Primary murine glioma stem cells (CSC2078 and CSC1589) were isolated from human glial fibrillary acidic protein (hGFAP)-Cre⁺ p53L/L phosphatase tensin homolog (Pten)L/⁺ mice as previously described (20); cell isolation was not conducted in the present study. The numbers of GSCs were encoded as the same as the code number of transgenic mice. The hGFAP-Cre transgene to delete p53 alone or in combination with Pten in all CNS lineages using conditional p53 and Pten alleles as previously described (20). Mice (n=42, 15-40 weeks old, 20-22 g, Dana-Farber Cancer Institute) were interbred and maintained on FvB/C57Bl6 hybrid background. 73% of hGFAP-Cre⁺; P53^lox/lox; Pten^lox/+ mice (between ages 15-40 weeks) presented with acute-onset neurological symptoms in clinical analysis. All 42 neurologically symptomatic mice harbored malignant gliomas were classified as anaplastic astrocytomas [World Health Organization (WHO) III, n=28, 66%] or GBM (WHO IV, n=14, 34%). In this study, all manipulations were performed with IACUC approval (20). Patient-derived glioma stem cells (GSCs) TS543 were a gift from Dr Cameron W. Brenner (Memorial Sloan Kettering Cancer Center) (21). CSC2078 and CSC1589 were cultured in neurobasal media (Dulbecco's Modified Eagle's medium/F12, HyClone; GE Healthcare Life Sciences) with epidermal growth factor (20 ng/ml, cat. no. 315-09, PeproTech, Inc.) and basic fibroblast growth factor (10 ng/ml, cat. no. 450-33, PeproTech, Inc.) or differentiation medium (neurobasal media with 1% FBS). TS543 cells were cultured in NeuroCult NS-A proliferation media (human; Stem Cell Technologies) supplemented with epidermal growth factor (20 ng/ml, cat. no. GMP100-15, PeproTech, Inc.) and basic fibroblast growth factor (10 ng/ml, cat. no. 100-18B, PeproTech, Inc.). Cells were cultured at 37°C in a 5% CO₂.

CSC2078 cells were seeded into a 6-well plate at 2×10⁵ cells per well overnight. The following day, CSC2078 were transfected with three independent Brd4-specific siRNAs (20 µM; siBrd4s-475, siBrd4s-1648, siBrd4s-3820) and a negative control siRNAs (20 µM) for 48 h. siRNAs was purchased from Gene Pharma and the sequences were described in Table SI. The transfection agent INTERFERin (Thermo Fisher Scientific, Inc.) was used according to the manufacturer's protocol. The efficiency of transfection was determined by western blotting.

To determine the dose response to JQ1, cells were seeded into a 96-well plate in triplicate at the density of 10,000 cells per well and adhered overnight. Cells were treated with serial dilutions of JQ1 at 37°C for 24 or 48 h. JQ1 was added by 2-fold serial dilutions (the concentration of JQ1 for CSC2078 and TS543: 0.05, 0.1, 0.2, 0.4 and 0.8 µM; the concentration of JQ1 for CSC1589: 0.0625, 0.125, 0.25, 0.5 and 1 µM). CSC2078 cells transfected with siRNAs for 24, 48 and 72 h. The Cell Counting Kit-8 (CCK-8) solution (10 µl, MedChem Express) was added to each well of the plate. Then, the plate was incubated at 37°C for 4 h. The absorbance was measured at 450 nm using a microplate reader. The Cell Counting Kit-8 (cat. no. HY-K0301) and JQ-1 (cat. no. HY-13030) was purchased from MedChem Express.

For the soft agar colony assay, 2,500 cells were plated in the upper layer for each well in 6-well plate in triplicate. The concentration of the lower layers of agar was 0.6% and the upper layers were 0.4%. The agar was mixed with the neurobasal medium or NeuroCult NS-A proliferation media at a ratio of 1:1. The number of colonies in five random microscopic fields were counted after 10-15 days of seeding GSCs and staining by 0.1% crystal violet for 1 h at room temperature. Images were captured using a light microscope at ×400 magnification in five random microscopic fields (BX71 Olympus Corporation).

