PCR primers and small interfering ribonucleic acids (siRNAs) were purchased from Sigma-Aldrich (St. Louis, MO, USA) or Qiagen (Hilden, Germany). siRNA sequences, Sigma-Aldrich: CITED2, sense r(CCAGGUUUAACAACUC CCA)dTdT, antisense r(UGGGAGUUGUUAAACCUGG) dTdT and sense r(CCUAAUGGGCGAGCACAUA)dTdT, antisense r(UAUGUGCUCGCCCAUUAGG)dTdT; FOXO3a, sense r(GAAUGAUGGGCUGACUGAA)dTdT, antisense r(UUCAGUCAGCCCAUCAUUC)dTdT. siRNA sequences, Qiagen: AllStars negative control, FOXO1, sense r(GGA GGU AUG AGU CAG UAU A)dTdT, antisense r(UAUACUGACU CAUACCUCC)dAdT, sense r(UCAUGUCUUGAUAAGU UAA)dTdT, antisense r(UUAACUUAUCAAGACAUGA) dGdG and sense r(CCAAGUAGCCUGUUAUCAA)dTdT, antisense r(UUGAUAACAGGCUACUUGG)dTdT; FOXO3a, sense r(GAAUGAUGGGCUGACUGAA)dTdT, antisense r(UUCAGUCAGCCCAUCAUUC)dAdG, sense r(GGAACG UGAUGCUUCGCAA)dTdT, antisense r(UUGCGAAGCAU CACGUUCC)dGdG and sense r(GAUUCAUGCGGGUCC AGAA)dTdT, antisense r(UUCUGGACCCGCAUGMUC) dGdA. Metafectene Pro was acquired from Biontex (Munich, Germany), Attractene from Qiagen, DEVD-amc (N-acetyl-Asp-Glu-Val-Asp-aminomethyl-coumarin) and Z-VAD-fmk (N-benzyloxycarbonyl-Val-Ala-Asp-fluoromethylketone) from Bachem (Weil am Rhein, Germany) and recombinant human TRAIL from PeproTech (Rocky Hill, NJ, USA). Antibodies used were anti-actin (polyclonal goat anti-human, sc1616; Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA), anti-FOXO3a (immunohistochemistry, monoclonal rabbit anti-human, clone EP1949Y), anti-FOXO3a (western blot analysis, monoclonal rabbit anti-human, clone EPR1950) (both from Epitomics, Burlingame, CA, USA), anti-HIF1α (BD Transduction Laboratories, San Jose, CA, USA), anti-phospho-AKT Ser473 (polyclonal rabbit anti-mouse), anti-phospho-FOXO1 Ser256 (polyclonal rabbit anti-human) (both from Cell Signaling Technology, Danvers, MA, USA), anti-phospho-FOXO3a Thr32 (polyclonal rabbit anti-human; Upstate, Merck Millipore, Darmstadt, Germany), anti-phospho-TP53 Ser15 (polyclonal rabbit anti-human; Cell Signaling Technology) and anti-TP53 (monoclonal mouse anti-human; Calbiochem, Merck Millipore). The 3HRE-pTK-luc reporter construct was a generous gift from Berra et al (23). The adenoviral FOXO constructs (24), plasmids encoding FOXO3a, FOXO3a-A3 and FOXO3a-green fluorescent protein (GFP) (25), the plasmid encoding FOXO1-GFP (26) and the constructs FOXO3a-TM and FOXO3a-TM-ER-dDB (27) were kindly provided by colleagues from other laboratories. The TP53-luc construct was purchased from Agilent Technologies (Santa Clara, CA, USA).

LNT-229 human malignant glioma cells were kindly provided by N. de Tribolet. Cells were cultured in Dulbecco’s modified Eagle’s Medium (4,500 mg/l glucose; Sigma-Aldrich), supplemented with 10% fetal calf serum (FCS; PAA, Pasching, Austria), 2 mM glutamine, 100 IU/ml penicillin and 100 μg/ml streptomycin, at 37°C and 5% CO₂. For some experiments, glucose was added to serum- and glucose-free medium to give final concentrations of 2 or 5 mM. A Binder CB53 incubator (Binder, Tuttlingen, Germany) was used for experiments in hypoxia.

