LncRNA HOXA‐AS3 promotes the malignancy of glioblastoma through regulating miR‐455‐5p/USP3 axis

Abstract Our objective was to determine the molecular mechanisms by which lncRNA HOXA‐AS3 regulates the biological behaviour of glioblastoma multiforme (GBM). We used an lncRNA microarray assay to identify GBM‐related lncRNA expression profiles. Qrt‐PCR was used to survey the levels of expression of long non‐coding RNA (lncRNA) HOXA‐AS3 and the target gene. Dual‐luciferase reporter assays were used to investigate the interaction of lncRNA HOXA‐AS3, the target gene and miRNA. Western blot analysis was used to examine the expression of USP3 and epithelial‐mesenchymal transition (EMT) genes. The MTT assay, transwell assay and wound healing assay were used to analyse the effects of lncRNA HOXA‐AS3 on GBM cell viability, mobility and invasiveness, respectively. Our results showed that lncRNA HOXA‐AS3 was significantly up‐regulated in GBM cells and could promote GBM cell proliferation, invasion and migration in vitro and in vivo. HOXA‐AS was found to be associated with poor survival prognosis in glioma patients. The dual‐luciferase reporter assay also revealed that lncRNA HOXA‐AS3 acts as a mir‐455‐5p sponge by up‐regulating USP3 expression to promote GBM progression. Western blot analysis showed that lncRNA HOXA‐AS3 could up‐regulate EMT‐related gene expression in GBM. Experiments showed mir‐455‐5p could rescue the effect of lncRNA HOXA‐AS3 on cell proliferation and invasion. The newly identified HOXA‐AS3/mir‐455‐5p/USP3 pathway offers important clues to understanding the key mechanisms underlying the action of lncRNA HOXA‐AS3 in glioblastoma.

as RNAs if they are longer than 200 nucleotides. Most lncRNAs are poorly annotated coding genes. 5 The expression of some lncRNAs is correlated with poor cancer prognosis. 6,7 LncRNAs regulate tumorigenesis and invasion by activating or silencing oncogenes via various mechanisms, such as epigenetic regulation, RNA decay and post-translational modification. 8 Many lncRNAs have high levels of expression in brain tumours such as GBM and astrocytoma, and these lncRNAs have been shown to rely on various molecular mechanisms. 9 In the current study, we identified a novel lncRNA, HOXA cluster antisense RNA 3 (HOXA-AS3). It was found to be one of the most up-regulated lncRNAs in GBM tissues. 10,11 HOXA-AS3 is a HOX cluster, and it is a set of transcription factor genes that regulate hematopoietic lineage differentiation and embryological development. 10,12 Members of the HOX cluster family, such as HOXA-AS2, have been shown to regulate cell proliferation and the formation of tumour-related vasculogenic mimicry in various tumours. 13,14 In contrast, there have been very few studies on the biological mechanism underlying the action of lncRNA HOXA-AS3 in tumours.
We here identified lncRNA HOXA-AS3 and observed characteristic up-regulation of its expression in malignant glioma tissue relative to normal brain tissues, which is consistent with the information in the TCGA database. Down-regulation of lncRNA HOXA-AS3 has been shown to significantly suppress glioblastoma cell (LN229 and U251) proliferation, invasion and migration. We also found that lncRNA HOXA-AS3 could act as a miRNA sponge to reduce miR-455-5p concentration, up-regulate USP3 expression and drive the epithelial-mesenchymal transition (EMT) process of GBM cells.
This study is the first to provide evidence of a positive correlation between lncRNA HOXA-AS3 and USP3. This study improved our understanding of the HOXA-AS3/miR-455-5p/USP3 signalling pathway and may facilitate future development of new treatments for glioblastoma.

| TCGA data analysis
The processed lncRNA expression data of 502 cases of GBM and five cases of normal brain tissues were download from the TCGA database (http://cance rgeno me.nih.gov/).

