Background
Hypoxia is well evidenced within most solid tumors [
1]. Acute, intermittent, and cycling hypoxia are associated with inadequate blood flow, whereas chronic hypoxia is the consequence of increased oxygen diffusion distance, resulting from tumor expansion [
2]. These hypoxic areas can either promote cell death or provoke an adaptive response, leading to the selection for death resistance [
3‐
5]. Once tumor cells adapt to hypoxia, they are more resistant to apoptosis and less responsive to cancer therapy. We recently showed that cycling hypoxia and chronic hypoxia are important tumor microenvironment phenomena that limit tumor response to chemotherapy in glioblastoma multiforme (GBM) [
6]. Therefore, several potential mechanisms, including the lack of oxygen that is available for anti-tumor drugs to act, DNA over-replication, increased genetic instability, the anti-proliferative effect of hypoxia [
7], increased multidrug resistance (MDR) linked to adenine triphosphate (ATP)-binding cassette (ABC) transporter [
8,
9], are thought to play a role in cycling hypoxia-induced chemoresistance. However, the exact mechanisms triggered by cycling hypoxia, leading to resistance to apoptosis, remain unclear.
Previous studies indicated that the repetitive periods of hypoxia and reoxygenation could lead to an increased production of reactive oxygen species (ROS) [
10,
11]. Moreover, NADPH oxidase subunit 4 (Nox4) is a critical mediator involved in cycling hypoxia-mediated ROS production, tumor progression, and resistance to cytotoxic therapies in GBM [
12‐
14]. Although ROS play an important role in apoptosis induction under both physiologic and pathologic conditions, ROS have also anti-apoptotic effects in cancer cells through unknown mechanisms [
15,
16]. A recent study reported that ROS stimulates NF-κB signaling, resulting in a B-cell lymphoma extra-long (Bcl-xL)-mediated resistance to drug-induced cell death [
17]. Bcl-xL is a member of the Bcl-2 family of proteins and acts as a pro-survival protein by preventing the release of mitochondrial contents and caspase activation [
18]. NF-κB can bind directly to Bcl-xL promoter to regulate its expression [
19]. Furthermore, Bcl-xL is also one of hypoxia-inducible factor-1 (HIF-1) target genes because its promoter contains a hypoxia-responsive element (HRE) [
20]. These observations point to the possibility that Bcl-xL is a critical contributor to cycling hypoxia-mediated anti-apoptosis and resistance to cytotoxic therapies.
In the present study, we determined the impact and mechanism of cycling hypoxia on anti-apoptosis and chemoresistance in GBM. Our results show that cycling hypoxic stress significantly increases resistance to temozolomide (TMZ) via Bcl-xL upregulation. ROS-mediated HIF-1α and NF-κB activation plays an essential role in cycling hypoxia-mediated Bcl-xL induction. Moreover, pretreatment with a ROS scavenger, Tempol, in intracerebral glioblastoma-bearing mice demonstrated a synergistic suppression of tumor growth and increased survival rate in TMZ chemotherapy, suggesting that ROS blockade before drug administration and in combination with chemotherapy may be an effective approach to suppress tumor microenvironment-induced chemoresistance and further improve the efficacy of chemotherapy in GBM.
Methods
Cell culture
U251, U87 and GBM8401 glioblastoma cells were cultured in DMEM (Life Technologies) supplemented with 10 % fetal bovine serum (FBS), 10 mM HEPES, and 1 % penicillin–streptomycin.
In vitro hypoxic treatments
Cells were treated in Biospherix C-Chamber (Biospherix) inside a standard culture chamber by means of exhausting and gassing with 95 % N
2 and 5 % CO
2 to produce oxygen concentrations of 0.5–1 % at 37 °C to achieve hypoxic condition. Cells were treated with or without in vitro non-interrupted hypoxic or cycling hypoxic stress as previously described [
6]. Briefly, cell cultures were exposed to 3 cycles consisting of 0.5–1 % O
2 for 1 h interrupted by 5 % CO
2 and air for 30 min for cycling hypoxic treatment and to persistent 0.5–1 % O
2 for 4 h for non-interrupted hypoxic treatment.
