GAS5 Attenuates Malignant Progression of Glioma Stem-like Cells via Regulating E-cadherin

Background: It is commonly understood that glioma stem-like cells (GSCs) play a vital role in the malignant progression of glioma. Recent studies have reported that long non-coding RNAs (lncRNAs) closely associate with glioma development, however, the underlying molecular regulatory mechanisms on GSCs have not been fully clari�ed. Methods: LncRNA GAS5 expression level was analyzed by bioinformatics and qRT-PCR assays. GSCs were separated from glioma tissues and identi�ed using immuno�uorescence and �ow cytometry. Cell proliferation ability was measured by EdU assays. Cell invasion and migration ability was evaluated by transwell assays. Cell apoptosis was measured by �ow cytometry. Subcutaneous tumorigenesis assays were performed to explore the GSCs growth in vivo. Luciferase reporter assays were used to identify the direct molecular interaction. E-cadherin expression level was measured by qRT-PCR assays and western blot. Results: In this study, we established two highly malignant glioma stem-like cells from clinical surgical specimens, and found that lncRNA GAS5 was downregulated in GSCs and high-grade glioma tissues, and showed a correlation with patient survival. Functional assays showed that depleting GAS5 promoted the proliferation, invasion, migration, stemness and tumorigenicity, and inhibited apoptosis of GSCs. Mechanistically, GAS5 directly sponge with miR-23a which acts as an oncogene by regulating E-cadherin (CDH1). In addition, rescue experiments demonstrated that GAS5 modulate both the expression and function of E-cadherin via the dependent manner of miR-23a. Conclusions


Introduction
Gliomas are the most commonly occurring primary intracranial tumors in the central nervous system, and glioblastoma multiforme (GBM) is the most prevalent, malignant and aggressive in gliomas 1 . The overall survival and prognosis of GBM are poor, although the therapeutic regimens have been optimizing continuously 2 . Many researchers attribute the poor prognosis to the glioma stem-like cells (GSCs) which harbor remarkable cellular heterogenicity and are responsiable for tumor progression including initiation, metastasis, treatment resistance and so on 3 . But the molecular mechanisms of GSCs have not been elucidated, which may be urgently needed for improving glioma treatment e cacy 3,4 .
Long non-coding RNAs (lncRNAs) have been revealed to play critical roles in various pathophysiological processes via activating or silencing the corresponding on transcription and/or post-transcription level 5 .
The increasing evidence supports that lncRNAs are abnormally expressed in various tumors and are involved in multiple malignant phenotypes of tumors 6 . Growth Arrest-Speci c 5 (GAS5), located at chromosome 1q25, was frequently downregulated and acted as a potential tumor-suppressor gene in various malignancies, such as breast cancer 7 , hepatocellular carcinoma 8 , and ovarian cancer 9 . It has also been reported that GAS5 could curbs the proliferation and metastasis of glioma cells, implicating that it might be a promising threptic treatment of gliomas 10 . However, there has been little study of the function and the related molecular mechanisms of GAS5 in glioma stem-like cells (GSCs), which needs further investigation.
CDH1 is widely recognized as a tumor suppressor involved in epithelial-to-mesenchymal transition (EMT) and inhibits metastasis, which encodes the cell-cell adhesion protein E-cadherin 11 . EMT enables polarized epithelial cells to mesenchymal cells, exhibiting characteristics of enhanced invasion and migration capacities 12 . The loss of E-cadherin (CDH1) expression is the crucial activation factor of EMT, and the downregulation of CDH1 in glioma contributes to tumor progression 13 .
In the present study, we analyzed the expression of GAS5 in glioma of TCGA and CGGA databases and found GAS5 was negatively associated with the malignancy of glioma. Based on integrated analysis of both function assays and further molecular regulation assays, it was identi ed that GAS5-mediated CDH1 suppression via sponging miR-23a inhibited the malignant behavior of GSCs.

Materials And Methods
Tumor specimens and cell culture Clinical cancer and adjacent normal brain tissue specimens derived from glioma patients collected from the Department of Neurosurgery, A liated Hangzhou First People's Hospital, Zhejiang University School of Medicine, after obtaining informed consent. The whole process was consistent with the requirement of Ethics Committee of the Zhejiang University School of Medicine.

