Long Noncoding RNA NEAT1, Regulated by the EGFR Pathway, Contributes to Glioblastoma Progression Through the WNT/β-Catenin Pathway by Scaffolding EZH2

Glioma is the most common and malignant tumor in the central nervous system. Maximal safe resection following radiotherapy and chemotherapy is the most popular therapeutic regimen for patients who have a malignant brain tumor. However, the overall survival of glioblastoma (GBM) patients is approximately 15 months from the time of diagnosis (1). EGFR alteration is an important marker in glioblastoma. EGFRvIII, the most common EGFR mutant, is found in 50%–60% cases, strongly indicating a poor survival prognosis (2–4). In 2014, EGFR module (EM)/PDGFRA (platelet-derived growth factor receptor A) (PM)-based molecular classification was performed and provided a molecular diagnostic framework to expedite the search for new glioma therapeutic targets (5). Furthermore, the molecular mechanism underlying glioma progression remains poorly understood. Therefore, a better understanding of gliomagenesis and progression are essential for the development of diagnostic markers and novel effective therapies for glioma patients. Long noncoding RNAs, non-protein coding transcripts longer than 200 nucleotides, have recently emerged as critical regulators of gene expression. Bromodomain protein BRD4 promotes the expression of HOTAIR to regulate the cell cycle of glioma cells (6). In a hypoxic microenvironment, HIF1A-AS2 facilitates the maintenance of mesenchymal glioblastoma stem-like cells (7). A lncRNA transcribed from the 5-prime end of the HOXA transcript HOXA11-AS contributes to the malignant progression of glioblastoma (8). In this study, we compared the lncRNA profiles of glioma tissues from different grades using RNA microarrays from the CGGA, Rembrandt, TCGA, and GSE16011 projects. We discovered that nuclear enriched abundant transcript 1 (NEAT1) displayed specific overexpression in the EGFR module, but not in the PDGFR module. NEAT1 is an abundant intranuclear lncRNA containing two transcripts, NEAT1_1 (3.7 kb) and NEAT1_2 (23 kb), and the latter is a core component of paraspeckles, which is a major complex in RNA nuclear retention that participates in precursor RNA splicing (9). Both NEAT1_1 and NEAT1_2 are induced in an anoxic environment (10). A recent study reports that p53 induces the formation of NEAT1 lncRNA-containing paraspeckles that modulate the replication stress response and chemosensitivity (11). NEAT1 is also a p53-dependent lncRNA in chronic lymphocytic leukemia (12). NEAT1 can also be induced by poly I:C triggered by the toll-like receptor3-p38 pathway after infection with virus (13). Previous studies have shown that NEAT1 is an oncogene in various cancers, such as urinary bladder cancer, colorectal cancer, and nasopharyngeal carcinoma (14–17). In addition to lncRNA, many tumor pathways have been shown to play an important role in glioma progression and invasion. Abnormal activation of the canonical WNT/β-catenin pathway has been reported to participate in malignant progression and has been implicated in the poor prognosis of malignant tumors (18) by regulating multiple aspects of cancer formation and growth. Axin2, GSK3B, and ICAT (CTNNBIP1), as WNT/β-catenin negatively regulated factors, play an important part in the canonical WNT/β-catenin pathway by promoting the degradation of β-catenin and repressing T-cell factor/lymphoid enhancer factor (TCF/LEF)–β-catenin transcriptional activity in vitro (19–21). However, the relationship between NEAT1 and the WNT/β-catenin pathway is rarely reported. In this study, we found that the EGFR/NEAT1/EZH2/β-catenin axis, conferring an oncogenic function in glioblastoma, might be a novel therapeutic target in glioblastoma.

MRNA expression data sets and the associated clinical information were downloaded from the following websites: CGGA (http://www.cgga.org.cn), Rembrandt (http://caintegrator.nci.nih.gov/rembrandt/), GSE16011 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE16011), and TCGA (http://cancergenome.nih.gov/). DNA methylation and copy number data of GBM were downloaded from TCGA website (http://xena.ucsc.edu/getting-started/). The population characteristics of the 4 data sets are in Supplementary Tables S1 to S4.