Real-time quantitative PCR (Q-PCR)
Q-PCR analysis was performed as described previously [
21]. The primers for quantitative analysis of
Bcl-
xL and the housekeeping gene 60S acidic ribosomal proteins were:
Bcl-
xL (F) 5′- GATCCCCATGGCAGCAGTAAAGCAAG -3′ and (R) 5′- CCCCATCCCGGAAGAGTTCATTCACT -3′ and the house keeping gene 60S acidic ribosomal protein (F) 5′-ACGAGGTGTGCAAGGAGGGC-3′ and (R) 5′-GCAAGTCGTCTCCCATCTGC-3′.
Western blot analysis
Cells and tissues were lysed and extracts were prepared as described previously [
21]. Nuclear and cytoplasmic lysates were prepared with the CelLytic Nuclear Extraction Kit (Sigma-Aldrich) according to the manufacturer’s protocol. HIF-1α, p65 and Bcl-xL proteins in human cells were detected in 150 µg of cell extract using monoclonal anti-HIF-1α antibody (diluted 1:650; Novus), anti-p65 antibody (diluted 1:500; Novus) and anti-Bcl-xL antibody (1:600; Novus). Western blots were normalized using a monoclonal anti-β-actin antibody (diluted 1:10,000; Sigma-Aldrich) for cell extracts and a monoclonal anti- TATA box binding protein (TPB) (diluted 1:1,000; Sigma-Aldrich) for nuclear extracts.
Reporter assays
The HIF-1α-luciferase reporter plasmid derived from our previous study [
10] and NF-κB-luciferase reporter plasmid (Clontech) were utilized to determine HIF-1 and NF-κB-dependent transactivation of luciferase activities, respectively. In the measurement of HIF-1 and NF-κB-dependent transactivation of luciferase activities, the dual-luciferase reporter assay system (Promega) was used. U251 and U87 cells were transfected with each reporter construct and the TK-Renilla luciferase plasmid was used as a transfection control. Luciferase detection was performed 48 h after reporter construct transfection. Expression was calculated as the relative Firefly luciferase activity normalized with respect to the activity of transfection control Renilla luciferase. To determine the role of Tempol, YC-1 or Bay 11-7082 in cycling hypoxia-induced transcriptional activation of Bcl-xL, the stably Bcl-xL promoter-driven luciferase reporter-transfected U251 and U87 cells were incubated with Tempol (4 mM; Sigma-Aldrich), YC-1 (10 μM; Sigma-Aldrich) and Bay 11-7082 (5 μM; Sigma-Aldrich) together with in vitro cycling hypoxic stress for 4 h. Firefly luciferase activities were assayed and normalized to those of extract protein concentrations measured with the Bio-Rad protein assay kit (Bio-Rad). Luciferase activity was determined by mixing 10 μL of extracts from 1 × 10
5 cells and 100 μL of luciferase assay reagent (Promega) according to the manufacturer’s instructions.
ROS levels analysis
ROS levels were assessed by using carboxy-2′7′-dihydrodichlorofluorescein diacetate (H2DCFDA, Molecular Probes) to assess total ROS. Cells were incubated with 5 μg/mL of H2DCFDA for 30 min, then washed with PBS, trypsinized and collected in 1 mL of PBS. Fluorescent stained cells were transferred to polystyrene tubes with cell-strainer caps (Falcon) and subjected to FACScalibur instrument and FACSDiva 6.0 software (BD Bioscience) for acquisition and analysis.
Vector constructions and viral transduction
The lentiviral vector pLKO AS2 (National RNAi Core Facility, Taiwan) was used as the backbone to generate a lentiviral reporter vector. The multiple cloning sites (MCS) of pTA-Luc vector (Clontech) was inserted with the cDNA fragment bearing −1075 to +617
Bcl-
xL promoter to drive the expression of firefly luciferase gene. The
Bcl-
xL promoter driven reporter gene cassette was amplified from promoter to SV40 ploy A on the constructed pTA-Luc vector using PCR and inserted into pLKO AS2 as pLKO AS2- Bcl-xL-p by XhoI and MluI restriction enzymes. The pGreenFire1-SFFV [
12] was used to generate glioblastoma reporter cells bearing SFFV promoter-driven a dual optical reporter gene encoding both green fluorescence protein (GFP) and luciferase (Luc). Lentiviral vectors carrying short hairpin RNAs (shRNA)-targeting HIF-1α (5′- TGCTCTTTGTGGTTGGATCTA-3′) and p65 (5′- CGGATTGAGGAGAAACGTAAA -3′) and scrambled shRNA (
http://rnai.genmed.sinica.edu.tw/file/vector/C6-7/17.1.pLAS.Void.pdf) were provided by National RNAi core facility, Academia Sinica in Taiwan. The lentiviral vector pLVCT-tTR-KRAB (Addgene) was used to express Bcl-xL shRNA (Sigma-Aldrich) following the manufacturer’s protocol. Lentivirus production and cell transduction were carried out according to protocols described elsewhere [
22,
23]. All constructs were confirmed by DNA sequencing. The U251, U87, GBM8401 cells bearing the SFFV promoter-driven a dual optical reporter gene and Bcl-xL promoter-driven Luc reporter gene were termed GBM8401/SFFV-LucGFP, U251- Bcl-xL-P-Luc and U87- Bcl-xL-P-Luc respectively.