To silence FOXO3a gene expression, short hairpin RNA (shRNA) sequences were cloned into the pSUPER-puro vector (28). FOXO3a shRNA sequences were sense GATCCCCGGAACGTGATGCTTCGCAATT CAAGAGATTGCGAAGCATCACGTTCCTTTTTGGAAA, antisense TCGATTTCCAAAAAGGAACGTGATGCTTCG CAATCTCTTGAATTGCGAAGCATCACGTTCCGGG and sense GATCCCCGAATGATGGGCTGACTGAATTCAAG AGATTCAGTCAGCCCATCATTCTTTTTGGAAA, anti-sense TCGATTTCCAAAAAGAATGATGGGCTGACTGA ATCTCTTGAATTCAGTCAGCCCATCATTCGGG, respectively. The scrambled shRNA sequence has been published previously (29). Glioma cells were transfected using Attractene, stable cell lines were generated by puromycin selection (5 μg/ml). LNT-229 cells stably transfected with an shRNA targeting HIF1α and its control (Sima) as well as HIF2α and HIF1α-HIF2α double-knockdown LNT-229 cells were a kind gift of Henze et al (30).

A crystal violet staining assay was used for the determination of cell density. The dye was resolubilized in 0.1 M sodium citrate, the absorbance measured at 560 nm (Multiskan™ EX; Thermo Fisher Scientific, Langenselbold, Germany). Cell proliferation was quantified based on the incorporation of bromodeoxyuridine (BrdU Cell Proliferation ELISA; Roche Diagnostics, Mannheim, Germany). For the analysis of cell death, after treatment e.g. under hypoxic conditions, adherent and non-adherent cells were collected, washed in phosphate-buffered saline (PBS), stained with 1 μg/ml propidium iodide (PI) and analyzed by flow cytometry (BD Canto II, Heidelberg, Germany). PI-positive cells were regarded as dead cells. Caspase activity was assessed using the fluorescent substrate DEVD-amc as previously described (31) and a Mithras LB 940 microplate reader (Berthold Technologies, Bad Wildbad, Germany).

Following treatment e.g. under hypoxic conditions, cells were washed twice with PBS, incubated for 20 min at 37°C in the presence of 10 μM H₂DCFDA (Invitrogen, Carlsbad, CA, USA) washed with PBS and analyzed by flow cytometry.

Glucose concentrations of cell-free supernatants were measured on a Hitachi 917 analyzer (Roche Diagnostics). Furthermore, cells were incubated for 10, 75 and 150 min, respectively, in medium supplemented with 100 μM 2-NBDG (Invitrogen) under normoxic or hypoxic conditions. Afterwards, cells were washed twice with PBS, detached, resuspended in PBS containing 1 μg/ml PI and analyzed by flow cytometry at 465/540 nm. PI-positive (dead) cells were excluded from the analysis.

Cells were seeded in glass dishes. After 24 h medium was changed to pre-incubated treatment medium and the dishes were sealed air-tight. After another 48 h, oxygen concentration in the culture medium was measured in an ABL-80 FLEX Blood Gas Analyzer (Radiometer, Willich, Germany). Oxygen consumption was then normalized to protein concentration (32). Furthermore, oxygen consumption was determined using an XF96 extracellular flux analyzer (Seahorse Bioscience, North Billerica, MA, USA).

Using Metafectene Pro, cells were cotransfected with a 3HRE-pTK-luc (23) and, for normalization of transfection efficiency, a pRL-CMV Renilla luciferase construct. Activities of Firefly and Renilla luciferase (33) were analyzed in an Infinite^® M200 PRO microplate reader (Tecan, Maennedorf, Switzerland).