Self-renewal capacity was measured by neural sphere formation assays. The cells were cultured into 6-well plates (100/well) in triplicate and treated with JQ1 (0.1, 0.2, 0.25 and 0.5 µM) for 24 h and transfected with siRNAs for 48 h at 37°C. The neural spheres were cultured in NSC proliferation media (HyClone; GE Healthcare Life Sciences) containing growth factors (epidermal growth factor, 20 ng/ml; basic fibroblast growth factor, 10 ng/ml) for 10 days at 37°C. After 10 days, neurospheres containing >50 cells were scored (22). The number of neural spheres in each well was counted under an IX71 light microscope (Olympus Corporation) at ×100 and 200 magnification. Data were presented as the percentage of sphere-forming cells relative to control.

Cells were seeded in 6-well plates (1×10⁵ cells/well) and treated with JQ1 (0.05, 0.1, 0.125, 0.2, 0.25 and 0.5 µM) and/or linifanib (5 µM, cat. no. HY-50751, MedChem Express) for 24 h. CSC2078 cells were transfected with siRNAs for 48 h. We fixed cells with 1 ml 70% ethanol overnight at 4°C. The collected cells were detected using a propidium iodide (PI)/RNase Staining Solution (cat. no. CY2001-P Sungene Biotech) according to the manufacturer's protocols. The cell cycle was analyzed using a flow cytometer (BD Biosciences) and ModFit LT 5.0 (Verity Software House).

Cells were seeded in 6-well plates overnight at a density of 1×10⁵ cells/well. Then, cells were treated with JQ1 and/or linifanib for 24 h and were transfected with siRNAs for 48 h. The vehicle-, JQ1/linifanib and JQ1 + Linifanib-treated cells were stained with Annexin V-fluorescein isothiocyanate and PI according to the instructions provided with the Apoptosis Detection Kit (cat. no. WLA001, Wanleibio Co., Ltd.). The vehicle-treated cells were treated with dimethyl sulfoxide (DMSO; maximum final concentration of 0.01%). The apoptosis rate was calculated as the sum of early apoptotic and late apoptotic cells Apoptosis was analyzed using flow cytometry.

A TUNEL assay was conducted using the DeadEnd™ Colorimetric TUNEL System (Promega Coporation) and DeadEnd™ Fluorometric TUNEL System (Promega Corporation). The DeadEnd™ Colorimetric TUNEL System was used to detect the apoptosis of CSC2078 treated with JQ1. Cells were treated with JQ1, or vehicle (DMSO at a maximum final concentration of 0.01%) for 24 h. The DeadEnd™ Fluorometric TUNEL System was used to detect the apoptosis of tumors of CSC2078-subcutaneous xenograft mice as described below. The TUNEL assay was performed according to the manufacturer's instructions. Images of cells and tumor tissues were captured using an Olympus BX53 microscope (light and fluorescence) at ×400 magnification. For quantitation, the TUNEL-positive cells were counted in five random area images for each sample using ImageJ 1.46r (National Institutes of Health).

CSC1589 cells were treated with JQ1 for 24 h. Then, the CSC1589 were fixed with 4% paraformal-dehyde for 5 min. Then, CSC1589 were stained with 1 mg/ml Hoechst (Beyotime Biotechnology Inc., Nantong, China) for 10 min at room temperature in the dark. Finally, the cells were washed three times with PBS and imaged by Olympus BX53 microscope (fluorescence) at ×400 magnification.

Total RNA of CSC2078 and TS543 cells was extracted using TRIzol (Thermo Fisher Scientific, Inc.) according to the manufacturer's protocols. Following isolation, total RNA was reverse transcribed using the Revert Aid First Strand cDNA Synthesis Kit (cat. no. K1622; Thermo Fisher Scientific, Inc.). Briefly, RNA (2 µg) was incubated for 60 min at 42°C and for 5 sec at 70°C. qPCR was conducted using the SYBR^® Premix Ex Taq™ II (Tli RNaseH Plus, cat. no. RR820A, Takara Bio, Inc.) according to the manufacturer's protocols. The thermocycling conditions were as follows: 95°C initial denaturation for 30 sec; followed by 40 cycles of denaturation at 95°C for 5 sec, and annealing and extension at 60°C for 30 sec) using CFX96 (Bio-Rad Laboratories, Inc.). After being normalized to the expression levels of GADPH, the relative mRNA expression levels were determined by the 2^−ΔΔCq method (23). The primers used are described in Table SII.