Immunocytochemical Staining
GSCs were xed with methanol for 20 min at room temperature, permeabilized with 0.25% Triton X-100 (Beyotime, China), and blocked with blocking buffer for 60 min. The xed cells were then incubated rst with primary antibodies of SOX2, OCT4 (CST, US) for 60 min, respectively. After washing with phosphate buffer saline (PBS) for 3 times, the second antibody (Beyotime, China) was applied for 30 min at room temperature in the dark. Finally, the cells were stained with DAPI, and observed under uorescence microscope (Zeiss, Germany) at a magni cation of 200×.

Flow cytometry
Single-cell suspensions of GSCs were suspended with 100 µl of PBS and incubated with 5 µl primary antibodies of anti-SOX2 and anti-OCT4 (dilution 1:200) for 1 h at room temperature, after digestion and centrifugation. Then the cells were washed with PBS, and incubated with secondary antibody (Beyotime, China) for 1 h at room temperature. After centrifugation and washes with PBS, the cells were resuspended in 200 µl of PBS and nally analyzed with ow cytometry with Cytexpert 2.0 software.

Vector construction and cell transfection
The overexpression vector of GAS5, the short hairpin RNA (shRNA) targeting GAS5 (shGAS5-1 and shGAS5-2) and the corresponding negative control, the miR-23a mimics, inhibitors, and corresponding negative control, were all constructed by GenePharma (Shanghai, China). The vectors and corresponding controls were transfected into GSCs according to the manufacturer's protocol.
Quantitative real-time reverse transcription PCR (qRT-PCR) Total RNA was extracted from cells with TRIzol (Invitrogen, US) and reverse transcribed to cDNA using the reverse transcription kit (Thermo, US). qRT-PCR was performed to measure the expression levels of GAS5, CDH1 and miR-23a. The expression of GAPDH and U6 served as control. The expression level was analyzed by 2 −ΔΔCt method.

5-Ethynyl-20-deoxyuridine (EdU) assay
Single-cell suspensions of GSCs were seeded into 24-well plates pre-coated with Matrigel (50µl, 1:100 dilution, BD, US) for adherence in an incubator containing 5% CO 2 at 37°C. 50 µM EdU (RiboBio, China) was added to each well for incubation 2 h. Then the cells were xed in 4% polyformaldehyde for 20 min and stained by Apollo dye solution for 25 min. Cell nuclei were stained with DAPI for 20 min. The proportion of EdU-positive cells were analyzed under a uorescence microscope.

Transwell assay
Transwell assays were performed to evaluate the invasion and migration of GSCs. Some upper chambers were coated with Matrigel (50µl, 1:8 dilution, BD, US) in an incubator overnight for invasion assays. Other upper chambers were pre-coated with Matrigel (50µl, 1:100 dilution) for migration assays, in which the Matrigel was aspirated after 20min. The subsequent protocol was performed as previously described 14 . GSCs sphere formation assay Single-cell suspensions of GSCs were seeded on 24-well plates at a density of 100 cells per well. The sphere numbers of GSCs (>50 µm) in each well was counted under a microscope. The sphere formation e ciency (SFE) of GSCs was calculated by the spheres numbers/100 ×100% method.

Cell apoptosis assay
The cell apoptosis rate of GSCs was detected by the kit of annexin V-APC/7-AAD (BD, US). Single-cell suspensions of GSCs were washed by PBS and re-suspended by binding buffer including. 5µl annexin V-APC and 5µl 7-AAD were added to the suspensions of GSCs (15 min each), respectively, in the dark at room temperature. The cell apoptosis rate was analyzed by ow cytometry within 1 h after staining.
Tumorigenicity assay 4-6 weeks old athymic Balb/c nude mice (15-20 g) were bred in the animal center under speci c pathogen-free (SPF) conditions. Single cell suspensions of 1×10 6 GSCs with up/downregulation of GAS5 and the corresponding negative control were injected subcutaneously into the right ank of each mouse, respectively. 3 weeks later, the volume and weight of the tumors were calculated by length ×width 2 ×0.5).

