groups and according to the median of CA9 ex-

| Immune landscape related to CA9 expression level
CIBERSORT is a method for characterizing the immune cell composition using gene expression data to define 22 immune cell subtypes. 15 The results include a P-value for each sample of global deconvolution. Each sample in the data set will get a P-value, that is used to select the samples for further studies if their P-values are less than .05.
The immune and stromal scores were calculated by ESTIMATE. 16,17

| Weighted gene co-expression network analysis
The WGCNA 18 package was used to identify key modules, associated with tumour grade and based on CA9 expression levels. The module eigengenes of the clinical features were hierarchically clustered into different colour modules. GO analysis was used to evaluate the molecular functions of the co-expressed genes. Hub genes, that are related to tumour grade, were identified by Cytoscape.

| Evaluating the prognostic significance of carbonic anhydrases in TSCC
The expression levels of carbonic anhydrases are shown in Figure 1A.
Twenty anhydrases, that are related to prognosis and overall survival prediction, were identified using LASSO Cox analysis and univariate and multivariate Cox regression analyses (Table 1; Figure 1B,C). CA9 was eventually identified based on the Kaplan-Meier survival analysis (P = .002; Table 1). The CIBERSORT analysis demonstrated that follicular helper T

cells, regulatory T cells and eosinophils were significantly infiltrated
in the different groups ( Figure 2F). However, the immune score and stromal score showed no significance in the high and low CA9 expression groups ( Figure 2G,H).

| Clinical characteristics and risk scores of CA9 in TSCC
CA9 expression levels are shown in Figure 3A. The transcription levels of CA9 in the tumours were significantly higher than those in normal tissues. According to the Kaplan-Meier analysis method, a high expression level of CA9 in TSCC was associated with poorer overall survival (P = .0017; Figure 3B). Samples, containing complete clinical information (age, gender, grade and stage), were divided into groups of low and high around the median of each CA9 expression.
Univariate and multivariate Cox regression analyses indicated that CA9 expression correlates with tumour grade (Table 2). In view of the prognostic value of CA9 in TSCC, a nomogram was constructed for predicting 3-and 5-year survival. The result illustrated that CA9 expression (group) shares the largest contribution to overall survival, followed by the grade and stage ( Figure 3C).

| Identify key modules associated with patients' grade based on the expression levels of CA9 in TSCC
Based on the results shown in Table 2, the weighted gene co- were identified ( Figure 4E).

| D ISCUSS I ON
Carbonic anhydrase 9 is a type of zinc metalloenzyme that is able to catalyse the reversible hydration of carbon dioxide. 19 It can control the intracellular pH and protect cancer cells from hypoxia-induced apoptosis. 20 In many malignancies, CA9 is highly related to hypoxia and is regulated by the transcription factor HIF-1α. 21 Knockdown of CA9 abolished the chemoresistance resulting from ZEB1 overexpression and prevented maintenance of pHi mediated by overexpression of ZEB1 in tongue cancer cells. ZEB1 and CA9 protein expression was associated with a poor prognosis for patients in TSCC (Table   S2). 22 Meanwhile, high CA9 levels reduce the chemosensitivity and facilitate the survival of various cancer cells by enhancing the acidification of the cancer microenvironment. 23 In renal clear cell carcinoma, CA9 could be an independent prognostic tumour marker and a diagnostic biomarker. Similarly, our analysis suggested that CA9 expression level was associated with prognosis and could regulate signalling pathways in TSCC, including hypoxia and cell cycle. In oral squamous cell carcinoma, a high CA9 mRNA level was associated with an increased risk of locoregional recurrence, while the expression level of CA9 was related to tumour grade and clinical stages in TSCC patients. 24 However, TCGA data showed that there was no significant difference between the mRNA expression level of CA9 and the prognosis of patients with cancers occurring in the lips and the other sites of oral cavity (Table S2). Therefore, our research suggested that CA9 could be used as a potential predictor of tumour grade and prognosis in TSCC. The regulation of T cells might also be due to the expression of CA9specific protein epitopes in TSCC.
Using COX regression model, we found that CA9 was highly related to tumour grade in tongue cancer patients. Thus, we analysed tumour grade co-expressed genes in high and low expression groups using WGCNA. We found that genes, which correlated with tumour grade, were highly enriched in cell differentiation and negative regulation of cell proliferation using GO analysis. In sinonasal papilloma, CA9 expression correlated with cancer cells differentiation and proliferation and emerged as a marker of cancer recurrence. 26 In pancreatic ductal adenocarcinoma, CA9 expression was significantly lower in cases with well-differentiated adenocarcinoma. 27 It is possible that CA9 influenced TSCC grade by regulating neoplastic cell differentiation and proliferation. In addition, hug genes, such as DSC2, CSTA, PKP1 and DYNLL1, also appear to be involved in this process. In the esophageal squamous cell carcinoma, DSC2 displayed a role in tumour differentiation, metastasis and patients' clinical outcomes. 28 PKP1 was inversely associated with the histological grade of human primary oropharyngeal squamous cell carcinoma. 29 It could be hypothesized that the role of CA9 in cancer cell differentiation may also be a result of the activation of multiple co-expressed genes that affect TSCC differentiation.
In conclusion, CA9 high expression correlates with poor prognosis and tumour grade in TSCC. The activation of tumour-associated pathways and cell differentiation-related genes plays a crucial part in these biological processes. To test these hypotheses, it will be necessary to undertake cytological and animal experiments studies in the future. These investigations are currently under way.

ACK N OWLED G EM ENTS
This work was supported by the National Natural Science

CO N FLI C T O F I NTE R E S T
We have no conflicts of interest.

AUTH O R CO NTR I B UTI O N S
Guan Chenyu and Daiqiao Ouyang involved in most of the experiments, collected and analysed data, and drafted the manuscript. * Yongjie Qiao, Kan Li, Guangsen Zheng, Xiaomei Lao and Sien Zhang assisted in analysed data and drafted the manuscript. Guiqing Liao and Yujie Liang conceived and designed the experiments, oversaw the collection of results, data interpretation and wrote the manuscript. All authors read and approved the paper.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data included in this study are available upon request by contact with the corresponding author.