GSCs were lysed in phenylmethylsulfonyl fluoride and radioimmunopre-cipitation assay lysis buffer (1:100, Wuhan Boster Biological Technology, Ltd.), and the collected lysates were treated by sonication at 4°C, then centrifuged at 12,000 × g at 4°C for 15 min. The supernatants were used as cell lysates. The protein concentration was analyzed by a BCA protein assay kit (cat. no. PP1001, Bioteke Corporation). Samples containing 30 µg of protein were separated by SDS-PAGE (10, 12 and 15%) and transferred onto polyvinylidene difluoride membranes which were probed with antibodies. Prior, the membranes were blocked with 5% non-fat milk in Tris-buffered saline containing 0.05% Tween-20 at room temperature for 1 h. Subsequently, the membranes were incubated overnight at 4°C with the following antibodies (1:1,000 dilutions) against BRD4 antibody [EPR5150(2); ab128874) from Abcam; phosphorylated (phosphor)-AKT (Ser473, cat no. 4060) and phospho-retinoblastoma (Rb; Ser807/811, cat. no. 9308S) from Cell Signaling Technology, Inc.; total AKT (cat. no. 10176-2-AP), c-Myc (cat. no. 10828-1-AP), Bcl2-associated X protein (BAX; cat. no. 50599-2-lg), BCL2 (cat. no. 12789-1-AP), caspase 3/p17/p19 (cat. no. 19677-1-AP), P27 (cat. no. 25614-1-AP), E2F1 (cat. no. 12171-1-AP), matrix metalloproteinase 2 (MMP2; cat. no. 10373-2-AP), MMP9 (cat. no. 10375-2-AP) and Actin antibody from ProteinTech Group, Inc.; CCND1 (A11022) from ABclonal Biotech Co., Ltd.; proliferating cell nuclear antigen (PCNA; PC10, SC-56) from Santa Cruz Biotechnology, Inc.; Rb (G3-245) from BD Pharmingen (BD Biosciences); VEGF (cat. no. 05-443) from Upstate Biotechnology; Histone H2A.X (D17A3) XP (H2AX, cat. no. 7631), phospho-Histone H2A.X (Ser139, 20E3, γH2AX; cat. no. 9718), VEGFR2 (cat. no. 55B11) and phospho-VEGF receptor 2 (Tyr1175, cat. no. 19A10) from Cell Signaling Technology, Inc. AKT (phospho Thr308) from Immunoway Biotechnology. The secondary antibodies, horseradish peroxidase (HRP)-conjugated Affinipure Goat Anti-Mouse IgG (H+L) (cat. no. SA00001-1; 1:2,000) and HRP-conjugated Affinipure Goat Anti-Rabbit IgG (H+L) (cat. no. SA00001-2, 1:2,000) antibodies were purchased from ProteinTech Group, Inc. After visualization using the enhanced chemiluminescence (Thermo Fisher Scientific, Inc.), the density of the bands was analyzed using image analysis software (Image J 1,46r).

CSC1589 cells were cultured in 12-well plates at density of 1×10⁵ cells/well and treated with JQ1 for 0.125 or 0.25 µM. Single cell suspensions were seeded on coverslips pre-coated with fibronectin (1 µg/ml, cat. no. F1141-5MG, Sigma-Aldrich; Merck KGaA) and ornithine (50 mg/ml, P3655-50MG, Sigma-Aldrich; Merck KGaA) overnight. CSC1589 cells were fixed with 4% paraformaldehyde at room temperature for 10-15 min and permeabilized with 0.3% Triton X-100 (Sigma-Aldrich; Merck KGaA) in PBS for 5 min. Then, potential non-specific binding sites were blocked with antibody dilution buffer [10% fetal bovine serum and (Sigma-Aldrich; Merck KGaA) and 1% bovine serum albumin (Gentihold) in PBS]. Sections was incubated with an antibody against Nestin (ab6320, 1:50; Abcam) overnight at 4°C. After washing the sections with PBS for 5 min, secondary antibodies were applied [Donkey anti-Mouse IgG (H+L) ReadyProbes™ Secondary Antibody, Alexa Fluor^® 488, R37114, 1:5,000; Invitrogen; Thermo Fisher Scientific, Inc.] for 30 min at room temperature before mounting in water soluble mounting medium (Wuhan Boster Biological Technology, Ltd.) with DAPI (room temperature, 2 min). After the final washing step by PBS, the glass coverslips were mounted upside-down and visualized with an Olympus BX53 microscope at ×400 magnification.