| Tumour tissues samples
Malignant glioma specimens (n = 25) and normal brain tissues (NBT; n = 10) were purchased from January 2009 to December 2019 at the People's Hospital affiliated with Jiangsu University. This study protocol was approved by the ethics committee of Jiangsu University Affiliated People's Hospital, and the procedures were performed in accordance with the approved guidelines. All tumours were characterized by two pathologists. Glioblastoma multiforme cell lines (U251, LN229, U87 and U138) were   obtained from the American Type Culture Collection. T98G and SNB19 were purchased from the BeNa culture collection (BNCC), and normal human astrocytes (NHAs) were obtained from Sciencell Research Laboratories. All cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Life Technologies) supplemented with 10% foetal bovine serum (FBS; Life Technologies) and maintained in a humidified atmosphere containing 5% CO 2 at 37°C. Lentivirus transfection was purchased from GenePharma (Shanghai) and used to establish LN229 cell line labelled with green fluorescent protein (GFP).

| RNA extraction and lncRNA microarray analysis
Total RNA was isolated using TRIzol reagent (Invitrogen). Qualified total RNA samples were amplified by cDNA according to the Low Input Quick Amp WT Labeling Kit (Agilent) and standard operating procedures (SOPs). The expression profile of lncRNA in five malignant glioma and five NBT were screened by using lncRNA microarray (Shanghai Biotechnology Corporation). Tissue sample preparation and microarray hybridization were separately performed according to manufacturer's recommendations. qRT-PCR was used to further determine the nuclear/ cytoplasmic level of lncRNA HOXA-AS3 in glioma cells. GAPDH and U6 were used as a nuclear and cytoplasmic control. The primer sequences for the amplification of the lncRNA HOXA-AS3 and GAPDH used were listed in Table S1 2.5 | Quantitative RT-PCR (qRT-PCR) SYBR TM Green PCR Master Mixs (ABI, USA) were designed for quantitative RT-PCR. A LightCycler 490 Probes Master kit was also used to detect the relative expression of gene, and GAPDH was applied as the internal control for normalization. qRT-PCR results were shown using the 2-ΔΔCt method.

| Oligonucleotide transfection
The small interfering RNA (siRNA) and vector were purchased from Shanghai Gene Pharmaceutical Company. These plasmids were transfected into LN229 and U251 cells with Lipofectamine 3000 (Invitrogen) to obtain lentivirus soups complying with the manufacturer's instructions. siRNA sequence was shown in Table S1 2.7 | Inducible shRNA knock-down Short hairpin RNA (shRNA) targeting lncRNA HOXA-AS3 (sh-HOXA-AS3) is used to silence lncRNA HOXA-AS3 in vivo. Lentiviral particles were packaged into LN229 cells by co-transfection with plasmids. All shRNA sequences were shown in Table S1.

| MTT assays
Cell proliferation was evaluated with an MTT kit (Promega) complying with manufacturer's instructions. Thereafter, 3 × 10 3 cells were seeded in each well of a 24-well plate and maintained in medium containing 10% FBS for 2 weeks, during which time the medium was replaced with MTT reagent every 4 days. Cell proliferation and viability were measured with absorbance at 490 nm.

| Cell invasion assays
Cell invasion was analysed using Transwell chambers (Corning). In

| Cell migration assays
The scratch assay was used for the study of cell motility. After 24 hours, the cells were seeded on 6-well plates at an initial density of 5 × 10 5 in DMEM supplemented with 10% FBS. After the cells reached confluence, a scuffed wound was introduced with a sterile pipette tip into each well. The injuries were observed using phase contrast microscopy on an inverted microscope.

| Dual-Luciferase reporter assays
For luciferase reporter generation, the target genes of lncRNA HOXA-AS3 were predicted by using bioinformatics analysis. USP3 3'UTR was cloned into pmirGLO-dual-luciferase reporter as described previously. 15 Mutant and wild-type plasmids were acquired to preventing their binding with miR-455-5p from GenePharma

| Western blotting
The proteins of the cultured GBM cells were extracted, separated by electrophoresis and transferred to PVDF membranes (Millipore).
Protein concentrations were blocked with bovine serum albumin and then incubated with primary antibodies at 4°C overnight. The membranes were then incubated with secondary antibody at room temperature for one hour. Immunoreactivity was detected using the ECL chemiluminescent detection system (Thermo Fisher Scientific).

