miR‐363‐3p induces EMT via the Wnt/β‐catenin pathway in glioma cells by targeting CELF2

Abstract In our previous study, we reported that CELF2 has a tumour‐suppressive function in glioma. Here, we performed additional experiments to elucidate better its role in cancer. The expression profile of CELF2 was analysed by the GEPIA database, and Kaplan–Meier curves were used to evaluate the overall survival rates. Four different online databases were used to predict miRNAs targeting CELF2, and the luciferase assay was performed to identify the binding site. The biological effects of miR‐363‐3p and CELF2 were also investigated in vitro using MTT, Transwell, and flow cytometry assays. Western blotting, qPCR, and TOP/FOP flash dual‐luciferase assays were performed to investigate the impact of miR‐363‐3p and CELF2 on epithelial‐to‐mesenchymal transition (EMT) and the Wnt/β‐catenin pathway. The effect of miR‐363‐3p was also tested in vivo using a xenograft mouse model. We observed an abnormal expression pattern of CELF2 in glioma cells, and higher CELF2 expression correlated with better prognosis. We identified miR‐363‐3p as an upstream regulator of CELF2 and confirmed its direct binding to the 3′‐untranslated region of CELF2. Cell function experiments showed that miR‐363‐3p affected multiple aspects of glioma cells. Suppressing miR‐363‐3p expression inhibited glioma cell proliferation and invasion, as well as promoted cell death via attenuating EMT and blocking the Wnt/β‐catenin pathway. These effects could be abolished by the downregulation of CELF2. Treatment with ASO‐miR‐363‐3p decreased tumour size and weight in nude mice. In conclusion, miR‐363‐3p induced the EMT, which resulted in increased migration and invasion and reduced apoptosis in glioma cell lines, via the Wnt/β‐catenin pathway by targeting CELF2.

been achieved in the therapy, especially in the treatment of high-grade glioma. To date, most patients with advanced glioma receive standard care that includes surgical resection followed by temozolomide plus radiation, but the 5-year overall survival remains poor. 4 The pathogenesis of glioma is multifactorial, and it is considered that both genetic and environmental factors contribute to its emergence, as in the cases of other complex diseases. 5 Identification of the underlying pathogenic mechanisms involved in the initiation and progression of glioma is critical for developing effective treatments.
CELF2 is an RNA-binding protein belonging to the CELF family of proteins, which binds to double-or single-stranded RNA and regulates the expression of thousands of transcripts. 6,7 Comprehensive genomic analysis has shown that RNA-binding proteins, including CELF2, are predominantly downregulated in tumours compared with their expression in normal tissues and play a crucial role in tumour development. 8 Guo et al. 9 showed that CELF2 inhibits ovarian cancer progression by stabilizing FAM198B. Piqué et al. 10 pointed out that CELF2 was targeted by promoter hypermethylation-associated transcriptional silencing in human cancer, and the presence of CELF2 hypermethylation was associated with shorter overall survival in patients with breast cancer. CELF2 was also shown to suppress nonsmall cell lung cancer growth through the PI3-K pathway. 11 MicroRNAs (miRNAs) are a family of endogenous, singlestranded, small non-coding RNAs that act as post-transcriptional regulators of gene expression. 12 By binding to the 3′-untranslated regions (3-UTR) of target mRNAs, miRNAs repress translation or accelerate the degradation of mRNA and thus regulate biological processes, including cellular differentiation, proliferation, and apoptosis. [13][14][15] Accumulating evidence has shown that dysregulation of miRNA levels plays an important role in various types of cancer.
Several studies on the effects of miR-363-3p have been conducted in various cancers, and they reached divergent conclusions. Mohamed et al. reported that miR-363-3p increased chemoresistance of an ovarian cancer line to taxane. 16 Dorbna et al. 17 demonstrated that upregulated miR-363-3p prevented apoptosis and promoted growth of leukemic cells in vitro. In contrast, miR-363 was shown to be a tumour suppressor in lung cancer, osteosarcoma, and colorectal cancer. [18][19][20] At present, only two studies examined the role of miR-363-3p with glioma cells. Bi et al. 21 reported that HNF1A-AS1 promotes glioma cell growth by acting as a miR-363-3p sponge. Xu et al. 22 reported that miR-363-3p promotes cell proliferation, protects against apoptosis, and enhances invasion by directly targeting PDHB in glioma cells. The mechanism behind miR-363-3p modulating glioma cells needs further investigation.
Hence, we focussed on the relationship between CELF2 and miR-363-3p in glioma cells. We identified miR-363-3p as an upstream regulator of CELF2 expression, evaluated its biological functions in vitro and in vivo, as well as investigated the pathways involved in these processes. For the first time, we provided evidence that miR-363-3p induces the epithelial-to-mesenchymal transition (EMT) in glioma cells via the Wnt/β-catenin pathway by targeting CELF2. Our findings suggest that miR-363-3p may be a novel therapeutic target for glioma therapy.