Total RNA of CSC2078 cells was isolated using the TRIzol, after which the quality, concentration and integrity were determined using a NanoDrop spectrophotometer (NanoDrop Technologies; Thermo Scientific Fisher, Inc.). RNA (3 µg) was used for the RNA sample preparations. Sequencing libraries were generated using the TruSeq RNA Sample Preparation Kit (Illumina, Inc.). Briefly, mRNA was purified from total RNA using poly-T oligoattached magnetic beads. Fragmentation was carried out using divalent cations under elevated temperature (94°C, 5 min) in an Illumina proprietary fragmentation buffer. First strand cDNA was synthesized using random oligonucleotides and SuperScript II (25°C for 10 min; 42°C for 15 min; 70°C for 15 min, Illumine, Inc.) reverse transcriptase. Then, the second strand cDNA synthesis was performed by DNA Polymerase I and RNase H (16°C for 1 h, Illumine, Inc.). Remaining overhangs were converted into blunt ends via exonuclease/polymerase activities and the enzymes were removed. After adenylation of the 3′ends of the DNA fragments, Illumina PE adapter oligonucleotides were ligated to prepare for hybridization. To select cDNA fragments of the preferred 200 bp in length, the library fragments were purified using the AMPure XP system (Beckman Coulter, Inc.). DNA fragments with ligated adaptor molecules on both ends were selectively enriched using Illumina PCR Primer Cocktail (Illumina, Inc.) in a 15 cycle PCR reaction (Applied Biosystems; Thermo Fisher Scientific, Inc.). The thermocycling conditions were as follows: 98°C initial denaturation for 30 sec; followed 98°C denaturation for 10 sec, and 60°C annealing for 30 sec; and 72°C extension for 30 sec. Products were purified (AMPure XP system) and quantified using the Agilent high sensitivity DNA assay on a Bioanalyzer 2100 system (Agilent Technologies, Inc.). The sequencing library was then sequenced on a Hiseq platform (Illumina, Inc.) by Shanghai Personal Biotechnology Co. Ltd.

We used HTSeq(0.9.1) statistics (https://htseq.readthedocs.io/en/release_0.11.1/history.html#version-0-9-1) to compare the Read Count values on each gene as the original expression of the gene, and then used fragments per kilobase of transcript per million to standardize the expression. Then we used DESeq (1.30.0) (24) to analyze the genes of difference expression with screened conditions as follows: Expression difference multiple |log2FoldChange| >1, significance, P<0.05. In addition, we used R language Pheatmap (1.0.8) software package (https://cran.r-project.org/web/packages/pheatmap/index.html) to perform bi-directional clustering analysis of all different genes of samples. We mapped all the genes to terms in the GO database and calculated the numbers of differentially enriched genes in each term. Based on the whole genome, terms with significant enrichment of differentially enriched genes were calculated by hypergeometric distribution. We counted the number of differentially expressed genes following KEGG pathway enrichment analysis (https://www.genome.jp/kegg/); the metabolic pathways and signaling pathways that differentially expressed genes mainly participate in were identified.