| RNA pull-down assay
A pull-down test was used to examine the potential association of lncRNA HOXA-AS3 with miRNA. The cells were treated and quantified with 1 mL of cellular nuclear lysate buffer for 3 days. Next, the cells were incubated with magnetic beads coated with streptavidin (Sigma) at 25°C for 2 hours, and the withdrawal test was carried out in an RNA complex coupled to biotin. After removal, the beads were washed in lysis buffer and RNA complexes bound to the beads were collected by centrifugation. The RNA in the beads was eluted and detected by qRT-PCR.

| Statistical analysis
All statistical analyses were performed by SPSS 21.0 (IBM). Each experiment was repeated three or more times, as mentioned in each of the figure legends. Count data were expressed as a percentage or as a ratio, and the chi-square test was applied for comparison.
The results were analysed using Student's t test or unidirectional analysis of variance (ANOVA). Data represent mean ± SD. *P < .05, **P < .01 and ***P < .001 were considered statistically significant.

| LncRNA HOXA-AS3 expression was upregulated in GBM tissue and cell lines
LncRNA microarray assay was used to detect abnormal expression of thousands of lncRNAs in malignant glioma tissues ( Figure 1A).
LncRNA HOXA-AS3 showed the most up-regulated of 11.3-fold change, which was selected as a candidate for further examination. In addition, the lncRNAs expression level of GBM patients from TCGA database showed that lncRNA HOXA-AS3 was a significant up-regulated lncRNA in GBM tissue. ( Figure 1B). The expression of lncRNA HOXA-AS3 in malignant glioma patients' tissues and normal brain tissues was validated by RT-PCR ( Figure 1C). Then, Malignant glioma clinical samples were divided into two subsets with low expression versus high expression level. Kaplan-Meier analysis showed that patients with higher levels of lncRNA HOXA-AS3 had significantly poorer overall survival rates than those with lower expression of this lncRNA in GBM F I G U R E 1 LncRNA HOXA-AS3 was up-regulated in glioblastoma patients and glioblastoma cells. A, Volcano plots show differentially expression levels of lncRNA in glioblastoma multiforme (GBM) tissues compared with normal tissues analysis. B, A heatmap is shown the expression of lncRNA for GBM samples (n = 502) compared to normal tissue samples (n = 5). Red represents lncRNAs that were up-regulated in GBM. C, QRT-PCR analysis of expression levels of lncRNA HOXA-AS3 in GBM tissue and normal brain tissues (**P < .01). D, The survival rate between higher expression of lncRNA HOXA-AS3 (n = 14) and lower expression of lncRNA HOXA-AS3 (n = 11) in GBM patient was calculated by Kaplan-Meier curve. E, QRT-PCR analysis revealed lncRNA HOXA-AS3 obviously up-regulated in glioma cell lines compared with NHAs (*P < .05, **P < .01). F, lncRNA HOXA-AS3 was mainly located in the cytoplasm of the GBM cells F I G U R E 2 Knock-down of lncRNA HOXA-AS3 reduces glioblastoma multiforme cell tumorigenicity and EMT process in vitro. A, The efficiency of lncRNA HOXA-AS3 knock-down was detected by qRT-PCR (**P < .01). B, MTT experiments suggested that lncRNA HOXA-AS3 knock-down attenuated the proliferative capacity of LN229 cells (**P < .01, ***P < .001). C, MTT experiments suggested that lncRNA HOXA-AS3 knock-down attenuated the proliferative capacity of U251 cells (**P < .01, ***P < .001). D, Colony formation assay manifested lncRNA HOXA-AS3 knock-down reduces LN229 and U251 cells tumorigenicity (**P < .01). E, Transwell experiments revealed knock-down of lncRNA HOXA-AS3 inhibited LN229 and U251 cell invasion (**P < .01). F, The wound healing assay showed a significant decrease of LN229 and U251 cells migration after transfected lncRNA HOXA-AS3 inhibitor (**P < .01). G, Western blot detected the different expression levels of EMT-related gene (E-cadherin N-cadherin and vimentin) in si-HOXA-AS3 group and blank control group glioma cells (**P < .01, ***P < .001) patients ( Figure 1D). Furthermore, qRT-PCR was used to assess the

