The human glioma cell lines T98G and U87 were purchased from the Chinese Academy of Sciences Cell Bank and identified by the STR site detection assay (Microread Genetics Co., Beijing, China). In addition, 76 human glioma tissue samples including 32 low-grade gliomas (4 grade I tumors and 28 grade II tumors), 44 high-grade gliomas (16 grade III tumors and 28 grade IV tumors), and 2 normal brain tissues from decompression operation were obtained from the Department of Neurosurgery of Qilu Hospital of Shandong University from October 2013 to June 2015. The glioma specimens were verified and classified according to the WHO classification standard of tumors by two experienced clinical pathologists. Our study was approved by the Institutional Review Board of Shandong University. Written informed consent was obtained from all the patients, and the hospital ethics committee approved the experiments.

The cells were cultured in DMEM (Invitrogen) supplemented with 10% FBS (Gibco) and maintained at 37°C with 5% CO₂ in a humidified chamber. Hypoxic conditions were induced by incubating the cells in a modular incubator chamber flushed with a gas mixture containing 1% O₂, 5% CO₂, and 94% N₂ at 37°C.

Human glioma tissue samples or solid tumors removed from sacrificed mice were fixed with 4% formaldehyde. Paraffin-embedded tumor tissues were sectioned to 5-μm thickness and mounted on positively charged microscope slides, and 1 mM EDTA (pH 8.0) for HIF-1α or citrate solution (pH 6.0) for other antigens was used for antigen retrieval. Endogenous peroxidase activity was quenched by incubating the slides in methanol containing 3% hydrogen peroxide, followed by washing in PBS for 6 min. The sections were incubated for 2 h at room temperature with normal goat serum and subsequently incubated at 4°C overnight with primary antibodies (Abcam, 1:200 HIF-1α, 1:400 LC3B, and 1:300 CREB3. Cell Signaling Technology, 1:300 cleaved caspase 3). The sections were then rinsed with PBS and incubated with horseradish peroxidase-conjugated goat anti-rabbit or anti-mouse antibodies, followed by reaction with diaminobenzidine and counterstaining with Mayer's hematoxylin. H&E staining was performed free of charge by the pathology department of Qilu Hospital of Shandong University. Evaluation of the staining reaction was performed in accordance with the immunoreactive score (IRS): IRS = SI (staining intensity) x PP (percentage of positive cells). An SI value of 0 was negative; 1, weak; 2, moderate; and 3, strong. A PP value of 0 was negative; 1, 10% positive cells; 2, 11–50% positive cells; 3, 51–80% positive cells; and 4, >80% positive cells. Five visual fields from different areas of each tumor were used for the IRS evaluation.

A T98G GFP-LC3 stable cell line was established by transient transfection of the pSELECT-GFP-LC3 lentiviral vector (Invivogen, USA). GFP-LC3 puncta formation under normoxic or hypoxic conditions was determined by capturing images using a DP71 CCD digital camera microscope (Olympus). To quantify autophagic cells after treatment, we counted the number of autophagic cells as determined by the presence of GFP-LC3 puncta (≥20 puncta = a positive cell) in 200 cells.

After the desired treatment, cells were washed twice with cold PBS and harvested with a rubber scraper. Cell pellets were lysed and kept on ice for ≥30 min in a buffer containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.5% Nonidet P-40, 50 mM NaF, 1 mM Na₃VO₄, 1 mM phenylmethylsulfonyl fluoride, and 1 mM PMSF. The lysates were cleared by centrifugation, and the supernatants were collected. Cell lysates were then separated by SDS-PAGE and subjected to western blot analysis with primary antibodies and horseradish peroxidase-conjugated secondary antibodies. Antibodies against the following were used for western blotting: P62 (Cell Signaling Technology, 5114S), LC3B (Abcam, 63817), GAPDH (Cell Signaling Technology, 3683S), caspase 3 (Cell Signaling Technology, 9668T), cleaved caspase 3 (Cell Signaling Technology, 9664S), CREBRF (Santa Cruz Biotechnology, sc-133747), CREB3 (Abcam, ab42454), and ATG5 (Abcam, ab108327). GAPDH served as the loading control in all experiments involving hypoxic treatment due to the upregulation of β-actin under hypoxia.

CREB3 and negative control siRNAs were synthesized by Rio-Bio (China). The sequences of CREB3 siRNAs were as follows: CREB3-1299, 5′-GCA GUC AGA AGU GCC GAA ATT-3′ (sense) and 5′-TTC GUC AGU CUU CAC GGC UUU-3′ (antisense). The sequences of negative control were as follows: 5′-UUC UCC GAA CGU GUC ACG UTT-3′ (sense) and 5′-ACG UGA CAC GUU CGG AGA ATT-3′ (antisense). The siRNAs were transfected into T98G and U87 cells for 48 h using Lipofectamine 2000 according to the protocol of the manufacturer.