N5, N9, and N33 patient-derived cells were obtained from Professor Fan from BNU Beijing Normal University. N5, N9, and N33 cells were grown in DMEM/F12 (1:1; Gibco) and supplemented with 10% FBS and 1% penicillin–streptomycin. Cells were grown at 37°C in 5% CO₂. For cell treatments in several experiments, 15 ng/mL EGF was mixed with DMEM/F12 (1:1) without 10% FBS for 24 hours. All glioblastoma cells except in vivo cultures were maintained for less than eight passages. Sequencing and clinical data are listed in Supplementary Tables S5 and S6.

Lentiviruses, such as Cas9, GFP vector, RFP vector, and luciferase lentivirus, which were for in vivo experiments, were obtained from Genechem. Lentivirus such as EGFRvIII, HOTAIR, HOTAIR 5′ domain, and the HOTAIR 5′ domain mutant were purchased from Shanghai Genepharma. siRNAs for EZH2 were purchased from Shanghai Genepharma. sgRNAs for NEAT1 were purchased from Shanghai Genepharma. All sequences are provided in Supplementary Tables S7 and S8.

The CCK8 (Cell Counting Kit-8) assay was used to evaluate cell proliferation. Cells were seeded at 2 × 10² cells per well in 96-well plates. CCK8 was added at 0-, 24-, 48-, 72-, and 96-hour time-points. For colony formation assays, 250 cells were seeded in each well of six-well plates and were incubated for 14 days. The colonies were fixed with 4% paraformaldehyde, stained with crystal violet and counted. Invasion assays were performed using Boyden chambers (Corning) and 8-μm-sized pore membranes coated with Matrigel. Glioblastoma cells transfected with sgRNA or scramble were harvested 48 hours after transfection by 0.25% trypsinization. After the double staining with FITC-Annexin V and propidium iodide (PI), the cells were analyzed with flow cytometry (FACScan; BD Biosciences) equipped with CellQuest software (BD Biosciences).

Total RNA was extracted using TRIzol reagent (Sigma). The nuclear and cytoplasmic components were separated using 0.5% NP-40 (Solarbio) with an RNAase inhibitor (Promega), followed by extraction using TRIzol reagent (Sigma). Thereafter, cDNAs were synthesized using the PrimeScript RT Reagent Kit (Promega) according to the manufacturer's instructions. qRT-PCR was conducted in triplicate with LightCycler2.0 (Bio-Rad), and the expression was normalized to GAPDH as the total RNA and cytoplasmic RNA endogenous control. U6 was used as the nuclear RNA endogenous control. PCR primers were designed and synthesized using a primer designing tool (http://www.ncbi.nlm.nih.gov/tools/primer-blast/), and the primer sequences are listed in Supplementary Table S9. The relative quantification value for each target gene was expressed as 2−ΔΔC_t.

The pGL4.49[luc2P/TCF-LEF RE/Hygro] vector was purchased from Promega. The vectors were transfected into N5, N9, and N33 cells, and vector Renilla luciferase (Promega) served as an internal control. After 48 hours, cells were lysed and luciferase activities were measured using the Dual Luciferase Assay Kit (Promega) following the manufacturer's instructions.

Total genomic DNA was extracted using the TIAN amp Genomic DNA Kit (Tiangen). Conventional PCR and gPCR were performed using the Bio-Rad system, and 1% agarose gel electrophoresis (AGE) was used to detect the PCR products.

Western blot analysis was performed according to the manufacturer's instructions as previously described (22). Primary antibodies rabbit anti-p53, anti-PDGFR, anti-EGFR, rabbit anti-EGFRvIII, rabbit anti-β-catenin, rabbit anti-phospho-β-catenin (Ser675), mouse anti-H3, rabbit anti-Axin2, rabbit anti-EZH2, rabbit anti-STAT3, rabbit anti-Phospho-STAT3 (Ser705), rabbit anti-NF-κB (p65), and rabbit anti-phospho-NF-κB (p65) (Ser536) were purchased from Cell Signaling Technology. Primary antibodies rabbit anti-ICAT and mouse anti-GSK3B were purchased from Abcam. Primary antibody rabbit anti-TCF-4 was purchased from Santa Cruz Biotechnology. Mouse anti-GAPDH antibody was purchased from ProteinTech.