| Knock-down of lncRNA HOXA-AS3 reduces GBM cell tumorigenicity and EMT process
To evaluate the functionality of lncRNA HOXA-AS3 in vitro and in vivo, we reduced lncRNA HOXA-AS3 expression in LN229 and U251 cells by using siRNA. The efficiency of lncRNA HOXA-AS3 downregulation was tested by qRT-PCR ( Figure 2A). Functionally, colony formation and MTT assay showed that lncRNA HOXA-AS3 lncRNA degradation significantly reduced cell proliferation for LN229 and U251 cells after si-HOXA-AS3 transfection ( Figure 2D). Using transwell assays, we demonstrated that silencing lncRNA HOXA-AS3 could decrease invasion and migration functions of GBM cells ( Figure 2E). We also found that knock-down of lncRNA HOXA-AS3 expression strongly inhibited flattening and spreading in a cellwounding assay, suggesting that lncRNA HOXA-AS3 expression dramatically affects key tumorigenesis gene signatures ( Figure 2F).

F I G U R E 3
LncRNA HOXA-AS3 depression inhibits glioblastoma tumorigenesis in vivo. A, The volume tumours formed by the LN229/ sh-HOXA-AS3 cells were significantly smaller relative to those formed by the LN229/sh-NC cells at the end of the experiment (*P < .05, **P < .01 and ***P < .001). B, Images of dissected tumours after the sh-NC cells and sh-HOXA-AS3 cells were subcutaneous injection into the two groups of nude mice. C, The average weight significantly differed between the two groups (**P < .01). D, lncRNA HOXA-AS3 expression of LN229/sh-HOXA-AS3 group was significantly lower relative to LN229/sh-NC group in tumour tissues (**P < .01) Western blotting analysis was performed to verify EMT-related biomarkers (E-cadherin, N-cadherin and vimentin) expression. These results showed knock-down of lncRNA HOXA-AS3 can increase E-cadherin protein expression and reduce inhibit N-cadherin and vimentin protein expression ( Figure 2G).

| Overexpression of lncRNA HOXA-AS3 promotes glioblastoma cell proliferation and invasion
To assess whether lncRNA HOXA-AS3 can also promote glioblastoma cell proliferation, we used a simulator to regulate lncRNA

| Deregulation of lncRNA HOXA-AS3 suppresses GBM cell proliferation and invasion in GBM orthotopic xenografts
To further evaluate the potential therapeutic value of lncRNA HOXA-AS3 inhibition in vivo, we transfected shRNA lncRNA HOXA-AS3 plasmid vectors into LN229 cells (sh-HOXA-AS3/ LN229) for orthotopic GBM xenografts. Then, we injected nude mice subcutaneously with sh-HOXA-AS3/LN229 cells and measured the tumour volume every 4 days thereafter for 32 days. The quantification of the tumour volume showed that tumour growth was slowed down in the mice injected with sh-HOXA-AS3 cells compared to the sh-NC cells (P < .01) ( Figure 3A,B). As shown in Figure 3C, treatment with sh-HOXA-AS3 also resulted in a significant reduction in tumour weight. We further extracted RNA from tumour tissues for qRT-PCR analysis, and the results showed that the expression of lncRNA HOXA-AS3 in sh-HOXA-AS3 tissues was significantly lower relative to control tissues ( Figure 3D). These results suggest that suppression of lncRNA HOXA-AS3 may have therapeutic potential for established tumours.