Fluorescence-activated cell-sorting (FACS) analyses
Tumor tissues were disaggregated with an enzyme cocktail containing collagenase type III (Sigma), hyaluronidase (Sigma), and collagenase type IV (Sigma), washed several times, and resuspended in phosphate-buffered saline (PBS) to produce a single cell suspension. Fluorescence was measured using a FACScalibur instrument and FACSDiva 6.0 software (BD Bioscience). Tumor cells were gated according to DsRed expression and side scatter (SSC). The hypoxic subpopulations were further gated or isolated based on the analysis of Hoechst 3342 and GFP fluorescence in dot plots. The control cells are derived from disaggregated the orthotopic GBM8401 or U87 xenografts, which are both Hoechst 3342 and GFP-negative, and were set in the lower left quadrant of the plot. Same setting conditions were used thereafter, cell populations located outside of this quadrant of the plot were defined as either Hoechst 3342− and GFP+ cells (chronic hypoxic cells), Hoechst 3342+ and GFP+ cells (cycling hypoxic cells) or Hoechst 3342+ and GFP− cells (Normoxic cells).
Caspase-3 activity and apoptosis assays
U251 and U87 cells expressing Tet-inducible Bcl-xL shRNA were pretreated for 48 h with Dox (0.04 μg/mL) to induce Bcl-xL knockdown and then exposed to in vitro hypoxic stress, either non-interrupted hypoxic or cycling hypoxic stress, before TMZ treatment. The caspase-3-like protease activities were determined by the Caspase-3 Colorimetric Activity Assay Kit (Sigma-Aldrich) according to the manufacturer’s protocol. Briefly, Cell lysates were incubated with 2 mM Caspase-3 substrate (Ac-DEVD-pNA) in 1× assay buffer (20 mM HEPES, pH 7.4, 2 mM EDTA, 0.1 % CHAPS, 5 mM DTT) for 90 min at 37 °C. The absorbance was read at 405 nm and the results were calculated using a p-nitroaniline calibration curve. Annexin V and propidium iodide staining was performed to determine cell apoptosis using the Annexin V-FITC Apoptosis Detection Kit (Sigma-Aldrich) for 10 min at room temperature according to the manufacturer’s instructions, and then flow cytometric analysis was performed.
Cytotoxicity assay
Cells were treated with DMSO (vehicle control), Tempol (4 mM; Sigma-Aldrich), TMZ (250 μM; Sigma-Aldrich) and the combination of Tempol and TMZ after in vitro hypoxic stress or cell sorting. After incubation at 37 °C for 48 h, the medium was removed from each well, 15 μL 3-94,5-dimethyl-2-yl-2,5-diphenyl-tetrazolium (MTT) (Sigma-Aldrich) solution (2 mg/mL) were added and the plates were incubated at 37 °C for 4 h. The reaction was stopped by the addition of 100 μL of isopropanol/HCl, and the absorbance at 570 nm was recorded on a μQuant plate reader (Bio-Tek).
Animal models
Eight-week-old male athymic nu/nu mice were used to establish animal tumor models. For the subcutaneous GBM xenograft model, 5 × 106 GBM8401- Bcl-xL-P-Luc cells were injected subcutaneously into the dorsal flank and small (80 ± 16.0 mm3) subcutaneous tumors developed 14 days later were used for animal imaging studies. For the orthotopic GBM xenograft model, 2 × 105 GBM8401/SFFV-LucGFP cells were harvested by trypsinization and injected into the right basal ganglia of anesthetized mice. The tumors developed at 12 days after tumor implantation for evaluating the efficiency of therapy studies. All animal studies were conducted according to the Institutional Guidelines of China Medical University and approved by the Institutional Animal Care and Use Committees of China Medical University (approval number 102-54-N).
