Downregulation of MEG3 and upregulation of EZH2 cooperatively promote neuroblastoma progression

Abstract Neuroblastoma (NB), an embryonic tumour originating from sympathetic crest cells, is the most common extracranial solid tumour type in children with poor overall prognosis. Accumulating evidence has demonstrated the involvement of long non‐coding RNA (lncRNA) in numerous biological processes and their associations with embryonic development and multiple diseases. Ectopic lncRNA expression is linked to malignant tumours. Previous studies by our team indicate that MEG3 attenuates NB autophagy through inhibition of FOXO1 and epithelial‐mesenchymal transition via the mTOR pathway in vitro. Moreover, MEG3 and EZH2 negatively regulate each other. In present study, we first collected 60 NB tissues and 20 adjacent tissues for Quantitative real‐time polymerase chain reaction (Q‐PCR) experiments and performed clinical correlation analysis of the results. At the same time, nude mice were used for subcutaneous tumour formation to detect the effect of MEG3 in vivo. Two NB cell lines, SK‐N‐AS and SK‐N‐BE(2)C, were overexpressed MEG3 and rescued with EZH2 and then were subjected to proliferation, migration, invasion, apoptosis and autophagy experiments. RNA‐binding protein immunoprecipitation (RIP) and Co‐Immunoprecipitation (Co‐IP) experiments were performed to explore the molecular mechanism of MEG3 and EZH2 interaction. Q‐PCR revealed that MEG3 expression was negatively correlated with INSS stage and risk grade of NB. Moreover, MEG3 overexpression was associated with inhibition of NB growth in vivo. MEG3 exerted an anti‐cancer effect via stimulatory effects on EZH2 ubiquitination leading to its degradation. Conversely, EZH2 interacted with DNMT1 and HDAC1 to induce silencing of MEG3. The EZH2 inhibitor, DZNep, and HDAC inhibitor, SAHA, displayed synergistic activity against NB. Combined treatment with DZNep and SAHA inhibited proliferation, migration and invasion of NB through suppression of the PI3K/AKT/mTOR/FOXO1 pathway. In conclusion, downregulation of MEG3 and upregulation of EZH2 forms a feedback loop that concertedly promotes the development of NB. Combined blockage of EZH2 and HDAC1 with the appropriate inhibitors may therefore present an effective treatment strategy for NB cases with low MEG3 and high EZH2 expression.


| INTRODUC TI ON
Neuroblastoma (NB) is an embryonal tumour arising in the developing sympathetic nervous system with typical presentation in adrenal glands and/or sympathetic ganglia. 1,2 As the most common extracranial solid tumour type in children, NB is usually diagnosed in the first year of life with an average of 25-50 cases per million individuals, accounting for disproportionate morbidity and mortality among paediatric tumours. 3,4 Although the outcomes of NB have improved owing to updated and effective interventions, long-term survival rates in children with high-risk NB remain relatively poor. 1,5 Accumulating studies on the mechanisms underlying pathogenesis suggest that abnormalities at the genome, epigenome and transcriptome levels are involved in the occurrence of NB. [6][7][8] However, our understanding of the complex pathogenic pathways underlying tumorigenesis of NB is still evolving and further research is required for effective diagnosis and identification of therapeutic targets.
With the rapid development of high-throughput sequencing, in addition to abnormal expression of protein-coding genes, dysregulation of non-coding RNAs, in particular, long non-coding RNAs (lncRNAs), has been shown to play important roles in tumour development. [9][10][11] Similar to protein-coding genes, lncRNAs possess oncogenic and tumour suppressive activities. 12,13 However, limited studies to date have investigated the involvement of lncRNAs in the development of childhood tumours, including NB.
The lncRNA, maternally expressed gene 3 (MEG3), acts as a tumour suppressor in various cancer types, including liver, lung, nasopharyngeal and stomach cancer. 14,15 Current research on MEG3 in relation to NB is mainly focused on genetic polymorphisms and genetic susceptibility. Xia and colleagues identified two MEG3 polymorphisms, rs7158663 G>A and rs4081134 G>A, in a study on 392 children with NB and 783 control subjects via the Taqman method.
Stratified analysis revealed that subjects carrying the rs4081134 AG/AA genotype were susceptible to NB in subgroups older than 18 months and clinical stage III + IV. In a comprehensive investigation of MEG3 gene polymorphisms, children over 18 months of age simultaneously carrying both these risk genotypes were more likely to develop NB than those with only one or no risk genotype. 16 However, the specific functions and mechanisms of action of MEG3 in NB remain to be established.
Polycomb repressive complex 2 (PRC2) comprises EZH2, EED and SUZ12, among which, EZH2 is the only active subunit with a critical role. [17][18][19] The EZH2 catalytic SET domain catalysing histone 3 lysine 27 tri-methylation (H3K27me3) is reported to bind and silence specific tumour suppressor genes. 20,21 Earlier, Kamijo and co-workers demonstrated upregulation of EZH2 in NB in association with poorer prognosis and overall survival. Furthermore, EZH2 was shown to affect NB differentiation through regulation of NTRK1. 22 Another study by Bownes et al. revealed decreased NB proliferation and in vivo tumour growth following inhibition of EZH2. 23 The role of the lncRNA MEG3 in NB was previously explored by our team. Consistent with findings from other studies, our experiments supported a tumour suppressor role of MEG3, showing a negative correlation between MEG3 expression and development of NB in vitro. While our earlier results suggest that MEG3 and EZH2 mutually regulate and jointly promote progression of NB by forming a negative feedback loop, the precise pathways remain to be clarified. 24 The main objective of the present study was to uncover the mechanistic pathways between MEG3 and EZH2 and explore the potential therapeutic utility of combinations of drugs targeting both genes against NB.