The cells were lysed using Western and IP lysis buffer (Beyotime Biotechnology) and incubated with 40 μL of protein-A/G PLUS-Agarose beads (Millipore), 1 μg of rabbit anti-β-catenin (Cell Signaling Technology), and anti-TCF-4 (Santa Cruz Biotechnology) at 4°C overnight. After washing three times with RIPA buffer, the samples were analyzed by Western blot analysis.

Antibodies against phospho-β-catenin (Ser675), β-catenin, and ICAT (1:300 dilution) and the secondary antibodies Alexa Fluor 488 and 594 (Life Technologies Corp.) were employed in immunofluoresence staining. Microscopy analysis was performed using an FV-1200 laser scanning confocal microscope.

All chromatin immunoprecipitation (ChIP) experiments were executed using the Millipore EZ-Magna ChIP Kit (catalog no. 17-371). Briefly, 2–3 × 10⁵ cells were cross-linked with 1% formaldehyde for 10 minutes at room temperature. The cross-linking was then quenched with 10 × glycine. Chromatin was sonicated in lysis buffer to 200–1,000 bp, and the extraction of ChIP DNA was performed as per the kit's protocol. The antibodies utilized included H3K27Me3, STAT3, and NFκB (p65) (Cell Signaling Technology, 1:50 dilution). The primer sequences are given in Supplementary Table S9.

RNA binding protein (RBP) immunoprecipitation (RIP) experiments were performed using the RNA Immunoprecipitation Kit (BersinBi) according to the manufacturer's instructions. Rabbit anti-EZH2 and anti-DNMT1 antibodies (1:50; Cell Signaling Technology) were used for experiments. Lysates were incubated with the antibody for 4 hours in 4°C. The coprecipitated RNAs were detected by real-time PCR.

Chromatin isolation using the RNA purification (ChIRP) experiments were performed using the ChIRP Kit (BersinBi). Complementary DNA oligonucleotides against NEAT1 were constructed using BLAST and 3′-end biotinylated to capture DNA–RNA hybrids on streptavidin-coated magnetic beads. Cells were incubated in 1% glutaraldehyde and were then rinsed with phosphate-buffered saline. Probes (100 pmol) were added to 1 mg of precleared chromatin and were mixed by end-to-end rotation at 65°C for 10 minutes, 25°C for 30 minutes, 50°C for 5 minutes, and 25°C for 90 minutes. The beads were resuspended in 95 μL of RNA PK buffer at pH 7.0. Next, 5 μL of proteinase K was added and incubated at 50°C for 45 minutes with end-to-end shaking, and Trizol reagent was used to collect RNA. For protein, 100 μL of washed/blocked streptavidin magnetic beads were added to the samples (30 minutes). The beads were treated with Urea CHAPS buffer for 1 hour at room temperature, and proteins were analyzed by Western blot analysis. Sequences of biotin probes were listed in Supplementary Table S10.

In situ hybridization histochemistry (RNA ISH) experiments were performed using the RNA In Situ Hybridization Kit (BersinBi) according to the manufacturer's instructions. Briefly, glioblastoma cells were fixed in 4% paraformaldehyde for 20 minutes and washed with distilled water. Fixed cells were treated with pepsin (1% in 10 mmol/L HCl), and the cells were further incubated with 20 nmol/L ISH probe in hybridization buffer (100 mg/mL dextran sulfate, 10% formamide in 2 × SSC) at 90°C for 3 min. Hybridization was performed at 37°C for 18 hours, and the slide was washed and incubated with digoxin antibodies for 1 hour. Finally, the slide was mounted with diaminobenzidine (DAB) for detection. RNA ISH probes were designed and synthesized by BersinBi, and the probe sequence is listed in Supplementary Table S8.

