Epigenetic downregulation of STAT6 increases HIF-1α expression via mTOR/S6K/S6, leading to enhanced hypoxic viability of glioma cells

Cycloheximide, 5-aza-2′-deoxycytidine (5-Aza), trichostatin A (TSA), and rapamycin were purchased from Sigma-Aldrich (St. Louis, MO, USA). The indicated primary antibodies against the following proteins were used in this study: STAT1/2/3/4/5/6 (#9172/4594/12640/5097/9363/9362), DNMT1/3a/3b (#5119/3598/67259), cleaved-caspase 3/7 (#9664/9491), cleaved PARP (#5625), S6K (#2708), phospho-S6K (T389) (#9205), S6 (#2217), phospho-S6 (Ser235/236) (#2211), mTOR (#2972), 4E-BP1 (#9644), phospho-4E-BP1 (T70) (#9455), eIF4E (#2067), phospho-eIF4E (S209) (#9741), eIF4G (#2469), Rheb (#13879), TSC2 (#4308), p-TSC2 (T1462) (#3617), p-TSC2 (S1387) (#5584), ICAM1 (#4915), JAK2 (#3230), NFATC2 (#4389) and Myc taq (#2278) (Cell Signaling Technology, Danvers, MA, USA); anti-HIF-1α (610958) (BD Biosciences, San Jose, CA, USA); and anti-actin (sc-1616) and GAPDH (sc-48,167) (Santa Cruz Biotechnology, Santa Cruz, CA, USA). Secondary antibodies used were anti-goat IgG HRP (81–1620; Invitrogen, Carlsbad, CA, USA), anti-mouse IgG HRP (G-21040; Invitrogen), and anti-rabbit IgG HRP (111–035-003; Jackson Laboratories, Bar Harbor, ME, USA).

The human glioblastoma cell lines U87MG and U373MG were obtained from the Korean Cell Line Bank (Seoul, Korea) and U251 and LN229 were kindly provided by Dr. Hee Young Kim (Seoul National University, Seoul, Republic of Korea). The cancer cell lines were routinely grown in Dulbecco’s Modified Eagle Medium (DMEM; Sigma-Aldrich) containing 10% fetal bovine serum (FBS; Gemini, West Sacramento, CA, USA) and 0.1% antibiotic-antimycotic solution (Gibco, Carlsbad, CA, USA).

Fetal normal human astrocytes (NHA) were purchased from ABM (Richmond, BC, Canada) and were cultured according to manufacturer’s direction in Prigrow X Series Medium (ABM). Glioma stem-like cells (GSCs) were established from freshly resected tumors and were cultured in neurobasal media (Gibco) supplemented with N2 (Gibco) and B27 (Invitrogen). Cultures were supplemented with 20 ng/mL of epidermal growth factor (EGF) (Invitrogen) and basic fibroblast growth factor (bFGF) (Millipore, Billerica, MA, USA) every 2–3 days.

All cells were incubated at 37 °C in a humidified atmosphere of 95% air and 5% CO₂. For hypoxia treatment, cells were incubated in an oxygen control hypoxia chamber (Coy Laboratory Products, Grass Lake, MI, USA) at 37 °C in a humidified 5% CO₂ environment, with the balance provided by N₂.

Following informed consent, glioma and normal brain tissues were obtained from patients undergoing surgery at the Ajou University Hospital in accordance with Institutional Review Boards protocols. The samples were snap-frozen in liquid nitrogen and stored at − 80 °C until analysis. Detailed information of patients is provided in Additional file 2: Table S1.

Cells were lysed in RIPA buffer (50 mM Tris-HCl, 1% NP-40, 0.25% sodium deoxycholate, 150 mM NaCl, 1 mM EDTA) supplemented with protease inhibitor cocktail and phosphatase inhibitor cocktail. The protein concentration in lysates was determined using the Bio-Rad protein assay (Bio-Rad Laboratories, Hercules, CA, USA). Proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membranes. Primary antibodies were applied overnight at 4 °C. Peroxidase-conjugated secondary antibodies were applied at room temperature for 1 h, and immunoreactive protein were visualized using an enhanced chemiluminescence detection kit (WESTSAVE Gold; AbFrontier, Seoul, Korea), followed by exposure to X-ray film. Band intensities were quantified using ImageJ software (National Institutes of Health, Bethesda, MD, USA).