BALB/c male nude mice (age, 5-8 weeks) were obtained from Beijing Vital River Laboratory Animal Technology. A total of 10 mice each weighing 18-22 g were used. Mice were housed under a 12 h light/dark cycle in an air-conditioned room at 22±2°C with free access to food and water. In xenograft implantation experiments, nude mice were injected subcutaneously in the right flank with CSC2078 cells (1×10⁷) suspended in 100 µl PBS. After 2 days, nude mice of the JQ1-treated group were administered intraperitoneally with JQ1 (50 mg/kg/day) for 17 days. DMSO-containing JQ1 was diluted with physiological saline to obtain a final solution of 5 mg/ml in JQ1-treated groups. Mice of the control group were injected an equal amount of physiological saline containing 5% DMSO per day. We measured the weight and tumor size every day. Tumor size was measured with calipers. The tumor volume was calculated with the following formula: V=(length) 2× (width)/2. The humane endpoints were judged by mouse activity assessment (hunching, stationary, poor grooming and ruffling) or the mouse weight loss (>20% of total body weight). No animals were sacrificed due to meeting these endpoints; all mice were sacrificed via carbon dioxide after 17 days. The 10 mice were anesthetized by an intraperitoneal injection of 10% chloral hydrate at a dosage of 300 mg/kg body weight prior to carbon dioxide euthanasia; 100% carbon dioxide was applied 10-30% volume per minute and the mice were euthanized after exposure to carbon dioxide for 5-6 min. Mice death was confirmed when no spontaneous breathing for 2-3 min and no blinking reflex were observed. No secondary method was applied to verify mice death. After the mice were euthanized, tumor tissues were extracted were divided into two parts: One part was frozen at -80°C for protein and RNA extraction, and the other part was fixed in 4% paraformaldehyde (room temperature, 24 h) for immunofluorescent staining. The splanchnic tissues (heart, liver, spleen, lung and kidney) were excised for hemoxylin and eosin (H&E) staining. The largest tumor in the control group had a diameter of 13 mm and the JQ1-treated group had a diameter of 7 mm. Animal experiments were approved by the Animal Ethics Committee of Jilin University [approval no. 2018;(54)], and by the Institutional Committee for the Care and Use of Laboratory Animals of the Experimental Animal Center of Jilin University.

Animals of the vehicle- and JQ1-treated CSC2078 xenograft model groups were sacrificed and the tumors were rinsed in PBS, followed by fixation with 4% paraformaldehyde for 24 h at room temperature. The tumors were cut into five-µm thick sections and mounted on poly-D-lysine-coated slides. The sections were then deparaffinized with xylene and rehydrated via alcohol, followed by 100% alcohol for 5 min, 95% for alcohol for 5 min, 75% alcohol for 5 min and 50% alcohol for 5 min. Then, the sections were washed with PBS (1X) for 5 min twice at room temperature. Antigen retrieval was performed by application of citrate buffer pH 6.00 for 20 min. The slides were incubated with primary antibodies overnight at 4°C, against c-Myc (cat. no. 10828-1-AP, 1:100), BAX (cat. no. 50599-2-lg, 1:100), BCL2 (cat. no. 12789-1-AP, 1:100), MMP2 (cat. no. 10373-2-AP, 1:100) and MMP9 (cat. no. 10375-2-AP, 1:100) from ProteinTech Group, In. Subsequently, HRP secondary antibodies were applied for 40 min at 37°C (SV0002; Wuhan Boster Biological Technology, Ltd.) according to the manufacturer's protocols. DAB (DA1010 Beijing Solarbio Science & Technology Co., Ltd.) was applied at room temperature for 15 min and analysis was performed using a BX53 microscope (Olympus Corporation).

Statistical analyses were performed with Microsoft Excel (Microsoft Corporation) and GraphPad Prism 6 software (GraphPad Software, Inc.). Comparison of the two sets of data was using an unpaired Student's t-test. For comparison of more than two sets, one-way ANOVA with a Newman-Keuls post-hoc test was conducted. All quantifications were performed with at least three independent experiments. Data are expressed as the mean ± standard error of the mean. P<0.05 was considered to indicate a statistically significant difference.

To evaluate the role of epigenetic genes in GSCs, we acquired GSCs from hGFAP-Cre⁺ p53L/L PtenL/⁺ mice (20). We used siRNAs to silence the expression of Brd4 to detect the effects of Brd4 on the activity, proliferation and self-renewal ability of GSCs. Then, western blotting was performed to detect the protein expression of Brd4 in CSC2078 transfected with the Brd4-specific siRNAs. Compared with the control group, we found that the expression of Brd4 was significantly inhibited in the CSC2078 cells treated with siBrd4s, especially siBrd4s-475 and siBrd4s-3820. This demonstrated the effectiveness of siRNAs targeting Brd4 (Fig. 1A and B). We next explored the effects of silencing Brd4 on the proliferation and self-renewal of GSCs. We detected the cell viability of CSC2078 transfected with siBrd4s using a CCK-8 assay. The results showed that silencing of Brd4 significantly reduced cell viability (Fig. 1C). A cell growth curve and a soft agar colony formation assays demonstrated that the three independent Brd4 siRNAs had significant anti-proliferative effects on CSC2078 cells compared with the control (Figs. 1D and E, and S2A). Then, we conducted neural sphere formation assays to detect the effects of the most effective siRNAs on the self-renewal ability of CSC2078 cells. The results revealed that the number of neural spheres in the siBrd4 groups were significantly lower than the control group (Fig. 1F). The results illustrated that siRNAs targeting Brd4 effectively inhibited the self-renewal ability of CSC2078 cells (Figs. 1F and S2E). Taken together, these results suggested that Brd4 had important functions in mediating an anti-tumor effect on GSCs, and that Brd4 could be a potential therapeutic target of in ablating GSCs.