BALB/c-A nude mice at 3 to 4 weeks of age were purchased. To establish the intracranial tumor model, 3 × 10⁵ N33 cells were separately implanted stereotactically using cranial guide screws. Bioluminescence imaging was used to detect intracranial tumor growth mice on days 7, 14, and 21. In the meantime, overall mouse survival time was monitored. Twenty-five days after implantation, mice were randomly sacrificed in each group. The brains of the mice were carefully extracted and fixed in 10% formalin. Next, they were embedded in paraffin for hematoxylin and eosin (H&E) and IHC. At the end of the experiment, all the mice were dead, and the mouse overall survival curves were recorded according to the Kaplan–Meier method. In addition, 3 × 10⁵ mixed cells labeled with GFP and RFP were implanted stereotactically. After 21 days, brains of the mice were carefully extracted, fixed in 10% formalin and dehydrated in 30% sucrose. Next, they were embedded in optimal cutting temperature (OCT compound) and sliced for microscope observation.

The chi-squared test was used to determine whether there was a significant difference between the two groups. Survival curves were drawn using Kaplan–Meier survival plots, and the log-rank test was used to test significance. The correlation in tissues was examined using two-sided Pearson correlation. Statistical significance was determined using Student t test or ANOVA for functional analysis. Pathway and GO analysis were performed using DAVID (http://david.abcc.ncifcrf.gov/). Heat maps were constructed using Gene Cluster 3.0 and Gene Tree View software. Gene Set Enrichment Analysis (GSEA) was used. All statistical analysis were performed using SPSS 22.0 software and GraphPad Prism 6. The error bars in the figures represented SD. A P value < 0.05 was regarded as statistically significant.

To identify the lncRNAs that were involved in gliomagenesis, we selected 202 lncRNAs, most of which were according to Chiara Pastori, Zhou Du, and Brian J. Reon (6, 23, 24) and overlapped with glioma genome atlas public databases, such as the CGGA, Rembrandt, TCGA, and GSE16011 data sets (Fig. 1A). We placed these overlapping lncRNAs into the CGGA (n = 81), Rembrandt (n = 474), TCGA (n = 702), and GSE16011 (n = 246) data sets and classified them based on EM/PM (5). Next, we found that the NEAT1 lncRNA was highly expressed in the EM in the CGGA, Rembrandt, TCGA, and GSE16011 data sets (Fig. 1B). The ROCs of four data sets indicated that NEAT1 belonged to the EGFR module (Supplementary Fig. S1A). Status of EGFR (or EGFRvIII), p53, and PDGFR were validated through Western blot analysis, as shown in Supplementary Fig. S1B. Thus, we hypothesized that NEAT1 could be induced by EGFR activity. Sequencing data of patient-derived cells were displayed (Fig. 1C). In our study, the NEAT1_1 and NEAT1_2 expression levels were higher in lenti-EGFRvIII and the EGF (15 ng/mL) stimulation groups than in the control groups in N5 and N9 cells (Fig. 1D). Compared with the control groups, the EGFRvIII and EGF stimulation groups showed elevated levels of p-STAT3 (Tyr705) and p-NF-κB (Ser536) (Fig. 1E). The ChIP assay showed that STAT3 and NF-κB could bind to the NEAT1 promoter region (Fig. 1F). ChIP-qPCR showed that STAT3 and NF-κB binding activity were much higher in the lenti-EGFRvIII and EGF-stimulation groups than in the control groups (Fig. 1G). These results indicated that NEAT1 could be regulated by EGFR pathway activity.

To investigate the potential mechanisms of NEAT1 in glioblastoma cells, correlation analysis was performed in 109 and 172 samples with GBM mRNA databases from the CGGA and TCGA, and a cluster (|r| > 0.4, P < 0.01) of NEAT1-associated genes was obtained (Fig. 2A). Next, pathway and GO analysis were performed. Pathway and GO analysis using the NEAT1 positive correlation genes were conducted. We found that NEAT1 was associated with EGFR-driven tumor signaling (Supplementary Fig. S1C and S1D). In the meantime, we found that the negative-associated gene expression profiles were associated with the WNT/β-catenin pathway (Fig. 2B and C). GSEA analysis showed that NEAT1 negative-associated genes were primarily involved in the WNT/β-catenin signaling pathway (Fig. 2D). Further, correlation analysis revealed that the WNT/β-catenin negative regulatory factors Axin2, ICAT, and GSK3B had a significant negative correlation with NEAT1 in the CGGA data set (Fig. 2E).

