Inverse Correlation of miR-27a-3p and CDH5 Expression Serves as a Diagnostic Biomarker of Proliferation and Metastasis of Clear Cell Renal Carcinoma

Background: Cadherin-5 (CDH5) is aberrantly expressed in a variety of human cancers and plays an important role in angiogenesis. But its role in renal clear cell carcinoma (ccRCC) remains to be understood. Methods: The dysregulation of CDH5 expression in ccRCC and its association with clinicopathological characteristics were analyzed using the TCGA database. A meta-analysis was performed to verify the alteration of CDH5 expression in ccRCC using the GEO database. Quantitative RT-PCR and immunohistochemical staining were applied to assess the transcriptional and protein levels of CDH5. TargetScan and Tarbase were employed to predict the miRNAs with the potential to target mRNA of CDH5. Results: The mRNA level of CDH5 in ccRCC was signicantly higher than in normal tissue. CDH5 mRNA expression could therefore serve as a potential diagnostic biomarker for ccRCC (AUC=0.844). However, the reduced CDH5 transcription levels were signicantly correlated with patients in the T3-4 stage, lymph node and distant metastasis, as well as with a worse clinical outcome. We further observed that CDH5, at the protein level, was almost absent in ccRCC samples. In addition, a few databases screen showed that mir-27a-3p, is a highly conserved miRNA targeting CDH5. The expression of mir-27a-3p was signicantly elevated in ccRCC tissues in contrast to normal tissues. Importantly, it was positively associated with the T3-4 stage and M stage, respectively, suggesting that the expression level of mir-27a-3p could serve as a diagnostic biomarker for ccRCC (AUC=0.775). Conclusion: Our data suggest that the reduced translational levels of CDH5 in ccRCC was related to the overexpression of mir-27a-3p. The higher mir-27a-3p and lower CDH5 expression signicantly correlated with advanced clinical stages for ccRCC patients.


Background
The current incidence of renal cell carcinoma (RCC) increases by 2% per year. [1] By the end of 2020, it is estimated that there will be 73,750 new cases of RCC and about 14,830 new deaths [2,3]. The common type of RCC is clear cell renal cell carcinoma (ccRCC), and the mortality rate of ccRCC was about 4.1 per 100,000. [4] Partial and radical nephrectomy is the main method of therapy for localized ccRCC. However, even after surgery , 30% of the patients have distant metastases within ve years, which is the main cause of treatment failure. [5] ccRCC had been shown largely refractory to conventional cytotoxic chemo and hormone-therapy, while the effect of radiotherapy is a limited due to a low response rate. [6] It is necessary to identify new molecular biomarkers for early diagnosis and precise treatment.
Cadherins are cell adhesion molecules that are responsible for the normal structure of epithelial tissues. They also play a vital role in cell differentiation and tissue remodeling. [7] The alteration of cadherin expression leads to tumor initiation and progression. [8] The aberrant expression of E-cadherin (CDH1) and N-cadherin (CDH2) is closely related to the epithelial-to-mesenchymal transition (EMT) in various tumors, such as lung, colorectal and prostate cancer and others. [9][10][11] Expression of these cadherins also serves as reliable prognostic biomarkers.
CDH5, also known as VE-cadherin, was mainly expressed in the inter-endothelia junction. The CDH5 protein participates in maintaining the integrity of endothelial and vascular tissues, functioning as a regulator in signaling of endothelial cell during vessel morphogenesis and vascular development. [12] Therefore, the dysregulation of CDH5 expression promotes tumor progression by stimulating angiogenesis. [13] In breast cancer, the levels of CDH5 protein expression can be used to distinguish breast cancer patients from non-breast cancer patients and its high expression is predictive of metastasis in breast cancer. [14] In gastric cancer, the up-regulation of CDH5 protein expression is associated with poor outcomes for patients. [15] The altered expression of CDH5 protein in glioblastoma multiforme is suggested to contribute to the tumorigenic process by facilitating new blood vessel formation. [16] In highly aggressive melanoma cells, CDH5 protein acts as a co-regulatory element of an EphA2 complex to control tumor neovascularization. [17] However, the clinical signi cance of CDH5 mRNA expression levels has not been studied yet.