| Western blot
Cell lines or tissues were lysed in the RIPA buffer (Saierbio, Tianjin, China). Total protein in each sample was collected and measured.
Sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) was performed according to the manufacturer's instructions.
Total protein was degenerated by boiling for 5 min with a loading buffer and cooled to room temperature. Protein samples were loaded into SDS-PAGE wells for electrophoresis at 100 V. After transferring, the Hybond nitrocellulose membrane was blocked with the western blotting buffer for at least 60 min before incubation with the primary antibody. After the exposure to the primary antibody, the membrane was washed three times with the western blotting buffer and incubated with the secondary antibody. GAPDH expression was used as loading control. Proteins were detected using chemiluminescence substrates.

| MTT assay
Cells in the logarithmic growth phase were treated by trypsin and centrifuged to obtain a compound cell suspension of 1 × 10 4 /ml.

| Cell apoptosis assay
The cells were trypsinized and centrifuged at 1500 rpm for 3 min.

| Statistical analysis
Statistical significance analysis was performed using the GraphPad Student's t test was used to determine the significance of differences between the two groups in pairwise comparisons. All results are shown as the mean ±standard error of the mean. Effects were considered statistically significant if p < 0.05. All assays were repeated in triplicate.

| Low expression of the CELF2 gene in GBM samples
We analysed CELF2 expression levels and mutation rates in various cancer types. As shown in Figure 1A, 1B, CELF2 was expressed at different levels in cancer and normal tissues but was not extensively mutated in cancer. Figure 1C shows lower CELF2 expression level in GBM samples than in LGG and normal brain tissue samples.
The mean expression level in LGG was lower than that in normal brain tissue, but the difference was not statistically significant. The Kaplan-Meier survival analysis showed that higher CELF2 expression was associated with better prognosis ( Figure 1D).
We screened out five candidate miRNAs that were revealed in all four subsets (Figure 2A). We measured the expression of CELF2 and five candidate miRNAs by qRT-PCR in 18 glioma samples collected from the Department of Neurosurgery of the Second Affiliated Hospital of the Hebei Medical University ( Figure 2B). We calculated the Spearman's rank correlation coefficient to analyse the relationship between expression levels of CELF2 and each of the candidate miRNAs. We found that expression levels of miR-96 and miR-363-3p had a significant negative relationship with that of the CELF2 mRNA.
After referring to the literature, we selected miR-363-3p for further investigation ( Figure 2C). We identified one conserved binding site for miR-363-3p in the 3′-UTR of CELF2 by TargetScan and constructed a vector for miR-363-3p wt/mut ( Figure 2D). To determine whether a targeted combination existed, we conducted the dual-luciferase assay. As shown in Figure 2E, increased expression of miR-363-3p inhibited luciferase activity. When the CELF2 3′-UTR was mutated, the miR-363-3p mimic no longer affected luciferase activity ( Figure 2E). Hence, we confirmed that miR-363-3p could directly bind to the 3′-UTR of CELF2 mRNA. Furthermore, when we increased the expression level of miR-363-3p in glioma cell lines, we observed decreased level of CELF2 mRNA ( Figure 2F). These results indicated that miR-363-3p suppressed post-transcriptionally CELF2 mRNA expression.

| Effects of miR-363-3p on biological properties of glioma cell lines
To elucidate the role of miR-363-3p in glioma cells, we performed a series of in vitro experiments. Figure 3A shows that the expression of miR-363-3p could be efficiently modulated by transfecting glioma cells with mimic or ASO. We found that miR-363-3p mimic signifi-

| Downregulation of CELF2 expression eliminates the effects of ASO-miR-363-3p on glioma cell lines
From the above experiments, we concluded that miR-363-3p is critical for glioma development. We subsequently designed rescue assays to confirm that miR-363-3p regulated glioma cell biological properties by targeting CELF2. CELF2 siRNAs were transfected into glioma cells to specifically inhibit the increase in CELF2 expression induced by ASO-miR-363-3p. Figure 4A shows that CELF2 siRNA effectively suppressed CELF2 expression. Furthermore, treatment with CELF2 siRNA completely abolished the effects of ASO-miR-363-3p on U87/U118MG cells ( Figure 4B-4D). In particular, treatment with CELF2 siRNA significantly restored the S arrest induced by ASO-miR-363-3p in U87 cells. However, this phenomenon was not observed in U118MG cells ( Figure 4E). These results showed that miR-363-3p influenced the proliferation, invasion, and apoptosis of glioma cells by targeting CELF2.