LncRNAs have widely varying subcellular distributions; some reside predominantly in the nucleus, cytoplasm, or in both compartments. The RNAi-based gene-knockdown method is not effective to knockdown nuclear lncRNA (25). In our experiment, we proved that NEAT1 was a nuclear-enriched abundant lncRNA in N5, N9, and N33 cells (Supplementary Fig. S2A). The CRISPR/Cas9 system has been reported to successfully knockdown various lncRNAs (26–28). Therefore, we designed a strategy to knockdown NEAT1 by dual sgRNAs (Supplementary Fig. S2B). After sgRNAs infection of N5, N9 and N33 cells, we first amplified the dual-gRNA-targeted region by genomic PCR using primers outside of two gRNAs to confirm the knockdown of NEAT1. We next detected the deletion band (Supplementary Fig. S2C). qRT-PCR was then performed to detect NEAT1 expression. The results demonstrated a substantial reduction of NEAT1_1 and NEAT1_2 in N5, N9 and N33 cells (Supplementary Fig. S2D). We used qRT-PCR to detect the NEAT1_1 and NEAT1_2 levels in the nucleus and cytoplasm, respectively, and found that NEAT1_1 and NEAT1_2 dramatically decreased according to the dual-sgRNA-CRISPR/Cas9 system both in the nucleus and cytoplasm (Supplementary Fig. S2E). NEAT1 staining was attenuated in the nuclei of dual-sgRNA transfected cells, which highlighted that NEAT1 was a nuclear-enriched abundant lncRNA in glioblastoma and could be inhibited by the CRISPR-Cas9 system (Supplementary Fig. S2F). The malignant behaviors of glioblastoma, including proliferation and invasion, were inhibited after NEAT1 was knocked down (Supplementary Fig. S3A–S3C). To mimic the complexity of this heterogeneous disease, we used co-culture of scramble (e.g., GFP) with sgRNA1+2 group cells (e.g., RFP; for all three cells: N5, N9, N33) in an in vitro invasion assay. The results showed that NEAT1 KO cells are less invasive in vitro (Supplementary Fig. S3D). Knockdown of NEAT1 also induced cell apoptosis (Supplementary Fig. S3E).