The CCK-8 assay (Dojindo, Japan) was used to test cell viability. Tumor cells in medium containing 10% fetal bovine serum were seeded into 96-well, flat-bottomed plates at 5×10³ cells/well and incubated at 37°C overnight. After the desired treatment, the cells were incubated for an additional 4 h with 100 μl of serum-free DMEM and 10 μl of CCK-8 at 37°C. The absorbance at 450 nm was measured using a microplate reader.

To assess the degree of apoptosis of different groups of glioma cells, the extent of Annexin V-FITC/propidium iodide (PI) staining was determined by flow cytometry using the Annexin V/PI staining kit from Bender MedSystems (Vienna, Austria). Samples were measured using an Epics XL-MCL flow cytometer (Beckman Coulter, Brea, CA, USA) and analyzed with WinMDI 2.8 software.

Glioma cells were plated on glass slides in 24-well culture plates at 2×10⁵ cells/well for 24 h and subsequently treated with drugs for an additional 48 h in serum-free DMEM. Glass slides with glioma cells and paraffin-embedded tumor sections were stained by the TUNEL technique using a TACS^®2 TdT-Fluor in situ apoptosis detection kit (Trevigen, Inc.) according to the instructions of the manufacturer. TUNEL-positive cells were counted from ≥10 random fields under a fluorescence microscope.

Total RNA was extracted using TRIzol according to the manufacturer's protocol. Then, total RNA (50 ng) was reverse-transcribed with miR stem-loop RT primers or with U6 RT primers using a ReverTra Ace qPCR RT kit according to the manufacturer's protocol to generate cDNA. Real-time PCR was performed using a SYBR Premix Ex Taq™ kit with each primer. The reactions were performed using a LightCycler 2.0 Instrument. U6 expression was used as the endogenous control. The absolute expression levels were calculated as concentration ratios using a Roche LightCycler^® 2.0 system.

Data analyses were conducted with SPSS 16.0 (SPSS, IL, USA) and GraphPad-Prism5 (GraphPad, CA, USA). Descriptive statistics including means ± SD, Student's t-test, non-parametric Kruskal-Wallis tests for multiple comparison, the Mann-Whitney U test for two-group comparison, Kaplan-Meier plots, log-rank tests, one-way ANOVAs, and Pearson's correlation test were used to analyze significant difference; ^*P<0.05, ^**P<0.01, and ^***P<0.001 were considered statistically significant.

To analyze CREBRF/CREB3 expression in clinical samples, total RNA was extracted from surgically removed glioma tissues of 20 patients (2 normal brain tissue samples, 9 patients with WHO I–II and 9 patients with WHO III–IV gliomas), and RT-PCR was performed. Interestingly, significant differences in CREBRF expression were observed between low-grade (I–II) and high-grade (III–IV) patients. CREBRF expression was significantly reduced in high-grade glioma tissues compared with low-grade glioma tissues or normal brain tissues (Fig. 1A, left). Consistent with its negative regulating property, CREB3 expression was significantly reduced in low-grade glioma tissues compared with high-grade glioma tissues (Fig. 1A, right) with a significant negative correlation (Pearson correlation index, −0.6112; P-value=0.0042) (Fig. 1B). To further examine whether CREB3 expression correlates with the WHO grades of human glioma, we performed immunohistochemical staining to detect CREB3 expression in 76 human glioma specimens with different grades and 2 normal brain tissues (Table I). As shown in Fig. 1C, CREB3 expression positively correlated with the WHO grade of glioma (Fig. 1D). Furthermore, patients with clinical survival information were analyzed by Kaplan-Meier estimation. We found that glioma patients with high CREB3 expression exhibited significantly poor postoperative survival time (Fig. 1E). The above findings raised the intriguing possibility that CREBRF acts as a tumor suppressor and downstream CREB3 acts as a prognostic biomarker of glioma. The underlying mechanisms by which CREBRF suppresses the malignant progression of glioma remain uncertain.

Extensive autophagy in the hypoxic areas of tumors has been described (14). However, the association among CREB3 and LC3B levels and hypoxia levels in gliomas has not been reported, and the CREB3-mediated upregulation of autophagy in hypoxic areas is also unknown. To examine whether CREB3 expression correlates with autophagy and hypoxia levels in human glioma, we detected HIF-1α (a marker of hypoxia) and LC3B (a marker of autophagy) expressions in 76 human glioma specimens, which represent different grades and 2 normal brain tissue samples, using immunochemical staining. As shown in Fig. 2A, HIF-1α and LC3B expression levels strongly correlated with the WHO grade of glioma (Fig. 2B). In addition, HIF-1α, and LC3B expression positively correlated with CREB3 levels in glioma (Fig. 2C). Taken together, these results demonstrate that CREB3 expression correlates with the density of autophagic cells and HIF-1α levels in gliomas.