Luciferase reporter assay
The wild-type and mutant fragments with miR-23a binding sequences in the 3′-untranslated region (UTR) of GAS5 and CDH1 were inserted into pMIR-REPORT vectors. Then the pMIR-REPORT vectors, together with miR-23a mimics or the corresponding negative control, were transfected into GSCs for 48h. The dualluciferase reporter assay system (Promega, US) was used to evaluate the luciferase activity of GSCs.

Western blotting
Total protein from GSCs was extracted by RIPA buffer (Beyotime, China). Protein concentration was determined by the BCA Protein Assay Kit (Beyotime, China), after which 20µg total protein were separated by 10% SDS-PAGE, transferred to PVDF membrane, and then incubated with primary antibodies against CDH1 (CST, US) and GAPDH (CST, US) overnight at 4℃. The membranes were then incubated with secondary antibody for 60min at room temperature. Enhanced chemiluminescence (ECL) method was used for visualization for quantitative analysis.

Statistical analysis
All data in this study were presented as mean ±SD and analysed with the GraphPad Prism software (Version 8.0.2, (CA, USA). One-way/two-way analysis of variance (one/two-way ANOVA) was used to determine the differences among at least three groups. Student's t test was used to evaluate the differences between two groups. P value <0.05 was accepted as statistically signi cant (*p < 0.05; **p < 0.01; ***p < 0.001; **** p < 0.0001).

Results
1. The expression of GAS5 was negatively correlated with the malignancy of glioma.
The expression of GAS5 was analyzed in different subtype of glioma in TCGA and CGGA database, we found that the GAS5 expression was negatively correlated with the malignancy and overall survival of glioma in databases of CGGA and TCGA. Based on IDH mutation status classi cation of glioma, remarkable differences of GAS5 expression between IDH mutant subtype and IDH wildtype subtype were also analyzed. The expression of GAS5 in IDH wildtype subtype was signi cantly decreased compared with IDH mutant subtype in CGGA_693 datasets and TCGA_GBMLGG datasets (Fig. 1A, B). The similar expression pro le showed that the expression of GAS5 in 1p/19q non-codeleted subtype was signi cantly lower than that of 1p/19q codeleted subtype (Fig. 1C, D). Further analysis found that the expression of GAS5 in IDH wildtype subtype was signi cantly lower than that of IDH mutant combined with 1p/19q codeleted subtype and IDH mutant combined with 1p/19q non-codeleted subtype (Fig. 1E, F). Overall survival analysis of gliomas showed that the survival rate of patients with low GAS5 expression decreased signi cantly (Fig. 1G, H).