The WNT/β-catenin signal is transduced through several cytoplasmic relay components to β-catenin, which enters the nucleus and forms a complex with T-cell factor/lymphoid enhancer factor (TCF/LEF) to activate the transcription of WNT target genes (29). Correlation analysis implied that the expression of NEAT1 was associated with the WNT/β-catenin negative regulatory factors Axin2, ICAT, and GSK-3B (Fig. 3A). ICAT has been reported to inhibit glioblastoma cell proliferation by suppressing WNT/β-catenin activity, which is significantly downregulated in high-grade glioma patients compared with normal brain and low-grade glioma patients (19). Axin2 participates in a negative feedback loop that could serve to limit the duration of the WNT/β-Catenin signal. GSK-3B phosphorylates β-catenin, thereby flagging it for recognition and destruction by the ubiquitin/proteasome machinery (30). The NEAT1 knockdown group showed that the mRNA levels of Axin2, ICAT, and GSK3B were increased in N3, N5, and N9 cells (Fig. 3B; Supplementary Fig. S4A). The protein levels of Axin2, ICAT, and GSK3B were also increased in the downregulated NEAT1 groups (Fig. 3C; Supplementary Fig. S4B). Confocal results showed that ICAT levels increased after knockdown of NEAT1 (Fig. 3D; Supplementary Fig. S4C). Furthermore, the total protein levels of β-catenin and p-β-catenin (Ser675) were decreased in the NEAT1 downregulated group (Fig. 3E; Supplementary Fig. S5A). The TCF-LEF reporter system was used to examine the activity of WNT signaling. Our results showed that downregulated NEAT1 could attenuate WNT/β-catenin signaling activity (Fig. 3F; Supplementary Fig. S5B). In the meantime, we measured the cellular distribution of β-catenin and p-β-catenin (Ser675). Downregulated NEAT1 triggered a reduction of nuclear β-catenin and p-β-catenin (Ser675; Fig. 3G; Supplementary Fig. S5C). Moreover, the immunofluorescence results demonstrated that β-catenin and p-β-catenin (Ser675) expression in the nucleus and cytoplasm was markedly decreased after NEAT1 was knocked down (Fig. 3H; Supplementary Fig. S5D). Coimmunoprecipitation assay verified that the NEAT1 knockdown group was able to attenuate β-Catenin binding activity with TCF-4 (Fig. 3I; Supplementary Fig. S5E). The above results demonstrate that knockdown of NEAT1 expression regulated WNT/β-catenin through repressing WNT/β-catenin pathway negative regulatory factors such as Axin2, ICAT, and GSK3B and helping β-catenin nuclear transport in vitro. Next, we performed Pearson correlation analysis for the association between Axin2, ICAT, and GSK3B methylation levels and NEAT1 expression in glioblastoma. However, we did not find an obvious positive association between NEAT1 and the methylation levels of the WNT/β-catenin–positive regulatory factors Axin2, ICAT, and GSK3B (r < 0.2, Supplementary Fig. S6A–S6C). Moreover, we observed that NEAT1 expression did not have a negative correlation with gene copy numbers (Supplementary Fig. S6D–S6F). In this situation, we suspected that NEAT1 may perform like a scaffold with histone-modifying enzymes to regulate Axin2, ICAT, and GSK3B expression.

Evolutionary conservation is commonly used as an indicator of the functional significance of newly discovered genes. LncRNA conservation includes four dimensions: sequence, structure, function, and expression from syntenic loci (31). Through a sequence alignment analysis, we found that NEAT1 was an evolutionarily conserved lncRNA (Supplementary Fig. S7A). Some lncRNAs affect the expression of protein-coding genes by interacting with the PRC2 subunit EZH2 (32–34). In starBase v2.0, we found that EZH2 in murine embryonic stem cells could bind to Neat1 (http://starbase.sysu.edu.cn/; Supplementary Fig. S7B). In addition, we predicted the interaction probabilities of NEAT1 and RBPs via RNA–protein interaction prediction (http://pridb.gdcb.iastate.edu/RPISeq/) and found that lncRNAs such as HOTAIR and MALAT1 (metastasis associated with lung adenocarcinoma transcript 1), NEAT1_1, NEAT1_2, Neat1_1, and Neat1_2 could potentially bind to EZH2 (Supplementary Fig. S7C). Thus, we hypothesized whether NEAT1 could bind to EZH2 in glioblastoma cells. To investigate the potential interaction between lncRNA NEAT1 and EZH2, RIP, and ChIRP (Fig. 4A), assays were carried out. We found that both NEAT1_1 and NEAT1_2 were significantly enriched with the EZH2 antibody compared with IgG (control antibody; Fig. 4B), but did not interact with DNMT1 (Supplementary Fig. S8A) in N9 and N33 cells; HOTAIR and MALAT1 were used as positive controls. The downregulated NEAT1 group attenuated NEAT1-binding activity (Fig. 4C). Moreover, the ChIRP results showed that NEAT1 could be pulled out with biotinylated probes and was markedly enriched with EZH2 compared with the negative control (Fig. 4D and E). Next, we amplified HOTAIR, HOTAIR 5′ domain (35, 36), and HOTAIR 5′ domain mutants (Supplementary Fig. S8B). We found no observable fold change in the expression of MALAT1, NEAT1_1, and NEAT1_2 (Supplementary Fig. S8C). However, RIP experiments showed that the HOTAIR and HOTAIR 5′ domain overexpression groups impaired the binding activity of NEAT1_1, NEAT1_2, and MALAT1 (Fig. 4F). These results further indicated that NEAT1 was an EZH2-binding lncRNA.