We further verified CREBRF and the CREB3 levels in hypoxia T98G and U87 cells using quantitative real-time PCR, and our results are consistent with results from human glioma tissue samples. CREBRF expression significantly decreased in glioblastoma cells after 24 h of hypoxia treatment, and CREB3 increased accordingly (Fig. 3A). It is generally accepted that hypoxia activates autophagy as an evolutionarily conserved cellular catabolic process (8,28). To verify this pivotal phenomenon, we performed a GFP-LC3B puncta-formation assay and an LC3B conversion assay (29). Using GBM cells (T98G) that stably express a GFP-LC3B fusion protein, the localization of GFP-LC3B was examined by fluorescent microscopy. GFP-LC3B puncta that appear in the cytoplasm reflect the recruitment of LC3B proteins to autophagosomes. As shown in Fig. 3B, a significant increase in GFP-LC3B puncta was noted in hypoxic cells, and this result was confirmed by the quantification of GFP-LC3B dots per cell. Moreover, we detected the conversion of LC3B-I to LC3B-II and P62 (autophagy marker) expression together with CREBRF and CREB3 protein levels by western blot analysis under different hypoxia treatment times. Consistent with the GFP-LC3B puncta-formation assay, hypoxia led to a significant time-dependent upregulation of LC3B-II and downregulation of P62 (Fig. 3C). Also consistent with the quantitative real-time PCR results, hypoxia downregulated CREBRF and upregulated CREB3 protein levels in a time-dependent manner (Fig. 3C). Thus, both assays suggested that hypoxia induces autophagosome accumulation and that CREBRF/CREB3 may be involved in this process.

To further support the fact that endogenous CREB3 is involved in hypoxia-induced autophagy, we examined LC3B-II and P62 protein levels after attenuating CREB3 activity in hypoxic glioblastoma cells using RNA interference (RNAi) (Fig. 4A, left). Western blot analysis clearly indicated that CREB3 knockdown inhibited the hypoxia-induced LC3B-II increase and P62 degradation in glioblastoma cells (Fig. 4A, right). Using T98G cells transiently transfected with GFP-LC3B, our study revealed that the inhibition of endogenous CREB3 inhibited autophagy under hypoxic conditions (Fig. 4B). The data indicate that a blockade of endogenous CREB3 repressed autophagy in hypoxic glioblastoma cells.

To elucidate the mechanisms of CREB3-induced autophagy, several autophagy-related genes (ATGs) were examined by quantitative real-time PCR in T98G cells, revealing a significant decrease in ATG5 after CREB3 knockdown (Fig. 5A) that was confirmed by western blot analysis in both T98G and U87 cell lines (Fig. 5B). To further investigate the relationship between CREB3 and ATG5 in hypoxic glioblastoma cells, we collected 3 fresh, surgically removed WHO grade IV glioma tissue samples for fast-frozen sectioning, and staining revealed the co-localization of CREB3, ATG5, and LC3B proteins in high-grade glioma tissues (Fig. 5C).

In recent years, accumulating evidence has demonstrated that autophagy may serve as a protective mechanism in tumor cells and that therapy-induced apoptosis can be potentiated by autophagy inhibition (30,31). To determine the biological significance of CREB3-induced autophagy on apoptotic cell death, CREB3 interfering RNA was utilized in glioblastoma cells. A CCK-8 assay was utilized to determine cell viability. TUNEL staining assays and Annexin V-FITC/ PI were performed to examine the level of apoptosis in glioblastoma cells. As shown in Fig. 6A, CREB3 knock-down significantly suppressed glioblastoma cells exclusively under hypoxic conditions. Consistent with this observation, hypoxia-induced apoptotic cell death was augmented in the presence of CREB3 interfering RNA, which was demonstrated by the TUNEL assay (Fig. 6B) and the Annexin V-FITC/PI assay (Fig. 6C). In the process of apoptosis, caspase 3, as the final effector molecule, can be cleaved and activated as the most common characteristic of apoptosis. Subsequently, we examined cleaved caspase 3 protein levels by western blot analysis, and the data further indicated that CREB3 knock-down led to the activation of caspase 3 apoptotic signaling pathways (Fig. 6D). This type of crosstalk between autophagy and apoptosis provides a possible avenue for interference with hypoxia-induced autophagy of glioblastoma cells through anti-CREB3 adjuvant therapy to treat glioma patients.