SOX9‐activated PXN‐AS1 promotes the tumorigenesis of glioblastoma by EZH2‐mediated methylation of DKK1

Abstract Increasing evidence has validated the essential regulation of long non‐coding RNAs (lncRNAs) in the biological process of tumours. LncRNA PXN‐AS1 has been discovered to be as a tumour suppressor in pancreatic cancer; however, its function and mechanism remain greatly unknown in glioblastoma (GBM). Our present study indicated that PXN‐AS1 was highly expressed in GBM tissues and cells. Besides, the knock‐down of PXN‐AS1 was closely associated with the inhibitory proliferation and inducing apoptosis of GBM cells. PXN‐AS1 inhibition was also found to restrain GBM tumour growth. Importantly, SOX9 functioned as a transcription factor and activated PXN‐AS1 expression, and overexpressed PXN‐AS1 rescued the inhibitory role of down‐regulated SOX9 in GBM cell growth. Subsequently, it was discovered that PXN‐AS1 activated Wnt/β‐catenin pathway. DKK1 was widely known as an inhibitor gene of Wnt/β‐catenin pathway, and its expression was negatively associated with PXN‐AS1 and SOX9. Interestingly, we found that PXN‐AS1 could recruit EZH2 to mediate the H3K27me3 level of DKK1 promoter. Restoration experiments manifested that DKK1 knock‐down counteracted PXN‐AS1 depletion‐mediated repression in GBM cell growth. All facts pointed out that PXN‐AS1 might be of importance in exploring the therapeutic strategies of GBM.

underlying GBM tumorigenesis and explore the potential mechanism which might be useful in improving therapeutic strategies.
Long non-coding RNAs (lncRNAs) are one type of transcripts whose length are more than 200 nucleotides and have restriction in the ability of protein-coding. 5 Owing to their modulation in biological behaviours, increasing lncRNAs have attracted people's attention. 6 Extended evidence has confirmed that lncRNAs are closely associated with the progression of human diseases, particularly with cancers, through multiple mechanisms, including chromatin modification, transcriptional and post-transcriptional processing, and interacting with specific proteins.7-10 Over the past few years, it has been validated that lncRNAs could be diagnostic or prognostic biomarkers in diverse cancers. For example, lncRNA LINC00978 functioned as a diagnostic biomarker and promotes cancer growth in gastric cancer.11 LncRNA FOXP4-AS1 acts as a prognostic factor in colorectal cancer and modulates colorectal cancer cell proliferation and apoptosis.12 A wide range of lncRNAs have been reported in GBM. For example, LINC00470 mediated GBM cell autophagy by activating AKT. 13 LncRNA HOTAIRM1 was overexpressed in GBM and promotes cell invasion and tumour growth via up-regulating HOXA1.14 LncRNA PXN antisense RNA 1 (PXN-AS1) was reported to be down-regulated in pancreatic cancer cells and suppressed cancer progression. 15 However, the expression of PXN-AS1 was found to be up-regulated in GBM tissues through GEPIA database. Therefore, we wondered whether PXN-AS1 exerted reverse functions in GBM.
In this study, we explored the biological function and molecular mechanism of PXN-AS1 in GBM. It was found that PXN-AS1 was up-regulated in GBM cells and silenced PXN-AS1 resulted in the repressive cell proliferation and induced cell apoptosis.

| Quantitative real-time PCR (qRT-PCR)
Total RNA was extracted from U251 and U87 cells following the manual of RNeasy Mini Kit (Qiagen) for synthesizing cDNA templates using Reverse Transcription Kit (Toyobo). Gene expression was quantified by Bio-Rad SYBR Green Super Mix (Bio-Rad) and calculated by 2 −ΔΔCT method, relative to GAPDH or U6. The primers used here were shown in Table S1.

| Colony formation
The transfected U251 and U87 cells were plated at 500 cells/well in the 6-well plates. Following 14 days of incubation, clones were treated with 0.5% crystal violet in 4% paraformaldehyde.
Later, U251 and U87 cells were washed in PBS and fixed by 4% paraformaldehyde. Subsequently, Apollo was utilized for staining cells.
Before cell nucleus was observed, the stained cells were processed by Hoechst for half an hour. Then, cell proliferation was observed using a high-power microscope and photographed on MetaXpress software (Molecular Devices).

| JC-1 assay
Recently, JC-1 assay has been usually carried out to examine early apoptosis event in GBM cells under different conditions since this experiment could detect changes in mitochondrial membrane potential, which is a hallmark of early cell apoptosis. 16 When the mitochondrial membrane potential is high (namely apoptosis rate is low), JC-1 accumulates in the matrix of mitochondria, thus producing red fluorescence. On the contrary, JC-1 cannot concentrate in the matrix of mitochondria when the mitochondrial membrane potential is low (cells with high apoptosis rates), finally resulting in green fluorescence. In this study, JC-1 assay was implemented in GBM cells in line with the instruction of JC-1 detection kit (Beyotime).
Following fluorescence labelling, samples were rinsed in PBS for analysis with EnSpire Reader (PerkinElmer). The JC-1 ratio was evaluated as the rate of red signals relative to green signals.