To reveal how NEAT1 regulated WNT/β-catenin through EZH2, we separately designed three paired primers across the promoter region (2000 bp) of Axin2, ICAT, and GSK3B and then performed the ChIP-qPCR assay to examine the regulatory mechanisms. The results showed that knockdown of NEAT1 group decreased the H3K27Me3 binding ability in N33 cells (Fig. 5A). Then, we knocked down EZH2 through siRNA and found that the β-catenin level decreased (Fig. 5B). RIP and ChIRP assays showed that the interactivity between NEAT1 and EZH2 decreased in siEZH2 group (Fig. 5C and D). The mRNA and protein levels of Axin2, ICAT, and GSK3B increased again in the siEZH2 groups (Fig. 5E and F). Downregulation of EZH2 decreased the H3K27Me3 binding ability of the promoter region of Axin2, ICAT, and GSK3B in N33 cells (Fig. 5G). These results showed that NEAT1 was the scaffold of EZH2, which played an important role in regulating the WNT/β-catenin pathway in glioblastoma cells.

To further verify the role of NEAT1 in vivo, we constructed orthotopic mouse models using the N33 cells. In these models, the NEAT1 downregulated group showed a significant reduction of the intracranial tumor volume compared to the control groups (Fig. 6A). H&E staining displayed that the tumors in the control groups exhibited extensive branch-like growing patterns that spread into the surrounding tissue (Fig. 6B). We used coinjection of the scramble (e.g., GFP) with sgRNA1+2 (e.g., RFP) group cells (for all three cells: N5, N9, N33) to mimic the complexity of this heterogeneous disease. The results showed that NEAT1 KO group cells were less proliferative and invasive in vivo (Fig. 6C). Meanwhile, NEAT1 expression was markedly decreased (Fig. 6D) in N33 tumors. Downregulated NEAT1 in N33 tumors was associated with prolonged survival (Fig. 6E). Furthermore, IHC showed that knockdown of NEAT1 decreased the expression of β-catenin, p-β-catenin, and Ki-67 but increased the Axin2, ICAT, and GSK3B levels (Fig. 6F) in N33 tumors, consistent with the in vitro results.

In recent years, by RNA sequencing and annotation by the GENECODE project, thousands of lncRNAs have been discovered, the functions of which have not been established. EGFR module/PDGFR module–based glioma classification forms a framework towards establishing molecular diagnostic tools and identifying novel therapeutic targets to combat gliomas (5). These results indicate that NEAT1 functions as an oncogene in glioblastoma. In this study, we compared the lncRNA profiles of glioma tissues from different grades using public data sets. We discovered that the lncRNA NEAT1 was driven by EGFR pathway activity. NEAT1 is an essential architectural component of paraspeckle nuclear bodies. The tumor anoxia microenvironment can activate NEAT1 and induce nuclear paraspeckle formation through HIF-2α, leading to breast cancer cell survival (10). In prostate cancer, NEAT1 has an oncogenic character and alters the epigenetic landscape of target gene promoters to favor transcription (37). NEAT1 showed the malignant progression in glioma stem cells (38) and clinical significance (23, 24, 39). The results described above strongly prove the oncogenic role of NEAT1. In our study, after knocking down NEAT1 through the dual-CRISPR/Cas9 system, we found that the malignant behaviors of glioblastoma were inhibited.