| miR-363-3p induces EMT and activates the Wnt/β-catenin pathway by targeting CELF2
It is well established that the phenomenon EMT is strongly linked with tumorigenesis. We explored whether miR-363-3p affected the EMT. The expression levels of CELF2 and EMT markers were examined by western blotting and qPCR. We observed that E-cadherin

| ASO-miR-363-3p inhibits glioma cell growth in vivo
To verify whether miR-363-3p affects glioma cells in vivo, we screened U87 cells transfected with ASO-miR-363-3p (experimental group) or ASO-NC (control group) and injected them into the back of nude mice to construct U87 xenograft models ( Figure 6A). The experimental and control groups each contained nude mice of similar size and weight. The growth rate of tumours in the experimental group was significantly lower than that in the control group ( Figure 6B, 6C). Four weeks later, nude mice were euthanized, and the tumours were dissected out. The volume and weight of tumours in the experimental group were significantly lower than those of the tumours in the control group ( Figure 6D). IHC staining demonstrated that xenograft tumours derived from the ASO-miR-363-3p transfected cells had significantly higher CELF2 expression than tumours in the control group ( Figure 6E). Thus, our data showed that ASO-miR-363-3p inhibits glioma cell growth in vivo.

| DISCUSS ION
Glioma is the most common primary brain tumour that accounts for approximately 30% of all types of tumours in the central nervous system. 23 miRNAs have been shown to function as tumour suppressors or proto-oncogenes during tumorigenesis in various human cancers. 24,25 In our previous study, we demonstrated a tumoursuppressive function of CELF2 in glioma cells. 26 This study was designed to uncover the role of miR-363-3p as a negative modulator of CELF2 expression in the development of glioma.
By using an online database, we found that CELF2 was expressed at low levels in most cancers, whereas its mutation rate was low. We hypothesized that CELF2 expression is suppressed posttranscriptionally during tumorigenesis. According to the data in the TCGA database, GBM had lower CELF2 expression levels than LGG or normal brain tissue. Furthermore, CELF2 levels negatively correlated  Figure 3B, the proliferation of U87 and U118MG cells was decreased after their transfection with ASO-miR-363-3p. In the experiments illustrated in Figure 3E, we found that the proportion of cells in the S phase was significantly increased in the cell cycle after transfection with ASO-miR-363-3p, which meant that more cells were stuck in the S phase and could not proceed to the next phase to complete mitosis. Given that, we concluded that cell the results in U87 and U118MG cells were contradictory ( Figure 4E).
One explanation of this discrepancy could be that U87 and U118MG have different mechanisms to regulate cell cycle. In U87 cells, CELF2 plays a key role in regulating the cell cycle. CELF2 siRNA significantly suppressed CELF2 expression and thereby profoundly affected the cell cycle. However, the cell cycle of U118MG cells might be controlled by another protein, which is also targeted by miR-363-3p.
Thus, ASO-miR-363-3p led to the arrest in the S-phase, but treatment with si-CELF2 could not reverse this effect.
In our previous study, we showed that miR-95-3p promoted cell proliferation and migration but inhibited apoptosis by targeting CELF2, 26  Accumulating evidence indicates that EMT is closely involved with the metastatic ability of cancer cells, as it confers the loss of cell-to-cell adhesion and gain of migratory properties. 30 In fact, EMT participates in various normal physiological processes, including embryonic development, organ fibrosis, tissue regeneration, but also regulates tumour stemness, tumour initiation and malignant progression, cell migration and invasion, intravasation to the blood, metastasis, and resistance to therapy. [31][32][33] When EMT is activated, the expression of E-cadherin decreases. In addition, increased ex-  CELF2, as an RNA-binding protein, usually exerts its function of regulating protein transcription by binding mRNAs. Yeung et al. 11 reported that CELF2 interacts with PREX2 and reduces the association of PREX2 with PTEN. Guo et al. 9 showed that CELF2 expression increased the stability of its target, FAM198B, by binding to the AU/Urich elements in FAM198B 3′-UTR in ovarian cancer. Therefore, it is reasonable to believe that there might be a protein downstream of CELF2, participating in this regulation process. Direct evidence is needed to demonstrate how CELF2 interacts with the EMT or Wnt/ β-catenin pathway.
In conclusion, our experiments in glioma cell lines showed that miR-363-3p induced the EMT, which resulted in increased migration and invasion and reduced apoptosis via the Wnt/β-catenin pathway by targeting CELF2. This evidence suggests that miR-363-3p may be a novel therapeutic target for glioma.

ACK N OWLED G EM ENTS
We wish to thank Weisong Wang, Yue Yang, Pengxiang Zhang, Xiaofeng Cheng, and Hongbo Tui for their help.

CO N FLI C T O F I NTE R E S T
The authors confirm that there are no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.