To explore the effects of JQ1 on tumorigenic properties, we assessed cell proliferation and self-renewal in TS543, CSC2078 and CSC1589. CCK-8 assays revealed that cell viability was significantly reduced compared with the control after 24 and 48 h of treatment with JQ1 in a concentration-dependent manner (Figs. 2A and B, and S1A). Next, we analyzed cell proliferation using cell growth curves following 3 days of treatment of CSC2078, TS543 and CSC1589 with JQ1. JQ1 significantly inhibited the proliferation of all three GSC lines compared with the control (Figs. 2C and D, and S1B). In addition, JQ1 significantly reduced the clone formation capacity of GSCs compared with the control. Specifically, the number of cell clones was significantly lower after 10 days of treatment with JQ1. This further supported that JQ1 effectively inhibited GSCs proliferation (Figs. 2E and F, S1C and D, and S2A). In addition, neural sphere formation assays revealed that neural sphere numbers were significantly decreased by treatment with various concentrations of JQ1 for 10 days compared with the control (Figs. 2G and H, and S2B-D).

These results demonstrated that the treatment of GSCs with JQ1 has similar effects as silencing of Brd4. The anti-neoplastic effects of JQ1 effectively inhibited the proliferation and self-renewal of GSCs.

To further explore the mechanisms by which JQ1 and siBrd4 inhibit cell proliferation, we evaluated their effects on the cell cycle and apoptosis of GSCs. We analyzed the cell cycle via flow cytometry of CSC2078 and TS543 cells treated with JQ1 for 24 h, and CSC2078 transfected with siBrd4 for 48 h. We found that JQ1 significantly increased the percentage of cells in G1 and reduced that in S phase compared with the control; notable effects on the percentage of cells in G2/M phase were observed (Figs. 3A and B, and S3). We observed a similar G1/S cell cycle arrest effect in CSC2078 transfected with siBrd4 (Figs. 3C and S3). Collectively, these data suggested that JQ1 impeded cell cycle progression.

Treatment with JQ1 and siBrd4s also induced apoptosis in GSCs. We assessed the apoptosis of CSC2078 and TS543 cells via flow cytometry using Annexin V/PI staining. The number of apoptotic cells was significantly increased in both cells lines treated with JQ1 compared with the control (Figs. 3D and E, S4, S7 and S8A-D). Similarly, silencing of Brd4 using siRNAs significantly induced the apoptosis of CSC2078 cells compared with the control (Figs. 3F, S4, and S8E-G). Consistent with the aforementioned results, Hoechst 33342 staining illustrated that the number of apoptotic CSC1589 cells was significantly with JQ1 treatment than in the control group (Fig. S5A and B). The results of TUNEL staining also confirmed the ability of JQ1 to promote the apoptosis of CSC2078 cells (Fig. S5C and D). These data suggested that JQ1 reduces the viability and inhibits the proliferation of GSCs by inducing cell cycle arrest and apoptosis.