A tissue microarray (TMA) containing glioblastoma and normal brain samples (GL806; US Biomax,Rockville, MD, USA) was used for high throughput analysis of gene expression. The TMA contains 35 cases of glioblastoma and 5 normal brain tissues, represented as duplicate cores per case. Tissue microarray slides were deparaffinized in xylene followed by rehydration. For antigen retrieval, slides were boiled in 10 mM citric acid buffer (pH 6.0) for 20 min, followed by blocking of endogenous peroxidase by incubating in 0.3% H₂O₂ in methanol for 30 min. Slides were blocked by incubating for 1 h in 10% normal goat serum, and then incubated overnight at 4 °C with a primary antibody against STAT6 (1:100; Abcam, Cambridge, MA, USA) and then with biotinylated goat anti-rabbit IgG secondary antibody (1:400) for 1 h. The antigen-antibody reaction was visualized using the Vector SG peroxidase substrate in a Vectastain Elite ABC kit (Vector Laboratories, Burlingame, CA, USA). Slides were counterstained with Harris hematoxylin solution (Sigma), subjected to a dehydration process, then briefly washed in xylene and mounted with VectaMount Permanent Mounting Medium (Vector Laboratories). For the quantitative analysis, a Histo score (H score) was calculated based on the staining intensity and percentage of stained cells. The intensity score was defined as follows: 1, weak staining; 2, intermediate staining; 3, strong staining. The H score was determined by the formula: 3 × percentage of strong staining + 2 × percentage of moderate staining + percentage of weak staining, giving a range of 0 to 300 for the H scores.

Total RNA was isolated using RNAiso Plus (TaKaRa, Otsu, Japan), and cDNA was synthesized using avian myeloblastosis virus reverse transcriptase (New England Biolabs) and oligo (dT) primers (Promega, Madison, WI, USA), according to the manufacturer’s instructions. For qPCR, amplification reactions were performed using a Thermal Cycler Dice Real-Time System (TaKaRa) with SYBR Premix Ex Taq master mix (TaKaRa) according to the manufacturer’s instructions. The primers used for RT-qPCR (Bioneer, Daejeon, Korea) are described in Additional file 2: Table S2.

The methylation status of STAT6 promoter CpG islands was analyzed using methylation-specific PCR (MSP) and pyrosequencing. Genomic DNA (gDNA) was extracted from human glioma and normal brain tissue and U373MG and U87MG cell lines using DNeasy Blood & Tissue kits (Qiagen Hilden, Germany). Bisulfite conversion of gDNA was performed using the EpiTect Bisulfite Kit (Qiagen) and then the converted DNA was amplified with primers specific to methylated or unmethylated DNA using an EpiTect MSP Kit (Qiagen), according to the manufacturer’s instructions. The methylation-specific primer sequences in STAT6 promoter CpG islands were 5′-GCG TCG AGT TAA TTT TTT TC-3′ (forward) and 5′-CGC TTA ATA ACC TAA ACT CGC-3′ (reverse); the unmethylation-specific primer sequences were 5′-GGT GTT GAG TTA ATT TTT TTT-3′ (forward) and 5′-CCA CTT AAT AAC CTA AAC TCA C-3′ (reverse). After the reaction was complete, products were analyzed by electrophoresis on a 2.5% agarose gel.