| USP3 is the target gene of miR-455-5p in GBM
To identify miR-455-5p-mediated downstream regulators of cell growth and invasion in glioma, we applied three miRNA target prediction algorithms (Targetscan, miRDB and miRanda). We found 34 genes clustered in the aggressive and proliferation process ( Figure 5A). Among the many miR-455-5p targets, we focused on ubiquitin specific protease 3 (USP3) for further investigation due to miR-455-5p could bind to the 3′-UTR region of USP3 ( Figure 5B). Therefore, we constructed USP3 3'UTR-WT and USP3 3'UTR-MUT.
Both of plasmids were performed for luciferase reporter assay.
These results showed that miR-455-5p mimic significantly inhibited the luciferase activity of USP3 3'UTR-WT, while USP3 3'UTR-MUT was not affected, suggesting that USP3 is a direct target of miR-455-5p ( Figure 5C). We further found that USP3 expression was positively correlated with lncRNA HOXA-AS3 expression and negatively correlated with miR-455-5p expression in glioma tissues F I G U R E 5 USP3 is the target gene of miR-455-5p in glioblastoma multiforme (GBM). A, The Venn diagram shows the miR-455-5p target gene predicted by miRDB (red), Targetscan (green) and miRanda (blue). B, According to the binding site information provided by Targetscan, design USP3-MUT or USP3-WT to explore the interaction between USP3 and miR-455-5p. C, The luciferase activities of USP3-WT and USP3-MUT were measured in GBM cells transfected miR-455-5p mimics or miR-NC with dual-luciferase reporter assay (**P < .01). D, Pearson correlation showed the correlation between lncRNA HOXA-AS3 and USP3 in GBM (**P < .01). E, Pearson correlation showed the correlation between miR-455-5p and USP3 in GBM. F, Effect of lncRNA HOXA-AS3 and miR-455-5p interactions on USP3 expression was quantified by Western blot | 11765 CHEN Et al.
( Figure 5D,E). At the same time, the effect of lncRNA HOXA-AS3 and miR-455-5p interaction on USP3 expression was quantified by Western blot analysis ( Figure 5F). These evidences indicate that lncRNA HOXA-AS3 can regulate the expression of USP3 in glioma through miR-455-5p.

| Overexpression miR-455-5p suppresses the effect of lncRNA HOXA-AS3 on proliferation and invasion of GBM
To validate the role of the lncRNA HOXA-AS3/miR-455-5p/USP3 axis, a rescue test was used. The results showed that miR-455-5p  Figure 6D). Previous studies have shown that USP3 raised GBM progression by regulating EMT. 18 We explored the expression of EMT-related genes in GBM orthotopic xenografts from different treatment groups. Western blot analysis showed that down-regulation of lncRNA HOXA-AS3 inhibits USP3 expression and EMT in vivo ( Figure 6E).  23,24 MiR-455-5p has been considered as either an oncogene or a tumour suppressor in certain types of tumour. For example, miR-455-5p acts as an oncogene in lung cancer. 25 In contrast, miR-455-5p exerts tumour-suppressive effects in prostate cancer. 26 However, the role of miR-455-5p in GBM has not been revealed.

| D ISCUSS I ON
Using bioinformatics analysis, we found that miR-455-5p may be a promising target miRNA for lncRNA HOXA-AS3. To further validate the predictions, using qRT-PCR, we revealed a negative correlation between lncRNA HOXA-AS3 and miR-455-5p in LN229, U251 cells and GBM tissues. In addition, we demonstrated the direct binding of lncRNA HOXA-AS3 and miR-455-5p using a dual-luciferase reporting assay, suggesting that lncRNA HOXA-AS3 promotes GBM progression by sponging miR-455-5p. Therefore, we predict that lncRNA HOXA-AS3 can regulate GBM processes through sponge miR-455-5p. To validate this prediction, we performed a rescue experiment, and the results showed that miR-455-5p could significantly rescue the effect of lncRNA HOXA-AS3 on GBM proliferation and metastasis. Using bioinformatics analysis, we found that USP3 may be a promising target gene for miR-455-5p. The dual-luciferase report assay has shown that miR-455-5p can bind to the 3'-UTR region of USP3. These results indicate that lncRNA HOXA-AS3 can promote glioma progression by regulating the miR-455-5p/ USP3 axis.
In conclusion, we found an associated lncRNA network that regulates the tumour cell development network, consisting of lncRNA HOXA-AS3, miR-455-5p and USP3. Our research not only contributes to the intensive efforts to elucidate the mechanism of lncRNA HOXA-AS3, but also specifically encourages the development of clinic-based lncRNA anti-glioma diagnostics and therapeutics.

ACK N OWLED G EM ENTS
All authors appreciate the help from members who have made a contribution to this paper.

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 sets used or analysed in this study may be obtained from appropriate authors upon reasonable request.