| Flow cytometer of apoptosis
After transfection, cells were first fixed for 1 hours on ice and then treated with Annexin V-FITC/PI (BD Biosciences, Erembodegem, Belgium) for 15 minutes in the dark. Cell apoptosis rate was monitored by flow cytometry (BD Biosciences).

| Animal study
10 six-week-old male BALB/c nude mice were procured from Shanghai SIPPR-BK Laboratory Animal (Shanghai, China) and housed under SPF condition, with the approval from the Use Committee for Animal Care of Renji Hospital. Each mouse was subcutaneously inoculated with 5 × 10 6 GBM cells. Tumour volume was monitored every 4 days according to the calculating formula: Volume = 0.5 × length × width 2 . Tumour samples were excised from mice 28 days later for weight assessment.

| Immunohistochemistry
The tumour tissue samples from animal study were fixed with 4% paraformaldehyde and then embedded in paraffin for cutting.

| In situ hybridization assay
The fresh tumour tissue samples were cultured in 0.5% Triton X-100 and PBS for incubation with the PXN-AS1 FISH probe (RiboBio) in hybridization solution. After DAPI staining, samples were analysed with fluorescence microscope.

| Chromatin immunoprecipitation
After cross-linking, the chromatin samples were prepared for sonication to 200-1000-bp fragments and then immunoprecipitated with the 2 μg of specific antibodies (Millipore) and 30 μL of magnetic beads. qRT-PCR was followed finally.

| Subcellular fractionation
The GBM cells in pre-chilled PBS were centrifuged for incubation in cell fractionation buffer for cell cytoplasm and cell disruption buffer for cell nuclei. Isolated RNAs were finally detected by qRT-PCR.

| RNA immunoprecipitation
Magna RNA-binding protein immunoprecipitation kit (Millipore) was employed to perform RNA immunoprecipitation (RIP) assays.
In brief, the cultured GBM cells were first collected from RIP lysis buffer and then incubated in RIP buffer adding the beads bound the antibodies against human Ago2 (ab32381, Abcam) and normal rabbit IgG (ab172730, Abcam) for 2 hours. The final precipitated RNAs were subjected to qRT-PCR.

| Statistical analyses
Results were expressed as the mean ± standard deviation (SD) of bio-triple repeats. All data differences in this study were analysed through Student's t test or one-way analysis of variance using GraphPad Prism 6 software (GraphPad Software, Inc). The statistical significance was specified as the value of P < .05. Each experiment was repeated three times.
Flow cytometry analysis further confirmed the inhibitory role of silenced PXN-AS1 in cell apoptosis ( Figure 1H). All data indicated that PXN-AS1 was overexpressed in GBM cells and enhanced cell proliferation and restrained cell apoptosis.

| PXN-AS1 facilitates tumour growth in GBM
Next, we noticed the function of PXN-AS1 on GBM tumour growth in vivo, and U251 cells transfected with sh-PXN-AS1 or sh-NC were subcutaneously injected into nude mice. Observed in 28 days, the tumours were removed, and the weight was measured. As expected, the tumour growth rate was slower, and the final volume and weight in sh-PXN-AS1 group were lower than those in sh-NC group

| SOX9 interacts with PXN-AS1 promoter
Subsequently, we investigated the mechanism that correlated with

| PXN-AS1 activates WNT signalling pathway
SOX9 was reported to activate Wnt/β-catenin pathway and drive the tumour progression in non-small-cell lung cancer. 19 Hence, we explored the interaction between SOX9 and Wnt/β-catenin pathway in GBM. As shown in Figure S3A, two well-known downstream targets of Wnt/β-catenin pathway, including cyclin D1 and c-myc, demonstrated as remarkably decreased whereas CTNNB1 level unchanged in SOX9 down-regulated cells. Besides, subcellular fractionation assay plus Western blot revealed that SOX9 knock-down restrained the translocation of β-catenin into nuclear ( Figure S3B). And the suppressed nuclear translocation of β-catenin under SOX9 silence was further confirmed via IF assay ( Figure S3C). According to TOP/FOP flash assay, we found that the activity of Wnt/β-catenin pathway was also repressed by sh-SOX9 transfection ( Figure S3D). Owing to the activated role of SOX9 in Wnt/β-catenin pathway, we speculated that PXN-AS1 exerted the same function on Wnt/β-catenin pathway. Consistently, PXN-AS1 silence hampered the mRNA expression of cyclin D1 and c-myc but not that of CTNNB1, whereas restrained the protein levels of all the three genes ( Figure 4A-B). From subcellular fractionation plus Western blot analysis, nuclear translocation of β-catenin was alleviated in cells with the transfection of sh-PXN-AS1 ( Figure 4C-D), so were the results in IF staining ( Figure S3E). TOP/ FOP flash assay further confirmed the repressive effect of PXN-AS1 deficiency on Wnt/β-catenin pathway ( Figure 4E).