Pyrosequencing was performed by Genomictree (Daejeon, South Korea) using standard protocols. Briefly, bisulfite-modified gDNA was prepared using an EZ DNA Methylation-Lighting kit (Zymo Research, USA), and then converted gDNA was amplified by PCR. The reaction conditions were as follows: denaturation at 95 °C for 10 min, followed by 45 cycles of 95 °C for 30 s, 58 °C for 30 s, 72 °C for 30 s, and a final extension at 72 °C for 5 min. Pyrosequencing was performed on a PyroMark ID system using a Pyro Gold reagents kit (Qiagen) according to the manufacturer’s instruction without further optimization. The methylation percentage was calculated based on the average degree of methylation at 2–5 CpG sites formulated in pyrosequencing. Each primer was designed using Pyrosequencing Assay Design Software v2.0 (Qiagen). Primer sequences are listed in Additional file 2: Table S3.

For knockdown of specific proteins, cells were transfected for 72 h with siRNA (final RNA concentration, 25 nM) using DharmaFECT 2 reagents (Dharmacon), according to the manufacturer’s instructions. SMART pool siRNA against STAT6 (L-006690-00) (Dharmacon Thermo Scientific) and non-targeting control siRNAs (D-001810-10) were purchased from Dharmacon (Lafayette, CO, USA); siRNA against DNMT1 (1786–1), DNMT3b (1789–1), HIF-1α (3091–1) and a corresponding negative control (SN-1002) were purchased from Bioneer (Korea). siRNA against DNMT3a was chemically synthesized by Bioneer (Korea). The siRNA sequences are as follows: 5′-CAG GAG AUG AUG UCC AAC CC-3′ (sense) and 5′-GGG UUG GAC AUC AUC UCC UG-3′ (antisense).

STAT6 was overexpressed by transiently transfecting U373MG cells with pCMV6-Myc-DDK-STAT6 (RC210065; Origene Technologies) or mock control vector (PS100001; Origene Technologies) using TurboFectin 8.0 (OriGene Technologies, Inc.) according to the manufacturer’s recommendations. All assays were performed at least 48 h after transfection. STAT6 deletion mutants were generated from human STAT6 cDNA (RC210065, Origene) by PCR amplification using the appropriate primers. STAT6 deletion mutants (Δ1–260 (d1), Δ261–430 (d2), Δ431–522(d3), and Δ523–622 (d4)) were cloned into the pCMV6-entry vector (PS100001, Origene). Primer sequences are listed in Additional file 2: Table S4.

U373MG cells cultured on poly-d-lysine–coated coverslips were fixed with 4% (v/v) paraformaldehyde at 4 °C for 15 min and permeabilized by incubating for 5 min in phosphate-buffered saline (PBS) containing 0.1% Triton X-100. Fixed cells were incubated with anti-STAT6 (611,290, BD Biosciences) and HIF-1α (ab51608, Abcam) antibodies at 4 °C overnight, followed by incubation with Alexa Flour 488- or Alexa Flour 543-conjugated secondary antibodies (Molecular Probes, Eugene, OR, USA) for 2 h. Cells were mounted with DAPI-containing mounting solution and observed under a confocal microscope (LSM510; Carl Zeiss, Jena, Germany).

Newly synthesized proteins were detected using the Click-IT method (Life Technologies), according to the manufacturer’s instructions. Briefly MOCK- or full-length STAT6-transfected U373MG cells or control siRNA- or STAT6 siRNA-transfected U87MG cells were exposed to hypoxia for 18 h. Cells were subsequently washed with PBS, incubated in methionine-free DMEM for 1 h, then pulsed with 50 μM of the methionine analog, L-azidohomoalanine (AHA) (Cat # C10102; Life Technologies), for 2 h, after which cells were harvested. AHA-incorporated proteins were labeled with biotin using a Click-iT Biotin Protein Analysis Detection Kit (Cat# C33372; Invitrogen). Biotin-labeled proteins were pulled down using a streptavidin-agarose bead slurry (Pierce, Rockford, IL, USA) and analyzed by immunoblotting using an anti-HIF-1α antibody.