| Cell culture
The human NB cell lines, SK-N-BE(2)-C and SK-N-AS, were kind gifts

| Quantitative real-time polymerase chain reaction
Sixty NB tissues and twenty control tissues were obtained from patients subjected to surgery at Children's Hospital of Fudan University.
Informed consent was acquired from every patient. Total RNA was isolated from the above tissues or cell lines using TRIzol reagent (Takara) according to the manufacturer's instructions. cDNA was generated from reverse-transcribed RNA using a specific reverse transcription kit, in keeping with the manufacturer's protocol. Quantitative realtime polymerase chain reaction (Q-PCR) was performed using Hieff UNICON ® Universal Blue qPCR SYBR Green Master Mix (Yeasen) with a Roche instrument for determination of relative RNA levels.

| Animal assays
Control and overexpression MEG3 SK-N-AS cells were cultured and washed three times using PBS, resuspended, and then counted in PBS. The cell density was adjusted to 5 × 10 7 cells/ml. Twelve nude mice (BALB/c) were randomly divided into two groups and subcutaneously inoculated with 5 × 10 6 cells (100 μl cell suspension), then tumour volume was recorded once a week. After 4 weeks, mice were sacrificed, the tumours were taken out, photographed, weighed and fixed in formalin solution (only 50% of mice successfully grow tumours). The animal assays were approved by the Ethics Committee of Children's Hospital of Fudan University.

| Western blot
Total proteins were extracted using NP40 lysis buffer (Beyotime) containing PMSF (Beyotime) to protect against degradation of protein. Proteins were quantified using the Bradford assay with the Coomassie Brilliant Blue G250 reagent kit (Beyotime). After denaturation of the extracted protein by boiling at 95°C for 10 min, equal amounts of all protein samples were separated via SDS-PAGE (New Cell & Molecular Biotech) according to molecular weight of target protein and transferred to membranes. Next, membranes were blocked with 8% skimmed milk for 1 h and incubated overnight with corresponding primary antibodies at 4°C. On day 2, all reactions were operated at room temperature. After washing with TBST three times, membranes were incubated with HRP-conjugated secondary antibodies for 1 h. Immunoblot signals were obtained from an imaging system using an Enhanced Chemiluminescent Reagent kit (New Cell & Molecular Biotech). GAPDH or β-actin was selected as the loading control. All the antibodies used in this study are listed in Table S1.

| Immunohistochemical analysis
Human NB tumour and paired control tissues were collected from the Children's Hospital of Fudan University. Samples were soaked, embedded, dewaxed and incubated in citric acid antigen retrieval buffer, followed by 3% BSA for blocking. Slides were incubated overnight at 4°C with primary antibodies specific for Ki67 (1:200; Servicebio) and EZH2 (1:50, CST). On day 2, slides were washed three times and incubated with secondary antibody (1:200; Servicebio) for 1 h at room temperature. DAB colour developing solution was added, followed by haematoxylin staining. Finally, slides were dehydrated and mounted and images were obtained via microscopy (Thermo).