| PXN-AS1 recruits EZH2 to epigenetically suppress DKK1 expression
On basis of regulatory role of PXN-AS1 in DKK1 expression, we investigated the underlying regulatory mechanism of PXN-AS1 on DKK1. Firstly, we detected the distribution of PXN-AS1 in GBM cells.
As a result, PXN-AS1 was mainly localized in the nucleus of U251 and U87 cells ( Figure 5A), which implied that PXN-AS1 could participate in transcriptional regulation. Abundant literatures revealed that lncRNAs exerted regulatory functions via binding to histone modification enzymes. Recent evidence has showed that lncRNAs could modulate downstream genes via interacting with PRC2, which is a methyltransferase that represses the transcription of specific genes by trimethylating H3K27. Therefore, we carried out RIP assay to probe whether PXN-AS1 interacted with PRC2. Three key components of PRC2 (EZH2, SUZ12 and EED) were used in the study.
The results delineated that both PXN-AS1 and DKK1 promoter F I G U R E 5 PXN-AS1 recruits EZH2 to epigenetically suppress DKK1 expression. A, The distribution of PXN-AS1 in GBM cells was determined subcellular fractionation assay. B, The enrichment of PXN-AS1 or DKK1 promoter by EZH2/SUZ12/EED was determined by the use of RIP assay or ChIP assay, respectively. C, EZH2 knock-down efficiency was evaluated by qRT-PCR in GBM cells. D, DKK1 mRNA and protein levels were tested through qRT-PCR and Western blot upon EZH2 depletion. (E) ChIP assay revealed the interaction between EZH2 and DKK1 promoter, and the interaction and H3K27me3 enrichment were tested in sh-PXN-AS1 and sh-NC transfected cells. F, DKK1 expression was examined by qRT-PCR and Western blot after transfecting overexpression plasmids. G, Cell proliferation was measured via colony formation assay in cells treated indicated plasmids. H, Cell apoptosis was assessed through flow cytometry analysis in cells after transfecting indicated plasmids. *P < .05, **P < .01 could bind to EZH2 in U251 and U87 cells ( Figure 5B). Besides, we transfected sh-EZH2 into GBM cells for down-regulating its expression ( Figure 5C). Further, we found that DKK1 mRNA and protein levels were enhanced in sh-EZH2 transfected cells ( Figure 5D).
As illustrated in Figure 5G, the treatment of LiCl countered DKK1 overexpression-mediated suppression in GBM cell proliferation.

| D ISCUSS I ON
In recent years, extensive reports have elucidated that lncRNAs undergo the main responsibility for triggering the dysregulation of vari- Interestingly, the function of PXN-AS1 changes in different cancer types, which might be attributed to tumour heterogeneity to some extent. In detail, PXN-AS1 played a promoting role in the tumorigenesis of lung cancer23 and nasopharyngeal carcinoma,24 similar to the findings observed in current study. However, a recent report also suggested that PXN-AS1 served as a tumour suppressor in pancreatic cancer.15 As an important biological participator, Wnt/β-catenin pathway exerts huge functions in the process of tumour initiation and progression.25-27 Among which, its functional contribution to ln-cRNAs-mediated cancer development has attracted people's attention.28 Through researches of numerous years, it was confirmed that enhanced nuclear translocation of β-catenin indicates the activation of Wnt/β-catenin pathway. 29 The transcription factor SOX9 has been reported to activate Wnt/β-catenin pathway in non-smallcell lung cancer. 19 In this regard, we investigated whether SOX9 and SOX9-mediated PXN-AS1 could act the similar effect on GBM progression. The results showed that SOX9 knock-down effectively inhibited the expression of relative genes in Wnt pathway, and the nuclear translocation of β-catenin was also restrained by down-reg- LncRNA CASC15 epigenetically regulates PDCD4 expression by recruiting EZH2 and promotes melanoma progression. 36 In this study, we found that PXN-AS1 interacted with EZH2, and EZH2 silence decreased DKK1 expression. Importantly, PXN-AS1 knock-down reduced EZH2 interaction and the enrichment of H3K27me3 in DKK1 promoter region. Further, it was discovered that the treatment of LiCl rescued the promoting effect of DKK1 overexpression on GBM cell growth. As observed in restoration assays, up-regulated DKK1 counteracted PXN-AS1 silence-mediated suppression on GBM cell growth.
Conclusively, PXN-AS1 acts as an oncogene in GBM, and SOX9activated PXN-AS1 promotes GBM progression by epigenetically silencing DKK1, suggesting a helpful revelation for the exploration of novel GBM diagnostic and therapeutic strategies.

ACK N OWLED G EM ENT
We appreciate all the participants who provide supports for the study.

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