Cells were lysed in CHAPS buffer (40 mM HEPES pH 7.4, 120 mM NaCl, 1 mM EDTA, 0.3% CHAPS) supplemented with protease inhibitor cocktail and phosphatase inhibitor cocktail. Lysates (300 μg) were incubated with antibody (1 μg of anti-mTOR, STAT6, eIF4E or isotype control antibodies) at 4 °C overnight. Magnetic protein G Dynabeads (Invitrogen) were then added, and tubes were rotated for an additional 2 h. After washing five times with lysis buffer, proteins were eluted from beads with 2X Laemmli buffer at 95 °C for 10 min, followed by resolution by SDS-PAGE.

Cell viability was assessed using the EZ-Cytox Cell Viability Assay kit (Daeil Lab Service, Seoul, Korea). Control siRNA- or STAT6 siRNA-transfected U87MG cells were exposed to hypoxia for 5 days. The Ez-Cytox Kit reagent was then added to the medium and cells were incubated at 37 °C for 2 h. The plate was read on an iMark microplate absorbance reader (Bio-Rad) at a wavelength of 450 nm.

U373MG cells were stained with propidium iodide (PI) (Sigma-Aldrich) for cell cycle analysis, or with PI and annexin V-FITC (BD Biosciences) for detection of apoptosis. The distribution of cells in the different phases of the cell cycle (based on differences in DNA content) and apoptosis-positive cells were determined by flow cytometry using a FACSCalibur flow cytometer (BD Biosciences).

To explore the function and significance of STAT proteins in gliomas, we first examined the expression levels of individual STAT proteins in human glioma specimens. Western blot analyses demonstrated that, whereas STAT2 and STAT4 showed no prominent or consistent changes in expression, STAT1, STAT3 and STAT5 expression were increased in glioma tissues compared with neighboring non-tumor tissue. In contrast, STAT6 expression, which was apparent in non-tumor tissue, was decreased in glioma tissues, exhibiting an inverse relationship with tumor grade (Fig. 1a). Although it is well known that STAT3 is upregulated in glioma and plays a role in tumor progression and survival, little is known about STAT6 in this context. To obtain further evidence of STAT6 downregulation in gliomas, we performed immunohistochemical analyses using a brain glioma tissue array (see Materials and Methods). Compared to non-tumor tissue, GBM tissue exhibited a significantly lower percentage of positive staining for STAT6 (Fig. 1b). Interestingly, in a primary GBM specimen in which STAT6 downregulation was not observed, STAT6 expression was clearly reduced following recurrence of glioma (Fig. 1c). STAT6 was also downregulated in astrocytoma, but not in oligodendroglioma (Fig. 1d). A transcript analysis revealed that this downregulation of STAT6 protein expression resulted from decreased mRNA production (Fig. 1e). In line with our results, a published microarray-based, high-throughput assessment (GEO DataSet record GDS1962) showed that STAT6 mRNA is reduced in glioma tissues compared with non-tumor tissue (Fig. 1f). Decreased STAT6 mRNA expression in GBM was also shown in TCGA brain in Oncomine database (Fig. 1g). Taken together, these observations demonstrate that STAT6 transcripts are downregulated in glioma tissues, leading to reduced STAT6 protein expression.

To establish the mechanisms underlying downregulation of STAT6 transcript levels in human glioma, we first investigated the possibility of epigenetic silencing via CpG island hypermethylation. One CpG island upstream of the transcription start site of STAT6 gene (+ 756 to + 1033 bp), containing 14 CpG sites, was predicted using MethPrimer (http://​www.​urogene.​org/​methprimer/​) (Fig. 2a). An examination of methylation status in non-tumor tissue and glioma tissue using methylation-specific PCR (MSP) revealed that CpG islands in the STAT6 promoter were more frequently methylated in high-grade gliomas than low grade gliomas, while relatively weak methylation was found in non-tumor tissue (Fig. 2b). To validate these findings, we analyzed 14 CpG sites located at the core of the CpG islands in non-tumor tissue and glioma tissue using bisulfite sequencing, which has increased sensitivity compared with MSP. In line with the results of MSP analyses, samples from high-grade gliomas showed prominent methylation (Fig. 2c), demonstrating a strong correlation between STAT6 downregulation and methylation of the STAT6 promoter in glioma.