Cells were lysed in a buffer containing 50 mM Tris-HCl, 120 mM NaCl, 5 mM EDTA, 0.5% Nonidet P-40, 2 μg/ml aprotinin, 10 μg/ml leupeptin, 100 μg/ml phenylmethylsulfonyl fluoride, 50 mM NaF, 200 μM NaVO₅ and phosphatase inhibitor cocktails I and II (Sigma-Aldrich) and protein content was quantified using a Bradford assay (Bio-Rad, Hercules, CA, USA). Total protein (20 μg) was separated under reducing conditions by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and electroblotted on nitrocellulose (Amersham, Braunschweig, Germany). Membranes were blocked in Tris-buffered saline containing 5% skim milk and 0.1% Tween-20 and incubated with the appropriate primary (dilution 1:1,000) and secondary (dilution 1:3,000) antibodies. Immune complexes were detected by enhanced chemiluminescence (Pierce, Rockford, IL, USA). Densitometric analysis was performed using ImageJ according to http://rsb.info.nih.gov/ij/docs/menus/analyze.html#gels (34).

Extraction of total RNA was performed using TRIzol™ (Invitrogen) and the RNeasy™ system (Qiagen), cDNA synthesis using 1.5 μg of RNA and SuperScript VILO™ (Invitrogen). RNA originating from normal human brain was used with institutional review board approval (University Cancer Center Frankfurt). Real-time PCR was carried out in an iQ5 real-time PCR detection system (Bio-Rad, Munich, Germany) with ABsolute™ Blue qPCR SYBR-Green Fluorescein Mix (Thermo Fisher Scientific, Waltham, MA, USA). Gene expression data were normalized to the internal controls 18S ribosomal RNA (18S rRNA) and succinate dehydrogenase complex, subunit A, flavoprotein variant (SDHA) using the ddCT method and the iQ5 software (Bio-Rad). Primer sequences: 18S rRNA, forward 5′-CGGCTACCACATCC AAGGAA-3′ and reverse 5′-GCTGGAATTACCGCGGCT-3′; BIM (35), CITED2 (36), FOXO1 forward 5′-ATACATATGGCC AATCCAGCATG-3′ and reverse 5′-CTTCAAGAGTCCAGGC GCAC-3′; FOXO3a forward 5′-GAAGTTCCCCAGCGACT TGG-3′ and reverse 5′-CCAACCCATCAGCATCCATG-3′; glucose transporter 1 (GLUT1), forward 5′-GATTGGCTCCT T CTCTGTGG-3′ and reverse 5′-TCAAAGGACTTGCCCAG TTT-3′; manganese superoxide dismutase (MnSOD2) forward 5′-AAGGGAGATGTTACAGCCCAGATA-3′ and reverse 5′-TCC AGAAAATGCTATGATTGATATGAC-3′ (RTPrimerDB ID 2593); SDHA forward 5′-TGGGAACAAGAGGGCAT CTG-3′ and reverse 5′-CCACCACTGCATCAAATTCATG-3′; solute carrier family 16, member 3 (monocarboxylic acid transporter 4) (SLC16A3) (37), TRAIL (35), vascular endothelial growth factor (VEGF), forward 5′-CTACCTCCACCATGCCAAGT-3′ and reverse 5′-ATGTTGGACTCCTCAGTGGG-3′.

Glioma tissue specimens were derived from the University Cancer Center tumor bank (Goethe University, Frankfurt am Main, Germany) and fixed in 4% paraformaldehyde (formalin). The use of patient material was endorsed by the local ethics committee (Goethe University). Immunohistochemistry for FOXO3a and HIF1α was performed as previously described (38) using freshly cut 3 μm thick slides from formalin-fixed, paraffin-embedded tissue samples on the automated IHC staining system Discovery XT (Roche/Ventana, Tucson, AZ, USA).

LNT-229 cells stably expressing FOXO3a-GFP were seeded on glass coverslips coated with poly-L-lysine and allowed to adhere overnight. After treatment, cells were washed with PBS, fixed in 2% paraformaldehyde, washed twice with PBS, stained with 4′,6-diamidino-2-phenyl-indole (DAPI, 1 μg/ml in PBS) and analyzed by fluorescence microscopy (BZ-8000; Keyence, Osaka, Japan).