Primary culture of GSCs derived from clinical surgical specimens
To investigate the function of GAS5 in GSCs, we cultured two human GSCs (GSC1 and GSC2), which respectively derived from two patients diagnosed glioblastoma. The CSC1 and GSC2 not only exhibited typical sphere-like cell clusters, but also could adherently grow on laminin coated plates ( Fig. 2A). To further verify the cells were GSCs, we detected the GSCs makers of GSC1 and GSC2. Immuno uorescence assays indicated that the GSCs makers (SOX2 and OCT4) were positive in GSC1 and GSC2 (Fig. 2B). And ow cytometric analysis showed the positivity rates of the GSC markers (SOX2 and OCT4) in GSC1 and GSC2 were all more than 70% (Fig. 2C).
GAS5 expression was further detected in the GSCs (GSC1 and GSC2), glioma cell lines (T98G, U251, A172), and glioma tissue, which disclosed GAS5 downregulation in GSCs and glioma cells and tissue (Fig. 3A, B). To evaluate the function of GAS5 in GSCs, GSC1 and GSC2 were transfected with GAS5 overexpression plasmid and the corresponding negative control, the transfection e ciency was validated by qRT-PCR (Fig. 2C). EdU assays indicated that GAS5 upregulation decreased the proliferation ability of GSC1 and GSC2 cells (Fig. 2D, E). Transwell assays showed that overexpression of GAS5 resulted in signi cant decline of invasion and migration ability of GSC1 and GSC2 cells (Fig. 2F-I). And sphere formation assays indicated that upregulation of GAS5 remarkably weakened the sphere formation e ciency of GSC1 and GSC2 cells (Fig. 2J-K). Furthermore, ow cytometric assays veri ed that GAS5 upregulation in GSC1 and GSC2 cells led to obvious increase in cell apoptosis rate (Fig. 2L-M). 4. Silencing GAS5 promoted proliferation, invasion, migration and stemness, and inhibited apoptosis of GSCs in vitro.
To further verify the function of GAS5 in GSCs, silencing GAS5 expression in GSC1 and GSC2 cells was achieved via transfection of short hairpin RNAs (shRNAs) targeting GAS5 (including shGAS5-1 and shGAS5-2) and the corresponding negative control, and veri ed by qRT-PCR (Fig. 4A). EdU assays showed GAS5 downregulation enhanced the proliferation capacity of GSC1 and GSC2 cells (Fig. 4B-D). The results of Transwell assays indicated that silencing GAS5 expression promoted the invasion and migration of GSC1 and GSC2 cells ( Figures 4E-H). Sphere formation assays suggested that downregulation of GAS5 signi cantly increased the sphere formation e ciency of GSC1 and GSC2 cells (Fig. 2I, J). And ow cytometric assays suggested that GAS5 downregulation in GSC1 and GSC2 cells led to increase in cell apoptosis rate (Fig. 2K, L).

GAS5 inhibited the growth of GSCs in vivo.
To evaluate the effect of GAS5 on GSCs growth in vivo, tumorigenicity assays were performed. Subcutaneous injection of GSC1 cells transfected with shGAS5 or the corresponding negative control was performed, and the results showed that GAS5 downregulation of GSCs led to increase in both tumor volume and weight, compared with the control group ( Fig. 5A-C). In addition, GSC2 cells transfected with TUG overexpression vector or the corresponding negative control were subcutaneous implanted, and the results showed that overexpression of GAS5 in GSCs led to lower tumor volume and weight, compared with the control group ( Fig. 5D-F).

GAS5 acted as a sponge for miR-23a in GSCs
To explore the mechanisms on regulating biological function of GSCs, bioinformatics analysis was applied to identify the possible miRNA targets of GAS5. Potential bonding sites between GAS5 and miR-23a were predicted via the online database StarBase (Fig. 6A). Further veri cation with qRT-PCR found upregulation expression of miR-23a in GSCs, glioma cell lines and glioma tissue, compared with normal human astrocytes (NHAs) and normal brain tissue (Fig. 6B, C). Speci cally, miR-23a was negatively regulated by GAS5 in GSCs (Fig. 6D, E). To clarify the direct interaction between miR-23a and GAS5, wild type (WT) and mutant type (MUT) vector of GAS5 were constructed for luciferase activity assays, which showed that miR-23a signi cantly inhibited the luciferase activity of GAS5-WT ( Figures 6F, G), indicating that miR-23a was the direct target of GAS5. 7. Silencing miR-23a inhibited the proliferation, invasion and migration, and promoted apoptosis of GSCs by targeting CDH1.
According to bioinformatic analysis, CDH1 was one of the putative downstream targets of miR-23a (Figures 7A). And to further validate the direct bonding between miR-23a and CDH1, CDH1-WT and CDH1-MUT vectors were constructed for luciferase activity assays ( Figure 7A). To further explore the regulatory relationship between miR-23a and CDH1, both of miR-23a inhibitor and CDH1 shRNA were transfected into GSC1/2 cells. The assays of qRT-PCR and western blot indicated that miR-23a downregulation can increase CDH1 expression, and this effect can be partially offset by shRNA transfection of CDH1 (Fig. 7B-D). And luciferase reporter assays suggested that miR-23a signi cantly decreased the luciferase activity of CDH1-WT, compared with CDH1-MUT (Fig. 7E, F). Furthermore, EdU assays showed silencing miR-23a weakened the proliferation of GSCs, and the function can be reversed by downregulation of CDH1 (Fig. 7G-I). Transwell assays indicated that the weakened invasion and migration ability of GSCs induced by silencing miR-23a were partially offset by downregulation of CDH1 (Fig. 7J-M). And ow cytometric assays suggested that reduced miR-23a in GSCs led to increase in cell apoptosis rate, which can be reversed by downregulation of CDH1 (Fig. 7N, O).
8. Silencing GAS5 promoted the proliferation, invasion and migration, and inhibited apoptosis of GSCs by targeting miR-23a.
To elucidate the mechanism which GAS5 downregulation promoted malignancy of GSCs via regulating miR-23a/CDH1 axis, GSC1 and GSC2 cells were transfected with shGAS5 or shGAS5 together with miR-23a inhibitors. qRT-PCR and western blot assays suggested that shGAS5 can inhibit CDH1 expression, which was partly offset by miR-23a inhibitors (Fig. 8A-C). Besides, EdU assays indicated shGAS5 enhanced the proliferation capacity of GSCs, which could be reversed by miR-23a inhibitors (Fig. 8D-F). Consistent with results of EdU assays, the increased ability of invasion and migration of GSCs induced by shGAS5 were partly rescued by miR-23a inhibitors as well (Fig. 8G-J). Flow cytometric analysis suggested that shGAS5 inhibited the apoptosis level of GSCs, which could be reversed by miR-23a inhibitors (Fig. 8K-L).