GO enrichment analysis revealed that treatment of CSC2078 with JQ1 induced multiple changes of biological activities, among which 'cell differentiation' was particularly prominent (Fig. S6A). RNA-seq analysis of differentially expressed genes (DEGs) illustrated that the neural stem cell (NSC) markers Nestin and ciliary neurotrophic factor were significantly downregulated in the JQ1-treated groups (Fig. S6B). We detected the expression of Nestin via immuno-fluorescence staining of CSC1589 cells. The results indicated that protein expression levels of Nestin were significantly downregulated in JQ1-treated cells (Fig. S6C and D). Then, we detected the mRNA expression of Nestin by RT-qPCR in CSC2078. The results indicated that the mRNA expression levels of Nestin were significantly downregulated after treatment with JQ1 for 24 h (Fig. S6E). In addition, the astrocyte marker S-100β was upregulated, whereas the oligodendrocyte marker platelet-derived growth factor α (PDGFα) was strongly downregulated (Fig. S6B). The neuronal precursor marker class III b-tubulin was markedly upregulated in JQ1-treated cells. The results revealed that JQ1 can inhibit the expression of NSC markers and promote GSC differentiation to astrocytes under proliferative conditions (Fig. S6B).

To further explore the mechanism by which Brd4 regulates the growth of GSCs, we performed RNA-seq comparisons of the CSC2078 cells treated with JQ1 for 24 h. The results revealed that 1,327 and 1,484 genes were upregulated and downregulated, respectively (Fig. 4A and B). KEGG is a collection of databases containing genomes, biological pathways, diseases. Compared with the whole genomic background, these databases can identify pathways of significant enrichment between candidate target genes (25). The top 20 pathways of candidate target gene enrichment were shown in the enrich scatter diagram, and enrichment factors, q-value and gene number were used to report KEGG enrichment degree. Our results showed that PI3K/AKT signaling pathway was significantly enriched and the q-value of this pathway was close to zero; the DEG numbers were higher (Fig. 5). Therefore, we used western blotting to investigate the relationship between Brd4 and the PI3K/AKT signal pathway. The results demonstrated that inhibition of Brd4 using siRNAs or JQ1 inhibited PI3K and phospho-AKT (Ser473 or Thr308); thus, we speculated that Brd4 regulates GSC growth through PI3K/AKT signaling (Fig. 4C-H).

It has been reported that PI3K signaling is activated by various growth factors, including VEGF (26). We mainly analyzed the downregulated growth factors in CSC2078 cells treated with JQ1 from the results of RNA-seq analysis It was found that VEGF was inhibited to the highest degree (Fig. 6A). VEGF is released by cancer cells to induce tumor angiogenesis and can be secreted from the extracellular matrix (ECM) by MMP9 to promotes tumor growth (27). Thus, we used western blotting to detect the expression of VEGF and MMP in CSC2078 cells treated with JQ1. The results demonstrated that the expression of VEGF, MMP2 and MMP9 were significantly inhibited by JQ1 (Fig. 6B and C). Therefore, we preliminarily speculated that JQ1 exhibits anti-tumor effects via the VEGF/PI3K/AKT signaling pathway. In addition, we proposed that MMP2 and MMP9 proteins downstream of VEGF can be downregulated by JQ1, controlling the angiogenesis and metastasis of GBM and inhibiting tumor development.

In order to further verify that JQ1 plays an anti-tumor role in GSCs through the VEGF/PI3K/AKT signaling pathway, we selected the VEGR2 inhibitor linifanib for further experiments. Linifanib is a novel tyrosine-kinase inhibitor, which selectively targets VEGFR and PDGFR, and has low off-target inhibitory activity (28). We analyzed the cell cycle via flow cytometry of CSC2078 treated with JQ1 and/or linifanib for 24 h. We found that JQ1, linifanib alone or in combination induced the cell cycle arrest of CSC2078 (Fig. 6D and E). In addition, treatment with JQ1 and/or Linifanib significantly induced CSC2078 cell apoptosis. Importantly, the apoptosis rate in JQ1-treated group was not significant different compared with that linifanib-treated group or the combination group (Figs. 6F and G, and S9). Consistent with our observations, there was a reduction in the levels of phospho-VEGFR2, PI3K, phospho-AKT (Ser473), Cyclin B1 and Bcl-2 in CSC2078 treatment with JQ1, Linifanib or in combination for 24 h. However, the expression of Bax was increased and VEGFR2 and AKT was not significantly altered (Fig. 6H and I). Compared with treatment with each agent alone, co-treatment with JQ1 and linifanib did not cause greater attenuation of PI3K/AKT activity and further induce cell cycle arrest/apoptosis. These results further indicated that JQ1 exerts anti-tumor effects on GSCs through the VEGF/PI3K/AKT signaling pathway.