Experiments were performed at least three times with similar results. Data analysis was carried out with Excel (Microsoft, Seattle, WA, USA). Significance was tested using the two-tailed Student’s t-test.

In order to assess expression and localization of FOXO3a in gliomas, we performed immunohistochemistry on tissue samples of a panel of gliomas with different WHO grades. We observed that expression increased with WHO grade and accentuated in perinecrotic and perivascular tumor regions (Fig. 1A). To identify hypoxic areas, we additionally performed immunohistochemical staining of HIF1α. This revealed a similar distribution pattern of FOXO3a and HIF1α especially in perinecrotic ‘pseudopalisading’ glioma cells (Fig. 1B) where restricted availability of oxygen and nutrients is expected (39). Subsequently, we compared FOXO3a mRNA and protein expression in different established glioma cell lines and normal brain. All glioma cell lines exhibited elevated FOXO3a expression compared to normal brain (Fig. 1C) (data not shown).

To test our immunohistochemical observations, we analyzed FOXO3a expression under hypoxic conditions in vitro. Both mRNA and protein levels increased in LNT-229 glioma cells (Fig. 1D). However, shRNA-mediated silencing of the HIF1α gene did not result in reduced FOXO3a levels indicating that HIF1α was not the primary mediator of elevated FOXO3a expression under hypoxic conditions (Fig. 1E). As HIF2α may compensate for HIF1α deficiency, we also analyzed FOXO3a levels in HIF2α knockdown as well as in HIF1α-HIF2α double-knockdown LNT-229 cells. Both in normoxia and in hypoxia, depletion of HIF2α or simultaneous depletion of HIF1α and HIF2α did not affect FOXO3a expression (data not shown).

First, we verified that inhibition of PKB activity resulted in dephosphorylation of FOXO1 and FOXO3a in our glioma cell line (Fig. 2A). To monitor changes in FOXO3a subcellular localization, we carried out immunofluorescence studies employing FOXO3a-GFP cells (25). Transient expression of FOXO3a-GFP induced cell death in LNT-229 cells and this cell death was more pronounced than when introducing FOXO1-GFP (Fig. 2B). As both FOXO3a and FOXO1 alone may act pro-apoptotically (40,41), we analyzed FOXO1 expression after transient transfection of FOXO3a constructs. Enhanced FOXO3a expression did not affect FOXO1 mRNA levels (Fig. 2C). Using a 4-hydroxytamoxifen (4-OHT) inducible, constitutive nuclear FOXO3a-TM-construct (27) and the corresponding control FOXO3a-TM-ER-dDB without DNA-binding domain, we observed a strong increase in TRAIL expression induced by FOXO3a (Fig. 2D). However, cell death occurring upon transfection with constructs encoding FOXO3a and a constitutive nuclear FOXO3a-A3 was inhibitable by Z-VAD-fmk only to a lesser extent whereas Z-VAD-fmk completely reversed TRAIL-induced cell death which served as a positive control (data not shown). Presumably, like TP53, FOXO3a may trigger both apoptosis and autophagy (42) and cell death following FOXO3a overexpression has frequently been described (41).

Counteracted by phosphatase and tensin homolog (PTEN), insulin and growth factor signaling activates PKB that in turn phosphorylates FOXO3a at three conserved residues and thereby causes its cytoplasmic localization (5,43–45). To study changes in the subcellular localization of FOXO3a, we generated LNT-229-cells stably expressing FOXO3a-GFP by repeated selection and fluorescence-activated cell sorting (FACS). Using these cells, we observed FOXO3a shifting to the nucleus after implementation of different stressors: serum deprivation, hypoxia and hydrogen peroxide (H₂O₂) (Fig. 2E). Extent and duration of this shift varied depending on the kind of stress. Pharmacological blockade of PKB via AKT inhibitor VIII caused an immediate and durable translocation to the nucleus that could partially be reverted by addition of 10% FCS to the medium (data not shown). Addition of 0.5 mM H₂O₂ forced FOXO3a to permanently stay in the nucleus even in the presence of FCS. Nuclear shifting of FOXO3a after serum starvation was strongly enhanced by hypoxia and always a temporary event of short duration (<4 h) (Fig. 2F).