Next, we explored whether promoter methylation directly mediates the silencing of STAT6 in glioma tissue. First, we examined whether STAT6 expression was increased by treatment with 5-aza-2-deoxycytidine (5-Aza), a DNMT inhibitor. In the GBM cell lines, U373MG and LN229, STAT6 expression was barely detectable and methylation was more frequent compared with U87MG cells (Fig. 2d). In U373MG cells, 5-Aza induced a concentration-dependent increase in STAT6 mRNA and protein (Fig. 2e). However, the histone deacetylase inhibitor trichostatin A (TSA), had little effect on STAT6 expression (Additional file 1: Figure S1), suggesting that methylation and not acetylation of DNA plays a dominant role in STAT6 silencing in glioma.

To further confirm the involvement of DNMT, we examined STAT6 expression level following small interfering RNA (siRNA)-mediated knockdown of the major mammalian DNMTs, DNMT1, DNMT3a and DNMT3b. In U373MG cells, STAT6 expression was restored by silencing DNMT1 with DNMT1si, but not by silencing either DNMT3a or DNMT3b (Fig. 2f). We further found that the extent to which STAT6 expression increased in U373MG cells was dependent on the concentration of DNMT1si (Fig. 2g and h). Similar effects were observed in experiments using LN229 cells (Fig. 2i). Taken together, these results suggest a role for DNMT1 in STAT6 silencing in glioma.

Next, we explored the functional relevance of STAT6 downregulation in GBM cells. Given the characteristic hypoxic microenvironment of glioma, we initially assessed whether STAT6 is related to HIF-1α expression, since HIF-1α is the critical mediator to hypoxic adaptation and survival. Experiments were performed using U87MG cells with high STAT6 expression and low methylation levels. Interestingly, while siRNA-mediated knockdown of STAT6 did not induce significant differences in HIF-1α mRNA levels under hypoxia (Fig. 3a), HIF-1α protein levels were clearly increased (Fig. 3b).

We then tested whether STAT6 affects HIF-1α accumulation through regulation of its protein degradation. In the presence of cycloheximide (CHX) to block de novo protein synthesis, the half-life of HIF-1α was similar in control and STAT6 knockdown cells (Fig. 3c), indicating that STAT6 is not involved in HIF-1α protein degradation. An assay kit that detects only newly synthesized protein confirmed that de novo synthesis of HIF-1α was increased in the absence of STAT6 (Fig. 3d).

To validate this finding, we performed STAT6 gain-of-function experiments in U373MG cells with barely detectable STAT6 and high promoter methylation levels. While HIF-1α mRNA levels showed no differences under hypoxic conditions (Fig. 3e), HIF-1α protein expression was increased in mock-transfected cells under hypoxia and conversely reduced in STAT6-overexpressing cells (Fig. 3f). This result was supported by immunofluorescence (IF) staining showing decreased HIF-1α intensity in STAT6-expressing cells (Fig. 3g). Assessment of protein stability using cycloheximide revealed a similar half-life of HIF-1α in control and STAT6-overexpressing cells (Fig. 3h), further confirming the non-involvement of STAT6 in HIF-1α protein degradation. These findings were supported using an assay kit that specifically detects newly synthesized protein (Fig. 3i).