| Cell proliferation and colony formation
Cell proliferation ability was measured using Cell Counting Kit-8

| Migration and invasion assays
Migration and invasion assays were performed using 24-well plates containing 8 μm pore size transwell filter inserts with or without precoated diluted matrigel (1:5; Becton Dickinson). SK-N-AS cells at a density of 1 × 10 5 (migration) and 2 × 10 5 (invasion) and SK-N-BE (2)C cells at a density of 2 × 10 5 (migration) and 4 × 10 5 (invasion) diluted in serum-free medium were placed in the upper chamber and medium containing 30% FBS added to the lower chamber. After incubation for 48 h at 37°C, cells on the underside of the membrane were fixed with 4% PFA for 15 min and stained with 0.1% crystal violet solution within 20 min for further analysis. Penetrating cells from five random fields were counted under the microscope.

| Apoptosis detection
Cells were collected in a 6 cm culture dish, washed twice with PBS and digested with trypsin without EDTA. Next, 5 μl PE and 7-amino actinomycin D staining solution were added after fixation for 15 min in the dark. Samples were subjected to flow cytometry and analysed with FlowJo software.

| Plasmid construction
All short hairpin RNAs were designed using the website of and ΔSET EZH2 plasmids were synthesized by Shanghai Generay Biotech Co., Ltd. DNMT1 and HDAC1 overexpression plasmids were purchased from Shanghai Genomeditech and Shandong WZ Biotech Co., Ltd, respectively.

| RNA-binding protein immunoprecipitation
Cell pellets cultured in a 10 cm dish were collected, and the same volume of RNA-binding protein immunoprecipitation (RIP) lysis buffer was added to the tube. Samples were incubated overnight at −80°C after splitting on ice for 5 min. Next, protein A/G magnetic beads (Thermo) were washed five times with NT2 buffer and incubated with 5 μg antibody. After 2 h, samples were re-washed with NT2 buffer and mixed with supernatant fractions of the lysates overnight at 4℃. The next day, supernatants were washed five times with NT2 buffer and proteinase K buffer added for 30 min at 55℃. RNA was extracted with TRIzol reagent, and reverse transcription and Q-PCR were performed as described above.

| Co-immunoprecipitation experiments
NP40 lysates (1 ml) were added to 10 cm dish cells and proteins extracted using the conventional WB method. After incubation with 2 µg antibody for 2 h at 4°C, 30 µl protein A/G agarose beads (Santa Cruz) was added and inverted overnight. The next day, samples were washed thoroughly with NP40 lysis buffer (Regal Biology) five times, incubated in 30 µl of 2×SDS-PAGE sample loading buffer (Yeasen) and boiled at 95°C for 5 min for subsequent WB experiments.

| ChIP-seq and RNA-seq
For ChIP-seq, cells in a 10 cm dish were resuspended in 10 ml medium. Next, 270 μl of 37% formaldehyde solution was added at room temperature for 10 min, followed by 540 μl of 2.5 mol glycine for

| Statistical analysis
Every assay was repeated independently at least three times. Results were presented as the mean ± SD. Groups were compared using ttest, and p < 0.05 was considered statistically significant.

| MEG3 inhibits NB growth in vivo and EZH2 is highly expressed in NB
Previous experiments by our team support anti-tumour activity of MEG3 in NB. The present study was conducted on an expanded clinical sample size. Q-PCR analyses showed significant downregulation of MEG3 in NB compared to adjacent adrenal tissues ( Figure 1A).
In clinical correlation analysis, MEG3 was negatively correlated with INSS stage and risk grade of NB (Table 1, Table S2)

| Upregulation of EZH2 rescues the tumour inhibitory effects of MEG3
MEG3 has been shown to inhibit NB cell proliferation, migration, invasion and autophagy and promote apoptosis in our previous studies. 24 In the CCK-8 assay, wild-type EZH2, but not EZH2 depleted of the SET domain, facilitated cell proliferation (Figure 2A,B).
Similarly, EZH2, but not ΔSET EZH2, rescued colony formation ability  Figure 2J-N). In addition, electron microscopy results indicated that autophagy inhibited by MEG3 is increased by ectopic EZH2 and that depletion of SET leads to loss of this function ( Figure 2O). Therefore, we propose that MEG3 exerts anti-cancer activity through negatively regulating EZH2 in NB cells. Furthermore, the SET domain appears indispensable for EZH2 to exert its oncogenic effects.