Discussion
Glioblastoma multiforme (GBM) is one of the most aggressive and common primary tumor in central nervous system, although comprehensive therapies have been improving 15 . within the bulk tumor, glioma stem cells (GSCs) exist as a minor subpopulation, but actively contribute to dismal prognosis including tumor recurrence and chemo-and radio-therapy resistance, which can self-renew and ourish in an unfavorable tumor microenvironment 16 . GSCs not only exhibit potent tumor-initiating or tumorpropagating characteristics but also promote malignant invasion 17 . In consequence, therapies targeting GSCs have been crucial for improving GBM treatment and overcoming therapeutic resistance 18 . In the present study, we constructed two GSCs, derived from clinical specimens of glioma patients, and identi ed the marker of neural stem cells, to investigate the crucial therapeutic target gene of GSCs which was more closer to the real conditions.
Recently, an increasing number of studies has suggested that lncRNAs play an important role in the occurrence and progression of gliomas 19 . And investigating the molecular regulation mechanisms of lncRNAs may provide a promising therapeutic target in gliomas. Growth arrest-speci c 5 (GAS5), located on chromosome 1q25, plays an anti-oncogene role in tumors. Low expression of GAS5 is associated with poor survival and cisplatin resistance in cervical cancer through regulation of PDCD4 20 . In addition, GAS5 can enhance radio-sensitivity of esophageal squamous cell carcinoma by upregulating RECK 21 . It is also reported that GAS5 acts as a molecular switch for proliferation regulation in CD133+ cells via inhibiting glucocorticoid receptor (GR) mediated cell cycle control in pancreatic cancer 22 . Moreover, lung cancer cell derived exsomal GAS5 was regard as a biomarker to identify early-stage in non-small cell lung cancer 23 . It was also reported that GAS5 could effectively inhibit the growth and invasion capacity of glioma cells via targeting GSTM3 24 . However, the function and relative mechanisms of GAS5 in GSCs remains unclear up to now. In our experiments, we con rmed the GAS5 e ciently inhibit the proliferation, invasion, migration and tumorgenicity, and promote apoptosis of GSCs through regulating miR-23a/CDH1. miRNAs have recently emerged as key players in tumor progression. MiR-23a is associated with the regulation of various cellular processes, including cell proliferation, apoptosis and metastasis, and also play crucial roles in several types of tumors with different effects. It was reported that miR-23a was upregulated and could promote proliferation and suppress apoptosis by targeting PDCD4 in gastric cancer 25 . The upregulation of miR-23a also promotes esophageal squamous cell carcinoma through targeting TRAF5 26 . In contrast, miR-23a was downregulated and functioned as a anti-oncogene in breast cancer 27 . It was also found that miR-23a could promote progression in glioma via suppressing PTEN 28 . However, the role of miR-23a played in GSCs has never been reported previously. According to our research, down-regulated miR-23a can inhibit the proliferation, invasion, migration, and promote apoptosis of GSCs via targeting CDH1. CDH1, a crucial component in EMT, is widely known as a pivotal protein for holding epithelial cells tight by maintaining cell to cell junctions 29 . Downregulation of CDH1 could decrease the cellular adhesion and promoted the invasion and metastatic of tumor 30 . It was reported that BMP4 suppressed cell invasion of GBM through upregulation of CDH1 31 . And Univariate Cox regression analysis showed that low expression of CDH1 was associated with shorter progression-free survival in ependymoma 32 . The results of our experiments suggested that the expression of CDH1 was promoted by GAS5 via sponging miR-23a.