Mammalian target of rapamycin (mTOR) regulates numerous components of protein synthesis, including ribosomal proteins and initiation and elongation factors (Fig. 4a). Since our data suggest that STAT6 inhibits HIF-1α at the translational level, we assessed the involvement of mTOR signaling using rapamycin, a chemical inhibitor of mTOR. Rapamycin effectively suppressed hypoxia-induced HIF-1α in both control and STAT6 knockdown cells (Fig. 4b). In view of the finding that mTOR pathways comprise S6K-S6 and 4E-BP1 (Fig. 4a), we further examined the activation status of downstream effectors of mTOR. Phosphorylation levels of both S6K and its target S6, which were decreased in a hypoxic microenvironment, were partially restored in the absence of STAT6 while the total protein levels of S6K and S6 showed no changes (Fig. 4c). In the case of 4E-BP1, another key target of mTOR, both phosphorylated and total protein levels were increased in hypoxia, which were suppressed in the absence of STAT6 (Fig. 4d). The corresponding increase in 4E-BP1 mRNA expression in hypoxia and its suppression by STAT6si (Fig. 4e) indicated that phosphorylation and protein levels reflect transcriptional patterns. In STAT6 knockdown cells, restoration of the suppressed mTOR signaling in hypoxia facilitated eIF4E binding to eIF4G (Fig. 4f).

In line with STAT6 knockdown experiments, hypoxia-induced suppression of p-S6K-S6K signaling was further decreased in STAT6-overexpressing U373MG cells (Additional file 1: Figure S2A). However, protein and transcript levels of 4E-BP1 and p-4E-BP1 displayed no significant changes following STAT6 overexpression (Additional file 1: Figure S2B and C).

The mechanism underlying STAT6-mediated inhibition of mTOR activity under hypoxia was further investigated. The TSC2-Rheb signaling axis induced suppression of mTOR activity under hypoxia (Fig. 5a). However, STAT6 did not affect activity of the upstream kinase AKT (Fig. 5b). No significant differences in mRNA induction of genes regulating DNA damage and development, including REDD 1, BNIP3 and PML (Fig. 5c), or TSC2 phosphorylation at T1462 or S138 (Fig. 5d) were evident between control-, and STAT6 siRNA-transfected U87MG in hypoxic conditions.

Accordingly, we considered the possibility that STAT6 interferes with Rheb-mTOR interactions, thereby suppressing mTOR activity. Interestingly, hypoxia-induced Rheb-mTOR dissociation was maintained in STAT6 knockdown cells (Fig. 5e). To further examine the effects of STAT6 on Rheb-mTOR binding, we analyzed the interactions of STAT6 with Rheb or mTOR. STAT6 interacted with Rheb (Fig. 5g) but not mTOR (Fig. 5f), with a peak representing optimal binding at 3–6 h after hypoxia (Fig. 5h). Moreover, under reoxygenation conditions, hypoxia-induced STAT6 binding to Rheb was suppressed (Fig. 5i), indicating the necessity of hypoxic conditions for STAT6-Rheb interactions. For further analysis of STAT6-Rheb interactions, we mapped the Rheb-binding region of STAT6 using deletion mutants of each domain. Deletion of either the Linker Domain (LD) or SH2 domain (SH) abolished STAT6 binding to Rheb under hypoxic conditions (Fig. 5j). Upon inhibition of STAT6-Rheb interactions (via transfection of d3 or d4), S6 phosphorylation was restored, even in hypoxic conditions (Fig. 5j, INPUT), signifying inhibitory effects of STAT6 on mTOR signaling through interactions with Rheb.

To address the functional significance of STAT6-regulated HIF-1α expression in GBM cells, we initially examined the impact of STAT6 on cell viability under hypoxic microenvironments. In U87MG cells, siRNA-mediated knockdown of STAT6 led to increased cell viability (Fig. 6a and b) and decreased expression of apoptosis markers (Fig. 6c). Next, we examined the interrelationships among STAT6, HIF-1α and apoptosis by examining the expression patterns of HIF-1α and apoptosis markers at 18 h and 4 days after hypoxia, respectively. Hypoxia-induced cell death was decreased in the presence of STAT6si, which was reversed in double knockdown cells. In parallel, hypoxia-induced cell death was potentiated by HIF-1αsi and, conversely, attenuated in double knockdown cells, indicating that HIF-1α expression is dependent on STAT6 (Fig. 6d).

In STAT6 gain-of function experiments, STAT6-overexpressing U373MG cells exhibited decreased viability (Additional file 1: Figure S3A) along with increased expression of apoptosis markers (Additional file 1: Figure S3B and C).