ROS are byproducts of aerobic metabolism and involved in diverse cellular functions. On the one hand and at ‘lower levels’, they support cell proliferation through the activation of growth-related signaling pathways, on the other hand and at ‘higher levels’, they promote apoptosis. To determine whether FOXO3a regulation contributes to the resistance of malignant glioma cells under conditions of low glucose, low oxygen and high ROS levels, we used four different siRNA sequences targeting FOXO3a and yielding comparable results. MnSOD2, one of the main antioxidant enzymes and a FOXO3a target gene (46), was upregulated following stimulation with H₂O₂. Transient siRNA-mediated suppression of FOXO3a significantly reduced MnSOD2 expression (Fig. 2G). LNT-229-shFOXO3a cells (Fig. 2H) were employed to determine intracellular ROS by flow cytometric analysis of the ROS-sensitive dye dichlorodihydrofluorescein diacetate (H₂DCFDA). An increase in intracellular ROS was observed after administration of 0.5–1 mM H₂O₂ in LNT-229-shFOXO3a cells compared to the control cells (Fig. 3A). Loss of FOXO3a not only affected the intracellular ROS homeostasis but also elevated cell death rates after addition of H₂O₂. Addition of 10% FCS to the cell culture medium increased cell death in both LNT-229-shFOXO3a and control cells. Presumably, more ROS were generated by accelerated metabolic processes in the presence of serum, exceeding the detoxification capacity of both cell lines. However, also in this setup, control cells displayed better survival rates than LNT-229-shFOXO3a cells (Fig. 3B). LNT-229-shFOXO3a cells accumulated more intracellular ROS than their control cells, and this was even more pronounced under hypoxia. Severe hypoxia amplified ROS production in both cell lines, again more in LNT-229-shFOXO3a than in control cells (Fig. 3C). Addition of serum also boosted ROS production (data not shown).

To simulate conditions of the ‘perinecrotic niche’, we subjected LNT-229 cells to serum-free medium supplemented with 2 mM glucose and normoxia (21% O₂), hypoxia (1% O₂) or severe hypoxia (0.1% O₂). Interestingly, FOXO3a gene suppression conferred robust protection from cell death occurring under starvation conditions (Fig. 3D). Experiments with three different siRNAs targeting FOXO3a revealed the same phenotype (data not shown). Additionally, we employed an adenoviral dominant-negative FOXO3a construct (AdFOXO3aDN) and its control (PADEG) (24). This approach also confirmed the protective effect of FOXO3a inhibition under starvation conditions (Fig. 3E). We measured caspase-3 activity and found neither caspase-3 activation in hypoxia nor a difference between LNT-229-shFOXO3a and control cells. Therefore, in this in vitro paradigm, cell death under hypoxic conditions was non-apoptotic (47) and not affected by a pro-apoptotic function of FOXO3a (Fig. 3F).

To figure out the processes leading to cell death under starvation conditions, we determined glucose concentrations in culture supernatants of LNT-229-shFOXO3a and control cells. Glucose consumption generally increased in hypoxia and decreased in LNT-229-shFOXO3a cells (Fig. 4A). Similar effects were observed in AdFOXO3aDN cells (Fig. 4B). This difference was not due to a lower proliferation rate of LNT-229-shFOXO3a cells, as assessed by BrdU-incorporation assays (Fig. 4C) and crystal violet staining (Fig. 4D). We also analyzed the cells’ glucose flux using the fluorescent D-glucose analog 2-NBDG (48) and flow cytometry. Again, LNT-229-shFOXO3a cells took up less 2-NBDG than the corresponding control cells both in normoxia and in hypoxia (Fig. 4E).