In vivo treatment
For the bioluminescence imaging of in vivo transcriptional activation of Bcl-xL,
Mice bearing U87- Bcl-xL-P-Luc xenograft were received with vehicle, Tempol (250 mg/kg i.p.), YC-1 (15 mg/kg i.p.) or Bay 11-7082 (10 mg/kg i.p.) followed by in vivo cycling hypoxic treatment. The procedure for in vivo cycling hypoxic treatment was carried out following published methods [
14,
24]. Briefly, the tumor-bearing mice were exposed to continuous flow of a humidified gas mixture to induce in vivo hypoxia in 6-liter hypoxia chambers. The mice were exposed to normal air (control) or 12 cycles of 10 min 7 % O
2 breathing interrupted by 10 min periods of normal air breathing for cycling hypoxic treatment. For the therapeutic study, intracerebral gioblastoma-bearing mice were randomly assigned to four different therapeutic groups: control (vehicle treatment), pretreatment of Tempol (250 mg/kg i.p.), TMZ (5 mg/kg i.p.) or pretreatment of Tempol (250 mg/kg i.p.) + TMZ (5 mg/kg i.p.). The pretreatment of Tempol was performed at day 13 after tumor cell injection. Systemic TMZ treatment was performed for 5 days, starting on day 14 after tumor cell injection. Tumor progression was monitored by bioluminescence imaging and mice were monitored daily for survival. Animals were killed at the onset of neurologic signs or any type of distress.
Bioluminescent imaging (BLI)
Mice were imaged with the IVIS Imaging System 200 Series (Caliper) to record bioluminescent signal emitted from the engrafted tumors. Mice were anesthetized with isoflurane and received intraperitoneal injection of D-Luciferin (Caliper) at a dose of 270 µg/g body weight. Imaging acquisition was performed at 15 min after intraperitoneal injection of luciferin. For BLI analysis, regions of interest encompassing the intracranial area of signal were defined using Living Image software, and the total number of photons per second per steradian per square centimeter were recorded. To facilitate comparison of growth rates, each mouse’s luminescence readings were normalized against its own luminescence reading at day 12, thereby allowing each mouse to serve as its own control.
Statistical analysis
One-way analysis of variance with post hoc Scheffe analyses and Kaplan–Meier Survival Analysis with Tarone-Ware statistics were carried out using the SPSS package (version 18.0). The differences between control and experimental groups were determined by the two-sided, unpaired Student t test. P < 0.05 was considered significant.
Discussion
There is abundant evidence to suggest that cycling hypoxia as well as uninterrupted hypoxia play roles in many aspects of tumor growth and development [
27,
28]. However, some results demonstrated that cycling hypoxia would utilize different intracellular signaling pathways than those utilized by uninterrupted hypoxia. For instance, ROS might play a vital role in cycling hypoxia-induced alterations in the carotid body function [
29]. In addition, the expression of immediate early response genes and transcription factors, AP-1 and HIF-1, were activated by cycling hypoxia [
29]. These results suggest that signaling pathways associated with transcriptional regulation by cycling hypoxia are distinct from those utilized by uninterrupted hypoxia. Consistent with these findings, our data showed that ROS production is only significantly triggered by cycling hypoxia in glioblastoma cells. Therefore, the ROS-mediated HIF-1/NF-κB/Bcl-xL pathway is activated only in cycling hypoxia-treated glioblastoma cells. Furthermore, cycling hypoxia induced more chemoresistance than uninterrupted hypoxia in U87 or U251 glioblastoma cells.
Highly aggressive tumors are known to be exposed to hypoxia, which occurs as a result of inadequate blood supply [
30]. Since hypoxia can affect certain gene regulatory mechanisms and signal transduction pathways, including tumor cell apoptosis, metastasis, and tumor angiogenesis, deciphering the hypoxic tumor cell response is essential to understand tumor progression. HIF-1α is a master transcription factor that regulates both physiological and pathophysiological responses of mammalian cells to hypoxia [
31]. In addition, it is also regulated by other factors, including oncogenes, growth factors, and free radicals [
32]. It is well known that HIF-1α overexpression driven by hypoxia and free radical species contributes to therapy resistance [
2,
10]. Several signaling pathways have been proposed to be regulated by hypoxia. For example, hypoxia could activate the JNK and p38 stress kinases in human squamous carcinoma cells and further leads to phosphorylation of transcription factor ATF-2 [
33]. Moreover, HIF-1α signaling associated with p38 involves mitochondrial-derived ROS [
34]. Furthermore, we previously demonstrated that cycling hypoxia induces HIF-1 activity via ROS-mediated HIF-1α synthesis and stabilization in U87 glioblastoma cells [
10].