| EZH2 promotes FOXO1-mediated autophagy and mTOR mediates epithelialmesenchymal transition
In addition to cell phenotype, we explored whether the signalling pathway inhibited by MEG3 could be reactivated by EZH2. Western blot experiments showed that EZH2, but not ΔSET EZH2, could reactivate FOXO1-mediated autophagy and mTOR-induced EMT ( Figure 3A,C,E). However, EZH2 exerted no effect on the autophagy markers ATG3 and ATG12. Next, EZH2 was depleted via short hairpin RNA, which led to suppression of both autophagy and EMT. Notably, FOXO1 and mTOR pathways were also inhibited upon downregulation of EZH2. In accordance with data from EZH2 silencing experiments, DZNep, an EZH2 inhibitor, induced significant suppression of FOXO1 and mTOR pathways and decreased that of autophagy and EMT markers. Conversely, upregulation of EZH2 in NB cells increased autophagy and EMT markers and activated the FOXO1 and mTOR pathways ( Figure 3B,D,F). The collective findings demonstrate that

MEG3-mediated inhibition of autophagy and EMT through FOXO1
and mTOR is achieved via EZH2, in particular, the SET domain.

| UCHL1 serves as a bridge mediating regulation of EZH2 by MEG3
Previously, we demonstrated that MEG3 promotes degradation of EZH2 via ubiquitination and interacts with the deubiquitinase UCHL1, based on CHIRP experiments ( Figure 5A). UCHL1 is a thiol protease containing a total of 223 amino acids (among which aa 83-176 comprise the catalytic site) that hydrolyses a peptide bond at the C-terminal glycine of ubiquitin. Using information from the database, we speculated that MEG3 (31-92 bp) binds UCHL1 at the 83-176 aa region ( Figure S1C). To examine this theory, truncation and deletion mutants of UCHL1 were generated ( Figure 5B). RIP experiments confirmed interactions between the domain comprising 83-176 aa of UCHL1 and MEG3 ( Figure 5C,D). Notably, MEG3 lacking 31-92 bp lost the ability to bind UCHL1 ( Figure 5E,F). Data from the Co-IP assay suggested that UCHL1 may bind the SET domain of EZH2 ( Figure 5G,H) and immunofluorescence experiments revealed co-localization of EZH2 and UCHL1 in NB cells ( Figure S1D). Downregulation of UCHL1 via short hairpin RNA or its inhibitor, LDN57444, induced a decrease in EZH2 and, conversely, its upregulation led to increased EZH2 expression ( Figure 5I). Furthermore,  suggest that EZH2, DNMT1 and HDAC1 form a complex that inhibits MEG3 expression ( Figure 6N).

| Combined treatment with DZNep and SAHA inhibits the malignant biological behaviour of NB through the PI3K/AKT/mTOR/FOXO1 pathway
Compared with control and single-drug treatment groups, cotreatment with DZNep and SAHA induced a more significant decrease in EZH2, DNMT1 and HDAC1 expression ( Figure 7A).

| DISCUSS ION
Previous experiments by our group showed that MEG3 inhibits NB autophagy through FOXO1 and EMT through mTOR in vitro. inhibits the PI3K/AKT pathway via EZH2, which is upstream of mTOR.
We identified the SET domain as a key player in EZH2 activity based on the finding that its depletion led to loss of the rescue effects of EZH2 inhibitor development. 42 The HDAC inhibitor, Vorinostat (SAHA, developed by Merck) has not only been approved as a treatment agent for T-cell lymphoma but is also undergoing phase 2 and 3 clinical trials for breast cancer and non-small cell lung cancer. 43 Since inhibitors of DNMTs, EZH2 or HDACs are all non-specific with regulatory effects on numerous genes, combinations of these drugs may effectively enhance efficacy and reduce side effects. 44 In the present study, we showed that compared with the control and sin- In conclusion, MEG3 plays a tumour suppressor role in NB.

Concomitant downregulation of MEG3 and upregulation of EZH2
promotes the occurrence and development of NB. Combined blockage of EZH2 and HDAC1 may be effective in treatment of NB cases with low MEG3 and high EZH2 expression.

ACK N OWLED G EM ENTS
Not applicable.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest to declare.

E TH I C A L A PPROVA L
The studies involving human tissues and animal experiments were reviewed and approved by The Ethics Committee of Childrens' Hospital of Fudan University. And all of the patients signed informed consent forms before study.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data from this study are available from the corresponding author.