MiRNAs, are highly conserved non-coding single-stranded RNAs of 22 nucleotides length, which recognize the 3′ untranslated region (3′-UTR) of target mRNAs and either prevent their translation or facilitate their degradation, depending on partial or complete complementarity to the targeted nucleotide sequence. In other words, miRNAs regulate gene expression at the post-transcription level. [18] Several members of the CDH family genes has already been shown to be regulated by miRNAs and this mechanism was shown to be involved in cancer progression and metastasis. For example, CDH1 is a direct target for MiR-217, it was shown that miR-217 overexpression can enhance gastric cancer cell proliferation and reduce exosome CDH1 levels. [19] In non-small cell lung cancer, miR-124 can inhibit CDH2 expression and regulate the EMT. [20] The clinical signi cance of CDH5 mRNA expression and its control by a miRNA expression in ccRCC has not been shown yet.
In this study, we aim to analyze the expression of CDH5 at the gene and the protein level and its regulation by expression of miRNA-27a-3p in ccRCC. We also evaluated their potential signi cance role as diagnostic and prognostic biomarkers. In addition, the expressional levels of the CDH5 gene and miRNA-27a-3p were analyzed to correlate with the clinical characteristics of patients with ccRCC.

Materials And Methods
Analysis of the TCGA database on the transcriptional level of CDH5 The TCGA database was downloaded from the UCSC Cancer Genomics Browser (https://xena.ucsc.edu/). The RNA sequencing data for 533 ccRCC patients and 72 noncancerous tissues together with complete clinical information on age, gender, pathologic stage and others for all subjects from the TCGA database were analyzed. Subsequently, the correlation between the mRNA level of CDH5 and clinical characteristics of ccRCC patients was analyzed.
A meta-analysis based on the GEO datasets The alteration of CDH5 mRNA level in ccRCC was further investigated using the GEO database, with search terms as follows: 1) "renal cancer", 2) "kidney OR renal AND cancer OR carcinoma OR tumor OR neoplasm* OR malignant* OR ccRCC OR RCC OR RCCC". The microarrays data was used to compare the CDH5 mRNA level in ccRCC and normal tissues, by meta-analysis. The criteria of the microarrays data inclusion into the analysis were: 1) more than three tumor samples and three normal controls on the microarray, 2) the human tissues served as a source for microarrays. The STATA 12 software was used for meta-analysis.

Immunohistochemistry (IHC) staining assay
An SP-9000 detection kit (ZSGB-Bio, Beijing, China) was used for immunohistochemical staining on para n-embedded tissue slices. Tissues were depara nized and rehydrated in fresh xylene and gradient concentrations of alcohol, followed by antigenic thermal repair, followed by incubation with endogenous peroxidase blockers for 10 min. Primary antibodies against CDH5 (MABT129, Sigma, USA, 1:350 dilution) were added to the samples and incubated at 37° for 1.5h. Next, samples were incubated for 10 min with secondary antibody coupled to horseradish peroxidase-labeled streptomycin for 10 min, followed by incubation with the chromogenic horseradish peroxidase substrate 3,3′diaminobenzidine (DAB) reagent (ZLI-9018, ZSGB-BIO, Beijing, China). Hematoxylin was used for counterstaining. Images were acquired with an Olympus C-5050 microscope (Olympus, Japan).