Conclusions
In summary, our study not only uncovered the crucial effects of GAS5 as a tumor suppressor in GSCs but also explored the molecular regulation mechanisms through which miR-23a contributed to the progression of GSCs and identi ed CDH1 (E-cadherin) as a direct target, which might provide new experimental evidence for potential treatment strategy targeting on GSCs in glioma.

Declarations
Ethics approval and consent to participate: The study involving human participants were reviewed and approved by Ethics Committee of the Zhejiang University School of Medicine.

Consent for publication:
The patients/participants provided written informed consent to participate in this study, and had consent for publication.
Availability of data and materials: The datasets generated during and analysed during the current study are available in the TCGA and CGGA repository, [https://portal.gdc.cancer.gov; http://cgga.org.cn/index.jsp] Competing interests: No con ict of interest exists in the submission of the manuscript which is approved by all authors for publication.

Funding:
Agricultural and social development project of Hangzhou (NO. 2020ZDSJ0900) Authors' contributions: QD is responsible for the design of the study; HW is responsible for experimental implementation; DW is responsible for collection and analysis of data; YS and CS are responsible for the writing the manuscript; QH and LJ are responsible for reviewing the manuscript. All authors have read and approved the manuscript. doi:10.1016/j.humpath.2018.07.018 (2018). Figure 1  (A, B) The difference expression of GAS5 between IDH mutant and IDH wildtype subtypes in CGGA_693 and TCGA_GBMLGG datasets, respectively. ****p < 0.0001, Student's t test. (C, D) The difference of GAS5 expression between 1p/19q codeleted and 1p/19q non-codeleted subtypes in in CGGA_693 and TCGA_GBMLGG datasets, respectively. ****p < 0.0001, Student's t test. (E, F) The difference of GAS5 expression among IDH mutant and 1p/19q codeleted, IDH mutant and 1p/19q non-codeleted, and IDH wildtype subtypes in CGGA_693 and TCGA_GBMLGG datasets, respectively. ****p < 0.0001, one-way ANOVA. (G, H) Overall survival rate of glioma patients in low GAS5 group and high GAS5 group. The GSCs makers were detected by Immuno uorescence assays in GSC1 and GSC2. (C) The positivity rates of the GSC markers were analyzed by ow cytometric in GSC1 and GSC2. Figure 3 (A, B) GAS5 expression was analyzed by qRT-PCR in glioma tissue, glioma cell lines and GSCs. (C) GAS5 expression was analyzed with qRT-PCR in GSCs transfected with GAS5 or NC. *p < 0.05, **p < 0.01, Student's t test. (D, E) Proliferation ability was evaluated after GAS5 upregulation in GSCs using EdU assays. **p < 0.01, Student's t test. (F, I) Invasion and migration capacity was assessed after GAS5 upregulation in GSCs by transwell assays. **p < 0.01, ***p < 0.001, Student's t test. (J, K) The stemness was evaluated after GAS5 upregulation in GSCs by sphere formation assays. (L, M) Cell appotosis was evaluated via ow cytometric assays after GAS5 upregulation in GSCs. Figure 4 (A) qRT-PCR analysis of GAS5 expression in GSCs after transfection of the shNC, shGAS5-1 or shGAS5-2.