To investigate whether glioma cells compensate their reduced glucose uptake by enhanced oxidative phosphorylation, we determined oxygen consumption by ABL-80 FLEX blood gas analyzer. Conversely, FOXO3a knockdown resulted in a decrease in oxygen consumption (Fig. 4F). These results were confirmed by experiments conducted with the XF96 extracellular flux analyzer (data not shown).

Analysis of primary tumor tissue revealed a perinecrotic expression of FOXO3a matching the regions with higher levels of HIF1α (Fig. 1B). We therefore aimed to explore potential interactions between FOXO3a and HIF1α. HIF-specific transcriptional activity, examined by luciferase reporter assay (3HRE-pTK-luc construct), increased under hypoxic conditions. FOXO3a knockdown augmented this effect on HRE reporter gene activity (Fig. 5A). To confirm an inhibitory effect of FOXO3a on HIF1α, we transiently overexpressed a constitutive nuclear FOXO3a-A3 where all three PKB phosphorylation sites were mutated to alanine (49). This resulted in a marked reduction of HRE reporter gene activity (Fig. 5B). Additionally, we analyzed the expression of several well-known HIF1 target genes using RT-qPCR (50). The expression of vascular endothelial growth factor (VEGF) and the monocarboxylate transporter solute carrier family 16 member 3 (SLC16A3) increased under hypoxic conditions and this was further enhanced by FOXO3a knockdown, suggesting that FOXO3a inhibited HIF1 transcriptional activity in LNT-229 glioma cells (Fig. 5C). To study whether the protective effect of FOXO3a knockdown on survival under hypoxia was mediated by HIF1α, we applied siRNA targeting FOXO3a under starvation conditions in LNT-229 cells stably transfected with an shRNA targeting HIF1α or its control (Sima), respectively (30). Loss of HIF1α abolished the protection conferred by FOXO3a knockdown in hypoxia (Fig. 5D).

The expression of GLUT1, another long-known HIF1 target gene, decreased following FOXO3a knockdown. Though induced under hypoxic conditions (51) both in LNT-229-shFOXO3a and control cells, GLUT1 expression of control cells always exceeded that of cells lacking FOXO3a (Fig. 5E). To verify the RT-qPCR data, we analyzed surface GLUT1 levels by flow cytometry. This technique also revealed higher GLUT1 amounts in hypoxia and lower amounts in LNT-229-shFOXO3a cells (Fig. 5F). The expression of HIF1 target genes was thus differentially affected by FOXO3a knockdown in our glioma cell line.

As reported by Bakker et al, FOXO3a is induced in hypoxia and activates the transcription of CITED2 (36). CITED2 in turn impairs HIF1 activity by competing with HIF1 for binding to the transcriptional coactivator CBP/p300 (52). We hypothesized that CITED2 also could be the mediator of the phenotype observed in our cell line. Induction of CITED2 in hypoxia was suppressed by FOXO3a knockdown (Fig. 6A). Nevertheless, a CITED2 knockdown did not display the protective phenotype of the FOXO3a knockdown under hypoxic conditions (Fig. 6B). Regarding the glucose consumption of cells transiently transfected with siRNA, a CITED2 knockdown increased glucose consumption instead of decreasing it (Fig. 6C). Obviously, CITED2 was not the missing link between FOXO3a and HIF1 and did not account for either the diminished GLUT1 expression or finally the protection under starvation conditions.

As GLUT1 expression is known to be downregulated by TP53 (53) and FOXO3a can promote the translocation of TP53 to the cytoplasm (54), we reflected on TP53 potentially being involved in the reduced glucose consumption associated with FOXO3a knockdown. Indeed, also in our glioma cell line, FOXO3a knockdown increased TP53 transcriptional activity (Fig. 6D) as well as its phosphorylation at serine 15 (Fig. 6E) that is considered necessary for TP53 function (55). Considering our previous studies on the role of TP53 in LNT-229 glioma cells (32), the interplay of FOXO3a and TP53 may partially explain the phenotype observed under starvation conditions because antagonizing TP53 enhances cell death when availability of glucose and oxygen is restricted. Elevated TP53 activity in LNT-229-shFOXO3a cells could therefore improve cell survival in unfavorable environments.