In recent years, hypomethylating agents, such as 5-azacitidine and decitabine (5-Aza), have been clinically applied for treatment of myelodysplasia. Interestingly, these agents exert DNA methyltransferase (DNMT) inhibition effects at significantly lower concentrations than those required for cancer cell killing effects. Moreover, effects are sustained even after washout of drugs. In our cell culture model, the apoptotic cell content was significantly induced 3 days after washout of 5-Aza at concentrations ≥100 nM, which continued to increase until 6 days (Additional file 1: Figure S4).

To test whether epigenetic restoration of STAT6 mediates apoptotic effects of 5-Aza, STAT6-silenced U373MG cells were treated with 5-Aza. Transient 5-Aza treatment increased apoptotic cell death to 29 ± 0.5%, which was decreased to 18 ± 2.5% in the presence of STAT6si (Fig. 6e). Effects of STAT6 knockdown on apoptosis were further confirmed by reduced cleavage of capase-3, caspase-7 and poly-(ADP-ribose) polymerase (PARP) (Fig. 6f). Additionally, 5-Aza induced arrest of cells in the G2/M phase under the same conditions as in Fig. 6e, which was significantly reduced upon STAT6 knockdown (Fig. 6g). Collectively, these results suggest that the effects of transient, low-dose 5-Aza on cell death and the cell cycle are primarily, but not exclusively, dependent on STAT6 re-expression in glioma cells.

In addition to its cell-killing effects, transient exposure to low doses of 5-Aza is reported to exert sustained antitumor effects through induction of immune signaling [18, 47]. We further examined whether epigenetic re-expression of STAT6 by 5-Aza is the mechanism mediating expression of immune-related genes. Indeed, several immune-related genes, including interleukin (IL)-1β, C-C motif chemokine ligand (CCL)2, CCL20, IL-8, ICAM1 intercellular adhesion molecule 1 (ICAM1), nuclear factor of activated T cells 2 (NFATC2) and Janus kinase 2 (JAK2), that were significantly induced by low doses of 5-Aza were suppressed upon STAT6 knockdown (Additional file 1: Figure S5), clearly demonstrating dependence of expression on STAT6. Our results support the utility of 5-Aza as an effective chemotherapeutic agent with immune modulating capacity as well as cell-killing effects.

Glioma stem-like cells (GSCs) have been implicated in glioma initiation, progression, therapeutic resistance and tumor recurrence, and are thus critical targets for therapy [22]. Therefore, we determined whether our findings could be extended to GSCs. To this end, we established and characterized two patient-derived GSC cell lines, GSC1 and GSC2 (Additional file 1: Figure S6). An examination of STAT protein expression levels showed that STAT6 levels were significantly lower in GSCs compared with non-tumor tissue, whereas levels of STAT1, STAT3 and STAT5 were significantly higher (Fig. 7a). Notably, STAT6 levels in GSCs were even lower than those in high-grade glioma tissues (Fig. 7a). Next, we examined STAT6 promoter methylation status in GSCs. MSP and bisulfite sequencing showed that the STAT6 promoter was highly methylated in GSCs (Fig. 7b). Furthermore, 5-Aza treatment and DNMT1 knockdown successfully restored STAT6 expression in GSCs (Fig. 7c and d), demonstrating promoter-methylation–dependent STAT6 silencing in GSCs. Intriguingly, although STAT6 silencing was only dependent on DNMT1 in GBM cell lines (Fig. 2), in case of GSC1, both DNMT3a and DNMT1 contributed to STAT6 silencing (Fig. 7d). Finally, we examined the role of STAT6 in HIF-1α accumulation and 5-Aza effects on GSCs. STAT6 overexpression significantly reduced HIF-1α levels in hypoxic GSCs (Fig. 7e and f), indicating that STAT6 downregulation may facilitate GSC survival under hypoxic conditions. In addition, 5-Aza–induced apoptosis were significantly attenuated by STAT6 knockdown.