All sections were evaluated by two pathologists independently on the clinicopathological characteristics of the samples. The ambiguous cases were discussed and classi ed based on consensus. Each slice was randomly divided into 5 areas and evaluated in multiples elds at 20 × 10 magni cation. The scoring criteria were as previously described. [21] Prediction of miRNAs potentially targeting the CDH5 mRNA Potential miRNAs targeting the CDH5 mRNA were predicted using three publicly available programs: TargetScan (http://www.targetscan.org), Tarbase (http://carolina.imis.athena-innovation.gr) and miRBase (http://www.mirbase.org/). [22][23][24] Statistical analysis Statistical analysis was performed using the Statistical Program for Social Sciences (SPSS) 22.0 software. GraphPad Prism 6 software was employed for data imaging. The students' t-test was used to examine the correlation between CDH5 mRNA and miRNA expression data and clinicopathological parameters extracted from TCGA data on ccRCC. The signi cance of a difference between CDH5 mRNA expression values and the protein expression in cancer and normal tissues was evaluated with the use of the paired t-test. Diagnostic value for CDH5 mRNA expression in ccRCC was obtained after the analysis of receiver operating characteristic (ROC). A cut-off point was de ned as the maximum Youden index. A P-value <0.05 was considered to be statistically signi cant.

Results
The expression of CDH5 mRNA is altered in ccRCC.
We compared the mRNA expression values for CDH5 in 533 samples of ccRCC and 72 normal samples in the TCGA database. We detected that CDH5 mRNA expression was elevated signi cantly in ccRCC as compared to normal tissues (p<0.001; Figure 1a).
To further explore the consistency of data, regarding the aberrant expression of CDH5 mRNA, we performed qRT-PCR analysis for CDH5 expression with cell lines established from kidney and ccRCC tissues as a source. We showed that, in contrast to human cells 293T, the expression level of CDH5 mRNA in renal carcinoma cells was signi cantly higher (p<0.05, Figure 1b). We obtained consistent results comparing the 15 paired ccRCC and adjacent normal tissues in qRT-PCR analysis (p=0.0019, Figure 1c). This was con rmed further by a meta-analysis of cDNA microarrays using the GEO database. Totally 30 microarray studies were included for the meta-analysis, which in total contained 1,141 ccRCC and 651 normal samples. In 28 out of 30 microarrays, a higher value was observed in ccRCC tissues as compared to normal tissues. To obtain the result we used a random-effects model, which provides the pooled Standard Mean Difference (SMD) value as 1.07, corresponding to 95% of con dence interval (CI, 0.96-1.19, Figure 2a), and less than 0.001(I2=86.8%) the p-value, after the heterogeneity test. The sensitivity analysis indicated that no single study led to signi cant bias in overall merger results (Figure 2b). In addition, no publication bias was found in the study (Begg's test: p=0.116, Figure 2c). The relevant information related to these 30 microarrays is provided in Table 1.
Surprisingly, immunohistochemistry (IHC) analysis for CDH5 protein expression of ccRCC samples demonstrated the opposite results to CDH5 mRNA expression level in ccRCC. As shown in Figure 3a, we did not observe CDH5 in the cell membrane or cytoplasm of ccRCC, On the contrary, CDH5 was mainly distributed on the proximal convoluted tubules of adjacent normal kidney tissue samples, even in the distal convoluted tubules of normal kidney samples while in ccRCC samples CDH5 protein expression was not been identi ed. Furthermore, the positive expression rate for CDH5 in non-tumorous renal tissues was much higher than that in the ccRCC tissues, and CDH5 staining was strongly positive in almost all non-tumor specimens (Figure 3b).
Low CDH5 mRNA expression corresponds to a tumor stage/metastasis stage of ccRCC.
We analyzed whether the CDH5 mRNA expression changes with ccRCC development (Table 2). To our surprise, we found that the transcribed CDH5 was closely correlated with patients in T3-4 stages (p=0.016), N stage (p=0.008), and M stage (p=0.001), respectively. This means that CDH5 expression tends to decrease with the increase of tumor size and metastasis, however, CDH5 expression does not fall lower than that in normal tissues.
Although the expression of CDH5 protein normal tissue can be clearly observed in our IHC results, the CDH5 protein expression levels did not signi cantly correlate with the clinical parameters (Table 3). We infer, that this result might be due to our small sample size.