Constitutive activation of NF-κB frequently occurs in various cancer types and is important for cancer cells to escape apoptosis and survive in the presence of apoptotic stimuli [
35]. Therefore, cancer cell resistance to chemotherapeutic agents can be partly explained by NF-κB deregulation. Inhibition of NF-κB activity could sensitize hepatocellular carcinoma cells to doxorubicin-induced apoptosis [
36]. In addition, NF-κB is activated by hypoxia and plays a central role in gene regulation under hypoxia/reoxygenation [
37,
38]. However, to date, the involvement of ROS in NF-κB signaling is controversial. For example, some reports suggested that NF-κB is not a sensor of oxidative stress and ROS does not influence NF-κB activation during intermittent hypoxia/reoxygenation [
39,
40]. In contrast, our results show that cycling hypoxia induces HIF-1α and NF-κB activation through ROS generation and this activation dramatically decreases after Tempol treatment in both U87 and U251 glioblastoma cells. A recent review article also highlights that ROS have various inhibitory or stimulatory roles in NF-κB signaling, suggesting the complexity of ROS-mediated NF-κB signaling [
41].
GBM is the most malignant form of brain cancer with a high mortality rate and this aggressive capability may be partly due to tumor hypoxia [
42,
43]. It contains multiple hypoxic areas that exhibit elevated HIF-1 signal transduction activity, resulting in the deregulated expression of downstream target genes that contribute to GBM malignancy [
44]. The expression of BNIP3, a Bcl-2 family member, correlates with HIF-1α expression levels in various types of tumor cells. Its expression was predominantly detected in the nucleus under hypoxic stress and contributes to cell survival in GBM [
45]. The Bcl-xL and Bcl-2 proteins are dominant inhibitors of apoptotic cell death. Recent studies demonstrated that Bcl-xL is one of the downstream target genes of HIF-1α and NF-κB through transcriptional regulation [
19,
20,
25]. Therefore, we proposed that the activation of HIF-1α and NF-κB induced by ROS under cycling hypoxia could result in Bcl-xL induction and allow the survival of glioblastoma cells. Our results clearly indicate that cycling hypoxia induces Bcl-xL expression in vivo and in vitro via HIF-1α or NF-κB activation and these effects can be inhibited by Tempol, HIF-1α inhibitor, or NF-κB inhibitor treatment. ROS and their downstream transcription factors, HIF-1α and NF-κB, are critical mediators involved in cycling hypoxia-mediated Bcl-xL induction. In addition, unlike uninterrupted hypoxia treatment, cycling hypoxia induces prosurvival effects in glioblastoma cells in response to Bcl-xL induction and caspase-3 inhibition. Moreover, cycling hypoxia significantly increases chemoresistance to TMZ compared with normoxic conditions in U87 or U251 glioblastoma cells. This chemoresistance induced by cycling hypoxic stress was suppressed by a ROS scavenger as well as by HIF-1α and NF-κB inhibitors. Therefore, ROS mediated HIF-1α and NF-κB activation is a crucial mechanism involved in cycling hypoxia-induced anti-apoptosis and chemoresistance in glioblastoma cells. These findings are important in the selection of optimal strategies for anticancer therapy in GBM.
In conclusion, the present study provides insightful information regarding the differential regulatory mechanisms involved in cycling hypoxia and uninterrupted hypoxia in tumor chemosensitivity. In addition, our data also highlight the putative mechanisms of cycling hypoxia in tumor cell chemoresistance in glioblastoma and suggest that ROS are attractive therapeutic targets to counteract cycling hypoxia-induced chemoresistance. Our previous studies also suggests that ROS within GBM cells act as second messengers in intracellular signaling cascades, which contribute to cycling hypoxia-mediated tumor progression [
12,
14]. Since cycling hypoxia is now a well-recognized phenomenon in the tumor microenvironment [
2], ROS blockade should be used before and with chemotherapy to suppress cycling hypoxia-induced chemoresistance and tumor progression and to further enhance the therapeutic efficiency of cytotoxic therapies in GBM.
Authors’ contributions
CHH designed the study and interpreted the results. WLC and CHH wrote the manuscript. CCW, YJL and CPW performed experiments, analyzed the results, and assisted with writing the manuscript. All authors read and approved the final manuscript.