The cell cycle was significantly enriched in KEGG pathway enrichment analysis. To further investigate the mechanism by which Brd4 regulates cell cycle progression in GSCs, we used RT-qPCR to detect cell cycle-related genes, namely c-Myc, cyclin D1, p21, and p27, in CSC2078 and TS543 cells treated with JQ1. We found that the mRNA expression levels of the c-Myc and CyclinD1 were significantly inhibited, while P21 and P27 were significantly upregulated (Figs. 7A and 8A). Then, we detected the protein levels of cell cycle proteins including Cyclin D1, CDK inhibitor (CDKI) p27, Rb, phospho-Rb, E2F1 and c-Myc after treatment CSC2078 with JQ1, or siBrd4 by western blotting. We found that the protein expression of the c-Myc and CyclinD1 was significantly inhibited, while P27 was significantly upregulated. JQ1 or siBrd4 reduced the expression of Rb protein, but the expression level of p-Rb protein did not significantly change; thus, Rb protein was not activated. Therefore, E2F1 cannot be released from the RB/E2F1 complex to promote transcription of downstream genes. The Rb/E2F1 complex was deactivated resulting in a cycle arrest in GSCs treated with JQ1 or siRNAs (Fig. 7B-D).

As AKT can activate apoptosis in many types of cancer, we investigated the effect of Brd4 on apoptotic genes in GSCs. We used RT-qPCR to verify the effects of JQ1 and siBrd4 on important apoptotic genes in PI3K/AKT signaling, including Bim, BAX, Bak, Bcl-2, and Bcl-xl in CSC2078 and/or TS543 cells (Figs. 7A and 8A). We also investigated the effects of JQ1 or siBrd4 on apoptotic proteins, including BAX, Bcl-2, and cleaved caspase 3 in GSCs via western blotting. We found that the protein expression of cleaved caspase 3 and Bax were significantly upregulated, while Bcl-2 was inhibited. In addition, we found that the protein expression of γH2AX were significantly upregulated. JQ1 or siBrd4 caused DNA damage as indicated by the ratio of γH2AX/H2AX, which further promoted the apoptosis of GSCs. From our results, the trend was consistent between JQ1- and siBrd4s-treated cells. The results indicated that siBrd4 or JQ1 could exhibit pro-apoptotic effects through the VEGF/PI3K/AKT signal axis in GSCs (Fig. 8B-E).

In addition, we assessed the therapeutic effects of JQ1 on a glioblastoma tumor xenograft model. Nude mice bearing CSC2078 subcutaneous tumors were treated with 50 mg/kg of JQ1 via an intraperitoneal injection once daily and sacrificed on the 17th day (Fig. 9A). The results demonstrated that the tumor volume was significantly smaller in the JQ1-treated group than that in the control group (Fig. 9B and C). Apoptosis in tumor tissues obtained from control and JQ1-treated mice was detected using TUNEL staining. The number of apoptotic cells was markedly increased in JQ1-treated tumors (Fig. 9D). The results of H&E staining indicated that tumor tissues displayed obvious nuclear shrinkage in JQ1-treatment group (Fig. 9E). The results were consistent with the in vitro findings. Immunohistochemical staining revealed that c-Myc, PCNA, Bcl-2, MMP2 and MMP9 expression was decreased in tumor tissues from JQ1-treated mice, whereas BAX expression was elevated (Fig. 9F). To detect the toxicity of JQ1 in mice, H&E staining was performed in the organs of mice. There were no obvious histopathological findings in the JQ1-treated mice, as shown in Fig. 10A. Compared with the findings in control tumors, the protein expression levels of Brd4 and AKT were markedly unchanged in JQ1-treated tumors, whereas PI3K and phosphor-AKT (Ser473) was downregulated. The results were consistent with the in vitro findings (Fig. 10B and C). We observed significant reductions in c-Myc, Cyclin D1, and Bcl-2 levels in JQ1-treated mice compared with the control. By contrast, BAX and γH2AX expression was significantly increased (Fig. 10D and E). Together, these data suggested that JQ1 could effectively inhibit tumorigenesis and the development of GBM. The therapeutic effects of JQ1 may warrant a clinical trial.