The expression level of CDH5 mRNA can be used as a molecular marker for the diagnosis and prognosis of ccRCC development.
A ROC curve built based on the analysis of CDH5 mRNA expression in 533 cases of ccRCC and 72 cases of normal kidney tissue (Figure 4a). This allowed to estimate the area under the curve (AUC) value as equal to 0.844 ((p<0.001), which suggests that CDH5 mRNA expression could serve as a molecular marker to distinguish the tumor tissues from the non-tumor ones, (the sensitivity and speci city as 0.73 and 0.931, respectively). In addition, we have evaluated the overall survival (OS) and relapse-free survival (RFS) time using the Kaplan-Meier plotter, which revealed that low CDH5 mRNA expression was significantly associated with the shorter OS (Figure 4b, p=0.013). While, no statistically signi cant difference was observed with regards to RFS time in ccRCC (Figure4c, p=0.133).
MiRNA 27a-3p is involved in the control of CDH5 mRNA expression in ccRCC.
In order to investigate the discrepancies between low CDH5 protein and high CDH5 mRNA expression in ccRCC, we searched for putative miRNAs targeting CDH5 at the post-transcriptional level using TargetScan, Tarbase, and miRbase data. We found that CDH5 mRNA contains a seed site for miR-27a-3p (Figure 5a), which suggests that miR-27a-3p may be involved in the regulation of CDH5 mRNA expression. Previous studies have shown that mir-27a can promote cell proliferation and metastasis in ccRCC. [25] We downloaded the data regarding miR-27a-3p expression in TCGA and compared 240 ccRCC with 70 normal cases. The miR-27a-3p was overexpressed in ccRCC tumors compared to normal (Figure 5b, p<0.001). Although miR-27a-3p may not directly affect the transcription level of CDH5, it may be signi cantly related to the down-regulation of CDH5 protein level in ccRCC. We analyzed the relationship between miR-27a-3p miRNA and clinical information. We found that there was a similar to CDH5 mRNA expression alterations trend in ccRCC tissues (9.54±0.53) and in the normal kidney tissues (8.96±0.63), with statistical signi cance (p<0.001). Interestingly, in contrast to the CDH5 mRNA expression level, the miR-27a-3p miRNA expression level was positively correlated with stage T3-4 (p=0.026), stage M (p<0.001), and clinicopathological stage (p<0.007) in patients with ccRCC (Table 4). The AUC value, based on the ROC curve analysis 0.775 (p<0.001), for the CDH5 miRNA also allowed to distinguish tumor from non-tumor tissue (Figure 6a). Therefore, we inferred that miR-27a-3p and CDH5 play an important role, perhaps jointly, in the occurrence and development of ccRCC. The analysis of miR-27a-3p expression, using the Kaplan-Meier plotter, did not allow to distinguish between ccRCC and normal tissue samples in regards to either OS or RFS times (Figure 6b, 6c).