Next, we sought to identify the underlying molecular mechanisms by which NEAT1 regulated downstream effectors in glioblastoma. To examine this, we analyzed the differential expression of genes in the CGGA and TCGA data sets. Here, we discovered that the WNT/β-catenin pathway was affected by NEAT1. The WNT/β-catenin signaling pathway is associated with diverse processes that are involved in early embryonic patterning and regulation of stem cell self-renewal and differentiation (40). It is not strange that the WNT/β-catenin signal is found in diverse cancers because the aberrant regulation of WNT has a dramatic effect on cell differentiation and fate decisions. Our previous study showed that β-catenin/TCF-4 activity regulated AKT1 expression by binding to the AKT1 promoter, thereby increasing glioblastoma cell growth (41). In 2012, our laboratory used ChIP-seq assay of TCF4 and STAT3 and data mining of patient cohorts to derive the molecular subtypes of GBM (42). Moreover, the lncRNA MALAT1 has been reported to be regulated in non-canonical WNT/Ca²⁺/signaling (43). In this study, we disclosed that a high level of NEAT1 negatively regulated WNT negative signaling regulation factors (Axin2, ICAT, and GS3KB). A previous study demonstrated that ICAT is a WNT/β-catenin pathway inhibitor (44). Axin2 itself is a direct target of the WNT signaling pathway and therefore serves to control the duration and/or intensity of WNT signaling through a negative feedback loop (45). GSK3B could form a destruction complex with APC, AXIN, and CSNK1A1, so that β-catenin is phosphorylated at Ser33, Ser37, and Thr41, leading to its ubiquitination by bTRCP and degradation by the proteasome (46). In head and neck squamous cell carcinoma (HNSCC), the loss of GSK3B expression results in the stabilization and accumulation of c-MYC and aggressive HNSCC (47). Our TCF reporter activity, Western blot analysis and immunofluorescence results confirmed that knockdown of NEAT1 repressed the WNT/β-catenin pathway activity through down-regulating the mRNA and protein levels of ICAT, GSK3B, and Axin2.

To investigate how NEAT1 regulated the WNT/β-catenin pathway, we found that NEAT1 was a conserved lncRNA in different species and EZH2 was a potential NEAT1-binding protein. It was discovered that the lncRNA HOTAIR mediates its effect by interacting with PRC2 though 1 to 500′, enhancing the methylation of histone H3 lysine 27 (H3K27) and leading to the silencing of glioblastoma suppressor genes (48). The expression of EZH2 is high in high-grade glioma, promoting glioma progression, invasion, and metastasis (49). Thus, it was necessary to investigate whether NEAT1 could bind to EZH2 to silence suppressor gene expression. As we expected, our RNA immunoprecipitation and ChIRP results showed that NEAT1 interacted with EZH2 in glioblastoma cancer cell lines in vitro. Knockdown of NEAT1 decreased the trimethylation modification of H3K27 in the Axin2, ICAT, and GSK3B promoter region. Finally, our in vivo experiments showed tumor growth and invasion were inhibited and mouse lifespan was prolonged after NEAT1 was knocked down.

In summary, our current work revealed that the glioblastoma-associated lncRNA NEAT1 was an oncogenic factor that was regulated by EGFR pathway, promoting tumorigenesis by serving as a scaffold and recruiting the chromosome modification enzyme EZH2 to silence target-specific genes (Axin2, ICAT and GSK3B) to promote β-catenin nuclear transport. Thus, the WNT/β-catenin pathway was rekindled in glioblastoma (Fig. 6G). In conclusion, we highlighted that the EGFR/NEAT1/EZH2/β-catenin axis in glioblastoma conferred an oncogenic function in glioblastoma that might offer a novel therapeutic target.

No potential conflicts of interest were disclosed.

Conception and design: C. Kang, M. Li, C. Jiang

Development of methodology: Q. Chen, Y. Wang, C. Kang, C. Jiang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Q. Chen, J. Cai, C. Jiang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Q. Chen, J. Cai, J. Yang, M. Li, C. Jiang

Writing, review, and/or revision of the manuscript: Q. Chen, J. Cai, Q. Wang, M. Liu, J. Yang, C. Kang, M. Li, C. Jiang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Q. Chen, J. Cai, Q. Wang, Y. Wang, J. Zhou, C. Kang, C. Jiang

Study supervision: C. Kang, M. Li, C. Jiang

The authors thank Professor Xiaolong Fan for the patient-derived cells and Professor Tao Jiang for providing support of data and bioinformatics analysis. This research work was supported by the Research Special Fund for Public Welfare Industry of Health (no. 201402008), the National Key Research and Development Plan (no. 2016YFC0902502), and the National Natural Science Foundation of China (no. 81572701, 81372700, 81702972).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.