Discussion
Previously it was shown that high CDH5 expression is associated mainly with angiogenesis during cancer. [26][27][28] However, we found a more complicated relationship between CDH5 mRNA expression and cancer progression. The expression of CDH5 mRNA is increased only at the early stage of ccRCC. During the progress of ccRCC, we detected a gradual loss of CDH5 mRNA expression. We infer that high CDH5 mRNA expression may be associated with the initiation of ccRCC, but further tumor proliferation and metastasis is associated with a gradual loss of CDH5 expression. We suggest that the loss of CDH5 mRNA expression may be due to increased miR-27a-3p expression. Post-transcriptional control of CDH5 expression is under intensive investigation currently. [29] Other miRNAs reported being functional during ccRCC development, are miR-106a-5p, which was shown to bind and destroy PAK5 mRNA in ccRCC, resulting in a decrease of expression of an oncogene PAK5 protein. [30] Mir-148a targets AKT2 to inhibit cell growth, colony formation, and tumor growth of ccRCC. [31] We found that rise in miR-27a-3p expression is associated with the higher T stage and metastasis in ccRCC, which coincides with the low expression of CDH5 mRNA in ccRCC with the same clinicopathological features. We suggest that miR-27a-3p affects the occurrence and development of the disease by targeting CDH5 mRNA. It has been shown that miR-27a-3p is an important regulatory factor in a variety of cancers. It plays a carcinogenic role in renal cell carcinoma, cervical cancer, gastric adenocarcinoma, and breast cancer. [25,[32][33][34] Mir-27a-3p has been identi ed to promote tumor invasion and metastasis in gastric, colon, pancreatic, and nasopharyngeal cancers. [35][36][37][38] On the other hand, a tumor suppressor role has also been shown for miR-27a-3p in some cancers. For instance, it inhibits the proliferation of non-small cell lung cancer cells by targeting HOXB8. [39] In hepatocellular carcinoma (HCC), the miR-27a-3p inhibits tumor metastasis by binding to 3'UTR of CDH5 mRNA, thus, inhibiting EMT. [40]. We also observed a negative correlation between CDH5 mRNA and miR-27a-3p expression in ccRCC. But we infer the role of miR-27a-3p as a proto-oncogene in ccRCC. The role of CDH5 in ccRCC pathogenesis needs further investigation.
CDH5 plays a central role in contact inhibition by transmitting information on cell density to the nucleus. Thus, it is the main guardian of vascular integrity. [41] Previous studies show that the expression of CDH5 has a profound in uence on the invasion and metastasis in malignant tumors. CDH5 can promote tumor angiogenesis and tumor cell proliferation. [42] CDH5 mRNA is overexpressed at the early stages of cancer, which is consistent with our study. [43] Subsequently, the CDH5 mediated process of tumor cells in ltration into adjacent tissues facilitates penetration of the vascular endothelial barrier, and systemic dissemination tumor cells through blood vessels and lymphatics, thus contributing to acquisition of an EMT phenotype. [44] CDH5 can also affects the EMT process indirectly. [45] These studies support our ndings and warrant further investigation of the interrelation between CDH5 mRNA and miR-27a-3p in ccRCC pathogenesis.
In the current study, we also found an inverse relationship between CDH5 and miR-27a-3p expression in ccRCC clinical samples. In addition, our exploration of the TCGA database, TargetScan, TarBase and miRbase data, provide compelling evidence supporting the relation between miR-27a-3p and CDH5. In the past, most of the studies on tumor markers were devoted to the study of single mRNAs or miRNAs alone. Our research aims to study the effect of interaction between mRNA and miRNA, to investigate possible new approaches to nd clinically applicable tumor biomarkers of ccRCC.

Conclusions
Decreased CDH5 mRNA and increased miR-27a-3p expression was found to correlate signi cantly with the advanced stages of tumor development in ccRCC patients. The mir-27a-3p and CDH5 mRNA expression can be used as biomarkers and targets during diagnosis and treatment for ccRCC patients. The mir-27a-3p/CDH5 mRNA axis in the pathogenesis of ccRCC requires further investigation. The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.  The ROC curve was used to assess the diagnostic e cacy of CDH5 mRNA, the prognostic value of CDH5 mRNA was evaluated by OS and DFS. (a). The ROC curve for evaluating the diagnostic performance of CDH5 in 533 cases of ccRCC and 72 cases of normal kidney tissues based on the TCGA database. The AUC was 0.844. (b)(c). Potential prognostic value of CDH5 mRNA expression in ccRCC patients.

Figure 6
The ROC curve was used to assess the diagnostic e cacy of has-27a-3p, prognostic value was evaluated by OS and DFS. (a) The ROC curve for evaluating the diagnostic performance of miRNA in 240 cases of ccRCC and 70 cases of normal kidney tissues based on the TCGA database. The AUC was 0.775. (b)(c). Potential prognostic value of CDH5 miRNA expression in ccRCC patients in the TCGA database.